The identification of tumor antigen-specific T cells in cancer patients has important immunotherapeutic implications. In this study, we analyzed more than 12,000 tumor antigen-specific CD8+ T cells harvested from mouse models of TNBC and showed that CD39+CD8+ T cells represent tumor antigen-specific CD8+ T cells. We then demonstrated that significantly improved clinical outcomes were more likely for cases of TNBC infiltrated by high frequencies of CD39+CD8+ T cells. Improved OS also correlated with the frequency of CD39+CD68+ macrophages but not that of CD39+FOXP3+ Tregs. Furthermore, CD39+CD8+ T cell scoring in a small TNBC cohort receiving ICI therapy demonstrated the exceptional significance of this cell population in terms of response to ICI therapy.
Tumor mutation burden is higher in TNBC than in other BC subtypes, and the level of lymphocyte infiltration is greater. Furthermore, genomic instability produces high numbers of somatic mutations and causes the frequent occurrence of neoantigens34. Unlike the specific KRAS mutations identified in NSCLC or CRC, tumor-specific T cell detection in limited TNBC samples is hindered by the varied and unique autologous cellular antigens in TNBC patients35,36. To overcome these restrictions, we used the well-established 4T1 and 4T1.2 mouse breast adenocarcinoma cell lines as models that very closely mimic the tumor growth of human TNBC23. Primary tumor growth properties are similar in the two models, but 4T1.2 is a single-cell clone of 4T1 with a high propensity to metastasize spontaneously to bone. The env product of endogenous murine leukemia virus, gp70, is universally expressed in different murine cancer cell lines of disparate histological origin, such as CT26 (colorectal carcinoma), B16 (melanoma), and NBL 947.4 (neuroblastoma), as well as the breast cancer cell lines 4T1 and 4T1.2, but silent in normal mouse tissues; as such, it constitutes a highly suitable antigen target for the study of tumor-specific T cells25. In this study, the proportions of CD39− and gp70 tet− double-negative CD8+ T cells among gp70 tet− single negative CD8+ T cells was 80% in the 4T1 TNBC model and 58% in the 4T1.2 model; moreover, Bayesian posterior probability calculations confirmed that more than 99% of the gp70 tet+ tumor-specific CD8+ T cells were CD39+, which was sufficient to support the potential role of CD39+CD8+ T cells as a potential surrogate marker of tumor-specific T cells.
The correlations between tumor-specific T cell accumulation in the TME and the clinical outcomes demonstrate the critical influence of high-frequency tumor-specific T cells on antitumor immune responses and prognosis in BC patients37–39. Preclinical neoadjuvant immunotherapy studies in murine models of metastatic breast cancer revealed that long-term survival rates were improved by increased frequencies of tumor-specific T cells among PBMCs and within tumors; furthermore, tumor-specific T cells were actively dividing and displayed effector/memory phenotypes, with production of IFNγ and TNF10. In HPV-associated oropharyngeal squamous cell carcinoma tumors, dendritic cells have been shown to interact with T cells to form microaggregates localized within the tumor beds, thereby promoting the presentation of tumor antigens to tumor-specific T cells40. In addition, tumor-specific T cells produce chemokines that attract lymphoid cells, stimulating the continued formation of dendritic cells–T cell microaggregates and, thus, contributing to a positive feedback loop that promotes tumor rejection41.
Accumulating evidence implicates CD39 as a marker of tumor-specific CD8+ T cells11,15,16. In both colorectal and lung tumors, populations of bystander CD8+ TILs are defined by the absence of CD3911, while in treatment-naïve NSCLC, CD39+CD8+ immune cells have been defined as tumor antigen-specific CD8+ T cells13. In high-grade serous ovarian cancer, highly activated CD8+ TILs were characterized by co-expression of CD39, CD103, and PD-117. In HPV-related cervical squamous cell carcinoma, vulvar squamous cell carcinoma, and oropharyngeal squamous cell carcinoma, high-dimensional flow cytometry and single-cell RNA analyses identified HPV-specific T cells as CD39+ 42.
Enrichment of CD39+CD8+ T cells within tumors has shown disparate prognostic value in terms of response to immune checkpoint blockade, depending on the type of cancer. In high-grade serous ovarian cancer, CD39+CD8+ T cells are associated with high levels of prognostic significance and represent attractive targets for combination immunotherapies including PD-1, CD39, and TIGIT17. In addition, in human colorectal and lung tumors, CD39+CD8+ T cell frequencies correlate with key clinical parameters such as the mutational status of lung tumor epidermal growth factor receptors11,13. A study by Weinberg et al. also supported the correlation of CD39+CD013+CD8+ T frequencies with increased OS in HNSCC patients. Furthermore, analysis of The Cancer Genome Atlas datasets for HNSCC, lung adenocarcinoma, and lung squamous cell carcinoma revealed better 3-year OS in CD39high-patients than in CD39low-patients15. In this study, we demonstrated the exceptional prognostic significance of CD39+CD8+ T cells in terms of OS in TNBC patients. Furthermore, in a small cohort of TNBC patients receiving immunotherapy, we found that high CD39+CD8+ T cell density was associated with immunotherapy response. Therefore, our findings indicate that CD39+CD8+ T cells represent an alternative resource to predict clinical outcomes as well as response to immunotherapy in TNBC patients.
In this study, we focused mainly on CD39+ subpopulations within TNBC tumors, as CD39 is reportedly expressed at high levels in tumor-infiltrating immune cells such as effector T cells, Tregs, and myeloid cells31. In addition to investigating CD8+ T cells, we also examined the infiltration of FOXP3+ Tregs and CD68+ TAMs within TNBC tumors. Tregs have an essential immunosuppressive role in tumorigenesis and cancer development43; however, reports of the prognostic significance of infiltrating Tregs in TNBC are conflicting. Some studies have shown that TNBC patients bearing tumors with high frequencies of Tregs within the tumor but not the stroma experienced significantly longer OS and DFS, compared with individuals characterized by low Treg densities44. A study by Park. et al. suggested that high levels of FOXP3+ staining in Tregs (detected by IHC) served as an independent prognostic factor for OS and progression-free survival43. By contrast, another study showed that low frequencies of FOXP3+ TILs correlated significantly with favorable relapse-free survival and OS in TNBC 45.
Substantial progress has been made in elucidating the biology of TAMs, which can be manipulated to improve disease control in a substantial fraction of patients across different cancer types46. In general, TAMs are identified by the expression of the pan-macrophage marker CD68. Despite reports of higher frequencies of CD68+ macrophages in TNBC, compared with other types of BC47, results have been inconsistent. For example, CD68+ macrophages were identified as an independent prognostic indicator of reduced survival in a 287-patient TNBC cohort48; however, there was no significant correlation with survival in a cohort of 285 patients who had undergone primary tumor resection before any systemic treatment49 or a cohort of 203 patients treated with adjuvant chemotherapy50. In our naïve Asian TNBC cohort, high densities of CD68+ infiltrates were associated with OS but not with DFS. It can be speculated that these conflicting conclusions might be accounted for by variations in the cohorts studied and the use of CD68 as a pan-macrophage marker unable to distinguish between macrophage subpopulations. The relative spatial distribution of TAMs in tumor cell nests and tumor stroma may also have an important influence on the results51. In our study, because of the limited area of tumor tissue available for analysis, we quantified macrophage infiltrates across the entire tumor area to reduce possible sampling bias.
TNBC is characterized by an immunosuppressive TME enriched with myeloid cells. As a potent damage-associated molecular pattern released by damaged cells, extracellular ATP functions as a physiological adjuvant with multiple proinflammatory effects. The release of ATP is triggered by cellular stress and tissue damage52, as well as hypoxia-induced tumor cell death and dysregulation of cellular processes53. ATP can be inactivated by the ectoenzymic activity of CD39 (nucleoside triphosphate diphosphohydrolase-1 [NTPDase 1]), which degrades ATP to AMP. Cell surface expression of ectonucleotidases has been reported for various tumor types, along with adenosine production11,13,15,54. Adenosine is detectable in the TME/interstitial fluid of solid tumors, reportedly at concentrations capable of suppressing antitumor immune responses55. By triggering a proinflammatory response, ATP release paves the way for effective wound healing and/or the restoration of homeostasis; in addition, ATP promotes M1 macrophage polarization and enhances the tumoricidal potential of macrophages56. During the inflammatory process, ATP released by necrotic tumor cells and various inflammatory cells is swiftly broken down into adenosine, while CD39 expression is increased in inflammatory cells during chronic inflammation and tumorigenesis. Thus, in terms of the TME, CD39 expression indicates the presence of tumor-specific cells and, at the tumor site, serves as a hallmark of chronic antigenic stimulation.
Some limitations of our study should be noted. As our cohort included only six patients with low frequencies of FOXP3+ TILs, we were unable to perform meaningful multivariate and Kaplan–Meier analyses of FOXP3+ Tregs. However, between-patient variability was low, and work is ongoing to obtain additional patient samples to verify the prognostic importance of FOXP3+ Tregs in TNBC patients. In addition, the TNBC cohort receiving ICI therapy was small; further studies are, therefore, required to accurately determine the proportion of tumor-specific T cells in an immunotherapy-treated TNBC cohort and identify trajectory changes potentially predictive of general responsiveness to immunotherapy and disease progression.
In conclusion, we have identified CD39+CD8+ T cells as a potential surrogate marker of tumor-specific T cells in treatment-naïve TNBC and provided evidence that this population is associated with prognosis and immunotherapy response in these patients.
Table 1
Multivariate analysis of CD39+ immune infiltrates in TNBC patients with different survival outcomes.
Biomarkers | HR | 95% CI | P-value |
Overall survival |
CD39+ immune infiltrates | 0.25 | 0.09–0.88 | 0.0154* |
CD39+ CD8+T cells | 0.50 | 0.27–0.96 | 0.0317* |
CD39+ CD68+ macrophages | 0.38 | 0.20–1.04 | 0.0023* |
CD39+ FOXP3+ Tregs | 0.49 | 0.24–1.04 | 0.0522 |
Disease- free survival |
CD39+ immune infiltrates | 1.59 | 0.90–2.69 | 0.0970 |
CD39+ CD8+T cells | 0.71 | 0.43–1.19 | 0.1792 |
CD39+ CD68+ macrophages | 2.84 | 0.46–9.55 | 0.1564 |
CD39+ FOXP3+ Tregs | 1.47 | 0.89–2.41 | 0.1283 |
HR: hazards ratio; 95% CI: 95% confidence intervals