Clinicopathologic implications of the miR-197/PD-L1 axis in oral squamous cell carcinoma

Immune escape of a tumor from tumor-infiltrating lymphocytes (TILs) is induced by PD-L1, which is suppressed by miR-197. We investigated the clinicopathologic implications of the miR-197/PD-L1 axis and its effects on TILs and the clinicopathologic features of oral squamous cell carcinoma (OSCC). We used RT-PCR and immunohistochemistry in 68 OSCC patients to analyze the correlations between tumoral expression of miR-197 and PD-L1 and the degree of tumoral invasion by TILs (CD3+, CD4+, CD8+, PD-1+, FoxP3+, and CD20+ lymphocytes). PD-L1 levels correlated inversely with miR-197 but correlated positively with TILs. The aggressive features of OSCC, including high stage, angiolymphatic invasion, perineural invasion, and death, were associated with TIL depletion. High T stage (T4) tumors also had low PD-L1 but had high miR-197 expression. In a univariate survival analysis of the full cohort, high miR-197 was associated with poor overall survival, whereas high PD-L1 expression (2+) associated with good overall survival. In a multivariate analysis stratified based on miR-197 (median), high PD-L1 expression (2+) was an independent favorable prognostic factor for overall survival (P = 0.040) in the miR-197high subgroup but not the miR-197low subgroup. These findings may have clinicopathologic implications for the miR-197/PD-L1 axis and TILs in OSCC.


INTRODUCTION
Oral cancer accounts for 1% to 5% of human malignancies, and its occurrence has increased for more than a decade [1]. Oral squamous cell carcinoma (OSCC) is the predominant type and comprises approximately 90% of all oral cancers [2]. OSCC occurs mostly in males with long-term exposure to tobacco and alcohol who are in their fifth to seventh decade of life [3]. The prognosis of OSCC remains poor because of the tendency of the cancer to metastasize [4]. Despite recent advances in combined surgery, chemotherapy, and radiotherapy, the survival rates of patients with OSCC has shown no improvement over the past few decades [2]. Therefore, an understanding of the molecular mechanism underlying progression of OSCC in the context of the tumor microenvironment is needed for improvement of the therapeutic strategies for the disease.
Programmed cell death ligand-1 (PD-L1) is a cell-surface glycoprotein that induces T cell anergy and apoptosis by activating the PD-1 receptors on T lymphocytes [5]. Normally, PD-L1 is a factor in maintaining immunologic homeostasis, but in many cancers PD-L1 is overexpressed on tumor cells as well as on subsets of immune cells, including T cells, B cells, macrophages, and dendritic cells. Blocking the PD- Research Paper www.impactjournals.com/oncotarget L1/PD-1 pathway reverses the immune escape of the tumor and improves the anticancer immune responses in microenvironments containing tumor-infiltrating lymphocytes (TILs) [6]. PD-L1 has gained attention for its benefits in clinical trials of cancer immunotherapy. However, the number of studies on the biologic mechanisms and effects of PD-L1 in OSCC is limited [7,8], and the clinicopathologic implications, including prognostic values and associations with TILs, have not been clarified.
MicroRNAs (miRNAs) are single-stranded, noncoding short RNAs consisting of 20 to 22 nucleotides that inhibit the translation of genes by disrupting specific messenger RNAs. In recent studies, the expression profiles of miRNAs in OSCC were examined and some miRNAs were found to correlate with tumorigenesis and cancer progression [9,10]. MiR-197 is associated with a broad range of pathologic conditions, including malignancies and various non-neoplastic diseases [11]. Studies suggest an miR-197induced regulating mechanism of PD-L1 in tumor cells via the miR-197/CKS1B/STAT3 pathway in non-small cell lung cancer (NSCLC) [12,13]. In those studies, downregulation of miR-197 was suggested as the promoter of chemoresistance, having therapeutic potential by miR-197 replacement, particularly in PD-L1-positive NSCLC patients.
In the present study, we contribute to a better understanding of the clinicopathologic implication of the miR-197/PD-L1 axis and TILs in OSCC. We analyzed the expression of miR-197 and PD-L1 and analyzed the number of recruited TILs (CD3+, CD4+, CD8+, PD-1+, FoxP3+, and CD20+ lymphocytes) and the correlations with various clinicopathologic features and prognosis in OSCC patients.

The effects of miR-197 and PD-L1 expression and TILs on prognosis in OSCC
The association between conventional clinicopathologic factors and prognosis in OSCC patients was evaluated (Table 4 and Figure 3). Univariate analysis revealed that T stage (T3-T4; P < 0.001) and angiolymphatic invasion (P = 0.001) were predictors of poor OS in OSCC patients. Similarly, T stage (T3-T4; P < 0.001), angiolymphatic invasion (P = 0.001), and N stage (P = 0.007) were associated with poor DFS. High expression of miR-197 was weakly associated with poor OS (P = 0.033), and high PD-L1 expression in tumor cells (2+) predicted better OS (P = 0.039; Figure 3), as expected from the inverse correlation between miR-197 and PD-L1. In multivariate analysis, T stage (T3-T4) and angiolymphatic invasion were independent poor prognostic factors for OS (P < 0.001 and P = 0.002), and DFS (P = 0.001 and P = 0.003), but miR-197 or PD-L1 was not associated with OS or DFS.

Survival analysis in miR-197 low and miR-197 high subgroups
Because our results suggested that miR-197 inhibited PD-L1 expression in OSCC, we investigated correlations of PD-L1 with survival according to miR-197 level (Table 5 and Figure 3). In the miR-197 high subgroup (>median value of miR-197; n=34), prognostic factors for OS in univariate analysis, namely, PD-L1, T stage (T3-T4), and angiolymphatic invasion, were incorporated into the multivariate analysis, resulting in PD-L1 and angiolymphatic invasion as independent prognostic factors for OS (P = 0.040 and P = 0.035), whereas angiolymphatic invasion was the only prognostic factor for DFS (P = 0.009). In contrast, multivariate analysis of the miR-197 low subgroup (<median value of miR-197; n = 34) showed that T stage (T3-T4) and angiolymphatic invasion were independently associated with OS (P = 0.005 and P = 0.040), and T stage (T3-T4) was an independent prognostic factor for DFS (P = 0.001), with no prognostic significance of PD-L1 for OS or DFS in the miR-197 low subgroup. Therefore, high PD-L1 expression (2+) was a favorable prognostic factor for OS only in the miR-197 high   subgroup, suggesting that prognostic implication of PD-L1 is influenced by the miR-197 level.

DISCUSSION
In the tumor microenvironment, the PD-L1/ PD-1 axis has become the therapeutic locus in various cancers. In this study, we focused on the clinicopathologic implication of the miR-197/PD-L1 axis and the profiles of recruited TILs in OSCC. We observed the inverse correlation between miR-197 and PD-L1, and the positive correlation between PD-L1 and TILS, suggesting an active network of miR-197/PD-L1/TILs. Aggressive clinicopathologic features were associated with depleted TILs, and high T stage (T4) disease was associated with low PD-L1 but high miR-197 expression. Moreover, high PD-L1 expression (2+) was an independent favorable prognostic factor for overall survival (OS) (P = 0.040) in the miR-197 high subgroup, but not in the miR-197 low subgroup.
MiR-197 is upregulated in lung, liver, and thyroid cancers and targets a variety of genes promoting cell proliferation and inhibiting apoptosis [11,22,23]. MiR-197 has an indirect effect on PD-L1, which has already been investigated in a study of NSCLC [12]. In the NSCLC study, the miR-197/CKS1B/STAT3-induced PD-L1 network leads to tumor progression, and miR-197 is an inverse indicator of PD-L1 expression and predicts shorter OS [12]. Consistent with this result, we observed that miR-197 level inversely correlated with PD-L1 expression detected by immunohistochemistry in OSCC, suggesting that miR-197, at least in part, suppresses PD-L1 expression in OSCC. The proposed mechanism in NSCLC explains that miR-197 has a direct inhibitory effect on the cyclin-dependent kinase CKS1B, which promotes PD-L1 expression by activating STAT3 [12]. Although the detailed mechanism of PD-L1 expression remains to be clarified, our observations suggest that the putative miR-197-mediated negative regulation of PD-L1 might also have clinical significance in OSCC.  The mechanisms of PD-L1 expression in tumor cells have been suggested to include two major pathways. The extrinsic signaling pathway is activated by interferon-γ (IFN-γ) produced by TILs, resulting in JAK/STAT activation in tumor cells, and the intrinsic pathway is activated by PI3K/AKT/mTOR signaling [24,25]. In this context, miR-197 possibly suppress the downstream signaling of the extrinsic pathway of PD-L1 expression via CKS1B/STAT3 pathway, and therefore, the resultant PD-L1 expression level might reflect the signals from IFN-γ-producing T cells modified by miR-197, as well as signals from the intrinsic pathway. Considering the inverse correlation of miR-197 and PD-L1 in our study, PD-L1 expression in OSCC might be dependent on the miR-197- mediated negative regulation, which could provide the different prognostic significance of PD-L1 according to the miR-197 level.
We divided our OSCC cohort into miR-197 high and miR-197 low subsets, reflecting the mechanism of PD-L1 expression. PD-L1 had prognostic significance only in the miR-197 high subset, although it is not yet clear how high expression of miR-197 influences prognostic significance of PD-L1 in addition to expression of PD-L1. It is possible that OSCC with high miR-197 expression at a sufficient level might have an inhibitory effect on the signal from the extrinsic pathway induced    by STAT3, while sparing the intrinsic signal. This skewing effect of the PD-L1 pathway with co-activated Akt/mTOR-related downstream signals could influence the biology of OSCC in a complex manner, which remains to be clarified further.
Attempts have been made to identify the relation between PD-L1 expression and the effectiveness of immune-induced host defense against the tumor measured by TILs and prognosis of patients [26][27][28]. In this study, we observed that the higher density of TILs was associated with high PD-L1 expression in tumor cells of OSCC. Consistent with our results, previous studies reported the correlation between PD-L1 expression of tumor and microenvironmental TILs in cancers of the colon, breast, larynx, and thymus [28][29][30][31]. Moreover, studies that detected the PD-L1 mRNA expression showed that high PD-L1 expression is associated with increased TILs and favorable prognosis, as shown in our study [29][30][31][32]. These results seem paradoxical, given the inhibitory role of PD-L1 in immune responses. Considering the conflicting results of the PD-L1 association with TILs and prognosis according to tumor types or even within the same type of tumors [27,33,34], PD-L1 expression of tumor or recruited number of TILs might not be a surrogate marker for the final status of effective immune evasion, but a marker of active immune response by tumor cells to evade anti-tumor immune attack by host TILs, regardless of its effectiveness. Therefore, high PD-L1 expression and accumulated TILs in our study might reflect unsuccessfully skewed immune evasion [35], which remains to be clarified further.
In the present study of OSCC, high miR-197 was associated with higher tumor stage (T4) and poor prognosis, whereas high PD-L1 expression in tumors was associated with good prognosis. Although TILs did not directly correlate with survival of OSCC patients, low numbers of various TILs were associated with aggressive features of tumors, including high stage, angiolymphatic invasion, and perineural invasion. The phase of anti-tumor immune reaction might be different between early tumors and full-blown tumors because of different exposure time and amount of tumor antigens, and the mechanism of associations between miR-197/ PD-L1/TILs and stages of disease must be investigated further. Considering that many clinicopathologic factors and outcomes can influence or be influenced by antitumor immune reactions [36], comprehensive analyses of miR-197/PD-L1 and TILs could be more useful than individual analysis of PD-L1. Studies suggest that the miRNA is a factor in controlling the PD-1/PD-L1 signaling for cancer immunotherapy [18,37]. Our approach with miR-197, PD-L1, and TILs could be effective in recognizing the clinicopathologic implication of miR-197/PD-L1 and TILs, and selecting patients for immunotherapeutic strategy and avoiding immunerelated adverse reactions in OSCC.
Studies related to the expression of PD-L1 and its prognostic effects in different cancer types show discordant results [38][39][40]. Various clinicopathologic factors, including cancer types and individual host factors could influence pattern, degree, and effectiveness of anti-tumor immune reactions. Furthermore, different methods of detecting PD-L1 protein or mRNA have been developed, including IHC that utilizes various antibodies [41], in situ hybridization assay [32], DNA microarrays [29], and quantitative real-time PCR [30,31]. In addition, pre-operative neoadjuvant treatment might influence expression of PD-L1, as well as miR-197 and TILs. In this study, neoadjuvant treatment was administered in only 4.4% of patients, with no significant correlation with clinicopathologic variables, suggesting limited effects of neoadjuvant treatments. Because these various factors and techniques might make the results of PD-L1 expression widely variable, previous clinicopathologic studies must be interpreted carefully [42]. In this context, our miR-197/PD-L1-paired approach can contribute to understanding the mechanism and clinical significance of PD-L1-associated anti-tumor host immune function in OSCC. This study showed inverse correlation and prognostic effects between miR-197 and PD-L1 expression in OSCC. In addition, we observed that PD-L1 expression on IHC is associated with increased TILs and favorable prognosis in miR-197 high subgroup. Our clinicopathologic analysis of miR-197/PD-L1 and TILs could contribute to understanding of clinical significance of anti-tumor immune reactions in OSCC.

Patients, samples, clinical data, and study design
A total of 68 OSCC patients who underwent curative surgery between 2003 and 2011 at Seoul National University Bundang Hospital were enrolled. We reviewed hematoxylin-eosin (H&E), immunohistochemically stained slides, pathology reports, and medical records to establish the clinicopathologic features of the tumors. TNM staging was determined according to the seventh edition of the American Joint Committee on Cancer guidelines [43]. Tissue microarrays were constructed by collecting single-tissue cores (2 mm in diameter) from the most morphologically representative areas of formalin-fixed, paraffin-embedded (FFPE) tissue specimens, as previously described [44]. In the analysis of the clinical data, disease-free survival (DFS) was defined as survival time from curative surgical resection to the last follow-up date of diseasefree state or to death. Overall survival (OS) was defined as survival time from curative surgical resection to death by any cause. The time intervals for the regular followup was 3 months after surgery for the first 2 years, and then every 6 months for the next 3 years. The followup period changed if the patient had symptoms or signs that possibly correlated with disease or treatment at the regular visit. Mean follow-up duration was 44.3 months (range, 2.1-122.0 months). During the follow-up period, 28% (19/68) of patients had recurrence and 34% (23/68) of patients died. The remainder of the patients were considered censored in survival analysis for DFS and OS. The primary objective was to determine the correlation among miR-197, PD-L1, TILs, and clinicopathologic factors. The secondary objective was to determine the prognostic effects of miR-197, PD-L1, and TILs on OS and DFS.

RNA extraction and quantitative PCR for miR-197 expression analysis
Total RNA was extracted from 10-μm-thick FFPE tissue sections by use of a RecoverAll Total Nucleic Acid Isolation Kit for FFPE samples (Applied Biosystems, Foster City, CA, USA), and stored at −80° C until the time of use after measuring the concentration with a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). To measure the relative expression level of hsa-miR-197 (Catalog no. #4427975, Applied Biosystems), reverse transcription and real-time polymerase chain reaction were conducted by use of 10 ng of total RNA, Universal PCR Master Mix and a TaqMan microRNA Reverse Transcription kit (Catalog no. 4366596, Applied Biosystems) with U6 snRNA as an internal reference gene and reactive tonsil tissue as a normal control, as previously described [45]. The relative level of miR-197 in OSCC was calculated as 2 -ΔΔCt , where ΔCt = Ct (miR-197) -Ct (U6) and ΔΔCt = ΔCt(tumor) -ΔCt (normal).

Interpretation of immunohistochemical staining
All immunostained slides were blindly evaluated by two experienced pathologists (H.A. and J.H.P.). PD-L1 IHC was defined as membranous or by cytoplasmic positivity and scored as follows: 0 indicates no staining or staining in <10% of the tumor cells, 1+ indicates staining in ≥10% of the tumor cells with weak positivity, and 2+ indicates moderate to strong positivity, which was modified from several previous studies [46][47][48]. The numbers of CD3+, CD4+, CD8+, PD-1+, FoxP3+, and CD20+ TILs around the tumor cells were counted in three representative areas under a high-power optical microscope (400x). The average absolute number was recorded. www.impactjournals.com/oncotarget

Statistical analysis
Statistical analysis was done mainly by SPSS Statistic Version 21.0 software package (IBM Corp., Armonk, NY, USA). Chi-square tests, Student's t-test, Pearson's correlation test, and Spearman's rho test were used to examine the relations among expression of miR-197, PD-L1, number of TILs, and clinicopathologic variables. Survival analysis was performed by use of the Cox regression hazard model. For the multivariate survival analysis, clinicopathologic variables with significant Pvalues (<0.05) in univariate analysis were entered into the multivariate Cox model, and P-values of <0.05 were considered significant. R packages (Version 3.1.2) (http:// www.r-project.org) were used in the graphical dot-plot analysis [49][50][51].