Impact of Tumor-intrinsic Molecular Features on Survival and Acquired Tyrosine Kinase Inhibitor Resistance in ALK-positive NSCLC

Abstract While tyrosine kinase inhibitors (TKI) have shown remarkable efficacy in anaplastic lymphoma kinase (ALK) fusion-positive advanced non–small cell lung cancer (NSCLC), clinical outcomes vary and acquired resistance remains a significant challenge. We conducted a retrospective study of patients with ALK-positive NSCLC who had clinico-genomic data independently collected from two academic institutions (n = 309). This was paired with a large-scale genomic cohort of patients with ALK-positive NSCLC who underwent liquid biopsies (n = 1,118). Somatic co-mutations in TP53 and loss-of-function alterations in CDKN2A/B were most commonly identified (24.1% and 22.5%, respectively in the clinical cohort), each of which was independently associated with inferior overall survival (HR: 2.58; 95% confidence interval, CI: 1.62–4.09 and HR: 1.93; 95% CI: 1.17–3.17, respectively). Tumors harboring EML4-ALK variant 3 (v3) were not associated with specific co-alterations but were more likely to develop ALK resistance mutations, particularly G1202R and I1171N (OR: 4.11; P < 0.001 and OR: 2.94; P = 0.026, respectively), and had inferior progression-free survival on first-line TKI (HR: 1.52; 95% CI: 1.03–2.25). Non-v3 tumors were associated with L1196M resistance mutation (OR: 4.63; P < 0.001). EML4-ALK v3 and somatic co-alterations in TP53 and CDKN2A/B are associated with inferior clinical outcomes. v3 status is also associated with specific patterns of clinically important ALK resistance mutations. These tumor-intrinsic features may inform rational selection and optimization of first-line and consolidative therapy. Significance: In a large-scale, contemporary cohort of patients with advanced ALK-positive NSCLC, we evaluated molecular characteristics and their impact on acquired resistance mutations and clinical outcomes. Our findings that certain ALK variants and co-mutations are associated with differential survival and specific TKI-relevant resistance patterns highlight potential molecular underpinnings of the heterogenous response to ALK TKIs and nominate biomarkers that may inform patient selection for first-line and consolidative therapies.


Introduction
The identification of anaplastic lymphoma kinase (ALK) fusions as an oncogenic driver in non-small cell lung cancer (NSCLC) has led to the development remains sparse.Co-occurring genomic alterations can define molecular subgroups within a specific NSCLC subtype with distinct biology and therapeutic sensitivities (8).Consistent with other oncogenic addicted NSCLC subtypes such as EGFR, co-occurring TP mutations appear to be associated with worse prognosis in ALK-positive NSCLC (9)(10)(11).However, other co-alterations in important tumor suppressor genes such as CDKNA/B and SMARCA have also been associated with poorer response to TKI in other oncogene-driven tumors (12) but not yet characterized in ALK-positive NSCLC.Furthermore, the pattern of ALK acquired resistance mutations may be influenced by specific ALK variant subtypes (13), though the impact of these variants on clinical outcomes remains uncertain.
Expanding our understanding of these tumor-intrinsic molecular features may facilitate improved patient stratification for the multiple FDA-approved first-line ALK TKIs and potential combination approaches for treatment intensification.In this study, we sought to comprehensively characterize the clinical impact of co-mutations, variants, and other clinical features using a cohort of patients with ALK-positive NSCLC from two major U.S. cancer centers and genomic data from a large-scale international cohort of ALK-positive patients who underwent commercial circulating tumor DNA (ctDNA) testing.

Clinical Cohort
We retrospectively identified 309 patients with advanced NSCLC at two academic institutions (Johns Hopkins Sidney Kimmel Comprehensive Cancer Center and Memorial Sloan Kettering Cancer Center between January 1, 2010 and December 31, 2022 and March 1, 2005 and August 31, 2021, respectively) harboring somatic ALK fusions from genomic testing of primary tumor specimens or liquid biopsies (Supplementary Fig. S1).ALK fusions were detected utilizing tissue or liquid biopsy-based targeted next-generation sequencing (NGS), or through IHC or FISH performed in Clinical Laboratory Improvement Amendments (CLIA)-certified laboratories.Clinical records were reviewed to determine baseline characteristics, treatment history, and clinical outcomes.
The primary clinical endpoints of interest included progression-free survival (PFS) on first-line TKI (defined as time to clinical or radiographic disease progression or death due to any cause following initiation of first-line ALK TKI administered for stage IV NSCLC, excluding those who received first-line chemotherapy) and OS (defined as time from stage IV diagnosis to death due to any cause).This retrospective study was approved as exempt by the Institutional Review Boards at Johns Hopkins and Memorial Sloan Kettering (New York, NY).All research was performed in accordance with the U.S. Revised Common Rule.
We assessed the prevalence of co-existing individual clinically relevant somatic alterations, including nonsynonymous missense and nonsense singlenucleotide variants (SNV), insertions, deletions, and frameshift mutations in TP, PIKCA, APC, and CTNNB; copy-number loss or loss-of-function mutations in PTEN, CDK/, CDKNA/B, and CCNE; and copy-number alterations (CNA) of MYC and MET (14).These variants were assessed through Patients who underwent tissue-or plasma-based NGS that did not cover a mutation of interest were excluded from the analysis of said mutation.The presence of ALK resistance mutations were noted for patients who underwent additional tissue-or plasma-based NGS at the time of disease progression.If multiple instances of NGS testing was pursued at progression (either for the same line of progression or for multiple lines of progression), the union of each unique resistance ALK mutation was attributed to the patient.We applied Cox proportional hazards survival models to assess effect sizes and generate HR based on ALK variant status and associated somatic mutations.Differences between categorical variables were assessed using Fisher exact test.

Liquid Biopsy Cohort
To further characterize somatic co-mutations and ALK resistance mutations, results from the Guardant Health deidentified clinical database of plasma samples processed between January 1, 2017 and April 1, 2022 were selected for patients with lung cancer with detectable EML-ALK-activating fusion (n = 1,350), as defined by the Guardant360 CDx or laboratory developed (LDT) panel tests (collectively G360; ref. 19).The specific timeframe was selected to capture a contemporary cohort of patients likely treated with second generation or later TKIs.G360 specimens were analyzed for clinically relevant somatic co-mutations and ALK resistance mutations as in the clinical cohort.The G360 CDx and LDT assays report results for SNVs, insertion-deletions, and fusions across up to 83 genes at an average coverage depth of 10,000x, including all ALK exonic regions (17).Somatic profiling of copy-number loss alterations was not included in all liquid biopsy samples.Samples were filtered to include only the most recent testing timepoint for all patients in the selected population (n = 1,118).Fusion breakpoints for each detected mutation were retained for variant subtyping, and fusions were classified into variants based on the exon-exon annotation (20).Relative clonal structure grouping within co-occurring resistance mutations was assessed by recording the maximum and minimum mutant allele frequency (MAF) of all mutations of interest.

Data Availability
Data were generated by the authors but are not publicly available due to patient confidentiality and protection of private health information.Deidentified data may be provided upon reasonable request from the corresponding author.

Baseline Characteristics and EML4-ALK Variant Status
The clinical cohort included a total of 309 patients (Table 1), 273 of whom underwent tissue NGS at baseline at the discretion of their treating oncologist (Supplementary Fig. S2).In this group, the median age at diagnosis was 53 (range: 16-91).Patients were predominantly female (57.9%), without history of smoking (72.8%), and with stage IV disease (80.7% amongst those with staging information available) at diagnosis.The most common tumor histology was adenocarcinoma (96.8%).Of the 230 patients (74.4%) with known ALK fusion breakpoint, 92.2% harbored an EML-ALK fusion.Of the 201 EML-ALK tumors with known variant, v1 was most common (43.8%) followed by v3 (35.8%;Fig. 1).A total of 7.8% of tumors had ALK fusions involving partners other than EML.Most patients received first-line TKI therapy alectinib (57.9%) followed by crizotinib (36.2%).A minority (6.5%) of patients received combination chemotherapy as the first-line treatment for metastatic disease.Of

Somatic Co-mutations and ALK Resistance Mutations in ALK-positive NSCLC
In addition to characterizing driver EML-ALK fusions, we also assessed the prevalence of clinically relevant co-occurring somatic mutations in our two  cohorts.In the clinical cohort, cell cycle and tumor suppressor alterations were most common, driven in large part by TP and CDKNA/B loss co-mutations in 24.1% and 22.5% of patients who underwent NGS, respectively.The most common somatic co-mutation in the liquid biopsy cohort was TP, occurring in 39.5% of specimens (Fig. 1B).There was no correlation between v1 or v3 status and co-occurring somatic mutation in the clinical cohort, while MYC amplification was enriched in v1 patients in the liquid biopsy cohort (Fisher test OR 0.47; P = 0.02; Supplementary Table S2).
We next sought to explore the relationship between EML-ALK variant type and the emergence of putative ALK resistance mutations in response to TKI treatment.Of the 309 patients in the clinical cohort, 50 patients underwent liquid-or tissue-based NGS at time of progression on TKI therapy, of whom 26 patients (52.0%) were found to have on-target ALK resistance mutations (Fig. 2).Acquired ALK resistance mutations were more likely to occur in v3 patients (Fisher test OR 7.08; P = 0.019).The most common resistance mutation were G1202R [seen in 30.8% of patients with resistance ALK mutations and enriched in patients harboring EML-ALK v3 (Fisher test OR 39.2; P < 0.001)], and L1196M (identified in 26.9% of patients who had resistance ALK mutations; Fig. 2; Supplementary Table S3).
In the liquid biopsy cohort, we identified 202 patients (18.1%) with an aggregate of 330 on-target ALK resistance mutations.G1202R (59.4%),L1196M (36.1%), and I1171N (13.9%) were the most common ALK kinase domain mutations detected (Fig. 2).Resistance mutations were most commonly identified in v3 patients (59.9%), followed by v1 (25.2%).Specifically, ALK G1202R and I1171N were both significantly associated with v3 compared with all other variant types (Fisher test OR 4.11; P < 0.001 and OR: 2.94; P = 0.026, respectively), while ALK L1196M was more common among non-v3 patients (Fisher test OR 0.22; P < 0.001).We found no other significant associations between v3 and all other variant types for other ALK resistance mutations for which we had sufficient power to test (Supplementary Table S3).We also assessed the association of ALK resistance mutations with co-existing somatic mutations (Supplementary Table S4).We observed that baseline PIKCA co-mutations were associated with development of ALK resistance mutations (Fisher test OR 2.12; P = 0.028).Collectively, Wnt/β-catenin/PIK3CA pathway mutations were enriched in ALK-resistance positive specimens (Fisher test OR 2.17; P = 0.001), agnostic of variant subtype.No other somatic co-mutations were associated with acquired on-target ALK resistance.

Multiple Resistance Mutations in ALK-positive NSCLC
In our clinical cohort, of the 26 cases found to have acquired ALK resistance mutations, 3 (11.5%)were found to have multiple ALK resistance mutations.
In the liquid biopsy cohort, which included 202 patients with ALK resistance mutations, 80 samples (39.7%) were found to harbor multiple ALK resistance mutations (Supplementary Fig. S3).Notably, D1203N was observed almost exclusively in the context of multiple co-occurring ALK resistance mutations.We found no significant difference in the prevalence of multiple ALK resistance mutations between EML-ALK v1 or v3, the two variants for which we had adequate power for comparison (Supplementary Fig. S3B).
In addition, we assessed whether these multiple ALK mutations in the liquid biopsy cohort were likely to be clonal or nonclonal events (Supplementary Fig. S4).As an estimation, clonal events were defined in this context by a mutation cluster with a maximum MAF range of <5% across resistance mutations.
Of the patients with multiple mutations, a majority (n = 63, 78.8%) had putative clonal resistance events.However, we did not find any association between

Impact of Genomic and Molecular Features on Clinical Outcomes
Finally, we sought to determine the impact of these genomic and molecular features on clinical outcomes.Median OS [84.9 months (95% confidence interval, CI: 73.  and 4B).Furthermore, in a multivariate model accounting for factors with known PFS significance (brain metastasis and use of first-generation TKI crizotinib), the presence of TP co-mutation remained highly predictive of poorer first-line TKI PFS (adjusted HR: 2.72; 95% CI: 1.48-4.97;Supplementary Table S5B).In the subgroup of patients who received first-line alectinib or brigatinib, numerically worse PFS was observed for those that harbored concurrent  S5C; Supplementary Fig. S5B-S5C].

Discussion
In this study, we leveraged the largest cohort to date of patients with advanced ALK-positive NSCLC with genomic profiling data to evaluate molecular characteristics and their impact on acquired resistance mutations and  With clinico-genomic data from a more contemporary cohort of patients primarily treated with second-generation TKIs, we were able to affirm and extend prior observations about the propensity of EML-ALK v3 to develop specific resistance mutations, including strong associations not previously described involving specific resistance mutations I117N and L1196M (13).We found that v3 is not only associated with the highly resistant G1202R acquired mutation but also with I1171N, which is uniquely resistant to alectinib but sensitive to other second-generation TKIs like brigatinib and ceritinib (21).In addition, non-v3 tumors (primarily v1) are associated with L1196M, an acquired gatekeeper mutation that is similarly resistant to alectinib but sensitive to other second-generation TKIs (21,22).We observed that multiple ALK resistance mutations were present at a prevalence of about 40% at the time of acquired ALK-dependent resistance, affirming a prior single-institution series of post-lorlatinib tissue NGS results (23).These co-occurring resistance mutations appear to primarily be clonal events suggestive of compound resistance mutations that are likely to arise as a result of sequential ALK TKI therapy (24).
Consistent with exploratory analyses from the ALTA-1 L and ALEX randomized phase III studies, we found that EML-ALK v3 was associated with shorter first-line PFS in our clinical cohort (25,26).Similar to ALTA-1L, there was no difference in OS of v3 patients, suggesting the effectiveness of later lines of therapy, such as lorlatinib, in addressing differentially acquired resistance mutations such as G1202R and I1171N (27).The underlying biology for this EML-ALK v3 difference in TKI sensitivity remains incompletely understood.
Compared with non-v3 variants, alternative splicing in v3 appears to result in increased isoform heterogeneity that may be associated with worse clinical reponse to crizotinib (28).Provocatively, cell line inhibition experiments have shown decreased crizotinib sensitivity to the v3a isoform compared with v3b and enrichment of v3a-expressing cells with crizotinib, certinib, or alectinib exposure (28)(29)(30).Supported by these prior data, our clinical findings have implications in a setting where the optimal sequencing of ALK TKIs remains an open question.Lorlatinib, a third-generation TKI, has more recently become available as a first-line TKI option for advanced ALK-positive NSCLC, though concerns about toxicity have led to significant reticence in its adoption until more mature survival data emerge (31).Our clinical findings suggest that EML-ALK variant type may be an important predictive biomarker to guide first-line TKI selection or risk stratification in future prospective studies.In this framework, patients with v3 might be prioritized for a more potent third-generation TKI such as lorlatinib, whereas non-v3 patients may still have durable first-line PFS with a less toxic, L1196M-sensitive second-generation TKI (e.g., brigatinib, ceritinib, or ensartinib).
In our clinical cohort, we identified TP mutations and CDKNA/B loss as independent adverse prognostic factors.TP co-mutations occur in up to 40% of patients with advanced ALK-positive NSCLC and is associated with inferior OS and PFS (9)(10)(11)25).CDKNA/B loss is less commonly observed in NSCLC, with inactivating mutations or focal deletions in chromosome 9p occurring in about 12% of advanced NSCLC (32).The significance of somatic CDKNA/B loss in patients with ALK-positive NSCLC is less well described, but has recently been reported to confer worse survival and higher risk of brain metastases (33).The latter finding has been supported by work showing enrichment of CDKNA/B loss in metastatic brain lesions compared with primary tumor specimens in NSCLC (34).Interestingly, the deleterious effect of CDKNA/B loss is not seen with our PFS analysis, especially in those who received second-generation TKIs, perhaps owing to the improved central nervous system protection offered by these agents.
Our liquid biopsy cohort identified PIKCA mutations and, more broadly, PIK3CA/Wnt/β-catenin pathway alterations to be associated with ALK resistance mutations.These mutations have not been well characterized in the landscape of ALK-positive disease, but do not appear to be enriched in progression specimens from patients with EGFR-mutated NSCLC (35).The potential relevance of these mutations in mediating ALK TKI resistance and other adverse clinical outcomes warrants further investigation, especially in light of previous work showing cooperative, non-redundant functions in these pathways in promoting EGFR-mutated tumor progression (14).
Our group and others have shown that ALK-positive NSCLC has relatively high tumor PD-L1 expression (36,37).This is thought to be due to the intrinsic upregulation of PD-L1 via EML-ALK effects on downstream PIK-AKT and MEK-ERK signaling (38) and, importantly, not indicative of anti-PD(L)1 immune checkpoint inhibitor sensitivity (39).While high PD-L1 expression has been associated with primary resistance to TKIs in EGFR-mutated NSCLC (40), its reported impact on TKI clinical outcomes in ALK-positive NSCLC appears more heterogenous (41,42).In our current study, we observed that patients with PD-L1 high tumors on first-line alectinib or brigatinib have a numerically shorter PFS.This trend was attenuated when accounting for variant type, TP mutational status, and brain metastasis in our multivariate analysis, suggesting that PD-L1 status is not an independent determinant of ALK TKI outcomes.
A limitation of our observational clinical cohort is the heterogeneity of tissue NGS platforms included, resulting in variable coverage of the somatic alterations of interest.Particularly, reporting of CNAs is not standardized across NGS platforms; as a result, for example, loss of CDKNA/B may be underestimated in both the clinical and liquid biopsy cohorts.In addition, the decision to perform NGS testing at disease progression was made at treating physician's discretion, resulting in only a subset of patients undergoing such testing in our clinical cohort.This may bias the acquired resistance alterations observed in our clinical cohort, potentially with overrepresentation of patients with aggres-sive or higher burden of disease more readily detected by the liquid biopsy testing most commonly used upon progression (as opposed to tissue NGS).We also acknowledge the limitation in standardized assessment of endpoints such as progression from retrospective data, highlighting the need for prospective studies assessing the contribution of variant status, concurrent somatic mutation status, and PD-L1 expression in outcomes on ALK TKI.

Conclusions
In this large clinical cohort of patients with ALK-positive NSCLC primarily treated with alectinib, we found that EML-ALK v3 status and somatic co-mutations in TP and CDKNA/B were associated with inferior clinical outcomes and v3 status is associated with specific, TKI-relevant ALK resistance patterns.As ALK-positive NSCLC is often diagnosed solely by FISH or IHC, our study highlights the importance of comprehensive genomic tumor profiling at the time of diagnosis to identify patients at most risk for developing early disease progression and provide an opportunity to optimize first-line and potential consolidative therapy in ALK-positive NSCLC.
Authors' Disclosures institutional or commercial CLIA-certified panel NGS platforms.Most commonly utilized tissue NGS platforms included the 341-gene Memorial Sloan Kettering-Integrated Mutation Profiling of Actionable Cancer Targets (MSK-IMPACT; ref. 15), 1,000-gene Johns Hopkins Solid Tumor NGS Panel (16), 50-gene Johns Hopkins Limited Solid Tumor NGS Panel (17), 324-gene Foun-dationOne CDx (F1CDx; ref. 18) and 592-gene Caris Life Sciences Molecular Intelligence Tumor Seek (Caris Life Sciences); all of which use hybridization capture-based tissue NGS, except for the targeted PCR-based Johns Hopkins University NGS Limited platform.

FIGURE 1 A
FIGURE 1 A, EML4-ALK variant type observed in clinical and liquid biopsy cohorts.B, Clinically relevant somatic co-mutations observed in clinical and liquid biopsy (LB) cohorts.Copy-number deletion/loss not reported in the liquid biopsy cohort; loss-of-function mutations in CDKN2A/B, PTEN, and CTNNB1 are included.

FIGURE 2
FIGURE 2 ALK resistance mutations observed in clinical (n = 26; A) and liquid biopsy (n = 202; B) cohorts, by patient (columns) and grouped by EML4-ALK variant type.Filled regions of the oncoprint indicate presence of the ALK resistance mutation designated by the row name for a given patient.Resistance mutation percentages indicate share of each among total patients, by cohort.Both G1202R and I1171N are associated with EML4-ALK v3 and L1196M is associated with EML4-ALK non-v3.

FIGURE 3
FIGURE 3 Kaplan-Meier curve for OS and PFS on first-line TKI by EML4-ALK v3 status (A).B, OS and PFS subgroup analysis by Cox regression.HR, unadjusted hazard ratio; CI, confidence interval.resistance mutation structural clonality and variant type (χ 2 = 0.061; P = 0.805), indicating that the emergence of multiresistance clones is likely not related to the EML-ALK isoform differences.

FIGURE 4
FIGURE 4 Kaplan-Meier curve for OS (A) and PFS (B) on first-line TKI by TP53 mutation and CDKN2A/B loss.

TABLE 1
Baseline characteristics of patients with ALK-positive NSCLC in clinical cohort by EML4-ALK v1 and v3 subgroups Harada reports personal fees from Roche, Takeda, and Pfizer during the conduct of the study; personal fees from BMS, Lilly, AstraZeneca, J&J, Merck, Sanofi, and MSD outside the submitted work.A. Bubie reports other from Guardant Health, Inc during the conduct of the study; other from Guardant Health, Inc outside the submitted work.L.A. Kiedrowski reports personal fees from Guardant Health during the conduct of the study; personal fees from Guardant Health outside the submitted work.J.C.Murray reports personal fees and non-financial support from Regeneron and personal fees from Johnson & Johnson outside the submitted work.K.A. Marrone reports personal fees from Daiichi-Sankyo, Regeneron, AstraZeneca, Janssen, Amgen; grants and personal fees from Mirati and Bristol Meyers Squibb outside the submitted work.S.C.Scott reports grants from Janssen and Mirati; personal fees from AstraZeneca, Regeneron, Tempus, Foundation Medicine, and Genentech outside the submitted work.S. Houseknecht reports personal fees from As-traZeneca and Aptar Pharma, and other from Pfizer outside the submitted work.J. Feliciano reports grants from AstraZeneca, Pfizer, Bristol Myers Squibb;personal fees and other from Regeneron; personal fees from AstraZeneca, Coherus, Eli Lilly, Genentech, Takeda, Jansen, and Daiichi outside the submitted work.C.L. Hann reports grants and personal fees from AstraZeneca, Daiichi, BMS; grants from Amgen and AbbVie; personal fees from Puma BioTechnology and Janssen during the conduct of the study.V. Anagnostou reports grants from AstraZeneca, Bristol Myers Squibb, Delfi Diagnostics, Personal Genome Diagnostics; personal fees from Neogenomics, AstraZeneca, and Foundation Medicine outside the submitted work; in addition, V. Anagnostou has a patent to cancer genomic analyses, ctDNA therapeutic response monitoring and immunogenomic features of response to immunotherapy (63/276,525; 17/779,936; 16/312,152; 16/341,862; 17/047,006; 17/598,690) issued.P.M. Forde reports grants from AstraZeneca, BMS, Novartis, Regeneron, BioNTech, Nextpoint and other from Ascendis, AstraZeneca, BMS, Curevac, Novartis, Regeneron, G1, Genelux, Genentech, Gritstone, Merck, Janssen, F Star, Sanofi, Amgen, Fosun, Teva, Synthekine, Flame, Iteos, Xencor, Tavotek, Teva outside the submitted work.J.R. Brahmer reports personal fees from Roche during the conduct of the study; grants and personal fees from Bris- G. tol Myers Squibb, AstraZeneca, RAPT Therapeutics; grants from Merck; AACRJournals.orgCancer Res Commun; 4(3) March 2024 793