Prognostic implication of PD‐L1 polymorphisms in non‐small cell lung cancer treated with radiotherapy

Abstract Background To investigate the impact of programmed death‐ligand 1 (PD‐L1) polymorphisms on the prognosis of non‐small cell lung cancer (NSCLC) patients treated with curative radiotherapy. Methods Four single nucleotide polymorphisms (SNPs) (rs822336G>C, rs822337T>A, rs822338C>T, and rs2297136A>G) in the PD‐L1 gene were evaluated in 124 NSCLC patients. Clinical stage was I in 28, II in 17, and III in 79 patients. Fifty‐seven patients received radiotherapy alone, including 28 patients who received stereotactic body radiotherapy. Sixty‐seven patients received sequential or concurrent chemoradiotherapy. Risk factors for survival outcomes were analyzed with the log‐rank test and multivariate Cox proportional hazards models. Results The rs822336GC+CC genotype was associated with better overall survival (OS) (hazard ratio [HR] = 0.60, 95% confidence interval [CI] = 0.37–0.97, p = 0.036) and regional failure‐free survival (RFFS) (HR = 0.32, 95% CI = 0.14–0.76, p = 0.009), compared with rs822336GG genotype. The rs822337TA+AA genotype was associated with better OS (HR =0.54, 95% CI = 0.34–0.88, p = 0.014), progression‐free survival (PFS) (HR = 0.64, 95% CI = 0.41–0.99, p = 0.046), and RFFS (HR = 0.38, 95% CI = 0.17–0.81, p = 0.013), compared with rs822337TT genotype. Three SNPs (rs822336, rs822337, and rs822338) were in linkage disequilibrium. Combined GTC and GTT (GT*) haplotype was associated with significantly worse OS (p = 0.018), PFS (p = 0.044), and RFFS (p = 0.038), compared with those with other combined haplotypes. Patients with diplotypes of two GT* haplotypes showed significantly worse OS (p = 0.023) and RFFS (p = 0.014) than those with other diplotypes. Conclusions These findings suggest that PD‐L1 polymorphisms could be predictive markers for NSCLC patients receiving radiotherapy.


| INTRODUCTION
Radiotherapy has been used to treat early to advanced non-small cell lung cancer (NSCLC) with a curative intent. To be brief, while radiotherapy can be an alternative to surgery for early stage NSCLC patients who cannot undergo surgery for any reason, radiotherapy combined with chemotherapy is recommended for locally advanced NSCLC patients. Recently, a randomized controlled trial revealed that consolidation immunotherapy after concurrent chemoradiotherapy increased overall survival (OS) of locally advanced NSCLC patients. 1,2 Along with the growing interest in immune checkpoint inhibitors in cancer treatment, many investigations in the field of radiotherapy have focused on programmed death-ligand 1 (PD-L1) protein which plays an important role for cancer cells to escape immune surveillance. 3 Some researchers explored the impact of baseline PD-L1 expression on post-radiotherapy outcomes in NSCLC, but the results were contradictory. 4,5 Others reported that changes in PD-L1 expression and density of CD8+ tumorinfiltrating lymphocytes after radiotherapy were related to the prognosis of NSCLC patients treated with preoperative concurrent chemoradiotherapy, with no association between the baseline PD-L1 status and changes after radiotherapy. 6,7 Considering that upregulation of PD-L1 expression after irradiation led to radioresistance in animal tumor models, 8,9 it can be presumed that the capability of PD-L1 expression after radiotherapy would be important to determine the prognosis, rather than the baseline status. PD-L1 is encoded by the PD-L1 gene located on chromosome 9 at position p24.1, whose polymorphisms have been reported to be predictive markers in NSCLC patients who received chemotherapy or surgery. [10][11][12][13] However, there has been no report about the influence of PD-L1 polymorphisms on the post-radiotherapy prognosis in any type of cancer.
Therefore, we hypothesized that PD-L1 polymorphisms may affect the prognosis of NSCLC patients receiving radiotherapy. To examine this hypothesis, we examined the relationship between PD-L1 polymorphisms and treatment outcomes in NSCLC patients treated with radiotherapy.

| Patients
From November 2010 to May 2018, 305 patients with pathologically confirmed clinical stage I-III NSCLC were treated with curative radiotherapy in our institution. Clinical TNM stage was evaluated according to the AJCC 8th staging system. 14 Of them, 152 patients had available genomic DNA samples for single nucleotide polymorphisms (SNPs) genotyping. After excluding the patients who received a total equivalent dose in 2 Gy fractions of less than 54 Gy (N = 8), undertook surgical resection after radiotherapy without the evidence of disease recurrence (N = 1), or had follow-up information of less than 12 months without the evidence of disease recurrence (N = 19), this study enrolled 124 patients for analyses ( Figure 1). The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Institutional Review Board of the Kyungpook National University Chilgok Hospital (2019-01-025). The need for informed consent was waived in consideration of the retrospective study design.

| SNP selection and genotyping
Among five PD-L1 SNPs which were selected in a previous study, 12 four SNPs (rs822336G>C, rs822337T>A, rs822338C>T, and rs2297136A>G), which were applicable to the iPLEX ® Assay and MassARRAY ® System (Agena Bioscience), were genotyped. The linkage disequilibrium (LD) status was determined with Haploview ver. 4.2 software. 15 Then, the haplotype frequencies were estimated using the Phase ver. 2.1.1 software package. 16

| Statistical analysis
The distribution of clinicopathologic factors according to genotype was compared with Pearson's chi-square test, Fisher's exact test, Student's t-test, and Mann-Whitney U-test as appropriate. Survival rates were estimated from the first day of treatment to the date of the event or the last follow-up with the Kaplan-Meier method: overall survival (OS), progression-free

| Characteristics
Patient and tumor characteristics are shown in Table 1.
The median age was 70 years (range: 45-87); 107 patients were male. TNM stage was I in 28, II in 17, and III in 79 patients. Twenty-eight patients with cT1-4N0 received stereotactic body radiotherapy (SBRT) without any adjuvant treatment. Ninety-six patients with stage I-III received intensity-modulated radiotherapy or three-dimensional conformal radiotherapy (referred to as the non-SBRT subgroup). Combination therapy in the non-SBRT subgroup was radiotherapy alone in 29, sequential chemoradiotherapy in 24, and concurrent chemoradiotherapy in 43 patients. The most common chemotherapy regimen for sequential or concurrent chemoradiotherapy was paclitaxel-cisplatin doublet. The details of radiotherapy and chemotherapy are summarized in Table S1. None of the patients received immune checkpoint inhibitors after radiotherapy without evidence of recurrence.

| Clinical factors and outcomes
With a median follow-up of 29 (range: 4-116) months, OS, PFS, LFFS, RFFS, and DMFS rates of all patients were 58.9%, 29.4%, 57.8%, 66.4%, and 56.7% at 2 years, respectively. In all patients, age (≤70 years vs. >70 years), sex, TNM stage (I-II vs. III), histologic type (adenocarcinoma vs. others), and radiotherapy modality (SBRT vs. non-SBRT) were significant risk factors for at least one of the survival outcomes in the univariate analyses (Table S2). Age and sex were significantly associated with RFFS; TNM stage with OS, PFS, RFFS, and DMFS; histologic type with OS and LFFS; and radiotherapy modality with PFS, LFFS, and DMFS. Chemotherapy was not associated with any of the survival outcomes.

| Allele frequencies of PD-L1 SNPs
The frequencies of the four SNPs are shown in Table 2.
The distribution of clinical factors including sex, age, TNM stage, histologic type, total radiation dose, radiotherapy modality, and chemotherapy was not related to any of the four SNPs, except total radiation dose for rs822337 (data not shown).

| PD-L1 polymorphisms and outcomes
In the multivariate analyses adjusted for age, sex, TNM stage, tumor histology, radiotherapy modality, and chemotherapy, rs822336 and rs822337 were significantly related to outcomes (Table 3 and Table S4). The rs822336GC+CC genotype was associated with better OS (hazard ratio As for the haplotypes of rs822336G>C-rs822337T>A-rs822338C>T, the combined GTC and GTT (GT*) haplotype was related to worse OS (p = 0.018), PFS (p = 0.044), and RFFS (p = 0.038), compared with those with combined other haplotypes (Table 4). GT* were defined as bad haplotypes, while others were defined as good haplotypes. Patients with at least one of the good haplotypes showed better OS (HR = 0.58, 95% CI = 0.36-0.93, p = 0.023) and RFFS (HR = 0.39, 95% CI = 0.19-0.83, p = 0.014) than those with two bad haplotypes (Table 4). Thus, diplotypes with two GT* haplotypes, GTC/GTC, and GTC/GTT, were classified as high-risk diplotypes (vs. low-risk diplotypes for others). Survival curves according to the risk groups of diplotypes are presented in Figure 2G-I.

| Subgroup analyses
The differences in survival outcomes between the risk groups of diplotypes were analyzed in the SBRT and non-SBRT subgroups, respectively. In the SBRT subgroup, the patients with high-risk diplotypes showed significantly worse PFS and RFFS in the multivariate analyses adjusted for sex, age, histologic type, and cT stage, with a tendency toward worse DMFS ( Figure S1). In the non-SBRT subgroup, high-risk diplotypes had borderline significance for OS and RFFS in the multivariate analyses adjusted for sex, age, TNM stage, histologic type, and use of chemotherapy ( Figure S2).

| DISCUSSION
This study investigated whether PD-L1 polymorphisms could predict the prognosis in NSCLC patients treated with radiotherapy. Among four SNPs evaluated, rs822336 and rs822337 were significantly related to treatment outcomes. In diplotype analyses of the three SNPs with LD (rs822336, rs822337, and rs822338), the patients with highrisk diplotypes showed significantly worse OS and RFFS than those with other diplotypes. These findings imply that the PD-L1 polymorphisms could be utilized as predictive markers for NSCLC patients receiving radiotherapy. Currently, immunogenic cell death is considered an important mechanism of tumor cell death after radiotherapy, Note: All the results were from multivariate analyses adjusted for sex, age, TNM stage, histologic type, radiotherapy modality, and the use of chemotherapy. Haplotypes of GT* (GTC and GTT) were defined as bad haplotypes (htB), while others (GAT and CAT) were defined as good haplotypes (htG). Abbreviations: CI, confidence interval; DMFS, distant metastasis-free survival; HR, hazard ratio; LFFS, local failure-free survival; OS, overall survival; PFS, progression-free survival; RFFS, regional failure-free survival.
in addition to direct DNA damage. 3,17,18 Damage-associated molecular patterns released by irradiation activate dendritic cells, presenting tumor neoantigens, and activating CD8+ T cells. [18][19][20] Radiotherapy also promotes T-cell infiltration into tumors by upregulating the expression of adhesion molecules on endothelial cells and the release of cytokines. 21 The activated CD8+ T cells are known to be important to reduce or eradicate the primary tumor, or distant metastasis after radiotherapy. 18,22 However, IFNγ produced by CD8+ T cells after radiotherapy can upregulate PD-L1 expression on tumor cells, which in turn leads to radioresistance. 8,9 The programmed cell death 1 (PD-1)/PD-L1 axis has an important role in immune evasion of tumor cells. 3 PD-L1 expressed on tumor cells binds to PD-1 on effector T cells, resulting in suppressing the cytotoxic activity of T cells. 23 However, the significance of the baseline expression of PD-L1 remains controversial in NSCLC patients. 4 PD-L1 expression at baseline has been reported to be either associated with no prognostic significance, better prognosis, or worse prognosis after surgery, chemotherapy, or radiotherapy. 7,11,12,[24][25][26] In addition, the clinical importance of radiation-induced upregulation of PD-L1 expression is controversial, even though the expression of PD-L1 has been reported to increase after radiotherapy in patients with various tumors including NSCLC. 6,7,[27][28][29] In soft tissue sarcoma, the rate of positive PD-L1 expression (>1%) in tumor cells and tumor-associated macrophages increased after preoperative radiotherapy, and positive PD-L1 expression on tumor-associated macrophages was significantly related to worse DMFS. 28 In cervical cancer patients, patients with positive PD-L1 expression (≥1%) after 12 Gy of carbon-ion radiotherapy showed a significantly better PFS compared to those without PD-L1 expression. 29 In addition, the PD-L1 expression level after preoperative chemoradiotherapy (≥50% vs. <50%) was not associated with OS after surgery in NSCLC. 7 In the current study, the rs822336GG genotype, rs822337TT genotype, and high-risk diplotypes of rs822336-rs822337-rs822338 in the PD-L1 gene were significantly related to worse OS and RFFS. An rs822336G-rs822337T haplotype of the PD-L1 gene was reported to show a significantly increased promoter activity than an rs822336C-rs822337A haplotype in a luciferase assay, suggesting rs822336G-rs822337T is associated with an increased PD-L1 expression. 12 The poor prognosis of our patients with high-risk diplotypes might be caused by having a pair of the rs822336G-rs822337T haplotype. Fujimoto et al. 6 revealed that NSCLC patients with increased PD-L1 expression after preoperative chemoradiotherapy showed a significantly worse OS than those with unchanged or decreased PD-L1 expression. Therefore, further studies are needed to investigate the relationship between the genotypes of the three SNPs and PD-L1 expression, especially radiotherapy-induced expression level change, along with the prognostic significance. It would be worthy to investigate whether PD-L1 polymorphisms may be utilized in identifying patients who would benefit from the combination of radiotherapy and PD-1/PD-L1 inhibitors.
This study has some limitations as a retrospective study. This study enrolled a relatively small number of patients, whose stages and treatment modalities were various. The status of PD-L1 expression at baseline was not tested in most patients. However, the consistent significance of the effect of PD-L1 genotypes on the prognosis, regardless of patient-, tumor-, and treatment-related factors, could support the reliability of our results. In particular, the patients with high-risk diplotypes experienced significantly poorer regional control in both the SBRT and non-SBRT subgroups.
In summary, our results suggest that PD-L1 polymorphisms could be predictive markers for NSCLC patients receiving radiotherapy. As far as we know, this is the first study to report the prognostic value of PD-L1 polymorphisms in NSCLC patients treated with radiotherapy. Further studies are required to confirm our findings and to investigate the possible mechanisms in the relationship between PD-L1 polymorphisms and failures after radiotherapy.

ETHICAL APPROVAL STATEMENT
The study was approved by the Institutional Review Board of the Kyungpook National University Chilgok Hospital (2019-01-025).