Immunotherapy by immune checkpoint inhibition has achieved critical success in treating advanced clear cell renal cell carcinoma (ccRCC) during the past 5 years [1]; however, not all patients benefited from treatment. Although investigations of immune checkpoint genes and their products in RCC tumors have been conducted in the past, the prognostic role of soluble immune checkpoint markers has not been extensively explored, especially among localized cancer patients.

Previous studies have implicated high expression of inhibitory immune checkpoint genes in T cells strongly linked to T cell exhaustion and inefficient control of infections and tumors [2]. Giraldo et al. reported that in a group of 40 localized ccRCC cases the presence of CD8⁺PD-1⁺TIM3⁺LAG3⁺ tumor infiltrative lymphocytes (TILs) with CD4⁺ICOS⁺ T-reg cells identified patients with deleterious prognosis [3]. In a study involving 135 primary ccRCC cases and 51 metastatic ccRCC cases, PD-L2 expression in tumor cells and LAG3 expression in TILs were identified as poor prognostic factors in ccRCC patients [4]. In another cohort of RCC patients from Japan, high expression of immune checkpoint molecules in TILs correlated with poor overall and recurrence-free survival [5]. Thus, the expression of immune checkpoint genes in both immune and tumor cells may correlate with anti-tumor immunity in the tumor microenvironment (TME), pointing to the markers’ prognostic or therapeutic potential in ccRCC.

Only a few studies have examined the association between soluble immune checkpoint-related proteins and cancer outcomes. Circulating T cell regulatory proteins, some of which might modulate immune checkpoints, could be released from immune and tumor cells [6]. High level of circulating PD-L1 was associated with impaired immunity and poor outcomes in aggressive RCC, diffused large B-cell lymphoma and pancreatic cancer [7–9]. A spliced variant of PD-L1 was also reported to be secreted in the blood and to induce immune suppression in multiple cancers [10]. He Y et al. reported that low sLAG3 was associated with advanced stage in non-small-cell lung cancer (NSCLC) [11]. Taken together, these results suggest soluble immune checkpoint-related molecules may play a prognostic role in RCC and other cancers.

To identify soluble immune checkpoint-related proteins that can predict outcomes of ccRCC, patients, we implemented a three-stage strategy. First, we systematically assessed the level of soluble immune checkpoint-related proteins and their association with recurrence and survival in 182 ccRCC cases from an ongoing case-control study at The University of Texas MD Anderson Cancer Center (MDACC). Second, we evaluated the expression of immune genes in ccRCC tumors from the MDACC cohort, and we analyzed their association with cancer outcomes. Third, we analyzed immune gene expression in an external set of RCC tumor data from The Cancer Genome Atlas (TCGA). In short, this is an integrated, multi-stage investigation focus on peripheral immune checkpoint-related proteins and further supported by tumoral data from MDACC and TCGA cohorts.

In this study, we identified a panel of soluble immune checkpoint-related proteins being associated with clinical outcomes of ccRCC patients. We demonstrated that circulating levels of sLAG3 are associated with risk of advanced disease; sPD-L2 level associated with risk of recurrence; sTIM3 and sBTLA levels correlated with risk of death in ccRCC patients. We generated a risk score combining the two biomarkers associated with survival indicating the soluble immune checkpoint-related proteins jointly predict the death risk of ccRCC. Furthermore, sLAG3 and sCD28 levels negatively correlated with the number and cytolytic activity of T cells in ccRCC tumors. These results highlighted the prognostic value of these soluble immune checkpoint-related proteins and unveiled potential biological mechanisms in ccRCC development.

We identified sPD-L2 as the most significant biomarker associated with ccRCC recurrence in this study. Previous work has described that sPD1 enhanced anti-tumor immunity by blocking PD-L1 in tumor cells [16], whereas sPD-L1 predicted poor prognosis in aggressive diffuse large B-Cell lymphoma [8]. In our study, sPD-L2 but not sPD-L1 was predictive of recurrence risk in ccRCC, and sPD-L2 level seems to be higher than that of sPD-L1, which is consistent with the protein expression of these two markers in TILs from another study [4]. This finding is partially supported by one previous study suggesting that high PD-L2 expression in tumor is associated with reduced cancer free survival in RCC patients [17]. The source of sPD-L2 may be derived from tumor exosomes [18] or alternatively activated macrophages [19] to inhibit T cell-mediated anti-tumor response. Therefore, high sPD-L2 could be a predictive biomarker of recurrence risk in ccRCC patients, although the finding warrants further confirmation in independent populations and exploration of the underlying biological mechanisms.

High density of LAG3+ T cells is a signature of T cell exhaustion in tumors [20]. Our results indicated that high level of sLAG3 is associated with advanced tumor stage in ccRCC patients. This is consistent with Camisaschi’s study that LAG3 were highly expressed in Treg cells in peripheral blood, tumor-involved lymph nodes and within tumor tissues isolated from patents with advanced (stage III and IV) melanoma and colorectal cancer [21]. Also, sLAG3 was also found marginally associated with poor survival (P = 0.07), which was supported by another study that sLAG3 was associated with poor survival in chronic lymphocytic leukemia, and sLAG3 could promote leukemic cell activation and anti-apoptotic effects [22]. However, another study of breast cancer showed that sLAG3 could serve as “Th1” (type I T helper cell) marker and that high level of sLAG3 predicted better OS [23]. This discrepancy may be due to the differential regulatory role of sLAG3 in mediating interaction between LAG3 and MHC-II, or the distinct immune landscapes of different cancer sites [20, 24]. Furthermore, we found that levels of sLAG3 negatively correlated with CD8A (T cell marker) expression and T cell cytolytic activity in tumors. Therefore, we propose that high level of sLAG3 may be indicative of T cell suppression in the TME, which in turn lead to advanced development of ccRCC.

TIM3 (HAVCR2) is an inhibitory receptor expressed on T cell and tumor cell surfaces which regulates Th1 and cytotoxic T cell responses [25, 26]. High TIM3 and PD1 expressions on T cells and in tumors are the signature of “deeply” exhausted status, which are frequently observed in ccRCC [27, 28]. The soluble TIM3 is shown to be generated by A Disintegrin and Metalloprotease (ADAM)-mediated ectodomain shedding from both T cells and tumor cells [29]. Our results revealed that sTIM3 is associated with advanced disease and increased death risk of ccRCC. This finding further confirmed previous studies that TIM3 + PD1+ TILs exhibited exhausted phenotype in the TME, thereby resulting in poor prognosis of cancer patients [3, 30]. The mechanisms of how sTIM3 interacts with its ligands, other inhibitory checkpoint proteins and TCR signaling pathway in T cells remain unclear necessitating further independent research. Interestingly, we found the association of sTIM3 with death risk was significant only in early-stage (I and II) patients suggesting the confounding effect of staging on survival, although our overall Cox proportional hazard analyses have been adjusted for clinical stage. The utility of sTIM3 as an early prognostic biomarker requires further confirmation in independent prospective studies.

BTLA is another inhibitory checkpoint protein that interacts with HVEM and LIGHT, a group of costimulatory molecules, resulting in suppression of T cell immunity [31]. In this study, sBTLA level (>2269 pg/mL) was identified as a predictor of poor OS in ccRCC patients indicating that sBTLA may play a similar role as membranous BTLA in suppressing T cell response. This is in line with Benjamin et al’ study reported that sBTLA (>1910 pg/mL) could predict poor survival in pancreatic cancer patients [9]. BTLA expression in gastric cancer and lymphoma is also reported associate with poor prognosis [32, 33]. However, the immune function of sBTLA may varies by cancer sites, which warrants further research.

CTLA4 and CD28 play opposite roles during T cell activation [34]. However, in our study sCTLA4 and sCD28 were both associated with poor outcome of ccRCC patients, though not significant after multiple testing adjustment. sCTLA4 has been shown as an extrinsic suppressive factor of T cell activation, which could be prominently secreted by T-reg cells [35]. Our results revealed that sCTLA4 is associated with increased risk of recurrence. Meanwhile, sCD28 is also associated with death risk in our ccRCC cohort. We identified negative correlation between sCD28 and cytolytic activity in ccRCC tumors. This result is in line with one previous study showing sCD28 with inhibitory role in T cell proliferation in autoimmune diseases [36]. All of the above findings suggest the potential interactions between sCD28/sCTLA4 and anti-tumor immunity mediating their association with ccRCC outcomes.

There are several strengths to our study including the prospective high-quality cohort with relatively long follow-up time, the multiplex profiling of soluble immune checkpoint-related proteins and immune gene expressions in tumors, and the correlation analysis of soluble immune checkpoint-related proteins and T cell functions to provide biological validity. Despite these strengths, we also acknowledge some limitations. First, we have limited sample size with relatively small number of outcome events, which might constrain the power of our study. Limited tumor tissues available in our cohort and partial clinical outcome information in TCGA dataset might lead to some uncertainties in our results. Additional validation within larger independent cohort is necessary. After consideration of multiple testing, some of the associations we identified may not be significant. Nevertheless, the associations of sLAG3 with advanced (stage III) disease; sPD-L2 with recurrence; sTIM3 and sBTLA with survival remain significant after the stringent Bonferroni adjustment suggesting less likelihood for false discovery. Second, we did not conduct mechanistic studies to determine the functional impact of soluble immune checkpoint-related proteins. Instead, we evaluated the associations between soluble immune checkpoint-related proteins and T cell functional gene expressions to decipher potential mechanisms. Third, evaluation of immune checkpoint expression in peripheral blood leukocytes may be informative to examine any correlation between soluble protein levels and peripheral immune gene expression. Nevertheless, our inquiry is supported by one previous study correlating tumoral and peripheral blood T cell phenotypes with aggressiveness of ccRCC [3].

In this study, we identified a panel of circulating immune checkpoint-related proteins that are associated with clinical outcomes and T cell phenotypes in ccRCC patients. Individually and jointly, the soluble immune checkpoint-related proteins may assist risk stratification of ccRCC patients to identify those at high risk of recurrence or poor survival for more intensive surveillance and/or treatment. Future study may apply these markers to test their predictive value for treatment outcome in immunotherapy-treated patients.