Proteomic Heterogeneity of the Extracellular Matrix Identifies Histologic Subtype-Specific Fibroblast in Gastric Cancer

Abstract

Gastric cancer (GC) is a highly heterogeneous disease regarding histologic features, genotypes, and molecular phenotypes. Here, we investigate extracellular matrix (ECM)-centric analysis, examining its association with histologic subtypes and patient prognosis in human GC. We performed quantitative proteomic analysis of decellularized GC tissues that characterizes tumorous ECM, highlighting proteomic heterogeneity in ECM components. We identified 20 tumor-enriched proteins including four glycoproteins, serpin family H member 1 (SERPINH1), annexin family (ANXA3/4/5/13), S100A family (S100A6/8/9), MMP14, and other matrisome-associated proteins. In addition, histopathological characteristics of GC reveals differential expression in ECM composition, with the poorly cohesive carcinoma-not otherwise specified (PCC-NOS) subtype being distinctly demarcated from other histologic subtypes. Integrating ECM proteomics with single-cell RNA sequencing, we identified crucial molecular markers in the PCC-NOS-specific stroma. PCC-NOS-enriched matrisome proteins and gene expression signatures of adipogenic cancer-associated fibroblasts (CAF_adi) are closely linked, both associated with adverse outcomes in GC. Using tumor microarray analysis, we confirmed the CAF_adi surface marker, ATP binding cassette subfamily A member 8 (ABCA8), predominantly present in PCC-NOS tumors. Our ECM-focused analysis paves the way for studies to determine their utility as biomarkers for patient stratification, offering valuable insights for linking molecular and histologic features in GC.

Keywords: gastric cancer; poorly cohesive carcinoma; proteomics; tumor microenvironment.

Graphical abstract

Fig. 1

Characterization of patient-derived decellularized ECM.A, scheme of the workflow. Proteomic profiles of patient-derived tissue ECM (pdECM) were identified by tandem mass tag (TMT) mass spectrometry. B, clinical data of patients and samples. A histology, a stage of tumor, an anatomical region, and a depth of tumor are shown. C, representative image of native tissue and pdECM. The scale bar represents 100 μm. D, DNA quantification of native tissue and pdECM. E, average intensity of detected proteins in equally mixed reference samples. The counts of proteins within each category are shown in each parenthesis. F, the cellular component term of gene ontology of 100 proteins which have the highest intensities were analyzed. The bar graph shows the counts of proteins and the dots shows the statistical significance of each category. ECM, extracellular matrix; pdECM, patient-derived tissue ECM.

Fig. 2

Matrisome-focused proteomic profiles of patient-derived normal and tumor ECM.A, matrisome composition of each pdECM indicated by hierarchical clustered heatmap and bar plot. ∗p < 0.05. B, composition of proteins ranked by relative percent composition (RPC). The counts of proteins that cover 90% of total intensities are shown by each group. C, most abundant 20 matrisome proteins in each group. Bar plot shows the average RPC of each protein in each tissue type. RPC of each protein of each sample are shown as dots in the bar plot. D, PCA plot with matrisome protein intensities of all samples. PCA, principal component analysis; pdECM, patient-derived tissue ECM.

Fig. 3

Differentially expressed matrisome proteins (DEPs) between adjacent normal and tumor ECM.A, volcano plot of differentially expressed matrisome proteins between adj. normal and tumor ECM. Dotted lines show thresholds for DEPs. (|log₂FC| > 0.5 and p-value <0.05). B, representative immunohistochemistry (IHC) image of SERPINH1 and HAPLN1 which are tumor-enriched and normal-enriched matrisome each. C, paired analysis of DEPs. Gray dots show the log₂FC of each patient and red dots show the median values. Dots in red box show the patients with the opposite tendency. ECM, extracellular matrix; FC, fold change; HAPLN1, hyaluronan and proteoglycan link protein 1; SERPINH1, serpin family H member 1.

Fig. 4

PCC-NOS specific ECM profile.A, PCA plot of tumor ECM with matrisome protein expression profile. B, box plot of proteoglycan expression in PCC-NOS type and non-PCC-NOS type. C, volcano plot of differentially expressed matrisome proteins (DEPs) between PCC-NOS type and non-PCC-NOS type. Dotted lines show thresholds for DEPs. (log₂FC > 0.5 and p-value <0.05). D, heatmap of DEPs between PCC-NOS type and non-PCC-NOS type. PCC-NOS enriched matrisome proteins were defined as PEMs. E, paired analysis of selected PEMs. The changes of log₂ (intensity) value are shown by histology. ECM, extracellular matrix; PCA, principal component analysis; PCC-NOS, poorly cohesive carcinoma-not otherwise specified; PEM, PCC-NOS-enriched matrisome protein.

Fig. 5

Identifying PEM mainly expressing cells.A, defining cellular origin of proteins with single-cell RNA sequencing. The cellular origin of proteins was identified with two criteria. (1) an expression of gene that encoding the protein is cell type specific. (2) among the cell types, the average expression of the gene is highest in the cellular origin. The cellular origins of PEMs were identified from the data of Kumar et al. (12). B, lollipop plot of 20 most correlative genes with the PEM score. Lines show the Pearson’s correlation coefficient and circles show the proportion of positive fibroblasts expressing the genes. C, dot plot of single-cell level gene expression of 20 most correlative genes with the PEM score. D, the gene expression of 20 most correlative genes in the isolated fibroblasts. E, the gene expression of 20 most correlative genes and ABCA8 on pan-cancer single cell analysis from the data of Luo, et al. (13) Most of them shows the enriched expression on the adipogenic CAFs (CAF_adi) region. F, scatter plot of the correlation between CAF_adi score and PEM score. TCGA STAD data (n = 375) was used for the scoring. G and H, survival analysis with PEM score and 30 CAF_adi markers. The mean expression of selected genes was used to separate the high and low expression of patient group. ABCA8, ATP binding cassette subfamily A member 8; CAF_adi, adipogenic cancer-associated fibroblast; PEM, PCC-NOS-enriched matrisome protein; STAD, stomach adenocarcinoma; TCGA, the Cancer Genome Atlas.

Fig. 6

ABCA8 positive fibroblasts are specific to PCC-NOS subtype and predict poor prognosis.A, the bar plot shows the correlation of the histological subtype and ABCA8 grade in tumor microarray IHC analysis. B, the representative image of ABCA8 positive fibroblasts in PCC-NOS subtype and ABCA8 negative tissue in tubular adenocarcinoma. The scale bar represents 100 μm. C, survival analysis with three groups of patients based on stromal expression of ABCA8 protein. Diffuse-ABCA8+ groups showed poorer survival rate compared with the other diffuse type group and the other histology type group. D, survival analysis with ABCA8 gene expression. ABCA8 high expressing group showed poorer prognosis. PCC-NOS, poorly cohesive carcinoma-not otherwise specified; IHC, immunohistochemistry; ABCA8, ATP binding cassette subfamily A member 8.