A systems biology approach for the investigation of the heparin/heparan sulfate interactome

Abstract

A large body of evidence supports the involvement of heparan sulfate (HS) proteoglycans in physiological processes such as development and diseases including cancer and neurodegenerative disorders. The role of HS emerges from its ability to interact and regulate the activity of a vast number of extracellular proteins including growth factors and extracellular matrix components. A global view on how protein-HS interactions influence the extracellular proteome and, consequently, cell function is currently lacking. Here, we systematically investigate the functional and structural properties that characterize HS-interacting proteins and the network they form. We collected 435 human proteins interacting with HS or the structurally related heparin by integrating literature-derived and affinity proteomics data. We used this data set to identify the topological features that distinguish the heparin/HS-interacting network from the rest of the extracellular proteome and to analyze the enrichment of gene ontology terms, pathways, and domain families in heparin/HS-binding proteins. Our analysis revealed that heparin/HS-binding proteins form a highly interconnected network, which is functionally linked to physiological and pathological processes that are characteristic of higher organisms. Therefore, we then investigated the existence of a correlation between the expansion of domain families characteristic of the heparin/HS interactome and the increase in biological complexity in the metazoan lineage. A strong positive correlation between the expansion of the heparin/HS interactome and biosynthetic machinery and organism complexity emerged. The evolutionary role of HS was reinforced by the presence of a rudimentary HS biosynthetic machinery in a unicellular organism at the root of the metazoan lineage.

FIGURE 1.

Topological and functional analysis of heparin/HS-interacting network. A, the extracellular heparin/HS-interacting network (“Ec_hepint”; blue) was extracted from a data set of extracellular protein-protein interactions, and it was compared with the non-heparin/HS-interacting network (“Ec_not-hepint”; red) and the whole extracellular interactome (“Ec”; green) (see “Experimental Procedures”). The position of the nodes in the networks and the length of edges are arbitrary and only have a graphical purpose. The properties and topological parameters of the network analyzed are summarized in B. “Proteins” indicate the number of proteins (nodes) that form each network, and “PPI” (protein-protein interactions) indicates the number of interactions (edges) connecting them. The “Average degree” indicates the mean number of neighbors per node in the network. The “Characteristic path length” is the average over the shorter distances (number of links) separating all pairs of nodes in the network and offers a measure of the overall navigability of a network. The clustering coefficient is defined as the number of links connecting the first neighbors of a given node divided by the total possible number of connections between them. It is a measure of the tendency of nodes to form highly interconnected modules. The “Avg. clustering coefficient” for a network is calculated as the mean of the clustering coefficients for each node having a degree ≥2. The “Ec_hepint-random” network was generated from the extracellular heparin/HS-interacting network by applying a degree-preserving random shuffle of the edges (1320 shuffles). The “Ec_random” network was generated by randomly selecting a network of the same size as the extracellular heparin/HS-interacting network from the total extracellular interactome. The procedure was iterated 50 times, and mean network parameters are shown with S.D. in parentheses. The average clustering coefficient of the extracellular heparin/HS-interacting network is six standard deviations higher than the average value calculated for randomly picked networks. In C, nodes are binned according to their degree, and the average (Avg.) clustering coefficient for each bin is plotted applying the same color code used in A. The distribution of the clustering coefficients is characterized by the typical slope of protein-protein interaction networks, which indicates the presence of hierarchical modularity.

FIGURE 2.

Examples of highly clustered modules of heparin/HS-interacting network. HBPs with a high clustering coefficient were extracted together with their first neighbors from the extracellular heparin/HS-interacting network. The node color indicates the clustering coefficient of each node according to the legend. The node label indicates the UniProt short name of the HBP. Protein-protein interactions are represented as green edges. The highest clustering coefficients represented were 1.0 for VEGFB and vascular endothelial growth factor receptor-1 (FLT1) (A), 0.67 for fibrillin-2 (FBN2) (B), 0.60 for coagulation factor IX (F9) (C), and 0.40 for TGFβ2 (D). The graphs were generated using Cytoscape v.2.6.3 (32). VWF, von Willebrand factor; CTGF, connective tissue growth factor; ELN, elastin; VTN, vitronectin; APP, amyloid precursor protein. NRP1, neuropilin-1; LTBP-1, latent-transforming growth factor beta-binding protein-1; BMP2, bone morphogenetic protein-2.

FIGURE 3.

GO biological process terms enriched in heparin/HS interactome are shown as nodes connected by directed edges that indicate hierarchies and relationships between terms. The node size is proportional to the number of HBPs belonging to the functional category. The node color indicates the corrected p value (Benjamini-Hochberg false discovery rate correction) for the enrichment of the term according to the legend. For clarity, only highly significant terms are displayed (p < 1e−21). The graphs were generated using Cytoscape v.2.6.3 (32) and its plugin BiNGO v2.3 (42).

FIGURE 4.

Correlation between abundance of superfamilies associated with heparin/HS interactome and organism complexity. The PCCs describing the association between superfamily abundance and organism complexity were extracted from Vogel and Chothia (50). The PCCs for superfamilies enriched in the heparin/HS interactome (“Hepint”), enriched in the heparin/HS interactome and annotated as extracellular in Vogel and Chothia (50) (“Ec_hepint”), annotated as extracellular (“Ec”), and annotated as extracellular but not enriched in the heparin/HS interactome (“Ec_not-hepint”) are plotted along the y axis. For each group, a horizontal bar indicates the mean PCC. The dashed line indicates the threshold for strong correlation between superfamily abundance and organism complexity.

FIGURE 5.

Occurrence of heparin/HS biosynthetic enzymes across tree of life. The occurrence of heparin/HS biosynthetic enzymes across 630 organisms was analyzed using STRING 8.2 (52). The UniProt Accession numbers of the human HS biosynthetic enzymes were used as input. The conservation of each gene across different species is indicated by squares colored according to the sequence homology detected by STRING. For clarity, some species (e.g. bacteria) were grouped in collapsed nodes colored in gray, and the number indicated in parentheses reports the number of species grouped in each node. In these cases, split squares report the highest and lowest score for the given gene within the grouped species.

FIGURE 6.

Choanoflagellate M. brevicollis possesses rudimentary HS biosynthetic machinery. Orthologs of the human HS biosynthetic enzymes were established by applying the reciprocal Blast best hit criterion (43) against the non-redundant protein sequences databases of C. elegans, M. brevicollis, Fungi, D. discoideum, and Plantae. The figure reports the Blastp score of the best hits only in the cases when the reciprocal best hit criterion was satisfied. The “Linker” enzymes are responsible for the synthesis of the protein-GAG linker tetrasaccharide, which is shared between HS and other GAGs. The “HS” group includes enzymes that are specific for the HS-specific biosynthetic pathway that follow the formation of the linker tetrasaccharide. The full report of the Blastp search including M. brevicollis hits is provided in supplemental Table 6.