Prostate cancer (PCa) remains the second most commonly diagnosed cancer and the fifth cause of cancer death among men worldwide despite PCa screening and early effective treatments (1, 2). This may be due in part to the late systemic recurrences occurring after years of apparent PCa control in aging men who live longer, thanks to reduced competing cause of mortality. The management of PCa is challenging because this cancer is a highly heterogeneous disease (3, 4). Therapeutic options will vary considerably for patients with low-risk indolent PCa and those with high-risk life-threatening PCa. One major challenge lies in the adequate risk stratification to help select the most appropriate therapeutic strategy and avoid under or overtreatment (5). So far, the stratification of risk is based on clinico-pathological factors such as tumor grade (Gleason score) and stage (TNM), PSA level at diagnosis and presence of adverse pathological features at prostatectomy (6, 7). However, the risk of recurrence and progression and the response to treatments vary significantly between patients with otherwise similar clinico-pathological characteristics. Therefore, new biomarkers are needed to refine the prognostication and improve the management of PCa (8). Moreover, the most common endpoint of biomarker studies in PCa has been biochemical recurrence (BCR), but few studies have been able to relate biomarkers with long-term outcomes such as metastasis and mortality by PCa. It is even more challenging to identify biomarkers in the primary PCa that could lead to potential early interventions to reduce late PCa mortality.

The influence of the tumor microenvironment (TME) on the development and progression of cancer has gained greater interest during the last decade (9). The TME influences cancer evolution through diverse mechanisms including tumor differentiation, stimulation of angiogenesis and promotion of immune evasion (10–12) Tumor-infiltrating lymphocytes (TIL) participate in the host defense against tumor cells but their antitumor activity is greatly hampered by the immunosuppressive factors found within the TME which favors immune evasion (13). Several studies have shown that the density, phenotype and localization of TIL are associated with clinical outcomes in various types of solid cancers and that their detailed analysis can provide important prognostic information (14, 15). The prognostic potential of TIL has been notably well demonstrated in colorectal cancers (CRC). Studies by Galon et al. have shown that TIL could predict better than TNM disease-free survival and overall survival of CRC patients (16, 17). These studies led to the development of the “Immunoscore” to complement the current TNM in order to better predict the risk of recurrence and even response to therapy of CRC patients (18–20).

While the prognostic potential of TIL has been well established in CRC and some other solid cancers, the study of the relationship between TIL and the clinical outcomes of PCa has provided inconsistent results (21–33). These inconsistencies may be due in part to differences in study designs, to various subtypes of TIL studied and their localization within the specimen and to the different methodologies used to measure them. The types of patients’ cohorts and the length of their follow-up are also very important parameters to assess the clinical relevance of the findings. The most recent immunohistochemical (IHC) studies have focused on regulatory T (T_reg) cells using antibodies against the transcription factor FoxP3, the most specific biomarker of the T_reg cells. These studies showed an association between FoxP3⁺ TIL and higher risk of BCR or death from PCa (25, 31, 34). However, very few studies have characterized the infiltration of prostate tumors by CD45RO⁺ cells. CD45RO is a marker of central and effector memory T (T_mem) cells and is therefore associated with an effective immune response. In CRC, the combined analysis of CD45RO⁺ T_mem cells and CD8⁺ T cells in specific tumor areas helped predict cancer recurrence and patients’ survival (17).

To further characterize the intra-tumoral immune response against PCa and assess the potential prognostic value of TIL, we analyzed the infiltration of CD3⁺, FoxP3⁺ and CD45RO⁺ cells in the tumor core, tumor margin and the adjacent normal-like epithelium area in radical prostatectomy specimens from 98 PCa patients at increased risk of recurrence and with very mature long-term clinical follow-up. We also analyzed the expression of 31 genes associated with TIL by qRT-PCR using TaqMan^® Low Density Array (TLDA) technology in another set of 50 similar patients with fresh frozen radical prostatectomy. The combined results show that the balance between T_mem and T_reg infiltrating cells within the cancer is a strong independent predictor of systemic cancer progression and lethality by PCa. They also identify the immune checkpoint TIM-3 as a potential target to reverse T_reg dominant immunosuppression in PCa.

T (CD3⁺) and B (CD20⁺) lymphocytes infiltrate prostate cancers, but it was previously shown that B lymphocytes are not associated with clinical outcomes (24). We thus deliberately omitted the analysis of B lymphocytes in this study. CD3⁺ TIL are the effectors of the adaptative cellular immune response and therefore their number, phenotype and localization provide important information on the antitumor response specifically directed against tumor antigens and have been shown to predict clinical outcomes in many types of cancers (19, 35–37). Their presence reflects the immunogenicity of the tumor caused by the presence of either shared tumor antigens or mutated antigens. Although the mutational burden of most PCa is low compared to melanomas or lung cancers, some PCa show some high levels of TIL which might reflect some specific characteristics of these tumors (38, 39). It has been notably shown that tumor with deficient mismatch repair mechanisms are associated with a higher density of TIL due to the creation of neoepitopes resulting from the unrepaired DNA replication errors (40). Similarly, it was found that advanced PCa with inactivating CDK12 mutations, which affect the expression of genes involved in DNA damage response, had elevated neoantigen burden and were more infiltrated with T cells which even showed some clonal expansion (41, 42). Other characteristics were associated with a higher density of TIL in PCa. Vidotto et al. and Kaur et al. showed that tumors with loss of the PTEN tumor suppressor gene had higher density of FoxP3⁺ cells (33, 43). It was also found that higher TIL density was more frequent in ERG positive tumors (24, 33, 44). Hence, there is growing evidence of a relationship between genomic alterations and immune cell infiltration in PCa and thus the level of TIL subtypes may reflect distinct biological behaviors and prognoses.

In this study, we analyzed the infiltration of TIL by immunohistochemistry using CD3 as a marker to identify the entire T cell population, FoxP3 to identify the T_reg cells and CD45RO to identify T cells with a memory phenotype. We studied these T cell subtypes in three compartments of the tissue sections in order to assess whether the localization of these immune cells had a significant impact on the prognosis. Contrary to many other immunohistochemical studies that analyzed TIL using tissue microarrays (TMA) (22–25, 31, 33, 34) our study was performed on whole sections of FFPE tumors which allowed us to assess more accurately the number of TIL in the different tissue compartments. We showed that no matter the phenotype, the number of CD3⁺, FoxP3⁺ or CD45RO⁺ cells was highly variable from one tumor to the other and these cells were on average more frequent at the tumor margin, although the difference was not statistically significant ( Figure 2 ). This is consistent with another study reported by Yuan et al. which found that the majority of immune cells were found in the normal-like or pre-invasive tissue areas rather than in the tumor core (45). The higher density of immune cells at the tumor margin may reflects the properties of the tumor microenvironment making access to the tumor core more difficult. Such accumulation of immune cells outside the tumor core might provide important prognostic information. However, we found that it was the density of the T cells within the tumor area and not in the peritumor area (margin and adjacent normal-like areas) that was the most informative on the late clinical outcomes suggesting that T cells within the tumor core have a more important impact on the tumor evolution ( Figure 4 , Supplementary Table S2 ).

Our results showed that a high density of CD3⁺ cells in the peritumor area was associated with a lower risk of BCR while a low density of these cells in the tumor area increased the risk of lethal PCa. This is in contrast with some studies that concluded that a higher TIL density was associated with poorer prognosis. For example, McArdle et al. performed a study in a cohort of 80 patients with PCa, in which they analyzed on whole tumor sections the infiltration of CD4⁺ and CD8⁺ lymphocytes within the tumor (including the cancer cell nests and surrounding stroma) but excluding any infiltration in the surrounding normal-like epithelium. The analysis revealed that a higher infiltration of CD4⁺ T lymphocytes in the tumor was associated with a shorter PCa specific-free survival in both univariate (HR= 2.03, 95% CI=1.15–3.59, p=0.015) and multivariate (HR=2.29, 95% CI=1.25-4.22, p=0.008) analyses. Infiltration by CD8⁺ T lymphocytes was however not predictive of PCa specific-free survival (21). A Norwegian study carried out on 535 primary PCa displayed on TMA, showed that in multivariate Cox regression analysis, a higher density of CD8⁺ lymphocytes in both tumor epithelial and tumor stromal areas taken as one single value, was an independent negative prognostic factor for BCR-free survival (HR= 1.565, CI 95%= 1.132–2.165, p= 0.007). Also, when the infiltration of CD8⁺ cells was assessed only in the tumor area, the association remained significant (HR=1.445, CI 95%1.028–2.032, p=0.032) (23). Flammiger et al. reported the analysis of a large TMA encompassing 3261 PCa samples. They found that both a low and high infiltration by CD3⁺ lymphocytes were associated with a shorter BCR‐free survival when compared with an intermediate level of infiltration in Kaplan-Meier analyses (p=0.019), a phenomenon known as the Goldilocks effect (24). This Goldilocks effect along with the study design including the characteristics of the tumor series, the choice of TMA vs whole sections to analyze the cell density in the different tumor compartments, the choice of the antibodies as well as the known heterogeneity of PCa could all contribute to generate variable results from one study to the other. Moreover, looking at the whole T lymphocyte population with markers such as CD3 or even CD4 and CD8 knowing that these cell populations are also heterogeneous as they contain subpopulations with anti-tumor, but also tumor-promoting phenotypes might not provide accurate prognosis. Looking at TIL corresponding to T cell subpopulation with more define characteristics such as T_reg and T_mem is expected to provide more informative data.

We showed in this study that the ratio of immune cells within the tumor area have important independent prognostic values. We showed that low ratio of FoxP3⁺/CD3⁺ or of FoxP3⁺/CD45RO⁺ cell density was associated with a lower risk of definitive ADT. Consistent with the previous finding, to the opposite a high ratio of FoxP3⁺/CD3⁺ cell density was associated with a higher risk of definitive ADT or lethal PCa and a high ratio of FoxP3⁺/CD45RO⁺ was associated with a higher risk of BCR. Our results are in concordance with results of other studies looking at the prognostic value of T_reg. Davidsson et al. conducted a study using a TMA comprising tumor samples from 1367 men. The results inferred that neither infiltration of total CD4⁺ cells nor of CD8⁺ cells was associated with lethal PCa. However, a higher risk of lethal PCa was found when comparing the highest with the lowest quartile of FoxP3⁺ cells (odds ratio=1.98; 95% CI= 1.15–3.40) (25). Likewise, an Italian study aiming to characterize TIL in tumor and peripheral stroma areas of tumors of 22 men treated by radical prostatectomy and salvage radiotherapy concluded that a low infiltration by CD45⁺ and FoxP3⁺ cells in the peripheral stroma was correlated to a prolonged BCR-free survival and a better overall survival. However, there was no correlation between the infiltration by CD3⁺ and CD4⁺ TIL and clinical outcomes. Higher risks of dying from prostate PCa was found when comparing the highest with the lowest quartiles of FoxP3⁺ cells (odds ratio=1.98; 95% CI= 1.15–3.40) (26). Kaur et al. also found after analysis of the tumors of 144 African-American men an association between increased FoxP3⁺ cell density and a higher risk of metastasis in multivariate analysis (HR=12.89 (1.59-104.40) p=0.02) (33). More recently, Andersen et al. analyzed the prognostic potential of different immune cells in two large cohorts of radical prostatectomy specimens available in TMA (34). They observed that T_reg and M2 macrophages in stroma and epithelium, respectively, were adverse predictors of BCR in multivariate Cox regression analyses. A similar association of T_reg with BCR was found at the mRNA level in a third cohort therefore thus further supporting the association of T_reg with poorer prognosis (34). Using the same cohort as in this study, we recently reported the analysis of the prognostic value of the infiltration of tumors by macrophages and dendritic cells (DC) (46). We also found M2 macrophages were associated with a poorer prognosis. Indeed, a higher infiltration of CD163⁺ M2 macrophages in the normal adjacent epithelium as well as a higher infiltration of CD209⁺ immature DC at the tumor margin were associated with lethal PCa and BCR, respectively. Deeper analyses showed that the ratio of CD209⁺ immature DC over CD83⁺ mature DC was even more predictive of late adverse events, showing that the proportion of these immune cells must be taken into account to fully evaluate their prognostic value.

Our study also shows that CD45RO⁺ cell density and the ratio of CD45RO⁺/CD3⁺ cells in the tumor area was associated with good prognosis since a high density of CD45RO⁺ cells was associated with a lower risk of BCR and a high ratio of CD45RO⁺/CD3⁺ was associated with a lower risk of lethal PCa. To the contrary, a low ratio of CD45RO⁺/CD3⁺ density was associated with a higher risk of BCR. To our knowledge, our study appears to be the first to determine the prognostic value of CD45RO⁺ cells by immunohistochemistry in the microenvironment of PCa. There was one previous report of CD45RO⁺ cell infiltration in normal prostate but not in tumors (47). The infiltration of tumors by CD45RO⁺ cells was shown to be associated with clinical outcomes in various types of tumors (35, 48, 49). The independent prognostic value of CD45RO⁺ cells highlighted by this study underscore the importance to include this biomarker in future studies looking at TIL in PCa. Our data also show that the balance between CD45RO⁺ and FoxP3⁺ cells in the tumor area has an important prognostic value.

Supporting the results of the immunohistochemistry study, we found that high expression of FoxP3 gene as measured using TLDA technology was associated with a shorter lethal PCa-free survival in Kaplan-Meier analyses ( Figure 7 ) but also in univariate Cox regression analysis ( Supplementary Table S4 ). Amongst the genes we studied, CTLA-4 was the one that was the most highly correlated to FoxP3. The high expression of CTLA-4 by effector T_reg is well known and the use of anti-CTLA-4 is a relevant approach to kill T_reg or at least attenuate their suppressive activity (50–52). Beside CTLA-4, effector T_reg also express other ICP such as PD-1, TIM-3, LAG-3, GITR and OX40 that are also potential targets to modulate T_reg functions (50–53). In our analysis of the TLDA results we found that high expression of CTLA-4, PD-1, TIM-3 and LAG-3 genes together was associated with a shorter lethal PCa-free survival consistent with the immunosuppressive action of these molecules on T cells. However, when taken separately CTLA-4, PD-1 and LAG-3 were not significantly associated with the clinical outcomes but a high expression of TIM-3 was significantly associated with a shorter definitive ADT-free survival ( Figure 7 ). Inhibition of TIM-3 might be an approach to consider to counteract the immunosuppressive activity of FoxP3⁺ cells in PCa as suggested by some authors (54–56).

This study has however some limitations. The first one concerns the technical approach. We used in this study a standard IHC technique in which each slide is stained with a single antibody against a marker specific to a cell population instead of performing a multiplex analysis for the simultaneous detection of different markers and corresponding cell populations on a single slide. We selected the standard IHC approach for practical reasons and to ease an eventual clinical application as this approach is still the most frequently used in clinical pathology laboratories. However, when the objective is to determine a ratio between cell populations, this approach is less adapted and tedious compared to the use of a multiplex assay. Moreover, the multiplex assay also have the advantage to detect various cell phenotypes i.e. cell expressing combinations of markers. For example, it was reported that some T_reg cells may also express CD45RO (57–59). The use of a multiplex assay would allow to assess the prognostic value of the density of such double positive FoxP3⁺CD45RO⁺ cell population in addition of that of the single positive FoxP3⁺ or CD45RO⁺ cell populations. In this study, we could not detect this double positive cell population and so we don’t know the frequency of these cells in the series of tumors we analyzed. But since the FoxP3⁺ cell population is about 8 times less numerous than the CD45RO⁺ cell population, the occurrence of some cells expressing both markers is not expected to significantly modify the conclusion of the study. A second limitation of the study is the size of the cohorts. In several occasions, association with the outcomes did not reach statistical significance because of the limited size of the cohorts. The results of this study should be validated in larger cohorts.

In conclusion, we showed in this study that the infiltration of the tumor area by FoxP3⁺ T_reg and CD45RO⁺ T_mem cells are highly predictive of late clinical events when they are related to the CD3⁺ cell population or related to one another using ratio of cell density. Using ratio of FoxP3⁺ and CD45RO⁺ over CD3⁺ cell density or to one another is a way to normalize the frequency of these cells in the tissue which hopefully might offer more reproducible results. Our gene expression analysis by TLDA supported the association between a high expression of FoxP3 gene and a higher risk of lethal PCa. It also identified a correlation between a high TIM-3 gene expression with a higher risk for definitive ADT. These results let us to suggest that inhibition of TIM-3 might be a relevant approach to counter the immunosuppressive functions of T_reg in order to improve the anti-tumor immune response against PCa.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving humans were approved by Research ethics committee of the CHU de Québec-Université Laval. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

OM: Visualization, Validation, Methodology, Investigation, Formal analysis, Data curation, Conceptualization, Writing – review & editing, Writing – original draft. HL: Writing – review & editing, Supervision, Resources, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization. DS: Writing – review & editing, Methodology, Formal analysis. HH: Writing – review & editing, Resources, Investigation. BV: Writing – review & editing, Investigation, Formal analysis. BT: Writing – review & editing, Investigation, Funding acquisition. VF: Writing – review & editing, Funding acquisition. LL: Writing – review & editing, Methodology, Funding acquisition, Conceptualization. AB: Writing – review & editing, Writing – original draft, Supervision, Resources, Funding acquisition. YF: Writing – review & editing, Writing – original draft, Supervision, Resources, Project administration, Methodology, Funding acquisition, Formal analysis, Conceptualization.