Association of PD‐L1 expression with driver gene mutations and clinicopathological characteristics in non‐small cell lung cancer: A real‐world study of 10 441 patients

Abstract Background Programmed death ligand‐1 (PD‐L1) expression is a well‐known predictive biomarker of response to immune checkpoint blockade in non‐small cell lung cancer (NSCLC). However, there is limited evidence of the relationship between PD‐L1 expression, clinicopathological features, and their association with major driver mutations in NSCLC patients in Latin America. Methods This retrospective study included patients from Argentina with advanced NSCLC, and centralized evaluation of PD‐L1 expression concurrently with genomic alterations in the driver genes EGFR, ALK, ROS1, BRAF, and/or KRAS G12C in FFPE tissue samples. Results A total of 10 441 patients with advanced NSCLC were analyzed. Adenocarcinoma was the most frequent histological subtype (71.1%). PD‐L1 expression was categorized as PD‐L1 negative (45.1%), PD‐L1 positive low‐expression 1%–49% (32.3%), and PD‐L1 positive high‐expression ≥50% (22.6%). Notably, current smokers and males were more likely to have tumors with PD‐L1 tumor proportion score (TPS) ≥50% and ≥ 80% expression, respectively (p < 0.001 and p = 0.013). Tumors with non‐adenocarcinoma histology had a significantly higher median PD‐L1 expression (p < 0.001). Additionally, PD‐L1 in distant nodes was more likely ≥50% (OR 1.60 [95% CI: 1.14–2.25, p < 0.01]). In the multivariate analysis, EGFR‐positive tumors were more commonly associated with PD‐L1 low expression (OR 0.62 [95% CI: 0.51–0.75], p < 0.01), while ALK‐positive tumors had a significant risk of being PD‐L1 positive (OR 1.81 [95% CI: 1.30–2.52], p < 0.01). Conclusions PD‐L1 expression was associated with well‐defined clinicopathological and genomic features. These findings provide a comprehensive view of the expression of PD‐L1 in patients with advanced NSCLC in a large Latin American cohort.


INTRODUCTION
Treatment with immune checkpoint inhibitors (ICIs) directed against the PD-1/PD-L1 axis has revolutionized cancer treatment, especially achieving substantial success in the management of patients with advanced non-small cell lung cancer (NSCLC). 1 The expression of PD-L1, measured by immunohistochemistry as the tumor proportion score (TPS), is defined as the percent of PD-L1-positive tumor cells in the tumor tissue.PD-L1 expression is the primary clinically essential predictive biomarker for anti-PD-1/PD-L1 treatment efficacy in NSCLC.3][4][5] However, PD-L1 expression is incomplete and imperfect as a stand-alone biomarker since only a subgroup of patients has long-term clinical benefit and survival when treated with immune checkpoint inhibitors. 68][9] However, the main factors associated with its expression at baseline are not fully understood.
It is well known that NSCLC is a heterogeneous disease.Nonsquamous tumors in particular are characterized by subsets of driver genomic alterations capable of being druggable by tyrosine-kinase inhibitors, including EGFR, KRAS, BRAF, MET, and ERBB2 mutations or ALK, ROS1, RET, and NTRK genomic rearrangements. 10Critically, the intricate interplay between genomic alterations and PD-L1 expression in NSCLC is currently under investigation.2][13] Nevertheless, the association between these factors has only been examined in a limited number of studies, leading to a gap in understanding.Furthermore, inconsistencies have been observed in certain meta-analyses that have attempted to evaluate this relationship.8][19] However, validating this hypothesis has proven challenging as most studies investigating the clinicopathological features of NSCLC and PD-L1 expression have primarily focused on patients from North America, Europe, and Asia, while Latin American and African populations have been notably underrepresented. 20,21iven that there is a significant interest in a better understanding of the role of immunotherapy in oncogenic driven-NSCLC, it becomes crucial to comprehensively assess the factors linked to PD-L1 expression.This understanding could provide valuable insights into the mechanisms underlying primary response or resistance to immunotherapy.Thus, our study aimed to explore the potential association between PD-L1 expression and major driver gene alterations, including EGFR, KRAS, BRAF, ALK, and ROS1, as well as clinicopathological features.This investigation was conducted on a large cohort of patients with advanced NSCLC from a Latin American country.

Study population
This retrospective study included consecutive patients with advanced NSCLC, and effective evaluation of PD-L1 expression concurrently with analysis of EGFR, ALK, ROS1, BRAF, and/or KRAS mutations from available formalin-fixed paraffin-embedded (FFPE) samples.Patients were selected from February 2018 to September 2021 from the institutional databases of Biomarkers Inc., which centrally analyzed lung cancer tissue samples from multiple hospitals and cancer centers in Argentina.Data, including basic demographic as well as pathological characteristics, were collected from the same database based on the reports provided by clinicians.

PD-L1 assay
PD-L1 immunohistochemistry (IHC) testing was performed using the PD-L1 clone 22C3, and the pharmDx kit, and Dako Automated Link 48 platform (Dako).PD-L1 expression measured as TPS was calculated as the percentage of positive cells in at least 100 viable tumor cells with complete or partial membrane staining assessed by four experienced pathologists (LL, MB, MAB, and GGR). 22In instances of discrepancy, a consensus meeting involving a fifth senior pathologist was convened to reach an agreement, and the kappa coefficient was used.A PD-L1 TPS <1% was defined as negative, and a PD-L1 TPS ≥1% was considered positive.Additionally, PD-L1 positive samples were stratified as low PD-L1 expression (PD-L1 TPS 1%-49%) and high PD-L1 expression (PD-L1 TPS ≥50%).

Driver mutation analyses
The assessment of EGFR, BRAF, and KRAS p.G12C genomic alterations was performed by extracting genomic DNA from FFPE tumor tissue using a QIAMP mini DNA kit (Qiagen) at the QIAcube instrument (Qiagen), according to the manufacturer's instructions.EGFR mutations were detected using the AmoyDX EGFR 29 mutations detection kit (AmoyDx) at the Rotor-Gene Q instrument (Qiagen), and EGFR mutation analysis kit (Entrogen) at Cobas z480 instrument (Roche). 23,24Both kits are designed for real-time PCR assays for the qualitative detection of mutations of the EGFR gene (LRG_304t1) (Supplementary Methods).BRAF mutations were assessed using BRAF codon 600 mutation analysis real-time PCR kit (Entrogen) of exon 15 (Supplementary Methods). 25KRAS p.G12C mutation was tested by AmoyDX KRAS mutation detection kit, realtime PCR Kit (AmoyDx) at the Cobas z480 instrument (Roche). 26ALK fusion testing was performed with a fully automated IHC assay using a D5F3 clone (Ventana Roche).D5F3 was additionally assessed using the OptiView enhanced detection and amplification system.ALK-positive cases were interpreted using Ventana ALK (D5F3) CDx Assay (Roche). 27ROS1 fusion testing was performed in ALK-negative cases.ROS1 fusion was analyzed by IHC (D4D6 clone, Cell Signaling Technology) and confirmed by FISH (ZytoLight SPEC ROS1 Dual Color Break Apart Probe) (Supplementary Methods). 28,29

Statistical analysis
Categorical variables are summarized using frequency and percentage, while continuous variables are described by their median, standard deviation, or interquartile range (IQR).Associations with qualitative variables were assessed using the Chi-square or Fisher's exact test, and for quantitative variables, analysis of variance (ANOVA) or the Kruskal-Wallis test was employed.Pairwise comparisons between groups were conducted using the Wilcoxon test, and p-values were adjusted using the Holm method.The multivariate analysis of PD-L1 expression utilized a logistic regression model, and the results are reported as adjusted odds ratios (OR).All statistical analyses were performed using R software (version 4.3.0).Two-tailed tests and p-values <0.05 were used to determine statistical significance.
Tumors with KRAS, BRAF, and ROS1 genomic alterations were excluded from the multivariate analysis due to the limited number of cases with available data.However, a univariate association between PD-L1 expression and KRAS status was found since KRAS-mutated tumors were more likely PD-L1-high than PD-L1 low (19.9% vs. 12.3%, p = 0.003).Contrarily, no statistical associations were found between ROS1 or BRAF and PD-L1 status (p = 0.052 and p = 0.240, respectively).

DISCUSSION
The administration of immune checkpoint inhibitors targeting anti-PD-1 or PD-L1 has demonstrated enhanced survival outcomes in individuals with metastatic NSCLC.However, it is important to note that treatment benefits have only been observed in a specific subgroup of patients, and the identification of predictive factors for immunotherapy response remains under intensive research. 4Although imperfect, PD-L1 expression is the only biomarker currently utilized in clinical practice to guide treatment decisions regarding immunotherapy in advanced NSCLC.Nonetheless, the expression of PD-L1 in NSCLC exhibits considerable variability, and a comprehensive understanding of the factors influencing its expression is yet to be achieved.
Numerous studies have investigated the association between PD-L1 expression and clinicopathological features in NSCLC.5][16] To the best of our knowledge, our study is unique and represents the largest single-region real-world cohort of a centralized PD-L1 analysis in advanced NSCLC.
1][32][33] Remarkably, the predominant subgroup observed was the PD-L1 negative category.5][36][37] Notably, in line with our findings, although not entirely validated, certain studies have reported a positive correlation between elevated PD-L1 expression and nonadenocarcinoma histology. 38,39ased on our findings, patients who were current smokers exhibited a higher likelihood of having elevated PD-L1 expression.Consistent with this observation, previous studies have reported similar results, corroborating our findings. 40Notably, tobacco smoking is commonly linked to T cell exhaustion and the upregulation of PD-1, which ultimately contributes to immune evasion. 41,42Lung cancer cases in smokers have been noted to exhibit a substantial load of neoantigens, heightened immunogenicity, and upregulation of PD-L1. 43[46][47] In our study, a remarkably high PD-L1 expression (≥80%) was notably more prevalent among males than females, potentially attributable to the higher incidence of cigarette smoking in males.While this explanation appears to be the most plausible, the relationship between sex and PD-L1 expression remains inadequately elucidated.Notably, a meta-analysis conducted by Zhang et al. revealed that PD-L1 expression was elevated in males, pooling the results from 11 444 patients. 16Opposite, no significant correlation between positive PD-L1 expression and gender was found in a pooled analysis including 3128 cases performed by Yang et el. 15Conversely, the correlation between PD-L1 expression and gender could potentially be influenced by sex hormones, as emerging evidence suggests that these hormones have the capacity to regulate numerous immune-related genes, including those involved in the PD-1/PD-L1 pathway. 48,49he validation of the association between elevated PD-L1 expression and male patients holds significant clinical relevance, as multiple phase III studies investigating first-line immune checkpoint inhibitors in advanced NSCLC have demonstrated that anti-PD-1/anti-PD-L1 monotherapy exhibits greater efficacy in men compared to women. 50onhomogenous PD-L1 expression between primary tumor and metastatic sites has previously been reported. 51,52he multivariate analysis of our study showed that tissue samples from distant nodes were more likely to have high PD-L1 expression than those from the primary tumor and regional lymph nodes.Given it was not tested in paired primary and metastatic samples, it is not possible to have strong methodological conclusions.However, it reinforces that PD-L1 expression in lung cancer could be heterogeneous and dynamic, hence the reliability and feasibility of the PD-L1 expression on a single biopsy specimen, as a reference for immuno-oncology treatment, remains controversial. 53,54n this new era of genomic characterization of NSCLC, a deeper understanding of the molecular factors associated with PD-L1 expression can help elucidate mechanisms of primary response and resistance to immunotherapy.In this context, evidence has characterized that EGFR-mutated tumors have a lower tumor mutation burden (TMB), but the association with PD-L1 expression remains unclear. 38,55- 59Evans et al. analyzed the PD-L1 expression among 10 005 patients with NSCLC in the UK and found that classical EGFR mutations were associated with lower rates of PD-L1 expression, and nonclassical EGFR mutations were associated with higher rates. 60Contrarily, a meta-analysis conducted by Zhang et al., including 47 studies and 11 444 patients, showed that high PD-L1 expression was associated with EGFR mutations. 16In another meta-analysis performed by Li et al., analyzing 50 studies and 11 383 patients, the pooled results revealed that PD-L1 expression was related to EGFR wild-type tumors. 40Taking advantage of our large homogenous cohort, the multivariate analysis demonstrated that EGFR-mutated tumors more likely had low PD-L1 expression with a very low median TPS.
Likewise, preclinical studies have demonstrated that ALK translocation and its downstream signaling pathways can drive PD-L1 expression. 13,613][64] Our results suggest that ALK-positive tumors have a significantly higher risk of PD-L1 positive expression.
Our study revealed that KRAS p.G12C-mutant tumors were more likely to have a high PD-L1 expression in the univariate analysis.Studies on cell lines revealed that KRASmutated NSCLC can be regulated by MAPK and partially by STAT3 signaling pathways. 9,65,66As a consequence, similar to our findings, several studies and meta-analyses confirmed the positive correlation between PD-L1 expression and KRAS mutation in NSCLC. 38,40,67,68inally, the analysis of uncommon driver mutations is usually limited by patient numbers.A small number of studies have revealed that BRAF mutation, particularly p.V600E, is associated with a high level of PD-L1 expression. 68,69Concerning ROS1 fusion, no association was found in our study in line with previous reports. 59,70r results should be analyzed with caution considering study limitations.First, the retrospective nature of our analysis resulted in incomplete data for some patients.Second, PD-1 was unsuccessfully evaluated in 14% of cases as a result of poor tissue quantity and quality.The unsuccessful evaluation of PD-L1 expression was estimated at around 10% in the real-world setting and 5% in clinical trials. 34,37,71,72Third, analysis of uncommon driver alterations was usually limited by the low number of patients.In our cohort, not all cases were tested for the entire mutational profile (EGFR, KRAS, ALK, ROS1, and BRAF) which might have affected the multivariate analysis.The main reason for this discrepancy was the heterogeneous biomarker testing reimbursement for each case.Fourth, given that PD-L1 expression in NSCLC could be heterogeneous and dynamic, the association between a potentially changing variable (PD-L1 expression), with a constant variable (the mutational profile), might limit the reliability, and reproducibility of the results.However, our study had an advantage over other studies since all the samples were processed and read in the same institution with a consistent antibody, technique, and experienced pathologists.
In conclusion, this is the largest and most homogeneous study analyzing PD-L1 expression and its association with clinicopathological and genomic alterations in a Latin American cohort.In summary, we found that males and current smokers, as well as tumors with nonadenocarcinoma histology, KRAS mutations, and tissue samples from distant nodes, were associated with high PD-L1 expression.In contrast, tumors with EGFR mutations were more likely to have low PD-L1 expression.This study, together with the current evidence, is ultimately intended to understand the potential associations between PD-L1 expression with clinicopathological relevance and genomic alterations.
Patient characteristics and driver alterations according to PD-L1 expression.
F I G U R E 1 Flow chart of included patients and biomarker analysis.*Patientsexcludedrepresented samples nontested or not evaluable.TA B L E 1Note: χ 2 -test, One-way ANOVA, and Fisher's exact test.p-value was calculated for patients with available data.Abbreviations: ANOVA, analysis of variance; NSCLC NOS, non-small cell lung not otherwise specified; PD-L1, programmed death ligand-1; SD, standard deviation.a Percentages were calculated considering the available data and molecular test performed.Nonsmoker was defined as those who have never smoked, or who have smoked less than 100 cigarettes in their lifetime.
Frequency of cases harboring concomitant alteration.T A B L E 2 Impact of driver alterations status on the PD-L1 expression.Multivariate logistic regression.Multivariate logistic regression model; data presented by adjusted odds ratio.KRAS, BRAF, and ROS1 analyses were excluded given the low number of cases with complete data.Abbreviations: CI, confidence interval; NOS, not otherwise specified; OR, odds ratio; PD-L1, programmed death ligand 1. Note: