NetSlim: high-confidence curated signaling maps

Criteria for NetSlim

In generating NetSlim, we applied the following inclusion criteria to extract a high confidence set of reactions from pathways in NetPath: (i) reactions supported by at least two research articles; (ii) reactions shown to occur in multiple cell types (iii) PPIs reported through high-throughput analyses, as well as induced PTMs identified using mass spectrometry are included if they are supported by any other experimental method; (iv) components of well established protein complexes even if they do not meet other criteria; and (v) molecules or reactions described in expert review articles. Schematic representation of the NetSlim criteria is provided in Figure 1. Reactions that fulfilled one or more of the above criteria were included in NetSlim.

Graphical representation of NetSlim

The signaling pathway maps of NetSlim were generated using the PathVisio tool which facilitates editing and display of biological pathways (13). PathVisio is an improved visualization tool which incorporates all the features of GenMAPP (14). All molecular reactions that occur downstream of activated receptors are represented by solid edges, which depict direct interactions/reactions, or by dashed edges, which depict indirect/induced events. Signaling nodes are depicted based on the ability of molecules to interact with their partners, and on the availability of protein regulators for cross-talk with other signaling modules. The topology of the pathway, which determines whether a given molecule lies upstream or downstream from another reference molecule, is generally dependent on the biological information obtained from studies using mutants or inhibitors of specific molecules. The edges representing various PPIs, enzymatic reactions or protein translocation events are distinguished from each other by distinct colors. NetSlim maps with detailed legends are available at http://www.netpath.org/netslim. Pathway maps with nodes hyperlinked to specific NetPath molecule page and the edges linked to their corresponding citations can be obtained by clicking on ‘map with citation’ option in the NetSlim pathway pages. The browsing features and the information provided in NetSlim are shown in Figure 2 with TGF beta pathway as an example.

NetSlim map of the TGF-β pathway as a model

The signal transduction mechanisms underlying the pathophysiological activities of transforming growth factor-β (TGF-β) have been extensively studied since its discovery nearly 30 years ago. Mammals express three TGF-β isoforms (i.e. TGF-β1, TGF-β2 and TGF-β3) that are encoded by distinct genes in a tissue-specific and developmentally regulated manner (15). In addition, numerous studies have clearly established TGF-β as a multifunctional cytokine that plays essential roles in regulating virtually all aspects of mammalian development and differentiation, and in maintaining mammalian tissue homeostasis (15–18). Even more remarkably, malignant cells have evolved a variety of complex mechanisms capable of circumventing the tumor suppressing activities of TGF-β, and in doing so, typically convert the functions of TGF-β to that of a tumor promoter, particularly the induction of carcinoma epithelial–mesenchymal transition, invasion, and dissemination to distant organ sites (15–18). This peculiar conversion in TGF-β function is known as the ‘TGF-β Paradox’, which underlies the lethality of TGF-β in metastatic cancer cells. Thus, elucidating the effectors and signaling modules activated by TGF-β may offer new insights into the development of novel neoadjuvants capable of effectively targeting the TGF-β pathway to significantly improve the clinical course of patients with cancer, fibrosis or immunologic disorders.

TGF-β is secreted from cells as a latent homodimeric polypeptide that becomes tethered to the extracellular matrix by latent-TGF-β-binding proteins. Mature TGF-β isoforms are activated and liberated from extracellular matrix depots by a variety of mechanisms, including proteolysis, reactive oxygen species, changes in pH and physical interactions with integrins, thrombospondin-1 or SPARC (19–22). Once activated, mature TGF-β initiates transmembrane signaling by binding to two distinct transmembrane Ser/Thr protein kinases, termed TGF-β type I (TβR-I) and type II (TβR-II) receptors (23–25). In some cells and tissues, TGF-β also binds to a third cell surface receptor, TGF-β type III (TβR-III), which transfers TGF-β to TβR-II and TβR-I (26). Full activation of these cytokine:receptor ternary complexes transpires upon TβR-II-mediated transphosphorylation and activation of TβR-I (27, 28), which then phosphorylates and activates the latent transcription factors, Smad2 and Smad3 (29, 30). Afterward, phosphorylated Smad2/3 interact physically with Smad4, with the resulting heterotrimers translocating into the nucleus to regulate the expression of TGF-β-responsive genes (29, 30). These Smad-dependent events are subject to fine-tuning and crosstalk regulation in the cytoplasm by their interaction with a variety of adapter molecules, including SARA, Hgs, PML and Dab2 (31–34), and with Smad7 (35, 36), whose inhibitory activity is modulated by STRAP, AMSH2 and Arkadia (37–39); and in the nucleus by their interaction with a variety of transcriptional activators and repressors that occur in a gene- and cell-specific manner (40). In addition to activating canonical Smad2/3-dependent signaling, accumulating evidence clearly links the development of a variety of human pathologies to aberrant coupling of TGF-β to its noncanonical effector molecules. Included in this ever expanding list of non-canonical signaling molecules stimulated by TGF-β are PI3K (41), AKT (42), mTOR (43), integrins (44) and focal adhesion kinase (45), and members of the MAP kinase [e.g. ERK1/2 (46), JNK (47) and p38 MAPK (48)] small GTP-binding proteins (e.g. Ras, Rho and Rac1) (18).

The TGF-β-signaling pathway constructed in NetPath currently contains 226 proteins that participate in 482 reactions comprised of 340 molecular association events, 105 enzymatic events and 37 translocation events. Graphical representation of these molecules in their entirety is quite cumbersome, thus making their meaningful interpretation difficult for biologists users of the NetPath resource. Further many NetPath molecules and their reactions are supported by either single publications or derived from high-throughput experiments lacking secondary experimental validation. While the entire network is useful, especially to gain coverage in gene expression analyses, users often want to see only the high confidence regions of a pathway and this is currently difficult to construct. We have now involved experts from the field to contribute their knowledge to NetSlim with regard to the inclusion or exclusion of molecules in high confidence datasets. For instance, the presented NetSlim model of TGF-β pathway possesses 121 proteins involved in 252 reactions with 173 molecular associations, 60 enzyme catalysis reactions and 19 translocation events, almost half the size of the full NetPath pathway. Review articles contributed to the inclusion of ∼30% of the reactions in TGF beta NetSlim pathway map. There were many reactions which were derived from high-throughput experiments but were not included because no further validation was conducted.

As such, this TGF-β NetSlim model represents the ‘core TGF-β pathway’ parsed from the less stringent and all inclusive resource, NetPath. Despite its trimmed appearance, our NetSlim model still presents more molecules and reactions as compared to major publicly available pathway resources containing the same pathway, such as Kyoto Encyclopedia of Genes and Genomes (KEGG) (49), National Cancer Institute’s Pathway Interaction Database (PID) (50), Science's Signal Transduction Knowledge Environment (STKE) Connections Maps (51), Reactome (52) or PANTHER (53). Some of the pathway resources developed by commercial entities have been customized to better reflect their available product line and reagents, thus presenting a skewed and distorted view of the underlying signal transduction pathways. Our new high confidence map of the TGF-β signaling system is shown in Figure 3. The NetSlim criteria have greatly diminished the complexity of the current pathway map and we hope that they will prove useful to the community.

Impact of NetSlim criteria

The NetSlim criteria used to extract high confidence reactions from NetPath have considerably reduced the complexity of the pathway maps. However, as mentioned above, these pathway maps contain more molecules and reactions than other pathway resources available in the public domain. We have captured reactions described in the review articles while many others are still not included in reviews of the respective pathways in spite of being high confidence. The reactions identified in multiple cell types could be captured from single or multiple research publications and contribute to the list of high-confidence reactions. A comparison between the total number of molecules and the reactions in NetSlim and NetPath is provided in Figure 4.