Identification of potential pathway mediation targets in Toll-like receptor signaling

Abstract

Recent advances in reconstruction and analytical methods for signaling networks have spurred the development of large-scale models that incorporate fully functional and biologically relevant features. An extended reconstruction of the human Toll-like receptor signaling network is presented herein. This reconstruction contains an extensive complement of kinases, phosphatases, and other associated proteins that mediate the signaling cascade along with a delineation of their associated chemical reactions. A computational framework based on the methods of large-scale convex analysis was developed and applied to this network to characterize input-output relationships. The input-output relationships enabled significant modularization of the network into ten pathways. The analysis identified potential candidates for inhibitory mediation of TLR signaling with respect to their specificity and potency. Subsequently, we were able to identify eight novel inhibition targets through constraint-based modeling methods. The results of this study are expected to yield meaningful avenues for further research in the task of mediating the Toll-like receptor signaling network and its effects.

Figure 1. ihsTLR v1.0 reconstruction process.

(A) Flowchart illustrating the necessary steps to convert the Kitano-TLR map into a stoichiometric, mass-balanced model that can be functionally characterized using COBRA method and FBA. Using these computational tools, it was possible to determine a set of critical network reactions that are highly-specific candidates for TLR signaling mediation as changes in their activity attenuate the flux through their corresponding discrete input–output signaling (DIOS) pathways but have no adverse effect on the TLR network reactions. (B) The transfer of functional groups, such as phosphate groups, is very common in signaling pathways. We accounted for proteins explicitly in the corresponding network reaction. This created cycles that are artifacts of the modeling and decouple the phosphorylation/dephosphorylation reactions. Panel B illustrates such a case. The dephosphorylation of PK* to PK can run completely independently of the phosphorylation reaction of AP to AP* since AP is recovered in a subsequent step. The downstream signaling output is thus not dependent on the presence of PK*. (C) This panel illustrates how we circumvented this issue during the modeling by creating a sink reaction for AP* and thus interrupting the cycle formerly present. Since the modeling is only qualitative, the simulation result (e.g., signal yes/no) is not affected by this trick.

Figure 2. The input–output (I/O) relationships of the TLR network at the ligand-receptor-output level.

There were a total of 49 ligands (see Table S3 for complete list), 14 receptors, and 6 outputs. Because the ligands are already well characterized with respect to their receptor specificity, it is unnecessary to carry out the input–output analysis at the level of the external ligand. Rather, the inputs can simply be considered to be signals from the receptors—this reduces the number of inputs from forty-nine to a mere fourteen. NF-kappa-B, CRE/AP-1, and ROS production were all highly redundant targets as almost all of the receptor inputs activated these outputs. IRF3 and IRF7 were much less redundant and were only activated in the case of a small subset of receptor inputs.

Figure 3. Node connectivity in ihsTLR v1.0.

The rank-ordered results were separated for metabolic and non-metabolic species. The non-metabolic species include: ligands, receptors, signaling proteins, outputs (see also Table 1). The three most highly connected species were ATP, ADP, and H+, which participated in 57, 57, and 68 reactions, respectively. In contrast, no non-metabolic species participated in more than 24 reactions. The node connectivity distribution of metabolic and non-metabolic species followed a power law distribution. The fact that the higher connectivities were associated with metabolites illustrates the importance of mass- and charge- balanced network reconstructions for biological accuracy.

Figure 4. An overview of the discrete signaling (DIOS) pathways defined in the TLR network.

There were a total of ten pathways that signaled from input receptor signals to output transcription-level objectives. These ten pathways shared fourteen receptor signals and five output objectives. The most redundant objective was NF-kappa-B activation, which was the target for a majority of the pathways. Indeed, four of the pathways—RIP1, NOD1, NOD2, and RIP2/TRIP6/TRAF2—signaled only to NF-kappa-B. However, also note that IL-1 and a large subset of the TLRs signaled to multiple objectives through a variety of pathways such as PI3K, IL-1, and MyD88. Overall, this receptor-pathway-output format allowed for a better understanding of the TLR network and its input–output relationships, and also for the calculation of essential reactions as candidates for signaling mediation. Red: A summary of the eight critical network reactions identified through our analysis (see text). These control points were located within the ROS production, IL-1, and MyD88 pathways. Although some essential network reactions were identified for the other discrete signaling pathways, they were unsuitable for selective inhibition due either to their role in other signaling processes or their lack of specificity to a particular pathway.

Figure 5. A simplified illustration of the reactive oxygen species (ROS) production DIOS pathway.

The three critical network reactions are highlighted in red. Although there were over forty reactions in the ROS production pathway, most were associated with the TLR-induced activation of various phox proteins by the protein kinases PDK1 and PKCz. However, because PDK1 and PKCz work in parallel, none of these reactions could control the flux through the entire pathway, and therefore were not critical network reactions. On the other hand, the three critical network reactions Vav1-Rac1/GDP binding, Rac1 phosphorylation, and gp91-p22 binding, produced the two other components that comprised the final phox protein complex, and were therefore critical to the overall output ROS production. Note also that these critical network reactions were localized to the ROS production pathway and did not interfere with other cellular processes. Thus, they represent ideal targets for mediation of TLR-induced ROS production.

Figure 6. A simplified illustration of the IL-1 DIOS pathway.

There were three critical network reactions that controlled the IL-1 induced activation of NF-kappa-B. Uninhibited IL-1 signaling induced formation of a TRAF6/Ajuba/PKCz/SQST1 complex followed by autophosphorylation at the Thr-560 residue of PKCz. This activated complex then signaled downstream to NF-kappa-B via IKK phosphorylation. The three critical network reactions inhibited IL-1 induced NF-kappa-B activation by preventing the formation and subsequent autophosphorylation of the TRAF6/Ajuba/PKCz/SQST1 complex. Unlike inhibitors such as IL-1R2 and soluble IL-1R, which mediate IL-1 signaling by preventing the activation of the IL-1 receptors, the three critical network reactions worked by disrupting other components of the IL-1 DIOS pathway and did not affect the activation of IL-1 receptors.

Figure 7. A simplified depiction of the MyD88 DIOS pathway.

The two critical network reactions MyD88 dimerization and TRAF6/IRAK1 ubiquitination are highlighted in red. Formation of the MyD88 homodimer favors recruitment of IRAK1 into a complex with TRAF6 . The MyD88 dimer then dissociated from this complex to be either degraded or reused. The second critical network reaction, TRAF6/IRAK1 ubiquitination, occurred via the ubiquitin-conjugating enzymes Ubc13 and Uev1A, and was necessary for activation of NF-kappa-B and AP-1 through canonical IKK phosphorylation. Either of the two critical network reactions could completely abrogate the flux through the MyD88 pathway even though the TIR- or TIRAP-dependent TLR signaling was almost always active.