A gene combination and evaluation system and kit for CAR-T cell function identification and efficacy monitoring

Abstract

The application discloses a gene combination and an evaluation system and kit for CAR-T cell function identification and curative effect monitoring. The gene combination comprises endogenous function genes BATF, PRDM1, SLC2A1, SOCS3, MCL1, BCL2L1, and CAR specific sequences, GAPDH internal reference genes and ACTB internal reference genes. The application first clearly proposes the CAR-T cell endogenous function gene combination composed of BATF, PRDM1, SLC2A1, SOCS3, MCL1 and BCL2L1, comprehensively evaluates the intrinsic function activity of CAR-T cells from three core dimensions of activation, proliferation and survival, breaks through the limitation of traditional detection which only focuses on exogenous CAR genes or cell surface markers, and fills the industry blank of CAR-T cell functional quality control.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary fields of cell therapy, molecular diagnostics and nucleic acid detection, and specifically relates to a gene combination and evaluation system and kit for CAR-T cell function identification and efficacy monitoring. Background Technology

[0002] Chimeric antigen receptor T-cell (CAR-T cell) immunotherapy has become an important treatment for relapsed and refractory hematological malignancies, especially lymphoma. Its clinical efficacy and long-term survival prognosis depend primarily on the intrinsic functional activity of CAR-T cells, rather than solely on the integration of exogenous CAR genes.

[0003] Currently, there are still many technical shortcomings in the detection and quality control technology of CAR-T cells that need to be addressed, as follows:

[0004] 1. Limited detection dimensions, failing to reflect intrinsic cellular function: Current conventional detection methods mainly focus on quantifying the copy number of exogenous CAR genes or flow cytometry detection of cell surface markers. These methods can only determine the presence and quantity of CAR-T cells, but cannot assess the cells' endogenous activation status, proliferative potential, and anti-apoptotic survival ability. This often leads to phenomena such as "CAR gene positive but cells have no anti-tumor activity, and cell number meets the target but clinical efficacy is poor," resulting in a disconnect between quality control results and actual clinical efficacy.

[0005] 2. Limitations of detection methods: Flow cytometry can only detect cell surface protein expression and cannot penetrate the cell level to assess the activation level of intracellular signaling pathways, metabolic activity and long-term survival potential; Conventional quantitative PCR (qPCR) has problems such as limited multiple detection channels, weak resistance to sample matrix interference and poor accuracy in quantifying low-abundance targets, making it difficult to achieve simultaneous detection and accurate quantification of multiple genes, and thus failing to meet the clinical and industrial needs of CAR-T cell functional assessment.

[0006] 3. Lack of standardized functional evaluation system: At present, the industry has not yet established a unified and highly specific combination of endogenous functional genes for CAR-T cells, making it impossible to scientifically grade the functional activity of CAR-T cells, and it is also difficult to predict in advance the in vivo expansion capacity, anti-tumor effect and recurrence risk after cell infusion, which is not conducive to the standardized production and clinical application of CAR-T cell products.

[0007] Therefore, there is an urgent need in this field for a gene combination and application scheme that is highly specific, has clear functional significance, can realize simultaneous detection of multiple genes, and can comprehensively and accurately assess the intrinsic functional activity of CAR-T cells, filling the gap in existing technologies and promoting the standardization and normalization of CAR-T cell therapy. Summary of the Invention

[0008] The purpose of this invention is to address the shortcomings of existing technologies by providing a set of CAR-T cell endogenous functional characteristic genes, combined with a single-well multiplex detection system, to achieve integrated CAR-T cell identification, activation status assessment, proliferation potential grading, survival assessment, quality control, and clinical efficacy monitoring. This solves the core problem of existing technologies that "only look at quantity, not function." Currently, there is no CAR-T functional activity assessment scheme that integrates four dimensions—identity recognition, endogenous activation, proliferation metabolism, and anti-apoptotic survival—into a single gene combination and detection system. The implementation of this invention fills this technological gap.

[0009] To achieve the above objectives, the present invention adopts the following technical solution: a gene combination for CAR-T cell function identification and efficacy monitoring, comprising: endogenous functional genes BATF, PRDM1, SLC2A1, SOCS3, MCL1, BCL2L1, and CAR-specific sequences, internal reference genes GAPDH and ACTB.

[0010] An assessment system for CAR-T cell function identification, comprising:

[0011] The detection module is used to detect the expression levels of the above gene combinations in peripheral blood or bone marrow samples from patients. The standardized expression levels of each endogenous functional gene and CAR-specific sequence are obtained through GAPDH and ACTB dual internal reference gene correction.

[0012] The identification module is used to determine whether a cell is a CAR-T cell based on whether the detection module detects a CAR-specific sequence.

[0013] The activity analysis module is used to compare the standardized expression levels of each endogenous functional gene with preset thresholds to determine the expression level of each endogenous functional gene, and analyze the activity level of CAR-T cells based on the expression level.

[0014] Furthermore, the detection module includes specific primers and probes for detecting the expression levels of gene combinations. The specific primers are shown in the amino acid sequences SEQ ID NO.1 to SEQ ID NO.18, and the specific probes are shown in the amino acid sequences SEQ ID NO.19 to SEQ ID NO.27.

[0015] Further detection modules also include premixed solutions and reverse transcription reagents required for digital PCR reactions.

[0016] Further detection modules obtained the standardized expression levels of each endogenous functional gene and CAR-specific sequence by using a dPCR instrument to read the fluorescence signal of gene combinations, counting the number of positive droplets, calculating the absolute copy number of gene combinations, and correcting using GAPDH and ACTB dual internal reference genes, followed by sqrt(GAPDH²+ACTB²) normalization. The normalized expression levels of each endogenous functional gene and CAR-specific sequence are obtained by transforming the expression level by (normalized expression level × 10³ + 1).

[0017] The further identification module determines that the cell is a CAR-T cell when the detection module detects a CAR-specific sequence.

[0018] Further, the activity level of CAR-T cells was classified into three levels—high activity, intermediate activity, and low activity—based on the expression level of each endogenous functional gene.

[0019] The expression level of each endogenous functional gene is determined based on the comprehensive score S and the threshold according to formula (1):

[0020] (1)

[0021] In the formula, the letters of each endogenous functional gene represent the concentration of each gene. The intercept term and the coefficients of each gene are obtained by fitting the training set through sklearn Logistic Regression. The weighted linear combination is the comprehensive score S.

[0022] The P33 and P66 percentiles of the S-score distribution were used as activity grading thresholds: S _low =-2.19, S _high =-1.64, classifying it into three levels: high, medium, and low activity;

[0023] Highly active CAR-T cells showed high expression of BATF, PRDM1, BCL2L1, and MCL1, and low expression of SLC2A1 and SOCS3.

[0024] Active CAR-T cells are characterized by the expression of various genes;

[0025] Low-activity CAR-T cells showed low expression of BATF, PRDM1, BCL2L1, and MCL1, and high expression of SLC2A1 and SOCS3.

[0026] A kit for identifying the function of CAR-T cells includes: specific primers and probes for detecting the expression levels of the above-mentioned gene combinations, as well as a premixed solution and reverse transcription reagent required for digital PCR reaction. The specific primers are shown in the amino acid sequences SEQ ID NO.1 to SEQ ID NO.18, and the specific probes are shown in the amino acid sequences SEQ ID NO.19 to SEQ ID NO.27.

[0027] A method for assessing the functional activity of CAR-T cells includes the following steps:

[0028] (a) The expression levels of the nine targets in the above gene combination were detected, corrected using GAPDH and ACTB as dual internal reference genes, and normalized by sqrt(GAPDH²+ACTB²). (normalized expression level × 10³ + 1) transformation is used to obtain the normalized expression level of each gene;

[0029] (b) The normalized expression levels of each gene are used to obtain the comprehensive score S according to equation (1):

[0030] (1)

[0031] (c) The comprehensive score S of the six endogenous functional genes was determined according to the activity grading threshold S. _low = -2.19、S _high = -1.64 is used to classify expression into three levels: high expression, medium expression, and low expression;

[0032] (d) Determine the activity level of CAR-T cells based on the expression level of endogenous functional genes and output the determination results.

[0033] The above-mentioned gene combinations, evaluation systems, kits, or methods are used in the preparation of CAR-T cell functional activity detection reagents, quality control products, or efficacy monitoring products.

[0034] This invention, through extensive experimental screening, identified a highly specific combination of CAR-T cell endogenous functional genes. This combination comprises six core endogenous functional genes, combined with CAR exogenous specific sequences and dual internal reference genes, forming a four-dimensional integrated detection combination of "identity recognition—endogenous activation—proliferation metabolism—anti-apoptotic survival," specifically structured as follows:

[0035] 1. Endogenous activation-related genes (2)

[0036] (1) BATF gene: It belongs to the basic leucine zipper transcription factor family and is a key gene that regulates the activation, differentiation and effector function of T cells. It is significantly highly expressed in activated CAR-T cells. Its expression level can directly reflect the degree of activation of CAR-T cells and their effector ability to kill tumor cells. It is a core molecular marker for evaluating the anti-tumor activity of CAR-T cells.

[0037] (2) PRDM1 gene: namely positive regulatory domain zinc finger protein 1, is a key transcription factor that regulates the terminal differentiation of T cells, the formation of effector memory cells and the exercise of cytotoxicity. Its expression level is closely related to the strength of CAR-T cell cytotoxicity and the persistence of effector function.

[0038] 2. Genes related to proliferation metabolism and functional exhaustion (2)

[0039] The SLC2A1 gene (also known as the GLUT1 gene) encodes glucose transporter 1, mediating glucose transmembrane transport. Under continuous antigen stimulation or tumor microenvironment stress, this gene is continuously driven by the mTORC1 and HIF-1α signaling pathways and maintains high expression, reflecting that cells have entered a state of glycolytic reprogramming and functional exhaustion. In this model, it serves as a negative exhaustion marker, with a weighting coefficient of -0.629. It should be noted that SLC2A1 in a short-term activated state can positively reflect energy supply, but under continuous stimulation, it transitions to an exhaustion-increasing trajectory. Therefore, its relationship with the kill rate is initially positive and then reversed, exhibiting a non-monotonic relationship. This model uses its negative weighting under steady-state conditions.

[0040] (2) SOCS3 gene: namely, cytokine signaling inhibitor 3, can negatively regulate cytokine signaling pathways, balance the proliferation and activation state of CAR-T cells, inhibit cell exhaustion caused by excessive activation, and maintain the long-term functional stability of CAR-T cells. Under the continuous action of chronic cytokine signals (such as IL-6, IL-10, TGF-β), SOCS3, as a classic negative feedback inhibitor, is continuously upregulated, forming an exhaustion loop of cytokine-SOCS3-JAK / STAT signaling inhibition, leading to T cells entering a functional inactivation state (Akhtar et al., J Immunother Cancer, 2020; Palmer & Restifo, Nat Rev Immunol, 2009). The validation data of this invention show that the SOCS3 copy number in low-activity CAR-T samples is upregulated by about 1.5 times compared with medium-activity samples and by about 2.5 times compared with high-activity samples, which is consistent with the expected trend of the above exhaustion loop. Therefore, in the weighted scoring model, this gene also acts as a negative predictor, consistent with the classical inhibitory function of SOCS3 and the clinical phenotype of T cell exhaustion.

[0041] 3. Survival and anti-apoptosis related genes (2)

[0042] (1) MCL1 gene: Myeloid leukemia 1 gene, is a key member of the Bcl-2 family of anti-apoptotic subfamily. It can inhibit the apoptosis process of CAR-T cells and directly determine the survival time of CAR-T cells in vitro, the duration of in vivo and the long-term anti-tumor surveillance ability.

[0043] (2) BCL2L1 gene: namely Bcl-2-like protein 1 gene, which can resist activation-induced cell death, prolong the duration of CAR-T cell function in vivo, and reduce the efficacy decay and tumor recurrence risk caused by apoptosis after CAR-T cell infusion.

[0044] 4. Identification and Internal Reference Calibration Genes (2)

[0045] (1) CAR-specific sequence: It is a characteristic fragment of the exogenous CAR gene, used to confirm the successful integration of the CAR gene into T cells. It can effectively distinguish CAR-T cells from ordinary T cells, empty transfected T cells and other immune cells, and realize the specific identity recognition of CAR-T cells.

[0046] (2) GAPDH gene and ACTB gene: as dual internal reference genes, used to correct errors in the sample extraction process, fluctuations in the PCR amplification process and differences between different samples, to ensure the stability, reliability and comparability of the quantitative results of each target gene.

[0047] The gene combination described in this invention is not randomly selected; the genes exhibit clear functional synergy, enabling a comprehensive and precise assessment of the intrinsic functional state of CAR-T cells from four core dimensions. Its overall functional significance is as follows:

[0048] 1. Endogenous activation dimension: The expression levels of BATF and PRDM1 genes can directly reflect whether CAR-T cells are effectively activated and whether they have normal effector functions to kill tumor cells. High expression of both genes indicates that CAR-T cells are in an activated effector state with strong anti-tumor activity; low expression indicates that the cells are in a quiescent state or have low function and cannot effectively exert anti-tumor effects.

[0049] 2. Proliferation and Metabolic Dimension: The SLC2A1 gene and SOCS3 gene work synergistically to determine the in vitro expansion efficiency, in vivo proliferation capacity, and metabolic adaptability of CAR-T cells. The SLC2A1 gene provides the energy basis required for proliferation and effector function, while the SOCS3 gene balances proliferation and activation, avoiding excessive cell depletion. Together, they ensure that CAR-T cells can proliferate rapidly and maintain long-term functional stability.

[0050] 3. Anti-apoptotic survival dimension: MCL1 and BCL2L1 genes are core anti-apoptotic factors. Their expression levels are directly related to the in vivo survival time of CAR-T cells, long-term anti-tumor surveillance ability, and tumor recurrence risk. High expression of both genes can significantly prolong the survival time of CAR-T cells, enhance the anti-apoptotic ability of cells, and reduce the poor efficacy and recurrence risk caused by rapid cell apoptosis after infusion.

[0051] 4. Overall significance of the combination: This gene combination can comprehensively, accurately and stably assess the intrinsic functional activity of CAR-T cells through the synergistic detection of four core dimensions: activation, proliferation and survival. It completely solves the technical pain point of traditional detection that can only determine the number of cells but cannot assess cell function, and provides standardized molecular basis for the functional classification, quality control and clinical efficacy prediction of CAR-T cell products.

[0052] This invention further provides a weighted scoring model based on the expression levels of the aforementioned six endogenous functional genes, used for the objective and quantitative determination of CAR-T cell activity levels. This model uses the expression levels of these six endogenous genes... Normalized expression level was used as the input variable, and the comprehensive score S was calculated through a weighted linear combination, as shown in equation (1). In the equation, the normalized expression level was calculated by dividing the copy number of each gene by the square root of the sum of the squared copies of GAPDH and ACTB. Transformation form is (Standardized expression level × 10³ + 1).

[0053] (1)

[0054] The weighting coefficients in Equation (1) were obtained by fitting logistic regression of CAR-T samples, and L2 regularization was used to ensure the robustness of the model under small sample conditions. Analysis of experimental data revealed that the weighting coefficients of BATF, PRDM1, and BCL2L1 are positive, while the weighting coefficients of SLC2A1, SOCS3, and MCL1 are non-positive. The aforementioned weighting directions correspond intrinsically to the regulatory mechanisms of activation, proliferation metabolism, and anti-apoptosis dimensions of endogenous functional genes in CAR-T cells. The negative weighting coefficients of SOCS3 and SLC2A1 are consistent with the biological characteristic of this gene showing a reverse increase during the terminal differentiation and functional exhaustion stages of CAR-T cells due to continuous driving by the mTORC1 / HIF-1α signaling pathway and the negative feedback loop of cytokines. This constitutes a dual molecular marker of CAR-T functional exhaustion. This original discovery provides a new dimension for CAR-T functional activity assessment that differs from existing technologies.

[0055] Activity levels were determined using a dual-threshold method, with the p33 and p66 percentiles of the S-score distribution used to define the threshold S. _low = -2.19、S_high = -1.64; where S < S _low Determined to be of low activity, S _low ≤ S < S _high Determined as moderately active, S ≥ S _high The sample was classified as highly active. The above thresholds were calculated from the S-score distribution of CAR-T samples. If the distribution shifted between different batches of samples, it could be recalibrated using the same percentile method. The correspondence between in vitro specific kill rate and activity level is as follows: high activity corresponds to a kill rate of not less than 60%, medium activity corresponds to a kill rate of 30% to 60%, and low activity corresponds to a kill rate of less than 30%.

[0056] This invention employs digital PCR (dPCR) technology to construct a single-well multiplex detection system. Leveraging the advantages of dPCR technology, such as absolute quantification, high sensitivity, resistance to sample matrix interference, and stable detection of low-abundance targets, it enables the simultaneous detection of all targets (6 endogenous functional genes + 1 CAR-specific sequence + 2 internal reference genes) in a single reaction well, eliminating the need for multiple tubes. This effectively improves detection efficiency, saves sample volume, reduces detection costs, minimizes experimental errors, and ensures the accuracy and repeatability of detection results. It is suitable for the industrial-scale quality control of CAR-T cell products and routine clinical follow-up testing.

[0057] The weighted scoring model described in this invention significantly outperforms single-gene methods. Single-gene methods have limited ability to discriminate CAR-T functional activity, while the weighted scoring model of this invention integrates the synergistic expression information of six endogenous functional genes, significantly improving the overall discriminative ability for CAR-T functional activity, with an AUC index superior to any single-gene method. This result demonstrates that the multi-gene weighted scoring framework outperforms the simple superposition of single-gene methods, validating the non-simple superposition characteristics of the four-dimensional integrated gene combination.

[0058] The gene combination and detection system described in this invention can be widely applied to the entire process of CAR-T cell therapy, with specific application scenarios as follows:

[0059] 1. Quality control of CAR-T cell preparations: Used for functional activity grading after CAR-T cell preparation, screening for high-quality CAR-T cell preparations with high activation, high proliferation, and high anti-apoptotic survival ability, unifying quality standards between product batches, reducing fluctuations in clinical efficacy caused by differences in cell function, and improving clinical treatment response rate.

[0060] 2. Optimization of in vitro CAR-T cell preparation process: This is used to evaluate the effects of different culture conditions, cytokine combinations, matrix support systems, and culture times on the expression of functional genes in CAR-T cells, screen the optimal in vitro preparation process, and improve the overall quality and functional stability of CAR-T cell formulations.

[0061] 3. Predictive efficacy before clinical infusion: By detecting the expression level of this gene combination before CAR-T cell infusion, the in vivo expansion capacity, anti-tumor activity and long-term survival potential of CAR-T cells after infusion can be predicted in advance, providing a scientific basis for clinicians to formulate personalized infusion dosage and combination drug regimens, and reducing treatment risks.

[0062] 4. Clinical in vivo dynamic efficacy monitoring: After CAR-T cell infusion, peripheral blood or bone marrow micro-samples were collected from patients at preset time points (D7 / D14 / D28, M2 / M3 / M6 / M9 / M12 after infusion). The expression levels of this gene combination were detected to simultaneously monitor the in vivo number, endogenous activation status, proliferation capacity, and survival of CAR-T cells. This was combined with the results of tumor minimal residual disease (MRD) detection (flow cytometry, sensitivity 10). -4 Or NGS, sensitivity 10 -6 This allows for a comprehensive assessment of treatment efficacy, timely warnings of cell depletion, efficacy attenuation, and the risk of tumor recurrence, and provides molecular evidence for adjusting clinical treatment plans.

[0063] Compared with the prior art, the present invention has the following significant technical advantages and application value:

[0064] 1. Highly innovative and filling a gap in the industry: For the first time, it clearly proposes a combination of endogenous functional genes for CAR-T cells, consisting of BATF, PRDM1, SLC2A1, SOCS3, MCL1, and BCL2L1. It comprehensively evaluates the intrinsic functional activity of CAR-T cells from four core dimensions: identity recognition, endogenous activation, proliferation and metabolism, and anti-apoptotic survival. This breaks through the limitations of traditional detection methods that only focus on exogenous CAR genes or cell surface markers, and fills the gap in the quality control of CAR-T cell function.

[0065] 2. Clear functional significance and outstanding clinical value: Each member of this gene combination is directly related to the core function of CAR-T cells. Their expression level can directly reflect the anti-tumor activity, in vivo expansion capacity, long-term survival potential and recurrence risk of CAR-T cells. It provides molecular-level functional grading standards and efficacy prediction basis for CAR-T cell therapy, which can significantly improve the precision and effectiveness of clinical treatment.

[0066] 3. Highly efficient and stable detection, suitable for industrialization needs: Based on the dPCR single-well multiplex detection system, it can realize the simultaneous quantitative detection of multiple genes. It has high detection efficiency, low sample volume, low cost, strong anti-interference ability, accurate quantification, and good repeatability. It is suitable for the needs of industrial batch quality control of CAR-T cell products and routine clinical follow-up detection, and has good prospects for industrial application.

[0067] 4. High sensitivity and wide applicability: Relying on the high sensitivity of dPCR technology, it can accurately detect CAR-T cells with extremely low abundance in vivo, meeting the needs of long-term dynamic monitoring after CAR-T cell infusion; at the same time, this gene combination and detection system is not only applicable to lymphoma-targeted CAR-T cells, but can also be extended to the quality control and efficacy monitoring of CAR-T products for various hematological malignancies and solid tumors. The technology has strong versatility and huge market application value. Attached Figure Description

[0068] Figure 1 This is a schematic diagram of the BATF target detection results in an example.

[0069] Figure 2 This is a schematic diagram of the detection results of the target PRDM1 in the example.

[0070] Figure 3 This is a schematic diagram of the detection results of the target SLC2A1 in the embodiment.

[0071] Figure 4 This is a schematic diagram of the detection results of the target SOCS3 in the example.

[0072] Figure 5 This is a schematic diagram of the detection results of the target MCL1 in the embodiment.

[0073] Figure 6 This is a schematic diagram of the detection results of the target BCL2L1 in the embodiment.

[0074] Figure 7 This is a schematic diagram of the target CAR detection results in an example.

[0075] Figure 8 This is a schematic diagram of the detection results of the target GAPDH in the example.

[0076] Figure 9 This is a schematic diagram of the detection results of the target ACTB in the example.

[0077] Figure 10 This is the correlation between endogenous gene expression levels and kill rate in the examples.

[0078] Figure 11 This is a comparison of the ROC curves of six endogenous genes in Example 1.

[0079] Figure 12 This is a comparison chart of the score and lethality of Example S.

[0080] Figure 13 This is the confusion matrix diagram of Example 3×3. Detailed Implementation

[0081] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0082] It should be noted that the terms "comprising" and "having" and any variations thereof in the specification, claims and accompanying drawings of this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to such process, method, product or device.

[0083] Example 1: Establishment and validation of a single-well 9-fold dPCR detection system

[0084] Primer and probe design

[0085] For each target gene (BATF, PRDM1, SLC2A1, SOCS3, MCL1, BCL2L1, CAR-specific sequence, GAPDH, ACTB) in the gene combination described in this invention, specific primers and probes were designed. All primers are designed to span exons, which can effectively avoid genomic DNA contamination and ensure the specificity of detection results. The probes are labeled with different fluorescent groups to adapt to single-well multiplex detection and avoid fluorescence signal interference. Specific example sequences are shown in Table 1:

[0086] Table 1.9 Target Primer and Probe Sequences

[0087]

[0088] Sample processing and testing procedures

[0089] 1. Sample collection: Collect CAR-T cell in vitro culture preparation, peripheral blood samples or bone marrow samples from patients. Peripheral blood samples are anticoagulated, and bone marrow samples are separated into mononuclear cells by gradient centrifugation for later use.

[0090] 2. Nucleic acid extraction: Total RNA was extracted from the above samples using a dedicated nucleic acid extraction kit. The RNA purity was detected by Nanodrop (OD260 / 280 value controlled between 1.9 and 2.1, OD260 / 230 value not lower than 1.8). Subsequently, the total RNA was reverse transcribed into cDNA using a reverse transcription kit.

[0091] 3. Preparation of dPCR reaction system:

[0092] After removing the reagents, thaw them on ice for 15 min, vortex for 5 sec to mix thoroughly, and centrifuge at 3000 rpm for 30 sec before use. Thaw the templates on ice in the same way, vortex for 3 sec and centrifuge briefly for 5 sec before use. Prepare premixed solutions for each well according to Table 2. Set up one well for positive control and one well for negative control for each batch of tests. Prepare (n+2) portions of premixed solution according to the total number of samples n, with one extra portion prepared to compensate for aliquot loss.

[0093] Table 2 dPCR reaction system

[0094]

[0095] 4. Droplet generation and PCR amplification:

[0096] The prepared reaction system was added to the droplet generation chip. After uniform water-in-oil droplets were generated by the instrument, they were transferred to the PCR instrument and amplified according to the procedure in Table 3.

[0097] Table 3 dPCR amplification program

[0098]

[0099] 5. Signal Reading and Data Analysis:

[0100] After the PCR reaction is complete, a pre-optimized scan file is loaded for signal acquisition.

[0101] Positive determination rule: If an independent droplet cluster appears in the dual fluorescence channels corresponding to a target, and this cluster is positive in both channels simultaneously (i.e., located in the double-positive region), then the target is determined to be present in the sample. If there is no droplet cluster in the double-positive region, or only a single-channel positive cluster appears, then it is determined to be negative. Automatic thresholding of boundaries allows for precise cutting even in the presence of abnormal droplets.

[0102] Interpretation process: After each well scan, fluorescence signal overflow data of each singleton system in non-target channels are collected to establish a fluorescence crosstalk subtraction correction model. The median fluorescence intensity of the negative / positive droplet populations in each channel is used as the anchor point to delineate the double-positive region; the upper and lower thresholds are automatically generated by the template-free control (NTC) droplet populations on the same plate.

[0103] The number of positive droplets was counted using the instrument's software, the absolute copy number of each target gene was calculated, and the standardized expression level of each target gene was obtained through GAPDH and ACTB dual internal reference gene correction.

[0104] The probe combinations for the nine targets are assigned using a two-color orthogonal coding method, as shown in Table 4.

[0105] Table 4. Fluorescence channel allocation of target probes

[0106]

[0107] result

[0108] Single-plasm detection was performed on the positive plasmids of the nine targets, and fluorescence signal overflow data of each single-plasm system in the non-target channels were collected to establish a fluorescence crosstalk subtraction and correction model. After adjustment, the droplet clusters of the nine targets in each channel showed independent distribution without interference on the 2D scatter plot, which can be accurately segmented by an automatic thresholding algorithm. Figure 1-9 ).

[0109] Result determination

[0110] 1. Identification and determination: If a CAR-specific sequence is detected in the sample and the expression of endogenous functional genes conforms to the characteristic pattern of CAR-T cells, it can be identified as a CAR-T cell; if no CAR-specific sequence is detected, or the expression of endogenous functional genes does not conform to the characteristics of CAR-T cells, it can be identified as a non-CAR-T cell (ordinary T cells, empty vector transfected T cells, etc.).

[0111] 2. Functional activity grading: Based on the weighted scoring model of formula (1) and the activity grading threshold, CAR-T cell activity is divided into three levels: high activity, medium activity and low activity.

[0112] (1)

[0113] The specific threshold is S. _low = -2.19、S _high = -1.64.

[0114] A weighted scoring model was used to objectively and quantitatively determine the activity level of CAR-T cells. This model uses six endogenous genes... Normalized expression level was used as the input variable, and the comprehensive score S was calculated through a weighted linear combination, as shown in equation (1). In the equation, the normalized expression level was calculated by dividing the copy number of each gene by the square root of the sum of the squared copies of GAPDH and ACTB. Transformation form is (Standardized expression level × 10³ + 1).

[0115] 3. Clinical efficacy assessment: In clinical sample testing, combining the expression levels of each gene with the tumor MRD test results, if the number of CAR-T cells is sufficient and the expression of each functional gene is maintained at a high activity level, and the MRD test is negative, it indicates that the treatment is effective and the risk of recurrence is low; if the number of CAR-T cells decreases or the expression of functional genes drops to a medium to low level, and the MRD test is positive, it indicates that the efficacy is declining and the risk of recurrence is increasing, and the treatment plan needs to be adjusted in a timely manner.

[0116] Example 2: Validation of the specificity and sensitivity of the detection system

[0117] Based on the system established in Example 1, the detection specificity and sensitivity of the 9 targets were systematically verified.

[0118] I. Specificity Validation

[0119] Three types of negative control samples were collected: ① ordinary T cells derived from peripheral blood mononuclear cells (PBMCs) of healthy donors; ② T cells transfected with empty vector (Mock-T); ③ template-free control (NTC, nuclease-free water). Each type of control was prepared in triplicate, and the system was prepared according to Table 2 of Example 1 and tested.

[0120] Results: No independent positive droplet clusters were observed in any of the three negative controls across all nine target channels, resulting in a 100% negative detection rate (27 / 27 replicates). The number of positive droplets in each channel was ≤2 (below the lower threshold). Simultaneously, the above negative samples were compared with high-concentration positive plasmids (10... 6 When the samples (copies / μL) were mixed at a 1:1 ratio and then detected, the positive signals of the nine targets could still be clearly detected, with no non-specific crossover or signal suppression. The upper limit of the NTC noise background was ≤2 positive droplets, based on the Poisson distribution (λ=0.5, P(≤2)=0.999).

[0121] II. Sensitivity Verification (LOD)

[0122] The positive plasmids for the nine targets were diluted to 5 × 10³ copies / μL stock solution, and then serially diluted 10-fold to 5 × 10³ copies / μL. 2 5×10 1 5×10 0 Four concentration gradients were established (copies / μL). Each gradient had three replicates (n=3), and the systems were prepared according to Table 2 of Example 1 for instrumental analysis. Dilution was performed on an ice box at 4 °C, and a new pipette tip was used after each dilution to avoid interference from high-concentration residues in the low-concentration wells.

[0123] Standard curve and sample well arrangement order: Columns 1-3 of the plate are reserved for the standard curve of 9 target positive plasmids (5×10⁻⁶). 0 —5×10³ copies / μL, 4 concentration gradients × 3 replicates), column 4 is NTC (nuclease-free water), columns 5-12 are the samples to be tested. After sealing, centrifuge at 1000 rpm for 30 sec to remove air bubbles from the bottom of the wells.

[0124] Linear regression was performed on the data of 4 concentrations × 3 replicates (n=12) for each target, with the logarithm of copy number as the x-axis and the measured copy number as the y-axis; R² > 0.99 was considered acceptable for linearity. The LOD for each target was determined using a 95% positive detection rate as the criterion: all 3 replicates for each concentration gradient were positive, indicating detection at that concentration. The LOD and linear regression parameters for the 9 targets are shown in Table 5, indicating that the linear validation results for all 9 targets were acceptable.

[0125] Table 5. Target LOD and Linear Regression Parameters

[0126]

[0127] Example 3: Correlation between 9-fold dPCR expression profile and killing efficacy in CAR-T cells

[0128] This embodiment uses the 9-fold dPCR system established in Example 1 to simultaneously quantify 9 targets (6 endogenous genes, CAR, GAPDH, and ACTB) in 28 CAR-T cells (10 with high activity, 9 with medium activity, and 9 with low activity), analyzing the association between the expression levels of the 6 endogenous genes and the in vitro specific killing rate. An additional 10 independent validation cohorts were established for external validation in Example 5, in chronological order.

[0129] Activity is defined by in vitro specific killing rate: high activity ≥ 60%, medium activity 30-60%, low activity < 30% (see Table 6).

[0130] Table 6 Definition of lethality

[0131]

[0132] 1) Total RNA extraction: 1×10⁻⁶ RNA was extracted from each case. 6 Cells were lysed with 350 μL of RLT lysis buffer, and total RNA was extracted. Oligo(dT) was reverse transcribed into cDNA and stored at -20°C.

[0133] 2) cDNA quality control: Take 2 μL of cDNA and quantify it using GAPDH singlet dPCR. The absolute copy number should be ≥ 10e3 copies / μL, A260 / A280 1.9-2.1, A260 / A230 ≥ 1.8; if any one of them fails, re-extract.

[0134] 3) Single-well 9-fold dPCR system: Take 2 μL of cDNA and prepare a 7 μL reaction system (containing 6 endogenous genes, CAR, GAPDH, ACTB, a total of 9 probes) according to Table 2 of Example 1. Perform PCR amplification according to the procedure in Table 3.

[0135] 4) Signal reading: Two-color orthogonal encoding, 2D scatter thresholding, crosstalk correction.

[0136] 5) Output: Raw readings of 9 target copies / μL, normalized by sqrt(GAPDH²+ACTB²) and logarithmically transformed for downstream analysis.

[0137] Take another 1×10 from the same batch 6 Cells were washed once with PBS and resuspended in 1 mL of RPMI 1640 containing 2% FBS. Calcein-AM labeled lymphoma target cells (1×10⁻⁶) 5 Take 1 mL of the calcein-AM (E:T = 10:1) and mix it with 1 mL of effector cells in a 24-well plate. Incubate at 37°C with 5% CO2 for 4 h. Calcein-AM release (RFU) is read by a fluorescent microplate reader. Specific killing rate = (experimental release - spontaneous release) / (maximum release - spontaneous release) × 100%.

[0138] Kill rate quality control: Spontaneous release rate should be < 10% of maximum release; the difference between maximum release and spontaneous release should be > 5000 RFU, and the kill rate CV (3 replicates) should be < 20%.

[0139] result

[0140] (I) Raw readings of 9 target copies / μL. 28 training samples and 10 independent validation samples were simultaneously quantified by single-well 9-fold dPCR. The raw readings (unnormalized) are shown in Table 7 (copies / μL range):

[0141] Table 7.9 Original Readings of Targets

[0142] (II) Correlation between 6 endogenous genes and kill rate. Linear regression analysis was performed on the kill rate for each of the 6 endogenous genes, and the results are as follows. Positively regulated genes (BATF, PRDM1, SLC2A1, MCL1, BCL2L1) were positively correlated with the kill rate, while the negatively regulated gene SOCS3 was negatively correlated with the kill rate (inverse prediction, consistent with its role as an exhaustion signal) (Table 8). Figure 10 However, the R² range for a single gene is 0.06-0.59, which cannot independently predict the activity level and requires multi-gene weighting (see the S-score model in Example 4).

[0143] Table 8. Correlation between endogenous genes and kill rate

[0144]

[0145] Example 4: Weighted S-score of 6 endogenous genes and 4-dimensional functional activity model

[0146] This embodiment establishes a multi-gene weighted activation score S based on the 28 training data and 10 independent validation data provided in Example 3, and uses a "4-dimensional functional activity model" as the biological interpretation framework. The S score is composed of 6 endogenous genes. The normalized values ​​were obtained by weighted summation using logistic regression fitting. The coefficient β was fitted from the training cohort (n=28), with 6 genes as the independent variable. Normalized value, dependent variable = high activity vs. low activity.

[0147] I. dPCR Operation Procedure

[0148] This embodiment establishes an S-score model based on a training set of 28 cases; external validation of the 10-case independent validation cohort is described in Example 5. The training set cDNA samples are the same as in Example 3. The dPCR system, preparation, amplification procedure, and signal reading are the same as in Example 1, and the simultaneous quantification of the 9 targets is the same as in Example 3.

[0149] II. 4-Dimensional Functional Activity Model

[0150] The six endogenous genes described in this invention are functionally categorized into three dimensions: endogenous activation (BATF, PRDM1, reflecting TCR signal activation), proliferation metabolism (SLC2A1, SOCS3, reflecting energy metabolism and depletion balance), and anti-apoptotic survival (MCL1, BCL2L1, reflecting CAR-T cell survival ability). Combined with CAR probes (identification), this constitutes a four-dimensional integrated detection system. CARs are not included in the activity model and are treated as an independent QC dimension.

[0151] III. S-score formula (6-coefficient logistic weighting)

[0152] Activation score S is composed of 6 endogenous genes Weighted summation of normalized values:

[0153] (1)

[0154] The coefficient β was obtained by fitting the logistic regression (L2, C=0.1) to 28 training samples using sklearn. The dependent variable was high activity (kill rate ≥ 60%) vs. low activity, and the independent variable was 6 genes. Normalized values. See Table 9 for complete β, SE, p-values, and 95% CI.

[0155] Table 9. Logistic Fit Table for Gene Coefficients

[0156]

[0157] This model employs L2 regularization (C=0.1) to suppress overfitting. The S-value predicted by the six coefficients has an AUC of 0.917 on the training set, indicating that multi-gene weighting is superior to single-gene weighting. The interpretability of the S-value score is higher than that of simple mean or single-gene weighting: the coefficients of positively regulated genes (BATF, PRDM1, BCL2L1) are positive, while the coefficients of negatively regulated genes (SLC2A1, SOCS3, MCL1) are negative, consistent with the biological roles of each gene and in line with the biological expectation of increased SLC2A1 / SOCS3 exhaustion and downregulation of MCL1 anti-apoptosis. These results demonstrate the biological rationale for the four-dimensional weighted model. Furthermore, the p-value is not significant with a small sample size; the robustness of the coefficients will be further validated by expanding the training set (n≥100).

[0158] IV. Scoring and Grade Determination of the 28 Training Set S

[0159] The dPCR test results of 28 samples were normalized and analyzed. After logarithmic transformation, the S score is obtained by substituting into equation (1). Using binary classification (high vs. non-high), ROC analysis was performed on the 28 training data, and the Youden threshold was -1.489. Further analysis showed that the training set acc = 89.3% and AUC = 0.744. However, due to the actual needs of quality control, the single cut point obtained by binary classification is prone to overfitting when the small sample is imbalanced, and batch-to-batch jumps are likely to occur in industrial quality control. P33 / P66 is cut according to the distribution, is not sensitive to the sample composition, does not depend on a certain boundary sample, and is robust to small samples. S_high corresponds to the high activity boundary, and S_low corresponds to the medium and low boundary. The two thresholds are consistent with the monotonicity direction of the S score (the larger S is, the higher the activity), and the physical meaning is clear. This threshold method is more engineering operable than ROCYouden in industrial CAR-T quality control scenarios (batch standardization, not dependent on external gold standards). Therefore, P33 / P66 is used to grade and evaluate the thresholds of the three levels of activity.

[0160] The training set S-score P33 / P66 thresholds (P33, P66) = (-2.19, -1.64) are used for the three-level classification. The original S-scores and classification results are shown in Table 10.

[0161] Table 10 Training Set S-Score Three-Level Classification Results

[0162]

[0163] The training set accuracy was 78.6%, κ = 0.678, and AUC = 0.917. Of the 28 cases, 22 were correctly classified and 6 were misclassified, reflecting that the S score and kill rate are basically consistent. Approximately 15-25% of the variance comes from factors independent of the 6 genes, such as donor variability, batch variation, and E:T bias. Figure 11 ).

[0164] Example 5: Activity grading confirmation of 10 independent validation cohorts

[0165] In this embodiment, 10 independent batches of targeted CAR-T cells were collected (4 with high activity, 3 with medium activity, and 3 with low activity). The collection time was 8-12 weeks later than the training cohort. The generalizability of the S-score method was independently verified by applying Equation (1) and (P33, P66) thresholds from Example 4.

[0166] dPCR operation procedure

[0167] RNA extraction, cDNA quality control, preparation of single-well 9-fold dPCR system, amplification program, and signal reading for 10 validation samples followed the same procedure as in Example 1. dPCR detection values ​​are expressed as follows: Perform normalization, then perform After logarithmic transformation, the S-score was calculated by substituting into equation (1) (Table 11). The dPCR validation data and S-scores are as follows:

[0168] Table 11 Validation data and S scores of 10 dPCR cases

[0169]

[0170] The model's classification performance was evaluated using a validation set containing 10 samples, with an overall accuracy of 80.0% (95% confidence interval 50.0%–100.0%). This study is an ordered three-class classification task, and the calculated ordered weighted Kappa coefficient κ = 0.697, indicating good consistency in the classification results (κ = 0.61–0.80 represents a high degree of consistency). This evaluation used a one-to-many strategy to calculate the AUC for each activity level; due to the small size of the validation set, the AUC for all classifications was 1.00, and this result is only used as a reference indicator for this dataset. Of the 10 samples, 8 classification results were correct (V01, V02, V03, V04, V05, V06, V07, V08); 2 samples were misclassified: low-activity samples V09 and V10 were both classified as medium-activity, and both misclassifications were due to misclassification from low-activity levels to intermediate levels. The results above suggest that there is good consistency between the S-score and the sample killing activity grading (Table 12, Figures 12-13).

[0171] Table 12 3×3 Confusion Matrix of 10 Examples in the Validation Set

[0172]

[0173] IV. Conclusions and Working Thresholds

[0174] This study determined the S-score grading threshold as (S _low , S _high The S-score is calculated as (-2.19, -1.64). In the training cohort (n=28), the overall model accuracy is 78.6%, with a Wilson-Score accuracy of 62.8-94.4% confidence interval and an ordered weighted Kappa coefficient of 0.678. On an independent external validation set with 10 samples, the classification accuracy reaches 80.0%, with a Wilson-Score 95% confidence interval of 50.0-100.0% and an ordered weighted Kappa coefficient of 0.697. The consistency between the training and independent validation cohorts demonstrates the good applicability of the S-score grading threshold.

[0175] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A gene combination for CAR-T cell function identification and efficacy monitoring, characterized in that it comprises: Endogenous functional genes BATF, PRDM1, SLC2A1, SOCS3, MCL1, BCL2L1, and CAR-specific sequences, as well as internal reference genes GAPDH and ACTB.

2. An evaluation system for identifying CAR-T cell function, characterized in that... include: The detection module is used to detect the expression levels of the gene combination as described in claim 1 in a patient's peripheral blood sample or bone marrow sample, and obtains the standardized expression levels of each endogenous functional gene and CAR-specific sequence through GAPDH and ACTB dual internal reference gene correction. The identification module is used to determine whether a cell is a CAR-T cell based on whether the detection module detects a CAR-specific sequence. The activity analysis module is used to compare the standardized expression levels of each endogenous functional gene with preset thresholds to determine the expression level of each endogenous functional gene, and analyze the activity level of CAR-T cells based on the expression level.

3. The evaluation system for CAR-T cell function identification according to claim 2, characterized in that: The detection module includes specific primers and probes for detecting the expression levels of gene combinations. The specific primers are shown in the amino acid sequences SEQ ID NO.1 to SEQ ID NO.18, and the specific probes are shown in the amino acid sequences SEQ ID NO.19 to SEQ ID NO.

27.

4. The evaluation system for CAR-T cell function identification according to claim 2, characterized in that: The detection module also includes premixed solution and reverse transcription reagent required for digital PCR reaction.

5. The evaluation system for CAR-T cell function identification according to claim 2, characterized in that: The detection module obtains the standardized expression levels of each endogenous functional gene and CAR-specific sequence by reading the fluorescence signal of gene combinations using a dPCR instrument, counting the number of positive droplets, calculating the absolute copy number of gene combinations, and correcting using GAPDH and ACTB dual internal reference genes, followed by sqrt(GAPDH²+ACTB²) normalization. The normalized expression levels of each endogenous functional gene and CAR-specific sequence are obtained by transforming the expression level by (normalized expression level × 10³ + 1).

6. The evaluation system for CAR-T cell function identification according to claim 2, characterized in that: When the detection module detects a CAR-specific sequence, the identification module determines that the cell is a CAR-T cell.

7. The evaluation system for CAR-T cell function identification according to claim 2, characterized in that: The activity level of the CAR-T cells is divided into three levels: high activity, medium activity, and low activity, based on the expression level of each endogenous functional gene. The expression level of each endogenous functional gene is determined based on the comprehensive score S and the threshold according to formula (1): （1） In the formula, the letters of each endogenous functional gene represent the concentration of each gene. The intercept term and the coefficients of each gene are obtained by fitting the training set through sklearn Logistic Regression. The weighted linear combination is the comprehensive score S. The P33 and P66 percentiles of the S-score distribution were used as activity grading thresholds: S _low = -2.19, S _high = -1.64, classifying it into three levels: high, medium, and low activity; Highly active CAR-T cells showed high expression of BATF, PRDM1, BCL2L1, and MCL1, and low expression of SLC2A1 and SOCS3. Active CAR-T cells are characterized by the expression of various genes; Low-activity CAR-T cells showed low expression of BATF, PRDM1, BCL2L1, and MCL1, and high expression of SLC2A1 and SOCS3.

8. A kit for identifying the function of CAR-T cells, characterized in that... include: The specific primers and probes for detecting the expression level of the gene combination described in claim 1, as well as the premixed solution and reverse transcription reagent required for digital PCR reaction, wherein the specific primers are shown in the amino acid sequences SEQ ID NO.1 to SEQ ID NO.18, and the specific probes are shown in the amino acid sequences SEQ ID NO.19 to SEQ ID NO.

27.

9. A method for assessing the functional activity of CAR-T cells, characterized in that... Includes the following steps: (a) Detect the expression levels of the nine targets of the gene combination described in claim 1, corrected using GAPDH and ACTB as dual internal reference genes, and normalized by sqrt(GAPDH²+ACTB²). (normalized expression level × 10³ + 1) transformation is used to obtain the normalized expression level of each gene; (b) The normalized expression levels of each gene are used to obtain the comprehensive score S according to equation (1): （1） (c) The comprehensive score S of the six endogenous functional genes was determined according to the activity grading threshold S. _low = -2.19、S _high = -1.64 is used to classify expression into three levels: high expression, medium expression, and low expression; (d) Determine the activity level of CAR-T cells based on the expression level of endogenous functional genes and output the determination results.

10. The use of the gene combination of claim 1, the evaluation system of any one of claims 2-7, the kit of claim 8, or the method of claim 9 in the preparation of CAR-T cell functional activity detection reagents, quality control products, or efficacy monitoring products.

Images