Hox and Wnt pattern the primary body axis of an anthozoan cnidarian before gastrulation

Hox gene transcription factors are important regulators of positional identity along the anterior–posterior axis in bilaterian animals. Cnidarians (e.g., sea anemones, corals, and hydroids) are the sister group to the Bilateria and possess genes related to both anterior and central/posterior class Hox genes. Here we report a previously unrecognized domain of Hox expression in the starlet sea anemone, Nematostella vectensis, beginning at early blastula stages. We explore the relationship of two opposing Hox genes (NvAx6/NvAx1) expressed on each side of the blastula during early development. Functional perturbation reveals that NvAx6 and NvAx1 not only regulate their respective expression domains, but also interact with Wnt genes to pattern the entire oral–aboral axis. These findings suggest an ancient link between Hox/Wnt patterning during axis formation and indicate that oral–aboral domains are likely established during blastula formation in anthozoan cnidarians. Hox genes regulate anterior–posterior axis formation but their role in cnidarians is unclear. Here, the authors disrupt Hox genes NvAx1 and NvAx6 in the starlet sea anemone, Nematostella vectensis, showing antagonist function in patterning the oral–aboral axis and a link to Wnt signaling.

H ox genes are a specific family of homeobox-containing transcription factors that have been studied extensively in several clades of bilaterally symmetrical animals ("bilaterians", Fig. 1a). Originally discovered in Drosophila 1 , Hox genes play an important role in establishing segment identity along the anterior-posterior (A-P) axis during development and are conserved in all bilaterian lineages including vertebrates 2 . Hox genes are often clustered together along contiguous stretches of an animal's genome 3 , and are classified into three paralogy groups: anterior (Hox1-3), central (Hox4-8), and posterior (Hox9-13) (Fig. 1b). Early comparisons of vertebrate and insect Hox sequences revealed that homologous Hox genes occupy similar locations in the 3′-5′ topology of the cluster 4 and gene expression analysis further revealed that a gene's position in the cluster correlated with its expression profile along the A-P axis. Although there are derivations in the organization and distribution of Hox genes among bilaterians, Hox genes are consistently expressed in anterior-to-posterior territories as reflected by their phylogenic ancestry [5][6][7][8][9][10][11] . Furthermore, duplications (vertebrates) and fragmentation of Hox clusters (invertebrates) likely had a strong impact on the radiation of animals 8,12 , although the origin and function of this gene family within the first invertebrates is less clear [13][14][15][16] .
Phylogenetic reconstructions of the bilaterian Hox gene compliment indicate that the common ancestor of most extant animals had a Hox cluster containing anterior, central, and posterior Hox genes 17,18 (Fig. 1b, c). Although clustering is not an essential component of Hox functionality, it is widely associated with diverse animal clades from cnidarians and throughout Bilateria. Cnidarians (e.g., corals, anemones, and hydroids) are the sister group to the Bilateria 19 (Fig. 1a) and possess bona fide Hox genes related to anterior and central/posterior Hox genes 14,15,[20][21][22][23] . The cnidarian common ancestor likely had a Hox cluster consisting of at least two types of Hox genes, related to anterior and central/ posterior 21,22 (Fig. 1d, Supplementary Fig. 1). Definitive central class genes have yet to be found in a cnidarian. They first appear in the Xenoacoelomorph bilaterian lineage 24 and likely expanded in protostome/deuterostome lineages (Fig. 1b, c, Supplementary  Fig. 1). Currently, it is uncertain whether central class Hox genes arose in the bilaterian lineage from an ancestral central/posterior Hox gene, or were lost in the cnidarian lineage. Bona fide Hox genes have yet to be found in the genomes of earlier branching animal clades [25][26][27][28] , although some evidence suggests gene loss may have impacted this gene family 16 and a greater effort needs to be made to survey the diverse group of extant cnidarians. Efforts to functionally characterize the cnidarian Hox complement are crucial for determining the pre-bilaterian role of these important developmental regulators.
Functional studies in the fly 1 and mouse 29 first showed that Hox genes are causally involved in establishing adult body structures from the region in which they are expressed, thus controlling regional identity along the A-P axis. Hox genes also interact indirectly with one another in overlapping domains, with posterior Hox genes having functional "dominance" over more anterior genes; however, flies also exhibit examples of anterior dominance 30 . mRNA expression studies of Hox genes in multiple cnidarian species suggest a discrete role in late larval patterning, yet the developmental function remains untested 13,15,20,21,31,32 . The Hox cluster of the starlet sea anemone Nematostella consists of two neighboring non-Hox homeobox genes (NvHlxB9 and NvEve), along with putative orthologs to the anterior Hox genes Hox1 (NvAx6) and Hox2 (NvAx7 and NvAx8) 14,15 . Four additional Hox genes exist in the genome, consisting of: two central/ posterior genes (NvAx1 and NvAx1a), a pseudogene (NvAx9), and an additional anterior-like gene of indeterminate orthology (NvAx6a) (Fig. 1e). The assemblage of Hox clusters from different cnidarian genomes indicate that anthozoan cnidarians had a relatively intact Hox cluster, which has been fragmented in different species (Supplementary Fig. 1). Corals arguably have the most complete cluster consisting of both anterior and central/ posterior orthologs 22 (Supplementary Fig. 1c). Representative anemone species (Nematostella and Aiptasia) appear to have diverged from a cluster seen in some corals with a duplicated Hox2 ortholog (Ax7 and Ax8) ( Supplementary Fig. 1d-e). An ortholog of NvAx1 is present in all cnidarian genomes, yet there is no evidence of it being linked to a cnidarian cluster, suggesting it diverged long ago. In the case of Hydra, a high rate of cluster fragmentation is correlated with an expansion of central/posterior genes and less conservation among anterior-like genes 14,15 ( Supplementary Fig. 1f). Upcoming genome projects of other Hydrozoan cnidarians, Clytia hemisphaerica and two Hydractinia species, may reveal more details about the distribution of these genes. All of these findings are based on the idea that the ancestral condition was a state of Hox clustering, which is generally assumed to be the representative condition among bilaterian clades. A lack of functional data only allows us to speculate on the impact of cluster fragmentation within Cnidaria, and additional work is needed to resolve the impact of the cnidarian Hox diversification.
Previous in situ hybridization studies conducted on larval and juvenile stages of Nematostella revealed that the majority of NvHox genes are expressed in a staggered domain along the primary (oral-aboral) axis 15 with the anterior Hox gene (NvAx6) expressed in a region corresponding to the pharyngeal nerve ring at the oral pole, and the central/posterior Hox gene (NvAx1) expressed in the larval apical tuft at the aboral tip of the planula (Fig. 1e). The rest of the N. vectensis Hox genes are expressed later in development at planula stages asymmetrically along one side of the endoderm within the oral-aboral domain separating NvAx6 and NvAx1 15,20 . These data along with other studies in medusazoan cnidarians that possess complex metagenic lifecycles have created uncertainty as to whether cnidarian Hox genes have any bilaterian-like attributes. Early work using the cnidarian model system, Hydra, found a role in adult oral-aboral patterning 31,33,34 , while more recent findings concluded that the axial patterning role of Hox genes arose after the cnidarian bilaterian split 13,21,32 . Here we begin to dissect for the first time the functional role and hierarchy of Hox genes during cnidarian embryonic development and generate a presumptive molecular network for oral-aboral patterning.

Results
Hox genes are expressed prior to germ layer segregation. We analyzed the expression of Hox and neighboring non-Hox homeobox genes in Nematostella throughout early development (Fig. 2a), using a highly sensitive method (quantitative PCR) and verified expression by in situ hybridization. mRNA of both anterior and central/posterior Hox genes were detected early in development (Fig. 2b). Maternal mRNA of NvAx1 and NvAx6a could be detected by qPCR in fertilized eggs (Fig. 2b, 0hpf); however, transcriptional activation of the extended Hox gene cluster begins during early blastula formation (Fig. 2b, 12hpf) with expression of the anterior Hox gene NvAx6. Deployment of the Hox cluster continues during the blastula to gastrula transition, with the activation of NvEve (Fig. 2b, 16hpf) and two genes related to Hox2 (NvAx7 and NvAx8) (Fig. 2b, 18hpf). Lastly, activation of the central/posterior Hox gene (NvAx1a), which is fragmented from the predicted ancestral cnidarian Hox gene cluster (Fig. 1d) 22 , begins near the onset of gastrulation (Fig. 2b,  24hpf).
Two zones of spatial activation were found, with genes being expressed in oral (NvAx6, NvAx6a, NvEve, NvAx7,) or aboral (NvAx1, NvAx8) domains before and during early gastrulation prior to any asymmetries along the directive axis (Fig. 2b). Notably, NvAx6 and NvAx1 were the first to be detected by in situ hybridization, occupying complimentary oral and aboral domains during early blastula formation (12hpf) (Fig. 2c). NvHlxB9 and NvEve are expressed along the oral pole of the animal during early gastrulation 35 , and in agreement with the qPCR analysis (Fig. 2b), the NvHox genes appear to be expressed along a similar temporal timeline. Respective oral and aboral expression of NvAx6 and NvAx1 continues through the blastula to gastrula transition. At this stage NvAx6 is initially expressed at the site of gastrulation (animal pole) in the presumptive endomesoderm before becoming restricted to the pharyngeal endoderm ( Fig. 3a-e) associated with the pharyngeal nerve ring 15,20 . Conversely, NvAx1 maintains a broad aboral expression domain throughout blastula and gastrula stages before becoming refined to the most aboral domain of the apical tuft region during early planula stages ( Fig. 3f-j). The broad complimentary expression domains of the orally expressed anterior Hox gene NvAx6 and aborally expressed NvAx1 genes during blastula stages suggests they may play an important role in oral-aboral patterning during early development.
Unequal opposing role of Hox genes in oral-aboral patterning. NvAx6 and NvAx1 are expressed in opposite territories throughout development and their function was manipulated by microinjection of mRNA or antisense oligonucleotides into uncleaved zygotes. Knockdown of the orally expressed anterior Hox gene, NvAx6, results in embryos that have defective gastrulation, compared with controls ( Fig. 3k), and only initially form an inner endomesodermal plate (Fig. 3l). Overexpression of the aborally expressed Hox gene, NvAx1, by mRNA microinjection, results in a loss of gastrulation (Fig. 3m). Overexpression of the anterior Hox gene NvAx6 or morpholino knockdown of NvAx1, each interfere with pharyngeal development and produce external tissue with a pronounced asymmetry at the blastopore that fails to invaginate (Fig. 3n, o). At early gastrula stages (24hpf), knockdown of the orally expressed Hox gene, NvAx6, and the aborally expressed Hox gene, NvAx1, caused an expansion of NxAx1 and NvAx6 expression respectively ( Fig. 3p-q, s, t). However, overexpression of the anterior Hox gene NvAx6 exhibits only a slight expansion of NvAx1 expression (Fig. 3u). Conversely, overexpression of aboral Hox gene, NvAx1, completely abolishes NvAx6 expression at the oral pole (Fig. 3r) and inhibits oral development (Fig. 3m). This suggests that NvAx6 and NvAx1 maintain respective oral and aboral expression domains through antagonism, but that NvAx1 appears to have functional dominance over NvAx6, or that other factors contribute to the lack of antagonism in NvAx6 overexpression experiments (Fig. 3u).
The morphological changes found during early development and gastrulation are consistent with phenotypes identified during larval and polyp formation (Fig. 4). Three days after injection, ectopic tissue around the larval mouth is still present in animals with overexpression of NvAx6, where overexpression of NvAx1 inhibits larval development (Fig. 4a-c). The aboral sensory apical tuft is only affected by overexpression or knockdown of NvAx1 ( Fig. 4d- (Fig. 4k), where knockdown of NvAx1 results in deformed polyps with elongated body columns with less oral features, such as tentacles (Fig. 4l).
These findings suggest that altering the Hox program during early development results in changes along the oral-aboral axis that are maintained throughout development. To understand what changes occur as a result of manipulation of Hox expression, we analyzed a diverse set of developmental patterning genes and their relationship to NvAx6 and NvAx1 during early development. Control injections of dextran or control morpholinos do not impair expression of any genes analyzed herein ( Supplementary  Fig. 2). The endomesodermal marker, NvSnailA, is reduced in NvAx6 knockdown experiments and is lost in NvAx1 overexpression treatments (Fig. 5a, c-e). In these two treatments, embryos continue to gastrulate ( Fig. 3l-m), fail to form the pharynx, and do not express the pharyngeal marker NvFoxA or the oral marker NvBrachyury that is expressed at the ectodermal/ pharyngeal boundary (Fig. 5a), and which has been shown to regulate pharynx formation in anthozoans 36,37 . NvAx6 knockdown shows minor expansion of aboral markers such as NvFGF2A, NvSfrp1/5, NvDkk1/2/4, and NvSix3/6 ( Fig. 5b), but does not appear to effect aboral development, as embryos treated with NvAx6 anti-sense morpholino eventually form a swimming planula larva with an apical tuft at the aboral pole (Fig. 4e). Surprisingly, NvAx1 overexpression reduces aboral markers expression ( Fig. 5b-c, e) and the apical tuft in this region does not form (Fig. 4f). NvSix3/6 appears both diminished and disorganized, expanding into multiple regions of the embryo ( Fig. 5b-c, e).
Overexpression of the anterior Hox gene NvAx6 or morpholino knockdown of NvAx1 produce ectopic tissue that expresses the pharyngeal/oral markers NvFoxA and NvBrachyury (Fig. 5a, c, f, g). NvSnailA is expressed normally in cells of the endomesodermal plate indicating that these defects are related to  pharyngeal patterning and not endomesoderml plate specification (Fig. 5a). Aboral markers are lost in NvAx1 knockdown treatments, suggesting that the NvAx1 expression is necessary for aboral specification (Fig. 5b-c, g). However, aboral marker expression and apical tuft formation are not disturbed by NvAx6 overexpression (Figs 4g, 5b, c, f).
Role of NvAx6 and NvAx1 patterning during early development. Disruption of oral or aboral Hox genes expand opposing territories (Fig. 3q, t). Ectopic aboral Hox (NvAx1) expression restricts gastrulation and oral specification, resulting in the lack of endomesoderm formation throughout development (Fig. 5a). Similarly, knockdown of the oral Hox gene NvAx6 produces a severe defect in gastrulation (Fig. 3l) and results in an expansion of NvAx1 expression toward the oral pole (Fig. 3t). This suggests that stalled gastrulation as a result of NvAx6 knockdown is, in part, a result of the upregulation and expansion observed by the NvAx1 transcript, which antagonizes oral development (Fig. 3m). To test this hypothesis, we co-injected both NvAx6 (oral) and NvAx1 (aboral) morpholinos into N. vectensis zygotes and assessed changes in late gastrulae (48hpf). Controls were injected with either NvAx6 or NvAx1 morpholinos with a standard control morpholino and produced phenotypes identical to those seen with normal single gene-specific morpholino injections (Fig. 6). Embryos treated with both NvAx6 and NvAx1 morpholinos undergo gastrulation to form an outer ectoderm (Ec) and an inner endomesoderm (En) that expresses the maker NvSnailA, but fail to form a pharynx(P) or express the pharyngeal markers NvFoxA and NvBrachyury (Fig. 6). Aboral markers assessed in this study were lost in double injection (NvAx6 + NvAx1) experiments, including a complete loss of NvSix3/6 ( Fig. 6), which is reduced but not lost in NvAx1 morpholino treatments. Co-injection of NvAx6 and NvAx1 morpholinos has no effect on oral marker expression, including the pharyngeal marker NvFoxA at earlier gastrula stages (24hpf) ( Supplementary  Fig. 3), suggesting that patterning cues downstream of the oral Hox gene, NvAx6, are required after the onset of gastrulation. However, the aboral marker NvSix3/6 is lost at 24hpf in treated embryos ( Supplementary Fig. 3), demonstrating that NvAx1 expression has a patterning role at these earlier gastrula stages.
NvAx6 and NvAx1 regulate oral-aboral axis through Wnt. The oral-aboral axis of cnidarians is thought to be patterned by restricted Wnt domains beginning at the onset of gastrulation and progressing through larval development [38][39][40][41][42][43] . NvWnt1, NvWnt3, NvWnt4, and NvWntA are expressed around the blastopore/oral   pole and disappear following both oral Hox NvAx6 knockdown and aboral Hox NvAx1 overexpression (Fig. 5a). NvWnt2 is expressed in a band along the ectodermal midline of the embryo and serves as an important marker for defining oral and aboral territories. Knockdown of Hox NvAx6 expression results in shift of NvWnt2 expression toward the oral pole, while overexpression of NvAx6 has no observable effect on NvWnt2 expression (Fig. 5a). NvWnt2 expression is lost in aboral Hox (NvAx1) overexpression treatments, while knockdown of NvAx1 resulted in a robust upregulation and aboral expansion of the NvWnt2 domain (Fig. 5a). A more pronounced phenotype was produced in CRISPR cas9-mediated knockout of NvAx1, resulting in upregulation of NvWnt2 throughout the entire ectoderm and endomesoderm ( Supplementary Fig. 4). NvAx1 knockouts also expanded the orally expressed NvWnt1 into the endomesoderm ( Supplementary Fig. 4). These findings suggest that NvWnt2 is the intermediary between the ectodermal oral-aboral Hox gene boundary and that territorial Wnt expression is mediated by blastula Hox patterning.

Discussion
Since the first cnidarian genome became available 44 it has become clear that much of the developmental "toolkit" deployed during the development of bilaterian species remains present in the cnidarian lineage. Although there are aspects of cnidarian Hox clusters that have diverged from the ancestral condition in virtually all lineages (e.g., loss or duplication of Hox2 homologs, diversification of central/posterior genes in Hydra, Supplementary Fig. 1), it appears that the common ancestor of cnidarians and bilaterians utilized Hox genes to pattern their primary (oral-aboral) body axis.  that anterior and central/posterior Hox genes play important roles specifying oral and aboral territories, respectively, prior to the onset of gastrulation. During gastrulation and on through planula development, a complex developmental program is initiated that involves Wnt/β-catenin signaling that patterns the oral-aboral axis. During this time, Bmp agonists/antagonists activate Hox genes in the endomesoderm layer along a second axis, perpendicular to the oral-aboral axis, called the directive axis (may or may not be homologous to the dorsal-ventral axis of bilaterians) [47][48][49] . The directive axis is likely established around the onset of gastrulation, after establishment of the oral-aboral axis at the site of gastrulation. If the evolutionary origins of distinct clades of Hox genes (e.g, anterior class Hox genes) pattern similar regions of cnidarian body plans, as they do in bilaterians (i.e., anterior body regions), our findings would argue that the oral pole is homologous to the bilaterian anterior pole and the posterior end would be equivalent to the cnidarian aboral pole. This is consistent with numerous studies that show the cnidarian animal pole gives rise to the oral pole 50-54 , just as it does in virtually all other bilaterians [55][56][57][58] and in another early branching basal animal lineage, the ctenophores 59,60 . If the phylogenetic relationship of these genes also coincides with axial positioning, these findings would suggest that the oral-aboral axis is homologous to the bilaterian A-P axis, with the adult mouth corresponding to anterior in both cases. The alternatives are that A-P positioning never existed prior to bilaterians 13 or it could be argued that early animals may have possessed only "anterior-like" Hox genes without true central/posterior class genes 14 . The later scenario appears to be less likely, as medusazoan cnidarians like Clytia and Hydra appear to have true central/posterior class genes 21 , although greater taxonomic sampling is needed. Despite what we know about Hox genes in cnidarians, some components of the molecular patterning of the apical tuft neurosensory cilia region appear to be similar in different invertebrate larvae [61][62][63] . Further characterization of this transient larval apical tuft region across cnidarians is warranted, because some of the most robust indicators like the homeobox gene, Six3/6, is expressed orally in other cnidarians 64 . Of course, additional studies are required to determine with certainty the molecular basis for axial development in animal evolution.
In this report, we have characterized for the first time in any cnidarian embryo the existence of early domains of reciprocal Hox gene expression. In the starlet sea anemone, Nematostella vectensis, an anterior Hox gene is expressed at the beginning of blastula stages, extending through gastrulation and functionally interacts with a maternally expressed central/posterior Hox gene in oral-aboral axis formation and tissue specification. Although suggested to be primarily a vertebrate phenomenon 65 , elements of the hierarchal prevalence among Hox genes appears to be functioning in Nematostella. NvAx6 and NvAx1 act in a partial antagonistic fashion to pattern both oral and aboral territories (Fig. 7a, b), which appears to be mediated through interactions with Wnt signaling. Through these studies, we can begin to understand the molecular basis of oral-aboral patterning in Nematostella and develop a working model of cnidarian Hox function (Fig. 7c).
During early cleavage stages, the stabilization of β-catenin and components of both canonical and PCP Wnt signaling pathways establish the site of gastrulation/future oral pole in Nematostella 66,67 (Fig. 7c; I). Recent findings have shown that Nvβ-catenin activity is upstream of NvBrachyury 50 and that NvBrachyury is upstream of Wnt signaling 51 at the oral pole, a process that is partially mediated by the homeobox gene NvSix3/6 68 ( Fig. 7c; II). Our findings suggest that the maternal aboral marker NvAx1 also suppresses activation of Wnt signaling at the oral pole, when overexpressed, potentially through blocking activation of NvAx6 and downstream targets (e.g., NvBra, NvWntA, NvWnt1, NWnt3, and NvWnt4) (Fig. 7c; IV). These results are consistent with crosstalk between Nvβ-catenin and aboral patterning 68 , where oral Nvβ-catenin is necessary to initiate aboral patterning. Therefore, the loss of aboral genes as a result of NvAx1 overexpression may be a result of the inactivation of oral β-catenin, in turn effecting both oral and aboral specification. Recent work began to establish the gene regulatory network driving  endomesoderm formation in Nematostella, and highlighted the blastopore ring domains along the oral/animal pole 35 . Our findings add greater detail to this network and suggest that the NvWnt2 expression domain could be the boundary region between dedicated oral-aboral territories. Hyper activation of βcatenin signaling through pharmacological treatment was shown to shift NvWnt2 boundaries toward the aboral pole, while suppressing NvAx1 43 . Nvβ-catenin knockdown abolishes NvWnt2 expression 68 yet the functional link with oral and aboral Hox genes has yet to be resolved. The formation of drastically elongated polyps in aboral Hox (NvAx1) knockdown (Fig. 4l) points to a captivating similarity between the functions of vertebrate and anthozoan central/posterior (and posterior-like) Hox genes in axial elongation. Similar phenotypes were also witnessed as a result of mis-regulation of Nvβ-catenin signaling 66 , providing greater evidence for a link between these two molecular pathways. Results from the double knockdown of NvAx6 and NvAx1 provides support for the proposed role of NvAx1 in restricting oral specification, possibly through the restriction of Wnt signaling, as parallel knockdown of NvAx1 was sufficient to rescue the loss of gastrulation and oral specification phenotype produced in NvAx6 morpholino treatments. From these results, we propose that NvAx1 is an inhibitor of Wnt/β-catenin signaling ( Fig. 7c; IV) similar to what has recently been described during vertebrate development 69 . Therefore, the loss of NvAx1 expression in oral territories is necessary for establishment and progression of oral patterning and the core regulatory relationship between central/posterior (posterior-like) Hox expression and Wnt signaling may be a highly conserved characteristic of posterior Hox function. Further investigation into the mechanism of this relationship and the hierarchal nature of the other NvHox genes will help determine if the paradigm of posterior dominance is truly conserved between bilaterians and cnidarians, and will help resolve if NvAx6 exhibits any form of anterior dominance.
The phylogenetic origin of the Hox patterning system is a highly controversial matter among evolutionary developmental biologists. The cnidarian clade has the greatest diversity of Hox genes among the four basal animal lineages (sponge and ctenophore genomes do not contain Hox genes 25,26,28 ). Genomic resources from a diverse group of cnidarians suggest rearrangements and fragmentations of genomic clusters are present among cnidarian species. The axial patterning role of cnidarian Hox genes was long debated 21,31-34 , yet the functional analysis of Hox genes in Nematostella indicate that some aspects of genomic organization and A-P patterning were present in the cnidarian ancestor. It will be interesting if the study of other cnidarian species reveal an additional role for Hox genes during early development, which will ultimately help resolve the shared and derived characters of cnidarian and bilaterian Hox genes. Together these data suggest that cnidarians retain components of a simple ancestral Hox system that was originally deployed during early developmental stages and functioned to organize the primary axis in the cnidarian ancestor.

Methods
Animal care. Adult Nematostella vectensis were raised in 1/3× seawater at 16°C in constant dark. Animals were fed artemia once a week and fed oyster 48 h before spawning, which was engendered by an 8-h light box cycle at 16°C. Distinct groups of animals were spawned once every 3-4 weeks. Fertilized embryos were collected and placed in a 4% cysteine wash (4% cysteine in 1/3 filtered seawater, pH 7.4) for 15 min to remove the outer jelly layer. Eggs were then washed (3×) with 1/3 filtered seawater before conducting experiments. Treated embryos were raised to collection at 16°C. Embryos used were from randomized individuals from different genetic individuals to eliminate genetic variability. Mixed genetic pools of individual animals created a blind test of the effectiveness of each treatment regardless of genetic ancestry.  Microinjection knockdown overexpression and knockout. We disrupted NvAx6 and NvAx1 Hox gene expression using gene-specific antisense translation blocking morpholinos or CRISPR/Cas9-mediated genome editing using multiple guide RNAs (gRNAs) against each gene 36,49 . Genes were over expressed by injecting in vitro transcribed mRNA created using mMessage in vitro transcription kits (Ambion AM1348). Morphological and molecular analysis of phenotypes was conducted during late gastrula (48hpf), planula larvae (96hpf) stages of development, and 1-2-week-old polyps. Control injections using a standard control morpholino and dextran lineage tracer (or Cas9 protein without gRNAs in knockout injections) produced normal gastrula stage embryos (Fig. 3k) possessing an outer ectoderm(Ec), inner endomesoderm(En), and pharynx(P) (dashed white line) (Supplementary Fig. 2). Knockdown and knockout treatments produced similar molecular phenotypes ( Supplementary Fig. 4), thus validating the efficacy of the NvAx6 and NvAx1 morpholinos. Treated embryos were raised to collection at 16°C. Translation blocking morpholinos against NvAx6 and NvAx1 and a standard control morpholino have been developed through Gene Tools, LLC. Philomath OR, 97370. Morpholino sequences are as follows (Standard Control 5′-CCTCTTACCT CAGTTACAATTTATA-3′; NvAntHox6 5′-ACCGCCGCTCATGCCCAAA TGTGTC-3′; NvAntHox1 5′-TTGACTGCATGATGTGCGCTCTAGT-3′). Expression constructs for NvAx6 and NvAx1 were generated using the gateway cloning system (SPE3-Ax1-Rvenus, SPE3-Ax6-RmCherry, and SPE3-Ax6a-RmCherry). Morpholinos and mRNA were first injected in a dilution series ranging in concentration from 0.1 and 0.9 mM and appropriate concentrations (0.3 mM for NvAx6 MO, 0.5 mM NvAx6 mRNA, 0.9 mM for NvAx1, and 0.3 mM NvAx1 mRNA) were selected to achieve an optimal relationship between toxicity and phenotype penetrance.

NvSnailA
CRISPR/Cas9 genome editing in Nematostella has recently become a useful tool for genetic manipulation 36,49 . To perform knockouts of Hox genes, target sequences 18-20 bp in length fitting the gRNA target profile of 5′-G(G-A)-N(16-18)-NGG-3′ were identified within the coding sequence of the NvAx6 and NvAx1 locus (Supplementary Figs 5a, 6a) using a web-based program called ZiFiT (http:// zifit.partners.org). Six target sequences were selected for each locus bases on their location (choosing sites scattered throughout the coding sequence) and their gRNA efficiency score, which was calculated using the web-based CRISPR Efficiency Predictor provided by the DRSC at Harvard Medical School (http://www.flyrnai. org/evaluateCrispr/). Sequences and efficiency scores are listed in Supplementary  Table 1. Additionally, potential target sequences were blasted against the publically available Nematostella genome (http://genome.jgi.doe.gov) to insure that there were minimal off-target sites related to the sequence, limiting partial off-target sites to having no more than 15 base pairs in common with the 18-20 base pair target sequence. Oligonucleotides were generated integrating the target sequence (5′-G (G-A)-N(16-18)-NGG-3′) into a CRISPR RNA (crRNA) sequence containing a T7 (5′-AATTAATACGACTCACTATA-3′) or Sp6 (5′-AATATTTAGGTGACA CTATA-3′) promoter. Full-length gRNA was generated from a PCR assembled trans-activating crRNA (tracrRNA) oligonucleotide (Supplementary Table 1), followed by in vitro transcription (NEB Highscribe™ T7 and Sp6 RNA synthesis kits; Cat# E2040S and E2070S), and RNA purification (Zymogen RNA Clean and Concentrator™-25; Cat# R1017). gRNAs at a final concentration of 400 ng/µl (consisting of equal concentrations of each gRNA) were injected with Lyophilized bacterial type II Cas9 protein (PNA Bio, Thousand Oaks, CA) (1 μg/µl) reconstituted in 50% glycerol and 0.1 mM DTT to generate frame shift mutations and deletions within the target locus. Genomic DNA was isolated from 8 individual 18hpf embryos per treatment group for PCR analysis of the targeted region using a DNA extraction kit, according to the manufacturers protocol (Qiagen 69504) ( Supplementary Figs 5b, 6b). Additionally, in situ hybridization was performed on treated blastula stage embryos to ensure that gene expression was lost (Supplementary Figs 5c, 6c). Treated embryos were collected at 48hpf for in situ hybridization studies to assess changes in marker gene expression ( Supplementary  Fig. 4).
Quantitative PCR. Quantitative PCR was performed using the Roche LightCycler ® 480 Instrument II and LightCycler 480 SYBR Green I Master mix (Cat# 04707516001, Roche, Inc.). qPCR samples were standardized with NvGADPH and NvRiboPro and primers for other genes were designed using MacVector (www. macvector.com) to amplify 75-150 base-pair fragments of the desired gene. These primers were then back-blasted against the Nematostella genome to make sure they only will amplify a single region from the genome. We checked each primer efficiency with a dilution curve (10 −1 -10 −5 ) to make sure their range was within the negligible value of 1.9-2.0. Individual stages were collected from pooled samples of embryos consisting of roughly 100 embryos. These stages represent a single biological sample, and were confirmed through technical replication. Total RNA from each sample was stored in TRIzol (15596-026) at −80°C until processed. RNA was isolated according to manufacturer's specification. RNA was DNase treated using Qiagen RNase-free DNase #79254 for 15 min at 37°C. cDNA was standardized by using 1 µg of total RNA and synthesis was conducted using the Advantage RT-PCR kit (Clontech, 639506) for qPCR analysis. Samples were run using a Roche LightCycler 480 qPCR platform. Roche SYBR Green 1 Master mix (Roche 04887352001) was used for all qPCR. Relative fold change values were calculated in Microsoft Excel.
In situ hybridization. All in situ hybridizations were conducted as described below. Fixations were done in 1% gelatin-coated dishes to prevent tissue from sticking to the plastic (sticking to plastic causes tissue damage and non-specific staining). Embryos were fixed in ice cold 4% paraformaldehyde with 0.2% glutaraldehyde in 1/3× seawater for 2 min, followed by 4% paraformaldehyde in 1/3× seawater for 1 h at 4°C. Embryos were stored in 100% methanol at −20°C until used. Samples were rehydrated through a methanol series (75% methanol/25% PBS + 0.1% Tween 20, 50% methanol/50% PBS + 0.1% Tween 20, 25% methanol/75% PBS + 0.1% Tween 20) and then washed 3× in PBS + 0.1% Tween 20 (Ptw). Embryos were treated with 0.1 mg/ml Proteinase K in Ptw for 5 min, followed by 2× washes in Ptw + 2 mg/ml glycine. Embryos were moved to 24-well tissue culture plates, then incubated in 0.1% triethanolamine in Ptw for 5 min, followed by two washes with 0.1% triethanolamine in Ptw + 3 µl/ml of acetic anhydride. Samples were then washed 2× in Ptw, followed by a 1 h fixation in 4% PFA in Ptw at room temperature. Afterward samples were washed 5× in Ptw and then incubated overnight in hybridization buffer.
Probe sequences ranging from 550-1200 bps were cloned from cDNA using the pGEM ® -T vector system and DIG-labeled RNA probes were generated using MEGAscript in vitro transcription kits (Invitrogen, AM1333M). If available, cloned marker gene sequences were selected from a probe stock originally created for a previous study 35 . Probes were hybridized at 64°C for 2 days and developed with the enzymatic reaction of NBT/BCIP as substrate for the alkaline phosphataseconjugated anti-DIG antibody (Cat# 11093274910, Roche, Inc.). Wild-type samples were developed for an equal amount of time and if no expression was visible, a subset of samples remained in developing solution for at least 1 day to determine if lower levels of expression was present. When developing in situs of treated embryos, development time was based on the signal seen in control samples. All experiments exhibited greater than 50% penetrance of phenotypes, with 75% of the data containing greater than 75% penetrance. Due to variability during injection, embryo size and development time of different genes, greater than 50% in replicates, were considered the dominant phenotype created by genetic manipulations. Furthermore, to obtain material to test all the genes, samples were collected from multiple days of injection and pooled for in situ hybridization. Replicate attempts of in situ hybridization were attempted at least twice on pooled samples to obtain sample sizes generally greater than 30 embryos per replicate.
Immunostaining. Methods for immunostaining were adapted from previously published protocols 51 . Embryos collected for immunostaining were fixed in ice cold 4% paraformaldehyde with 0.2% glutaraldehyde in 1/3× seawater for 2 min, followed by 4% paraformaldehyde in 1/3× seawater for 1 h at 4°C. Samples were then washed five times (5 min each) in PBT (1% Triton-x and 1% BSA in PSS) and stored in PBS at 4°C for up to 1 month. Samples were then washed three times (30 min each) in PBT followed by a 1-h block in blocking solution (5% normal goat serum in PBT) at room temperature. Samples were then treated with the primary antibody (monoclonal anti-α-Tubulin; SigmaT9026) diluted 1:500 in blocking solution and incubated overnight (12 h) at 4°C. Primary antibody was removed and samples were rinsed five times (10 min each) in PBT before adding the secondary antibody (goat-anti-mouse-568; A-11004) diluted 1:250 in blocking solution and incubated over night at 4°C. Tissues were then washed at least three times (15 min each) in PBT before adding counterstain. DAPI was used (0.1 μg/µl in PBS) was used to label cell nuclei. Fluorescent phalloidin (Invitrogen A12379) was used at a concentration of 1:200 in PBT to stain filamentous actin, incubating for no more than 3 h at room temperature. Samples were washed at least three times (15 min each) in PBS before clearing. Samples were cleared in 80% glycerol or using Murray's Clear (1:1 benzyl benzoate and benzyl alcohol).
Microscopy and imaging. Results from in situ hybridization studies were imaged using a Zeiss Axio Imager Z1 with a Zeiss HRc color digital camera run by Zeiss Zen 2012 software. Fluorescently labeled embryos were imaged using a Zeiss LSM-710 confocal microscope. Images of polyp stage results and whole wells of 24-well plates (as pictured in Fig. 4 and Supplementary Figs 5-6) were imaged using A Zeiss SteREO Discovery.V8 stereoscopic microscope with a Cannon EOS 5D MarkII full frame DSLR camera. Images were processed and scale bars were added using the Fiji distribution of ImageJ (http://fiji.sc).
Data availability. The authors declare that all data supporting the findings of this study are available within the article and its Supplementary Information