Introduction

Lung cancer is the leading cause of cancer-related death worldwide1, with adenocarcinoma representing the main histological subtype. Immune checkpoint inhibitors (ICI) could be offered to stage IV lung adenocarcinoma patients according to the tumor genotype, PD-L1 expression, patient’s performance status and comorbidities, with a substantial improvement. Several biomarkers have been described to predict immunotherapy response. However, benefits of ICI are only seen in ~15% of cancer patients and there is no strong validated predictor of response used in clinical practice. Identification and development of predictive biomarkers of ICI response are critical.

Somatic mutations in the NF1 tumor suppressor gene are found in 5% to 15% of lung adenocarcinoma2. NF1 encodes neurofibromin an inhibitor of the RAS/MAPK and PI3K/AKT/mTOR pathways. Several studies have shown that NF1-mutated lung adenocarcinoma is a distinct clinical and molecular subtype3,4. ICI sensitivity signals have been reported in NF1-mutated tumors, mainly in melanoma5,6. Only few published clinical data—limited to two case reports7,8—showed ICI sensitivity in NF1-mutated lung cancer. The imputability of NF1 mutation in ICI response was not demonstrated in these reports.

We sought to investigate whether NF1 could be a predictive biomarker of response to ICI in a large cohort of lung adenocarcinoma.

In 14 TCGA cohorts, 341 out of 4181 lung adenocarcinoma patients (8.2%) presented a NF1 mutation. TMB was significantly higher in NF1-mutated versus NF1 wild-type (WT) lung adenocarcinoma (p < 0.0001) with a mean TMB at 14.1 mut/Mb [0.7–65.7] in NF1-mutated tumors versus 6.5 mut/Mb [0.0–96.5] in NF1 WT tumors (Fig. 1, Supplementary Fig. 1). NF1-mutated tumors also showed a higher mean TMB (14.1 mutations/Mb) than TP53-, KRAS-, KEAP1-, and STK11-mutated tumors (mean TMBs at 10.5, 8.6, 11.2, and 8.8 respectively) (Fig. 1 and Supplementary Table 1). The factors significantly associated with an elevated TMB according to a multivariate analysis are: smoking status (p = 0.018) and mutational status of NF1 (p = 0.0183), TP53 (p < 0.0001) and KEAP1 (p = 0.0101) (Table 1).

Fig. 1: TMB in lung adenocarcinoma according to mutation status.
figure 1

a TMB in NF1 mutated lung adenocarcinomas (N = 341 tumors) versus NF1 WT lung adenocarcinomas (N = 3840 tumors) showed a statistically significant difference with a mean TMB at 14.1 mut/Mb in NF1-mutated tumors vs 6.5 mut/Mb in NF1 WT tumors (p < 0.0001). b NF1-mutated lung adenocarcinomas showed a higher mean TMB vs TP53-, KRAS-, KEAP1-, and STK11-mutated lung adenocarcinomas. Numbers of TP53, KRAS, KEAP1 and STK11 co-mutations for the 341 NF1-mutated samples are available in Supplementary Table 1. ***P ≤ 0.001 and ****P ≤ 0.0001.

Full size image
Table 1 Univariate and multivariate logistic regression analysis of factors associated with a high TMB
Full size table

We compared the mRNA levels of nine immune genes in NF1-mutated (N = 66; 9.6%) vs NF1 WT (N = 620; 90.4%) lung adenocarcinomas (Fig. 2). We observed a significantly higher expression of CXCL9 (p = 0.008), PD-L1 (p = 0.010), PD-L2 (p = 0.011), and CD8A (p = 0.006) in NF1-mutated samples vs NF1 WT samples. Lung adenocarcinoma samples with a TMB ≥ 10 mut/Mb also had a higher expression of these same genes compared to samples with a TMB < 10 mut/Mb. TP53-mutated samples had a significantly higher expression of all investigated immune markers than TP53 WT samples. Conversely, KRAS-mutated samples showed no increase in the expression of immune genes compared to KRAS WT tumors. KEAP1 and STK11-mutated tumors showed a significant decrease in the expression of the nine immune genes vs WT tumors (Supplementary Table 2).

Fig. 2: NF1-mutated lung adenocarcinoma immune infiltrate profiles.
figure 2

a–j show the mRNA expression levels of nine genes implicated in inflammation and immune checkpoint inhibitors response, in NF1-mutated (n = 66) and wild-type (n = 620) lung adenocarcinomas: CD4 (a), CD8A (b), CD8B (c), CTLA4 (d), CXCL9 (e), CXCL13 (f), IDO1 (g), PD1 (h), PDL1 (i), and PDL2 (j). Each plot represents the mRNA level of key immune genes (y axis) in NF1-mutated (in red) and WT (in blue) tumors (x axis). A significant increase in CD8A, PDL1, CXCL9, and PDL2 expression is observed in NF1-mutated vs WT tumors. k shows the immune infiltrate abundances (y axis) for B cells, CD4+ T cells, CD8+ T cells, dendritic cells, macrophages, and neutrophils (x axis) in NF1-mutated (in red) and WT (in blue) tumors. A significant CD8+ and macrophage infiltration increase is observed in NF1-mutated vs WT tumors. WT: wild-type for NF1. *p ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001 and ****P ≤ 0.0001.

Full size image

Using the 542 lung adenocarcinoma samples available in TIMER database, we estimated the abundance of six different cell types: B cells, CD4+ T cells, CD8+ T cells, dendritic cells, macrophages, and neutrophils. NF1 mutations were associated with significantly higher CD8 + T cells (p = 0.03) and macrophage (p = 0.02) infiltrations in lung adenocarcinoma compared to NF1 WT tumors (Fig. 2). TP53 mutations were associated with significantly higher CD8 + T cells (p = 0.01), neutrophils (p = 0.001), and dendritic cells (p = 0.03) infiltrations. Conversely, KRAS mutations were associated with significantly lower B cells (p = 0.02) and dendritic cells (p = 0.04) infiltrations. STK11 and KEAP1 mutations were associated with significantly lower infiltrations of all six markers (all p < 0.03) (Supplementary Fig. 2).

A heatmap representing the mRNA expression z-score in 686 lung adenocarcinomas with available expression data (NF1 mutant: N = 66, 9.6%; NF1 WT: N = 620, 90.4%) is shown (Fig. 3). We identified three clusters of tumors according to the level of expression of these immune genes, that we named “hot tumors”, “warm tumors”, and “cold tumors”. A significant enrichment of NF1-mutated tumors was identified in “hot tumors”: OR[IC95] = 1.84[1.07, 3.18] (Fisher p = 0.02015). No NF1-mutated tumors enrichment was identified in “warm” (OR[IC95] = 0.74[0.40, 1.32], Fisher p = 0.3416) or in “cold” tumors (OR[IC95] = 0.67[0.35, 1.24], Fisher p = 0.2106). TMB was higher in “hot” tumors than in “warm” (p = 0.013) and “cold” (p < 0.001) tumors. All NF1-mutated samples with available expression data had a TMB > 10 mut/Mb.

Fig. 3: Lung adenocarcinoma clustering according to the mRNA expression of immune genes (N = 686).
figure 3

Heatmap shows the expression of PD-L1 (encoded by CD274), PD-L2 (encoded by PDCD1LG2), CXCL9, and CD8A. NF1 mutation status (upper) and TMB (lower) are indicated. The number of NF1 patients and the TMB [mean (95CI)] based on a three-part clustering: “hot”, “warm” and “cold” tumors are indicated. TMB: Tumor Mutational Burden (mut/Mb). 95CI: 95% confidence interval.

Full size image

In a 2023 single-center retrospective study, Wang et al. showed an enhanced proliferation and immune activity of macrophages and NK cells in case of germline NF1 mutations9 in patients with juvenile myelomonocytic leukemia. Data are available for NF1-mutated sporadic cancers—which mainly include melanoma. Johnson et al. conducted a study on 65 patients with advanced melanoma treated with ICI. The subgroup of patients with NF1-mutated melanomas had a higher TMB and better response rates (74%) than those with BRAF/NRAS-mutated and wild-type melanomas5. Furthermore, a retrospective multicenter analysis revealed a significantly better median OS (p = 0.0154) when receiving first-line immune checkpoint inhibitor treatment for NF1-mutated (n = 80) than for wild-type (n = 432) melanomas6. However, there are currently no clinical trials evaluating ICI efficacy that enrolled patients according to a stratified randomization on their NF1 status2. In the context of lung adenocarcinomas, only two case reports showed durable responses to ICI in two patients with sporadic NF1-mutated cancers. The first one showed a progression free survival of 95.4 weeks with pembrolizumab7. The second reported a stable disease for 8 months in a stage IV lung adenocarcinoma patient treated with pembrolizumab8. These observations support the need to explore NF1 as a predictive biomarker of ICI sensitivity in the context of lung adenocarcinoma.

MSK-IMPACT study showed that response to ICI is associated with a high TMB, whatever tumor histological subtype. In lung cancers, an association was identified between an elevated TMB and durable clinical benefits of ICI10,11,12. Three studies presented at recent congresses confirmed a significantly higher TMB (P < 0.001, p < 0.0001 and p < 0.0001) and a higher expression of PD-L1 expression (p < 0.01, p = 0.05 and p < 0.0001) in NF1-mutated versus NF1-WT lung cancers13,14,15. Our in silico analysis of 686 samples confirmed these results: NF1 mutation was associated with a higher TMB (p < 0.0001) and a higher expression of key immune genes. We showed a significant enrichment of NF1-mutated samples in “hot tumors” (p = 0.02) and an increased CD8 + T-cell infiltration (p = 0.03).

Given that NF1-mutated lung adenocarcinoma is frequently co-mutated (and especially with TP53), one cannot exclude an impact of these co-mutations on the TMB and immune infiltration. However, our multivariate analysis shows that NF1 status is significantly associated (p = 0.0183) with high TMB independently of the co-mutation status. TP53 mutations have been reported to be a biomarker associated with ICI benefit response, in particular in case of KRAS co-mutations10. A 2017 clinical study aimed to identify novel biomarkers for immune check-point inhibitor responses in 904 patients with lung adenocarcinoma16. Patients with TP53-mutated lung adenocarcinoma were characterized by a significant increased expression of PD-L1 (p < 0.05) and TMB (p < 0.001), and a higher tumor infiltration by CD8 + T cells (p < 0.05). In 2020, a study of 637 patients with non-small cell lung cancer (NSCLC) confirmed a high TMB (p = 0.007) and high PD-L1 expression (p < 0.001) in TP53-mutated samples10. NF1 mutation status was not investigated in these studies. Here, we confirmed that TP53 mutations were correlated with higher TMB (p < 0.0001) and higher expression of immune markers. Pilar et al. showed that the combination of multiple biomarkers (PD-L1 immunohistochemistry, T-cell infiltration, and TMB) increased performance, compared with the three biomarkers alone17. Our results suggest that a mixed score including NF1 should be considered.

The presence of NF1 mutations was correlated with high TMB. This sole observation did not allow to determine if NF1 mutations could be a consequence of genetic instability. The transcriptional and immune analyses were limited to small datasets. Because of this limited size, we could not check whether the subgroup of NF1-mutated patients with low TMB expressed markers of immune infiltration. In addition, one of the limitations of in silico analysis is the lack of available clinical data. We did not have sufficient clinical outcomes of patients with NF1-mutated lung adenocarcinoma and treated with ICI in these TCGA public databases. We were unable to make a multivariate model to predict the efficacy of ICI. We therefore looked for factors associated with a high TMB, that is correlated with durable clinical benefits of ICI. To study the impact of genes on ICI response, a progression free survival analysis would have been necessary. Additional studies are warranted to specify the biological effect of NF1 mutations on immune environment in lung adenocarcinoma.

We showed that NF1-mutated lung adenocarcinomas are associated with a high TMB, high immune genes expression, and CD8 + T-cells infiltration. To date in clinical practice, besides PD-L1 expression that is actually use to drive the choice of immunotherapy regimens in non-small cell lung cancer that lacks a driver mutation, there is no strong predictive factor of immune checkpoint blockade. We suggest the potential of NF1 alterations as a novel biomarker for ICI, warranting further investigations alone or in combination with other ICI predictive biomarkers. Our results highlight the need for clinical trials evaluation ICI efficacy specifically dedicated to NF1-mutated lung adenocarcinomas with a randomization based on NF1 status.

Methods

Data acquisition

We studied in silico data using 14 publicly available datasets with lung adenocarcinoma samples on the cBioPortal website (The Cancer Genome Atlas [TCGA] data, https://www.cbioportal.org/): MSKMind (Nature Cancer 2022, n = 247 samples), Broad (Cell 2012, n = 183 samples), CPTAC (Cell 2020, n = 110 samples), MSK (2021, n = 186 samples), MSK (J Thorac Oncol 2020, n = 604 samples), MSK (NPJ Precision Oncology 2021, n = 426 samples), MSK (Science 2015, n = 35 samples), OncoSG (Nat Genet 2020, n = 305 samples), TCGA (Firehose Legacy, n = 586 samples), TCGA (Nature 2014, n = 230 samples), TCGA (PanCancer Atlas, n = 566 samples), TSP (Nature 2008, n = 163 samples), NCI (Nature Genetics 2021, n = 232 samples) and MSK (Cancer Discov 2017, n = 915 samples). After removing non-lung adenocarcinoma samples and duplicates, 4181 lung adenocarcinoma samples were included in the analysis.

Tumor mutational burden and genetic alterations

We analyzed the presence and type of NF1 alterations, the level of NF1 mRNA expression (log2 RNA seq V2 RSEM), and the tumor mutational burden (TMB) defined as the number of mutations per megabase (mut/Mb).

We compared the NF1-mutated lung adenocarcinomas profile with TP53- and KRAS-mutated tumors which are likely associated with ICI response10,16,18, and KEAP1- and STK11-mutated tumors which are likely known to be resistant to ICI18,19,20,21. For KRAS-mutated tumors, we only included tumors with KRAS hotspot mutations.

Tumor immune infiltrates

We determined the tumor immune infiltrates using the mRNA expression level of markers known to be predictive of ICI response17,22,23: PD-1, PD-L1, PD-L2, CTLA4, IDO, CXCL9, CXCL13, CD4, and CD8. Only five datasets had mRNA expression data: Firehorse Legacy (n = 341 lung adenocarcinoma), PanCancer Atlas (n = 196 lung adenocarcinoma), Nature 2014 (n = 49 lung adenocarcinoma), Nat Genet 2020 (n = 304 lung adenocarcinoma), and Cell 2020 (n = 110 lung adenocarcinoma). After removing duplicates and samples without mRNA expression of the tested immune markers, the mRNA expression analysis was performed in 686 samples.

Using TIMER (Tumor Immune Estimation Resource, https://cistrome.shinyapps.io/timer/), a public web server for systematical analysis of immune infiltrates, we estimated the abundance of six immune cell populations including B cells, CD4+ T cells, CD8+ T cells, dendritic cells, macrophages, and neutrophils in 542 lung adenocarcinoma samples.

Statistical analyses

Comparisons were performed with Student t-test or Wilcoxon for quantitative variables. ANOVA with false discovery rate correction (two-stage step-up method) was performed for multi-group comparison. Clustering analyses and heatmap visualization were generated using the “ComplexHeatmap” R package. Samples clustering was performed according to the mRNA expression level of significant immune genes. Enrichment score was calculated using the Fisher exact test. We set up univariate and multivariate logistic regression models to determine the factors associated with a high TMB (defined as >10 Mut/Mb). We included as clinical and genomic criteria: age, gender, smoking status (cut-off: 20 pack-years) and the mutational status of NF1, KRAS, TP53, KEAP1 and STK11. All p-values were two-sided, and the level of significance was set at p < 0.05. Statistical analyses were performed with R statistical software (version 4.2.2) and GraphPad Prism (version 10.0.0).

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.