Integrated proteogenomic characterization of localized prostate cancer identifies biological insights and subtype-specific therapeutic strategies

Abstract

Localized prostate cancer (PCa) is highly variable in their response to therapies. Although a fraction of this heterogeneity can be explained by clinical factors or genomic and transcriptomic profiling, the proteomic-based profiling of aggressive PCa remains poorly understood. Here, we profiled the genome, transcriptome, proteome and phosphoproteome of 145 cases of localized PCa in Chinese patients. Proteome-based stratification of localized PCa revealed three subtypes with distinct molecular features: immune subgroup, arachidonic acid metabolic subgroup and sialic acid metabolic subgroup with highest biochemical recurrence (BCR) rates. Further, we nominated NANS protein, a key enzyme in sialic acid synthesis as a potential prognostic biomarker for aggressive PCa and validated in two independent cohorts. Finally, taking advantage of cell-derived orthotopic transplanted mouse models, single-cell RNA sequencing (scRNA-seq) and immunofluorescence analysis, we revealed that targeting NANS can reverse the immunosuppressive microenvironment through restricting the sialoglycan-sialic acid-recognizing immunoglobulin superfamily lectin (Siglec) axis, thereby inhibiting tumor growth of PCa. In sum, we integrate multi-omic data to refine molecular subtyping of localized PCa, and identify NANS as a potential prognostic biomarker and therapeutic option for aggressive PCa.

Fig. 1. Proteomic subgroups and molecular characteristics of PCa.

A Study schematic. The left panel shows the multi-omic experimental design of the discovery cohort. The right panel shows the type of validation experiments, including formalin fixed paraffin-embedded (FFPE) proteome and immunohistochemistry (IHC) in two validation cohorts. LC-MS/MS represents liquid chromatography tandem mass spectrometry. WES represents whole exome sequencing, IF represents immunofluorescence. Created with BioRender.com, released under a Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International license. B Heatmap visualizes the three proteomic subgroups of PCa based on the non-negative matrix factorization (NMF) unsupervised clustering algorithm and their corresponding clinical features in the discovery cohort. PSA represents prostate-specific antigen. BCR represents biochemical recurrence. NA represents not available. C Comparison of the proteomic subtyping (NMF-based) and two kinds of transcriptomic subtyping (NMF-based and PAM50-based) in the discovery cohort. NA represents not available. D Gene sets enrichment analysis (GSEA) showing distinct molecular characteristics among three proteomic subgroups of the discovery cohort (one subgroup vs the union of other two subgroups). E Kaplan-Meier curves showing the biochemical recurrence-free survival (bRFS) of patients in three proteomic subgroups of the discovery cohort. P value is calculated by Log-rank test. F Bar plots comparing the clinicopathological features among three proteomic subgroups of the discovery cohort, including prostate-specific antigen (PSA), Gleason score (GS) and lymph node metastasis (N stage). P values are determined using one-way ANOVA test. G GSEA showing the activation of the androgen response pathway in subgroup 2 compared to the union of subgroup 1 and 3 of the discovery cohort. Source data are provided as a Source Data file.

Fig. 2. Multi-omics profile of three PCa subgroups.

A Comparison of tumor mutational burden (TMB) and somatic copy number aberration (SCNA) burden among three proteomic subgroups of the discovery cohort (n = 26/40/61 for subgroup 1/2/3). The middle lines represent the median. Lower/upper hinges denote 25–75% IQR with whiskers extending to 1.5 IQR. P-values are determined using two-tailed Student’s t test. B Bar plot comparing the FOXA1 mutation frequency among three proteomic subgroups of the discovery cohort (top panel). Numbers at the top of the bar represent counts of each subgroup, with corresponding proportions at the bottom of the bar. Prevalence of top 30 recurrent mutation genes in three proteomic subgroups of the discovery cohort (bottom panel). P values are determined using one-way ANOVA test. C Prevalence of specific SCNAs and gene fusions in three proteomic subgroups of the discovery cohort (top panel). NA represents not available. Bar plots comparing frequency of specific SCNAs and gene fusions among three proteomic subgroups of the discovery cohort (bottom panel). Numbers at the top of the bar represent counts of each subgroup, with corresponding proportions at the bottom of the bar. P values are determined using one-way ANOVA test. D Kinase-substrate enrichment analysis (KSEA) showing significantly up-regulated and down-regulated kinases in subgroup 2 patients of the discovery cohort (subgroup 2 vs subgroup 1 and subgroup 3 combined). Red bars refer to FDR < 0.05 and z score > 0, blue bars refer to FDR < 0.05 and z score ≤ 0. Source data are provided as a Source Data file.

Fig. 3. Higher sialic acid metabolism and NANS expression in subgroup 2 tumors associated with poor prognosis.

A GSEA showing top 10 up-regulated metabolic pathways in subgroup 2 of the discovery cohort (subgroup 2 vs subgroup 1 and 3). P values are determined using two-sided Fisher’s exact test. B Differential expression of key enzymes in the sialic acid synthesis pathway among three subgroups of the discovery cohort. C Comparison of N-Acetylneuraminic acid abundance among three subgroups of the discovery cohort based on metabolomics (n = 7/7/8 for subgroup 1/2/3). The middle lines represent the median. Lower/upper hinges denote 25-75% IQR with whiskers extending to 1.5 IQR. P values are determined using two-sided Wilcoxon rank-sum test. D Kaplan-Meier curves showing bRFS of patients in three subgroups of the validation cohort 1. P value is calculated by Log-rank test. E GSEA showing the up-regulation of the amino sugar and nucleotide sugar metabolism pathway in subgroup 2 of the validation cohort 1 (subgroup 2 vs subgroup 1 and 3). F Comparison of NANS protein abundance among tumors of three subgroups in the discovery cohort (n = 31/46/68 for subgroup 1/2/3). The middle lines represent the median. Lower/upper hinges denote 25-75% IQR with whiskers extending to 1.5 IQR. P-value is determined using two-sided Wilcoxon rank-sum test. G Immunohistochemistry staining for NANS in tumors of three subgroups in the discovery cohort (n = 37/44/38 for subgroup 1/2/3). Scale bar represents 100um. Data are presented as mean ± standard deviation (SD). P values are determined using two-tailed Student’s t test. H Comparison of NANS proteomic abundance among tumors of three subgroups in the validation cohort 1 (n = 37/44/38 for subgroup 1/2/3). The middle lines represent the median. Lower/upper hinges denote 25–75% IQR with whiskers extending to 1.5 IQR. P values are determined using two-sided Wilcoxon rank-sum test. I Kaplan-Meier curves showing bRFS of patients in the discovery cohort. P value is calculated by Log-rank test. (J)Kaplan-Meier curves showing bRFS of patients in the validation cohort 1. P-value is calculated by Log-rank test. K Kaplan–Meier curves showing bRFS of patients in the validation cohort 2 grouped by immunostaining intensity of NANS. P value is calculated by Log-rank test. Source data are provided as a Source Data file.

Fig. 4. NANS increases the sialic acid anabolism and sialylation level accounting for immunosuppressive microenvironment of PCa.

A IHC staining and quantification for α2,3 and α2,6-sialylation modifications in tumor regions of three proteomic subgroups in the discovery cohort (n = 37/44/38 for subgroup 1/2/3). The scale bar represents 100 um. Data are presented as mean ± SD. P values are determined using two-tailed Student’s t test. B Heatmap visualizes the proteomic subtypes of 8 PCa cell lines based on the nearest template prediction (NTP) algorithm. Three replicates for each cell line. C Knockout (KO) of NANS in 22Rv1 and Myc-CaP cells via single guide RNA (sgRNA) as determined by qRT-PCR. Empty vector is used as control. n = 3 biological replicates for each group. Data are presented as mean ± SD. P values are determined using two-tailed Student’s t test. D KO of NANS in 22Rv1 and Myc-CaP cells via sgRNA as determined by Western blot. Empty vector is used as control. P values are determined using two-tailed Student’s t test. E Comparison of the relative abundance of N-Acetylneuraminic acid in 22Rv1 and Myc-CaP cells between sgNANS and control group based on the targeted metabolomics analysis. The middle lines represent the median. Lower/upper hinges denote 25–75% IQR with whiskers extending to 1.5 IQR. n = 5 biological replicates for each group. P values are determined using two-sided Wilcoxon rank-sum test. F Heatmap showing differentially expressed sialylated sites in 22Rv1 and Myc-CaP cells between sgNANS and control group. G EcoTyper analysis dissecting the tumor immune microenvironment of patients in three proteomic subgroups of the discovery cohort. H Multiplex immunofluorescence (mIF) staining and quantification of immune cells in three proteomic subgroups of the discovery cohort. n = 10 different samples for each group. The scale bar represents 50um. Data are presented as mean ± SD. P values are determined using two-tailed Student’s t test. Source data are provided as a Source Data file.

Fig. 5. Inhibition of NANS reverses the immunosuppressive microenvironment and suppresses the tumor growth of PCa in orthotopic transplanted mouse model.

A Uniform manifold approximation and projection (UMAP) visualization of 29515 single cells from 6 mice tumor tissue samples, colored by cell type annotations. B UMAP visualization of epithelial cells, colored by cell type annotations. C Comparison of the Ro/e (the ratio of observed to expected cell numbers) in 4 epithelial cell subtypes before and after Nans knockout. P-values are determined using two-sided Chi-square test. D Bubble plot showing differentially activated hallmark pathways and sialic acid-related pathway among 4 epithelial cell subtypes. E Comparison of the Ro/e in all cell types before and after Nans knockout. P values are determined using two-sided Chi-square test. F UMAP visualization of macrophages, colored by cell type annotations. G Comparison of the Ro/e in 4 macrophage subtypes before and after Nans knockout. P values are determined using two-sided Chi-square test. H Dot plot depicting the expression of detected Siglec genes in each identified cell subtype, where dot size and color represent percentage of marker gene expression (pct. exp) and the averaged scaled expression (avg. exp. scale) value, respectively. I Heatmap depicting the enrichment of classical immune-related pathways in each identified macrophage subtype. J IF staining and quantification of sialylation level (SNA/MALII) and immune cells in murine tumor tissues of the sgNans and control group. n = 3 for each group. The scale bar represents 50um. Data are presented as mean ± SD. P values are determined using two-tailed Student’s t test. K Ex vivo imaging of murine tumor tissues of the sgNans and control group. n = 5 for each group. Data are presented as mean ± SD. P values are determined using two-tailed Student’s t test. L Comparion of tumor volume and weights between sgNans and control group. n = 5 for each group. Data are presented as mean ± SD. P values are determined using two-tailed Student’s t test. Source data are provided as a Source Data file.