Hox gene expression in postmetamorphic juveniles of the brachiopod Terebratalia transversa

Background Hox genes encode a family of homeodomain containing transcription factors that are clustered together on chromosomes of many Bilateria. Some bilaterian lineages express these genes during embryogenesis in spatial and/or temporal order according to their arrangement in the cluster, a phenomenon referred to as collinearity. Expression of Hox genes is well studied during embryonic and larval development of numerous species; however, relatively few studies focus on the comparison of pre- and postmetamorphic expression of Hox genes in animals with biphasic life cycle. Recently, the expression of Hox genes was described for embryos and larvae of Terebratalia transversa, a rhynchonelliformean brachiopod, which possesses distinct metamorphosis from planktonic larvae to sessile juveniles. During premetamorphic development, T. transversa does not exhibit spatial collinearity and several of its Hox genes are recruited for the morphogenesis of novel structures. In our study, we determined the expression of Hox genes in postmetamorphic juveniles of T. transversa in order to examine metamorphosis-related changes of expression patterns and to test whether Hox genes are expressed in the spatially collinear way in the postmetamorphic juveniles. Results Hox genes are expressed in a spatially non-collinear manner in juveniles, generally showing similar patterns as ones observed in competent larvae: genes labial and post1 are expressed in chaetae-related structures, sex combs reduced in the shell-forming epithelium, whereas lox5 and lox4 in dorso-posterior epidermis. After metamorphosis, expression of genes proboscipedia, hox3, deformed and antennapedia becomes restricted to, respectively, shell musculature, prospective hinge rudiments and pedicle musculature and epidermis. Conclusions All developmental stages of T. transversa, including postmetamorphic juveniles, exhibit a spatial non-collinear Hox genes expression with only minor changes observed between pre- and postmetamorphic stages. Our results are concordant with morphological observation that metamorphosis in rhynchonelliformean brachiopods, despite being rapid, is rather gradual. The most drastic changes in Hox gene expression patterns observed during metamorphosis could be explained by the inversion of the mantle lobe, which relocates some of the more posterior larval structures into the anterior edge of the juveniles. Co-option of Hox genes for the morphogenesis of novel structures is even more pronounced in postmetamorphic brachiopods when compared to larvae. Electronic supplementary material The online version of this article (10.1186/s13227-018-0114-1) contains supplementary material, which is available to authorized users.

Many animals are characterized by an indirect life cycle in which embryos develop through a larval stage and subsequent metamorphosis, during which the larval body is reshaped into the adult one (e.g., [42,43]). As larvae and adults can significantly differ in their morphology, the transition process might be quite dramatic and hence attracted attention of many researchers as one of the pivotal moments of the animal development [44][45][46]. Although the process of metamorphosis has puzzled numerous developmental biologists, there are relatively few studies regarding shifts of Hox gene expression accompanying it [15,[47][48][49][50][51][52]. In some animals, both larvae and adults show canonical spatial collinearity, which often correlates with the gradual type of metamorphosis. This can be exemplified by investigated annelid species, in which both life stages exhibit spatial collinearity of most of the Hox genes, yet there are shifts in the combinations of genes defining particular body regions before and after metamorphosis [47,48]. On the other hand, in other animals (especially those with the more pronounced metamorphosis) only one of the developmental stages exhibits canonical spatial collinearity of Hox genes expression, whereas the remaining stage shows either a non-collinear expression or does not express Hox genes at all. For instance, in the tunicate Ciona intestinalis Hox genes exhibit spatially collinear expression in the nervous system of larvae, whereas in juveniles only the three posterior genes are expressed in the intestine [15]. Conversely in pilidiophoran nemertean Micrura alaskensis and in indirectly developing enteropneust Schiozcardium californicum the specialized larvae develop without expressing any of the Hox genes, which, in turn, are expressed in the canonical collinear way only in the rudiments of juvenile worms developing either inside larval body (pilidiophorans) or as the posterior extension of late larva (enteropneusts) [49,50]. A somehow similar situation is found in the indirectly developing sea urchin Strongylocentrotus purpuratus, in which only two Hox genes (hox7 and hox 11/13b) take part in the larva formation, whereas the rudiments of adult animal, developing inside the larval body, show collinear expression of five Hox genes (hox7, hox8, hox9/10, hox11/13a and hox11/13b) in the extra-axial mesoderm [51,[53][54][55]. Yet another type of the metamorphosis-related Hox genes expression shifts is found in scaphopod Antalis entalis in which only the mid-trochophore stage exhibits staggered Hox genes expression, whereas both competent larvae and postmetamorphic juveniles lack spatial collinearity [52]. Some of the scaphopod Hox genes partially retain their expression profiles throughout metamorphosis (hox2, hox5, lox5), whereas other substantially changes their expression domains (hox3, lox4, post1, post2) or are expressed only before (hox1) or after (hox4) metamorphosis [52]. It is therefore evident that the metamorphosis-related shifts in Hox gene expression and function vary a lot from one animal clade to another, as a result of diverse evolutionary and developmental processes, which shape the ontogeny of each particular group [56].
One of the animal groups with a distinct metamorphosis event are rhynchonelliformean brachiopods, represented by T. transversa for which Schiemann et al. recently described Hox genes expression in embryos and larvae [19]. Brachiopods, along with phoronids and possibly ectoprocts, constitute the clade Lophophorata (Fig. 1A, [57,58]), which, together with, for example, annelids, mollusks, flatworms, nemerteans and rotifers, belongs to a large clade of protostome animals called Spiralia (Fig. 1A, [58][59][60][61]). Extant brachiopods are traditionally divided into three groups: Rhynchonelliformea,  [58]), metamorphosis of Terebratalia transversa (B, based on Freeman [68]) and detailed morphology of competent larva (C, based on Santagata [70]). S stands for Spiralia, L for Lophophorata. 1 Competent planktonic larva, anterior to the right; 2 Larva settles on the substrate; 3 Inversion of the mantle lobe in the settled larva; 4 Juvenile; note that over the course of metamorphosis the internal surface of the larval mantle lobe becomes external, shell-covered surface of juvenile animal, external surface of the mantle lobe becomes inner surface of the mantle, whereas anterior lobe contributes to the lophophore rudiment developing inside mantle cavity. Musculature in C is depicted in green, nervous system in red and excretory organs in orange. al larval anterior lobe, an anterior nerves, anr anterior nerve ring, cm circular muscle, crs chaetal sac retractor muscle, csm chaetal sac musculature, csn chaetal sac nerve, ds dorsal shell, lgm longitudinal gut-related muscles, lpm lateral pedicle muscle, lr lophophore rudiment, m mantle, ml larval mantle lobe, mo mouth, mpm medial pedicle muscle, np neuropil, nr nephridium rudiment, pcn paraxial nerve cord, pe pedicle, pl larval pedicle lobe, pne pedicle nerve, vmm ventral mantle lobe lateral muscle, vs ventral shell Craniiformea and Linguliformea, the two latter forming sister clades [18,58,61,62], historically united into group Inarticulata. As all brachiopods, adults of T. transversa are filter feeding animals with external anatomy superficially similar to bivalves-most of the body, including lophophore, a filtering organ, is enclosed in the two-valved shell, which covers the dorsal and ventral surfaces of the body. The clade Rhynchonelliformea is further characterized by the set of morphological features, including posterior soft-tissued pedicle (by which animal attaches to the substrate), blind gut devoid of anus and articulated valve-hinge [63]. Additionally, rhynchonelliformean larvae (Fig. 1B1, C) differ from those found in other brachiopods by possessing three distinct body regions-anterior lobe, mantle lobe (bearing four chaetal sacs) and the most posterior pedicle lobe [63][64][65][66][67]. The rhynchonelliformean larva settles by adhering to the substrate with the posterior tip of the pedicle lobe ( Fig. 1B2) and undergoes a specific metamorphosis, which in case of T. transversa is relatively rapid (few hours to 1 day [64]) and involves inversion of the mantle lobe ( Fig. 1B2-3, [64,68]). The latter results in profound relocation of some larval tissues-in competent larvae the mantle lobe partially covers the pedicle one and its chaetae projects posteriorly, after metamorphosis mantle lobe with chaetae projects anteriorly, its former interior surface becomes exposed and produce protegulum (the first rudiment of the shell), whereas its former exterior surface constitutes walls of the mantle cavity of the juvenile (Fig. 1B2-4) [64,68,69]. Therefore, the rapid transition from larvae to the juvenile involves profound reshaping of the entire body, which poses a question to which extent are those two stages continuous [70].
Schiemann et al. investigated genomic order of Hox genes of T. transversa and Hox genes expression in embryos and larvae of T. transversa and craniiformean Novocrania anomala [19]. T. transversa has a split Hox cluster comprising of 10 Hox genes in three independent parts. One scaffold contains two anterior Hox genes, labial (lab) and proboscipedia (pb). A separate scaffold contains the longest section of the Hox complex, containing genes hox3, deformed (dfd), sex combs reduced (scr), lox5, antennapedia (antp), lox4 and post2, whereas the most posterior gene post1 is located in the third independent scaffold [19]. A disorganization of the Hox cluster has also been reported for linguliformean brachiopod, Lingula anatina, in which although all Hox genes are in the single cluster post1, post2, lox4 and antp have been translocated upstream to the lab [18]. In embryos and larvae of T. transversa, but also of craniiformean N. anomala, detected expression pattern of Hox genes does not show the canonical spatial collinearity [19]. However, as stated before, in some indirectly developing animals, larvae and juveniles can show collinear expression of Hox genes in patterning of one of the life stages, while the other develops without evident Hox expression collinearity.
Therefore, in this study, we supplemented findings of Schiemann et al. [19] by examination of the postmetamorphic Hox gene expression in T. transversa juveniles 2 days after metamorphosis. The main questions, which we were aiming to answer, were: (1) If and how is Hox genes expression pattern shifted during metamorphosis in rhynchonelliformean brachiopods? (2) Is there any staggered Hox genes expression along the A-P axis emerging after metamorphosis as a result of displacement of larval Anlagen and their development into definite adult structures?
Two days after metamorphosis, the juveniles of T. transversa already resemble the adult animal in their general shape ( Fig. 2A-C). The body is clearly divided into main part covered by the two-valved juvenile shell and a posterior pedicle (pe, Fig. 2A-C), by which the juvenile is attached to the substrate.
Optical sections through the animal show the narrow mantle cavity (mc, Figs. 2B, C, 3C'-G'), which expands ventro-medially to about two-thirds of the length of the animal body and is lined with the ciliated cells (vmc, Fig. 2C, F, I). The remnant of the larval anterior lobe, from which the prospective lophophore will develop [66,75], is situated inside the mantle cavity (lr, Figs At this stage, the lophophore rudiment is poorly developed and consists of two, scarcely ciliated lobes without tentacles (lr, Figs. 2B, C, 3B, B' , C, C'). The lobes are penetrated by numerous, fine tyrosinated-tubulin immunoreactive (tTIR) fibers (lrf, Fig. 2C, E), which communicate with the nervous system and probably represent the developing innervation of the prospective lophophore.
Two prominent tTIR and aTIR longitudinal structures are present in the ventro-lateral part of the animal (nd, Fig. 2C, F, I), extending along the ventral surface from the mid-posterior region to the ventro-posterior part of the mantle cavity. Dorsally those structures have numerous finger-like projections (ns, Figs. 2H, I, 3G, G'), which contact nuclei-free regions (as revealed by DAPI staining, Fig. 2H). We suggest that those structures represent metanephridia (composed of nephrostome and nephridial duct) of the juveniles, which connect the developing coelom with the mantle cavity. Their form and position are similar to what has been described for the metanephridia of relatively closely related Terebratulina retusa [76]. Although metanephridia in brachiopods are considered to be responsible only for release of gametes and not for excretion [76], they are present (albeit initially as non-functional rudiments) already in the early juveniles of N. anomala [77]. It is possible that aTIR structures described by Santagata [70] as larval protonephridia in T. transversa (nr, Fig. 1C) actually represent rudiments of the metanephridial ducts or nephrostomes which acquire their final form during or soon after metamorphosis.
The DAPI staining revealed an empty cavity inside the body of the juvenile with two pairs of anterior and posterior branches (cr, Figs. 2B, C, 3), which most probably represents the developing coelom in which some of the forming muscles are freely positioned (Fig. 3E, E' , H, H') [78]. Its two anterior branches extend along digestive tract and penetrate the lophophore rudiment (Figs. 2B, 3C-F, C'-F'). A similar arrangement of the coelom in the lophophore rudiment of postmetamorphic juveniles has been described for relatively closely related rhynchonelliformean Calloria inconspicua [79].
In addition to the already mentioned musculature related to the mantle margin, we identified rudiments of all the muscle groups (pedicle adjustors, shell diductors as well as anterior and posterior shell adductors, respectively, vpa, dms, aam, pam, Figs. 2D, G, 3D-H, D'-H') described for the older juveniles of T. transversa [71] with the only exception of the lophophore-related tentacle muscles (which correlates with lack of the lophophore tentacles 2 days after metamorphosis). In different specimens, the particular groups of muscles were developed to different degree corroborating the observation of the extensive and rapid remodeling of muscular tissue in the postmetamorphic juveniles [70].

In situ hybridization of Hox genes
The expression of the Hox genes in juvenile T. transversa (2 days after metamorphosis) was examined with colorimetric (CISH; hox3 is expressed in two paired domains posteriorly to the most lateral projections of the shell (arrowheads, Figs. 4C, 5E), where prospective hinge rudiments will form in the older juveniles [69]. CISH investigation showed additional broad weak staining in the posterior part of the body (Fig. 4C), which was not reproduced with FISH (Fig. 5E) and which might result from unspecific probe binding in the posterior shell as shown by sense probe staining (Additional file 1: Fig. S1B and C).
Expression of the gene scr is restricted to the mantle margin (Figs. 4E, 5I). Signal from probes against scr in CISH seems to be diversified into smaller domains with strong signal interspaced by wider regions of relatively weaker expression (respectively, arrowheads and asterisks, Fig. 4E), indicating an unequal expression of the gene along mantle margin. However, this diversification is not visible in FISH examination (compare Figs. 4E, 5I). The uniform signal from probes against scr in FISH might be an effect of the specific staining of the scr expressing cells and unspecific binding of the probe at the mantle margin (as in fluorescent stainings against lab and pb, where unspecific signal is visible along mantle margin; compare Fig. 5I with 5A, C).
The gene lox5 is expressed in the continuous dorsoposterior domain, which extend from posterior region of the shell-covered body (asterisks, Figs. 4F, 5J) to the pedicle tissues (arrowheads, Figs. 4F, 5J) and its expression is restricted to the dorsal epidermal cells as revealed by FISH (double arrowheads, Fig. 5K).
antp has a distinct expression domain only in the epidermis of the pedicle, as revealed by both CISH (arrowhead, Fig. 4G) and FISH (arrowhead, Fig. 5N-P). The signal in the CISH staining developed for the long time and before it became evident the strong staining had appeared in some specimens also in the dorso-posterior part of the shell-covered body. However, the control with sense probe showed that this staining results from unspecific binding of the probe in the dorsal protegulum (larval shell rudiment; asterisks, Additional file 1: Figure S1B) and on the borders between the dorsal protegulum and the remaining parts of the shell (arrowheads, Additional file 1: Figure S1B). The strong dorsal band was also visible in FISH staining (double arrowheads, Fig. 5N), but combined staining with DAPI showed that it is restricted to the surface area and does not penetrate the epidermis (arrowhead, Additional file 1: Fig. S1D, E), supporting our finding that it represents an unspecific probe binding by shell components. on all panels, on the right panels outline of the body is depicted in light blue, musculature in green, body cavities in gray, nervous system in red and excretory organs in orange. Scale bars on all images represent 20 μm. aam anterior shell adductor muscle, asc anterior supraesophageal commissure, cr coelom rudiment, csm chaetal sac muscle, dc dorsal commissure, dm dorsal mantle, dms shell diductor muscle, dt digestive tract, lr lophophore rudiment, mc mantle cavity, mm mantle margin muscle, np nephropore, ns nephrostome, pam posterior shell adductor muscle, psc posterior supraesophageal commissure, vm ventral mantle, vpa ventral pedicle adjustor muscle We did not manage to detect expression of post2 with in situ hybridization, which corresponds to the reported overall low level of post2 transcription in postmetamorphic juveniles of T. transversa [19]. After long developmental time, CISH staining yielded signal in the dorso-posterior part of the shell-covered body (asterisk, Fig. 4I); however, the control with the sense probe showed that this results from unspecific binding of the probe in the larval dorsal protegulum (Additional file 1: Figure S1C). The FISH staining only revealed a signal at the borders of the larval protegulum and the remaining parts of the shell (arrows, Additional file 1: Fig. S1F) and, similarly as in case of antp, FISH combined with DAPI staining revealed that this signal is restricted to the surface (shell components) and does not penetrate to the cellular epidermal layer (arrowhead, Additional file 1: Figure S1G). The unspecific binding of some probes by the larval protegulum has been already reported for T. transversa larvae [80], and apparently this phenomenon can also pose a problem in investigation of postmetamorphic animals.
Expression of the most posterior Hox gene post1 is detected along mantle margin (arrowheads, Figs. 4J, 5R), showing a relatively equal strength of signal with both CISH and FISH.

Double fluorescent in situ hybridization of the chaetae-related genes
In addition to the investigation of Hox genes in postmetamorphic juveniles, we performed double FISH of genes lab, post1 and arx (Aristaless-related homeobox) at the early developmental stages of T. transversa in order to better understand the relation of the expression patterns to the chaetal sac formation. The two former Hox genes have been proposed as related to chaetae formation in Brachiopoda [19], whereas arx is expressed in the chaetal sac musculature of annelid Platynereis dumerilli [81] and in the developing chaetal sacs of T. transversa [19]. We did a double fluorescent staining of post1 and labial (Fig. 6A) as well as post1 and arx (Fig. 6B) in the late gastrula stage and lab and arx in the early trilobed larva (Fig. 6C).
In the late gastrulae, gene post1 is co-localized with lab (Fig. 6A3), which shows extremely weak expression at this developmental stage (asterisks, Fig. 6A2). This is concordant with the CISH results from Schiemann et al. [19]), but our results show that it is also expressed in some post1-negative cells in between chaetal sac Anlagen (asterisks, Fig. 6A3). Additionally post1-positive cells of the late gastrulae strongly express gene arx (Fig. 6B). In the early trilobed larvae lab is expressed in the chaetal sac-related cells (Fig. 6C1), whereas arx expression is restricted to the subpopulation of the cells of the inner mantle lobe epithelium (Fig. 6C2) and the two genes are not co-expressed by any cells (Fig. 6C3).

Metamorphosis and Hox gene expression in Rhynchonelliformea
Comparison of the expression of Hox genes between late, competent larva and postmetamorphic juvenile of T. transversa (Fig. 7) shows that in both stages almost all Hox genes (with the exception of hox3, post2 and post1) are expressed in the corresponding organs and body regions: lab in chaetal sacs, pb and dfd in mesoderm, scr in the shell growth zone, whereas lox5, antp and lox4 are expressed in the dorso-posterior ectoderm. Most of the observed differences and shifts in the expression domains can be explained by the inversion of the mantle lobe, which constitutes the most profound process during the whole metamorphosis in Rhynchonelliformea (Fig. 1B, [64]). Another factor, which contributes to the observed changes, is the restriction of the expression of some Hox genes from broad, less specific larval domains to the particular structures of the juvenile, which emerge during or after metamorphosis. For example, pb is generally expressed in the anterior mesoderm in late larvae but in juveniles its expression becomes restricted only to particular mesodermal structures, i.e., newly formed anterior shell adductors muscles.
The comparison of the Hox gene expression between larvae and juveniles allows the identification of the Anlagen of adult structures in the larva. For example, the expression patterns of the Hox genes before and after metamorphosis suggest that only the posterior part of the larval pedicle lobe contributes to the pedicle of the adult, whereas the more anterior part becomes the posterior region of the shell-covered body, as it has been proposed by Stricker and Reed [64,72]. Among six Hox genes expressed in the pedicle lobe of the late larvae of T. transversa, only lox5 and antp are expressed in the pedicle of the postmetamorphic juvenile (Fig. 7), both of them being expressed in the most posterior part of the larval pedicle lobe [19].
Next to Rhynchonelliformea, two inarticulate clades belong to Brachiopoda: Craniiformea and Linguliformea [82], both possessing a planktonic larvae, which undergoes more or less pronounced metamorphosis [63,70,[83][84][85][86][87]. In Linguliformea, the metamorphosis itself is extended over time with some of the juvenile traits present already in the planktotrophic larvae [67,70,83,88], and the most advanced larval stages are even commonly considered as representing planktonic juveniles or paralarvae [63,70,88]. One can therefore speculate that as larval and adult body plans in Linguliformea are continuous, their patterning by Hox genes should be similar as is a case in T. transversa. On the other hand, there are two competing hypotheses about nature of the rearrangement of the larval body plan during metamorphosis of Fig. 7 Comparison of the Hox genes expression between the late competent larva (A, based on Schiemann et al. [19]) and juvenile (B) of Terebratalia transversa. Animals are shown in the dorso-ventral view (right panels) and in the lateral view with dorsal to the right (left panels). Anterior to the top on all panels. Bars on the right show antero-posterior Hox gene expression gradients in ectoderm and mesoderm of each developmental stage craniiformean brachiopods [63,84,85,89,90]. The main controversy regards whether the N. anomala larva, which lacks the distinct pedicle lobe, attaches to the substrate with its dorso-posterior side [84] or with the posterior tip of the posterior lobe [85]. Expression of the Hox genes is relatively similar between embryos and larvae of N. anomala and corresponding stages of T. transversa [19], indicating a conserved nature of Hox genes patterning between Craniiformea and Rhynchonelliformea. lox5 and antp, which after metamorphosis are expressed in the pedicle of T. transversa juveniles are expressed in the posterior tip of the posterior lobe of N. anomala larvae [19] favoring interpretation that posterior tip of N. anomala larvae corresponds to the pedicle of Rhynchonelliformea [85,90]. Further investigation of the postmetamorphic expression of Hox genes, especially lox4 and antp, in N. anomala could support this hypothesis.
Unlike some bilaterians in which metamorphosis seems to be related to highly different Hox gene expression between larvae and adults (e.g., tunicates [15], Bryozoa [91], scaphopods [52]) or in which Hox genes are not expressed in the larvae and only pattern adult body (pilidiophoran nemerteans [49], indirectly developing Hemichordates [50], sea urchins [51,53,54]), rhynchonelliformean brachiopods exhibit continuity in the patterning of larval and adult body plans. Consequently, in regard to Hox gene expression, metamorphosis in T. transversa is similar to the condition found in another spiralian clade, Annelida. Although there are some shifts in expression patterns of particular Hox genes between annelid larvae and juvenile worms [47,48], those differences are mostly related to restriction of some of the genes from broader larval to more specific adult domains [47]. This similarity can be explained if one assumes that, same as in Annelida, the metamorphosis of rhynchonelliformean larvae is not as drastic as it might seem and instead represents a relatively gradual process [67]. In T. transversa, several of the adult structures, including shell secreting epithelium [64,68] or pedicle muscles [70,72], are already present in the competent larvae as the Anlagen. Thus, even though transition from larva to juvenile poses large ecological change, from the morphological point of view the mantle lobe inversion is related mostly to tissue relocation and not to the degeneration or formation of entire body regions, as is the case in pilidiophoran nemerteans, indirectly developing hemichordates and sea urchins or ascidians.
From the phylogenetic and developmental point of view, it would be interesting to compare shifts of Hox genes expression observed during metamorphosis between T. transversa and Phoronida. Phoronids are closely related to brachiopods [57,58,92,93] (in past even proposed as specialized clade belonging to Brachiopoda [62,82]) and their rapid metamorphosis involves drastic rearrangements of the larval body plan [67,[94][95][96][97], which is much more complicated than the transition found in Rhynchonelliformea and sometimes referred to as catastrophic or cataclysmic metamorphosis [94,96,98]. The recent analysis of the body region-specific transcriptomes revealed that in adults of Phoronis austarlis, which possesses an organized Hox cluster, Hox gene expression does not exhibit spatial collinearity [18]. Unfortunately, data on the spatial expression of Hox genes in early developmental stages of any phoronid species are still lacking [98], preventing analysis of metamorphosis-related Hox genes expression shifts. Nevertheless, it is possible that in phoronids the larvae and juveniles exhibit pronounced differences in the Hox genes expression as is a case in some other animals with catastrophic and extensive metamorphosis [15,49,50].

Germ layer-specific expression of Hox genes
In most of the investigated Bilateria, Hox genes are predominantly expressed in the ectodermal domains and often their antero-posterior staggered expression is especially evident in the neuroectoderm, which lead to the assumption that at least one of the original roles of Hox genes was patterning of the developing nervous system along A-P axis [27,32,99]. Interestingly, we did not find any of the Hox genes expressed in the nervous system of postmetamorphic juveniles of T. transversa. This could be explained by the fact that in juveniles the main nervous structures are brain and lophophore nerves (Fig. 2C) both related to the larval anterior lobe and postmetamorphic lophophore rudiment, which represent derivatives of the head and hence do not express Hox genes (the same has been shown for the phoronid lophophor [18]). Schiemann et al. [19] also did not describe expression of any of the Hox genes in neuroectoderm of earlier developmental stages of T. transversa. However, as co-expression of neuroectoderm markers has not been tested in that work, it is difficult to ascertain whether T. transversa really lack Hox genes expression in neuroectoderm on all developmental stages.
Hox genes can be also expressed in particular mesodermal domains in almost all investigated bilaterians, with the exception of Hemichordates (where their expression is restricted to ecto-and endoderm [29,50]), rotifers (expression exclusively in the nervous system [32]) and Nemerteans (expression in ecto-and neuroectoderm [41]). Whether Hox genes were ancestrally expressed in the bilaterian mesoderm remains an open question. Nevertheless, taking into account that set of Hox genes expressed in the mesodermal derivatives differs substantially from one animal group to another and that their transcription in mesodermal tissues can happen on very different developmental stages, it seems plausible that Hox genes have been recruited many times independently to act in mesoderm development and specification [99].
In Brachiopods, three of the Hox genes (pb, hox3 and dfd) show mesodermal expression albeit all of them are also expressed in ectodermal domains at some point of development [19]. Prior to the metamorphosis, hox3 and dfd are expressed both in the mesodermal and ectodermal structures and after metamorphosis, due to the restriction of broader domains into specific structures, hox3 remained expressed only in the ectoderm, whereas dfd become restricted to the mesoderm (Fig. 7).
Orthologs of those three genes are reported as mesodermally expressed in some other spiralian species as well. For instance, pb exhibits mesodermal expression domains in gastropod Haliotis asinina [36] and two annelids-Chaetopterus variopedatus [33] and Alitta virens [48], hox3 and dfd are expressed in the mesoderm of scaphopod Antalis entalis [52] and hox3 is mesodermally expressed in annelid Capitella teleta [47]. However, lack of evidence that all three of those three genes are expressed in the mesoderm of any single non-brachiopod spiralian species as well as different timing of their mesodermal expression in particular animals indicates that expression of pb, hox3 and dfd in developing mesoderm might represent apomorphic feature of Brachiopoda or Lophophorata (investigation of phoronids and ectoprocts is needed to ascertain).

Expression of Hox genes during the morphogenesis of brachiopod-specific structures in T. transversa
Although Hox genes are believed to originally be responsible for antero-posterior patterning [1,2,11], in certain animal lineages some of them were co-opted for morphogenesis of evolutionary novel structures [100][101][102][103]. Among Spiralia, such phenomenon has been reported in, e.g., conchiferan molluscs [36][37][38]52] and annelids [34,35,47,48], whereas recently Schiemann et al. suggested that in brachiopod larvae 4 out of 10 Hox genes have been recruited for patterning of chaetae (lab and post1) and shell fields (scr and antp) [19]. Our results generally support findings of Schiemann et al.-although we did not find evidence for the expression of antp in the shell field-and show that co-option of Hox genes for morphogenesis of novel structures is even more pronounced in juveniles of T. transversa than it is in the larvae.
lab and post1 are recruited for the morphogenesis of chaetae in the embryos and larvae of T. transversa [19]. The gene lab is constantly expressed in the chaetal sacs from formation of their early Anlagen up to the latest larval stage, whereas post1 is only briefly expressed during short time window, when the Anlagen are formed. In our study, we detect expression of lab in the chaetal sacs of juveniles as well, but surprisingly we found that post1 is also expressed in the postmetamorphic juveniles. Moreover, its expression is not only restricted to the chaetal sacs but instead could be detected in the entire marginal zone of the mantle. This finding, however, makes sense when one takes into consideration that as adults T. transversa, as most of the rhynchonelliformean brachiopods, possess numerous chaetae along the mantle margin [88,104]. We therefore propose that although both lab and post1 are involved in the chaetae formation in Rhynchonelliformea, they play different roles: post1 is expressed in the regions where prospective chaetae will develop, possibly stimulating epidermal cells to differentiate into chaetal sacs before its expression decays. A similar role has been suggested for post1 in annelids, whose chaetae are considered homologous to brachiopod ones based on morphological [105] and molecular [19] similarities. In annelids, post1 is expressed in the cells of developing chaetae-bearing parapodia, but the expression becomes more faint over the time of development and is not detectable in the already formed parapodia [34,35,47]. lab, on the other hand, is possibly involved in the patterning of the growth of the chaetae itself, remaining expressed long after onset of chaetal sac formation. This hypothesis needs to be tested in the future by functional gene inference and the examination of older juveniles or adults, in which, accordingly, we would expect lack of post1 expression and broad expression of lab along the entire mantle margin.
Additionally, our investigation of the expression of chaetae-related genes in the earlier developmental stages of T. transversa shows that process of chaetal sacs formation is complicated and involves cell types, which spatially and temporarily differ in their gene expression profiles. At the late gastrula stage, lab, post1 and arx are all expressed in the two pairs of cell clusters, which have been interpreted as chaetal sacs Anlagen by Schiemann et al. [19]. In the later larval stage, only expression of lab is retained in the chaetal sacs-related cells, post1 is not expressed anymore, whereas expression of arx is shifted to the inner mantle lobe epithelium, which secrets protegulum (the larval shell rudiment). Interestingly, arx is not only expressed in the chaetal sacs Anlagen of annelids [81] and brachiopods and in the protegulum secreting epithelium of brachiopods but also in the radula formative tissue of the gastropod Tylomelania sarasinorum [106]. This indicates that among lophotrochozoans arx is generally expressed in the tissues forming various hard structures and cannot be unambiguously related to only single type of them.
The two-valved shell and posterior pedicle represent two distinct apomorphies of brachiopods, and we found four out of ten Hox genes expressed in the structures related to those morphological novelties. Our results indicate that scr is likely co-opted for the juvenile shell formation, as the gene is expressed in the mantle margin in the region specialized for shell secretion [69]. This finding corresponds to the results of Schiemann et al. [19], who found expression of scr in the epithelial cells forming larval shell rudiment.
Both shell and pedicle require sets of specialized muscles, which constitute an important part of the brachiopod body. In the late larvae of T. transversa, the genes pb and dfd are likely responsible for A-P patterning of mesoderm [19], yet during postembryonic development they seem to be recruited into morphogenesis of specific muscular structures that drive the biomechanics of, respectively, shell and pedicle. Additionally, hox3, another gene that seems to play a role in mesoderm patterning during earlier developmental stages [19], is expressed in the regions where future hinge rudiments will develop [69], suggesting that it could be involved in the morphogenesis of this autapomorphic rhynchonelliformean feature.

Conclusions
All developmental stages of T. transversa, including juveniles, express Hox genes in a spatially non-collinear manner [19]. Most of the patterns observed in the late larvae seem to persist throughout metamorphosis and are retained in juveniles, corroborating morphological observations that metamorphosis, despite being rapid, is of gradual type and most of the adult organs are present as Anlagen in the competent larvae. The most drastic shifts in Hox gene expression patterns observed during metamorphosis can be explained by: (1) the inversion of the mantle lobe which relocates some of the more posterior larval structures into the anterior edge of the juveniles and (2) restriction of the broad expression domains, present in larvae, to the specific structures in juveniles.
Concordantly to the previous study on larvae of T. transversa, we found that certain Hox genes have been evolutionary co-opted for morphogenesis of specialized structures in brachiopods. In both larvae and juveniles, lab is expressed in the chaetal sacs, whereas post1 marks the area where prospective chaetae will develop. In juveniles, four out of the ten Hox genes are expressed in the epidermal (scr, hox3) and muscular (pb, dfd) tissues related to shell and pedicle, two autapomorphic features of Brachiopoda.

Animal collection and fixation
Gravid adults of T. transversa (Sowerby 1846) were collected near San Juan Island, Washington, USA. Eggs obtained from the animals were fertilized, and developing larvae were cultured following previously published protocols (e.g., [19,64,68]) up to the metamorphosis. Two days after metamorphosis, juvenile animals were gently scraped from the bottom of the dish with a razor blade, relaxed with MgCl 2 , fixed in 3.7% formaldehyde and washed in phosphate buffer. Fixed animals were stored in 100% methanol.

In situ hybridization
Probes against Hox genes were synthesized using the same plasmid clones as used in Schiemann et al. [19], where the gene orthology assessment has been performed. Single whole-mount in situ hybridization was performed following an established protocol [107]. dUTP-digoxigenin-labeled probes were hybridized at a concentration of 1 ng/μl at 67 °C for 72 h, detected with anti-digoxigenin-AP antibody in 1:5000 concentration in blocking buffer and visualized with nitroblue tetrazolium chloride and 5-bromo-4-chloro-3-indolyl phosphate (in colorimetric in situ hybridization) or detected with anti-digoxigenin-POD antibody in 1:200 concentration in blocking buffer and visualized with TSA-Cy5-Plus (in fluorescent in situ hybridization). Additionally, animals prepared for FISH were stained for 30 min in DAPI to visualize cell nuclei. Stained juveniles where mounted in 70% glycerol and examined with Zeiss Axiocam HRc connected to a Zeiss Axioscope Ax10 using bright-field Nomarski optics (CISH) or scanned in Leica SP5 confocal laser scanning microscope (FISH). Double fluorescence in situ hybridization was conducted as described elsewhere [108].

Immunohistochemistry
For investigation of juvenile morphology, mouse primary monoclonal antibodies against tyrosinated-tubulin (Sigma, T9028) and acetylated-tubulin (Sigma, T6793) were used in 1:500 concentration. To visualize the primary antibodies, secondary goat anti-mouse antibodies (Life Technologies) conjugated with fluorochrome (Alex-aFluor647) were applied in 1:50 concentration. F-actin was visualized with AlexaFluor555-labeled phalloidin, and cell nuclei were stained with DAPI. Stained juveniles were mounted in 80% glycerol and scanned in Leica SP5 confocal laser scanning microscope.

Image processing and figure preparation
Z-stacks of confocal scans were projected into 2D images and 3D reconstructions in IMARIS 9.1.2. Both light micrographs and CLSM images were adjusted in Adobe Photoshop CS6 and assembled in Adobe Illustrator CS6. All the schematic drawings were done with Adobe Illustrator CS6.