KRAS is the most common oncogenic driver in lung adenocarcinoma (LUAC). We previously reported that STK11/LKB1 (KL) or TP53 (KP) comutations define distinct subgroups of KRAS-mutant LUAC. Here, we examine the efficacy of PD-1 inhibitors in these subgroups. Objective response rates to PD-1 blockade differed significantly among KL (7.4%), KP (35.7%), and K-only (28.6%) subgroups (P < 0.001) in the Stand Up To Cancer (SU2C) cohort (174 patients) with KRAS-mutant LUAC and in patients treated with nivolumab in the CheckMate-057 phase III trial (0% vs. 57.1% vs. 18.2%; P = 0.047). In the SU2C cohort, KL LUAC exhibited shorter progression-free (P < 0.001) and overall (P = 0.0015) survival compared with KRASMUT;STK11/LKB1WT LUAC. Among 924 LUACs, STK11/LKB1 alterations were the only marker significantly associated with PD-L1 negativity in TMBIntermediate/High LUAC. The impact of STK11/LKB1 alterations on clinical outcomes with PD-1/PD-L1 inhibitors extended to PD-L1–positive non–small cell lung cancer. In Kras-mutant murine LUAC models, Stk11/Lkb1 loss promoted PD-1/PD-L1 inhibitor resistance, suggesting a causal role. Our results identify STK11/LKB1 alterations as a major driver of primary resistance to PD-1 blockade in KRAS-mutant LUAC.
Significance: This work identifies STK11/LKB1 alterations as the most prevalent genomic driver of primary resistance to PD-1 axis inhibitors in KRAS-mutant lung adenocarcinoma. Genomic profiling may enhance the predictive utility of PD-L1 expression and tumor mutation burden and facilitate establishment of personalized combination immunotherapy approaches for genomically defined LUAC subsets. Cancer Discov; 8(7); 822–35. ©2018 AACR.
See related commentary by Etxeberria et al., p. 794.
This article is highlighted in the In This Issue feature, p. 781
Despite improvements in overall survival (OS) and clinical responses of unprecedented duration with the use of therapeutic mAbs that target programmed cell death-1 (PD-1) or programmed cell death-1 ligand (PD-L1), the majority of patients with non–small cell lung cancer (NSCLC) fail to respond to PD-1/PD-L1 axis inhibitors (1–8). The landscape of primary resistance to PD-1 blockade in NSCLC is largely unknown, with no single factor capable of accurately segregating responders from nonresponders. Expression of PD-L1 on the membrane of tumor and immune cells is associated with enhanced objective response rates (ORR) to PD-1/PD-L1 inhibition, but is neither sensitive nor specific (1–3, 7, 9–12). A higher burden of nonsynonymous somatic mutations [tumor mutation burden (TMB)] further correlates with increased likelihood of clinical benefit and is undergoing evaluation as a predictive biomarker in many tumor types (4, 13–15).
KRAS mutations are the most prevalent oncogenic driver in NSCLC, accounting for approximately 25% of lung adenocarcinoma (LUAC; refs. 16, 17). We previously reported that co-occurring genomic alterations in KRAS and the STK11/LKB1 (KL) or TP53 (KP) tumor suppressor genes define subgroups of KRAS-mutant LUAC with distinct biology, therapeutic vulnerabilities, and immune profiles (18). STK11/LKB1 encodes a serine threonine kinase with an established role in the regulation of cellular metabolism/energy homeostasis, growth, and polarity through phosphorylation of adenosine monophosphate-activated protein kinase (AMPK) and 12 AMPK-related kinases (19). Inactivation of STK11 by mutational or nonmutational mechanisms is associated with an inert or “cold” tumor immune microenvironment, with reduced density of infiltrating cytotoxic CD8+ T lymphocytes in both human tumors and genetically engineered murine models (18, 20, 21). On the basis of these findings, we hypothesized that STK11/LKB1 genomic alterations may predict for lack of clinical benefit from PD-1/PD-L1 blockade in KRAS-mutant LUAC and conducted a study to address this hypothesis and examine the interrelationship between individual genetic alterations, TMB, and PD-L1 expression.
One hundred seventy-four patients who met the prespecified eligibility criteria were included in the Stand Up To Cancer (SU2C) dataset [MD Anderson Cancer Center (MDACC; N = 62), Memorial Sloan Kettering Cancer Center (MSKCC; N = 56), Dana-Farber Cancer Institute/Massachusetts General Hospital (DFCI/MGH; N = 56); Table 1]. The overall cohort was representative of the general population of patients with KRAS-mutant LUAC with median patient age of 66 years (range, 42–87), high percentage of current/former smokers (88.5%), and typical frequencies of distinct KRAS-mutant alleles (Supplementary Fig. S1A and S1B; refs. 16–18, 22). Across the entire cohort, 31% of tumors were classified as KL, 32% were KP, and 37% K-only (Supplementary Fig. S1C). The majority of patients received PD-1 inhibitor monotherapy (165/174, 95%) and the remainder received combination with CTLA4 blockade (9/174, 5%; Supplementary Fig. S1A). Demographic and clinical characteristics were generally well balanced between the comutation-defined subgroups (Supplementary Fig. S1A and S1D). STK11/LKB1 mutations were in their overwhelming majority predicted to be deleterious (Supplementary Fig. S2).
|Cohort .||.||Stand Up To Cancer (SU2C) .||.||.|
|.||Foundation Medicine (FM) .||MDACC .||MSKCC .||DFCI/MGH .||Checkmate-057 (CM-057) .||MDACC (PD-L1 ≥ 1%) .|
|N .||924 (346 KRASMUT) .||174 .||.||.|
|.||.||62 .||56 .||56 .||44 .||66 .|
|anti-PD(L)-1 + anti-CTLA4||NA||9||NA||3|
|Cohort .||.||Stand Up To Cancer (SU2C) .||.||.|
|.||Foundation Medicine (FM) .||MDACC .||MSKCC .||DFCI/MGH .||Checkmate-057 (CM-057) .||MDACC (PD-L1 ≥ 1%) .|
|N .||924 (346 KRASMUT) .||174 .||.||.|
|.||.||62 .||56 .||56 .||44 .||66 .|
|anti-PD(L)-1 + anti-CTLA4||NA||9||NA||3|
aOne patient with STK11/LKB1–mutant tumor was treated with nivolumab and NKTR-214 (CD122-based agonist) and one patient with STK11/LKB1 wild-type tumor was treated with pembrolizumab and OX40 agonist.
Comutations in STK11/LKB1 Are Associated with Inferior Clinical Outcome with PD-1 Blockade in Multiple Independent Cohorts of KRAS-mutant LUAC
The objective response rates to PD-1 inhibition in KL, KP, and K-only groups were significantly different (P < 0.001, Fisher exact test; Fig. 1A and C). KL tumors were mostly resistant to PD-1 axis blockade (ORR 7.4% overall), with consistently low response rates seen in each of the three independent datasets (MDACC: 9.1%, MSKCC: 9.1%, DFCI/MGH: 4.8%). In contrast, KP LUAC were more sensitive to PD-1 inhibitors (ORR 35.7% overall). K-only tumors with no identifiable mutations in either STK11/LKB1 or TP53 had an intermediate response rate (28.6%). Assessment of additional co-occurring genetic alterations in the few KL tumors that responded to PD-1 blockade did not identify any obvious unifying molecular features (Supplementary Fig. S3).
To replicate these findings in the context of a randomized clinical trial, we further analyzed the impact of STK11/LKB1 and TP53 genetic alterations on clinical outcomes in 44 patients with KRAS-mutant NSCLC (96% LUAC) with available whole-exome sequencing (WES) data that were randomly assigned to treatment with nivolumab (n = 24) or docetaxel (n = 20) in the CheckMate-057 randomized phase III clinical trial (NCT01673867). In agreement with data from the SU2C cohort, ORR differed significantly between the KL, KP, and K-only subgroups in the nivolumab arm of CM-057 (P = 0.047), with KL tumors being refractory (ORR: 0%, 0/6) and KP more sensitive (ORR: 57.1%, 4/7) to nivolumab (Fig. 1B). Although ORR did not differ significantly among the three subgroups in the docetaxel arm (P = 0.65), it is relevant to note that the ORR in the KL subgroup was 0% [0/3; ORR was 0% (0/6) and 18.2% (2/11) in KP and K-only subgroups, respectively]. Given the relatively small numbers within subgroups, it cannot be determined whether STK11/LKB1 mutation is prognostic or predictive of treatment outcomes in the CM-057 dataset.
Progression-free survival (PFS) differed among the three groups in the SU2C cohort (P = 0.0018), with significantly shorter PFS for patients with KL compared with either KP [HR 1.77; 95% confidence interval (CI), 1.16–2.69; P = 0.0072] or K-only tumors (HR 1.98; 95% CI, 1.33–2.94; P < 0.001) in pair-wise comparisons (Fig. 2A, left). In contrast, patients with KP and K-only tumors had similar PFS. Because STK11/LKB1 abrogation likely determines immunotherapy resistance in this context, we further compared PFS in patients with STK11/LKB1 wild-type and mutant tumors by merging the KP and K-only cohorts. PFS was significantly shorter in KL tumors compared with KRAS-mutant LUAC with wild-type STK11/LKB1 (HR 1.87; 95% CI, 1.32 to 2.66; P < 0.001; Fig. 2A, right). The CM-057 study had limited power to detect PFS or OS differences due to the small size of subgroup cohorts, and no significant differences were seen in PFS or OS in either arm (Supplementary Figs. S4 and S5).
OS also varied significantly among the three groups in the SU2C cohort (P = 0.0045; Fig. 2B, left). Median OS was 6.4 months in KL compared with 16.0 months in KP and 16.1 months in K-only LUACs. In the two-group comparison, OS was significantly shorter in STK11/LKB1-mutant compared with wild-type tumors (HR 1.99; 95% CI, 1.29–3.06; P = 0.0015; Fig. 2B, right). KRAS subgroup remained a significant independent predictor of OS on multivariate analysis (P = 0.00055). Notably, STK11/LKB1 mutation or deficiency were not associated with worse OS in The Cancer Genome Atlas (TCGA) cohort, arguing against a purely prognostic role for STK11/LKB1 inactivation in this setting of predominantly early-stage, surgically resected tumors (Supplementary Fig. S6), in agreement with previous studies in metastatic tumors (23–25).
Because nonmutational mechanisms can also account for STK11/LKB1 inactivation in LUAC (19), we further assessed expression of STK11/LKB1 by IHC in a subset of tumors for which archival tissue was available (26). KRASMUT;STK11/LKB1MUT (KL) tumors expressed low to undetectable levels of LKB1, whereas KRASMUT;STK11/LKB1WT tumors displayed variable levels of STK11/LKB1 expression, with 17.6% having a LKB1 H-score of zero (Fig. 3A). Patients bearing STK11/LKB1-deficient tumors (STK11/LKB1MUTor STK11/LKB1WT and LKB1 H-score zero) exhibited significantly shorter PFS (HR 1.80; 95% CI, 1.15–2.82; P = 0.0094; Fig. 3B, left) and OS (HR 2.03; 95% CI, 1.13–3.65; P = 0.016; Fig. 3B, right) compared with those harboring STK11/LKB1-proficient tumors (STK11/LKB1WT and LKB1 H-score > 0).
STK11/LKB1 Mutations Are Significantly Enriched among TMB Intermediate/High, PD-L1–Negative Tumors
In a parallel, unbiased analysis, we sought to identify candidate genomic drivers of absent PD-L1 expression (as an indicator of a “cold” or non-T cell–inflamed immune microenvironment) in LUAC using the large Foundation Medicine (FM) dataset (Supplementary Fig. S7). We focused on TMB intermediate and high (TMBI/H) tumors and excluded TMB low (TMBL) LUAC because low TMB has been associated with impaired response to PD-1 axis inhibitors in retrospective studies, likely due to poor tumor immunogenicity (4, 13). We then compared the prevalence of individual genomic alterations in PD-L1 negative (PD-L1Neg; TMBI/H) versus high positive (PD-L1HP; TMBI/H) tumors (Fig. 4A). This analysis identified STK11/LKB1 as the only significantly enriched gene in the PD-L1–negative group (adjusted P < 0.001). Further interrogation of the PD-L1/TMB landscape indicated that STK11/LKB1 alterations are most prominently enriched in TMBI; PD-L1Neg samples, and LUAC-bearing STK11/LKB1 alterations are less likely to be either PD-L1HP or TMBL (Fig. 4B). Furthermore, STK11/LKB1 was significantly enriched (P < 0.001) for PD-L1Neg; TMBI/H-negative status even when the analysis was restricted to KRAS-mutant samples. Thus, we conclude that STK11/LKB1 is associated with higher likelihood of absent PD-L1 expression among TMBI/H tumors irrespective of KRAS status.
We further analyzed PD-L1 expression and TMB in the KL/KP/K-only subgroups and their KRAS wild-type counterparts (Fig. 4C; Supplementary Fig. S8). PD-L1 expression varied significantly between the KRAS subgroups (Fig. 4C, P < 0.001), with KL least likely to be PD-L1HP (P < 0.001; Supplementary Fig. S8). Among KRAS wild-type tumors, STK11/LKB1 alterations were also associated with lower likelihood of high PD-L1 expression (Supplementary Fig. S8). KP LUAC exhibited the highest rate of PD-L1 positivity (56.3% PD-L1–positive, 31.3% PD-L1HP) followed by the TP53-altered, KRAS wild-type group (32.3% PD-L1-positive, 11% PD-L1HP; Supplementary Fig. S8). In contrast, median TMBs across samples with KRAS, STK11/LKB1, or TP53 alterations were comparable, ranging between 8.1 and 11.7 mutations/Mb (Fig. 4D; Supplementary Fig. S8).
Consistent with the observed association between STK11/LKB1 genomic alterations and low PD-L1 expression in the FM cohort, significant difference in the rate of PD-L1 positivity (PD-L1 ≥ 1%) was also noted among KL, KP, and K-only tumors in the SU2C and CheckMate-057 cohorts (Fig. 4C; Supplementary Fig. S9), with KL exhibiting the lowest frequency of PD-L1–positive (13.6% in the SU2C and 11.1% in the CM-057 cohort) and PD-L1HP tumors (0% in both cohorts; Fig. 4C; Supplementary Fig. S9).
We further directly interrogated the composition of the tumor immune microenvironment in surgically resected LUAC specimens (PROSPECT cohort) with available WES and automated quantitative IHC-based immune profiling (21, 27). In agreement with lower tumor cell PD-L1 expression, STK11/LKB1-mutated tumors exhibited lower densities of infiltrating CD3+ (P = 0.0019) and CD8+ (P = 0.0072) T lymphocytes but not FOXP3+ cells (P = 0.7648; Supplementary Fig. S10A). Furthermore, in the TCGA dataset STK11/LKB1-mutated LUAC (or STK11/LKB1-deficient LUAC as determined using a previously validated gene expression signature; ref. 28) exhibited lower T-cell signature scores (ref. 29; Supplementary Fig. S10B) and expressed lower CD274 (encoding PD-L1) mRNA levels (Supplementary Fig. S10C). Thus, we provide compelling evidence from multiple independent cohorts that STK11/LKB1 genomic alterations are associated with, and may actually promote, a non-T cell–inflamed immune microenvironment with lack of tumor cell PD-L1 expression, despite an intermediate or high TMB.
STK11/LKB1 Genomic Alterations Are Associated with Primary Resistance to PD-1 Axis Inhibitors in PD-L1–Positive NSCLC
In view of the strong association between STK11/LKB1 genomic alterations and lack of PD-L1 expression on tumor cells, we also sought to examine the impact of STK11/LKB1 mutations on clinical responses to PD-1 axis blockade in PD-L1 positive (≥1%) nonsquamous NSCLC. For this analysis, we identified a distinct cohort of 66 patients with nonsquamous NSCLC (irrespective of KRAS status) treated with PD-1/PD-L1 inhibitors at MDACC, with available STK11/LKB1 genomic profiling and PD-L1 expression (assessed using the FDA-approved 22C3 pharmDx assay). Within this PD-L1–positive group, STK11/LKB1–mutated tumors exhibited significantly lower ORR to PD-1/PD-L1 blockade compared with NSCLC with intact STK11/LKB1 (ORR 0% vs. 34.5%, P = 0.026, Fisher exact test; Fig. 5A), despite inclusion of PD-L1–high-expressing tumors in the STK11/LKB1-mutant group (Fig. 5B). Importantly, STK11/LKB1 mutations were associated with dramatically shorter PFS (HR 4.76; 95% CI, 2.0–11.1, P = 0.00012, log-rank test) and OS (HR 14.3; 95% CI, 3.4–50.0, P < 0.0001, log-rank test) with PD-1 axis blockade (Fig. 5C and D). The effect of STK11/LKB1 genomic alterations on PFS and OS with PD-1/PD-L1 blockade did not differ significantly across PD-L1 high (PD-L1 ≥ 50%) and low (PD-L1 < 50%) groups (Pinteraction = 0.48 for PFS and Pinteraction = 0.59 for OS; Supplementary Fig. S11). We therefore conclude that STK11/LKB1 genomic alterations affect response to PD-1/PD-L1 blockade at least partially independently of PD-L1 status and that their effect likely extends to the entire population of nonsquamous NSCLC regardless of KRAS status. Extension of the effect of STK11/LKB1 inactivation to the broader population of nonsquamous NSCLC is further supported by data from a separate cohort of patients with TMBI/H nonsquamous NSCLC (without available PD-L1 expression) treated with anti–PD-1/PD-L1 therapy, whereby STK11/LKB1 alterations (regardless of KRAS status) were associated with significantly shorter time on drug (HR 2.91; 95% CI, 1.22–6.92; P = 0.0156; Supplementary Fig. S12).
TP53 Comutations May Affect Response to PD-1 Inhibitors in PD-L1–Negative KRAS-Mutant LUAC
As part of an exploratory analysis we interrogated the impact of KRAS comutations on clinical benefit from PD-1 blockade in PD-L1–negative tumors. Among PD-L1–negative KRAS-mutant LUAC in the SU2C cohort (n = 46), disease control rate differed significantly among the subgroups (P = 0.034) and was highest (70%) in KP tumors (Supplementary Fig. S13). The difference in ORR also favored the KP subgroup (30%), but did not reach statistical significance (P = 0.11, Supplementary Fig. S13). This result further supports the notion that the predictive utility of comutations may extend beyond that of PD-L1 expression.
Stk11/Lkb1 Ablation Induces De Novo Resistance to PD-1 Blockade in a Syngeneic Murine Model of KRAS-Mutant LUAC
In order to establish whether primary resistance to immunotherapy is causally linked with STK11/LKB1 inactivation, we generated Stk11/Lkb1-proficient/deficient isogenic derivatives of the LKR13 Kras-mutant murine LUAC cell line (previously established from a spontaneously arising LUAC in the KrasLA1/+ model) using CRISPR/Cas9-mediated biallelic disruption of the Stk11/Lkb1 locus. Upon confirmation of Stk11/Lkb1 knockout (by immunoblotting for the STK11/LKB1 protein; Supplementary Fig. S14), isogenic cell lines were implanted into the right flank of syngeneic recipient mice, and cohorts of tumor-bearing mice were randomized to treatment with anti–PD-L1 mAb or IgG control. Treatment with anti–PD-L1 mAb potently suppressed LKR13-derived tumors, but growth of Stk11/Lkb1-deficient LKR13 knockout (LKR13KO) continued unabated (Fig. 6A). In agreement with findings in human STK11/LKB1-deficient tumors, lower numbers of CD3+CD8+ and CD3+CD8+/PD1+ T lymphocytes were present in Stk11/Lkb1-deficient LKR13KO tumors compared with their Stk11/Lkb1-proficient counterparts, whereas numbers of CD45+ and CD3+CD4+ cells were not significantly different (Supplementary Fig. S15). Furthermore, we did not observe enrichment of tumor-associated neutrophils in the microenvironment of Stk11/Lkb1-deficient tumors in this model (Supplementary Fig. S15), contrary to a previous report (20). Similar results were obtained using a second syngeneic tumor model based on the LKR10 Kras-mutant murine LUAC cell line in response to treatment with anti–PD-1 mAb or isotype control (Fig. 6B). Thus, Stk11/Lkb1 loss directly promotes primary resistance to PD-1/PD-L1 blockade and fosters establishment of a non-T cell–inflamed tumor immune microenvironment in immunocompetent murine models of KRAS-mutant LUAC.
In this study, we identify genomic alterations in STK11/LKB1 as a tumor cell–intrinsic determinant of primary resistance to PD-1 axis blockade in three independent retrospective cohorts of KRAS-mutant LUAC, a fourth cohort of PD-L1–positive NSCLC regardless of KRAS status, as well as in patients with KRAS-mutant NSCLC treated with nivolumab in the pivotal CheckMate-057 randomized phase III clinical trial. Somatic mutations in STK11/LKB1 are prevalent in LUAC (16.7% in the large FM cohort), particularly among KRAS-mutant tumors (25.4% in the combined FM/SU2C cohort) and foster establishment of a non-T cell–inflamed tumor immune microenvironment with frequently undetectable tumor cell PD-L1 expression (18, 20, 21). Furthermore, we show that genetic ablation of Stk11/Lkb1 directly promotes resistance to anti–PD-1/anti–PD-L1 therapy in two immunocompetent Kras-mutant LUAC murine models. Therefore, STK11/LKB1 inactivation represents a major driver of immune escape and innate resistance to PD-1 blockade in KRAS-mutant LUAC.
Our work demonstrates that alterations in STK11/LKB1 are associated with lack of PD-L1 expression in tumor cells across multiple independent cohorts, despite the presence of intermediate or high TMB. This finding is consistent with lower densities of infiltrating CD8+ CTLs in both human and murine STK11/LKB1-deficient tumors. However, the negative impact of STK11/LKB1 genomic alterations on clinical response to PD-1 axis inhibitors also extends to PD-L1–positive NSCLC, indicating that it is at least partially independent of PD-L1 expression. In addition, in an exploratory analysis among PD-L1–negative KRAS-mutant LUAC, KP tumors exhibited more favorable response to PD-L1 blockade, with partial response/stable disease achieved in 7 of 10 patients. Therefore, analysis of STK11/LKB1 and TP53 comutations may help refine response prediction algorithms in both PD-L1 positive and negative tumors as well as in cases where tumor biopsy for assessment of PD-L1 expression is unavailable or impractical but profiling of circulating tumor DNA (liquid biopsy) has been obtained.
It is important to note that nonmutational mechanisms may also account for STK11/LKB1 inactivation in a subset of LUAC (19). Quantitative IHC for STK11/LKB1 can capture STK11/LKB1-deficient tumors in the absence of STK11/LKB1 genomic alterations (26). Therefore, evaluation of LKB1 expression by IHC may further enhance the predictive utility of a composite biomarker panel encompassing PD-L1 expression, TMB, and STK11/LKB1 genomic alterations.
Although our study primarily examined clinical response to PD-1 axis inhibitors in KRAS-mutant tumors, we anticipate that the effect of STK11/LKB1 inactivation extends to the entire LUAC population, regardless of KRAS status. This hypothesis is supported by (i) poor ORR and shorter PFS and OS with PD-1/PD-L1 blockade in STK11/LKB1-mutant tumors in a cohort of PD-L1–positive NSCLC encompassing both KRAS-mutant and wild-type tumors; (ii) shorter time on PD-1 inhibitor in a separate cohort of patients with STK11/LKB1-altered TMBI/H tumors; and (iii) evidence of a “cold” immune microenvironment in STK11/LKB1-altered LUAC irrespective of KRAS status, and was further proposed in a recent separate study (30). However, application to the wider population of nonsquamous NSCLC will require further validation in larger datasets.
The mechanistic basis of T-cell exclusion in STK11/LKB1-deficient tumors is under active investigation and did not constitute a focus of the current study. However, based on established and emergent STK11/LKB1 functions, a number of possibilities are proposed including altered cytokine/chemokine milieu (20), metabolic restriction of effector T cells (31), or impaired antigenicity, possibly as a result of STK11/LKB1-dependent changes in the epigenetic landscape of tumor cells (32). In a set of elegant in vivo experiments, inducible expression of MYC in KrasG12D-driven murine lung adenomas triggered rapid expulsion of CD3+ T lymphocytes [as well as B cells and natural killer (NK) cells] from the tumor microenvironment via induction of IL23 and CCL9 (33). STK11/LKB1 loss has been reported to promote transcriptional upregulation of MYC via the MZF1 transcription factor (34). In a colorectal cancer murine isograft model, T-cell exclusion and suppression of Th1 cell differentiation were mediated by TGFβ signaling (35), which has also been shown to be subject to modulation by STK11/LKB1 (36). Furthermore, loss of PTEN, which, similar to STK11/LKB1 alterations, results in mTOR pathway activation, has been associated with impaired CD8+ T-cell recruitment in melanoma (37). In this tumor type, prior seminal work highlighted active WNT/β-catenin signaling as a key molecular driver of the non-T cell–inflamed phenotype, via ATF3-mediated suppression of CCL4 production and impaired recruitment of CD103+ dendritic cells to the tumor immune microenvironment (29); interestingly, STK11/LKB1 deficiency has previously been associated with WNT pathway activation (38). Finally, it was recently demonstrated that tumor cell–derived prostaglandin E2 (PGE2) can also impair recruitment of conventional type 1 dendritic cells (cDC1) to the tumor microenvironment both directly, through downregulation of chemokine receptor expression in cDC1, and indirectly, via attenuation of NK-cell viability and function (39). It was previously reported that expression of COX2, which catalyzes the conversion of arachidonic acid to prostaglandins, is enhanced in STK11/LKB1-deficient NSCLC cells via activated CRTC1 (40). Thus, multiple and potentially nonoverlapping mechanisms may underpin establishment and maintenance of a “cold” tumor immune microenvironment in STK11/LKB1-deficient NSCLC, and further work is required to elucidate nodal downstream effectors and signaling cascades.
Delineation of pathways and mechanisms of immune escape downstream of STK11/LKB1 inactivation is also critical in order to inform rational combination therapeutic approaches aimed at invigorating antitumor immunity. Several strategies to convert non-T cell–inflamed tumors into T cell–inflamed tumors have been proposed and are undergoing preclinical and clinical evaluation, including activation of innate immune recognition with STING agonists, TLR agonists, ionizing radiation, or expression of LIGHT in tumor cells (41–44). Such approaches, as well as efforts that tackle specific STK11/LKB1 loss–dependent immunosuppressive cascades, will require prospective evaluation in patients with STK11/LKB1–deficient NSCLC.
In contrast, evidence of preexisting CD8+ T-cell infiltrate and adaptive immune resistance in the majority of KP (and possibly K-only STK11/LKB1-proficient) tumors supports simultaneous targeting of multiple immune inhibitory pathways in this subgroup (42).
Taken together, our data reveal a novel, frequent driver of de novo resistance to PD-1 blockade in KRAS-mutant LUAC and potentially the entire LUAC population. More broadly, somatic genomic alterations in individual genes may modulate the efficacy of PD-1/PD-L1 inhibitors in given tumor types and across tumor types. Although the fully integrated set of determinants of response to these agents is not yet completely defined, our results suggest that the development of tailored immunotherapy approaches for NSCLC may be facilitated by genomic profiling to allow simultaneous characterization of specific somatic alterations including KRAS, STK11/LKB1, and TP53, in addition to TMB and PD-L1 expression.
Patients with stage IV KRAS-mutant LUAC who received at least one cycle of PD-1 inhibitor therapy or combined PD-1/PD-L1 and CTLA4 blockade, were alive for ≥14 days thereafter, and had available molecular profiling of KRAS, STK11/LKB1, and TP53 were identified by retrospective electronic medical record review. Three independent cohorts were studied from members of the Stand Up To Cancer/American Cancer Society Lung Cancer Translational Research Dream Team: MDACC, MSKCC, and a combined cohort from DFCI/MGH, cumulatively forming the SU2C cohort. A fourth cohort of 66 patients with PD-L1–positive (≥1%) nonsquamous NSCLC (regardless of KRAS status) from MDACC with available tumor molecular profiling was assessed to determine the impact of STK11/LKB1 genomic alterations on clinical outcomes with anti–PD-1/PD-L1 therapy specifically in PD-L1–positive tumors. The study was conducted in accordance with the ethical standards outlined in the Declaration of Helsinki and its subsequent amendments. All participating patients at each institution provided written informed consent for the collection of clinical, demographic, and molecular data as well as the use of tissue for IHC and molecular studies, which proceeded in accordance with IRB-approved protocols at each of the participating institutions.
A fifth independent cohort of 44 patients with KRAS-mutant NSCLC (24 treated with nivolumab and 20 treated with docetaxel) with available STK11/LKB1 and TP53 mutational status and tumor cell PD-L1 expression from the CheckMate-057 (CM-057) international phase III randomized controlled trial (NCT01673867) was also analyzed (45).
Finally, a separate large cohort of 924 unselected patients with LUAC who submitted samples to FM for hybrid capture-based comprehensive genomic profiling (CGP) were included in an integrated analysis of TMB, PD-L1 expression, and genomic alterations of individual cancer-related genes. Duration of therapy (“time on drug”) was known for a subset of patients with CGP that received PD-1/PD-L1 inhibitors. Approval for this study, including a waiver of informed consent and a HIPAA waiver of authorization, was obtained from the Western Institutional Review Board (protocol no. 20152817).
Tumor response was assessed by dedicated thoracic radiologists (MDACC, MSKCC, DFCI) or the study investigators (MGH) using Response Evaluation Criteria in Solid Tumors, version 1.1 (RECISTv1.1). ORR was defined as the percentage of patients achieving a confirmed or unconfirmed complete or partial response. Attribution of stable disease as best overall response to therapy required a minimum interval of ≥30 days between the first day of the first cycle of treatment (C1D1) and radiologic evaluation. Patients who died before radiologic reassessment were deemed to have progressive disease. PFS was defined as the time from C1D1 to the date of disease progression or death from any cause. OS was defined as the time from C1D1 to the date of death from any cause. Efficacy endpoints for patients included in CM-057 were evaluated as described previously (45).
In the SU2C cohort, tumor cell PD-L1 expression was assessed in the most recent preimmunotherapy formalin-fixed paraffin-embedded tumor biopsy tissue at each institution using the PD-L1 E1L3N XP rabbit mAb (Cell Signaling Technology), and quantified as the percent of tumor cell membranes exhibiting specific staining of any intensity. PD-L1 expression in CM-057 samples was assessed using the Validated 28-8 pharmDx Assay (Dako; ref. 3). PD-L1 staining of tumor samples submitted to FM was performed using the VENTANA PD-L1 (SP142) assay. Tumors were characterized as PD-L1 negative (PD-L1 < 1%), low positive (1% ≤ PD-L1 < 50%), or high positive (PD-L1 ≥ 50%). Assessment of PD-L1 expression in the separate cohort from MDACC (N = 66) was based on the FDA-approved 22C3 pharmDx assay (Dako).
Molecular Profiling Platforms and Study Group Definitions
CGP of tumor and/or circulating cell-free tumor DNA utilized CLIA-certified assays available in each of the participating institutions (46–51). Samples from CM-057 underwent WES according to previously described methodology (4). Samples submitted to Foundation Medicine were processed at a CLIA-certified laboratory as described previously (52). TMB was measured by Foundation Medicine as described previously (53). Raw TMB values were measured in units of mutations per Mb and characterized as low (TMB < 6), intermediate (6 ≤ TMB < 20), or high (TMB ≥ 20).
KRAS-mutant LUAC bearing nonsynonymous somatic mutations in STK11/LKB1 and/or mono- or biallelic loss of the STK11/LKB1 locus were denoted as KL. KRAS-mutant LUAC harboring nonsynonymous somatic mutations in TP53 and/or mono- or biallelic loss of the TP53 locus were classified as KP. KRAS-mutant tumors with intact STK11/LKB1 and TP53 were referred to as K-only (these tumors include a multitude of additional genetic alterations in addition to mutant KRAS). Triple-mutant tumors (KRAS;TP53;STK11) were classified as KL (18). In CM-057 tumors bearing nonsynonymous somatic mutations in STK11/LKB1 were denoted as KL. In the FM cohort, a KRAS-mutant LUAC sample was considered altered in STK11/LKB1 (KL) or TP53 (KP), if there was detection of a known nonsynonymous somatic mutation, any truncating alteration, or biallelic loss.
LKR10/LKR10KO or LKR13/LKR13KO Kras-mutant murine LUAC cells (2 × 106) were injected subcutaneously into the right flank of syngeneic recipient male mice (129Sv genetic background). Mice bearing tumors ≥ 200 mm3 were randomly assigned to intraperitoneal treatment with: (i) six doses of 200 μg anti–PD-1 (clone RMPI-14; BioXCell) or isotype control antibody (clone 2A3; BioXCell) administered twice weekly (n = 5–8 mice per group; LKR10/LKR10KO isogenic system) or (ii) six doses of 200 μg anti–PD-L1 (mIgG1-D265AFc clone 80) or IgG control antibody administered twice weekly (n = 8–9 mice per group; LKR13/LKR13KO isogenic system). Tumor caliper measurements were obtained twice weekly. Mice were sacrificed when tumor volume reached 1,500 mm3 (LKR10/LKR10KO) or 2,000 mm3 (LKR13/LKR13KO) or when moribund. Single-cell suspension was established from excised tumors using a commercially available Tumor Dissociation Kit (Miltenyi Biotec) and the gentleMACS Dissociator (Miltenyi Biotec), and cell suspensions were prepared corresponding to 40 mg of gross tumor per 100 μL 1 × PBS/0.05 mmol/L EDTA. One hundred microliters of cell suspension per sample was stained with an antibody cocktail including CD3-FITC (clone 17A2), CD4-PerCP55 (clone RM4-5), CD8-PECy7 (clone 53-6.7), CD45-AF700 (clone 30-F11), CD11b-FITC (clone M1/70), Ly-6G-PerCP/Cy5.5 (clone 1A8), and Zombie dye according to the manufacturer's protocol (BioLegend). Cells were analyzed using the BD FACSCanto multicolor flow cytometer and BD FACSDIVA software. The animal study was approved by the MD Anderson Institutional Animal Care and Use Committee. The LKR13 and LKR10 murine cell lines were generously provided by Dr. Tyler Jacks in 2005. All cell lines tested negative for Mycoplasma in March 2017 using the MycoAlert Mycoplasma Detection Kit (Lonza, LT-07-118). Cells were used in in vivo experiments within 10 passages from thawing.
The significance of the association between the KL, KP, and K-only subgroup allocations and objective response to PD-1 axis blockade was assessed using the Fisher exact test. The Kaplan–Meier method was used to estimate PFS and OS. For the analysis of PFS, data for patients who were alive and had no evidence of disease progression at the time of the PFS data lockout (December 31, 2016) or who were lost to follow-up were censored at the time of the last radiologic tumor assessment. For the analysis of OS, data for patients who were alive or lost to follow-up at the time of the OS data lockout (April 25, 2017) were censored at the time of the last documented patient contact. Analysis of both PFS and OS in the separate cohort of PD-L1–positive patients from MDACC was based on a January 15, 2018 data lockout. Differences between groups in PFS and OS were assessed on the basis of the log-rank test. Bonferroni-adjusted P values were employed to account for multiple comparisons. HRs and the corresponding 95% CIs were computed using the Cox proportional hazards model. The Wald test was applied for testing the HR of 1. P ≤ 0.05 was considered statistically significant for all comparisons, unless stated otherwise.
Analysis of clinical endpoints in CM-057 was conducted as previously described and corresponds to a February 18, 2016, database lock (45).
For TMB and PD-L1 analysis, statistics were calculated using R version 3.3.2 (2016-10-31). The enrichment analysis for LUAC in the PD-L1/TMB landscape was limited to genes altered in >1% of samples (100 genes).
Disclosure of Potential Conflicts of Interest
F. Skoulidis has received speakers bureau honoraria from Bristol-Myers Squibb. M.E. Goldberg has ownership interest (including patents) in Foundation Medicine, Inc. M.D. Hellmann is a consultant/advisory board member for Bristol-Myers Squibb, Merck, Genentech/Roche, AstraZeneca, Janssen, Mirati, and Shattuck Labs. M.M. Awad reports receiving a commercial research grant from Bristol-Myers Squibb and is a consultant/advisory board member for Merck, Bristol-Myers Squibb, and AstraZeneca. J.F. Gainor is a consultant/advisory board member for Bristol-Myers Squibb, Pfizer, Ariad/Takeda, Loxo, Array Biopharma, Amgen, Novartis, Genentech/Roche, and Theravance. R.J. Hartmaier has ownership interest (including patents) in Foundation Medicine. L. Gay has ownership interest (including patents) in Foundation Medicine, Inc. S.M. Ali is a Scientific Advisory Board member at Incsysus. J.A. Elvin has ownership interest (including patents) in Foundation Medicine, Inc. G. Singal has ownership interest (including patents) in Foundation Medicine. J.S. Ross has ownership interest (including patents) in Foundation Medicine. S. Kirov has ownership interest (including patents) in Bristol-Myers Squibb. J. Szustakowski has ownership interest (including patents) in Bristol-Myers Squibb. S.-H.I. Ou is a consultant/advisory board member for Foundation Medicine. N. Peled has received speakers bureau honoraria from Roche, Bristol-Myers Squibb, MSD, Pfizer, and Guardant360 and is a consultant/advisory board member for Guardant, Roche, MSD, Bristol-Myers Squibb, Takeda, Guardant360, AZ, and BI. M. Nishino reports receiving a Merck investigator studies program grant to institution, a Toshiba Medical Systems research grant to institution, and an AstraZeneca research grant to institution; is a consultant/advisory board member for WorldCare Clinical, Toshiba Medical Systems, and Daiichi Sankyo; and has received honoraria from Roche and Bayer. J.L. Sauter has ownership interest (including patents) in Merck. J. Zhang is a consultant/advisory board member for AstraZeneca. V.A. Miller is on the Board of Directors at Revolution Medicines and has ownership interest (including patents) in Foundation Medicine and Revolution Medicines. G.M. Frampton has ownership interest (including patents) in Foundation Medicine. J.D. Wolchok reports receiving a commercial research grant from Bristol-Myers Squibb and is a consultant/advisory board member for Bristol-Myers Squibb, Medimmune, and Genentech. P.A. Jänne reports receiving commercial research grants from Astellas Pharmaceuticals, AstraZeneca, Daiichi Sankyo, and PUMA; has ownership interest (including patents) in LabCorp; and is a consultant/advisory board member for AstraZeneca, Boehringer Ingelheim, Roche, Pfizer, Merrimack Pharmaceuticals, Chugai, LOXO Oncology, Ignyta, and Ariad Pharmaceuticals. P.J. Stephens has ownership interest (including patents) in FMI. C.M. Rudin is a consultant/advisory board member for AbbVie, Genentech, Seattle Genetics, Bristol-Myers Squibb, AstraZeneca, Celgene, and Harpoon. L.A. Albacker has ownership interest (including patents) in Foundation Medicine Inc. J.V. Heymach is a consultant/advisory board member for Bristol-Myers Squibb, AstraZeneca, Merck, Genentech, EMD Serono, Boehringer Ingelheim, Spectrum, Lilly, Novartis, and GSK. No potential conflicts of interest were disclosed by the other authors.
Conception and design: F. Skoulidis, M.E. Goldberg, D.M. Greenawalt, J.A. Elvin, J.S. Ross, D. Fabrizio, P.M. Szabo, H. Chang, J. Szustakowski, R. Edwards, N. Sharma, N. Peled, B.W. Carter, W.L. Denning, K.-K. Wong, P.J. Stephens, W.J. Geese, L.A. Albacker, J.V. Heymach
Development of methodology: F. Skoulidis, M.E. Goldberg, M.D. Hellmann, R.J. Hartmaier, L. Gay, J.A. Elvin, J.S. Ross, D. Fabrizio, P.M. Szabo, H. Chang, S. Kirov, P. Vitazka, N. Peled, B.W. Carter, H. Hamdi, J. Rodriguez-Canales, N. Kalhor, A.A. Jungbluth, G.M. Frampton, L.A. Albacker, J.V. Heymach
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): F. Skoulidis, M.E. Goldberg, M.D. Hellmann, M.M. Awad, J.F. Gainor, A.B. Schrock, J.A. Elvin, J.S. Ross, D. Fabrizio, H. Chang, P. Vitazka, J.A. Bufill, N. Sharma, S.-H.I. Ou, N. Peled, H. Rizvi, J. Erasmus, D.F. Halpenny, A.J. Plodkowski, M. Nishino, W.L. Denning, A. Galan-Cobo, H. Hamdi, T. Hirz, J. Rodriguez-Canales, N. Kalhor, A.A. Jungbluth, M. Mino-Kenudson, R. Azimi, Y.Y. Elamin, V.A. Papadimitrakopoulou, I.I. Wistuba, G.M. Frampton, A.T. Shaw, P.A. Jänne, C.M. Rudin, W.J. Geese, L.A. Albacker, J.V. Heymach
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): F. Skoulidis, M.E. Goldberg, D.M. Greenawalt, M.D. Hellmann, M.M. Awad, J.F. Gainor, A.B. Schrock, R.J. Hartmaier, S.E. Trabucco, S.M. Ali, J.A. Elvin, G. Singal, J.S. Ross, D. Fabrizio, P.M. Szabo, H. Chang, A. Sasson, S. Srinivasan, S. Kirov, J. Szustakowski, N. Peled, D.R. Spigel, B.W. Carter, A.J. Plodkowski, N.M. Long, M. Nishino, P. Tong, J. Wang, P.A. Villalobos, E.R. Parra, N. Kalhor, L.M. Sholl, J.L. Sauter, M. Mino-Kenudson, J. Zhang, F. Jiang, K.-K. Wong, J.J. Lee, V.A. Papadimitrakopoulou, V.A. Miller, G.M. Frampton, J.D. Wolchok, A.T. Shaw, P.J. Stephens, C.M. Rudin, W.J. Geese, L.A. Albacker, J.V. Heymach
Writing, review, and/or revision of the manuscript: F. Skoulidis, M.E. Goldberg, D.M. Greenawalt, M.D. Hellmann, M.M. Awad, J.F. Gainor, A.B. Schrock, L. Gay, S.M. Ali, J.A. Elvin, G. Singal, J.S. Ross, D. Fabrizio, P.M. Szabo, H. Chang, J. Szustakowski, R. Edwards, N. Sharma, S.-H.I. Ou, N. Peled, D.R. Spigel, E.J. Aguilar, B.W. Carter, J. Erasmus, D.F. Halpenny, A.J. Plodkowski, N.M. Long, M. Nishino, E.R. Parra, L.M. Sholl, J.L. Sauter, M. Mino-Kenudson, J. Zhang, K.-K. Wong, J.J. Lee, V.A. Papadimitrakopoulou, I.I. Wistuba, V.A. Miller, G.M. Frampton, J.D. Wolchok, A.T. Shaw, P.A. Jänne, P.J. Stephens, C.M. Rudin, W.J. Geese, L.A. Albacker, J.V. Heymach
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.E. Goldberg, G. Singal, J.S. Ross, H. Chang, E.J. Aguilar, A.J. Plodkowski, J.L. Sauter, G.C. Leonardi, J.V. Heymach
Study supervision: J.S. Ross, H. Chang, L.A. Albacker, J.V. Heymach
We thank the patients and their families who participated in this study. We would also like to thank Dr. Patrick Hwu for critical review of the manuscript and Emily Roarty for editorial assistance. This research was supported by a Stand Up To Cancer - American Cancer Society Lung Cancer Dream Team Translational Research Grant (grant number: SU2C-AACR-DT17-15). Stand Up To Cancer (SU2C) is a program of the Entertainment Industry Foundation. Research grants are administered by the American Association for Cancer Research, the scientific partner of SU2C. Research was also supported by a Sabin Family Foundation Fellowship (to F. Skoulidis), a Khalifa Bin Zayed Al Nahyan Foundation Scholar Award (to F. Skoulidis), the Jane Ford-Petrin Fund, the Gil and Dody Weaver Foundation, the Bill and Katie Weaver Charitable Trust, a Cancer Prevention Research Institute of Texas Multi-Investigator Research Award (RP160652; to J.V. Heymach), 1R01 CA205150 (to J.V. Heymach), a V Foundation Translational Award (to J.V. Heymach), generous philanthropic contributions to The University of Texas MD Anderson Lung Cancer Moonshots Program, Lung SPORE grant 5 P50 CA070907, MD Anderson Cancer Center Support Grant P30 CA016672, MSKCC Core Grant P30 CA008748, and 1R01 CA203636 (M. Nishino).