Abstract

Introduction

Somatic KRAS mutations are identified in 25% of patients with non-small cell lung cancers (NSCLC). Patients with these mutations have shorter survival compared to patients with EGFR-mutant NSCLC or KRAS wild type tumors.¹ Since there is significant clinical heterogeneity in patients with KRAS-mutant NSCLC, defining clinically relevant subsets of KRAS-mutant NSCLC is important. In small cohorts, investigators have found specific KRAS point mutations (such as G12V and C12R), were associated with poorer outcomes.² However in a large retrospective analysis of nearly 700 patients with metastatic disease, no apparent differences in outcome based on KRAS mutation subtype were identified.³

In patients with lung cancer, the predictive utility of KRAS mutations as a marker of response to both targeted therapy and standard cytotoxic chemotherapy has been of particular interest. The presence of a KRAS mutation suggests lack of response to EGFR tyrosine kinase inhibitors,^4,5 likely because EGFR and KRAS mutations only rarely occur together. Patients with KRAS codon 13 mutations appear to have poorer outcomes with adjuvant cisplatin-based chemotherapy following resection of early stage disease,⁶ but in the metastatic setting KRAS mutations do not appear to independently predict response or resistance to chemotherapy treatments.^7,8

While the identification of subsets of NSCLC with oncogenic drivers has transformed the treatment of this disease, these advances have thus far largely been limited to patients with mutations in EGFR⁹ or oncogenic fusions involving ALK¹⁰, RET¹¹, or ROS1¹² kinases. There has been progress in the development of compounds that selectively target KRAS G12C, but efforts to specifically target mutant KRAS in the clinic have thus far been largely unsuccessful. Clinical testing of agents targeting downstream pathways, such as MEK and PI3K-AKT, in patients with KRAS-mutant tumors has yielded relatively low response rates.^13,14 This may be due to the fact that there is a significant molecular diversity in KRAS-mutant tumors compared to other known driver events and it is these underlying mechanisms that drive divergent biologic and clinical behavior.¹⁵ However, even KRAS G12C mutant cell lines exhibit a range of responses to pharmacologic inhibition of KRAS G12C, a finding that further highlights the molecular diversity of KRAS-mutant lung cancers.¹⁶

We hypothesized that broad next-generation sequencing may allow an in-depth description of clinically heterogenous group of patients and offer prognostic and predictive markers based on the presence of co-occurring mutations in patients with KRAS-mutant NSCLC. To evaluate this hypothesis, we investigated the effect of commonly co-occurring genomic alterations, clinical characteristics on survival and treatment response in patients with advanced KRAS-mutant NSCLC.

Methods

Results

Clinical Characteristics

We identified 550 patients with lung cancer and KRAS mutations on NGS testing between January of 2014 and October 2016. Of that cohort of patients, 330 had metastatic or recurrent lung cancer. Seventy-two percent of patients had metastatic disease at the time of initial diagnosis (n=240) while 28% (n=90) had recurrent disease. The predominant histology was adenocarcinoma (n=298, 90%). Patient demographics and histology are noted in Table 1. The most common KRAS mutation observed was G12C (44%) (Figure 1A).

Open in a separate window

Open in a separate window

Figure 1

KRAS genotype analysis. Figure 1A: Type of KRAS Codons: KRAS point mutations in dataset. Mutations occurring in less than 1% of patients grouped into “other” category. Figure 1B: Distribution of three most frequently co-occurring mutations as depicted in a proportional Venn diagram.

Table 1

Baseline patient characteristics

Characteristic	N=330 (%)
Age at diagnosis (stage IV)
Median (range)	61 (45–80)
Sex
Women	195 (59)
Men	135 (41)
KPS (%)
≥80	241 (73)
<80	72 (22)
Not recorded	17 (5)
Smoking History Category
Current	37 (11)
Former	271 (82)
Never	22 (7)
Smoker History, pack-years
Median Pack-year	30
Range	0–135
Unknown	2
Pathology
Adenocarcinoma	294 (89)
Squamous	11 (3)
Adenosquamous	4 (1)
Other	21 (6)

Open in a separate window

Co-occurring Mutations and Survival

The median follow-up among the 177 patients alive at the data cutoff of January 2017 was 12 months (range 1–114). The median overall survival (mOS) for all patients with KRAS-mutant advanced lung cancers in this cohort was 17 months (95% CI: 14–25). Patients with and without concurrent mutation in TP53 had a similar overall survival (HR 0.9, 95% CI 0.6–1.2, p=0.5). Patients with a concurrent mutation in STK11 (KRAS/STK11) were found to have a shorter overall survival (OS) (HR 1.7, 95% CI 1.1–2.4, p=0.002, Figure 2A). In addition, patients with concurrent mutation in KEAP1 or NFE2L2 (KRAS/KEAP1/NFE2L2) were also found to have a shorter OS (HR 2.1, 95% CI 1.4–3.1, p<.0001, Figure 2B). A multivariable Cox proportional hazards model was used to assess the independent effect of the most frequently identified co-occurring mutations (TP53, STK11, KEAP1/NFE2L2), adjusting for age, gender, performance status, and smoking history. After adjustment for these clinical variables, only KEAP1/NFE2L2 was independently associated with shorter OS (HR 1.96, 95% CI 1.3-3-2.92, p<0.001, Table 3)

Open in a separate window

Open in a separate window

Open in a separate window

Figure 2

Associations of co-occurring genomic alterations and KRAS with overall survival from time Stage IV diagnosis A, STK11 B, KEAP1 or NFE2L2, C, TP53

Table 3

Multivariate Analysis of Overall Survival in patients with KRAS-mutant NSCLC

Variable	HR (95% CI of HR)	p-Value
TP53 (+ vs −)	1.11 (0.77 to 1.58)	0.581
STK11 (+ vs −)	1.30 (0.86 to 1.97)	0.216
KEAP1/NFE2L2 (+/−)	1.96 (1.33 to 2.92)	<0.001
Gender (male vs female)	1.20 (0.86 to 1.68)	0.284
Age at diagnosis	1.03 (1.01 to 1.05)	0.003
Smoking (former/current vs never)	1.43 (0.86 to 1.68)	0.402
KPS at diagnosis (KPS 80–100 vs <80)	0.88 (0.60 to 1.29)	0.514

Open in a separate window

Immunotherapy and co-mutation

To evaluate the effects of co-mutations on outcomes after immunotherapy, we obtained treatment history of 86 patients that underwent therapy with an immune checkpoint inhibitor as monotherapy (nivolumab or pembrolizumab). In a univariate analysis, neither the presence of TP53, SKT11, nor KEAP1/NFE2L2 was associated with a difference in duration of immune therapy (Supplemental Figure 2). These results were confirmed in a multivariate analysis adjusting for these three co-mutations as well as, age, gender, performance status, and line of therapy. Patients with a co-occurring mutation in KEAP1 or NFE2L2 were found to have a shorter overall survival from the start of immune checkpoint inhibitor in both a univariate (Figure 3) and multivariate analysis (HR 3.54, 95% CI 1.55–8.11, p=0.003, Supplemental Table 3) adjusting for co-mutations as well as, tumor mutation burden, age, gender, performance status, and line of therapy. The presence of a co-occurring mutation in neither STK11 nor TP53 was associated with a significant difference in mOS from start of therapy. Tumor mutation burden was associated with difference in overall survival from time of initiation of immunotherapy as patients with higher mutation burdens were found to have longer survival from start of treatment (HR 0.9, 95% CI 0.83–0.99, p=0.025, Supplemental Table 3).

Open in a separate window

Open in a separate window

Figure 3

Overall survival (from time of start of immune checkpoint inhibitor therapy) for treatment of Stage IV disease based on presence of A, KEAP1/NFE2L2 co-mutation and B, STK11 co-mutation

Discussion

We report the landscape of co-occurring genomic alterations in a series of 330 patients with advanced KRAS-mutant lung cancer, highlighting the most common events and identifying mutations in the KEAP1/NFE2L2 pathway as a significant, independent negative prognostic factor for patients with KRAS-mutant NSCLC. We go on to demonstrate that this association with poor overall survival is also associated with shorter duration of therapy with platinum-doublet chemotherapy and overall survival after immunotherapy. These data suggest that the observed clinical heterogeneity in patients with KRAS-mutant NSCLC is likely due in part to differences in co-occurring molecular events.

Concurrent mutations are common in patients with KRAS-mutant NSCLC, with the most common events in STK11 and TP53. Although TP53 mutations are common, our analysis concurs with that of others, reporting that concurrent TP53 mutation in KRAS-mutant NSCLC is not prognostic.¹⁹ Multiple prior analyses have reported STK11 mutations are more frequent in KRAS-mutant NSCLC as opposed to KRAS wild type tumors.²⁰ Preclinical data has suggested that loss of STK11 leads to a more aggressive tumor phenotype.²¹ However there has been conflicting evidence whether STK11 mutations have prognostic or predictive implications.^15,22 Our initial univariate analysis also demonstrated that concurrent STK11 mutations were associated with shorter overall survival in patients with advanced KRAS-mutant NSCLC. However, when a multivariable analysis was conducted adjusting for concurrent mutations in KEAP1/NFE2L2, TP53 and other clinical variables, concurrent mutation in STK11 was no longer associated with difference in overall survival.

The differences seen in the univariate analysis compared to multivariable analysis is likely due to significant overlap between patients with STK11 and KEAP1 concurrent mutations (also observed by others¹⁵). Our analysis suggests, however, that it is not the concurrent STK11 mutation that is associated with adverse outcomes in these patients, but rather the presence of concurrent KEAP1 or NFE2L2 mutation. This observation is limited as we evaluated only somatic mutations and not protein expression, and therefore it is possible that we did not capture tumors that have suppressed STK11 mRNA expression through a mechanism other than STK11 mutation which were observed in the subsets described by Skoulidis et al.¹⁵ These results highlight the importance of multivariable analyses, incorporating not just clinical features but also genomic features, in order to more accurately describe prognostic features in the genomically complicated landscape of KRAS-mutant NSCLC.

Mutations in the KEAP1/NFE2L2 pathway identify 27% of patients with KRAS-mutant lung cancer as having an independent negative prognostic factor. KEAP1 and NFE2L2 mutations have been best described in squamous cell lung cancer. Activation of the NRF2 pathway (through activation of NFE2L2 or inactivation of KEAP1) was found to be altered in 34% of squamous cell cancers of the lung when evaluated by TCGA investigators.²³ Solis et al. also showed that increased nuclear expression of Nrf2 and decreased or absent expression of KEAP1 in 38% and 46% of patients with squamous cell carcinoma, respectively and less commonly in adenocarcinoma (18%).²⁴ While low cytoplasmic KEAP1 expression was associated with worse overall survival in squamous cell carcinomas, no association was found between KEAP1 expression and outcomes in patients with adenocarcinoma. That analysis, similar to the TCGA analysis, is heavily weighted towards patients with early stage resected disease as opposed to advanced stage disease which is the focus of our report.

The poor prognosis of patients with concurrent KEAP1 or NFE2L2 mutations may be due in part to the observation that activation of the Nrf2 pathway may be associated with resistance to chemotherapy. Mutations in KEAP1 affect the repressive activity of KEAP1, stimulating nuclear accumulation of Nrf2, and induce constitutive expression of cytoprotective enzymes.²⁵ Cell lines expressing lower levels of KEAP1 or KEAP1 mutant cells demonstrated greater resistance to cisplatin than cell lines with normal KEAP1.²⁶ Concordant with this, we observed that patients with a concurrent KEAP1 or NFE2L2 mutation who were treated with a platinum/pemetrexed combination had more rapid disease progression (as suggested by shorter duration of therapy) compared to KEAP1 or NFE2L2 wild type patients. Our analysis was limited to patients with KRAS-mutant NSCLC, suggesting a cooperativity between these mutations in KEAP1 or NFE2L2 and KRAS. Moreover, the frequency of KEAP1/NFE2L2 in KRAS mutant NSCLC is much higher than seen in other lung cancer with other oncogenic drivers.

Treatment with immune check point inhibitors has been a significant advance in the treatment of NSCLC,^27,28 particularly for patients without targetable oncogenic drivers. Despite these advances, single-agent response rates in unselected populations remain relatively low (10–30% in reported clinical trials). Various biomarkers have been reported that may potentially predict response to immune checkpoint inhibitors in solid tumors including PD-L1 expression²⁹ and tumor mutation burden or neoantigen load^30,31. We attempted to explore the impact of concurrent genomic alterations in patients who received single agent immune checkpoint inhibitors during the treatment of their disease. Our analysis of clinical benefit with immunotherapy was limited in size, since only a subset of the patients in our larger analysis received this therapy. Keeping these limitations in mind, our analysis suggests that patients with KRAS-mutant advanced NSCLC and a concurrent mutation in KEAP1 or NFE2L2, have significantly shorter overall survival from initiation of immune checkpoint therapy. No statistically significant differences in OS from start of therapy were observed based on concurrent STK11 or TP53 mutations. Formal response assessment was available on a subset of these patients (data not shown), and no significant difference in response was observed based on concurrent mutation in either KEAP1/NFE2L2, STK11, or TP53.

It is possible that mutations in KEAP1/NFE2L2 are associated with other features that have been associated with outcomes after immunotherapy, such as PD-L1 expression status which was not available in this series of patients. PD-L1 expression has been reported in KRAS-mutant NSCLC and was more frequently observed in smokers,³² however relationship to concurrent mutation is thus far unknown. Skouldolis et al. described that the cluster of KRAS-mutant NSCLC with low STK11 expression demonstrated a lack of immune system engagement however concurrent mutations in KEAP1 were also common in this cluster.¹⁵ In a follow up analysis, they reported that patients with concurrent mutations in STK11 had lower objective response rates to immunotherapy agents,³³ a finding we did not observe in our larger series of patients.

Higher nonsynonymous mutation burden in tumors has been associated with improved objective response, durable clinical benefit, and progression free survival in patients treated with the PD-1 antibody pembrolizumab.³⁰ More recently, normalized mutation count calculated from results of routine next-generation sequencing was reported to be predictive of response to nivolumab in NSCLC.^34,35 Our multivariable analysis did incorporate normalized tumor mutation count and demonstrated that tumor mutation burden was associated with longer overall survival from initiation of immune checkpoint inhibitor. However, presence of concurrent mutation in KEAP1 or NFE2L2 was associated with significantly shorter mOS independent of tumor mutation burden.

While the findings we have described are consistent with other reports analyzing the molecular characteristics of patients with KRAS-mutant lung cancer, there are some limitations to our analysis. All patients were identified based on molecular testing at a single institution. Although molecular analysis is offered routinely for all patients, some patients may have had inadequate biopsy specimens, precluding next-generation sequencing analysis. Patients received diverse treatments and therefore the analyses of initial platinum/pemetrexed chemotherapy and immunotherapy involve a much smaller subset of patients. Duration of therapy was used as a surrogate marker for progression free survival when describing treatment history which has potential pitfalls as patients may discontinue therapy for other reasons (e.g. toxicity) as opposed to progression.

While it has been evident that there is significant clinical heterogeneity in patients with KRAS-mutant NSCLC, it is now becoming clear that this clinical heterogeneity is like due to biologic heterogeneity. Skouldolis et al. have previously reported molecular stratification of KRAS-mutant lung adenocarcinomas using RNA sequencing expression data from a subset of lung adenocarcinomas in the TCGA. These results were further validated in other small cohorts, however this dataset focused on primarily early stage disease with only a subset of patients analyzed having Stage IV disease (and all of these from the BATTLE-2 clinical trial and therefore platinum refractory).¹⁵

Our analysis shows that routine next-generation sequencing can not only provide information regarding potential actionable mutations, but also suggest prognostic and predictive features of a patient’s cancer by exploring the presence of various co-occurring mutations. Considering any given mutation in the context of other mutations and using multivariate analyses is crucial in evaluating the significance of somatic genetic events. Prospective identification of patients with concurrent KEAP1 or NFE2L2 mutations in patients with KRAS-mutant NSCLC should be considered as a prognostic factor in clinical trials evaluating therapies for these patients. Our results indicate that patients with concurrent KRAS and KEAP1/NFE2L2 have a clinically distinct behavior and may require stratification in trials.

Supplementary Material

Acknowledgments

References