Proteogenomic landscape of squamous cell lung cancer

Abstract

How genomic and transcriptomic alterations affect the functional proteome in lung cancer is not fully understood. Here, we integrate DNA copy number, somatic mutations, RNA-sequencing, and expression proteomics in a cohort of 108 squamous cell lung cancer (SCC) patients. We identify three proteomic subtypes, two of which (Inflamed, Redox) comprise 87% of tumors. The Inflamed subtype is enriched with neutrophils, B-cells, and monocytes and expresses more PD-1. Redox tumours are enriched for oxidation-reduction and glutathione pathways and harbor more NFE2L2/KEAP1 alterations and copy gain in the 3q2 locus. Proteomic subtypes are not associated with patient survival. However, B-cell-rich tertiary lymph node structures, more common in Inflamed, are associated with better survival. We identify metabolic vulnerabilities (TP63, PSAT1, and TFRC) in Redox. Our work provides a powerful resource for lung SCC biology and suggests therapeutic opportunities based on redox metabolism and immune cell infiltrates.

Fig. 1

Identification of three proteomic subtypes of SCC. One hundred eight patient tumors are displayed as columns, and the 1000 most variable proteins by absolute median deviation are displayed as rows. The patient tumors were organized by consensus clustering into five clusters corresponding to three biological subtypes (Inflamed, Redox, and Mixed). There is partial concordance with the Wilkerson et al. mRNA-based classifiers of these same samples, but the primitive group is not recapitulated. Mutation status and copy alterations of commonly mutated SCC genes/loci are shown directly above the heatmap. We identified five groups of proteins from the heatmap clustering. We took the proteins from each group and searched against GO: Biological Processes to yield a list of pathways (Enrichr). The topmost enriched pathway with P_adj ≤ 0.05 was used to label the protein clustering in the heatmap. The mean transcript-protein correlations for these pathways using matched RNAseq expression were: 0.46 for neutrophil degranulation, 0.56 for extracellular matrix organization, 0.35 for platelet degranulation, 0.64 for glutathione metabolic process, and 0.60 for bicarbonate transport

Fig. 2

The Inflamed subtype is immune rich. a Pathway enrichment of significantly different proteins (±1.5 fold-change and Wilcoxon P_adj ≤ 0.05 compared to the rest of the cohort). The Inflamed subtype was enriched for immune pathways including neutrophil degranulation. b–d Immune, Stromal, and ESTIMATE scores from the ESTIMATE algorithm. The Inflamed subtype had the highest median Immune score. e–o Box plots showing CIBERSORT results for each immune subtype compared across the three proteomic subtypes and between Inflamed A and Inflamed B. Inflamed had significantly higher proportions of memory B-cells (Wilcoxon P = 5.73E-03), monocytes (Wilcoxon P = 3.78E-04), and neutrophils (Wilcoxon P = 0.021) compared to the other subtypes, and plasma cells were significantly lower (Wilcoxon P = 0.048). M2 macrophages were significantly higher in Redox (Wilcoxon P = 0.016), and resting NK cells were higher in Mixed (Wilcoxon P = 0.027). Regulatory T-cells were significantly higher in Inflamed B (Wilcoxon P = 0.039), and neutrophils were significantly higher in Inflamed A (Wilcoxon P = 0.025).  box plots Significance was denoted using the following: * = P < 0.05, ** = P < 0.01, *** = P < 0.001. The center line indicates the median, the bounds of the box indicate the interquartile range (IQR: defined as the difference between the 75th and 25th percentiles), the topmost and bottom-most horizontal lines indicate the most extreme points less than 1.5 times the IQR below the 25th or above the 75th percentile, black points indicate outliers, and red points indicate individual values

Fig. 3

Scoring intratumoral and stromal myeloid lineage cells in Inflamed A and B. The IHC for case 1 at low (a) and high power magnification (b) highlights minimal CD33 + stromal neutrophils (in red) with a score of 1+ while background CD8+ stromal lymphocytes are highlighted in brown. The IHC for case 2 at low (c) and high power magnification (d) highlights a moderate CD33+ neutrophil population within the stroma with a score of 2 + with rare intratumoral CD33+ neutrophils. The IHC for case 3 at low (e) and high power magnification (f) illustrates the marked CD33+ neutrophil population within the stroma with a score of 3+. Scoring legend: 0 = virtual absence, 1 = low (<25%), 2 = moderate (25–50%), and 3 = marked increase (>50%)

Fig. 4

Tertiary lymph nodes (TLN) are enriched in the Inflamed subtype and are associated with better outcomes. a, b Proteomic subtypes are not associated with overall or recurrence-free survival. c TLN, indicated by red boxes, are observable in H&E stained slides under low magnification. d TLN can be observed in all three proteomic subtypes but are enriched in Inflamed (Cochran-Mantel-Haenszel P = 0.036). The height of each rectangle is proportional to the TLN score in a given subtype, and the width of each column is proportional to the number of samples in each subtype. e, f Presence of TLN provided a benefit to both overall and recurrence-free survival. g, h Immunohistochemical staining shows TLN primarily consist of CD20+ memory B-cells

Fig. 5

The Redox subtype is enriched for oxidative stress pathways and NFE2L2/KEAP1 alterations. a Pathway enrichment of significantly different proteins (±1.5 fold-change and Wilcoxon P_adj ≤ 0.05 compared to the rest of the cohort). The Redox subtype was enriched for oxidation-reduction and keratinization pathways. b 43 Redox samples (84%) had at least one genomic alteration of NFE2L2 or KEAP1. c Stratifying patients by NFE2L2 or KEAP1 mutation (NFE2L2/KEAP1 wild type patients vs. NFE2L2 mutant and KEAP1 mutant patients) revealed enrichment of oxidation-reduction and xenobiotic metabolism pathways in NFE2L2/KEAP1-altered patients (Enrichr). d, e We used DNA copy number to protein correlation >0.5 and correlation of RNA to protein >0.5 to identify 290 genes for subsequent analysis. f Highly correlated genes were intersected with significantly elevated proteins from the Redox subtype and genes that adversely impacted SCC viability in the Project DRIVE shRNA screen. g–l Project DRIVE screen results for squamous cell lung cancer cell lines. RSA < −3 is considered significant. Bar plots represent individual values

Fig. 6

The Mixed subtype has biology associated with Wnt/stromal signaling. a Pathway enrichment of significantly different proteins (±1.5 fold-change and P_adj < 0.05 compared to the rest of the cohort). There were no significantly elevated pathways in Mixed, but several pathways were decreased including immune and oxidation-reduction pathways. b Principal component analysis of the three subtypes reveal distinct clustering of Inflamed to the left, Redox on the right, and Mixed interspersed throughout. c Lollipop plot of APC alterations in the cohort with PFAM domain annotation (https://pfam.xfam.org/) extracted from cBioPortal (http://www.cbioportal.org). Mutations with black lines were nonsynonymous and red lines were truncating. Inflamed had no APC mutations