Outcomes in patients treated with frontline immune checkpoint inhibition (ICI) for advanced NSCLC with KRAS mutations and STK11/KEAP1 comutations across PD-L1 levels

Abstract

Introduction:

In patients with advanced NSCLC (aNSCLC), the impact of KRAS mutations (m) and comutations with STK11 and KEAP1 on outcomes across different PD-L1 levels remains incompletely understood. We aimed to investigate the frequency of KRAS mutations and comutations across PD-L1 levels, and the association between these mutations and survival, stratified by PD-L1 expression.

Methods:

We conducted a nationwide cohort study of patients diagnosed with aNSCLC between 2016 and 2021 treated with frontline (chemo)immunotherapy, who underwent molecular genotyping including KRAS, STK11, and KEAP1. Real-world overall survival (OS) and progression-free survival (rwPFS) were estimated using Kaplan-Meier methodology. Cox multivariable regressions were used to evaluate the association between KRASm and survival across different PD-L1 strata, and to assess whether the association between KRASm and survival differed by PD-L1 level. Finally, within subgroups defined by PD-L1 expression, we used interaction terms to assess whether co-mutations with STK11 and KEAP1 moderated the association between KRAS mutation and survival.

Results:

Of our 2593-patient cohort, 982 (37.9%) were KRASm and 1611 (62.1%) KRASwt. KRASm were enriched in the PD-L1 ≥50% cohort (334/743, 45%), but within patients with KRASm, co-mutations with STK11 and KEAP1 were enriched in the PD-L1 0% cohort. KRASm was associated with significantly worse OS in the PD-L1 0% cohort compared to the PD-L1≥50% cohort (P for interaction=0.008). On adjusted analyses stratified by PD-L1, KRASm was associated with worse survival only in the PD-L1 0% group (OS HR 1.46, p=0.001). KEAP1 and STK11 comutations were most strongly associated with worse OS in the PD-L1 0% subgroup; patients with triple KRASm/KEAPm/STK11m PD-L1 0% NSCLC experienced the worst outcomes.

Conclusions:

KRASm are associated with worse overall survival in PD-L1 negative NSCLC; however, this association is largely driven by comutations with STK11 and KEAP1, which are enriched in PD-L1 negative tumors.

1. Introduction

KRAS mutations(m) are the most common oncogenic alterations in NSCLC, occurring in 20–40% of lung adenocarcinomas^1–3. While KRAS has been historically considered undruggable, two KRAS G12C inhibitors are now FDA approved^4,5 and there is a suite of emerging KRAS-directed agents in development⁶. Understanding the relative therapeutic benefit of these novel agents, as well as designing clinical trials with prognostic balance between arms, will require a thorough understanding of outcomes and treatment responsiveness of KRASm aNSCLC across subgroups defined by KRASm subtype, PD-L1 status, and comutations.

While immune checkpoint inhibitors (ICI) are now part of standard frontline treatment for wild-type NSCLC, different oncogenic alterations in NSCLC impart disparate treatment responses. Whereas driver alterations such as EGFR and ALK are associated with poor response to checkpoint inhibition, KRASm aNSCLC are characterized by higher association with tobacco exposure and responsiveness to ICI^7,8, attributable to high tumor mutational burden and tumor-infiltrating lymphocytes as well as enrichment for high PD-L1 expression^9–12. However, there is substantial heterogeneity within KRASm aNSCLC with subsets defined by co-occurring genomic alterations exhibiting distinct biology and therapeutic vulnerabilities¹³. Prior studies have shown worse immunotherapy response and survival in patients with KRASm and comutations in genes including TP53, STK11, KEAP1, and SMARCA4^14–17. In particular, STK11 alterations are enriched among PD-L1 negative tumors, associated with low tumor-infiltrating CD8+ T cells, and implicated as a driver of primary resistance to PD-1 blockade in KRASm lung adenocarcinoma^15,17. Loss of KEAP1, which is involved in oxidative damage response, has also been associated with an immunosuppressive tumor microenvironment as well as resistance to chemotherapy^18,19.

Both STK11 and KEAP1 mutations seem to confer resistance to PD(L)1 inhibition uniquely in KRASm but not KRASwt lung adenocarcinoma¹⁵, and are associated with (and potentially drive) lower PD-L1 expression^15,17. However, the clinical impact of these comutations across different PD-L1 expression strata, as well as the implication of triple KRAS/STK11/KEAP1 mutations, have not been fully elucidated. Since PD-L1 expression is an important predictor of immune checkpoint inhibitor (ICI) efficacy, we hypothesized that the association between KRASm (and comutations) and survival with ICI therapy differs across and is mediated by PD-L1 expression level. We used a real-world clinicogenomic observational database to investigate the association between KRAS mutation and comutations on progression-free and overall survival, stratified by PD-L1 expression.

2. Methods

We conducted a real-world study using the longitudinal electronic health record (EHR)-derived Flatiron-Foundation Medicine Clinico-Genomic Database (CGDB). This database includes de-identified patient-level structured and unstructured data curated via technology-enabled abstraction, originating from approximately 280 cancer clinics (~800 sites of care) throughout the United States²⁰. Local institutional review board approval of the study protocol and informed consent waiver were obtained. The study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Our study cohort included adult patients (≥18 years) with a new diagnosis of advanced/metastatic NSCLC between 2016 and 2021 who received frontline (chemo)immunotherapy. Eligible patients underwent comprehensive molecular genotyping that included KRAS, STK11, and KEAP1 within 3 months of starting therapy. Patients with alterations in EGFR, ALK, and ROS1 were excluded.

2.1. Definition of Variables and Endpoints

We described patient demographics (age, sex, race, smoking status), ECOG performance status (PS), tumor histology and PD-L1 expression, practice setting, insurance type, and treatment type (ICI monotherapy, chemoimmunotherapy, or anti-PD(L)1 and anti-CTLA4). Patients were classified on the basis of PD-L1 expression (negative 0%, low 1–49%, high ≥50%).

KRAS status was defined as wild type or mutant based on Foundation Medicine (FMI) testing. FMI test results were also used to define KRAS mutation subtype as well as co-mutations with STK11 and KEAP1. Mutations were defined based on single variant findings only (copy number variation and rearrangements were excluded). PD-L1 status was defined based on a combination of the Flatiron Health abstracted biomarker table and FMI reported results.

The primary endpoint was real-world overall survival (OS), defined as the time from initiation of frontline therapy to death or censoring at the last follow-up. Secondary endpoint was real-world progression-free survival (rwPFS), defined as the time from initiation of frontline therapy to progression, death, or censoring at the last follow-up. Real-world progression events included both RECIST and non-RECIST progression events from chart abstraction²¹. Death was defined from a composite mortality variable developed by Flatiron Health using structured and unstructured EHR-derived data linked to a commercial death data source and U.S. Social Security Death Index²².

2.2. Statistical Analysis

Baseline and demographic characteristics were compared between subgroups of interest using t-test for continuous variables and chi-square test for categorical variables. OS and rwPFS were estimated using Kaplan-Meier methodology; multivariable Cox regressions were used to evaluate associations with survival controlling for relevant covariates. Prior to fitting models, we checked for evidence of dependent left truncation via Kendall’s tau; as no significant dependence was found, we used Cox models with risk-set sampling. All statistical tests were conducted using R Software (version 4.2).

In our primary analysis, we constructed several Cox proportional-hazards (PH) regression models to examine the association between KRASm and OS following first line therapy including adjustment for PD-L1 status (negative, low, high, unknown) and other potentially confounding covariates, including smoking history (never, current or former), age (continuous), treatment site (academic, community), treatment type (ICI monotherapy, chemoimmunotherapy, or anti-PD(L)1 and anti-CTLA4), gender (M/F), race (Black, White, Other/unknown), year of diagnosis (2016–2017, 2018–2019, 2020–2021), histology (non-squamous, squamous, NSCLC NOS), and ECOG PS at time of diagnosis (0–1, 2+, unknown).

We conducted two additional analyses: We formally explored the interaction between KRAS status and PD-L1 level using an interaction term in order to assess whether the association between KRASm and survival differed across PD-L1 subgroups. Because a significant interaction was detected, we then separately analyzed subgroups defined by PD-L1 expression (0%, 1–49%, ≥50%), and assessed association between KRAS mutation and overall survival using Cox multivariable regressions adjusting for covariates as described above.

Secondly, within subgroups defined by PD-L1 expression (0%, 1–49%, ≥50%), we assessed whether comutations with STK11 and KEAP1 moderated the association between KRAS mutation and survival. This was done by including interaction terms (i.e. main effects of each mutation, and an interaction between the two mutations) in a Cox multivariable regression. Pairwise comparisons between mutation groups (e.g. wt/wt vs double mut) were calculated via linear combinations of main effect terms and interaction terms.

3. Results

3.1. Patient characteristics

Our overall cohort consisted of 2593 patients, 982 (37.9%) KRASm and 1611 (62.1%) KRASwt (Supplemental Figure 1, Table 1). Mean age was 69 years, 47.4% were female, 70.0% were White, 67.7% had ECOG performance status 0–1, 92.5% had history of smoking, and 74.3% had non-squamous histology. Amongst 1997 patients (77.0%) with known PD-L1 status, PD-L1 status was distributed evenly between 0% (25.1%), 1–49% (23.3%), and ≥50% (28.7%). Among patients with KRASm (982/2593), most common KRASm subtypes were G12C (40.6%), G12V (19.9%), and G12D (13.5%). Chemoimmunotherapy was more common overall (62.6%) than ICI monotherapy (35.2%); as expected, ICI monotherapy use was higher in patients with PD-L1≥50% (60.3%) than in patients with PD-L1 0% (19.4%) or 1–49% (22.6%) (Supplemental Table 1).

Table 1.

Cohort Characteristics.

	Total (N=2593)	KRAS m (N=982)	KRAS wt (N=1611)	p-value1
Age, mean (SD)	69.3 (9.3)	69.0 (9.6)	69.4 (9.2)	0.221
Gender				< 0.001
Female	1230 (47.4%)	558 (56.8%)	672 (41.7%)
Male	1363 (52.6%)	424 (43.2%)	939 (58.3%)
Race				0.906
White	1816 (70.0%)	690 (70.3%)	1126 (69.9%)
Black or African American	175 (6.7%)	68 (6.9%)	107 (6.6%)
Other/unknown	602 (23.2%)	224 (22.8%)	378 (23.5%)
ECOG Performance Status				0.255
0/1	1756 (67.7%)	657 (66.9%)	1099 (68.2%)
2+	416 (16.0%)	151 (15.4%)	265 (16.4%)
Unknown	421 (16.2%)	174 (17.7%)	247 (15.3%)
Year of Diagnosis				0.379
2016–2017	878 (33.9%)	337 (34.3%)	541 (33.6%)
2018–2019	1277 (49.2%)	492 (50.1%)	785 (48.7%)
2020–2021	438 (16.9%)	153 (15.6%)	285 (17.7%)
Smoking Status				< 0.001
History of smoking	2399 (92.5%)	934 (95.1%)	1465 (90.9%)
No history of smoking	194 (7.5%)	48 (4.9%)	146 (9.1%)
Practice type				0.223
Academic	193 (7.4%)	81 (8.2%)	112 (7.0%)
Community	2400 (92.6%)	901 (91.8%)	1499 (93.0%)
Histology				< 0.001
Nonsquamous	1926 (74.3%)	899 (91.5%)	1027 (63.7%)
Squamous	550 (21.2%)	43 (4.4%)	507 (31.5%)
NSCLC histology NOS	117 (4.5%)	40 (4.1%)	77 (4.8%)
PD-L1 expression				< 0.001
0%	651 (25.1%)	211 (21.5%)	440 (27.3%)
1–49%	603 (23.3%)	233 (23.7%)	370 (23.0%)
≥50%	743 (28.7%)	334 (34.0%)	409 (25.4%)
Unknown	596 (23.0%)	204 (20.8%)	392 (24.3%)
KRAS Subtype
G12C		399 (40.6%)
G12V		195 (19.9%)
G12D		133 (13.5%)
Other		255 (26.0%)
Treatment type				0.264
ICI monotherapy	913 (35.2%)	348 (35.4%)	565 (35.1%)
Chemoimmunotherapy	1622 (62.6%)	618 (62.9%)	1004 (62.3%)
Anti-PD-(L)1 + anti-CTLA-4	58 (2.2%)	16 (1.6%)	42 (2.6%)

¹Unpaired T-test for continuous variables, Fisher’s chi square test for categorial variables.

KRAS, Kirsten rat sarcoma virus; SD, standard deviation; ECOG, Eastern Cooperative Oncology Group; ICI, immune checkpoint inhibitor; PD-L1, programmed death-ligand 1; CTLA, cytotoxic T-lymphocyte antigen

Distribution of KRASm and co-mutations across PD-L1 strata are shown in Figure 1. KRASm were more common in PD-L1 high (334/743, 45.0%) than in PD-L1 low (233/603, 38.6%) or PD-L1 negative (211/651, 32.4%) tumors. However, amongst patients with KRASm, co-mutations with STK11 and KEAP1 were enriched in the PD-L1 negative cohort. Over half of patients with PD-L1 negative KRASm tumors had comutation with STK11 (50/211), KEAP1 (21/211), or both (57/211); whereas the majority of patients with PD-L1 high KRASm tumors had KRASm alone without STK11 or KEAP1 comutations (271/334, 81.1%).

Fig. 1.

Frequency of KRAS mutation and comutations with STK11 and KEAP1 by PD-L1 level. Abbreviations: PD-L1, programmed death ligand 1; KRAS, kirsten rat sarcoma gene; wt, wild type; mt, mutant.

3.2. Association between KRAS status and survival within PD-L1 Subgroups

On multivariable regression including all patients, a significant interaction was seen between KRAS status and PD-L1 level, whereby KRASm was associated with significantly worse OS in the PD-L1 negative cohort compared to PD-L1 high cohort (p=0.008). Other factors associated with worse OS included male sex, squamous or other histology, ECOG performance status ≥2, KRASm status, PD-L1 0%, and treatment with ICI monotherapy compared to chemoimmunotherapy (Supplemental Table 2). Within adjusted analyses stratified by PD-L1 level, KRASm was associated with worse OS and rwPFS only in the PD-L1 negative group (OS HR 1.46, p=0.001; rwPFS HR 1.23, p=0.035) (Table 2, Figure 2).

Table 2.

Prognostic impact of KRAS and comutations by PD-L1 level

	PD-L1 0% (n=651)			PD-L1 1–49% (n=603)			PD-L1 ≥50% (n=743)
	HR	95% CI	p-value	HR	95% CI	p-value	HR	95% CI	p-value
HR for OS, KRASm vs wt1	1.46	1.18-	0.001	1.25	0.98-	0.068	0.96	0.78-	0.675
HR for rwPFS, KRASm vs wt1	1.23	1.01-	0.035	1.11	0.90-	0.342	0.99	0.83-	0.944
Adjusted pairwise comparisons for OS1
Single KRASm vs KRASwt	1.15	0.87-	0.322	1.12	0.84-	0.456	0.95	0.76-	0.613
KRASm/STKUm vs KRASwt/STK11wt	1.81	1.37-	<0.001	2.32	1.65-	<0.001	1.38	0.89-	0.145
KRASm/STK11m vs single KRASm	1.57	1.14-	0.006	2.08	1.46-	<0.001	1.46	0.94-	0.092
KRASm/KEAP1m vs KRASwt/KEAP1wt	2.37	1.74-	<0.001	2.15	1.46-	<0.001	1.34	0.90-	0.143
KRASm/KEAP1m vs single KRASm	2.03	1.45-	<0.001	1.77	1.20-	0.004	1.54	1.04-	0.032
Triple KRASm/STK11m/KEAP1m vs single KRASm	2.73	1.75-	<0.001	2.64	1.62-	<0.001	2.35	1.25-	0.008
Triple mut vs KRASm/STK11m	2.01	1.24-	0.005	1.37	0.78-	0.271	2.29	0.97-	0.060
Triple mut vs KRASm/KEAP1m	1.67	0.86-	0.130	2.11	0.92-	0.080	1.72	0.81-	0.157

¹All HR adjusted for treatment type, demographics (sex, race, age), year of diagnosis, histology, smoking status, ECOG performance status, practice type.

HR, hazard ratio; OS, overall survival; rwPFS, real world progression-free survival; m, mutant; wt, wild type

Fig. 2.

Overall survival and real-world progression free survival by KRASm vs KRASwt, PD-L1 0 %/1–49 %/≥50 % groups.

3.3. Association between STK11/KEAP1 co-mutations and survival within PD-L1 Subgroups

Comutations with KEAP1 and STK11 were strongly associated with worse OS in the PD-L1 negative and PD-L1 low subgroups; these associations were weaker or absent in the PD-L1 high subgroup (Table 2). Adjusted pairwise comparisons of mutation combinations showed that in the PD-L1 negative group, KRASm in isolation (i.e. STK11 and KEAP1 wt) was no longer associated with worse survival compared to triple wt. However, double KRASm/STK11m was associated with worse survival compared to KRASwt/STK11wt (HR 1.81) as well as KRASm alone (HR 1.57). In the PD-L1 low subgroup, KRAS/STK11 double mutation was similarly associated with worse survival compared to double wt (HR 2.32) and KRAS single mutation (HR 2.08). However, in the PD-L1 high subgroup, double mutations did not impart significantly worse prognosis compared to either double wt or single KRASm. Similarly, KRAS/KEAP1 double mutations were much more strongly associated with worse survival in the PD-L1 negative and PD-L1 low groups, than in the PD-L1 high subgroup (Table 2).

Within patients with KRASm and PD-L1 negative tumors, triple mutation (KRAS/KEAP1/STK11) was associated with worse survival than both KRAS single mutation (HR 2.50) and KRAS/STK11 double mutation (HR 1.97) (Table 2, Figure 3). In PDL1 low and high subgroups, triple mutation was associated with worse survival than KRAS single mutation (HR 2.2–2.7), but did not differ significantly compared to KRAS/STK11 double mutant.

Fig. 3.

Survival by comutation status, PD-L1 negative KRASm subgroup. A) overall survival. B) real-would progression free survival.

4. Discussion

This real world clinicogenomic database analysis of over 2,500 patients with aNSCLC treated with frontline ICI-based therapy further defines the impact of KRASm and comutations with STK11 and KEAP1, with an emphasis on differing prevalence and impact of comutations across PD-L1 strata. Of note, the 38% prevalence of KRASm seen in our cohort is slightly higher than that of prior reports^1–3, largely due to the up-front exclusion of patients with EGFR, ALK, and ROS1 alterations. Our data confirms prior reports that KRAS/STK11 and KRAS/KEAP1 comutations define specific poor-prognosis subsets of patients who could benefit from novel or intensified treatment approaches, and shows that this effect is most prevalent and most pronounced in the PD-L1 negative cohort.

On overall adjusted analysis, there was a significant interaction between KRASm status and PD-L1 expression, wherein KRAS mutation was most strongly associated with worse survival in the PD-L1 negative cohort. However, further analysis showed that this association was largely driven by comutations with STK11 and KEAP1 which were highly enriched in patients with PD-L1 negative tumors. Indeed, in adjusted pairwise comparisons, the association between KRAS single mutation and survival was no longer significant in the PD-L1 negative subgroup, suggesting that comutations accounted for much of the association between KRASm and survival. Within PD-L1 negative patients, KRAS comutations with STK11 and KEAP1 were associated with particularly poor outcomes, with patients with KRASm/STK11m/KEAPm triple mutations (which was actually the most common comutational profile) experiencing the worst survival. Owing both to (1) enrichment of STK11 and KEAP1 comutations in PD-L1 negative tumors and (2) stronger association with worse survival imparted by these comutations in PD-L1 negative tumors, patients with PD-L1 negative and KRASm tumors represent a treatment-resistant group with poor outcomes who may particularly benefit from treatment intensification and novel therapeutics.

KRASm have been recognized as adverse prognostic indicators in mNSCLC even prior to the ICI era^23,24; thus, it remains challenging to discern the prognostic vs predictive roles of these alterations²⁵. Nevertheless, our findings add to the body of literature showing that KRAS mutations in NSCLC impart variable predictive significance with ICI therapy depending on comutation status, and provides further insight on differences across PD-L1 levels, still a primary biomarker for treatment selection. Multiple studies have established that KRAS/STK11 and KRAS/KEAP1 tumors have poor outcomes with immunotherapy^15,17, leading to close attention to efficacy of KRAS inhibitors^4,14 as well as intensification strategies such as dual checkpoint blockade²⁶ in these comutation subsets.

Our findings reinforce prior reports that STK11 alterations are enriched in PD-L1 negative (particularly TMB-intermediate or high) cohorts, irrespective of KRAS status, and may in fact be causally related to a non-T cell inflamed immune microenvironment with low PD-L1 expression despite intermediate or high TMB¹⁷. This study also reported that STK11 alterations are associated with lower response rate and survival with checkpoint inhibition even in PD-L1 positive patients, suggesting that STK11 alterations also mediate checkpoint inhibitor resistance independently of PD-L1 status¹⁷. While our study corroborates a correlation (albeit weaker) between STK11 alterations and poorer ICI outcomes even in PD-L1 positive patients, we show that the effect is much more prominent in the PD-L1 negative population. Our study also supports the finding that alterations in KEAP1, which are characterized by an immunosuppressive TME and low CD8T cell infiltration^19,27, are also associated with decreased immunotherapy efficacy in the KRASm population^15,28, particularly in PD-L1 negative tumors.

In our study, patients with triple mutations in KRAS/STK11/KEAP1 had the worst survival outcomes, particularly in PD-L1 negative but also in PD-L1 positive tumors. In preclinical models, concurrent STK11 and KEAP1 mutations have been shown to lead to aggressive tumor growth and treatment resistance through mechanisms such as ferroptosis protection²⁹. KEAP1 comutations with other genes including STK11 has been previously shown to lead to reduced immunotherapy efficacy¹⁹. Moreover, whereas STK11 is thought to drive primary resistance to PD(L)1 inhibition, KEAP1 loss and resultant changes in the oxidative stress response may confer resistance to multiple treatment modalities including chemotherapy, immunotherapy, and radiation³⁰. As patients in our cohort were treated with either immunotherapy or chemoimmunotherapy, it is perhaps unsurprising that patients with multiple mechanisms of therapeutic resistance imparted by KEAP1/STK11 comutations had poorer outcomes, although the nature of this interaction (ex, additive or synergistic) is unknown.

There are important limitations to this retrospective study. While we included patients treated with immunotherapy based treatment, both patients treated with immunotherapy alone and with chemo-immunotherapy were included in our analysis. While this factor was adjusted for in our survival analysis, we did not investigate whether the negative impact of STK11/KEAP1 comutations was driven by differences in response to immunotherapy or chemo-immunotherapy. In addition, a moderate proportion of our total cohort was excluded from our analysis due to unknown PD-L1 expression, since this was a focus of our analysis. These excluded patients had similar characteristics to the rest of the cohort. Finally, our models did not include an exhaustive list of other biomarkers, including KRAS mutation subtype, tumor mutational burden, and other comutations (TP53, SMARCA4) that have variably shown prognostic or predictive utility, and thus we cannot rule out residual confounding.

In conclusion, our study uses a large clinicogenomic database to show that while KRAS mutations are enriched in PD-L1 high aNSCLC, comutations with KEAP1 and STK11 are enriched in PD-L1 negative cohorts, and that these PD-L1 negative, comutation-positive patients represent a particularly poor-prognosis group in need of further study to understand resistance mechanisms and therapeutic strategies to overcome them. Importantly, this data also provides additional confirmation for the important role of molecular testing in management of patients with aNSCLC and supports the use of a broad based NGS platform at the time of diagnosis.

Highlights.

• A US nationwide clinicogenomic database was used to investigate KRASm and survival across PD-L1 in mNSCLC undergoing frontline ICI-based treatment

• A significant interaction was seen between KRASm and PD-L1 status

• KRASm was associated with worse survival only in PD-L1 0% tumors

• This association was driven by KEAP1 and STK11 comutations, which were enriched in PDL1 0%