Abstract
Background/Aim: Colorectal cancer (CRC) remains the leading cause of cancer-related mortality worldwide and necessitates the development of novel therapeutic strategies. The tumor immune microenvironment (TME) critically influences disease progression and the response to immune checkpoint inhibitors (ICIs). Tumor-infiltrating lymphocytes (TILs) are key components of the TME with established prognostic and predictive significance. Nevertheless, detailed TIL characterization using flow cytometry has not been fully investigated in CRC.
Materials and Methods: We analyzed TILs from 90 fresh CRC specimens using multicolor flow cytometry to investigate the association between specific T cell subsets and clinical outcomes. Patients were classified into Hot and Cold groups based on hierarchical clustering of TIL marker expression.
Results: The Hot group demonstrated significantly better overall survival (OS) compared to the Cold group (5-year OS: 86.7% vs. 63.9%, p=0.006), although recurrence-free survival (RFS) was not significantly different (5-year RFS: 79.5% vs. 66.1%, p=0.24). CITRUS analysis revealed that PD-1+Tim-3+CD103+CD8+ T cells were enriched in hot tumors (32.1% vs. 6.1%, p<0.001) and correlated with a favorable prognosis. Importantly, multivariate analysis demonstrated that a low frequency of PD-1+Tim-3+CD103+ cells among CD8+ T cells was an independent prognostic factor for OS [hazard ratio (HR)=3.36, 95% confidence interval (CI)=1.20-9.34, p=0.02].
Conclusion: A high frequency of PD-1+Tim-3+CD103+ CD8+ T cells is associated with better survival in CRC, highlighting their potential as a prognostic biomarker and therapeutic target.
Introduction
Colorectal cancer (CRC) is the third most common malignancy worldwide, with approximately 1.9 million new cases diagnosed annually (1). Its incidence is increasing worldwide, emphasizing the urgent need for novel therapeutic strategies. Among these, immune checkpoint inhibitors (ICIs) have been approved for various cancer types. However, their clinical application remains limited to patients with microsatellite instability (MSI)-high CRC, and ICIs have not yet been approved for those with microsatellite-stable (MSS) CRC (2). The tumor microenvironment (TME) has gained significant attention in recent years because it critically influences disease progression and therapeutic responses to ICIs (3). Therefore, a comprehensive evaluation of the TME is essential for identifying patients with CRC who may benefit from immunotherapy.
Tumors are often classified as “hot” or “cold” based on their immune activity. “Hot tumors” are characterized by abundant tumor-infiltrating lymphocytes (TILs), high expression of immune checkpoint molecules, and a pro-inflammatory microenvironment (4-6). These features are usually associated with high mutational burden, which enhances neoantigen presentation and immune recognition. In contrast, “cold tumors” lack TILs, display immunosuppressive features, frequently harbor a low mutational burden, impaired antigen presentation, and many suppressive cells such as regulatory T cells (Tregs). The infiltration of TILs is commonly evaluated using the Immunoscore (6), and their density and composition have been reported to predict survival (7-11). A high density of CD8+ TILs is associated with better overall and disease-free survival, reflecting their contribution to anti-tumor immunity. In contrast, an increased proportion of Tregs, which suppress immune responses, is associated with a poor prognosis.
Flow cytometry enables the precise quantification and phenotypic characterization of immune cell subsets, offering advantages over immunohistochemistry, which is limited in its ability to assess multiple markers simultaneously (12). Multimarker analyses, including those of PD-1 and Tim-3, have provided detailed insights into the activation state, functionality, and diversity of TIL subsets. Our group has applied flow cytometric analysis to various cancers, identifying prognostic markers and molecular subtypes based on TILs (4, 8, 9, 13-15). As mentioned above, the immune classification of CRC using the Immunoscore has been well established, but few studies have comprehensively characterized TILs using flow cytometry or clarified their association with clinical outcomes in CRC.
Therefore, in this study, we performed detailed multi-color flow cytometric analysis of TILs from CRC specimens and characterized their phenotypes. Importantly, this analysis utilized fresh viable cells processed immediately after surgical specimens, enabling the accurate assessment of immune cell phenotypes without artifacts caused by tissue fixation or prolonged storage. This approach enabled the capture of the immune landscape of the CRC microenvironment. We aimed to classify the TME based on the TIL phenotypes of CRC, evaluate their prognostic relevance, and identify TIL subsets associated with clinical outcomes.
Materials and Methods
Patients. Fresh CRC tumor samples were collected from 90 patients who underwent surgery at the University of Osaka Hospital between 2016 and 2021. Postoperative surveillance included blood examinations at three-month intervals and chest and abdominal CT imaging every 6 months. Clinicopathological factors and prognoses were retrospectively analyzed. The study protocol was approved by the Ethics Committee of the University of Osaka Hospital (no. 13266-36), and written informed consent was obtained from all participants in accordance with the Declaration of Helsinki.
Flow cytometry. To isolate TILs from fresh tumor samples, we used a Tumor Dissociation Kit with a gentleMACS Dissociator (Miltenyi Biotec, Bergisch Gladbach, Germany) according to the manufacturer’s guidelines. Isolated cells were incubated with Human TruStain FcX Fc Receptor-blocking solution (BioLegend, San Diego, CA, USA), followed by staining with fluorescent-conjugated antibodies and a live/dead fixed yellow dead cell staining kit (Life Technologies, Carlsbad, CA, USA). Stained cells were analyzed using an LSR Fortessa (BD Biosciences, San Jose, CA, USA), and the frequencies of cell populations were analyzed using FACSDiva software (BD Biosciences). An isotype control for each fluorophore was used to confirm positive staining. The gating strategy is illustrated in Figure 1.
Representative flow cytometry plots. OX40, TNF receptor superfamily member 4; PD-1, programmed cell death 1; Tim-3, T-cell immunoglobulin and mucin domain-containing protein; ICOS, inducible T-cell costimulator.
Antibodies. For flow cytometry, the following antibodies were used: CD45RA-FITC (clone HI100), CD25- PE (BC96), 4-1BB (TNFRSF9)-BV421 (4B4-1), CD8-BV510 (RPA-T8), CD103-BV605 (Ber-ACT8), CD4-BV711 (OKT4), CD45-BV786 (HI30), Tim-3-APC (F38-2E2), CD3-Alexa Fluor 700 (UCHT1), and IgG1 isotype control (MOPC-21), from BioLegend. Inducible T-cell co-stimulator (ICOS)-PE-Cy5.5 (ISA-3) and IgG1 isotype control (P3.6.2.8.1) were obtained from eBioscience (San Diego, CA, USA). OX40-PE-CF594 (ACT35), PD-1-PE-Cy7 (EH12.1), and IgG1 isotype control (X40) were obtained from BD Biosciences.
Heat mapping and cluster identification. Individual data were transformed to Z-scores for standardization, and the expression profiles of T-cell surface markers were visualized using heat maps. Hierarchical clustering was performed using Ward’s method in the JMP software version 17.2 (SAS Institute Inc., Cary, NC, USA). To identify distinct immune cell populations, we utilized Cluster Identification, Characterization, and Regression (CITRUS), a supervised machine learning algorithm integrated into the Cytobank platform (Beckman Coulter, Brea, CA, USA). CITRUS enables single-cell clustering and comparisons between different patient outcome groups by identifying significantly distinct populations. In this study, the CITRUS algorithm was applied to raw flow cytometry data containing 11 surface markers (CD3, CD4, CD8, CD45RA, PD-1, Tim-3, ICOS, 4-1BB, OX40, CD25, and CD103) to classify immune cell subpopulations. For CITRUS analysis, 5,000 events were randomly sampled per patient to ensure equal representation across samples. All 11 markers were used as clustering channels, and Hot and Cold groups were used as association variables. The minimum cluster size was set at 5% of total events. A nearest shrunken centroid predictive model (PAMR) was applied to identify clusters associated with clinical outcomes. Model performance was evaluated using 5-fold cross-validation to reduce overfitting. Multiple hypothesis testing was controlled using a false discovery rate (FDR) threshold of 1%. Samples were acquired across multiple days under standardized instrument settings. Instrument stability was ensured through daily quality control using calibration beads. To assess potential batch effects, event distributions and cluster compositions were visually examined across acquisition dates, and no systematic batch-related bias was observed.
Statistical analysis. The clinicopathological variables of the patients were analyzed using chi-square tests for categorical data, while continuous variables were compared using Student’s t-test. Nonparametric tests, specifically the Mann–Whitney U-test, were used to assess the significance of differences between groups. Recurrence rates were exclusively assessed in patients who achieved R0 resection. Kaplan–Meier curves were used to analyze recurrence-free survival (RFS) and overall survival (OS), and differences between the curves were assessed using the log-rank test. Patients with stage IV and R2 resection were excluded from the RFS analysis. Univariate and multivariate analyses of prognostic factors were conducted using the Cox proportional hazards model, providing hazard ratios and 95% confidence intervals. To determine the optimal cut-off value, the area under the receiver operating characteristic (ROC) curve was calculated. Statistical analyses were conducted using the JMP software version 17.2.
Results
Patients’ characteristics. Table I summarizes the patient backgrounds. The median age of the patients at the time of surgery was 67.5 years. The tumors were located in the right colon in 25 patients and in the left colon, including the rectum, in 65 patients. Pathological staging revealed that 12 patients had stage I, 32 had stage II, 28 had stage III, and 18 had stage IV. Among those who underwent curative surgery, 18 patients (20.0%) developed recurrence during the follow-up period.
Clinicopathological characteristics of patients in this study.
Heat mapping and prognostic analysis in the Hot group and Cold group. All patients were examined using hierarchical clustering of TIL surface marker expression and classified into two groups: Hot and Cold (n=45 each) (Figure 2A). The pN and pM statuses tended to be higher in the Cold group (p=0.098 and p=0.063, respectively), while other clinicopathological factors did not differ significantly between the two groups (Table II). Regarding prognosis, overall survival rates (OS) were significantly higher in the Hot group than in the Cold group (5-year OS rate of 86.7% vs. 63.9%, p=0.006) (Figure 2B). Among patients who underwent R0 resection, recurrence-free survival rates (RFS) were not significantly different between the two groups, although the Hot group exhibited a higher 5-year RFS compared with the Cold group (5-year RFS rate: 79.5% vs. 66.1%, p=0.24). Furthermore, multivariate analyses identified Cold group (p<0.03) and pStageIII or IV (p<0.01) as independent prognostic factors for OS (Table III).
Clustering and prognosis based on TIL phenotypes in CRCs. (A) Clustering of CRC patients using a hierarchical clustering method with 11 surface markers of TILs. (B) Kaplan-Meier curves for overall survival based on the Hot (red) and Cold (blue) groups. (C) Kaplan-Meier curves for recurrence-free survival based on the Hot (red) and Cold (blue) groups. p-Values were calculated using the log-rank test. TIL: Tumor-infiltrating lymphocytes; CRC: colorectal cancer.
Background characteristics of the Hot and Cold groups.
Univariate and multivariate analysis of overall survival according to Hot and Cold groups.
Comparison of T-cell marker expression between the two groups. The frequencies of TIL subpopulations and expression of surface molecules on TILs in the two groups are shown in Figure 2. The Hot group exhibited a higher frequency of CD8+ T cells (40.6% vs. 27.4%, p<0.0001) and a lower frequency of CD4+ T cells (48.2% vs. 61.6%, p<0.0001) compared with the Cold group (Figure 3A). In CD8+ T cells, all markers except CD45RA were significantly upregulated in the Hot group compared to those in the Cold group (Figure 3B). In CD4+ T cells, the expression of most markers, except CD45RA and CD103, was higher in the Hot group (Figure 3C). These findings indicate that the Hot group exhibited higher expression of immune checkpoint molecules, suggesting an activated immune status in hot tumors.
TIL phenotype based on surface marker expression between the Hot and Cold tumors. Box-and-whisker plots showing (A) the frequency of CD45+ cells in viable cells, lymphocytes in CD45+ cells, CD3+ cells in lymphocytes, CD8+ cells in CD3+ T cells, CD4+ cells in CD3+ T cells, (B) the expression frequency of CD45RA, PD-1, Tim-3, 4-1BB, OX40 and CD25 in CD8+ T cells, and (C) the expression frequency of CD45RA, PD-1, Tim-3, ICOS, 4-1BB, OX40 and CD25 in CD4+ T cells. Differences were assessed using the nonparametric Mann–Whitney U-test (*p<0.05, **p<0.01), and only statistically significant comparisons were annotated. TIL: Tumor-infiltrating lymphocytes.
CITRUS analysis between the Hot and Cold group. To identify the distinct cell populations between the two groups, CITRUS analysis was performed using TIL surface marker expression data obtained from flow cytometry (Figure 4A). Cluster 21496 was identified as a characteristic of the Hot group, with higher expression of CD8, PD-1, CD103, and Tim-3 (Figure 4B, C). Flow cytometry analysis confirmed that the frequency of PD-1+Tim-3+CD103+ cells among CD8+ T cells was significantly higher in the Hot group than in the Cold group (32.1% vs 6.1%, p<0.001) (Figure 4D-E).
CITRUS analysis comparing the Hot and Cold groups. (A) Schematic overview of the CITRUS analysis. (B) CITRUS identified cluster 21496 and its frequency was compared between the Hot and Cold groups. (C) Expression patterns on T-cell markers in cluster 21469 relative to all TILs. (D) Representative flow cytometry plots showing PD-1+Tim-3+CD103+CD8+ cells in the Hot and Cold groups. (E) Frequency of PD-1+Tim-3+CD103+ cells among CD8+ T cells in the Hot and Cold groups. PD-1, Programmed cell death 1; Tim-3, T-cell immunoglobulin and mucin domain-containing protein 3; TIL: tumor-infiltrating lymphocytes. Differences between groups were analyzed using the Mann–Whitney U-test. **p<0.01.
The clinical relevance of PD-1+Tim-3+CD103+ CD8+ T cells. To evaluate the prognostic value of PD-1+Tim-3+CD103+CD8+ T cells, receiver operating characteristic (ROC) curve analysis was performed to determine the optimal cut-off value for the 5-year OS. Based on a cut-off value of 11.7%, patients were divided into two groups: high (n=50) and low (n=40). No significant differences in the background characteristics were observed between the two groups (Table IV). Regarding prognosis, patients with a high frequency of PD-1+Tim-3+CD103+ cells among CD8+ T cells had significantly better OS than those with a low frequency (5-year OS: 86.2% vs. 61.3%, p=0.004) (Figure 5). In univariate analyses for OS, pStageIII or IV (p<0.01), postoperative chemotherapy (p=0.01), and a low frequency of PD-1+Tim-3+CD103+ cells among CD8+ T cells (p=0.01) were associated with poor prognosis (Table V). Furthermore, multivariate analyses identified a low frequency of PD-1+Tim-3+CD103+ cells among CD8+ T cells (p=0.02) and pStageIII or IV (p<0.01) as independent prognostic factors for OS.
Background characteristics of patients with high and low frequency of PD-1+Tim-3+CD103+ cells among CD8+ T cells.
Kaplan-Meier curves for overall survival based on PD-1+Tim-3+CD103+ CD8+ T cells. Overall survival of patients with high (red) and low (blue) frequency of PD-1+Tim-3+CD103+ cells among CD8+ T cells. p-Values were calculated using the log-rank test. PD-1, Programmed cell death 1; Tim-3, T-cell immunoglobulin and mucin domain-containing protein 3.
Univariate and multivariate analysis of overall survival according to frequency of PD-1+Tim-3+CD103+ cells among CD8+ T cells.
Discussion
In this study, we performed comprehensive flow cytometry profiling of TILs in CRCs and classified CRC tumors into hot and cold phenotypes based on TIL density and activation. Hot tumors are characterized by a higher proportion of CD8+ T cells and elevated expression of immune checkpoint molecules, including PD-1, Tim-3, and CD103. Among these, CD8+ T cells co-expressing PD-1, Tim-3, and CD103 were markedly enriched in hot tumors and were significantly associated with improved overall survival, highlighting their potential as prognostic biomarkers and reflecting a functionally active anti-tumor immune response.
We classified colorectal cancer tumors into hot and cold phenotypes based on the density and activity of TILs obtained using multicolor flow cytometry. Recent studies have shown that cancer progression is influenced by the host immune system. The immune classification of tumors was proposed based on a simple immune score that quantifies the density and location of immune cells within the tumor (16, 17). Patients with high immune scores exhibited significantly better survival than those with low immunoscores, highlighting that tumors with high densities of CD3+ and CD8+ T cells (hot tumors) were associated with more favorable outcomes. These findings highlight the pivotal role of the tumor immune microenvironment in shaping disease progression and patient survival. Our results are consistent with these observations, reinforcing the importance of immune profiling in CRC and its potential use in guiding therapeutic strategies and improving prognosis.
TILs in CRC comprise heterogeneous subpopulations with distinct effects on tumor progression and prognosis. High densities of CD8+ T cells are consistently associated with improved survival, emphasizing their central role in anti-tumor immunity (16, 18, 19). However, chronic antigen exposure can induce dysfunctions characterized by the expression of PD-1 and Tim-3 (20-22). Although these molecules are often regarded as exhaustion markers, recent evidence indicates that PD-1+Tim-3+CD8+ T cells may retain effector potential and respond to ICIs. In non-small cell lung cancer, CD8+ T cells co-expressing PD-1 and Tim-3 demonstrate higher cytokine production, indicating a transient dysfunctional state rather than full exhaustion (23, 24). In our study, CD8+ T cells co-expressing PD-1, Tim-3, and CD103 were associated with a favorable prognosis. CD103 defines tissue-resident memory T (TRM) cells that persist locally, secrete IFN-γ and TNF, and exert durable cytotoxicity. CD103+CD8+ TILs have been reported in multiple cancers, including ovarian, lung, and hepatocellular carcinoma, where they are linked to better outcomes and ICI responsiveness (25-30). Therefore, the association with improved survival suggests that these PD-1+Tim-3+CD8+ T cells not only represent activated populations but may also reflect antigen-experienced, tumor-reactive T cells enriched within the local microenvironment. Overall, the prognostic significance of PD-1+Tim-3+CD103+ CD8+ T cells in CRC likely reflects the presence of TRM-like CD8+ T cells that maintain anti-tumor activity despite expression of exhaustion markers, highlighting the complex dynamics of TIL biology (31, 32).
Although the overall efficacy of ICIs in CRC is generally limited, the identification of hot tumors in this study, as determined by flow cytometric analysis of TILs, and their association with a favorable prognosis, may represent a potential approach for stratifying patients who could benefit from ICIs (33, 34). Other approaches, including transcriptomic, spatial, and histological analyses, have also been reported to delineate immunologically active subgroups within MSS-CRC (35-38). Collectively, integrating diverse analytical methods will be essential for identifying pMMR patient subsets that may benefit from ICI therapy.
In this study, we applied a profiling analysis of the CITRUS algorithm to identify the specific cell population that differed significantly between the Hot and Cold groups. CITRUS is an unsupervised, high-dimensional analytical approach that reduces the bias inherent in manual gating and using comprehensive cytometric marker panels to reveal statistically significant cellular subsets associated with tumor-specific environments (4, 39, 40). This method effectively identifies clinically relevant cell subsets, facilitates biomarker discovery and therapeutic target identification, and enables robust statistical comparison of immune cell populations across conditions, highlighting subtle yet important shifts in immune profiles (41, 42). Moreover, integrating CITRUS into a discovery-validation workflow enhances its utility. In our study, CITRUS identified cluster 21496 as a characteristic population in the Hot group comprising CD8+ T cells with high expression of PD-1, CD103, and Tim-3, which were subsequently subjected to functional analysis of these subsets. Thus, CITRUS is a useful tool for unbiased identification and characterization of distinct immune cell populations in complex tissues.
Study limitations. The cohort included 90 patients from a single institution, underscoring the need for validation in larger multicenter studies. As routine clinical testing for MSI and TMB in CRC had not yet been widely implemented at the time of this study, the MSI status and TMB, which are both established determinants of tumor immunogenicity and predictors of ICI response, were not assessed. Future studies incorporating genomic and transcriptomic analyses will be essential to clarify how these molecular characteristics interact with immune phenotypes identified by flow cytometry, and to further refine patient stratification for immunotherapy.
Conclusion
We identified hot CRC tumors, in which a high frequency of PD-1+Tim-3+CD103+ CD8+ T cells was associated with a favorable prognosis. These CRC tumors may represent potential targets with higher responsiveness to ICIs. Future studies should investigate the mechanisms that regulate their activity and explore therapeutic strategies to enhance their anti-tumor functions.
Footnotes
Authors’ Contributions
Daijiro Matoba contributed to conceptualization, methodology, software development, data curation, investigation, validation, formal analysis, visualization, project administration, and writing of the original draft. Takuro Saito contributed to conceptualization, methodology, investigation, supervision, funding acquisition, project administration, and writing of the original draft and review and editing. Takehiro Noda contributed to conceptualization, methodology, investigation, supervision, visualization, project administration, and writing of the original draft and review and editing. Azumi Ueyama contributed to software development, data curation, investigation, formal analysis, and review and editing. Mamoru Uemura, Hisashi Wada, Yuichiro Doki, and Hidetoshi Eguchi contributed to supervision, resources, and review and editing. Funding was acquired by Takuro Saito and Hisashi Wada.
Data Availability Statement
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.
Conflicts of Interest
The Authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this article.
Funding
This work was supported by institutional funding from Osaka University. The Department of Clinical Research in Tumor Immunology, Graduate School of Medicine, Osaka University has a joint research laboratory with Shionogi & Co., Ltd.
Artificial Intelligence (AI) Disclosure
No artificial intelligence (AI) tools, including large language models or machine learning software, were used in the preparation, analysis, or presentation of this manuscript.
- Received January 26, 2026.
- Revision received March 3, 2026.
- Accepted March 9, 2026.
- Copyright © 2026 The Author(s). Published by the International Institute of Anticancer Research.
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.
References
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.