Androgen conspires with the CD8+ T cell exhaustion program and contributes to sex bias in cancer

Abstract

Sex bias exists in the development and progression of nonreproductive organ cancers, but the underlying mechanisms are enigmatic. Studies so far have focused largely on sexual dimorphisms in cancer biology and socioeconomic factors. Here, we establish a role for CD8⁺ T cell-dependent antitumor immunity in mediating sex differences in tumor aggressiveness, which is driven by the gonadal androgen but not sex chromosomes. A male bias exists in the frequency of intratumoral antigen-experienced Tcf7/TCF1⁺ progenitor exhausted CD8⁺ T cells that are devoid of effector activity as a consequence of intrinsic androgen receptor (AR) function. Mechanistically, we identify a novel sex-specific regulon in progenitor exhausted CD8⁺ T cells and a pertinent contribution from AR as a direct transcriptional transactivator of Tcf7/TCF1. The T cell-intrinsic function of AR in promoting CD8⁺ T cell exhaustion in vivo was established using multiple approaches including loss-of-function studies with CD8-specific Ar knockout mice. Moreover, ablation of the androgen-AR axis rewires the tumor microenvironment to favor effector T cell differentiation and potentiates the efficacy of anti-PD-1 immune checkpoint blockade. Collectively, our findings highlight androgen-mediated promotion of CD8⁺ T cell dysfunction in cancer and imply broader opportunities for therapeutic development from understanding sex disparities in health and disease.

Fig 1.. CD8⁺ T cell immunity mediates sex differences in murine bladder cancer aggression.

(A) BBN-induced carcinogenesis model. Mice were exposed ad libitum to 0.1% BBN in drinking water for the first 14 weeks to induce bladder cancer formation and then monitored for a total of 302 days. Percent survival, genotype/sex of each experimental group, and p values from the log rank test are shown. N = 13-20 mice per group. (B) MB49 tumor growth in mice with indicated genotypes. (C) Antibody-mediated depletion of CD4⁺ and/or CD8⁺ cells in mice challenged with MB49. (D) RT-PCR for qualitative detection of Actb and Y chromosome-encoded Sry gene transcripts from MB49 cells. DNA extracted from tails of male and female mice are included as controls. (E) Diagram representation of Four Core Genotype (FCG) mouse model. BKL171 was generated from a BBN-induced bladder tumor of a XXM FCG mouse. (F) BKL171 tumor growth in mice with indicated genotypes after subcutaneous implantation. Mean tumor area (mm²) ± SEM are reported, with statistical significance determined using the repeated measures two-way ANOVA. n = 4-10 mice per group. M, male; F, female; Ab, antibody; ns, not significant. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001. p-values in A and F were corrected for multiple testing using the Bonferroni procedure.

Fig 2.. Superior effector function of tumor-infiltrating T cells in female mice underlies sex-biased cancer outcomes.

(A) Flow cytometric analysis of IFNγ, TNFα and GZMB expression in CD8⁺ T cells from the spleens and tumors (TILs) of male and female mice 9 days post subcutaneous MB49 challenge. Cells were stimulated ex vivo with 50 ng/mL PMA, 1 μg/mL Ionomycin and 1X Brefeldin A for 2 hours. Statistical significance was determined using Student’s t test. (B) Impact of donor and recipient sex on the adoptive transfer therapy of T cells. MB49 tumor growth in Tcrb/Tcrd KO mice post adoptive transfer on Day -1 with 5 x 10⁵ CD8⁺ T cells, which were prepared separately from the draining lymph nodes of Day 14 MB49 tumor in WT mice. Mean tumor area (mm²) ± SEM are reported, with statistical significance determined using the repeated measures two-way ANOVA. n = 4-10 mice per group. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001.

Fig 3.. Sex differences in CD8⁺ T cell immunity at single-cell resolution in the tumor microenvironment.

(A) Uniform Manifold Approximation and Projection (UMAP) of 26,698 CD8⁺ TIL (3 female mice – 9,955; 3 male mice – 16,743) scRNA-seq profiles, colored by cluster. Cells were sorted by FACS from Day 10 MB49 tumors. 9,955 cells are shown here for each sex for the purpose of easier visualization. Enclosed clusters 1, 2, 7 and 9 show male-biased frequencies as outlined in (B). (C) Summary of key CD8⁺ T cell genes that show sex-biased expression level in each cell cluster. Genes with Bonferroni-adjusted p-values ≤ 0.05 from the sex-based differential expression analyses using Wilcoxon rank-sum test were considered male (blue) or female (purple) biased depending on the directionality of their average log fold change. (D) Enrichment pattern of published PE and effector-like CD8⁺ T cell signatures (7). (E) Ontogeny inference. Slingshot algorithm was applied to order single cells in clusters 1, 2, 6, 7, 9 and 10 in pseudotime. Tcf7, Cd44, Gzmb, Prf1, Havcr2 and Pdcd1 expression levels in these clusters are indicated. (F) Quantification of pseudotime trajectories of male and female Tcf7⁺ CD8⁺ TILs from E. Statistical significance was determined using Wilcoxon rank-sum test. ****p ≤ 0.0001.

Fig 4.. Male-biased CD8⁺ progenitor exhausted T cell frequency in the tumor microenvironment.

(A) UMAP of CD8⁺ TILs from MB49 tumors as assessed by spectral flow cytometry at indicated time points, colored by cluster. Enclosed cluster 2 shows a male-biased frequency as indicated. (B) Indicated protein expression on UMAP. (C) Percentage of the CD8⁺ TILs at indicated or all combined time points using the PE score, which was generated by UCell (38) based on protein-level expression of (TCF1, SLAMF6, BCL2, CD44, CD69) or (CD62L, PD-1, LAG3, TIM3, CTLA4, TOX) that are positively and negatively associated with PE cells, respectively. (D) MB49 growth in Rag2 KO mice that were adoptively transferred with 1.75 x 10³ TIM3⁻SLAMF6⁺CD8⁺ T cells from Day 12 MB49 tumors of WT mice. Tumor weights from Day 14 are reported. (E) Flow cytometric analysis of TIM3 and TCF1 expression in donor TILs. Statistical significance for % PE cells was determined by Wilcoxon rank-sum test. Remaining statistical significance was determined by Student’s t test. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001. p-values in D and E were corrected for multiple testing using the Bonferroni procedure.

Fig 5.. T cell intrinsic AR signaling contributes to male bias in tumor growth and accumulation of CD8⁺ progenitor exhausted T cells.

MB49 growth and circulating free testosterone levels in (A) Four Core Genotype or (B) WT mice three weeks post sham or surgical castration (cast). (C) PCA of spectral flow cytometry data on CD8⁺ TILs and (D) percentage of CD8⁺ PE T cells from B. (E) MB49 growth in castrated WT mice after intraperitoneal injection with IgG control or anti-mouse CD8 antibodies. depl, Depletion. (F) Immunoblot of AR and β-actin in stimulated CD8⁺ T cells from Ar^fl/(fl) control or E8iCre-Ar^fl/(fl) mouse spleens. LNCaP lysates were used as the positive AR control. Rep, replicate. (G) MB49 growth in Ar^fl/(fl) or E8iCre-Ar^fl/(fl) mice. (H-I) Percentage of male PE (H) and terminally exhausted (I) CD8⁺ T cells from G. (J) MB49 growth in WT mice after daily treatment of vehicle or enzalutamide (enz) starting on Day 5. (K) MB49 growth in WT mice with or without castration (ADT, androgen deprivation therapy) and anti-PD-1 antibodies at the indicated time points. For tumor growths, mean area (mm²) ± SEM are reported, with statistical significance determined using the repeated measures two-way ANOVA. n = 6–11 mice per group. Remaining statistical significance was determined by Student’s t test. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 0.0001. p-values in A, B, D, E, G, J and K were corrected for multiple testing using the Bonferroni procedure.

Fig 6.. AR signaling in CD8⁺ progenitor exhausted T cells.

(A) Enrichment of androgen response (red) and type I IFN (blue) signatures in CD8⁺ TILs from Day 10 MB49 tumors. (B and C) Pseudotime analysis of lymphocytic choriomeningitis virus specific CD8⁺ T cells (35), as colored by the relative enrichment of androgen response signature. Boxplots indicate expression of androgen-responsive genes in effector-like (“Eff”) and progenitor exhausted (“Ex”) cells. (D) Log₂-transformed ratio of ATAC-seq signals for androgen-responsive genes in progenitor (SLAMF6⁺TIM3⁻) over terminally (SLAMF6⁻TIM3⁺) exhausted CD8⁺ T cells (7). (E) Multiplexed immunofluorescent analysis of nuclei, CD3, CD8, TIM3, TCF1 and AR in BBN-induced bladder tumors collected from male mice at the time of sacrifice for morbidity. Scale bar, 1000 μm (top left), 100 μm (top right), 15 μm (bottom). (F) Quantification of CD8⁺ T cells in normal bladder versus tumor (left) and frequency of AR-expressing CD8⁺ T cell in the indicated cell subsets (right). (G) qPCR analysis of FACS-sorted CD8⁺ T cell subsets (red box) from Day 14 MB49 tumors. Statistical significance was determined by Student’s t test. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001. p-values in F (right) were corrected for multiple testing using the Bonferroni procedure.

Fig 7.. AR modulation of Tcf7 and its transcriptional network.

(A) Workflow of Integrated Cell-type-specific Regulon Inference Server from single-cell RNA-Seq (IRIS3). See methods for details. (B) Sex differences in Tcf7 centered regulon in the CD8⁺ PE TILs. Enclosed transcription factors have AREs within their respective promoters. (C) Correlation in expression between indicated transcription factors and Tcf7 in male-biased scRNA-seq clusters. (D) qPCR analysis of indicated genes in effector memory (“EFF”), central memory (“CM”) or naïve CD8⁺ T cells. Naïve CD8⁺ T cells were stimulated with testosterone, type I IFN or both for 6 hours. Blue, Male; Pink, Female. (E) Luciferase reporter assay. Renilla and Red Firefly luciferase activity, the latter of which was regulated by a 1 kb long WT or mutant Tcf7 promoter sequence lacking putative AREs (fig. S6A), were simultaneously measured in control or EGFP-C1-AR overexpressing HEK293FT post 24h stimulation with 500 ng/mL testosterone (T) or 50 nM DHT. (F) AR ChIP-qPCR of EV- and AR-Jurkats post 24h of 100 nM DHT stimulation compared to vehicle control. (G) Schematic representation of sex differences CD8⁺ T cell fate in the tumor microenvironment. Statistical significance was determined by Student’s t test. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001.