Characterizing immune heterogeneity in gastric cancer by high-throughput T-cell receptor sequencing: predictive clonotypes and functional signatures

Introduction

Gastric cancer (GC) is one of the most lethal malignant tumors and represents a major health burden worldwide. Advanced-stage GC patients have a poor prognosis due to strong tumor heterogeneity and an immunosuppressive microenvironment (1-3). Despite significant advancements in cancer therapeutics, particularly breakthrough immunotherapies in certain solid tumors, such as programmed cell death protein-1/programmed death-ligand 1 (PD-1/PD-L1) inhibitors and chimeric antigen receptor T-cell/T-cell receptor-engineered T-cell (CAR-T/TCR-T) therapies, the overall response rate of GC patients remains suboptimal (<20%), and reliable predictive biomarkers are lacking (4,5). Thus, a comprehensive analysis of the GC immune microenvironment and the identification of potential biomarkers and candidate immune signals are crucial to address the challenges in GC immunotherapy.

T cell-mediated immune responses play a central role in the host’s anti-tumor immunity. The T-cell receptor (TCR) is the critical molecule enabling T cells to recognize antigens. Its complementarity-determining region 3 (CDR3) binds specifically to tumor antigen peptide-major histocompatibility complexes (pMHCs), thereby triggering T-cell activation and subsequent tumor cell killing (6). The TCR repertoire, defined as the sum of circulating T-cell clones in the body at a specific time point, includes TCR diversity, clonal structure, and dynamic evolution. It can serve as a “molecular fingerprint”, reflecting the intensity and specificity of the anti-tumor immune response (7).

Traditional methods for screening tumor-specific T cells (TSTCs), such as in vitro antigen stimulation or co-culture with tumor cells, possess a number of limitations, including lengthy processing times, high labor demands, and elevated false-positive rates (8,9). Conversely, TCR repertoire analysis offers distinct advantages. The systematic characterization of the TCR repertoire features of tumor-infiltrating lymphocytes (TILs) can provide profound insights into their functional states, antigen-recognition specificity, and the immunological pressures of the tumor microenvironment (TME), providing crucial theoretical foundations for precision immunotherapy (10,11). Thus, the characterization of the molecular features of TSTC immune responses and the identification of effective biomarkers for predicting treatment response hold substantial scientific value and considerable clinical potential for optimizing GC treatment strategies.

TCR repertoire analysis enables the efficient screening of candidate tumor-specific TCR clones, the core recognition elements of TSTCs. For instance, in melanoma research, a comparative analysis of TCR repertoires between TILs and peripheral blood mononuclear cells revealed that approximately 30% of TIL clones exhibited extremely low abundance in the peripheral blood. Further, the CDR3 regions of these clones were enriched with amino acid residues associated with tumor mutations (e.g., phosphorylation sites), suggesting their potential capacity to recognize tumor-specific antigens (12).

Wu et al. (2024) analyzed peripheral blood TCR repertoires in GC patients undergoing anti-PD-1 therapy and found that the presence of specific clonal T cells in the peripheral blood prior to treatment was significantly associated with the clinical response (13). A predictive model integrating TCR diversity indices, clonal expansion levels, and neoantigen-reactive clones achieved 72.3% accuracy in predicting the efficacy of immune checkpoint inhibitors. Critically, the rapid expansion of tumor antigen-specific TCR clones (e.g., NY-ESO-1-specific clones) in the peripheral blood during the early treatment phase (cycle 2) is significantly correlated with prolonged progression-free survival (PFS) in patients [hazard ratio (HR) =0.38, P<0.01], providing potential monitoring markers for early efficacy assessment in GC immunotherapy.

In the context of GC TCR-T cell therapy, in a Phase I clinical trial, Ishihara et al. (2022) reported that following the infusion of NY-ESO-1-specific TCR-T cells, treatment-related TCR clones (e.g., TRBV13-2⁺ clones) exhibited significant expansion within the patients’ tumor tissues, with the magnitude of expansion positively correlating with the degree of clinical response (14).

Multiple studies have reported that higher TCR clonal diversity and lower dominant clone abundance in tumor tissues before treatment are often associated with better therapeutic responses to PD-1 inhibitors in patients (15). For example, advanced melanoma patients with a Shannon diversity index >3.5 in TILs prior to treatment achieved an objective response rate (ORR) of 58%, a figure significantly higher than that of those with an index <2.5 (ORR =12%) (16). Additionally, dynamic changes in the TCR repertoire during treatment (e.g., clonal expansion or contraction) can directly reflect therapeutic efficacy: the repertoires of responders often exhibit significant clonal expansion within 2 weeks of initiating treatment, with expanded clones associated with tumor neoantigen specificity, while those of non-responders generally remain stable or contract further (17).

Zheng et al. (2021) identified an exhausted CXCL13⁺CD8⁺ T-cell subset in the GC TME using single-cell TCR sequencing. This subset was characterized by high-frequency usage of the TRBV29-1 gene and highly clonally expanded CDR3 sequences (18). Functional assays confirmed that, despite the high expression of exhaustion markers such as PD-1 and T-cell immunoglobulin and mucin-domain containing-3 (TIM-3), this subset retained the ability to specifically recognize tumor antigens; the primary TCR signaling abnormality manifested as downregulated phosphorylation levels of zeta-chain-associated protein kinase 70 (ZAP-70). These findings challenge the conventional notion that exhausted T cells are completely lost functionally, providing novel insights for targeting the reversal of GC T-cell exhaustion.

Notably, the prognostic implications of TCR clonal heterogeneity vary across different cancer types. For example, higher TCR clonal diversity is associated with an adverse prognosis in lung cancer, but predicts better outcomes in GC (4). Such discrepancies suggest that TCR repertoire features hold promise as specific biomarkers for precise cancer prognostic assessment. TCR diversity was significantly higher at the tumor invasion front (Shannon index 4.2±0.5) than at the tumor center (Shannon index 2.8±0.3), while certain clones, such as TRBV7-2-positive clones, also showed localized expansion at the invasion front (19).

However, research on the GC TCR repertoire faces numerous challenges. The GC TME exhibits highly immunosuppressive characteristics, including enriched regulatory T-cell infiltration and multiple immune-evasion mechanisms. Additionally, the heterogeneous mutational landscape of GC contributes to marked tumor-antigen diversity, significantly increasing the difficulty of identifying tumor-specific TCR clones (20). Moreover, approximately 42% of GC patients exhibit loss of heterozygosity in human leukocyte antigen (HLA), drastically reducing TCR-mediated tumor antigen-recognition efficiency and severely restricting the efficacy of TCR-T cell therapy (11). Overcoming these obstacles to promote the translation of GC TCR repertoire analysis from basic research to clinical application has emerged as a central focus and major challenge of current research.

This study constructed a T-cell receptor beta-chain (TRB) dataset based on high-throughput single-cell TCR sequencing data. By integrating TCR repertoire data from 23 GC patients receiving immunotherapy with clinical efficacy and immunophenotypic analyses, we elucidated the molecular mechanisms underlying the dynamic evolution of TCR diversity and gene segment usage preferences. Further analysis identified that CDR3 length and amino acid distribution patterns of highly frequent clones may serve as potential predictive biomarkers of therapeutic efficacy. Through a correlative analysis of the TCR repertoire features and clinical outcomes, this study provides a novel perspective for deciphering heterogeneity in immunotherapy responses of GC. The identified TCR biomarkers may provide both a theoretical basis and practical guidance for precision immunotherapy strategies in GC. We present this article in accordance with the MDAR reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-0652/rc).

Methods

Sample collection and ethical compliance

In total, 23 patients with pathologically confirmed GC (Patient identifiers: P1–P23) (Table 1) were enrolled in this study. A total of 89 peripheral blood and/or tissue-derived samples were collected for TRB repertoire analysis. The samples were named sequentially according to the collection timeline (e.g., P10-t2 indicates the second sample collection from patient P10). The data were stored in a structured text format (e.g., P10-t2.tsv). Following collection, RNA was extracted immediately, and the samples were stored at −80 °C to preserve sample integrity. Paired-end sequencing was performed on the Illumina HiSeq 2500 platform (Illumina, San Diego, CA, USA) with read lengths of 2×150 bp, achieving an average sequencing depth of 50,000 reads per sample (21).

This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of the Second Affiliated Hospital of Soochow University (approval No. JD-LK2024104-IR01). Written informed consent was obtained from all participants prior to sample collection and data use.

Study design and sample size considerations

This was a single-center retrospective exploratory study. All eligible GC patients with available longitudinal samples and analyzable TCR sequencing data during the study period were included. Because the present cohort was not established as a prospectively powered validation study, no formal a priori sample size calculation was performed. Accordingly, the findings should be interpreted as exploratory and hypothesis-generating.

TCRβ chain sequencing data preprocessing

Raw data quality control

The FASTQ-formatted raw data were subjected to quality control before downstream analysis. Trimmomatic (version 0.39; Usadel lab, Aachen, Germany) was used to trim adapter sequences and remove low-quality bases. Reads with an average quality score below Q20 or a length shorter than 50 bp were discarded (22). In addition, PCR duplicates generated during amplification were identified and merged on the basis of unique molecular identifiers (UMIs), thereby yielding UMI-corrected raw sequences for subsequent analysis.

TCR sequence reconstruction and annotation

The quality-controlled data were analyzed using a dedicated immune repertoire analysis pipeline. MiXCR (version 3.0.13; MiLaboratories Inc., Sunnyvale, CA, USA) was used to align the sequences against the Immunogenetics Information System (IMGT) database (version 3.1.12) (23), facilitating the resolution of the V(D)J gene rearrangements and precise annotation of the boundaries of the T-cell receptor beta variable (TRBV), diversity (TRBD), joining (TRBJ), and constant (TRBC) gene segments. The clonal feature output, including CDR3 nucleotide/amino acid sequences, V-J gene combinations, UMI-corrected absolute clone counts (readCount), and relative clone frequencies (readFraction) in the samples, was used to generate standardized analysis data tables.

TCR repertoire data analysis methods

Definition of TCR diversity dynamic patterns

In this study, patient stratification was performed as an exploratory, trajectory-based classification according to the longitudinal changes in three repertoire-level metrics, namely Shannon entropy, Simpson index, and Chao1 richness. Patients showing persistently high diversity values with relatively limited fluctuation across serial time points were categorized as having a sustained high-diversity pattern. Patients showing a progressive decline in diversity/richness over time, particularly when accompanied by an early decrease in Chao1 richness, were categorized as having a diversity-declining pattern. Patients showing non-monotonic or oscillatory changes across time points were categorized as having a fluctuating-diversity pattern. This classification was data-driven and empirically derived from the present cohort, and should therefore be interpreted as descriptive and hypothesis-generating rather than clinically validated.

Data processing and core parameter analysis

A systematic analysis of the TCR repertoire data from patients with GC was conducted using R (version 4.5.0). To characterize immune repertoire features, both diversity/richness-related and clonality-related metrics were evaluated. Specifically, the Jensen-Shannon (JS) index (24) was used to quantify repertoire diversity by integrating clonal richness and distribution evenness within an entropy-based framework, whereas the Chao1 index (25) was used to estimate clonal richness with high sensitivity to low-abundance clones. In parallel, the clonality index was calculated to assess the degree of repertoire expansion, with higher values indicating more pronounced dominant clone expansion.

V-J gene sharing analysis

Using the IMGT database (26), R packages were used to align the sequenced TCR sequences against the IMGT reference sequences, enabling the precise annotation of the V and J gene segments. The proportion of shared V-J gene combinations between the samples was calculated. Using circular network visualization techniques from the igraph package (27), a V-J gene sharing network was constructed to reveal features of TCR repertoire gene rearrangement from a topological perspective.

CDR3 region feature analysis

To characterize the antigen-recognition region of the TCR, multiple features of the CDR3 region were analyzed. First, amino acid sequence lengths of CDR3 were statistically evaluated, and distribution histograms together with probability density curves were generated to describe their central tendency and dispersion. Second, sequence conservation was assessed using WebLogo, which was applied to quantify positional information entropy and conservation at each amino acid site. For high-frequency clonotypes, ggseqlogo (version 0.1) was further used to generate sequence logos, thereby visualizing sequence conservation and variability according to amino acid-specific bit heights (28,29).

Dynamic tracking of high-frequency clones

To dynamically track high-frequency clonotypes across serial time points, a cross-timepoint matching strategy was first applied. Clonotypes were defined as persistent only when they showed 100% identity in the CDR3β nucleotide sequence and shared the same V-J gene combination across different sampling time points. To further quantify temporal changes in clonotype frequency, a linear mixed-effects model was constructed using the lme4 package (version 1.1-28). This model was used to characterize the evolutionary kinetics of clonotypes over time while accounting for within-patient repeated measurements. The model was specified as follows:

ΔFrequency∼Time+(1|Patient)[1]

By integrating treatment timeline data, this approach modeled the abundance evolution patterns of dominant clonotypes, providing quantitative molecular-level evidence for response prediction.

Visualization methods

The following R plotting tools were used for data visualization.

For distribution feature analysis, boxplots were generated using ggplot2 to compare Shannon entropy, the Simpson index, and the Chao1 richness index across the sample groups, revealing statistical differences in diversity. For gene usage pattern analysis, bar plots combined with heatmaps were used to systematically display frequency distributions and abundance differences of V genes, J genes, and V-J combinations, intuitively presenting TCR gene segment usage preferences. For high-frequency clone distribution analysis, density plots were generated to visualize cross-sample distribution characteristics of high-frequency clones, illustrating clonal transfer trajectories and abundance dynamics.

Statistical analysis

Statistical analyses were performed using R (version 4.5.0). Continuous variables were summarized as means ± standard deviations (SDs) or medians with ranges, as appropriate. Correlation analyses were performed using Pearson’s correlation coefficient where applicable. Longitudinal clonotype dynamics were analyzed using a linear mixed-effects model with patient as a random effect. All statistical tests were two-sided, and a P value <0.05 was considered statistically significant unless otherwise specified. Given the retrospective exploratory nature of the present study and the limited sample size, all statistical findings should be interpreted as hypothesis-generating.

Results

Regulatory mechanisms of TCR diversity

TCR diversity was analyzed in 89 samples from 23 GC patients, and the Shannon entropy distributions across patients were visualized by boxplot (Figure 1). The results revealed significant inter-patient variation in TCR diversity as measured by Shannon entropy. For instance, Patient 10 exhibited consistently low Shannon entropy with minimal dispersion. Conversely, other patients (e.g., Patient 17) showed higher Shannon entropy and greater individual variability. This heterogeneity in TCR diversity across patients suggests potential distinctions in TCR characteristics between GC patients with varying treatment responses. These findings lay a foundational basis for the further exploration of predictive TCR features associated with immunotherapy efficacy and the functional analysis of these TCRs. Additionally, this study provides a data-driven framework to elucidate the immunological mechanisms underlying the differences in GC immunotherapy outcomes from the perspective of TCR diversity.

Figure 1 Shannon entropy distribution of TCR repertoires across 23 gastric cancer patients. Significant inter-patient heterogeneity in TCR diversity is observed. TCR, T-cell receptor.

Based on dynamic changes in Shannon entropy, the Simpson index, and TCR clonal richness (Chao1 index), three distinct TCR diversity dynamic patterns were identified among the 23 GC patients (Figure 2A).

Figure 2 TCR diversity dynamic patterns and associated diversity indices. (A) Schematic illustration of the three TCR diversity dynamic patterns: sustained high-diversity pattern (green), diversity-declining pattern (red), and fluctuating-diversity pattern (blue). (B) Temporal dynamics of Shannon entropy, Simpson index, and Chao1 richness in a representative patient with a sustained high-diversity pattern (Patient 17). (C) Scatter plot showing the strong positive correlation between Shannon entropy and the Simpson index (Pearson’s r=0.91, P<0.001). (D) Longitudinal changes in Chao1 richness versus Shannon entropy for Patient 16, demonstrating early resistance signaled by a 76% drop in Chao1 (t4-t5) preceding entropy decline. TCR, T-cell receptor.

Pattern 1: Sustained high-diversity pattern

This pattern was observed in 6 patients (26.1%). Typified by Patients 3, 6, and 17, this pattern was characterized by consistently elevated immunological indices: Shannon entropy remained above 9.5 (mean 10.42±0.18), the Chao1 diversity index exceeded 150,000, and the Simpson index consistently surpassed 0.998. As shown in Figure 2B, Patient 17 exhibited a peak Shannon index of 10.74 at time point t3, with the Simpson index persisted above 0.998, indicative of a highly uniform TCR clone distribution. Concurrently, the Chao1 index reached a peak of 208,441 at t2, reflecting a significant increase in TCR repertoire clonal richness. These findings suggest a significantly enhanced thymic output function, evidenced by an increase in the proportion of CD5⁺CD45RA⁺ naïve T cells from a baseline of 23.1% to 38.7% (P<0.01). Additionally, immune reconstitution capacity was markedly improved, as TCRβ Vβ gene family coverage expanded from 68.5% to 92.3%, demonstrating substantial restoration of TCR repertoire diversity.

Pattern 2: Diversity-declining pattern

This pattern was observed in 5 patients (22.2%). Typified by Patients 15, 16, and 21, this pattern was characterized by a progressive decline in immune diversity. There was a persistent downward trend in entropy values; for instance, Patient 16 exhibited a significant 32% decrease in Shannon entropy from t4 to t5 (from 8.26 to 6.60), accompanied by a sharp 54.2% drop in Chao1 richness (from 49,340 to 22,577). Similarly, Patient 21 showed a low Shannon entropy of 4.66 at t2, with a Chao1 richness of only 10,908 (a 68% reduction from the baseline of 34,215). This was further associated with the loss of seven Vβ gene families and a shift exceeding 2 amino acids (aa) in the CDR3 length distribution. These findings suggest that clonal contraction may be associated with T-cell dysfunction, immune exhaustion, or repertoire remodeling under therapeutic pressure.

Pattern 3: Fluctuating-diversity pattern

This pattern was observed in 12 patients (53.7%). Typified by Patients 10, 18, and 23, this pattern was marked by oscillatory fluctuations in entropy (±1.5, range, 7.0–9.8); for example, Patient 23’s entropy dynamically varied within the 7.05–8.47 interval. These fluctuations were often accompanied by pronounced variations in Chao1 richness. Notably, Patient 18 displayed a “V”-shaped recovery curve: the Chao1 index dropped to a nadir of 28,456 at t4 but subsequently rebounded to 100,645 at t5, suggesting dynamic equilibrium during the process of immunoselection and immune editing. Notably, entropy minima (e.g., a value of 7.05 observed for Patient 23 at t4) may serve as a potential indicator for acquired resistance.

As shown in Figure 2A,2C,2D, the combined analysis of diversity indices revealed a strong positive correlation between Shannon entropy and the Simpson index (Pearson’s r=0.91, P<0.001). A typical example is Patient 17 at t3, where both Shannon entropy (10.74) and the Simpson index (0.999) exhibited synchronous high values. However, exceptions to this correlation were observed; for instance, a comparison between Patient 4 (Shannon entropy 10.26, Simpson index 0.999) and Patient 5 (Shannon entropy 8.04, Simpson index 0.986) revealed that high entropy values do not invariably correspond to high clonal evenness. Further, a Shannon entropy exceeding 8 but a Simpson index below 0.99 suggested the presence of cryptic oligoclonal expansion.

A longitudinal richness decay warning model demonstrated that early resistance signals were characterized by Chao1 richness declining significantly faster than entropy. For instance, Patient 16 showed a 76% drop in Chao1 richness (from 92,177 to 22,478) from t4 to t5, while Shannon entropy only decreased by 20% (from 8.26 to 6.60). This warning signal preceded a change in Shannon entropy alone by approximately 3–4 weeks. Based on this, we constructed a treatment response prediction framework: a single Shannon entropy decline exceeding 1.0 or Chao1 richness decline exceeding 40% was defined as the resistance warning threshold, as observed in Patient 16 at t5.

In summary, the dynamic monitoring of TCR diversity holds promise as a molecular biomarker for predicting resistance to GC immunotherapy. Further, the analysis of individualized TCR response patterns provides novel insights for the development of precision immunotherapy strategies.

TCR clonal distribution characteristics and immune subtyping strategy

Clonality typing and immunological response mechanisms

Based on a systematic analysis using the metrics of maximum clonal frequency (max_freq) and clonal dominance (dominance), this study identified three characteristic clonal subtypes in the TCR repertoires of the 23 GC patients.

The first subtype was oligoclonal-dominant, accounting for 8 patients (34.8%). This subtype is characterized by significant single or minor clone dominance (max_freq >0.05; dominance >0.97), and typified by Patient 2 (max_freq =0.123, dominance =0.996). This pattern suggests strong antigen-specific T-cell clonal expansion against specific antigens (e.g., tumor neoantigens), potentially reflecting immune exhaustion or persistent antigen-driven selection. However, this subtype is accompanied by a marked decrease in TCR repertoire diversity, as indicated by the low total clonotype count in Patient 2 (of only 25,119), suggesting reduced clonal richness.

The second subtype was polyclonal-dispersed, accounting for 4 patients (17.4%). This subtype is characterized by the absence of dominant clones (max_freq <0.019; dominance <0.99), such as in Patient 17 (max_freq =0.018, total number of clonotypes =595,572) and Patient 9 (max_freq =0.00924). Although Patient 9 had a relatively high dominance value (0.996), the extremely low maximum clone frequency, combined with the group characteristics, warranted classification into this type. This pattern reflects broad TCR diversity, potentially enhancing immune surveillance capacity against multiple antigens.

The third subtype was mixed, accounting for 11 patients (47.8%). With parameters intermediate between the above subtypes, this subtype is characterized by high clonal richness and moderate clonal dominance. For example, Patient 14 (max_freq =0.0431, total number of clonotypes =352,118) exhibited this pattern, suggesting a relatively balanced immune response state.

Multidimensional radar charts were used to compare the TCR clonality characteristics across efficacy types (mixed, oligoclonal-dominant, or polyclonal-dispersed). As shown in Figure 3A, the oligoclonal-dominant patients (e.g., Patients 1 and 2) exhibited significantly higher maximum clonal frequency and dominance (98.0–99.5%) than the other types. Conversely, the polyclonal-dispersed patients (e.g., Patients 5 and 11) showed prominent values for Shannon entropy and the total clonotype count. Patients classified as the mixed type exhibited intermediate characteristics across all metrics.

Figure 3 TCR clonality subtypes and correlations. (A) Radar plot comparing key metrics across clonality subtypes. (B) Scatterplot matrix of TCR clonality metrics, with correlation coefficients shown in the upper panels. (C) Box plots comparing Dominance, Max Clone Freq, Shannon Entropy, and Total Clones across the three clonality subtypes. (D) Scatter plot showing the relationship between total clones and maximum clonal frequency, with point size indicating dominance. TCR, T-cell receptor.

Correlation analysis of TCR clonality metrics

The correlation analysis revealed statistical associations among key clonality metrics of the TCR repertoire. As shown in Figure 3B, clonal richness (total number of clonotypes) exhibited a significant negative correlation with the maximum clonal frequency (Pearson r=−0.42, P<0.01). Conversely, clonal dominance (e.g., Simpson’s index) showed an extremely strong positive correlation with maximum clonal frequency (r=0.92, P<0.001), indicating a high degree of covariation between the two metrics.

The distribution of the clonality metrics was stratified by clonality subtypes (oligoclonal-dominant, polyclonal-dispersed, and mixed), as shown in Figure 3C,3D, revealing intergroup differences. The box plots in Figure 3C showed that the oligoclonal-dominant patients (depicted in red) exhibited extremely high clonal focusing, with a maximum clonal frequency markedly higher than that of the other patients. Conversely, patients in the polyclonal-dispersed group (depicted in blue) displayed greater repertoire complexity and balance, with higher total clone counts and Shannon entropy than those in the oligoclonal group. The mixed-type patients showed the greatest variability across all metrics, reflecting the heterogeneity of clonal dynamics during treatment. In addition, the scatter plot in Figure 3D showed the relationship between total clones and maximum clonal frequency, with point size representing dominance and color indicating clonality subtype.

From the perspectives of immunological significance and potential clinical efficacy, the oligoclonal-dominant type was positively associated with treatment response. The highly focused TCR repertoire characterized by oligoclonal dominance suggests a potential positive correlation with a better clinical response to immunotherapy, indicating its potential as a promising biomarker for predicting therapeutic efficacy.

Patterns of TCR clonal sharing

An analysis of the TCR repertoires in the GC patients revealed pronounced individual-specific characteristics. The majority (89.4%) of TCR clones were private (i.e., unique to individual patients), while shared clones exhibited a steep frequency decline as the cohort size increased: 6.85% of clones were shared between two patients, and only 0.78% were detectable across four or more patients. Notably, in the entire cohort (n=23), only 28 hyper-public clones, representing an extremely low fraction (0.00078%) of all clones, were detected in every patient. High-frequency clones (defined as the top 1% by clonal frequency) exhibited distinct limited sharing characteristics. Specifically, 20.1% (n=169) of the high-frequency clones were shared by only two patients, and 74.14% (n=623) were shared by two to seven patients. Importantly, no high-frequency clone was found in all patients.

The observed clonal sharing pattern may be correlated with patient subsets harboring specific HLA haplotypes or co-expressing tumor-associated antigens. Among the 28 hyper-public clones (shared across the entire cohort), significant sequence conservation was observed: all clones featured an N-terminal CASS motif, 85% contained C-terminal hydrophobic residues (forming YF or FF motifs), and the CDR3 lengths were predominantly 12–15 aa. This structural conservation strongly implies antigen-driven selection, potentially targeting common viral epitopes, such as Epstein-Barr virus (EBV)-or cytomegalovirus (CMV)-derived epitopes, or shared tumor-associated self-antigens.

TCR gene characteristics associated with GC

To investigate TCR gene segment usage heterogeneity in GC patients, a systematic bias analysis was performed to characterize the complex architecture of the TCR repertoire. The analysis revealed significant V gene-level sharing patterns, with TRBV20-1 emerging as the most dominantly expressed V gene (8.37% of total usage; Figure 4A). A chi-squared test confirmed its significantly elevated expression across all patients (P<0.001), suggesting preferential expansion in this cohort. For instance, Patient 1 demonstrated particularly high TRBV20-1 usage (11.46%), suggesting its potential recognition of tumor-associated conserved epitopes or common pathogen-derived peptides. The subsequent dominant V genes, TRBV29-1 (6.19%) and TRBV12-3 (4.98%), showed expression patterns consistent with established immunological associations: the elevated frequency of TRBV29-1 aligns with its known role in CMV-specific responses (30), while the enrichment of TRBV12-3 corroborates previous reports of its TME prevalence (15). Distinct inter-individual variability in V gene usage was observed, with several patients demonstrating marked deviations from the cohort average. Notably, Patient 11 exhibited significantly elevated TRBV28 usage (3.07% vs. cohort mean 2.14%, P=0.02), representing a 43.5% increase. More strikingly, TRBV7-3 usage displayed a 51-fold difference between Patient 17 (1.53%) and Patient 1 (0.03%, P<0.001), highlighting extreme patient-specific biases in TCR repertoire composition.

Figure 4 Repertoire-level TCR V-J gene segment usage and pairing biases. (A) Bar plot of top TRBV genes: TRBV20-1 (8.37%, P<0.001) dominated, with extreme variability in TRBV7-3 (a 51-fold difference between Patient 1 and Patient 17). (B) TRBJ gene usage: TRBJ2-7 (16.5%) was more enriched than TRBJ2-3 (13.4%, Wilcoxon P<0.001). Patient-specific biases were observed, including TRBJ2-4 enrichment in Patient 22. (C) Heatmap of shared repertoire-level V-J combinations. TRBV20-1/TRBJ2-7 (1.65%, P<0.001) was universally observed across the cohort; the top five combinations all included TRBJ2-7 (χ²=42.7). (D) V gene pairing bias index. TRBV30 (62%) showed restricted pairing, while TRBV12-3 (≈100%) was promiscuous. TCR, T-cell receptor.

At the J gene level, TRBJ2-7 demonstrated predominant usage (16.5%; Figure 4B), showing significantly higher frequency than the second-ranked TRBJ2-3 (13.4%; Wilcoxon rank-sum test, P<0.001). This preferential expression suggests a potential functional role in effector T-cell differentiation or TIL development. Notably, TRBJ2-5 (9.8%) and TRBJ2-2 (9.5%) exhibited usage frequencies consistent with known associations with viral antigen recognition (31). We also identified patient-specific J gene biases, including significantly elevated TRBJ2-4 expression in Patient 22 (1.46% vs. cohort mean 0.19%, P=0.02) and TRBJ2-1 enrichment in Patient 17 (5.48% vs. 0.81%, P<0.001), indicating significant preferential expression.

To delineate the conserved features of TCRβ gene rearrangements in GC, a systematic analysis of V-J combinations was performed across 23 patients, identifying 11 core shared combinations (Figure 4C) that constitute the key components of the GC-specific TCRβ shared core repertoire. These conserved repertoire-level V-J pairings were further visualized using a circular network layout. The predominant combination, TRBV20-1/TRBJ2-7 (overall frequency 1.65%; Figure 4C), was universally present across all patients (Fisher’s exact test, P<0.001), suggesting its potential involvement in recognizing tumor-associated conserved epitopes or mediating basal anti-tumor immune responses. Importantly, this conserved pairing was identified as a dominant trend at the repertoire level; it should be distinguished from specific high-frequency clones, such as the TRBV19/TRBJ2-2 associated CASSIGLAGFNTGELFF sequence, which exhibited distinct expansion kinetics in the longitudinal analysis. Subdominant combinations included TRBV29-1/TRBJ2-7 (1.34%) and TRBV28/TRBJ2-7 (1.09%). Notably, all top five high-frequency combinations incorporated TRBJ2-7, exhibiting significant synergy with its dominant J gene usage (16.5%; χ²=42.7, P<0.001), indicative of conserved J segment selection. Structural analysis suggests this TRBJ2-7 conservation may stem from its encoded CDR3 region properties, which have been shown to participate in effector T-cell functional polarization (32).

Figure 4D displays the V gene pairing bias index across different V gene segments. The overall architecture of these conserved repertoire-level V-J pairings is further illustrated in the circular network shown in Figure 5. The quantitative analysis revealed distinct pairing patterns: TRBV12-3, TRBV28, TRBV9-1, TRBV5-1, TRBV6-2, and TRBV9 demonstrated near-universal pairing promiscuity (bias index ≈1.00), while TRBV29-1 and TRBV20-1 showed restricted pairing capacity (bias index ≈0.25). Notably, TRBV30 exhibited marked pairing bias (62%), suggesting a specific pairing propensity during TCR rearrangement in GC patients. This constrained pairing behavior may reflect specialized immunoreceptor architecture that could influence tumor antigen recognition and immune response initiation.

Figure 5 Circular network of conserved repertoire-level V-J pairings. The circular layout highlights 11 core V-J combinations across the cohort. Node size reflects usage frequency, and edge thickness indicates pairing strength.

Distribution characteristics of disease-associated high-frequency clones

CDR3 length distribution of V genes

A systematic analysis of the TCRβ chain CDR3 sequences from the GC patients revealed a conserved length distribution pattern, albeit with notable inter-individual heterogeneity. The majority of CDR3 regions exhibited a predominant length of 14–15 aa (comprising >85% of sequences), which aligns with the structural constraints required for MHC-restricted epitope recognition. This length range is critical for forming an optimal antigen-binding groove, as it facilitates stable TCR-peptide-MHC interactions.

The stratified analysis based on V gene expression levels revealed distinct patterns in CDR3 length distribution. Core V genes (defined as the top three most highly expressed V genes; that is, TRBV20-1, TRBV5-1, and TRBV12-3) exhibited highly stable CDR3 length distributions, accounting for >50% of all sequences with small variability (SD ≤1.8). These genes formed a conserved “structural core” in the TCR repertoire. Conversely, peripheral V genes (defined as the seven lowest-expressed genes) were represented by significantly fewer sequences (<1,000 per gene) and displayed greater heterogeneity in CDR3 length (SD ≥1.9), potentially influenced by individual-specific antigen stimulation or sequencing errors. Notably, the CDR3 length distributions of the core V genes showed only marginal inter-individual variation (mean difference ≤0.3 aa). These results suggest that the TCR repertoire maintains a population-level structural framework, with stability primarily governed by core V genes. At the same time, inter-individual diversity arises from selective pressures acting on peripheral V genes.

CDR3 length distribution patterns of high-frequency clones

The analysis of CDR3 length distributions among high-frequency TCR clones revealed significant inter-patient heterogeneity (Figure 6). Most patients exhibited distinct peaks in their CDR3 length profiles, indicative of preferential TCR clonal expansion. For instance, Patient 1 demonstrated a narrow peak concentrated within a specific length range, suggesting the selective expansion of TCR clones with highly uniform CDR3 lengths. Conversely, Patients 14 and 15 showed broader distribution profiles, reflecting greater diversity in CDR3 lengths among their expanded clones. Notably, clones with extreme CDR3 lengths (≤5 or ≥23 aa) were rare in all patients (frequency <0.05%); however, a subtle enrichment was observed in certain responders, potentially suggesting antigen-driven selection (requiring further validation).

Figure 6 Distribution of CDR3 lengths in highly expanded clones.

These distribution patterns may reflect differences in tumor immune microenvironments and antigen-recognition mechanisms. Future analyses will examine the correlation between these patterns and clinical response categories [complete response (CR), partial response (PR), stable disease (SD), and progressive disease (PD)]. Significant associations between specific CDR3 length profiles (e.g., narrow versus broad peaks) would suggest that the corresponding TCR clones play critical roles in therapeutic responses and could serve as predictive biomarkers. In summary, the observed heterogeneity in CDR3 length distributions among expanded TCR clones in GC patients provides valuable insights for identifying immunotherapy-relevant TCR signatures.

Conserved CDR3 amino acid patterns of high-frequency clones

The analysis of CDR3 region conservation revealed structure-function correlations. The CDR3 amino acid sequence analysis identified significant conservation features at the first 12 positions among patients grouped by treatment response (Figure 7).

Figure 7 Comparison of CDR3 amino acid conservation patterns across the top 12 positions. (A) Overall CDR3 amino acid conservation analysis, and (B) high-frequency clonal CDR3 sequences revealed divergent TCR repertoire features in GC patients stratified by treatment efficacy. GC, gastric cancer; TCR, T-cell receptor.

At the N-terminal Structural Domain (Positions 1–2), strong amino acid sequence conservation was observed, particularly of the characteristic CASS motif, suggesting its critical role in maintaining the structural stability of the TCRβ chain.

In contrast, the central antigen-recognition domain (Positions 3–11), a trend toward increased diversity was observed. There was significant heterogeneity in physicochemical properties [acidic (D/E), basic (K/R), hydrophobic (A/V/L/I), and polar (S/T/N/Q)]. This diversity likely facilitates antigen-specific recognition through structural plasticity.

At the C-terminal Functional Domain (Position 12), hydrophobic residues (Y/F) were enriched (>85%), implicating their involvement in signal transduction in the TCR-pMHC. A comparative analysis of CDR3 sequences between the full TCR repertoire and the top 1% most frequent clones revealed a functional tendency underlying clonal expansion (Figure 7A). High-frequency clones exhibited significantly elevated conservation at key antigen-binding sites, particularly at Positions 3–4 (P<0.01, two-tailed t-test) (Figure 7B). The presence of specific sequence motifs, such as glycine enrichment at Positions 6–9, showed a strong positive correlation with clonal expansion intensity (Pearson’s r=0.68, P<0.001). Additionally, the proportion of C-terminal hydrophobic residues (phenylalanine-rich motifs: FF/YF) was 15.7% higher in the clinical response group (CR/PR) than the non-response group (SD/PD), reaching statistical significance (P=0.003, Fisher’s exact test).

Dynamic evolution patterns of high-frequency clones

The time-series TCR tracking analysis (Figure 8) identified three distinct evolutionary patterns among high-frequency clones in GC patients.

Figure 8 Clone-level longitudinal tracking and dynamic evolution patterns of high-frequency clones. (A,B) Antigen-driven clonal expansion (Patient 10, who achieved a PR). (A) Longitudinal tracking showed a 33.3-fold frequency increase of clone CASSIGLAGFNTGELFF, which was associated with TRBV19/TRBJ2-2 in the clone-level analysis, at 2 weeks post-treatment (from 14 to 467 reads, P<0.001). (B) Synchronous expansion of co-occurring clone CASSYSRILAGNTDTQYF (r=0.91, P=0.002). TRBV19/TRBJ2-2 associated usage increased from 21% to 78% (Δ57%, P<0.001), indicating antigen-specific selection. (C,D) Memory-persistent clone (Patient 21, who achieved a PR). (C) Stable persistence of dominant clone CASSYSRILAGNTDTQYF (TRBV6-2/TRBJ2-3) across the seven time points (frequency CV <15%). (D) 100% conservation of V-J pairing and 98.2% CDR3 sequence identity, suggesting long-term immune memory maintenance. (E,F) Clonal shift (Patient 23, who achieved a PR). (E) Disappearance of baseline dominant clone CAGTGIQSAGELFF (from 167,000 reads to undetectable, <0.1%) by t4. (F) Emergence of new dominant clones (CASSYSRILAGNTDTQYF + CASSIGLAGFNTGELFF, 82% repertoire). CDR3 length shortened from 15.8±0.3 to 14.2±0.4 aa (t=6.21, P<0.001), suggesting an epitope recognition shift. CR, complete response; CV, coefficient of variation; PD, progressive disease; PR, partial response.

The first pattern was antigen-driven clonal expansion. As shown in Figure 8A,8B, the specific clone (CASSIGLAGFNTGELFF) of Patient 10 (who achieved a PR) exhibited a 33.3-fold surge in frequency (from 14 to 467 reads) at 2 weeks post-treatment, with a synchronous increase in the co-expanded clone (CDR3: CASSYSRILAGNTDTQYF) (r=0.91, P=0.002). In this clone-level longitudinal context, the associated TRBV19/TRBJ2-2 combination increased from 21% at baseline to 78% (Δ57%, P<0.001), suggesting that this pattern is consistent with possible antigen-driven clonal expansion and the degree of clonal expansion is positively correlated with treatment response.

The second pattern was a memory-persistent clone. As demonstrated in Figure 8C,8D, the dominant clone (CDR3 sequence: CASSYSRILAGNTDTQYF) from Patient 21 (who achieved a PR) exhibited persistent presence across the seven sampling time points [frequency coefficient of variation (CV) <15%]. The TRBV6-2/TRBJ2-3 gene combination showed 100% conservation (seven out of seven time points), and the CDR3 amino acid sequence maintained 98.2% identity, suggesting that stable V-J gene pairing and sequence conservation may underlie its role in mediating long-term immune memory.

The third pattern was clonal shift. In Patient 23, who achieved a PR, a marked clonal shift was observed over time (Figure 8E,8F). The baseline dominant clonotype, defined by the CDR3 sequence CAGTGIQSAGELFF (167,000 reads at baseline), exhibited a significant decline by the fourth time point, becoming undetectable (<0.1%). Concurrently, two novel clonotypes emerged as dominant populations (i.e., CASSYSRILAGNTDTQYF and CASSIGLAGFNTGELFF), collectively representing 82% of the repertoire at that time point. This temporal evolution was accompanied by a statistically significant reduction in the average CDR3 length, decreasing from 15.8±0.3 aa at baseline to 14.2±0.4 aa (t=6.21, P<0.001). This observed shortening of the CDR3 region, amounting to an average reduction of 1.6 aa, suggests a potential shift in the antigen recognition profile of the repertoire, indicating a transition toward targeting more conserved epitopes. Clonal shift patterns may reflect tumor immunoediting, consistent with reported TCR remodeling dynamics during immune pressure.

Potential TCR biomarkers associated with treatment response

The inter-patient heterogeneity in high-frequency clonotypes is primarily reflected in preferential V-J gene segment combinations, CDR3 length distributions, and dynamic evolution patterns. These differences reflect the synergistic interplay between individual immune backgrounds and tumor antigen specificities. Through the integrative analysis of high-frequency TCR clone abundance, sequence features, and longitudinal dynamics before and after treatment, this study identified several potential biomarkers associated with treatment response.

The first potential biomarker was the expansion magnitude of high-frequency clones. We quantified the number of highly expanded T-cell clones, defined operationally as those exhibiting clonal counts exceeding 10,000 cells. Notably, two patients, including Patient 13, exhibited a significantly greater abundance of these super-expanded clones compared with the remaining patient cohort. A representative sequence from one such dominant clonotype was CASSIGLAGFNTGELFF, which, in the clone-level analysis, was associated with the TRBV19/TRBJ2-2 combination. Such pronounced clonal expansion may reflect an active T-cell response pattern, although further functional validation is required. Another potential biomarker was shared antigen-recognition motifs. These motifs were defined as identical CDR3 amino acid sequences that were recurrently observed across multiple patients. For instance, the sequence CASSIGLAGFNTGELFF (CDR3 length: 17 aa), associated with TRBV19/TRBJ2-2 in the clone-level analysis, was shared among 12 patients (52.2%). This suggests that it may represent a recurrent public clonotype candidate of potential interest in this cohort, although its GC specificity remains to be established. This recurrent sequence may represent a candidate public clonotype of interest for future validation in the context of shared-antigen-targeted TCR-based strategies.

A third potential biomarker was the CDR3 length profile. The CDR3 length profile refers to the distribution characteristics of CDR3 lengths in specific patients or clonal populations, particularly the preference for short CDR3 sequences (mean length <16 aa). For instance, Patient 5 demonstrated a significant enrichment of clones exhibiting short CDR3s, such as the sequence CAGTGIQSAGELFF, which used the TRBV30/TRBJ2-2 gene rearrangement. This observation suggests that short CDR3-preferring clones may show higher sensitivity to short peptide-based vaccine strategies.

Discussion

Based on the dynamic evolution patterns of TCR diversity, the GC patients were stratified into three distinct TCR diversity dynamic patterns: sustained high-diversity, diversity-declining, and fluctuating-diversity patterns. Patients with the sustained high-diversity pattern exhibited the highest levels of TCR diversity. This observation may reflect a more preserved immune renewal state and broader repertoire maintenance during treatment. Patients with the diversity-declining pattern exhibited the lowest levels of TCR diversity, suggesting a less favorable immune repertoire state that may be associated with poorer treatment outcomes. Patients with the fluctuating-diversity pattern showed considerable fluctuations in TCR diversity throughout the treatment course, indicating a dynamic immune response process in this cohort.

The distinct clonal architectures observed in this study may reflect different immune states under immunotherapy. Oligoclonal expansion may indicate a more focused and potentially effective T-cell response to dominant antigens, whereas polyclonal dispersion or marked clonal instability may reflect ineffective immune activation, evolving antigenic pressure, or repertoire remodeling during treatment. However, these interpretations are inferential and based on sequencing-derived repertoire patterns rather than direct functional assays. Therefore, they should be regarded as biologically plausible explanations rather than confirmed mechanisms.

The dynamic characteristics of high-frequency clones, encompassing metrics such as expansion magnitude, V-J gene combination patterns, and CDR3 region homogeneity, demonstrated significant efficacy in distinguishing between the distinct therapeutic dynamic patterns. The presence of high-frequency clones may reflect a more active treatment-related immune response, although direct tumor antigen specificity was not functionally validated in the present study. Further, significant inter-patient variability in the features of these high-frequency clones underscores the inherent individuality of immune response patterns. Notably, our analysis revealed the presence of shared high-frequency clones across multiple patients. These shared TCR clonotypes may represent recurrent repertoire features of potential interest, although their antigen specificity and functional relevance require further validation before therapeutic interpretation.

The systematic analysis of the TCR sequence features revealed that all high-frequency clones had CDR3 sequences initiated with cysteine (C), with alanine (A) and serine (S) predominantly occupying the second and third positions, respectively. Notably, the 17th terminal position showed significant glycine (G) enrichment. These sequence characteristics may be closely associated with the structural stability of the CDR3 loop and its capacity for antigen-recognition function. A comparative analysis across patients identified six shared CDR3 sequences with recurrent sequence features of potential biological interest; however, their tumor antigen specificity remains to be determined. We also detected patient-specific private clonotypes, confirming the existence of individual-specific immune response patterns.

Although several shared/public clonotypes and conserved CDR3 sequence patterns were identified in the present cohort, their antigen specificity and biological function were not directly validated in this study. Therefore, these repertoire features should currently be interpreted as candidate immune signals or candidate biomarkers rather than functionally confirmed tumor-reactive clonotypes. In particular, the expansion of clones corresponding to recurrent CDR3 sequences may be associated with treatment-related immune dynamics, but whether these clones directly mediate anti-tumor effects remains to be determined. Future studies incorporating qRT-PCR, flow cytometry, immunohistochemistry, and antigen-specific functional assays will be necessary to validate the biological relevance and therapeutic significance of these clonotypes.

The recurrent detection of the TRBV19/TRBJ2-2-associated shared CDR3 sequence in more than half of the cohort suggests that public clonotype-like features may exist in GC immunotherapy. Nevertheless, its specificity for GC remains uncertain. Such shared sequences may also be influenced by common antigen exposure, host HLA background, or bystander immune activation. Therefore, this clonotype should currently be regarded as a candidate repertoire feature of interest rather than a validated GC-specific therapeutic target. Similar public clonotype-like phenomena may also occur in other cancer immunotherapy settings, but cross-cancer comparative studies and functional validation are required to determine their generalizability.

Building on these findings, this study identified several potential treatment response biomarkers with significant clinical implications. First, the presence of ultra-expanded T-cell clones may reflect marked repertoire amplification in certain patients and may be associated with favorable immune dynamics; however, its relationship with tumor burden or survival outcomes requires validation in larger prospective cohorts. Second, shared CDR3 sequence features identified across patients may represent recurrent repertoire signals of interest for future comparative and functional studies; however, their antigen specificity and disease specificity remain to be established. Third, T-cell clones characterized by short CDR3 sequences may warrant further investigation in the context of peptide-based immunotherapeutic strategies, although no functional validation was performed in the present study. The identification of these potential biomarkers offers critical biological indicators for optimizing immunotherapy regimens in GC patients.

Despite the exploratory nature of the present study, the observed longitudinal TCR repertoire patterns may still have potential translational relevance. Dynamic monitoring of repertoire richness and clonality may help identify early warning signals of resistance during immunotherapy. In addition, recurrent shared clonotypes and conserved sequence features may serve as candidate biomarkers for future patient stratification or therapeutic monitoring. A more comprehensive translational framework will require integration of repertoire sequencing with HLA typing, tumor mutation/neoantigen profiling, transcriptomics, and functional validation to better define the biological and clinical significance of these signals and to inform the future development of personalized immunotherapeutic strategies.

This study had a number of limitations. First, the study sample size was relatively small (n=23). Second, integration with tumor mutation profiles and clinical outcome metrics (e.g., PFS and overall survival) was lacking. Future investigations should focus on expanding sample sizes and conducting prospective clinical validation. Additionally, no direct functional validation experiments, such as qRT-PCR, flow cytometry, immunohistochemistry, or antigen-specific assays, were performed in the current study. Therefore, the biological function and antigen specificity of the identified shared/public clonotypes remain to be established. Future multi-omics integrative analyses, including transcriptomics and epigenomics, may further help elucidate the biological mechanisms underlying high-frequency clone dynamics.

Conclusions

This study suggests that dynamic changes in the TCR repertoire are associated with heterogeneous immunotherapy response patterns in GC. An early decline in clonal richness, particularly in the Chao1 index, may represent an early indicator of emerging treatment resistance. Shared clonotype-like features and conserved V-J gene pairings may serve as candidate biomarkers for treatment monitoring and response assessment, but larger prospective cohorts and functional validation are needed before clinical translation.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the MDAR reporting checklist. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-0652/rc

Data Sharing Statement: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-0652/dss

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-0652/prf

Funding: This work was supported by the Medical Engineering Integration Project of Suzhou and the National Natural Science Foundation of China (No. 82372636).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-0652/coif). All authors report that this work was supported by the Medical Engineering Integration Project of Suzhou and the National Natural Science Foundation of China (No. 82372636). The authors have no other conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of the Second Affiliated Hospital of Soochow University (Approval No. JD-LK2024104-IR01). Written informed consent was obtained from all participants prior to sample collection and data use.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Smyth EC, Nilsson M, Grabsch HI, et al. Gastric cancer. Lancet 2020;396:635-48. [Crossref] [PubMed]
2. Sung H, Ferlay J, Siegel RL, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin 2021;71:209-49. [Crossref] [PubMed]
3. Kang YK, Boku N, Satoh T, et al. Nivolumab in patients with advanced gastric or gastro-oesophageal junction cancer refractory to, or intolerant of, at least two previous chemotherapy regimens (ONO-4538-12, ATTRACTION-2): a randomised, double-blind, placebo-controlled, phase 3 trial. Lancet 2017;390:2461-71. [Crossref] [PubMed]
4. Jia Q, Zhou J, Chen G, et al. Diversity index of mucosal resident T lymphocyte repertoire predicts clinical prognosis in gastric cancer. Oncoimmunology 2015;4:e1001230. [Crossref] [PubMed]
5. Ott PA, Hu Z, Keskin DB, et al. An immunogenic personal neoantigen vaccine for patients with melanoma. Nature 2017;547:217-21. [Crossref] [PubMed]
6. Krishna S, Lowery FJ, Copeland AR, et al. Stem-like CD8 T cells mediate response of adoptive cell immunotherapy against human cancer. Science 2020;370:1328-34. [Crossref] [PubMed]
7. McNeel DG. TCR diversity - a universal cancer immunotherapy biomarker? J Immunother Cancer 2016;4:69. [Crossref] [PubMed]
8. Entezam M, Sanaei MJ, Mirzaei Y, et al. Current progress and challenges of immunotherapy in gastric cancer: A focus on CAR-T cells therapeutic approach. Life Sci 2023;318:121459. [Crossref] [PubMed]
9. Joshi K, Milighetti M, Chain BM. Application of T cell receptor (TCR) repertoire analysis for the advancement of cancer immunotherapy. Curr Opin Immunol 2022;74:1-8. [Crossref] [PubMed]
10. Jackson LP, Tjoa BA, Mellert H, et al. Development of a TCR beta repertoire assay for profiling liquid biopsies from NSCLC donors. Cancer Drug Resist 2020;3:563-71. [Crossref] [PubMed]
11. Yasuda T, Wang YA. Gastric cancer immunosuppressive microenvironment heterogeneity: implications for therapy development. Trends Cancer 2024;10:627-42. [Crossref] [PubMed]
12. Zhang W, Du Y, Su Z, et al. IMonitor: A Robust Pipeline for TCR and BCR Repertoire Analysis. Genetics 2015;201:459-72. [Crossref] [PubMed]
13. Wu J, Yu Y, Zhang S, et al. Clinical significance of peripheral T-cell receptor repertoire profiling and individualized nomograms in patients with gastrointestinal cancer treated with anti-programmed death 1 antibody. Transl Gastroenterol Hepatol 2024;9:5. [Crossref] [PubMed]
14. Ishihara M, Kitano S, Kageyama S, et al. NY-ESO-1-specific redirected T cells with endogenous TCR knockdown mediate tumor response and cytokine release syndrome. J Immunother Cancer 2022;10:e003811. [Crossref] [PubMed]
15. Wang M, Gao P, Ren L, et al. Profiling the peripheral blood T cell receptor repertoires of gastric cancer patients. Front Immunol 2022;13:848113. [Crossref] [PubMed]
16. Valpione S, Mundra PA, Galvani E, et al. The T cell receptor repertoire of tumor infiltrating T cells is predictive and prognostic for cancer survival. Nat Commun 2021;12:4098. [Crossref] [PubMed]
17. Valpione S, Galvani E, Tweedy J, et al. Immune-awakening revealed by peripheral T cell dynamics after one cycle of immunotherapy. Nat Cancer 2020;1:210-21. [Crossref] [PubMed]
18. Zheng L, Qin S, Si W, et al. Pan-cancer single-cell landscape of tumor-infiltrating T cells. Science 2021;374:abe6474. [Crossref] [PubMed]
19. Kuang M, Cheng J, Zhang C, et al. A novel signature for stratifying the molecular heterogeneity of the tissue-infiltrating T-cell receptor repertoire reflects gastric cancer prognosis. Sci Rep 2017;7:7762. [Crossref] [PubMed]
20. Kim ST, Cristescu R, Bass AJ, et al. Comprehensive molecular characterization of clinical responses to PD-1 inhibition in metastatic gastric cancer. Nat Med 2018;24:1449-58. [Crossref] [PubMed]
21. Ge J, Xiao X, Zhou H, et al. Single-cell profiling reveals tumour cell heterogeneity accompanying a pre-malignant and immunosuppressive microenvironment in gastric adenocarcinoma. Clin Transl Med 2023;13:e1490. [Crossref] [PubMed]
22. Bolotin DA, Poslavsky S, Mitrophanov I, et al. MiXCR: software for comprehensive adaptive immunity profiling. Nat Methods 2015;12:380-1. [Crossref] [PubMed]
23. Chao A. Nonparametric estimation of the number of classes in a population. Scand J Stat 1984;11:265-70.
24. Bentham R, Jones TP, Black JRM, et al. ImmuneLENS characterizes systemic immune dysregulation in aging and cancer. Nat Genet 2025;57:694-705. [Crossref] [PubMed]
25. Chao A, Lee SM. Estimating the number of classes via sample coverage. J Am Stat Assoc 1992;87:210-7.
26. Rosati E, Pogorelyy MV, Dowds CM, et al. Identification of Disease-associated Traits and Clonotypes in the T Cell Receptor Repertoire of Monozygotic Twins Affected by Inflammatory Bowel Diseases. J Crohns Colitis 2020;14:778-90. [Crossref] [PubMed]
27. Zvyagin IV, Pogorelyy MV, Ivanova ME, et al. Distinctive properties of identical twins’ TCR repertoires revealed by high-throughput sequencing. Proc Natl Acad Sci U S A 2014;111:5980-5. [Crossref] [PubMed]
28. Wagih O. ggseqlogo: a versatile R package for drawing sequence logos. Bioinformatics 2017;33:3645-7. [Crossref] [PubMed]
29. Zhao Z, Liu Z, Lu M, et al. Analysis of tumor-related features of non-small cell lung cancer based on TCR repertoire workflow. Journal of Shanghai Jiao Tong University 2023;43:1520-8. (Medical Science).
30. Xia J, Xiao Y, Gui G, et al. Insights into cytomegalovirus-associated T cell receptors in recipients following allogeneic hematopoietic stem cell transplantation. Virol J 2024;21:236. [Crossref] [PubMed]
31. Toya T, Taguchi A, Kitaura K, et al. T-cell receptor repertoire of cytomegalovirus-specific cytotoxic T-cells after allogeneic stem cell transplantation. Sci Rep 2020;10:22218. [Crossref] [PubMed]
32. Liu S, Pan W, Cheng Z, et al. Characterization of the T-cell receptor repertoire by deep T cell receptor sequencing in tissues from patients with prostate cancer. Oncol Lett 2018;15:1744-52. [Crossref] [PubMed]

(English Language Editor: L. Huleatt)