Methods for classifying cells

Abstract

Reported herein is a method for sorting cells in a cell mixture, the mixture comprising T cells and B cells, the method comprising the following steps: first, applying at least labeled antibodies that bind to F-actin, MHCII, and CD3 to the cell mixture to obtain a labeled cell mixture, wherein the antibodies are each labeled with a dye, and the dyes have different (non-overlapping) emission wavelengths; second, acquiring at least one image of the cell mixture; and third, sorting the cells in the cell mixture into single cells if the cells are single cells that are F-actin positive, MHCII positive, and CD3 negative, or MHCII negative and CD3 positive, or sorting the cells in the cell mixture into cell doublets or multiplets if the cells are aggregates of two or three cells that are F-actin positive, MHCII positive, and CD3 positive.

Description

[Technical Field]

[0001] The present invention relates to the field of analytical techniques. More specifically, a method is reported herein for classifying cells in a mixture of B and T cells into isolated cells, multiplets of cells without signaling, and multiplets of cells with signaling based on differential markers. Such classification allows for the characterization of therapeutic agents that interfere with the formation of cell signaling. [Background technology]

[0002] Background of the Invention Therapeutic antibodies are widely used to treat serious diseases. Most of them modify immune cells and act within the immunological synapse, an essential cell-cell interaction for directing humoral immune responses. Although many antibodies have been designed and evaluated, high-throughput tools for systematic antibody characterization and functional prediction are lacking.

[0003] The formation of the immunological synapse is the first event in the adaptive immune response, induced by the interaction between a T cell and its corresponding antigen-presenting cell (APC). This rapidly formed cell-cell interface is initiated by the recognition of peptide-loaded MHC complexes by the T cell receptor (TCR). It involves the rearrangement of cytoskeletal actin filaments and the recruitment of signaling, costimulatory, co-inhibitory, and adhesion molecules to the nascent synapse [1, 2]. This process is crucial for triggering and fine-tuning T cell responses and ensuring an intact immune response. Dysfunctional immunological synapse formation has been observed in several immune-related disorders [3-8] and is therefore considered a potential target for stimulating or inhibiting immune responses by modulating its assembly or function [9-11]. For example, various therapeutic antibodies have been developed that alter immunological synapse formation to treat cancer and autoimmune diseases [12-15]. Although significant progress in the development of immunological synapse-targeting agents has been achieved in recent years [9], there is still a need to further refine these compounds, particularly to improve their efficacy. It has been shown that antibody size and format [16, 17], dose, and target expression

[18] may be important parameters for immunological synapse formation and its effect on T cell function.

[0004] However, no methods have been reported to systematically quantify and characterize the morphology of the immunological synapse, examine its correlation to T cell responses, or identify properties that predict antibody efficacy in vitro.

[0005] A key technology for high-throughput data acquisition for this purpose is imaging flow cytometry (IFC), which combines the advantages of conventional flow cytometry with deep multichannel imaging at the single-cell level. Recently, IFC has been successfully applied to visualize and quantify the immunological synapse of primary human T:APC cell conjugates [19-21], but none of these studies investigated the formation of the immunological synapse in the context of T cell function.

[0006] Recent studies have demonstrated the potential of machine learning algorithms for more robust and accurate analysis of high-throughput imaging data, an approach that has been demonstrated to overcome the limitations of traditional gating strategies [22-24]. Leveraging machine learning for IFC data analysis has also enabled the identification of subcellular morphological patterns, combined analysis of RNA and protein data, and the implementation of predictive models [22-26]. While there are limited open-source software implementations available designed for IFC data analysis [26, 27], they either rely on additional software that complicates the analysis pipeline or focus solely on predictive performance, lacking interpretability.

[0007] The immunological synapse has previously been studied using high-content cell imaging on human cell lines and primary cells with an artificial APC system utilizing plate-bound ICAM-1 and stimulatory antibodies

[37] . German et al. convincingly demonstrated the power of their pipeline by profiling the immunological synapse, but did not investigate whether these profiles could be used to predict drug efficacy

[37] . Other studies have also investigated the potential for synapse formation with CAR T cell therapies, where researchers used the average intensity of staining per cell, such as F-actin and P-CD3 zeta, tumor antigen clustering, and polarization of perforin-containing granules, as measures of synapse formation quality. These characteristics differed between different CAR T cells and correlated with their efficacy in vitro and in vivo as well as clinical outcomes [39, 40].

[0008] M. Chen et al. have disclosed that heparin-binding EGF-like growth factors regulate the bidirectional activation of CD4+ T cells and dendritic cells independently of the epidermal growth factor receptor (Am. J. Resp. Crit. Care, 2018, Meeting Abstracts. A5826).

[0009] BH Hosseini et al. disclose that immune synapse formation determines the interaction force between T cells and antigen-presenting cells as measured by atomic force microscopy (Proc. Natl. Acad. Sci USA 106 (2009) 17852-17857).

[0010] F. Ahmed et al. disclose that numbers are important for quantitative and dynamic analysis of immunological synapse formation using imaging flow cytometry (J. Immunol. Meth. 347 (2009) 79-86).

[0011] G. Wabnitz et al. have disclosed that influx microscopy of human leukocytes is a tool for quantitative analysis of actin reorganization in the immune synapse (J. Immunol. Meth. 423 (2015) 29-39).

[0012] US Patent Application Publication No. 2021 / 270812 discloses a method for analyzing immune cells. Summary of the Invention

[0013] Herein, we report a method for sorting cells in a cell mixture using single-cell imaging flow cytometry.

[0014] Herein, we further report a method for classifying cells in a cell mixture using single-cell imaging flow cytometry in combination with artificial intelligence (scifAI) for preprocessing, feature engineering, and explainable predictive machine learning on imaging flow cytometry (IFC) data. The algorithm flowchart is shown in Figure 40.

[0015] The method of the present invention allows for the analysis of class frequencies and morphological changes under different immune stimuli. The applicability of the method of the present invention is demonstrated by analyzing T cell cytokine production across multiple donors and therapeutic antibodies. The characteristics are quantitatively predicted in vitro, demonstrating the potential to link morphological features with function and significantly influence antibody design.

[0016] The method according to the invention is universally applicable to IFC data and, given its modular structure, is easy to integrate into existing workflows and analytical pipelines, e.g., for rapid antibody screening and functional characterization.

[0017] Therefore, the present invention encompasses the following embodiments.

[0018] 1. A method for sorting cells in a cell mixture, the mixture comprising T cells and activated B cells or antigen-presenting cells, the method comprising: a) applying at least labeled antibodies that bind to F-actin, MHCII, and CD3 to the cell mixture to obtain a labeled cell mixture, wherein the antibodies are each labeled with a dye, the dyes having different, optionally non-overlapping, emission wavelengths; b) acquiring at least one image of the cell mixture; c) removing the cells in the cell mixture; i) the cells are single cells and F-actin positive; - MHCII positive and CD3 negative, or - If MHCII negative and CD3 positive, classify as isolated cells, ii) classifying the cells into cell doublets or multiplets if the cells are aggregates of two or more cells and are F-actin positive, MHCII positive and CD3 positive; A method comprising:

[0019] 1a. A method for sorting cells in a cell mixture, the mixture comprising T cells and activated B cells or antigen presenting cells, the method comprising: a) applying at least labeled antibodies that bind to F-actin, MHCII, and CD3 to the cell mixture to obtain a labeled cell mixture, wherein the antibodies are each labeled with a dye, the dyes having different, optionally non-overlapping, emission wavelengths; b) acquiring at least one image of the cell mixture; c) classifying the images as follows: i) The cells in the image are F-actin positive, - MHCII positive and CD3 negative, or - Classify the image as containing isolated cells if MHCII negative and CD3 positive; ii) classifying the image as containing cell doublets or multiplets if the cells are aggregates of two or more cells, are F-actin positive, MHCII positive and CD3 positive; A method comprising:

[0020] 2. step a) applying to the cell mixture at least labeled antibodies that bind to F-actin, MHCII, CD3, and P-CD3 zeta, wherein the antibodies are each labeled with a dye, and the dyes have different (non-overlapping) emission wavelengths; Step c) comprises: i) if the cells are F-actin positive, MHCII positive, CD3 negative and P-CD3 zeta negative, they are classified as single B cells or antigen-presenting cells; ii) if the cells are F-actin positive, MHCII negative, CD3 positive and P-CD3 zeta negative, they are classified as single T cells without signal transduction; iii) if the cells are F-actin positive, MHCII negative, CD3 positive and P-CD3 zeta positive, they are classified as single T cells with signaling; iv) If the doublet is F-actin positive, MHCII positive, CD3 positive, and P-CD3 zeta negative, it is classified as a doublet that forms a synapse between a B cell or an antigen-presenting cell and a T cell without signal transduction; v) When the doublets or multiplets are F-actin positive, MHCII positive, CD3 positive, and P-CD3 zeta positive, classifying them into doublets or multiplets that form synapses involving signal transduction between one or more B cells or antigen-presenting cells and one or more T cells. 2. The method of embodiment 1, wherein

[0021] 2a. step a) applying to the cell mixture at least labeled antibodies that bind to F-actin, MHCII, CD3, and P-CD3 zeta, wherein the antibodies are each labeled with a dye, and the dyes have different (non-overlapping) emission wavelengths; step c) classifying the images as follows: i) classifying an image as containing a single B cell or antigen-presenting cell if the cells in the image are F-actin positive, MHCII positive, CD3 negative, and P-CD3 zeta negative; ii) classifying the image as containing a single T cell without signaling if the cell in the image is F-actin positive, MHCII negative, CD3 positive, and P-CD3 zeta negative; iii) classifying the image as containing a single T cell with signaling if the cell in the image is F-actin positive, MHCII negative, CD3 positive, and P-CD3 zeta positive; iv) classifying the image as containing a doublet of B cells or antigen-presenting cells and T cells that forms a synapse without signal transduction if the cell doublet in the image is F-actin positive, MHCII positive, CD3 positive, and P-CD3 zeta negative; v) classifying the image as containing a doublet or multiplet of one or more B cells or antigen-presenting cells and one or more T cells that form a synapse with signal transduction, if the cell doublet or multiplet in the image is F-actin positive, MHCII positive, CD3 positive, and P-CD3 zeta positive. The method of embodiment 1a, wherein

[0022] 3. The method according to any one of embodiments 1 to 2a, wherein step b) is a step of acquiring an image of the cell mixture using an imaging flow cytometer.

[0023] 4. The method of any one of embodiments 1 to 3, wherein the acquired images are images showing single cells or isolated doublets or multiplets, respectively.

[0024] 5. A method for sorting cells in a cell mixture, the mixture comprising T cells and activated B cells or antigen-presenting cells, the method comprising: a) acquiring at least one image of the cell mixture; b) generating a feature extraction pipeline for deriving biologically interpretable features from at least one image; c) determining whether the cells are: Class i) single B cells or antigen-presenting cells; Class ii) single T cells without signaling; Class iii) single T cells with signal transduction; Class iv) Doublets that form synapses between B cells or antigen-presenting cells and T cells without signal transduction. Class v) Doublets or multiplets that form synapses involving signal transduction between one or more B cells or antigen-presenting cells and one or more T cells a step of predicting that the classification result belongs to one of the following: A method comprising:

[0025] 5a. A method for sorting cells in a cell mixture, the mixture comprising T cells and activated B cells or antigen-presenting cells, the method comprising: a) acquiring at least one image of the cell mixture; b) generating a feature extraction pipeline for deriving biologically interpretable features from at least one image; c) based on the derived biologically interpretable features, cells in the image are i) a single B cell or antigen-presenting cell; ii) a single T cell without signaling; iii) single T cells with signaling; iv) Doublets that form synapses between B cells or antigen-presenting cells and T cells without signal transduction; v) Doublets or multiplets that form synapses involving signal transduction between one or more B cells or antigen-presenting cells and one or more T cells. and predicting that A method comprising:

[0026] 6. The method of any one of embodiments 1 to 5a, wherein cell doublets and multiplets (within the image) are classified into synapses based on morphology, labeling intensity, co-localization of labels, texture and synaptic features.

[0027] 7. Cell doublets and multiplets (in the image) are characterized by the following additional features: - co-localization of CD3 and MHCII markers, and / or - Co-localization of MHCII and P-CD3 zeta labeling, and / or - MHCII texture, and / or -CD3 label texture, and / or -Intensity of P-CD3 zeta labeling 7. The method of any one of embodiments 1 to 6, wherein the synapses are classified based on one or more of:

[0028] 8. The method of any one of embodiments 1 to 7, wherein features are determined for each cell or doublet or multiplet or image based on the ratio of labeling signal intensity in the synaptic region to the whole cell.

[0029] 9. The method according to any one of embodiments 1 to 4 and embodiments 6 to 8, wherein the labeled cell mixture is an intracellular labeled cell mixture obtained by fixing the cells, permeabilizing the cells, and applying a labeled antibody.

[0030] 9a. Step d) below d) counting the number of cells or cell images of each class and calculating the relative frequency of cells of each class in the cell mixture.

[0031] 10. The method according to any one of embodiments 1 to 9a, wherein dead, deformed or excised cells (aggregates of four or more cells) are removed before step b), or images of dead, deformed or excised cells (aggregates of four or more cells) are not recorded, or images of dead, deformed or excised cells as well as out-of-focus images are removed before step c), or images of dead, deformed or excised cells as well as out-of-focus images are not analyzed in step c), or images of dead, deformed or excised cells as well as out-of-focus images are not counted in step d).

[0032] 11. Step b) further comprises: b-1) a substep of gating in-focus live+CD3+MHCII+ cells; b-2) a substep of selecting images representing a single CD3+ T cell and a single MHCII+ B cell or antigen-presenting cell from the population obtained in step b-1) using area and aspect ratio features; b-3) a substep of measuring the signal intensity of labeled CD3 within a synaptic mask (the synaptic mask is defined as a combination of the morphological CD3 and MHCII masks with an extension of 3) and gating synapses that show a CD3 signal within the mask; b-4) A substep of excluding (images of) T cells and B cells or antigen-presenting cells in one layer by using height and area features of bright field (BF); 11. The method of any one of embodiments 1 to 10, comprising:

[0033] 12. Step c) c) constructing a model based on the biologically interpretable features derived based on the XGBoost classifier, and based on the model, determining whether the cell or image is: Class i) single B cells or antigen-presenting cells; Class ii) single T cells without signaling; Class iii) single T cells with signal transduction; Class iv) Doublets that form synapses between B cells or antigen-presenting cells and T cells without signal transduction. Class v) Doublets or multiplets that form synapses involving signal transduction between one or more B cells or antigen-presenting cells and one or more T cells 12. The method according to any one of embodiments 5 to 11, comprising predicting that the target gene belongs to one of the following:

[0034] 13. The method of any one of embodiments 1 to 12, wherein the mixture comprises B cells or antigen-presenting cells and T cells in a cell ratio of about 4:3.

[0035] 14. The method of any one of embodiments 1 to 13, wherein the T cells are CD4-positive memory T cells or CD8-positive T cells or a mixture thereof.

[0036] 15. The method of any one of embodiments 1 to 14, wherein the T cells are CD4-positive memory T cells.

[0037] 16. The method of any one of embodiments 1 to 15, wherein after mixing, the mixture of cells is centrifuged.

[0038] 17. The method according to any one of embodiments 1 to 16, wherein step b) further comprises compensating the image using a compensation matrix derived from a stained single cell.

[0039] 18. A method for sorting cells in a cell mixture, the mixture comprising T cells and activated B cells or antigen-presenting cells, the method comprising: a) a mixture of labeled cells; a-1) aliquoting the cell mixture into at least two aliquots; a-2) applying to a first aliquot of the mixture an antibody that binds to one or more cell surface targets present on one or both of the cells of the mixture, and applying to a second aliquot of the mixture an antibody that has the same structure as the antibody applied to the first aliquot but that does not bind to one or more cell surface targets present on one or both of the cells of the mixture; a-3) incubating the aliquot obtained in step a-2); a-4) applying at least labeled antibodies that bind to F-actin, MHCII, and CD3 to the incubated aliquot of the cell mixture obtained in step a-3) to obtain a labeled cell mixture, wherein the antibodies are each labeled with a dye, and the dyes have different (non-overlapping) emission wavelengths; and preparing the compound by b) acquiring at least one image of each aliquot of the labeled cell mixture; c) Images of cells in each aliquot of the cell mixture separately i) the image is of a single cell and is F-actin positive; - MHCII positive and CD3 negative, or - If it contains cells that are MHCII negative and CD3 positive, it is classified as containing isolated cells, ii) classifying the image as containing cell doublets or multiplets if the image contains cells that are aggregates of two or three cells and are F-actin positive, MHCII positive and CD3 positive; d) counting the number of images of cells of each class for each aliquot and calculating the relative frequency of cells of each class in the cell mixture; e) determining the difference in class frequencies between the first aliquot and the second aliquot; A method comprising:

[0040] 19. Step a-4) is a step of applying at least labeled antibodies that bind to F-actin, MHCII, CD3, and P-CD3 zeta to the incubated aliquot of the cell mixture, wherein the antibodies are each labeled with a dye, and the dyes have different (non-overlapping) emission wavelengths; Step c) comprises: i) if the cells are F-actin positive, MHCII positive, CD3 negative and P-CD3 zeta negative, they are classified as single B cells or antigen-presenting cells; ii) if the cells are F-actin positive, MHCII negative, CD3 positive and P-CD3 zeta negative, they are classified as single T cells without signal transduction; iii) if the cells are F-actin positive, MHCII negative, CD3 positive and P-CD3 zeta positive, they are classified as single T cells with signaling; iv) If the doublet is F-actin positive, MHCII positive, CD3 positive, and P-CD3 zeta negative, it is classified as a doublet that forms a synapse between a B cell or an antigen-presenting cell and a T cell without signal transduction; v) When the doublets or multiplets are F-actin positive, MHCII positive, CD3 positive, and P-CD3 zeta positive, classifying them into doublets or multiplets that form synapses involving signal transduction between one or more B cells or antigen-presenting cells and one or more T cells. 19. The method of embodiment 18, wherein

[0041] 19a. step a-4) is a step of applying at least labeled antibodies that bind to F-actin, MHCII, CD3, and P-CD3 zeta to the incubated aliquot of the cell mixture, wherein the antibodies are each labeled with a dye, and the dyes have different (non-overlapping) emission wavelengths; step c) classifying the images as follows: i) classifying an image as containing a single B cell or antigen-presenting cell if the cells in the image are F-actin positive, MHCII positive, CD3 negative, and P-CD3 zeta negative; ii) classifying the image as containing a single T cell without signaling if the cell in the image is F-actin positive, MHCII negative, CD3 positive, and P-CD3 zeta negative; iii) classifying the image as containing a single T cell with signaling if the cell in the image is F-actin positive, MHCII negative, CD3 positive, and P-CD3 zeta positive; iv) classifying the image as containing a doublet that forms a synapse between a B cell or an antigen-presenting cell and a T cell without signal transduction if the doublet in the image is F-actin positive, MHCII positive, CD3 positive, and P-CD3 zeta negative; v) classifying the doublets or multiplets in the image as including doublets or multiplets that form synapses involving signal transduction between one or more B cells or antigen-presenting cells and one or more T cells, if the doublets or multiplets in the image are F-actin-positive, MHCII-positive, CD3-positive, and P-CD3 zeta-positive. 19. The method of embodiment 18, wherein

[0042] 20. The method according to any one of embodiments 18 to 19a, wherein step b) is a step of acquiring an image of the cell mixture using an imaging flow cytometer.

[0043] 21. The method according to any one of embodiments 18 to 20, wherein the acquired images are images showing single cells or isolated doublets or multiplets, respectively.

[0044] 22. The method according to any one of embodiments 18 to 21, wherein cell doublets and multiplets or images thereof are classified into synapses based on morphology, labeling intensity, co-localization of labels, texture and synaptic features.

[0045] 23. Cell doublets and multiplets in the image are further characterized by: - co-localization of CD3 and MHCII markers, and / or - Co-localization of MHCII and P-CD3 zeta labeling, and / or - MHCII texture, and / or -CD3 label texture, and / or -Intensity of P-CD3 zeta labeling 23. The method of any one of embodiments 18 to 22, wherein the synapses are classified based on one or more of:

[0046] 24. The method according to any one of embodiments 18 to 23, wherein features are determined for each cell or doublet or multiplet or image based on the ratio of labeling signal intensity in the synaptic region to the whole cell.

[0047] 25. The method according to any one of embodiments 18 to 24, wherein the labeled cell mixture is an intracellular labeled cell mixture obtained by fixing the cells, permeabilizing the cells, and applying a labeled antibody.

[0048] 26. The method according to any one of embodiments 18 to 25, wherein dead, deformed or excised cells (aggregates of four or more cells) are removed before step b), or images of dead, deformed or excised cells (aggregates of four or more cells) are not recorded, or images of dead, deformed or excised cells, as well as out-of-focus images, are removed before step c), or images of dead, deformed or excised cells, as well as out-of-focus images, are not analyzed in step c), or images of dead, deformed or excised cells, as well as out-of-focus images, are not counted in step d).

[0049] 27. Step b) further comprises: b-1) a substep of gating in-focus live+CD3+MHCII+ cells; b-2) a substep of selecting images representing a single CD3+ T cell and a single MHCII+ B cell or antigen-presenting cell from the population obtained in step b-1) using area and aspect ratio features; b-3) a substep of measuring the signal intensity of labeled CD3 within a synaptic mask (the synaptic mask is defined as a combination of the morphological CD3 and MHCII masks with an extension of 3) and gating synapses that show a CD3 signal within the mask; b-4) A substep of excluding T cells and B cells or antigen-presenting cells in one layer by using the height and area features of bright field (BF); 27. The method of any one of embodiments 18 to 26, comprising:

[0050] 28. The method of any one of embodiments 18 to 27, wherein the mixture comprises B cells or antigen-presenting cells and T cells in a cell ratio of about 4:3.

[0051] 29. The method of any one of embodiments 18 to 28, wherein the T cells are CD4-positive memory T cells or CD8-positive T cells or a mixture thereof.

[0052] 30. The method of any one of embodiments 18 to 29, wherein the T cells are CD4-positive memory T cells.

[0053] 31. The method of any one of embodiments 18 to 30, wherein after mixing, the mixture of cells is centrifuged.

[0054] 32. The method according to any one of embodiments 18 to 31, wherein step b) further comprises compensating the image using a compensation matrix derived from a stained single cell.

[0055] 33. A method for ranking antibodies in a large number of antibodies, the method comprising: 1) carrying out the method of any one of embodiments 18 to 31 for each antibody of a multiplicity of antibodies individually or for all antibodies together, wherein each antibody is applied to a separate aliquot of the mixture; 2) an antibody, i) A single B cell or antigen-presenting cell class (image); ii) a single T cell (image) class without signaling; iii) a single T cell (image) class with signaling; iv) A class of doublets (images) that form synapses between B cells or antigen-presenting cells and T cells without signal transduction; v) ranking based on changes in the frequency of one or more classes of (image) doublets or multiplets that form synapses involving signal transduction between one or more B cells or antigen-presenting cells and one or more T cells; A method comprising:

[0056] 34. The method of embodiment 33, wherein the ranking of the antibody in the plurality of antibodies is due to a reduced stimulation of the immune response.

[0057] 35. The method of any one of embodiments 33-34, wherein ranking of antibodies in a multiplicity of antibodies reduces the frequency of cells or images that fall into doublets or multiplets that form signaling synapses with one or more B cells or antigen-presenting cells and one or more T cells.

[0058] 36. The method of any one of embodiments 33 to 35, wherein the ranking of antibodies in a multiplicity of antibodies is further based on decreasing the signal intensity of labeled F-actin, labeled P-CD3 zeta, and labeled MHCII in the synaptic region.

[0059] 37. The method of embodiment 33, wherein the ranking of the antibody in the plurality of antibodies is by reduced inhibition of the immune response.

[0060] 38. The method of any one of embodiments 33 and 37, wherein ranking of antibodies in a multiplicity of antibodies reduces the frequency of cells or images that fall into doublets or multiplets that form signaling synapses with one or more B cells or antigen-presenting cells and one or more T cells.

[0061] 39. The method of any one of embodiments 33 and 37-38, wherein ranking of antibodies in a multiplicity of antibodies is further by increasing the frequency of cells or images classified as single T cells without signaling.

[0062] 40. The method of any one of embodiments 33 and 37 to 39, wherein ranking of antibodies in a multiplicity of antibodies is further by reducing the frequency of cells or images classified as single T cells with signaling.

[0063] 41. The method of any one of embodiments 33 and 37 to 40, wherein the ranking of antibodies in a large number of antibodies is further achieved by increasing the frequency of the average signal intensity of labeled F-actin, increasing the signal intensity of labeled P-CD3 zeta in the synaptic region, and declustering the signals for the T cell receptor.

[0064] 42. Use of the method according to any one of embodiments 1 to 41 for characterizing the morphology of synapses formed between T cells and B cells or antigen-presenting cells.

[0065] 43. Use of the method according to any one of embodiments 33 to 41 for determining the correlation between antibody concentration and T cell response.

[0066] 44. Use of the method according to any one of embodiments 33 to 41 for predicting the therapeutic mechanism of action of an antibody.

[0067] 45. Use of the method according to any one of embodiments 33 to 41 for predicting the efficacy of an antibody.

[0068] 46. ​​The use of F-actin, MHCII and CD3 to sort T cells in a mixture containing T cells and B cells or activated B cells or antigen-presenting cells.

[0069] 47. The use of F-actin, MHCII and CD3 to sort B cells in a mixture containing T cells and B cells or activated B cells or antigen-presenting cells.

[0070] 48. The image contains an isolated cell or the cell is classified as an isolated cell, i.e., the classification of the cell is that the cell is a single cell and is F-actin positive; - MHCII positive and CD3 negative, or -MHCII negative and CD3 positive 48. The use of the method according to any one of embodiments 46 to 47, wherein the cell is an isolated cell.

[0071] 49. The use according to any one of embodiments 46 to 47, wherein the image comprises a doublet or multiplet of cells or the cells are classified as a doublet or multiplet of cells, i.e., the classification of the cells is that the cells are a doublet or multiplet of cells if they are an aggregate of two or three cells and are F-actin positive, MHCII positive and CD3 positive.

[0072] 50. The use according to any one of embodiments 46 to 49, further comprising P-CD3 zeta.

[0073] 51. The use according to any one of embodiments 46 to 50, wherein the image includes a single B cell or antigen-presenting cell, or the cell is classified as a single B cell or antigen-presenting cell, i.e., the classification of the cell is that the cell is a single B cell or antigen-presenting cell if the cell is F-actin positive, MHCII positive, CD3 negative and P-CD3 zeta negative.

[0074] 52. The use of any one of embodiments 46 to 51, wherein the image includes a single T cell without signaling or the cell is classified as a single T cell without signaling, i.e., the classification of the cell is that the cell is a single T cell without signaling if the cell is F-actin positive, MHCII negative, CD3 positive and P-CD3 zeta negative.

[0075] 53. The use of any one of embodiments 46 to 52, wherein the image includes a single T cell without signaling or the cell is classified as a single T cell with signaling, i.e., the classification of the cell is that the cell is a single T cell with signaling if the cell is F-actin positive, MHCII negative, CD3 positive and P-CD3 zeta positive.

[0076] 54. The use according to any one of embodiments 46 to 53, wherein the image includes a doublet of B cells or antigen-presenting cells and T cells that forms a synapse without signal transduction, or the cells are classified as a doublet of B cells or antigen-presenting cells and T cells that forms a synapse without signal transduction, i.e., the cells are classified as a doublet of B cells or antigen-presenting cells and T cells that forms a synapse without signal transduction if the doublet is F-actin positive, MHCII positive, CD3 positive and P-CD3 zeta negative.

[0077] 55. Use according to any one of embodiments 46 to 54, wherein the image comprises doublets or multiplets that form synapses with signal transduction between one or more B cells or antigen-presenting cells and one or more T cells, or the cells are classified as doublets or multiplets that form synapses with signal transduction between one or more B cells or antigen-presenting cells and one or more T cells, i.e., the classification of the cells is such that if the doublets or multiplets are F-actin positive, MHCII positive, CD3 positive and P-CD3 zeta positive, the cells are doublets or multiplets that form synapses with signal transduction between one or more B cells or antigen-presenting cells and one or more T cells.

[0078] 56. The method or use according to any one of embodiments 1 to 55, wherein the multiplet is a multiplet of two B cells or antigen-presenting cells and one T cell.

[0079] 57. The method or use according to any one of embodiments 1 to 55, wherein the multiplet is a multiplet of one B cell or antigen-presenting cell and two T cells.

[0080] In addition to the various embodiments depicted and claimed, the presently disclosed subject matter is directed to other embodiments having other combinations of the features disclosed and claimed herein. Thus, specific features presented herein may be combined with each other in other manners within the scope of the presently disclosed subject matter, such that the presently disclosed subject matter includes any suitable combination of features disclosed herein. The foregoing descriptions of specific embodiments of the presently disclosed subject matter have been presented for purposes of illustration and description. They are not intended to be exhaustive or to limit the disclosed subject matter to the disclosed embodiments. DETAILED DESCRIPTION OF THE INVENTION

[0081] General definition It should be noted that as used in this specification and the appended claims, the singular forms "a," "an," and "the" include plural referents unless the context clearly dictates otherwise. Thus, for example, a reference to "a cell" includes a plurality of such cells and equivalents thereof known to those skilled in the art, and so forth. Similarly, the terms "a" (or "an"), "one or more," and "at least one" may be used interchangeably herein. It should also be noted that the terms "comprising," "including," and "having" may be used interchangeably.

[0082] The term "about" refers to a range of + / - 20% of the preceding numerical value. In certain embodiments, the term "about" refers to a range of ±10% of the preceding numerical value. In certain embodiments, the term "about" refers to a range of ±5% of the preceding numerical value.

[0083] As used herein, the terms "comprise(s)," "having / has," "can," "containing," and variations thereof are intended to be open-ended transitional phrases, terms, or words that do not exclude the possibility of additional acts or structures. The term "comprising" also encompasses the term "consisting of." The present disclosure also contemplates other embodiments that "comprising," "consisting of," and "consisting essentially of" the embodiments or elements presented herein, whether or not explicitly stated.

[0084] antibody General information regarding the nucleotide sequences of human immunoglobulin light and heavy chains is given in Kabat, EA, et al., Sequences of Proteins of Immunological Interest, 5th ed., Public Health Service, National Institutes of Health, Bethesda, MD (1991).

[0085] The term "antibody" as used herein is used in the broadest sense and encompasses a variety of antibody structures, including, but not limited to, full-length antibodies, monoclonal antibodies, multispecific antibodies (e.g., bispecific antibodies), and antibody-antibody fragment-fusions and combinations thereof.

[0086] The term "natural antibody" refers to naturally occurring immunoglobulin molecules with various structures. For example, native IgG antibodies are heterotetrameric glycoproteins of approximately 150,000 daltons composed of two identical light chains and two identical heavy chains that are disulfide-bonded. From the N-terminus to the C-terminus, each heavy chain has a heavy chain variable region (VH) followed by three heavy chain constant domains (CH1, CH2, and CH3), thereby positioning a hinge region between the first heavy chain constant domain and the second heavy chain constant domain. Similarly, from the N-terminus to the C-terminus, each light chain has a light chain variable region (VL) followed by a light chain constant domain (CL). The light chains of antibodies can be assigned to one of two types, called kappa (κ) and lambda (λ), based on the amino acid sequence of their constant domains.

[0087] The term "full-length antibody" refers to an antibody having a structure substantially similar to that of a natural antibody. A full-length antibody comprises two full-length antibody light chains, each comprising, from N- to C-terminus, a light-chain variable region and a light-chain constant domain, and two full-length antibody heavy chains, each comprising, from N- to C-terminus, a heavy-chain variable region, a first heavy-chain constant domain, a hinge region, a second heavy-chain constant domain, and a third heavy-chain constant domain. In contrast to natural antibodies, full-length antibodies may comprise additional immunoglobulin domains, such as one or more additional scFvs, or heavy- or light-chain Fab fragments, or scFabs conjugated to one or more of the ends of the different chains of the full-length antibody (but only one fragment at each end). These conjugates are also encompassed by the term full-length antibody.

[0088] The "class" of an antibody refers to the type of constant domain or constant region, preferably the Fc region, possessed by the heavy chain. There are five major classes of antibodies: IgA, IgD, IgE, IgG, and IgM, some of which may be further divided into subclasses (isotypes), e.g., IgG1, IgG2, IgG3, IgG4, IgA1, and IgA2. The heavy chain constant domains corresponding to the different classes of immunoglobulins are called α, δ, ε, γ, and μ, respectively.

[0089] The term "heavy chain constant region" refers to the region of an immunoglobulin heavy chain comprising the constant domains, i.e., the CH1 domain, hinge region, CH2 domain, and CH3 domain. In a specific embodiment, a human IgG constant region extends from Ala118 to the carboxyl terminus of the heavy chain (numbering according to the Kabat EU index). However, the C-terminal lysine (Lys447) of the constant region may or may not be present (numbering according to the Kabat EU index). The term "constant region" refers to a dimer comprising two heavy chain constant regions that may be covalently linked to each other via hinge region cysteine ​​residues that form interchain disulfide bonds.

[0090] The term "heavy chain Fc region" refers to the C-terminal region of an immunoglobulin heavy chain, including at least a portion of the hinge region (middle and lower hinge regions), CH2 domain, and CH3 domain. In a specific embodiment, the human IgG heavy chain Fc region extends from Asp221 or Cys226 or Pro230 to the carboxyl terminus of the heavy chain (numbering according to the Kabat EU index). Thus, the Fc region is smaller than the constant region, but the C-terminal portion is identical to it. However, the C-terminal lysine (Lys447) of the heavy chain Fc region may or may not be present (numbering according to the Kabat EU index). The term "Fc region" refers to a dimer comprising two heavy chain Fc regions, which can be covalently linked to each other via hinge region cysteine ​​residues that form interchain disulfide bonds.

[0091] The constant region of an antibody, more precisely the Fc region (and similarly the constant region), is directly involved in complement activation, C1q binding, C3 activation and Fc receptor binding. The effect of an antibody on the complement system depends on the specific conditions, but binding to C1q is caused by a defined binding site in the Fc region. Such binding sites are known in the prior art and are described, for example, in Lukas, TJ et al., J. Immunol. 127 (1981) 2555-2560; Brunhouse, R. and Cebra, JJ, Mol. Immunol. 16 (1979) 907-917; Burton, DR et al., Nature 288 (1980) 338-344; Thommesen, JE et al., Mol. Immunol. 37 (2000) 995-1004; Idusogie, EE et al., J. Immunol. 164 (2000) 4178-4184; Hezareh, M. et al., J. Virol. 75 (2001) 12161-12168; Morgan, A. et al., Immunology 86 (1995) 319-324, and European Patent No. 0307434. Such binding sites are, for example, L234, L235, D270, N297, E318, K320, K322, P331, and P329 (numbering according to the EU index of Kabat). Antibodies of the subclasses IgG1, IgG2, and IgG3 generally exhibit complement activation, C1q binding, and C3 activation, whereas IgG4 does not activate the complement system, does not bind C1q, and does not activate C3. The term "Fc region of an antibody" is well known to those skilled in the art and is defined based on papain cleavage of an antibody.

[0092] As used herein, the term "monoclonal antibody" refers to an antibody obtained from a population of substantially homogeneous antibodies; i.e., the individual antibodies comprising the population are identical and / or bind the same epitope, except for possible variant antibodies containing, for example, naturally occurring mutations or arising during production of the monoclonal antibody preparation, and such variants are generally present in minor amounts. In contrast to polyclonal antibody preparations, which typically include different antibodies directed against different determinants (epitopes), each monoclonal antibody of a monoclonal antibody preparation is directed against a single determinant on an antigen. Thus, the modifier "monoclonal" indicates the character of the antibody as being obtained from a substantially homogeneous population of antibodies and should not be construed as requiring production of the antibody by any particular method. For example, monoclonal antibodies can be produced by a variety of techniques, including, but not limited to, hybridoma methods, recombinant DNA methods, phage display methods, and methods utilizing transgenic animals containing all or part of the human immunoglobulin loci.

[0093] The term "valent" as used within this application refers to the presence of a specific number of binding sites on an antibody. Thus, the terms "bivalent," "tetravalent," and "hexavalent" refer to the presence of two binding sites, four binding sites, and six binding sites, respectively, on an antibody.

[0094] A "monospecific antibody" refers to an antibody that has a single binding specificity, i.e., that specifically binds to one antigen. Monospecific antibodies can be prepared as full-length antibodies or antibody fragments (e.g., F(ab')2), or combinations thereof (e.g., full-length antibodies with additional scFv or Fab fragments). Monospecific antibodies need not be monovalent; that is, they may contain two or more binding sites that specifically bind to one antigen. For example, naturally occurring antibodies are monospecific but bivalent.

[0095] A "multispecific antibody" refers to an antibody that has binding specificities for at least two different epitopes on the same antigen or two different antigens. Multispecific antibodies can be prepared as full-length antibodies or antibody fragments (e.g., F(ab')2 bispecific antibodies), or combinations thereof (e.g., full-length antibodies plus additional scFv or Fab fragments). Multispecific antibodies are at least bivalent, i.e., contain two antigen-binding sites. Furthermore, multispecific antibodies are at least bispecific. Thus, bivalent bispecific antibodies are the simplest form of multispecific antibodies. Engineered antibodies with two, three, or more (e.g., four) functional antigen-binding sites have been reported (see, e.g., U.S. Patent Application Publication No. 2002 / 0004587).

[0096] In certain embodiments of all aspects and embodiments of the present invention, the antibody is a multispecific antibody, e.g., at least a bispecific antibody. In certain embodiments, one of the binding specificities is for a first antigen and the other is for a different second antigen. In certain embodiments, the multispecific antibody can bind to two different epitopes of the same antigen. Multispecific antibodies may also be used to localize cytotoxic agents to cells expressing one or more antigens.

[0097] Multispecific antibodies can be prepared as full-length antibodies or antibody-antibody fragment fusions.

[0098] Techniques for producing multispecific antibodies include, but are not limited to, recombinant co-expression of two immunoglobulin heavy chain-light chain pairs with different specificities (see Milstein, C. and Cuello, A.C., Nature 305 (1983) 537-540; WO 93 / 08829; and Traunecker, A., et al., EMBO J. 10 (1991) 3655-3659) and "knobs-in-holes" engineering (see, e.g., U.S. Pat. No. 5,731,168). Multispecific antibodies can also be produced by manipulating electrostatic steering effects to create antibody Fc heterodimeric molecules (WO 2009 / 089004), cross-linking two or more antibodies or fragments (see, e.g., U.S. Pat. No. 4,676,980, and Brennan, M., et al., Science 229 (1985) 81-83), producing bispecific antibodies using leucine zippers (see, e.g., Kostelny, SA, et al., J. Immunol. 148 (1992) 1547-1553), using general light chain technology to circumvent light chain mispairing problems (see, e.g., WO 98 / 50431), using specific techniques to generate bispecific antibody fragments (see, e.g., Holliger, P., et al., Proc. Natl. Acad. Sci. USA 90 (1993) 6444-6448), and for example, by the preparation of trispecific antibodies in Tutt, A., et al., J. Immunol. 147 (1991) 60-69.

[0099] Also included herein are engineered antibodies with three or more antigen-binding sites, including, for example, "Octopus antibodies," or DVD-Igs (see, e.g., WO 2001 / 77342 and WO 2008 / 024715). Other examples of multispecific antibodies with three or more antigen-binding sites can be found in WO 2010 / 115589, WO 2010 / 112193, WO 2010 / 136172, WO 2010 / 145792, and WO 2013 / 026831. Bispecific antibodies or antigen-binding fragments thereof also include "dual-acting Fabs" or "DAFs" (see, e.g., U.S. Patent Application Publication Nos. 2008 / 0069820 and WO 2015 / 095539).

[0100] Multispecific antibodies can also be provided in an asymmetric manner with domain crossovers in one or more binding arms of the same antigen specificity, i.e., by exchanging VH / VL domains (see, e.g., WO 2009 / 080252 and WO 2015 / 150447), CH1 / CL domains (see, e.g., WO 2009 / 080253), or complete Fab arms (see, e.g., WO 2009 / 080251, WO 2016 / 016299; see also Schaefer et al., Proc. Natl. Acad. Sci. USA 108 (2011) 1187-1191, and Klein at al., MAbs 8 (2016) 1010-1020). In certain embodiments of all aspects and embodiments of the invention, the multispecific antibody comprises a Cross-Fab fragment. The term "Cross-Fab fragment" refers to a Fab fragment in which either the variable or constant regions of the heavy and light chains have been exchanged. A Cross-Fab fragment contains a polypeptide chain composed of a light chain variable region (VL) and a heavy chain constant region 1 (CH1), and a polypeptide chain composed of a heavy chain variable region (VH) and a light chain constant region (CL). Asymmetric Fab arms can also be engineered by introducing charged or uncharged amino acid mutations at the domain interface to direct proper pairing of the Fab heavy chain fragment with the cognate light chain. See, for example, WO 2016 / 172485.

[0101] The antibody or fragment may also be a multispecific antibody as described in WO 2009 / 080254, WO 2010 / 112193, WO 2010 / 115589, WO 2010 / 136172, WO 2010 / 145792 or WO 2010 / 145793.

[0102] The antibody or fragment thereof may also be a multispecific antibody as disclosed in WO 2012 / 163520.

[0103] Various additional molecular formats of multispecific antibodies are known in the art and can be produced using the cells according to the invention (see, e.g., Spiess et al., Mol. Immunol. 67 (2015) 95-106).

[0104] Bispecific antibodies are generally antibody molecules that specifically bind to two different, non-overlapping epitopes on the same antigen or to two epitopes on different antigens.

[0105] In certain embodiments of all aspects and embodiments, the antibody comprises: Domain-swapped full-length antibodies (i.e., a multispecific IgG antibody comprising a first Fab fragment and a second Fab fragment, wherein in the first Fab fragment: a) only the CH1 and CL domains have been replaced with each other (i.e. the light chain of the first Fab fragment comprises the VL and CH1 domains and the heavy chain of the first Fab fragment comprises the VH and CL domains), b) only the VH and VL domains have been replaced with each other (i.e. the light chain of the first Fab fragment comprises the VH and CL domains and the heavy chain of the first Fab fragment comprises the VL and CH1 domains), or c) the CH1 and CL domains are replaced with each other and the VH and VL domains are replaced with each other (i.e., the light chain of the first Fab fragment comprises the VH and CH1 domains and the heavy chain of the first Fab fragment comprises the VL and CL domains); the second Fab fragment comprises a light chain comprising a VL and a CL domain and a heavy chain comprising a VH and a CH1 domain; A domain-swapped full-length antibody can comprise a first heavy chain comprising a CH3 domain and a second heavy chain comprising a CH3 domain, both CH3 domains being engineered to be complementary by respective amino acid substitutions, e.g., as described in WO 96 / 27011, WO 98 / 050431, EP 1870459, WO 2007 / 110205, WO 2007 / 147901, WO 200 a multispecific IgG antibody that supports heterodimerization of a first heavy chain and an engineered second heavy chain as disclosed in WO 9 / 089004, WO 2010 / 129304, WO 2011 / 90754, WO 2011 / 143545, WO 2012 / 058768, WO 2013 / 157954 or WO 2013 / 096291 (which are incorporated herein by reference); Full-length antibody with domain swapping and an additional heavy chain C-terminal binding site (BS) (i.e., a multispecific IgG antibody, a) a full-length antibody comprising two pairs of full-length antibody light chains and two pairs of full-length antibody heavy chains, wherein a binding site formed by each pair of full-length heavy chains and full-length light chains specifically binds to a first antigen; b) one additional Fab fragment fused to the C-terminus of one heavy chain of the full-length antibody, wherein the binding site of the additional Fab fragment specifically binds to a second antigen; an additional Fab fragment that specifically binds to a second antigen i) a multispecific IgG antibody in which a) the light chain variable domain (VL) and the heavy chain variable domain (VH) are replaced with each other, or b) the light chain constant domain (CL) and the heavy chain constant domain (CH1) are replaced with each other, or ii) is a single chain Fab fragment; Single-arm single-chain antibodies (i.e., an antibody comprising a first binding site that specifically binds to a first epitope or antigen and a second binding site that specifically binds to a second epitope or antigen, wherein the individual chains are as follows: -Light chain (variable light domain + constant light kappa domain) -Light / heavy chain combination (variable light domain + light chain constant domain + peptide linker + variable heavy domain + CH1 + hinge + CH2 + CH3 with knob mutation) - a heavy chain (variable heavy domain + CH1 + hinge + CH2 + CH3 with hole mutation), Two-armed single-chain antibodies (i.e., an antibody comprising a first binding site that specifically binds to a first epitope or antigen and a second binding site that specifically binds to a second epitope or antigen, wherein the individual chains are as follows: Light Chain / Heavy Chain 1 Combination (VLCD + LCH Constant Domain + Peptide Linker + VHC Domain + CH1 + Hinge + CH2 + CH3 with Hole Mutation) - an antibody having a light chain / heavy chain 2 combination (variable light domain + light chain constant domain + peptide linker + variable heavy domain + CH1 + hinge + CH2 + CH3 with knob mutation); Common light chain bispecific antibodies (i.e., an antibody comprising a first binding site that specifically binds to a first epitope or antigen and a second binding site that specifically binds to a second epitope or antigen, wherein the individual chains are as follows: -Light chain (variable light domain + constant light domain) -Heavy Chain 1 (VH domain + CH1 + Hinge + CH2 + CH3 with hole mutation) - heavy chain 2 (variable heavy domain + CH1 + hinge + CH2 + CH3 with knob mutation), T cell bispecific antibody (TCB) (i.e., a full-length antibody with an additional domain-swapped heavy chain N-terminal binding site, - first and second Fab fragments, each binding site of the first and second Fab fragments specifically binding to a first antigen; - a third Fab fragment, wherein the binding site of the third Fab fragment specifically binds to a second antigen, and wherein the third Fab fragment comprises a domain crossover such that the variable light domain (VL) and the variable heavy domain (VH) are replaced by one another; and an Fc region comprising a first Fc region polypeptide and a second Fc region polypeptide; the first Fab fragment and the second Fab fragment comprise a heavy chain fragment and a full-length light chain, respectively; the C-terminus of the heavy chain fragment of the first Fab fragment is fused to the N-terminus of the first Fc region polypeptide; a full-length antibody in which the C-terminus of the heavy chain fragment of the second Fab fragment is fused to the N-terminus of the variable light chain domain of a third Fab fragment, and the C-terminus of the CH1 domain of the third Fab fragment is fused to the N-terminus of a second Fc region polypeptide; Antibody-multimer fusion (i.e., a multimeric fusion protein, (a) an antibody heavy chain and an antibody light chain; (b) a first fusion polypeptide comprising, from N-terminus to C-terminus, a first portion of a non-antibody multimeric polypeptide, an antibody heavy chain CH1 domain or an antibody light chain constant domain, an antibody hinge region, an antibody heavy chain CH2 domain, and an antibody heavy chain CH3 domain; and a second fusion polypeptide comprising, from N-terminus to C-terminus, a second portion of a non-antibody multimeric polypeptide, and an antibody light chain constant domain if the first polypeptide comprises an antibody heavy chain CH1 domain, or an antibody heavy chain CH1 domain if the first polypeptide comprises an antibody light chain constant domain; (i) the antibody heavy chain of (a) and the first fusion polypeptide of (b), (ii) the antibody heavy chain of (a) and the antibody light chain of (a), and (iii) the first fusion polypeptide of (b) and the second fusion polypeptide of (b), are each independently covalently linked to each other by at least one disulfide bond; The composite (multispecific) antibody is selected from the group of composite (multispecific) antibodies consisting of a multimeric fusion protein in which the variable domains of an antibody heavy chain and an antibody light chain form a binding site that specifically binds to an antigen.

[0106] The "knob-into-hole" dimerization module and its use in antibody engineering is described in Carter P.; Ridgway JBB; Presta LG: Immunotechnology, Volume 2, Number 1, February 1996, pp. 73-73(1).

[0107] The CH3 domains of antibody heavy chains can be modified using the "knob-into-hole" technique, which is described in detail with some examples in, for example, WO 96 / 027011, Ridgway, JB, et al., Protein Eng. 9 (1996) 617-621, and Merchant, AM, et al., Nat. Biotechnol. 16 (1998) 677-681. In this method, the interaction surfaces of two CH3 domains are altered to increase heterodimerization of these two CH3 domains, thereby increasing heterodimerization of polypeptides containing them. Each of the two CH3 domains (of the two heavy chains) can be a "knob," and the other can be a "hole." The introduction of disulfide bridges further stabilizes the heterodimer (Merchant, AM, et al., Nature Biotech. 16 (1998) 677-681; Atwell, S., et al., J. Mol. Biol. 270 (1997) 26-35) and increases the yield.

[0108] The mutation T366W in the CH3 domain (of an antibody heavy chain) is designated as a "knob mutation" or "mutated knob," and the mutations T366S, L368A, and Y407V in the CH3 domain (of an antibody heavy chain) are designated as "hole mutations" or "mutated hole" (numbering according to the EU index of Kabat). Additional interchain disulfide bridges between CH3 domains (Merchant, AM, et al., Nature Biotech. 16 (1998) 677-681) can also be used, for example, by introducing a S354C mutation in the CH3 domain of a heavy chain bearing a "knob mutation" (designated "knob-cys-mutation" or "mutated knob-cys") and a Y349C mutation in the CH3 domain of a heavy chain bearing a "hole mutation" (designated "hole-cys-mutation" or "mutated hole-cys") (numbering according to the EU index of Kabat).

[0109] The term "domain crossover" as used herein refers to deviations in domain sequence from that of a native antibody in that, in a pair of antibody heavy chain VH-CH1 fragment and its corresponding cognate antibody light chain, i.e., antibody Fab (fragment-antigen binding), at least one heavy chain domain is replaced by the corresponding light chain domain, or vice versa. There are three general types of domain crossovers: (i) crossovers of CH1 and CL domains, where the domain crossover in the light chain results in a VL-CH1 domain sequence and the domain crossover in the heavy chain fragment results in a VH-CL domain sequence (or a full-length antibody heavy chain having a VH-CL-hinge-CH2-CH3 domain sequence); (ii) domain crossovers of VH and VL domains, where the domain crossover in the light chain results in a VH-CL domain sequence and the domain crossover in the heavy chain fragment results in a VL-CH1 domain sequence; and (iii) domain crossovers of a complete light chain (VL-CL) and a complete VH-CH1 heavy chain fragment ("Fab crossover"), where the domain crossover results in a light chain with a VH-CH1 domain sequence and the domain crossover results in a heavy chain fragment with a VL-CL domain sequence (all domain sequences listed above are in the N-terminal to C-terminal direction).

[0110] As used herein, the term "replaced by one another" with respect to corresponding heavy and light chain domains refers to the domain crossover described above. Thus, when the CH1 and CL domains are "replaced by one another," the term refers to the domain crossover described in item (i) and the resulting heavy and light chain domain sequences. Thus, when the VH and VL are "replaced by one another," the term refers to the domain crossover described in item (ii), and when the CH1 and CL domains are "replaced by one another" and the VH and VL domains are "replaced by one another," the term refers to the domain crossover described in item (iii). Bispecific antibodies containing domain crossovers have been reported, for example, in WO 2009 / 080251, WO 2009 / 080252, WO 2009 / 080253, WO 2009 / 080254, and Schaefer, W. et al., Proc. Natl. Acad. Sci USA 108 (2011) 11187-11192. Such antibodies are generally referred to as CrossMabs.

[0111] In certain embodiments of all aspects and aspects of the present invention, the multispecific antibody comprises at least one Fab fragment comprising a domain crossover between the CH1 domain and the CL domain, or between the VH domain and the VL domain, or between the VH-CH1 domain and the VL-VL domain. In multispecific antibodies with domain crossover, Fabs that specifically bind to the same antigen are constructed to have the same domain sequence. Therefore, when two or more Fabs with domain crossover are contained in a multispecific antibody, the Fabs specifically bind to the same antigen.

[0112] A "humanized" antibody refers to an antibody that comprises amino acid residues from non-human HVRs and amino acid residues from human FRs. In certain embodiments, a humanized antibody comprises substantially all of at least one, and typically two, variable domains, in which all or substantially all of the HVRs (e.g., CDRs) correspond to those of a non-human antibody and all or substantially all of the FRs correspond to those of a human antibody. A humanized antibody may optionally comprise at least a portion of an antibody constant region derived from a human antibody. A "humanized form" of an antibody, e.g., a non-human antibody, refers to an antibody that has undergone humanization.

[0113] The term "recombinant antibody," as used herein, means all antibodies (chimeric, humanized, and human) that are prepared, expressed, generated, or isolated by recombinant means, such as using cells according to the invention. This includes antibodies isolated from recombinant cells, such as NS0, HEK, BHK, amniotic cells, or CHO cells, that have been modified according to the invention.

[0114] As used herein, the term "antibody fragment" refers to a molecule other than an intact antibody that contains a portion of an intact antibody that binds to the same epitope of the same antigen as the intact antibody, i.e., it is a functional fragment. Examples of antibody fragments include, but are not limited to, Fv, Fab, Fab', Fab'-SH, F(ab')2, bispecific Fab, diabody, linear antibody, and single-chain antibody molecules (e.g., scFv or scFab).

[0115] Specific embodiments of the method according to the invention Here, scifAI, a machine learning framework for efficient and explainable analysis of high-throughput imaging data based on modular implementation, is reported.

[0116] The methods according to the present invention are shown herein to have potential for (i) predicting immunologically relevant cell class frequencies, (ii) systematic morphological profiling of immunological synapses, (iii) investigating inter-donor and inter- and intra-experiment variability, and (iv) characterizing the mechanism of action of therapeutic antibodies, and (v) predicting their functionality in vitro.

[0117] The present invention is based, at least in part, on the discovery that high-throughput imaging of the immunological synapse using IFC, combined with specific data pre-processing and machine learning, enables the screening of novel antibody candidates and improved evaluation of lead molecules with respect to functionality, insight into mode of action, and antibody properties such as affinity, avidity, and format.

[0118] The present invention is exemplified below using specific antibodies and techniques, which are presented only as examples of the functionality of the invention and should not be construed as limiting, with the true scope of the invention being set forth in the appended claims.

[0119] Comprehensive multi-channel imaging flow cytometry dataset of the immunological synapse Using high-throughput IFC, a comprehensive dataset was generated for the systematic analysis of the immunological synapse of T cell / B cell conjugates (TB conjugates) (see Figures 1 and 2). Human memory CD4+ T cells isolated from peripheral blood of different donors were cocultured with superantigen (Staphylococcus aureus enterotoxin A, SEA)-pulsed EBV-transformed lymphoblastoid B cells (B-LCLs) expressing high levels of the costimulatory molecules CD86 and CD80, or were left untreated (Figures 3-6). To investigate the functional immune synapse, we selected P-CD3 zeta (Y142), the highest titer of SEA (100 ng / mL), and a 45-minute time point as a readout of early T cell activation (Figures 7 and 8). In total, we screened nine donors in four independent experiments (Figure 2), and acquired 1,182,782 images (±SEA, Figure 3).

[0120] A suitable multichannel panel for analysis and classification was found to consist of bright field (BF), F-actin (cytoskeleton), MHCII, CD3, and P-CD3 zeta (TCR signaling), which allowed us to capture a wide range of biologically motivated features of the immunological synapse (Figure 1).

[0121] Dead, deformed, out-of-focus or clipped cells were removed using a multi-step pipeline (see Examples).

[0122] Furthermore, a set of 5,221 images from seven randomly selected donors was classified by an experienced immunologist into nine classes organized into two levels (Figures 3 and 9). The first level represented the number of cells present in the image: singlet (n = 1), doublet (n = 2), and multiplet (n > 2). The second level characterized the cell types, their interactions, and the presence of TCR signaling. Singlets consisted of the "single B-LCL" class, the "single T cell signaling" class, and the "single T cell with signaling" class. Hereinafter, "without" will be abbreviated as "w / o" and "with" as "w / ." Doublets included the "T cell with small B-LCL" class, the "B-LCL and T cell in one layer" class, the "synapse without signaling" class, the "synapse with signaling" class, and the "no cell-cell interaction" class. The class "multisynaptic" includes three or more cells and at least one B-LCL and T cell.

[0123] Without being bound by this theory, it is assumed that the "T cells with small B-LCL" and "no cell-cell interactions" classes were experimental artifacts, although they were annotated to increase the predictive power of the classification model, and were subsequently excluded and not used in further analyses (see Examples).

[0124] ScifAI: An explainable AI framework for the analysis of multi-channel imaging flow cytometry data Herein, we report the single-cell imaging flow cytometry AI (scifAI) module.

[0125] The module, universally applicable to single-cell imaging projects, provides functions for importing and preprocessing input data, several feature engineering pipelines including implementations of biologically motivated features and feature sets generated by autoencoders (see examples), and methods for efficient and meaningful feature selection.

[0126] Furthermore, this module implements several machine learning and deep learning models to train supervised image classification models for the prediction of cellular organizations, such as immunological synapses. Following the principles of multi-instance learning, this module also implements the functionality to regress a set of selected images against downstream sequential readouts, such as cytokine production.

[0127] Extensive documentation, as well as example code in the form of Jupyter notebooks, is provided online at https: / / github.com / marrlab / scifAI / and https: / / github.com / marrlab / scifAI-notebooks.

[0128] ScifAI for high-throughput profiling of the immunological synapse To unbiasedly characterize immunological synapses, a set of biologically motivated, interpretable features was first designed and computed using the scifAI module. These features were based on morphology, intensity, colocalization, texture, and synaptic features extracted from five-panel staining images and their corresponding masks (see Examples and Figures 10-11). Synaptic features were implemented based on the ratio of signal intensity of each fluorescent channel within the synaptic region to that of the entire cell. Taking advantage of a large amount of unlabeled data, a multi-channel autoencoder was implemented to learn a second set of data-driven features from images in an unsupervised manner

[24] . The autoencoder was designed to encode images into a 128-dimensional abstract feature space by reconstructing the input image (see Examples).

[0129] We then used scifAI to construct a supervised machine learning pipeline for the classification of 5,221 annotated images across nine immunologically relevant cell classes. We trained and benchmarked a series of supervised machine learning models for the prediction of all nine classes using both interpretable feature spaces and abstract autoencoder features across all donors and experimental conditions. The models included an XGBoost classifier for interpretable features and a multiclass logistic regression (LR) for interpretable and data-driven features. A feature preselection pipeline using an ensemble of different methods was implemented to preselect features and reduce dimensionality (see Examples and Figures 12-13). For comparison, we trained several convolutional neural network (CNN) architectures, including Resnet18, ResNet34, DeseNet121, and DeepFlow, which have previously been shown to be successful in classification tasks on imaging flow cytometry data [22, 24, 28]. CNN architectures inherently learn feature representations based on input images and their corresponding labels. All models were trained on a stratified subset containing 2923 (70%) annotated images. To benchmark the classification model and feature space combination, we compared the macro F1 scores against the remainder of the images as a holdout test set containing 1567 (30%) annotated images (see Examples). The XGBoost model using the interpretable feature set performed best among all classifiers (F1-macro=0.93±0.01, mean±standard deviation bootstrap, n=1000).

[0130] Therefore, in particular embodiments of all aspects and embodiments of the method according to the invention, an XGBoost model that uses an interpretable feature set is used.

[0131] The XGBoost model was followed by the convolutional neural networks ResNet34 (0.92 ± 0.01), ResNet18 (0.91 ± 0.01), DeepFlow (0.90 ± 0.01), DenseNet121 (0.90 ± 0.02), multiclass logistic regression using an interpretable feature set (0.89 ± 0.02), and logistic regression using a data-driven feature set (0.83 ± 0.02).

[0132] The XGBoost model was found to offer the best compromise between performance and explainability, and was therefore selected as the final classifier for label augmentation to the full dataset (Figure 4).

[0133] Examination of the confusion matrix of the model on the holdout set revealed that misclassification occurred primarily within the signaling properties of cell classes, while all other classes showed good agreement overall (see Figure 14).

[0134] After training the XGBoost classifier, we investigated which underlying features drove class prediction. These features were therefore ranked by their respective Gini index (see Figure 5). The most predictive features were based on CD3 and MHCII colocalization, MHCII and P-CD3 zeta colocalization, MHCII and CD3 texture, and P-CD3 zeta intensity.

[0135] In certain embodiments of all aspects and embodiments of the methods according to the invention, cell doublets and multiplets are classified in addition to being synaptic based on one or more of the following: - the (correlated) distance between MHCII and CD3 markers, - the distance between the centers of MHCII and CD3 labels, - Manders overlap coefficient of MHCII and P-CD3 zeta labels, - uniformity of CD3 labeling, -MHCII labeling contrast, -kurtosis intensity of P-CD3 zeta labeling, - Manders overlap coefficient of MHCII and CD3 labels, and / or -Maximum intensity of CD3 labeling.

[0136] In one preferred embodiment of all aspects and embodiments of the method according to the invention, cell doublets and multiplets are classified in addition to being synaptic based on one or more of the following: - co-localization of CD3 and MHCII markers, and / or - Co-localization of MHCII and P-CD3 zeta labeling, and / or - MHCII texture, and / or -CD3 label texture, and / or -Intensity of P-CD3 zeta labeling.

[0137] Without being bound by this theory, it is hypothesized that based on the class characteristics and definitions, (i) the texture of the CD3 and MHCII labels can be used to detect the presence of T cells and B-LCL cells in an image, (ii) the co-localization of the CD3 and MHCII labels can be used to detect different doublet types, and (iii) the intensity of the P-CD3 zeta label and the co-localization of the MHCII and P-CD3 zeta labels can be used to detect whether it is a signaling T cell (see Figure 15).

[0138] Annotated data and a subset of available IFC channels are sufficient for high classification performance How many annotated samples were needed to achieve reasonable classification performance remains to be investigated. Therefore, we repeatedly trained the model using stratified subsets of the training data and evaluated the F1 macros on the test set. The results showed that by using 1500 images (45% of the training data), we could achieve 90% of the F1 macros on the test set (Figure 16).

[0139] Furthermore, we investigated which channels were sufficient to reach high performance. Therefore, we kept the BF channel and trained the model using all possible combinations of the fluorescence channels.

[0140] Channels BF, MHCII and P-CD3 zeta were found to be sufficient to reach the F1 macro as well as using all channels (Figure 17).

[0141] Characterization of the effects of therapeutic antibodies on synaptogenesis The effect of therapeutic antibodies on the formation of immunological synapses and their morphological profile are being further investigated to better characterize. This analysis included the investigation of potential class frequency changes and feature differences.

[0142] Two antibodies were used in the study: one immune response activator and one immune response inhibitor. An activating T cell bispecific (TCB) antibody was designed to target the B cell coreceptors CD3 and CD19

[29] (see Figure 18). The inhibitory antibody teplizumab has been described to bind only to CD3 (see Figure 19) and has been shown to attenuate T cell responses [30, 31]. For each antibody, appropriate controls (Ctrl-TCB and isotype, respectively) were run within the same experiment and donor. Because teplizumab required an existing immune response for subsequent inhibition, T cells were first stimulated using SEA (see Figure 18). The same setup was used for the isotype control. Six donors across two experiments for CD19-TCB and seven donors across three experiments for teplizumab were measured (see Figures 20-23). To examine class frequency changes between antibodies and their controls, we used the previous XGBoost classifier (see Figures 4 and 15) to predict the class of all images based on interpretable features (see Examples and Figure 24). To ensure that the pre-trained XGBoost model was transferable from ±SEA to antibody experiments, experts annotated randomly selected subsets of 396 images for CD19-TCB and 227 images for teplizumab. The high agreement between the expert annotations and XGBoost predictions in the new experiments (macro F1 score = 0.86 for TCB and 0.85 for teplizumab) confirmed that the trained model was generalizable and therefore usable for further analysis (see Figure 25).

[0143] To compactly represent the changes in class frequencies, log2-fold change values ​​were calculated between antibodies and their respective controls.

[0144] To examine the differences in synaptic features under antibody stimulation, we selected images predicted as "synapses with signaling" for each donor and compared features interpretable from the fluorescence channels alone, including texture, synaptic features, morphology, intensity, and colocalization, between the antibody and their controls. We did not include the BF channel because its intensity is difficult to interpret and its morphological features are captured by other fluorescence channels (see Methods).

[0145] In certain embodiments of all aspects and embodiments according to the invention, the method is for examining class frequency changes in the presence of a therapeutic antibody, examining the number or frequency of synaptic doublets and multiplets with signaling in the absence and presence of a therapeutic antibody, and / or comparing features interpretable from the fluorescence channels including texture, synaptic features, morphology, intensity and co-localization between the antibody and their controls.

[0146] CD19-TCB increases the formation of stable immune synapses Stimulation of the immune response with CD19-TCB resulted in a significant increase in doublet and multiplet frequencies, with the "synapse with signaling" class showing the highest increase (median log_2(CD19-TCB / Ctrl-TCB) = 2.6, n = 6 donors, p = 0.036), followed by "multiplets of two B cells and one T cell forming synapses with signaling" (median = 1.94, p = 0.036), "B-LCL and T cell in one layer" (median = 1.77, p = 0.036), and "doublets or multiplets of one or two B cells and one T cell forming synapses without signaling" (median = 0.44, p = 0.036). For singlets, the overall trend was a decrease in class frequencies of "single B-LCL" (median=-0.29, p=0.036) and "single T cell without signaling" (median=-0.78, p=0.036) (see Figure 26).

[0147] We further analyzed the differences in synaptic features induced by CD19-TCB (see Methods) and compared 210 interpretable features from all fluorescence channels. Of 210 × 6 = 1,260 possible combinations of features and donors, we found that 210 features significantly increased and 163 features significantly decreased (see Figure 27). All donors responded similarly to stimulation with CD19-TCB. On average, 27 ± 4 features significantly decreased and 33 ± 7 features significantly increased per donor (lower dashed line in Figure 27). From these features, several features were identified that had similar changes in at least four of the six donors (Figure 27 and Table 1).

[0148] Table 1: Significant features induced by CD19-TCB. Features that were significantly changed for at least four donors after addition of CD19-TCB are shown. This table represents a list of consistent features from Figure 27. 1 represents a significant increase (red in Figure 27), -1 represents a significant decrease (blue in Figure 27), and 0 represents no significant change (gray in Figure 27). TIFF2025535744000002.tif239159TIFF2025535744000003.tif165159

[0149] Furthermore, similar to SEA stimulation, an increase in the "mean intensity of P-CD3 zeta" was identified, resulting in a higher enrichment within the synaptic region (see Figures 28-29 and Table 1).

[0150] Furthermore, higher concentrations of F-actin and MHCII were found at the synapses (FIGS. 31 to 34).

[0151] Thus, the addition of therapeutic antibodies resulted in an increased frequency of doublets and multiplets, as well as a higher enrichment of F-actin and MHCII in the synaptic region. Without being bound by theory, this may indicate enhanced formation of a tight immunological synapse that translates into efficient TCR signaling. These observations are generally consistent with the mechanism of action previously described for TCBs, which promote stable interactions between tumor cells and T cells [32-33].

[0152] Teplizumab alters synaptogenesis and TCR signaling In contrast to CD19-TCB, the presence of teplizumab significantly reduced the frequency of doublets and multiplets (Figure 35). The highest reduction was observed in the class of "doublets of one B cell and one T cell forming synapses with signaling" (median log_2(teplizumab / isotype) = -0.75, n = 7, p = 0.018), followed by "multiplets of two B cells and one T cell forming synapses with signaling" (median = -0.51, p = 0.031), "doublets of B cells and T cells forming synapses without signaling" (median = -0.44, p = 0.018), and "B-LCLs and T cells in one layer" (median = -0.18, p = 0.018). Thus, "single T cells with signaling" (median=0.66, p=0.018) and "single B-LCLs" (median=0.07, p=0.018) were significantly increased compared to the isotype. Surprisingly, the frequency of the "T cells without signaling" class was significantly decreased (median=-0.35, p=0.018) (see Figure 35), likely due to the significant increase in "single T cells with signaling."

[0153] We analyzed the differences in the characteristics of teplizumab-induced synapses in seven donors. From the "B cell-T cell doublets forming synapses without signaling," we extracted 132 image features based on F-actin, MHCII, and P-CD3 zeta and their colocalization. Because of interference between teplizumab and anti-CD3 staining antibodies, we could not include CD3 features in the analysis. Therefore, we used anti-CD4 staining antibodies to identify T cells. Of 132 × 7 = 924 possibilities, 131 significantly increased features and 169 significantly decreased features were identified (Figure 30 and Table 2).

[0154] Table 2: Significant features induced by teplizumab. Features that changed significantly for at least six donors after the addition of teplizumab are shown. The table represents a list of consistent features from Figure 30. 1 represents a significant increase (red in Figure 30), -1 represents a significant decrease (blue in Figure 30), and 0 represents no significant change (gray in Figure 30). TIFF2025535744000004.tif212159

[0155] Notably, teplizumab resulted in, on average, 25 ± 8 significantly decreased features and 19 ± 17 significantly increased features per donor (Figure 30, bottom dashed line). Donor 6 showed the fewest changes, with five significantly increased features. In contrast, donor 4 showed the greatest number of increased features, with 50 features. A set of significantly increased or decreased features was identified for at least five of the seven donors (Figures 36-37). A decrease in the mean intensity of F-actin was identified, whereas donors 2 and 4 showed significant increases (Figures 36 and 38). This opposite response in the two donors could also be detected for other F-actin-related features (Table 2). In addition to changes in F-actin features, a significant decrease in P-CD3 zeta intensity within the synapse was also detected, and stronger clustering of TCR signaling around the entire T cell was observed (Figures 37 and 39).

[0156] Thus, the method according to the present invention identified a decrease in the number of synapses and alterations in F-actin reorganization and P-CD3 zeta signaling to synapses, providing new insight into the immunosuppressive mechanism of action of teplizumab.

[0157] conclusion The method according to the invention allows for the analysis of the mechanism of action of therapeutic antibodies.

[0158] Furthermore, the method according to the invention makes it possible to predict the functionality of therapeutic antibodies in vitro.

[0159] The present invention is based, at least in part, on the generation of morphological profiles of the immunological synapse that allow characterization of the mechanism of action of therapeutic antibodies early after the initiation of an immune response, which allows prediction of the associated downstream T cell responses.

[0160] These findings and methods differ from previous studies [21, 35], which did not consider inter-experimental effects on synaptogenesis.

[0161] The present invention is based, at least in part, on the detection and identification of changes in the immunological synapse. This is achieved, at least in part, by incorporating interpretable features extracted from fluorescence images into a machine learning framework. This makes the method scalable, produces reproducible results, and facilitates deployment in existing workflows, unlike previous studies [27, 36, 37] that use a combination of analytical steps.

[0162] Without being bound by this theory, it is hypothesized that the combination of interpretable features and explainable machine learning enabled the identification of relevant morphological classes, such as immunological synapses, with accuracy comparable to or even better than state-of-the-art methods, thereby enabling the analysis of the morphological profile of immunological synapses in an unbiased manner, as well as the characterization of antibody mechanisms of action in a biologically relevant context.

[0163] Therefore, our method is an improvement compared to known methods [26, 38] that primarily focus on performance rather than interpretability.

[0164] The power of the method according to the invention is demonstrated by analyzing the effect of two therapeutic antibodies on the immunological synapse, CD19-TCB and teplizumab, both of which bind to CD3 and have been described to activate and suppress T cell responses, respectively [29-31].

[0165] The method according to the invention has been found to result in the formation of more stable immune synapses in the presence of CD19-TCB, as indicated by a higher concentration of MHCII and F-actin within the synapse, paralleled by a higher intensity of P-CD3 zeta labeling.

[0166] The present method demonstrated that the presence of teplizumab reduced synapse formation, prevented F-actin reorganization, and localized P-CD3 zeta to synapses. These findings provide new insight into the immunosuppressive mechanism of action of teplizumab, which has not been extensively studied in vitro [30, 31].

[0167] Without being bound by this theory, it can be hypothesized that the decreased P-CD3 zeta intensity in the synaptic region and the observed non-polarized distribution of P-CD3 zeta signals around the entire T cell may indicate alterations in TCR signaling that may lead to reduced T cell effector function. Higher numbers of peripheral P-CD3 zeta microclusters have been reported in autoreactive T cells with altered synapse formation and abnormal T cell responses [6].

[0168] Unexpectedly, the method according to the invention allowed the identification of features within synapse classes that reveal donor-to-donor variability upon stimulation with different antibodies.

[0169] Thus, the method according to the invention allows rapid screening of responders in vitro and pre-selection of suitable patients for clinical trials.

[0170] In summary, by applying the method according to the present invention it was possible to fully investigate the mechanism of action of therapeutic antibodies based on significant features and also to gain more insight into donor-to-donor variability that may lead to different functional outcomes in vivo.

[0171] Our method improves on state-of-the-art methods by incorporating biologically motivated features such as texture, intensity statistics and synapse-related features.

[0172] For the first time, interpretable features of the immunological synapse were used to predict the efficacy of therapeutic antibodies on T cell cytokine production, making it possible to predict the functional outcome of unseen antibodies and identify the facilitating factors required for prediction.

[0173] For example, in the case of TCB, MHCII intensity and F-actin morphology were found to be the most prominent features in predicting cytokine readout.

[0174] The ability to predict unseen antibodies allows for different antibody formats to be explored to better understand mechanistically how different formats may affect T cell responses and help guide format selection.

[0175] Methods according to the present invention encompass data acquisition and analysis that can be tailored to explore various hypotheses and develop a variety of applications based on imaging flow cytometry data.

[0176] For example, in a recent example, memory CD4+ T cells were analyzed as poised to exhibit faster immune responses and higher synaptic propensity compared to naive T cells

[49] , while imaging and analysis of CD8+ T cells as key players in cytotoxicity could similarly elaborate how synaptic features correlate with the killing efficiency of therapeutic antibodies against tumor cells.

[0177] The method according to the present invention can also be used in the design of IFC experiments to optimize the number and type of stains and the total number of images acquired per donor.

[0178] The methods according to the invention can be used to improve the quality and speed of antibody development, for example to provide new insight into the mechanism of action of a particular candidate molecule or to predict in vitro efficacy in a high-throughput manner. Identification of lead molecules and better prioritization with respect to epitope, affinity, avidity and antibody format will have a major impact on the decision-making process.

[0179] Moreover, the methods according to the present invention may further help to identify responders among a patient population and predict their clinical outcome.

[0180] All references cited herein are expressly incorporated by reference in their entirety.

[0181] The following examples and figures are provided to aid the understanding of the present invention, the true scope of which is set forth in the appended claims. It is understood that modifications can be made in the procedures set forth without departing from the spirit of the invention. [Brief explanation of the drawings]

[0182] [Figure 1]Schematic of the data generation and analysis pipeline. To systematically analyze the immunological synapse of TB cell conjugates, 1,182,782 images were acquired using an imaging flow cytometer. These can then be manually classified (right) or scifAI (left) can be used to extract morphological features and train machine learning models to profile the immunological synapse and characterize the functionality of therapeutic antibodies. [Figure 2] Gating strategy to identify single interacting TB-LCL synapses using the IDEAS software of the imaging flow cytometer. [Figure 3] A subset of 5221 images was manually annotated by experts into nine immunologically relevant classes that could be categorized as singlets (either B or T cells), doublets (containing one B cell and one T cell), and multiplets (containing three or more cells). Cell images are shown in bright field (BF, scale bar = 2.4 μm), F-actin (cytoskeleton), MHCII, CD3, and P-CD3 zeta (markers of TCR signaling). [Figure 4] We benchmarked six different approaches to training predictive machine learning models for identifying immunologically relevant classes, combining different classification algorithms and feature engineering strategies. These approaches included interpretable (interp.) features combined with an explainable classifier, an autoencoder for generating data-driven features, an explainable classifier, and three convolutional neural networks. Interpretable features combined with an XGBoost classifier provided the best tradeoff between interpretability and classification performance. [Figure 5] A list of donors, their ages, sex and experiment numbers used in this experiment. [Figure 6] List of experiments and donors with and without SEA. [Figure 7]Testing assay conditions using conventional FACS. Primary memory CD4+ T cells isolated from PBMCs of healthy donors were stimulated with B-LCL cells in the presence of different concentrations of SEA (0.1–100 ng / mL) or left untreated (-SEA). The frequencies of P-CD3ζ+ (P-CD3 zeta positive), TNFα+ (TNF alpha positive), and CD69+CD4+ (CD69 and CD4 positive) T cells were determined at various time points. The small FACS histograms in the bar graphs show the expression levels of the three markers after 60 minutes, comparing the highest concentration of SEA (100 ng / mL) with the untreated control (-SEA). Data shown represent one experiment using T cells from three different donors. [Figure 8] Percentage of single TB-LCL synapses and P-CD3ζ+CD4+ (P-CD3 zeta and CD4 positive) T cells measured by imaging flow cytometry between two different SEA concentrations (10 and 100 ng / mL) after 45 and 120 minutes. Data represent two donors. [Figure 9] Number of expert labelings per donor. [Figure 10] A visual representation of each multi-channel image and the corresponding mask. The mask was exported along with the image from IDEAS software. [Figure 11] List of interpretable features. Morphology, intensity statistics, texture, synaptic features are based on one channel. Colocalization features are based on two channels. ScifAI automatically detects existing channels and generates the specified features. [Figure 12] A feature preselection pipeline to reduce the dimensionality of the feature space and remove multicollinearity. First, highly correlated features were removed. Then, an ensemble of different classifiers was trained on the data and their top-k features were selected. Finally, hierarchical clustering was performed on the union of features to account for multicollinearity. [Figure 13]The number of selected features before passing them to the XGBoost classifier. To obtain the best top-k, the data selection pipeline + XGBoost was trained using a stratified, randomly selected 85% of the training set and tested using the remaining 15%. [Figure 14] Confusion matrix of data selection pipeline (top-k=211) + XGBoost based on predictions on the test set. [Figure 15] The top eight features for cell class detection were ranked based on the Gini index. These features include colocalization of MHCII, CD3, and P-CD3ζ (P-CD3 zeta), texture, and intensity. Exemplary images are from donor 7, sampled from the 5th, 50th, and 95th percentiles of each feature's distribution. [Figure 16] Number of annotated images versus classification performance. [Figure 17] The classifier was trained using an XGBoost model. The training data was used for this evaluation using 5-fold cross-validation. At each step, features based on the selected channel were used to train the classifier. Brightfield (BF) is the unstained channel and is therefore always retained in the data. Combinations are ranked based on the F1-macro. [Figure 18] Schematic of teplizumab's mechanism of action. [Figure 19] Schematic diagram of the mechanism of action of CD19-TCB. [Figure 20] Donors and their respective experiments used for class frequency analysis and feature difference analysis. [Figure 21] Donors and their respective experiments used for class frequency analysis and feature difference analysis. [Figure 22] List of experiments and donors using TCB and its controls. [Figure 23] List of experiments and donors with teplizumab and isotype controls. [Figure 24] Class frequencies for the major classes described in Methods. Each dot represents a donor color-coded by experiment. [Figure 25] Confusion matrices for classification of CD19-TCB and teplizumab based on 396 and 227 expert-annotated images, respectively. The pre-trained model (Figure 4) reached macro F1 scores of 0.86 and 0.85, respectively, with both datasets. [Figure 26] Differences in class frequencies shown as log2 fold change between CD19-TCB and its corresponding control (Ctrl TCB). Each dot represents a donor, color-coded as in Figure 21. The vertical black lines are the median across donors for each class. [Figure 27] Systematic comparison of 210 relevant features between CD19-TCB and Ctrl-TCB across images predicted as "synapses with signaling" across six donors. Each row represents a feature, and each column represents a donor. For each donor, significantly increased features are shown in dark gray / black, and significantly decreased features are shown in light gray. Donors are sorted based on the number of significantly changed features. The bar graph below shows the number of increased or decreased features per donor. [Figure 28] Statistical and visual inter-donor comparison of the representative feature "Mean Intensity P-CD3 Zeta" between CD19-TCB and Ctrl-TCB. For visualization, the feature is mapped between 0 and 1 separately for each donor. [Figure 29] A visual representation of the representative feature "mean intensity P-CD3 zeta" randomly sampled for both Ctrl-TCB and CD19-TCB from donor 9 was found to be consistent with the statistical results (scale bar = 2.4 μm). [Figure 30] Systematic comparison of 132 relevant features between teplizumab and isotype across images predicted as "synapses with signaling" among all six donors. Each row represents a feature and each column represents a donor. For each donor, significantly increased features are shown in dark gray / black, and significantly decreased features are shown in light gray. Donors are sorted based on the number of significantly changed features. The bar graph below shows the number of increased or decreased features per donor. [Figure 31]Statistical and visual inter-donor comparison of the representative feature "F-actin enrichment at synapses" between CD19-TCB and Ctrl-TCB. For visualization, the feature is mapped between 0 and 1 separately for each donor. [Figure 32] Visual representation of a representative feature, “F-actin enrichment at synapses,” randomly sampled for both Ctrl-TCB and CD19-TCB from donor 9, and found to be consistent with statistical results (scale bar = 2.4 µm). [Figure 33] Statistical and visual inter-donor comparison of the representative feature "MHCII enrichment at synapses" between CD19-TCB and Ctrl-TCB. For visualization, the feature is mapped between 0 and 1 separately for each donor. [Figure 34] Visual representation of a representative feature, "MHCII enrichment at synapses," randomly sampled for both Ctrl-TCB and CD19-TCB from donor 9, found to be consistent with statistical results (scale bar = 2.4 µm). [Figure 35] Difference in class frequency shown as log2 fold change between teplizumab and its corresponding control (isotype). Each dot represents a donor, color-coded as in Figure 22. The vertical black line is the median across donors for each class. [Figure 36] Statistical and visual inter-donor comparison of F-actin signatures between teplizumab and its isotype. [Figure 37] Statistical and visual inter-donor comparison of the characteristic P-CD3 zeta between teplizumab and its isotype. [Figure 38] Visual representations of the characteristic F-actin were randomly sampled from donor 3 for both isotype and teplizumab and were found to be consistent with the statistical results (scale bar = 2.4 μm). [Figure 39] A visual representation of the feature P-CD3 zeta was randomly sampled from donor 3 for both isotype and teplizumab and was found to be consistent with the statistical results (scale bar = 2.4 μm). [Figure 40] 1 is an algorithmic flowchart of a method for classifying cells in a cell mixture using single cell imaging flow cytometry combined with artificial intelligence (scifAI) according to the present invention.

[0183] Materials and Methods cell culture EBV-transformed B-lymphoblastoid cell lines (B-LCLs) from donor 333 were obtained from Astarte Biologics (#1038-3161JN16). Cells were cultured in RPMI-1640 medium (PAN-Biotech, Catalog #P04-17500) containing 10% FBS (Amprotech, Catalog #AC-SM-0014Hi) and 2 mM L-glutamine (PAN-Biotech, Catalog #P04-80100). Z138 (MCL, kindly donated by the University of Leicester) and Nalm-6 (ALL, DSMZ ACC 128) tumor cells were cultured in RPMI-1640 containing 10% FBS and 1% Glutamax (Invitrogen / Gibco #35050-038).

[0184] Immune synapse formation and imaging flow cytometry To analyze the immune synapse, human memory CD4+ T cells were isolated from PBMCs of nine healthy human donors using the Stemcell Technologies Negative Selection EasySep Enrichment Kit (Cat. No. 19157). Live / dead staining of T cells and B-LCL cells was performed separately for 15 minutes at room temperature using the fixable viability dye eF780 (eBioscience, Cat. No. 65-0865-14). Cells were then resuspended in RPMI-1640 medium supplemented with 10% FBS (Amprotech, Cat. No. AC-SM-0014Hi), 5% penicillin-streptomycin (Gibco, Cat. No. 15140-122), and 2 mM L-glutamine (PAN-Biotech, Cat. No. P04-80100). B-LCL cells were then transferred to wells of a 96-well round-bottom plate (300,000 cells / well) and either preincubated with the superantigen Staphylococcal enterotoxin A (SEA) (Sigma-Aldrich, Cat. No. S9399) for 15 min at 37°C or left untreated. mem The cells were added to the B-LCL cells (250,000 cells / well) prepared above at a final ratio of 4:3 (B-LCL:T mem ), followed by infusion of the appropriate in-house generated compounds (10 μg / mL isotype Ctrl or teplizumab and 1 μg / mL (5 nM) Ctrl-TCB or CD19-TCB) into B-LCL-T mem To enhance conjugate formation between B-LCL cells and T cells, the cells were centrifuged at 300 x g for 30 seconds and then transferred to a 37°C incubator for 45 minutes. The medium in each well was then carefully aspirated with a pipette, and the cells were immediately fixed at room temperature for 12 minutes and then permeabilized with Foxp3 / Transcription Factor Staining Buffer (eBioscience, Catalog No. 00-5523-00).

[0185] Intracellular staining was performed in permeabilization buffer containing fluorescently labeled antibodies CD3-BV421 (clone UCHT1, Biolegend, catalog no. 300433), HLA-DR-PE-Cy7 (clone L243, Biolegend, catalog no. 307616), phalloidin AF594 (ThermoFisher, catalog no. A12381), and P-CD3ζY142-AF647 (K25-407.69, BD catalog no. 558489) for 40 min at 4°C.

[0186] After washing, cells were suspended in FACS buffer (PBS supplemented with 2% FBS) and acquired using an Amnis ImageStreamX Mark II Imaging Flow Cytometer (Luminex) equipped with five lasers (405, 488, 561, 592, and 640 nm). On average, approximately 55,000 images were collected per sample at 60x magnification on the slow setting. IDEAS software (version 6.2.187.0, EMD Millipore) was used for data analysis and cell labeling.

[0187] To identify immune synapses using IDEAS software, the gating strategy shown in Figure 2 was implemented. Cells were first gated on in-focus live CD3+MHCII+ cells. Within this population, area and aspect ratio features were used to select images showing single CD3+ T cells and single MHCII+ B-LCL cells. Next, CD3 intensity within a self-generated synaptic mask was measured to exclude non-interacting cells. As used herein, the term "mask" refers to the outer silhouette of an overlay of all images (BF and all labeled antibodies) obtained for a cell, doublet, or multiplet, respectively. The synaptic mask was defined as a combination of the morphological CD3 and MHCII masks and an extension of 3. Only synapses showing CD3 signal within the mask were gated. Finally, bright-field (BF) height and area features were used to exclude T+B-LCL cells within one layer and analyze single T-LCL synapses.

[0188] Intracellular staining of cytokines using conventional flow cytometry For intracellular cytokine staining, cells were first treated with GolgiPlug (BD Biosciences, Catalog No. 555029) and GolgiStop (BD Biosciences, Catalog No. 554724) for at least 2–4 h before staining. After incubation, live / dead staining was performed using the fixable viability dye eF780 (eBioscience, Catalog No. 65-0865-14) for 20 min at 4°C. Cells were then fixed and permeabilized using the eBioscience Foxp3 / Transcription Factor Staining Buffer Set (Cat. No. 00-5523-00) as described for the synaptogenesis assay. Intracellular staining was performed in permeabilization buffer containing fluorescently labeled antibodies TNFα-APC (clone MAb11, BD Biosciences, catalog no. 554514), IFN-γ-PE (clone B27, BD Biosciences, catalog no. 554701), and Granzyme B-PE-Cy7 (clone QA16A02, Biolegend, catalog no. 372214) for 30 min at 4°C. Finally, cells were suspended in FACS buffer (PBS supplemented with 2% FBS and 1 mM EDTA) and acquired on a BD Biosciences FACS Celesta.

[0189] Tumor cell lysis assay (in vitro) B cell-depleted PBMCs from healthy donor blood were prepared using standard density gradient isolation followed by B cell depletion with CD20 microbeads (Miltenyi, catalog number 130-091-104). The B cell-depleted PBMCs were then incubated with tumor targets (Z-138 or Nalm-6) at a 5:1 ratio in the presence or absence of CD19-TCB for 24 hours. Tumor cell lysis was calculated based on LDH release (LDH Cytotoxicity Detection Kit, Roche Applied Science) and normalized to spontaneous release (PBMCs + untreated targets = 0% tumor cell lysis) and maximum release (tumor target lysis with Triton X-100 = 100% lysis).

[0190] Quantification of CD19 expression CD19 expression on B-LCL cells was examined using anti-human CD19-AF647 (Biolegend #302220) antibody and the corresponding isotype control muIgG2b (Biolegend #400330) with the Bangs Laboratories Quantum™ Alexa Fluor® 647 MESF Kit (Cat. No. 647) according to the manufacturer's instructions. Quantification of CD19 molecules on tumor target cell lines Nalm-6 and Z-138 was performed with the Dako QiFi Kit (Cat. No. K0078) according to the manufacturer's instructions using anti-human CD19 purified (BD #555410) antibody and the corresponding isotype control muIgG2b (BD #557351).

[0191] 2. Preparation of Imaging Datasets for Analysis A total of 2,899,575 imaging flow cytometry images were recorded. The dataset consisted of nine different donors across four independent experiments. Donor 1 and Donor 2 were used twice. Different conditions were measured, including -SEA (total number of images = 625,001), +SEA (557,781), Ctrl-TCB (330,000), CD19-TCB (324,020), isotype (405,000), and teplizumab (403,375). Images included brightfield (BF), F-actin, MHCII, CD3, P-CD3ζ, and live / dead staining. The live / dead staining was used only to remove dead cells. For each experiment, images were compensated using a compensation matrix derived from a single stained cell. After compensation, raw images (16-bit) and their corresponding segmentation masks for each channel were exported from IDEAS software and saved in HDF5 format. To enable parallelization, each image and its corresponding mask were saved separately.

[0192] Interpretable feature engineering from images To investigate the immunological synapse, we extracted a set of 296 biologically motivated features. These features included morphology, intensity, colocalization, texture, and synapse-related values ​​(see Figures 5, 6, 9, 22, and 32). Morphological features were calculated based on segmentation masks from each channel. Features included "area," "bounding box area," "convex area," "eccentricity," "equivalent diameter," "Euler number," "extent," "maximum ferret diameter," "minimum ferret diameter," "filled area," "major axis length," "minor axis length," "Hu moment," "orientation," "perimeter," "Crofton perimeter," "solidity," and "weighted Hu moment." All morphological features were extracted using the scikit-image library

[41] . For intensity features, cells were first segmented using their corresponding masks. Intensity features included "minimum," "sum," "mean," "standard deviation," "skewness," "kurtosis," "maximum," and "Shannon entropy." We also calculated percentiles of intensity values, including the 10th, 20th, ..., and 90th percentiles. All intensity features were calculated based on NumPy

[42] and SciPy

[43] functionalities. For colocalization features, we implemented the Dice distance and Jaccard distance, and calculated the mask overlap between two channels using the SciPy

[43] library. We also calculated the correlation distance

[43] , Euclidean distance

[43] , Manders overlap coefficient

[44] , intensity correlation quotient

[44] , structural similarity

[41] , and Hausdorff distance

[41] . For texture features, we used gray-level co-occurrence matrix (GLCM) features

[45] , including contrast, dissimilarity, homogeneity, ASM, energy, and correlation. The synapse-related features were defined as "Ch enrichment (mean)" = (Ch intensity at synapse) / m(Ch intensity), "Ch enrichment (total)" = sum(Ch intensity at synapse) / sum(Ch intensity), and "Ch enrichment (max)" = max(Ch intensity at synapse) / mean(Ch intensity)

[35] . Finally, "background average" and "gradient RMS" were performed for image quality control.All these features were implemented using NumPy (version = 1.18.5), Pandas (1.1.5), SciPy (1.8.0), scikit-image (0.19.2), and scikit-learn (1.0.2)

[46] .

[0193] Autoencoder Feature Extraction To take advantage of the large amount of unlabeled data, we implemented and trained a multi-channel autoencoder

[24] . This autoencoder contained a separate encoder for each channel. The encoder was designed to map each channel to a 32-dimensional vector. Concatenation of these vectors resulted in a 5*32-dimensional space. These features were then mapped to a 128-dimensional feature vector. To reconstruct the original image, we implemented a decoder on top of the concatenated vector. We used the L2 norm as the reconstruction loss. The augmentations used to train the autoencoder included random rotation, random scaling, random flipping, and random Gaussian noise.

[0194] Feature Preselection Given the large number of features, a feature preselection pipeline was implemented to select the most relevant features following the work of Haq et al.

[47] (see Figures 12–14). First, the Pearson correlation between features was measured. If at least two features were highly correlated (|corr| > 0.95), only one of them was retained (randomly) and the rest were removed. In the next step, features were ranked using six different methods. These methods included mutual information, linear support vector machines, logistic regression with L1 regularization, logistic regression with L2 regularization, random forests, and XGBoost. The top k features (selected hyperparameters) from each method were selected and their union was used. After this reduction, the Spearman correlation matrix between features was calculated, and spectral clustering was performed on the correlations. Then, m clusters were created and one feature was randomly selected per cluster. A final step was performed to account for multicollinearity between features.

[0195] classification There are three main techniques used to train supervised learning algorithms, feature-based techniques, and deep learning.

[0196] Classical supervised learning models Two different algorithms were used to train the machine learning models. One was a boosting method called XGBoost

[48] that uses an ensemble of trees (n_trees=100) on the data. The second model was logistic regression. The advantage of using these models was that they provided explainability after training.

[0197] Convolutional Neural Networks To train the supervised deep learning models, well-known architectures in the field of computer vision, such as ResNet18, Resnet34, ResNet50, ResNet152, DeseNet121, and DeepFlow, were used [22, 24, 28]. All models were pre-trained on ImageNet. Considering that the models were designed for three-channel input, the first convolutional layer, which has three input channels versus six, was removed. In addition, the classification layer also needed to be adjusted to nine classes. However, the remaining networks retained their predefined ImageNet weights. We used a multiclass cross-entropy loss for training. The learning rate (lr) was set to 0.001, and a decreasing adaptation strategy with a plateau of 10 epochs was used. The augmentations used to train the autoencoder included random rotation, random scaling, random flipping, and random Gaussian noise.

[0198] Classification feature importance Feature preselection filtering was used to reduce the number of features, and then an XGBoost classifier was trained on the annotated data. XGBoost can provide feature importance using the Gini index, but these importances can be biased for various reasons, including correlation between preselected features, the number of features, the preselection process, and outliers. To account for this, we randomly split the training data into five folds (stratified) and trained the XGBoost classifier five times, each time using four of the five folds. This process was repeated 100 times, resulting in 500 different models. Each training run used a random number of preselected features (top-k) between 30 and 200 features. Finally, a set of Gini indices was obtained for all features. Features were ranked using the median Gini index of each feature.

[0199] Classification staining importance To determine which stains contribute most to the prediction, we used recursive channel reduction. In all runs, we kept the BF unstained. We then trained "Interpretable Features + XGBoost" based on the features of the selected channels.

[0200] Class Frequency Analysis For each donor, we first used a trained XGBoost classifier to predict the class of all images, then we filtered out images using this data cleaning protocol. 1. Filter images containing dead cells with "average live / dead intensity" >= "average live / dead intensity (90th percentile)" (using live / dead staining) 2. Filter dead cells (using live / dead staining) for "Average Live / Dead Intensity" > "Average Live / Dead Intensity (90th Percentile)" 3. Filter out-of-focus images with these conditions: "Gradient RMS BF" > "Gradient RMS BF (2nd percentile)" and "Gradient RMS BF" < "Gradient RMS BF (90th percentile)". 4. Using XGBoost predictions, we filter images based on high entropy (entropy > 1.0). Entropy was calculated using the SciPy package. This step is performed to filter out images for which the classifier is most uncertain about its prediction. 5. Filter images predicted as "B-LCL" where "mean MHCII intensity" < "mean MHCII intensity (5th percentile)." This step ensures that images predicted as "B-LCL" contain the minimum MHCII intensity. 6. Filter images predicted as "B-LCL" where "MHCII area" < "MHCII area (10th percentile)." This step ensures that images predicted as "B-LCL" contain cells of appropriate size and reduces artifacts. 7. Filter images predicted as "T cells" with "CD3 mean intensity" < "CD3 mean intensity (1st percentile)". This step ensures that images predicted as "T cells" contain the minimum CD3 intensity. 8. Filter images predicted as "B-LCL and T cells in one layer" when "MHCII area" < "MHCII area (20th percentile)." This step is performed to exclude "B-LCL and T cells in one layer" with small "B-LCL." 9. Image filtering based on isolation forest outlier detection. The main parameters used were n_estimators=100, max_samples='auto', contamination='auto', max_features=20. To reduce the execution time, only the top 30 features were used based on the Gini index from the XGBoost training. 10. Filter the image based on uniform manifold approximation and projection (UMAP).

[0201] First, all images were z-transformed into a 2D dimensional space using UMAP. Features were normalized using the mean and standard deviation of each feature. To reduce the execution time, only the top 30 features based on the Gini index from the XGBoost training were used. Then, the DBSCAN algorithm was run with eps=0.09 and min_samples=5. The resulting clustering was filtered if (number of images in cluster) / (total number of images)<0.0001.

[0202] All these steps were performed based on the scikit-learn implementation. Unless otherwise specified, all parameters were set using the default values ​​in scikit-learn. After data cleaning, the frequency of each class was calculated for each donor and condition using the formula: F_C = (number of images predicted as C) / (total number of images). To address the compositional nature of the data, the fold change in frequency was compared using log_2(F_C_antibody / F_C_control). This transformation has the advantage that the sum of frequencies does not equal a constant value. After calculating the log_2 fold change, the Wilcoxon rank-sum test was used to analyze the effect of antibodies on class frequencies. The Wilcoxon rank-sum test tests whether two samples are likely to come from the same population. To account for multiple testing, the Benjamini-Hochberg correction was used for +SEA / -SEA, CD19-TCB / control TCB, and teplizumab / isotype, respectively. Because the experiments were performed independently, only each of these comparisons was corrected separately.

[0203] Feature Difference Analysis The effects of perturbation by the presence of CD19-TCB and teplizumab, respectively, on signaling synapses were analyzed.

[0204] First, we selected images predicted to be "synapses with signaling." Bright-field (BF) features were removed because BF intensity has no biological meaning. Additionally, BF morphological features were already captured based on the F-actin mask. Therefore, this information was redundant. Without being bound by this theory, we hypothesize that this feature reduction was also necessary because it reduces the number of tests and increases the likelihood of finding meaningful p-values ​​after correcting for multiple tests.

[0205] This procedure yielded 210 features for comparison of SEA and TCB based on F-actin, MHCII, CD3, and P-CD3 zeta.

[0206] In the case of teplizumab, the number of features was further reduced. Without being bound by this theory, this reduction was necessary due to the use of CD4 in recording the teplizumab image instead of CD3, which is used for CD19TCB. Therefore, a meaningful comparison between teplizumab and its control based on CD3 was not feasible. Therefore, 132 features extracted from F-actin, MHCII, and P-CD3 zeta were analyzed.

[0207] After feature selection, features were compared using the Mann-Whitney U test for each condition and its control.

[0208] To understand the direction of change, the difference in the median values ​​of the characteristics for each condition and its control was used. To account for multiple testing, the Benjamini-Hochberg procedure with α = 0.05 was used. Because the conditions were independent, the p-values ​​for each condition and its control were corrected separately.

[0209] Granzyme B prediction and feature ranking To predict granzyme B, we used images predicted as "synapses without signaling" and "synapses with signaling" for each condition. While not being bound by this theory, it is assumed that synapses lead to cytokine production. Given that thousands of images were available for each donor and condition, we used an aggregation pipeline to create feature vectors corresponding to each donor and condition. To reduce the number of features, we only used consistent feature changes in CD19-TCB (Figure 3). For each donor and condition, we aggregated features using the 5th, 50th, and 95th percentiles to obtain the extreme values ​​and averages of all features.

[0210] After deriving the aggregate features, we trained a linear regression model with LassoLars using leave-one-donor cross-validation. The most important features were based on coefficient magnitude.

[0211] Visualization For plots and images, Python's matplotlib (version = 3.3.2) and seaborn (0.11.2) were used.

[0212] References 1. Huppa, J. B., and Davis, M. M., Nat. Rev. Immunol. 3, 973 - 983 (2003). 2. Dustin, M. L., Cancer Immunol. Res. 2, 1023 - 1033 (2014). 3. Kearney, C. J., et al., Crit. Rev. Immunol. 35, 325 - 347 (2015). 4. Kallikourdis, M., et al., Front. Immunol. 6, 433 (2015). 5. Laura, R. D., et al., Jci Insight 3, e120757 (2018). 6. Schubert, D. A., et al., J. Exp. Med. 209, 335 - 352 (2012). 7. Calvez, R., et al., Haematologica 96, 1415 - 1423 (2011). 8. Alan, G. R., et al., J. Clin. Invest. 118, 2427 - 2437 (2008). 9. Francesca, F. and Baldari, C. T., Pharmacol. Res. 134, 118 - 133 (2018). 10. Tai, Y., et al., Front. Pharmacol. 9, 642 (2018). 11. Scott, M., S.-T., et al., Ann. N.Y. Acad. Sci. 1183, 123 - 148 (2010). 12. Wetzel, S. A., et al., J. Immunol. 169, 6092 - 6101 (2002). 13. Cremasco, F. et al., PLoS One 16, e0241091 (2021). 14.Amita, J. et al., J. Clin. Pharmacol. 46, 10-20 (2006). 15.Lee,JYet al.,Nat.Commun.7,13354(2016). 16.Dickopf,S.,et al.,Comput.Struct.Biotechnol.J.18,1221-1227(2020). 17.Cartwright,ANR,et al.,Nat.Commun.5,5479(2014). 18.Vyver,AJVD,et al.,Mol.Cancer.Ther.20,357-366(2021). 19.Wabnitz,GH,et al.,J.Immunol.Meth.423,29-39(2015). 20.Wabnitz,GH,et al.,Eur.J.Immunol.41,3157-3169(2011). 21.Ahmed,F.,et al.,J.Immunol.Meth.347,79-86(2009). 22.Lippeveld,M.,et al.,Cytom.Part A 97,308-319(2020). 23. Eulenberg, P., et al., Nat. Commun. 8, 463 (2017). 24.Kranich,J.,et al.,J.Extracell.Vesicles 9,1792683(2020). 25.Chlis,N.-K.,et al.,Nucleic Acids Res.48,gkaa926-(2020). 26.Doan,M.,et al.,Nat.Protoc.16,3572-3595(2021). 27.Hennig,H.,et al.,Methods,San Diego Calif 112,201-210(2017). 28.Eulenberg,P.,et al.,Biorxiv 081364(2017)doi:10.1101 / 081364. 29.Leclercq,G.,et al.,J.Immunother.Cancer 10,e003766(2022). 30.Xu,D.,et al.,Cell Immunol.200,16-26(2000). 31.Herold,K.C.,et al.,J.Clin.Invest.111,409-418(2003). 32.Cremasco,F.,et al.,Plos One 16,e0241091(2021). 33.Bacac,M.,et al.,Clin.Cancer Res.22,3286-3297(2016). 35.Wabnitz,G.,et al.,J.Vis.Exp.(2019)doi:10.3791 / 55345. 36.Carpenter,A.E.,et al.,Genome Biol.7,R100-R100(2006). 37.German,Y.,et al.,Cell Reports 36,109318(2021). 38.Perakis,A.,et al.,Arxiv(2021)doi:10.1007 / 978-3-030-87589-3_58. 39.Naghizadeh,A.,et al.,Plos Comput.Biol.18,e1009883(2022). 40.Xiong,W.,et al.,Mol.Ther.26,963-975(2018). 41.Walt,S.van der,et al.,Peerj 2,e453(2014). 42.Harris,C.R.,et al.,Nature 585,357-362(2020). 43.Virtanen,P.,et al.,Nat.Methods 17,261-272(2020). 44.Adler,J.and Parmryd,I.,Cytom.Part A 77A,733-742(2010). 45.Haralick,R.M.,et al.,Ieee Transactions Syst.Man.Cybern.SMC-3,610-621(1973). 46.Buitinck,L.,et al.Arxiv(2013). 47.Haq,A.U.,et al.,Ieee Access 7,151482-151492(2019). 48.Krishnapuram,B.,et al.,Proc.22nd Acm.Sigkdd Int.Conf.Knowl.Discov.Data Min.785-794(2016)doi:10.1145 / 2939672.2939785.

Claims

1. 1. A method for sorting cells in a cell mixture, the mixture comprising T cells and activated B cells or antigen presenting cells, the method comprising the steps of: a) applying at least labeled antibodies that bind to F-actin, MHCII, and CD3 to the cell mixture to obtain a labeled cell mixture, wherein the antibodies are each labeled with a dye, and the dyes have different emission wavelengths; b) acquiring at least one image of the cell mixture; c) separating the cells in the cell mixture; i) the cells are single cells and F-actin positive; - MHCII positive and CD3 negative, or -MHCII negative and CD3 positive If so, a step of classifying isolated cells, ii) classifying the cells into cell doublets or multiplets if they are aggregates of two or more cells that are F-actin positive, MHCII positive and CD3 positive; A method comprising:

2. step a) applying at least labeled antibodies that bind to F-actin, MHCII, CD3, and P-CD3 zeta to the cell mixture, wherein the antibodies are each labeled with a dye, and the dyes have different emission wavelengths; Step c) comprises: i) sorting the cells into single B cells or antigen-presenting cells if they are F-actin positive, MHCII positive, CD3 negative and P-CD3 zeta negative; ii) if the cells are F-actin positive, MHCII negative, CD3 positive and P-CD3 zeta negative, classifying them into single T cells without signal transduction; iii) if the cells are F-actin positive, MHCII negative, CD3 positive and P-CD3 zeta positive, classifying them into single T cells with signal transduction; iv) if the doublet is F-actin positive, MHCII positive, CD3 positive and P-CD3 zeta negative, classifying it as a doublet that forms a synapse between a B cell or an antigen-presenting cell and a T cell without signal transduction; v) When the doublet or multiplet is F-actin positive, MHCII positive, CD3 positive, and P-CD3 zeta positive, classifying it into a doublet or multiplet that forms a synapse involving signal transduction between one or two B cells or antigen-presenting cells and one T cell. That is, The method of claim 1.

3. The method according to any one of claims 1 to 2, wherein step b) is a step of acquiring an image of the cell mixture using an imaging flow cytometer.

4. The method according to any one of claims 1 to 3, wherein the acquired images are images showing a single cell or an isolated doublet or multiplet, respectively.

5. The method of any one of claims 1 to 4, wherein the mixture comprises B cells or antigen-presenting cells and T cells in a cell ratio of about 4:

3.

6. The method of any one of claims 1 to 5, wherein the T cells are CD4-positive memory T cells.

7. The method of any one of claims 1 to 6, wherein after said mixing, said mixture of cells is centrifuged.

8. 1. The use of F-actin, MHCII, and CD3 to classify cells in a mixture comprising T cells and B cells or activated B cells or antigen-presenting cells, comprising: The classification is the cell is a single cell, F-actin positive, - MHCII positive and CD3 negative, or -MHCII negative and CD3 positive If the cell is an isolated cell, Is it something like that? or, The classification is If the cells are aggregates of two or three cells, are F-actin positive, MHCII positive and CD3 positive, the cells are cell doublets or multiplets. That is, use.

Images