Abstract
Background:
PD-L1 expression is the only FDA-approved biomarker for immune checkpoint inhibitors (ICIs) in patients with lung adenocarcinoma, but sensitivity is modest. Understanding the impact of molecular phenotype, clinical characteristics, and tumor features on PD-L1 expression is largely unknown and may improve prediction of response to ICI.
Methods:
We evaluated patients with lung adenocarcinoma in whom PD-L1 testing and targeted next-generation sequencing (MSK-IMPACT) was performed on the same tissue. Clinical and molecular features were compared across PD-L1 subgroups to examine how molecular phenotype associated with tumor PD-L1 expression. In patients treated with anti-PD-(L)1 blockade, we assessed how these interactions impacted efficacy.
Results:
1586 patients with lung adenocarcinoma had paired PD-L1 testing and targeted next-generation sequencing. PD-L1 negativity was more common in primary compared to metastatic samples (p <0.001). The distribution of PD-L1 expression (lymph nodes enriched for PD-L1 high; bones predominantly PD-L1 negative) and predictiveness of PD-L1 expression on ICI response varied by organ. Mutations in KRAS, TP53, and MET significantly associated with PD-L1 high expression (each p < 0.001, q < 0.001) and EGFR and STK11 mutations associated with PD-L1 negativity (p <0.001, q = 0.01; p = 0.001, q < 0.001, respectively). WNT pathway alterations also associated with PD-L1 negativity (p = 0.005). EGFR and STK11 mutants abrogated the predictive value of PD-L1 expression on ICI response.
Conclusion:
PD-L1 expression and association with ICI response varies across tissue sample sites. Specific molecular features are associated with differential expression of PD-L1 and may impact predictive capacity of PD-L1 for response to ICIs.
Keywords: PD-1, PD-L1, Immunotherapy, NSCLC
Introduction
Tumor expression of programmed death-ligand 1 (PD-L1) is an important but imperfect predictive biomarker for treatment with immune checkpoint inhibitors (ICIs). In patients with advanced lung adenocarcinoma, PD-L1 expression is the only FDA-approved biomarker for treatment with anti-programmed cell death protein 1 (PD-1) therapies. High PD-L1 expression (tumor cell expression ≥50%) is associated with improved responses to ICIs.[1–3] Nevertheless, a majority of PD-L1 high patients do not respond to ICIs and an important, albeit small, proportion of PD-L1 low/negative patients do respond to ICIs.[1–3] Improved understanding of the interrelationships between clinical and molecular features and tumor PD-L1 expression may lay the groundwork for optimizing biomarkers, understanding the biology of the tumor-immune interface, and facilitating more precise use of ICIs in the future.
Our current knowledge of PD-L1 expression is primarily derived from large randomized controlled clinical trials.[1, 2, 4–6] These trials tend to have uniform procedures for tissue collection and selective eligibility criteria, which may not fully reflect the routine patient population. In the ‘real world,’ there is more variation in tumor sampling, as well as more heterogeneity in clinical, pathologic, and molecular characteristics.[7] Few studies have assessed if clinical features influence PD-L1 expression and prior data on whether tumor sampling procedures such as type of sample (e.g. biopsy versus resection) and sample site influence PD-L1 expression are conflicting.[8–10]
Furthermore, emerging data suggests that specific molecular alterations correlate with PD-L1 expression and may influence the predictive utility of PD-L1 expression.[8, 11–13] However, these analyses have primarily focused on single-gene alterations or specific driver oncogenes in lung cancer.[5, 8, 11–14] Prior pre-clinical studies have shown a complex interplay between the molecular and immunologic pathways of the tumor microenvironment,[15] but a systematic exploration of the impact of molecular phenotypes on tumor PD-L1 has not been performed. We hypothesized that a comprehensive molecular analysis could reveal distinct subsets of somatic tumor features associated with patterns of PD-L1 expression.
To address this hypothesis, we examined more than 1500 patients with lung adenocarcinoma who had tumor PD-L1 testing and targeted next-generation sequencing (MSK-IMPACT) performed on the same tissue sample. Clinical and molecular features were compared across PD-L1 subgroups to examine how the molecular phenotype may interact with the immunologic phenotype of a tumor. A cohort of patients treated with anti-PD-(L)1 was also used to assess how these interactions may impact the likelihood of response to these therapies.
METHODS:
Patients
Following Memorial Sloan Kettering Cancer Center (MSK) institutional review board approval, patients diagnosed with lung adenocarcinoma at MSK in whom PD-L1 IHC testing and targeted next-generation sequencing via MSK-IMPACT[16] were performed on the same tissue sample (irrespective of receipt of PD-(L)1 therapy) were identified from the pathology database. Tumor tissue samples tested between April 2014 and February 2019 were included. Patients’ medical and pathology records were reviewed to extract demographic and pathology specific data.
To assess response to anti-PD-(L)1 therapy, we identified all patients with lung adenocarcinoma who began treatment (anti-PD-(L)1 +/− anti-CTLA-4) from April 2011 through December 2017 (database lock of April 1, 2019) with available PD-L1 expression (irrespective of available MSK-IMPACT testing) on a tissue sample prior to starting therapy (n = 148 with PD-L1 IHC only; n=346 with PD-L1 and MSK-IMPACT). This “ICI cohort” encompassed additional patients beyond our “biomarker cohort” (which was not limited to those who received immunotherapy) in order to maximize sample size for analysis (Supplementary Figure 1). Scans were reviewed by a thoracic radiologist (A.P, R.P-J, P.S) and efficacy was assessed by Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1; measurements were done prospectively in patients treated in clinical trials and retrospectively in patients treated as standard of care. In patients who received standard of care anti-PD-(L)1, the timing of imaging tests was not pre-specified but typically occurred approximately every two months, similar to patients on trial. Progression-free survival (PFS) and overall survival (OS) were calculated from the date the patient began treatment to the date of progression or date of death, respectively. Patients who had not progressed at the time of database lock were censored at the date of their last scan. Patients who had not died were censored at date of last contact.
PD-L1 testing
PD-L1 expression scores were reported as the percentage of tumor cells with membranous staining as determined by thoracic pathologists (J.L.S., N.R., W.D.T.). PD-L1 subgroups were defined as negative (PD-L1 < 1%), intermediate (PD-L1 1 – 49%), and high (PD-L1 ≥ 50%). PD-L1 IHC testing was performed using the E1L3N antibody (Cell Signaling) in the biomarker cohort. In the cohort of patients treated with anti-PD-(L)1, several antibodies were used, including E1L3N (n = 327), 22C3 (n = 89, DAKO), 28–8 (n = 70, DAKO), SP142 (n = 5, Ventana) and SP263 (n = 3, Ventana).
MSK-IMPACT Sequencing
The MSK-IMPACT assay was performed as previously described,[16, 17] for sequencing all exons and select introns of a custom gene panel of 341 (n = 13), 410 (n = 198), or 468 (n = 1375) genes. Somatic alterations, including mutations and copy number alterations, were identified. Somatic nonsynonymous tumor mutation burden (TMB) was normalized across panels of various sizes by dividing the total mutations by the coding region captured in each panel (0.897 megabases (Mb) for 341-, 1.017 Mb for 410-, and 1.139 Mb for 468-gene panel). The fraction of copy number altered genome (FGA) and whole genome doubling (WGD) were inferred from sequencing data using the FACETS algorithm (version v0.5.6) as previously described.[18, 19] Samples with purity estimates below 20% were excluded for FGA and WGD analyses.
Gene and Pathway Analysis
Individual genes were queried for distribution and enrichment among the predefined PD-L1 subgroups. Unless otherwise noted, analysis included previously described oncogenic/likely oncogenic variants and variants of unknown significance. Pathway lists were curated by comparing previously reported gene lists[20–24] with overlapping genes covered in the MSK-IMPACT panel (Supplementary Table 1).
Statistical Analysis
The Fisher’s exact test or Pearson chi-square was used to compare proportions. Between-group differences were examined using the Mann-Whitney U test (2 groups) and the Kruskal-Wallis exact test (3 groups). Frequencies of gene alterations within PD-L1 subgroups were considered significant with a p-value < 0.05 and, to reduce false discovery in multiple testing, q-value < 0.15. For subsampling analysis, gene alteration frequencies in PDL1-high were compared to randomly sub-sampled sets of 528 PDL1-negative cases. Q-values were determined for all genes for each comparison. Analysis was repeated 100,000 times and distribution of q-values determined in this empirical analysis are plotted for select genes. Correlations were examined by the Spearman rank correlation coefficients. For survival analysis, Kaplan-Meier curves were compared using the log-rank test. All reported p-values are two-sided. Statistical analyses were performed with GraphPad Prism software version 7 (La Jolla, CA, www.graphpad.com) and R version 3.3.3 software (www.r-project.org).
RESULTS:
Clinical features and PD-L1 expression
We identified 1586 patients with lung adenocarcinoma within our “biomarker cohort” and 494 patients within our “ICI cohort” (Supplementary Figure 1). The clinical characteristics of the patients stratified by PD-L1 expression (negative, intermediate, high) were generally similar in both cohorts (Table 1, Supplementary Table 2). PD-L1 expression did not significantly differ by age, sex, or smoking status.
Table 1.
Clinical characteristics of patients with paired PD-L1 + MSK-IMPACT testing.
| PD-L1 expression | High (n = 245) | Intermediate (n = 284) | Negative (n = 1056) | P-value | |||
|---|---|---|---|---|---|---|---|
| N = 1586 | No. | (%) | No. | (%) | No. | (%) | |
| Age (median, range) | 67 | (30–93) | 67 | (25–92) | 68 | (27–89) | 0.63 |
| Gender | |||||||
| Female | 154 | (63) | 170 | (60) | 678 | (64) | |
| Male | 91 | (37) | 114 | (40) | 379 | (36) | 0.41 |
| Smoking status | |||||||
| Ever | 185 | (76) | 196 | (69) | 730 | (69) | |
| Never | 60 | (24) | 88 | (31) | 327 | (31) | 0.13 |
| Procedure | |||||||
| Biopsy | 164 | (67) | 167 | (59) | 534 | (51) | |
| Resection | 81 | (33) | 118 | (41) | 523 | (49) | <0.001 |
| Sample type | |||||||
| Primary | 136 | (56) | 181 | (64) | 789 | (75) | |
| Metastatic | 109 | (44) | 103 | (36) | 268 | (25) | <0.001 |
Tumor sampling characteristics and PD-L1 expression
PD-L1 expression was significantly lower in tissue derived from the primary tumor versus a metastatic site (p < 0.001, Table 1). Among patients with primary tumor tissue sampling, there was lower PD-L1 expression when resection specimens were used compared to biopsy samples (p < 0.001, Table 1).
Further investigation into PD-L1 expression by anatomic site of tissue sampling revealed the distribution of PD-L1 expression varied modestly by organ (Figure 1a). Among metastatic sites, the proportion of cases with PD-L1 high expression was greatest in lymph node (PD-L1 high = 30%) and least in bone (PD-L1 high = 16%). Tumor purity, determined by visual assessment by pathologist or computationally estimated from molecular results using FACETS, did not vary with PD-L1 expression (Supplementary Figure 2).
Figure 1. PD-L1 expression and ICI response by anatomic site.
(A) The distribution of PD-L1 expression (high [≥50], intermediate [1–49%], negative [<1%]) by tissue sampling site. (B) Progression free survival (PFS) in PD-L1 high versus negative cases in which PD-L1 testing was performed in lung (log-rank HR 0.52 [95% CI 0.32–0.71], p<0.001) or distant (log-rank HR 0.56 [95% CI 0.36–0.88], p=0.005) metastasis tissue. (C) Progression free survival (PFS) in PD-L1 high versus negative cases in which PD-L1 testing was performed in lymph node (log-rank HR 0.67 [95% CI 0.37–1.21], p=0.16) or bone (log-rank HR 0.92 [95% CI 0.42–2.01], p=0.83) metastasis tissue.
Response to ICIs in PD-L1 subgroups by site of tumor samples was explored. PD-L1 status (PD-L1 high vs PD-L1 negative) was predictive, as expected, for response to ICI treatment from lung and distant metastatic sites (PFS HR 0.56, 95% CI: 0.43–0.74, p < 0.01 and HR 0.48, 95% CI: 0.31–0.73, p < 0.01, respectively, Figure 1b), but had decreased predictiveness from lymph node and bone (PFS HR 0.63, 95% CI:0.38–1.04, p = 0.04 and HR 0.79, 95% CI:0.40–1.58, p = 0.50, respectively Figure 1c). Similar to other reports, the predictiveness of PD-L1 from lymph nodes was further decreased (PFS: HR 0.69, 95% CI: 0.41–1.17, p = 0.10) when patients with ALK- and EGFR-mutations were removed from the analysis (Supplementary Table 3). [25]
Molecular features and PD-L1 expression
Summary molecular features
PD-L1 expression was modestly associated with increased TMB as both a categorical (chi-square p < 0.001) and continuous variable (spearman rho = 0.167, p < 0.001) (Figure 2a). PD-L1 expression did not correlate with FGA (spearman rho = 0.061, p = 0.056, Figure 2b) or rates of WGD (p = 0.07, Figure 2c).
Figure 2. Molecular features of PD-L1 subgroups.
(A) Comparison of TMB and PD-L1 expression as continuous variables (above, dots represent individual tumor samples; Spearman rho=0.167, p<0.001) and as categorical variables (below, donut plots characterize proportion of patients who are PD-L1 high [represented in green] within TMB subgroups [TMB low < 8 mt/mb, TMB intermediate ≥ 8 mt/mb and < 20 mt/mb, TMB high ≥ 20 mt/mb]. (B) Comparison of fraction of genome altered (FGA) and PD-L1 expression. Dots represent individual tumor samples. (Spearman rho 0.061, p=0.056) (C) PD-L1 expression by whole genome duplication (WGD) status (PD-L1 high green, PD-L1 intermediate dark blue, PD-L1 negative light blue).
Individual molecular alterations
Multiple altered genes were significantly associated with PD-L1 high or PD-L1 negative samples (Figure 3a–b, Supplementary Figure 3). Mutations in KRAS, TP53, and MET were enriched in PD-L1 high compared to PD-L1 negative samples (p < 0.001, q = 0.001; p < 0.001, q < 0.001; p < 0.001, q < 0.001; Figure 3a–b). Other notable genes enriched in the PD-L1 high compared to PD-L1 negative samples included ARID2, ARID1A, ARID1B, and ATM (p < 0.001, q = 0.03; p < 0.001, q = 0.03; p = 0.003, q = 0.07; p = 0.004, q = 0.11 Figure 3a–b). By contrast, mutations in EGFR and STK11 were associated with PD-L1 negativity (p < 0.001, q = 0.01; p = 0.001, q < 0.001 Figure 3a–b). Among commonly altered genes in lung adenocarcinoma, the proportion of PD-L1 high patients was greatest among MET alterations and least among STK11 mutations (Figure 3b). Even after subsampling PD-L1 negative cases in multiple simulations, these primary gene-level associations with PD-L1 remained (Supplementary Figure 4).
Figure 3. Differential PD-L1 expression based on alterations in individual genes and pathways.
(A) Frequency of altered individual genes within PD-L1 high vs PD-L1 negative expression subgroups. Genes labeled red were associated with significantly differential PD-L1 expression (q value <0.15). (B) Distribution of PD-L1 expression by commonly altered genes in lung adenocarcinoma. (C) Percentage of tumors harboring an alteration of individual pathways within PD-L1 subgroups (PD-L1 intermediate subgroup not shown).
Pathway alterations
Pathway analysis by PD-L1 subgroup (PD-L1 high vs PD-L1 negative) was performed to assess trends in alterations by PD-L1 subgroup. Tumors with alterations of the WNT signaling and PI3K pathways were more frequently PD-L1 negative (p = 0.005, p = 0.04), while alterations in antigen presentation, DDR, NOTCH, and SWI/SNF pathways were enriched with PD-L1 high expression (p = 0.02, p < 0.001, p = 0.02, p = 0.005, Figure 3c).
Interaction of molecular features and PD-L1 expression on benefit of ICI
Response to ICIs in PD-L1 subgroups (PD-L1 high vs PD-L1 negative) was influenced by the presence of somatic alterations (Figure 4, Supplementary Table 4). The predictive value of PD-L1 expression (high vs. negative) on benefit of ICI treatment was lost in patients with EGFR (Figure 4a), minimized in patients with STK11 mutant tumors (Figure 4b), while remained strong in KRAS and TP53 mutant tumors (Figure 4c–d). Patients with alterations in WNT pathway lost predictive value of PD-L1 expression, while those with RTK RAS, DDR, and PI3K pathway alterations retained the expected increased benefit of ICI treatment in the PD-L1 high vs negative subgroups (Figure 4e–f).
Figure 4. ICI outcomes by PD-L1 expression within mutant genes and altered pathway subgroups.
(A) Objective response rate (ORR), PFS, and OS within patients with EGFR mutations. (B) Objective response rate (ORR), PFS, and OS within patients with STK11 mutations. (C) Objective response rate (ORR), PFS, and OS within patients with KRAS mutations. (D) Objective response rate (ORR), PFS, and OS within patients with TP53 mutations. (E) Forest plot of progression-free survival within subgroups of individual altered pathways, comparing outcomes of patients with PD-L1 high vs negative tumors. (F) Forest plot of overall survival within subgroups of individual altered pathways, comparing outcomes of patients with PD-L1 high vs negative tumors.
Alterations in the gene encoding PD-L1 (CD274) were rare (6/1586). These included 2 amplifications, 2 deep deletions, and 2 mutations, all of which had negative PD-L1 expression except for 1 amplification (PD-L1 TPS = 50%). Nevertheless, 3 of 3 patients with CD274 alterations (2 amplifications and 1 mutation) who received immunotherapy have had ongoing partial responses to anti-PD-L1 treatment (PFS: 29 mos, 12mos, 3 mos).
DISCUSSION
This is the largest series to examine the clinical and molecular correlates of PD-L1 expression and their relationship to ICI response in lung adenocarcinomas. We found that distinct clinical and molecular features were associated with differential PD-L1 expression and ICI response. These findings may be used to guide clinical decisions regarding anatomic site of tissue sampling used to assess PD-L1 expression and interpretation of PD-L1 status in the context of genomic abnormalities.
Although most clinical characteristics had similar distributions of PD-L1 expression, tumor sampling site was an important exception. Metastatic tumor sampling was associated with high PD-L1 expression. This aligns with prior studies that have shown that advanced disease may be more likely to express PD-L1 than earlier stage disease,[8, 9, 26] yet our study also distinguishes PD-L1 expression among metastatic sites of disease not just the stage of the patient. In doing so, we found that PD-L1 is differentially expressed among metastatic sites and particularly enriched in lymph nodes, which may influence the predictive capacity of PD-L1.[25] It is possible that lymph node sampling may require a different scale to differentiate those patients who may most likely to benefit. Others have also demonstrated that metastatic sites such as the brain may have differential PD-L1 expression and decreased tumor lymphocyte infiltration than primary tumors.[26, 27] Further investigation is required to determine why there is variation among metastatic sites, and whether this may reflect the varied innate immune characteristics of these anatomic sites.[28] In cases where multiple tissue samples exist, or potential biopsy sites of equivalent safety are being considered, clinicians may consider prioritizing metastatic sites to maximize capturing PD-L1 expression and avoiding sites such as bone where PD-L1 expression is less reliably predictive.
In our examination of the relationship between tumor molecular phenotype and PD-L1 expression, summary molecular metrics such as TMB and aneuploidy appear largely independent of PD-L1 status. TMB has previously been shown to be an independent predictor ICI response[29–31] and we find that PD-L1 expression was only modestly increased in patients with high TMB. Separately, although aneuploidy has been shown to correlate with decreased responses to immunotherapy in melanoma and lung cancer, it is not broadly associated with PD-L1 expression.[29, 32, 33]
Importantly, we identified several molecular alterations that associated with differential PD-L1 expression. Alterations of the lung cancer molecular driver KRAS and the tumor-suppressor TP53 had the strongest correlations with high PD-L1 expression. In contrast, we found that STK11 and EGFR and WNT pathway alterations were associated with low PD-L1 expression and abrogate the predictive value of PD-L1 for ICI. Each of these findings align closely with recent independent results demonstrating similar molecular associations with PD-L1 expression,[34] affirming the strength of these conclusions.
Beyond to the consistency of these observations in clinical data, pre-clinical data also supports the potential biological connections between these molecular and immunological phenotypes in this report. For example, loss of TP53 was shown to generate adaptive activity via NF-kB signaling and production of inflammatory cytokines.[35] Similarly, RAS mutations induced changes in cytokine expression and immune-stimulatory transcription programs in multiple tumor models.[36, 37] Meanwhile, pre-clinical and pre-clinical data support that STK11[11, 38–41] and WNT/β-catenin[15, 42, 43] signaling alterations are associated with an immunosuppressed tumor phenotype. Loss of STK11 stimulates production of G-CSF, CXCL7, and IL-6 by the tumor, thereby recruiting neutrophils and blocking anti-tumoral cytotoxic T cells.[15, 38, 44] Recently, it was additionally shown that patients with STK11 alterations who received platinum chemotherapy with immunotherapy may not benefit from the addition of immunotherapy.[39] Overall, the results of this report bolsters emerging mechanistic hypotheses related to these pathways and provides groundwork for future functional studies.
There are important limitations to this work. The distribution of PD-L1 expression in our study is enriched for PD-L1 negativity, which differs from many clinical trial reports but is more generalizable across the ‘real-world’ clinical experience. The reports of PD-L1 distribution in ‘real world’ settings generally suggest a similarly high rate of PD-L1 negativity.[7, 9, 13, 45, 46] For example, a large European analysis from multiple medical centers (ETOP Lungscape cohort)[13] similarly found a rate of PD-L1 negativity of 57% and 3 smaller real-world cohorts described rates of PD-L1 negativity from 42%−72%.[9, 45, 46] Furthermore, site of tissue sampling, stage of disease, and prior treatments all appear to relate to PD-L1 expression.[8, 9, 26] Since the majority of our cohort represents primary tumors rather than metastatic disease, this may contribute to the high PD-L1 negativity rate in our study. Most but not all studies have described concordance across PD-L1 antibodies, including with E1L3N used in this report.[16, 46–49] Despite our PD-L1 distribution, we still confirmed several previously described and biologically grounded associations, which adds confidence to the results. Furthermore, a robust random subsampling analysis was performed to further explore results in the context of the enriched PD-L1 negativity in our population and overall confirmed the primary conclusions. Finally, and importantly, similar conclusions in our report have also been found in two recent independent analyses in which PD-L1 was more evenly distributed across negative, low, and high groups.[25, 34]
Our report highlights that patterns of PD-L1 expression were distinctly tied to clinical and molecular factors. We found that certain anatomic sites of tissue sampling correlate with PD-L1 high or PD-L1 negative expression and may influence the predictive capacity of PD-L1 as a biomarker for response to ICIs. Leveraging the largest cohort of lung adenocarcinoma patients with PD-L1 testing and targeted next-generation sequencing (MSK-IMPACT), we were also able to identify which molecular alterations relate to PD-L1 expression. Further, this analysis revealed that molecular alterations such as KRAS, TP53, EGFR, and STK11 differentially impact PD-L1 prediction of ICI response. Building on the associations identified, mechanistic studies are needed to determine causal relationships between the molecular phenotype and immunologic phenotype of tumors. Genomic profiling and other factors such as tumor sampling site profoundly impact the prognostic value of PD-L1 and the integration of this data may enhance the predictive value of the PD-L1 biomarker in lung adenocarcinoma in the future.
Supplementary Material
Patients in the “biomarker cohort” compared to patients in the “ICI cohort”.
Comparison of PD-L1 expression and tumor purity estimated (A) by visual histopathologic assessment or (B) by computational inference by FACETS.
Frequency of altered individual genes within (A) PD-L1 <50% vs ≥50% expression subgroups and (B) PD-L1 0% vs ≥1% expression subgroups. Genes labeled in red were significantly enriched (q value <0.15).
Half the PDL1 negative cohort size (n=528) was randomly subsampled 100,000 times. In each simulation, gene enrichment analysis was performed comparing the molecular landscape of PD-L1 high tumors to the randomly subsampled PD-L1 negative cohort (identical methods as for Figure 3a) and computed the q-values for all genes. Plots show the histogram of the q-values for the four genes of interest (TP53, KRAS, STK11, EGFR) from each simulation. The genes of interest remained frequently significant. The vertical blue-dashed line is the q-value: 0.15 and the arrow and the % indicate the fraction of simulations in which the significant association was found.
Forest plot of progression-free survival and overall survival within subgroups of patients with commonly altered genes in lung adenocarcinoma, comparing outcomes of patients with PD-L1 high vs negative tumors.
Highlights.
Distinct clinical and molecular features are associated with differential PD-L1 expression and ICI response in lung adenocarcinoma.
The anatomic site sampled for PD-L1 testing may influence its capacity as a biomarker for ICIs.
Molecular alterations in KRAS, TP53, EGFR, STK11, and the WNT pathway are tied to PD-L1 expression and modulate predictiveness of PD-L1.
PD-L1 must be interpreted within the context of these features, which modulate the prediction of response to ICIs.
Acknowledgement of research support:
Supported by Memorial Sloan Kettering Cancer Center Support Grant/Core Grant (P30 CA008748) and the Druckenmiller Center for Lung Cancer Research at MSK; MDH is a Damon Runyon Clinical Investigator supported (in part) by the Damon Runyon Cancer Research Foundation (CI-98-18) and is a member of the Parker Institute for Cancer Immunotherapy.
Footnotes
Disclosures: JLS reports stock ownership in the following companies: Pfizer, Thermo Fischer Scientific, Inc., Merck & Co Inc and Chemed Corp. KCA has been a compensated consultant for AstraZeneca. MAE is a consultant for AstraZeneca and received support from Astrazeneca, Invivoscribe, and Raindance Technologies. ML has received advisory board compensation from Boehringer Ingelheim, AstraZeneca, Bristol-Myers Squibb, Takeda, and Bayer, and research support from LOXO Oncology and Helsinn Healthcare. CML is a consultant for AbbVie, Amgen, Ascentage, AstraZeneca, Bicycle, Celgene, Chugai, Daiichi Sankyo, Genentech/Roche, GI Therapeutics, Loxo, Novartis, PharmaMar, and Seattle Genetics; serves on the scientific advisory boards of Elucida and Harpoon; and reports personal fees from Bristol‐Myers Squibb and Ipsen. GJR has research funding to his institution from Pfizer, Novartis, Takeda, and Roche. MDH receives research funding from Bristol-Myers Squibb; is paid consultant to Merck, Bristol-Myers Squibb, AstraZeneca, Genentech/Roche, Janssen, Nektar, Syndax, Mirati, and Shattuck Labs; receives travel support/honoraria from AstraZeneca and BMS; and a patent has been filed by MSK related to the use of tumor mutation burden to predict response to immunotherapy (PCT/US2015/062208), which has received licensing fees from PGDx.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
REFERENCES:
- 1.Reck M, Rodríguez-Abreu D, Robinson AG et al. Pembrolizumab versus Chemotherapy for PD-L1–Positive Non–Small-Cell Lung Cancer. 2016; 375: 1823–1833. [DOI] [PubMed] [Google Scholar]
- 2.Mok TSK, Wu YL, Kudaba I et al. Pembrolizumab versus chemotherapy for previously untreated, PD-L1-expressing, locally advanced or metastatic non-small-cell lung cancer (KEYNOTE-042): a randomised, open-label, controlled, phase 3 trial. Lancet 2019; 393: 1819–1830. [DOI] [PubMed] [Google Scholar]
- 3.Garon EB, Rizvi NA, Hui R et al. Pembrolizumab for the Treatment of Non–Small-Cell Lung Cancer. 2015; 372: 2018–2028. [DOI] [PubMed] [Google Scholar]
- 4.Gandhi L, Rodríguez-Abreu D, Gadgeel S et al. Pembrolizumab plus Chemotherapy in Metastatic Non–Small-Cell Lung Cancer. 2018; 378: 2078–2092. [DOI] [PubMed] [Google Scholar]
- 5.Sabari JK, Leonardi GC, Shu CA et al. PD-L1 expression, tumor mutational burden, and response to immunotherapy in patients with MET exon 14 altered lung cancers. Ann Oncol 2018; 29: 2085–2091. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Socinski MA, Jotte RM, Cappuzzo F et al. Atezolizumab for First-Line Treatment of Metastatic Nonsquamous NSCLC. 2018; 378: 2288–2301. [DOI] [PubMed] [Google Scholar]
- 7.Velcheti V, Patwardhan PD, Liu FX et al. Real-world PD-L1 testing and distribution of PD-L1 tumor expression by immunohistochemistry assay type among patients with metastatic non-small cell lung cancer in the United States. PLoS One 2018; 13: e0206370. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Boothman AM, Scott M, Ratcliffe M et al. Impact of Patient Characteristics, Prior Therapy, and Sample Type on Tumor Cell PD-L1 Expression in Patients with Advanced NCSLC Screened for the ATLANTIC Study. J Thorac Oncol 2019. [DOI] [PubMed] [Google Scholar]
- 9.Kim S, Koh J, Kwon D et al. Comparative analysis of PD-L1 expression between primary and metastatic pulmonary adenocarcinomas. Eur J Cancer 2017; 75: 141–149. [DOI] [PubMed] [Google Scholar]
- 10.Ilie M, Long-Mira E, Bence C et al. Comparative study of the PD-L1 status between surgically resected specimens and matched biopsies of NSCLC patients reveal major discordances: a potential issue for anti-PD-L1 therapeutic strategies. Ann Oncol 2016; 27: 147–153. [DOI] [PubMed] [Google Scholar]
- 11.Skoulidis F, Goldberg ME, Greenawalt DM et al. STK11/LKB1 Mutations and PD-1 Inhibitor Resistance in KRAS-Mutant Lung Adenocarcinoma. 2018; 8: 822–835. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Hastings K, Yu HA, Wei W et al. EGFR mutation subtypes and response to immune checkpoint blockade treatment in non-small-cell lung cancer. Annals of Oncology 2019. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Kerr KM, Thunnissen E, Dafni U et al. A retrospective cohort study of PD-L1 prevalence, molecular associations and clinical outcomes in patients with NSCLC: Results from the European Thoracic Oncology Platform (ETOP) Lungscape Project. Lung Cancer 2019; 131: 95–103. [DOI] [PubMed] [Google Scholar]
- 14.Offin M, Guo R, Wu SL et al. Immunophenotype and Response to Immunotherapy of RET-Rearranged Lung Cancers. JCO precision oncology 2019; 3: 10.1200/PO.1218.00386. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Wellenstein MD, de Visser KE. Cancer-Cell-Intrinsic Mechanisms Shaping the Tumor Immune Landscape. Immunity 2018; 48: 399–416. [DOI] [PubMed] [Google Scholar]
- 16.Cheng DT, Mitchell TN, Zehir A et al. Memorial Sloan Kettering-Integrated Mutation Profiling of Actionable Cancer Targets (MSK-IMPACT): A Hybridization Capture-Based Next-Generation Sequencing Clinical Assay for Solid Tumor Molecular Oncology. J Mol Diagn 2015; 17: 251–264. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Zehir A, Benayed R, Shah RH et al. Mutational landscape of metastatic cancer revealed from prospective clinical sequencing of 10,000 patients. Nat Med 2017; 23: 703–713. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Shen R, Seshan VE. FACETS: allele-specific copy number and clonal heterogeneity analysis tool for high-throughput DNA sequencing. Nucleic Acids Res 2016; 44: e131. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Bielski CM, Zehir A, Penson AV et al. Genome doubling shapes the evolution and prognosis of advanced cancers. Nat Genet 2018; 50: 1189–1195. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Gao J, Shi LZ, Zhao H et al. Loss of IFN-gamma Pathway Genes in Tumor Cells as a Mechanism of Resistance to Anti-CTLA-4 Therapy. Cell 2016; 167: 397–404.e399. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Knijnenburg TA, Wang L, Zimmermann MT et al. Genomic and Molecular Landscape of DNA Damage Repair Deficiency across The Cancer Genome Atlas. Cell Rep 2018; 23: 239–254.e236. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Miao D, Margolis CA, Gao W et al. Genomic correlates of response to immune checkpoint therapies in clear cell renal cell carcinoma. Science 2018; 359: 801–806. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Patel SJ, Sanjana NE, Kishton RJ et al. Identification of essential genes for cancer immunotherapy. Nature 2017; 548: 537–542. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Sanchez-Vega F, Mina M, Armenia J et al. Oncogenic Signaling Pathways in The Cancer Genome Atlas. Cell 2018; 173: 321–337.e310. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Hong L, Dibaj S, Negrao MV et al. Spatial and temporal heterogeneity of PD-L1 and its impact on benefit from immune checkpoint blockade in non-small cell lung cancer (NSCLC). Journal of Clinical Oncology 2019; 37: 9017–9017. [Google Scholar]
- 26.Mansfield AS, Aubry MC, Moser JC et al. Temporal and spatial discordance of programmed cell death-ligand 1 expression and lymphocyte tumor infiltration between paired primary lesions and brain metastases in lung cancer. Ann Oncol 2016; 27: 1953–1958. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Mansfield AS, Ren H, Sutor S et al. Contraction of T cell richness in lung cancer brain metastases. Sci Rep 2018; 8: 2171. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Pao W, Ooi C-H, Birzele F et al. Tissue-Specific Immunoregulation: A Call for Better Understanding of the “Immunostat” in the Context of Cancer. Cancer Discovery 2018. [DOI] [PubMed] [Google Scholar]
- 29.Rizvi H, Sanchez-Vega F, La K et al. Molecular Determinants of Response to Anti–Programmed Cell Death (PD)-1 and Anti–Programmed Death-Ligand 1 (PD-L1) Blockade in Patients With Non–Small-Cell Lung Cancer Profiled With Targeted Next-Generation Sequencing. 2018; 36: 633–641. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Rizvi NA, Hellmann MD, Snyder A et al. Cancer immunology. Mutational landscape determines sensitivity to PD-1 blockade in non-small cell lung cancer. Science 2015; 348: 124–128. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Makhnin A, Drilon A, Martinez A et al. A Prospective Study of Circulating Tumor DNA to Guide Matched Targeted Therapy in Lung Cancers. 2018. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Davoli T, Uno H, Wooten EC, Elledge SJ. Tumor aneuploidy correlates with markers of immune evasion and with reduced response to immunotherapy. Science 2017; 355. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Roh W, Chen PL, Reuben A et al. Integrated molecular analysis of tumor biopsies on sequential CTLA-4 and PD-1 blockade reveals markers of response and resistance. Sci Transl Med 2017; 9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.L G, L Y, S L. et al. Clinicopathological and genomic correlates of PD-L1 expression in nonsquamous nonsmall cell lung cancer. In 2019 World Conference On Lung Cancer Barcelona, Spain: 2019. [Google Scholar]
- 35.Meylan E, Dooley AL, Feldser DM et al. Requirement for NF-kappaB signalling in a mouse model of lung adenocarcinoma. Nature 2009; 462: 104–107. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Ji H, Houghton AM, Mariani TJ et al. K-ras activation generates an inflammatory response in lung tumors. Oncogene 2006; 25: 2105–2112. [DOI] [PubMed] [Google Scholar]
- 37.Pylayeva-Gupta Y, Lee KE, Hajdu CH et al. Oncogenic Kras-induced GM-CSF production promotes the development of pancreatic neoplasia. Cancer Cell 2012; 21: 836–847. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Skoulidis F, Byers LA, Diao L et al. Co-occurring genomic alterations define major subsets of KRAS-mutant lung adenocarcinoma with distinct biology, immune profiles, and therapeutic vulnerabilities. Cancer Discov 2015; 5: 860–877. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Skoulidis F, Arbour KC, Hellmann MD et al. Association of STK11/LKB1 genomic alterations with lack of benefit from the addition of pembrolizumab to platinum doublet chemotherapy in non-squamous non-small cell lung cancer. Journal of Clinical Oncology 2019; 37: 102–102. [Google Scholar]
- 40.Arbour KC, Jordan E, Kim HR et al. Effects of Co-occurring Genomic Alterations on Outcomes in Patients with KRAS-Mutant Non-Small Cell Lung Cancer. Clin Cancer Res 2018; 24: 334–340. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Shen R, Martin A, Ni A et al. Harnessing Clinical Sequencing Data for Survival Stratification of Patients With Metastatic Lung Adenocarcinomas. JCO Precision Oncology 2019; 1–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Spranger S, Bao R, Gajewski TF. Melanoma-intrinsic beta-catenin signalling prevents anti-tumour immunity. Nature 2015; 523: 231–235. [DOI] [PubMed] [Google Scholar]
- 43.Luke JJ, Bao R, Sweis RF et al. WNT/beta-catenin Pathway Activation Correlates with Immune Exclusion across Human Cancers. Clin Cancer Res 2019; 25: 3074–3083. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Koyama S, Akbay EA, Li YY et al. STK11/LKB1 Deficiency Promotes Neutrophil Recruitment and Proinflammatory Cytokine Production to Suppress T-cell Activity in the Lung Tumor Microenvironment. Cancer Res 2016; 76: 999–1008. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Humphries MP, McQuaid S, Craig SG et al. Critical Appraisal of Programmed Death Ligand 1 Reflex Diagnostic Testing: Current Standards and Future Opportunities. J Thorac Oncol 2019; 14: 45–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Tsao MS, Kerr KM, Kockx M et al. PD-L1 Immunohistochemistry Comparability Study in Real-Life Clinical Samples: Results of Blueprint Phase 2 Project. J Thorac Oncol 2018; 13: 1302–1311. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Rimm DL, Han G, Taube JM et al. A Prospective, Multi-institutional, Pathologist-Based Assessment of 4 Immunohistochemistry Assays for PD-L1 Expression in Non-Small Cell Lung Cancer. JAMA Oncol 2017; 3: 1051–1058. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.Adam J, Le Stang N, Rouquette I et al. Multicenter harmonization study for PD-L1 IHC testing in non-small-cell lung cancer. Ann Oncol 2018; 29: 953–958. [DOI] [PubMed] [Google Scholar]
- 49.Gaule P, Smithy JW, Toki M et al. A Quantitative Comparison of Antibodies to Programmed Cell Death 1 Ligand 1. JAMA Oncol 2017; 3: 256–259. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Patients in the “biomarker cohort” compared to patients in the “ICI cohort”.
Comparison of PD-L1 expression and tumor purity estimated (A) by visual histopathologic assessment or (B) by computational inference by FACETS.
Frequency of altered individual genes within (A) PD-L1 <50% vs ≥50% expression subgroups and (B) PD-L1 0% vs ≥1% expression subgroups. Genes labeled in red were significantly enriched (q value <0.15).
Half the PDL1 negative cohort size (n=528) was randomly subsampled 100,000 times. In each simulation, gene enrichment analysis was performed comparing the molecular landscape of PD-L1 high tumors to the randomly subsampled PD-L1 negative cohort (identical methods as for Figure 3a) and computed the q-values for all genes. Plots show the histogram of the q-values for the four genes of interest (TP53, KRAS, STK11, EGFR) from each simulation. The genes of interest remained frequently significant. The vertical blue-dashed line is the q-value: 0.15 and the arrow and the % indicate the fraction of simulations in which the significant association was found.
Forest plot of progression-free survival and overall survival within subgroups of patients with commonly altered genes in lung adenocarcinoma, comparing outcomes of patients with PD-L1 high vs negative tumors.