Machine learning methods for predicting cell phenotype using holographic imaging

Abstract

Provided herein are methods for determining a cell phenotype, such as activation state, of a population of T cells. In some aspects, the provided methods are label-free methods. In some aspects, the provided methods can be used for monitoring a cell phenotype, such as activation state, of a culture of T cells. Also provided herein are computing devices for use in performing the provided methods.

Description

MACHINE LEARNING METHODS FOR PREDICTING CELL PHENOTYPE USING HOLOGRAPHIC IMAGINGCross-Reference to Related Applications

[0001] This application claims priority to U.S. Provisional Application No. 63 / 431,635, filed December 9, 2022, entitled “MACHINE LEARNING METHODS FOR PREDICTING CELL PHENOTYPE USING HOLOGRAPHIC IMAGING,” and U.S. Provisional Application No. 63 / 537,467, filed September 8, 2023, entitled “MACHINE LEARNING METHODS FOR PREDICTING CELL PHENOTYPE USING HOLOGRAPHIC IMAGING,” the contents of which are incorporated by reference in their entirety for all purposes.Incorporation By Reference of Sequence Listing

[0002] The present application is being filed along with a Sequence Listing in electronic format. The Sequence Listing is provided as a file entitled 735042024740SeqList.xml, created December 7, 2023, which is 38,941 bytes in size. The information in the electronic format of the Sequence Listing is incorporated by reference in its entirety.Field

[0003] The present disclosure relates to methods for determining a cell phenotype, such as activation state, of a population of T cells. In some aspects, the provided methods are label- free methods. In some aspects, the provided methods can be used for monitoring a cell phenotype, such as activation state, of a culture of T cells. Also provided herein are computing devices for use in performing the provided methods.Background

[0004] Existing methods for determining a cell phenotype, such as activation state, of T cells, for example during in vitro or ex vivo culture of the T cells, can be time-consuming or require specialized reagents, instrumentation, or trained operators. In some instances, existing methods involve directly manipulating or labelling T cells, for example in a manner involving incubation with immunoaffinity-based reagents that can interfere with the quality or function of the T cells. Such methods can pose risks of contamination of or damage to cells prior toany downstream processing. Improved methods for accurately determining a cell phenotype, such as activation state, of T cells are needed. Such methods are useful, for example, in the production of cell products, for instance cell therapy products that contain T cells. Provided herein are methods and computing devices that meet such needs.Summary

[0005] Provided herein in some embodiments is a method for determining a cell phenotype of a population of T cells, comprising determining, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by T cells with a cell phenotype of interest, wherein the population-level output measure is determined based on a plurality of population-level statistics, wherein: each population-level statistic is of one or more input measures for a cellular feature of a plurality of cellular features derived from holographic information obtained for the population of cells, the plurality of population-level statistics comprising one or more population-level statistics for each of the plurality of cellular features; and each input measure is from an individual cell of the population of cells.

[0006] In some of any embodiments, the method is for determining the activation state of the population of T cells, and the marker is expressed by activated T cells. In some of any embodiments, the marker is CD137 (4-1BB).

[0007] In some of any embodiments, the method is for determining the memory phenotype of the population of T cells, and the marker, e.g., CCR7, is expressed by T cells having a central memory phenotype or a stem cell memory phenotype.

[0008] Also provided herein in some embodiments is a method for determining the activation state of a population of T cells, comprising determining, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by activated T cells, wherein the population-level output measure is determined based on a plurality of population-level statistics, wherein: each population-level statistic is of one or more input measures for a cellular feature of a plurality of cellular features derived from holographic information obtained for the population of cells, the plurality of population-level statistics comprising one or more population-level statistics for each of the plurality of cellular features; and each input measure is from an individual cell of the population of cells.

[0009] In some of any embodiments, the method is for determining recombinant receptor expression of the population of T cells, and the marker is a recombinant receptor introducedinto T cells of the population of cells prior to when the holographic information is obtained. In some of any embodiments, the recombinant receptor is a chimeric antigen receptor (CAR) or an engineered T cell receptor (TCR).

[0010] Also provided herein in some embodiments is a method for determining recombinant receptor expression of a population of T cells, comprising determining, for a population of cells comprising T cells, a population-level output measure of expression of a recombinant receptor introduced into T cells of the population of T cells, wherein the population-level output measure is determined based on a plurality of population-level statistics, wherein: each population-level statistic is of one or more input measures for a cellular feature of a plurality of cellular features derived from holographic information obtained for the population of cells, the plurality of population-level statistics comprising one or more population-level statistics for each of the plurality of cellular features; and each input measure is from an individual cell of the population of cells.

[0011] In some of any embodiments, the method comprises determining the plurality of population-level statistics from the one or more input measures for each of the plurality of cellular features.

[0012] In some of any embodiments, the method comprises determining the one or more input measures for each of the plurality of cellular features from the holographic information.

[0013] In some of any embodiments, the method comprises obtaining the holographic information.

[0014] Also provided herein in some embodiments is a method for determining a cell phenotype of a population of T cells, comprising: (a) obtaining holographic information for a population of cells comprising T cells; (b) determining one or more input measures for each of a plurality of cellular features derived from the holographic information, wherein each input measure is from an individual cell in the population of cells; (c) determining a plurality of population-level statistics, wherein each population-level statistic is of the one or more input measures for a cellular feature of the plurality of cellular features, and the plurality of population-level statistics comprises one or more population-level statistics for each of the plurality of cellular features; and (d) determining, based on the plurality of population-level statistics, a population-level output measure of expression of a marker expressed by T cells with a cell phenotype of interest for the population of cells.

[0015] In some of any embodiments, the method is for determining the activation state of the population of T cells, and the marker is expressed by activated T cells. In some of any embodiments, the marker is CD137 (4-1BB).

[0016] Also provided herein in some embodiments is a method for determining the activation state of a population of T cells, comprising: (a) obtaining holographic information for a population of cells comprising T cells; (b) determining one or more input measures for each of a plurality of cellular features derived from the holographic information, wherein each input measure is from an individual cell in the population of cells; (c) determining a plurality of population-level statistics, wherein each population-level statistic is of the one or more input measures for a cellular feature of the plurality of cellular features, and the plurality of population-level statistics comprises one or more population-level statistics for each of the plurality of cellular features; and (d) determining, based on the plurality of population-level statistics, a population-level output measure of expression of a marker expressed by activated T cells for the population of cells.

[0017] In some of any embodiments, the method is for determining recombinant receptor expression of the population of T cells, and the marker is a recombinant receptor introduced into T cells of the population of cells prior to when the holographic information is obtained. In some of any embodiments, the recombinant receptor is a chimeric antigen receptor (CAR) or an engineered T cell receptor (TCR).

[0018] Also provided herein in some embodiments is a method for determining recombinant receptor expression of a population of T cells, comprising: (a) obtaining holographic information for a population of cells comprising T cells; (b) determining one or more input measures for each of a plurality of cellular features derived from the holographic information, wherein each input measure is from an individual cell in the population of cells; (c) determining a plurality of population-level statistics, wherein each population-level statistic is of the one or more input measures for a cellular feature of the plurality of cellular features, and the plurality of population-level statistics comprises one or more populationlevel statistics for each of the plurality of cellular features; and (d) determining, based on the plurality of population-level statistics, a population-level output measure of expression for the population of cells of a recombinant receptor introduced into T cells of the population of cells.

[0019] In some of any embodiments, the method further comprises engineering the population of T cells following the determining of the population-level output measure.

[0020] In some of any embodiments, the holographic information is obtained by differential digital holographic microscopy (DDHM). In some of any embodiments, the obtaining the holographic information comprises imaging the population of cells using DDHM.

[0021] In some of any embodiments, the one or more population-level statistics for at least one, optionally each, of the plurality of cellular features comprise one or more quantiles of the one or more input measures of the cellular feature. In some of any embodiments, the one or more population-level statistics for each of the plurality of cellular features comprise one or more quantiles of the one or more input measures of the cellular feature. In some of any embodiments, the one or more quantiles are selected from (compise one or more of) the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of the one or more input measures of the cellular feature.

[0022] In some of any embodiments, the one or more population-level statistics for at least one, optionally each, of the plurality of cellular features are determined by applying a distribution-based pooling filter to the one or more input measures of the cellular feature. In some of any embodiments, the one or more population-level statistics for each of the plurality of cellular features are determined by applying a distribution-based pooling filter to the one or more input measures of the cellular feature.

[0023] In some of any embodiments, the population-level output measure is the number of cells, the percentage of cells, the proportion of cells, or the density of cells of the population of cells that express the marker.

[0024] In some of any embodiments, the population of cells is from a culture of cells being cultured in vitro or ex vivo.

[0025] In some of any embodiments, the holographic information is obtained during the in vitro or ex vivo culture of the culture of cells.

[0026] In some of any embodiments, the population-level output measure is of expression of the marker during the in vitro or ex vivo culture of the culture of cells.

[0027] In some of any embodiments, the in vitro or ex vivo culture is under conditions to expand T cells of the culture of cells.

[0028] In some of any embodiments, the in vitro or ex vivo culture is in a bioreactor.

[0029] In some of any embodiments, the population of cells are incubated under T cell stimulating conditions prior to when the holographic information is obtained. In some of any embodiments, the method comprises incubating the population of cells under T cell stimulating conditions prior to when the holographic information is obtained.

[0030] In some of any embodiments, the incubation is prior to the in vitro or ex vivo culture.

[0031] In some of any embodiments, the T cell stimulating conditions comprise incubation in the presence of T cell stimulatory agents that induce a primary activation signal and a costimulatory signal in T cells. In some of any embodiments, the T cell stimulatory agents comprise an anti-CD3 antibody or antibody fragment. In some of any embodiments, the T cell stimulatory agents comprise an anti-CD28 antibody or antibody fragment. In some of any embodiments, the T cell stimulatory agents are immobilized on a bead.

[0032] In some of any embodiments, the T cell stimulatory agents are immobilized on an oligomeric streptavidin mutein reagent.

[0033] In some of any embodiments, a recombinant receptor is introduced into T cells of the population of cells prior to when the holographic information is obtained. In some of any embodiments, the method comprises introducing a recombinant receptor into T cells of the population of cells prior to when the holographic information is obtained.

[0034] In some of any embodiments, the introducing comprising contacting the population of cells with an agent comprising a polynucleotide encoding the recombinant receptor.

[0035] In some of any embodiments, the recombinant receptor is a chimeric antigen receptor (CAR) or an engineered T cell receptor (TCR).

[0036] In some of any embodiments, the population of cells is enriched for T cells. In some of any embodiments, at least 50, 55, 60, 65, 70, 75, 80, 85, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, or 100% of the population of cells are T cells.

[0037] In some of any embodiments, the population-level output measure is determined by providing the plurality of population-level statistics as input to a machine learning model trained to predict population-level output measures of expression of the marker based on population-level statistics of the plurality of cellular features.

[0038] In some of any embodiments, the machine learning model is trained using a dataset of reference population-level statistics, wherein: for each of a first plurality of referencepopulations of cells comprising T cells, the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of the plurality of cellular features, wherein: each reference population-level statistic is of one or more reference input measures for a cellular feature of the plurality of cellular features derived from holographic information obtained for the reference population of cells; and each reference input measure is from an individual cell of the reference population of cells.

[0039] In some of any embodiments, the machine learning model is trained using a dataset of reference population-level output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures comprises a reference population-level output measure of expression of the marker for the reference population of cells, wherein: the first and second pluralities of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo', and reference populations of cells from the first plurality of reference populations of cells are from the same reference culture of cells as reference populations of cells from the second plurality of reference populations of cells.

[0040] Also provided herein in some embodiments is a method for training a machine learning model that predicts a cell phenotype of a population of T cells, comprising training a machine learning model using: (i) a dataset of reference population-level statistics, wherein for each of a first plurality of reference populations of cells comprising T cells, the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of a plurality of cellular features, wherein: each reference population-level statistic is of one or more reference input measures for a cellular feature of the plurality of cellular features derived from holographic information obtained for the reference population of cells; and each reference input measure is from an individual cell of the reference population of cells; and (ii) a dataset of reference population-level output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures comprises a reference population-level output measure of expression of a marker expressed by T cells with a cell phenotype of interest for the reference population of cells, wherein: the first and second pluralities of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo', and reference populations of cells from the first plurality of reference populations of cells are from the samereference culture of cells as reference populations of cells from the second plurality of reference populations of cells; whereby the machine learning model is trained to predict population-level output measures of expression of the marker based on population-level statistics of the plurality of cellular features.

[0041] In some of any embodiments, the method is for determining the activation state of the population of T cells, and the marker is expressed by activated T cells. In some of any embodiments, the marker is CD137 (4-1BB).

[0042] Also provided herein in some embodiments is a method for training a machine learning model that predicts the activation state of a population of T cells, comprising training a machine learning model using: (i) a dataset of reference population-level statistics, wherein for each of a first plurality of reference populations of cells comprising T cells, the dataset of reference population-level statistics comprises one or more reference populationlevel statistics for each of a plurality of cellular features, wherein: each reference populationlevel statistic is of one or more reference input measures for a cellular feature of the plurality of cellular features derived from holographic information obtained for the reference population of cells; and each reference input measure is from an individual cell of the reference population of cells; and (ii) a dataset of reference population-level output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures comprises a reference population-level output measure of expression of a marker expressed by activated T cells for the reference population of cells, wherein: the first and second pluralities of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo', and reference populations of cells from the first plurality of reference populations of cells are from the same reference culture of cells as reference populations of cells from the second plurality of reference populations of cells; whereby the machine learning model is trained to predict populationlevel output measures of expression of the marker based on population-level statistics of the plurality of cellular features.

[0043] In some of any embodiments, the method is for determining recombinant receptor expression of the population of T cells, and the marker is a recombinant receptor introduced into T cells of the population of cells prior to when the holographic information is obtained.In some of any embodiments, the recombinant receptor is a chimeric antigen receptor (CAR) or an engineered T cell receptor (TCR).

[0044] Also provided herein in some embodiments is a method for training a machine learning model that predicts recombinant receptor expression of a population of T cells, comprising training a machine learning model using: (i) a dataset of reference populationlevel statistics, wherein for each of a first plurality of reference populations of cells comprising T cells, the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of a plurality of cellular features, wherein: each reference population-level statistic is of one or more reference input measures for a cellular feature of the plurality of cellular features derived from holographic information obtained for the reference population of cells; and each reference input measure is from an individual cell of the reference population of cells; and (ii) a dataset of reference population-level output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures comprises a reference population-level output measure of expression for the reference population of cells of a recombinant receptor introduced into T cells of the reference population of cells, wherein: the first and second pluralities of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo', and reference populations of cells from the first plurality of reference populations of cells are from the same reference culture of cells as reference populations of cells from the second plurality of reference populations of cells; whereby the machine learning model is trained to predict population-level output measures of expression of the recombinant receptor based on population-level statistics of the plurality of cellular features.

[0045] In some of any embodiments, the plurality of cellular features comprise one or of intensity skewness, intensity correlation, intensity homogeneity, intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean.

[0046] In some of any embodiments, the plurality of cellular features are selected from (comprises one or more of) intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean. Also provided herein in some embodiments is a method for determining the activation state of a population of T cells, comprising determining, for a population of cells comprising T cells, expression of a marker expressed by activated T cells, wherein the marker is CD 137, and the determining is based onone or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features are selected from (comprises one or more of) intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean; and each input measure is from an individual cell of the population of cells.

[0047] In some of any embodiments, the plurality of cellular features comprises intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean.

[0048] Provided herein in some embodiments is a method for determining the activation state of a population of T cells, comprising determining, for a population of cells comprising T cells, expression of a marker expressed by activated T cells, wherein the marker is CD137 (4- IBB), and the determining is based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean; and each input measure is from an individual cell of the population of cells.

[0049] Also provided herein, in some embodiments, is a method for determining the activation state of a population of T cells, comprising determining, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by activated T cells, wherein the marker is CD137 (4-1BB), and the population-level output measure is determined based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean; and each input measure is from an individual cell of the population of cells.

[0050] In some of any embodiments, the plurality of cellular features comprises one or more of intensity skewness, intensity correlation, and intensity homogeneity. In some of any embodiments, the plurality of cellular features comprises intensity skewness, intensity correlation, and intensity homogeneity.

[0051] Also provided herein in some embodiments is a method for determining the activation state of a population of T cells, comprising determining, for a population of cellscomprising T cells, expression of a marker expressed by activated T cells, wherein the marker is CD137 (4-1BB), and the determining is based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises intensity skewness, intensity correlation, and intensity homogeneity; and each input measure is from an individual cell of the population of cells.

[0052] Also provided herein, in some embodiments, is a method for determining the activation state of a population of T cells, comprising determining, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by activated T cells, wherein the marker is CD137 (4-1BB), and the population-level output measure is determined based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises intensity skewness, intensity correlation, and intensity homogeneity; and each input measure is from an individual cell of the population of cells.

[0053] In some of any embodiments, the plurality of cellular features comprises one or more of intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean. In some of any embodiments, the plurality of cellular features comprises intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean.

[0054] In some of any embodiments, the plurality of cellular features comprises one or more of peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height. In some of any embodiments, the plurality of cellular features comprises peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height.

[0055] In some of any embodiments, the plurality of cellular features comprises one or more of cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter. In some of any embodiments, the plurality of cellular features comprises cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter.

[0056] Also provided herein, in some embodiments, is a method for determining the memory phenotype of a population of T cells, comprising determining, for a population ofcells comprising T cells, expression of a marker expressed by central memory T cells or stem cell memory T cells, wherein the marker is CCR7, and the determining is based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises one or more of cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter; and each input measure is from an individual cell of the population of cells.

[0057] In some of any embodiments, the plurality of cellular features comprises intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean. In some of any embodiments, the plurality of cellular features comprises one or more of peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height.

[0058] In some of any embodiments, the plurality of cellular features comprises peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height.

[0059] In some of any embodiments, the plurality of cellular features comprises one or more of cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter. In some of any embodiments, the plurality of cellular features comprises cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter.

[0060] Also provided herein, in some embodiments, is a method for determining the memory phenotype of a population of T cells, comprising determining, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by central memory T cells or stem cell memory T cells, wherein the marker is CCR7, and the population-level output measure is determined based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises one or more of cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter; and each input measure is from an individual cell of the population of cells.

[0061] In some embodiments, the plurality of features comprises cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter.

[0062] Also provided herein in some embodiments is a method for determining recombinant receptor expression of a population of T cells, comprising determining, for a population of cells comprising T cells, expression of a recombinant receptor introduced into T cells of the population of cells, wherein the determining is based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height; and each input measure is from an individual cell of the population of cells.

[0063] In some of any embodiments, the one or more input measures for at least one, optionally each, of the plurality of cellular features are determined by providing the holographic information for individual cells of the population of cells to a convolutional neural network, wherein: the plurality of cellular features are cellular features extracted by the convolutional neural network; and the one or more input measures are determined from the convolutional neural network. In some of any embodiments, the one or more input measures for each of the plurality of cellular features are determined by providing the holographic information for individual cells of the population of cells to a convolutional neural network, wherein: the plurality of cellular features are cellular features extracted by the convolutional neural network; and the one or more input measures are determined from the convolutional neural network.

[0064] In some of any embodiments, the one or more reference input measures for at least one, optionally each, of the plurality of cellular features are determined by providing the holographic information for individual cells of the reference population of cells to a convolutional neural network, wherein: the plurality of cellular features are cellular features extracted by the convolutional neural network; and the one or more reference input measures are determined from the convolutional neural network. In some of any embodiments, the one or more reference input measures for each of the plurality of cellular features are determined by providing the holographic information for individual cells of the reference population of cells to a convolutional neural network, wherein: the plurality of cellular features are cellular features extracted by the convolutional neural network; and the one or more reference input measures are determined from the convolutional neural network.

[0065] In some of any embodiments, the convolutional neural network is trained using a dataset of reference holographic information, wherein for each of a first plurality of reference populations of cells comprising T cells, the dataset of reference holographic information comprises holographic information obtained for individual cells of the reference population of cells.

[0066] In some of any embodiments, the one or more reference population-level statistics for at least one, optionally each, of the plurality of cellular features comprise one or more quantiles of the one or more reference input measures for the cellular feature. In some of any embodiments, the one or more reference population-level statistics for each of the plurality of cellular features comprise one or more quantiles of the one or more reference input measures for the cellular feature. In some of any embodiments, the one or more quantiles are selected from the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of the one or more reference input measures for the cellular feature.

[0067] In some of any embodiments, the one or more reference population-level statistics for at least one, optionally each, of the plurality of cellular features are determined by applying a distribution-based pooling filter to the one or more reference input measures for the cellular feature. In some of any embodiments, the one or more reference population-level statistics for each of the plurality of cellular features are determined by applying a distribution-based pooling filter to the one or more reference input measures for the cellular feature.

[0068] Also provided herein in some embodiments is a method for training a machine learning model that predicts a cell phenotype of a population of T cells, comprising: (a) training a convolutional neural network using a dataset of reference holographic information, wherein for each of a first plurality of reference populations of cells comprising T cells, the dataset of reference holographic information comprises holographic information obtained for individual cells of the reference population of cells; (b) determining, from the convolutional neural network, one or more reference input measures for each cellular feature of a plurality of cellular features derived from the holographic information, wherein the plurality of cellular features are cellular features extracted by the convolutional neural network, and each reference input measure is from an individual cell of a reference population of cells; (c) determining a dataset of reference population-level statistics, wherein the dataset of referencepopulation-level statistics comprises one or more reference population-level statistics for each of the plurality of cellular features, each reference population-level statistic determined by applying a distribution-based pooling filter to the one or more reference input measures for a cellular feature of the plurality of cellular features; and (d) training a machine learning model using the dataset of reference population-level statistics and a dataset of reference populationlevel output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures comprises a reference population-level output measure of expression of a marker expressed by T cells with a cell phenotype of interest for the reference population of cells, wherein: the first and second pluralities of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo', and reference populations of cells from the first plurality of reference populations of cells are from the same reference culture of cells as reference populations of cells from the second plurality of reference populations of cells; whereby the machine learning model is trained to predict population-level output measures of expression of the marker based on population-level statistics of the plurality of cellular features.

[0069] In some of any embodiments, the method is for determining the activation state of the population of T cells, and the marker is expressed by activated T cells. In some of any embodiments, the marker is CD137 (4-1BB).

[0070] Also provided herein in some embodiments is a method for training a machine learning model that predicts the activation state of a population of T cells, comprising: (a) training a convolutional neural network using a dataset of reference holographic information, wherein for each of a first plurality of reference populations of cells comprising T cells, the dataset of reference holographic information comprises holographic information obtained for individual cells of the reference population of cells; (b) determining, from the convolutional neural network, one or more reference input measures for each cellular feature of a plurality of cellular features derived from the holographic information, wherein the plurality of cellular features are cellular features extracted by the convolutional neural network, and each reference input measure is from an individual cell of a reference population of cells; (c) determining a dataset of reference population-level statistics, wherein the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of the plurality of cellular features, each reference population-level statistic determined byapplying a distribution-based pooling filter to the one or more reference input measures for a cellular feature of the plurality of cellular features; and (d) training a machine learning model using the dataset of reference population-level statistics and a dataset of reference populationlevel output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures comprises a reference population-level output measure of expression of a marker expressed by activated T cells for the reference population of cells, wherein: the first and second pluralities of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo', and reference populations of cells from the first plurality of reference populations of cells are from the same reference culture of cells as reference populations of cells from the second plurality of reference populations of cells; whereby the machine learning model is trained to predict population-level output measures of expression of the marker based on populationlevel statistics of the plurality of cellular features.

[0071] In some of any embodiments, the method is for determining recombinant receptor expression of the population of T cells, and the marker is a recombinant receptor introduced into T cells of the population of cells prior to when the holographic information is obtained. In some of any embodiments, the recombinant receptor is a chimeric antigen receptor (CAR) or an engineered T cell receptor (TCR).

[0072] Also provided herein in some embodiments is a method for training a machine learning model that predicts recombinant receptor of a population of T cells, comprising: (a) training a convolutional neural network using a dataset of reference holographic information, wherein for each of a first plurality of reference populations of cells comprising T cells, the dataset of reference holographic information comprises holographic information obtained for individual cells of the reference population of cells; (b) determining, from the convolutional neural network, one or more reference input measures for each cellular feature of a plurality of cellular features derived from the holographic information, wherein the plurality of cellular features are cellular features extracted by the convolutional neural network, and each reference input measure is from an individual cell of a reference population of cells; (c) determining a dataset of reference population-level statistics, wherein the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of the plurality of cellular features, each reference population-level statistic determined byapplying a distribution-based pooling filter to the one or more reference input measures for a cellular feature of the plurality of cellular features; and (d) training a machine learning model using the dataset of reference population-level statistics and a dataset of reference populationlevel output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures comprises a reference population-level output measure for the reference population of cells of expression of a recombinant receptor introduced into T cells of the reference population of cells, wherein: the first and second pluralities of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo', and reference populations of cells from the first plurality of reference populations of cells are from the same reference culture of cells as reference populations of cells from the second plurality of reference populations of cells; whereby the machine learning model is trained to predict population-level output measures of expression of the recombinant receptor based on population-level statistics of the plurality of cellular features.

[0073] In some of any embodiments, the one or more input measures for at least one, optionally each, of the plurality of cellular features are obtained from a fully connected layer of the convolutional neural network. In some of any embodiments, the one or more input measures for each of the plurality of cellular features are obtained from a fully connected layer of the convolutional neural network.

[0074] In some of any embodiments, the holographic information for the first plurality of reference populations of cells is obtained by DDHM.

[0075] In some of any embodiments, the holographic information comprises phase information and intensity information.

[0076] In some of any embodiments, the reference population-level output measure is the number of cells, the percentage of cells, the proportion of cells, or the density of cells of the reference population of cells that express the marker.

[0077] In some of any embodiments, the dataset of reference population-level statistics and the dataset of reference population-level output measures are time-matched.

[0078] In some of any embodiments, the in vitro or ex vivo culture of the reference cultures of cells is performed under the same or similar conditions as the in vitro or ex vivo culture of the culture of cells.

[0079] In some of any embodiments, the holographic information for the first plurality of reference populations of cells is obtained during the in vitro or ex vivo culture of the reference cultures of cells.

[0080] In some of any embodiments, the dataset of reference population-level output measures are of expression of the marker during or after the in vitro or ex vivo culture of the reference cultures of cells.

[0081] In some of any embodiments, the dataset of reference population-level output measures is determined using fluorescence imaging of the second plurality of reference populations of cells. In some of any embodiments, the fluorescence imaging is by flow cytometry.

[0082] In some of any embodiments, the method is for determining the activation state of the population of T cells, and the marker is expressed by activated T cells.

[0083] In some of any embodiments, the marker is CD137 (4-1BB).

[0084] In some of any embodiments, the method is for determining the memory phenotype of the population of T cells, and the marker is expressed by T cells having a central memory phenotype or a stem cell memory phenotype. In some of any embodiments, the marker is expressed by T cells having a central memory phenotype. In some of any embodiments, the marker is expressed by T cells having a stem cell memory phenotype. In some of any embodiments, the marker is CCR7.

[0085] In some of any embodiments, the method is for determining recombinant receptor expression of the population of T cells, and the marker is a recombinant receptor introduced into T cells of the population of cells prior to when the holographic information is obtained. In some of any embodiments, the recombinant receptor is a chimeric antigen receptor (CAR) or an engineered T cell receptor (TCR).

[0086] In some of any embodiments, the first plurality of reference populations of cells are incubated under T cell stimulating conditions prior to when the holographic information for the first plurality of reference populations of cells is obtained.

[0087] In some of any embodiments, the incubation of the first plurality of reference populations of cells is performed under the same or similar conditions as the incubation of the population of cells.

[0088] In some of any embodiments, the incubation of the first plurality of reference populations of cells is prior to the in vitro or ex vivo culture of the reference cultures of cells.

[0089] In some of any embodiments, the first and / or second plurality of reference populations of cells are enriched for T cells. In some of any embodiments, the first and second plurality of reference populations of cells are each enriched for T cells.

[0090] In some of any embodiments, at least 50, 55, 60, 65, 70, 75, 80, 85, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, or 100% of the first and / or second plurality of reference populations of cells are T cells. In some of any embodiments, at least 50, 55, 60, 65, 70, 75, 80, 85, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, or 100% of each of the first and second plurality of reference populations of cells are T cells.

[0091] Also provided herein in some embodiments is a method for monitoring a cell phenotype of a culture of T cells, comprising determining, for a population of cells from a culture of cells comprising T cells that is being cultured in vitro or ex vivo, a population-level output measure of expression of a marker expressed by T cells with a cell phenotype of interest, wherein the population-level output measure is determined according to some of any of the provided methods.

[0092] In some of any embodiments, the method is for monitoring the activation state of the population of T cells, and the marker is expressed by activated T cells. In some of any embodiments, the marker is CD137 (4-1BB).

[0093] In some of any embodiments,

[0094] Also provided herein in some embodiments is a method for monitoring the activation state of a culture of T cells, comprising determining, for a population of cells from a culture of cells comprising T cells that is being cultured in vitro or ex vivo, a populationlevel output measure of expression of a marker expressed by activated T cells, wherein the population-level output measure is determined according to some of any of the provided methods.

[0095] In some of any embodiments, the method is for monitoring recombinant receptor expression of the population of T cells, and the marker is a recombinant receptor introduced into T cells of the population of cells prior to when the holographic information is obtained. In some of any embodiments, the recombinant receptor is a chimeric antigen receptor (CAR) or an engineered T cell receptor (TCR).

[0096] Also provided herein in some embodiments is a method for monitoring recombinant receptor expression of a culture of T cells, comprising determining, for a population of cells from a culture of cells comprising T cells that is being cultured in vitro or ex vivo, a population-level output measure of expression of a recombinant receptor introduced into T cells of the culture of cells, wherein the population-level output measure is determined according to some of any of the provided methods.

[0097] Also provided herein in some embodiments is a computing device comprising instructions in memory for performing some of any of the provided methods, the instructions comprising instructions for: (a) receiving the holographic information for individual cells of a population of cells comprising T cells, the one or more input measures for each of the plurality of cellular features for individual cells of a population of cells comprising T cells, or the plurality of population-level statistics for a population of cells comprising T cells; and (b) determining, according to the method, the population-level output measure for the population of cells from the holographic information, the one or more input measures for each of the plurality of cellular features, or the plurality of population-level statistics.

[0098] In some of any embodiments, the computing device further comprises in memory a machine learning model trained according to some of any of the provided methods, wherein the population-level output measure for the population of cells is determined using the machine learning model.Brief Description of the Drawings

[0099] FIG. 1 and FIG. 2 show validation results for a machine learning method for determining the overall activation state (e.g., CD137 positivity) of T cell populations using holographic imaging. The T cell populations for FIG. 1 and FIG. 2 were subjected to different stimulation processes.

[0100] FIG. 3 shows validation results for a machine learning method for determining recombinant receptor expression of T cell populations using holographic imaging.

[0101] FIG. 4 shows validation results for a machine learning method for determining the memory phenotype of T cell populations using holographic imaging.Detailed Description

[0102] Provided herein are methods for determining a cell phenotype, such as activation state, of a population of cells containing T cells. In some aspects, the provided methods do not require labeling the cells. In some aspects, the provided methods do not comprise labeling the T cells. In some aspects, the provided methods are label-free methods. In some aspects, the provided methods can be used for monitoring a cell phenotype, such as activation state, of a culture of cells containing T cells, for instance during in vitro or ex vivo culture and prior to downstream processing steps for the T cells. In some embodiments, the determination of a cell phenotype, such as activation state, is performed using machine learning. Also provided herein are methods for training a machine learning model to predict a cell phenotype, such as activation state, of a population of cells containing T cells, as well as computing devices for use in performing any of the provided methods.

[0103] Existing methods for determining a cell phenotype, such as activation state, of T cells, for example during in vitro or ex vivo culture of the T cells, can be time-consuming or require specialized reagents, instrumentation, or trained operators. In some instances, existing methods involve directly manipulating or labelling T cells, for example in a manner involving incubation with immunoaffinity-based reagents, that may interfere with the quality or function of the T cells. Such methods may risk contamination of or damage to cells prior to any downstream processing.

[0104] The provided methods allow for the label-free determination and monitoring of a cell phenotype, such as activation state, of T cells. In some aspects, the provided methods involve the determination of activation state based on expression in T cells of CD 137 (4- 1BB). The results demonstrated herein show that CD137 is a particularly useful marker for predicting overall activation state, as the results are consistent with a finding that CD 137 expression is associated with changes in characteristics that can be monitored with holographic imaging, such as the size of the activated T cells. In some aspects, and without wishing to be bound by theory, the degree to which CD 137 expression dynamically varies over time during activation contributes to its usefulness as a marker for predicting activation state during the monitoring of cells by holographic imaging. In some such embodiments, variation in CD137 expression over time is associated with changes in cell size. In some such embodiments, variation in CD137 expression is associated one or more of cell size, intensity,texture (or smoothness), and circularity. In some embodiments, an increase in CD137 expression is associated with one or more of (e.g., at least one of, at least two of, at least three of, or all of ) increased intensity, increased size (e.g., increased cell area and / or increased cell radius), increased texture (decreased cell smoothness), and lower circularity. In some embodiments, an increase in CD 137 expression is associated with one or more of (e.g., at least one of, at least two of, at least three of, or all of ) decreased intensity, decreased size (e.g., cell area and / or cell radius), greater cell smoothess, and greater circularity.

[0105] The provided methods also allow for the label-free determination and monitoring of another cell phenotype, such as the memory phenotype, of T cells. In some aspects, the provided methods involve the determination of the memory phenotype of the population of T cells. In some embodiments, the determination is based on expression in T cells of CCR7, where CCR7-positive cells are typically considered to be of an earlier memory phenotype, such as a central memory or stem cell memory phenotype, while CCR7-negative cells are typically considered to have an effector memory phenotype. See Blaeschke et al., Cancer Immunol Immunother, 67: 1053-1066 (2018). Early memory phenotypes, such as a stem cell memory or a central memory phenotype, are reported to result in sustained in vivo response of CAR T cell therapies given their proliferative capacity and effector capabilities in both hematological and solid tumor environments. See Gargett et al., Cytotherapy, 21: 593-602 (2019). This makes CCR7 a particularly relevant marker for therapeutic populations of T cells. The results demonstrated herein show that CCR7 expression is associated with changes in characteristics that can be monitored with holographic imaging. In some such embodiments, variation in CCR7 expression is associated one or more of cell size, intensity, texture (or smoothness), and circularity. In some embodiments, an increase in CCR7 expression is associated with one or more of (e.g., at least two of, at least three of, at least four of, or all of) the features of cell area, intensity mean, perimeter, equivalent peak diameter, and peak area normalized. In some embodiments, an increase in CCR7 expression is associated with one or more of (e.g., at least two of, at least three of, at least four of, or all of) the features of (i) greater cell size, (ii) smaller mean intensity, (iii) larger phase correlation, and (iv) smaller normalized radius variance. In some embodiments, greater cell size is indicated by greater area, longer perimeter, and / or larger diameter.

[0106] In some aspects, the provided methods further comprise engineering the T cells for which the phenotype is determined and / or monitored to produce a cell therapy product.

[0107] In some aspects, the provided methods involve determining, for a population of cells containing T cells, a population-level output measure of expression of a marker expressed by T cells, such as a marker expressed by activated T cells. For instance, in some embodiments, the population-level output measure is the number of cells, the percentage of cells, the proportion of cells, or the density of cells of the population of cells that express the marker.

[0108] In some aspects, the population-level output measure is determined based on holographic information obtained for individual cells of the population of cells. In some aspects, the population-level output measure is determined based on cellular features derived from the holographic information for the individual cells. In some aspects, the populationlevel output measure is determined based on population-level statistics of the cellular features. In some aspects, the population-level statistics summarize input measures of the cellular features across individual cells of the population of cells. For example, in some embodiments, the population-level statistics include quantile values of the input measures across individual cells of the population of cells for the cellular features.

[0109] In some aspects, population-level statistics of cellular features derived from holographic information obtained for individual cells of reference populations of T cells are used to train a machine learning model to predict population-level output measures of expression of the marker. In some embodiments, the machine learning model is trained using reference population-level output measures of expression of the marker for reference populations of cells. In some embodiments, the reference population-level output measures are determined using fluorescence imaging, such as by immunoaffinity-based fluorescent labelling for the marker.

[0110] As demonstrated herein, marker expression of individual cells, such as activation marker expression, was not required or used for developing the provided methods, thereby obviating the need for any specialized equipment or reagents, e.g., for capturing matched holographic and fluorescent images of individual cells. In addition, because the cells that were holographically imaged did not have to be fluorescently labelled as part of the provided methods, the impact on cellular feature measurements and model predictive accuracy of anyrandom or systematic morphometric changes that can occur in labelled cells, compared to unlabelled cells, was avoided.

[0111] In addition, the same holographic imaging system used in accordance with the provided methods for obtaining holographic information for training the machine learning model can also be used for obtaining holographic information for prediction using the trained machine learning model. This is in contrast to other methods in which different imaging systems may be used for obtaining holographic information for model training and use, for instance a dual fluorescence-holographic imaging system for model training and a holographic imaging-only system for model use. Thus, in some aspects, the machine learning models of the provided methods are highly transferrable for use in subsequent applications of the models following model training, for instance compared to other methods involving use of different imaging systems before and after model training.

[0112] In some embodiments, holographic information is provided as input to a convolutional neural network. In some embodiments, the selection or design of cellular features derived from holographic information is not required prior to model training. Instead, in some embodiments, cellular features can be automatically extracted from holographic information as part of model training, for instance extracted by a convolutional neural network. In some embodiments, cellular features that are automatically extracted include those that may not be detectable by humans or may not be known to be predictive of expression of the marker. In some embodiments, cellular features that are automatically extracted are not detectable by humans based on visual inspection of holographic images. In some embodiments, one or more of the cellular features that are automatically extracted are not known to be predictive of expression of the marker (e.g., the activation marker). Thus, in some aspects, the provided methods are accurate, efficient, objective, and unbiased methods for predicting cell phenotype, such as activation state.

[0113] In some aspects, the methods described herein can be used to monitor and characterize T cell phenotype, such as activation state or memory status, using a nondestructive, non-damaging approach in which T cells (e.g., unlabelled T cells) can be imaged without damaging them (e.g., cells can be drawn from a bioreactor and returned to the bioreactor in an undamaged state). The provided methods can be used to monitor the dynamics of T cell phenotype, such as activation state or memory status, during cultivationwithout frequent cell sampling or arduous analytical techniques. In some aspects, the provided methods can be used to identify relationships between a cell phenotype (such as activation state or memory status) and outcomes, for instance how the cells evolve over time.

[0114] In some embodiments, the provided methods can be used to predict the quality of T cells subjected to a manufacturing process, e.g., the quality of T cells during or after the manufacturing process. In some aspects, a cell phenotype (such as activation state or memory status) as predicted by the provided methods can be used as a readout during manufacturing, e.g., of manufacturing success or of the success of a manufacturing step (e.g., of cell stimulation). In some embodiments, the provided methods can be used to monitor whether T cells are sufficiently activated, such as for expansion of the T cells to a desired threshold number, for instance to numbers needed for clinical doses of the T cells for a T cell therapy.

[0115] In some aspects, the activation of T cells can lead to the differentiation of T cells. Higher proportions of early memory T cells, such as naive-like T cells, in T cell therapies can improve patient outcomes (see, e.g., Jiang et al., Journal of Pharmaceutical Sciences (2021) 110:1871-1876). In some embodiments, the provided methods can be used to monitor the memory status of the T cells, either directly or by monitoring the activation state of the T cells. In some embodiments, the activation state of the T cells is monitored to predict the memory status of the T cells.

[0116] In some aspects, the cell phenotype information obtained by the provided methods can be used during process development to optimize the duration or other conditions of the manufacturing process or steps thereof in order to improve the quality of processed T cells. In some instances, this information can be used to develop process control strategies in which, for example, when a predicted cell phenotype, such as activation state, falls outside a determined range, conditions of one or more (e.g., the current or a subsequent) manufacturing steps can be altered, e.g., the duration of the current or subsequent manufacturing step can be altered, to improve the final quality of the T cells being manufactured. For example, when a predicted cell phenotype, such as activation state, falls outside a determined range during cultivation, subsequent cultivation can, in some instances, be performed under perfusion conditions and / or in the presence of small molecules for, e.g., modulating T cell phenotype towards desired profiles.

[0117] In some aspects, the cell phenotype information obtained by the provided methods can be used to assess or reduce batch-to-batch variability of T cells subjected to the manufacturing process. In some aspects, the cell phenotype information can be used to assess or reduce batch-to-batch variability of a drug product produced using the manufacturing process. For instance, by ensuring that T cells across different cell therapy manufacturing runs are at comparable activation states, the differentiation and memory status of the T cells can be kept consistent. This can reduce variability (e.g., patient-to-patient variability) in the resulting T cell therapies (see, e.g., Jiang et al., Journal of Pharmaceutical Sciences (2021) 110:1871-1876).

[0118] In some embodiments, the provided methods can be used to monitor the activation state of T cells prior to or following the engineering of the T cells. In some aspects, transgene expression can be higher in activated vs. non-activated T cells, such as following the viral transduction of the T cells (see, e.g., Ghassemi et al., Nature Biomedical Engineering (2022) 6:118-128). In some aspects, electroporation efficiency for engineering can be higher in activated vs. non-activated T cells (see, e.g., Zhang et al., BMC Biotechnology (2018) 18:4). In some embodiments, the T cells are monitored in accordance with the provided methods prior to engineering, for instance so that engineering can be initiated once the provided methods predict that the T cells are sufficiently activated for improved transgene expression. In some embodiments, the T cells are monitored in accordance with the provided methods following engineering, for instance to determine whether the T cells are or remain sufficiently activated following engineering to improve transgene expression.

[0119] All publications, including patent documents, scientific articles, and databases, referred to in this application are incorporated by reference in their entirety for all purposes to the same extent as if each individual publication were individually incorporated by reference. If a definition set forth herein is contrary to or otherwise inconsistent with a definition set forth in the patents, applications, published applications, and other publications that are herein incorporated by reference, the definition set forth herein prevails over the definition that is incorporated herein by reference.

[0120] The section headings used herein are for organizational purposes only and are not to be construed as limiting the subject matter described.I. METHODS FOR DETERMINING T CELL PHENOTYPE

[0121] In some embodiments, the provided methods involve determining a cell phenotype, such as activation state, of a population of cells that contains T cells. In some embodiments, the provided methods are for determining a cell phenotype, such as activation state, of a population of cells that contains T cells. Exemplary populations of cells are described in Section II. In some embodiments, the provided methods involve performing any of the cell processing steps described in Section II for the population of cells.

[0122] Provided herein, in some embodiments, is a method for determining a cell phenotype of a population of T cells, comprising determining, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by T cells with a cell phenotype of interest, wherein the population-level output measure is determined based on a plurality of population-level statistics, wherein: each population-level statistic is of one or more input measures for a cellular feature of a plurality of cellular features derived from holographic information obtained for the population of cells, the plurality of population-level statistics comprising one or more population-level statistics for each of the plurality of cellular features; and each input measure is from an individual cell of the population of cells.

[0123] Also provided herein, in some embodiments, is a method for determining a cell phenotype of a population of T cells, comprising: (a) obtaining holographic information for a population of cells comprising T cells; (b) determining one or more input measures for each of a plurality of cellular features derived from the holographic information, wherein each input measure is from an individual cell in the population of cells; (c) determining a plurality of population-level statistics, wherein each population-level statistic is of the one or more input measures for a cellular feature of the plurality of cellular features, and the plurality of population-level statistics comprises one or more population-level statistics for each of the plurality of cellular features; and (d) determining, based on the plurality of population-level statistics, a population-level output measure of expression of a marker expressed by T cells with a cell phenotype of interest for the population of cells. In some embodiments, the term “determining” means predicting.

[0124] Accordingly, in some embodiments, provided herein is a method for predicting a cell phenotype of a population of T cells, comprising: (a) obtaining holographic informationfor a population of cells comprising T cells; (b) predicting one or more input measures for each of a plurality of cellular features derived from the holographic information, wherein each input measure is from an individual cell in the population of cells; (c) predicting a plurality of population-level statistics, wherein each population-level statistic is of the one or more input measures for a cellular feature of the plurality of cellular features, and the plurality of population-level statistics comprises one or more population-level statistics for each of the plurality of cellular features; and (d) predicting, based on the plurality of population-level statistics, a population-level output measure of expression of a marker expressed by T cells with a cell phenotype of interest for the population of cells.

[0125] Also provided herein, in some embodiments, is a method for training a machine learning model that predicts a cell phenotype of a population of T cells, comprising training a machine learning model using: (i) a dataset of reference population-level statistics, wherein for each of a first plurality of reference populations of cells comprising T cells, the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of a plurality of cellular features, wherein: each reference population-level statistic is of one or more reference input measures for a cellular feature of the plurality of cellular features derived from holographic information obtained for the reference population of cells; and each reference input measure is from an individual cell of the reference population of cells; and (ii) a dataset of reference population-level output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures comprises a reference population-level output measure of expression of a marker expressed by T cells with a cell phenotype of interest for the reference population of cells, wherein: the first and second pluralities of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo', and reference populations of cells from the first plurality of reference populations of cells are from the same reference culture of cells as reference populations of cells from the second plurality of reference populations of cells; whereby the machine learning model is trained to predict population-level output measures of expression of the marker based on population-level statistics of the plurality of cellular features.

[0126] Also provided herein, in some embodiments, is a method for determining the activation state of a population of T cells, comprising determining, for a population of cellscomprising T cells, a population-level output measure of expression of a marker expressed by activated T cells, wherein the marker is CD137 (4-1BB), and the population-level output measure is determined based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean; and each input measure is from an individual cell of the population of cells.

[0127] Also provided herein, in some embodiments, is a method for determining the activation state of a population of T cells, comprising determining, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by activated T cells, wherein the marker is CD137 (4-1BB), and the population-level output measure is determined based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises intensity skewness, intensity correlation, and intensity homogeneity; and each input measure is from an individual cell of the population of cells.

[0128] In some embodiments, the term “determining” means predicting.

[0129] Accordingly, in some embodiments, provided herein is a method for predicting the activation state of a population of T cells, comprising predicting, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by activated T cells, wherein the marker is CD137 (4-1BB), and the population-level output measure is predicted based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean; and each input measure is from an individual cell of the population of cells.

[0130] Also provided herein, in some embodiments, is a method for predicting the activation state of a population of T cells, comprising predicting, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by activated T cells, wherein the marker is CD137 (4-1BB), and the population-level output measure is predicted based on one or more input measures for each of a plurality of cellularfeatures derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises intensity skewness, intensity correlation, and intensity homogeneity; and each input measure is from an individual cell of the population of cells.

[0131] Also provided herein, in some embodiments, is a method for determining the memory phenotype of a population of T cells, comprising determining, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by central memory T cells or stem cell memory T cells, wherein the marker is CCR7, and the population-level output measure is determined based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises one or more of cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter; and each input measure is from an individual cell of the population of cells. In some embodiments, the term “determining” means predicting.

[0132] Accordingly, provided herein, in some embodiments, is a method for predicting the memory phenotype of a population of T cells, comprising predicting, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by central memory T cells or stem cell memory T cells, wherein the marker is CCR7, and the population-level output measure is predicted based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises one or more of cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter; and each input measure is from an individual cell of the population of cells.

[0133] Also provided herein, in some embodiments, is a method for training a machine learning model that predicts a cell phenotype of a population of T cells, comprising: (a) training a convolutional neural network using a dataset of reference holographic information, wherein for each of a first plurality of reference populations of cells comprising T cells, the dataset of reference holographic information comprises holographic information obtained for individual cells of the reference population of cells; (b) determining, from the convolutional neural network, one or more reference input measures for each cellular feature of a plurality of cellular features derived from the holographic information, wherein the plurality of cellularfeatures are cellular features extracted by the convolutional neural network, and each reference input measure is from an individual cell of a reference population of cells; (c) determining a dataset of reference population-level statistics, wherein the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of the plurality of cellular features, each reference population-level statistic determined by applying a distribution-based pooling filter to the one or more reference input measures for a cellular feature of the plurality of cellular features; and (d) training a machine learning model using the dataset of reference population-level statistics and a dataset of reference populationlevel output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures comprises a reference population-level output measure of expression of a marker expressed by T cells with a cell phenotype of interest for the reference population of cells, wherein: the first and second pluralities of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo', and reference populations of cells from the first plurality of reference populations of cells are from the same reference culture of cells as reference populations of cells from the second plurality of reference populations of cells; whereby the machine learning model is trained to predict population-level output measures of expression of the marker based on population-level statistics of the plurality of cellular features.

[0134] Also provided herein, in some embodiments is a method for determining recombinant receptor expression of a population of T cells, comprising determining, for a population of cells comprising T cells, expression of a recombinant receptor introduced into T cells of the population of cells, wherein the determining is based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height; and each input measure is from an individual cell of the population of cells. In some embodiments, the term “determining” means predicting.

[0135] Accordingly, provided herein, in some embodiments, is a method for predicting recombinant receptor expression of a population of T cells, comprising predicting, for a population of cells comprising T cells, expression of a recombinant receptor introduced into T cells of the population of cells, wherein the determining is based on one or more inputmeasures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height; and each input measure is from an individual cell of the population of cells.

[0136] In some embodiments, the cell phenotype is expression of a marker by cells of the population of cells. In some embodiments, the marker is expressed on the surface of T cells. In some embodiments, the provided methods involve determining the presence or absence of marker-expressing cells in the population of cells, the degree to which marker-expressing cells are present in the population of cells, or the degree to which marker-expressing cells are present in T cells of the population of cells. In some embodiments, the presence or absence of marker-expressing cells in the population of cells is determined. In some embodiments, the degree to which marker-expressing cells are present in the population of cells is determined. In some embodiments, the degree to which marker-expressing cells are present in T cells of the population of cells is determined.

[0137] In some embodiments, the cell phenotype is based on expression of one or more of a combination of markers by cells of the population of cells. In some embodiments, the cell phenotype is based on a population-leve output measure of expression of one or more of a combination of markers by cells of the population of cells. In some embodiments, the markers are expressed on the surface of T cells. In some embodiments, the cell phenotype is expression of some markers of the combination of markers and non-expression of the remaining markers of the combination of markers. In some embodiments, the cell phenotype is a population-level output measure of expression of some markers of the combination of markers and non-expression of the remaining markers of the combination of markers. In some embodiments, the cell phenotype is expression of each of the combination of markers. In some embodiments, the cell phenotype is a population-leve output measure of expression of each of the combination of markers.

[0138] In some embodiments, reference herein to “marker” also refers to a combination of markers, such as any combination of the markers described herein. In some embodiments, reference herein to “expression of the marker”, “marker-expressing cells”, “cells expressing the marker”, or similar also refers to combination marker expression and / or non-expressionconsistent with the cell phenotype of interest. For instance, for the cell phenotype CCR7+CD45RA-, “marker-expressing cells” can refer to CCR7+CD45RA- cells. Likewise, for the cell phenotype CCR7+CD27+, “marker-expressing cells” can refer to CCR7+CD27+ cells.

[0139] In some embodiments, the provided methods involve determining the activation state of the population of cells. In some embodiments, “activation state” refers to the presence or absence of activated T cells in the population of cells, the degree to which activated T cells are present in the population of cells, or the degree to which activated T cells are present in T cells of the population of cells. In some embodiments, the presence or absence of activated T cells in the population of cells is determined. In some embodiments, the degree to which activated T cells are present in the population of cells is determined. In some embodiments, the degree to which activated T cells are present in T cells of the population of cells is determined. In some embodiments, such determinations are predictions.

[0140] In some embodiments, the provided methods involve determining the memory status of the population of cells. In some embodiments, “memory status” refers to the presence or absence of T cells of a particular memory phenotype in the population of cells, the degree to which T cells of a particular memory phenotype are present in the population of cells, or the degree to which T cells of a particular memory phenotype are present in T cells of the population of cells. In some embodiments, the presence or absence of T cells of a particular memory phenotype in the population of cells is determined. In some embodiments, the degree to which T cells of a particular memory phenotype are present in the population of cells is determined. In some embodiments, the degree to which T cells of a particular memory phenotype are present in T cells of the population of cells is determined.

[0141] In some embodiments, the memory phenotype is that of naive-like T cells. In some embodiments, the presence or absence of naive-like T cells in the population of cells is determined. In some embodiments, the degree to which naive-like T cells are present in the population of cells is determined. In some embodiments, the degree to which naive-like T cells are present in T cells of the population of cells is determined. In some embodiments, the memory phenotype is a central memory phenotype or a stem cell memory phenotype. In some embodiments, the presence or absence of T cells having a central memory phenotype or a stem cell memory phenotype in the population of cells is determined. In someembodiments, the degree to which T cells having a central memory phenotype or a stem cell memory phenotype are present in the population of cells is determined. In some embodiments, the degree to which T cells having a central memory phenotype or a stem cell memory phenotype are present in T cells of the population of cells is determined. In some embodiments, the determination is a prediction.

[0142] In some embodiments, the memory phenotype is that of central memory T cells. In some embodiments, the presence or absence of central memory T cells in the population of cells is determined. In some embodiments, the degree to which central memory T cells are present in the population of cells is determined. In some embodiments, the degree to which central memory T cells are present in T cells of the population of cells is determined.

[0143] In some embodiments, the memory phenotype is that of effector memory T cells. In some embodiments, the presence or absence of effector memory T cells in the population of cells is determined. In some embodiments, the degree to which effector memory T cells are present in the population of cells is determined. In some embodiments, the degree to which effector memory T cells are present in T cells of the population of cells is determined.

[0144] In some embodiments, the memory phenotype is that of terminally differentiated T cells. In some embodiments, the presence or absence of terminally differentiated T cells in the population of cells is determined. In some embodiments, the degree to which terminally differentiated T cells are present in the population of cells is determined. In some embodiments, the degree to which terminally differentiated T cells are present in T cells of the population of cells is determined.In some embodiments, the provided methods involve determining, for the population of cells, a population-level output measure indicative of the cell phenotype, e.g., activation state, of the population of cells. In some embodiments, the population-level output measure indicates the presence or absence of marker-expressing cells in the population of cells. In some embodiments, the population-level output measure indicates the degree to which marker-expressing cells are present in the population of cells. In some embodiments, the population-level output measure indicates the degree to which marker-expressing cells are present in T cells of the population of cells.

[0145] In some embodiments, the population-level output measure is the number of marker-expressing cells, the percentage of marker-expressing cells, the proportion of markerexpressing cells, or the density of marker-expressing cells in the population of cells. In someembodiments, the population-level output measure is the number of marker-expressing cells, the percentage of marker-expressing cells, the proportion of marker-expressing cells, or the density of marker-expressing cells in T cells of the population of cells. In some embodiments, the population-level output measure is the percentage of marker-expressing cells in T cells of the population of cells.

[0146] In some embodiments, the marker is expressed on the surface of T cells.

[0147] In some embodiments, the marker is CD3. In some embodiments, the marker is CD4. In some embodiments, the marker is CD8.

[0148] In some embodiments, the marker is a recombinant protein. In some embodiments, the recombinant protein is a recombinant receptor. In some embodiments, the recombinant receptor is a chimeric antigen receptor (CAR). Exemplary CARs are described in Section II- C-2.

[0149] In some embodiments, the recombinant receptor is a T cell receptor (TCR). Exemplary TCRs are described in Section II-C-3.

[0150] In some embodiments, the provided methods involve determining, for the population of cells, a population-level output measure indicative of the activation state of the population of cells. In some embodiments, the population-level output measure indicates the presence or absence of activated T cells in the population of cells. In some embodiments, the population-level output measure indicates the degree to which activated T cells are present in the population of cells. In some embodiments, the population-level output measure indicates the degree to which activated T cells are present in T cells of the population of cells.

[0151] In some embodiments, the population-level output measure is the number of activated T cells, the percentage of activated T cells, the proportion of activated T cells, or the density of activated T cells in the population of cells. In some embodiments, the population-level output measure is the number of activated T cells, the percentage of activated T cells, the proportion of activated T cells, or the density of activated T cells in T cells of the population of cells. In some embodiments, the population-level output measure is the percentage of activated T cells in T cells of the population of cells.

[0152] In some embodiments, the provided methods involve determining, for the population of cells, a population-level output measure of expression of a marker indicative of T cell activation. In some embodiments, “activation state” refers to the presence of T cellsexpressing the marker in the population of cells, the degree to which T cells expressing the marker are present in the population of cells, or the degree to which T cells expressing the marker are present in T cells of the population of cells.

[0153] In some embodiments, the marker is a marker that is expressed by activated T cells (an “activation marker”). In some embodiments, “activated T cell” refers to a T cell expressing such a marker. In some embodiments, the marker is CD137 (4-1BB). In some embodiments, the marker is CD25. In some embodiments, the marker is CD69. In some embodiments, the marker is a combination of two more activation markers, e.g., two or more of CD137, CD25, and CD69. In some embodiments, the marker comprises CD137. In some embodiments, the marker comprises CD 137 and one or more further activation markers. In some embodiments, the one or more further activation markers comprise CD25 and / or CD69.

[0154] In other embodiments, the marker is a marker that is expressed by non-activated T cells. In some embodiments, “activated T cell” refers to a T cell not expressing such a marker.

[0155] In some embodiments, the provided methods involve determining, for the population of cells, a population-level output measure indicative of the memory status of the population of cells. In some embodiments, the population-level output measure indicates the presence or absence of T cells of a particular memory phenotype in the population of cells. In some embodiments, the population-level output measure indicates the degree to which T cells of a particular memory phenotype are present in the population of cells. In some embodiments, the population-level output measure indicates the degree to which T cells of a particular memory phenotype are present in T cells of the population of cells. Early memory phenotypes, such as a stem cell memory or a central memory phenotype, are reported to result in sustained in vivo response of CAR T cell therapies given their proliferative capacity and effector capabilities in both hematological and solid tumor environments. See Gargett et al., Cytotherapy, 21: 593-602 (2019). CCR7 positive cells are typically considered to be of an earlier memory phenotype.

[0156] In some embodiments, the population-level output measure is the number of T cells of a particular memory phenotype, the percentage of T cells of a particular memory phenotype, the proportion of T cells of a particular memory phenotype, or the density of T cells of a particular memory phenotype in the population of cells. In some embodiments, thepopulation-level output measure is the number of T cells of a particular memory phenotype, the percentage of T cells of a particular memory phenotype, the proportion of T cells of a particular memory phenotype, or the density of T cells of a particular memory phenotype in T cells of the population of cells. In some embodiments, the population-level output measure is the percentage of T cells of a particular memory phenotype in T cells of the population of cells.

[0157] In some embodiments, the memory phenotype is that of naive-like T cells. In some embodiments, the population-level output measure is the number of naive-like T cells, the percentage of naive-like T cells, the proportion of naive-like T cells, or the density of naive- like T cells in the population of cells. In some embodiments, the population-level output measure is the number of naive-like T cells, the percentage of naive-like T cells, the proportion of naive-like T cells, or the density of naive-like T cells in T cells of the population of cells. In some embodiments, the population-level output measure is the percentage of naive-like T cells in T cells of the population of cells.

[0158] In some embodiments, the memory phenotype is that of central memory phenotype or stem cell memory phenotype.

[0159] In some embodiments, the memory phenotype is that of central memory T cells. In some embodiments, the population-level output measure is the number of central memory T cells, the percentage of central memory T cells, the proportion of central memory T cells, or the density of central memory T cells in the population of cells. In some embodiments, the population-level output measure is the number of central memory T cells, the percentage of central memory T cells, the proportion of central memory T cells, or the density of central memory T cells in T cells of the population of cells. In some embodiments, the populationlevel output measure is the percentage of central memory T cells in T cells of the population of cells.

[0160] In some embodiments, the memory phenotype is that of stem cell memory T cells. In some embodiments, the population-level output measure is the number of stem cell memory T cells, the percentage of stem cell memory T cells, the proportion of stem cell memory T cells, or the density of stem cell memory T cells in the population of cells. In some embodiments, the population-level output measure is the number of stem cell memory T cells, the percentage of stem cell memory T cells, the proportion of stem cell memory T cells,or the density of stem cell memory T cells in T cells of the population of cells. In some embodiments, the population-level output measure is the percentage of stem cell memory T cells in T cells of the population of cells.

[0161] In some embodiments, the memory phenotype is that of effector memory T cells. In some embodiments, the population-level output measure is the number of effector memory T cells, the percentage of effector memory T cells, the proportion of effector memory T cells, or the density of effector memory T cells in the population of cells. In some embodiments, the population-level output measure is the number of effector memory T cells, the percentage of effector memory T cells, the proportion of effector memory T cells, or the density of effector memory T cells in T cells of the population of cells. In some embodiments, the populationlevel output measure is the percentage of effector memory T cells in T cells of the population of cells.

[0162] In some embodiments, the memory phenotype is that of terminally differentiated T cells. In some embodiments, the population-level output measure is the number of terminally differentiated T cells, the percentage of terminally differentiated T cells, the proportion of terminally differentiated T cells, or the density of terminally differentiated T cells in the population of cells. In some embodiments, the population-level output measure is the number of terminally differentiated T cells, the percentage of terminally differentiated T cells, the proportion of terminally differentiated T cells, or the density of terminally differentiated T cells in T cells of the population of cells. In some embodiments, the population-level output measure is the percentage of terminally differentiated T cells in T cells of the population of cells.

[0163] In some embodiments, the provided methods involve determining, for the population of cells, a population-level output measure of expression of a marker indicative of T cell memory status. In some embodiments, “memory status” refers to the presence of T cells expressing the marker in the population of cells, the degree to which T cells expressing the marker are present in the population of cells, or the degree to which T cells expressing the marker are present in T cells of the population of cells.

[0164] In some embodiments, the marker is a marker that is expressed by naive-like T cells (a “naive-like marker”). In some embodiments, “naive-like T cell” refers to a T cellexpressing such a marker. In some embodiments, the marker is CCR7. In some embedments, the marker is CD27. In some embodiments, the marker is CD45RA.

[0165] In other embodiments, the marker is a marker that is expressed by non-naive-like T cells. In some embodiments, “naive-like T cell” refers to a T cell not expressing such a marker.

[0166] In some embodiments, the combination of markers is any combination of the markers described herein.

[0167] In some embodiments, the combination of markers is a combination of markers indicative of the memory phenotype of T cells. In some embodiments, the combination of markers is a combination of markers that is expressed by T cells of a particular memory phenotype.

[0168] In some embodiments, the memory phenotype is of an earlier memory phenotype. In some embodiments, the earlier memory phenotype is a central memory phenotype or a stem cell memory phenotype. In some embodiments, the earlier memory phenotype is a central memory phenotype. In some embodiments, the earlier memory phenotype is a stem cell memory phenotype. In some embodiments, the marker that is expressed by the earlier memory phenotype, e.g., the central memory phenotype or the stem cell memory phenotype, is CCR7.

[0169] In some embodiments, the memory phenotype is that of naive-like T cells. In some embodiments, the naive-like T cells are CCR7+CD27+. In some embodiments, the naive-like T cells are CCR7+CD45RA+. In some embodiments, the naive-like T cells are CD27+CD45RA+.

[0170] In some embodiments, the memory phenotype is that of central memory T cells. In some embodiments, the central memory T cells are CCR7+CD45RA-.

[0171] In some embodiments, the memory phenotype is that of effector memory T cells. In some embodiments, the effector memory T cells are CCR7-CD45RA-.

[0172] In some embodiments, the memory phenotype is that of terminally differentiated T cells. In some embodiments, the effector memory T cells are CCR7-CD45RA+.

[0173] In some embodiments, the population-level output measure indicates the presence of T cells expressing the marker in the population of cells. In some embodiments, the population-level output measure indicates the degree to which T cells expressing the markerare present in the population of cells. In some embodiments, the population-level output measure indicates the degree to which T cells expressing the marker are present in T cells of the population of cells.

[0174] In some embodiments, the population-level output measure is the number of cells, the percentage of cells, the proportion of cells, or the density of cells in the population of cells that express the marker. In some embodiments, the population-level output measure is the number of cells, the percentage of cells, the proportion of cells, or the density of cells in T cells of the population of cells that express the marker. In some embodiments, the populationlevel output measure is the percentage of cells in T cells of the population of cells that express the marker.

[0175] In some embodiments, the cell phenotype, e.g., activation state, is determined based on holographic information obtained for the population of cells, e.g., for individual cells of the population of cells. In some embodiments, the population-level output measure is determined based on holographic information obtained for the population of cells, e.g., for individual cells of the population of cells. In some embodiments, the provided methods involve obtaining holographic information for the population of cells, e.g., for individual cells of the population of cells. In some embodiments, the holographic information includes holographic information for one or more individual cells of the population of cells. In some embodiments, the holographic information includes holographic information for each of a plurality of individual cells of the population of cells. Exemplary holographic information and methods for obtaining same are described in Section I-A.

[0176] In some embodiments, the cell phenotype, e.g., activation state, is determined also based on non-holographic information obtained for the population of cells. In some embodiments, the population-level output measure is determined also based on nonholographic information obtained for the population of cells. In some embodiments, the nonholographic information includes non-holographic information for one or more individual cells of the population of cells. In some embodiments, the non-holographic information includes non-holographic information for each of a plurality of individual cells of the population of cells.

[0177] In some embodiments, the cell phenotype, e.g., activation state, is determined based on one or more input measures for each of one or more cellular features. In someembodiments, the population-level output measure is determined based on one or more input measures for each of one or more cellular features. In some embodiments, “cellular feature” refers to a characteristic of an individual cell. In some embodiments, the one or more cellular features are derived from holographic information. In some embodiments, each cellular feature is derived from holographic information obtained for an individual cell.

[0178] In some embodiments, “input measure” refers to a value that is determined, e.g., measured or calculated, for a feature, e.g., a cellular feature. In some embodiments, the provided methods involve determining one or more input measures for each of one or more cellular features. In some embodiments, each input measure is determined from holographic information obtained for the population of cells, e.g., for individual cells of the population of cells. In some embodiments, each input measure is from an individual cell of the population of cells. In some embodiments, each input measure is determined from holographic information obtained for an individual cell of the population of cells. In some embodiments, each input measure is measured from the holographic information. In some embodiments, each input measure is calculated using the holographic information. Exemplary cellular features and methods for determining input measures for same are described in Section I-B.

[0179] In some embodiments, the cell phenotype, e.g., activation state, is determined also based on one or more other input measures for each of one or more other features derived from non-holographic information. In some embodiments, the population-level output measure is determined also based on one or more other input measures for each of one or more other features derived from non-holographic information. In some embodiments, each of the other features is a characteristic of an individual cell. In some embodiments, each of the other features is derived from non-holographic information obtained for an individual cell.

[0180] In some embodiments, the cell phenotype, e.g., activation state, is determined based on at least one population-level statistic. In some embodiments, the population-level output measure is determined based on at least one population-level statistic. In some embodiments, each population-level statistic is of one or more input measures for a cellular feature of the one or more cellular features. In some embodiments, each population-level statistic describes the one or more input measures. In some embodiments, each population-level statistic describes the distribution of the one or more input measures. In some embodiments, theprovided methods involve determining each population-level statistic from the one or more input measures. In some embodiments, each population-level statistic is calculated using the one or more input measures. Exemplary population-level statistics are described in Section I- C. Exemplary methods for determining the population-level output measure based on the at least one population-level statistic are described in Section I-D.

[0181] In some embodiments, the cell phenotype, e.g., activation state, is determined based on one or more other population-level statistics. In some embodiments, the population-level output measure is determined based on one or more other population-level statistics. In some embodiments, the one or more other population-level statistics are of one or more other input measures for each of the one or more other features derived from non-holographic information.

[0182] In some embodiments, any number of the steps of the provided methods are performed in a closed system. In some embodiments, any number of the steps of the provided method are automated.A. Holographic Information

[0183] In some embodiments, the cell phenotype, e.g., activation state, of the population of cells is determined based on holographic information obtained for the population of cells, e.g., for individual cells of the population of cells. In some embodiments, the population-level output measure is determined based on holographic information obtained for the population of cells, e.g., for individual cells of the population of cells. In some embodiments, the holographic information obtained for the population of cells includes holographic information for each of a plurality of individual cells of the population of cells. In some embodiments, the holographic information for each of the plurality of individual cells is obtained simultaneously or near simultaneously.

[0184] In some embodiments, the holographic information is obtained using methods that do not involve recording the projected image of the population of cells. In some embodiments, the holographic information includes a hologram obtained of the population of cells. In some embodiments, the holographic information is obtained using a light wavefront of interest and a reference wavefront. In some embodiments, the holographic information includes phase information and intensity information obtained for the population of cells, for instance as described in Alm et al. (2013), “Cells and Holograms - Holograms and DigitalHolographic Microscopy as a Tool to Study the Morphology of Living Cells” in Holography: Basic Principles and Contemporary Applications. In some embodiments, the phase information and intensity information are obtained using the light wavefront of interest and the reference wavefront.

[0185] In some embodiments, an image of the population of cells can be reconstructed from the holographic information. In some embodiments, an image of the population of cells can be reconstructed from the hologram. In some embodiments, an image of the population of cells can be reconstructed from the phase information and the intensity information. In some embodiments, the reconstruction is performed using a computer and numerical algorithms.

[0186] In some embodiments, the provided methods involve obtaining the holographic information for the population of cells, e.g., for individual cells of the population of cells. In some embodiments, the holographic information is obtained by microscopy. In some embodiments, the holographic information is obtained by imaging the population of cells, e.g., individual cells of the population of cells, using microscopy. In some embodiments, the holographic information for individual cells of the population of cells is obtained by segmenting holographic information obtained for the population of cells.

[0187] In some embodiments, the holographic information is obtained by iterative imaging of cells, e.g., individual cells or samples of cells drawn from a bioreactor that holds a population of cells (e.g., a population of cells in suspension).

[0188] In some embodiments, “holographic information” refers to a 2D image of a population of cells or of an individual cell that is obtained using holographic imaging. In some embodiments, the holographic information includes multiple 2D images of the population of cells or the individual cell that are obtained using holographic imaging. In some embodiments, the holographic information includes a 2D phase image of the population of cells or the individual cell. In some embodiments, the holographic information includes a 2D intensity image of the population of cells or the individual cell. In some embodiments, the holographic information includes a 2D phase image and a 2D intensity image of the population of cells or the individual cell.

[0189] In some embodiments, holographic information itself can be used to train machine learning models, e.g., convolutional neural networks, and holographic information can be provided as input to a machine learning model, e.g., convolutional neural network, inaccordance with any of the provided methods. In other embodiments, cellular features derived from holographic information, such as any described in Section I-B, can be used to train machine learning models and provided as input to a machine learning model in accordance with any of the provided methods.

[0190] The imaging technique may use a digital device to acquire and, for example, record the output (e.g., results) of the imaging process. In some embodiments, the digital device is a charge-coupled device (e.g., a CCD camera). In some embodiments, the digital device is a complementary metal-oxide semiconductor device (e.g., a CMOS camera). Thus, in some embodiments, image data is obtained using a digital device. In some embodiments, the digital device interfaces with a computer to store and / or analyze the results (e.g., output) of the imaging process.

[0191] To obtain holographic information associated with T cells in a non-damaging, nondestructive manner, T cells may be contained in a liquid, such as a culture media. In some embodiments, the T cells may be suspended in a liquid, e.g., a culture media, for imaging. Therefore, in some embodiments, the imaging technique is capable of imaging T cells contained and / or suspended in a liquid.

[0192] In some embodiments, the holographic information is obtained by interferometric microscopy, optical coherence tomography, diffraction phase microscopy, or digital holographic microscopy (DHM). In some embodiments, the holographic information is obtained by imaging the population of cells using interferometric microscopy, optical coherence tomography, diffraction phase microscopy, or DHM.

[0193] In some embodiments, the holographic information is obtained by DHM (see, e.g., Carl et al., Applied Optics (2004) 43(36):6536-6544; and Marquet et al., Optics Letters (2005) 30(5):468-470). In some embodiments, the holographic information is obtained by imaging the population of cells, e.g., individual cells of the population of cells, using DHM. In some embodiments, the DHM is traditional DHM, in-line DHM, or differential DHM (DDHM).

[0194] DHM is a technique which can allow for recording of a 3D sample or object without the need to scan the sample layer-by-layer. In this respect, DHM can be a superior technique to confocal microscopy. In DHM, holographic information can be recorded by a digital camera such as a CCD- or a CMOS-camera, which can subsequently be stored or processedon a computer. Different DHM techniques, including DDHM, as well as microscope configurations and elements are described in US-7362449, US-9684281-B2, US- 10578541- B2, US-2015248109-A1, US-9846151-B2, US-2014193850-A1, US-2016184817-A1, US- 11067379-B2, US-2014195568-A1, and US-2021142472-Al.

[0195] To make a holographic representation, or hologram, traditionally a highly coherent light source such as laser-light can be used to illuminate a sample. The light from the source can be split into two beams, an object beam and a reference beam. The object beam can be sent via an optical system to the sample to interact with it, thereby altering the phase and amplitude of the light depending on the object’s optical properties and 3D shape. The object beam which has been reflected on or transmitted through the sample can then be made (e.g., by set of mirrors and / or beam splitters) to interfere with the reference beam to result in an interference pattern that can be digitally recorded. Since the hologram can be more accurate when object beam and reference beam have comparable amplitude, an absorptive element can be introduced in the reference beam, which can decrease its amplitude to the level of the object beam without altering the phase of the reference beam, or at most changing the phase globally. The recorded interference pattern can contain information on the phase and amplitude changes which depend on the object's optical properties and 3D shape.

[0196] An alternative way of making a hologram is by using an in-line holographic technique. In some aspects, in-line DHM is similar to traditional DHM, but does not split the beam, at least not by a beam splitter or other external optical element. In-line DHM can be used to look at a solution of particles, e.g. cells, in a fluid such that, for example, some part of the at least partially coherent light will pass through the sample without interacting with the particles (reference beam) and interfere with light that has interacted with the particles (object beam) to give rise to an interference pattern which can be recorded digitally and processed. In-line DHM can be used in transmission mode.

[0197] Another DHM technique is DDHM. In some aspects, DDHM is different to the other techniques in its use of reference and object beams. In a preferred set-up of DDHM, the sample can be illuminated by illumination means which include at least partially coherent light in reflection or in transmission mode. The reflected or transmitted sample beam can be sent through an objective lens and subsequently split in two by a beam splitter and sent along different paths in a differential interferometer, e.g., of the Michelson or Mach-Zehnder type.In one of the paths, a beam-bending element or tilting means can be inserted, e.g., a transparent wedge. The two beams can then be made to interfere with each other in the focal plane of a focusing lens, and the interference pattern in this focal plane can be recorded digitally and stored by, e.g., a CCD-camera connected to a computer. Due to the beambending element, the two beams can be slightly shifted in a controlled way, and the interference pattern can depend on the amount of shifting. Then, the beam-bending element can be turned, thereby altering the amount of shifting. The new interference pattern can also be recorded. This can be done a number of times (N), and from these N interference patterns, the gradient (or spatial derivative) of the phase in the focal plane of the focusing lens can be approximately computed. This is called the phase- stepping method, but other methods of obtaining the phase gradient are also known, such as a Fourier transform data processing technique. The gradient of the phase can be integrated to give the phase as a function of position. The amplitude of the light as a function of position can be computed from the possibly but not necessarily weighted average of the amplitudes of the N recorded interference patterns. Since phase and amplitude are thus known, the same information can be obtained as in a direct holographic method (using a reference and an object beam), and a subsequent 3D reconstruction of the object can be performed.

[0198] In DDHM, an illumination means can include spatially and temporally partially coherent light. In some aspects, this is in contrast with other DHM methods that might use highly correlated laser light. Spatially and temporally partially coherent light can be produced by, e.g., a LED. A LED can be cheaper than a laser and produce light with a spectrum centered around a known wavelength, which can be spatially and temporally partially coherent, e.g., not as coherent as laser light, but still coherent enough to produce holographic images of sufficient quality for the applications at hand. In some aspects, LEDs can also have the advantage of being available for many different wavelengths and can be very small in size and easy to use or replace if necessary. Therefore, in some aspects, methods that use spatially and temporally partially coherent light for obtaining holographic images can lead to more cost-effective devices for implementing such methods. In some embodiments, the illumination means is a red LED.

[0199] The holographic images may undergo object segmentation and further analysis to obtain a plurality of features that quantitatively describe the imaged objects (e.g., T cells,cellular debris). As such, various features, such as cellular features described in Section I-B, may be directly assessed or calculated from DDHM using, for example, steps of image acquisition, image processing, image segmentation, and feature extraction. In some embodiments, a digital recording device is used to record holographic images. In some embodiments, a computer including algorithms for analyzing holographic images may be used. In some embodiments, a monitor and / or computer may be used for displaying the results of the holographic image analysis. In some embodiments, the analysis is automated (e.g., capable of being performed in the absence of user input)

[0200] Any type of DHM can be used in accordance with the provided methods. In some embodiments, the DHM is traditional DHM. In some embodiments, the DHM is in-line DHM. In some embodiments, the DHM is differential DHM.

[0201] Exemplary DHM systems for use in the provided methods, for instance the Ovizio iLine F (Ovizio Imaging Systems NV / SA, Brussels, Belgium), include those that can be used in conjunction with a bioreactor for the incubation of T cells. In some embodiments, the population of cells for which holographic information is obtained is from a culture of cells being cultured in vitro or ex vivo, for instance as described in Section II-D. In some embodiments, the in vitro or ex vivo culture is in a bioreactor. In some embodiments, the population of cells is moved from the bioreactor for imaging. In some embodiments, the imaging is performed using a digital holographic microscope that is connected to the bioreactor. The microscope can be any that is capable of obtaining phase information of a fluid sample and having illumination means. The microscope can be any that is used for DHM, including any for traditional DHM, in-line DHM, or DDHM.

[0202] In some embodiments, one or more fluidic systems capable of guiding fluid from the bioreactor to the microscope are connected to the bioreactor and microscope. In some embodiments, at least one fluidic system contains one or more tubes which may come in direct contact with fluid from the bioreactor. In some embodiments, at least one tube contains a part which is at least partially transparent for the illumination means of the microscope for obtaining holographic information of said fluid sample. In some embodiments, the tube contains a part which is at least partially transparent for the illumination means of the microscope and which contains a flow cell and / or a microfluidic system. In some embodiments, the flow cell and / or microfluidic system contains a cross section in which theheight and / or width varies along the cross section. In some aspects, this allows for obtaining clear holographic images for a variety of concentrations of objects suspended in the fluid. A high concentration of suspended objects could lead to a large number of objects being stacked on top of one another and could lead to difficulties in obtaining a holographic image, especially if the microscope works in transmission mode. A low concentration could result in the microscope obtaining holographic images of the fluid medium only and not of an object suspended in that medium. If the concentration is high, a holographic image can be obtained at the position where the height or width is small, thereby ensuring that not too many objects are stacked in the illumination beam. If the concentration is small, a holographic image can be obtained at the position where the height or width is large, thereby ensuring that at least one suspended object is in the illumination beam. In some embodiments, the microfluidic system contains a branching of the tube into multiple tubes of different cross sections, diameters, heights, and / or widths. Such an arrangement can allow for obtaining clear holographic images for a variety of concentrations of objects suspended in the fluid. In some embodiments, the cross section, diameter, height, and / or width of the flow cell and / or microfluidic system is chosen as a function of the size of the suspended objects and / or the size of the illumination beam of the microscope. In some embodiments, the narrowest dimension in a cross section of the flow cell and / or microfluidic system is larger than 10 micrometer, more preferably larger than 30 micrometer, even more preferably larger than 50 micrometer, and / or the largest dimension in a cross section of the flow cell and / or microfluidic system is smaller than 5000 micrometer, more preferably smaller than 3000 micrometer, even more preferably smaller than 2500 micrometer. In some embodiments, the microfluidic system is attached on a substrate, for instance to ease manufacturing and / or provide stability to the microfluidic system.

[0203] Although it is not necessary, it may be desired that a fluid flow is present in at least one of the fluidic systems. This can allow for sampling of the contents of the bioreactor in time and monitoring of different samples to obtain a better knowledge of the state and / or reactions of the bioreactor. A fluid flow may be present due to natural phenomenon such as convection, conduction, or radiation, by density or pressure differences induced by, e.g., the reactions taking place in the bioreactor or heat gradients, by gravity, etc. If a fluid flow is desired, but is not occurring spontaneously, or if the flow needs to be controlled, one or morepumping systems may be connected to the fluidic systems in order to induce a flow in said systems. Therefore, in some embodiments, at least one pumping system is connected to one or more fluidic systems and is capable of inducing a fluid flow in said fluidic systems.

[0204] In some embodiments, at least one fluidic system contains a fluid-tight flexible part which, when compressed, pulled, and / or pushed, results in a fluid flow in the fluidic system. As such, a fluid flow can be induced in the fluidic system without a high risk of leaks and without contamination of the actuator of the flow. In some embodiments, a pumping system is connected to the fluidic system and is capable of pulling and / or pushing the fluid-tight flexible part to induce a fluid flow in the fluidic system.

[0205] In some aspects, the contents of the bioreactor can be non-destructively monitored. Thereby, it is possible to re-introduce the populations of cells which are observed in the microscope to the bioreactor. Therefore, in some embodiments, at least one fluidic system forms a closed circuit between the bioreactor and the microscope and back to the bioreactor, e.g., the fluidic system is capable of guiding fluid from the bioreactor to the microscope and back.B. Cellular Features

[0206] In some embodiments, the cell phenotype, e.g., activation state, of the population of cells is determined based on one or more input measures for each of one or more cellular features. In some embodiments, the population-level output measure is determined based on one or more input measures for each of one or more cellular features. In some embodiments, the one or more cellular features are derived from holographic information. In some embodiments, each input measure is determined from holographic information obtained for an individual cell of the population of cells. In some embodiments, determining the one or more input measures for each of the one or more cellular features involves one or more steps selected from holographic information acquisition, holographic information processing, holographic information segmentation, and cellular feature extraction. In some embodiments, image segmentation is used to determine holographic information associated with an individual cell. The process of segmentation is known in the art, for instance, from watershed treatment, clustering-based image threshold and using neural networks. Various algorithms exist for watershed treatment, including Meyer's flooding algorithm and optimal spanningforest algorithms (watershed cuts). Cluster-based image thresholding may employ Otsu’s method. In some embodiments, the holographic information includes the segmented images.

[0207] In some embodiments, the one or more input measures are a plurality of input measures. In some embodiments, the activation state is determined based on a plurality of input measures for each of one or more cellular features. In some embodiments, the population-level output measure is determined based on a plurality of input measures for each of one or more cellular features.

[0208] In some embodiments, the one or more cellular features are a plurality of cellular features. In some embodiments, the activation state is determined based on one or more input measures for each of a plurality of cellular features. In some embodiments, the populationlevel output measure is determined based on one or more input measures for each of a plurality of cellular features.

[0209] In some embodiments, the activation state is determined based on a plurality of input measures for each of a plurality of cellular features. In some embodiments, the population-level output measure is determined based on a plurality of input measures for each of a plurality of cellular features.

[0210] In some embodiments, input measures for some or all of the one or more cellular features are obtained for an individual cell. In some embodiments, input measures for each of the one or more cellular features are obtained for an individual cell.

[0211] In some embodiments, the one or more input measures for a cellular feature of the one or more cellular features include input measures from one or more individual cells of the population of cells. In some embodiments, the one or more input measures for a cellular feature of the one or more cellular features include input measures from a plurality of individual cells of the population of cells. In some embodiments, input measures for a first cellular feature are for a first plurality of individual cells, and input measures for a second cellular feature are for a second, distinct plurality of individual cells that can contain some or none of the first plurality of individual cells. In some embodiments, input measures for a first cellular feature are for a first plurality of individual cells, and input measures for a second cellular feature are also for the first plurality of individual cells. In some embodiments, the input measures for all of the cellular features are from the same plurality of individual cells.

[0212] In some embodiments, the one or more input measures for a cellular feature of the one or more cellular features include input measures from some or all individual cells of the population of cells. In some embodiments, the one or more input measures for a cellular feature of the one or more cellular features include input measures from live cells only. In some embodiments, the live cells are classified as live using an automated method. In some embodiments, the live cells are classified as live by analysis of holographic images of cells of the population of cells, for instance, as exemplified herein (e.g., using the OsOne software for the classification of live cells). In some embodiments, the one or more input measures for a cellular feature of the one or more cellular features include input measures only from cells of at least a certain size. In some embodiments, size is determined by the radius mean cellular feature. In some embodiments, size is determined by the cell area cellular feature. In some embodiments, the one or more input measures for a cellular feature of the one or more cellular features include input measures only from cells of sufficient circularity, for instance as determined using the circularity cellular feature. In some embodiments, the one or more input measures for a cellular feature of the one or more cellular features include input measures only from cells having input measures for the intensity smoothness cellular feature below a certain threshold. In some embodiments, the one or more input measures for a cellular feature of the one or more cellular features include input measures only from cells satisfying all of the above criteria.

[0213] In some embodiments, the one or more input measures for a cellular feature of the one or more cellular features include one or more input measures only from cells that meet one or more (e.g., one, two, three, four, or all) of the following criteria: (i) classified as live, (ii) having radius mean cell feature measurements > 5, (iii) having intensity smoothness cell feature measurements < 0.03, (iv) having cell area cell feature measurements > 60, and (v) having circularity cell feature measurements > 0.5. In some embodiments, the cells meet all of the foregoing criteria.

[0214] In some embodiments, the one or more input measures for a cellular feature of the one or more cellular features include input measures from each of the individual cells of the population of cells.

[0215] In some embodiments, a cellular feature of the one or more cellular features is derived from phase information of the holographic information. In some embodiments, acellular feature of the one or more cellular features is derived from intensity information of the holographic information. In some embodiments, a cellular feature of the one or more cellular features is derived from the phase information and the intensity information. In some embodiments, a cellular feature of the one or more cellular features is derived from an image of an individual cell that is reconstructed from the holographic information. Exemplary methods for determining cellular features from holographic information are described in US- 9684281, US-20140193850, US-20140195568-A1, US-10578541-B2, US-9904248-B2, US- 11067379-B2, and US-2021142472-Al. Exemplary software for determining cellular features from holographic information includes Ovizio OsOne (Ovizio Imaging Systems NV / SA, Brussels, Belgium).

[0216] In some embodiments, the one or more cellular features includes one or more morphological features, one or more optical features, one or more intensity texture features, one or more phase texture features, or a combination of any of the foregoing. In some embodiments, the one or more cellular features includes one or more morphological features, one or more intensity texture features, and one or more phase texture features. Exemplary cellular features and descriptions thereof are described in Table El.

[0217] In some embodiments, the one or more cellular features include one or more morphological features. In some embodiments, the one or more morphological features describe one or more characteristics of an individual cell’s physical shape. In some embodiments, the one or more morphological features are selected from aspect ratio, cell area, circularity, compactness, elongatedness, elongation, diameter, hu moment 1, hu moment 2, hu moment 3, hu moment 4, hu moment 5, hu moment 6, hu moment 7, perimeter, radius mean, radius variance, and normalized radius variance. In some embodiments, the one or more morphological features include aspect ratio. In some embodiments, the one or more morphological features include cell area. In some embodiments, the one or more morphological features include circularity. In some embodiments, the one or more morphological features include compactness. In some embodiments, the one or more morphological features include elongatedness. In some embodiments, the one or more morphological features include elongation. In some embodiments, the one or more morphological features include diameter. In some embodiments, the one or more morphological features include hu moment 1. In some embodiments, the one or moremorphological features include hu moment 2. In some embodiments, the one or more morphological features include hu moment 3. In some embodiments, the one or more morphological features include hu moment 4. In some embodiments, the one or more morphological features include hu moment 5. In some embodiments, the one or more morphological features include hu moment 6. In some embodiments, the one or more morphological features include hu moment 7. In some embodiments, the one or more morphological features include perimeter. In some embodiments, the one or more morphological features include radius mean. In some embodiments, the one or more morphological features include radius variance. In some embodiments, the one or more morphological features include normalized radius variance.

[0218] In some embodiments, the one or more morphological features include cell area and radius mean.

[0219] In some embodiments, the one or more cellular features include one or more optical features. In some embodiments, the one or more optical features describe one or more optical properties of an image of an individual cell. In some embodiments, the one or more optical features are selected from refraction peak diameter, intensity maximum, mean intensity, intensity minimum, mass excentricity, optical height maximum (radians), optical height maximum (pm), mean optical height (radians), mean optical height (pm), normalized optical height, optical height minimum (radians), optical height minimum (pm), optical volume, refraction peak surface, normalized peak area, aggregate size, refraction peak intensity, and normalized peak height. In some embodiments, the one or more optical features include refraction peak diameter. In some embodiments, the one or more optical features include intensity maximum. In some embodiments, the one or more optical features include mean intensity. In some embodiments, the one or more optical features include intensity minimum. In some embodiments, the one or more optical features include mass excentricity. In some embodiments, the one or more optical features include optical height maximum (radians). In some embodiments, the one or more optical features include optical height maximum (pm). In some embodiments, the one or more optical features include mean optical height (radians). In some embodiments, the one or more optical features include mean optical height (pm). In some embodiments, the one or more optical features include normalized optical height. In some embodiments, the one or more optical features include optical height minimum(radians). In some embodiments, the one or more optical features include optical height minimum (pm). In some embodiments, the one or more optical features include optical volume. In some embodiments, the one or more optical features include refraction peak surface. In some embodiments, the one or more optical features include normalized peak area. In some embodiments, the one or more optical features include aggregate size. In some embodiments, the one or more optical features include refraction peak intensity. In some embodiments, the one or more optical features include normalized peak height.

[0220] In some embodiments, the one or more cellular features include one or more intensity texture features. In some embodiments, the one or more intensity texture features describe one or more properties of the intensity information of an individual cell. In some embodiments, the one or more intensity texture features are selected from intensity variance, intensity average contrast, intensity average entropy, intensity average, intensity average uniformity, intensity contrast, intensity correlation, intensity entropy, intensity homogeneity, intensity uniformity, intensity skewness, and intensity smoothness. In some embodiments, the one or more intensity texture features include intensity variance. In some embodiments, the one or more intensity texture features include intensity average contrast. In some embodiments, the one or more intensity texture features include intensity average entropy. In some embodiments, the one or more intensity texture features include intensity average. In some embodiments, the one or more intensity texture features include intensity average uniformity. In some embodiments, the one or more intensity texture features include intensity contrast. In some embodiments, the one or more intensity texture features include intensity correlation. In some embodiments, the one or more intensity texture features include intensity entropy. In some embodiments, the one or more intensity texture features include intensity homogeneity. In some embodiments, the one or more intensity texture features include intensity uniformity. In some embodiments, the one or more intensity texture features include intensity skewness. In some embodiments, the one or more intensity texture features include intensity smoothness.

[0221] In some embodiments, the one or more intensity texture features include maximum intensity, minimum intensity, intensity entropy, and intensity contrast.

[0222] In some embodiments, the one or more cellular features include one or more phase texture features. In some embodiments, the one or more phase texture features describe oneor more properties of the phase information of an individual cell. In some embodiments, the one or more phase texture features are selected from optical height variance (radians), optical height variance (pm), phase average contrast, phase average entropy, phase average, phase average uniformity, phase contrast, phase correlation, phase entropy, phase homogeneity, phase skewness, phase smoothness, and phase uniformity. In some embodiments, the one or more phase texture features include optical height variance (radians). In some embodiments, the one or more phase texture features include optical height variance (pm). In some embodiments, the one or more phase texture features include phase average contrast. In some embodiments, the one or more phase texture features include phase average entropy. In some embodiments, the one or more phase texture features include phase average. In some embodiments, the one or more phase texture features include phase average uniformity. In some embodiments, the one or more phase texture features include phase contrast. In some embodiments, the one or more phase texture features include phase correlation. In some embodiments, the one or more phase texture features include phase entropy. In some embodiments, the one or more phase texture features include phase homogeneity. In some embodiments, the one or more phase texture features include phase skewness. In some embodiments, the one or more phase texture features include phase smoothness. In some embodiments, the one or more phase texture features include phase uniformity.

[0223] In some embodiments, the cell phenotype is that of activated T cells.

[0224] In some embodiments, the plurality of cellular features includes one or more of (e.g., any combination of two, three, four, five, six, seven, eight, nine, or all of) intensity skewness, intensity correlation, intensity homogeneity, intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean. In some embodiments, the plurality of cellular features includes intensity skewness, intensity correlation, intensity homogeneity, intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean.

[0225] In some embodiments, the plurality of cellular features are selected from (comprises one or more of) maximum intensity, minimum intensity, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean. In some embodiments, the plurality of cellular features includes maximum intensity, minimum intensity, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean.

[0226] In some embodiments, the one or more cellular features includes one or more of (e.g., any combination of two, three, four, five, six, or all of) intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean.

[0227] In some embodiments, the plurality of cellular features includes one or more of (e.g., any combination of two or all of) intensity skewness, intensity correlation, and intensity homogeneity. In some embodiments, the plurality of cellular features includes intensity skewness, intensity correlation, and intensity homogeneity.

[0228] In some embodiments, the cell phenotype is a memory phenotype. In some embodiments, the memory phenotype is an earler memory phenotype, such as a stem cell memory phenotype or a central memory phenotype. In some embodiments, the cell phenotype is a central memory phenotype. In some embodiments, the cell phenotype is a stem cell memory phenotype. In some embodiments, the marker is CCR7. In some embodiments, the cell phenotype is a memory phenotype, e.g., a stem cell memory phenotype or a central memory phenotype, and the marker is CCR7.

[0229] Accordingly, in some embodiments, the disclosed method is for determining the memory phenotype of the population of T cells, and the marker is expressed by T cells having a central memory phenotype or a stem cell memory phenotype. In some embodiments, the marker is expressed by T cells having a central memory phenotype. In some embodiments, the marker is expressed by T cells having a stem cell memory phenotype. In some embodiments, the marker is CCR7.

[0230] In some embodiments, the plurality of cellular features comprises one or more of cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter. In some embodiments, the plurality of cellular features comprises cell area. In some embodiments, the plurality of cellular features comprises perimeter. In some embodiments, the plurality of cellular features comprises mean intensity. In some embodiments, the plurality of cellular features comprises normalized peak area. In some embodiments, the plurality of cellular features comprises equivalent peak diameter. In some embodiments, the plurality of cellular features comprises cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter.

[0231] In some embodiments, the cell phenotype is recombinant receptor expression, e.g., CAR or TCR expression.

[0232] In some embodiments, the plurality of cellular features includes one or more of (e.g., any combination of two, three, four or all of) peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height. In some embodiments, the plurality of cellular features includes peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height. In some embodiments, the plurality of cellular features comprises one or more of larger features reflective of cell size (area, perimeter, diameter), smaller mean intensity features, larger phase correlation features, and smaller radius variance normalized features. In some embodiments, the plurality of cellular features comprises one or more of larger cell area, larger cell perimeter, larger cell diameter, smaller intensity mean, larger phase correlation, and smaller radius variance normalized.

[0233] In some embodiments, the plurality of cellular features are cellular features extracted by a machine learning model. In some embodiments, the extraction is automatic, e.g., without prior design or engineering of the cellular features. In some embodiments, the one or more input measures for at least one of the plurality of cellular features are determined by providing the holographic information for individual cells of the population of cells to the machine learning model. In some embodiments, the one or more input measures for each of the plurality of cellular features are determined by providing the holographic information for individual cells of the population of cells to the machine learning model. In some embodiments, the one or more input measures are determined from the machine learning model.

[0234] In some embodiments, the machine learning model is a deep learning model. In some embodiments, the machine learning model is a convolutional neural network. Advantages of convolutional neural networks include their ability to automatically detect and extract features important for classification or regression without human supervision and to detect features that may not be detectable by humans.

[0235] In some embodiments, the one or more input measures for at least one of the plurality of cellular features are obtained from a fully connected layer of the convolutional neural network. In some embodiments, the one or more input measures for each of the plurality of cellular features are obtained from a fully connected layer of the convolutional neural network.

[0236] Architectures of convolutional neural networks suitable for feature extraction can be identified and designed by one of ordinary skill in the art. In some embodiments, the convolutional neural network is a LeNet, AlexNet, ResNet, GoogleNet / InceptionNet, MobileNetVl, ZfNet, Depth-based, Highway Network, Wide ResNet, VGG, PolyNet, Inception v2, Inception v3, Inception v4, Inception-ResNet, DenseNet, Pyramidal Net, Xception, Channel-Boosted, Residual Attention Neural Network, Attention-based, Feature- Map Exploitation-based, Squeeze-and-Excitation, or Competitive-Squeeze-and-Excitation convolutional neural network.

[0237] In some embodiments, the machine learning model, e.g., convolutional neural network, is trained using a dataset of reference holographic information. In some embodiments, for each of a first plurality of reference populations of cells containing T cells, the dataset of reference holographic information contains holographic information obtained for the reference population of cells, e.g., for individual cells of the reference population of cells. Exemplary reference populations of cells are described in Section I-D-l.

[0238] In some embodiments, the machine learning model, e.g., convolutional neural network, is trained using non-holographic images. In some embodiments, the machine learning model, e.g., convolutional neural network, is trained using holographic and nonholographic images.C. Population-Level Statistics

[0239] In some embodiments, the cell phenotype, e.g., activation state, is determined based on at least one population-level statistic. In some embodiments, the population-level output measure is determined based on at least one population-level statistic. In some embodiments, each population-level statistic is of one or more input measures for a cellular feature of the one or more cellular features. In some embodiments, each population-level statistic describes the one or more input measures for the cellular feature of the one or more cellular features. In some embodiments, each population-level statistic describes the distribution of the one or more input measures for the cellular feature of the one or more cellular features.

[0240] In some embodiments, the at least one population-level statistic includes one or more population-level statistics for each of the one or more cellular features. In some embodiments, the at least one population-level statistic is a plurality of population-level statistics. In some embodiments, the plurality of population-level statistics includes one ormore population-level statistics for each of the one or more cellular features. In some embodiments, the plurality of population-level statistics includes multiple population-level statistics for each of the one or more cellular features.

[0241] In some embodiments, the one or more population-level statistics for a first cellular feature are different statistics than the one or more population-level statistics for a second cellular feature. For instance, in some embodiments, the one or more population-level statistics for a first cellular feature include the mean of the one or more input features for the first cellular feature, and the one or more population-level statistics for a second cellular feature do not include the mean of the one or more input features for the second cellular feature. In some embodiments, the one or more population-level statistics for a first cellular feature are the same statistics as the population-level statistics for a second cellular feature. In some embodiments, the one or more population-level statistics are the same statistics for all cellular features.

[0242] Exemplary population-level statistics are described in this section and are known in the art. The one or more population-level statistics for each of the one or more cellular features can be independently selected from any such population-level statistics, e.g., from any of the described exemplary population-level statistics. In some embodiments, the one or more population-level statistics for each of the one or more cellular features are the same and are selected from any of the described exemplary population-level statistics.

[0243] In some embodiments, the one or more population-level statistics for a cellular feature of the one or more cellular features include a population-level statistic that summarizes the one or more input measures of the cellular feature. In some embodiments, the population-level statistic is a summary statistic.

[0244] In some embodiments, the one or more population-level statistics for a cellular feature of the one or more cellular features include a measure of the central tendency of the one or more input measures of the cellular feature. In some embodiments, the one or more population-level statistics include one or more of a mean, median, and mode of the one or more input measures of the cellular feature.

[0245] In some embodiments, the one or more population-level statistics for a cellular feature of the one or more cellular features include a measure of the statistical dispersion of the one or more input measures of the cellular feature. In some embodiments, the one or morepopulation-level statistics include one or more of the standard deviation, interquartile range, range, mean absolute difference, median absolute deviation, average absolute deviation, and distance standard deviation of the one or more input measures of the cellular feature.

[0246] In some embodiments, the one or more population-level statistics for a cellular feature of the one or more cellular features include a measure of the shape of the distribution of the one or more input measures of the cellular feature. In some embodiments, the one or more population-level statistics include one or both of the skewness and kurtosis of the one or more input measures of the cellular feature.

[0247] In some embodiments, the one or more population-level statistics for a cellular feature of the one or more cellular features include one or more quantiles of the one or more input measures of the cellular feature. In some embodiments, the one or more quantiles are selected from (include one or more of) the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of the one or more input measures of the cellular feature. In some embodiments, the one or more quantiles include the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of the one or more input measures of the cellular feature. In some embodiments, the one or more population-level statistics for each cellular feature of the one or more cellular features include the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of the one or more input measures of the cellular feature.

[0248] In some embodiments, the one or more population-level statistics for a cellular feature of the one or more cellular features are determined by applying a pooling filter to the one or more input measures of the cellular feature. In some embodiments, the one or more population-level statistics for each cellular feature of the one or more cellular features are determined by applying a pooling filter to the one or more input measures of the cellular feature.

[0249] In some embodiments, the pooling filter summarizes the one or more input measures of the cellular feature. In some embodiments, the pooling filter is a point estimatebased pooling filter. In some embodiments, the pooling filter is a mean pooling filter. In some embodiments, the pooling filter is a maximum pooling filter.

[0250] In some embodiments, the pooling filter is a distribution-based pooling filter. In some embodiments, the distribution-based pooling filter is based on the estimated marginal distribution of the one or more input measures of the cellular features, such as described inOner et al., arXiv:2006.01561. In some embodiments, the estimated marginal distribution is calculated using kernel density estimation, such as with a Gaussian kernel.D. Determining Population-Level Output Measures

[0251] In some embodiments, the cell phenotype, e.g., activation state, of the population of cells is determined based on the population-level output measure determined for the population of cells. In some embodiments, the population-level output measure is determined based on the one or more input measures for each of the one or more cellular features. In some embodiments, the population-level output measure is determined based on the at least one population-level statistic determined for the population of cells.

[0252] In some embodiments, each population-level statistic is compared to a corresponding threshold value. For threshold values for some population-level statistics, a population-level statistic above the threshold value can indicate the presence of markerexpressing cells. For threshold values for other population-level statistics, a population-level statistic below the threshold value can indicate the presence of marker-expressing cells in the population of cells. In some embodiments, the threshold value for a cellular feature of the one or more cellular features is a value exhibited by at least 2%, 4%, 6%, 8%, 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, or 90% of a plurality of reference populations of cells containing T cells. In some embodiments, each of the plurality of reference populations of cells contains marker-expressing cells. In some embodiments, each of the plurality of reference populations of cells contains marker-expressing T cells. In some embodiments, at least 2%, 4%, 6%, 8%, 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, or 90% of each of the plurality of reference populations of cells are marker-expressing cells. In some embodiments, at least 2%, 4%, 6%, 8%, 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, or 90% of each of the plurality of reference populations of cells are marker-expressing T cells. In some embodiments, at least 2%, 4%, 6%, 8%, 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, or 90% of T cells of each of the plurality of reference populations of cells are marker-expressing T cells. Exemplary reference populations of cells are described in Section I-D- 1.

[0253] In some embodiments, the population-level output measure indicates whether at least one population-level statistic indicates the presence of marker-expressing cells in the population of cells. In some embodiments, the population-level output measure indicates whether at least 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, or 90% of the population-levelstatistics indicate the presence of marker-expressing cells in the population of cells. In some embodiments, the population-level output measure indicates whether the majority of population-level statistics indicate the presence of marker-expressing cells in the population of cells. In some embodiments, the population-level output measure indicates whether each population-level statistic indicates the presence of marker-expressing cells in the population of cells.

[0254] In some embodiments, each population-level statistic is compared to a corresponding threshold value. For threshold values for some population-level statistics, a population-level statistic above the threshold value can indicate the presence of activated T cells. For threshold values for other population-level statistics, a population-level statistic below the threshold value can indicate the presence of activated T cells in the population of cells. In some embodiments, the threshold value for a cellular feature of the one or more cellular features is a value exhibited by at least 2%, 4%, 6%, 8%, 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, or 90% of a plurality of reference populations of cells containing T cells. In some embodiments, each of the plurality of reference populations of cells contains activated T cells. In some embodiments, each of the plurality of reference populations of cells contains T cells expressing the marker indicative of T cell activation. In some embodiments, at least 2%, 4%, 6%, 8%, 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, or 90% of each of the plurality of reference populations of cells are activated T cells. In some embodiments, at least 2%, 4%, 6%, 8%, 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, or 90% of each of the plurality of reference populations of cells are T cells expressing the marker. In some embodiments, at least 2%, 4%, 6%, 8%, 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, or 90% of T cells of each of the plurality of reference populations of cells are T cells expressing the marker. Exemplary reference populations of cells are described in Section I-D- 1.

[0255] In some embodiments, the population-level output measure indicates whether at least one population-level statistic indicates the presence of activated T cells in the population of cells. In some embodiments, the population-level output measure indicates whether at least 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, or 90% of the population-level statistics indicate the presence of activated T cells in the population of cells. In some embodiments, the population-level output measure indicates whether the majority of population-level statistics indicate the presence of activated T cells in the population of cells. In some embodiments, thepopulation-level output measure indicates whether each population-level statistic indicates the presence of activated T cells in the population of cells.

[0256] In some embodiments, the population-level output measure is determined using a trained machine learning model. Exemplary machine learning models and methods of training same are described in Hastie et al., The Elements of Statistical Learning (2016); and Abu-Mostafa et al., Learning from Data (2012). Exemplary machine learning models are also described in Hastie et al., The Elements of Statistical Learning (2016); and Abu-Mostafa et al., Learning from Data (2012).

[0257] In some embodiments, the population-level output measure is or is derived from an output of the trained machine learning model. In some embodiments, the one or more input measures for each of the one or more cellular features are provided as input to the trained machine learning model or to a process that includes the trained machine learning model. In some embodiments, the at least one population-level statistic for each of the one or more cellular features is provided as input to the trained machine learning model or to a process that includes the trained machine learning model. In some embodiments, as part of being provided as input to the trained machine learning model, the input measures or populationlevel statistics undergo one or more preprocessing steps. In some embodiments, the preprocessing steps include normalization steps. In some embodiments, the preprocessing steps include dimensionality reduction steps.

[0258] In some embodiments, the machine learning model is an unsupervised machine learning model. In some embodiments, the machine learning model is a semi-supervised machine learning model. In some embodiments, the machine learning model is a supervised machine learning model.

[0259] In some embodiments, the machine learning model is trained to predict the presence or absence of marker-expressing cells in populations of cells containing T cells, the degree to which marker-expressing cells are present in populations of cells containing T cells, or the degree to which marker-expressing cells are present in T cells of populations of cells containing T cells. In some embodiments, the machine learning model is trained to predict the presence or absence of marker-expressing T cells in populations of cells containing T cells, the degree to which marker-expressing T cells are present in populations of cells containing Tcells, or the degree to which marker-expressing T cells are present in T cells of populations of cells containing T cells.

[0260] In some embodiments, the machine learning model is trained to predict the presence or absence of activated T cells in populations of cells containing T cells, the degree to which activated T cells are present in populations of cells containing T cells, or the degree to which activated T cells are present in T cells of populations of cells containing T cells. In some embodiments, the machine learning model is trained to predict the presence or absence of T cells expressing the marker indicative of T cell activation in populations of cells containing T cells, the degree to which T cells expressing the marker indicative of T cell activation are present in populations of cells containing T cells, or the degree to which T cells expressing the marker indicative of T cell activation are present in T cells of populations of cells containing T cells.

[0261] In some embodiments, the machine learning model is trained to predict based on input measures for the one or more cellular features. In some embodiments, the machine learning model is trained to predict based on population-level statistics for the one or more cellular features. In some embodiments, the machine learning model is trained to predict based on features that include the one or more cellular features as well as other features, for instance features of populations of cells that are not based on the characteristics of individual cells.

[0262] In some embodiments, the population-level output measure is determined by providing the plurality of population-level statistics as input to the machine learning model. In some embodiments, the machine learning model is trained to predict population-level output measures of expression of the marker based on population-level statistics of the plurality of cellular features.

[0263] In some embodiments, the machine learning model is a classification model. In some embodiments, the machine learning model is a regression model. For any of the exemplary machine learning models described herein, both classification and regression versions of the machine learning model are disclosed.

[0264] In some embodiments, the machine learning model is a linear model. In some embodiments, the machine learning model is a non-linear model. In some embodiments, the machine learning model is a Bayesian model.

[0265] In some embodiments, the machine learning model is a regularized model. In some embodiments, the machine learning model is a lasso-regularized model. In some embodiments, the machine learning model is a ridge-regularized model. In some embodiments, the machine learning model is an elastic-net-regularized model.

[0266] In some embodiments, the machine learning model is an artificial neural network. In some embodiments, the machine learning model is a support vector machine. In some embodiments, the machine learning model is an ensemble model. In some embodiments, the machine learning model includes decision trees. In some embodiments, the machine learning model includes boosted decision trees. In some embodiments, the machine learning model is a random forest model. In some embodiments, the machine learning model is an ensemble model that includes any combination of the machine learning models described herein. In some embodiments, a first machine learning model of the ensemble model is trained to predict a population-level output measure for a first marker, such as any described herein, and a second machine learning model of the ensemble model is trained to predict a populationlevel output measure for a second marker, such as any described herein, that is different from the first marker. In some embodiments, a first machine learning model of the ensemble model is trained to predict a population-level output measure for a marker, such as any described herein, and a second machine learning model of the ensemble model is trained to predict a population-level output measure for the same marker.

[0267] In some embodiments, the machine learning model is a multioutput model. In some embodiments, the multioutput model is a deep learning model, such as any described herein, for instance any of the convolutional neural networks described in Section I-B. In some embodiments, the multioutput model is trained to predict a population-level output measure for each of a plurality of different markers, which can be independently selected from any of the markers described herein.1. Reference Datasets

[0268] In some embodiments, the threshold values are determined based on a dataset of reference input measures. In some embodiments, the machine learning model is trained using a dataset of reference input measures. In some embodiments, for each of a first plurality ofreference populations of cells containing T cells, the dataset of reference input measures includes one or more reference input measures for each of the one or more cellular features.

[0269] In some embodiments, the threshold values are determined based on a dataset of reference population-level statistics. In some embodiments, the machine learning model is trained using a dataset of reference population-level statistics. In some embodiments, for each of a first plurality of reference populations of cells containing T cells, the dataset of reference population-level statistics includes one or more reference population-level statistics for each of the one or more cellular features. In some embodiments, each reference population-level statistic is of one or more reference input measures for a cellular feature of the plurality of cellular features.

[0270] In some embodiments, the provided methods are for training a machine learning model that predicts the cell phenotype, e.g., activation state, of a population of T cells. In some embodiments, the provide methods involve training a machine learning model using a dataset of reference input measures. In some embodiments, for each of a first plurality of reference populations of cells containing T cells, the dataset of reference input measures includes one or more reference input measures for each of one or more cellular features. Exemplary machine learning models are described in Section I-D.

[0271] In some embodiments, the provided methods involve training a machine learning model using a dataset of reference population-level statistics. In some embodiments, for each of a first plurality of reference populations of cells containing T cells, the dataset of reference population-level statistics includes one or more reference population-level statistics for each of the one or more cellular features. In some embodiments, each reference population-level statistic is of one or more reference input measures for a cellular feature of one or more cellular features. Exemplary machine learning models are described in Section I-D.

[0272] In some embodiments, the threshold values are determined based on a dataset of reference population-level output measures. In some embodiments, the machine learning model is trained using a dataset of reference population-level output measures. In some embodiments, the provided methods involve training the machine learning model using a dataset of reference population-level output measures. In some embodiments, for each of a second plurality of reference populations of cells, the dataset of reference population-leveloutput measures includes a reference population-level output measure for the reference population of cells.

[0273] In some embodiments, the machine learning model is a multiple instance learning model. In some embodiments, the machine learning model is a deep multiple instance learning model.

[0274] In some embodiments, the provided methods involve (a) training a convolutional neural network using a dataset of reference holographic information, wherein for each of a first plurality of reference populations of cells containing T cells, the dataset of reference holographic information contains holographic information obtained for individual cells of the reference population of cells; (b) determining, from the convolutional neural network, one or more reference input measures for each cellular feature of a plurality of cellular features derived from the holographic information, wherein the plurality of cellular features are cellular features extracted by the convolutional neural network, and each reference input measure is from an individual cell of a reference population of cells; (c) determining a dataset of reference population-level statistics, wherein the dataset of reference population-level statistics contains one or more reference population-level statistics for each of the plurality of cellular features; and (d) training a machine learning model using the dataset of reference population-level statistics and a dataset of reference population-level output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures contains a reference population-level output measure of expression of a marker for the reference population of cells. In some embodiments, the marker is a marker expressed by activated T cells.

[0275] In some embodiments, the machine learning model is trained to predict populationlevel output measures of expression of the marker based on population-level statistics of the plurality of cellular features.

[0276] In some embodiments, the first and second pluralities of reference populations of cells are the same. In some embodiments, the first and second pluralities of reference populations of cells are different from one another. In some embodiments, the dataset of reference population-level statistics and the dataset of reference population-level output measures are time-matched, e.g., include population-level statistics and population-level output measures that were measured at the same times or within one hour, 30 minutes, 20minutes, 10 minutes, or 5 minutes of one another, whether from the same or different reference populations of cells.

[0277] In some embodiments, for each of the first plurality of reference populations of cells, each reference input measure is derived from holographic information obtained for the reference population of cells, e.g., for individual cells of the population of cells. In some embodiments, each reference input measure is from an individual cell of the reference population of cells.

[0278] In some embodiments, the first and / or second plurality of reference populations of cells includes at least 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 95, or 100 reference populations of cells. In some embodiments, each of the first and / or second plurality of reference populations of cells is any of the populations of cells described in Section II. In some embodiments, the provided methods involve performing any of the cell processing steps described in Section II for each of the first and / or second plurality of reference populations of cells. In some embodiments, the first and / or second plurality of reference populations of cells are each enriched for T cells. In some embodiments, at least 50, 55, 60, 65, 70, 75, 80, 85, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, or 100% of each of the first and / or second plurality of reference populations of cells are T cells.

[0279] In some embodiments, the first and / or second plurality of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo. In some embodiments, the first and second pluralities of reference populations of cells include reference populations from the same reference culture of cells. In some embodiments, the provided methods involve culturing the reference cultures of cells. Exemplary methods and conditions for culture are described in Section II-D. In some embodiments, the in vitro or ex vivo culture of the reference cultures of cells is performed under the same or similar conditions as the in vitro or ex vivo culture of the culture of cells.

[0280] In some embodiments, the first and / or second plurality of reference populations of cells are incubated under T cell stimulating conditions. In some embodiments, the incubation is prior to when the holographic information for the first plurality of reference populations of cells is obtained. In some embodiments, the provided methods involve incubating the first and / or second plurality of reference populations of cells under T cell stimulating conditions. Exemplary methods and conditions for stimulating T cells are described in Section II-B. Insome embodiments, the incubation of the first and / or second plurality of reference populations of cells is performed under the same or similar conditions as the incubation of the population of cells.

[0281] In some embodiments, the first and / or second plurality of reference populations of cells are genetically engineered to express a recombinant protein. In some embodiments, the provided methods involve genetically engineering the first and / or second plurality of reference populations of cells to express a recombinant protein. Exemplary engineering methods are described in Section II-C. In some embodiments, the engineering of the first and / or second plurality of reference populations of cells is performed under the same or similar conditions as the engineering of the population of cells. In some embodiments, the first and / or second plurality of reference populations of cells are engineered to express the same recombinant protein that the population of cells is engineered to express.

[0282] In some embodiments, the holographic information for the first plurality of reference populations of cells is obtained according to any of the methods described in Section I-A. In some embodiments, the provided methods involve obtaining the holographic information for the first plurality of reference populations of cells. In some embodiments, the holographic information for the first plurality of reference populations of cells is obtained during the in vitro or ex vivo culture of the reference cultures of cells. In some embodiments, multiple reference populations of cells of the first plurality of reference populations of cells are from the same reference culture of cells. In some embodiments, the holographic information for the multiple reference populations of cells is obtained at multiple timepoints during the in vitro or ex vivo culture of the reference culture of cells.

[0283] In some embodiments, the holographic information for the first plurality of reference populations of cells is obtained according to the method used to obtain the holographic information for the population of cells. In some embodiments, the holographic information for the first plurality of reference populations of cells is obtained by DHM.

[0284] In some embodiments, the reference input measures are determined according to any of the methods described in Section I-B. In some embodiments, the provided methods involve determining the reference input measures. In some embodiments, the dataset of reference input measures or reference population-level statistics includes only the one or more cellular features for which input measures or population-level statistics are obtained forthe population of cells. In some embodiments, the dataset of reference input measures or reference population-level statistics includes other features in addition to the one or more cellular features. The other features can be determined from holographic or non-holographic information obtained for the first plurality of reference populations of cells. The other features can be based or not based on the characteristics of individual cells of the first plurality of reference populations of cells.

[0285] In some embodiments, the one or more reference population-level statistics for each of the one or more cellular features are any of the population-level statistics described in Section I-C. In some embodiments, the provided methods involve determining the one or more reference population-level statistics for each of the one or more cellular features. In some embodiments, for a cellular feature of the one or more cellular features, the one or more reference population-level statistics is the same as the one or more population-level statistics for the population of cells. In some embodiments, the one or more reference population-level statistics for a cellular feature of the one or more cellular features include one or more quantiles of the one or more reference input measures of the cellular feature. In some embodiments, the one or more quantiles are selected from the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of the one or more reference input measures of the cellular feature. In some embodiments, the one or more quantiles include the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of the one or more reference input measures of the cellular feature. In some embodiments, the one or more reference population-level statistics for each cellular feature of the one or more cellular features include the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of the one or more reference input measures of the cellular feature.

[0286] In some embodiments, the one or more reference population-level statistics for a cellular feature of the one or more cellular features are determined by applying a pooling filter to the one or more reference input measures for the cellular feature. In some embodiments, each reference population-level statistic is determined by applying a distribution-based pooling filter to the one or more reference input measures for a cellular feature of the plurality of cellular features. In some embodiments, the pooling filter is a point estimate-based pooling filter. In some embodiments, the pooling filter is a mean pooling filter. In some embodiments, the pooling filter is a maximum pooling filter. In some embodiments, the pooling filter is a distribution-based pooling filter.

[0287] In some embodiments, for each of the second plurality of reference populations of cells, the reference population-level output measure indicates the cell phenotype of the reference population of cells. In some embodiments, the reference population-level output measure indicates the presence or absence of marker-expressing cells in the reference population of cells. In some embodiments, the reference population-level output measure indicates the degree to which marker-expressing cells are present in the reference population of cells. In some embodiments, the reference population-level output measure indicates the degree to which marker-expressing cells are present in T cells of the reference population of cells.

[0288] In some embodiments, the reference population-level output measure is the number of marker-expressing cells, the percentage of marker-expressing cells, the proportion of marker-expressing cells, or the density of marker-expressing cells in the reference population of cells. In some embodiments, the reference population-level output measure is the number of marker-expressing cells, the percentage of marker-expressing cells, the proportion of marker-expressing cells, or the density of marker-expressing cells in T cells of the reference population of cells. In some embodiments, the reference population-level output measure is the percentage of marker-expressing in T cells of the reference population of cells.

[0289] In some embodiments, for each of the second plurality of reference populations of cells, the reference population-level output measure indicates the activation state of the reference population of cells. In some embodiments, the reference population-level output measure indicates the presence or absence of activated T cells in the reference population of cells. In some embodiments, the reference population-level output measure indicates the degree to which activated T cells are present in the reference population of cells. In some embodiments, the reference population-level output measure indicates the degree to which activated T cells are present in T cells of the reference population of cells.

[0290] In some embodiments, the reference population-level output measure is the number of activated T cells, the percentage of activated T cells, the proportion of activated T cells, or the density of activated T cells in the reference population of cells. In some embodiments, the reference population-level output measure is the number of activated T cells, the percentage of activated T cells, the proportion of activated T cells, or the density of activated T cells in T cells of the reference population of cells. In some embodiments, the reference population-level output measure is the percentage of activated T cells in T cells of the reference population of cells.

[0291] In some embodiments, the reference population-level output measure is of expression of the marker indicative of T cell activation by T cells of the reference population of cells. In some embodiments, the reference population-level output measure indicates the presence of T cells expressing the marker in the reference population of cells. In some embodiments, the reference population-level output measure indicates the degree to which T cells expressing the marker are present in the reference population of cells. In some embodiments, the reference population-level output measure indicates the degree to which T cells expressing the marker are present in T cells of the reference population of cells.

[0292] In some embodiments, the reference population-level output measure is the number of cells, the percentage of cells, the proportion of cells, or the density of cells of the reference population of cells that express the marker. In some embodiments, the reference populationlevel output measure is the number of cells, the percentage of cells, the proportion of cells, or the density of cells in T cells of the reference population of cells that express the marker. In some embodiments, the reference population-level output measure is the percentage of cells in T cells of the reference population of cells that express the marker.

[0293] In some embodiments, the dataset of reference population-level output measures are of expression of the marker during or after the in vitro or ex vivo culture of the reference cultures of cells. In some embodiments, multiple reference populations of cells of the second plurality of reference populations of cells are from the same reference culture of cells. In some embodiments, the reference population-level output measures for the multiple reference populations of cells is of expression of the marker at multiple timepoints during the in vitro or ex vivo culture of the reference culture of cells. In some embodiments, for a reference population of cells, the reference population-level output measure is of expression of the marker at the same timepoint that the holographic information for a reference population of cells of the first plurality of reference populations of cells is obtained.

[0294] In some embodiments, the dataset of reference population-level output measures is determined using fluorescence imaging of cells of the reference cultures of cells. In some embodiments, the provided methods involve determining the dataset of reference population-level output measures using fluorescence imaging of the second plurality of reference populations of cells. In some embodiments, the fluorescence imaging is by flow cytometry.

[0295] In some embodiments, one or more preparation and / or non-affinity-based cell separation steps are carried out prior to labelling the cells for flow cytometry. In some embodiments, the cells are washed, centrifuged, and / or incubated in the presence of one or more reagents, for example, to remove unwanted components, enrich for desired components, or lyse or remove cells sensitive to particular reagents. In some embodiments, the cells are separated based on one or more properties, such as density, adherent properties, size, sensitivity, and / or resistance to particular components. In some embodiments, the methods include density -based cell separation methods.

[0296] In some embodiments, the cells are labelled with one or more fluorescent markers (e.g., one or more fluorophores) that produce a fluorescent signal that can be measured by a flow cytometer. In some embodiments, the cells are labelled by incubating the cells with one or more staining reagents, in which each staining reagent contains a fluorescent signal or marker (e.g., fluorophore). The staining reagent can be any reagent for characterization, selection, or isolation of a particular cell type or subtype of cells. In some embodiments, the staining reagent contains an immunoaffinity -based reagent, such as an antibody.

[0297] In some embodiments, the staining reagent stains the cells based on the cells’ expression or expression level of the marker, e.g., the marker indicative of T cell activation. In some embodiments, the staining reagent contains one or more fluorescent markers that may be attached, such as by chemical conjugation, to a binding agent that is able to bind, such as specifically bind, to the marker. In some embodiments, the binding agent is a protein. In some embodiments, the binding agent is an antibody or an antigen-binding fragment. The fluorescent marker may be conjugated to the binding agent, e.g., antibody, by any method known in the art.

[0298] As is well known in the art, an “antibody” is an immunoglobulin (Ig) molecule capable of specific binding to a target, such as a carbohydrate, polynucleotide, lipid, or polypeptide, through at least one epitope recognition site, located in the variable region of the Ig molecule. As used herein, the term encompasses not only intact polyclonal or monoclonal antibodies, but also fragments thereof, such as dAb, Fab, Fab', F(ab')2, Fv), single chain (scFv), synthetic variants thereof, naturally occurring variants, fusion proteins comprising anantibody portion with an antigen-binding fragment of the required specificity, chimeric antibodies, nanobodies, and any other modified configuration of the immunoglobulin molecule that comprises an antigen-binding site or fragment (epitope recognition site) of the required specificity. Minibodies comprising an scFv joined to a CH3 domain are also included herein.

[0299] A binding agent, such as an antibody, that "specifically binds" or "preferentially binds" (used interchangeably herein) to a marker is a term well understood in the art. A molecule is said to exhibit "specific binding" or "preferential binding" if it reacts or associates more frequently, more rapidly, with greater duration, and / or with greater affinity with a particular marker than it does with alternative markers. An antibody specifically binds or preferentially binds to a target if it binds with greater affinity, avidity, more readily, and / or with greater duration than it binds to other substances. It is also understood that specific binding or preferential binding does not necessarily require (although it can include) exclusive binding. Methods to determine such specific or preferential binding are also well known in the art, e.g., an immunoassay.

[0300] Antibodies for flow cytometry can be selected based on the marker being detected in the plurality of reference populations of cells, e.g., based on the marker that is indicative of T cell activation. In some embodiments, the marker is CD137 (4-1BB). Exemplary CD137- binding agents, such as anti-CD137 antibodies, for flow cytometry include clone 4B4-1 available from Miltenyi Biotec and clone 17B5 available from ThermoFisher Scientific.

[0301] In some embodiments, the marker is CD3. In some embodiments, the marker is CD4. In some embodiments, the marker is CD8. Exemplary CD3-, CD4-, and CD8-binding agents, such as antibodies, are described in Section II-A.

[0302] In some embodiments, the binding agent, such as antibody, is conjugated to a fluorescent marker, such as a fluorophore. For instance, the cells may be incubated with one or more fluorescently labeled antibodies. In some embodiments, any fluorescent marker or fluorophore suitable for use with flow cytometry analysis can be used. Exemplary fluorescent markers include fluorescent proteins (e.g., GFP, YFP, RFP), fluorescent moieties (e.g., fluorescein isothiocyanate (FITC), Phycoerythrin (PE), allophycocyanin (APC), and Alexa Fluor (AF)), nucleic acid colorants (e.g., 4 ', 6-diamidino-2-phenylindole (DAPI), SYT016, and propidium iodide (PI)), cell membrane stain (e.g., FMI-43), cell functional dyes (e.g.,Fluo-4 and Indo-1), and synthetic dyes (e.g., Brilliant Violet (BV)). Exemplary fluorophores include hydroxycoumarin, Cascade Blue, Dylight 405 Pacific Orange, Alexa Fluor 430, Fluorescein, Oregon Green, Alexa Fluor 488, BODIPY 493, 2,7-Diochlorofluorescien, ATTO 488, Chromeo 488, Dylight 488, HiEyte 488, Alexa Fluor 532, Alexa Fluor 555, ATTO 550, BODIPY TMR-X, CF 555, Chromeo 546, Cy3, TMR, TRITC, Dy547, Dy548, Dy549, HiEyte 555, Dylight 550, BODIPY 564, Alexa Fluor 568, Alexa Fluor 594, Rhodamine, Texas Red, Alexa Fluor 610, Alexa Fluor 633, Dylight 633, Alexa Fluor 647, APC, ATTO 655, CF633, CF640R, Chromeo642, Cy5, Dylight 650, Alexa Fluor 680, IRDye 680, Alexa Fluor 700 (AF700), Cy5.5, ICG, Alexa Fluor 750, Dylight 755, IRDye 750, Cy7, PE-Cy7, Cy7.5, Alexa Fluor 790, Dylight 800, IRDye 800, BV421, BV510, BV570, BV605, BV650, BV711, BV750, BV785, Qdot® 525, Qdot® 565, Qdot® 605, Qdot® 655, Qdot® 705, and Qdot® 800.

[0303] In some embodiments, the cell staining for flow cytometry involves incubation with the staining reagent, which in some embodiments is followed by washing steps and separation of cells bound to the staining reagent from those cells not bound to the staining reagent. In some embodiments, a volume of cells is mixed with an amount of a desired staining reagent and incubated under conditions for staining of the cells. In some embodiments, the staining is carried out at a temperature between 0°C and 25°C, such as at or about 4°C. In some embodiments, the staining is carried out for greater than 5 minutes, typically greater than 15 minutes. In some embodiments, the staining is carried out for between 15 minutes and 6 hours, such as between 30 minutes and 2 hours. In some embodiments, the staining is carried out, for example, at or about 15 minutes, 30 minutes, 1 hour, 1.5 hours, 2 hours, 2.5 hours, 3 hours, or any value between any of the foregoing. In some embodiments, one or more wash steps are carried out prior to introducing the cells into the flow cytometer for analysis. In some embodiments, the stained cells are introduced into a flow cytometer.

[0304] In some embodiments, the cells are prepared by suspending single cells at a density of 1 x 106to 1 x 107cells / mE in order to allow the cells to pass through the flow cytometer for reading. In some embodiments, this concentration of cells is called the fluid sheath. In some embodiments, the fluid sheath influences the rate of flow sorting, which typically progresses at around 2,000-20,000 cells per second. The cell sample's fluid sheath can bemade of a phosphate buffered saline solution, but other solutions are available, as will be known and understood by those skilled in the art.II. T CELL POPULATIONS

[0305] Exemplary populations of cells containing T cells and cell processing steps involving the populations of cells are described in this section. In some embodiments, each of the reference populations of cells referred to in Section I-D-l is individually selected from any of the described populations of cells containing T cells. In some embodiments, the provided methods involve performing any of the described cell processing steps for each of the plurality of reference populations of cells. In some embodiments, the same or similar cell processing steps are performed for each of the plurality of reference populations of cells. In some embodiments, the same or similar cell processing steps are performed for each of the plurality of reference population of cells and for the population of cells.

[0306] In some embodiments, at least 2%, 4%, 6%, 8%, 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, or 90% of the population of cells is T cells. In some embodiments, the population of cells is enriched for T cells. In some embodiments, at least 50, 55, 60, 65, 70, 75, 80, 85, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, or 100% of the population of cells is T cells. Exemplary methods for the selection of T cells from, for example, mixed populations of cells are described in Section LA.

[0307] In some embodiments, the population of cells are incubated under T cell stimulating conditions. In some embodiments, the provided methods involve incubating the population of cells under T cell stimulating conditions. Exemplary T cell stimulating conditions are described in Section ILB.

[0308] In some embodiments, the population of cells are genetically engineered to express a recombinant protein. In some embodiments, the provided methods involve genetically engineering the population of cells to express a recombinant protein. Exemplary engineering methods are described in Section ILC.

[0309] In some embodiments, the population of cells is from a culture of cells being cultured in vitro or ex vivo. In some embodiments, the provided methods involve culturing the culture of cells. Exemplary conditions for in vitro or ex vivo culture are described in Section ILD.

[0310] In some embodiments, the holographic information for the individual cells of the population of cells is obtained during the in vitro or ex vivo culture of the culture of cells. In some embodiments, the cell phenotype, e.g., activation state, of the population of cells during the in vitro or ex vivo culture is determined. In some embodiments, the population-level output measure indicates the cell phenotype, e.g., activation state, of the population of cells during the in vitro or ex vivo culture. In some embodiments, expression of the marker, e.g., the marker indicative of T cell activation, by the population of cells during the in vitro or ex vivo culture is determined. In some embodiments, the population-level output measure is of expression of the marker during the in vitro or ex vivo culture of the culture of cells.

[0311] In some embodiments, the culture of cells is incubated under T cell stimulating conditions. In some embodiments, the provided methods involve incubating the culture of cells under T cell stimulating conditions. Exemplary T cell stimulating conditions are described in Section II-B.

[0312] In some embodiments, the incubation under T cell stimulating conditions is prior to when the holographic information for the individual cells of the population of cells is obtained. In some embodiments, the incubation under T cell stimulating conditions is subsequent to when the holographic information for the individual cells of the population of cells is obtained. In some embodiments, the incubation under T cell stimulating conditions is prior to and subsequent to when the holographic information for the individual cells of the population of cells is obtained.

[0313] Sections II-A to II-C describe exemplary methods and reagents for selecting, stimulating, and culturing populations of cells containing T cells, respectively. In some embodiments, the provided methods involve one or more of the described steps of selecting, stimulating, and culturing T cells. In some embodiments, the provided methods involve steps of stimulating T cells. In some embodiments, the provided methods involve steps of culturing T cells. In some embodiments, the provided methods involve steps of stimulating and culturing T cells.

[0314] In some embodiments, a step of selecting T cells is performed prior to steps of stimulating or culturing T cells. In some embodiments, a step of selecting T cells is performed subsequent to steps of stimulating or culturing T cells.

[0315] In some embodiments, the incubation under T cell stimulating conditions is prior to the in vitro or ex vivo culture. In some embodiments, at least a portion of the in vitro or ex vivo culture is under T cell stimulating conditions. In some embodiments, all of the in vitro or ex vivo culture is under T cell stimulating conditions. In some embodiments, the incubation under T cell stimulating conditions is subsequent to the in vitro or ex vivo culture.

[0316] In some embodiments, some or all of the steps of selecting, stimulating, and culturing T cells are carried out at a temperature that is above room temperature, for instance at a physiological temperature. In some embodiments, the temperature is between about 35 °C and about 39°C, such as at or about 37°C.

[0317] In some embodiments, any number of the steps of the provided methods are performed in a closed system. In some embodiments, any number of the steps of the provided method are automated.A. Selection

[0318] In some embodiments, the population of cells is from a biological sample. In some embodiments, the population of cells are primary cells. In some embodiments, the primary cells are primary cells from a human subject. The biological samples can include tissue, fluid, and other samples taken directly from the subject. The biological sample can be a sample obtained directly from a biological source or a sample that is processed. Exemplary biological samples include body fluids, such as blood, plasma, serum, cerebrospinal fluid, synovial fluid, urine, and sweat; tissue; and organ samples, including processed samples derived therefrom. Exemplary biological samples also include whole blood, peripheral blood mononuclear cells (PBMCs), leukocytes, bone marrow, thymus, tissue biopsy, tumor, leukemia, lymphoma, lymph node, gut associated lymphoid tissue, mucosa associated lymphoid tissue, spleen, other lymphoid tissues, liver, lung, stomach, intestine, colon, kidney, pancreas, breast, bone, prostate, cervix, testes, ovaries, tonsil, or other organ, or cells derived therefrom.

[0319] In some examples, cells from the circulating blood of the subject are obtained by, e.g., apheresis or leukapheresis. The biological samples can contain lymphocytes, including T cells, monocytes, granulocytes, B cells, other nucleated white blood cells, red blood cells, and / or platelets, and in some aspects contain cells other than red blood cells and platelets.

[0320] In some embodiments, the biological sample is a sample containing T cells. In some embodiments, the biological sample is a whole blood sample, a buffy coat sample, a peripheral blood mononuclear cell (PBMC) sample, an unfractionated T cell sample, a lymphocyte sample, a white blood cell sample, an apheresis product, or a leukapheresis product. In some embodiments, the biological sample is an apheresis product. In some embodiments, the biological sample is a leukaphresis product.

[0321] In some embodiments, the cells obtained from the subject are washed to, e.g., remove the plasma fraction and to place the cells in an appropriate buffer or media for subsequent processing steps. In some embodiments, the cells are washed with phosphate buffered saline (PBS). In some embodiments, the wash solution lacks calcium, magnesium, and / or many or all divalent cations. In some aspects, a washing step is accomplished using a semi- automated “flow-through” centrifuge (for example, the Cobe 2991 cell processor, Baxter) according to the manufacturer's instructions. In some aspects, a washing step is accomplished by tangential flow filtration (TFF) according to the manufacturer's instructions. In some embodiments, the cells are resuspended in a variety of biocompatible buffers after washing, such as Ca2+ / Mg2+free PBS. In some embodiments, components of a blood cell biological sample are removed, and the cells are directly resuspended in culture media. In some embodiments, the biological sample (e.g., an apheresis product or a leukapheresis product) is washed in order to remove one or more anti-coagulants, such as heparin, added during apheresis or leukapheresis.

[0322] In some embodiments, the selection of cells includes one or more preparation and / or non-affinity based cell separation steps. In some examples, the cells are washed, centrifuged, and / or incubated in the presence of one or more reagents, for example, to remove unwanted components, enrich for desired components, or lyse or remove cells sensitive to particular reagents. In some examples, the cells are separated based on one or more properties, such as density, adherent properties, size, sensitivity, and / or resistance to particular components. In some embodiments, the methods involve density-based cell separation methods, such as the preparation of white blood cells from peripheral blood by lysing the red blood cells and centrifugation through a Percoll or Ficoll gradient.

[0323] In some embodiments, a cryopreserved and / or cryoprotected apheresis product or leukapheresis product is thawed. In some embodiments, the thawed cell composition issubjected to dilution (e.g., with a serum-free medium) and / or wash (e.g., with a serum-free medium), which in some cases can remove or reduce unwanted or undesired components. In some cases, the dilution and / or wash removes or reduces the presence of a cryoprotectant, e.g. DMSO, contained in the thawed sample, which otherwise may negatively impact cellular viability, yield, or recovery upon extended room temperature exposure. In some embodiments, the dilution and / or wash allows media exchange of a thawed cryopreserved product into a serum-free medium, e.g., one described in US-20210207080.

[0324] Exemplary methods and reagents for the selection of T cells for producing the populations of cells containing T cells, including by positive or negative selection, are described in US-11400115, US-20190112576, US-10228312, US-20200354677, US- 20200384025, US-20210163893, US-20220002669, US-5985658, US-9023604, US- 20030175850, US-20040082012, US-20080255004, US-20130059288, and US-9678061.

[0325] In some embodiments, at least a portion of the selection step includes incubation of cells with a selection reagent. In some embodiments, the incubation is with a selection reagent or reagents, e.g., as part of selection methods which may be performed using one or more selection reagents for selection of one or more different cell types based on the expression or presence in or on the cell of one or more specific molecules, such as surface markers, e.g., surface proteins, intracellular markers, or nucleic acid. In some embodiments, any known method using a selection reagent or reagents for separation based on such markers may be used. In some embodiments, the selection reagent or reagents result in a separation that is affinity- or immunoaffinity-based separation. For example, the selection in some aspects includes incubation with a reagent or reagents for separation of cells and cell populations based on the cells’ expression or expression level of one or more markers, typically cell surface markers, for example, by incubation with a binding partner, e.g., antibody, that specifically binds to such markers, followed generally by washing steps and separation of cells having bound the binding partner, from those cells having not bound to the binding partner.

[0326] In some aspects of such processes, a volume of cells is mixed with an amount of a desired affinity-based selection reagent. The immunoaffinity-based selection can be carried out using any system or method that results in a favorable energetic interaction between the cells being separated and the molecule specifically binding to the marker on the cell, e.g., thebinding partner on a solid surface, e.g., particle. In some embodiments, methods are carried out using particles such as beads, e.g., magnetic beads, that are coated with a selection agent (e.g., antibody) specific to the marker of the cells. The particles (e.g., beads) can be incubated or mixed with cells in a container, such as a tube or bag, while shaking or mixing, with a constant cell density-to-particle (e.g., bead) ratio to aid in promoting energetically favored interactions.

[0327] In some embodiments, the total duration of the incubation with the selection reagent is from or from about 5 minutes to 6 hours, such as 30 minutes to 3 hours, for example, at least or about at least 30 minutes, 60 minutes, 120 minutes, or 180 minutes.

[0328] In some embodiments, after the incubation and / or mixing of the cells and selection reagent and / or reagents, the incubated cells are subjected to a separation to select for cells based on the presence or absence of the particular selection reagent or reagents. In some embodiments, after incubation with the selection reagents, incubated cells, including cells in which the selection reagent has bound are transferred into a system for immunoaffinity-based separation of the cells. In some embodiments, the system for immunoaffinity-based separation is or contains a magnetic separation column.

[0329] Such separation steps can be based on positive selection, in which the cells having bound the selection reagents, e.g., antibody, are retained for further use, and / or negative selection, in which the cells having not bound to the selection reagent, e.g., antibody, are retained. In some examples, both fractions are retained for further use. In some embodiments, the process steps further include negative and / or positive selection of the incubated and cells, such as using a system or apparatus that can perform an affinity-based selection. In some embodiments, isolation is carried out by enrichment for a particular cell population by positive selection, or depletion of a particular cell population, by negative selection. In some embodiments, positive or negative selection is accomplished by incubating cells with one or more antibodies or other binding agents that specifically bind to one or more surface markers expressed or expressed (marker+) at a relatively higher level (marker111811) on the positively or negatively selected cells, respectively.

[0330] The separation need not result in 100 % enrichment or removal of a particular cell population or cells expressing a particular selection marker. For example, positive selection of or enrichment for cells of a particular type, such as those expressing a selection marker,refers to increasing the number or percentage of such cells, but need not result in a complete absence of cells not expressing the selection marker. Likewise, negative selection, removal, or depletion of cells of a particular type, such as those expressing a selection marker, refers to decreasing the number or percentage of such cells, but need not result in a complete removal of all such cells.

[0331] In some examples, multiple rounds of separation steps are carried out, where the positively or negatively selected fraction from one step is subjected to another separation step, such as a subsequent positive or negative selection. In some examples, a single separation step can deplete cells expressing multiple markers simultaneously, such as by incubating cells with a plurality of antibodies or other binding partners, each specific for a marker targeted for negative selection. Likewise, multiple cell types can simultaneously be positively selected by incubating cells with a plurality of antibodies or other binding partners expressed on the various cell types. In certain embodiments, separation steps are repeated and or performed more than once, where the positively or negatively selected fraction from one step is subjected to the same separation step, such as a repeated positive or negative selection. In some examples, a single separation step is repeated and / or performed more than once, for example to increase the purity of the selected cells and / or to further remove and / or deplete the negatively selected cells from the negatively selected fraction. In certain embodiments, one or more separation steps are performed two times, three times, four times, five times, six times, seven times, eight times, nine times, ten times, or more than ten times. In certain embodiments, the one or more selection steps are performed and / or repeated between one and ten times, between one and five times, or between three and five times.

[0332] In some embodiments, T cells are separated from a PBMC, apheresis, or leukapheresis sample by negative selection of markers expressed on non-T cells, such as B cells, monocytes, or other white blood cells, such as CD 14. In some aspects, a CD3+ selection step is used to generate a population enriched in CD3+ T cells from a starting sample, such as a PBMC, apheresis, or leukapheresis sample, wherein the starting sample has not been subjected to positive or negative selection based on another marker such as CD4 and / or CD8. In some embodiments, the starting sample is selected for CD3 without any previous, concurrent, or subsequent selection for another marker (e.g., CD4 and / or CD8) in order to generate a population enriched in CD3+ T cells. In some embodiments, thepopulation of cells is enriched in CD3+ T cells, which population is not subjected to a further selection, e.g., CD4+ and / or CD8+ selection, before being subjected to a step of, for example, stimulating the population of cells. In certain embodiments, the population enriched in CD3+ T cells has a ratio of between 1:10 and 10:1, between 1:5 and 5:1, between 4:1 and 1:4, between 1:3 and 3:1, between 2:1 and 1:2, between 1.5:1 and 1:1.5, between 1.25:1 and 1:1.25, between 1.2:1 and 1:1.2, between 1.1:1 and 1:1.1, or about 1:1 or 1:1 CD4+ T cells to CD8+ T cells.

[0333] In some embodiments, T cells are separated from a PBMC, apheresis, or leukapheresis sample by negative selection of markers expressed on non-T cells, such as B cells, monocytes, or other white blood cells, such as CD 14. In some aspects, a CD4+ or CD8+ selection step is used to separate CD4+ helper and CD8+ cytotoxic T cells. Such CD4+ and CD8+ populations can be further sorted into sub-populations by positive or negative selection for markers expressed or expressed to a relatively higher degree on one or more naive-like, memory, and / or effector T cell subpopulations.

[0334] In particular embodiments, a biological sample, e.g., a PBMC, apheresis, or leukapheresis sample, is subjected to selection of CD4+ T cells, where both the negative and positive fractions are retained. In certain embodiments, CD8+ T cells are selected from the negative fraction. In some embodiments, a biological sample is subjected to selection of CD8+ T cells, where both the negative and positive fractions are retained. In certain embodiments, CD4+ T cells are selected from the negative fraction.

[0335] In some aspects, a CD8-based positive selection step is used to generate a population enriched in CD8+ T cells from a starting sample, such as a PBMC, apheresis, or leukaphresis sample, wherein the starting sample has not been subjected to selection based on another marker such as CD3+ and / or CD4+. In some embodiments, both the negative and positive fractions from the CD8 positive selection step are retained, and the CD8-negative fraction is further subjected to a CD4-based positive selection step in order to generate a population enriched in CD4+ T cells. In some embodiments, cells from the population enriched in CD8+ T cells and cells from the population enriched in CD4+ T cells are mixed, combined, and / or pooled to generate a population containing CD4+ T cells and CD8+ T cells.

[0336] In some aspects, a CD4-based positive selection step is used to generate a population enriched in CD4+ T cells from a starting sample, such as a PBMC, apheresis, orleukaphresis sample, wherein the starting sample has not been subjected to selection based on another marker such as CD3+ and / or CD8+. In some embodiments, both the negative and positive fractions from the CD4 positive selection step are retained, and the CD4-negative fraction is further subjected to a CD8-based positive selection step used to generate a population enriched in CD8+ T cells. In some embodiments, cells from the population enriched in CD4+ T cells and cells from the population enriched in CD8+ T cells are mixed, combined, and / or pooled to generate a population containing CD8+ T cells and CD4+ T cells.

[0337] In certain embodiments, the population enriched in CD4+ T cells and the population enriched in CD8+ T cells are pooled, mixed, and / or combined prior to stimulating cells, e.g., culturing the cells under stimulating conditions such as described in Section I-B. In certain embodiments, the pooled, mixed, and / or combined cells or populations have a ratio of between 1:10 and 10:1, between 1:5 and 5:1, between 4:1 and 1:4, between 1:3 and 3:1, between 2:1 and 1:2, between 1.5:1 and 1:1.5, between 1.25:1 and 1:1.25, between 1.2:1 and 1:1.2, between 1.1:1 and 1:1.1, or about 1:1, or 1:1 CD4+ T cells to CD8+ T cells. In certain embodiments, the cells or populations are pooled, mixed, and / or combined in order to have a ratio of or of about 1:1 CD4+ T cells to CD8+ T cells in the pooled, mixed, and / or combined cell composition.

[0338] In some embodiments, a selection agent that specifically binds CD4 and a selection agent that specifically binds CD8 are used to generate a population enriched in CD4+ T cells and a population enriched in CD8+ T cells, respectively. In some embodiments, the capacities of the CD4-specific selection agent and the CD8-specific selection agent are the same or substantially the same, for example, a unit volume or unit weight of the selection agents (e.g., ClinicMACS CD4 selection reagent and CD8 selection reagent) can be used to select CD4 or CD8 cells from the same number of total cells. In some embodiments, a greater amount of CD4-specific selection agent can be used than the CD8-specific selection agent with the same or substantially the same capacity. For example, the volumes or weights of the CD4-specific selection agent and the CD8-specific selection agent can be at a ratio of about 5:1, 4:1, 3:1, 2:1, or 1:5:1.

[0339] In some aspects, the incubated sample or population of cells to be separated is incubated with a selection reagent containing small, magnetizable, or magnetically responsive material, such as magnetically responsive particles or microparticles, such as paramagneticbeads (e.g., Dynabeads or MACS® beads). The magnetically responsive material, e.g., particle, generally is directly or indirectly attached to a binding partner, e.g., an antibody, that specifically binds to a molecule, e.g., surface marker, present on the population of cells that it is desired to separate, e.g., that it is desired to negatively or positively select.

[0340] In some embodiments, the magnetic particle, e.g., bead, contains a magnetically responsive material bound to a specific binding member, such as an antibody or other binding partner. Many well-known magnetically responsive materials for use in magnetic separation methods are known, e.g., those described in US-4452773 and in EP-452342. Colloidal sized particles, such as those described in US-4795698 and US-5200084 may also be used.

[0341] The incubation can be carried out under conditions whereby the antibodies or other binding partners, such as secondary antibodies or other reagents, which specifically bind to such antibodies or other binding partners, which are attached to the magnetic particle, e.g., bead, specifically bind to cell surface molecules if present on cells within the sample.

[0342] In certain embodiments, the magnetically responsive particles are coated in primary antibodies or other binding partners, secondary antibodies, lectins, enzymes, or streptavidin. In certain embodiments, the magnetic particles are attached to cells via a coating of primary antibodies specific for one or more markers. In certain embodiments, the cells, rather than the beads, are labeled with a primary antibody or binding partner, and then cell-type specific secondary antibody- or other binding partner (e.g., streptavidin)-coated magnetic particles, are added. In certain embodiments, streptavidin-coated magnetic particles are used in conjunction with biotinylated primary or secondary antibodies.

[0343] In some aspects, separation is achieved in a procedure in which the sample is placed in a magnetic field, and those cells having magnetically responsive or magnetizable particles attached thereto will be attracted to the magnet and separated from the unlabeled cells. For positive selection, cells that are attracted to the magnet are retained; for negative selection, cells that are not attracted (unlabeled cells) are retained. In some aspects, a combination of positive and negative selection is performed during the same selection step, where the positive and negative fractions are retained and further processed or subject to further separation steps.

[0344] In some embodiments, the affinity-based selection is via magnetic-activated cell sorting (MACS) (Miltenyi Biotech, Auburn, CA). Magnetic Activated Cell Sorting (MACS),e.g., CliniMACS systems, are capable of high-purity selection of cells having magnetized particles attached thereto. In certain embodiments, MACS operates in a mode wherein the non-target and target species are sequentially eluted after the application of the external magnetic field. That is, the cells attached to magnetized particles are held in place while the unattached species are eluted. Then, after this first elution step is completed, the species that were trapped in the magnetic field and were prevented from being eluted are freed in some manner such that they can be eluted and recovered. In certain embodiments, the non-target cells are labelled and depleted from the heterogeneous population of cells.

[0345] In some embodiments, the separation and / or isolation steps are carried out using magnetic beads in which immunoaffinity reagents are reversibly bound, such as via a peptide ligand interaction with a streptavidin mutein as described in US-20170037369. Exemplary of such magnetic beads are Streptamers®. In some embodiments, the separation and / or steps is carried out using magnetic beads, such as those commercially available from Miltenyi Biotec.

[0346] In some embodiments, the T cells are isolated, selected, or enriched by chromatographic isolation, such as by column chromatography including affinity chromatography or gel permeation chromatography. Such methods may be described as (traceless) cell affinity chromatography technology (CATCH) and may include any of the methods or techniques described in US- 10228312 and US-20170037369. Generally, a chromatographic method is a fluid chromatography, typically a liquid chromatography. In some aspects, the chromatography can be carried out in a flow through mode in which a fluid sample containing the cells is applied, for example, by gravity flow or by a pump on one end of a column containing the chromatography matrix and in which the fluid sample exists the column at the other end of the column. In addition, the chromatography can be carried out in an “up and down” mode in which a fluid sample containing the cells to be isolated is applied, for example, by a pipette on one end of a column containing the chromatography matrix packed within a pipette tip and in which the fluid sample enters and exists the chromatography matrix / pipette tip at the other end of the column. Alternatively, the chromatography can also be carried out in a batch mode in which the chromatography material (e.g., stationary phase) is incubated with the sample that contains the cells, for example, under shaking, rotating, or repeated contacting and removal of the fluid sample, for example, by means of a pipette.

[0347] In some embodiments, the selection agent is contained in a chromatography column, e.g., bound directly or indirectly to the chromatography matrix (e.g., stationary phase). In some embodiments, the selection agent is present on the chromatography matrix (e.g., stationary phase) at the time the sample is added to the column. In some embodiments, the selection agent is capable of being bound indirectly to the chromatography matrix (e.g., stationary phase) through a reagent, e.g., selection reagent. In some embodiments, the selection reagent is bound covalently or non-covalently to the stationary phase of the column. In some embodiments, the selection reagent is reversibly immobilized on the chromatography matrix (e.g., stationary phase). In some cases, the selection reagent is immobilized on the chromatography matrix (e.g., stationary phase) via covalent bonds. In some aspects, the selection reagent is reversibly immobilized on the chromatography matrix (e.g., stationary phase) non-covalently.

[0348] In some embodiments, the selection agent may be present, for example bound directly to (e.g., covalently or non-covalently) or indirectly via a selection reagent, on the chromatography matrix (e.g., stationary phase) at the time the sample is added to the chromatography column (e.g., stationary phase). Thus, upon addition of the sample, T cells can be bound by the selection agent and immobilized on the chromatography matrix (e.g., stationary phase) of the column. Alternatively, in some embodiments, the selection agent can be added to the sample. In this way, the selection agent binds to the T cells in the sample, and the sample can then be added to a chromatography matrix (e.g., stationary phase) containing the selection reagent, where the selection agent, already bound to the T cells, binds to the selection reagent, thereby immobilizing the target cells on the chromatography matrix (e.g., stationary phase).

[0349] In some embodiments, one or both of a first and / or second selection can employ a plurality of affinity chromatography matrices and / or antibodies, whereby the plurality of matrices and / or antibodies are serially connected. In some embodiments, the affinity chromatography matrix or matrices employed in selection adsorb or are capable of selecting or enriching at least about 50 x 106cells / mL, 100 x 106cells / mL, 200 x 106cells / mL or 400 x 106cells / mL. In some embodiments, the adsorption capacity can be modulated based on the diameter and / or length of the matrix. In some embodiments, the culture-initiating ratio of the selected or enriched population is achieved by choosing a sufficient amount of matrixand / or at a sufficient relative amount to achieve the culture-initiating ratio assuming based on, for example, the adsorption capacity of the matrix or matrices for selecting cells.

[0350] In some aspects, the chromatography matrix / stationary phase is a non-magnetic material or non-magnetisable material. Such material may include derivatized silica or a crosslinked gel. A crosslinked gel (which is typically manufactured in a bead form) may be based on a natural polymer, such as a crosslinked polysaccharide. Suitable examples include agarose gels or a gel of crosslinked dextrans. A crosslinked gel may also be based on a synthetic polymer, e.g., on a polymer class that does not occur in nature. Usually, such a synthetic polymer on which a stationary phase for cell separation is based is a polymer that has polar monomer units and which is therefore in itself polar.

[0351] Illustrative examples of suitable synthetic polymers are polyacrylamides, a styrene- divinylbenzene gel, and a copolymer of an acrylate and a diol or of an acrylamide and a diol. An illustrative example is a polymethacrylate gel, commercially available as a Fractogel®. A further example is a copolymer of ethylene glycol and methacrylate, commercially available as a Toyopearl®. In some embodiments, a stationary phase may also include natural and synthetic polymer components, such as a composite matrix or a composite or a co-polymer of a polysaccharide and agarose, e.g., a polyacrylamide / agarose composite, or of a polysaccharide and N,N'-methylenebisacrylamide. An illustrative example of a copolymer of a dextran and N,N'-methylenebisacryl-iamide is the above-mentioned Sephacryl® series of material. A derivatized silica may include silica particles that are coupled to a synthetic or to a natural polymer. Examples of such embodiments include polysaccharide grafted silica, polyvinyl-ipyrrolidone grafted silica, polyethylene oxide grafted silica, poly(2- hydroxyethylaspartamide) silica, and poly(N-isopropylacrylamide) grafted silica.

[0352] In some embodiments, the isolation and / or selection results in one or more populations of enriched T cells, e.g., CD3+ T cells, CD4+ T cells, and / or CD8+ T cells. In some embodiments, two or more separate population of enriched T cells are isolated, selected, enriched, or obtained from a single biological sample. In some embodiments, separate populations are isolated, selected, enriched, and / or obtained from separate biological samples collected, taken, and / or obtained from the same subject.

[0353] In certain embodiments, the isolation and / or selection results in one or more populations of enriched T cells that includes at least 60%, at least 65%, at least 70%, at least75%, at least 80%, at least 85%, at least 90%, at least 95%, at least 98%, at least 99%, at least 99.5%, at least 99.9%, or at or at about 100% CD3+ T cells. In particular embodiment, the population of enriched T cells consists essentially of CD3+ T cells.

[0354] In certain embodiments, the isolation and / or enrichment results in a populations of enriched CD4+ T cells that includes at least 60%, at least 65%, at least 70%, at least 75%, at least 80%, at least 85%, at least 90%, at least 95%, at least 98%, at least 99%, at least 99.5%, at least 99.9%, or at or at about 100% CD4+ T cells. In certain embodiments, the population of CD4+ T cells includes less than 40%, less than 35%, less than 30%, less than 25%, less than 20%, less than 15%, less than 10%, less than 5%, less than 1%, less than 0.1%, or less than 0.01% CD8+ T cells, and / or contains no CD8+ T cells, and / or is free or substantially free of CD8+ T cells. In some embodiments, the population of enriched T cells consists essentially of CD4+ T cells.

[0355] In certain embodiments, the isolation and / or enrichment results in a populations of enriched CD8+ T cells that includes at least 60%, at least 65%, at least 70%, at least 75%, at least 80%, at least 85%, at least 90%, at least 95%, at least 98%, at least 99%, at least 99.5%, at least 99.9%, or at or at about 100% CD8+ T cells. In certain embodiments, the population of CD8+ T cells contains less than 40%, less than 35%, less than 30%, less than 25%, less than 20%, less than 15%, less than 10%, less than 5%, less than 1%, less than 0.1%, or less than 0.01% CD4+ T cells, and / or contains no CD4+ T cells, and / or is free of or substantially free of CD4+ T cells. In some embodiments, the population of enriched T cells consists essentially of CD8+ T cells.

[0356] In some embodiments, the selection marker may be CD4 and the selection agent specifically binds CD4. In some aspects, the selection agent that specifically binds CD4 may be selected from the group consisting of an anti-CD4 antibody, a divalent antibody fragment of an anti-CD4 antibody, a monovalent antibody fragment of an anti-CD4 antibody, and a proteinaceous CD4 binding molecule with antibody-like binding properties. In some embodiments, an anti-CD4 antibody, divalent antibody fragment, or monovalent antibody fragment (e.g. anti-CD4 Fab fragment) can be derived from antibody 13B8.2 or a functionally active mutant of 13B8.2 that retains specific binding for CD4. For example, exemplary mutants of antibody 13B8.2 or ml3B8.2 are described in US-7482000, US-20140295458, US-10228312, and Bes et al., J Biol Chem (2003) 278:14265-14273. The mutant Fabfragment termed "m!3B8.2" carries the variable domain of the CD4 binding murine antibody 13B8.2 and a constant domain containing constant human CHI domain of type gamma for the heavy chain and the constant human light chain domain of type kappa, as described in US-7482000. In some embodiments, the anti-CD4 antibody, divalent antibody fragment, or monovalent antibody fragment, e.g., a mutant of antibody 13B8.2, contains the amino acid replacement H91A in the variable light chain, the amino acid replacement Y92A in the variable light chain, the amino acid replacement H35A in the variable heavy chain, and / or the amino acid replacement R53A in the variable heavy chain, each by Kabat numbering. In some aspects, compared to variable domains of the 13B8.2 Fab fragment in ml3B8.2, the His residue at position 91 of the light chain (position 93 in SEQ ID NO: 1) is mutated to Ala, and the Arg residue at position 53 of the heavy chain (position 55 in SEQ ID NO: 2) is mutated to Ala. In some embodiments, the reagent that is reversibly bound to the anti-CD4 antibody or fragment thereof is commercially available or derived from a reagent that is commercially available (e.g., catalog No. 6-8000-206, 6-8000-205, or 6-8002-100; IBA GmbH, Gottingen, Germany). In some embodiments, the selection agent comprises an anti-CD4 Fab fragment. In some embodiments, the anti-CD4 Fab fragment contains a variable heavy chain having the sequence set forth in SEQ ID NO: 2 and a variable light chain having the sequence set forth in SEQ ID NO: 1. In some embodiments, the anti-CD4 Fab fragment contains the CDRs of the variable heavy chain having the sequence set forth in SEQ ID NO: 2 and the CDRs of the variable light chain having the sequence set forth in SEQ ID NO: 1.

[0357] In some embodiments, the selection marker may be CD 8 and the selection agent specifically binds CD8. In some aspects, the selection agent that specifically binds CD8 may be selected from the group consisting of an anti-CD8 antibody, a divalent antibody fragment of an anti-CD8 antibody, a monovalent antibody fragment of an anti-CD8 antibody, and a proteinaceous CD8 binding molecule with antibody-like binding properties. In some embodiments, the anti-CD8 antibody, divalent antibody fragment, or monovalent antibody fragment (e.g., anti-CD8 Fab fragment) can be derived from antibody OKT8 (e.g., ATCC CRL-8014) or a functionally active mutant thereof that retains specific binding for CD8. In some embodiments, the reagent that is reversibly bound to anti-CD8 or a fragment thereof is commercially available or derived from a reagent that is commercially available (e.g., catalog No. 6-8003 or 6-8000-201; IBA GmbH, Gottingen, Germany). In some embodiments, theselection agent contains an anti-CD8 Fab fragment. In some embodiments, the anti-CD8 Fab fragment contains a variable heavy chain having the sequence set forth in SEQ ID NO: 3 and a variable light chain having the sequence set forth in SEQ ID NO: 4. In some embodiments, the anti-CD8 Fab fragment contains the CDRs of the variable heavy chain having the sequence set forth in SEQ ID NO: 3 and the CDRs of the variable light chain having the sequence set forth by SEQ ID NO: 4.

[0358] In some embodiments, the selection marker may be CD3 and the selection agent specifically binds CD3. In some aspects, the selection agent that specifically binds CD3 may be selected from the group consisting of an anti-CD3 antibody, a divalent antibody fragment of an anti-CD3 antibody, a monovalent antibody fragment of an anti-CD3 antibody, and a proteinaceous CD3 binding molecule with antibody-like binding properties. In some embodiments, the anti-CD3 antibody, divalent antibody fragment, or monovalent antibody fragment (e.g., anti-CD3 Fab fragment) can be derived from antibody OKT3 (e.g., ATCC CRL-8001; see, e.g., Stemberger et al., PLoS One (2012) 7(4):e35798) or a functionally active mutant thereof that retains specific binding for CD3. In some embodiments, the reagent that is reversibly bound to the anti-CD3 antibody or a fragment thereof is commercially available or derived from a reagent that is commercially available (e.g., catalog No. 6-8000-201 or 6-8001-100; IBA GmbH, Gottingen, Germany). In some embodiments, the selection agent contains an anti-CD3 Fab fragment. In some embodiments, the anti-CD3 Fab fragment contains a variable heavy chain having the sequence set forth in SEQ ID NO: 5 and a variable light chain having the sequence set forth in SEQ ID NO: 6. In some embodiments, the anti-CD3 Fab fragment contains the CDRs of the variable heavy chain having the sequence set forth in SEQ ID NO: 5 and the CDRs of the variable light chain having the sequence set forth in SEQ ID NO: 6.

[0359] In any of the above examples, the divalent antibody fragment may be an F(ab’)2 fragment or a divalent single-chain Fv fragment. In some embodiments, the monovalent antibody fragment may be selected from the group consisting of a Fab fragment, an Fv fragment, and a single-chain Fv fragment (scFv). In any of the above examples, the proteinaceous binding molecule with antibody-like binding properties may be an aptamer, a mutein based on a polypeptide of the lipocalin family, a glubody, a protein based on the ankyrin scaffold, a protein based on the crystalline scaffold, an adnectin, or an avimer.B. Stimulation

[0360] In some embodiments, the population of cells are incubated under T cell stimulating conditions. In some embodiments, the provided methods involve incubating the population of cells under T cell stimulating conditions.

[0361] Exemplary methods and stimulatory reagents for the stimulation of T cells are described in US-6040177, US-6352694, US-11400115, US-20190112576, US-20190136186, US-11274278, US-20210032297, US-20200354677, US-20200384025, US-20210163893, and US-20220002669.

[0362] In some embodiments, the conditions for T cell stimulation can include one or more of particular media, temperature, oxygen content, carbon dioxide content, time, agents, e.g., nutrients, amino acids, antibiotics, ions, and / or stimulatory factors, such as cytokines, chemokines, antigens, binding partners, fusion proteins, recombinant soluble receptors, and any other agents, designed to stimulate the T cells.

[0363] In some embodiments, the incubation is in basal media. In some embodiments, the basal media is serum-free. In some embodiments, the basal media is free of serum derived from human. In some embodiments, the basal media contains a mixture of inorganic salts, sugars, amino acids, and, optionally, vitamins, organic acids, and / or buffers or other well known cell culture nutrients. In addition to nutrients, the basal media can also help maintain pH and osmolality. A wide variety of commercially available basal media are well known to those skilled in the art and include Dulbeccos' Modified Eagles Medium (DMEM), Roswell Park Memorial Institute Medium (RPMI), Iscove modified Dulbeccos' medium, and Hams medium. In some embodiments, the basal media is Iscove's Modified Dulbecco's Medium, RPMI- 1640, or a-MEM.

[0364] In some embodiments, the basal media is a balanced salt solution (e.g., PBS, DPBS, HBSS, or EBSS). In some embodiments, the basal media is selected from Dulbecco's Modified Eagle's Medium (DMEM), Minimal Essential Medium (MEM), Basal Medium Eagle (BME), F-10, F-12, RPMI 1640, Glasgow's Minimal Essential Medium (GMEM), alpha Minimal Essential Medium (alpha MEM), Iscove's Modified Dulbecco's Medium, and M199. In some embodiments, the base media is a complex medium (e.g., RPMI-1640 or IMDM). In some embodiments, the base media is OpTmizer™ CTS™ T-Cell Expansion Basal Medium (ThermoFisher).

[0365] In some embodiments, the basal media is supplemented with additional additives. In some embodiments, the basal media is not supplemented with any additional additives. Additives to cell culture media include nutrients, sugars, e.g., glucose, amino acids, vitamins, and additives such as ATP and NADH.

[0366] In some embodiments, the incubation is in a serum- free medium, such as one described in US-20210207080.

[0367] In some embodiments, the incubation is in the presence of one or more recombinant cytokines. In some embodiments, the one or more recombinant cytokines are human recombinant cytokines. In certain embodiments, the one or more recombinant cytokines bind to and / or are capable of binding to receptors that are expressed by and / or are endogenous to T cells. In particular embodiments, the one or more recombinant cytokines include a member of the 4-alpha-helix bundle family of cytokines. Members of the 4-alpha-helix bundle family of cytokines include interleukin-2 (IL-2), interleukin-4 (IL-4), interleukin-7 (IL-7), interleukin-9 (IL-9), interleukin 12 (IL-12), interleukin 15 (IL-15), granulocyte colony-stimulating factor (G-CSF), and granulocyte-macrophage colony-stimulating factor (GM-CSF). In some embodiments, the one or more recombinant cytokines are selected from IL-2, IL- 15, and IL- 7. In some embodiments, the incubation is in the presence of IL-2, IL- 15, and IL-7.

[0368] In certain embodiments, the amount or concentration of the one or more recombinant cytokines are measured and / or quantified with International Units (IU). International units may be used to quantify vitamins, hormones, cytokines, vaccines, blood products, and similar biologically active substances. In some embodiments, IU are or include units of measure of the potency of biological preparations by comparison to an international reference standard of a specific weight and strength, e.g., WHO 1st International Standard for Human IL-2, 86 / 504. International Units are the only recognized and standardized method to report biological activity units that are published and are derived from an international collaborative research effort. In particular embodiments, the IU for population, sample, or source of a cytokine may be obtained through product comparison testing with an analogous WHO standard product. For example, in some embodiments, the lU / mg of a population, sample, or source of human recombinant IL-2, IL-7, or IL-15 is compared to the WHO standard IL-2 product (NIBSC code: 86 / 500), the WHO standard IL- 17 product (NIBSC code: 90 / 530), and the WHO standard IL- 15 product (NIBSC code: 95 / 554), respectively.

[0369] In some embodiments, the biological activity in lU / mg is equivalent to (ED50 in ng / mL)-l x 106. In particular embodiments, the ED50 of recombinant human IL-2 or IL-15 is equivalent to the concentration required for the half-maximal stimulation of cell proliferation (XTT cleavage) with CTLL-2 cells. In certain embodiments, the ED50 of recombinant human IL-7 is equivalent to the concentration required for the half-maximal stimulation for proliferation of PHA-activated human peripheral blood lymphocytes. Details relating to assays and calculations of IU for IL-2 are discussed in Wadhwa et al., Journal of Immunological Methods (2013) 379 (1-2): 1-7 and Gearing and Thorpe, Journal of Immunological Methods (1988) 114 (l-2):3-9. Details relating to assays and calculations of IU for IL-15 are discussed in Soman et al., Journal of Immunological Methods (2009) 348 (1- 2):83-94.

[0370] In some embodiments, the cells are stimulated or subjected to stimulation in the presence of a recombinant cytokine, e.g., a recombinant human cytokine, at a concentration of between 1 lU / mL and 1,000 lU / mL, between 10 lU / mL and 50 lU / mL, between 50 lU / mL and 100 lU / mL, between 100 lU / mL and 200 lU / mL, between 100 lU / mL and 500 lU / mL, between 250 lU / mL and 500 lU / mL, or between 500 lU / mL and 1,000 lU / mL.

[0371] In some embodiments, the cells are stimulated or subjected to stimulation in the presence of recombinant IL-2, e.g., human recombinant IL-2, at a concentration between 1 lU / mL and 500 lU / mL, between 10 lU / mL and 250 lU / mL, between 50 lU / mL and 200 lU / mL, between 50 lU / mL and 150 lU / mL, between 75 lU / mL and 125 lU / mL, between 100 lU / mL and 200 lU / mL, or between 10 lU / mL and 100 lU / mL. In particular embodiments, cells are stimulated or subjected to stimulation in the presence of recombinant IL-2 at a concentration at or at about 50 lU / mL, 60 lU / mL, 70 lU / mL, 80 lU / mL, 90 lU / mL, 100 lU / mL, 110 lU / mL, 120 lU / mL, 130 lU / mL, 140 lU / mL, 150 lU / mL, 160 lU / mL, 170 lU / mL, 180 lU / mL, 190 lU / mL, or 100 lU / mL. In some embodiments, the cells are stimulated or subjected to stimulation in the presence of or of about 100 lU / mL of recombinant IL-2, e.g., human recombinant IL-2.

[0372] In some embodiments, the cells are stimulated or subjected to stimulation in the presence of recombinant IL-7, e.g., human recombinant IL-7, at a concentration between 100 lU / mL and 2,000 lU / mL, between 500 lU / mL and 1,000 lU / mL, between 100 lU / mL and 500 lU / mL, between 500 lU / mL and 750 lU / mL, between 750 lU / mL and 1,000 lU / mL, orbetween 550 lU / mL and 650 lU / mL. In particular embodiments, the cells are stimulated or subjected to stimulation in the presence of IL-7 at a concentration at or at about 50 IU / mL,100 lU / mL, 150 lU / mL, 200 lU / mL, 250 lU / mL, 300 lU / mL, 350 lU / mL, 400 lU / mL, 450 lU / mL, 500 lU / mL, 550 lU / mL, 600 lU / mL, 650 lU / mL, 700 lU / mL, 750 lU / mL, 800 lU / mL, 750 lU / mL, 750 lU / mL, 750 lU / mL, or 1,000 lU / mL. In particular embodiments, the cells are stimulated or subjected to stimulation in the presence of or of about 600 lU / mL of recombinant IL-7, e.g., human recombinant IL-7.

[0373] In some embodiments, the cells are stimulated or subjected to stimulation in the presence of recombinant IL-15, e.g., human recombinant IL-15, at a concentration between 1 lU / mL and 500 lU / mL, between 10 lU / mL and 250 lU / mL, between 50 lU / mL and 200 lU / mL, between 50 lU / mL and 150 lU / mL, between 75 lU / mL and 125 lU / mL, between 100 lU / mL and 200 lU / mL, or between 10 lU / mL and 100 lU / mL. In particular embodiments, cells are stimulated or subjected to stimulation in the presence of recombinant IL- 15 at a concentration at or at about 50 lU / mL, 60 lU / mL, 70 lU / mL, 80 lU / mL, 90 lU / mL, 100 lU / mL, 110 lU / mL, 120 lU / mL, 130 lU / mL, 140 lU / mL, 150 lU / mL, 160 lU / mL, 170 lU / mL, 180 lU / mL, 190 lU / mL, or 200 lU / mL. In some embodiments, the cells are stimulated or subjected to stimulation in the presence of or of about 100 lU / mL of recombinant IL- 15, e.g., human recombinant IL- 15.

[0374] In some embodiments, the incubation is in the absence of recombinant cytokines.

[0375] In certain embodiments, the stimulation is performed under static conditions, such as conditions that do not involve centrifugation, shaking, rotating, rocking, or perfusion, e.g., continuous or semi-continuous perfusion of the media. In some embodiments, either prior to or shortly after, e.g., within 5, 15, or 30 minutes, of the initiation of the stimulation, the cells are transferred (e.g., transferred under sterile conditions) to a container such as a bag or vial, and placed in an incubator. In particular embodiments, the incubator is set at, at about, or at least 16°C, 24°C, or 35°C. In some embodiments, the incubator is set at 37°C, at about 37°C, or at 37°C ±2°C, ±1°C, ±0.5°C, or ±0.1 °C. In particular embodiments, the stimulation under static condition is performed in a cell culture bag placed in an incubator. In some embodiments, the culture bag is composed of a single-web polyolefin gas permeable film which enables monocytes, if present, to adhere to the bag surface.

[0376] In some embodiments, the T cell stimulating conditions involve incubation in the presence of T cell stimulatory agents. In some embodiments, the T cell stimulatory agents bind to molecules expressed on the surface of T cells. In some embodiments, a T cell stimulatory agent of the T cell stimulatory agents induces a primary activation signal in T cells. In some embodiments, a T cell stimulatory agent of the T cell stimulatory agents induces a costimulatory signal in T cells. In some embodiments, the T cell stimulatory agents induce a primary activation signal and a costimulatory signal in T cells.

[0377] In some embodiments, the T cell stimulatory agent that induces a primary activation signal binds to a member of a TCR / CD3 complex in T cells. In some embodiments, the T cell stimulatory agent binds to CD3. In some embodiments, the T cell stimulatory agent is an anti- CD3 antibody, a divalent antibody fragment of an anti-CD3 antibody, a monovalent antibody fragment of an anti-CD3 antibody, or a proteinaceous CD3 binding molecule with antibodylike binding properties. In some embodiments, the T cell stimulatory agent is an anti-CD3 antibody or antibody fragment. In some embodiments, the anti-CD3 antibody, divalent antibody fragment of an anti-CD3 antibody, or monovalent antibody fragment of an anti-CD3 antibody (e.g., anti-CD3 Fab fragment) is derived from antibody OKT3 (e.g., ATCC CRL- 8001; see, e.g., Stemberger et al., PLoS One (2012) 7(4):e35798) or a functionally active mutant thereof that retains specific binding for CD 3. In some embodiments, the T cell stimulatory agent is an anti-CD3 Fab. In some embodiments, the anti-CD3 Fab contains a variable heavy chain having the sequence set forth in SEQ ID NO: 5 and a variable light chain having the sequence set forth in SEQ ID NO: 6. In some embodiments, the anti-CD3 Fab contains the CDRs of the variable heavy chain having the sequence set forth in SEQ ID NO: 5 and the CDRs of the variable light chain having the sequence set forth in SEQ ID NO: 6.

[0378] In some embodiments, the T cell stimulatory agent that induces a costimulatory signal binds to a costimulatory molecule in T cells. In some embodiments, the costimulatory molecule is CD28, CD90 (Thy-1), CD95 (Apo- / Fas), CD137 (4-1BB), CD154 (CD40L), ICOS, LAT, CD27, 0X40, or HVEM.

[0379] In some embodiments, the costimulatory molecule is CD28. In some embodiments, the T cell stimulatory agent is an anti-CD28 antibody, a divalent antibody fragment of an anti-CD28 antibody, a monovalent antibody fragment of an anti-CD28 antibody, or aproteinaceous CD28 binding molecule with antibody-like binding properties. In some embodiments, the T cell stimulatory agent is an anti-CD28 antibody or antibody fragment. In some embodiments, the anti-CD28 antibody, divalent antibody fragment of an anti-CD28 antibody, or monovalent antibody fragment of an anti-CD28 antibody (e.g., anti-CD28 Fab fragment) is derived from antibody CD28.3 (deposited as a synthetic single chain Fv construct under GenBank Accession No. AF451974.1; see also Vanhove et al., Blood (2003) 102(2):564-570), the variable heavy and light chains of which contain the amino acid sequences set forth in SEQ ID NO: 7 and 8, respectively. In some embodiments, the T cell stimulatory agent is an anti-CD28 Fab. In some embodiments, the anti-CD28 Fab contains a variable heavy chain having the sequence set forth in SEQ ID NO: 7 and a variable light chain having the sequence set forth in SEQ ID NO: 8. In some embodiments, the anti-CD28 Fab contains the CDRs of the variable heavy chain having the sequence set forth in SEQ ID NO: 7 and the CDRs of the variable light chain having the sequence set forth in SEQ ID NO: 8.

[0380] In some embodiments, the costimulatory molecule is CD90. In some embodiments, the T cell stimulatory agent is an anti-CD90 antibody, a divalent antibody fragment of an anti-CD90 antibody, a monovalent antibody fragment of an anti-CD90 antibody, or a proteinaceous CD90 binding molecule with antibody-like binding properties. In some embodiments, the T cell stimulatory agent is an anti-CD90 antibody or antibody fragment. In some embodiments, the T cell stimulatory agent is an anti-CD90 Fab. In some embodiments, the anti-CD90 antibody, divalent antibody fragment of an anti-CD90 antibody, or monovalent antibody fragment of an anti-CD90 antibody (e.g., anti-CD90 Fab fragment) is derived from the anti-CD90 antibody G7 (Biolegend, cat. no. 105201).

[0381] In some embodiments, the costimulatory molecule is CD95. In some embodiments, the T cell stimulatory agent is an anti-CD95 antibody, a divalent antibody fragment of an anti-CD95 antibody, a monovalent antibody fragment of an anti-CD95 antibody, or a proteinaceous CD95 binding molecule with antibody-like binding properties. In some embodiments, the T cell stimulatory agent is an anti-CD28 antibody or antibody fragment. In some embodiments, the T cell stimulatory agent is an anti-CD95 Fab. In some embodiments, the anti-CD95 antibody, divalent antibody fragment of an anti-CD95 antibody, or monovalent antibody fragment of an anti-CD95 antibody (e.g., anti-CD95 Fab fragment) is derived frommonoclonal mouse anti-human CD95 CH11 (Upstate Biotechnology, Lake Placid, NY), anti- CD95 mAh 7C11, or anti-APO-1, such as described in Paulsen et al., Cell Death & Differentiation (2011) 18(4):619-631.

[0382] In some embodiments, the costimulatory molecule is CD137. In some embodiments, the T cell stimulatory agent is an anti-CD137 antibody, a divalent antibody fragment of an anti-CD137 antibody, a monovalent antibody fragment of an anti-CD137 antibody, or a proteinaceous CD 137 binding molecule with antibody-like binding properties. In some embodiments, the T cell stimulatory agent is an anti-CD137 antibody or antibody fragment. In some embodiments, the T cell stimulatory agent is an anti-CD137 Fab. In some embodiments, the anti-CD137 antibody, divalent antibody fragment of an anti-CD137 antibody, or monovalent antibody fragment of an anti-CD137 antibody (e.g., anti-CD137 Fab fragment) is derived from LOB 12, IgG2a, or LOB 12.3, IgGl as described in Taraban et al., Eur J Immunol. (2002) 32(12):3617-27. See also, e.g., US-6569997, US-6303121, and Mittler et al., Immunol Res. (2004) 29(1-3): 197-208.

[0383] In some embodiments, the costimulatory molecule is CD40. In some embodiments, the T cell stimulatory agent is an anti-CD40 antibody, a divalent antibody fragment of an anti-CD40 antibody, a monovalent antibody fragment of an anti-CD40 antibody, or a proteinaceous CD40 binding molecule with antibody-like binding properties. In some embodiments, the T cell stimulatory agent is an anti-CD40 antibody or antibody fragment. In some embodiments, the T cell stimulatory agent is an anti-CD40 Fab.

[0384] In some embodiments, the costimulatory molecule is CD40L. In some embodiments, the T cell stimulatory agent is an anti-CD40L antibody, a divalent antibody fragment of an anti-CD40L antibody, a monovalent antibody fragment of an anti-CD40L antibody, or a proteinaceous CD40L binding molecule with antibody-like binding properties. In some embodiments, the T cell stimulatory agent is an anti-CD40L antibody or antibody fragment. In some embodiments, the T cell stimulatory agent is an anti-CD40L Fab. In some embodiments, the anti-CD40L antibody, divalent antibody fragment of an anti-CD40L antibody, or monovalent antibody fragment of an anti-CD40L antibody (e.g., anti-CD40L Fab fragment) is derived from Hu5C8, as described in Blair et al., JEM (2000) 19(4):651- 660. See also, e.g., US-7563445, US20010026932, US7547438, and US-7172759.

[0385] In some embodiments, the costimulatory molecule is ICOS. In some embodiments, the T cell stimulatory agent is an anti-ICOS antibody, a divalent antibody fragment of an anti- ICOS antibody, a monovalent antibody fragment of an anti-ICOS antibody, or a proteinaceous ICOS binding molecule with antibody-like binding properties. In some embodiments, the T cell stimulatory agent is an anti-ICOS antibody or antibody fragment. In some embodiments, the T cell stimulatory agent is an anti-ICO Fab. In some embodiments, the anti-ICOS antibody, divalent antibody fragment of an anti-ICOS antibody, or monovalent antibody fragment of an anti-ICOS antibody (e.g., anti-ICOS Fab fragment) is derived from any of the antibodies described in US-20080279851 and Deng et al., Hybrid Hybridomics (2004) 23(3): 176-82.

[0386] In some embodiments, the costimulatory molecule is Linker for Activation of T cells (LAT). In some embodiments, the T cell stimulatory agent is an anti-LAT antibody, a divalent antibody fragment of an anti-LAT antibody, a monovalent antibody fragment of an anti-LAT antibody, or a proteinaceous LAT binding molecule with antibody-like binding properties. In some embodiments, the T cell stimulatory agent is an anti-LAT antibody or antibody fragment. In some embodiments, the T cell stimulatory agent is an anti-LAT Fab.

[0387] In some embodiments, the costimulatory molecule is CD27. In some embodiments, the T cell stimulatory agent is an anti-CD27 antibody, a divalent antibody fragment of an anti-CD27 antibody, a monovalent antibody fragment of an anti-CD27 antibody, or a proteinaceous CD27 binding molecule with antibody-like binding properties. In some embodiments, the T cell stimulatory agent is an anti-CD27 antibody or antibody fragment. In some embodiments, the T cell stimulatory agent is an anti-CD27 Fab. In some embodiments, the anti-CD27 antibody, divalent antibody fragment of an anti-CD27 antibody, or monovalent antibody fragment of an anti-CD27 antibody (e.g., anti-CD27 Fab fragment) is derived from any of the antibodies described in US-8481029.

[0388] In some embodiments, the costimulatory molecule is 0X40. In some embodiments, the T cell stimulatory agent is an anti-OX40 antibody, a divalent antibody fragment of an anti-OX40 antibody, a monovalent antibody fragment of an anti-OX40 antibody, or a proteinaceous 0X40 binding molecule with antibody-like binding properties. In some embodiments, the T cell stimulatory agent is an anti-OX40 antibody or antibody fragment. In some embodiments, the T cell stimulatory agent is an anti-OX40 Fab. In some embodiments,the anti-OX40 antibody, divalent antibody fragment of an anti-OX40 antibody, or monovalent antibody fragment of an anti-OX40 antibody (e.g., anti-OX40 Fab fragment) is derived from any of the antibodies described in US-9475880 and Melero et al., Clin Cancer Res. (2013) 19(5): 1044-53.

[0389] In some embodiments, the costimulatory molecule is HVEM. In some embodiments, the T cell stimulatory agent is an anti-HVEM antibody, a divalent antibody fragment of an anti-HVEM antibody, a monovalent antibody fragment of an anti-HVEM antibody, or a proteinaceous HVEM binding molecule with antibody-like binding properties. In some embodiments, the T cell stimulatory agent is an anti-HVEM antibody or antibody fragment. In some embodiments, the T cell stimulatory agent is an anti-HVEM Fab. In some embodiments, the anti-HVEM antibody, divalent antibody fragment of an anti-HVEM antibody, or monovalent antibody fragment of an anti-HVEM antibody (e.g., anti-HVEM Fab fragment) is derived from any of the antibodies described in WO-2006054961, US-8188232, and Park et al., Cancer Immunol Immunother. (2012) 61(2):203-14.

[0390] In some embodiments, the T cell stimulatory agents are immobilized on a solid support. In some embodiments, the solid support is a bead. In some embodiments, the bead is biocompatible, e.g., composed of a material that is suitable for biological use. In some embodiments, the beads are non-toxic to T cells.

[0391] In some embodiments, the bead has a diameter of greater than about 0.001 pm, greater than about 0.01 pm, greater than about 0.1 pm, greater than about 1.0 pm, greater than about 10 pm, greater than about 50 pm, greater than about 100 pm, or greater than about 1000 pm and no more than about 1500 pm. In some embodiments, the bead has a diameter of about 1.0 pm to about 500 pm, about 1.0 pm to about 150 pm, about 1.0 pm to about 30 pm, about 1.0 pm to about 10 pm, about 1.0 pm to about 5.0 pm, about 2.0 pm to about 5.0 pm, or about 3.0 pm to about 5.0 pm. In some embodiments, the bead has a diameter of about 3 pm to about 5 pm. In some embodiments, the bead has a diameter of at least or at least about or about 0.001 pm, 0.01 pm, 0.1pm, 0.5pm, 1.0 pm, 1.5 pm, 2.0 pm, 2.5 pm, 3.0 pm, 3.5 pm, 4.0 pm, 4.5 pm, 5.0 pm, 5.5 pm, 6.0 pm, 6.5 pm, 7.0 pm, 7.5 pm, 8.0 pm, 8.5 pm, 9.0 pm, 9.5 pm, 10 pm, 12 pm, 14 pm, 16 pm, 18 pm, or 20 pm. In certain embodiments, the bead has a diameter of or about 4.5 pm. In certain embodiments, the bead has a diameter of or about 2.8 pm.

[0392] In some embodiments, the bead has a density of greater than 0.001 g / cm3, greater than 0.01 g / cm3, greater than 0.05 g / cm3, greater than 0.1 g / cm3, greater than 0.5 g / cm3, greater than 0.6 g / cm3, greater than 0.7 g / cm3, greater than 0.8 g / cm3, greater than 0.9 g / cm3, greater than 1 g / cm3, greater than 1.1 g / cm3, greater than 1.2 g / cm3, greater than 1.3 g / cm3, greater than 1.4 g / cm3, greater than 1.5 g / cm3, greater than 2 g / cm3, greater than 3 g / cm3, greater than 4 g / cm3, or greater than 5g / cm3. In some embodiments, the bead has a density of between about 0.001 g / cm3and about 100 g / cm3, about 0.01 g / cm3and about 50 g / cm3, about 0.1 g / cm3and about 10 g / cm3, about 0.1 g / cm3and about .5 g / cm3, about 0.5 g / cm3and about 1 g / cm3, about 0.5 g / cm3and about 1.5 g / cm3, about 1 g / cm3and about 1.5 g / cm3, about 1 g / cm3and about 2 g / cm3, or about 1 g / cm3and about 5 g / cm3. In some embodiments, the bead has a density of about 0.5 g / cm3, about 0.5 g / cm3, about 0.6 g / cm3, about 0.7 g / cm3, about 0.8 g / cm3, about 0.9 g / cm3, about 1.0 g / cm3, about 1.1 g / cm3, about 1.2 g / cm3, about 1.3 g / cm3, about 1.4 g / cm3, about 1.5 g / cm3, about 1.6 g / cm3, about 1.7 g / cm3, about 1.8 g / cm3, about 1.9 g / cm3, or about 2.0 g / cm3. In certain embodiments, the bead has a density of about 1.6 g / cm3. In particular embodiments, the bead has a density of about 1.5 g / cm3. In certain embodiments, the bead has a density of about 1.3 g / cm3.

[0393] In some embodiments, a plurality of the beads has a uniform density. In some embodiments, a uniform density has a density standard deviation of less than 10%, less than 5%, or less than 1% of the mean bead density.

[0394] In some embodiments, the bead reacts in a magnetic field. In some embodiments, the bead is a magnetic bead. In some embodiments, the bead is paramagnetic. In particular embodiments, the bead is superparamagnetic. In certain embodiments, the bead does not display any magnetic properties unless it is exposed to a magnetic field.

[0395] In particular embodiments, the bead contains a magnetic core. In particular embodiments, the bead contains a paramagnetic core. In particular embodiments, the bead contains a superparamagnetic core. In some embodiments, the core contains a metal. In some embodiments, the metal can be iron, nickel, copper, cobalt, gadolinium, manganese, tantalum, zinc, zirconium, or any combinations thereof. In certain embodiments, the core contains metal oxides (e.g., iron oxides), ferrites (e.g., manganese ferrites, cobalt ferrites, and nickel ferrites), hematite, and / or metal alloys (e.g., CoTaZn). In some embodiments, the core contains one or more of a ferrite, a metal, a metal alloy, an iron oxide, and chromium dioxide.In some embodiments, the core contains elemental iron or a compound thereof. In some embodiments, the core contains one or more of magnetite ( FC O4), maghemite (yFc2O3), and greigite (Fe3S4). In some embodiments, the core contains an iron oxide (e.g., FCTCE).

[0396] In some embodiments, the bead contains at least one material at or near the bead surface that can be coupled, linked, or conjugated to an agent. In some embodiments, the T cell stimulatory agents are immobilized on the bead via this material. In some embodiments, the bead is surface functionalized, e.g., has functional groups that are capable of forming a covalent bond with a binding molecule, e.g., a polynucleotide or a polypeptide. In particular embodiments, the bead has surface-exposed carboxyl, amino, hydroxyl, tosyl, epoxy, and / or chloromethyl groups. In particular embodiments, the bead has surface-exposed agarose and / or sepharose. In some embodiments, the bead has surface-exposed protein A, protein G, or biotin.

[0397] In certain embodiments, the bead contains a magnetic, paramagnetic, and / or superparamagnetic core that is covered by a surface functionalized coat or coating. In some embodiments, the coat can contain a material that can include a polymer, a polysaccharide, a silica, a fatty acid, a protein, a carbon, agarose, sepharose, or a combination thereof. In some embodiments, the polymer can be a polyethylene glycol, poly (lactic-co-glycolic acid), polyglutaraldehyde, polyurethane, polystyrene, or a polyvinyl alcohol. In certain embodiments, the outer coat or coating comprises polystyrene. In particular embodiments, the outer coating is surface functionalized.

[0398] In some embodiments, the population of cells containing T cells is incubated with a stimulatory reagent at a ratio of beads to cells at or at about 3:1, 2.5:1, 2:1, 1.5:1, 1.25:1, 1.2:1, 1.1:1, 1:1, 0.9:1, 0.8:1, 0.75:1, 0.67:1, 0.5:1, 0.3:1, or 0.2:1. In particular embodiments, the ratio of beads to cells is between 2.5:1 and 0.2:1, between 2:1 and 0.5:1, between 1.5:1 and 0.75:1, between 1.25:1 and 0.8:1, or between 1.1:1 and 0.9:1. In particular embodiments, the ratio of beads to cells is about 1:1 or is 1:1.

[0399] In some embodiments, the T cell stimulatory agents are not immobilized on a solid support, e.g., not immobilized on a bead. In some embodiments, the T cell stimulatory agents are part of a stimulatory reagent that is in soluble form. Exemplary T cell stimulatory reagents in soluble form are described in US2021 / 0032297 (see also Poltorak et al., Scientific Reports (2020)).

[0400] In some embodiments, the T cell stimulatory agents are immobilized on an oligomeric streptavidin mutein reagent. In some embodiments, the T cell stimulatory agents are reversibly immobilized on the oligomeric streptavidin mutein reagent. In some embodiments, the oligomeric streptavidin mutein reagent is an oligomer or polymer of a streptavidin mutein.

[0401] In some embodiments, the T cell stimulatory agents include binding partners that are reversibly bound to a streptavidin mutein molecule of the oligomeric streptavidin mutein reagent. In some embodiments, the binding partners are fused to the C-terminus of a heavy chain of the T cell stimulatory agents.

[0402] In some embodiments, the binding partners reversibly bind to a biotin-binding site of the streptavidin mutein. In some embodiments, the binding of the binding partners to the streptavidin mutein is disrupted by the presence of biotin. In some embodiments, the biotin is D-biotin.

[0403] In some embodiments, one or both of the binding partners is biotin, a biotin analog, or a streptavidin-binding peptide. In some embodiments, one or both of the binding partners is a streptavidin-binding peptide. In some embodiments, the streptavidin-binding peptide comprises the sequence set forth in any of SEQ ID NO: 9-15. In some embodiments, the sequence of the streptavidin-binding peptide is set forth in any of SEQ ID NO: 9-15. In some embodiments, the streptavidin-binding peptide comprises the sequence set forth in SEQ ID NO: 15. In some embodiments, the sequence of the streptavidin-binding peptide is set forth in SEQ ID NO: 15.

[0404] In some embodiments, the oligomeric streptavidin mutein reagent comprises between or between about 2,000 and 3,000 tetramers of the streptavidin mutein. In some embodiments, the oligomeric streptavidin mutein reagent comprises about 2,400 tetramers of the streptavidin mutein.

[0405] In some embodiments, the oligomeric streptavidin mutein reagent has a radius of between or between about 90 and 110 nm. In some embodiments, the oligomeric streptavidin mutein reagent has a radius of about 100 nm. In some embodiments, the radius is a hydrodynamic radius.

[0406] In some embodiments, individual molecules of the oligomeric streptavidin mutein reagent are crosslinked by a bifunctional linker. In some embodiments, the bifunctional linkeris a heterobifunctional linker. In some embodiments, the bifunctional linker is an amine-to- thiol linker.

[0407] In some embodiments, the streptavidin mutein reversibly binds to biotin, a biotin analog, or a streptavidin-binding peptide. In some embodiments, the streptavidin mutein reversibly binds to a streptavidin-binding peptide. In some embodiments, the streptavidin- binding peptide comprises the sequence set forth in any of SEQ ID NO: 9-15. In some embodiments, the sequence of the streptavidin-binding peptide is set forth in any of SEQ ID NO: 9-15.

[0408] In some embodiments, the streptavidin mutein comprises one or more mutations compared to wild-type streptavidin. In some embodiments, the streptavidin mutein comprises one or more mutations compared to the sequence of amino acids set forth in SEQ ID NO: 16.

[0409] In some embodiments, the streptavidin mutein comprises one or more mutations compared to a minimal streptavidin. In some embodiments, the minimal streptavidin begins N-terminally in the region of the amino acid positions 10 to 16 and terminates C-terminally in the region of the amino acid positions 133 to 142 compared to the sequence set forth in SEQ ID NO: 16. In some embodiments, the streptavidin mutein begins N-terminally in the region of the amino acid positions 10 to 16 and terminates C-terminally in the region of the amino acid positions 133 to 142 compared to the sequence set forth in SEQ ID NO: 16.

[0410] In some embodiments, the sequence of the minimal streptavidin is from position Alal3 to Serl39 of the sequence of amino acids set forth in SEQ ID NO: 16. In some embodiments, the minimal streptavidin has an N-terminal methionine residue instead of Alal3.

[0411] In some embodiments, the sequence of the minimal streptavidin is set forth in SEQ ID NO: 17. In some embodiments, the streptavidin mutein comprises one or more mutations compared to the sequence of amino acids set forth in SEQ ID NO: 17.

[0412] In some embodiments, the sequence of the minimal streptavidin is set forth in SEQ ID NO: 18. In some embodiments, the streptavidin mutein comprises one or more mutations compared to the sequence of amino acids set forth in SEQ ID NO: 18.

[0413] In some embodiments, the streptavidin mutein comprises the amino acid sequence Val44-Thr45-Ala46-Arg47 (SEQ ID NO: 19) at sequence positions corresponding to positions 44 to 47 of the sequence of amino acids set forth in SEQ ID NO: 16. In someembodiments, the streptavidin mutein comprises the amino acid sequence Ile44-Gly45- Ala46-Arg47 (SEQ ID NO: 20) at sequence positions corresponding to positions 44 to 47 of the sequence of amino acids set forth in SEQ ID NO: 16.

[0414] In some embodiments, the streptavidin mutein further comprises the amino acid replacements Glut 17, Glyl20, and Try 121 at sequence positions corresponding to positions of the sequence of amino acids set forth in SEQ ID NO: 16.

[0415] In some embodiments, the streptavidin mutein comprises the sequence of amino acids set forth in any of SEQ ID NO: 21-28. In some embodiments, the streptavidin mutein comprises the sequence of amino acids set forth in SEQ ID NO: 21. In some embodiments, the sequence of the streptavidin mutein is set forth in SEQ ID NO: 21.

[0416] In some embodiments, the incubation under T cell stimulating conditions is performed for at least 12 hours. In some embodiments, the incubation under T cell stimulating conditions is performed for, for about, or for less than, 48 hours, 42 hours, 36 hours, 30 hours, 24 hours, 22 hours, 20 hours, 18 hours, 16 hours, or 12 hours. In some embodiments, the incubation is performed for between or between about 16 hours and 24 hours. In particular embodiments, the incubation is for between or between about 12 hours and 36 hours, 18 hours and 30 hours, or for or for about 24 hours. In some embodiments, the incubation is performed for, for about, or for less than 2 days. In some embodiments, the incubation is performed for, for about, or for less than one day.C. Genetic Engineering

[0417] In some embodiments, the population of cells are genetically engineered to express a recombinant protein. In some embodiments, the provided methods involve genetically engineering the population of cells to express a recombinant protein. In some embodiments, a heterologous or recombinant polynucleotide encoding the recombinant protein is introduced into cells of the population of cells. Any method of introducing a heterologous or recombinant polynucleotide that would result in integration of the polynucleotide encoding the recombinant protein into the genome of a cell such as a T cell may be used, including viral and non-viral methods of genetic engineering. Introduction of the polynucleotides, e.g., heterologous or recombinant polynucleotides, encoding the recombinant protein into the cell may be carried out using any of a number of known vectors. Exemplary vectors are described in Section II-C-1. Such vectors include viral, including lentiviral and gammaretroviral,systems. Exemplary methods include those for transfer of heterologous polynucleotides encoding the recombinant proteins, including via viral, e.g., retroviral or lentiviral, transduction.

[0418] Exemplary methods for the genetic engineering of cells are described in US- 11400115, US-20190112576, US-20190136186, US-11274278, US-20210032297, US- 20200354677, US-20200384025, US-20210163893, and US-20220002669.

[0419] In some embodiments, the heterologous or recombinant polynucleotide encoding the recombinant protein is introduced using a non-viral method, such as electroporation, calcium phosphate transfection, protoplast fusion, cationic liposome-mediated transfection, nanoparticles such as lipid nanoparticles, tungsten particle-facilitated microparticle bombardment, strontium phosphate DNA co-precipitation, or other approaches described in, e.g., US-10654928 and US-7446190. Transposon-based systems also are contemplated.

[0420] In particular embodiments, the cells are genetically engineered, transformed, or transduced after the cells have been stimulated, such as by any of the methods described herein, e.g., in Section II-B. In particular embodiments, the cells are genetically engineered, transformed, or transduced at, at about, or within 72 hours, 60 hours, 48 hours, 36 hours, 24 hours, or 12 hours, inclusive, from the initiation of the stimulation. In particular embodiments, the cells are genetically engineered, transformed, or transduced at, at about, or within 3 days, two days, or one day, inclusive, from the initiation of the stimulation. In certain embodiments, the cells are genetically engineered, transformed, or transduced between or between about 12 hours and 48 hours, 16 hours and 36 hours, or 18 hours and 30 hours after the initiation of the stimulation. In particular embodiments, the cells are genetically engineered, transformed, or transduced between or between about 18 hours and 30 hours after the initiation of the stimulation. In particular embodiments, the cells are genetically engineered, transformed, or transduced at or at about 16 hours, 18 hours, 20 hours, 22 hours, or 24 hours after the initiation of the stimulation.

[0421] In some embodiments, the engineering, e.g., transduction, is performed for between 24 and 48 hours, between 36 and 12 hours, between 18 and 30 hours, or for or for about 24 hours. In some embodiments, the engineering, e.g., transduction, is performed for or for about 24 hours, 48 hours, or 72 hours, or for or for about 1 day, 2 days, or 3 days, respectively. In particular embodiments, the engineering, e.g., transduction, is performed foror for about 24 hours ± 6 hours, 48 hours ± 6 hours, or 72 hours ± 6 hours. In particular embodiments, the engineering, e.g., transduction, is performed for or for about 72 hours, 72 ± 4 hours, or for or for about 3 days.

[0422] In certain embodiments, methods for genetic engineering are carried out by contacting or introducing one or more cells of a population with a nucleic acid molecule or polynucleotide encoding the recombinant protein. In certain embodiments, the nucleic acid molecule or polynucleotide is heterologous to the cells. In particular embodiments, the heterologous nucleic acid molecule or heterologous polynucleotide is not native to the cells. In certain embodiments, the heterologous nucleic acid molecule or heterologous polynucleotide encodes a protein, e.g., a recombinant protein, that is not natively expressed by the cell. In particular embodiments, the heterologous nucleic acid molecule or polynucleotide is or contains a nucleic acid sequence that is not found in the cell prior to the contact or introduction.

[0423] In some embodiments, the cells are engineered, e.g., transduced, in the presence of a transduction adjuvant. Exemplary transduction adjuvants include polycations, fibronectin or fibronectin-derived fragments or variants, and RetroNectin. In certain embodiments, the cells are engineered in the presence of polycations, fibronectin or fibronectin-derived fragments or variants, and / or RetroNectin. In particular embodiments, the cells are engineered in the presence of a polycation that is polybrene, DEAE-dextran, protamine sulfate, poly-L-lysine, or a cationic liposome. In particular embodiments, the cells are engineered in the presence of protamine sulfate.

[0424] In some embodiments, the genetic engineering, e.g., transduction, is carried out in any of the media described in Section II-B. In some embodiments, the genetic engineering, e.g., transduction, is carried out in serum free media, e.g, any as described in US- 20210207080.

[0425] In some embodiments, the genetic engineering, e.g., transduction, is carried out in the presence of one or more recombinant cytokines. In some embodiments, the genetic engineering, e.g., transduction, is carried out in the presence of any of the recombinant cytokines described in Section II-B and at any of the concentrations described in Section II- B.

[0426] In some embodiments, the cells are genetically engineered, transformed, or transduced in the presence of the same or similar media as was present during the stimulation. In some embodiments, the cells are genetically engineered, transformed, or transduced in media having the same cytokines as the media present during stimulation. In certain embodiments, the cells are genetically engineered, transformed, or transduced, in media having the same cytokines at the same concentrations as the media present during stimulation.

[0427] In some embodiments, genetically engineering the cells is or includes introducing the polynucleotide, e.g., the heterologous or recombinant polynucleotide, into the cells by transduction. In some embodiments, the cells are transduced or subjected to transduction with a viral vector. In particular embodiments, the cells are transduced or subjected to transduction with a viral vector. In some embodiments, the virus is a retroviral vector, such as a gammaretroviral vector or a lentiviral vector. Methods of lentiviral transduction are known. Exemplary methods are described in, e.g., Wang et al., J. Immunother. (2012) 35(9):689-701; Cooper et al., Blood (2003) 101:1637-1644; Verhoeyen et al., Methods Mol Biol. (2009) 506: 97-114; and Cavalieri et al., Blood. (2003) 102(2):497-505.

[0428] In some embodiments, the transduction is carried out by contacting one or more cells of a population with a nucleic acid molecule encoding the recombinant protein. In some embodiments, the contacting can be effected with centrifugation, such as spinoculation (e.g., centrifugal inoculation). Such methods include any of those as described in US- 10428351. Exemplary centrifugal chambers include those produced and sold by Biosafe SA, including those for use with the Sepax® and Sepax® 2 system, including an A-200 / F and A-200 centrifugal chambers and various kits for use with such systems. Exemplary chambers, systems, and processing instrumentation and cabinets are described, for example, in US- 6123655, US-6733433, US-20080171951, and US-US6733433. Exemplary kits for use with such systems include single-use kits sold by BioSafe SA under product names CS-430.1, CS- 490.1, CS-600.1 and CS-900.2.

[0429] In particular embodiments, genetic engineering, such as by transforming (e.g., transducing) the cells with a viral vector, further includes one or more steps of incubating the cells after the introducing or contacting of the cells with the viral vector. In some embodiments, cells, e.g., cells of the transformed cell population (also called “transformedcells”), are incubated subsequent to processes for genetically engineering, transforming, transducing, or transfecting the cells to introduce the viral vector into the cells.

[0430] In some embodiments, the cells, e.g. transformed cells, are incubated after the introducing of the heterologous or recombinant polynucleotide, e.g., viral vector particles, is carried out without further processing of the cells. In particular embodiments, prior to the incubating, the cells are washed, such as to remove or substantially remove exogenous or remaining polynucleotides encoding the heterologous or recombinant polynucleotide, e.g. viral vector particles, such as those remaining in the media after the genetic engineering process following the spinoculation.

[0431] In some embodiments, the further incubation is effected under conditions to result in integration of the viral vector into a host genome of one or more of the cells. For example, the further incubation provides time for the viral vector that may be bound to the T cells following transduction, e.g., via spinoculation, to integrate within the genome of the cell to delivery the gene of interest. In some aspects, the further incubation is carried out under conditions to allow the cells, e.g. transformed cells, to rest or recover in which the culture of the cells during the incubation supports or maintains the health of the cells. In particular embodiments, the cells are incubated under static conditions, such as conditions that do not involve centrifugation, shaking, rotating, rocking, or perfusion, e.g., continuous or semi- continuous perfusion of the media.

[0432] It is within the level of a skilled artisan to assess or determine if the incubation has resulted in integration of viral vector particles into a host genome, and hence to empirically determine the conditions for a further incubation. In some embodiments, integration of a viral vector into a host genome can be assessed by measuring the level of expression of a recombinant protein, such as a heterologous protein, encoded by a nucleic acid contained in the genome of the viral vector particle following incubation. A number of well-known methods for assessing expression level of recombinant molecules may be used, such as detection by affinity-based methods, e.g., immunoaffinity-based methods, e.g., in the context of cell surface proteins, such as by flow cytometry. In some examples, the expression is measured by detection of a transduction marker and / or reporter construct. In some embodiments, nucleic acid encoding a truncated surface protein is included within the vector and used as a marker of expression and / or enhancement thereof.

[0433] In certain embodiments, the incubation is performed under static conditions, such as conditions that do not involve centrifugation, shaking, rotating, rocking, or perfusion, e.g., continuous or semi-continuous perfusion of the media. In some embodiments, either prior to or shortly after, e.g., within 5, 15, or 30 minutes, the initiation of the incubation, the cells are transferred (e.g., transferred under sterile conditions) to a container such as a bag or vial, and placed in an incubator. In some embodiments, the cells are transferred into the container under closed or sterile conditions. In some embodiments, the container, e.g., the vial or bag, is then placed into an incubator for all or a portion of the incubation. In particular embodiments, incubator is set at, at about, or at least 16°C, 24°C, or 35°C. In some embodiments, the incubator is set at 37°C, at about 37°C, or at 37°C ±2°C, ±1°C, ±0.5°C, or ±0.1 °C.

[0434] In some embodiments, the incubation is performed in serum free media. In some embodiments, the serum free media is a defined and / or well-defined cell culture media. In certain embodiments, the serum free media is a controlled culture media that has been processed, e.g., filtered to remove inhibitors and / or growth factors. In some embodiments, the serum free media contains proteins. In certain embodiments, the serum-free media may contain serum albumin, hydrolysates, growth factors, hormones, carrier proteins, and / or attachment factors.

[0435] In some embodiments, the further incubation is carried out in any of the media described in Section II-B. In some embodiments, the further incubation is carried out in serum free media, e.g, any as described in US-20210207080.

[0436] In some embodiments, the further incubation is carried out in the presence of one or more recombinant cytokines. In some embodiments, the further incubation is carried out in the presence of any of the recombinant cytokines described in Section II-B and at any of the concentrations described in Section II-B.

[0437] In some embodiments, the further incubation is in the presence of the same or similar media as was present during the engineering. In some embodiments, the further incubation is in media having the same cytokines as the media present during engineering. In certain embodiments, the further incubation is in media having the same cytokines at the same concentrations as the media present during engineering.

[0438] In particular embodiments, the cells are further incubated in the absence of cytokines. In particular embodiments, the cells are further incubated in the absence of any recombinant cytokine. In particular embodiments, the cells are further incubated in the absence of recombinant IL-2, IL-7, and IL- 15.

[0439] In certain embodiments, the cells are further incubated after the introducing of the polynucleotide encoding the heterologous or recombinant protein, e.g., viral vector, for, for about, or for at least 18 hours, 24 hours, 30 hours, 36 hours, 40 hours, 48 hours, 54 hours, 60 hours, 72 hours, 84 hours, 96 hours, or more than 96 hours. In certain embodiments, the cells are further incubated after the introducing of the polynucleotide encoding the heterologous or recombinant protein, e.g., viral vector, for, for about, or for at least one day, 2 days, 3 days, 4 days, or more than 4 days. In some embodiments, the further incubating is performed for an amount of time between 30 minutes and 2 hours, between 1 hour and 8 hours, between 6 hours and 12 hours, between 12 hours and 18 hours, between 16 hours and 24 hours, between 18 hours and 30 hours, between 24 hours and 48 hours, between 24 hours and 72 hours, between 42 hours and 54 hours, between 60 hours and 120 hours between 96 hours and 120 hours, between 90 hours and between 1 days and 7 days, between 3 days and 8 days, between 1 day and 3 days, between 4 days and 6 days, or between 4 days and 5 days prior to the genetic engineering. In some embodiments, the further incubating is for or for about between 18 hours and 30 hours. In particular embodiments, the further incubating is for or for about 24 hours or for for for about one day.

[0440] In certain embodiments, the total duration of the further incubation is, is about, or is at least 12 hours, 18 hours, 24 hours, 30 hours, 36 hours, 42 hours, 48 hours, 54 hours, 60 hours, 72 hours, 84 hours, 96 hours, 108 hours, or 120 hours. In certain embodiments, the total duration of the further incubation is, is about, or is at least one day, 2 days, 3 days, 4 days, or 5 days. In particular embodiments, the further incubation is completed at, at about, or within 120 hours, 108 hours, 96 hours, 84 hours, 72 hours, 60 hours, 54 hours, 48 hours, 42 hours, 36 hours, 30 hours, 24 hours, 18 hours, or 12 hours. In particular embodiments, the further incubation is completed at, at about, or within one day, 2 days, 3 days, 4 days, or 5 days. In some embodiments, the total duration of the further incubation is between or between about 12 hour and 120 hours, 18 hour and 96 hours, 24 hours and 72 hours, or 24 hours and 48 hours, inclusive. In some embodiments, the total duration of the furtherIllincubation is between or about between 1 hour and 48 hours, 4 hours and 36 hours, 8 hours and 30 hours or 12 hours and 24 hours, inclusive. In particular embodiments, the further incubation is performed for or for about 24 hours, 48 hours, or 72 hours, or for or for about 1 day, 2 days, or 3 days, respectively. In particular embodiments, the further incubation is performed for 24 hours ± 6 hours, 48 hours ± 6 hours, or 72 hours ± 6 hours. In particular embodiments, the further incubation is performed for or for about 72 hours or for or for about 3 days.

[0441] In some embodiments, the further incubation is completed between or between about 24 hour and 120 hours, 36 hour and 108 hours, 48 hours and 96 hours, or 48 hours and 72 hours, inclusive, after the initiation of the stimulation. In some embodiments, the further incubation is completed at, about, or within 120 hours, 108 hours, 96 hours, 72 hours, 48 hours, or 36 hours from the initiation of the stimulation. In some embodiments, the further incubation is completed at, about, or within 5 days, 4.5 days, 4 days, 3 days, 2 dayrs, or 1.5 days from the initiation of the stimulation. In particular embodiments, the further incubation is completed after hours 24 hours ± 6 hours, 48 hours ± 6 hours, or 72 hours ± 6 hours after the initiation of the stimulation. In some embodiments, the further incubation is completed after or after about 72 hours or after or after about 3 days.1. Viral Particles

[0442] In some embodiments, the nucleic acid sequence encoding the recombinant protein is contained in a viral particle. In some embodiments, the viral particle is a recombinant infectious virus particle. In some embodiments, the viral particle is a viral vector, such as a vector derived from simian virus 40 (SV40), adenoviruses, or adeno-associated virus (AAV). In some embodiments, the viral particle is a recombinant lentiviral vector or retroviral vector, such as a gamma-retroviral vector (see, e.g., Koste et al., Gene Therapy (2014) doi: 10.1038 / gt.2014.25; Carlens et al., Exp Hematol (2000) 28(10): 1137-46; Alonso-Camino et al. (2013) Mol Ther Nucl Acids 2, e93; Park et al., Trends Biotechnol. 2011 November 29(11): 550-557. In some embodiments, the viral particle is a recombinant lentiviral vector.

[0443] In some embodiments, the retroviral vector has a long terminal repeat sequence (LTR), e.g., a retroviral vector derived from the Moloney murine leukemia virus (MoMLV), myeloproliferative sarcoma virus (MPSV), murine embryonic stem cell virus (MESV),murine stem cell virus (MSCV), spleen focus forming virus (SFFV), or adeno-associated virus (AAV). Many retroviral vectors are derived from murine retroviruses. In some embodiments, the retroviruses include those derived from any avian or mammalian cell source. The retroviruses can be amphotropic, meaning that they are capable of infecting host cells of several species, including humans. In one embodiment, the gene to be expressed replaces the retroviral gag, pol and / or env sequences. A number of illustrative retroviral systems have been described (e.g., U.S. Pat. Nos. 5,219,740; 6,207,453; 5,219,740; Miller and Rosman (1989) BioTechniques 7:980-990; Miller, A. D. (1990) Human Gene Therapy 1:5-14; Scarpa et al. (1991) Virology 180:849-852; Bums et al. (1993) Proc. Natl. Acad. Sci. USA 90:8033-8037; and Boris-Lawrie and Temin (1993) Cur. Opin. Genet. Develop. 3:102- 109.

[0444] The viral vector genome can be constructed in a plasmid form that can be transfected into a packaging or producer cell line. In some embodiments, the nucleic acid encoding a recombinant protein, such as a recombinant receptor, is inserted or located in a region of the viral vector, such as in a non-essential region of the viral genome. In some embodiments, the nucleic acid is inserted into the viral genome in the place of certain viral sequences to produce a vims that is replication defective.

[0445] Any of a variety of known methods can be used to produce retroviral particles whose genome contains an RNA copy of the viral vector genome. In some embodiments, at least two components are involved in making a virus-based gene delivery system: first, packaging plasmids, encompassing the structural proteins as well as the enzymes necessary to generate a viral vector particle, and second, the viral vector itself, e.g., the genetic material to be transferred. Biosafety safeguards can be introduced in the design of one or both of these components.

[0446] In some embodiments, the packaging plasmid can contain all retroviral, such as HIV-1, proteins other than envelope proteins (Naldini et al., 1998). In other embodiments, viral vectors can lack additional viral genes, such as those that are associated with vimlence, e.g., vpr, vif, vpu and nef, and / or Tat, a primary transactivator of HIV. In some embodiments, lentiviral vectors, such as HIV-based lentiviral vectors, contain only three genes of the parental vims: gag, pol and rev, which reduces or eliminates the possibility of reconstitution of a wild-type virus through recombination.

[0447] In some embodiments, the viral vector genome is introduced into a packaging cell line that contains all the components necessary to package viral genomic RNA, transcribed from the viral vector genome, into viral particles. Alternatively, the viral vector genome may contain one or more genes encoding viral components in addition to the one or more sequences, e.g., recombinant nucleic acids, of interest. In some aspects, in order to prevent replication of the genome in the target cell, however, endogenous viral genes required for replication are removed and provided separately in the packaging cell line.

[0448] In some embodiments, a packaging cell line is transfected with one or more plasmid vectors containing the components necessary to generate the particles. In some embodiments, a packaging cell line is transfected with a plasmid containing the viral vector genome, including the LTRs, the cis-acting packaging sequence and the sequence of interest, i.e. a nucleic acid encoding an antigen receptor, such as a CAR; and one or more helper plasmids encoding the virus enzymatic and / or structural components, such as Gag, pol and / or rev. In some embodiments, multiple vectors are utilized to separate the various genetic components that generate the retroviral vector particles. In some such embodiments, providing separate vectors to the packaging cell reduces the chance of recombination events that might otherwise generate replication competent viruses. In some embodiments, a single plasmid vector having all of the retroviral components can be used.

[0449] In some embodiments, the retroviral vector particle, such as lentiviral vector particle, is pseudotyped to increase the transduction efficiency of host cells. For example, a retroviral vector particle, such as a lentiviral vector particle, in some embodiments is pseudotyped with a VSV-G glycoprotein, which provides a broad cell host range extending the cell types that can be transduced. In some embodiments, a packaging cell line is transfected with a plasmid or polynucleotide encoding a non-native envelope glycoprotein, such as to include xenotropic, polytropic or amphotropic envelopes, such as Sindbis virus envelope, GALV or VSV-G.

[0450] In some embodiments, the packaging cell line provides the components, including viral regulatory and structural proteins, that are required in trans for the packaging of the viral genomic RNA into lentiviral vector particles. In some embodiments, the packaging cell line may be any cell line that is capable of expressing lentiviral proteins and producing functional lentiviral vector particles. In some aspects, suitable packaging cell lines include 293 (ATCCCCL X), 293T, HeLA (ATCC CCL 2), D17 (ATCC CCL 183), MDCK (ATCC CCL 34), BHK (ATCC CCL- 10) and Cf2Th (ATCC CRL 1430) cells.

[0451] In some embodiments, the packaging cell line stably expresses the viral protein(s). For example, in some aspects, a packaging cell line containing the gag, pol, rev and / or other structural genes but without the LTR and packaging components can be constructed. In some embodiments, a packaging cell line can be transiently transfected with nucleic acid molecules encoding one or more viral proteins along with the viral vector genome containing a nucleic acid molecule encoding a heterologous protein, and / or a nucleic acid encoding an envelope glycoprotein.

[0452] In some embodiments, the viral vectors and the packaging and / or helper plasmids are introduced via transfection or infection into the packaging cell line. The packaging cell line can produce viral vector particles that contain the viral vector genome. Methods for transfection or infection are well known. Examples include calcium phosphate, DEAE- dextran and lipofection methods, electroporation and microinjection.

[0453] When a recombinant plasmid and the retroviral LTR and packaging sequences are introduced into a special cell line (e.g., by calcium phosphate precipitation for example), the packaging sequences may permit the RNA transcript of the recombinant plasmid to be packaged into viral particles, which then may be secreted into the culture media. The media containing the recombinant retroviruses in some embodiments is then collected, optionally concentrated, and used for gene transfer. For example, in some aspects, after cotransfection of the packaging plasmids and the transfer vector to the packaging cell line, the viral vector particles are recovered from the culture media and titered by standard methods used by those of skill in the art.

[0454] In some embodiments, a retroviral vector, such as a lentiviral vector, can be produced in a packaging cell line, such as an exemplary HEK 293T cell line, by introduction of plasmids to allow generation of lentiviral particles. In some embodiments, a packaging cell is transfected and / or contains a polynucleotide encoding gag and pol, and a polynucleotide encoding a recombinant receptor, such as an antigen receptor, for example, a CAR. In some embodiments, the packaging cell line is optionally and / or additionally transfected with and / or contains a polynucleotide encoding a rev protein. In some embodiments, the packaging cell line is optionally and / or additionally transfected with and / orcontains a polynucleotide encoding a non-native envelope glycoprotein, such as VSV-G. In some such embodiments, approximately two days after transfection of cells, e.g. HEK 293T cells, the cell supernatant contains recombinant lentiviral vectors, which can be recovered and titered.

[0455] Recovered and / or produced retroviral vector particles can be used to transduce target cells using the methods as described. Once in the target cells, the viral RNA is reverse- transcribed, imported into the nucleus and stably integrated into the host genome. One or two days after the integration of the viral RNA, the expression of the recombinant protein, e.g. antigen receptor, such as CAR, can be detected.

[0456] In some embodiments, the vector is a viral vector, such as a retroviral vector. In some embodiments, the polynucleotide encoding the recombinant receptor and / or additional polypeptide(s) are introduced into the cell via retroviral or lentiviral vectors, or via transposons (see, e.g., Baum et al. (2006) Molecular Therapy: The Journal of the American Society of Gene Therapy. 13:1050-1063; Frecha et al. (2010) Molecular Therapy 18:1748- 1757; and Hackett et al. (2010) Molecular Therapy 18:674-683).

[0457] In some embodiments, the one or more polynucleotide(s) or vector(s) encoding a recombinant receptor and / or additional polypeptide(s) are introduced into cells, e.g., T cells, prior to elution, cultivating, and / or expansion. This introduction of the polynucleotide(s) or vector(s) can be carried out with any suitable retroviral vector. In some embodiments, following engineering, resulting genetically engineered cells can be liberated from the initial stimulus (e.g., anti-CD3 / anti-CD28 stimulus) and subsequently be stimulated in the presence of a second type of stimulus (e.g., via a de novo introduced recombinant receptor). This second type of stimulus may include an antigenic stimulus in form of a peptide / MHC molecule, the cognate (cross-linking) ligand of the genetically introduced receptor (e.g. natural antigen and / or ligand of a CAR) or any ligand (such as an antibody) that directly binds within the framework of the new receptor (e.g. by recognizing constant regions within the receptor). See, for example, Cheadle et al, “Chimeric antigen receptors for T-cell based therapy” Methods Mol Biol. 2012; 907:645-66 or Barrett et al., Chimeric Antigen Receptor Therapy for Cancer Annual Review of Medicine Vol. 65: 333-347 (2014).

[0458] In some cases, a vector may be used that do...

Claims

ClaimsWhat is claimed is:

1. A method for determining a cell phenotype of a population of T cells, comprising:(a) obtaining holographic information for a population of cells comprising T cells;(b) determining one or more input measures for each of a plurality of cellular features derived from the holographic information, wherein each input measure is from an individual cell in the population of cells;(c) determining a plurality of population-level statistics, wherein each populationlevel statistic is of the one or more input measures for a cellular feature of the plurality of cellular features, and the plurality of population-level statistics comprises one or more population-level statistics for each of the plurality of cellular features; and(d) determining, based on the plurality of population-level statistics, a populationlevel output measure of expression of a marker expressed by T cells with a cell phenotype of interest for the population of cells.

2. A method for determining a cell phenotype of a population of T cells, comprising determining, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by T cells with a cell phenotype of interest, wherein the population-level output measure is determined based on a plurality of population-level statistics, wherein: each population-level statistic is of one or more input measures for a cellular feature of a plurality of cellular features derived from holographic information obtained for the population of cells, the plurality of population-level statistics comprising one or more population-level statistics for each of the plurality of cellular features; and each input measure is from an individual cell of the population of cells.

3. The method of claim 2, wherein the method comprises determining the plurality of population-level statistics from the one or more input measures for each of the plurality of cellular features.

4. The method of claim 2 or claim 3, wherein the method comprises determining the one or more input measures for each of the plurality of cellular features from the holographic information.

5. The method of any one of claims 2-4, wherein the method comprises obtaining the holographic information.

6. The method of any one of claims 1-5, wherein the holographic information is obtained by differential digital holographic microscopy (DDHM).

7. The method of any one of claims 4-6, wherein the obtaining the holographic information comprises imaging the population of cells using DDHM.

8. The method of any one of claims 1-7, wherein the one or more populationlevel statistics for at least one, optionally each, of the plurality of cellular features comprise one or more quantiles of the one or more input measures of the cellular feature.

9. The method of claim 8, wherein the one or more quantiles comprise one or more of the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of the one or more input measures of the cellular feature.

10. The method of any one of claims 1-7, wherein the one or more populationlevel statistics for at least one, optionally each, of the plurality of cellular features are determined by applying a distribution-based pooling filter to the one or more input measures of the cellular feature.

11. The method of any one of claims 1-10, wherein the population-level output measure is the number of cells, the percentage of cells, the proportion of cells, or the density of cells of the population of cells that express the marker.

12. The method of any one of claims 1-11, wherein the population of cells is from a culture of cells being cultured in vitro or ex vivo.

13. The method of claim 12, wherein the holographic information is obtained during the in vitro or ex vivo culture of the culture of cells.

14. The method of claim 12 or claim 13, wherein the population-level output measure is of expression of the marker during the in vitro or ex vivo culture of the culture of cells.

15. The method of any one of claims 12-14, wherein the in vitro or ex vivo culture is under conditions to expand T cells of the culture of cells.

16. The method of any one of claims 12-15, wherein the in vitro or ex vivo culture is in a bioreactor.

17. The method of any one of claims 1-16, wherein the population of cells are incubated under T cell stimulating conditions prior to when the holographic information is obtained.

18. The method of any one of claims 1-17, wherein the method comprises incubating the population of cells under T cell stimulating conditions prior to when the holographic information is obtained.

19. The method of claim 17 or claim 18, wherein the incubation is prior to the in vitro or ex vivo culture.

20. The method of any one of claims 17-19, wherein the T cell stimulating conditions comprise incubation in the presence of T cell stimulatory agents that induce a primary activation signal and a costimulatory signal in T cells.

21. The method of claim 20, wherein the T cell stimulatory agents comprise an anti-CD3 antibody or antibody fragment.

22. The method of claim 20 or claim 21, wherein the T cell stimulatory agents comprise an anti-CD28 antibody or antibody fragment.

23. The method of any one of claims 20-22, wherein the T cell stimulatory agents are immobilized on a bead.

24. The method of any one of claims 20-22, wherein the T cell stimulatory agents are immobilized on an oligomeric streptavidin mutein reagent.

25. The method of any one of claims 1-24, wherein a recombinant receptor is introduced into T cells of the population of cells prior to when the holographic information is obtained.

26. The method of any one of claims 1-25, wherein the method comprises introducing a recombinant receptor into T cells of the population of cells prior to when the holographic information is obtained.

27. The method of claim 25 or claim 26, wherein the introducing comprising contacting the population of cells with an agent comprising a polynucleotide encoding the recombinant receptor.

28. The method of claim 26 or claim 27, wherein the recombinant receptor is a chimeric antigen receptor (CAR) or an engineered T cell receptor (TCR).

29. 'fhe method of any one of claims 1 -28, wherein at least 50, 55, 60, 65, 70, 75, 80, 85, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, or 100% of the population of cells are T cells.

30. The method of any one of claims 1-29, wherein the population-level output measure is determined by providing the plurality of population-level statistics as input to a machine learning model trained to predict population-level output measures of expression of the marker based on population-level statistics of the plurality of cellular features.

31. The method of claim 30, wherein the machine learning model is trained using a dataset of reference population-level statistics, wherein: for each of a first plurality of reference populations of cells comprising T cells, the dataset of reference population-level statistics comprises one or more reference populationlevel statistics for each of the plurality of cellular features, wherein: each reference population-level statistic is of one or more reference input measures for a cellular feature of the plurality of cellular features derived from holographic information obtained for the reference population of cells; and each reference input measure is from an individual cell of the reference population of cells.

32. The method of claim 31, wherein the machine learning model is trained using a dataset of reference population-level output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures comprises a reference population-level output measure of expression of the marker for the reference population of cells, wherein: the first and second pluralities of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo', and reference populations of cells from the first plurality of reference populations of cells are from the same reference culture of cells as reference populations of cells from the second plurality of reference populations of cells.

33. A method for training a machine learning model that predicts a cell phenotype of a population of T cells, comprising training a machine learning model using:(i) a dataset of reference population-level statistics, wherein for each of a first plurality of reference populations of cells comprising T cells, the dataset of referencepopulation-level statistics comprises one or more reference population-level statistics for each of a plurality of cellular features, wherein: each reference population-level statistic is of one or more reference input measures for a cellular feature of the plurality of cellular features derived from holographic information obtained for the reference population of cells; and each reference input measure is from an individual cell of the reference population of cells; and(ii) a dataset of reference population-level output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures comprises a reference population-level output measure of expression of a marker expressed by T cells with a cell phenotype of interest for the reference population of cells, wherein: the first and second pluralities of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo', and reference populations of cells from the first plurality of reference populations of cells are from the same reference culture of cells as reference populations of cells from the second plurality of reference populations of cells; whereby the machine learning model is trained to predict population-level output measures of expression of the marker based on population-level statistics of the plurality of cellular features.

34. The method of any one of claims 1-33, wherein the plurality of cellular features comprise one or of intensity skewness, intensity correlation, intensity homogeneity, intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean.

35. The method of any one of claims 1-34, wherein the plurality of cellular features comprises intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean.

36. A method for determining the activation state of a population of T cells, comprising determining, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by activated T cells, wherein the marker is CD137 (4-1BB), and the population-level output measure is determined based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean; and each input measure is from an individual cell of the population of cells.

37. The method of any one of claims 1-36, wherein the plurality of cellular features comprises intensity skewness, intensity correlation, and intensity homogeneity.

38. A method for determining the activation state of a population of T cells, comprising determining, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by activated T cells, wherein the marker is CD137 (4-1BB), and the population-level output measure is determined based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises intensity skewness, intensity correlation, and intensity homogeneity; and each input measure is from an individual cell of the population of cells.

39. The method of claim 38, wherein the plurality of cellular features comprises intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean.

40. The method of any one of claims 1-33, wherein the plurality of cellular features comprises one or more of peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height.

41. The method of any one of claims 1-33 and 40, wherein the plurality of cellular features comprises peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height.

42. The method of any one of claims 1-33, wherein the plurality of cellular features comprises one or more of cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter.

43. The method of any one of claims 1-33 and 42, wherein the plurality of cellular features comprises cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter.

44. A method for determining the memory phenotype of a population of T cells, comprising determining, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by central memory T cells or stem cell memory T cells, wherein the marker is CCR7, and the population-level output measure is determined based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises one or more of cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter; and each input measure is from an individual cell of the population of cells.

45. The method of claim 44, wherein the plurality of features comprises cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter.

46. A method for determining recombinant receptor expression of a population of T cells, comprising determining, for a population of cells comprising T cells, expression of a recombinant receptor introduced into T cells of the population of cells, wherein the determining is based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein:the plurality of cellular features comprises peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height; and each input measure is from an individual cell of the population of cells.

47. The method of any one of claims 1-33, wherein the one or more input measures for at least one, optionally each, of the plurality of cellular features are determined by providing the holographic information for individual cells of the population of cells to a convolutional neural network, wherein: the plurality of cellular features are cellular features extracted by the convolutional neural network; and the one or more input measures are determined from the convolutional neural network.

48. The method of any one of claims 31-33 and 47, wherein the one or more reference input measures for at least one, optionally each, of the plurality of cellular features are determined by providing the holographic information for individual cells of the reference population of cells to a convolutional neural network, wherein: the plurality of cellular features are cellular features extracted by the convolutional neural network; and the one or more reference input measures are determined from the convolutional neural network.

49. The method of claim 47 or claim 48, wherein the convolutional neural network is trained using a dataset of reference holographic information, wherein for each of a first plurality of reference populations of cells comprising T cells, the dataset of reference holographic information comprises holographic information obtained for individual cells of the reference population of cells.

50. The method of any one of claims 31-35, 37, 40-43, and 47-49, wherein the one or more reference population-level statistics for at least one, optionally each, of the pluralityof cellular features comprise one or more quantiles of the one or more reference input measures for the cellular feature.

51. The method of claim 50, wherein the one or more quantiles are selected from the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of the one or more reference input measures for the cellular feature.

52. The method of any one of claims 31-35, 37, 40-43, and 47-49, wherein the one or more reference population-level statistics for at least one, optionally each, of the plurality of cellular features are determined by applying a distribution-based pooling filter to the one or more reference input measures for the cellular feature.

53. A method for training a machine learning model that predicts a cell phenotype of a population of T cells, comprising:(a) training a convolutional neural network using a dataset of reference holographic information, wherein for each of a first plurality of reference populations of cells comprising T cells, the dataset of reference holographic information comprises holographic information obtained for individual cells of the reference population of cells;(b) determining, from the convolutional neural network, one or more reference input measures for each cellular feature of a plurality of cellular features derived from the holographic information, wherein the plurality of cellular features are cellular features extracted by the convolutional neural network, and each reference input measure is from an individual cell of a reference population of cells;(c) determining a dataset of reference population-level statistics, wherein the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of the plurality of cellular features, each reference population-level statistic determined by applying a distribution-based pooling filter to the one or more reference input measures for a cellular feature of the plurality of cellular features; and(d) training a machine learning model using the dataset of reference population-level statistics and a dataset of reference population-level output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-leveloutput measures comprises a reference population-level output measure of expression of a marker expressed by T cells with a cell phenotype of interest for the reference population of cells, wherein: the first and second pluralities of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo', and reference populations of cells from the first plurality of reference populations of cells are from the same reference culture of cells as reference populations of cells from the second plurality of reference populations of cells; whereby the machine learning model is trained to predict population-level output measures of expression of the marker based on population-level statistics of the plurality of cellular features.

54. The method of any one of claims 47-53, wherein the one or more input measures for at least one, optionally each, of the plurality of cellular features are obtained from a fully connected layer of the convolutional neural network.

55. The method of any one of claims 31-35, 37, 40-43, and 47-54, wherein the holographic information for the first plurality of reference populations of cells is obtained by DDHM.

56. The method of any one of claims 1-55, wherein the holographic information comprises phase information and intensity information.

57. The method of any one of claims 32-35, 37, 40-43, and 47-56, wherein the reference population-level output measure is the number of cells, the percentage of cells, the proportion of cells, or the density of cells of the reference population of cells that express the marker.

58. The method of any one of claims 32-35, 37, 40-43, and 47-57, wherein the dataset of reference population-level statistics and the dataset of reference population-level output measures are time-matched.

59. The method of any one of claims 32-35, 37, 40-43, and 47-58, wherein the in vitro or ex vivo culture of the reference cultures of cells is performed under the same or similar conditions as the in vitro or ex vivo culture of the culture of cells.

60. The method of any one of claims 32-35, 37, 40-43, and 47-59, wherein the holographic information for the first plurality of reference populations of cells is obtained during the in vitro or ex vivo culture of the reference cultures of cells.

61. The method of any one of claims 32-35, 37, 40-43, and 47-60, wherein the dataset of reference population-level output measures are of expression of the marker during or after the in vitro or ex vivo culture of the reference cultures of cells.

62. The method of any one of claims 32-35, 37, 40-43, and 47-61, wherein the dataset of reference population-level output measures is determined using fluorescence imaging of the second plurality of reference populations of cells.

63. The method of claim 62, wherein the fluorescence imaging is by flow cytometry.

64. The method of any one of claims 1-35, 37, and 47-63, wherein the method is for determining the activation state of the population of T cells, and the marker is expressed by activated T cells.

65. The method of any one of claims 1-35, 37, and 47-64, wherein the marker is CD137 (4-1BB).

66. The method of any one of claims 1-35 and 42-63, wherein the method is for determining the memory phenotype of the population of T cells, and the marker is expressed by T cells having a central memory phenotype or a stem cell memory phenotype.

67. The method of claim 66, wherein the marker is expressed by T cells having a central memory phenotype.

68. The method of claim 66, wherein the marker is expressed by T cells having a stem cell memory phenotype.

69. The method of any one of claims 1-35, 42-63, and 66-68, wherein the marker is CCR7.

70. The method of any one of claims 1-33 and 40-63, wherein the method is for determining recombinant receptor expression of the population of T cells, and the marker is a recombinant receptor introduced into T cells of the population of cells prior to when the holographic information is obtained.

71. The method of any one of claims 46-63 and 70, wherein the recombinant receptor is a chimeric antigen receptor (CAR) or an engineered T cell receptor (TCR).

72. The method of any one of claims 31-35, 37, 40-43, and 47-71, wherein the first plurality of reference populations of cells are incubated under T cell stimulating conditions prior to when the holographic information for the first plurality of reference populations of cells is obtained.

73. The method of claim 72, wherein the incubation of the first plurality of reference populations of cells is performed under the same or similar conditions as the incubation of the population of cells.

74. The method of claim 72 or claim 73, wherein the incubation of the first plurality of reference populations of cells is prior to the in vitro or ex vivo culture of the reference cultures of cells.

75. The method of any one of claims 31-35, 37, 40-43, and 47-74, wherein the first and / or second plurality of reference populations of cells are enriched for T cells.

76. The method of any one of claims 1 -35. 37, 40-43, and 47-75, wherein at least 50, 55, 60, 65, 70, 75, 80, 85, 90, 91 , 92, 93, 94, 95, 96, 97, 98, 99, or 100% of the first and / or second plurality of reference populations of cells are T cells.

77. A method for monitoring a cell phenotype of a culture of T cells, comprising determining, for a population of cells from a culture of cells comprising T cells that is being cultured in vitro or ex vivo, a population-level output measure of expression of a marker expressed by T cells with a cell phenotype of interest, wherein the population-level output measure is determined according to the method of any one of claims 1-32, 34, 35, 37, 40-43, 47-52, and 54-76.

78. A computing device comprising instructions in memory for performing the method of any one of claims 1-32, 34, 35, 37, 40-43, 47-52, and 54-76, the instructions comprising instructions for:(a) receiving the holographic information for individual cells of a population of cells comprising T cells, the one or more input measures for each of the plurality of cellular features for individual cells of a population of cells comprising T cells, or the plurality of population-level statistics for a population of cells comprising T cells; and(b) determining, according to the method, the population-level output measure for the population of cells from the holographic information, the one or more input measures for each of the plurality of cellular features, or the plurality of population-level statistics.

79. The computing device of claim 78, wherein the computing device further comprises in memory a machine learning model trained according to the method of any one of claims 33-35, 37, 40-43, and 47-76, wherein the population-level output measure for the population of cells is determined using the machine learning model.

Images