Differences in tumor-associated T-cell receptor repertoires between early-onset and average-onset colorectal cancer

Abstract

The incidence of colorectal cancer (CRC) among individuals younger than age 50 (early-onset CRC [EOCRC]) has substantially increased, and yet the etiology and molecular mechanisms underlying this alarming rise remain unclear. We compared tumor-associated T-cell repertoires between EOCRC and average-onset CRC (AOCRC) to uncover potentially unique immune microenvironment-related features by age of onset. Our discovery cohort included 242 patients who underwent surgical resection at Cleveland Clinic from 2000 to 2020. EOCRC was defined as younger than age 50 years at diagnosis (N = 126) and AOCRC as 60 years of age or older (N = 116). T-cell receptor (TCR) abundance and clonality were measured by immunosequencing of tumors. Logistic regression models were used to evaluate the associations between TCR repertoire features and age of onset, adjusting for sex, race, tumor location, and stage. Findings were replicated in 152 EOCRC and 1984 AOCRC cases from the Molecular Epidemiology of Colorectal Cancer Study. EOCRC tumors had significantly higher TCR diversity compared with AOCRC tumors in the discovery cohort (odds ratio [OR] = 0.44, 95% confidence interval [CI] = 0.32 to 0.61, P < .0001). This association was also observed in the replication cohort (OR = 0.74, 95% CI = 0.62 to 0.89, P = .0013). No significant differences in TCR abundance were observed between EOCRC and AOCRC in either cohort. Higher TCR diversity, suggesting a more diverse intratumoral T-cell response, is more frequently observed in EOCRC than AOCRC. Further studies are warranted to investigate the role of T-cell diversity and the adaptive immune response more broadly in the etiology and outcomes of EOCRC.

EOCRC tumors exhibit greater T-cell receptor diversity relative than AOCRC. This suggests a more diverse adaptive immune response, which may limit cancer interception and contribute to more aggressive disease.

With advances in screening and therapies, colorectal cancer (CRC) incidence has steadily declined over the past few decades. However, despite the overall decrease, the incidence of early-onset CRC (EOCRC; defined as diagnosis at age <50 years) has increased by 51% since 1994 in the United States (1) and is expected to continue to rise significantly (2-4). CRC has become the number 1 cause of cancer-related death in men aged 20-49 years (5). It is estimated that approximately 11% of colon cancers and 23% of rectal cancers will occur in adults younger than age 50 years by 2030 (4,6,7). Although a higher proportion of EOCRC is attributable to inherited cancer syndromes than average-onset CRC (AOCRC; ≥60 years), the majority of cases are sporadic. As there have not been substantial changes in the frequency of germline genetic mutations or cancer syndromes, it is likely that environmental or lifestyle factors, such as Western diet, physical activity, obesity, the gut microbiome, antibiotic use, or novel factors during early life and adolescence, are the driving force behind the rising incidence of EOCRC (8-11). However, the underlying multifactorial causes are not well understood.

Recent studies suggest that EOCRC may have unique clinical, pathological, and molecular features relative to AOCRC (12-17). A relevant feature that has been underexplored is host immunity (18-20). A strong tumor-infiltrating lymphocyte (TIL) response is a well-known prognostic indicator for better disease-specific survival (21-24) and potentially a predictive factor for responses to checkpoint inhibitor immunotherapy in CRC (25-27). Studies by our group and others also support the beneficial prognostic role of TILs independent of microsatellite instability (MSI) status (24,28-32), even though microsatellite instable (MSI-high) tumors are more likely to have high TILs than microsatellite stable (MSS) tumors (33). Aside from MSI, the factors that contribute to adaptive immune response heterogeneity across CRCs, even those diagnosed at the same stage, remain largely uncharacterized. Few prior studies have examined the immune landscape of EOCRCs, and they have been limited by small sample sizes and a narrow focus on immune-related gene expression (18-20). However, these studies provide evidence for distinct immune signatures of EOCRC by anatomic site and suggest that further studies are warranted to elucidate the roles of innate and adaptive immune surveillance in the etiology and outcomes of EOCRC.

T-cell clonality is a metric that summarizes the diversity of T-cell receptors (TCRs) present in a tumor sample. Specifically, T-cell clonality denotes the expansion of T cells with a specific TCR that recognizes a particular antigen. High clonality indicates a T-cell population targeted toward a specific antigen, resulting in a more homogeneous immune response. Conversely, lower clonality suggests a more diverse T-cell population with a broader repertoire of TCRs capable of recognizing a wider range of antigens but with less specificity toward any single antigen. For example, studies have shown that increased T-cell clonality within a tumor may be associated with a better response to cancer immunotherapy, as the immune system prioritizes targeting tumor-specific antigens (34-36). A recent review article has also summarized that higher clonality is associated with a worse prognosis in nasopharyngeal cancer, gastric cancer, and lymphoma (37). Because of the potential differences in immune features and poor prognosis in EOCRC patients (3), we hypothesized that EOCRCs demonstrate lower TCR abundance and greater T-cell diversity than AOCRCs, indicating more diverse T-cell responses targeted against specific tumor neoantigens. This would potentially suggest a less robust immune response, which may permit a precancerous lesion to develop into cancer earlier, and more aggressive disease progression among EOCRC patients. Using quantitative measures of TCR repertoires, we explored the contribution of intratumoral T-cell responses to earlier development of CRC and validated the results in a large population-based study.

Methods and materials

Discovery population

The discovery cohort included 242 patients with histologically confirmed stage I-IV CRC who underwent surgical resection at Cleveland Clinic from 2000 to 2020 and who consented to the Colorectal Cancer Tumor Biobank. To provide more separation in our comparison groups, EOCRC was defined as younger than 50 years of age at diagnosis (N = 126), and AOCRC as 60 years of age or older (N = 116). Clinical data including age at diagnosis, sex, race, ethnicity, body mass index, clinical and pathological stage, grade, anatomic site, and MSI status were abstracted from electronic medical records. The study received approval from the Cleveland Clinic Institutional Review Board under protocol number 4134. DNA was extracted from frozen tumors using standard protocols. TCR beta-chain repertoires were characterized using immunosequencing (survey-level immunoSEQ, assay version 4; Adaptive Biotechnologies, Seattle, WA) in the Center for Immunotherapy and Precision Immuno-Oncology Discovery Lab at Cleveland Clinic. Briefly, a multiplex PCR system is used to amplify the sequences of the TRB gene (TCR beta locus; https://www.genenames.org/data/gene-symbol-report/#!/hgnc_id/HGNC : 12155) in the third complementarity-determining region (CDR3β) from DNA samples (38) and then sequenced using an Illumina platform. Sequencing results are normalized on the basis of an assay targeting housekeeping genes with the same length amplicons. Two TCR repertoire metrics estimated from this assay were used in subsequent analyses: (1) TCR abundance (fraction_productive_of_cells from Adaptive), a measure of the normalized number of rearranged T-cell quantity relative to the total number of all nucleated cells in a tumor sample (ranging from 0 to 1); and (2) Simpson clonality (productive_simpson_clonality from Adaptive), a quantitative measure of the TCR repertoire diversity (ranging from 0 to 1), where a clonality equals 1 represents a repertoire dominated by a single clone. Because of the right-skewed distributions of the raw data, we performed log transformation of clonality and abundance data after excluding outliers (where total T-cell or sample cell counts were outside 3 standard deviations of the mean) for both the discovery and replication study separately.

Replication population

The Molecular Epidemiology of Colorectal Cancer Study (MECC) is a population-based study of incident CRC cases and healthy controls recruited in northern Israel from 1998 through 2017. Cases included patients with pathologically confirmed invasive colorectal adenocarcinomas. Controls are participants without a prior history of CRC selected from the same source population as cases and with individual matching on age, sex, Jewish ethnicity, and primary clinic site. Patients provided written informed consent, and the MECC study protocol was approved by the City of Hope Institutional Review Board under protocol number 19404. Baseline demographic and clinical characteristics of the CRC cases contributing to this study are described elsewhere (39,40). Out of 6006 CRC cases, 3865 cases that underwent pathology review had sufficient macrodissected tissue and extracted DNA available for immunosequencing. TCR repertoires were evaluated using survey-level immunoSEQ at the Adaptive Biotechnologies service facility (Version 2 and Version 4). Quality control procedures similar to the discovery cohort were implemented, and details have been described elsewhere (41). In brief, TCR abundance and Simpson clonality were log-transformed because of the skewed distribution of data. Log-transformed Simpson clonality (assay version-specific) was then z-score transformed before combining data across assay versions. A total of 2750 CRC cases had immunoSEQ data that passed quality control, but 614 CRC cases diagnosed between ages 50 and 60 were excluded to mirror the discovery inclusion criteria. In total, 152 EOCRC and 1984 AOCRC cases were included in the final analyses.

Investigation of clonality in paired tumor and normal tissue samples

To explore the possibility that a clonality-age association may not be tumor specific and that observed associations derive from highly expanded peripheral clones that may also be present in the tumor, we used 100 paired colorectal normal and tumor tissues from the Colonomics (CLX) study (https://www.colonomics.org) (42). The CLX study was approved by the Bellvitge University Hospital Ethics Committee (Protocol Number PR112/15). Out of 100 patients, 2 patients were diagnosed younger than age 50, 10 patients were diagnosed between ages 50 and 60, and the remaining 88 patients were diagnosed older than age 60. For consistency with inclusion criteria for the discovery cohort, only 90 patients (2 EOCRC and 88 AOCRC) were included in the analysis. We examined the correlation of clonality data, again generated with the survey-level immunoSEQ assay (43), between paired normal epithelium and tumor tissues stratified by age.

Statistical analyses

Descriptive statistics and univariate analyses using χ² and t tests were conducted for demographic and clinical characteristics between EOCRC and AOCRC groups. The relationship between TCR features and clinical and demographic characteristics was evaluated using an analysis of variance. Clinical and demographic factors that are both associated with age of onset and T-cell features were adjusted in the multivariate models. Logistic regression models were used to evaluate the associations between TCR repertoire features and age of onset (using AOCRC as the reference group), adjusting for sex, race/ethnicity, tumor location, and stage. To investigate if the association between T-cell feature and age of onset is independent of microsatellite status in tumors, we performed multivariate analyses by (1) incorporating MSI status (microsatellite stable, microsatellite instable, and missing) into the model as a confounder, and (2) restricting to MSS tumors as an effect modifier. For paired-sample analyses, we calculated the correlation coefficient between age of onset and clonality of normal and tumor tissues. Comparisons of clonality in normal and tumor tissues were examined using analysis of variance and F statistics. All analyses were performed using SAS 9.4 (SAS Institute, NC). All tests of statistical significance were two-sided.

Results

Demographic and tumor characteristics

Clinical and demographic characteristics of the discovery and replication cohorts are described in Table 1. In the discovery cohort, the average age of onset was 41 years for EOCRC and 73 years for AOCRC. A similar age distribution was observed for the replication cohort with averages of 42 years for EOCRC and 74 years for AOCRC. There was no significant difference (P = .3231) in the gender distribution between EOCRC and AOCRC in the discovery cohort. In the replication cohort, female cases were overrepresented among EOCRC cases (60.53%) when compared with the AOCRC group (46.12%, P = .0006). In line with previous studies, EOCRC patients were more likely to have left-sided tumors in both the discovery and replication cohorts (74.60% and 70.37%, respectively). Sixty-five percent of EOCRC cases were diagnosed at stage III and IV, compared with 51% of AOCRC in the discovery cohort (P = .045). Similarly, EOCRC cases were more likely to be diagnosed at advanced stages than AOCRC in the replication cohort (65% vs 31%, P < .0001). Approximately 11% of patients with MSI data available in the discovery cohort had MSI-H tumors; however, the majority (>50%) of AOCRC patients had unknown MSI status vs 13% among EOCRC patients with unknown MSI status. In the replication cohort, EOCRC had a slightly higher proportion with MSI-H phenotype when compared with that of AOCRC (17.76% vs 12.75%).

Table 1.

Demographic and clinical characteristics of study participants

	CCF Discovery Cohort (N = 242)			MECC Replication Cohort (N = 2136)
Characteristics	EOCRC (N = 126)	AOCRC (N = 116)	P		EOCRC (N = 152)	AOCRC (N = 1984)	P
Age at diagnosis (mean ± SD)	41.35 ± 6.57	73.2 ± 7.8			41.91 ± 6.94	74.20 ± 7.81
Sex
Male	67 (53.17%)	69 (59.48%)	.3231		60 (39.47%)	1069 (53.88%)	.0006**
Female	59 (46.83%)	47 (40.52%)			92 (60.53%)	915 (46.12%)
Race				Ethnicity
White	109 (86.51%)	95 (81.90%)		Jewish	85 (55.92%)	1732 (87.30%)
Non-White	17 (13.49%)	21 (18.10%)	.3246	Non-Jewish	67 (44.08%)	252 (12.70%)	<.0001**
Tumor location
Left	94 (74.60%)	65 (56.03%)			107 (70.37%)	1246 (62.80%)
Right	32 (25.40%)	51 (43.97%)	.0024**		45 (29.61%)	738 (37.20%)	.0612
Clinical stage a
I	13 (10.32%)	12 (10.34%)			13 (8.55%)	374 (18.85%)
II	31 (24.60%)	46 (39.66%)			74 (48.68%)	992 (50.0%)
III	47 (37.30%)	39 (33.62%)			29 (19.08%)	407 (20.51%)
IV	35 (27.78%)	19 (16.38%)	.045*		36 (23.68%)	211 (10.64%)	<.0001**
Differentiation
Well/Moderate	105 (88.24%)	86 (74.78%)			137 (90.13%)	1821 (92.16%)
Poor	14 (11.76%)	29 (25.22%)	.0079*		15 (9.87%)	155 (7.84%)	.4954
Microsatellite instability
MSI-High	14 (11.11%)	13 (11.21%)			27 (17.76%)	253 (12.75%)
MSS	96 (76.19%)	42 (36.20%)			118 (77.63%)	1589 (80.09%)
Missing	16 (12.70%)	61 (52.59%)	.0742		7 (4.61%)	142 (7.16%)	.1035

^aA total of 187 MECC patients with missing clinical stage information was excluded from the analysis. CCF = Cleveland Clinic; MECC = Molecular Epidemiology of Colorectal Cancer Study.

^* P < .05,

^** P < .005.

In the discovery cohort, TCR clonality was associated with tumor location (F = 9.677, P = .0021), grade (F = 13.876, P = .0002), and MSI status (F = 9.489, P = .0024), whereas TCR abundance was statistically significantly related to tumor location (F = 4.11, P = .0438), grade (F = 12.59, P = .002), and stage (F = 2.81, P = .04). Therefore, in the multivariate model, covariates such as sex, race, tumor location, and clinical stage were included. The same covariates, substituting Jewish ethnicity for race, were carried forward to the replication analysis for model consistency. The distributions of TCR clonality and abundance by tumor stage in the discovery and replication cohorts were included in Supplementary Figure 1 (available online).

Univariate association analysis: TCR repertoire features, age of CRC onset, and other covariates

EOCRC tumors had a nonstatistically significant lower TCR abundance (F = 3.40, P = .066) and a significantly lower TCR clonality (F = 35.86, P < .001) in the discovery cohort (Supplementary Figure 2, A and B, available online). In the MECC cohort, no difference in TCR abundance was found between EOCRC and AOCRC (F = 0.56, P = .4530), whereas EOCRC tumors presented with a statistically significantly lower TCR clonality when compared with AOCRC tumors (F = 9.29, P = .0023; Supplementary Figure 2, C and D, available online). Relationships between TCR features and other covariates including sex, race, tumor location, clinical stage, and MSI status are summarized in Supplementary Table 1 (available online).

Multivariate association analysis: TCR repertoire features and age of CRC onset

In both the discovery and replication cohorts, we found no statistically significant differences in TCR abundance between EOCRC and AOCRC tumors, after adjusting for sex, race/ethnicity, tumor location, and stage (Figure 2, A and C). In the discovery cohort, EOCRC tumors had significantly lower TCR clonality (higher T-cell diversity) compared with AOCRC tumors in a multivariable model adjusting for sex, race, tumor location, and stage (odds ratio [OR] = 0.38, 95% confidence interval [CI] = 0.25-0.56, P < .0001; Figure 2, B). Similarly, in the replication cohort, EOCRC was associated with lower TCR clonality (OR = 0.74, 95% CI = 0.62 to 0.89, P = .0013; Figure 2, D).

Figure 2.

Associations between age of onset (EOCRC vs AOCRC; AOCRC as reference) and TCR features. A) TCR abundance in CCF Discovery Cohort. B) TCR clonality in CCF Discovery Cohort. C) TCR abundance in MECC Replication Cohort. D) TCR clonality in MECC Replication Cohort.

Given the established differences in tumor molecular features between MSI-H and MSS tumors, we evaluated whether the observed association between lower clonality and younger age of onset was affected by MSI status. Adjustment for MSI status did not substantially change the association in either the discovery or replication cohort (OR for discovery cohort = 0.38, 95% CI = 0.25 to 0.56, P = .0023; OR for replication cohort = 0.73, 95% CI = 0.61 to 0.88, P = .0012; Supplementary Table 2, available online). Further, we examined the association between TCR clonality and age of onset restricted to the MSS tumors. Clonality remained statistically significantly associated with younger age of onset in the 138 patients with MSS tumors in the discovery cohort (OR = 0.41, 95% CI = 0.24 to 0.72, P = .0017; Supplementary Table 3, available online) and in the replication cohort (OR = 0.70, 95% CI = 0.65 to 0.97, P = .0257; Supplementary Table 3, available online).

Analysis of paired tumor and normal tissues

To further investigate whether the association between clonality and age of onset is a tumor-specific phenomenon that is not driven by highly expanded peripheral clones that accumulate with aging, we compared the TCR clonality from 90 paired colorectal normal and tumor tissues from the CLX study. There was no statistically significant difference between normal and tumor tissue TCR clonality among all samples (F = 3.58, P = .0603). We found a marginally statistically significant association between age of onset and TCR clonality in tumor tissue (correlation coefficient = 0.222, P = .0411; Figure 3, A). However, there was no correlation between TCR clonality and age of onset among normal tissues (correlation coefficient = 0.011, P = .9217; Figure 3, B). Thus, we found evidence supporting that the observed association between age of onset and TCR clonality is a tumor-specific phenomenon, which is not correlated with the peripheral repertoire clonality.

Figure 3.

Correlation of T-cell receptor (TCR) clonality and age of colorectal cancer onset in normal colon and tumor tissues from the Colonomics study. A) Correlation between age and clonality in tumor tissues. B) Correlation between age and clonality in normal colon tissues.

Discussion

Using quantitative T-cell diversity and abundance measures, we demonstrated in a single-institution cohort that higher TCR repertoire diversity, indicating a more diverse intratumoral T-cell response, is more commonly observed in EOCRC than AOCRC. This observation was replicated in a large independent sample from a population-based study.

Several studies have shown that EOCRC tumors may be clinically and pathologically distinct from AOCRC tumors. EOCRC tumors are more likely to occur in the distal colon and rectum, be poorly differentiated, and present at more advanced stages (8-15). However, the differences in clinical presentation between EOCRC and AOCRC are not fully explained by genomic differences.(16) There may be other biological elements, such as differences in antitumor immunity, that contribute to earlier age of presentation or potentially more aggressive features. Tumor-infiltrating lymphocytes have been associated with CRC progression, prognosis, and treatment responses (21-24,26,29-32). Therefore, it is possible that depressed immune surveillance in the gut may contribute to the accelerated etiology of EOCRC and adverse outcomes in a subset of cases. Altered host immune responses could play a contributory role in the more rapid development of disease, even if this modulation is driven by chronic inflammatory processes resulting from lifestyle factors or environmental exposures.

Recent research has focused on the differences in the tumor microenvironment between EOCRC and AOCRC (44-46). A study with 40 EOCRC (<50 years) and 39 AOCRC (>65 years) patients examined the mRNA expression of immune-related genes and found a distinct immune response signature in EOCRC (45). Using NanoString, they identified that expression of the complement genes C7 and CFD and the inflammatory gene SAA1 were associated with EOCRC and shaped a unique microenvironment for EOCRC. Another study including 20 EOCRC (<45 years) and 20 matched AOCRC (70-75 years) patients found no statistically significant difference in the infiltration of total T cells, conventional CD4⁺ and CD8⁺ T cells, and regulatory T cells using a multiplex immunofluorescence assay (44). They also found no significant difference was detected in a global analysis of 770 tumor immune genes using a NanoString Immune Profiling panel. Further, Ugai et al. conducted a study to examine the differential heterogeneity in immune cell infiltrates between incident tumors from 35 early-onset (<50 years), 73 intermediate-onset (50-54 years), and 1410 average-onset (>55 years) CRC cases from two prospective cohorts (Nurses’ Health Study and Health Professionals Follow-up Study) (46). Using multiplex immunofluorescence assays, digital imaging analyses, and machine learning algorithms, they found that EOCRC tumors were likely to exhibit lower levels of tumor-infiltrating lymphocytes (P = .013) and peritumoral lymphocytic reaction (P = .025). Intermediate-onset CRC tended to have lower densities of macrophages, mature macrophages, and regulatory T cells when compared with later-onset CRC. Their findings provide evidence supporting possible differences in the tumor immune microenvironment between EOCRC and AOCRC patients, as also observed in our study. However, with small samples, different age cutoffs for EOCRC, and variable technologies to investigate the tumor immune features, it is challenging to directly compare findings on differential tumor microenvironments in EOCRC and AOCRC.

Our study focused on the quantitative measures on TCR abundance and clonality through DNA immunosequencing to evaluate the heterogeneity in microenvironment cell responses between YOCRC and AOCRC. With large sample sizes in both the discovery and replication cohorts, we demonstrated that when compared with AOCRC tumors, EOCRC patients manifest greater TCR diversity within tumors, indicating more diverse intratumoral T-cell responses against tumor antigens (Figure 1). TCR clonality between EOCRC and AOCRC remained statistically significantly different in both cohorts after adjusting for confounding factors, including stage at diagnosis and MSI status. Our findings provide evidence that the tumor immune microenvironment, particularly TCR diversity, is distinctive between EOCRC and AOCRC.

Figure 1.

Conceptual framework for the relationship between T-cell repertoire and colorectal cancer age of onset.

Other studies have shown that peripheral T-cell repertoire clonality increases with age (47,48), with higher clonality often observed in the peripheral T-cell repertoire among older individuals (>60 years). Thus, we sought to examine whether our findings were tumor specific or potentially influenced by this well-characterized peripheral T-cell phenomenon. Using normal tissues as a proxy for peripheral blood, we compared the T-cell profiles of paired tumor and adjacent normal tissues from the CLX study. We found no statistically significant correlation of TCR clonality between tumor and normal tissue pairs. Further, there was no correlation between clonality and age in normal tissues. Older patients tended to have a slightly higher clonality in tumor tissues, which is in line with our discovery and replication findings. This indicates that our observed association between age of onset and TCR clonality is likely to be tumor specific and unlikely to be driven by peripheral T-cell repertoire expansion with age.

We acknowledge a few key limitations of this study. First, a majority of our study subjects were of European ancestry; hence, the findings cannot be generalized to other groups. Further validation studies including patients from diverse ancestral backgrounds are needed. Second, although tumor samples from both the discovery and replication cohorts underwent analysis using the same immunosequencing technology, there is the potential for measurement errors or batch effects. Third, our study focused only on the diversity and abundance of T cells, not on other innate or adaptive immune cells or T-cell subtypes. Studies with comprehensive features and subtypes of immune cells are warranted. We acknowledge the high missing rate of MSI status, particularly among the older CRC patients in the discovery cohort. Because testing for MSI status was not a routine practice for CRC patients during the earlier recruitment period, MSI information was unavailable. Another limitation of our study is the lack of data on infectious exposure. Given that cytomegalovirus (CMV) significantly influences TCR diversity, it would be valuable to include the infection status as a covariate in our models. Unfortunately, we do not have access to CMV or other infection history data for both discovery and replication cohorts, as these exposures were not collected within any of the epidemiologic instruments used. Finally, it is important to note that our approach does not directly investigate the causal relationship between TCR clonality and the timing of CRC onset. By categorizing CRC patients into EOCRC and AOCRC using age at diagnosis, we found that EOCRC patients exhibit lower TCR clonality in their tumors when compared with AOCRC patients. Rather than establishing a direct link between TCR clonality and the onset of CRC, our findings highlight the distinct differences in TCR clonality characteristics between EOCRC and AOCRC. Our analysis in the CLX study attempts to clarify the tumor-specific nature of TCR clonality differences by age of onset, but the possibility that this association is driven by broader clonality expansion with age cannot be definitively ruled out without a long-term prospective study of healthy young adults.

Our study is the first to systematically investigate TCR repertoire using quantitative measurements between EOCRC and AOCRC. Using a robust discovery-replication design, we demonstrated that EOCRC tumors exhibit higher TCR diversity than AOCRC tumors, even after controlling for tumor stage, primary tumor location, and MSI status. Greater TCR diversity implies a broader range of T-cell responses against tumor antigens, which may influence disease prognosis and response to therapy. Further research investigating the T-cell subsets that are specific to EOCRC tumors could shed light on early or accelerated tumor development in young adults.

Data availability

Individual-level TCR sequencing data for the MECC and CLX studies will be deposited to immuneACCESS, a repository of publicly available TCR and BCR (B-cell receptor) sequences (https://clients.adaptivebiotech.com/immuneaccess). Deidentified demographic, clinical, and genetic data for MECC subjects are available through dbGaP (phs001045.v1.p1; phs001856.v1.p1; phs001903.v1.p1). CLX clinical metadata can be found at https://www.colonomics.org/data-browser/. For the CCF discovery cohort, regulatory limitations due to restrictive informed consent language prevent the public sharing of clinical and TCR sequencing data.

Author contributions

Yayu Tsai, MS, PhD (Conceptualization; Data curation; Formal analysis; Investigation; Validation; Writing—original draft; Writing—review & editing), Alok A. Khorana, MD (Conceptualization; Data curation; Resources; Supervision; Writing—review & editing), Stephen B. Gruber, MD, PhD (Conceptualization; Data curation; Funding acquisition; Investigation; Methodology; Project administration; Resources; Supervision; Validation; Writing—review & editing), Victor Moreno, PhD (Data curation; Formal analysis; Funding acquisition; Methodology; Project administration; Resources; Supervision; Validation; Writing—review & editing), Gad Rennert, MD (Data curation; Funding acquisition; Methodology; Project administration; Resources; Supervision; Validation), Hedy S. Rennert, MD (Data curation; Project administration; Resources; Supervision; Validation), Joel K. Greenson, MD (Data curation; Project administration; Resources), Rebeca Sanz-Pamplona, PhD (Data curation; Formal analysis; Writing—review & editing), Gregory E. Idos, MD, MS (Project administration), Kevin J. McDonnell, MD (Project administration), David Liska, MD (Conceptualization; Data curation; Funding acquisition; Resources; Supervision), Joseph D. Bonner, PhD (Data curation; Formal analysis), Timothy A. Chan, MD, PhD (Conceptualization; Resources; Supervision), Nicole Osborne, MS (Data curation), Raghvendra M. Srivastava, PhD (Data curation), Christopher P. Walker, BS (Data curation; Project administration), Marilena Melas, PhD (Data curation), Suneel Kamath, MD (Data curation; Project administration; Writing—review & editing), Shao Xiang, MD (Data curation; Project administration), Shimoli V. Barot, MD (Data curation; Project administration), Kanika G. Nair, MD (Data curation; Project administration), Jonathan Mitchem, MD (Conceptualization; Writing—review & editing), Stephanie L. Schmit, PhD, MPH (Conceptualization; Data curation; Formal analysis; Funding acquisition; Investigation; Project administration; Resources; Supervision; Writing—original draft; Writing—review & editing).

Funding

AAK was supported by Sondra and Stephen Hardis Chair in Oncology Research. The MECC study was supported in part by R01 CA197350, R01 CA081488, R01 CA248931, P30 CA014089, and a generous gift from Daniel and MaryAnn Fong. VM had the support of the Agency for Management of University and Research Grants (AGAUR) of the Catalan Government grant 2017SGR723, the Instituto de Salud Carlos III, co-funded by FEDER funds –a way to build Europe– grants PI17-00092 and the Spanish Association Against Cancer (AECC) Scientific Foundation grant GCTRA18022MORE.

Conflicts of interest

GEI: receives/received research funding from Myriad Genetics and Laboratories. JKG: consultant for Guardant Health. VM: owns stock in Aniling. SBG: co-founder of Brogent International LLC with equity. SK: consulting and advisory role in Exelixis, Tempus, Guardant Health, and SeaGen. AAK: consulting honoraria from Janssen, Bayer, BMS, Sanofi, Pfizer, and Anthos. DL: consulting and advisory roles in Olympus Medical Systems and research funding from Merck. Others: none.