Molecular Mechanisms of the Toll-Like Receptor, STING, MAVS, Inflammasome, and Interferon Pathways

ABSTRACT Pattern recognition receptors (PRRs) form the front line of defense against pathogens. Many of the molecular mechanisms that facilitate PRR signaling have been characterized in detail, which is critical for the development of accurate PRR pathway models at the molecular interaction level. These models could support the development of therapeutics for numerous diseases, including sepsis and COVID-19. This review describes the molecular mechanisms of the principal signaling interactions of the Toll-like receptor, STING, MAVS, and inflammasome pathways. A detailed molecular mechanism network is included as Data Set S1 in the supplemental material.

The network includes transcription factor-target gene relationships for cytokines and PRR pathway proteins. These could support the modeling of transcription and translation and enable longer (in silico time) pathway simulations. These transcription factor-target gene relationships were constructed using data from two online databases. CytReg (downloaded 2 June 2021; http://cytreg.bu.edu) (23) was used to model transcription factor-mediated expression of cytokines. RegNetwork (downloaded 2 June 2021; http:// www.regnetworkweb.org) (24) was used to model transcription factor-mediated expression of PRR pathway proteins (except for cytokines). Proteins were not included if they were outside the PRR pathways (e.g., metabolic enzyme expression affected by the mitogen-activated protein kinase [MAPK]-activated transcription factors).

TLR PATHWAY
TLRs are a class of dimeric PRRs which sense a variety of extracellular and endosomal PAMPs and DAMPs and initiate signal transduction resulting in transcriptional FIG 2 The PRR pathway network. The PRR network (see Data Set S1) is depicted. The network includes both the human and the mouse interactions. Two subnetworks of this network (the TLR4 pathway and the cytosolic PRR pathways) are shown in Fig. S1 and S2. reprogramming or programmed cell death (25)(26)(27)(28). The vertebrate TLR family of genes contains at least 27 members, of which 13 are found in the mammalian TLR family (29).
As mentioned above, NEMO binds to M1-polyUb. This brings IKKa/b into close proximity with TAK1. TAK1 phosphorylates and activates IKKa and IKKb (55,56). Activation of IKKa/b is completed via intracomplex trans-autophosphorylation (55,56). In addition to activating the NF-k B pathway, IKKb also activates MAPK signaling via TPL2 (see Data Set S1 in the supplemental material for details), resulting in broad transcriptional reprogramming (58).
NF-k B pathway. NF-k B is a class of five dimeric transcription factors: NF-k B1, NFk B2, RelA, RelB, and c-Rel (54,59,62,63). Unlike the three Rel forms, NF-k B1 and NFk B2 lack a carboxy-terminal transactivation domain, and therefore, they cannot directly initiate transcription. Instead, the full-length forms of NF-k B1 and NF-k B2 (termed p105 and p100, respectively) each contain a carboxy-terminal ankyrin repeat domain (ARD). p105 processing into p50 is constitutive, whereas p100 processing into p52 is induced by its phosphorylation by IKKa (63). There are 28 NF-k B dimeric forms (Data Set S1).
The Ik B family consists of nine members: p105, p100, Ik Ba, Ik Bb, Ik B« , Ik Bh , Ik Bz , Ik BNS, and BCL3 (59,63). The Ik B family members are all characterized by their ARD. The NF-k B family (including both forms of NF-k B1 and NF-k B2) all contain an aminoterminal Rel homology region (RHR) (54, 59, 62, 63). The Ik B ARD regions can bind to NF-k B dimers (at the NF-k B RHR sequences) (54,63). This results in exactly 200 potential trimeric complexes (Data Set S1; not every combination is necessarily chemically possible). The cytosolic Ik B forms (p105, p100, Ik Ba, Ik Bb, and Ik B« ) inhibit NFk B by preventing its localization to the nucleus, whereas the nuclear Ik B forms (Ik Bh , Ik Bz , Ik BNS, and BCL3) can function by activating or inhibiting transcription (54,62,63). Phosphorylation of the classical Ik B forms (Ik Ba, Ik Bb, and Ik B« ) by either IKKa or IKKb results in dissociation of the Ik B from the NF-k B RHR (54,62,63). This enables nuclear import of the NF-k B and K48 polyubiquitination and degradation of the Ik B.
As described above, a key outcome of TLR activation is the reprogramming of gene expression (another TLR-mediated effect is programmed cell death, described below). Notably, TLR activation upregulates the production and secretion of cytokines (26,64). For example, NF-k B upregulates the expression of interleukin-1 (IL-1), IL-2, IL-6, IL-8, IL-12, and TNF, which upregulate proliferation, inflammation, and angiogenesis (76). TLR activation also upregulates the production and secretion of interferons, which can act in an autocrine or paracrine manner, resulting in the transcriptional reprogramming of a wide range of genes with diverse effects (74,77).
Programmed cell death. Successful human pathogens often express virulence factors to antagonize the host immune response (78). For example, Yersinia pestis can enter host cells and secrete YopJ into the host cell cytosol which inhibits TAK1, IKKs, and MKKs, thereby blocking the NF-k B and MAPK pathways (78,79). To prevent the host cell from becoming a means to immune evasion and/or pathogen proliferation, the TLR pathway activates receptor-interacting serine/threonine-protein kinase 1 (RIPK1), which will cause programmed cell death if it is not deactivated by the NF-k B and MAPK pathways.
For example, NLRP9b activation by double-stranded RNA requires DHX9, and NLRC4 activation by bacterial proteins (flagellin, and the rod and needle from type III secretion systems) requires NAIP. Finally, some ligands activate their receptor using a mechanism other than receptor ligation (e.g., Bacillus anthracis lethal factor cleaves NLRP1B).

DISCUSSION
The molecular mechanisms utilized by the TLR and cytosolic PRR pathways are highly diverse. Very characteristic are the oligomeric SMOCs and the branching chains of ubiquitin. It remains unclear what benefit, if any, is derived from the utilization of such large and complex molecular structures in the struggle against pathogens and their virulence factors. Modeling can be used to address these questions, but accurate modeling of these pathways remains challenging.
Molecular interaction networks of the TLR pathway have been published previously (7)(8)(9)(10)(11)(12)(13)(14), and this review includes a molecular mechanism network of the TLR and cytosolic PRR pathways (see Data Set S1 in the supplemental material). Despite having significant scope and detail, none of these networks are comprehensive, primarily because so many significant pathway components remain obscure. For example, although numerous alternative splice isoform sequences of TLR pathway proteins have been sequenced, many of their functions remain unclear. This is unfortunate because it is likely that many of these isoforms have important physiological roles. For example, it is known that a truncated isoform of MyD88 is dominant negative and plays an important role in preventing chronic inflammation (109). Other unresolved issues include fully determining the structure-function relationship of branched polyubiquitin and identifying the phosphatases that act on TLR pathway phosphoproteins.
Developing PRR pathway models at molecular interaction level for performing pathway simulations will aid the development of therapeutics for diseases related to the innate immune system (e.g., sepsis), but developing these models will require more than just molecular mechanism networks. Critically, the molecular reaction rates (k on , k off , and k cat ) and reactant concentrations are required. Targeted proteomics has been used to measure protein concentrations to support the development of numerous signaling pathway models (110). However, measuring the corresponding reaction rates remains challenging, but they can be predicted using bimolecular simulations (111,112). These two sets of values can be used as input parameters for pathway modeling and simulation (22,113). Importantly, pathway models can be trained using, for example, microscopy data, which is how we trained our model of the mouse macrophage chemotaxis signaling pathway (114). Although the development of PRR pathway models remains a formidable challenge, they will be necessary for accurately predicting the behavior of these pathways, especially for pathways stimulated by a diverse population of microbes and PAMPs.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.