True toll-like receptors are absent in Placozoa but TLR-like recognition may be organized differently
TLRs are transmembrane (TM) receptors that comprise extracellular leucine-rich-repeats (LRRs), capable of selectively binding foreign ligands like lipoproteins, lipopolysaccharides, peptidoglycans, ssRNA, dsRNA, (unmethylated) CpG-DNA or flagellin, depending on the receptor subtype [21]. Upon binding of the ligand, the cytoplasmic Toll/interleukin-1 receptor (TIR) homology domain associates with the TIR-domain of the adapter protein MyD88, which itself recruits an Interleukin-1 receptor-associated kinase (IRAK) via the association of both of the protein’s Death domains (DDs) [27]. Branches of downstream effectors terminate with the activation of the transcription factors NF-κB, AP-1 or Interferon regulatory factor (IRF), leading to the expression of inflammatory cytokines and other immune genes [21].
To determine if TLR signaling is present in Placozoa, we screened the predicted proteins of the Trichoplax sp. H2 genome for TLR pathway orthologs via BLAST and KEGG-, EggNOG- and OrthoMCL-mapping, and further validated putative orthologs by phylogenetic analyses and the presence of the necessary domains using the HMM based domain classification of InterProScan 5 [28]. The latter was also used to identify and characterize TIR-domain containing predictions. In addition, modeling with HHpred [29] was used for certain candidate genes as a more sensitive approach to verify the presence or absence of domains that are otherwise difficult to detect (e.g. because of a higher divergence from the consensus in placozoans).
We found 16 proteins containing a TIR-domain (Additional file 1: Figure S1). Eight of these belong to a class of “evolutionary conserved TIR-domain containing” (ecTIR-DC) proteins that appear to be absent in vertebrates but are widespread among invertebrates and for which a function as adaptors or regulators in immunity-related signaling has been hypothesized [30]. Among these are the only two placozoan TIR-domain proteins that also contain LRRs. Both gene products (classified as ecTIR-DC 14 II in [30]) are 3000 amino acids (AAs) large, lack a TM domain and contain several other domains, making any functional prediction speculative.
While these analyses detected no canonical TLRs in Trichoplax sp. H2, we further examined if recognition via TLRs may be similarly organized as in Hydra, where two separate membrane proteins appear to serve the same function: one containing the extracellular LRRs, the other containing the intracellular MyD88 recruiting TIR-domain [2, 31]. We found nine proteins with extracellular LRRs and a predicted single TM, another potential candidate additionally contained immunoglobulin-like domains (Additional file 1: Table S1). Modeling with HHpred revealed no further intra- or extracellular folds or domains, potentially pointing to a different function, and in two proteins the predicted secondary structure was found to be most similar to that of bilaterian TLRs. For the role of a cytoplasmic transducer we found no TIR-only prediction with a TM. Instead, we found a possible candidate in a SEFIR-domain containing transmembrane protein (Additional file 1: Figure S1). SEFIR-domains are related to TIR-domains and are also thought to be involved in homotypic interactions with other TIR/SEFIR-containing proteins [32]. However, SEFIR-domains are also typical for interleukin-17 (IL-17) receptors which are further characterized by extracellular fibronectin-III-like domains. In the case of the Trichoplax SEFIR-domain containing transmembrane protein, neither fibronectin-domains nor other immunoglobulin-like folds were detected by InterProScan and Blast searches did not reveal the slightest resemblance to IL-17 receptors. On the contrary, HHpred detected similarity to IL-17 receptors (and weaker similarity to a TLR) in the region of the SEFIR domain. A reported single hit of a fibronectin-domain to the extracellular part was far beyond significance (see Methods) but homology may nevertheless be considered suggestive, given the presence of the SEFIR domain. The relation of this placozoan gene to other known genes, as well as its function, thus remains difficult to deduce. For the potential role of MyD88 we identified two MyD88-like proteins containing a TIR-domain and the DD-related caspase-recruitment domain (CARD) and another MyD88-like protein containing a SEFIR-domain and a DD (Additional file 1: Figure S1). The latter domain composition resembles both, that of CIKS/CIKSL adaptor proteins involved in IL-17 signaling, and that of MyD88 [33]. HHpred modeling found significant similarity to MyD88 in the region of the DD and to interleukin-17 receptors in the region of the SEFIR domain, which, again, complicates speculations about the protein’s function. However, it is also possible that some of the placozoan proteins that have been found to contain only TIR-domains, or those that belong to the widespread ecTIR-DC proteins, serve as alternative signaling adaptors to MyD88 [21, 30, 34].
We can thus summarize that placozoans possess a variety of TIR/SEFIR-domain containing proteins and putative LRR-receptors that may act in the context of TLR recognition and signaling. A situation resembling that of Hydra’s bipartite TLRs. However, although Hydra’s atypical TLRs are also supported by functional assays [31], we have to note that their proposed TIR-domain containing transmembrane proteins (HyTRR-1, HyTRR-2) have been found to likely adopt an extracellular immunoglobulin-like fold and thus seem rather related to interleukin-1 receptors [30]. If this contradicts a role in TLR recognition or simply emphasizes a derived status of hydrozoans within Cnidaria (anthozoans possess true TLRs [2]), has yet to be resolved. In the case of placozoans, the recognition arm of TLR signaling remains speculative since true TLR orthologs are absent and functional data are missing. It is also possible that other, yet unknown, receptors provide the input for TLR signaling.
The presence of TLR pathway orthologs indicates primordial TLR signaling
We identified many orthologs of the downstream effectors of the canonical TLR pathway (Fig. 1; Additional file 1: Table S2; Additional file 2: Dataset S1; Additional file 3: Dataset S2), indicating the presence of components for primordial TLR signaling in Placozoa. There are, however, important exceptions: While the pathway leading to AP-1 via mitogen-activated protein kinases (MAPK) seems complete, the path to NF-κB lacks the NF-kappa-B inhibitor alpha (IκBα), which, after phosphorylation by inhibitor of nuclear factor kappa-B kinases (IKKs), enables translocation of NF-κB to the nucleus (reviewed in [35]). However, the only conserved domains present in IκBα and related proteins are ankyrin-repeats (ANK) and the detection of putative placozoan homologs may thus be impeded if they deviated substantially from the bilaterian consensus. It is also possible that an unrelated protein among the roughly 200 ANK containing proteins in Trichoplax sp. H2 fulfills this function. Furthermore, the here identified placozoan NF-κB-like protein (p100-subunit-like) deserves some attention. While InterProScan identified only ANK repeats, the analysis with HHpred supports complete Rel homology and Rel homology dimerisation domains (RHD, IPT). Interestingly, two adjacent predictions in the genome contain well recognizable RHDs only, which adds up to three clustered members of the NF-κB/Rel family in placozoans, but otherwise these two appear to have no relation to known members of this gene family. The lack of a DD in the NF-κB- p100-like protein further indicates that the function and regulation of NF-κB complexes in Placozoa deviates from canonical pathways.
Since sponges and Cnidaria do possess complete NF-κB (in current NCBI databases, cf. [36, 37]), the question if placozoans have a disrupted NF-κB or represent a stage before acquisition of the DD depends on the actual phylogeny at the base of the Metazoa, which at this point in time is controversial (e.g. see [38]). However, if Placozoa are basal in the phylogeny, then the evolutionary scenario would be that they represent the stage before acquisition of the DD. If on the other hand, Placozoa were intermediate to sponges and Cnidaria, then the inference would be a disrupted NF-κB.
The absence of TRIF and TBK1 orthologs further indicates that Placozoa lack the MyD88-independent pathway, consistent with the hypothesis of its emergence in chordates [39], but the most important deviation from the reference pathway is the unclear orthology of complete IRAK family genes. While we found several IRAK-related kinases via BLAST, KEGG- or EggNOG-mapping, all of them lack a DD which is thought to be responsible for association with MyD88 [27]. Many of these IRAK related placozoan protein kinases lie in a tight genomic cluster and thus seem to be placozoan specific paralogs produced by gene duplication. However, while Porifera possess reasonable IRAK orthologs with a DD (e.g. A. queenslandica: XP_011406220.2) this seems not to be the case for Cnidaria (c.f. [37, 40]) where all IRAK-like genes lack a DD, at least the closest cnidarian hits for bilaterian IRAKs in current databases such as UniProt, NCBI or Compagen. Nevertheless, it is generally assumed that TLR signaling is present in Cnidaria and functional analyses in Hydra have shown that several downstream targets of TLR signaling are affected after silencing of MyD88 [41], indicating that the pathway is functional even in the absence of a clear IRAK ortholog and the same could be true for placozoans. In the absence of IRAK, other kinases containing a DD could serve similar functions. In the predicted gene models of Trichoplax sp. H2 we were, however, only able to identify one kinase containing a Death-like domain (DLD).
Trichoplax sp. H2 contains a rich and unique scavenger receptor repertoire
Scavenger receptors (SR) are another class of cell surface receptors whose functional diversity has only recently been recognized. They are able to bind a large variety of ligands and their primary role seems to be the clearance from unwanted compounds like cellular debris, pathogens or other foreign particles by endocytosis [25]. SRs constitute a structurally heterogeneous group, but most receptors associated with scavenger receptor activity are characterized by the presence of either C-type lectin domains (CTLD), scavenger receptor cysteine-rich domains (SRCR) or the CD36 domain. Generally, the SR repertoire seems to be greater in invertebrates, likely because of the absence of an adaptive immune system [3, 42, 43]. Their capacity to manage internalization of extracellular ligands by endocytosis makes them suitable candidates not only for detecting and fending off invasive pathogens but also to mediate symbiosis with intracellular bacteria.
We found a large repertoire of transmembrane proteins containing SRCR domains or CTLDs: Together with a single CD36 ortholog, the Trichoplax sp. H2 genome harbors 82 putative scavenger receptors, 45 containing SRCR domains, 36 CTLDs (Fig. 2; Additional file 4: Dataset S3A). Like in Cnidaria [43], no overlap between SRCR domain and CTLD domain containing proteins was observed. As expected for invertebrates, the number of scavenger receptors containing the above domains exceeds by far that of mammals (e.g. 11 in human) and also that of most Cnidaria so far investigated (12–38), with the exception of Aiptasia pallida (72) where it is more or less the same [43]. However, the diversity of domain composition in placozoan SRs is greater than in the latter two animal groups. While the receptors containing either SRCR, CTLD or CD36 are composed of 6 different domains in mammals and 13 in Cnidaria [25, 43], more than 20 different domains are present in placozoan SRs and most combinations appear restricted to the phylum. Accordingly, most scavenger receptors found in Trichoplax sp. H2 cannot be assigned to well recognized mammalian classes, with the exception of Class B, E and I.
The majority of CTLD containing scavenger receptors in placozoans are G protein–coupled receptors
The SR complement of placozoans also differs in another unique aspect as 28 of 36 CTLD containing receptors bear seven transmembrane helices instead of a single transmembrane domain and are G protein–coupled receptors (GPCRs) of the Secretin/Adhesion family. To our knowledge these are the only described GPCRs whose extracellular domains consist of CTLDs. The one exception is Branchiostoma which harbors one adhesion GPCR containing an N-terminal CTLD, albeit together with other extracellular domains [44]. This suggests that CTLD mediated recognition of molecular patterns in Placozoa is linked to GPCR signaling, trafficking and endocytosis [45]. Furthermore, β-arrestin mediated endocytosis also couples GPCRs to the MAPK pathway. Hence, binding and internalization of targets could trigger a variety of cellular events, including defense or management of symbiotic microorganisms. In this context it is noteworthy that we found a large repertoire of arrestins (defined by the presence of N- and C-terminal Arrestin-like domains) in the Trichoplax sp. H2 genome (Additional file 4: Dataset S3B): in addition to one ortholog of β-arrestin, twenty three α-arrestins, related to the mammalian genes ARRDC2–4, are present, seventeen of which reside in a tight cluster on scaffold 76. Compared to mammals (two β-arrestins, two visual arrestins, six α-arrestins [46]), Nematostella (UniProt: one β-arrestin, three α-arrestins) and Amphimedon (UniProt: one β-arrestin, six α-arrestins) this gene number constitutes a unique expansion of the arrestin family in Placozoa. Even though the role of α-arrestins in GPCR signaling is less well understood, they are also known to interact with GPCRs in concert with β-arrestin [46]. This observation indicates that GPCR mediated endocytosis in placozoans plays an important role for defense in particular and GPCR signaling in general.
Secreted proteins containing fibrinogen-related domains appear to play an important role in extracellular defense
Before pathogens come into contact with cell-surface receptors, the line of first defense consists of secreted lectin-like proteins that are able to recognize carbohydrates associated with bacterial cell walls and either agglutinate pathogens, or coat and opsonise them to enhance phagocytosis. Possible candidates, among others, are secreted proteins that contain CTLDs [47] or fibrinogen-related domains (FReDs) [48]. Twelve predictions without a TM were identified harboring 1–3 CTLDs, eight of which contain a signal peptide (SP) and are thus likely to be secreted factors (Additional file 4: Dataset S3C). Fibrinogen-related domains on the other hand were previously claimed to be absent in placozoans [48]. However, using the InterProScan annotation we were able to detect 18 FReD containing gene models with SUPERFAMILY member SSF56496 (Fibrinogen, C-terminal globular domain; Additional file 4: Dataset S3C). While 2 of these are large proteins related to fibrillin, the remainder are around 250 AA in length, bear a single FReD and an SP (except one) and are mostly organized in three clusters in the genome. The PANTHER classification system [49] classifies them as intelectins, a novel type of soluble lectins (X-type lectin), which seem remotely related to ficolins and have been shown to selectively recognize bacterial glycans [50]. Moreover, an additional 15 gene models (11 with SP) are classified as intelectins by PANTHER and all of these, except 2, are genomic neighbors of the above, producing 5 clusters of 2–9 genes each. In 28 of these 31 genes, modeling with HHpred further supports the presence of a FReD (the remaining 3 predictions are probably truncated). This observation indicates that most of the putative intelectins belong to the same gene family but contain FReDs that are difficult to detect as a result of their phylogenetic distance to related sequences, which are the basis of canonical domain models. This could also be the reason why the identified FReDs in placozoan intelectins are comparatively short compared to vertebrate intelectins (up to 70 AAs versus 200 AAs).
Vertebrate and placozoan intelectins thus only align well in the region of the shorter placozoan FReD (restricted to the N-terminal portion of placozoan intelectins) while the latter also show high sequence similarity in the C-terminal portion (Additional file 5: Figure S2; Additional file 6: Dataset S4). Most placozoan sequences share 10 conserved cysteine residues (4–5 with vertebrate intelectins; complete alignment in Additional file 6: Dataset S4), suggesting that they are essential to form either intrachain or interchain disulfide bonds. The latter could serve to form intelectin oligomers, as is the case in human Intelectin-1 [50], which probably increases agglutination of trapped bacteria. While mammals possess up to 6 intelectins, teleosts have 9 and tunicates 22 intelectins [51], there are up to 31 intelectins in Trichoplax sp. H2. The apparent expansion of this gene family in placozoans suggests that intelectins play an important role in extracellular defense of the animal.
Some lectins, like the FReD containing ficolins and the CTLD containing mannose-binding protein (MBP), are known to initiate the lectin-complement-pathway [26] which represents an immediate response system for recognition and defense. In the cnidarian Aiptasia it has also been shown that the complement system is involved in the management of the cnidarian-dinoflagellate symbiosis [52]. It is thus tempting to speculate if the placozoan intelectins or secreted CTLD proteins could serve a similar function. However, intelectins, as well as the placozoan secreted CTLD proteins, lack collagen repeats or coiled-coils that could mediate oligomerization similar to ficolins or MBP, which is needed to trigger the lectin-pathway. Moreover, placozoans apparently lack a complement system at all since most of its factors are absent. For example, neither KEGG pathway mapping nor screening for the necessary domains revealed genes for the key factors Mannan-binding lectin serine protease (MASP) or the complement factors B, C2, C3 or C4. With respect to the factors C3/C4, the only members of the Alpha-2-macroglobulin family present in Trichoplax sp. H2 are two CD109 genes which are not involved in the complement system. We only identified a single C1q-like factor (Additional file 4: Dataset S3C) bearing a clear C1q domain, and collagen repeats that are only recognizable by structural homology prediction with HHpred. In vertebrates, C1q plays an important function in the classical complement pathway [26]. In invertebrates, which lack adaptive immunity, C1q factors are assumed to rather act as lectins in immune recognition and have undergone expansion in several taxa [53]. The latter, apparently, does not apply to Placozoa.
Intracellular defense in Placozoa does not conform to the classical NOD-like receptor pathway and RIG-I-like receptors or cGAS-STING signaling are absent
NLRs mediate recognition once pathogens or their components have entered the cytoplasm [22, 54]. These proteins contain a central NOD domain (NACHT or NB-ARC) for nucleotide binding and self-oligomerization and C-terminal LRRs for molecular pattern recognition. An N-terminal effector domain mediates protein-protein interactions with downstream targets. Effector domains are members of the Death clan like PYD, CARD or the Death Effector Domain (DED). In the canonical NOD pathway, the receptors oligomerize after binding of PAMPs and expose their CARDs, enabling interaction with the CARD containing serine-threonine kinase RIPK2 which activates NF-κB and MAPK signaling [22]. Other NLRs contain tetratricopeptide (TPR), WD40 or ANK repeats instead of LRRs and seem not involved in PAMP recognition (e.g. [54, 55]). RIG-I-like receptors detect viral RNA and are composed of a central DEAD/DEAH box helicase domain, an N-terminal CARD and a C-terminal regulatory domain [23]. A further mechanism for detecting foreign cytosolic DNA is the cGAS-STING pathway [24].
While genes for cGAS, STING or RIG-I-like receptors were not found in Trichoplax sp. H2, 42 predicted genes containing a central NACHT/NB-ARC-domain (NACHT 3, NB-ARC 39) were identified (Fig. 3; Additional file 4: Dataset S3D). None of these contain LRRs and are thus unlikely to be involved in recognition of invading pathogens. In addition, no RIPK1/2 ortholog could be found in the Trichoplax sp. H2 genome, which is similar to the situation in cnidarians [55, 56]. The C-terminal repeats of placozoan NLRs are always either WD40 or TPR type repeats. Effector domains are predominantly CARDs or Death-like domains (DLDs) and in two cases DDs. These domain combinations of placozoan NLRs mostly resembles that of Hydra, which also does not contain any NOD-like receptors with LRRs, in contrast to anthozoans [54, 55].
Apaf-1 and controlled suicide as a means of defense
The NOD-like repertoire of placozoans also differs remarkably in another aspect to bilaterians and cnidarians: while most animals possess only one homolog of the Apoptotic protease activating factor-1 (Apaf-1), placozoans harbor several paralogs in their genome. If we consider DLDs as difficult to detect putative CARDs, Trichoplax sp. H2 harbors twelve Apaf-1 homologs, possibly even more because several Apaf-1 genes reside in clusters of NB-ARC containing genes but some gene models are missing either WD40 repeats or a CARD/DLD (Fig. 3; Additional file 4: Dataset S3D).
Apaf-1 plays a key role for the initiation of the mitochondrial pathway of apoptosis [57]. It forms an oligomeric apoptosome and ignites the action of downstream caspases. The trigger is the released danger signal cytochrome c resulting from mitochondrial damage, a possible result of an infection. Apaf-1 thus enables controlled suicide of impaired cells. The death signal is passed from Apaf-1 to Caspase-9 via the association of both of the protein’s CARDs. Although H2 seems not to possess a clear Caspase-9 ortholog, it harbors two neighboring genes containing a caspase-like domain and an N-terminal CARD (Additional file 4: Dataset S3D), suggesting a similar function in apoptosis.
The presence of many Apaf-1 paralogs in placozoans suggests that controlled cell death after damage (e.g. by infection) plays a fundamental role in maintaining somatic and genetic integrity of the animal. Rather than triggering inflammatory responses, the cell is sacrificed once pathogens have escaped defense mechanisms and captured the cell. There are two reasons this strategy would make sense. First, the remarkable regenerative capacity of placozoans would be a contributing factor to how this hypothetical mechanism works and second, placozoans do not possess any organs whose accurate function has to be secured at all costs. Even the sacrifice of the entire animal seems plausible, if we assume that a local population mainly reproduces asexually and consists of genetically identical individuals. The selective advantage for a local genotype could thus lie in a rapid suicide of single infected animals to prevent spreading a disease to the entire population.