Structural aspects of nucleic acid-sensing Toll-like receptors

Introduction

The innate immune system is the first line of defense against pathogenic microorganisms in humans, and it rapidly responds to foreign invasion via various downstream responses. Microbial components are recognized by different sets of cell-surface and intracellular sensors, such as Toll-like receptors (TLRs), Nod-like receptors (NLRs), and RIG-I-like receptors (RLRs). Activation of these receptors triggers signaling cascades that induce the production of pro-inflammatory cytokines and interferons, leading to the activation of inflammatory responses, anti-viral responses, and cell-mediated immune responses. Nucleic acids from bacteria, viruses, and fungi in the form of single-stranded (ss) or double-stranded (ds) DNA and RNA, as well as their derivatives, have been extensively investigated as stimulators of the innate immune system. Several types of nucleic acid sensors have been identified, including members of the TLR, NLR, RLR, and cyclic GMP-AMP synthase protein families (Barbalat et al. 2011; Paludan 2015; Sparrer and Gack 2015). In this review, we summarize recent scientific progress on the recognition and regulation mechanisms of nucleic acid-sensing TLRs (TLR3, TLR8, TLR9, TLR13) based on data obtained from structural biology studies (Fig. 1).

TLR families and TLR signaling

To date, ten TLRs (TLR1–TLR10) and 12 TLRs (TLR1–TLR9 and TLR11–TLR13) have been identified in humans and mice, respectively. Each TLR recognizes a specific ligand known as a pathogen-associated molecular pattern, such as a specific lipopeptide, lipopolysaccharide, flagellin, or nucleic acid (Kawai and Akira 2010). Vertebrate TLRs can be classified into six major families: TLR1 (TLR1, TLR2, TLR6, TLR10), TLR3, TLR4, TLR5, TLR7 (TLR7, TLR8, TLR9), and TLR11 (TLR11, TLR12, TLR13) (Roach et al. 2005). The TLR1 and TLR11 families also encompass TLR14 and TLR21–TR23, respectively. Members of the TLR1, TLR4, and TLR5 families are expressed on the cell surface, while TLR3, TLR7, and TLR11 family members are expressed on endosomal membranes (Miggin and O'Neill 2006). TLRs are type I transmembrane receptors and consist of extracellular leucine-rich repeats (LRRs), a transmembrane domain, and a cytoplasmic TIR domain (Bell et al. 2003; Matsushima et al. 2007). The LRR domain is responsible for ligand recognition, while the cytoplasmic TIR domain initiates downstream signaling cascades by interacting with their adaptor proteins, such as MyD88, Mal, TRIF, and TRAM (Kawai and Akira 2010). The binding of a ligand to its specific TLR leads to activation of transcription factors, such as NF-κB, mitogen-activated protein kinases, IRF3, and IRF7. Based on crystallographic studies of TLR extracellular domains (Jin et al. 2007; Kang et al. 2009; Liu et al. 2008; Park et al. 2009; Yoon et al. 2012), the following TLR activation mechanism has been proposed: the free/inactive form of TLR exists as a monomer, and upon binding of its ligand, dimerization of TLR occurs and then TLR is activated. TLR dimers are typically ''m'' shaped, with the C-terminal regions of the two TLR protomers positioned in close proximity (Song and Lee 2012). Dimerization of the extracellular domain regions induced by ligand binding promote subsequent dimerization of the intracellular TIR domain, which is followed by recruitment of adaptor proteins that facilitate signal transduction.

Nucleic acid-sensing TLRs

Viral/bacterial nucleic acid is a potent stimulant of innate immunity. Members of the TLR3 and TLR7 families are nucleic acid-sensing TLRs. TLR3 is activated by dsRNA, TLR7/8 detects ssRNA, and TLR9 senses DNA that contains unmethylated cytosine–phosphate–guanosine (CpG) motifs. In addition, TLR13 in the TLR11 family detects bacterial 23S ribosomal RNA (Li and Chen 2012; Oldenburg et al. 2012), while TLR11 and TLR12 form heterodimers that recognize the protein profilin in Toxoplasma gondii (Andrade et al. 2013). Nucleic acid-sensing TLRs are localized within intracellular compartments such as endosomes and lysosomes, which allows them to distinguish between host and foreign nucleic acids.

TLR7 and TLR8 are closely related TLRs. Both recognize pathogen-derived ssRNA (Diebold et al. 2004; Heil et al. 2004; Lund et al. 2004), and it has suggested that they are also involved in the pathogenesis of autoimmune diseases via recognition of self-derived nucleic acids (Barrat et al. 2005). TLR7 is expressed primarily on B cells and plasmacytoid dendritic cells (DCs), whereas TLR8 is abundantly expressed on monocytes, myeloid DCs, and neutrophils. Murine TLR8 appears to be non-functional as TLR7-deficient mice are unable to elicit normal responses when stimulated with the TLR7/8 agonist imidazoquinoline and ssRNA (Heil et al. 2004; Jurk et al. 2002). However, several recent studies have demonstrated that TLR8 is able to function as an innate immune receptor (Cervantes et al. 2012). Unmethylated CpG sites on DNA have been shown to activate the immune system (Krieg et al. 1995). TLR9 has also been demonstrated to recognize unmethylated CpG containing DNA (Bauer et al. 2001; Hemmi et al. 2000).

In mammals, the CpG motif is usually methylated at the cytosine base (Smith and Meissner 2013) and does not induce immune activation. However, in bacteria, CpG sites are unmethylated and do induce immune-reactivity (Bauer et al. 2001; Krieg et al. 1995). TLR9 can therefore discriminate between self and non-self DNA via the methylation states at CpG sites.

In addition to the TLR3 and TLR7 families, mice have another endosomal TLR family which is entirely absent in humans, namely, the TLR11 family (TLR11, TLR12, and TLR13). TLR13 has recently been shown to be a receptor of the conserved ssRNA sequence CGGAAAGACC, which is derived from bacterial 23S rRNA (Li and Chen 2012; Oldenburg et al. 2012).

Members of the TLR7 family possess a long inserted loop region (otherwise known as the Z-loop) composed of approximately 40 amino acid residues. Accumulating evidence suggests that proteolytic processing of the Z-loop in the endolysosomes is required for the generation of functional, mature receptors (Ewald et al. 2008, 2011; Ishii et al. 2014; Park et al. 2008; Sepulveda et al. 2009). This cleavage requirement is assumed to prevent unwanted receptor activation by self-nucleic acids at other locations within the cell. The cysteine lysosomal protease, asparaginyl endopeptidase, and the pH-dependent endosomal protease, cathepsin, both participate in Z-loop cleavage of TLR9 and TLR7 (Ewald et al. 2011; Maschalidi et al. 2012; Park et al. 2008; Sepulveda et al. 2009).

Human TLR7 (hTLR7) processing has recently been shown to be mediated by the furin-like proprotein convertase (Hipp et al. 2013). In addition, the cleaved form of TLR8 has been found to be predominantly expressed in immune cells (Ishii et al. 2014). Although there is much debate regarding the association between the N- and C-terminal fragments of cleaved TLRs, recent studies have revealed that following hTLR7 proteolytic processing, the liberated N-terminal fragment remains bound to the C-terminus (Hipp et al. 2015; Onji et al. 2013). Consistent with this finding, recent structural studies demonstrate that after Z-loop cleavage, the N- and C-terminal halves of TLR8 and TLR9 associate with each other and that both fragments are involved in ligand binding (Ohto et al. 2015; Tanji et al. 2013).

Recognition mechanism of dsRNA by TLR3

Human TLR3 is the first TLR for which the three-dimensional structure was determined. The structure of the unliganded TLR3 was independently reported by two groups (Bell et al. 2005; Choe et al. 2005). The unliganded TLR3 is a monomer with a typical horseshoe-shaped structure consisting of 23 LRRs with N- and C-terminal cap domains (Fig. 1a). The first ligand-bound structure of nucleic acid-sensing TLRs was determined in mouse TLR3, which was complexed with a 46-bp dsRNA (Fig. 1b) (Liu et al. 2008).

Each TLR3 extracellular domain is bound to a dsRNA to form an ''m'' shaped dimer. The 46-bp dsRNA is located in center of the signaling complex and takes on an A-form DNA-like structure. As it is sandwiched between the two protomers of TLR3, it comes into extensive contact with the TLR3 domains. The 46-bp dsRNA spans approximately 130 Å, which just fits within the length of the overall complex (approx. 140 Å). Leonard et al. (2008) proposed that the minimum dsRNA length required for the formation of a stable signaling complex with TLR3 is approximately 40–50 bp, which correlates with results from structural studies. With longer dsRNA ligands, TLR3 forms a multivalent complex where multiple TLR3 molecules bind to a single dsRNA molecule (Leonard et al. 2008). Each TLR3 protomer interacts with the dsRNA at two distinct sites—one close to the C-terminus (CT; LRR19–LRR21) and the other close to the N-terminus (NT; LRR-NT–LRR3) (Figs. 2a and 3a).

As shown in the side views in Fig. 1b, the TLR protomers bend towards each other at one end to accommodate the dsRNA molecule. As a result, the two TLR3 molecules are only able to establish protein–protein interactions at their C-termini (Fig. 2a), which are approximately 25 Å apart. The homotypic interactions between the two protomers are essential for TLR3 activation, as confirmed by cell-based assays (Wang et al. 2010). TLR3 recognizes dsRNA mainly through electrostatic interactions at the ribose–phosphate backbone. The N-terminus of the dsRNA binding site on TLR3 contains several His residues which associate with the phosphate groups of dsRNA, and some of these interactions have been demonstrated to be essential for TLR3 activation. As protonation of these His residues occurs under acidic environments, successful binding of dsRNA to TLR3 requires acidic pH (Bell et al. 2005; Leonard et al. 2008).

The affinity of the 48-bp dsRNA for TLR3 varies greatly between pH 6.5 (Kd = 2,610 nM) and pH 5.5 (Kd = 14 nM). Similar to TLR7, TLR3 has also been reported to be proteolytically processed, which modulates its responsiveness to ligands. This process occurs in the inserted loop region at the LRR12 domain (Garcia-Cattaneo et al. 2012; Qi et al. 2012) and is mediated by cathepsins B and H. Interestingly, the loop region at which TLR3 is cleaved protrudes into the ascending lateral face in the middle region of the LRR domain and does not associate with either the dsRNA or the other protomer. Therefore, the structural significance of the proteolytic processing of TLR3 remains elusive.

Recognition of the chemical ligand by TLR8

While TLR7 and TLR8 generally recognize ssRNA, they can also be activated by small chemical ligands, such as imidazoquinolines and nucleoside analogs (Hemmi et al. 2002; Jurk et al. 2002). Therefore, these small molecules have attracted much attention due to their potential as therapeutics for various infectious diseases.

Zhu et al. (2009) reported the preformed dimer of full-length TLR8 in a cell-based assay. Consistently, the extracellular domain of hTLR8 alone consistently forms preformed dimer both in solution and in crystal structures (Tanji et al. 2013). TLR8 protein expressed in S2 cells was already cleaved at the Z-loop by an unknown protease, but the N- and C-terminal halves were still associated with each other. TLR8 consists of 26 LRRs which form a ring structure stabilized by several interactions. For example, the LRR-NT and LRR-CT of TLR8 interact directly with each other to strengthen the ring. LRR14 and LRR15 also form continuous β-sheets in the concave surface of the LRR domain. Furthermore, following proteolytic cleavage, the Z-loop at the C-terminal end forms ordered structures on the concave surface of TLR8, leading to extensive hydrophobic interactions with N-terminal half of TLR8.

While both unbound (Fig. 1c) and ligand-bound (Fig. 1d) TLR8s display an ''m''-shaped dimer structure, there are significant structural differences between these two forms of TLR8. The distances between the two LRR-CTs in the dimer are 50 and 30 Å in the unbound (inactive) and bound (active) forms, respectively. Upon binding to its ligand, the interactions between the two protomers of the TLR8 dimer are completely altered due to structural rearrangements, which then allow recruitment of the intracellular TIR domain for signal transduction.

Chemical ligands bind to TLR8 at two equivalent positions on the dimer, forming a symmetrical 2:2 complex. The ligand is positioned at the TLR8 dimer interface, between the N-terminal half of one protomer and the C-terminal half of the other protomer (Figs. 2b; 3b). Therefore, it also stabilizes the activated form of TLR8. Prior to structural studies of TLRs, it was not known whether the N-terminal fragment in the TLR7 family was required for ligand recognition following Z-loop processing. However, structural observations clearly demonstrated that both N- and C-terminal fragments of TLR8 are necessary for ligand recognition.

Recognition of ssRNA by TLR8

For a long time TLR8 was thought to be an ssRNA receptor, as cells that express TLR8 respond to ssRNA (Diebold et al. 2004; Heil et al. 2004; Lund et al. 2004). The possibility that the same receptor can be activated by both chemical ligands and ssRNAs was an intriguing research question, as these molecules differ drastically in size and chemical property. Subsequent structural studies of hTLR8–ssRNA complexes have unraveled this mystery (Tanji et al. 2015).

It was discovered that TLR8 is not simply a ssRNA receptor, but also an uridine receptor. The crystal structure of TLR8 and 20-mer ssRNA complexes revealed that rather than binding directly to the full-length ssRNA, TLR8 binds to RNA degradation products at two distinct sites (Fig. 1e). Electron density for uridine mononucleoside was observed at the first site, while that of a dinucleotide (UG) was observed at the second site (Figs. 2c; 3c). The first site is identical in structure to chemical ligand binding sites and accommodates the ssRNA degradation product uridine. The second binding site is located at the interior of the TLR8 ring structure and is sandwiched between the concave surface and the Z-loop to accommodate ssRNAs that are longer than 2 mer. The overall structure of the TLR8–ssRNA complex is essentially identical to that of the chemical ligand-bound form of TLR8, indicating that binding of uridine and dinucleotide to TLR8 induces the activated form of the TLR8 dimer. As the dinucleotide binding site is located outside of the TLR8 dimerization interface, binding at that site alone would not induce the activation of TLR8.

These observations from structural analyses are further confirmed by biophysical and biochemical analyses. Among the mononucleosides (uridine, guanosine, cytidine, thymidine, adenosine), uridine shows the highest affinity for TLR8 with a Kd value of 55 μM, as determined by isothermal titration calorimetry (ITC) analyses. Moreover, neither 5′-UMP nor 3′-UMP can bind to TLR8, suggesting that uridine must be in its mono-nucleoside form for binding to occur (Tanji et al. 2015). This notion is well supported by structural analyses, as the incorporation of phosphate groups at either the 5′ or 3′ position of uridine would clash with TLR8 molecules. Although TLR8 shows a preference for uridine, the affinity of uridine for TLR8 alone (Kd of 55 μM) is still weaker than its affinity for chemical ligands (e.g. Kd of 0.2 μM for R848). This difference is reconciled by synergistic associations that take place in the presence of ssRNA. It was observed that ssRNA binding at the second binding site potentiates the affinity of uridine for TLR8 by 50-fold (Kd of 1.0 μM). Among all mononucleosides, only uridine can prominently activate TLR8 when co-stimulated with ssRNA (Tanji et al. 2015). These results suggest that oligonucleotide binding at the second binding site promotes uridine binding at the first site, possibly via allosteric regulation.

Recognition of chemical ligands by TLR8 is mediated by three types of interactions: stacking interactions between TLR8 and aromatic moiety of ligands, hydrogen bonds, and interactions between the alkyl substituent on the ligands and hydrophobic pockets formed between the TLR8 protomers. The former two interactions are conserved in uridine recognition. The lack of the hydrophobic interaction between uridine and TLR8 may account for the weak affinity of uridine for TLR8; consequently, TLR8 activation requires further potentiation from ssRNA binding. Structural analyses are providing a better understanding of the interactions between TLR8 and chemical ligands (Tanji et al. 2013), which ultimately will enable further structural modifications and optimizations for TLR7 and TLR8 chemical ligands (Beesu et al. 2015; Ganapathi et al. 2015; Kokatla et al. 2014; Yoo et al. 2014).

Recognition of CpG DNA by TLR9

TLR9 is a member of the TLR7 family and has the same number of LRRs as TLR7/8 in addition to the Z-loop. Crystal structures of TLR9 take on three forms: unliganded, bound to CpG-DNA, and bound to inhibitory DNA (iDNA). These structures serve as a basis for understanding the signaling and functional mechanisms of TLR9 (Ohto et al. 2015).

Unlike TLR8, the extracellular domain of TLR9 protein expressed in S2 cells was not cleaved at the Z-loop. Therefore, the Z-loop of TLR9 was enzymatically cleaved by V8-protease treatment, and the resulting structure was biochemically analyzed. Gel-filtration and ultracentrifugation analyses indicated that dimerization of the cleaved TLR9 was induced upon binding to CpG-DNA. However, although TLR9 with an intact Z-loop was able to bind to CpG-DNA, dimerization was not induced. These results suggest that Z-loop processing of TLR9 does not regulate DNA binding but that it is required for CpG-DNA-dependent dimerization of TLR9 (Ohto et al. 2015).

The unliganded horse TLR9 is a monomer in both crystal and solution and thus represents the inactivated form. The ring-shaped monomer structure of TLR9 is similar to that of TLR8 (Tanji et al. 2013). Upon ligand binding, the TLR9 and CpG-DNA complex forms a 2:2 complex with the CpG-DNA wedged between the two TLR9 protomers in an extended conformation (Fig. 1f). The CpG-DNA acts as a stabilizer of this specific dimer structure. TLR9 dimers represent the activated form of the receptor as the two C-termini are in close proximity with each other (approx. 30 Å). This structure has clearly demonstrated that CpG-DNA induces TLR9 dimerization, leading to its activation. The TLR9–CpG-DNA interface is divided into two parts, denoted as interface 1 in TLR9 and interface 2 in TLR9* (Figs. 2d; 3d). The most important structural feature of TLR9 for ligand recognition is the presence of the CpG binding groove on interface 1, which is formed at the N-terminus of TLR9 (LRR-NT, LRR1, LRR2). The bases of the CpG motif, cytosine and guanosine, form multiple interactions with TLR9 within the binding groove. The flanking regions of the CpG dinucleotide also contribute to binding interactions. In contrast to interface 1, where many base–protein interactions take place, interface 2 mainly recognizes the backbone phosphates of CpG-DNA. The importance of these interfaces has been confirmed by NF-κB reporter assay.

The importance of unmethylated CpG dinucleotide was examined through ITC experiments by converting DNA molecules containing CpG dinucleotides into various sequences, such as methylated CpG. The original unmethylated CG-containing DNA yielded the highest affinity for TLR9 (Kd of 20 nM), while the altered sequences showed reduced affinities. For example, the Kd of methylated CpG-DNA for TLR9 was found to be 50 nM. The methylation of CpG-DNA also reduced TLR9 activity (Bauer et al. 2001; Krieg et al. 1995; Rutz et al. 2004).

Results from structural analyses reveal that the introduction of a methyl group at the fifth position of cytosine does not interfere with the position of TLR9 but that it may disrupt water clusters that mediate interactions between CpG and TLR9, thus resulting in the reduced affinity. Binding of CpG-DNA to TLR9 is pH dependent (Rutz et al. 2004), as demonstrated through partially purified recombinant proteins and ITC experiments (Ohto et al. 2015). As expected, these authors found that the binding affinity was stronger under acid conditions than under basic conditions. Structural studies have also revealed several histidine residues located around the DNA-binding region, which contribute to electrostatic interactions with the phosphate groups of DNA (Ohto et al. 2015), as predicted previously (Peter et al. 2009). Protonation of these His residues at acidic pH accounts for the pH-depend nature of TLR9 activation, similar to TLR3 (Liu et al. 2008).

Crystal analyses of TLR9 complexed to iDNA indicate that iDNA forms a stem-loop structure via intramolecular base pairing and that it occupies the interior of the TLR9 ring structure (Fig. 1h). The iDNA exhibits stronger binding affinity for TLR9 (Kd of 3 nM between iDNA4084 and horse TLR9) than does CpG-DNA. The antagonistic effect of iDNA is attributable to partial overlap in the binding interfaces between iDNA and CpG-DNAs.

Recognition of 23S rRNA by TLR13

The crystal structure of mouse TLR13 complexed with a 13-nt ssRNA (ssRNA13) derived from 23S rRNA has recently been published, revealing the sequences and conformation-specific recognition sites of ssRNA13 by TLR13 (Fig. 1h) (Song et al. 2015).

The unliganded form of TLR13 is a monomer, which dimerizes upon binding to ssRNA13, as revealed by gel-filtration chromatography. Similar to other TLRs, the TLR13–ssRNA13 complex exhibits an ''m''-shaped dimer structure. Unlike TLR8 (Tanji et al. 2013) and TLR9 (Ohto et al. 2015), however, which form ring structures, TLR13 is composed of 27 LRR units arranged into an oval shape. The LRR-NT of TLR13 interacts with LRR25–LRR27 and LRR-CT on the concave surface of the LRR domain. The joint region between the N- and C-terminus provides an interaction surface for ssRNA13 recognition. TLR13 and ssRNA form a 1:1 complex, which further forms a 2:2 complex, such as TLR9 (Figs. 2e; 3e). Single base pairing between G2057 and C2064 promotes the formation of a stem-loop structure in ssRNA13, while most of the interactions of the other bases of the loop region (AAAGAC: A2058 to C2063) with TLR13 are base-specific. A2058, A2060, and G2061 participate in three-layered stacking networks through extensive interactions with LRR22–LRR25 and specific hydrogen bonds with TLR13. The loop region of ssRNA13 is highly conserved among the different bacterial species. As was observed in TLR3 and TLR9, the interaction between TLR13 and ssRNA13 is also pH dependent, with a preference for acidic pH. The stem-loop conformation of ssRNA13 is essential for recognition by TLR13.

Mutations that disrupt base parings (G2057 and C2064) of the ssRNA13 stem region have been found to abolish both TLR13 dimerization and TLR13-induced cytokine production. Notably, the stem-loop structure of ssRNA13 is completely different from that observed in the Escherichia coli ribosome (Dunkle et al. 2010), where ssRNA13 is completely buried in the ribosome and inaccessible to TLR13, suggesting that ssRNA13 needs to be processed or degraded prior to participating in TLR13 activation. TLR13 can distinguish RNA from DNA, and studies have shown that ssDNA containing the same sequence as ssRNA13 is unable to induce TLR13 dimerization, possibly due to the 2'-OH groups in ribonucleic acids, which are involved in the maintenance of inter- and intramolecular hydrogen bonds in TLR13-ssRNA interface. Since recognition of ssRNA13 by TLR13 is independent of ribosomes, any ssRNA that forms ssRNA13-like stem-loop structures can potentially be a ligand for TLR13. This possibility was confirmed in studies showing that vesicular stomatitis virus (VSV) also induces TLR13-mediated immune responses (Shi et al. 2011). Accordingly, following its identification, the ssRNA–VSV16 sequence derived from VSV was shown to induce TLR13 activation and downstream immune responses (Song et al. 2015).

Similarities and differences among nucleic acid-sensing TLRs

Currently, four nucleic acid-sensing TLRs (TLR3 in TLR3 family, TLR8-9 in TLR7 family, TLR13 in TLR11 family) have been structurally characterized. TLR3 is the only TLR identified to date that recognizes double-stranded nucleic acids. Due to the rigidity and size of dsRNA, the recognition mode of dsRNA by TLR3 (Liu et al. 2008) differs drastically from that of other nucleic acid-sensing TLRs (Ohto et al. 2015; Song et al. 2015; Tanji et al. 2015). On the other hand, both the monomer and ligand-induced dimer structures of TLR8 and TLR9 recognizing single stranded nucleic acid are very similar to each other (Ohto et al. 2015; Tanji et al. 2013, 2015). Interestingly, despite the similarity between their overall structures, the ligand binding sites of TLR8 and TLR9 do differ (Figs. 2c, d; 3c, d). TLR8 has two ligand binding sites that synergistically function to respond to ssRNA-derived molecules, uridine, and oligonucleotide. Based on the sequence and functional similarities between TLR7 and TLR8, some structural features of TLR8 should be applicable to TLR7. Recently, a similar mode of receptor activation was found in TLR7; instead of uridine in TLR8, guanosine and its modified derivatives were identified as endogenous ligands for TLR7, and the interactions between TLR7 and guanosine were also enhanced by the presence of oligonucleotide (Shibata et al. 2015).

Ligand specificity between TLR7 and TLR8 has been reported; for example, imiquimod and loxoribine can specifically activate TLR7, but not TLR8 (Lee et al. 2003). The ability to discriminate between guanosine and uridine may be explained by certain structural differences in the ligand binding site between TLR7 and TLR8, and this mechanism should be further investigated in future crystallographic studies. As TLR7 and TLR8 have multi-binding sites that act in synergy, it was expected that TLR9, which belongs to the same family of receptors, also possesses comparable binding mechanisms. In this regard, recent biochemical analyses of TLR9 suggest that aside from the binding site, which is involved in CpG-DNA recognition, a secondary binding site may also be present that induces species-specific responses to CpG-DNA (Pohar et al. 2015a, b).

Results from structural studies strongly suggest that ligand processing is crucial for the TLR recognition in certain receptors (Fig. 4). For example, both TLR7 and TLR8 recognize degradation products derived from RNA. Likewise, TLR13 senses degradation products of 23S rRNA. Moreover, DNA processing is important for TLR9 activation via certain types of CpG-DNA (Chan et al. 2015). Nucleic acid-sensing TLRs are confined to lysosomes, where various hydrolytic enzymes reside, including proteases, lipases, nucleases, glycosidases, phospholipases, phosphatases, and sulfatases. Once the ligand is transported to the lysosome, it is degraded unless otherwise protected. It will be crucial to confirm the processing of these ligands in vivo and to identify the nucleases and phosphatases involved in this process.

Closing remarks

Until recently, structural information on the nucleic acid recognition mechanisms of TLRs has been limited, especially regarding the dsRNA receptor, TLR3. Recent structural elucidations of the TLR7 and TLR13 family members have enabled us to understand how these TLRs function. Moreover, our increasing understanding of these structures has provided remarkable insight into the regulatory mechanisms of nucleic acid sensing by TLRs, including the presence of multiple ligand-binding sites, proteolytic processing of the TLR extracellular domain, and nucleic acid processing, such as degradation. These features could be used as potential targets for therapeutic intervention.

Compliance with ethical standards