Immune function assessing of TIM3/CD28‐modified CD19 CAR‐T cells and general CD19 CAR‐T cells through a high‐throughput single‐cell microarray platform

Chimeric antigen receptor (CAR) T cells are widely used to treat hematological tumors due to their powerful ability to target and kill cancer cells, of which accurate function evaluation at the single‐cell level is crucial to ensuring the efficacy of diagnosis and treatment. Currently, a universal platform to evaluate the efficacy of immune single cells rapidly, efficiently, and systematically is urgently needed. Here, the cytotoxicity, proliferative potential, and persistence of TIM3/CD28‐modified CD19 CAR‐T cells are evaluated in comparison with ordinary CD19 CAR‐T cells through high‐performance and throughput graphene oxide quantum dot (GOQD)‐based single‐cell microfluidic chips. Overall secretory factor expression, immune‐therapy effect of different effector‐target ratios, spatial immune‐therapy effects, and subgroup type profiling are demonstrated to explicit the immunotherapy effect of TIM3/CD28‐modified CD19 CAR‐T cells. TIM3/CD28‐modified CD19 CAR‐T cells show stronger anti‐tumor ability and maintain excellent immunotherapy effects even at low effector‐target ratios and remote distances. TIM3/CD28 also strengthens the local targeting ability of TIM3/CD28‐modified CD19 CAR‐T. Importantly, TIM3/CD28‐modified CD19 CAR‐T exhibits more distinct Th1/Th2 long‐term persistent and potent killer subgroups, which is very helpful for personalized therapy. Overall, this study provides a valuable approach that can be widely implemented to analyze current CAR‐T combinations and evaluate the function of innovative CAR treatments in the future.


| INTRODUCTION
The success of CAR-T cell therapy in advanced malignant lymphoid tumors has revolutionized traditional cancer treatment, offering a potential cure for relapsed and refractory patients, 1,2 of which accurate functional evaluations at the single-cell level can effectively guarantee the effect of diagnosis and treatment.CD19 CAR-T cells have been extensively studied and reported, [2][3][4][5] but some patients may fail to enter remission or experience disease relapse after receiving CD19 CAR-T cell therapy, with the potential for relapse and common extraneous tumortargeting side effects. 3,4This is caused by the low persistence of CD19 CAR-T cells, 6 drug resistance of lymphoid cancer cells, 2,7 cytokine release syndrome (CRS) and neurotoxicity generated during the treatment process. 8,9With the introduction of costimulatory signals, some studies have shown that CARs containing the 4-1BB costimulatory domain are more durable than CARs containing the CD28 costimulatory domain, 4,10 while CD28 stimulates more effectors than the 4-1BB costimulatory domain. 11,124][15] However, when developing combinatorial multi-antigen targeting strategies, it is critical to ensure that an effective response is generated to each antigen, and targeting one antigen preferentially will not eliminate the problem caused by antigen-negative recurrence. 16Furthermore, combining CAR-T cell therapy with immune checkpoint inhibitors provides a synergistic approach to optimize the antitumor efficiency and durability of CAR-T cell therapy, such as costimulatory signaling combined with PD-1, 17 to improve the effectiveness of CAR-T cell therapy.TIM3 inhibitory ligands are ubiquitously expressed on the cell membrane surface of lymphoma cells, [18][19][20] and CAR-T cells with costimulatory signaling combined with TIM3 effectively reduce the possibility of lymphatic tumor immune escape and increase the effectiveness of CAR-T cell therapy.To verify the feasibility of these strategies, improve the efficacy of CAR-T therapy, and provide personalized solutions, it is extremely necessary to evaluate the function of CAR-T cells and their side effects; however due to the high response of CAR-T cell therapy in leukemia patients and the small number of nonresponders, it is difficult to systematically evaluate and establish parameters associated with various responses. 21AR-T cells have been evaluated at the genetic level in many studies. 22,23However, compared to gene sequencing, secreted proteins are more direct, faster, and more reliable due to the uncertainty in the transcription and translation process.Many studies have evaluated the effector levels in culture supernatants, 18 but there is a wide range of cell heterogeneity, and useful single-cell information is lost. 24n addition, due to the time needed for in vitro expansion of CAR-T therapy and because many patients do not survive until reinfusion, 3,4 it is necessary to shorten the manufacturing time to increase the likelihood that patients can receive treatment.Studies have shown that the dose of CAR-T cells necessary for effective treatment is very low, but this is not fixed. 25For patients at different stages and situations, it is necessary to evaluate the appropriate dose of CAR-T cells to maximize the efficacy of the treatment.Although the expansion and activation of cells after reinfusion is essential, 3,26 compared with a large number of target cells at the beginning of reinfusion, there are very few CAR-T cells 4 ; these cells inevitably produce the following different spatial modes in vivo: remote-distance non-contact mode and local-distance contact mode.The resulting differences in secretion, which we call spatial effects, have not been effectively assessed in vivo or in vitro.For different individuals, it is also necessary to evaluate CAR-T cells, define T-cell subgroups based on multiple cytokine expression profiles, and determine the ideal properties of CAR-T cells for different types of patients. 27ere, TIM3/CD28-modified CD19 CAR-T cells were designed that express TIM3/CD28 reversal receptors based on CD19 CAR-T cells and can convert the inhibitory effect of TIM3 inhibitory ligands on the lymphoma surface into the stimulatory effect of CD28 on the CAR-T cells. 18Compared with CD19 CAR-T cells, TIM3/CD28modified CD19 CAR-T cells exhibit stronger immune function due to the reversal of TIM3/CD28 inhibition-trans-stimulation.In order to evaluate its specific ability, we propose an accurate functional evaluation technology for CAR-T cells, and the overall secretory factor expression, immune-therapy effect of different effector-target ratios, and spatial immune-therapy effects of TIM3/ CD28-modified CD19 CAR-T and CD19 CAR-T targeting and killing lymphoma cells were evaluated at the singlecell level based on 15-plex secreted factors.For the first time, up to 15 secretory differences in activated CAR-T cells under different contact modes (spatial effects) were efficiently assessed in vitro.The results showed that, compared with CD19 CAR-T cells, TIM3/CD28-modified CD19 CAR-T cells retained the characteristics of Th1and Th2-like responses and increased endurance, proliferation and killing capabilities.Interestingly, TIM3/ CD28-modified CD19 CAR-T can be effective at lower concentrations and more remote distances.In the clinic with TIM3/CD28-modified CD19 CAR-T cell therapy, patients may not need CAR-T cell concentrations that are too high, saving time and money for the entire procedure and reducing the possibility of adverse reactions.Then, the subgroups of TIM3/CD28-modified CD19 CAR-T and CD19 CAR-T with different therapeutic effects were further studied, which can distinguish the Th1 immune response subgroup, Th2 immune response subgroup, durable killing subgroup, short killing subgroup, and super secretory subgroup; thus, an alternative treatment option is provided for the treatment of different categories of patients.Most importantly, this proposed strategy can be extended to the functional assessment of multiple Chimeric antigen receptor (CAR) therapies, providing a promising approach for future research on novel CARs.

| High-throughput multi- performance systematic evaluation platform for CD19 CAR-T and TIM3/CD28modified CD19 CAR-T immune single cells
To systematically evaluate the function of CAR-T cells at the single-cell level, CAR-T cell secretion expression profiles with different effector-target ratios and spatial effects must be analyzed.Then, further subgroup analysis must be performed for different types of patients to determine the ideal properties of CAR-T cells.To address this challenge, our proposed platform uses a single-cell microarray GOQD-based chip and a multi-factor detection GOQD-based chip (Figure S1).The platform accurately evaluates the multiple immune functions of CAR-T cells from the perspectives of overall secretory factor expression, effector-target ratios, spatial effects, and subgroups to provide a personalized treatment plan.High-throughput microchamber arrays of GOQDs bioaffinity can isolate external influences and ensure the activity of cells. 28The single-cell microarray chip is physically combined with the multi-factor detection chip, and each cell microchamber covers the entire barcode array, so secreted immune factors can be specifically captured by the antibodies on the barcode (Figure S2).PLL glass substrates are commonly used to lay antibody barcodes, and achieving excellent detection linearity and long-term stability is challenging. 29The GOQD glass substrate demonstrated excellent uniformity, long-term stability, and low background, and it was first applied to the evaluation of CAR-T single cells here.The immobilized antibody capacity and detection background of the GOQD substrate were compared with those of the PLL substrate in Figure S3.
Before single-cell chips are combined with antibody barcode chips, live CAR-T cells are mixed with target cells in equal proportions.Then, a hybrid single-cell printing method is used to generate a high-throughput single-cell mixed array with different proportions of effector cell and target cell microchambers based on a Poisson distribution (Figure 1A).After the two chips are combined and closed, cells can move freely in a single row in their respective closed microchambers, in which combinations of cells with different effector-target ratios and spatial effects naturally occur.The multi-indexed barcode arrays used for CAR-T assessment (Figure 1B) were divided into six groups of categories, assessing key immune functions (e.g., cytotoxic, 30 effector, 31 regulatory, 32,33 chemokine, 34 stimulatory, 35 and inflammatory 36,37 ) of CAR-T cells, which were microprinted with the help of an accompanying secreted factor detection chip.For the cytotoxic category, which mainly plays a role in destroying tumor cells, GZMB and Perforin were the prime effectors.The effector category primarily acts as the driver and effector of the Th1 effect, including IFNγ, TNF-α and IL-2.The regulatory category mainly regulates the Th2 effect and dampens the anti-tumor immune response, among which IL-4 is the driver and effector of the Th2 effect.The chemotactic category mainly regulates the migration and localization of immune cells, such as MCP-1 and MIP-1α.For the stimulating category, which represents the degree by which immune cells are stimulated, too high stimulation may indirectly lead to the CRS response in the body.The inflammatory category mainly plays a role in triggering inflammation and autoimmunity.
Here, the function of the newly designed TIM3/CD28modified CD19 CAR-T cells was mainly evaluated in comparison with that of the general CD19 CAR-T cells.TIM3/CD28-modified CD19 CAR-T cells are a kind of CAR-T cell that express TIM3/CD28 reversal receptors based on CD19 CAR-T cells, which can convert the inhibition of TIM3 into the stimulation of CD28 receptors.Their CAR chain structure is shown in Figure S4.To accurately evaluate their functions, quantitative formulas (Table S1) were first used to preliminarily analyze the six key immune functions of the two CAR-T cells in response to the same target cells (Figure 1C).For cytokine secretion detection, the GOQD substrate detection method used for CAR-T evaluation in this paper was compared with the traditional ELISA method in Figure S5.Compared with traditional ELISA and Flow cytometry methods, the method used in this paper has higher detection throughput because the multiplexed channels overcome spectral overlap, and the detection sensitivity is further improved due to the high antibody loading density and low detection background of GOQDs confirmed in Figure S3.Based on the difference in the number ratio of CAR-T cells and target cells in different patients, the effective function of CAR-T cells with different effectortarget ratios in the microenvironment was further evaluated (Figure 1D).In addition, based on the complex distance relationship between effector and target cells in vivo, the function of CAR-T cells at different distances in the local and remote spaces was evaluated in-depth (Figure 1E).More importantly, the immune signatures of unique potential subgroups were evaluated by screening and analyzing CAR-T cell expression profiles in multiple conditions (Figure 1F).Among them, many spatial and subgroup characteristics that cannot be detected by traditional gene, cell culture supernatant and serum analysis methods were obtained 23,38,39 ; the method is fast, comprehensive, and precise, demonstrating irreplaceable capabilities and providing a guarantee for personalized treatment.CAR-T cells in vivo treatment of mouse tumor experiment has been done in the previous work. 18This work evaluates the secretion microenvironment of CAR-T cells and target cells in vitro through a single-cell platform.

| Overall secretory factor expression evaluation of CD19 CAR-T and TIM3/ CD28-modified CD19 CAR-T single cells
To evaluate the overall effective function of the two types of CAR-T cells, TIM3/CD28-modified CD19 CAR-T and CD19 CAR-T were mixed with Namalwa target cells in equal proportions for printing, and secretion data of the single CAR-T cell and single target cell microchamber of the whole piece were extracted.The secretion factor detection map of TIM3/CD28-modified CD19 CAR-T cells targeting Namalwa tumor cells is shown in Figure 2A, the corresponding cell map is shown in Figure 2B, and the large-scale microchamber chip image and scanned fluorescence image are shown in Figure S6.The secretion factor detection map of CD19 CAR-T targeting Namalwa tumor cells is shown in Figure S7.In addition, TIM3/CD28-modified CD19 CAR-T cells targeting Raji tumor cells were used to study the effective function of TIM3/CD28-modified CD19 CAR-T against other target cells; the secretion profile is shown in Figure S8, and their overall immune function evaluation is shown in Figure S9.Unengineered T cells were used as blank controls, the secretion profile is shown in Figure S10, and their overall effective function evaluation is shown in Figure S11.The 6 key immune functions (cytotoxic, effector, regulatory, chemokine, stimulatory, inflammatory) were quantitatively analyzed by quantitative formulas (Table S1), which were fitted using the corresponding fluorescence values of the 15 recombinant factors with a 10-fold concentration gradient.Compared with CD19 CAR-T cells, TIM3/CD28 switch receptormodified TIM3/CD28-modified CD19 CAR-T cells secreted more Granzyme B (GZMB), IFN-γ, TNF-α, IL-2, IL-4, GM-CSF, IL-8, etc. (Figure 2C).At the same time, to study the immune response of TIM3/CD28-modified CD19 CAR-T to different target cells, the expression levels of TIM3/CD28-modified CD19 CAR-T targeting Namalwa and Raji tumor cells were also compared (Figure S12).It can be seen that the expression levels of the cytotoxic, effector, regulatory and inflammatory factors are basically the same.Compared with targeting Namalwa cells, the expression of chemokines that target Raji cells was enhanced, and the expression of stimulatory factors was weakened.Among these factors, GZMB, as a highly effective cytotoxic factor, reflects the stronger direct anti-tumor ability of TIM3/CD28-modified CD19 CAR-T cells.TNF-α has a slow-release killing effect, 40,41 and IL-2 is an important proliferation factor of CAR-T cells. 31,42TIM3/CD28-modified CD19 CAR-T cells expressed higher levels of TNF-α and IL-2 than CD19 CAR-T cells by 3-4-fold (Figure 2D), which was divided into four regions.The upper right corner shows that TNF-α þ IL-2 þ cells accounted for 85.4%, reflecting the more potent proliferation and slow-release killing of TIM3/CD28-modified CD19 CAR-T cells.
TIM3/CD28-modified CD19 CAR-T cells were very active in the Th1-like immune process, producing more Th1 effectors including IFN-γ, TNF-α and IL-2. 43,44As a negative regulator of anti-tumor immunity, TIM3 is overexpressed in activated CD4 Th1-type CAR-T cells, 45 and TIM3/CD28 receptors may activate these cells to express more Th1 effectors.Studies have shown that Th1/ Th2-type function is necessary to achieve long-term remission in patients. 22Therefore, we believe that the secretion level of the Th1 effector combination may represent the long-term persistence of CAR-T cells.TIM3/CD28-modified CD19 CAR-T cells were also found to be very active in the Th2-like immune process, producing more Th2 drivers and effector IL-4; as a result, their persistence was further enhanced.As a positive signal, CD28 is involved in chemokine production, 46,47 and compared with CD19 CAR-T, high expression of the chemokine MIP-1α in CART-28 was found.Notably, when TIM3/CD28-modified CD19 CAR-T is activated, it increases the expression level of will highly express corresponding dangerous cytokines, such as GM-CSF and IL8.Although IL-6 and MCP-1 stimulating factors were not detected in vitro, some adverse effects may occur, such as CRS and neurotoxicity, in vivo.Although the overall expression of other factors was not high, such as IL-10, MCP-1, IL-6, IL-17, and IL-1β, the heatmaps (Figure 2E,G) indicate that TIM3/CD28-modified CD19 CAR-T cells and CD19 CAR-T cells exhibit highly expressed subgroups.After the cells were mixed for heatmap analysis, it can be clearly seen that they are basically divided into the following types: multi-secretion TIM3/CD28-modified CD19 CAR-T cells and lowsecretion CD19 CAR-T cells (Figure S13).Similarly, by mixing the TIM3/CD28-modified CD19 CAR-T targeting Namalwa and Raji data for heatmap analysis, it was also  clearly divided into two categories due to some key secreted factors, such as MCP-1 (Figure S14).These factors have complex correlations (Figure 2F,H).For TIM3/ CD28-modified CD19 CAR-T cells, IFN-γ, TNF-α, IL-2, IL-4 and GM-CSF are highly correlated, especially Th1 effectors, which have an excellent correlation.For CD19 CAR-T cells, the correlation between perforin, IL-2, IL-4, and IL-10 is relatively high, and a certain synergy may occur between them.
The reason behind this complex correlation should be accurately evaluated from other perspectives.Overall, compared with CD19 CAR-T cells, TIM3/CD28 causes TIM3/CD28-modified CD19 CAR-T cells to secrete more GZMB, IFN-γ, TNF-α, IL-2 and IL-4, etc.; as a result, the cells exhibit higher cytotoxicity, proliferative ability, and longer persistence of antigen stimulation, leading to stronger anti-tumor capacity.However, the same TIM3/ CD28-modified CD19 CAR-T cells also secrete more dangerous stimulating factors such as GM-CSF and IL-8 than CD19 CAR-T cells, which may lead to some adverse effects, such as CRS and neurotoxicity, in vivo.

| Immune-therapy effect evaluation of different effector-target ratios for CD19 CAR-T and TIM3/CD28-modified CD19 CAR-T single cells
Evaluating CAR-T cells under different effector-target ratios is very important because different stages of proliferation and apoptosis occur in the process of immunotherapy, 48,49 resulting in different effector-target ratio microenvironments in vivo.The appropriate reinfusion dose in vitro must be accurately evaluated before reinfusion to ensure the therapeutic efficacy of different CAR-T cells, shorten the manufacturing time of CAR-T cells and provide personalized treatment.Cell microchambers with different effector-target ratios were further screened to accurately evaluate their functions (Figure 3A).The hybrid single-cell printing results were optimized by using a 1:1 ratio of mixed red and green stained cells (Figure S15, enabling thousands of 1:1 effector-target ratio single-cell printing, and the number of microchambers with different effector-target ratios is counted in Table S2.The number of multi-effector cells (>2:1) and multi-target cells (>2:1) in the microchambers of the two CAR-T cell groups showed different ratios after they were mixed and isolated in equal proportions (Figure 3B).The proportion of multi-effector cell microchambers (high concentration of CAR-T cells) in CD19 CAR-T decreased by 4% compared with TIM3/CD28modified CD19 CAR-T, and the proportion of multitarget cell microcavities (low concentration of CAR-T cells) in CD19 CAR-T increased by 5% compared with TIM3/CD28-modified CD19 CAR-T.The different tumor killing abilities of premixing the two types of CAR-T cells with target cells during activation caused this ratio difference.TIM3/CD28-modified CD19 CAR-T exhibits a stronger tumor killing ability, so the multi-effector cell ratio increases, while the multi-target cell ratio decreases.The change is small and occurred because TIM3/CD28modified CD19 CAR-T was premixed for 4 h before single cells were printed onto chips.The number of each effector-target ratio microchamber is shown in Table S2.Subsequently, the concentration of GZMB produced within 16 h was used to compare the cytotoxicity of TIM3/CD28-modified CD19 CAR-T and CD19 CAR-T targeting Namalwa at different effector-target ratios in Figure S16, and the results showed that TIM3/CD28modified CD19 CAR-T cells exhibited higher cytotoxicity than that of CD19 CAR-T cells and could be more effective at killing tumors; these results correspond well to the cytotoxicity verified using the LDH method with different effector-target ratios in cell populations in previous work. 18o further analyze the reasons for this result, several key cytokines (GZMB, IFN-γ, TNF-α, IL-2, IL-4, and GM-CSF) with different effector-target ratios of the two cells were quantitatively analyzed (Figure 3C), and other cytokines are shown in Figure S17.When there were only target cells (E:T = 0:1), almost no effective factors were detected.When the target cells were single cells and the number of CAR-T cells increased gradually from 1 to 5, the secretion of immune factors also increased gradually.For TIM3/CD28-modified CD19 CAR-T cells, in addition to the same proportion of GZMB increase and number increase, the proportion of IFN-γ, TNF-α, IL-2, IL-4, and GM-CSF increase (10-100 times) was much greater than the proportion of number increase (5 times).For CD19 CAR-T cells, the performance was not satisfactory, and the ratio by which the above factors increased was less than 10-fold.Some factors, such as INF-γ and IL-2, remained flat and lower than the trend observed for the number increase ratio, and even when the target cell ratio was 1:1, the concentration of secreted factors was 1-2-fold less than that of TIM3/CD28-modified CD19 CAR-T cells.This shows that, compared with CD19 CAR-T cells, TIM3/ CD28-modified CD19 CAR-T cells transformed with TIM3/CD28 exhibited better proliferation and killing ability at the same concentration (1:1) and high concentration (>1:1).In addition, when only effector cells were present (T:E = 0:1), a small number of effective factors were detected, which was due to premix activation before cell isolation.CD19 CAR-T shows the weakening trend that may occur when the effector cells are single cells and the number of target cells gradually increases from 1 to 5.

WANG ET AL.
However, it is gratifying that, except for IL-2, the secreted factors of TIM3/CD28-modified CD19 CAR-T are relatively stable.The other five factors showed exponential growth, even surpassing that of the multi-effector cell microchamber secreted factor of CD19 CAR-T cells.Thus, even at low concentrations (<1:1) in vivo, TIM3/CD28modified CD19 CAR-T cells can maintain good proliferation and killing ability, which can ensure therapeutic efficacy at very low in vitro amplification concentrations.TIM3/CD28-modified CD19 CAR-T cells can generate a better therapeutic effect than that of CD19 CAR-T cells at a very low concentration and exhibit a stable proliferation ability with increasing TIM3/CD28-modified CD19 CAR-T cell number.After exceeding the target cell, TIM3/CD28modified CD19 CAR-T can further achieve an exponential increase in therapeutic efficacy.However, to improve the therapeutic efficacy, CD19 CAR-T needs to be expanded to an effector-target ratio of 1:1 or even higher, and the therapeutic efficacy is less than the ratio of quantitative increase.In conclusion, TIM3/CD28-modified CD19 CAR-T cells engineered by TIM3/CD28 receptors exhibit excellent cytotoxicity at low and high concentrations, enabling them to maintain good immune function at all stages from initial infusion to the end of treatment, which increases the durability of immunotherapy.

| Evaluation on the spatial immune- therapy effect caused by CD19 CAR-T and TIM3/CD28-modified CD19 CAR-T single cells
Functional evaluation of CAR-T cells under the influence of spatial effects is also very important because CARs on CAR-T cells need to contact and bind with ligands on target cells to stimulate immune efficacy.However, the microenvironment in vivo is complex and changeable, and the number of CAR-T cells initially infused is limited.This is because low concentrations often result in remote spatial effects in vivo, which have not been effectively evaluated.Even though single TIM3/CD28modified CD19 CAR-T cells can exhibit excellent therapeutic efficacy at low concentrations, the impact of spatial effects on CAR-T cells in vivo needs to be considered.Since effector cells and target cells are smaller than the microchamber cross-section, they can move freely in the closed microchamber.At this location, the relative spatial position of the effector-target cells was recorded after single-cell printing was performed and the solution was left for 1 h.Through real-time photography of effector cells and target cells in the microchamber, the state in which the effector cell receptors and the target cell ligands are in contact is called "local", and the state in which they are separated is called "remote".Local contact and remote non-contact, two spatial forms of cell microchambers with an effector-target ratio of 1:1, occur and are screened and evaluated (Figure 4A).Through statistical analysis of nearly 1000 microchambers, it was found that the two spatial states of the two CAR-T cells had different proportions.TIM3/CD28-modified CD19 CAR-T cells had 18% more local contacts than CD19 CAR-T cells, and TIM3/CD28-modified CD19 CAR-T cells tended to be more targeted after being transformed by TIM3/CD28 (Figure 4B).The secreted factor radar profiles of the two spatial effects of the two CAR-T cells were quantitatively evaluated (Figure 4C,E), and the detailed changing magnification is shown in Table S3.
The two CAR-T cell secretion profiles were compared, revealing that TIM3/CD28-modified CD19 CAR-T shows an overall trend of expanding outward after local contact.Most factors, such as the killer factor GZMB, Th1 effectors IFN-γ, TNF-α and IL-2, the Th2 regulator IL-4, the chemokine MIP-1α, and the stimulators GM-CSF and IL-8, tended to expand by nearly or more than an order of magnitude.However, in the local contact of CD19 CAR-T, except for a small increase in effector TNF-α, chemokine MIP-1α and stimulatory factor GM-CSF, the overall trend remained unchanged, and some factors, such as GZMB, showed a trend of retraction.This demonstrates that the TIM3/CD28 receptor converts inhibitory signals into positive stimulating signals and causes TIM3/CD28modified CD19 CAR-T cells to secrete more effective factors after local contact, which has stronger immune efficacy.
Interestingly, TIM3/CD28-modified CD19 CAR-T cells can ensure that secreted factors are strongly secreted without contact, which may indicate that TIM3/CD28modified CD19 CAR-T cells can enhance secretion in other ways when target cells are not in contact with the surrounding microenvironment.Compared with CD19 CAR-T cells, TIM3/CD28-modified CD19 CAR-T cells secrete more GZMB, which is responsible for direct cell killing and the Th1 effectors IFN-γ, TNF-α and IL-2, and secrete less stimulatory factor GM-CSF when activated and noncontacting target cells at a remote distance.As a result, their killing ability, proliferation ability and persistence are increased and the irritation that may indirectly lead to cytotoxicity at the remote non-contact spatial distance that occurs at low concentrations is reduced.In the case of remote non-contact spatial distance, although the secretion of GM-CSF and IL8 is reduced, the activated TIM3/ CD28-modified CD19 CAR-T and CD19 CAR-T cells secrete certain cytokines that may cause CRS and neurotoxicity.Notably, another negative side of TIM3/CD28modified CD19 CAR-T has a good treatment effect in the future clinical transformation.
To illustrate this difference more clearly, we further statistically analyzed 8 key factors with obvious differences, namely, GZMB, IFN-γ, TNF-α, IL-2, IL-4, MIP-1α, GM-CSF, and IL-8 (Figure 4D,F).Other cytokines are shown in Figure S18.TIM3/CD28-modified CD19 CAR-T cells showed better therapeutic efficacy after local contact, whereas CART-19 cells showed an inhibition result in which secretion was unchanged or even decreased.In addition, although TIM3/CD28-modified CD19 CAR-T non-contact behavior similarly reduced some therapeutic efficacy, we found that it surpassed the therapeutic efficacy of CD19 CAR-T when in contact.This provides a new perspective and direction for evaluating the immune function of CAR-T cells engineered with other novel receptors based on microfluidic chips from space effects.

| Exploration of Th1/Th2 long-term persistent and potent killer CAR-T subgroups combining effector-target ratios and spatial effects
CD19 CAR-T cell therapy is highly effective in pediatric and adult patients with B-ALL.1][52] In addition, the expression ratio of CD4 and CD8 in the CAR-T-cell group is different, 53,54 so CAR-T cells have complex subgroups, which increases the difficulty of evaluation, and patientspecific CAR-T therapy is time-consuming and expensive.Therefore, a secreted factor subgroup evaluation standard is urgently needed to better evaluate the durability and efficacy of treatment.Although flow cytometry can quickly analyze the cell phenotype but it is only an endpoint cell attribute, 55 cannot represent the real process of CAR-T cells after contacting target cells, and the spectral overlap limits its throughput to detect the multifunctionality of CAR-T cells. 56Some studies have shown that Th1 function can improve cell persistence, Th2 function is very important in maintaining long-term remission, Th1 and Th2 functions coexist in patients with long-term remission, and there is a general and robust association between Th2 function deficiency and CD19positive relapse. 22,43,57o further explore unique subgroups suitable for personalized therapy, the complex and variable situations were considered in vivo, and subgroup systems that occur at different effector-target ratios and spatial effects were evaluated.Here, the K-means t-SNE strategy 58 was used to classify the subgroups of CAR-T cells, and the K value was determined from three times the number of highly expressed factors, which decreased until no identical subgroup was observed. 28Two CAR-T cells, TIM3/CD28modified CD19 CAR-T and CD19 CAR-T, were divided into a total of 26 subgroups according to the 15-plex secreted factors they expressed (Figure 5A).The proportion of each subgroup is shown in Figure 5B.The expression levels of 6 types of secreted factors in each subgroup are shown in Figure 5C through the shape of the violin diagram, and the correlations between the various subgroups are shown in the left branch of the hierarchical clustering tree.All CD19 CAR-T cells are classified into the weak immune function branch marked in green, and the low expression subgroups 1, 5 and 6 that occur in the TIM3/CD28-modified CD19 CAR-T cells are also classified into this branch, but no subgroup of CD19 CAR-T is classified into the strong immune function branch marked in blue.The secretory RadViz spectra of the two CAR-T cells are shown in Figure S19 and S20, in which all subpopulations are distinguished by different colors, and each subpopulation is basically radial and has its own unique secretion characteristics.In addition, subgroup assessments of CART-28-targeted Raji cells and unengineered T cells are shown in Figure S21 and S22, respectively.The Raji-targeted TIM3/CD28-modified CD19 CAR-T cells were divided into 13 subgroups, which also exhibited strong immune function.Weaker subgroups were also observed, such as subgroup-7.The secretory RadViz spectra of Raji-targeted TIM3/CD28modified CD19 CAR-T cells are shown in Figure S23.Each subgroup also has unique secretory signatures.The unengineered T cells were divided into 5 subgroups, and subgroup-2 accounted for 89.6% of the total, which basically did not secrete factors.
Furthermore, the results obtained for each subgroup of two CAR-T cells that target the same target, Namalwa, CD19 CAR-T and TIM3/CD28-modified CD19 CAR-T, were deeply analyzed and emphasized, and the minimum expression mean threshold for each factor was set to 10 pg/mL.Over the threshold was positive expression, such as IL-2 þ , and below the threshold was negative expression, such as IL-2 -.The threshold was applied to each subgroup of the following types of immune functions: cytotoxic, effector, regulatory, chemokine, stimulatory, and inflammatory, and TIM3/CD28-modified CD19 CAR-T and CD19 CAR-T cell subgroups were sorted from left to right according to the number of positively expressed immune factors (Figure 5D).Interestingly, TIM3/CD28-modified CD19 CAR-T subgroups 1, 4, and 5, classified as having weak immune function, were ranked last among the TIM3/CD28-modified CD19 CAR-T cell subgroups.Among the weak immune function CD19 CAR-T subgroups, only 2 subgroups (subgroups 4 and 5) expressed more than 8 immune factors compared with 9 subgroups of TIM3/CD28-modified CD19 CAR-T cells.As mentioned earlier, the Th1 immune response can greatly improve the viability of CAR-T cells and then improve persistence, which is accompanied by the secretion of a large number of effectors, such as IFN-γ, TNF-α, and IL-2. 43,44All subgroups of TIM3/CD28-modified CD19 CAR-T cells produced positive expression of more than two Th1 effectors, while only subgroup 11 (4.3%) of CD19 CAR-T cells showed a strong Th1 immune response.In addition, even though TIM3/CD28-modified CD19 CAR-T cells have many weakly secreted subgroup-1, -5 and -6 cells, they express key immune factors, such as the toxicity factor GZMB and the effector factors TNF-α and IL-2.These factors exhibit good direct killing and long-lasting killing effects and a proliferative capacity, and the absence of stimulatory and inflammatory factors reduces the risk of developing CRS.The Th2 immune response can further improve the persistence of CAR-T cells, of which a representative factor is IL-4. 59Compared with TIM3/ CD28-modified CD19 CAR-T, which all express IL-4 factors except for weak immune subgroups-1, -5, and -6, we found that CD19 CAR-T only has subgroup-4 and -5 express IL-4 factor.In addition, many immune-deficient subgroups were not found in TIM3/CD28-modified CD19 CAR-T cells, while the main proportion subgroups of CD19 CAR-T cells showed immune function deficiency, which showed obvious inhibition results.At the same time, perforin-deficient cells were found in these two CAR-T cells, and the stimulatory proinflammatory factor IL-8 was overexpressed in perforin-deficient TIM3/ CD28-modified CD19 CAR-T subgroup-2 and CD19 CAR-T subgroup-10 cells.
The number of various immune secretory types and proportions of the effector-target ratio and spatial effects in all subgroups were further analyzed statistically, and the results were consistent with those described above.Blue-green bubbles represent two kinds of CAR-T cells, and their size represents the size of the subgroup.Yellowred bubbles represent the proportion of single and 2-5 multiple CAR-T cells for a single target cell, and purplepink bubbles represent the proportion of remote and local spatial forms in the subgroup (Figure 5D).The detailed percentage values are shown in Table S4.It can be clearly seen that most of the multi-positive immune subgroups have a large proportion of multi-cells and local contact spatial form, showing obvious cell cooperation effects.In addition, it was found that the Th1 immune subgroup exhibits an obvious feature as the local contact spatial form occurs in a high proportion.Even for CD19 CAR-T, a small number of subgroups 11, 8, and 10 have this feature; their Th1 immunity is enhanced, and TIM3/CD28 will counteract the contact inhibitory effect, promote the increase in the proportion of TIM3/CD28-modified CD19  S2.Data visualization in (A, C, D) is carried out using Python.immune efficacy.
Through subgroup analysis, it was found that CART-28 cells have a unique Th2 subgroup in addition to more Th1 subgroups and long-term persistent subgroups coexisting with Th1 and Th2 functions, while CD19 CAR-T cells generally have many Th2 function-deficient subgroups, and Th1 subgroups do not exhibit high expression of the three effectors simultaneously.In addition, the killing ability of the CD19 CAR-T and TIM3/CD28modified CD19 CAR-T subgroups was estimated by the expression of the cytotoxic factors GZMB and Perforin.Further analysis of the effect caused by the effector-target ratio and spatial effect revealed many distinct subgroups, such as Th1 immune response subpopulations that would overcome the inhibitory effect to some extent and tend to be closer to target cells.These subgroup results, which were accurately evaluated, reveal numerous potential CAR-T immune subtypes, providing evidence and multiple options for the ideal properties of personalized CAR-T therapy in the future.However, due to the high cost of CAR-T products and the need for private customization, this evaluation method has not yet been clinically transformed.

| CONCLUSION
In this work, we propose a GOQD-based highthroughput single-cell chip-based method for evaluating the immunotherapy efficacy of CAR-T cells.From the perspective of single cells, the overall secretory factor expression, immune-therapy effect of different effectortarget ratios, spatial immune-therapy effects, and subgroup type of CD19 CAR-T and TIM3/CD28-modified CD19 CAR-T were systematically studied.TIM3/CD28 strengthened the long-term targeting anti-tumor ability of TIM3/CD28-modified CD19 CAR-T cells, and we further demonstrated that in the remote non-contact spatial form, single TIM3/CD28-modified CD19 CAR-T cells can maintain excellent immunotherapy effects; TIM3/CD28-modified CD19 CAR-T cells show more distinct Th1/Th2 long-term persistence and potent killer subgroups than that of CD19 CAR-T cells.Notably, this platform can not only evaluate these CAR-T combinations but also provide a reliable platform for the functional evaluation of novel CAR therapies in the future. 4| EXPERIMENTAL SECTION 4.1 | Microfluidic chip design and fabrication Two types of microfluidic chips were designed to print cells and antibody barcodes.Before the chip was fabricated, the designed pattern needed to be transformed into a mask and transferred to a silicon wafer by photolithography.Using the same photolithography and etching process parameters for the silicon templates of two chips, a positive photoresist AZ6130 was spin-coated on a clean silicon wafer to form a photoresist thickness of 2 μm.After prebaking, the samples were covered with the corresponding chrome mask and exposed to UV light.After development and hardening, the patterned silicon wafer was loaded into a dry etcher to etch the microchambers and microchannels to a thickness of 24 μm.Finally, the photoresist on the surface of the patterned silicon stencil was removed for long-term preparation of PDMS chips.

| Cell culture
Raji and Namalwa cells were cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum.Unengineered T, CD19 CAR-T and TIM3/CD28-modified CD19 CAR-T cells were cultured in TexMACS GMP medium supplemented with 10% fetal bovine serum and 50 IU/mL IL-2.These cells were cultured in a cell incubator at 37°C with 5% CO 2 .All CAR-T cells utilized in this study originated from the same T cell source.

| Construction of the chain
CD19 CARs specifically consisted of a single-chain antibody targeting CD19, IgG4 Fc spacer, CD8 transmembrane domain, 41BB endodomain and CD3ζ-chain of the T cell receptor complex, which were cloned and WANG ET AL.
of 17 inserted into pCDH-CMV-MCS-EF1-CopGFP (RRID: Addgene_99730) to generate the 19BBZ lentiviral vector.The cells with this vector were named CD19 CAR-T.We generated a second lentiviral vector named the TIM3/ CD28 fusion protein, which encodes the same 19BBZ CAR and binds to the chimeric switch receptor.The T2A sequence was the ribosome skipping sequence inserted between the CAR and the TIM3/CD28 fusion protein, and the cells with this vector were named TIM3/CD28modified CD19 CAR-T, as shown in Figure S4.

| GOQD self-assembly modification
The microchamber PDMS chip and antibody glass chip were treated with oxygen plasma for 5 min, immersed in 3-aminopropyl trimethoxysilane (APTES, Sigma-Aldrich) for 45 min, immersed in GOQD solution for 45 min after sonication, and ultrasonically cleaned again.

| Antibody barcode microprinting
The PDMS microfluidic chip designed for printing antibody barcodes with parallel microchannels was physically bonded with GOQDs functionalized glass substrates.Capture antibodies (2 μL each) were loaded into the inlet of microchannel, and the antibody was sucked out from the outlet of the other end by a vacuum pump to achieve microprinting.Peel off the PDMS chip from the glass slide to form a factor detection chip.Before use, the chip was treated with 3% BSA to avoid non-specific adsorption of factors.All antibodies used in this study are anti-human antibodies.

| Hybrid single-cell printing
First, CAR-T effector cells were resuspended at 1000 rpm for 4 min and mixed with target cells in a ratio of 1:1, incubated for 4 h, and suspended in 50 μL of medium at a concentration of 10 6 cells per mL.The cell solution was then printed in an S-shaped motion trajectory, and passed over each column of the GQOD-modified microchamber array chip in turn.After printing, the factor detection chip was immediately covered on the microchamber array chip; the covering process avoids the generation of air bubbles.After being clamped with a clamp, it was left to stand for 1 h in a cell culture incubator, allowing CAR-T cells move naturally and target cancer cells in each microchamber, and then scanned under a fluorescence microscope to record the location and number of CAR-T (stained by carboxyfluorescein succinimidyl ester dye) and target (stained by deep red cell proliferation dye) cells in each microchamber.Subsequently, it was moved to a cell incubator for incubation.

| Secreted immune factor detection
After 16 h of incubation in the incubator, the chip was removed, and the factor detection chip was peeled off in 3% BSA and rinsed with 1% BSA.A mixture of biotinylated detection antibodies at a concentration of 0.25 μg/mL (each antibody) was loaded onto the factor detection chip and incubated for 45 min.Then, the chip was rinsed with 1% BSA solution.Next, 300 μL of 1:100 APC dye-labeled streptavidin was added and incubated for 30 min.Finally, the chip was washed thoroughly with PBS and water.

F I G U R E 1
Schematic diagram for a multifunctional accurate evaluation of CAR-T cells by high-throughput single-cell chips.(A) The factors secreted by CAR-T cells mixed with target cells in the cell array chip and detected by the antibody barcode chip.(B) The barcoded 15-fold antibody panel includes the following groups of cytokines: cytotoxic, effector, regulatory, chemokine, stimulatory, and inflammatory cytokines.(C) The overall evaluation of TIM3/CD28-modified CD19 CAR-T and CD19 CAR-T effective function, in which the six areas divided by the red dotted lines correspond to the evaluation results of the 6 main immune functions respectively.(D) Microchamber screening for different effector-target ratios and target-effector ratios.(E) Assessment of the spatial effects of TIM3/ CD28-modified CD19 CAR-T and CD19 CAR-T cells, whose unique receptors are plotted.(F) Subgroup analysis of CAR-T cells.
WANG ET AL.
The underlying reason is that TIM3/CD28 receptors convert inhibitory signals into positive stimulatory signals.CD19 CAR-T is affected by TIM3 ligands and produces inhibitory signals, and TIM3/CD28-modified CD19 CAR-T is stimulated by TIM3 ligands to produce positive signals.This dual difference results in an increase in the proportion of multi-effector cell microchambers and a decrease in the proportion of multi-target cell microchambers for TIM3/ CD28-modified CD19 CAR-T cells.

F
I G U R E 3 TIM3/CD28 receptors enable TIM3/CD28-modified CD19 CAR-T cells to maintain high therapeutic efficacy at low and high effector-target ratios.(A) Images of cells with different effector-target ratios; red indicates target cells, and green indicates effector cells.(B) Proportion of microchambers of two types of CAR-T cells with different effector-target ratios (the value at the center of the circle is the statistical proportion that the majority of cells are more than two times larger than the minority of cells).(C) Statistical analysis of the immune factor secretion trends of GZMB, IFN-γ, TNF-α, IL-2, IL-4, and GM-CSF in the different effector-target ratio microchambers of TIM3/CD28-modified CD19 CAR-T cells and CD19 CAR-T cells.Data are presented by value (scatter), median (line), quartiles (box), and mean (red dot), which are connected by the red dotted line.The number of each effector-target ratio microchamber is shown in TableS2.All the statistics in this figure were performed for each effector-target ratio condition compared to that with only target or effector single cells (0:1).*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; two-tailed Student's t test.

F I G U R E 4
TIM3/CD28 receptors enhance the local targeting ability and remote teletherapy efficacy of TIM3/CD28-modified CD19 CAR-T cells.(A) Images of local contact and remote non-contact spatial forms, with target cells in red and effector cells in green.(B) Proportion of 1:1 effector-target ratio microchambers of two CAR-T cells with different spatial effects.(C) 15-plex secreted factor radar profiles of TIM3/CD28-modified CD19 CAR-T cells with different spatial effects.(D) Statistical analysis of 8 significantly different immune factors with different spatial effects in TIM3/CD28-modified CD19 CAR-T cells.Data are presented by value (scatter), median (line), quartiles (box), and mean (red dot), which are connected by the red dotted line.N is calculated as in (B).All the statistics in this figure are performed for each spatial effect condition compared to the remote spatial effect condition.*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; two-tailed Student's t test.(E) 15-plex secreted factor radar profiles of CD19 CAR-T cells with different spatial effects.(F) Statistical analysis of 8 significantly different immune factors with different spatial effects in CD19 CAR-T cells.N is calculated as in (B).All the statistics in this figure are calculated in the same way as in (D).

F I G U R E 5
TIM3/CD28-modified CD19 CAR-T displayed more Th1 and Th2 coexisting long-term persistent and potent killer subgroups than CD19 CAR-T.(A) Subgroups of TIM3/CD28-modified CD19 CAR-T and CD19 CAR-T cells.Cluster size N is indicated as in (D).(B) Subgroup proportions of TIM3/CD28-modified CD19 CAR-T and CD19 CAR-T cells.(C) Violin plot of the subgroup immune function profile and hierarchical cluster of subgroup classification for TIM3/CD28-modified CD19 CAR-T (marked in blue) and CD19 CAR-T (marked in green).(D) Analysis of TIM3/CD28-modified CD19 CAR-T-and CD19 CAR-T-positive secretory species, the number of various immune secretory types, and the proportion of effector-target ratio and spatial effects in all subgroups.Relevant specific proportions are shown in Table