PD-L1 expression and its correlation with clinicopathological and molecular characteristics in Chinese patients with non-small cell lung cancer

Little is known about the relationship between programmed cell death-ligand 1 (PD-L1) expression and histologic and genetic features in real-world Chinese non-small cell lung cancer patients. From November 2017 to June 2019, tumor tissues were collected from 2674 non-small cell lung cancer patients. PD-L1 expression was detected with immunohistochemistry using the 22C3 and SP263 antibodies, and patients were stratified into subgroups based on a tumor proportion score of 1%, 1% to 49%, and ≥ 50%. Genetic alterations were profiled using targeted next-generation sequencing. In the total population, 50.5% had negative PD-L1 expression (tumor proportion score < 1%), 32.0% had low-positive expression (1%–49%), and 17.5% had high-positive expression (≥50%). The PD-L1 positive rate was 39.0% in squamous cell carcinomas and 53.6% in adenocarcinomas. PD-L1 expression was higher in squamous cell carcinomas (P < .001) and lower in adenocarcinomas (P < .001). Of the overall patient population, 11.2% had Kirsten rat sarcoma viral oncogene (KRAS) mutations, 44.9% had epidermal growth factor receptor (EGFR) mutations, 2.1% had BRAF V600E mutations, 0.3% had MET exon 14 skipping mutations, 5.4% had anaplastic lymphoma kinase translocations, and 0.9% had ROS proto-oncogene 1 translocations. Patients carrying ROS proto-oncogene 1 translocations (P = .006), KRAS (P < .001), and MET (P = .023) mutations had significantly elevated expression of PD-L1, while those harboring EGFR (P < .001) mutations had lower PD-L1 expression. In our study, PD-L1 expression was significantly higher in squamous cell carcinomas and lower in adenocarcinomas, and was positively associated with MET and KRAS mutations, as well as the wild-type EGFR gene state. Nonetheless, additional studies are needed to further validate those associations and determine the clinical significance for immune checkpoint inhibitors of these factors.


Introduction
Lung cancer is a leading cause of cancer-related deaths in China, [1,2] with non-small cell lung cancer (NSCLC) accounting for approximately 85% of cases. [3]A recent implementation of immune checkpoint inhibitors (ICIs), such as antibodies against programmed cell death-1 (PD-1) and programmed cell death-ligand 1 (PD-L1), has shifted the therapeutic paradigm for NSCLC.[6][7] In the first-line setting, pembrolizumab monotherapy demonstrated superior efficacy compared to chemotherapy in patients with advanced or metastatic NSCLC who JG, HY, and YZ contributed equally to this work.

The authors have no funding and conflicts of interest to disclose.
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.
tested negative for epidermal growth factor receptor (EGFR) sensitizing mutations or anaplastic lymphoma kinase (ALK) fusions and exhibited a PD-L1 tumor proportion score (TPS) of ≥ 1%. [4,5]n light of the fact that PD-L1 expression is the only US Food and Drug Administration-approved predictive biomarker for anti-PD-1/PD-L1 treatment of NSCLC, it is prevalent to investigate factors that affect PD-L1 expression and analyze their application prospects for ICIs.A meta-analysis showed that PD-L1 expression in NSCLC patients positively correlated with smoking, male gender, an adenocarcinoma histology, and wild-type EGFR, ALK, ROS proto-oncogene 1 (ROS1), and Kirsten rat sarcoma viral oncogene (KRAS) genes. [8]However, another meta-analysis conducted by Huang et al demonstrated that survival benefits from ICIs in NSCLC were not significantly associated with histology, central nervous system metastases, age, gender, or performance status. [9]Several additional studies have also demonstrated that the regulation of PD-L1 expression in NSCLC was influenced by oncogenic drivers, including but not limited to EGFR, ALK, ROS1, KRAS, TP53, MET, and PIK3CA. [10,11]Apart from single gene alteration, some investigations have endeavored to identify potential correlations between tumor mutational burden (TMB), a quantitative indicator of the prevalence of genic mutations, and PD-L1 levels.However, no conclusive evidence has been obtained thus far. [12,13]lthough several western studies provided evidence regarding the relationship between molecular phenotypes, clinicopathological factors, and PD-L1 expression in NSCLC patients, it is widely acknowledged that the genomic profiling of Chinese NSCLC is significantly different from that of TCGA, especially the mutation rates of EGFR, which can be up to approximately 50%. [14,15]To date, there are few relevant studies and evidence regarding the relationship between the above factors and PD-L1 expression in Chinese NSCLC patients.In this work, we performed next-generation sequencing and PD-L1 immunohistochemistry (IHC) on tumor tissues obtained from 2674 Chinese NSCLC patients.The objective was to explore the correlation between clinicopathologic, genomic patterns and PD-L1 expression, as well as their prognostic value in Chinese patients.

Patient and samples
Patients diagnosed with NSCLC at our institution from November 2017 to June 2019 were retrospectively included (Fig. S1, Supplemental Content, http://links.lww.com/MD/L190, which demonstrates the patient selection flowchart).The inclusion criteria were patients between the ages of 18 and 90, male or female; patients histologically confirmed NSCLC; patients had PD-L1 and 381-genes NGS testing results.The exclusion criteria were as follows: Patients only with plasma sample tested and patients with formalin-fixed paraffin-embedded (FFPE) tumor tissue block that older than 5 years.After applying both inclusions and exclusion criteria, 2674 patients with NSCLC were included in the study.The study was approved by the Ethics Committee of the Institutional Review Boards at Shanghai Chest Hospital.

PD-L1 IHC
FFPE tumor samples were subjected to PD-L1 IHC using the PD-L1 IHC 22C3 pharmDx assay (Agilent Technologies; catalog number, SK006) or the PD-L1 IHC SP263 assay (Roche Diagnostics GmbH; catalog number,740-4907/07208162001).The staining for 22C3 was performed on a Dako Link-48 autostainer system at Teddy Clinical Research Laboratory (Shanghai) Limited, and the staining for SP263 was performed on a Roche BenchMark Ultra platform at QIAGEN (Suzhou) Clinical Lab, following the manufacturer instructions.The proportion of tumor cells showing membranous staining was scored as TPS by the technical experts and then classified as negative, lowpositive, or high-positive PD-L1 expression based on the TPS score of < 1%, 1% to 49%, or ≥ 50%.

Genetic profiles and TMB estimation
Next generation sequencing targeting 381 cancer-related genes was conducted at 3D Medicines, Inc.,.Genomic DNA was extracted from FFPE tumor and paired normal tissue samples.TMB was defined as the number of synonymous and nonsynonymous somatic mutations, stop gain/loss, and splicing variants and indels per Mb in examined coding regions.

Statistical analyses
The prevalence of PD-L1 positivity was compared between subgroups using Chi-square analysis or Fisher exact test, as appropriate.All analyses were performed using R version 3.6.1,and P values <.05 were considered statistically significant.
The prevalence of PD-L1-positive tumors was 83.9% for squamous cell carcinoma and 16.1% for adenocarcinoma (Table 1).In the squamous cell carcinoma subgroup, 123 patients (26.3%) had strong PD-L1 expression, and 216 (25.3%) had low expression.Patients with squamous cell carcinomas were more likely than those with non-squamous cell histology subtypes to express PD-L1 (P < .001).Among patients with adenocarcinoma, 344 patients (26.3%) exhibited highpositive PD-L1 expression.There were no significant differences in the proportions of PD-L1 expression in the adenosquamous and NSCLC-not otherwise specified subgroups.

EGFR and KRAS mutation subtypes
Since aberrations in EGFR exons 19 to 21 and KRAS G12X were most clinically relevant in NSCLC, the study subsequently focused on mutations occurring in these regions.Of 1201 NSCLC patients with EGFR mutations, exon 19 deletions, exon 21 L858R, and exon 20 T790M were identified in 543, 511, and 64 patients, respectively.PD-L1 expression was observed in 63.7% and 66.4% of patients carrying exon 19 deletion and L858R, respectively.Whereas the proportion of PD-L1 expression in patients carrying T790M was 95.2%, less than other EGFR mutation subtypes.Additionally, 8 forms of KRAS G12X variants were identified in this study: G12A, G12C, G12D, G12F, G12R, G12S, G12V and G12D.The correlation between PD-L1 expression and EGFR, KRAS hotspot mutations was summarized in Table 3.
The correlation between mutations in distinct signaling pathways and the level of PD-L1 expression was also investigated.The expression difference of PD-L1 was the most significant in the case of Wnt pathway mutations (3% with PD-L1 positive vs 10% with PD-L1 negative), suggesting that Wnt pathway inhibitors might enhance the efficacy of ICIs by promoting T cell infiltration into tumors (Fig. S3, Supplemental Content, http:// links.lww.com/MD/L192, which demonstrates the relationship between the mutations in different signaling pathways and the level of PD-L1 expression).
The mutation burden in lung and other cancers is also considered predictive of benefit for checkpoint inhibition.In this study, the median TMB of high-positive, low-positive, and negative PD-L1 expression subgroups was 8.1/MB, 6.7/MB, and 5.6/MB, respectively (Fig. S4, Supplemental Content, http://links.lww.com/MD/L193, which demonstrates TMB across PD-L1 expression in tumor).

Discussion
This work represents a large-scale investigation into the relationship between PD-L1 expression and clinicopathologic and molecular characteristics in Chinese NSCLC patients.Using TPS ≥ 1% as the cutoff, our cohort had a PD-L1-positive rate of 56.1%, which is consistent with previous reports on Asian populations. [16]Currently, pembrolizumab is the first-line treatment for metastatic NSCLC patients with PD-L1 TPS ≥ 50% in some countries, including China.In the present study, 17% of patients had a TPS of ≥ 50%, suggesting they may be eligible for pembrolizumab monotherapy.In the multinational KEYNOTE-001, −010, and −024 studies, which used the PD-L1 IHC 22C3 pharmDx assay, 67% and 28% of NSCLC patients exhibited PD-L1 TPS of ≥ 1% and ≥ 50%, respectively. [17]Although the rates observed in these clinical trial cohorts are slightly higher compared to the current study, our results are consistent with the rates reported in the Asia Pacific subpopulation of the global EXPRESS study, which used the 22C3 pharmDx assay, where TPS ≥ 1% and ≥ 50% were reported at 53% and 25%, respectively. [18]The difference in PD-L1 expression rates between our study and the clinical trial cohorts could potentially be explained by the higher prevalence of EGFR-mutated adenocarcinoma in Chinese NSCLC patients, as this subset is known to have lower PD-L1 levels.
The literature contains conflicting data on the association between NSCLC histology and PD-L1 expression, with some reports finding no significant differences, [19] while others report higher expression in adenocarcinoma, [8,20] or higher expression in squamous cell carcinoma. [10,16]In the present study, PD-L1 expression was more frequently observed in squamous cell carcinomas than in adenocarcinomas, particularly in the TPS ≥ 50% subgroup.This finding may help explain the results of the CHECKMATE-078 trial, which enrolled 504 NSCLC patients (451 from China), and showed that nivolumab extended overall survival compared to docetaxel in patients with squamous (hazard ratio [HR] = 0.61) and non-squamous (HR = 0.76) NSCLC. [21]Taken together, these findings suggest that squamous cell carcinoma may exhibit better sensitivity to anti-PD-1/ PD-L1 agents than adenocarcinomas in Chinese NSCLC.
EGFR mutations are the most frequent driver mutations in Chinese NSCLC, and a series of clinical studies have *Some patients had more than one above mutation and are represented in more than one column.EGFR = epidermal growth factor receptor, KRAS = Kirsten rat sarcoma viral oncogene, PD-L1 = programmed cell death-ligand 1. that ICIs may have limited efficacy in EGFR-mutant NSCLC. [22,23]Nonetheless, the relationship between EGFR mutation status and PD-L1 expression remains controversial. [16,24]In the current study, 28.5% of patients with PD-L1 expression carried mutated EGFR, whereas patients with wildtype EGFR showed higher positive and high-positive PD-L1 expression rates and higher TMB.Another study revealed that tumors carrying mutated EGFR tended to have low PD-L1 and CD8 + T cell levels, resulting in weaker immunogenicity. [25]his may partially explain why NSCLC patients with mutant EGFR respond poorly to ICIs.In our cohort, 11.3% of the EGFR-mutated patients had a PD-L1 TPS of ≥ 50%.A previous study found that high PD-L1 expression in NSCLC patients with mutant EGFR was associated with de novo resistance to EGFR-TKIs (tyrosine kinase inhibitors). [26]Our study showed that the majority (86.4%) of observed EGFR mutations were accounted for by exon 19 deletions and exon 21 L858R substitutions.In another study, it was found that compared with wildtype EGFR, exon 19 deletions were associated with significantly worse response rates and progression-free survival (PFS) after anti-PD-1/PD-L1 therapy in NSCLC patients, and the outcomes for patients carrying L858R EGFR and wild-type EGFR were similar, [27] which may be explained by the correlation of PD-L1 expression and different EGFR mutation types in our study.PD-L1 high-positive expression was lower in patients with the EGFR T790M mutations observed in our study.A recent report found that NSCLC patients with wild-type EGFR who were treated with nivolumab after disease progression during EGFR-TKI therapy had longer PFS than patients carrying the T790M mutation (median PFS: 2.1 vs 1.3 months; P = .099),possibly due to lower PD-L1 expression in patients carrying the T790M mutation. [28]While NSCLC patients with L858R or rare EGFR subtype mutations who exhibited high levels of PD-L1 may show a potential benefit from anti-PD-1/PD-L1 therapy, it is crucial to recognize that multiple factors, such as TMB, the presence of specific immune cell populations, can influence the effectiveness of immune therapy.Further larger-scale, prospective research is needed to understand the interplay between these factors and evaluate the clinical implications of these findings.
Mutations in KRAS or MET were associated with high PD-L1 expression in our cohort, with high PD-L1 levels being 3 times more common in patients carrying mutant MET than wildtype MET.KRAS or MET mutations were also associated with higher TMB, indicating that mutations in these genes may promote tumor immunogenicity in NSCLC.A meta-analysis noted that ICIs improved the overall survival of previously treated NSCLC patients with mutant KRAS (HR = 0.64; P = .03),but not those with wild-type KRAS. [29]Different KRAS mutations encode proteins with distinct functions that influence downstream signaling.Patients carrying KRAS G12C and G13D mutations expressed higher levels of PD-L1 in the present study, suggesting they may show a superior response to immunotherapy.Maybe prospective trials will be necessary to determine whether MET and KRAS-TKIs as combination options with PD-L1/PD-1 inhibitors.In contrast, we found no association between PD-L1 expression level and BRAF V600E mutations or ALK and ROS1 translocations.
Amplification of the chromosomal region 9p24.1, which contains genes coding PD-L1 (CD274), PD-L2 (PDCD1LG2), and JAK2 (JAK2), has been linked to the overexpression of PD-L1 in cancer. [30]The current investigation reports that 3 patients carried CD274 and PDCD1LG2 amplifications without other driver mutations, and all 3 expressed high PD-L1 levels.These findings imply that PD-L2 might also have predictive value for immunotherapy in NSCLC patients.
Although the association between PD-L1 expression and clinicopathological characteristics in NSCLC has been studied extensively, the question remains controversial.Our data demonstrated that high PD-L1 expression was significantly associated with female gender, biopsied tumor samples, metastatic tumor samples, and freshly prepared FFPE samples.The differential expression of PD-L1 in primary and metastatic sites is widely acknowledged.One study found that PD-L1 expression was less common in primary NSCLC tumor samples compared with metastases measured both in histologic and cytological cohorts. [31]These data suggest that assessing PD-L1 expression in freshly collected tissue and from metastatic sites before treatment with anti-PD-1/PD-L1 agents is important.
There are several limitations to our study.Firstly, it is important to note that our study design is retrospective, which inherently introduces limitations in terms of data collection and potential biases in patient selection.Second, the International Association for the Study of Blueprint Project [32] suggests that the Dako 22C3 and Ventana SP263 assays are equivalent; however, using 2 assays in the present study may have affected the expression estimates.Finally, some data on baseline characteristics were missing or incomplete.

Conclusions
In conclusion, PD-L1 expression in Chinese NSCLC varied in terms of sampling methods, histologic types, oncogene, and PD-L1 (CD274)/PD-L2 (PDCD1LG2) alteration status.Given the association with PD-L1 expression, our findings support the potential for combination immunotherapy with MET, KRAS inhibitors, and CD274/PDCD1LG2 amplification, which may support eligibility for treatment with anti-PD-1/PD-L1 agents for patients with NSCLC.

Table 1
Correlations between clinicopathological features and PD-L1 tumor proportion score.

Table 2
Correlations between oncogenic driver mutation status and PD-L1 tumor proportion score.

Table 3
Correlations between EGFR and KRAS mutation subtypes and PD-L1 tumor proportion score.