Extensive loss of Wnt genes in Tardigrada

Wnt genes code for ligands that activate signaling pathways during development in Metazoa. Through the canonical Wnt (cWnt) signaling pathway, these genes regulate important processes in bilaterian development, such as establishing the anteroposterior axis and posterior growth. In Arthropoda, Wnt ligands also regulate segment polarity, and outgrowth and patterning of developing appendages. Arthropods are part of a lineage called Panarthropoda that includes Onychophora and Tardigrada. Previous studies revealed potential roles of Wnt genes in regulating posterior growth, segment polarity, and growth and patterning of legs in Onychophora. Unlike most other panarthropods, tardigrades lack posterior growth, but retain segmentation and appendages. Here, we investigated Wnt genes in tardigrades to gain insight into potential roles that these genes play during development of the highly compact and miniaturized tardigrade body plan. We analyzed published genomes for two representatives of Tardigrada, Hypsibius exemplaris and Ramazzottius varieornatus. We identified single orthologs of Wnt4, Wnt5, Wnt9, Wnt11, and WntA, as well as two Wnt16 paralogs in both tardigrade genomes. We only found a Wnt2 ortholog in H. exemplaris. We could not identify orthologs of Wnt1, Wnt6, Wnt7, Wnt8, or Wnt10. We identified most other components of cWnt signaling in both tardigrade genomes. However, we were unable to identify an ortholog of arrow/Lrp5/6, a gene that codes for a Frizzled co-receptor of Wnt ligands. Additionally, we found that some other animals that have lost several Wnt genes and are secondarily miniaturized, like tardigrades, are also missing an ortholog of arrow/Lrp5/6. We analyzed the embryonic expression patterns of Wnt genes in H. exemplaris during developmental stages that span the establishment of the AP axis through segmentation and leg development. We detected expression of all Wnt genes in H. exemplaris besides one of the Wnt16 paralogs. During embryo elongation, expression of several Wnt genes was restricted to the posterior pole or a region between the anterior and posterior poles. Wnt genes were expressed in distinct patterns during segmentation and development of legs in H. exemplaris, rather than in broadly overlapping patterns. Our results indicate that Wnt signaling has been highly modified in Tardigrada. While most components of cWnt signaling are conserved in tardigrades, we conclude that tardigrades have lost Wnt1, Wnt6, Wnt7, Wnt8, and Wnt10, along with arrow/Lrp5/6. Our expression data may indicate a conserved role of Wnt genes in specifying posterior identities during establishment of the AP axis. However, the loss of several Wnt genes and the distinct expression patterns of Wnt genes during segmentation and leg development may indicate that combinatorial interactions among Wnt genes are less important during tardigrade development compared to many other animals. Based on our results, and comparisons to previous studies, we speculate that the loss of several Wnt genes in Tardigrada may be related to a reduced number of cells and simplified development that accompanied miniaturization and anatomical simplification in this lineage.

Several lines of evidence suggest that Tardigrada represents the outgroup of an arthropod + onychophoran lineage in a monophyletic Panarthropoda [67,68,71], although other relationships have been recovered in some analyses [69,72], making studies of tardigrades critical for resolving where in panarthropod phylogeny important roles of Wnt genes evolved. Additionally, the unique body plan and developmental mode of Tardigrada raises intriguing questions about the roles of Wnt genes in this lineage. Tardigrades have a highly compact body plan composed of a single-segment head and four leg bearing trunk segments. This compact body plan evolved in conjunction with miniaturization [73,74]. Embryonic expression analyses of Hox genes and other AP axis patterning genes in the tardigrade Hypsibius exemplaris revealed that tardigrades have lost a mid-trunk region [75,76]. The mid-trunk region that is missing in tardigrades develops by posterior growth in other panarthropods [75,77]. Tardigrades have lost posterior growth; all segments develop nearly simultaneously in these animals [78][79][80][81]. Additionally, the proximodistal (PD) axis of H. exemplaris legs is missing an intermediate domain defined by dachshund expression that is found in onychophorans, arthropods, and other animals [82]. The fact that Wnt genes regulate the development of both the AP axis and PD axis in other animals suggests that the evolution of the compact tardigrade body plan may be associated with modifications to the functions of Wnt genes in Tardigrada.
Here, we present the first study of Wnt genes in Tardigrada. We discovered that several Wnt genes and arrow, an ortholog of Lrp5 and Lrp6 in vertebrates, which codes for a co-receptor of Wnt ligands, have been lost in the tardigrade lineage. Based on comparisons to other metazoan genomes, it appears that the loss of several Wnt genes and arrow/Lrp5/6 are common features of genome evolution in secondarily miniaturized animals, like tardigrades. Expression patterns of Wnt genes in H. exemplaris embryos suggest that these genes play roles during establishment of the AP axis, segmentation, endomesodermal development, foregut development, and leg development. Interestingly, Wnt genes exhibit distinct expression patterns during segmentation and leg development, rather than similar expression patterns like in many other animals. This fact, along with the extensive loss of Wnt genes in Tardigrada, may indicate that tardigrade Wnt genes exhibit reduced combinatorial interactions compared to some other animals. We suggest that the extensive loss of Wnt genes in Tardigrada is associated with miniaturization and the associated reduction of cell number and simplified developmental mechanisms that accompany this process.

Phylogenetic and genomic analyses of tardigrade Wnt genes
We identified eight candidate Wnt genes in the H. exemplaris genome and seven Wnt genes in the genome of R. varieornatus by BLAST search. We confirmed that all candidate Wnt genes encode a Wnt superfamily domain by CD-search analysis [83]. Our phylogenetic analyses revealed that both H. exemplaris and R. varieornatus possess orthologs of Wnt4, Wnt5, Wnt9, Wnt11, Wnt16, and WntA ( Fig. 1; Table 1; Additional file 1: Fig. S1). These species each had two paralogs of Wnt16, which we refer to as Wnt16A and Wnt16B. Additionally, H. exemplaris possesses an ortholog of Wnt2. We did not detect orthologs of Wnt1, Wnt6, Wnt7, Wnt8, and Wnt10-Wnt genes that would be predicted to be present in tardigrade genomes [2,3,5,6,84]. Next, we investigated the arrangement of Wnt genes in the genomes of H. exemplaris and R. varieornatus [71,85]. In H. exemplaris, all Wnt genes were found on unique scaffolds (Table 1). Scaffolds with Wnt genes ranged in length from 95,696 to 1,074,739 nt in this species. In R. varieornatus, scaffolds with Wnt genes ranged in length from 1,744,794-9,333,084 nt.
Only Wnt9 and WntA were found on the same scaffold, but these genes were 829,930 nt apart on this scaffold.

Identification of conserved components of canonical Wnt signaling in Tardigrada
We were surprised that tardigrades were missing so many Wnt orthologs. This led us to wonder whether these losses were associated with modifications to the cWnt signaling pathway, a highly conserved pathway that utilizes Wnt ligands [1,4,11,16]. First, we identified orthologs of wntless, which codes for a transmembrane transport protein that is necessary for secretion of Wnt ligands, in both species of tardigrades (Table 1). Next, we investigated the complement of Frizzled (Fz) genes, which code for receptors of Wnt ligands [86,87].   [83], we confirmed that all candidate Fz genes encoded both a cysteine rich domain and a seven-transmembrane G protein-coupled receptor domain, domains that are indicative of Fz proteins. A phylogenetic analysis revealed that these candidates are orthologous to the four Fz genes that are found in other panarthropods ( Fig. 2; Table 1). We also identified orthologs of genes that encode all major internal components of the cWnt signaling pathway [11], including disheveled, the β-catenin gene armadillo (arm), three β-catenin destruction complex genes-axin, shaggy, and Adenomatous polyposis coli tumor suppressor-and pangolin, which codes for a transcription factor that regulates the expression of targets of cWnt signaling (Table 1). However, we were unable to identify an ortholog of arrow, called Lrp5/6 in many animals (Additional file 2: S1). This gene encodes a co-receptor of Wnt ligands, which forms a receptor complex with Fz receptors [88][89][90]. From N-terminus to C-terminus, arrow orthologs typically encode three clusters of low-density lipoprotein receptor repeat class B domains separated by calciumbinding EGF-like domains, followed by three low-density lipoprotein receptor class A domains (Fig. 3a, b).
This structure was not encoded by the best tardigrade arrow hits (Fig. 3c).

Analysis of Lrp sequences in other bilaterians that exhibit extensive loss of Wnt orthologs
To better understand the relationship between missing Wnt ligand coding genes and the absence of a clear arrow ortholog in tardigrades, we searched for arrow in the genomes of five additional non-parasitic bilaterian animals that, like tardigrades, are highly miniaturized or evolved from miniaturized ancestors [73,74,91,92], and are missing several Wnt ligand-coding genes [5,93]. We focused on free-living miniaturized animals because they may have evolved under similar selective regimes and evolutionary constraints as tardigrades. These animals included the roundworm Caenorhabditis elegans, the rotifer Adineta vaga [94], the flatworm Schmidtea mediterranea [95], the acoel Hofstenia miamia [96], and the meiobenthic annelid Dimorphilus gyrociliatus [97]. D. gyrociliatus possessed a clear ortholog of D. melanogaster arrow ( Fig. 3c; Additional file 2: S1). S. mediterranea encoded a gene that was a reciprocal best BLAST hit to D. melanogaster Arrow (Additional file 2: S1). This gene had been identified as an ortholog of arrow/Lrp5/6 in a previous study [98]. The best BLAST hits from the Table 1 Canonical Wnt signaling components in tardigrades. GenBank accession numbers and scaffold numbers are from previously published genome studies [71,85]. An ortholog of Arrow was not found remaining species did not encode the pattern of protein domains that is typical of Arrow (Fig. 3c). A previous study indicated that C. elegans has lost an ortholog of arrow/Lrp5/6, in agreement with our results [99]. Next we performed a phylogenetic analysis of Arrow and related sequences. The D. gyrociliatus candidate Arrow/ LRP5/6 ortholog was nested within the clade of Arrow/ LRP5/6 sequences. Of the remaining miniaturized animals that we investigated, only H. miamia encoded a sequence that was nested within the clade of Arrow sequences, although this sequence was on a long branch (Fig. 4). Bootstrap support and posterior probability support was low for the Arrow/LRP5/6 clade, likely due to the inclusion of sequences from miniaturized animals, which were generally recovered on long branches. The best matches to D. melanogaster Arrow in tardigrade genomes, as determined by BLAST search (Additional file 2: S1), were recovered in a well-supported clade of LRP1 sequences (Fig. 4).

Expression patterns of Wnt genes during tardigrade development
We analyzed expression of Wnt genes at four different developmental stages in H. exemplaris-stage 11, the elongation stage ( Fig. 5a, b), stage 12 and 13, when segmental features first appear ( Fig. 5c-f ), and stage 14, the leg bud stage (Fig. 5g, h) [100]. He-Wnt9 was broadly expressed across embryos at all embryonic stages that we investigated, but may exhibit strongest expression in the endomesodermal layer (Additional file 3: Fig.  S2e-h). We did not detect clear He-Wnt16A expression at any stage that we investigated (Additional file 3: Fig.  S2m-p). The AP axis forms by a process referred to as elongation during stage 11 in H. exemplaris-a process during which the entire AP axis forms simultaneously, rather than forming in anteroposterior order as would be indicative of posterior growth (Fig. 5a, b) [100]. We analyzed He-six3 expression, a marker of anterior identity [76,101], with in situ hybridization at stage 11 to distinguish between the anterior and posterior embryonic poles ( Fig. 6a, b). In stage 11 embryos, an internalized cell layer extended to the external ectodermal cell layer at one embryonic pole (Fig. 5a, b). Six3 was expressed at the opposite pole ( Fig. 6a'  After elongation, the AP axis is curved in a characteristic "C" shape. At stage 12, the developing foregut in the head and trunk endomesoderm is visible (Fig. 5c, d) [100]. We did not detect strong signal for He-Wnt2 at this stage or later stages (Additional file 3: Fig. S2a-c). He-Wnt4 was expressed in the endomesodermal layer of the developing trunk (Fig. 7b, b'). He-Wnt5 was expressed in the ventrolateral ectoderm of the posterior head (Fig. 7c, d). He-Wnt11 signal was detected broadly across embryos at stage 12, but the strongest expression was localized to the posteriormost region of developing embryos (Fig. 7e, e'; Additional file 3: Fig. S2j). He-Wnt16B was expressed broadly across the posterior half of the developing trunk, General pattern of conserved domains in Arrow/LRP5/6 orthologs. c Pattern of conserved domains in the best matches to Arrow/LRP5/6 in species that have lost several Wnt ligands. The lower case letter after the H. miamia sequence does not indicate specific orthology. The specific orthology of this sequence is unclear based on our analyses. The S. mediterranea sequence represents a highly derived Arrow/LRP5/6 ortholog based on functional analyses [98] excluding the posteriormost region (Fig. 7f, f '). He-WntA was expressed in the ectodermal layer in the anteriormost region of the developing trunk and posteriormost part of the head (Fig. 7g-h).
At stage 13, segmentation is clearly visible in the form of ectodermal furrows and endomesodermal pouches (Fig. 5e, f ) [100]. He-Wnt4 was expressed strongest in the endomesodermal layer of the developing trunk (Fig. 8a,   Fig. 4 Majority rule consensus tree of LRP sequences. Tardigrade sequences are in colored boxes. Bootstrap supports are shown as percentages out of 500 bootstrap replicates. For simplicity, only branch support values relevant to determining the identity of the candidate tardigrade LRP sequences are shown. Lower case letters after taxon names indicate unclear orthology. The orthology of the S. mediterranea LRP5/6 sequence is based on functional analyses [98].  b). At stage 13, He-Wnt5 expression looked very similar to its expression at stage 12. At this stage, He-Wnt5 was primarily expressed in a ventrolateral region of the ectodermal layer of the developing head (Fig. 8c, d). He-Wnt11 was expressed broadly across developing embryos (Additional file 3: Fig. S2k). He-Wnt16B was expressed in the lateral ectoderm of the third and fourth trunk segment, excluding the posteriormost region of the fourth trunk segment (Fig. 8e, f ). He-WntA was expressed in a dorsolateral stripe in the boundary between the head and the first trunk segment (Fig. 8g, g').
At stage 14, developing legs are clearly visible (Fig. 5g,  h). Other features, such as pharynx and trunk ganglia are also visible [100]. At this stage, He-Wnt4 was very weak or absent (Additional file 3: Fig. S2d). He-Wnt5 was expressed broadly across the inner surface of developing legs, between the developing hind legs, in the ventrolateral ectoderm of the head, and in the developing pharynx ( Fig. 9a-b''). He-Wnt11 exhibited strongest expression in the posteriormost region of the AP axis, between the developing hind legs, at this stage (Fig. 9c, d). We also detected staining for He-Wnt11 in other parts of the embryo (Additional file 3: Fig. S2l). At stage 14, He-WntA was expressed in a dorsolateral ectodermal stripe between the head and first trunk segment, and in the developing pharynx (Fig. 9g, g'). He-Wnt16B was expressed in a stripe in each developing leg (Fig. 9e, f ' , f ''). We also detected expression of this gene in the posteriormost region between the developing hind legs (Fig. 9e,  f ''), and near the posterior part of the pharynx, which likely represents the esophagus (Fig. 9f ).

Evolutionary dynamics of Wnt genes in Tardigrada
Orthologs of Wnt2, Wnt4, Wnt5, Wnt9, Wnt11, Wnt16, and WntA are conserved in Tardigrada (Fig. 1). By contrast, Wnt1, Wnt6, Wnt7, Wnt8, and Wnt10 are missing in the genomes of two tardigrade species, H. exemplaris and R. varieornatus. These genes were most likely lost specifically within the tardigrade lineage (Fig. 10). However, these two tardigrade species are fairly closely related [78]. Therefore, whether these genes were lost in an ancestor of all tardigrades, or within crown group Tardigrada is unclear. Likewise, it is difficult to place the duplication event that gave rise to Wnt16 paralogs. It is clear that this duplication event occurred somewhere in the tardigrade lineage, because tardigrade Wnt16 genes formed a well-supported monophyletic group in our phylogenetic analyses (Fig. 1). It is also clear that these genes evolved by a duplication event that occurred in an ancestor of both H. exemplaris and R. varieornatus, rather than by independent duplication events (Fig. 10). The fact that the closest relative of He-Wnt16A is Rv-Wnt16A and the closest relative of He-Wnt16B is Rv-Wnt16B supports this conclusion (Fig. 1). As with the losses of Wnt genes, this duplication event could have occurred in any common ancestor of H. exemplaris and R. varieornatus in the tardigrade lineage. Genomic data from more distantly related tardigrade species would enable a more precise phylogenetic resolution of tardigrade specific losses and duplication of Wnt genes. By contrast, we can more precisely resolve the absence of Wnt2 in the genome of R. varieornatus as a loss in the lineage leading to this species after it split from the lineage leading to H. exemplaris, because Wnt2 is retained in H. exemplaris and outgroups of Tardigrada (Fig. 10).
Additionally, our results suggest that Wnt genes are dispersed throughout a chromosome or several chromosomes in the tardigrade species that we investigated, rather than clustered like Wnt genes in some other animal genomes [2,5,59]. The dispersion of Wnt genes in the genome most likely represents a derived state and could be related to the extensive loss of Wnt genes. Dispersion of ancestrally clustered paralogs in tardigrade genomes has been previously suggested for Hox genes and NK homeobox genes [75,102]. Therefore, dispersion of ancestrally clustered paralogs of developmental genes could represent a general pattern of genome evolution in the tardigrade lineage. The evolution of rapid embryogenesis has been suggested to abrogate purifying selection that would otherwise maintain clustered Hox genes and Parahox genes [103], and may also explain the dispersion of Wnt genes and NK homeobox genes in tardigrade genomes. Additionally, dispersion of ancestrally clustered paralogous developmental genes has been suggested to represent a general outcome of miniaturization [74].
In contrast to the extensive loss of Wnt orthologs, we identified nearly all other components of the cWnt signaling pathway in tardigrades (Table 1). This was not surprising given that cWnt signaling regulates many developmental processes in other animals. However, we were unable to identify an ortholog of arrow (Additional file 2: S1), referred to as Lrp5/6 in many animals, which codes for a co-receptor of Wnt ligands [88][89][90]104]. The closest match to arrow in tardigrades is a gene that most likely codes for an Lrp1 ortholog (Figs. 3c, 4). Possibly, the Wnt ligands that were lost in Tardigrada represent the Wnt ligands that required Arrow as a co-receptor to activate cWnt signaling in the ancient ancestors of tardigrades. To test this hypothesis, more studies are required to determine exactly which Wnt ligands require Arrow for cWnt signaling broadly across Metazoa.

Potential correlates of Wnt gene loss in Tardigrada
Barring initial redundancy, which is a poor explanation for long-term retention of Wnt paralogs [3], the loss of Wnt genes in the tardigrade lineage is most likely associated with modifications to development in Tardigrada. However, identifying exactly how development has been modified in tardigrades is extremely difficult, especially given the relatively scarce functional data for most Wnt genes. Additionally, we should not draw strong conclusions based on the loss of several Wnt genes in Tardigrada, given that gene loss is very common in this lineage [97,105], and given that the loss of one or more Wnt gene(s) is fairly common across Metazoa [3]. Nonetheless, a comparative approach may help explain the extensive loss of Wnt genes in Tardigrada. First, we note that, in addition to tardigrades, several other secondarily miniaturized animals exhibit extensive loss of Wnt genes (Fig. 10, Caenorhabditis, Adineta, Schmidtea, . The extensive loss of Wnt genes in some secondarily miniaturized animals may be due to the evolution of simpler cell fate specification mechanisms related to reduction in cell number and simplified morphology that often accompanies miniaturization. Along these lines, our expression data suggest that combinatorial interactions among Wnt genes may play a much less important role during development of tardigrades compared to typical macroscopic animals. In arthropods, onychophorans, and annelids, several Wnt genes are coexpressed in the posterior growth zone and are expressed in similarly positioned segmentally reiterated stripes [3,28,108]. Although Wnt genes exhibit different expression patterns during appendage development between arthropods and onychophorans (see below), within each lineage, several Wnt genes exhibit very similar expression patterns during appendage development [3,28,63]. Similar expression patterns of different Wnt genes in arthropods, onychophorans, and annelids most likely reflect combinatorial interactions of these genes during development [3]. Necessary combinatorial interactions of several Wnt paralogs could partly explain their conservation across much of Metazoa. None of the Wnt genes in H. exemplaris exhibited highly similar expression patterns during segmentation or leg development. Therefore, combinatorial interactions between Wnt orthologs must be less important during development of tardigrades. In some secondarily miniaturized animals, like tardigrades, one or more Wnt genes may become superfluous or redundant with a reduced requirement of combinatorial interactions, ultimately leading to loss by neutral evolutionary processes. A second non-mutually exclusive possibility is that the loss of one or more Wnt genes in Tardigrada is related to the loss of posterior growth, a process that is regulated by cWnt signaling in many macroscopic animals [20,21,27,109,110]. As with tardigrades, posterior growth is reduced, absent, or highly modified in many secondarily miniaturized animals ( [79,100,111,112], Martín-Durán JM, pers. comm.), which have also lost several Wnt genes (Fig. 10, Caenorhabditis, Adineta, Schmidtea,  Dimorphilus, Hofstenia). Additionally, long germband insects may retain fewer ancestral Wnt genes than their short germband relatives that continue to utilize posterior growth (Fig. 10, compare the long germband insects Drosophila, Anopholes, and Apis to the shortband insect Tribolilum) [3]. Arrow/Lrp5/6 is also necessary for normal posterior growth in arthropods and vertebrates [54,[113][114][115]. Therefore, the loss of arrow/Lrp5/6 in tardigrades and several other miniaturized animals may represent an additional genomic signature of the loss of posterior growth in these animals. Although feasible, there are significant difficulties associating the loss of Wnt genes and arrow/Lrp5/6 with the loss of posterior growth. First, the loss of arrow/Lrp5/6 is not tightly associated with the loss of Wnts genes or posterior growth. For example, D. gyrociliatus retains an arrow/Lrp5/6 ortholog (Figs. 3c, 4; Additional file 2: S1), but is missing several Wnt genes (Fig. 10, Dimorphilus) [97] and lacks posterior growth (Martín-Durán JM, pers. comm.). Likewise, although long germband insects have lost several Wnt genes and do not utilize posterior growth, they retain arrow [88]. Therefore, arrow clearly plays important developmental roles besides regulating posterior growth. Evidence for an additional role of arrow/Lrp5/6 comes from studies of the flatworm S. mediterranea. In this species, this gene is required to regulate posterior cell fate and proliferation during AP axis regeneration [98]. Second, given the sparse data available in regards to which Wnt gene regulates posterior growth, and the apparent interchangeability of Wnt gene function in this process based on data that are available [26,58], it is not possible to associate with confidence the loss of any particular Wnt gene in tardigrades with the loss of posterior growth. Nonetheless, based on a comparative perspective, it remains possible that reduced combinatorial interactions and the loss of posterior growth, both associated with miniaturization, contributed to the loss of several Wnt genes and arrow/Lrp5/6 in Tardigrada.

Wnt genes and regionalized AP cell fate specification
Several Wnt genes appeared to be expressed in regionalized patterns along the developing AP axis during embryogenesis in H. exemplaris. Given the absence of segment markers during elongation, it is difficult to determine precisely where the Wnt genes are expressed during this stage relative to each other (Fig. 6). However, the relative order of expression along the AP axis of some Wnt genes appeared to be maintained between the elongation stage and later stages during which segmental features are developing (Figs. 7, 8, 9). In other bilaterians, Wnt genes play a critical role in establishing the AP axis by promoting posterior identity, in part, by repressing anterior identity [9,[15][16][17][18][19]. We did not detect expression of any Wnt genes at the anterior pole of embryos at the elongation stage, besides He-Wnt9 and He-Wnt11, genes that exhibited very broad expression patterns. Therefore it is possible that some Wnt genes are playing roles in promoting posterior identity in H. exemplaris and other tardigrades. Interestingly, several Wnt genes are expressed in regionalized AP patterns in the onychophoran Euperipatoides kanangrensis during early stages of segmentation [28]. Based on these regionalized patterns, it has been suggested that Wnt genes may specify segment identities in onychophorans [28]. Potentially, one or more Wnt genes are playing a role in specifying segment identities in tardigrades. Future functional studies may help resolve this issue.

Wnt genes and segment polarity
In many arthropods, wg, en, and hh, interact positively as part of a regulatory network to establish segment polarity [30-34, 43-45, 50, 52, 57]. A segmentally reiterated pattern of gene expression emerges, which includes a stripe Fig. 10 Summary of distribution of Wnt orthologs in metazoan genomes. "X" in white boxes indicates the loss of a Wnt gene. "?" in white boxes indicates that a fully sequenced genome is unavailable for the associated lineage, so it is unclear whether the ortholog has been lost, or is present, but unsequenced. Gray boxes associated with Hofstenia indicate unclear orthology of the four Wnt genes found in the genomes of representatives of this lineage. Dashed lines coming off of Hofstenia represent different hypotheses of the relationship of this lineage with other bilaterians. The interrelationships of panarthropod phyla (Arthropoda, Onychophora, Tardigrada) are depicted as a polytomy because they are not currently resolved [69]. Arthropoda = Drosophila-Ixodes; Onychophora = Euperipatoides; Hypsibius-Ramazzottius = Tardigrada of wg expression positioned immediately in front of a stripe of en and hh expression. Parasegmental boundaries develop between the stripes of wg and the stripes of en and hh expression. Parasegmental boundaries are then replaced by segmental boundaries, which develop posterior to the en and hh stripes. In spiders, a different Wnt gene most likely substitutes for wg to regulate segment polarity [58]. Segment polarity genes have been investigated in two onychophoran species, Euperapatoides kanangrensis [64,65] and E. rowelli [66]. The expression patterns of the segment polarity genes in onychophorans are remarkably similar to those in arthropods. However, in contrast to arthropods, in onychophorans, Wnt genes and other segment polarity genes, besides en in E. kanangrensis, are first expressed in stripes after the earliest signs of segmentation appear [64,66]. Therefore, unlike in arthropods, wg and other Wnt genes most likely do not play a role in segment formation in onychophorans [28,64,66]. Nonetheless, a conserved segment polarity network is most likely regulating intrasegmental patterning in onychophorans after segments develop [65,66].
In H. exemplaris, the first signs of segmentation are the formation of endomesodermal pouches [100]. He-En expression is first detected in the ectoderm at a later stage in segmentally reiterated stripes immediately anterior to where ectodermal furrows develop between the underlying endomesodermal pouches [116]. Our results suggest that wg has been lost in the tardigrade lineage. Therefore, if cWnt signaling regulates en expression to establish segment polarity in the tardigrades of our study, this interaction must be mediated though a different Wnt gene. No Wnt gene was expressed in stripes in the ectoderm at the earliest stages of the segmentation process in H. exemplaris (Figs. 7,8). It is possible that cWnt signaling is not required for maintaining en expression during segment formation in Tardigrada. However, cWnt signaling could be maintaining en expression during segment formation in Tardigrada via a slightly modified mechanism. He-Wnt4 was expressed broadly throughout the endomesodermal layer of the trunk during segment formation (Figs. 7b' , 8a, b), rather than in the ectoderm as would be expected for a Wnt gene interacting with en. Nonetheless, the ligand that He-Wnt4 encodes could be providing a signal to En expressing cells. No other Wnt gene was expressed in a segmentally reiterated pattern that would be indicative of a role in segment formation. The only Wnt gene that was expressed in a stripe-like pattern in H. exemplaris was He-Wnt16B, which was expressed later in developing legs (Fig. 9e, f ' , f ''). Although this gene is unlikely to be playing a role in segment formation, its stripe-like expression pattern may indicate a later acting segment polarity function. However, by the time that legs are visible, He-En is no longer expressed in segmentally reiterated stripes [116]. Taken together, our results raise the interesting possibility that Wnt genes may not play a role in regulating segment polarity via regulatory interactions with en in Tardigrada. The precise developmental roles of Wnt4, Wnt16B, and other Wnt genes need to be clarified before a strong conclusion can be drawn regarding a role of cWnt signaling in regulating segment polarity, or the lack thereof, in Tardigrada. Additionally, other commonly conserved components of the segment polarity network, such as hh, need to be investigated in tardigrades. Resolving these issues is critical for determining the antiquity of the roles that the segment polarity network plays during development in Arthropoda and Onychophora.

Wnt genes and leg development
In arthropods, Wnt genes also play important roles in regulating appendage development. First, wg activates expression of Distal-less (Dll) to initiate appendage outgrowth [46,51,117,118]. Dll most likely also regulates appendage growth in onychophorans [119,120]. Wg and several other Wnt genes are expressed in a distal pattern in developing appendages in E. kanangrensis, suggesting that wg, and potentially other Wnt genes, may also be regulating growth via Dll in onychophorans [28,120]. Dll is expressed broadly across the developing legs of H. exemplaris, and most likely plays a role in regulating leg outgrowth in Tardigrada [82]. The tardigrades we analyzed appear to have lost wg. Therefore, wg cannot be activating Dll expression in these species. It is possible that Wnt signaling is not required for activating Dll expression in Tardigrada. Alternatively, one or more other Wnt genes may play this role. Wnt4 is a potential candidate for regulating Dll expression in tardigrades. In H. exemplaris, this gene is expressed in the endomesodermal layer below where legs will develop in the overlying ectoderm ( Fig. 8a, b). Two additional strong candidates are Wnt5 and Wnt16B (discussed in more detail below), which are both expressed strongly in developing legs (Fig. 9a-b' , e, f ' , f ''). Functional studies are required to determine whether these Wnt genes or others are required to activate Dll expression in Tardigrada.
Later in appendage development in arthropods, wg specifies ventral appendage fate [40,42,46,51,[121][122][123]. As with wg, several other Wnt genes are expressed in the developing ventral leg domain in arthropods and likely play a redundant or combinatorial role in specifying ventral appendage fate [3,60,62,63]. Unlike in arthropods, no Wnt genes are expressed in the ventral appendage domain in onychophorans [28]. It has been suggested that ventral expression of Wnt genes in developing appendages evolved in the arthropod lineage [28]. The tardigrades that we studied lack a wg ortholog, so this gene cannot be playing a role in establishing ventral appendage fate in these species. However, in H. exemplaris, He-Wnt5 is expressed on the inner side of developing legs (Fig. 9a-b'). Although not definitive, we interpret the inner side of the developing legs as ventral (Fig. 5h). In this interpretation, He-Wnt5 may specify ventral appendage fate in H. exemplaris. Wnt5 is not expressed in a ventral stripe in developing legs of arthropods [3,58,63]. This may indicate that a role of Wnt5 in specifying ventral fates in developing H. exemplaris legs, even if present, is not homologous to the function that Wnt genes play in specifying ventral fates in arthropod appendages.
We detected expression of He-Wnt16B in a stripe in each developing leg (Fig. 9e, f ' , f ''). Although not definitive, these stripes may lay in the posterior region of each developing leg (Fig. 5g, h). The insects that have been studied most extensively, D. melanogaster and T. castaneum, lack a Wnt16 ortholog [3]. In chelicerates and the millipede Glomeris marginata, Wnt16 is expressed in the distal tip and a ventral stripe, or just in a ventral stripe, in developing legs [3,58,63]. The Wnt16 ortholog of the onychophoran E. kangrenesis is expressed at the distal tip and in a posterior stripe in developing legs and other appendages, besides the frontal appendage, in which it is only expressed in a posterior stripe [28]. It is possible that a posterior stripe of Wnt16 in developing appendages was inherited in tardigrades and onychophorans from the last common ancestor of these lineages. Depending on the interrelationships of the panarthropod phyla (reviewed in [69]), this ancestor could represent the last common ancestor of Panarthropoda, in which case, the absence of a posterior stripe of Wnt16 in developing arthropod appendages would represent a derived state of Arthropoda.

Conclusions
Studies of tardigrades hold the potential to help resolve the evolution of developmental mechanisms in Panarthropoda. Our study revealed interesting possibilities regarding the evolution of the roles of Wnt signaling in regulating the development of key features of Panarthropoda-segmentation and appendages. Although in many respects the anatomy of tardigrades may best represent the anatomy of the last common ancestor of Panarthropoda [124,125], we cannot assume that developmental mechanisms in Tardigrada represent ancestral panarthropod mechanisms. In fact, studies indicate that tardigrade development is highly derived with simplification representing a common theme of developmental evolution in this lineage. For example, tardigrades have lost several Hox genes, and thus the Hox code must be simpler in modern tardigrades then it was in the last common ancestor of Panarthropoda [75]. Tardigrades are also missing a dachshund ortholog, indicating that leg patterning is simpler in modern tardigrades than it was in the last common ancestor of Panarthropoda [82]. In this study, we discovered that tardigrades have lost several Wnt genes. Other miniaturized animals are also missing several Wnt genes. In fact, Tardigrada, Nematoda, and Rotifera, which are all miniaturized animals [73,74,91,92], exhibit the highest level of gene loss among animals with sequenced genomes [105]. In terms of the body plan, miniaturization is associated with anatomical simplification and reduction in cell number [73,74,91,92]. Development of the simple body plans of highly miniaturized animals would not be expected to require the complex mechanisms that control development of larger animals. As developmental mechanisms, such as posterior growth or cell fate specification mechanisms that require complex combinatorial interactions, are lost in association with miniaturization, the genes that once regulated these processes may also be lost. Therefore, we propose that independent cases of extreme miniaturization in animals explain remarkable examples of convergence in terms of genome and developmental evolution.

Identifying candidate genes and phylogenetic analyses
Reciprocal BLAST searches were performed to identify candidate genes. We collected sequences from a genome assembly for R. varieornatus [85], and a genome assembly [71], an embryonic transcriptome assembly [126], and an adult transcriptome assembly [127] for H. exemplaris. We also collected sequences from genome or transcriptome assemblies for A. vaga [94], S. mediterranea [95], H. miamia [96], D. gyrociliatus [97], and E. kanangrensis [65]. All other sequences that we included in our phylogenetic analyses were publicly available in GenBank. We confirmed that candidate genes encoded predicted protein domains by CD search analysis [83]. For phylogenetic analyses, protein sequences were aligned with MUSCLE [128]. For Wnt analyses, we aligned sequences to a previously published matrix [3]. Information about sequences used in phylogenetic analyses is available in Additional file 4: S2. Alignments were trimmed using Gblocks [129,130]. Alignments were visually inspected in Mesquite [131]. Alignments are available in Additional file 5: FASTA alignments. The LG model [132] was used for phylogenetic analyses, with an estimated proportion of invariable sites and an estimated gamma shape parameter with four substitution rate categories. Maximum likelihood analyses were performed with PhyML [133]; branch support was calculated by bootstrap (500 replicates) and the aLRT SH-like method. We produced majority rule consensus trees in Mesquite with required tree topologies set to 0.5 from three maximum likelihood trees [131]. Bayesian analyses were performed with MrBayes with Nchains = 4 [134]. Tracer was used to diagnose convergence [135]. Posterior probabilities were calculated from 4500 trees from the posterior tree distribution.

PCR and cloning
Primers were designed from H. exemplaris gene sequences (sequences available upon request). GoTaq Green Master Mix (Promega) was used to amplify fragments of Wnt genes from H. exemplaris embryonic cDNA. Fragments were cloned into the pCR4-TOPO TA vector (Invitrogen). This strategy worked for all Wnt genes besides He-Wnt16A and He-WntA. He-Wnt16A was amplified from a single exon from genomic template. A He-WntA fragment was synthesized by Integrated DNA Technologies. Sanger sequencing was performed by Eton Bioscience to confirm the identity of cloned or synthesized sequences.

In situ hybridization and imaging
In situ hybridization was performed by following a published protocol [136]. This protocol has been used successfully for several previous studies of tardigrade development [75,76,82]. After completion of the in situ hybridization protocol, embryos were mounted on slides in DAPI-Flouromount-G (SouthernBiotech). DIC and fluorescence images were captured on an Olympus FV1000 Fluoview confocal microscope. We used the fluorescence properties of the chromogenic in situ hybridization stain to capture confocal data [137]. Fluorescence data were collected using a UPlanSApo 100×/1.40 oil objective, a 405 nm laser to capture DAPI data, and a 635 nm laser using the Cy5 excitation and detection presets in the Olympus Fluoview software to capture in situ data. Brightness and contrast of confocal stacks were adjusted in ImageJ. Images were produced in the Volume Viewer plugin in ImageJ. Image levels were adjusted in Photoshop.