LRP1b loss predicts sensitivity to immunotherapy in patients with NSCLC: an analysis of the phase 3 Checkmate 026 randomized trial

Abstract

Purpose:

Low-density lipoprotein receptor-related protein 1b (LRP1b) is a cell surface receptor, commonly altered in many cancers and associated with improved responses, progression free survival (PFS) and overall survival (OS) with immune checkpoint inhibition (ICI).

Experimental Design:

LRP1b alterations were determined by whole exome sequencing (WES) and associated with PFS and objective response rates (ORR) in patients with non-small cell lung cancer (NSCLC) in post-hoc analysis of the randomized controlled phase 3 Checkmate-026 trial (CM026, NCT02041533) examining chemotherapy vs nivolumab, adjusting for tumor mutational burden (TMB) and clinical features. We separately evaluated a de-identified nationwide (US-based) NSCLC clinico-genomic database (CGDB) for associations of LRP1b alterations and progression-free survival with anti-PD-1 immunotherapy.

Results:

In the CGDB cohort of patients with NSCLC (N=18,369), LRP1b mutations were positively associated with TP53/KRAS alterations and inversely with EGFR/RET/MET/ERBB2 alterations and significantly improved PFS with PD-1 inhibition (n=1569, adjusted HR 0.86 p=0.014). In CM026, patients with LRP1b alterations had a statistically significant improvement in ORR to nivolumab vs. LRP1b wild-type (wt) patients (odds ratio 3.5; 95% CI 1.71–7.13; p=0.0006) but not with chemotherapy (odds ratio 0.63; 95% CI 0.32–1.26; p=0.19), adjusting for TMB, age, gender, histology, smoking and performance status. LRP1b mutations were associated with improved PFS with nivolumab (HR 0.66, p=0.04) but not chemotherapy (HR 1.26, p=0.25), also maintained in multivariate and TMB adjusted analysis.

Conclusions:

LRP1b mutations are a candidate predictive biomarker for ICI vs. chemotherapy in NSCLC. Further mechanistic characterization of LRP1b and prospective validation are warranted, and might enable future clinical use.

Translational relevance:

Identification of novel predictive biomarkers for response to immune-checkpoint inhibitors (ICI) in solid tumors is necessary for selection of patients who may benefit from these therapies and to inform development of new immunotherapeutic strategies. LRP1b is a gene commonly deleted or mutated in multiple tumor types. In this study we show that LRP1b alterations are enriched in NSCLC tumors with high TMB, TP53 alterations and less commonly associated with non-smoking associated genomic alterations such as EGFR, ERBB2, RET, and MET. Importantly, in both a national database analysis and a prospective, randomized phase 3 trial of patients with NSCLC treated with first-line PD-1 immunotherapy vs platinum-based chemotherapy, LRP1b alterations were significantly associated with improved radiographic responses and longer progression-free survival with nivolumab but not chemotherapy, even after adjusting for clinical variables and TMB status. These data suggest LRP1b mutations are a candidate predictive biomarker for ICI vs. chemotherapy in NSCLC.

Introduction

Immune checkpoint inhibitors (ICI) have dramatically improved the outcomes of patients with solid tumors in recent years. Notably, remarkable improvement in overall survival (OS) and landmark survival beyond 5 years in patients with advanced-stage non-small cell lung cancer (NSCLC) has been observed after the introduction of anti-PD-1 agents as front-line therapy either in combination with chemotherapy, combination immune therapy, or alone (1–6). ICI are now included in standard of care front-line treatment of advanced stage NSCLC without actionable molecular alterations (7). Many individuals with NSCLC derive benefit from ICI, but there remains a significant population with primary resistance to ICI as well as patients who have progressive disease after a short period of response (8).

Biomarkers to accurately identify patients who will respond or not respond to standard of care immune therapy options are needed. Program death receptor ligand 1 (PD-L1) proportion score on tumor cells is predictive of benefit from ICI for NSCLC. However, PD-L1 is not a binary biomarker and some tumors with < 1% PD-L1 respond to therapy, and some tumors with high PD-L1 do not respond to ICI. Tumor mutational burden (TMB), the number of mutations per million base pairs (Mb) of analyzed DNA, is an additional biomarker for response to ICI in NSCLC as well as other solid tumors (9–11). While widely used, PD-L1 and TMB are imperfect predictors of benefit from ICI. These tests do not fully represent the complexity of the tumor immune microenvironment (TIME) and recruitment of immune cells (12). TMB is particularly problematic in the lack of consistency in methodology of calculation and significance of a particular result across the commercially available testing platforms (13).

It has become clear that the adaptive anti-tumor immune response is extraordinarily complex, prompting the need to develop additional biomarkers to improve selection of appropriate treatment and provide insight into novel immunotherapeutic approaches(9, 12). Several mechanisms for primary resistance to ICI have been described including insufficient neoantigens, impaired processing of presentation of tumor antigens, impaired intra-tumoral immune infiltration, anti-inflammatory metabolic mediators, negative impact of immune regulatory cells, use of alternate immune checkpoints and severe T-cell exhaustion (9, 12). Existent biomarkers do not adequately examine all these processes, leading to patients receiving either inadequate treatment with ICI monotherapy or unnecessary toxicity with combination therapy.

Low-density lipoprotein receptor-related protein 1b (LRP1b) is a commonly mutated gene in multiple tumor types [14]. In multiple retrospective studies, LRP1b alterations correlated with response to single agent anti-PD-1 therapy in patients with a variety of solid malignancies, including NSCLC (14–17). For example, in one multicenter analysis, we found that LRP1b pathogenic alterations were identified in 36% of NSCLC patients, and across a range of solid tumors, particularly in NSCLC, LRP1b alterations were independently associated with improved ORR (OR 7.5, p=0.0009), PFS (HR 0.42 p=0.0003), and OS (HR 0.62 p=0.053 overall and HR 0.41 95% CI 0.19–0.91 in NSCLC) in patients treated with PD-1 based immunotherapy (16). These findings remained significant even after adjusting for TMB, given that LRP1b is a large gene and alterations have also been found to be associated with increased TMB across multiple tumor types (15, 18). LRP1b is a member of the low-density lipoprotein receptor (LDLR) family of cell-surface receptors, noted to bind a wide variety of ligands with function in hemostasis, endocytosis, cell proliferation and both the innate and adaptive immune system (19, 20). Preclinical work by Beer et al. demonstrated transfection of a plasmid containing murine LRP1b into A549 cells, a NSCLC cell line known to have altered LRP1b expression, will suppress cellular proliferation compared to cells transfected with empty plasmid. Furthermore, they showed siRNA directed at LRP1b in a NSCLC cell line containing wt LRP1b significantly increased cellular proliferation (21). These studies highlight the role of LRP1b signaling as an independent factor in control of cellular proliferation, but further study of this protein and its role in regulating the tumor immune microenvironment has been limited by its large gene size (16kbp) and lack of available mouse models (19).

As its current role in clinical practice is unclear, LRP1b is not included in many major commercial or academic next generation sequencing (NGS) platforms. Data from a large, prospective, controlled trial is needed to establish LRP1b as a predictive biomarker and to spur further development of mechanistic studies and preclinical models.

Our study retrospectively evaluated the frequency of LRP1b alterations in a large real-world NSCLC population using available genomic sequencing data from Foundation Medicine. We then evaluated the association of LRP1b alterations on immunotherapy outcomes in standard practice settings making use of a well-validated PFS measure from Flatiron Health. In order to determine predictive utility, we then conducted a post-hoc analysis of a large phase 3 randomized controlled trial using whole exome sequencing (WES) data from patients enrolled on Checkmate 026 (CM026) (3). This randomized, phase III clinical trial compared single-agent nivolumab to investigator choice chemotherapy in patients with previously untreated, recurrent, or metastatic NSCLC with PD-L1 expression of at least 1%.

Methods

Checkmate 026 analysis

Checkmate 026 was a phase 3 1:1 randomized controlled trial of patients with untreated stage IV or recurrent NSCLC and a PD-L1 tumor-expression level of 1% or more of either nivolumab administered intravenously at a dose of 3 mg per kilogram of body weight once every 2 weeks or platinum-based chemotherapy administered once every 3 weeks for up to six cycles. Patients who received chemotherapy could cross over to receive nivolumab at the time of disease progression. The primary end point of this trial was progression-free survival, as assessed by means of blinded independent central review, among patients with a PD-L1 expression level of 5% or more. A subgroup of patients enrolled on CM026 were included in the exploratory biomarker cohort for tumor whole exome sequencing (WES) on their initial research specimen for inclusion in this analysis. The inclusion criteria for this study have been previously published (5).

All studies involving human subjects were conducted in accordance with the US Common Rule and the Declaration of Helsinki. All subjects provided informed written consent for participation and genetic testing. All human investigations were conducted after IRB approval/exemption for this retrospective analysis.

Whole exome capture and sequencing: Checkmate 026

Genomic DNA (150 ng) was used for library preparation using the Agilent SureSelectXT reagent kit (Agilent Technologies, Santa Clara, USA) with the on-bead modifications of Fisher et al, 2011. A total of 500 ng of enriched library was used in the hybridization and captured with the SureSelect All Exon v5 (Agilent Technologies, Santa Clara, USA) bait. Following hybridization, the captured libraries were purified according to the manufacturer’s recommendations and amplified by polymerase chain reaction (11 cycles). Normalized libraries were pooled and DNA was sequenced on the Illumina HiSeq 2500 using 2 × 100-bp paired-end reads.

Mutation calling and tumor mutation burden calculation: Checkmate 026

Whole exome sequencing data were used to generate tumor mutation burden (total number of missense mutations). Missense mutations were identified from paired tumor/germline whole exome sequencing data using two mutation callers (Strelka and Sentieon). The union of calls was used to calculate the tumor mutation burden. Deleterious LRP1b alterations were identified as mutations resulting in frameshifts, stop gained, splicing site, and/or missense mutations with High or Moderate predicted impact to protein structure.

Statistical analyses: Checkmate 026

The primary endpoint of this analysis was to compare the ORR with nivolumab as compared to chemotherapy in NSCLC patients stratified by LRP1b alteration status, under the hypothesis that LRP1b alterations enrich for a greater objective radiographic response rate with nivolumab but not chemotherapy, and may thus suggest a predictive role for LRP1b in PD-1 responsiveness in lung cancer. Secondarily, we considered radiographic progression free survival as the time from enrollment to progression or death according to treatment and LRP1b alteration status. Hazard ratios (HRs) and odds ratios (ORs) with 95% confidence intervals (CIs) were estimated using a stratified Cox proportional hazards model and generalized linear model (Logit link function), respectively. Survival curves and rates were estimated using the Kaplan-Meier method stratified by LRP1b mutation status and treatment arms. Fishers exact testing (two-sided) was performed for differences in ORR proportions between LRP1b biomarker defined subgroups with a pre-specified alpha error of 0.05. Survival analysis was performed using an unstratified Cox proportional hazards model. Multivariate models were adjusted for age, sex, smoking status, histology, ECOG performance status and log10 transformed TMB. In both univariate and multivariate analyses that evaluated the association between LRPb1 status and outcome, the interaction term of treatment arm was included in the models. Due to significant crossover to nivolumab in patients receiving chemotherapy on original study (60%), overall survival was not considered as a statistical endpoint in this analysis. All statistical analyses were conducted using R software.

Real-world population next generation sequencing analysis

Study Design and Patient Selection: CGDB

The cohort consisted of patients with confirmed diagnosis of non-small cell lung cancer included in the U.S.-wide Flatiron Health (FH)-Foundation Medicine, Inc. (FMI) de-identified clinico-genomic database (CGDB) between January 2011- June 2024. All patients underwent genomic testing using FMI comprehensive genomic profiling (CGP) assays (described below) with molecular data linked directly.

De-identified clinical data originated from approximately 280 US cancer clinics (~800 sites of care). Retrospective longitudinal clinical data were derived from electronic health records (EHR), comprising patient-level structured and unstructured data, curated via technology-enabled abstraction of clinical notes and radiology/pathology reports, which were linked to genomic data derived from FMI testing by de-identified, deterministic matching(22). Clinical data included demographics, clinical and laboratory features, timing of treatment exposure, treatment progression, and survival. Lines of therapy in the database were oncology clinician-defined and rule-based. Patients with a Birth Year of 1938 or earlier may have an adjusted Birth Year in Flatiron datasets due to patient de-identification requirements.

Any patient with record of treatment with a single-agent anti-PD1 axis ICI were considered. Any line of therapy completely in immortal time was excluded, meaning we did not consider any patient’s progression-free survival records if they stopped ICI treatment before they received NGS testing, in order to avoid survivorship bias. Patients must have tissue-assessed TMB and LRP1b data available.

Comprehensive Genomic Profiling: CGDB

Hybrid capture-based next-generation sequencing (NGS) was performed on DNA extracted from patient tumor tissue specimens in a Clinical Laboratory Improvement Amendments (CLIA)–certified, College of American Pathologists (CAP)-accredited laboratory (Foundation Medicine, Cambridge, MA). Samples were evaluated for alterations as previously described (23). Tumor mutational burden (TMB) was determined on up to 1.1 Mb of sequenced DNA (24) with analytical validations reported in the Supplement of KEYNOTE-158 (11). Predominant genetic ancestry was determined by training a random forest classifier to distinguish the 5 ancestral superpopulations of the 1000 Genomes Project (25), as previously described (26).

Biomarker Definitions: CGDB

Liquid biopsy was not used in these assessments, to ensure confidence in a negative biomarker result. Foundation Medicine’s FoundationOne®CDx has an FDA-approved CDx indication to identify patients with TMB ≥ 10 mutations per megabase in solid tumors who may benefit from treatment with pembrolizumab. The same TMB algorithm was also used in the laboratory developed test (LDT) predecessor, FoundationOne®, which was used to test all patients in this cohort. LRP1bLRP1BLRP1BWhile TMB is a continuous factor, it has highly skewed distributions that are challenging to normalize, and we therefore made use of categorical ranges of TMB for Cox models. The cutoff of 10 was utilized for dichotomous analyses, and the TMB ranges described in a pan-tumor validation of the Foundation Medicine TMB measure (27) were used in alternate models. The FoundationOne LDT panel includes LRP1b, which was either characterized as “known / likely deleterious”, otherwise alterations were considered variants of unknown significance (VUS), and if the specimen did not have any LRP1b alterations, it was considered wild-type.

Time to Event Outcomes: CGDB

Progression-Free Survival (PFS) was calculated from index date until progression (due to any cause) or death. Patients not yet reaching next treatment line or death were censored at date of last clinical visit. Flatiron Health database mortality information is a composite derived from 3 sources: documents within the EHR, Social Security Death Index, and a commercial death dataset mining data from obituaries and funeral homes, with validations reported in comparison to the National Death Index (28). Flatiron Health PFS measure has been benchmarked against the mortality measure, as well as assessed for reliability and repeatability (29). The Flatiron Health and Foundation Medicine CGDB has replicated associations with biomarker subgroup analyses of randomized controlled trials(30–32).

Statistical Analysis: CGDB

Differences in time-to-event outcomes were assessed with the log-rank test and Cox proportional hazard (PH) models. Chi-square tests and Wilcoxon rank sum tests were used to assess differences between groups of categorical and continuous variables, respectively. Multiple comparison adjustments were not performed; p-values are reported to quantify the strength of association for biomarker and each outcome, not for null hypothesis significance testing, and interpretations adopted broadly considering consistency of multiple cohorts with no analysis standing on its own.

Missing values were handled by simple imputation with expected values determined using random forests with the R package ‘missForest.’ In subsequent analyses, imputed values were treated identically to measured values. R software was used for all statistical analyses.

Data Sharing Statement

The data that support the findings of this study were originated by Flatiron Health, Inc. and Foundation Medicine, Inc as well as Bristol Myers Squibb. Requests for data sharing by license or by permission for the specific purpose of replicating results in this manuscript can be submitted to PublicationsDataaccess@flatiron.com and cgdb-fmi@flatiron.com for the CGDB and to BMS through Virginia.IP@bms.com.

Results

Clinico-genomic database study

A total of 18,369 patients with NSCLC were included in the Foundation Medicine CGDB retrospective analysis to determine the associations of predicted pathogenic LRP1b alterations with tumor mutation burden and other NSCLC genotypes. Of these, 1570 patients with NSCLC in the advanced/metastatic setting were treated with single agent anti-PD-1 ICI therapy and had next-generation sequencing testing evaluable for LRP1b alterations (CONSORT diagram in Figure 1). The patient characteristics can be found in Table 1. LRP1b genomic alterations were detectable in 646/1570 patients with NSCLC (41.1%). Of these, 9.3% of patients had known or likely pathogenic protein deleterious or functionally disrupting alterations, while 31.8% harbored alterations of unknown function. Comparing patients with any LRP1b alteration vs. wild-type, strong differences were not observed in patient ECOG performance status, sex, age, socioeconomic status, histology, stage at diagnosis, or genetic ancestry. Consistent with prior reports, there was a strong association of LRP1b status with TMB level (p<0.001). Strong differences in the prevalence of established predictive biomarkers for systemic NSCLC therapies commonly observed in non-smokers were seen and enriched in LRP1b wild type patients, including EGFR alterations (2.7% vs 0.5%, p<0.001), ERBB2 mutations (2.8% vs. 0.4%, p = 0.004), RET rearrangements (1.6 vs 0.3%, p=0.025), and MET exon 14 skipping (2.4 vs 0.8%, p=0.026). However, ALK rearrangements did not significantly differ (1.1 vs 0.5%, p=0.18) by LRP1b status. TP53 mutations were also more common in the LRP1b altered group (77.2% vs. 61.6%, p < 0.001), and KRAS G12C mutations were more commonly observed in LRP1b altered patients (13.2 vs 9.4% p=0.064). LRP1b status was not associated with PD-L1 protein expression (p=0.119). These data suggests that LRP1b alterations are enriched in smoking related NSCLC tumors and tumors with high tumor mutation burden.

Figure 1.

ICI monotherapy cohort overview. Cohort selection diagram (A) and treatment diagram with index date (B).

Table 1.

ICI monotherapy cohort summary and patient clinical and genetic characteristics in the national clinical-genetic cohort study.

Variable	wild-type (N=924)	altered (N=646)	Total (N=1570)	p value
ECOG				0.643
0	193 (20.9%)	119 (18.4%)	312 (19.9%)
1	439 (47.5%)	323 (50.0%)	762 (48.5%)
2	158 (17.1%)	109 (16.9%)	267 (17.0%)
3+	134 (14.5%)	95 (14.7%)	229 (14.6%)
Sex				0.518
Female	463 (50.1%)	313 (48.5%)	776 (49.4%)
Male	461 (49.9%)	333 (51.5%)	794 (50.6%)
Age (per decade increase)				0.015
Median (Q1, Q3)	6.9 (6.1, 7.5)	6.8 (6.0, 7.4)	6.9 (6.1, 7.5)
Line of Therapy				0.08
1st	262 (28.4%)	216 (33.4%)	478 (30.4%)
2nd	448 (48.5%)	299 (46.3%)	747 (47.6%)
3rd+	214 (23.2%)	131 (20.3%)	345 (22.0%)
Socioeconomic Status				0.645
1 - Lowest SES	129 (15.3%)	102 (16.8%)	231 (15.9%)
2	174 (20.6%)	115 (19.0%)	289 (19.9%)
3	175 (20.8%)	141 (23.3%)	316 (21.8%)
4	193 (22.9%)	132 (21.8%)	325 (22.4%)
5 - Highest SES	172 (20.4%)	116 (19.1%)	288 (19.9%)
N-Miss	81	40	121
TMB range				< 0.001
TMB < 5	389 (42.1%)	38 (5.9%)	427 (27.2%)
TMB 5–10	308 (33.3%)	165 (25.5%)	473 (30.1%)
TMB 10–20	171 (18.5%)	249 (38.5%)	420 (26.8%)
TMB 20+	56 (6.1%)	194 (30.0%)	250 (15.9%)
LRP1B level				< 0.001
wild-type	924 (100.0%)	0 (0.0%)	924 (58.9%)
known/likely pathogenic	0 (0.0%)	146 (22.6%)	146 (9.3%)
other alteration	0 (0.0%)	500 (77.4%)	500 (31.8%)
Drug				0.709
Atezolizumab	56 (6.1%)	38 (5.9%)	94 (6.0%)
Avelumab	1 (0.1%)	0 (0.0%)	1 (0.1%)
Durvalumab	7 (0.8%)	3 (0.5%)	10 (0.6%)
Nivolumab	661 (71.5%)	452 (70.0%)	1113 (70.9%)
Pembrolizumab	199 (21.5%)	153 (23.7%)	352 (22.4%)
Histology				0.505
Non-squamous cell carcinoma	669 (72.4%)	455 (70.4%)	1124 (71.6%)
NSCLC histology NOS	33 (3.6%)	20 (3.1%)	53 (3.4%)
Squamous cell carcinoma	222 (24.0%)	171 (26.5%)	393 (25.0%)
Stage at Diagnosis				0.01
Stage I	61 (6.6%)	67 (10.4%)	128 (8.2%)
Stage II	63 (6.8%)	39 (6.0%)	102 (6.5%)
Stage III	174 (18.8%)	149 (23.1%)	323 (20.6%)
Stage IV	603 (65.3%)	377 (58.4%)	980 (62.4%)
Unknown/not documented	23 (2.5%)	14 (2.2%)	37 (2.4%)
Genetic Ancestry				0.529
AFR	93 (10.1%)	67 (10.4%)	160 (10.2%)
AMR	13 (1.4%)	8 (1.2%)	21 (1.3%)
EAS	26 (2.8%)	10 (1.5%)	36 (2.3%)
EUR	768 (83.1%)	547 (84.7%)	1315 (83.8%)
Other	24 (2.6%)	14 (2.2%)	38 (2.4%)
PD-L1				0.119
Median (Q1, Q3)	14.5 (0.0, 74.5)	29.5 (0.5, 74.5)	24.5 (0.5, 74.5)
N-Miss	733	492	1225
EGFR alteration				< 0.001
Exon 19 deletion	24 (2.7%)	3 (0.5%)	27 (1.8%)
Exon 20 insertion A763_Y764insFQEA	1 (0.1%)	0 (0.0%)	1 (0.1%)
Exon 20 insertion other	22 (2.5%)	2 (0.3%)	24 (1.6%)
G719X	3 (0.3%)	1 (0.2%)	4 (0.3%)
L858R	19 (2.1%)	2 (0.3%)	21 (1.4%)
L861Q	1 (0.1%)	1 (0.2%)	2 (0.1%)
Multiple	27 (3.0%)	6 (0.9%)	33 (2.2%)
Negative	796 (88.7%)	619 (97.5%)	1415 (92.4%)
Other/NOS	4 (0.4%)	1 (0.2%)	5 (0.3%)
N-Miss	27	11	38
ALK rearrangement				0.181
Negative	884 (98.9%)	625 (99.5%)	1509 (99.1%)
Positive	10 (1.1%)	3 (0.5%)	13 (0.9%)
N-Miss	30	18	48
ROS1 rearrangement				0.008
Negative	871 (100.0%)	611 (99.2%)	1482 (99.7%)
Positive	0 (0.0%)	5 (0.8%)	5 (0.3%)
N-Miss	53	30	83
BRAF mutation				0.12
Negative	834 (97.3%)	599 (98.5%)	1433 (97.8%)
V600E	23 (2.7%)	9 (1.5%)	32 (2.2%)
N-Miss	67	38	105
KRAS mutation				0.064
G12C	87 (9.4%)	85 (13.2%)	172 (11.0%)
Multiple	4 (0.4%)	4 (0.6%)	8 (0.5%)
Negative	620 (67.1%)	398 (61.6%)	1018 (64.8%)
Other	213 (23.1%)	159 (24.6%)	372 (23.7%)
MET exon 14 skipping				0.026
Negative	810 (97.6%)	591 (99.2%)	1401 (98.2%)
Positive	20 (2.4%)	5 (0.8%)	25 (1.8%)
N-Miss	94	50	144
MET amplification				0.357
Negative	809 (97.5%)	576 (96.6%)	1385 (97.1%)
Positive	21 (2.5%)	20 (3.4%)	41 (2.9%)
N-Miss	94	50	144
RET rearrangement				0.025
Negative	817 (98.4%)	594 (99.7%)	1411 (98.9%)
Positive	13 (1.6%)	2 (0.3%)	15 (1.1%)
N-Miss	94	50	144
NTRK rearrangement				0.396
Negative	827 (99.9%)	595 (100.0%)	1422 (99.9%)
Positive	1 (0.1%)	0 (0.0%)	1 (0.1%)
N-Miss	96	51	147
ERBB2 mutation				0.004
Negative	807 (97.2%)	592 (99.3%)	1399 (98.1%)
Positive	23 (2.8%)	4 (0.7%)	27 (1.9%)
N-Miss	94	50	144
TP53 mutation				< 0.001
Negative	355 (38.4%)	147 (22.8%)	502 (32.0%)
Positive	569 (61.6%)	499 (77.2%)	1068 (68.0%)
STK11 mutation				0.208
Negative	787 (85.2%)	535 (82.8%)	1322 (84.2%)
Positive	137 (14.8%)	111 (17.2%)	248 (15.8%)

LRP1b and TMB and Outcomes on ICI

Univariable associations of LRP1b subclass and TMB with PFS following anti-PD-1 inhibitor therapy are shown in Figure 2. In univariate analysis, LRP1b status, as either known/likely pathogenic or VUS/unknown function LRP1b alterations, associates with outcome on single-agent ICI (HR 0.68 to 0.73, p<0.0001). We found that TMB ≥10 has a stronger univariate association with ICI progression-free survival as compared to any LRP1b alteration (HR: 0.65, 95% CI: 0.59 – 0.72 vs. HR: 0.73 95%CI: 0.68 – 0.82). In the multivariable setting, adjusting for potential imbalances, TMB ≥10 (HR: 0.69, 95%CI: 0.61 – 0.78, p < 0.001) and LRP1b status (HR: 0.86, 95%CI: 0.76 – 0.97, p = 0.014) were both significant contributors to the model (Figure 3A). Evaluating both TMB with expanded range as well as LRP1b with either known/likely deleterious or VUS, TMB demonstrated decreasing hazards of progression along a biologically-expected gradient, and both subclasses of LRP1b point estimates suggest consistent decreased hazards of progression (Figure 3B). Progression-free survival was longer in tumors with TMB ≥ 10 and any alteration in LRP1b compared to LRP1b wild type (HR: 0.79, 95%CI: 0.67–0.93, p < 0.005, while there was no difference in PFS for pathogenic/likely pathogenic LRP1b altered NSCLC when TMB < 10 (HR 1.06, 95%CI: 0.75–1.5 p=0.73) (Supplementary Figure 1). This lack of association also did not change after adjusting for corticosteroid use nor EGFR/ROS/ALK mutation status (data not shown). LRP1b alterations were significantly associated with PFS only among high TMB patients, irrespective of likely pathogenic or VUS alterations, some of which may be pathogenic.

Figure 2.

Univariable associations of LRP1b and TMB in patients with NSCLC treated with single-agent anti-PD1 axis therapies. Outcome associations are visualized with Kaplan-Meier plots stratified by (A) any LRP1b alteration, (B) known or likely functional LRP1b status, (C) variant of unknown significance LRP1b status, and (D) by TMB10.

Figure 3.

Multivariable associations of LRP1b and TMB in patients with NSCLC treated with single-agent anti-PD1 axis therapies. Relative hazards of progression are estimated with Cox PH models with variables shown. TMB and LRP1b are included as dichotomous variables in (A) and with additional stratifications in (B).

Prospective validation: LRP1b in Checkmate-026 trial

We next sought to validate the predictive utility of LRP1b genetic alterations in a randomized controlled trial of ICI therapy vs chemotherapy in patients with NSCLC. To accomplish this, we evaluated Checkmate-026 (CM026) which included a total study size of 541 patients, of whom 312 patients (58%) were evaluable (CONSORT diagram in Figure 4). A total of 137 individuals (44%) with pathogenic LRP1b alterations were identified with 64 in the nivolumab arm and 73 in the chemotherapy arm. This is similar to the 41.1% prevalence of LRP1b alterations in the nationwide FH-FMI CGDB, inclusive of known/likely and VUS alterations, suggesting that WES and NGS testing arrive at similar estimates. In CM026, missense mutations were the most common (n=86, 62.8%), followed by copy number loss (n=19, 13.9%), stop-gain mutation (n=15, 10.9%), frame shift mutation (n=10, 7.35%) and splice site mutation (n=7, 5.1%). Consistent with the observation made based on Foundation Medicine research cohort (Supplementary Table 1), Supplementary Figure 1 demonstrates the clear relationship between TMB and LRP1b mutational status in the CM026 cohort, which showed a statistically significant (p<0.001) increase in TMB in individuals with LRP1b alterations.

Figure 4.

Checkmate 026 Clinical Trial CONSORT diagram.

^aSamples were not available for various reasons, including but not limited to lack of patient pharmacogenetic consent, samples exhausted for PD-L1 testing, or poor tissue sampling

^bSamples dropout due to internal quality control failure such as too few sequence reads, and low or uneven target region coverage

For the primary endpoint of the present biomarker analysis of CM026, we examined the associations of LRP1b with objective radiographic response rate by RECIST 1.1 criteria (ORR) in each treatment group from CM026. The ORR rate was significantly higher for patients with LRP1b alterations who received nivolumab treatment at 45.3% vs. 19.1% in patients with wild type LRP1b NSCLC (odds ratio 3.47; 95% CI 1.62–7.60l; p=0.0006). In the physician choice chemotherapy arm, however the ORR was lower for patients with LRP1b alterations at 25.0% as compared to 35.8% for LRP1b wt patients (odds ratio 0.64; 95% CI 0.30–1.36; p=0.23). In addition, univariate analysis showed an association between ORR and LRP1b alterations in nivolumab arm (odds ratio 3.5; 95% CI 1.72–7.13; p=0.0006) but not in the chemotherapy arm (odds ratio 0.63; 95% CI 0.32–1.26; p=0.19) (Figure 5a). Similar observations were made in multivariable analysis after adjustment for age, gender, smoking status, histology and ECOG performance status. The multivariable adjusted odds ratio for ORR with nivolumab for LRP1b mutated vs wild type NSCLC was 3.46 (95% CI 1.67–7.17, p=0.0008) while the adjusted ORR with chemotherapy for LRP1b mutated vs wild type NSCLC was not significant at 0.55 (95%CI 0.27–1.16, p=0.12) (Figure 5a). The association between LRP1b alterations and ORR remained significant after adjustment for TMB with an odds ratio for ORR of 3.25 (95%CI 1.53–6.90 p=0.002) for nivolumab for LRP1b altered vs wild type NSCLC, while no association was found for chemotherapy response based on LRP1b status adjusting for TMB (p=0.11) (Figure 5a). These prospective randomized data suggest that LRP1b alterations enrich and predict for anti-PD-1 immunotherapy response but not for platinum based chemotherapy in patients with NSCLC.

Figure 5.

Evaluation of predictive value of LRPb1 mutation in CM026. (A) Univariate and multivariate analyses of ORR in nivolumab vs chemotherapy arms. (B-C) Kaplan-Meier plots of nivolumab vs chemotherapy by LRP1b status for the outcome of PFS. (D) Univariate and multivariate analyses of PFS in nivolumab vs chemotherapy arms. Multivariate analysis adjusts for age, sex, smoking history, performance status and histology.

We next examined the impact of LRP1b status on the secondary endpoint of PFS by RECIST 1.1 by treatment group. Consistent with observations made in the clinico-genomic database study, PFS was improved with nivolumab in patients with LRP1b alterations compared to LRP1b wild type patients (median 6.4 vs 4.2 months, HR 0.69; 95% CI 0.47–1.02) but not in the chemotherapy arm (median 5.4 vs 7.1 months, HR 1.3 95% CI 0.88–1.94; Figures 5b-c). In univariate analysis, PFS was significantly improved with nivolumab in patients with LRP1b alterations compared to LRP1b wild type patients (HR 0.66, 95% CI 0.45–0.97 p=0.04), while PFS was numerically worse but not significantly different with chemotherapy in patients with vs those without LRP1b mutations (HR 1.26, 95% CI 0.85–1.87, p=0.25) (Figure 5d). These findings were similar in multivariable analysis adjusting for age, gender, smoking status, histology and ECOG performance status. The multivariable HR for PFS for nivolumab was 0.64 (95% CI 0.44–0.94; p=0.024) and was 0.65 (95%CI 0.43–0.98; p=0.04) after further adjusting for TMB, favoring patients with LRP1b altered NSCLC (Figure 5d). However, for chemotherapy, no differences were observed for PFS (multivariable HR 1.26 p=0.27; and HR 1.28 adjusting for TMB, p=0.27, Figure 5d). These data align with the ORR predictive data and suggest that LRP1b alterations predict for both radiographic responses and the durability of progression free survival with nivolumab treatment, but do not predict for response or PFS with chemotherapy.

We performed an exploratory analysis regarding LRP1b mutations and biomarkers associated with response and resistance to nivolumab therapy from the available Checkmate-026 data. An inclusion criteria for Checkmate-026 was tumor PDL-1 by Tumor Proportion Score (TPS) ≥ 1%. The incidence of LRP1b mutations was similar in tumors PDL-1 1–49% and ≥ 50% (Supplementary Figure 2). There was a similar incidence of tumors with the PD1 checkpoint antibody resistant KRAS mutations with STK11 and/or KEAP mutations in LRP1b wild type and LRP1b mutated tumors (Supplementary Figure 3). LRP1b mutations remained associated with improved response and PFS with nivolumab in tumors with KRAS mutations with STK11 and/or KEAP mutations, and in those same genotypes in multivariate analysis including TMB (Supplementary Figures 4-5).

Discussion

In the present study, we combined a real-world nationwide analysis of LRP1b genetic alterations in patients with NSCLC treated with immunotherapy, and separately analyzed LRP1b alterations in a prospective randomized controlled phase 3 clinical trial testing immunotherapy against chemotherapy. We found that LRP1b alterations were highly prevalent at 41–44% using a common commercial platform in a national clinico-genomic database or whole exome sequencing platform in CM026. LRP1b alterations, while associated with higher TMB, were clearly associated with TP53 and KRAS mutations and inversely associated with alterations in EGFR, RET, MET splice site, and HER2, which tend to occur more commonly in non-smokers. We show that LRP1b alterations are associated with significantly better progression-free survival times in a nation-wide clinico-genomic database cohort study of patients with NSCLC treated with anti-PD-1 immunotherapy. Importantly, we also found that LRP1b mutations, including those of both likely and unknown pathogenicity, were associated with significantly better objective radiographic responses and more durable progression free survival with nivolumab immunotherapy in patients with PD-L1+ NSCLC in the first line treatment setting in a randomized controlled trial, but were not predictive of platinum-based chemotherapy outcomes. These data remained consistent in multivariable analysis and after adjustment for tumor mutation burden, suggesting that LRP1b alterations may serve as an independent predictive biomarker for anti PD-1 responsiveness in NSCLC patients.

The frequency of LRP1b mutations in the clinico-genomic cohort was consistent with previously reported studies [22]. Mutations in LRP1b in this population appear to be enriched in individuals with smoking-related mutations such as KRAS, TP53 and STK11 [23, 24]. Tumors with oncogenic KRAS mutations combined with STK11 and/or KEAP mutation are genotypes highly predictive of resistance to PD-1 checkpoint therapy in NSCLC.[33, 34] In the Checkmate-026 data set, LRP1b mutations remained associated with improved response and PFS with nivolumab in tumors with KRAS mutations with STK11 and/or KEAP mutations despite those genotypes being highly resistant to immune therapy.(supplemental figure 2). The finding that LRP1b mutation remains predictive of response to nivolumab in KRAS driven tumors with STK11 and/or KEAP mutations suggests that LRP1b alterations may confer responsiveness to anti-PD-1 therapy rather than LRP1b mutations being more prevalent in tumors more likely to be responsive to immune therapy for other reasons. LRP1b mutations are not mutually exclusive of EGFR, HER2, c-MET, ROS1, RET, and ALK alterations in NSCLC, and further investigation should evaluate the predictive utility of LRP1b assessment in the setting of EGFR, ALK and other driver molecular alterations more commonly seen in never-smokers as these are associated with relative resistance to ICI monotherapy [25]. Within the clinico-genomic cohort, the frequency of mutations in LRP1b correlated with TMB, supporting prior studies demonstrating relationship between these genetic markers [17].

However, in CM026, while LRP1b was also associated with high TMB status, outcomes remained significant for LRP1b alterations and ORR and PFS despite adjustment for clinical variables and TMB status. In the clinico-genomic cohort, PFS was longer in tumors with TMB ≥ 10 with any alteration in LRP1b compared to LRP1b wild type supporting the conclusion that LRP1b alterations are associated with benefit from ICI independent of TMB (Supplementary Figure 1a, c, e). There was no difference in PFS for LRP1b altered NSCLC when TMB < 10 (Supplementary Figure 1b, d, and f). The reasons for the lack of predictive utility for LRP1b alterations in low TMB patients in the CG dataset are unclear, but may include differences in immune microenvironments and immune evasion mechanisms in these low TMB patients, other co-occurring genetic alterations, such EGFR/ROS/ALK/RET alterations commonly associated with non-smoking status and lack of LRP1b alterations, and the requirements for PD-L1 positivity in the CM026 study. These data suggest that knowledge of LRP1b status, which is not currently included in many commercial panels, may provide greater informed decision making around anti-PD-1 monotherapy treatment selection. The application of LRP1b as a predictive biomarker for the value of combined chemo-immunotherapy as is presently standard-of-care in NSCLC, is not known and worth investigating.

In this retrospective analysis, a unique population of patients with low to intermediate TMB and altered LRP1b can be observed in both real-world and investigational cohorts of patients with NSCLC. The frequency of LRP1b alterations in this cohort was 41–44% across both studies. These alterations were enriched in individuals with smoking-related mutations such as KRAS and TP53. However, there was a small population of patients with EGFR, ALK and other non-smoking related alterations as well as patients with low/intermediate TMB levels that additionally harbored a mutation in LRP1b. These are not mutually exclusive, and supports further investigation of LRP1b alterations to guide chemo-immunotherapy decisions in patients with NSCLC beyond TMB and other potential ICI biomarkers.

We additionally evaluated the relationship between LRP1b mutation and clinical outcome in patients enrolled in CM026 with evaluable exploratory biomarker data. In this population, LRP1b alterations were associated with increased TMB, which is consistent with prior studies in NSCLC and other tumor types [17, 18]. While TMB is associated with improved outcomes with anti-PD-1 therapies in NSCLC, this association remains imperfect, and additional candidate predictive biomarkers are needed to add precision to ICI benefit prediction, such as LRP1b.

The relationship between LRP1b mutational status and response to nivolumab in patients enrolled on CM026 suggests significantly improved ORR and PFS with nivolumab compared to LRP1b wild type patients, even in the TMB low setting, and an inverse relationship was seen with chemotherapy, suggesting that LRP1b may be predictive of immunotherapy but not chemotherapy benefit in NSCLC. We did not analyze overall survival due to the confounding impact of common cross-over therapy in the CM026 trial and this remains a limitation of the present analysis. Further validation in other contexts such as neoadjuvant or adjuvant NSCLC settings or in other chemo-ICI combination studies is warranted, and across other solid tumors where LRP1b alterations are common such as melanoma, esophageal and gastric cancer, head and neck cancer, and bladder and prostate cancer(16, 17). We have previously shown LRP1b to enrich for responsiveness to pembrolizumab in men with metastatic castration resistant prostate cancer (mCRPC) (4, 31) suggesting our data supports further study of LRP1b as a more general predictive biomarker of immune checkpoint blockade in solid tumors.

LRP1b mutant patients have higher TMB as expected based on previous studies. The predictive capacity of LRP1B in our analysis is maintained after adjusting for TMB. Therefore, LRP1b mutation is not surrogate simply for high TMB, but reflects its own independent predictive signal and potential biology. This difference in response between LRP1b mutated and wild type arms was not observed in the cohort of patients receiving investigator choice chemotherapy. These data suggest the clinical benefit of LRP1B altered NSCLC lies within response to ICI rather than overall more favorable disease biology. Tumors with low TMB would generally be postulated to have a lower number of neoantigens for T cells to recognize than high TMB tumors, and the biology of LRP1b in mediating immune evasion is as yet unknown.

These results are consistent with NSCLC subgroups analysis in prior study by Brown et al. (16). They evaluated a total of 101 solid tumor patients, including 41 with NSCLC, treated with single-agent anti PD-1 therapy found to have an alteration in LRP1b. Subjects were categorized as pathogenic/likely-pathogenic variants (P/LP) and variant of unknown significance (VUS). Their analysis showed the presence of a P/LP alteration in LRP1b was associated strongly with an increased overall response rate (ORR) and PFS with treatment with ICI across multiple tumor types, independent of microsatellite stability status and tumor mutational burden (TMB) (4). Further mechanistic and biologic studies of LRP1b loss of function in immune competent models and genetically engineered models of cancer are warranted to investigate how LRP1b loss contributes to immune evasion or exhaustion.

Together, these studies demonstrate there is a population of NSCLC patients with LRP1b alterations that derive meaningful clinical benefit from ICI monotherapy. Based on this analysis, LRP1b may have clinical utility as an additional biomarker for sensitivity for immunotherapy to better select individuals who can be spared the toxicity of chemotherapy without losing the clinical benefit. These data also imply that LRP1b has an independent role in modulating the immune response to cancer in the presence of ICI. Further understanding of its exact function could lead to better understanding of the anti-tumor immune response and development of novel immunotherapies.

In this new immunotherapeutic era, significant effort has been focused on understanding the way tumors can develop primary and acquired resistance to ICI beyond PD-L1 expression (33). T-cell related factors including the quantity and subclasses of TILs as well as upregulation of alternative immune checkpoints such as LAG3 and TIGIT are promising mechanisms of resistance that can be targeted with novel therapies currently approved or in development (34–36). The tumor immune microenvironment (TIME), a complex network of immune and stromal cells as well as their secreted cytokines and the extracellular matrix that supports them, plays a pivotal role in anti-tumor immunity and can be impacted by production of immunosuppressive cytokines by a tumor or infiltration with immunosuppressive tumor infiltrating lymphocytes (TILs) (34). A better understanding of complex anti-tumor immune response, including the processes of successful T-cell homing and trafficking to the tumor through modulation of the TIME, is likely to lead to further treatment options in the future (33). The exact underlying biological function of wt LRP1b as well as the impact of a likely pathogenic or a VUS mutation in this gene on oncogenesis and anti-tumor immunity is not completely understood currently. In our study, alterations in LRP1b that were both likely pathogenic and those with unknown pathogenicity appeared to be associated with PFS and response outcomes in patients with NSCLC receiving ICI but not chemotherapy. There is an increasing body of evidence that LRP1b may play a role in modulation of the tumor-immune microenvironment. A retrospective study performed by Xu et al evaluating a cohort of NSCLC patients in China identified mutations in LRP1b in 13.98% of 110 samples analyzed. While the small sample does not allow for meaningful comparison of clinical outcomes in this study, LRP1b alterations were enriched for multiple types of tumor-infiltrating lymphocytes (TILs), including CD8+ T cells, activated NK cells and M1 macrophages. LRP1b mutated tumors additionally had higher expression of PD-L1, PD-1 and LAG3 compared to wildtype. Finally, gene expression analysis in this study demonstrated upregulation of Notch, insulin, and mTOR signaling pathways as well as ubiquitin-mediated proteolysis in LRP1b mutated samples compared to wt (37).

Our results combined with the study above suggest that further investigation into the exact function of LRP1b is necessary to examine whether it could serve beyond just a biomarker in patients with mutations but as an additional target for drug development in patients with LRP1b wt cancers. Further understanding the function of LRP1b and its role in modulation of the TIME can lead to insight on additional strategies to manipulate the adaptive immune response to overcome primary and acquired ICI resistance.

While the retrospective nature of our study may be a limitation, the independent validation of LRP1b alterations in both a prospective randomized trial and in a real world standard of care national database of patients with NSCLC is a major strength. The most common first line standard of care in NSCLC typically involves a combination of chemotherapy with immunotherapy, and further studies are needed of LRP1b alterations to predict chemo-immunotherapy combination therapy benefits. In addition, in CM026, the exclusion of patients with 0% PD-L1 expression led to enrichment of a more inflammatory phenotype of NSCLC that was subsequently enriched for other markers of sensitivity to ICI. We do not have additional biologic correlative data on these samples including presence of TILs and their immunophenotype or gene expression.

Further prospective studies including LRP1b status as one component of a composite biomarker for PD-1 responsiveness are needed to better define its role in clinical practice (12), in the context of CD8 T cell infiltration, TMB, and other variables. Further studies should include evaluation of presence of tumor infiltrating lymphocytes as related to LRP1b mutation status in a prospective manner. Also, in an adequately powered trial, relationships between LRP1b and other genes as predictive markers for response to ICI could be elucidated not just in lung cancer but in many tumor types including completed trials that included tumor genetic data. Two studies identified a subset of patients with NSCLC bearing co-occurring mutations in FAT atypical cadherin (FAT) family genes and LRP1b associated with improved progression free survival on ICI (14, 38). A prospective trial to evaluate the relationship between LRP1b and common driver alterations such as EGFR and ALK in predicting response to ICI could be considered as well. The rapidly growing and increasingly complex role of multigene sequencing in development of predictive biomarkers in lung cancer also highlights the need for robust tumor banking programs and preservation of tumor genetic material to ensure adequate samples for translational investigation.

As we expanded our immunotherapy armamentarium, we will also need to expand our understanding of available predictive biomarkers. Further study of LRP1b as a possible predictive biomarker in patients receiving dual checkpoint inhibition with combination anti-PD-1 and CTLA-4 blockade as well as LAG-3 inhibition or TIGIT targeting agents currently under investigation would be meaningful.

In summary, our analysis of real-world NGS data from a national FH-FMI CGDB and WES data from the Checkmate 026 study suggests that mutations in LRP1b are common in NSCLC and associated with more favorable outcomes on ICI but not with chemotherapy. LRP1bLRP1bFurther mechanistic studies are needed to understand the exact role of LRP1b in modulation of the anti-tumor immune response and sensitization to ICI across malignancies, which could provide insight into new therapy development and its potential predictive utility. LRP1b