Integrative proteomic characterization of adenocarcinoma of esophagogastric junction

Abstract

The incidence of adenocarcinoma of the esophagogastric junction (AEG) has been rapidly increasing in recent decades, but its molecular alterations and subtypes are still obscure. Here, we conduct proteomics and phosphoproteomics profiling of 103 AEG tumors with paired normal adjacent tissues (NATs), whole exome sequencing of 94 tumor-NAT pairs, and RNA sequencing in 83 tumor-NAT pairs. Our analysis reveals an extensively altered proteome and 252 potential druggable proteins in AEG tumors. We identify three proteomic subtypes with significant clinical and molecular differences. The S-II subtype signature protein, FBXO44, is demonstrated to promote tumor progression and metastasis in vitro and in vivo. Our comparative analyses reveal distinct genomic features in AEG subtypes. We find a specific decrease of fibroblasts in the S-III subtype. Further phosphoproteomic comparisons reveal different kinase-phosphosubstrate regulatory networks among AEG subtypes. Our proteogenomics dataset provides valuable resources for understanding molecular mechanisms and developing precision treatment strategies of AEG.

Fig. 1. Multi-omics landscape of adenocarcinoma of the esophagogastric junction (AEG).

a Schematic overview of the experimental design and data acquisition process for proteomics, phosphoproteomics, WES, and RNA-seq. NAT indicates normal adjacent tissue. b The genomic profiles of AEG patients. The top panel shows the tumor mutation burden (TMB) in each patient. The top bars show the clinicopathological features of AEG patients. The middle panel is the oncoplot generated with maftools depicting the top 30 mutated cancer-related genes in the present AEG cohort. The bottom panel shows the proportion of different types of nucleotide substitutions in each patient. The right panel represents mutation types and frequencies for each gene. P, two-sided Wilcoxon’s rank test for age, sex, smoking, and alcohol, and Kruskal–Wallis rank sum test for Siewert type and tumor stages. c Overview of the proteomics profile in 103 AEG patients. d Overview of the phosphoproteomics profile of 206 samples from 103 AEG patients. e Overview of the RNA-seq profile of 83 tumors and paired NAT samples.

Fig. 2. Proteomic variations in AEG tumors.

a Volcano plot showing the difference in proteins between AEG tumor and paired NAT samples. Red circles represent upregulated proteins (FDR < 0.01 and log2(fold change) > 1), and blue circles indicate down-regulated proteins (FDR < 0.01 and log2(fold change) < −1). b Functional enrichment results of upregulated and downregulated proteins, respectively. c Heatmap showing the difference in the protein abundance of hallmarks between AEG tumor and paired NAT samples. Different font colors indicate different hallmark categories. FDR, adjusted P from Wilcoxon’s rank-sum test. d Comparisons of integrated abundances of “apical junction” and “KRAS signaling up” between AEG tumor and paired NAT samples (n = 103). Each box represents the IQR and median of the hallmark scores in each group, whiskers indicate 1.5 times IQR. P, two-sided Wilcoxon’s rank-sum test. e Kaplan–Meier survival curves comparing groups with high (n = 51) and low (n = 52) abundance of “apical junction” and “KRAS signaling up” gene sets, respectively. f Heatmap showing the difference in the proteins that are included in at least two sets of top 50 DEPs, top 50 proteins with the largest degree, closeness, or betweenness. Bubble plot on the right shows the degree, closeness, or betweenness of the corresponding proteins in the PPI network. ^*P < 0.05, ^**P < 0.01, ^***P < 0.001, two-sided Wilcoxon’s rank-sum test.

Fig. 3. Proteomic subtyping of AEG tumors.

a Heatmap showing the differentially expressed proteins among the three subtypes. Tiling bars above the heatmap show the distribution of different clinicopathological characteristics among the three subtypes. P, Fisher’s exact test. b Kaplan–Meier survival curve comparing patients in different subtypes (n = 40 for S-I, n = 23 for S-II, n = 40 for S-III). The hazards ratio (HR) with 95% confidence interval (CI) is also shown. c Volcano plot showing the difference in subtype-specific mutated genes. P, Fisher’s exact test. d The differences in integrated protein abundances of hallmarks comparing tumor and NAT samples in each subtype. e Comparison of the integrated abundance of “activity of G2M checkpoint” among three subtypes (n = 40 for S-I, n = 23 for S-II, n = 40 for S-III). P, two-sided Wilcoxon’s rank-sum test. f Comparison of the integrated abundance of “activity of MYC targets” gene set among the three subtypes (n = 40 for S-I, n = 23 for S-II, n = 40 for S-III). P, two-sided Wilcoxon’s rank-sum test. g Twelve signature proteins that are significantly associated with patient survival. The heatmap on the left shows the relative abundance of signature proteins in tumor and paired NAT samples of each subtype. The forest plot on the right shows the prognostic score for each protein in multivariate Cox regression analysis. The middle points indicate hazard ratios. The endpoints represent lower or upper 95% confidence intervals. Red indicates unfavorable proteins, while blue indicates favorable proteins. P, multivariate Cox proportional-hazards. In e and f, each box represents the IQR and median of integrated abundance in each subtype, whiskers indicate 1.5 times IQR.

Fig. 4. Clinical relevance and biological functions of FBXO44.

a Comparison of FBXO44 protein abundance between tumor and NAT samples in each subtype and tumors from different subtypes. FC indicates fold change (n = 40 for S-I, n = 23 for S-II, n = 40 for S-III). Each box represents the IQR and median of normalized protein abundance in normal or tumor samples of each subtype, whiskers indicate 1.5 times IQR. P, two-sided Wilcoxon’s rank-sum test. b The distribution of FBXO44 in an independent cohort of tumor and NAT samples from 251 AEG patients. P, chi-squared test. c Kaplan–Meier survival curve comparing FBXO44-high (n = 52) and -low (n = 51) abundance patients. P, log-rank test. d Kaplan–Meier survival curve of FBXO44 in an independent clinical cohort (n = 61 for FBXO44+, n = 190 for FBXO44−). P, log-rank test. e Cell proliferation assays of FBXO44 OE or KD OE19 and SK-GT-4 cell lines (n = 3 biological replicates). P, two-sided Student’s t test. f Transwell invasion assays of FBXO44 OE or KD OE19 and SK-GT-4 cell lines (n = 3 biological replicates). P, two-sided Student’s t test. g Cell migration assays of FBXO44 OE or KD OE19 and SK-GT-4 cell lines (n = 3 biological replicates). P, two-sided Student’s t test. h Representative bioluminescent images of mice bearing OE19 xenograft tumors harboring FBXO44 OE or FBXO44 KD or their corresponding controls at different time points after injection. i Representative bioluminescent images of mice bearing orthotopic OE19 tumors harboring FBXO44 OE or FBXO44 KD or their corresponding controls at different time points post implantation. j Representative images of liver metastasis in mice bearing orthotopic OE19 tumors. In e–g, error bars represent mean ± SDs.

Fig. 5. Comparisons of genomic features among the three proteomic subtypes.

Mutation signatures and the best match SBS signatures in the S-I (a), S-II (b), and S-III (c) subtypes. Somatic interactions of the top mutated genes in the S-I (d), S-II (e), and S-III (f) subtypes. Asterisks indicate statistical significance (Fisher’s exact test). g Fractions of affected pathways and samples by gene mutations in major oncogenic pathways. h Oncoplot of the RTK-RAS pathway in the three subtypes. Red font indicates tumor suppressor genes and blue font represents oncogenes.

Fig. 6. Immune infiltration across different proteomic subtypes.

a Heatmap showing the relative abundance of different cells across samples of the three AEG subtypes. The Kruskal–Wallis Rank Sum test was used to compare the differences between subtypes. P, Kruskal–Wallis rank sum test. b The difference in the relative abundance of different infiltrating cells in the three AEG subtypes. FDR, Wilcoxon’s rank-sum test. c Comparisons of aDC abundance between AEG tumor and NAT samples in the S-I, S-II, and S-III subtypes (n = 40 for S-I, n = 23 for S-II, n = 40 for S-III). FDR, Wilcoxon’s rank-sum test. d Comparisons of fibroblast abundance between AEG tumor and NAT samples in the S-I, S-II, and S-III subtypes (n = 40 for S-I, n = 23 for S-II, n = 40 for S-III). FDR, Wilcoxon’s rank-sum test. e H&E analysis of tumor cells, lymphoid cells, myeloid cells and fibroblasts across three AEG subtypes. Scale bars used for 0×, 40×, and 200× magnification were 1, 500, and 100 μm, respectively. f The differential significance of the protein expression of immune checkpoints across the three AEG subtypes. In c and d, each box represents the IQR and median of the relative cell abundance in normal or tumor samples of each subtype, whiskers indicate 1.5 times IQR.

Fig. 7. Phosphoproteomic analyses in three AEG subtypes.

a Volcano plot showing the differential significance of phosphorylation sites. Red circles represent sites with increased phosphorylation (FDR < 0.01 and log2(fold change) > 1) and blue circles indicate downregulated phosphorylation sites (FDR < 0.01 and log2(fold change) < −1). b Enriched biological processes of differentially phosphorylated sites in each subtype. c Kinase enrichment of differentially phosphorylated sites in each AEG tumor subtype. P, Fisher’s exact test. Asterisks (*) represent statistical significance (P < 0.05). Kinase-phosphosubstrate regulatory networks in tumors of the S-I (d), S-II (e), and S-III (f) subtypes.