Association of LRP1B Mutation With Tumor Mutation Burden and Outcomes in Melanoma and Non-small Cell Lung Cancer Patients Treated With Immune Check-Point Blockades

Background: Tumor mutation burden (TMB) have been served as the most prevalent biomarkers to predict immunotherapy response. LRP1B (low-density lipoprotein receptor-related protein 1B) is frequently mutated in melanoma, non-small cell lung cancer (NSCLC) and other tumors; however, its association with TMB and survival in patients with immunotherapy remains unknown. Methods: We curated somatic mutation data and clinicopathologic information from 332 melanoma immunotherapy samples for discovery and 113 NSCLC samples for further corroboration. Bayesian variants non-negative matrix factorization was used to extract tumor mutational signatures. Multivariate Cox and logistic regression models were applied to adjust confounding factors. The CIBERSORT and GSEA algorithm were separately used to infer leukocyte relative abundance and significantly enriched pathways. Results: Patients with LRP1B mutation were identified to be associated with prolonged survival in both immunotherapy cohort. Higher tumor mutation burden was found in LRP1B mutated patients, and the association remained significant after controlling for age, gender, stage, mutations in TP53 and ATR, and mutational signatures. Immune response and cell cycle regulation circuits were among the top enriched pathways in samples with LRP1B mutations. Conclusion: Our studies suggested sequencing even a single, frequently mutated gene may provide insight into genome-wide mutational burden, and may serve as a biomarker to predict immune response.

The low-density lipoprotein receptor-related protein 1B (LRP1B), which encoding endocytic LDL-family receptor, is among the top 10 significantly mutated genes in human cancer (15). It has been demonstrated that LRP1B could bind to multiple extracellular ligands, including fibrinogen and apoEcarrying lipoproteins. Frequently inactivation mutation of LRP1B was observed in melanoma (16), lung cancer (17), esophagus squamous-cell carcinoma (18), head and neck squamous cancer (19,20), gastric cacner (21), and so on. Owing to its large size (coding sequence, 16 kbp), LRP1B is often missed as a significantly mutated gene analysis, whereas its mutation still could have a functional consequence in tumorigenesis and heterogeneity.
The characteristic mutational signatures are the fingerprints of endogenous and exogenous factors that have acted over the course of tumor development and progression. Exogenous mutational signatures, such as ultraviolet radiation exposure (22,23), tobacco smoking (9) and endogenous mutational signatures, such as APOBEC family of cytidine deaminases (24) mismatch repair defiency (10) were all contributed to higher tumor mutation burden and immune responding.
Tumor microenvironment (TME) also associate with response to ICB therapy. Baseline levels of tumor-infiltrating CD8 + T cells, CD4 T + cells and NK cells were shown to be correlated with the likelihood of immune response (25)(26)(27). Mutations in genes involved in antigen presentation and interferon-related circuits were reported with immune (28,29). Recently, a T cell-inflamed gene expression profile (GEP) was shown to predict response to immunotherapy (30,31).
LRP1B is one of the most frequently mutated genes in tumor samples; however, its associations with TMB and prognosis remain unclear. In this study, we investigated whether LRP1B mutations are associated with TMB

Genomic Data and Clinical Information of Melanoma and NSCLC
Somatic mutations were acquired from previous WES studies totaling 332 melanoma cases (5-7, 23, 34) and 113 NSCLC cases (8,9,32). Whole exome capture libraries were constructed using the Agilent SureSelect All Exon V2/V4 or 50 Mb kit. Enriched exome libraries were sequenced on a HiSeq 2000, 2500, or 4000 platform (Illumina) to generate paired-end reads (2 × 76 bp/100 bp) to a goal of 178X mean target coverage (range 32-380). The detailed sequencing information of each cohort was collected and illustrated in Supplementary Table S1. Mutations were re-annotated by oncotator against hg19 reference genome (33). Clinicopathological information including age, gender, stage, PD-L1 expression, smoke, ICB types, immune response status and survival were curated from supplemental materials of these studies (Supplementary Table S2). The predicted MHC binding affinity scores, HLA types, and clinical features were also collected from the previous studies (5, 7-9, 32, 34) (melanoma, n = 224; NSCLC, n = 113). Patients with complete or partial responses were considered to be efficacious to ICB treatment and the rest were regarded as the non-responding. Gene expression data in melanoma were available in Riaz.N and Hugo.W cohorts (6,23). Gene expression data for NSCLC cohort were not available. Somatic mutations for samples in the TCGA datasets of SKCM (n = 467), NSLC (n = 998) were downloaded from Genome Data Commons (https://portal.gdc.cancer.gov).

Deciphering Mutational Signature Operative in the Genome
We used SignatureAnalyzer (https://software.broadinstitute.org/ cancer/cga/Home) proposed by Kim et al. (35) to extract mutational signatures from the aggregated somatic mutation data in melanoma and NSCLC cohorts. This framework is based on Bayesian variant non-negative matrix factorization and it can automatically determine the optimal number of extracted mutational signatures. The SignatureAnalyzer factorized the mutation portrait matrix into two non-negative matrices W and H, where W representing mutational processes and H representing the corresponding mutational activities. The number of mutational signatures is the number of columns of matrix W. The rows of matrix A are the 96 mutational contexts (i.e., C > A, C > G, C > T, T > A, T > C, T > G, and their 5 ′ and 3 ′ adjacent bases), and its columns are the samples of each cohorts (melanoma, n = 332; NSCLC, n = 113). Mutational signatures were annotated by calculating cosine similarity against 30 validated mutational signatures in the Catalog of Somatic Mutations in Cancer (COSMIC, v85) (36).

LRP1B Mutation With Tumor Mutation Burden
The extracted mutational signatures were stratified as binary variables (i.e., 0 and 1) in the multivariate model. The classified method is according to the previous study, which a signature was considered significant if it contributed to more than 100 substitutions or more than 25% of total mutation activities (37). As mutations in TP53 and ATR and some mutational signatures increase mutation rates in the cancer genome, we used Generalized Linear Models and Fit Proportional Hazards Regression Model to analyze associations between LRP1B mutation and TMB by including them as confounding factors. TMB was defined the number of non-synonymous alterations (SNVs or indels) using whole exome sequencing for each patient, and the median served as the cutoff value of high vs. low.

Gene set Enrichment Analysis
We partitioned 72 cases whose gene expression profiles were available into two groups according to mutation status of LRP1B. Specifically, read counts of gene expression data were normalized by calcNormFactors in R package edgeR and remove batch effect by removeBatchEffect functions in package limma. We then fed them into lmFit and eBayes functions in the R limma package, and used these statistics as input to R-function in ClusterProfile package to do gene set enrichment analysis (GSEA). The gene sets examined in GSEA of REACTOME pathways were obtained from MSigDB database (v6.2) (38).

Tumor Infiltrating Lymphocyte Cell Analysis
We curated gene expression profile of 72 pre-treatment melanoma samples from two studies of Riaz and Hugo et al. (6,23). The relative abundance of 22 tumor infiltrating lymphocyte cells (TILs) in different LRP1B mutation status were estimated by the CIBERSORT algorithm (39), a computational approach for inferring leukocyte representation in bulk tumor transcriptomes.

T Cell-Inflamed Gene Expression (RNA) Profiling (GEP) Scores
We applied and followed the T cell-inflamed Gene expression (RNA) profiling (GEP) proposed by Ayers et al. (31) to

Statistical Analyses
The statistical analyses in this study was performed with R software (version 3.2.3). The continuous variables between groups were compared by the Wilcoxon rank sum test. The association between proportion of mutated genes and immune therapy response was evaluated by Fisher's exact test. Kaplan-Meier survival analysis and Cox proportional hazards model were used to analyze the association between mutated genes and prognosis with the R package survminer and forestmodel. Association of LRP1B mutation with TMB was examined by the logistic regression by including confounding factors such age, gender, stage, PD-L1, TP53, ATR mutation and extracted mutational signatures. All comparisons were two-sided with a significance level of 0.05, and the Benjamini-Hochberg method was applied to control false discovery rate (FDR) for multiple hypothesis testing (40).

LRP1B Mutation Predictive of Immunotherapy Survival in Melanoma
In Kaplan-Meier survival analysis, the LRP1B mutation was significantly associated with a better immunotherapy survival outcome in the pooled melanoma cohort (log-rank test, P = 0.005; Figure 2A). The association between LRP1B mutation with survival remained statistically significant after taking into account age, gender, TMB status and stage (Cox proportional hazards model, HR, 0.63 [95%CI, 0.40-0.97], P = 0.037; Figure 2B). We also noticed a preferable immune response status in LRP1B mutant samples (Fisher's exact test, P = 0.008).

LRP1B Mutation Associated With TMB
Melanoma samples with LRP1B mutations had a significantly higher tumor mutation burden and neoantigen burden by Wilcoxon rank sum test (log 2 TMB, 9.5 vs. 7.3, P < 0.001; log 2 NB, 9.2 vs. 6.9, P < 0.001; Figures 3A,B). We also observed the parallel result in TCGA melanoma cohort (log 2 TMB, 9.2 vs. 7.7; Wilcoxon rank sum test, P < 0.001; Supplementary Figure 2). Tumor mutation burden is largely attributed to genomic instability, which is prevalent in melanoma. In these samples, we extracted 6 mutational signatures (Supplementary Figures 3A,B) (Supplementary Figure 3C). The mutational activity attributable to each mutational signature in each melanoma sample and variation were shown in Supplementary Table S3.
To rule out the possibility that associations between LRP1B mutations and TMB were affected by these confounding factors, we included mutation associated signatures (3,7,11,26) and mutations in TP53 and ATR in the multivariate model. Associations between LRP1B mutations and TMB remained statistically significant after adjustment (logistic regression model, OR, 10.29; 95% CI, 4.70-24.26, P < 0.001; Figure 3C).

Further Corroboration of LRP1B Mutations in the NSCLC Cohort
As LRP1B mutation was common in lung cancer, we aggregated whole exome sequencing data from three studies of NSCLC immunotherapy to corroborate the results. LRP1B was also frequently mutated (35 of 113 patients [31.0%]) in the  Figure 4B).

Significantly Enriched Pathways and Immunocytes Associated With LRP1B Mutations
As LRP1B play an important role in immunotherapy outcomes, we investigate the potential mechanism behind LRP1B mutation and immune response. Gene set enrichment analysis (GSEA) on Reactome gene sets revealed enrichment of genes involved in Cell Cycle Mitotic and Antigen Processing and Presentation pathways were significantly altered in samples with LRP1B mutations (p.adj < 0.001, Figure 6A,  Supplementary Figure 6B).
Moreover, we evaluated (with use of CIBER algorithm) the abundance of tumor-infiltrating lymphocyte cells in melanoma microenvironment using the gene expression data. We found that CD8 + T cells, activated CD4 memory T cells, activated NK cells, were more enriched in LRP1B mutant type group, nevertheless, Neurtrophils was enriched in wild-type group (Figure 6C).

DISCUSSION
We analyzed LRP1B mutation with immune response and outcome in 332 samples from the melanoma immunotherapy cohort and 113 samples from the NSCLC immunotherapy cohort for further corroboration. LRP1B was frequently mutated in melanoma and NSCLC, and its mutation was associated with higher TMB and better survival outcome. The association of LRP1B mutation with TMB was independent of a significant presence of mutational signatures and of mutations in TP53 and ATR. Samples with LRP1B mutations were characterized by upregulation of signaling pathways involved in immune system, antigen processing and presentation and cell cycle checkpoints. Tumor-infiltrating immune cells results manifested LRP1B mutated samples were infiltrated in CD8 + T cells, NK cells and activated CD4 memory T cells, which supported the previous observations that such leucocytes and pathways present predominately in the tumor microenvironment of immune responder and promote the immune response (27,41).
Besides melanoma and non-small cell lung cancer, LRP1B was also frequently mutated in multiple types of human cancer. Several observations favor LRP1B being a bona fide driver mutation in many tumors, including a very high frequency of homozygous deletions (42), and frequent point mutation in esophagus (18), oral (20), liver (43), colon (44), gastric (21), breast (45), thyroid (46), and pancreatic cancer (47). Recently, LRP1B has been implicated in antigen presentation and as a regulator of inflammation and progression in cancer (47,48). Owing to its large size, LRP1B was often excluded from lists of significantly mutated genes. Another common ignored gene was MUC16, but Li et al. found MUC16 mutation was strongly associated with tumor mutation load and prognostic outcome in gastric cancer patients, and may benefits for patients with immunotherapy (49). The mechanisms underlying the association between TMB and LRP1B is not entirely clear. A leading hypothesis suggests that the large size (coding sequence,16 kbp) and gene location near the FRA2F fragile site (50) may contribute to a higher tumor mutation rate.
The classical role of LRPs is to modulate the clearance of a numerous of extracellular ligands from the pericellular microenvironment (51,52). In recent years, LRPs emerged as an important regulator of the inflammatory response. LRP1 was supposed to modulate the microglial immune response via regulation of c-Jun N-terminal kinase (JNK) and NF-κB signaling pathways (53). Overexpression of LRP1 is associated with worsened prognosis and suppressive tumor immunity in renal clear-cell carcinoma (54). Deletion of LRP5 and LRP6 in dendritic cells markedly delayed tumor growth and enhanced host antitumor immunity by blocking the Wnt pathway (55). Two SNPs in atopy-related immunologic candidate genes LRP1B were associated with pancreatic cancer risk, even after adjustment for multiple comparisons (47).
Recent studies also revealed LRP1B may as a tumor suppressor in common chemotherpay with deletion mutation of LRP1B causing a chemoresistance and poor outcome (56,57). In TCGA database, patients with LRP1B mutation have no benefits on conventional chemotherapy in melanoma and NSCLC (Supplementary Figure 9), but displayed preferable clinical outcome in immune check-point blockades therapy, suggested the molecular marker prediction of selectivity and specificity in drug treatment.
The main limitation is using the public dataset from different cohort which are somewhat heterogeneous in data processing and patient population. Although we utilized multiple datasets in different tumor type for analysis, but lack of the independent dataset for validation in the same tumor types. The analyzing tools used in sequencing data may have been different between these studies and may introduce bias in the final mutation lists. Besides, the number of samples with expression data in the cohort was limited, which limits the ability to adjust statistical power. As a result, association data between mutational patterns and gene expression, including analysis in immune cell infiltration and oncogenic pathways need further validation and studies.
Our studies suggested that sequencing even a single, frequently mutated gene may provide insight into genome-wide tumor mutational burden and could be served as a biomarker to predict immune response. The mechanisms through which LRP1B induce high mutations rate in tumors are still unknown and require future investigation. An investigation of a possible role of LRP1B in response to immune check-point blockade treatment in other cancer types (HNSC, ESCC, STAD, etc.) is warranted to study. The full implication of LRP1B in immunotherapy prognostic and monitoring remains elusive and requires in-depth studies.

DATA AVAILABILITY
All relevant data and materials within this work are made available in this manuscript and TCGA databases.

ETHICS STATEMENT
All studies have been approved by the Institutional Research Board.