A Comprehensive Mutational and Histopathological Analysis of STK11-Mutant Non-Small Cell Lung Carcinomas

Abstract

Despite recent advances in the understanding of genomic and immunopathologic mechanisms of lung cancer, this disease remains the leading cause of cancer-related deaths in the United States. STK11 (LKB1) mutations are present in approximately 20% of non-small cell lung cancers (NSCLCs) and drive tumor progression through disruption of cellular metabolism, polarity, and stress responses ultimately leading to immune evasion and resistance to cancer therapy. Although these tumors are associated with poor prognoses, the clinicopathologic significance of different STK11 mutation subtypes and their associations with tumor histology, cellular behaviors, metastatic potential, and clinical outcomes remain incompletely understood. In this study, we retrospectively analyze a large cohort of STK11-mutant and STK11 wild-type NSCLCs using a combination of next generation sequencing, immunologic biomarkers, histopathologic characterization, and radiographic imaging. Overall, we demonstrate that STK11-mutant tumors display diverse molecular and morphologic features and are associated with high rates of aggressive histopathologic growth patterns, lymphovascular invasion, and spread through the airspaces (STAS). Among Stage 4 cases, STK11 mutations have notable differences in organotropism, with the STK11-loss cohort in particular demonstrating the highest rates of brain metastases at the time of initial diagnosis. Furthermore, among Stage 4 cases, while all STK11 mutation types result in decreased overall survival probability compared to the STK11 wild-type cohort, the effect appears most pronounced among the STK11-loss/KRAS-mutant group. These findings underscore the heterogeneity of STK11-mutant NSCLCs and highlight the opportunity for further investigation into specific STK11 mutation subtypes in guiding prognosis and therapeutic decision making for individuals with lung cancer.

INTRODUCTION

Lung cancer is the third most common type of cancer in the United States and remains the leading cause of cancer-related death.¹ Over the last several decades, significant advancements have been made towards understanding the genetic and immunopathologic underpinnings of lung cancer, leading to the development of mutation-targeted and immunology-based therapies, respectively. However, nearly half of non-small cell lung cancers (NSCLCs) harbor pathogenic mutations without any available targeted therapies.² One such currently non-targetable gene, STK11 (Serine/Threonine Kinase 11), also known as Liver Kinase B1 (LKB1) is mutated in approximately 20% of non-squamous NSCLCs. STK11 is a tumor suppressor gene critical for cell metabolism, growth, and immune regulation. Under homeostatic conditions, STK11 activates the AMP-kinase signaling pathway subsequently inhibiting mTOR signaling. Furthermore, STK11 has been shown to activate p53-mediated transcriptional activity critical for the regulation of cell proliferation and apoptosis. Loss of STK11 leads to reduced AMP-kinase phosphorylation, mTOR activation, and decreased p53 activity resulting in uncontrolled cellular proliferation. ³

STK11-mutant NSCLCs are aggressive tumors associated with impaired response to immunotherapy and chemotherapy, higher risk of disease progression and metastasis, and an overall poor prognosis.⁴ STK11 mutations often co-occur with KRAS (~50%) and KEAP1 (~30%) mutations, forming a distinct molecular subtype with poor prognosis as largely driven by resistance to upfront PD-1/PD-L1 immune checkpoint inhibitors in the treatment of advanced stage NSCLC and also to KRAS-targeted therapies in the presence of a KRAS-G12C actionable mutation.^5–7 Despite their aggressive clinical nature, the pathophysiology of STK11-mutant NSCLCs remains incompletely understood. Proper interpretation of the impact of a specific STK11 mutation on protein function is therefore potentially relevant to guiding clinical care, yet this can prove challenging as STK11 mutations are highly variable in both their specific alteration type and genomic location. Although in-vitro studies using functional tests, including STK11 autophosphorylation and luciferase-based TP53 transcriptional activation assays, have attempted to determine the pathogenic impact of a number of STK11-missense and splice site variants, many mutations remain incompletely characterized both in-vitro and in real world cases.⁸ Interestingly, a previous study demonstrated that the specific exons where STK11 mutations occur may influence their clinical phenotype among non-squamous NSCLCs, with STK11 exon 1-2 mutations conferring a stroma-derived signature and poorer prognosis.⁹ However, there currently remains a lack of correlation between specific STK11 genomic information and prognostically relevant histopathologic features of NSCLC, including architectural growth patterns, lymphovascular invasion (LVI), and spread through the airspaces (STAS).

To address this knowledge gap, in this study we comprehensively characterize a large cohort of STK11-mutant non-squamous NSCLCs by analyzing their genomic alterations and co-mutational profiles, histopathologic features, metastatic patterns, and overall survival outcomes.

MATERIALS AND METHODS

Study cohort

Under local IRB approval, the electronic medical record was queried to identify patients with a pathologically-confirmed diagnosis of non-squamous NSCLC receiving care at the Beth Israel Deaconess Medical Center (BIMDC, Boston, Massachusetts, USA) from 2013-2025. For this study, patients were selected based on STK11 mutational status as identified via next generation sequencing, and a randomized stage-matched cohort of STK11 wild-type cases from the same clinical cohort were selected as a control group. Patient demographics, tumor data, and comprehensive tumor molecular profiling results were compiled via retrospective chart review. Smoking status was defined by self-reported smoking history and divided into two groups. The smoker group included both current and former smokers with a history of >100 cigarettes in their lifetime. The nonsmoker group included both never smokers and individuals who smoked <100 cigarettes in their lifetime (equivalent to never smokers).

Next generation sequencing and PD-L1 analysis

Comprehensive tumor molecular profiling consisted of next generation sequencing (NGS) using the FoundationOne CDx and FoundationOne Liquid CDx tests and programmed cell death ligand (PD-L1) IHC (clone 22C3 pharmDx kit, Agilent Technologies, Santa Clara, CA), both performed by Foundation Medicine Inc. (Cambridge, MA). Both the Foundation Medicine and OncoKB database (https://www.oncokb.org/) available through Memorial Sloan Kettering Cancer Center were utilized to interpret the predicted oncogenic effect of different STK11 mutations. For NGS results, all genetic alterations were noted; however, only those that occurred at a ≥5% frequency within our cohort are reported. Loss is defined at the level of DNA based genomic alterations based on Foundation Medicine NGS testing and can be complete or partial as detected by copy number variation analysis. In order to preserve diagnostic accuracy for Tumor Mutation Burden (TMB) analysis, only solid tumors that generated a numeric result were included. Any blood specimens (n=23) or solid tumor specimens in which TMB was unable to be calculated due to specimen integrity (n=37) were excluded from TMB analysis. Variant Allele Frequency (VAF), defined as the fraction of sequencing reads at a given locus that contain the variant allele compared to the total number of reads covering that locus, were included from Foundation Medicine NGS reports. FoundationOne Liquid CDx results were not included in the VAF analysis.

Metastasis analysis

For the metastasis analysis, metastatic lesions were compiled based on the original radiology reports as reviewed by radiologists at the time of diagnosis. All data was gathered using imaging scans including Computed Tomography (CT), Positron Emission Tomography/Computed Tomography (PET/CT), and Brain Magnetic Resonance Imaging (MRI). All patients included in the metastasis analysis were Stage 4 and treatment-naïve. Individuals who had an incomplete radiologic workup were excluded from analysis.

Statistical and graphical analysis

Graphical analysis was performed using GraphPad Prism software version 10.3.1 (GraphPad Software, Boston, MA). Statistical analysis was performed using R version 4.3.1. Comparisons of numerical variables between groups were conducted using the t-test (age) or Kolmogorov-Smirnov test (tumor mutation burden, VAF) for two groups and the Kruskal-Wallis test for more than two groups (VAF by PD-L1 TPS category). Associations between numerical variables were analyzed using Pearson’s correlation coefficient. Associations between categorical variables were analyzed using Fisher’s exact test. All tests were two-tailed. Benjamini-Hochberg adjustment for multiple testing (adjusted p-value) was applied separately to clinicopathologic characteristics and co-mutation analyses. Kaplan-Meier survival plots were generated by calculating the number of months between the date of initial diagnosis and the date of death or last censored event on Stage 4 cases only. A log-rank (Mantel-Cox) test was utilized to determine statistical significance using GraphPad Prism software. Statistical significance level was set at p < 0.05. Within the figures, asterisks denote the following statistical values: * = p<0.05, ** = p<0.01, *** = p<0.001, **** = p<0.0001.

RESULTS

STK11 mutation type in NSCLC is highly variable and influences the co-mutation profile

The clinicopathologic demographics for the STK11-mutant NSCLC cohort (n=139) and stage matched STK11 wild-type controls (n=196) are summarized in Table 1. The cohorts did not exhibit any significant differences with respect to age, sex, or tumor stage. However, STK11-mutant tumors demonstrated higher rates of poorly differentiated tumors (NSCLC-NOS, p=0.024) and had slightly higher rates of NGS testing performed on primary tumors (p=0.038). Consistent with previous studies, the STK11-mutant cohort had a higher rate of smokers (96% vs. 71% in the STK11 wild-type cohort, p<0.0001), higher TMBs (average of 9.6 muts/mb vs. 7.3 muts/mb in the STK11 wild-type cohort, p<0.0001) and overall lower PD-L1 TPS scores (60% PD-L1 TPS score of 0 vs. 41% in the STK11 wild-type cohort, p<0.0001) compared to the STK11 wild-type cohort.^10–15 The STK11-mutant tumors also had less self-identified Asian individuals compared to the STK11 wild-type cohort (2% vs. 16%, p<0.0001). This is likely due to the fact that STK11 is a smoking associated mutation and there are higher rates of lung cancer in never-smokers of Asian descent (n=12/31 (39%) Asian smokers in the STK11 wild-type group vs. n=3/3 (100%) Asian smokers in the STK11-mutant group). Given the well-established high frequency of co-mutation with KRAS in STK11-mutant NSCLCs, we reexamined these clinicopathologic parameters sub-stratified by KRAS mutational status. Importantly, we observed no statistically significant differences between the groups, with the exception of TMB, which was higher in the STK11-mutant/KRAS wild-type group compared to the STK11-mutant/KRAS-mutant group (average TMB 12.8 muts/mb vs. 6.9 muts/mb, p<0.001, Supplementary Table S1).

Table 1:

Comparison of clinicopathologic demographics between the STK11 wild-type and STK11-mutant cohorts.

	Overall (n=335)	STK11 wild-type (n=196)	STK11-mutant (n=139)	p-value (non-adjusted)	p-value (adjusted)
Age (average, range)	67.7, 39-93	68.5, 39-93	66.7, 39-87	0.141	0.177
Sex				1	1
Male	n=150, 44.8%	n=88, 44.9%	n=62, 44.6%
Female	n=185, 55.2%	n=108, 55.1%	n=77, 55.4%
Smoker	n=274, 81.8%	n=140, 71.4%	n=134, 96.4%	<0.0001	<0.0001
Stage				0.811	0.901
1	n=32, 9.6%	n=19, 9.7%	n=13, 9.4%
2	n=35, 10.4%	n=19, 9.7%	n=16, 11.5%
3	n=71, 21.2%	n=39, 19.9%	n=32, 23.0%
4	n=197, 58.8%	n=119, 60.7%	n=78, 56.1%
Race (self-identified)				<0.0001	<0.001
White	n=246, 73.4%	n=135, 68.9%	n=111, 79.9%
African-American	n=31, 9.3%	n=16, 8.2%	n=15, 10.8%
Hispanic	n=7, 2.1%	n=5, 2.6%	n=2, 1.4%
Asian	n=34, 10.1%	n=31, 15.8%	n=3, 2.2%
Other/NA	n=17, 5.1%	n=9, 4.6%	n=8, 5.8%
Average TMB (muts/mb)	8.4	7.3	9.6	<0.0001	0.0003
TMB high (≥10 muts/mb)	n=79, 27.9%	n=36, 23.7%	n=43, 32.6%	0.111	0.159
PD-L1 TPS				<0.0001	0.0003
0	n=149, 49.3%	n=70, 40.7%	n=79, 60.3%
1-49	n=86, 28.5%	n=49, 28.5%	n=37, 28.2%
≥50	n=68, 22.5%	n=53, 30.8%	n=15, 11.5%
NGS testing substrate				0.038	0.063
Primary	n=147, 43.9%	n=77, 39.3%	n=70, 50.4%
Metastasis	n=165, 49.2%	n=101, 51.5%	n=64, 46.0%
Blood	n=23, 6.9%	n=18, 9.2%	n=5, 3.6%
Histopathologic subtype				0.024	0.048
Adenocarcinoma	n=296, 88.4%	n=180, 91.8%	n=116, 83.5%
NOS	n=39, 11.6%	n=16, 8.2%	n=23, 17.3%

To more comprehensively characterize any other differences in mutational profiles between STK11 wild-type and STK11-mutant tumors, we next analyzed the NGS testing results for any other genes that showed alterations with a prevalence of ≥5% in these cohorts. Some notable differences were found, including lower frequencies of ALK fusions (0% vs. 5%, p=0.012) and EGFR mutations (2% vs. 28%, p<0.0001) in the STK11-mutant cohort and higher frequencies of KRAS (54% vs. 36%, p=0.002), KEAP1 (34% vs. 5%, p<0.0001), and SMARCA4 (14% vs. 4%, p<0.001) mutations in the STK11-mutant cohort (Supplementary Table S2). When substratifying by KRAS mutation status, we observed that STK11-mutant/KRAS-mutant tumors have lower frequencies of TP53 (29% vs. 72%, p<0.0001), BRAF (1% vs. 13%, p=0.012), and NF1 (3 vs. 19%, p=0.003) mutations and higher frequencies of ATM mutations (15 vs. 2%, p=0.006, Supplementary Table S3).

While the overall genomic profiles of STK11-mutant NSCLCs have been previously examined, these studies most commonly bin all STK11-mutant tumors into a single group. However, STK11 mutations are highly variable, both with respect to their specific type of alteration and their genomic location. The potential pathologic impact of these STK11 variables in NSCLC remains unclear. To address this, we grouped STK11 mutations into three categories: 1.) Loss (defined as the partial or full loss of STK11 based on copy number variation), 2.) Missense, and 3.) Other (which includes frameshift, splice site, and nonsense mutations). Of note, no mutations detected in this cohort were predicted as neutral or likely neutral and all known oncogenic and likely oncogenic mutations in our cohort were classified as loss of function. The STK11-other cohort was the largest (67%), followed by STK11-missense (18%), and STK11-loss (15%) (Figure 1A, Supplementary Table S4). The breakdown of mutation type frequency did not significantly vary when subgrouping by KRAS mutation status, with similar frequencies of each mutation type in both the KRAS wild-type and mutant cohorts (Figure 1A, Supplementary Table S4). Conversely, the frequency of specific KRAS mutation subtypes varied between the STK11 wild-type cohort and the STK11-mutant cohort, with the STK11-mutant cohort demonstrating higher occurrences of both G12C and Q61 mutations (p=0.027, Supplementary Figure S1A). In order to investigate the STK11/KRAS co-mutation relationship further, we next analyzed the STK11 mutation subtype frequency among the KRAS G12C mutant cases compared to all other KRAS mutations. The G12C alteration is clinically relevant, not only as the most common KRAS mutation in NSCLC, but also the one that has FDA-approved specific targeted therapies including Sotorasib and Adagrasib.¹⁶ There was no difference in the frequency of specific STK11 mutation subtypes when comparing KRAS G12C and all other KRAS mutations (p=0.933, Supplementary Figure S1B).

Figure 1. Analysis of STK11 mutation type and frequency.

A.) Analysis of STK11 mutation type across the entire cohort (outer circle, n=139) shows the highest frequency of STK11-other mutations (67%) which includes frameshift (29%), stop (25%), and splice site (13%) mutations, followed by missense (18%) and loss (15%) mutations. No significant differences in mutation type frequency were noted when substratifying by KRAS-mutation status (inner circle). B.) While the TMB trends higher in the KRAS wild-type cohort across all STK11 mutation types, this difference is only statistically significant within the STK11-missense cohort (p=0.006 vs. p=0.191 in the loss and 0.082 in the other cohorts). C.) Analysis of STK11 mutation type by exon location shows a high proportion of frameshift and loss mutations in Exons 1-2 (52% and 18%, respectively vs. 25% and 2% in Exons 3-9) while Exons 3-9 show a high proportion of missense and splice site mutations (26% and 17%, respectively vs. 3% and 6% in Exons 1-2). Red=known oncogenic mutation, purple=likely oncogenic mutation, black=mutation effect unknown. Mutations falling within or spanning the catalytic kinase domain are denoted by the black dotted bar.

A previous study demonstrated that the genomic location of STK11 alterations may influence the clinical behavior of non-squamous NSCLCs through their observation that STK11 mutations located in exons 1-2 confer a more aggressive phenotype and shorter overall survival than those in exons 3-9.⁹ To investigate whether mutation type frequency is influenced by genomic location within our cohort, we next grouped the mutations by exons and found that the majority of cases fell within exons 3-9 (67%) followed by exon 1-2 mutations (24%) and multi-exon losses (9%). These frequencies did not significantly vary by KRAS mutation status (p=0.314). Of note, all missense mutations occurred within the catalytic kinase domain and all but one fall within the exon 3-9 cohort (Figure 1C).

Given the observed differences in co-mutational profiles between both STK11 wild-type and STK11-mutant tumors and STK11-mutant tumors by KRAS mutation status (Supplementary Tables S2 and S3), we sought to understand whether these co-mutational profiles were related to the specific underlying STK11 mutation type. Interestingly, we observed a trend towards higher mutation rates for TP53 (62%, p=0.262), CDKN2A/B (48%, p=0.132), NF1 (24%, p=0.046), and KEAP1 (48%, p=0.337) within the STK11-loss cohort. The STK11-missense cohort showed a trend towards the lowest co-mutation rates for CDKN2A/B (20%, p=0.132) and no instances of co-mutation with SMARCA4 (p=0.047). The frequency of TP53 mutations was increased in the KRAS wild-type cohort compared to the KRAS-mutant cohort for each mutation type (Supplementary Figure S1C, Supplementary Table S4). These findings suggest that the STK11 co-mutation profile varies depending on the specific underlying STK11 mutation type, thus conferring significant genomic heterogeneity to these tumors.

Influence of STK11 mutation type on VAF, TMB, and PD-L1 TPS

Given the wide variety of STK11 mutations, we next questioned whether mutation type influences their clonality/heterogeneity within NSCLCs. To address this, we analyzed the variant allele frequency (VAF) of STK11 non-loss mutations within our cohort and found that the mean STK11 VAF across all specimens was 34%. Both the STK11-missense and STK11-other cohort showed no significant differences in STK11 VAF between the KRAS wild-type and KRAS-mutant cohorts (p=0.16 and 0.859, respectively, Supplementary Figure S2A). The STK11 VAF also did not vary significantly based on sex, histopathologic subtype, PD-L1 TPS, or TMB (Supplementary Figures S2B–E).

As previously discussed and consistent with previously published data, STK11-mutant NSLCSs have higher average TMBs but lower PD-L1 TPS scores. We questioned whether these findings would remain similar regardless of the underlying STK11 mutation or KRAS co-mutational profile. Interestingly, across all STK11 mutation groups, the TMB trended higher in the KRAS wild-type groups compared to the KRAS-mutant groups; however, this difference was only statistically significant in the STK11-missense cohort (p=0.006, Figure 1B). STK11-missense mutations also had a trend towards the highest proportion of PD-L1 TPS high cases (24% vs. 9% average across the loss and other cohorts Supplementary Figure S1D), but the difference in the distributions of PD-L1 TPS group by mutation type was not statistically significant (p=0.317). The average STK11 VAF, TMB, and PD-L1 TPS score did not significantly vary based on exon location (data not shown). These data demonstrate that while overall STK11-mutant tumors may have high TMBs and low PD-L1 TPS scores, these important biomarkers for immunotherapy eligibility may vary with respect to the specific STK11 mutation type.

STK11-mutant tumors are associated with aggressive histopathologic features

Previous studies have correlated STK11-mutant NSCLCs with poor clinical outcomes.^17–19 Therefore, we next performed a comprehensive histopathological characterization of STK11-mutant lung adenocarcinoma surgical resection specimens (n=32) and compared them to an STK11 wild-type lung adenocarcinoma cohort (n=27) in order to gain insight regarding the cellular behaviors/patterns that contribute to their aggressive clinical nature. Of note, all treatment naïve resection cases were included and selected from our full cohort, which was based on the availability of NGS results.

We observed that STK11-mutant lung adenocarcinomas demonstrate a trend towards a higher proportion of aggressive histologic growth patterns (Grade 3) compared to the STK11 wild-type cohort (66% vs. 52%, p=0.303). Furthermore, when comparing STK11-mutant and STK11 wild-type tumors, we observed STK11-mutant tumors had higher rates of STAS (94% vs 67%, p=0.016) and a trend towards higher rates of multifocality (25% vs 11%, p=0.20) and lymph node metastases (60% vs 48%, p=0.44). The rates of LVI (69% vs 78%, p=0.56) and VPI (34% vs 30%, p=0.784) were comparable between the two cohorts (Figures 2, 3A, and Supplementary Figure S3A–B). We noted similar average tumor size across the two cohorts (3.4 cm in the STK11-mutant cohort vs. 3.2 cm in the STK11 wild-type cohort, p=0.71) Importantly, the aggressive features in the STK11-mutant tumors were observed in both the KRAS wild-type and KRAS co-mutated cohorts; however, there appeared to be an additive impact of STK11 and KRAS co-mutation towards enhanced rates of STAS in particular (Supplementary Figure S3C, p=0.011). Compared to previously published studies on NSCLC resection cases across all stages (independent of mutation profile), the average frequencies of STAS, LVI, and VPI are approximately 40%, 30-40%, and 20-25%, respectively.^20–27 In comparison, the rates of STAS, LVI, and VPI appear markedly higher within this STK11-mutant resection cohort, highlighting more aggressive histopathologic behaviors.

Figure 2. High risk histologic features of STK11-mutant lung adenocarcinomas.

Resected STK11-mutant lung adenocarcinomas demonstrate a high proportion of high-risk architectural growth patterns, including A.) solid and B.) micropapillary. C.) Spread of tumor through the airspaces (STAS) is frequently seen, as well as D.) visceral pleural invasion (white arrows highlight tumor cells transgressing the pleural elastic lamina). E.) Lymphovascular invasion is a common finding, along with F.) lymph nodal metastases in the thoracic lymph nodes harvested at the time of surgery. (Hematoxylin and Eosin stains, Verhoeff Elastic stain in D; A, B, F 400x original magnification, C-E 200x original magnification)

Figure 3. Analysis of histopathologic features, tumor mutation profile, and VAF for surgically resected STK11-mutant lung adenocarcinomas, stratified by KRAS status.

A.) STK11-mutant lung adenocarcinomas demonstrate high rates of aggressive growth patterns (micropapillary, complex glands, solid: gray-scale bars) as well as high rates of STAS, LVI, VPI, and lymph node metastases independent of KRAS co-mutation status. B.) Mutation profile and VAFs of STK11-mutant resected tumors.

While our cohort showed a high frequency of high-grade (solid, cribriform, and micropapillary) patterns overall, when sub-grouping by mutation type, the STK11-loss mutant-tumors had the highest proportions of grade 3 histology (75%), STAS (100%), LVI (75%), and VPI (50%) despite having the smallest average tumor size (1.9 cm vs. 3.4 cm in the STK11-missense and 3.6 cm in the STK11-other cohorts). When analyzing VAFs for the non-loss resection cases, we observed that tumors with greater proportions of high-grade growth patterns tended to also have higher STK11 VAFs. Interestingly, the high proportions of these aggressive histologic features appear to occur independently of KRAS, TP53, or KEAP1 co-mutation status (Figure 3B).

STK11 mutations promote aggressive behavior and brain metastases

Given the relatively poor prognoses of STK11-mutant NSCLCs, and our observed increase in aggressive histopathologic features, we next questioned whether STK11 mutation status and specific mutational type influences metastatic behavior as assessed at the time of initial clinical presentation. We observed that the STK11 VAF was higher in specimens sequenced from metastatic lesions compared to primary lung tumors (Figure 4A, p=0.001). Furthermore, the average STK11 VAF was lower in Stage 1 tumors compared to Stage 2-4 tumors (p=0.002, Supplementary Figure S4A). While the mean STK11 VAF was higher in metastatic lesions compared to primary tumors, we did not observe a significant difference in the frequencies of the different STK11 mutation types between primary tumors and metastases (p=0.345, Supplementary Figure S4B).

Figure 4. STK11 mutations and association with metastatic behaviors.

A.) NGS specimens analyzed from metastatic lesions demonstrate a higher STK11 VAF than primary tumors (41% vs. 27%, p=0.001, p adjusted=0.006). B.) Flowchart depicting strategy for analysis of STK11 metastases. All data was gathered from time of diagnosis, pre-treatment. C.) Comparison of metastasis locations in STK11 wild-type vs. STK11-mutant NSCLCs at time of diagnosis reveals an increased frequency of brain metastases in STK11-mutant tumors compared to STK11 wild-type (49% vs. 31%, p=0.012 p adjusted=0.09) and a decreased frequency of malignant pleural effusions in STK11-mutant tumors compared to STK11 wild-type (23% vs. 44%, p=0.006, p adjusted=0.09). D.) Sankey diagram demonstrating STK11 mutation type and frequency of metastatic lesions to the top five most prevalent metastatic sites.

To further investigate STK11 VAF we performed longitudinal sequencing analysis on patients with multiple rounds of NGS testing whose first round was performed on the primary lung tumor and subsequent round was performed on a lung recurrence or non-lung solid tissue metastasis (Supplementary Figure S4C). Across the lung recurrence cases, the VAF was higher in the recurrence (VAF 2) than the original primary tumor (VAF 1) in 6 of 7 cases. Across the cases in which the second round of NGS was performed on a metastatic lesion, the VAF was increased in the metastatic recurrence compared to the primary tumor in 5 of the 6 cases. These results, along with the elevated VAF observed in metastatic lesions, argue against mere sampling bias or tumor heterogeneity, but rather provide indirect evidence that STK11-mutant cells may have a selective advantage and a tendency to accumulate over time contributing to disease progression.

The increased mean VAF in STK11-mutant metastatic testing substrates led us to next question if they have distinct patterns of metastasis. To address this, we began by identifying all Stage 4 cases within our cohort, excluding those that had incomplete testing, were lost to follow up, or had multiple (active) primary adenocarcinomas at initial diagnoses. The remaining cases were then grouped by STK11 mutation status, and all imaging scans at the time of diagnosis in treatment-naïve patients were reviewed. All radiographic lesions consistent with metastatic disease were compiled for each patient and grouped by anatomic site (Figure 4B). We observed that STK11-mutant tumors have a significantly higher frequency of brain metastases (p=0.01) and lower frequency of malignant pleural effusions (p=0.006) compared to STK11 wild-type tumors (Figure 4C). The distribution of metastatic lesions in STK11 wild-type vs. STK11-mutant tumors does not significantly vary when substratifying by KRAS mutation status; however, for many sites including the brain, bone, and extrathoracic lymph nodes, there appears to be an additive trend towards higher rates of metastasis in the STK11-mutant/KRAS-mutant tumors compared to the STK11-mutant/KRAS wild-type tumors (p=0.045 for brain, p=0.153 for bone, p=0.013 for intrathoracic lymph nodes, p=0.213 for extrathoracic lymph nodes, Supplementary Figure S4D). Importantly the higher rate of brain metastases in STK11-mutant tumors appears to be independent of TP53 and KEAP1 co-mutational status (Supplementary Figures S4E–F). These data suggest that for STK11-mutant NSCLCs, KRAS co-mutations may have the strongest additive effect towards promoting brain metastases compared to co-mutation with other frequently mutated genes.

Finally, to understand whether metastatic site distribution frequency varies by STK11 mutation type, we examined the top five most frequent metastatic sites within our cohort: intrathoracic lymph nodes, brain, bone, extrathoracic lymph nodes, and intrapulmonary metastases. Interestingly, we observed that while the STK11 mutation types all had similar rates of extrathoracic lymph node and intrapulmonary metastases, the STK11-loss mutations had a trend towards the highest rates of brain (73% vs. 46% average across the other mutation types, p=0.189) and lowest rates of bone and intrathoracic lymph node metastases (bone: 27% vs. 49% average across the other mutation types, p=0.331; intrathoracic lymph node: 82% vs. 93% average across the other mutation types, p=0.174, Figure 4D). These findings highlight that among all STK11 mutation types, loss mutations in particular may have a distinct impact on aggressive clinical behavior in NSCLC.

STK11-mutant tumors are associated with lower survival probability independent of STK11 mutation subtype

STK11-mutant NSCLCs have been clinically established as aggressive tumors with poor response to immunotherapy and shorter progression-free and overall survival.^4,17–19 To understand whether STK11 mutation subtype influences survival, we performed an overall survival analysis on Stage 4 tumors in the STK11-mutant (n=78) and STK11 wild-type (n=119) cohorts. We chose to exclude Stage 1-3 tumors for several reasons. By including only Stage 4 cases, the cohort is more clinically comparable with respect to therapeutic relevance (both medical and surgical), there is reduction of lead-time bias, and more endpoint clarity (i.e. cancer-specific mortality vs. non-cancer causes of death).

First, we examined overall survival by STK11 mutation subtype and observed that the STK11-loss, STK11-missense, and STK11-other groups are all associated with lower probability of overall survival compared to STK11 wild-type tumors (p=0.001, Figure 5A). There was no statistically significant difference in survival amongst the three STK11 mutation subtypes (STK11-loss vs. STK11-missense p=0.198, STK11-loss vs. STK11-other p=0.431, STK11-missense vs. STK11-other p=0.332). Next, we examined the impact of KRAS co-mutation and overall survival, and found that STK11 wild-type/KRAS wild-type tumors demonstrate the best probability of overall survival, single mutant tumors (STK11 WT/KRAS-mutant and STK11-mutant/KRAS WT) demonstrate comparable worse probability of overall survival, and double mutants (STK11-mutant/KRAS-mutant) exhibit the worst probability of overall survival (Supplementary Figure S5A). Interestingly, the additive impact of KRAS co-mutation on STK11-mutant overall survival probability appears most pronounced in the STK11-loss cohort, with STK11-loss/KRAS-mutant demonstrating significantly worse probability of survival than the STK11-loss/KRAS wild-type cohort (p=0.014). The STK11 wild-type, STK11-missense, and STK11-other cohorts showed no statistically significant differences in overall survival probability when stratifying by KRAS mutation status (p=0.077, p=0.780, and p=0.563, respectively, Figure 5B). Within our cohort, we did not observe a statistically significant difference in overall survival between the STK11 exon 1-2 and STK11 exon 3-8 mutant cohorts (p=0.26, Supplementary Figure S5B). Altogether, these results provide important evidence that although Stage 4 STK11-mutant tumors demonstrate worse overall survival compared to STK11 wild-type tumors (independent of STK11 mutation subtype), the synergistic effects of co-mutation with KRAS on decreased survival appear most pronounced in the STK11-loss group.

Figure 5. Survival analysis of Stage 4 STK11-mutant tumors by mutation type and KRAS co-mutation status.

Kaplan-Meier plots demonstrating the overall survival probability and median survival in months of Stage 4 STK11-mutant (n=78) vs STK11 wild-type (n=119) tumors demonstrates A.) significantly worse survival in all STK11-mutant tumors compared to STK11 wild-type, independent of STK11 mutation subtype (p=0.001) and no statistically significant difference in survival between the three STK11 mutation subtypes (STK11-loss vs. STK11-missense p=0.198, STK11-loss vs. STK11-other p=0.431, STK11-missense vs. STK11-other p=0.332). n=13 STK11-loss, n=14 STK11-missense, n=51 STK11-other. B.) KRAS co-mutation is associated with poorer probability of survival in the STK11-loss cohort (p=0.014) but not the STK11-missense (p=0.780), STK11-other (p=0.563), or STK11 wild-type (p=0.077) cohorts.

DISCUSSION

Altogether, this study is the first to provide a comprehensive characterization of STK11-mutant NSCLCs: beginning with an examination of specific mutation types and co-mutation profiles, next analyzing patterns of aggressive histopathologic features on resection specimens, and concluding with a characterization of their metastatic patterns of spread and overall survival analysis for advanced stage tumors. To our knowledge, this is the first study to evaluate the genomic, histopathologic, and metastatic features of STK11-mutant NSCLCs based on the unique STK11 mutation type. This study shows that different types of STK11 mutations demonstrate variable co-mutational profiles and impart different stage-specific clinical behaviors.

Understanding the potential differences these mutations may confer on a biological level can help us to better characterize their clinical impact towards guiding lung cancer screening and treatment. For example, previous research has demonstrated that STK11-mutant NSCLCs exhibit poor response to immunotherapy; however, these studies have not specifically sub-stratified treatment response based on STK11-mutation type.^4,17,19 In our study, we observe that STK11-missense mutant tumors tend to have higher TMBs and PD-L1 TPS scores compared to other STK11 mutation types, which suggests that they may be more likely to respond to immunotherapy than other types of STK11-mutant tumors. Similarly, our data suggests that STK11-loss mutations may confer a different clinical phenotype compared to the other STK11 mutation types, with the highest co-mutation rates for many genes (TP53, CDKN2A/B, KEAP1, and NF1), the lowest rates of PD-L1 TPS high cases, and the highest rates of brain metastases. There appear to be a number of interesting other co-mutational patterns based on STK11 mutation subtype to include an absence of co-mutation with ATM and RB1 within the STK11-loss cohort, and an absence of co-mutation with SMARCA4 within the STK11-missense cohort. These findings, along with the observed differences in clinicopathologic behaviors, suggest that all STK11-mutant tumors are not created equal. Therefore, future clinical trials on STK11-mutant NSCLCs may benefit from classifying cases by their specific mutation type, as treatment response may vary based on the specific genomic alteration present.

Mutations in STK11, particularly when accompanied by KEAP1 alterations, promote a unique metabolic and immune-evasive tumor phenotype in advanced stage NSCLC, leading to resistance to anti-PD-1/PD-L1 monotherapy. This therapeutic resistance represents a significant unmet clinical need and underscores the importance of rational combination approaches. For example, Skoulidis et al. offer strong biomarker-driven evidence supporting the integration of CTLA-4 blockade into treatment regimens for STK11/KEAP1-mutant tumors.²⁸ Dual immune checkpoint blockade (CTLA-4 + PD-1), unlike PD-L1 inhibition alone, reprograms the tumor microenvironment by inducing iNOS-expressing, tumoricidal myeloid cells, and reactivating both CD4⁺ and CD8⁺ effector T cells. Clinically, this translated into meaningful benefit in the POSEIDON phase III trial, where adding Tremelimumab (CTLA-4 inhibitor) to Durvalumab (PD-L1 inhibitor) and chemotherapy nearly doubled overall survival for patients with STK11/KEAP1-mutant NSCLC, from approximately 7.3 months to 15.8 months.

In patients with advanced stage KRAS G12C-mutant NSCLC, combination treatment strategies that target both the primary oncogenic driver and mechanisms of adaptive resistance are under active investigation. These approaches are particularly important in the context of STK11 mutations, where loss of STK11 reduces AMPK activity, resulting in unchecked mTOR and MAPK signaling that synergizes with KRAS-driven oncogenesis to enhance tumor proliferation and survival.³ This metabolic reprogramming underlies intrinsic resistance to KRAS-targeted therapies.⁶ The KRYSTAL-2 trial (NCT04330664) is currently evaluating a combination of Adagrasib and TNO155, a SHP2 inhibitor, based on preclinical evidence that SHP2 inhibition blocks upstream RAS activation, thereby deepening MAPK pathway suppression and sustaining the efficacy of KRAS G12C inhibitors.^29,30 This trial will be crucial for determining the therapeutic potential of this combination in a genomically defined, treatment-resistant NSCLC subset. Together, these combination strategies, including dual checkpoint blockade, metabolic reprogramming, and targeted signal pathway inhibition, represent promising avenues in precision oncology for treating advanced stage NSCLC with STK11 mutations. Furthermore, interpreting these treatment responses in the context of the specific STK11 mutation present may prove useful towards developing a better understanding of the pathophysiology and clinical behavior of these tumors.

Our surgical resection data demonstrates that STK11-mutant tumors are strongly associated with many high-risk histopathologic characteristics, including elevated rates of STAS, LVI, and VPI, and these features are independent of KRAS, TP53, and KEAP1 co-mutation status. Future studies with larger cohorts of surgical resection cases will be helpful to understanding the impact of specific STK11 mutation type on cellular behaviors and growth patterns in NSCLCs, as our relatively small (n=32) set of STK11-mutant tumor resection cases precludes our ability to perform meaningful statistical analysis of these histopathologic features by mutation subtype. This limitation is likely multifactorial, both due to the relatively lower frequency of STK11 mutations among all NSCLCs (~20%) and the fact that these tumors most frequently present at later stages, at which point surgical resection is not an option. One caveat of this analysis, is that all cases in this study were selected based on availability of NGS results. Historically, NGS testing was reserved for NSCLCs that are advanced stage, fail first line therapy, and/or recur, although now it is frequently performed on earlier stage tumors as well for selection of adjuvant therapy following resection. As tumors from this cohort (both STK11 wild-type and STK11-mutant) date as far back as 2013, the NGS tested resection cohort likely includes more clinically aggressive tumors, which may explain why the STK11 wild-type tumors also have a high proportion of aggressive features compared to historic averages. Despite this enrichment, however, the STK11-mutant tumors still appear more histopathologically aggressive than the STK11 wild-type tumors, especially with respect to increased rates of STAS.

Analyzing the pattern of metastatic disease for STK11-mutant NSCLCs demonstrated higher rates of brain metastases compared to STK11 wild-type tumors, with the highest rates of brain metastases seen specifically in the STK11-loss cohort. This finding may be due in part to the association between STK11-loss and acquisition of a more mesenchymal phenotype, as the epithelial-to-mesenchymal transition is believed to play a significant role in the development of brain metastases from NSCLCs.³¹ One caveat of this analysis is that not all lesions radiologically deemed “metastatic” were confirmed pathologically and therefore comprehensive histopathologic comparative analysis of the metastatic sites is lacking. However, our findings are in keeping with a previous study that demonstrated individuals with both a KRAS mutation and LKB1 loss by IHC staining had higher rates of brain metastases at the time of diagnosis.³² Similarly, future studies examining LKB1 immunohistochemical spatial analysis on STK11-mutant NSCLC resection specimens would be highly informative in understanding the functional consequence of specific mutations towards the heterogeneity of these tumors and correlating LKB1 expression with areas of aggressive growth patterns, STAS, and LVI.

While previous research has demonstrated that STK11-mutant NSCLCs are clinically aggressive tumors with shorter progression-free and overall survival, these studies do not take into account specific STK11 mutation subtypes.^4,17–19 In our current study, we demonstrate that STK11-mutant tumors result in decreased overall survival probability compared to STK11 wild-type tumors among Stage 4 cases, independent of the STK11 mutation subtype. However, when further substratifying by KRAS mutation status, the STK11-loss/KRAS-mutant tumors have significantly worse probability of overall survival compared to the STK11-loss/KRAS wild-type tumors. In contrast, the STK11-missense and STK11-other cohorts do not show a significant difference in overall survival probability when examined by KRAS co-mutation status. These data demonstrate the clinical importance of stratifying out STK11 mutation subtype when interpreting survival data and treatment response in future clinical trials.

In conclusion, this study highlights the biologic and clinical heterogeneity of STK11-mutant NSCLCs and suggests that different mutation classes may confer distinct molecular co-mutational profiles, histopathologic behaviors, and metastatic tendencies, all of which have potential implications for prognosis and therapeutic decision-making. As such, future research and clinical trials involving STK11-mutant NSCLCs should incorporate mutation subclassification to more accurately predict tumor behavior and optimize personalized treatment strategies.