4 Discussion
The interactions of PD-L1 binding to its inhibitory receptor PD-1 inactivate the T cells recognizing the antigen of tumour cells, and consequently, the generation of population of cytotoxic T lymphocytes is reduced that provide an opportunity to cancer cells for escaping immune surveillance. This finding has triggered research to shift treatment from targeting molecules directly on the surface of cancer cells to non-contact methods for blocking the PD-L1/PD-1 pathway for better activation of the immune system4. An increasing number of studies have confirmed that the application of PD-1/PD-L1 treatment efficiently enhance antitumor immunity that reduces tumour growth and improves survival, whereas the prognostic roles of PD-L1 positivity in OSCC are controversial.
Our results revealed a significant association between PD-L1 expression levels and clinicopathological factors. The levels of PD-L1 expression in TCs were relatively high in female patients, non-smoker, non-drinkers, which in line with previous findings of meta-analysis reported by Lenouvel et al52. Consumption of alcohol and tobacco have been proved to be positively associated with OSCC recurrence and poor prognosis, but have no significant association with PD-L1 expression on TCs4. Previous study using PD-L1 antibody to detect the PD-L1 expression found that non-smokers or non-alcohol consumers express higher frequency of PD-L1 expression12. Interestingly, the protective effect of PD-L1 expression was observed in females for OS and DSS, which did not weaken with smoking, or drinking1. It has been reported that PD-L1 is overexpressed in never-smokers and never-drinkers (NSND) who undergo the immunosuppressive therapy with pembrozulimab. The enrichment of T-cell activation, interferon-γ(IFN-γ) and PD-1 signalling were observed in OSCC from NSND. Overexpression of PD-L1 induced by IFN-c, the unique feature in NSND, was associated with rate of TILs. It could be inferred that PD1/PD-L1 blockade enhanced the immunity response that provided potential benefit of PD1/PD-L1 inhibition in NSND39. In addition, our meta-analysis also revealed that PD-L1 upregulation in TC was associated with clinicopathological features involving advanced stage, N+ status, in favour of the findings of previous reports19,24,28,30.
To date, the prognostic value of PD-L1 in OSCC cells has been identified in several meta-analysis. Consensus of them indicated that expressions of PD-L1 in TCs had protective effects on OS with no significance52-56, which was different from our study identifying prognostic role of PD-L1 expression in TCs was insignificant and unfavorable. As for DFS, four meta-analysis suggested that expression of PD-L1 was associated with better survival regardless of no statistical significance53-56. However, a previous study hold the opposite perspective that expression of PD-L1 probably lead to reduced survival52. In terms of DSS, our results revealed that PD-L1 expression exhibited a trend towards worse survival on DSS, which in line with the results reported by Troiano et al and Lenouvel et al.52,56 There was a significant association for worse LRFS in tumors characterized by high PD-L1 expression on TCs in our study. Geum et al. suggested locoregional recurrence had a significantly lower survival rate, and simultaneously, was significantly correlated with PD-L1 expression3. The remarkable advantages of this study was not only the abundant studies that was already enrolled in previous meta-analysis, but also in the first exclusively comprehensive evaluation of the oral compartments and immunostaining patterns as confounding factors for prognostic role of PD-L1 expression in OSCC.
The role of PD-L1 in the prognosis of OSCC have been undetermined due to expectable disparities in tissue of origin, location of staining, variations in immunostaining patterns (ie, assay cutoffs, antibody clones)9. In our study, high PD-L1 expression in TCs was associated with an disoperative effect on prognosis in DFS when only Asian patients of OSCC were analyzed. In OS and DSS of Asian patients, high PD-L1 expression on TC also had a tendency to increase the risk of poor prognosis. This might be attributed to the discrepancies in diet and lifestyle between the Asia and the non-Asia.
Of the selected studies, it was not difficult to find that tongue as the subsite of OSCC was more common than other subsite, suggesting that tongue might be an important site affecting the prognosis and clinicopathological factor of OSCC. In this study, we firstly compared the PD-L1 expression in TCs between tongue and other sites of OSCC patients, and the results showed no significant difference of any clinicopathological factors between the two groups. Next, we pooled the studies restricted only to tongue of OSCC with respect to prognosis and, found a significant relationship between the high expression of PD-L1 in TCs and worse DSS and OS, and the heterogeneity simultaneously decreased from 81% to 0%. Previous study have demonstrated that PD-L1 overexpression in TCs of the tongue and the floor of the oral cavity was associated with with a worse OS1. It was reasonable to speculate that discrepancies of the results in regard to the role of PD-L1 expressions in prognosis was attributed to skewed distribution of cancer location.
The use of different antibodies clones for immunohistochemistry resulted in discrepant results of association between PD-L1 expression and prognosis in OSCC. A trial in lung cancer comparing four anti-PD-L1 antibodies (22C3, 28-8, SP142, E1L3N) found that the patients detected as PD-L1-positive cases was similar for all antibodies except for SP142 which was only 50% as much as the other three of them57. Interestingly, SP142 showed more sensitive detection efficiency than 22C3 in another trail lung cancer58. This study extracted antibody of PD-L1 including 22C3, 28-8, 5H1, E1L3N, SP142 as different subgroups. The results indicated that OS was worse in 5H1, but greater in 22C3 in OSCC patients with high PD-L1 expression in TCs. A study using two different anti-PD-L1 antibodies (clones E1L3N and 22C3) to evaluate the prognostic significance of tumor PD-L1 expression in OSCC. Consequently, significant results of DSS only appeared in 22C3 but not in E1L3N12, in contrast to our findings demonstrating DSS was more likely to get worse with high PD-L1 expression detecting by E1L3N. It was not ignored that the positive expression ratio of tumor PD-L1 expression in OSCC tested by E1L3N was lower than 22C3 whether the cut off is ≥ 1% or ≥ 10%12. Subgroup analysis of our study revealed that almost no intra-heterogeneity was found in group of E1L3N and 22C3 in aspects of OS, DFS. Combined with our results, in OSCC, it can be believed that 22C3 seemed to be more delicate than E1L3N, and the PD-L1 expression in TCs appeared to have the protective role in OS via 22C3 as well as a damaging effect on DSS via E1L3N. Therefore, using different antibodies to evaluate the relationship between PD-L1 expression and prognosis probably lead to various results. Due to the lack of studies comparing the various antibodies specific to OSCC, our results needed to be interpreted with caution and were expected to be validated by more high-quality studies.
We evaluated the influence of PD-L1 expression on prognosis by using separate scoring methods. In this study, both DSS and DFS showed a significant result at a 5% cut-off of PPC. As for OS, 5% cut-off of 4 scores cut-off of H score presented a reduced survival. The cut-off value of 5% was frequently chosen in many clinical trials focusing on targeted anti-PD-L1 therapies. PD-L1 expression was associated with worse prognosis for OS and LRFS when a 5% cut-off of positive cells was applied28. Previous study aimed to explore the effects of expression of PD-L1 on survival rates and showed that PD-L1 expression were unrelated to in DFS and OS with TPS using different cutoffs of 1%, 5% and 10%2, in support of the data reported by Wirsing et. al22. One study9, using cutoffs of 5% of TPS, PD-L1 positivity was significantly associated with a better prognosis in DSS and OS, but this association was disappeared when confounded other factors8. PPC is defined as PD-L1-staining cells (TCs or TILs) divided by the total number of each type of cells, of which difference from TPS or CPS is that the denominator of the calculation formula of them is the total number of viable TCs. Therefore, a lower PPC score than TPS or CPS did not accurately represent the effect of immune checkpoints on TCs under the same conditions. Unlike with TPS, CPS reflecting an aggregate score of TC and IC evaluated expression of PD-L1in TC, as well as the impact of different IC on the tumor microenvironment. A randomized three-arm phase III KEYNOTE-048 trial was conducted in head and neck cancers (HNSCC) immunotherapy59,60, PFS and OS were tested in three groups including the CPS ≥20, the CPS ≥1, and the total population. In comparison with EXTREME regimen, patients undergoing the treatment of pembrolizumab plus chemotherapy or pembrolizumab monotherapy significantly improved OS and PFS, and the protective effect was progressively declining in the three groups. The percentage of HNSCC tumor cells expressing PD-L1 was 85% of when CPS was ≥159, which decreased to 50% when measured using TPS61. The response rate of pembrolizumab seemed to be increased in higher levels of PD-L1 expression, and CPS was more excellent in ability to predict response to anti-PD1 therapy comparing with TPS62. This prompted the FDA to approve the use of pembrolizumab monotherapy against HNSCC only in patients with a CPS score ≥1, which drew interest from two studies focusing on the prognostic value of PD-L1 expression recorded by TPS or CPS in OSCC. One study11 revealed the protective effect of PD-L1 expression was found on DSS and OS when positivity based on TPS (>5%) was used, and was observed on OS with the positivity defined as CPS >1. In the other study9, no statistical significance was found in the association between PD-L1 and survival in either CPS or TPS. Patients with CPS>1 showed a trend towards improved survival, while TPS>1% seemingly represented opposite trend. It need to be emphasized that tumours were classified into into four groups based on the presence of PD-L1 positivity in TC and IC, and the total prevalence of PD-L1 expression in both TC and IC or only in TC account for 72% in OSCC and showed good response to immunotherapy. Consequently, it should consider the combination of PD-L1 expressing in both TC and IC when observing PD-L1 of response to anti-PD1 therapy.
Immune cells as components of the immune environment, including CD4+ ,CD8+ TILs, have shown to be correlated with PD-L1 expression and play an important role in the mechanisms of immune response evasion of OSCC63. Several studies reported PD-L1-expressing tumor cells correlated positively to increased infiltration of CD4+ and CD8+TILs2,8,22,40, whereas other studies did not show any correlation11,25. Our study indicated that the PD-L1 high expression group was significantly associated with high infiltration by PD-1, CD8+ TILs, strengthening that the activity of CD8+TILs was reduced through interaction of PD-L1 expression in TCs activating expression of PD-1 on CD8+ TILs. In recent years, TILs were gradually considered as a predictor for the response of solid tumors to anti-PD1 therapy64,65. However, the prognostic value of PD-L1 in TILs was not fully understood. Our study on prognosis showed that the PD-L1 expression on TILs did not reveal a relationship with survival, which was consistent with a previous study11. Inversely, one study reported that longer OS and LRFS were appeared in the group of overexpression of PD-L1 in TILs35. Subramaniam et al.49reported that low TIL PD-L1 expression was significantly correlated with reduced LRFS, but this association was disappeared confounded other factors. This was consistent with what was reported previously in a study that high expression of PD-L1 in TILs was associated with better OS of OSCC without the confounding factors of nodal metastases1. It has been called adaptive immune resistance that a high TIL infiltration appeared simultaneously with PD-L1 positive TCs. The strong correlation between TILs and tumor PD-L1 staining implied PD-L1 may have been induced via enhanced T cell production of IFN-γin the same way as in tumour cells7.
Our study has several limitations. First, the prognostic values of several studies were estimated rather than provided directly. The actual association between PD-L1 expression and prognosis was masked on account of potential confounding factors. Second, the evaluation of the role of PD-L1expression in TILs on DSS, DFS, PFS and LRFS were not available to performed as a result of the lack of relevant data.
Conclusions
Current study indicated the expression of PD-L1 in TCs was relatively high in female patients, non-smoker, non-drinkers. The PD-L1 upregulation in TCs in cancers of advanced stage, N+ status was observed. The significant association of high PD-L1 expression in TCs in prognosis was only seen in LRFS. Although high PD-L1 expression in TCs had a trend toward significance of worse prognosis in OS, DSS and DFS in the initial analysis, the associations became significant based on stratified analysis of hypothesized confounders. It was worth noting that high expression of PD-L1 in TCs of tongue of OSCC was associated with worse DFS, OS. Future research on the ability to predict response to anti-PD1 therapy should focus on methods for PD-L1 immunostaining and various compartments of the oral cavity. In addition, TILs were shown to be independent of clinicopathological factors and prognosis, and relevant studies were expected to further confirmed.