Cell counting

Abstract

The invention relates to a device for contactlessly counting the concentration of particles in suspension in a fluid comprising a container associated with a stirring means for achieving a uniform suspension of the particles, which comprises: a means for acquiring images, which means has a field configured to form an image containing a plurality of particles after sedimentation, the field being focused on the inner surface of the bottom of the container, and a light source for illuminating the inner surface of the bottom of the container, and a computer that acquires a time-stamped sequence of at least two images (Ii, ti) in the focal plane of sedimentation of the particles in suspension, counts, in each of the images (Ii, ti), the number of particles Pi,<sb / >, determines a set of values (Pi, ti), and determines the concentration CP of particles, at the time t of the last acquired image (Ii, ti), by applying a function F to the set of values (Pi, ti), the function F being dependent on a hydrodynamic drag.

Description

Cell counting Field of invention

[0001] The present invention relates to a method for online and in situ counting of cells in a biological culture medium, in particular in the field of personalized cellular immunotherapy and more particularly the collection and automated processing of cell subsets, in particular CART-T (Chimeric Antigenic Receptor - T) cells in peripheral blood with the aim of locally reinjecting these cells to repair tissues. The invention relates more particularly to automatic cell counting methods for monitoring the proliferation of T lymphocytes in culture, in a closed system and in real time. The objective is to be able to reach a target concentration of T lymphocytes (approximately 1 billion cells in 600 ml, i.e. a concentration of 1.66 M / ml) at which the following steps of the innovative therapy based on Car-t cells can be initiated.

[0002] Cell counting or counting corresponds to the determination of the number of cells contained in a volume is done in the state of the art on a sub-sample or sub-part of the culture and very largely involves at least one sampling step, in particular for the following reasons: in order to facilitate counting and save time: it seems indeed very tedious and time-consuming to count the millions / billions of cells in culture, counting is often intrusive for the cells and the idea is therefore to be able to apply a treatment that is possibly deleterious or modifies cellular homeostasis for a sub-sample of the culture without impacting the rest of the cells in culture.

[0003] For biologists, cell counting is a daily manipulation carried out in the context of cell cultures. Its accuracy is essential for the well-being of the cells in culture but also for any experiment involving cells in order to be able to adapt the concentrations of reagents or to be able to obtain images that can be analyzed in imaging, for example.

[0004] Several automated counters with imagers that contain a micro-camera and image processing software have been developed in recent years. These counters nevertheless involve a step of taking a part of the sample and the use of markers aimed at improving the contrast between living cells and noise to facilitate their automatic detection. Examples include Trypan blue, which, when used in excess, stains the membranes of living cells, improving the contrast (and completely stains dead cells blue), or the use of fluorescent protein and membrane probes capable of binding to living eukaryotic cells.

[0005] Regarding cell imaging, there are two main imaging technologies: the classic microscope coupled with a camera or holographic reconstruction which consists of the projection of the diffraction of cells onto a lensless CMOS sensor by a point light source. Holographic reconstruction has the advantage of imaging a large surface but the light source must be as direct as possible which is difficult to achieve for cultures (very high height of the medium, roughness of the walls of the pocket, high cell concentration, presence of air bubbles and aggregates).

[0006] Regarding image analysis to automatically count cells, there are two main methods: The image analysis technique using filters, thresholding and segmentation of varying complexity. There is no generic algorithm; each application requires a suitable algorithm. This method is not very effective for the classification of poorly differentiated elements, i.e. the identification of cells of the same size but with slightly different morphology. The supervised learning method based on thousands of manually labeled cell images. Learning requires a very high computing power (96GB GPU) but is only performed once. A hyperparameter model is then created to analyze images quickly and classify the detected elements. The computing power (CPU or GPU) is then lower and depending on the number of classes to be identified, the image can be processed live. State of the art

[0007] Known in the prior art is FR3034103 describing a device for measuring the concentration of microorganisms in a liquid sample, comprising a light source, a sample holder, an image sensor, an optical device located between said sample holder and said image sensor, a microprocessor and a counting program. The optical device consists of a single transparent lens. This document also describes a method for monitoring the development of microorganisms in a liquid, characterized in that it comprises the following steps: taking a sample of said liquid, introducing said sample into a device mentioned above, detecting an image of said sample by said image sensor, determining the concentration of microorganisms contained in said sample by counting microorganisms on said image, by means of said program.

[0008] Patent EP2025744 describes a method for counting cells online and in situ in a biological culture medium, comprising the following steps: a plurality of measurements of the capacitance of said medium or a plurality of measurements of the conductance of said medium, at distinct frequencies varying within a predetermined measurement frequency range, an extraction of information on variation in permittivity due to β-dispersion in said medium, from said capacitance measurements, and a processing of said information on variation in permittivity to deliver information on counting cells in said medium.

[0009] Patent application WO2018220198 describes a device and a method for detecting and imaging biocontaminant elements, comprising a matrix of detector pixels having an active face on which light is incident, a light-transparent support which covers the active face and lighting means which produce the incident light on the active face, display means. The invention is characterized in that the incident light is the light emitted by the biocontaminant elements and in that the device further comprises at least one filter, a polarizer and an analyzer, said polarizer and said analyzer being positioned in a crossed configuration to overcome laser granularities in the images. Disadvantages of the prior art

[0010] The prior art solutions require sampling the solution. The step of sampling and pipetting the cell culture allows in particular to minimize a technical obstacle well described in the literature which corresponds to the presence of counting artifacts (air bubbles, etc.) likely to deteriorate the precision of the counting method by pipetting steps. In our case, cell culture and cell counting in a bag and without sampling required direct agitation of the bag, which on the contrary facilitates the appearance of air bubbles or the appearance of other artifacts such as scratches on the bag which can drastically alter the precision of the counting method.

[0011] Similarly, the lack of labeling made counting and image analysis more complex (no contrast markers possible). An additional problem came from the nature of the cells (T lymphocytes) and the cell culture volumes sought for the purpose of automatic design of innovative drugs. Indeed, these cells are non-adherent (in suspension) and since the concentration and culture volume are high, the lack of dilution further complicated the cell counting method at these high concentrations.

[0012] In addition, T lymphocytes can become activated: they tend to form large cell clusters.

[0013] In addition, T lymphocytes have variable morphologies depending on the individual, the culture medium (5.8 to 8.14 µm) but also depending on the stage of the cell cycle and the culture time, which can drastically alter the accuracy of the automatic counting method. Indeed, the diameter of T lymphocytes doubles before division.

[0014] Finally, state-of-the-art solutions are not compatible with counting directly in a culture bag, which imposes various optical constraints. Solution provided by the invention

[0015] In order to overcome these drawbacks, the present invention relates to a device for contactless counting of the concentration of particles suspended in a fluid comprising a container associated with a stirring means for homogeneously suspending the particles, characterized in that it comprises: an image acquisition means having a field adapted to form an image comprising a plurality of particles after sedimentation focused on the inner surface of the bottom of said container, a light source for illuminating the inner surface of the bottom of the container and a computer. This computer executes a computer program to control: the acquisition of a time-stamped sequence of at least two images (I i , t i ) in the focal plane of sedimentation of said suspended particles, the counting, in each of said images (I i , t i ), of the number of particles P i the determination of a set of values ​​(Pi , t i ) the determination of the concentration CP of particles, at time t of the last image (I i , t i ) acquired by applying a function F to said set of values ​​(P i , t i ), the function F being a function of the hydrodynamic drag force. Advantageously, said light source is arranged on the side opposite the camera with respect to said bottom, said light source comprising a condenser focused on the surface of said bottom.

[0016] Advantageously, said light source comprises an LED source positioned at the bottom of a tube having anti-reflection grooves.

[0017] Preferably, the axis of said tube is horizontal, and a mirror returns the light beam at 90°, perpendicular to the plane of a transparent sample support surface.

[0018] According to an alternative embodiment, said image acquisition means comprises a microscope optic with an input lens and an image sensor. The assembly is supported by a motorized stage.

[0019] Detailed description of a non-limiting example of embodiment

[0020] The present invention will be better understood on reading the following description, concerning a non-limiting example of embodiment illustrated by the appended drawings where:

[0021] represents a schematic view of a device according to the invention

[0022] represents a schematic view of the image processing according to the invention device according to the invention

[0023] represents the curve of the variance of the Laplace transform as a function of the position (5M / ml)

[0024] represents the cell size distribution curve over 50 images.

[0025] represents the steps in the process of measuring cell concentration

[0026] represents the accumulation curve of the number of cells as a function of time. General principle of the invention

[0027] The invention relates to a method for counting cells on a bag for monitoring the proliferation of T lymphocytes throughout the culture (between 0.01 and 10 million cells per milliliter), which is implemented automatically and in a closed environment, without taking a sub-sample of the cell culture and without applying dilutions and / or cell labeling. The method is based on determining the sedimentation kinetics of the cells contained in the culture medium, and deducing therefrom the cell concentration – and therefore the number of cells in a reference volume – as a function of the dynamic viscosity of the culture medium and the average diameter of the cells.

[0028] the real-time cell culture counting method according to the invention: be fully automated, be carried out entirely in a closed system directly in a culture bag (without sampling, dilution, cell labeling), alter cellular homeostasis as little as possible, ensure a counting accuracy of 90 to 95% throughout the culture (for a wide range of T lymphocyte concentrations between 0.01 and 10 million cells per milliliter).

[0029] To this end, the invention is based on the measurement of cell proliferation by analysis of sedimentation during time intervals from images acquired using a microscope in a reference plane, generally the bottom of the bag containing the culture medium and the cells to be counted, and implements: A culture bench containing the culture medium and the cells to be counted, comprising a controllable means for resuspending the cells, for example a flexible bag, with at least one constriction to generate turbulence. an image acquisition device with a lighting source (100) and a microscope (200) having a motorized adjustable focus and sufficient magnification for viewing the cells and an optical condenser adapted to the culture bench, associated with an image sensor (210). The assembly has a thickness of less than 30 mm.and a computer running a program controlling the acquisition of images and their processing, and in particular: the control of the focusing of the automatic microscope objective the control of the lighting the detection of bubbles the calculation of concentration.

[0030] The principle of dynamic cell counting proposed by the invention is based on the sedimentation rate and Stockes' law. The method implemented consists of agitating the culture medium in a bag in order to resuspend the cells so as to have a homogeneous distribution of the cells in suspension.

[0031] A microscope focused on the bottom of the bag allows visualization of only the cells that settle there by sedimentation. Immediately after resuspension by stopping the agitation, the kinetics of cell sedimentation are measured to estimate the cell concentration as a function of the number of cells at the bottom of the bag over time. Optical device

[0032] The optical device consists of a low-height miniaturized microscope, less than 30 millimeters, whose architecture is illustrated. The device consists of a light source (100) and an image sensor (200).

[0033] The light source (100) comprises an LED source (110) positioned at the bottom of a tube (120) having anti-reflection corrugations (130). The axis of the tube (120) is horizontal, and a mirror (140) returns the light beam at 90°, perpendicular to the plane of a transparent sample support surface (150). The output of the illumination source comprises an anti-bubble dome (160) and creates a hollow on the top of the pocket to facilitate the removal of bubbles from the camera field.

[0034] The image acquisition block comprises a microscope optic (220) with an input lens (230), and an image sensor (210). The assembly is supported by a motorized stage (240).

[0035] The transparent sample support surface (150) constitutes an intermediate optical surface, located between the light source (and / or the imaging system) and the interior surface of the bottom of the container serving as a window or optical barrier to allow the passage of light while isolating the contents of the container or optimizing the imaging conditions. This transparent surface reduces optical artifacts (for example, stray reflection or scattering of light) and protects the optical system from the fluid medium.

[0036] The mirror (140) returns the light beam at 90°, the transparent sample support surface (150) is located just before or on the inner surface of the bottom of the container, acting as a final crossing point for the light. The light would pass through this surface before illuminating the deposited particles.

[0037] The transparent sample support surface (150) is functionally linked to the inner surface of the container bottom but they are two distinct elements. The first plays an optical or mechanical role to facilitate image acquisition or illumination, while the second is the actual surface where the particles sediment and are observed.

[0038] Program controlling the operation of the device

[0039] A computer controls the execution of the control programs for the optical block for automatic detection of sedimented cells as well as the resuspension of the cells in the culture medium.

[0040] The optical block control includes:

[0041] • a self-illumination function

[0042] • an auto-focus function.

[0043] The auto-illumination function controls the automatic adjustment of the LED power (110) in order to systematically have the same average gray level for each count. Indeed, the higher the concentration, the stronger the absorbance of the solution. The function consists of sweeping the LED power from 5 to 100% and then applying a polynomial regression for an estimation of the LED power at a set gray level. Visually, an average gray level of 130 (out of 256) allows good contrast, a parameter that has been tested with a concentration of 0.1 to 10M / ml of T lymphocytes,

[0044] The auto-focus function (10) allows the automatic adjustment of the focus of the objective (230) of the microscope (220). A first step controls the initialization of the position of the camera (210) by retracting to the maximum and then goes back to a defined value. The camera (210) is moved to the area to be scanned. Then for each step, an image is taken and the variance of the Laplace transform is calculated. The position, for which the variance is maximum is then the position of best focus. The area to be scanned must exclude the outer wall of the container (bag) which risks being detected as the optimal focal point if few cells are present at the bottom of the bag. L represents the curve of the variance of the Laplace transform as a function of the position (for a concentration of 5M / ml).Curve (1) represents the variation of the Laplace transform after sedimentation, as a function of the stage setting, in number of microsteps (inner surface of the cell-coated bag). Curve (2) represents the variation of the Laplace transform without sedimentation, as a function of the stage setting, in number of microsteps (outer surface of the bag).

[0045] As shown in, at high concentration, the maximum of the Laplacian allows to find the best focus automatically. Optionally, a Gaussian filtering step (kernel 7,7) level 2, allows to reduce the noise and facilitates the measurement of the best focus for low cell concentrations

[0046] The dynamic counting and image processing algorithm is repeated several times, 6 times for example, with different kinetics, to smooth the result and retain the value corresponding to the median of the successive dynamic counts.

[0047] The kinematics are determined by the time and speed of agitation for cell resuspension. With relatively high agitation, the disaggregation of cell clumps is promoted.

[0048] The time between stopping stirring and initiating kinetics measurement is typically between 10 seconds and 30 seconds.

[0049] Image acquisition to measure sedimentation kinetics is done every 2 seconds for 100 seconds (i.e. 50 images per kinetic), this acquisition time can be reduced for high concentrations in order to avoid having too many sedimented cells and making image analysis (cell counting) difficult, or conversely can be extended in the case of low concentrations to increase the number of sedimenting cells to make counting more reliable.

[0050] The automatic image analysis and cell counting processing algorithm is based on image processing by binarization (conversion to gray level, Gaussian filter (Kernel (7,7), level 5), extraction of edges by Scharr Filter (gradient by first derivative of the two axes x and y then fusion), gray level thresholding (>100 non-critical value), dilation (Kernel (7,7), level 1), closing of contours by elliptic morphological transformation of kernel (15,15) then recording of the image).

[0051] The performance of cell detection by shape detection algorithm is superior to that by circle detection.

[0052] The calculator then determines the measurement of the average cell size of the 50 images by kinetics then median of the 6 averages obtained for the 6 repeated experiments (detection of cell contours by vertical / horizontal / diagonal compression, sorting of contours by extraction of the surface and calculation of the diameter and selection of unit cell diameters and sorting by form factor (cf.). The curve of the statistical distribution in cell size on a series of images makes it possible to determine the median diameter.

[0053] Relationship between sedimentation kinematics, median cell diameter, culture medium viscosity, and cell number

[0054] The above-mentioned treatments make it possible to determine two parameters: The evolution over time of the number of cells N c (t) deposited on the bottom of the surface S of the pocket, allowing the sedimentation speed v to be determined sedThe median diameter d c of cellsCell concentration C in the solution in M ​​per mL

[0055] Image processing measures the median diameter of the cells and then counts the number of cells. The diameter allows us to trace the sedimentation rate which, coupled with the number of cells, allows us to trace the concentration

[0056] The number N c of cells deposited on the bottom of the pocket is a function of time t, according to a formula:

[0057] N c =tCS v sed + N c initial

[0058] Nc initial number of cells at t=0 S imaged surface

[0059] Furthermore, the sedimentation rate is a function of the difference between the density cells and density of the culture medium (in kg:m 3 ), of the diameter d c (in m) of cells and viscosity µ l dynamics of the environment, (in Pascal per second).

[0060] This relationship is determined, for example, by Stokes' law

[0061] The usual culture medium for Car-T culture is, for example, RPMI + 8% HSA. Proliferation of T lymphocytes in culture is achieved by CDC3 / CD28 activation in the presence of interleukin 2 (IL-2), or other serum-free culture media better suited to the production of Car-T. Functioning

[0062] The measurement cycle includes a series of processes, some of which are iterative. Camera connection Step (10) of positioning the stage (230) for focusing the optical block Start of a dynamic counting cycle (20) Step (21) Agitation to resuspend and break up clusters LED power adjustment Recursive bubble detection: LED power threshold Bubble removal if necessary (pallet rise / fall) Delta threshold (max-min gray level) and gray mean threshold Bubble removal if necessary (pallet rise / fall) Step (22) Taking images, for example every 10 seconds Step (23) determination of the median diameter of the cells and calculation of the parameter V sedStep (24) Determination of the number of cellsStep (25) Calculation of the cell concentrationCamera disconnectionStep (26) Focus adjustmentReturn for a new dynamic counting cycle (20), six iterations for exampleStep (30) Calculation of a statistical value of the number of cells

Claims

Device for non-contact counting of the concentration of particles suspended in a fluid comprising a container associated with a stirring means for homogeneously suspending the particles, characterized in that it comprises: an image acquisition means (210) having a field adapted to form an image comprising a plurality of particles after sedimentation focused on the inner surface of the bottom of said container, and a light source (100) for illuminating the inner surface of the bottom of the container a computer controlling the acquisition of a time-stamped sequence of at least two images (I i , t i ) in the focal plane of sedimentation of said suspended particles, the counting, in each of said images (I i , t i ), of the number of particles P i the determination of the median diameter d c cells of particles from said images, the determination of a set of values (P i , ti ) the determination of the concentration CP of particles, at time t of the last image (I i , t i ) acquired by applying a function F to said set of values (P i , t i ), the function F being a function of the hydrodynamic drag force. Contactless device for counting the concentration of particles suspended in a fluid according to claim 1, characterized in that said light source (100) is arranged on the side opposite said image acquisition means (210) relative to said bottom, said light source (100) comprising a condenser focused on the surface of said bottom. Contactless device for counting the concentration of particles suspended in a fluid according to claim 1, characterized in that said light source (100) comprises an LED source (110) positioned at the bottom of a tube (120) having anti-reflection grooves (130). Contactless device for counting the concentration of particles suspended in a fluid according to claim 3, characterized in that the axis of said tube (120) is horizontal, and a mirror (140) reflects the light beam at 90°, perpendicular to the plane of a transparent sample support surface (150). Contactless device for counting the concentration of particles suspended in a fluid according to claim 1, characterized in that said image acquisition means (210) comprises a microscope optic (220) with an input lens (230), and an image sensor, this assembly being supported by a motorized plate (240).

Images