br B br Figure Constituents and Functions of the
Figure 2. Constituents and Functions of the Core Cancer Secretome
(A) A heat-scatterplot presenting the log2FCs and corresponding significance (false discovery rate [FDR]-adjusted p values) for the 16 genes making up the top 1% of the non-directional core secretome. The color and size of the points correspond to the log2FC and log-transformed p values, respectively, from the DE analysis between tumor and paired-normal samples.
(B) The top 1% of the increased core secretome, obtained in the same manner as the non-directional set in (A), except the fold change direction was incorporated to identify secretome genes exhibiting increased Lovastatin across many cancer types.
(C) Gene sets found to be significantly enriched in the decreased (left column), non-directional (center column), or increased core secretome (right column), in which the top 20 most significant sets from each directional class are shown. The intensity of the color in the heatmap indicates the enrichment significance of the gene set. Gene set names are colored according to the Molecular Signatures Database (MSigDB) collection from which they originate: Hallmark, Kyoto Ency-clopedia of Genes and Genomes (KEGG), Reactome, GO biological process, and GO molecular function. A non-stacked bar plot to the left of the heatmap shows the sizes (number of genes) of the original gene sets (gray bars) and of the filtered gene sets containing only secretome genes (black bars).
See also Figure S2.
functions attributed to the MMP family, MMP11 is somewhat unique in that it is secreted in its active form and its ECM sub-strates differ from those commonly targeted by MMPs (Pei et al., 1994). MMP11 has been reported to enable tumor invasion by inducing de-differentiation of surrounding adipocytes and supporting the accumulation of peritumoral fibroblasts (Andara-wewa et al., 2005).
To investigate core secretome genes that exhibited pan-can-cer expression increases, the gene-ranking process was repeated, except that the direction of expression fold change was incorporated instead of using the absolute log2FC values. The set of 16 secretome genes with the highest directional PF ranks (top 1%) across the different cancer types exhibited a lower degree of coordination compared to the non-directional
set (Figure 2B). Regarding function, the majority of the core increased secretome genes were involved in the structure and composition (e.g., COL1A1, ACAN, ZP3) (Iozzo and Schaefer, 2015; Pickup et al., 2014; Rankin and Dean, 2000) or modifica-tion (e.g., metalloprotease MMPs and a disintegrin and metallo-proteinase with thrombospondin motifs [ADAM(TS)]) (Egeblad and Werb, 2002) of the ECM. Another function shared by many of the proteins was signaling, either as receptors or effectors. For example, EFNA4, NXPH4, and GPC2 facilitate signaling associated with neuronal and developmental events, which sup-ports essential tumor functions such as angiogenesis, cell adhe-sion, and motility (Kurosawa et al., 2001; Missler and Sudhof,€ 1998; Wilkinson, 2001). Other proteins with signaling-related functions included CTHRC1 and C1QTNF6, which are involved in vascular remodeling (Park et al., 2013; Takeuchi et al., 2011), and SPP1, which is known to facilitate cell-matrix interac-tions (Shevde and Samant, 2014). Overall, core secretome shifts contribute to diverse malignant processes, particularly those relating to ECM remodeling, or to a reduction in tumor-suppres-sive activity.
Enrichment of Functions in the Core Cancer Secretome Although analysis of the top-ranked core secretome genes offered insight into common functions that were downregulated (or upregulated) across the different cancer types, it excludes in-formation about the remaining 99% of secretome members. We therefore conducted a gene set analysis (GSA) to account for the PF ranks of all of the secretome genes in determining coordi-nated shifts in secretome function. The GSA was performed us-ing both non-directional and directional PF ranks.