The evolution of Hox genes in Spiralia

The decoding of genomes of a larger number of animal species have provided further insights into the genomic Hox gene organization and with this indicated the evolutionary changes during the radiation of several clades. The expansion of gene expression studies during development and life history stages of more species, complete the picture of the relationship between cluster organisation and temporal and spatial correlation of the Hox activity. Now these results open the opportunity to look deeper into the regulatory pathways that form these patterns and identify what exact changes caused the evolution of the application of this iconical gene set for the evolution of new larval forms and new structures. Here we review recent progress of Hox gene related research in the large clade Spiralia, that comprises Annelida, Mollusca, Lophophorata, Platyhelminthes, Nemertea and others. Albeit their relationship to each other is not resolved yet, there are emerging patterns that indicate that Hox genes are mainly used for patterning late, adult body parts and that Hox genes are often not expressed on the larval stages. Hox genes seem also often recruited for the formation of morphological novelties. Together with the emerging genomic information Hox genes show a much more dynamic evolutionary history than previously assumed. appears to correlate with the disintegration of the ancestral Hox cluster, the loss of a

3 broader taxonomic span of spiralian lineages, ultimately providing a better resolved picture of Hox gene evolution in this bilaterian clade. Currently, there is transcriptomic and/or genomic data on Hox genes for 11 of the 15 major animal groups that comprise Spiralia ( Figure 1), with data missing for Gnathostomulida, Micrognathozoa, Gastrotricha and Cycliophora. Together, these new datasets clarify the ancestral Hox gene complement of Spiralia-likely comprising Hox1/lab, Hox2/pb, Hox3, Hox4/Dfd, Hox5/Scr, Lox5, Antp, Lox4/Lox2, and at least one posterior Hox-, demonstrating that distinct patterns of central and posterior Hox evolution cooccurred with the phylogenetic split of Spiralia into Gnathifera and Lophotrochozoa ( Figure 1). Gnathifera comprises Chaetognatha, Gnathostomulida, Micrognathozoa, and Rotifera [10,11], but information on Hox complements and their expression only exist for Chaetognatha and Rotifera [12][13][14][15]. Our understanding of Hox complements in chaetognaths-arrow worms-is currently based on transcriptomic data and targeted searches [14], and thus whether Hox genes are organized in a cluster in this group is unknown. Genomic data is available for rotifers however, indicating that a Hox cluster is absent in this group [15], which correlates with a lack of temporal and spatial collinear expression of rotiferan Hox genes [12]. Both chaetognaths and rotifers share the presence of a unique type of Hox gene phylogenetically related to both Medial/Central and Posterior Hox genes in other non-gnathiferan taxa, referred to as MedPost genes, as well as a unique motif in the Hox6/Lox5 group [12,16]. Other than that, Hox gene complements differ significantly between these two lineages, with chaetognaths apparently lacking Hox2 and rotifers missing Hox7/Antp, Hox8/Lox4/Lox2, and posterior genes ( Figure 1). Altogether, the shared idiosyncratic signatures of Hox gene complements observed in chaetognaths and rotifers backs the phylogenetic relationships among gnathiferan clades, suggesting that lineage-specific diversification of Hox complements might underpin phenotypic evolution in Gnathifera [12].  Lophotrochozoa (sensu [10]) is the second major lineage of spiralian taxa, and its internal phylogenetic relationships are still debated [10,11,[17][18][19]. Gastrotricha, Platyhelminthes and Dicyemida have often been related in phylogenomic analyses [19,20]. These are molecularly fast-evolving lineages, which is also reflected in their highly divergent Hox gene complements ( Figure 1). While data for Gastrotricha is absent, genome sequencing of the dicyemid Dicyema japonicum uncovered only 4 Hox genes belonging to three orthogroups:  genomic information and expression data are not available for entoprocts, transcriptomic analyses indicate that all lophotrochozoan Hox orthogroups were present in their last common ancestor [39] ( Figure 1). Therefore, if the Mollusca + Entoprocta association holds stable, a full Hox gene repertoire was likely ancestral for this clade.
In Annelida, the species Capitella teleta, Alitta virens, Platynereis dumerilii [40-42] and Owenia fusiformis (unpublished data) have complete Hox gene complements and exhibit an ordered Hox cluster, except for Post1, which is separate ( Figure 1). This is likely the ancestral annelid condition, and signs of spatial and temporal In summary, the recent genomic characterization of a broader array of spiralian lineages has uncovered two different dynamics of Hox evolution in this group, concomitant to the Gnathifera/Lophotrochozoa split ( Figure   1). The presence of an ordered Hox cluster with 11 orthogroups exhibiting spatial collinearity and temporal collinearity in blocks-as observed in other major bilaterian lineages-is probably ancestral to Lophotrochozoa, yet the poor understanding of the internal phylogenetic relationships of this group makes reconstructing the exact evolutionary history of Hox genes difficult. Interestingly, Lox2 and Post1 have been repeatedly lost during spiralian evolution, which as discussed below might be associated with the cooption of these Hox genes to morphological novelties.  [52]. The level, to which Hox expression has been studied varies a lot from clade to clade and for some spiralians expression of only single orthogroups has been investigated thus far (e.g., for chaetognaths, dicyemids and bryozoans), while for others the expression of full, or almost full Hox complements has been described throughout several developmental stages (e.g., for rotifers, nemerteans, brachiopods, phoronids, annelids and mollusks; Fig. 2). Fortunately, the comprehensively studied species are widely spread across phylogeny, which allows insight into evolution of the Hox function in the morphologically diverse clade of Spiralia.
In  [57,69,70]. The latter function has been also reported in other annelids based on in situ RNA hybridization [71] and regeneration stage-specific transcriptomics [72]. Traces of the A-P staggered expression are also evident in rotifers, where Hox genes are expressed almost exclusively in the developing nervous system [12]. In mollusks, the Hox genes seem to have dual function (  Altogether, comparison of Hox gene expression across Spiralia (Fig. 2) shows that unprecedented morphological diversity of this clade correlates well with evolutionary lability of Hox gene expression and function. In several spiralian clades the Hox genes are expressed not only in ectoderm, but also in the mesoderm and its derivatives.
However, detailed comparison of mesodermal Hox expression shows that the set of Hox genes expressed in the mesodermal derivatives differs markedly from clade to clade and that their temporal transcription in the

Hox genes and spiralian novelties
The cooption of Hox genes into the morphogenesis of particular structures and organs is an important evolutionary mechanism well studied in numerous animals. The Hox genes might be re-wired into patterning of the preexisting organs but they can also contribute to the origin of morphologically and molecularly novel structures, the so-called evolutionary novelties [73,74]. The latter, as evident from the preceding summary of Hox expression studies in annelids revealed that expression of Post1 deviates from the neuroectodermal A-P expression witnessed for other Hox genes and instead the gene is expressed in the developing chaetal sacs [40,41,55], the morphological structures that secrete chaetae, stiff bristles used by annelids for locomotion and protection. Expression of Post1 in the chaetae-related territories has also been reported in brachiopods [7,36] in concert with the morphology-based hypothesis on the homology of chaetae in both clades [77][78][79]. In brachiopods another Hox gene, lab, is also expressed in the chaetal sacs [7,36], however, this cooption seems to be restricted only to the brachiopod lineage. Since chaetae-like structures are also present in some fossil mollusks [80], it seems plausible that chaetae were already present in the last common ancestor of annelids, mollusks and brachiopods and that Post1 was already coopted for the patterning of the chaetal sacs in the lineage leading to this hypothetical ancestral animal. Interestingly, Post1 is missing from the genomes of phoronids [30, 32], which -according to this evolutionary scenario -have secondarily lost the chaetae forming apparatus. This suggests that the gene, and the morphological structure patterned by it, had been lost in unison in the phoronid lineage.
Another possible example of cooption of Hox genes into patterning of evolutionary novelties can be found in the serpulid annelid, Spirobranchus lamarcki. As for other Serpulidae, S. lamarcki possess an operculum -an unpaired head appendage with a biomineralized shield used to plug the tube where the worm dwells [81]. The operculum is considered an evolutionary novelty of serpulids and many species are capable of its regeneration.
Analysis of gene expression during operculum regeneration in S. lamarcki indicates that among many homeotic genes expressed in the regenerating organ, there is also a Hox gene Antp [82]. While the role of Antp in operculum formation is still unclear, these observations highlight the plasticity of Hox gene expression in spiralians, even in those clades exhibiting a marked and conserved spatial collinearity. reduction of morphological characters in certain clades. Most importantly, the functional analyses, using e.g., RNAi or CRISPR gene editing, are still needed to confirm that expression of particular Hox genes in the developing morphological structures is really indispensable for their morphogenesis.

Hox genes and life history stages
Studies of Hox gene expression during development of animals with complex life cycles can also inform about the evolution of life histories [83]. Spiralia are a particularly interesting clade in this respect, since many of them develop through distinct larval types [84,85], some of which represent clade specific innovations (e.g. pilidium of pilidiophoran nemerteans, actinotrocha of phoronids or mitraria of oweniid annelids), while others are highly conserved over long phylogenetic distances (e.g. trochophore present in numerous annelids, kamptozoans and mollusks).
In Hox genes are also not expressed during development of the actinotrocha, the highly specialized phoronid larva [32]. Their expression is delayed until later larval stages with most of the Hox genes being expressed in the rudiment of the adult trunk and in other posterior structures, which contribute to the adult body after metamorphosis (Fig. 2). A similar expression dynamic of Hox genes is observed in the mitraria larva of the palaeoannelid Owenia fusiformis (Martín-Durán, unpublished data), which also undergoes catastrophic metamorphosis [86]. The broad expression of head-specific genes in the larval phoronid indicates that the actinotrocha represents a so-called head-larva [32]. This would explain why the actinotrocha forms without ancestor. Therefore, the future investigation of Hox gene expression during embryonic and larval development of other indirectly developing spiralians could help to resolve whether their larvae represent modification of the ancestral spiralian larva or more recent evolutionary innovations resulting from the intercalation of new headlarvae. This would be especially interesting in the cases of some strange spiralian larvae, whose evolution and homology to the other larval types remain obscure, such as the Müller's larva of polyclads, creeping larva of kamptozoans or cyphonautes of gymnolaemate bryozoans.

Conclusions
A high conservation of Hox clusters in some lineages, together with the dissociation of the cluster and loss of certain genes in some sublineages characterizes the evolution of Hox genes in Spiralia. It remains unclear however, which forces prevent the Hox cluster from atomizing in this animal group, yet genome regulatory aspects have likely played a major role. Despite the conservation of the Hox cluster in major spiralian lineages, neofunctionalization of Hox genes-even of those existing as single copies-is not uncommon. Therefore, Hox genes are often used for axial patterning, but also deployed in novel structures in later stages of development.
Contrary to what is generally observed in arthropods and chordates, spiralians exhibit many cases in which Hox clusters repeatedly disintegrate without reproducible patterns, nor are Hox genes connected to specific germ metamorphosis. Together, the evolution and diversification of Spiralia for more than 500 million years is a showcase of Hox gene evolution, where defining a general pattern is difficult. Moreover, Spiralia teaches us a lesson about the importance of using more taxon sampling to test-and sometimes reject-hypotheses based on observations on a few animal lineages. In this context, the diverse patterns of Hox expression and genomic organization that we find in Spiralia provide a novel resource to discover new mechanisms of genome regulation and organization, and the interplay between the two, along with the correlation of these phenomena with morphological evolution.