Laura Andersen

Aging is accompanied by a progressive decline in immune function, increasing susceptibility to infections, cancer, and poorer treatment outcomes. The diversity of the T cell receptor (TCR) repertoire, reflecting the capacity to recognize diverse antigens, is known to decline with age. However, how intrinsic and extrinsic factors jointly shape TCR diversity and its rate of decline, as well as the extent to which this diversity influences clinical outcomes in cancer, remain poorly characterized. Here, we integrate TCR sequencing data from 1.32 billion unique T cell clones, derived from 798 patients with cancer and 5052 individuals without cancer across 14 independent cohorts, to systematically characterize the effects of age, sex, and chronic cytomegalovirus (CMV) infection on peripheral blood repertoire diversity. We identify a nonlinear trajectory of diversity loss that accelerates after midlife, with a marked increase in the rate of decline beyond approximately 50 years of age, which is further modified by sex and CMV infection. In patients with cancer, higher peripheral TCR diversity was associated with improved overall survival across multiple cancer types, even when accounting for the impact of age, sex, and CMV on diversity. Together, these findings underline peripheral TCR repertoire diversity as a dynamic marker of immune aging shaped by demographic and chronic viral factors.

Supplementary Figure 1 | Cell type contribution to compartments grouped by tissue. Cell type ranks of cell types from the compartments endothelial, epithelial, germline, and stromal grouped by tissue in healthy individuals (n=30). This is defined from the Tabula Sapiens definition of cell types, which can originate from more than one tissue.

Supplementary Figure 2 | ctDNA-based analysis of the cancer cohorts. A. ctDNA detection status in the colorectal cancer (CRC) and muscle invasive bladder cancer (MIBC) cohorts. ND = not detected, D = detected, NA = not available. Number and % of patients in each group are indicated. B. Plasma ctDNA tumor fractions (%) in the ctDNA-positive subgroup of the CRC (red) and MIBC (yellow) cohorts. Dashed line indicates ctDNA TF = 1%.

Supplementary Figure 5 | Longitudinal dynamics of cfDNA-derived immune cell type ranks for individual patients. A. Longitudinal trajectories of classical monocyte ranks in patients with colorectal cancer (CRC) across treatment. Grey lines represent individual patients, the red line highlights a selected patient. Blue bar indicates the pre-treatment median ± interquartile range (IQR) or zero when IQR < 0. B. Same as A, for patients with muscle-invasive bladder cancer (MIBC). C. Same as A, showing CD4+, α-β T cells ranks in CRC. D. Same as C, for MIBC. ACT = adjuvant chemotherapy. NAC = neoadjuvant chemotherapy. pre-treat = pre-treatment. pre-op = pre-operatively. pre-RC = pre-radical cystectomy. post-op = post-operatively. y = year.

Supplementary Figure 4 | Comparison of immune model detection of cancer to ctDNA-based detection. A. Immune model prediction probability of patient relapse in colorectal cancer (CRC) compared across healthy individuals and ctDNA tumor fraction (TF) groups: not detected (ND), TF < 1%, TF > 1%. P-values were calculated using two-sided Wilcoxon rank-sum tests. B. Same as A for muscle invasive bladder cancer (MIBC). C. Comparison of ctDNA-based cancer detection and immune model detection in CRC. The number of patients detected (red) and not detected (grey) with the immune model among ctDNA-positive and ctDNA-negative patients are indicated. P-values calculated using Fisher's exact test. D. Same as C for MIBC. E. Kaplan–Meier curves showing 3-year recurrence-free survival (RFS) in patients with CRC, stratified by immune model detection: yes (red), blue (no). F. Same as E for MIBC. G. Same as E, but patients are stratified into three groups: ctDNA-negative (ctDNA–; grey), ctDNA-negative/Immune score-positive (ctDNA–/Immune score+; blue), and ctDNA-positive (ctDNA+; red). Immune score positivity was defined using a 95% specificity threshold from the LASSO model described in Figure 2I. RFS was calculated from the date of surgery. Shaded areas represent 95% confidence intervals. H. Same as G, for patients with MIBC.

Supplementary Figure 3 | Comparison of cell type ranks to plasma ctDNA tumor fraction. A. Cell type ranks of monocytes and lymphocytes from spleen, lymph node, blood, and thymus compared to the plasma ctDNA tumor fraction (TF) in colorectal cancer (CRC). Patients are stratified by ctDNA status into: not detected (ND), TF < 0.1%, 0.1% < TF < 1%, and TF > 1%. P-values were calculated using two-sided Wilcoxon rank-sum tests. B. Same as A for muscle invasive bladder cancer (MIBC).

Cancer-associated shifts in immune cell contributions to cfDNA. A, Log2 FC (logFC) in cell type ranks grouped by category between healthy individuals (blue, n = 30) and patients with colorectal cancer (CRC; red, n = 124). Higher ranks reflect greater inferred cfDNA contribution. B, Comparison of median cell type ranks for monocytes (left) and lymphocytes (right) from the spleen, lymph node, blood, and thymus across ctDNA TF categories in CRC: ctDNA ND, ctDNA TF 1%. P values were calculated using two-sided Wilcoxon rank-sum tests. C, Correlation between cfDNA-derived ranks for effector CD4+ and CD8+ α–β T cells from lymph node and blood and T-cell fraction estimated using ImuneLENS (16) from germline WGS. Spearman’s correlation coefficient (r) and P value are shown. D, Same as in A, comparing healthy individuals (blue, n = 30) and patients with MIBC (yellow, n = 102). E, Same as B, for MIBC. F, Same as C, for MIBC. G, Volcano plot showing logFC in cfDNA-derived ranks vs. false discovery rate (FDR)-adjusted P values for individual immune cell types (monocytes, lymphocytes, and granulocytes) between healthy individuals (blue) and patients with CRC (red). Cell types with the highest FC are labeled. For cell types present in multiple tissues, only the instance with the highest log FC is shown. P values were calculated using two-sided Wilcoxon rank-sum tests and adjusted for multiple testing using the FDR method. H, Same as G, comparing healthy individuals and patients with MIBC (yellow). I, Receiver operating characteristic (ROC) curves for LASSO-regularized logistic regression models classifying CRC (red) and MIBC (yellow) vs. healthy controls. Models were trained on cell type ranks from monocytes, lymphocytes, and granulocytes, and evaluated using LOOCV. Area under the curve (AUC) and 95% CIs are indicated. Eff., effector; NS, not significant; Prog., progenitor; thy.-der., thymus-derived.

CfDNA coverage at TSSs enables deconvolution of blood cell type contributions. A, Schematic overview of the analytic approach. Plasma cfDNA is subjected to WGS at a target depth of 20–30×. GC-corrected coverage at TSS of protein-coding genes is computed using Griffin (15). TSS coverage is then correlated with gene expression profiles from the Tabula Sapiens single-cell reference dataset (14) to generate cell type ranks, reflecting relative contributions to the cfDNA pool. B, Mean cfDNA coverage in healthy individuals across ±1,000 base pairs surrounding the TSS of the top 2,000 highly expressed (red) and unexpressed (blue) genes in blood. Expression categories were defined using the GTEx dataset, with highly expressed genes ranked by TPM and unexpressed genes defined as those with TPM = 0. C, Distribution of median TSS coverage for all protein-coding genes in healthy individuals, overlaid with gene expression categories from GTEx: unexpressed (blue), expressed (red), and all other genes (gray). D, Inferred cell type ranks in healthy individuals grouped by cell type categories in healthy controls. Higher ranks indicate greater relative cfDNA contribution. Prog., progenitor. E, Same as D, grouped by cell type compartment defined by Tabula Sapiens. F, Cell types contributing to the Immune compartment grouped by tissue. G, Cell type ranks for individual immune cell types. CRC, colorectal cancer; eff., effector; NK, natural killer; Sec., secreting; thy.-der., thymus-derived.

Genome-wide coverage patterns of plasma cell-free DNA (cfDNA) fragments reflect nucleosome positioning in the cells of origin, enabling noninvasive inference of cell type contributions and transcriptional activity. Whereas the majority of cfDNA originates from hematopoietic cells, the diagnostic and biological relevance of this fraction remains underexplored. In this study, we performed cfDNA-based deconvolution of blood cell types by integrating transcription start site (TSS) coverage profiles from plasma whole-genome sequencing with single-cell transcriptomic reference data. By correlating cfDNA TSS coverage with gene expression across 457 blood cell types, we ranked their relative contributions to the cfDNA pool. We analyzed 788 pretreatment and longitudinal plasma samples from patients with localized colorectal cancer, muscle-invasive bladder cancer (MIBC), as well as 30 samples from healthy controls. In healthy individuals, cfDNA TSS coverage profiles reflected blood gene expression, and the inferred cell type contributions recapitulated the known hematopoietic composition. In patients with cancer, we observed a significant increase in cfDNA contributions from lymphocytes, including T cells and plasma cells, and decreased contributions from monocytes and granulocytes. These immune-derived signatures distinguished patients with colorectal cancer (AUC = 0.793) and MIBC (AUC = 0.745) from healthy controls. Longitudinal analysis of immune cell type contributions revealed treatment-associated changes in the relative abundance of classic monocytes and plasma cells, although these temporal dynamics were not predictive of relapse or outcome. Together, these findings suggest that cfDNA-derived immune signatures may capture aspects of systemic immune remodeling in cancer, potentially providing a complementary noninvasive biomarker in liquid biopsies beyond tumor-derived signals.Significance:There is limited understanding of how plasma cfDNA reflects systemic immune changes in cancer. We show that cfDNA coverage profiles reflect blood cell gene expression and can be used to infer blood cell ranks that recapitulate known cell compositions. These cfDNA-derived ranks reveal cancer-associated immune alterations that distinguish patients with cancer from healthy controls. Our findings highlight cfDNA as a minimally invasive biomarker of systemic immune remodeling in cancer.

Clinical characteristics of the study cohorts.