Correlation between classic driver oncogene mutations in EGFR, ALK, or ROS1 and 22C3-PD-L1 ≥50% expression in lung adenocarcinoma

Abstract

Introduction

Targeted somatic genomic analysis (EGFR, ALK and ROS1) and PD-L1 tumor proportion score (TPS) by immunohistochemistry (IHC) are used for selection of 1^st-line therapies in advanced lung cancer; however, the frequency of overlap of these biomarkers in routine clinical practice is poorly reported.

Methods

We retrospectively probed the first 71 lung adenocarcinoma-patient pairs from our institution analyzed for PD-L1 IHC using the clone 22C3 pharmDx kit and evaluated co-occurrence of genomic aberrations along with clinical-pathologic characteristics.

Results

Surgical resection specimens, small biopsies (transbronchial or core needle), and cytology cell blocks (needle aspirates or pleural fluid) were tested. PD-L1 TPSs of ≥50% were seen in 29.6% of tumors. Of 19 tumors with EGFR-mutations, ALK-FISH positivity, or ROS1-FISH positivity, 18 had PD-L1 TPS <50% versus only 1 tumor with PD-L1 TPS ≥50% (p=0.0073). PD-L1 TPS ≥50% tumors were significantly associated with smoking status compared to PD-L1 TPS <50% tumors (p=0.0111); but not with patient gender, ethnicity, tumor stage, biopsy site, or biopsy type/preparation.

Conclusions

PD-L1 IHC can be obtained in routine clinical lung cancer specimens. TPS of ≥50% seldom overlaps with presence of driver oncogenes with approved targeted therapies. Three biomarker-specified groups of advanced lung adenocarcinomas can now be defined, each paired with a specific palliative first line systemic therapy of proven clinical benefit: 1) EGFR/ALK/ROS1-affected with matched tyrosine kinase inhibitor (~20% of cases), 2) PD-L1-enriched (TPS ≥50%) with anti-PD-1 pembrolizumab (~30% of cases), or 3) biomarker negative (i.e. EGFR/ALK/ROS1/PD-L1 negative) with platinum doublet chemotherapy with/without bevacizumab (~50% of cases).

INTRODUCTION

The landscape for management of advanced non-small-cell lung cancers (NSCLCs) has been revolutionized by the systematic identification of disease-specific characteristics that can be paired with precision systemic therapies. Testing for somatic mutations/rearrangements in the epidermal growth factor receptor (EGFR), anaplastic lymphoma kinase (ALK), and ROS proto-oncogene 1 (ROS1) are now mandated as part of routine pre-treatment evaluation in patients diagnosed with advanced lung adenocarcinomas (ACs). When paired with the appropriate oral tyrosine kinase inhibitors (TKIs), patients in these molecularly defined subsets experience improved and durable outcomes, lesser toxicity, and better quality of life as compared to conventional platinum doublet chemotherapy [1–3]. More recently, the anti-programmed death 1 (PD-1) immune checkpoint inhibitor, pembrolizumab, has been approved by the United States Food and Drug Administration (FDA) for first line systemic therapy in advanced NSCLCs with programmed death ligand 1 (PD-L1) tumor proportion scores (TPS) of ≥50% using the immunohistochemistry (IHC) clone 22C3 [4]. The systemic therapy of choice for advanced NSCLCs of adenocarcinoma histology without a classic driver oncogene aberration (i.e. EGFR/ALK/ROS1 negative) and with PD-L1 TPS <50% remains a platinum doublet with/without bevacizumab in patients with adequate performance status and end-organ function [1–3].

However, the use of genomic and immune biomarkers in routine clinical care has generated a number of questions not well answered by the large published clinical trials to date. What is the overlap of PD-L1 TPS ≥50% with the presence of classic targetable driver oncogene aberrations? How should one best select the appropriate first line therapy based on the tumor’s biomarker profile? Can diagnostic NSCLC specimens obtained in routine clinical practice (i.e. small biopsy or cytology cellblock) be used for immune biomarker analysis? We postulate that classic driver oncogene aberrations and high PD-L1 expression do not often coexist, generating distinct subgroups of patients which may allow for optimal pairing of systemic therapies with disease characteristics. The current report depicts our real-world experience with the first consecutive 71 tumor-patient pairs analyzed for PD-L1 IHC using the FDA-approved companion diagnostic, clone 22C3 pharmDx kit. Our findings support the aforementioned postulation and indicate that a variety of routine clinical specimens may be acceptable for successful biomarker analysis.

METHODS

Tumor and data collection

Patient and tumor pairs followed at Beth Israel Deaconess Medical Center (BIDMC) with a diagnosis of lung cancer were recorded through an ongoing Institutional Review Board-approved study, as previously reported [5–7]. Pathologic data, tumor genotype, PD-L1 TPS, and clinical characteristics were amassed from retrospective chart extraction. Data was collected and managed using the REDCap electronic data capture held at BIDMC. The cut-off for data collection was December 19, 2016.

Tumor genomic analyses

Tumor genotype was performed by analyzing EGFR (Sanger sequencing of exons 18–21), ALK (fluorescence in situ hybridization [FISH] break-apart probe), ROS1 (FISH break-apart probe), and KRAS (sequencing of codons 12–13) in tumor samples, as described previously [5, 7]. The aforementioned tests are bundled within the rapid tumor genotype panel outsourced to a commercial vendor by our hospital. More comprehensive genomic profiling was not directly evaluated within this analysis but methods were described previously [5, 7].

Tumor PD-L1 analysis

PD-L1 IHC testing was performed at Integrated Oncology/LabCorp (New York, NY) using the PD-L1 clone 22C3 pharmDx kit and Dako Automated Link 48 platform. PD-L1 TPS was calculated as the percentage of at least 100 viable tumor cells with complete or partial membrane staining. The TPS interpretation was provided by the commercial vendor pathologist.

Statistical methods

Fisher’s exact test was used to compare categorical variables. All p-values reported are two-sided, and tests were conducted at the 0.05 significance level.

RESULTS

Baseline clinical and pathologic characteristics in tumors with or without PD-L1 TPS ≥50%

Of the 71 lung AC tumor specimens analyzed, 21 (29.6%) had PD-L1 TPS ≥50% and 50 (70.4%) had PD-L1 TPS <50% (Table 1, Figure 1A). PD-L1 TPS ≥50% was seen significantly more frequently in lung ACs in smokers (former or current) as compared to never smokers (p=0.0111, Table 1). PD-L1 TPS ≥50% was not associated with patient gender, ethnicity, tumor stage, biopsy site, or biopsy type/preparation (Table 1).

Table 1.

Clinical, pathologic and genomic characteristics of lung adenocarcinomas tested for PD-L1 immunohistochemistry (IHC) using the clone 22C3 pharmDx kit by a commercial vendor

Characteristic (n)	PD-L1 IHC TPS ≥50% (n=21)	PD-L1 IHC TPS <50% (n=50)	p-value
Gender
Women	8	31	0.0743
Men	13	19
Smoking history
Never smoker	2	21	0.0111
Smoker	19	29
Ethnicity
White	16	35	0.7742
Non-white	5	15
Stage
I–III	5	13	1.0000
IV/recurrent	16	37
Biopsy tissue origin
lung	6	19	0.5882#
node (thoracic and extra)	5	12
pleura	4	6
bone/soft tissue	2	7
liver	0	2
brain	1	4
other	3	0
Biopsy type/preparation
surgical resection	7	14	0.7768^
small biopsy	7	18
FNA cell block	4	12
effusion cell block	3	6
Genomic aberrations
EGFR mutation	0	13	0.0073*
ALK FISH positive	1	3
ROS1 FISH positive	0	2
KRAS mutation	7	16
EGFR/ALK/ROS1/KRAS neg	13	16
PD-L1 IHC score
0%	0	30	<0.0001+
1–24%	0	17
25–49%	0	3
50–74%	10	0
75–100%	11	0

TPS, tumor proportion score; FNA, fine needle aspirate; neg, negative test result;

^#lung vs. other;

^{^}surgical specimen vs. other;

^*EGFR/ALK/ROS1 positive vs. negative;

⁺PD-L1 TPS <50% vs. ≥50%. All p-values using Fisher’s exact test.

Figure 1. PD-L1 tumor proportion score (TPS) immunohistochemistry (IHC) results in 71 lung adenocarcinomas.

A. Pie chart of PD-L1 IHC TPS. B. Representative Hematoxylin and Eosin (H&E) and PD-L1 IHC using clone 22C3 in different tumor materials: 1) cytology cell blocks from pleural fluid, 2) cytology cell blocks from lymph node aspirates, 3) small core biopsies, and 4) surgical resection specimens. TBNA, transbronchial needle aspirate; TPS, tumor proportion score.

Representative examples of PD-L1 IHC staining on different types of clinical specimens, including cytology cell blocks from pleural fluid, cytology cell blocks from lymph node aspirates, core needle biopsies and surgical specimens are displayed on Figure 1B.

PD-L1 TPS and presence of classic driver oncogene aberrations

Of the 71 lung AC specimens analyzed, 13 had EGFR mutations, 4 had ALK rearrangements and 2 had a ROS1 rearrangement (Table 1, Figure 2). Of these 19 tumors with EGFR mutations, ALK FISH positivity, or ROS1 FISH positivity, 18 had PD-L1 TPS <50% (Table 1, Figure 2A), as compared with only 1 tumor with PD-L1 TPS ≥50% (p=0.0073). The only tumor with PD-L1 TPS ≥50% and a classic driver oncogene aberration harbored an ALK rearrangement (Figure 2B). Other oncogene alterations are depicted on Figure 2.

Figure 2. Co-occurrence of genomic aberrations in EGFR, ALK, ROS1 and PD-L1 IHC TPS.

A. Graphic pie chart representation of the frequency of driver oncogene mutations (EGFR, or ALK, or ROS1) with approved therapies in tumors with PD-L1 tumor proportion score (TPS) of <50%. B. Graphic pie chart representation of the reported frequency of driver oncogene mutations (EGFR, or ALK, or ROS1) with approved therapies in tumors with PD-L1 TPS of ≥50%. * all tumors were tested for aberrations on EGFR, or ALK, or ROS1 (as detailed on the Methods) and only tumors without positive results are included in this group.

Overall expression of PD-L1 TPS

PD-L1 TPS results provided by the commercial vendor are depicted in Table 1. Notably, the most common PD-L1 TPS was 0% (n = 30), followed by 1–24% (n=17), and few tumors had intermediate low expression PD-L1 TPSs of 25–49% (Table 1). The proportion of tumors with PD-L1 TPS of ≥50% was divided in each of two quartiles: 50–74% (n = 10) and 75–100% (n = 11).

DISCUSSION

To the best of our knowledge, our single institution experience with targeted tumor genotyping in combination with PD-L1 IHC for characterization of NSCLCs in a non-investigational setting is one of the first reports to evaluate the day-to-day applications of vetted biomarkers in this patient population. Prior to now, most published experience with PD-L1 characterization in NSCLC using the 22C3 pharmDx assay has been in large clinical studies of pembrolizumab and restricted to tissue obtained from FFPE core or surgical biopsies; further, several of these studies excluded patients whose tumors harbored actionable driver oncogene mutations [4, 8, 9]. Our results are unique in demonstrating that the FDA-approved companion diagnostic for pembrolizumab - the PD-L1 clone 22C3 pharmDx kit - can generate actionable PD-L1 results in routine clinical specimens, including cytology specimens from FFPE cell blocks of fine needle aspirates or effusion specimens (Table 1 and Figure 1B). As many diagnostic specimens obtained for use in routine clinical practice may be of this variety (i.e. cytology specimens as opposed to surgical resection specimens or core biopsies), our findings merit further validation to optimize biomarker assessment and therapeutic planning in the real-world setting.

Our finding that 29.6% of lung ACs had a PD-L1 TPS of ≥50% is consistent with prior reports using this same antibody platform and in larger cohorts; PD-L1 TPS of ≥50% were reported in 24.9% to 30.2% of advanced NSCLCs in the phase I–III trials (KEYNOTE-001, 024 and 042) of pembrolizumab [4, 8, 9]. In one of the largest published screening cohorts for PD-L1 using the 22C3 pharmDx assay to date in the KEYNOTE-024 trial, the frequency of overlap between common driver oncogene aberrations (i.e. in EGFR or ALK) and PD-L1 TPS of ≥50% was just 6% (30/500) [4]. In our own single center experience, high PD-L1 expression and abnormalities in EGFR, ALK, and ROS1 seldom overlapped: only 1 tumor that was ALK-affected had concurrent PDL1 TPS of ≥50% (4.8% of overlap), and all specimens harboring abnormalities in EGFR or ROS1 had a PD-L1 TPS of <50% (Figure 2). It is likely that the true frequency of overlap in larger cohorts will be on the order of less than 10%. Appropriately, the FDA label for pembrolizumab was written to highlight that tumors with EGFR mutations or ALK rearrangements should not receive first line pembrolizumab due to the durable response rates that exceed 60–70% when patients with these tumors receive approved first line EGFR- or ALK- directed TKIs [1–3, 10–12]; these numbers far surpass the response rates published for conventional platinum doublet chemotherapy or immune checkpoint inhibitor monotherapy, even in tumors with the highest level of PD-L1 TPS of ≥50% (around 25–30% and 40–45%, respectively) [4].

PD-L1 IHC scores using different epitopes and measurement methods are also somewhat predictive of response to the PD-1 inhibitor nivolumab [13] and the PD-L1 inhibitor atezolizumab in the management of advanced lung ACs [14]. Notably, these other PD-L1 IHC platforms have yet to receive companion diagnostic designation by the FDA. The harmonization of PD-L1 testing remains a serious unmet need, as there are notable variations in the antibodies and staining platforms used, definitions of positivity, and assay methodologies (i.e. assessment of tumor cells versus tumor-infiltrating immune cells). To the latter end, the “Blueprint PD-L1 IHC Assay Comparison Project” provided initial results demonstrating that the extent and intensity of PD-L1-stained tumors was relatively comparable between the 22C3, 28-8 and SP263 IHC assays but not the SP142 IHC assay [15].

In summary, our real-world experience with the use of genomic and immunologic biomarkers in routine advanced lung AC samples demonstrates that multiple types of diagnostic clinical tissue samples can be successfully utilized, that the frequency of PD-L1 TPS ≥50% using the 22C3 pharmDx assay is approximately 30% in this population, and that PD-L1 TPS ≥50% seldom overlaps with the presence of actionable driver oncogenes (i.e. EGFR, ALK, ROS1). On the basis of this evolving paradigm of disease characterization, we can now define three distinct subsets of patients with advanced lung AC (Supplementary Figure 1A) and with highly relevant therapeutic implications (Supplementary Figure 1B): 1) EGFR/ALK/ROS1-affected (~20% of cases, based on large multi-institutional cohorts [1]), 2) PD-L1-enriched (TPS ≥50%) in ~30% of cases, and 3) biomarker negative (i.e. EGFR/ALK/ROS1/PD-L1 negative) in ~50% of cases.

Decades after the initial studies of cytotoxic chemotherapy first demonstrated benefit in physically fit patients with molecularly undefined tumors, nearly half of all patients with advanced lung AC now have the opportunity to receive a specific therapy which may afford brisk and/or durable response and with lesser toxicity as compared to conventional platinum doublet chemotherapy [1–3, 10–12]. Further translational research will help define additional predictive biomarkers in NSCLC that will help further improve outcomes for our patients - particularly for the remaining half of patients for whom no precision first line therapy is available.