PaxA, but not PaxC, is required for cnidocyte development in the sea anemone Nematostella vectensis

Background Pax genes are a family of conserved transcription factors that regulate many aspects of developmental morphogenesis, notably the development of ectodermal sensory structures including eyes. Nematostella vectensis, the starlet sea anemone, has numerous Pax orthologs, many of which are expressed early during embryogenesis. The function of Pax genes in this eyeless cnidarian is unknown. Results Here, we show that PaxA, but not PaxC, plays a critical role in the development of cnidocytes in N. vectensis. Knockdown of PaxA results in a loss of developing cnidocytes and downregulation of numerous cnidocyte-specific genes, including a variant of the transcription factor Mef2. We also demonstrate that the co-expression of Mef2 in a subset of the PaxA-expressing cells is associated with the development with a second lineage of cnidocytes and show that knockdown of the neural progenitor gene SoxB2 results in downregulation of expression of PaxA, Mef2, and several cnidocyte-specific genes. Because PaxA is not co-expressed with SoxB2 at any time during cnidocyte development, we propose a simple model for cnidogenesis whereby a SoxB2-expressing progenitor cell population undergoes division to give rise to PaxA-expressing cnidocytes, some of which also express Mef2. Discussion The role of PaxA in cnidocyte development among hydrozoans has not been studied, but the conserved role of SoxB2 in regulating the fate of a progenitor cell that gives rise to neurons and cnidocytes in Nematostella and Hydractinia echinata suggests that this SoxB2/PaxA pathway may well be conserved across cnidarians.


Background
Cnidocytes (stinging cells) are a cnidarian-specific cell type and an important diagnostic feature of cnidarians (jellyfish, corals, hydroids, etc.). Used in prey capture, defense, and locomotion, cnidocytes are epithelial cells that contain an extrusive organelle (the cnidocyst) that varies widely in both form and function across cnidarians. Two major types of cnidocyte are recognized: Nematocytes contain a piercing/penetrating organelle, whereas spirocytes and ptychocytes contain adhesive/ ensnaring organelles. While nematocytes are found in all cnidarians, spirocytes and ptychocytes are restricted to anthozoans (corals, sea anemones, etc.). This distribution of cnidocyte types across Cnidaria suggests that the ancestral cnidocyte was a nematocyte and that spirocytes and ptychocytes may have evolved by modification from this ancestral nematocyte in the lineage leading to Anthozoa [1]. While cnidocyte development has been fairly well characterized in hydrozoan cnidarians, other cnidarian lineages have received relatively little attention, making it difficult to find commonalities across cnidarians.
Studies of Hydra (Medusozoa:Hydrozoa) have revealed much about the cell dynamics of cnidogenesis in this organism [2,3]. In Hydra, nests of new nematocytes differentiate synchronously from a population of hydrozoan-specific progenitor cells called interstitial cells that are found throughout the mid-gastric region of the ectoderm. The cnidocyst develops in a post-Golgi vacuole during terminal differentiation of the cnidocyte from its interstitial cell progenitor [reviewed by 4]. The proteins that comprise the structural elements of the cnidocyst (tubule, harpoon, and capsule wall) are largely cnidocyte specific and include the minicollagens and nematogalectins [5][6][7][8]. These proteins undergo posttranslational modifications including cleavage from a preproprotein, alternative splicing, and extensive disulfide bonding to form the contiguous structure of the tubule/capsule [6,9,10]. Initial synthesis of the structures comprising the cnidocyst is followed by invagination of the eversible tubule and swelling of the capsule to generate the high intracapsular osmotic pressure that enables rapid discharge of the mature organelle [11]. This final step of cnidocyst maturation requires the synthesis of poly-γ-glutamate by enzymes found within the developing capsule [12][13][14]. The enzyme thought to be responsible for the synthesis of poly-γ-glutamate, γ-glutamyl transpeptidase (Ggt), has been isolated from discharged cnidocytes from hydrozoan, anthozoan, and cubozoan lineages, suggesting that Ggt may also be a ubiquitous marker of cnidocytes [7,13,15,16]. Further surveys of gene and protein expression in cnidocytes from across cnidarians have revealed that these novel cells are highly heterogeneous, expressing both novel (cnidarian-specific) and conserved genes [7,9,15,[17][18][19][20].
Cnidocyte development has received relatively little study in anthozoans, as compared with hydrozoans, and these studies have focused largely on the sea anemone, Nematostella vectensis [21,22]. N. vectensis is reported to have three types of cnidocyte (Fig. 1a-c): two types of nematocyte (basitrichous isorhizas and microbasic p-mastigophores) and spirocytes. Two different size classes of basitrichous isorhizas have been identified [23,24], though it is unclear whether size variation correlates with functional variation in this cell type. The distribution of cnidocytes in N. vectensis varies by tissue [23][24][25][26], and adult tissues in N. vectensis are populated by multiple different types of cnidocyte at the same time ( Fig. 1d-g). Examination of the spatiotemporal distribution of minicollagens in N. vectensis revealed that cnidocyte development begins early, coincident with the onset of gastrulation [22]. In contrast with hydrozoan models, Fig. 1 Type and distribution of cnidocytes in N. vectensis. a Microbasic p-mastigophores have a thick shaft which ends in a distinctive v-shaped notch (arrow). b Basitrichous isorhizas are variable in size (compare left and right DIC images) and have a thin shaft lacking a v-shaped notch and a tubule which is visible through the capsule (arrows). c Spirocytes can be identified by the regularly spaced coiled tubule and the lack of a visible capsule (arrow); a basitrichous isorhiza is shown for comparison (arrowhead). d, e The mesenterial filaments and pharynx are populated largely by microbasic p-mastigophores and large basitrichous isorhizas. f The body wall is populated by small basitrichs, and g the tentacle tips are populated by large basitrichous isorhizas (arrowheads in c, g) and spirocytes (arrows in c, g). Only nematocytes (basitrichous isorhizas and microbasic p-mastigophores) can be labeled using DAPI, indicated by yellow color in the cartoons (a, b). Scale bars in a-c represent 5 μm. All images are DIC micrographs cnidocytes in N. vectensis appear to differentiate locally, in the tissue where they are deployed, from a dispersed population of proliferating progenitor cells [27].
Transcription factors in the Sox (SRY-related HMG box) family are known to play diverse roles in the development and maintenance of cell identity [28]. A recent study of neurogenesis in N. vectensis revealed that transcription factor SoxB2, an ortholog of bilaterian SoxB genes [29], is expressed in a population of progenitor cells that give rise to both neurons and cnidocytes and that knockdown of SoxB2 results in loss of minicollagenexpressing cells [30]. The transcription factor Mef2 (myocyte enhancer factor 2) is also suspected to play a role in cnidogenesis in N. vectensis as knockdown of Mef2 splice variant IV (Mef2IV) also resulted in loss of minicollagenexpressing cells [31], but the relationship between SoxB2 and Mef2IV has not been characterized. Although several other genes have been suggested to play a role in cnidocyte development based on the timing and distribution of their mRNA expression, there have been no further studies of the gene regulatory network controlling cnidocyte development in N. vectensis.
Paired box (Pax) genes are a family of homeodomain transcription factors that play an important role in specifying sensory cell identity of diverse metazoan tissues [32]. Perhaps, the most well-studied among all Pax genes, Pax6, is known to play an essential role in both the development and maintenance of the photosensitive tissues including the eye in bilaterians [33]. The presence of Pax orthologs in ctenophores [34,35], sponges [36][37][38][39], and placozoans [40,41] and their absence from the choanoflagellate Monosiga brevicollis [42], suggests that this gene family evolved in the stem metazoan, long before the origin of eyes, though paired-like genes may have a more ancient origin [43]. Cnidarians possess orthologs of all Pax lineages found in bilaterians except Pax4/6 [42], suggesting that diversification of this gene family largely occurred before the common ancestor of cnidarians and bilaterians. Additionally, cnidarians appear to have undergone several lineage-specific duplication events, resulting in the cnidarian-specific PaxA and PaxC, and the multiple orthologs of PaxD found specifically in N. vectensis [44]. The function of Pax genes has been studied in several cnidarian taxa, both with and without eyes [41,42,45,46]. PaxB, an ortholog of Pax2/5/8, has been shown to play a vital role in the development of the camera-type eye in the cubozoan Tripedalia cystophora [45], and PaxA has been shown to regulate the development of the simple eyes of the hydrozoan Cladonema radiatum [42]. The genome of the scyphozoan Aurelia aurita encodes two Pax genes (PaxA and PaxB), neither of which seems to be involved in the development of the eyes in this species [47].
Seven Pax genes have been described from N. vectensis: PaxA, PaxB, PaxC, PaxD1-D4 [44,48]; based on their expression pattern, PaxA and PaxC have been hypothesized to play a role in the development of neurons and/or cnidocytes [44]. Using a combination of descriptive and functional techniques, we test the hypothesized role of PaxA and PaxC in the development of cnidocytes in N. vectensis. We show that PaxA is expressed exclusively in developing cnidocytes, that knockdown of PaxA results in decreased expression of many cnidocyte-specific genes, and that co-expression of Mef2IV in a subset of the PaxA-expressing cells defines a second lineage of developing cnidocyte. Together, these results suggest spatiotemporal variability in the expression of multiple conserved transcription factors may be responsible for generating cnidocyte morphological and functional diversity.

Cnidogenesis occurs continuously throughout the ectoderm in N. vectensis
To study the differentiation of cnidocytes in situ, we used a combination of markers that label distinct stages of cnidocyst development (Fig. 2). RNA probes were used to label cells undergoing transcription of minicollagen (Ncol), and antibodies were used to label cells that have progressed past the onset of NCOL protein translation, but have not yet completed polymerization of the developing capsule, at which time mature cnidocysts are no longer recognized by NCOL antibodies [22]. High-concentration (143 μM) DAPI was used to label poly-γ-glutamate in the matrix of the mature cnidocyst capsule [14]. Cnidocytes develop early during embryogenesis in N. vectensis, and new cnidocytes continue to develop exclusively in the ectoderm of all life stages (Fig. 2, 3). We confirm earlier reports that mature cnidocyst capsules first appear in the late planula stage of development and are abundant throughout the ectoderm at all stages thereafter ( Fig. 2a-f ) [21]. Cnidocyte differentiation, measured by the onset of expression of Ncol3 mRNA, however, begins much earlier [22]. Colocalization of Ncol3 mRNA and NCOL3 protein indicates that initial transcription of Ncol3 starts before the gastrula stage and continues throughout development in the ectoderm only ( Fig. 2g-t). In gastrula-stage embryos, early-stage cnidocytes expressing Ncol3 mRNA only and cells that have already begun translation of NCOL3 protein are both present throughout the ectoderm (Fig. 2g, h, m-o), suggesting that the onset of Ncol3 transcription occurs before gastrulation. In the early and late planula stages, many cells have downregulated expression of mRNA and are labeled with only α-NCOL3 antibody (Fig. 2i, j, p-r). Co-labeling with α-NCOL3 antibody and DAPI in the later stages (Fig. 2k, l, t) confirms that these two markers label distinct stages of cnidogenesis as no cnidocytes were ever labeled with both probes at the same time. Higher-magnification images illustrate the stages of capsule morphogenesis (Fig. 2m-p) and clearly show that developing cnidocytes progress through a curved U-or J-shaped stage as previously demonstrated for another sea anemone, Metridum senile [49]. Viewed from the surface of the ectoderm (perpendicular to the long axis of the developing cnidocyst), it is clear that cnidocytes develop asynchronously throughout the heterogeneous epithelium, such that adjacent cells are often at very different stages of cnidogenesis (Fig. 2q). Furthermore, although this is a simple epithelium composed of a single cell layer, early-stage cnidocytes appear shifted toward the basal side of the epithelium as RNA expression localizes basally (Fig. 2r, s). By contrast, late-stage cnidocytes vectensis. a-f Mature nematocyst capsules (yellow, DAPI) are first detected in the ectoderm in the late planula stage and are dense in the tentacle tips at the polyp stage (f, arrow). g-l Expression of Ncol3 mRNA (green) and NCOL3 protein (red) at the gastrula stage (h). New cnidocytes continue to develop in the embryonic ectoderm at all stages, even when mature cnidocytes (yellow, DAPI) are present in this tissue (k, l). m-s Multiple stages of cnidocytes are present in the ectoderm at the same time. Early-stage cnidocytes express Ncol3 mRNA only (m); later, many developing cnidocytes express mRNA and translated protein at the same time (n, o). Late-stage cnidocytes express protein only (P) and appear like hollow cylinders (arrows, q). r In the late planula, developing cnidocytes are dense in the ectoderm (Ec) adjacent to the blastopore (asterisk). Early-stage cnidocytes (black arrows) appear shifted toward the mesoglea side of the ectoderm (dotted line), and late-stage cnidocytes (white arrows) appear shifted toward the apical side of the ectoderm (solid line). (r corresponds to the boxed region in j.) s Cells expressing Ncol3 mRNA only continue to develop in the embryonic ectoderm, even in the presence of mature cnidocytes (arrows, s). (s corresponds to the boxed region in r). t NCOL3 protein (red) and DAPI (yellow) do not co-localize at any stage of development. u Cartoon summary of cnidocyte development as observed through Ncol3 and DAPI labeling. Colors correspond to the labeling in a-t. Nuclei are blue, DAPI. The blastopore/oral pole is oriented to the top unless otherwise indicated by asterisk for this and all subsequent figures. Images in a-l are 3D renderings from confocal z-stacks; images m-t are single optical sections lacking mRNA expression appear shifted toward the apical membrane as capsule maturation occurs apically. The different morphological stages of cnidogenesis and a stylized representation of the heterogeneous embryonic epithelium are presented in Fig. 2u, v. Using both α-NCOL3 antibody (Fig. 3a) and Ncol3 mRNA probe (Fig. 3b-e), we demonstrate that cnidogenesis continues into the polyp stage exclusively in the ectodermal component of polyp tissues: tentacles (Fig. 3b), pharynx (Fig. 3c), mesenterial filaments (Fig. 3d), and body wall (Fig. 3e). We also demonstrate a ring of mature cnidocytes encircling the physal pore (Fig. 3f ), an ephemeral opening in the aboral end of the polyp, which has not yet been described.

Minicollagens are spatially co-expressed but may vary temporally
The co-expression of minicollagens was examined by double-fluorescent in situ hybridization (dFISH; Fig. 4al), double-fluorescent immunohistochemistry (dFIHC; Fig. 4m-t), and a combination of the two techniques  Fig. 4v-ak). Ncol3 and Ncol4 mRNAs are co-expressed in all developing cnidocytes of the body wall and budding tentacles ( Fig. 4a-d); we therefore use Ncol3 and Ncol4 as ubiquitous markers of cnidocyte development throughout the rest of this study. Ncol1 mRNA appears to be expressed in only a portion of the cnidocytes expressing Ncol3 mRNA in the body wall and tentacle ectoderm (Fig. 4e-h), whereas Ncol1 and Ncol4 appear to be fully co-expressed in both tissues ( Fig. 4i-l). NCOL3 and NCOL4 proteins largely co-localize ( Fig. 4m-p), although several early-stage cnidocytes lack labeling for NCOL4 (Fig. 4n), while several late-stage cnidocytes lack labeling for NCOL3 (Fig. 4p). Similarly, NCOL4 and NCOL1 largely co-localize ( Fig. 4q-t) although several NCOL4-labeled late-stage cnidocytes seem to lack NCOL1 labeling (Fig. 4r). Using Imaris software (Bitplane, Concord, MA, USA), we counted the number of cells that were labeled with α-NCOL1, α-NCOL3, and α-NCOL4 Fig. 4 Minicollagen mRNAs are co-expressed, but the onset of NCOL protein localization is variable. a-d Ncol3 and Ncol4 mRNAs are co-expressed in all developing cnidocytes at the tentacle bud stage. e-h Ncol3 and Ncol1 mRNAs are also largely co-expressed, though cnidocytes that appear to be expressing Ncol3 exclusively can be seen in the ectoderm (arrows f, g). i-l Ncol4 and Ncol1 mRNAs are co-expressed in all cnidocytes. m-p NCOL3 and NCOL4 proteins largely co-localize, though early-stage capsules may express only NCOL3 (arrows n, o) and late-stage capsules may express only NCOL4 (arrowheads o, p). q-t NCOL1 and NCOL4 proteins also co-localize in all developing cnidocytes, though some late-stage capsules may express only NCOL4 (arrows r, t). u In gastrula-stage embryos, cnidocytes comprise approximately 10% of the cells, independent of which antibody is used (NCOL1: 9.39 ± 0.59, NCOL3: 11.01 ± 0.81, NCOL4: 9.11 ± 0.68; mean ± standard error). N = number of embryos analyzed. (V-AG) Co-labeling with mRNA probes and antibodies for the corresponding protein confirms that transcription is downregulated before cnidocyst capsules are fully polymerized as proteins can be detected in cells that no longer express mRNA for each antibody/probe combination (arrows w, aa, ae). ah-ak Co-labeling of Ncol3 mRNA with α-NCOL4 also reveals many late-stage cnidocytes labeled with α-NCOL4 only (arrows aj). Main images are 3D renderings from confocal z-stacks, high-magnification panels are single optical sections antibodies at the planula stage and found approximately equal numbers of cells were labeled with each antibody (Fig. 4u). Co-labeling with probe and antibody to detect mRNA and protein at the same time ( Fig. 4v-ak) confirms the morphological staging of cnidocyte development demonstrated in Fig. 2 for all three minicollagens.

Nematogalectin, not γ-glutamyl transpeptidase, is a ubiquitous marker of cnidogenesis
To identify other markers of terminally differentiated cnidocytes and compare the molecular nature of cnidocytes from anthozoans and hydrozoans, we searched the list of 51 nematocyte-specific genes identified from Hydra [20] for orthologs in N. vectensis. From this search, we identified two genes with expression patterns suggestive of a role in cnidogenesis: nematogalectin (Ngal; Nemve1|232015) and γ-glutamyl transpeptidase (Ggt; Nemve1|100407) ( Fig. 5). Both genes were expressed in scattered ectodermal cells of embryos of all stages following gastrulation (Fig. 5a-l). At later stages the expression patterns of these two genes differ: In addition to the abundant expression of Ngal in the tentacle buds and mature tentacles (Fig. 5e, f ), this gene is also expressed in the body wall ectoderm (inset, Fig. 5e), whereas Ggt expression is restricted to the tentacles in both stages (Fig. 5k, l). Ngal mRNA is co-expressed with Ncol3 mRNA in many, but not all cells (M-O); however, all cells expressing Ngal mRNA are also labeled with α-NCOL3 Expression of a-f nematogalectin (Ngal) mRNA and g-l γ-glutamyl transpeptidase (Ggt) mRNA during development. Onset of expression begins at/before the gastrula stage in scattered cells of the ectoderm (surface plane, insets) for both Ngal (b) and Ggt (h). Ngal is expressed in the body wall ectoderm throughout development (b-e, insets), and expression becomes dense in the developing tentacle buds (e) and tentacle tips of the polyp (f, arrow). Ggt is also expressed in the embryonic ectoderm (h-j) and presumptive tentacle buds but is absent from the body wall at later stages (k, inset) and is expressed in only few cells of the tentacle tips (l, arrow). m Ngal mRNA is co-expressed with Ncol3 mRNA in most developing cnidocytes in early-planula-stage embryos. n, o Few cells exhibit Ncol3 mRNA expression without Ngal (arrows), but p Ngal mRNA and NCOL3 protein are perfectly co-expressed. q-s Ggt mRNA is co-expressed with Ncol3 mRNA in many but not all developing cnidocytes. (Arrows in r-s indicate cells expressing Ncol3 only.) t Co-labeling of NCOL3 protein and Ggt mRNA confirms that some cnidocytes express Ncol3 without Ggt (arrows). Images in a-l are DIC micrographs, images m and q are 3D renderings from confocal z-stacks, and images n-p and r-t are individual optical sections antibody (Fig. 5p). These results suggest that Ncol3 and Ngal are both ubiquitous markers of cnidocyte development, but the former is transcribed earlier than the latter. Ggt and Ncol3 are also co-expressed in many, but not all, developing cnidocytes ( Fig. 5q-s), but co-labeling with Ggt mRNA and α-NCOL3 reveals a population of cells that express NCOL3 protein without Ggt mRNA.
Thus, Ggt appears to be expressed in only a subset of cnidocytes.

Knockdown of SoxB2 results in decreased expression of all cnidocyte markers
To characterize genes involved in the cnidogenesis pathway, we induced knockdown of SoxB2 using two nonoverlapping SoxB2 translation-blocking morpholinos (Table 1) and examined the effects on expression of cnidocyte-specific genes. Wild-type/uninjected embryos and embryos injected with a standard control morpholino (Ctrl MO; Table 1) express abundant NCOL4 protein in developing cnidocytes throughout the ectoderm at the gastrula stage ( Fig. 6a-b). By contrast, those injected with translation-blocking morpholinos directed against the SoxB2 start site (SoxB2 ATG MO) or a region of the 5′ untranslated region (SoxB2 5UTR MO) had far fewer developing cnidocytes in the ectoderm (Fig. 6c, d). Quantification of these patterns revealed that cnidocytes comprise approximately 9% of the cells in uninjected and control morpholino-injected embryos (8.97 ± 3.8, 9.11 ± 1.3, respectively) and approximately 4-5% of the cells in SoxB2 morphant embryos (ATG: 3.57 ± 1.84, 5UTR: 5.27 ± 0.33) (Fig. 6e). Using quantitative realtime PCR (qRT-PCR) in gastrula-stage embryos, we confirm that SoxB2 morphants exhibit a minor decrease in expression of two neural markers (RFamide and Elav) and one cnidocyte marker (Ncol3) as previously shown for the SoxB2 ATG MO [30]. Further, embryos injected with either SoxB2 ATG MO or SoxB2 5UTR MO exhibited significantly lower expression of all remaining cnidocyte markers examined (Ncol1, Ncol4, Ngal, and Ggt) compared to embryos injected with standard control morpholino (Fig. 6f ). Using in situ hybridization, we validate the results of qRT-PCR analysis and demonstrate that only few cells expressing Ncol1, Ncol4, Ngal, or Ggt remain in the ectoderm following injection of either SoxB2 morpholino, relative to embryos injected with control morpholino (Fig. 6g). Together, these results confirm the role of SoxB2 in regulating the differentiation of cnidocytes in N. vectensis, but suggest either some degree of inefficiency with SoxB2-mediated knockdown or, potentially, a SoxB2-independent cnidogenesis pathway, as we never observed a complete loss of cnidocytes in these experiments.

Knockdown of SoxB2 results in decreased expression of Mef2IV and PaxA, but not PaxC
Given that SoxB2 regulates the differentiation of both neurons and cnidocytes in N. vectensis [30], we aimed to identify transcription factors that were specific to the cnidocyte differentiation pathway downstream of SoxB2. Mef2IV was previously shown to be expressed in scattered ectodermal cells of the gastrula-stage embryo [50], and a subsequent study of its function in vivo demonstrated that knockdown of Mef2IV by translation-blocking morpholino resulted in a decrease in the number of Ncol3-expressing cells [31]. Similarly, the expression of PaxA and PaxC during embryogenesis in N. vectensis is consistent with a role for these two transcription factors in patterning cnidocytes [44]; however, the function of these two genes has not been studied. Embryos injected with control morpholino exhibited wild-type expression of Mef2IV, PaxA, and PaxC in gastrula-stage embryos, whereas embryos injected with either SoxB2 morpholino exhibited fewer Mef2IV-and PaxA-expressing cells (Fig. 7a). The number of cells expressing PaxC did not appear to change in SoxB2 morphant embryos. Consistent with this, qRT-PCR results demonstrate a significant downregulation of PaxA in both treatments and a small, nonsignificant increase in PaxC expression Table 1 Morpholinos used in this study a I1E2-targets the boundary between intron 1 and exon 2 (splice acceptor site) and E1I1-targets the boundary between exon 1 and intron 1 (splice donor site). ATG-targets the start codon. 5′UTR-targets the 5′ untranslated region. All morpholinos were designed by and purchased from GeneTools, LLC (USA)

Morpholino
Type Target  in both treatments (Fig. 7b). Curiously, Mef2 was downregulated in SoxB2 ATG morphants only, suggesting that the 5′ region of this transcript plays an important role in modulating transcription of the different splice variants. Thus, while Mef2IV and PaxA may be downstream of SoxB2, PaxC is likely expressed in a SoxB2-independent lineage of cells.

Mef2IV and PaxA, but not PaxC, are co-expressed in differentiating cnidocytes
We next examined the co-expression of Ncol3 mRNA with Mef2IV, PaxA, and PaxC (Fig. 8). Mef2IV (Fig. 8ac) and PaxA (Fig. 8d-f ) are co-expressed with Ncol3 in many developing cnidocytes, and co-labeling for both transcription factors shows that Mef2IV-expressing cells comprise a subset of the PaxA-expressing cells (Fig. 8gj), whereby all Mef2IV-expressing cells also express PaxA and all PaxA-expressing cells also express Ncol3, but many PaxA-expressing cells lack Mef2IV expression and many Ncol3-expressing cells lack PaxA expression. Co-labeling of Mef2IV and PaxA with NCOL3 protein confirms the presence of cells expressing NCOL3 protein that do not express Mef2IV or PaxA (Fig. 8k-r).
PaxC was never observed to be co-expressed with Ncol3 ( Fig. 8s-v) or PaxA (Fig. 8w-z). The morphology of the PaxC-expressing cells (elongated in the apico-basal axis; Fig. 8v) suggests these cells may be sensory cells. Finally, we examined the co-expression of SoxB2 and PaxA during cnidogenesis and determined that SoxB2 does not co-localize with PaxA at any time during development (Fig. 8aa-ad).

Knockdown of PaxA, not PaxC, results in loss of cnidocytes
To assess the influence of PaxA and PaxC on cnidogenesis, we knocked down endogenous Pax expression using splice-blocking and translation-blocking morpholinos (Fig. 9). Morpholinos (Table 1) were microinjected into fertilized eggs before first cleavage, and embryos were reared to the early planula stage for analysis. The PaxA splice-blocking morpholino (PaxA sp MO) was designed to block the splice acceptor site at the boundary between intron 1 and exon 2 (I1E2 boundary). We first confirmed that transcribed mRNAs were improperly spliced by PCR amplification and sequencing of the region affected by the morpholino. Using PCR, we amplified the full 1.8 kb PaxA coding sequence from cDNA isolated from embryos injected with control morpholino (Fig. 9a); the same primer set failed to amplify a product from embryos injected with PaxA splice-blocking morpholino. The PaxC splice-blocking morpholino (PaxC sp MO) was designed to block the E1I1 boundary, and when injected, this morpholino results in retention of intron 1, which encodes a stop codon 12 amino acids downstream of the E1I1 boundary. The morphant PaxC transcript is therefore longer than the wild-type transcript and encodes a wild-type paired domain but the homeobox is no longer in the open reading frame (Fig. 9b). Gel electrophoresis confirms the presence of both transcripts in morphant embryos. To further support the efficacy of splice-blocking morpholinos, we performed tandem knockdown experiments using translation-blocking morpholinos for both PaxA and PaxC. Knockdown of PaxA (via splice-or translation-blocking morpholino) results in a decrease in the number of developing cnidocytes in the ectoderm of the planula-stage embryo (Fig. 9c-e), but knockdown of PaxC appears to have little effect on cnidocyte abundance (Fig. 9f, g). Quantification of these results indicates that the abundance of cnidocytes in the ectoderm of morphant embryos falls from approximately 9% in embryos injected with control morpholino to 4-5% in PaxA morphants (Fig. 9h). Surprisingly, quantitative analysis of PaxC knockdown reveals a small but significant increase in cnidocyte abundance such that the ectoderm of PaxC morphant embryos comprised 10-12% cnidocytes.
To test the sufficiency of PaxA to induce ectopic cnidocytes and rescue morpholino phenotypes we microinjected exogenous PaxA mRNA alone and in combination with morpholinos and assayed the effects on cnidocyte abundance in the ectoderm (Fig. 9i). Embryos injected with control morpholino only, PaxA mRNA only, or a combination of the two did not differ in cnidocyte abundance, suggesting PaxA is not sufficient to induce ectopic cnidocytes. The abundance of cnidocytes also did not increase in embryos injected with a combination of PaxA mRNA and PaxA sp MO as compared with those injected with PaxA sp MO alone, and neither of these treatments differed from embryos injected with PaxA sp MO and control mRNA encoding histone H2B. Conversely, embryos injected with PaxA mRNA in combination with SoxB2 ATG translation-blocking morpholino experienced a partial rescue as cnidocyte abundance Finally, we examined the effect of PaxA knockdown on two newly identified markers of differentiated cnidocytes in N. vectensis-Ngal and Ggt (Fig. 10). We first demonstrate that PaxA is co-expressed with Ngal ( Fig. 10a-d) in developing cnidocytes, but some cells seem to express Ngal without PaxA (Fig. 10b-d).
Similarly, PaxA and Ggt are also largely co-expressed ( Fig. 10e-h) though cells expressing Ggt only can also be detected (Fig. 10f-h). Without antibodies directed against the proteins encoded by these genes, it is not possible to determine whether this variation is spatial or temporal. Knockdown of PaxA by splice-blocking morpholino reduces the number of cells expressing Ngal and Ggt (Fig. 10i) and results in decreased expression (via qRT-PCR) of Ncol1, Ncol3, Ncol4, Ngal, Ggt, and Mef2IV (Fig. 10j) relative to embryos injected with control morpholino. These results confirm that PaxA is required for wild-type expression of all cnidocyte markers identified thus far and suggest the cells expressing Ngal or Ggt in the absence of PaxA may reflect temporal differences in expression. The PaxC sp MO blocks the E1I1 boundary causing retention of intron 1. Both WT and morphant transcripts were amplified from PaxC MO-injected embryos. c-g Developing cnidocytes labeled with α-NCOL4 antibody in embryos injected with c control morpholino, d PaxA splice-blocking morpholino, e PaxA translation-blocking morpholino, f PaxC splice-blocking morpholino, and g PaxC translation-blocking morpholino. h Quantitative analysis of cnidocyte abundance in early-planula-stage embryos. Control morpholino-injected embryos exhibited wild-type abundance of cnidocytes, whereas PaxA sp morphants and PaxA tr morphants had significantly fewer (control MO: 9.11 ± 0.39, PaxA sp: 3.65 ± 0.63, PaxA tr: 4.90 ± 0.66; mean ± standard error). Both PaxC sp morphants and PaxC tr morphants exhibited an increase in cnidocyte abundance, albeit with a small effect size (PaxC sp: 10.58 ± 0.68, PaxC tr: 12.44 ± 0.47). i PaxA mRNA partially rescues SoxB2 ATG morphant phenotype but is not sufficient to induce ectopic cnidocytes. Asterisk indicates significant difference from Ctrl MO. Dagger indicates significant difference from SoxB2 ATG MO Our combined gene and protein expression data suggest that cnidogenesis in N. vectensis is both temporally and spatially dynamic (Fig. 11). The results of our temporal analysis of Ncol expression (summarized in Fig. 2u) and our spatial analysis of PaxA/Mef2IV co-expression (Fig. 8) suggest there are two hypotheses that are consistent with our observations. First, the developing cnidocytes we observed may comprise multiple distinct cell populations, differentiated by their unique molecular profiles. We detected cells in the gastrula-stage embryo coexpressing PaxA and Ncol, PaxA, Mef2IV, and Ncol, and cells expressing Ncol without PaxA or Mef2IV (Fig. 8a-j). We also detected late-stage cnidocytes (Fig. 2u, stage IV) expressing PaxA or PaxA and Mef2IV with NCOL protein (Fig. 8k-r). The simplest explanation for these results is that one population of cells upregulates PaxA before Ncol and that PaxA expression continues to be detectable through late capsule development (Fig. 11a, Population I), while a second population upregulates PaxA followed by Mef2IV, both of which remain detectable through late capsule development (Fig. 11a, Population II). The observation of Ncol-expressing cells that lack PaxA and Mef2IV expression suggests the presence of a third population of cells for which the transcription factor(s) that upregulate Ncol have yet to be identified (Fig. 11a,  Population III). Alternatively, the developing cnidocytes examined in this study may comprise a single, temporally dynamic population of cells, such that the various combinations of gene expression reported here mark different stages in the lifespan of a single developing cell (Fig. 11b). Two important observations make this hypothesis less plausible than the previous. First, we observed cells in the Mef2IV must be upregulated before Ncol, these results can only be explained by a downregulation of Mef2IV before downregulation of PaxA during early cnidogenesis (before stage IV of capsule development). Second, we detected late-stage cnidocytes expressing PaxA alone and those co-expressing PaxA/Mef2IV. The only way to attribute these observations to a single temporally dynamic population of cells is to infer that both PaxA and Mef2IV are expressed bimodally with two peaks of expression in early and late in cnidogenesis. Although this is possible, the presence of multiple cell populations seems a simpler explanation. At present, all cnidocytes are thought to arise by differentiation from a SoxB2-expressing progenitor cell that also gives rise to neurons (Fig. 11c). A committed population of cnidocyte-specific progenitor cells has not yet been identified, but the incomplete loss of cnidocytes (Fig. 6) and incomplete knockdown of cnidocyte-specific genes in response to SoxB2 and PaxA knockdowns (Figs. 6,9,10) suggest either that some cnidocytes develop independent of SoxB2, or that there are There is a single population of developing cnidocytes and all combinations of gene expression presented here result from temporal shifts in gene expression. In both cases, Ncol mRNA (purple) must be expressed before NCOL protein is detectable (blue) and, when expressed, PaxA (red) and Mef2IV (yellow) must be upregulated before Ncol mRNA. Horizontal lines in a, b represent an arbitrary threshold of detectable expression. c SoxB2-expressing progenitor cells give rise to both neurons and cnidocytes in N. vectensis. Post-mitotic neurons differentiate from cells expressing pro-neural genes, though the existence of "pro-cnido" genes (expressed in a progenitor that gives rise only to cnidocytes) is still only hypothesized (indicated by dotted lines). Three populations of cnidocytes may be present. Two populations differ only in the expression of Mef2IV and are hypothesized to become either basitrichous isorhizas and microbasic p-mastigophores or the two distinct size classes of basitrichous isorhizas. A third population expresses Ncol1, Ncol3, Ncol4, and Ngal, but the transcription factor that activates expression of these cnidocyte-specific genes is unknown. We hypothesize that this population of cnidocytes becomes spirocytes additional permissive steps in this pathway that remain to be characterized.

Discussion
Pax genes are known to play a role in the development of diverse sensory structures, but this study is the first to demonstrate a role for this conserved family of transcription factors in the development of a phylum-specific cell type. We show that PaxA, but not PaxC, is expressed in a population of developing cnidocytes that also express Ncol3 (Fig. 8) and that knockdown of PaxA results in loss of cnidocytes (Figs. 9, 10). Further, we demonstrate that knockdown of the neural progenitor cell-specific transcription factor SoxB2 results in loss of PaxA-expressing cells (Fig. 7). Together, these results suggest that PaxAexpressing cells differentiate from SoxB2-expressing cells to give rise to early-stage cnidocytes. PaxA has been shown to be expressed in tissues where cnidogenesis occurs in both hydrozoans [42,51] and scyphozoans [47,52], yet the potential role of PaxA in regulating cnidocyte development in other cnidarians has not been studied. Because they differentiate from the same population of progenitor cells that give rise to neurons, cnidocytes are considered by many to be "neural" cell types; perhaps then these data support an ancestral role for Pax genes in metazoan neural specification. Interestingly, PaxA and PaxC seem to have arisen from a cnidarian-specific duplication of the ancestral gene that gave rise to poxneuro in bilaterians [44,47]. If the ancestral function of this gene was neural specification, perhaps the duplication event that gave rise to PaxA and PaxC enabled the neofunctionalization of one of these paralogs for a novel role in patterning cnidocytes in the stem cnidarian. If true, PaxA may well play a role in the specification of cnidocytes across cnidarians. A recent study of the hydrozoan Hydractinia echinata demonstrated that SoxB2 is required for cnidogenesis [29]. Characterizing the shared and divergent parts of the cnidogenesis pathways in multiple divergent lineages of cnidarians would be a powerful way to identify the mechanisms that generate morphological and functional diversity in this cell lineage.
Besides PaxA, the function of the other Pax genes in N. vectensis remains largely unknown. The expression of PaxC in isolated ectodermal cells in two different species of anthozoans (N. vectensis and Acropora digitifera) led to early hypotheses that this gene may also be involved in neurogenesis [44,46,53]. The morphology of the PaxCexpressing cells in N. vectensis is consistent with the morphology of neurons (Figs. 7, 8); however, our functional experiments suggest that PaxC is not under the control of SoxB2 as PaxC expression did not change following knockdown of SoxB2 (Fig. 7). If these PaxC-expressing cells really are neurons, then SoxB2-expressing cells must comprise only a subset of the complete population of "neural" progenitor cells in N. vectensis. Finally, knockdown of PaxC resulted in a minor increase in the number of cnidocytes in the gastrula-stage embryo (Fig. 9), suggesting that the abundance of these two cell types (PaxC-expressing cells and developing cnidocytes) may be linked by some kind of negative feedback interaction. These observations are intriguing, and understanding the relationships among SoxB2, PaxC, and developing cnidocytes is likely to reveal a cryptic level of cell diversity in N.

vectensis.
A previous study of Mef2 function in N. vectensis demonstrated both a causal relationship between Mef2IV and Ncol3 expression in the planula stage and an earlier role (pre-gastrula) for maternally expressed Mef2 splice variants in patterning the endoderm [31]. In this study, we did not observe any Mef2IV-expressing cells that did not also express PaxA and Ncol and none of the Mef2IV-expressing cells were found in the endoderm (Fig. 8). These results are consistent with a cnidocytespecific role for Mef2IV in N. vectensis as has been suggested previously [50] and further demonstrate that Mef2IV regulates the development of only a subset of developing cnidocytes (Fig. 11). In the coral A. millepora, Mef2 is expressed in the blastopore during gastrulation, where it is thought to play a role in endoderm specification [54]; thus, both within and across species, the role of Mef2 in patterning cnidocytes seems to be variable. Like Pax genes, orthologs of Mef2 are present in the genomes of ctenophores [55], sponges [36], and placozoans (NCBI accession: XP_002110582.1), but the function of this transcription factor in non-bilaterian metazoans is largely undescribed. If Mef2 were shown to play a role in regulating the development of the muscle cells in ctenophores, this could support the hypothesized homology of muscle cells across metazoa. Independent of its ancestral function, Mef2 appears to have been co-opted to play a role in cnidocyte specification in N. vectensis. Unlike Pax genes, however, Mef2 functional diversification appears to be the result of alternative splicing rather than duplication/divergence, as this gene is present in only a single copy across non-bilaterian genomes.
Quantitative investigation of the co-expression of PaxA, Mef2IV, and Ncol3 suggests a cryptic level of cnidocyte diversity in N. vectensis. Zenkert et al. [22] reported that over 90% of the cnidocytes present in planula-stage larvae are basitrichous isorhizas. In this study, many, but not all, of the Ncol3-expressing cells also expressed PaxA (Fig. 8d-f ) and approximately half of the cells expressing PaxA also expressed Mef2IV (Fig. 8g-j). These data are consistent with three distinct molecular fingerprints for different lineages of developing cnidocytes in N. vectensis (Fig. 11a) and suggest there may be variation that has gone unnoticed in previous studies of cnidocyte development in this species. Careful examination of the mature cnidocytes from larvae may in fact reveal morphological variation that correlates with the molecular variation we describe here.
Beyond the role of Pax genes in cnidocyte specification, we have demonstrated that the cell dynamics of cnidogenesis in N. vectensis differs in many ways from Hydra and other hydrozoan model systems. Using markers of different stages of cnidogenesis, we demonstrate that cnidocytes develop individually and asynchronously in the ectoderm of N. vectensis in all life stages (Figs. 2, 3). It is, therefore, common to find early-and late-stage cnidocytes adjacent to each other in the same epithelium at the same time in N. vectensis. This is in contrast with the mode of cnidocyte development in hydrozoans, whereby groups of cnidocytes differentiate synchronously from a centralized population of progenitor cells before migrating to the tissue in which they will be deployed [56]. As such, cnidocyte development is highly position dependent in Hydra with progressive stages of developing cnidocytes occurring at regular and predictable distances from the progenitor cell population [57,58]. This is not the case in N. vectensis where progenitor cells are scattered throughout the epithelium in which new cnidocytes are differentiating individually [27]. Furthermore, we have shown that Ncol1, Ncol3, and Ncol4 are all co-expressed in all cnidocytes in N. vectensis (Fig. 4), rather than being spatially restricted as previously reported. Our data suggest that the differences in minicollagen expression across cnidocyte types may be temporal, as differences in mRNA expression are not corroborated by differences in protein expression. While transcription begins at approximately the same time (as evidenced by dFISH), the timing of translation appears to vary slightly. In particular, it appears that NCOL3 may undergo translation slightly earlier than NCOL4 or NCOL1 and that the latter two may continue to be detectable by α-NCOL antibodies even after NCOL3 is not.
Minicollagens are known to be essential for cnidocyte development across cnidarians, but several recent studies suggest there are additional features that may be shared across cnidocytes. Indeed, chondroitin sulfate has been shown to play a role in scaffolding minicollagens during capsule construction in both Hydra and N. vectensis [59]. Furthermore, Ngal is known to be expressed in cnidocysts from hydrozoans [6,20], cubozoans [15], anthozoans (this study), and even the polar capsules of myxozoans [8]. Thus, Ncol, Ngal, and chondroitin sulfate together seem to comprise a minimal kernel of genes necessary for synthesis of the cnidocyst across cnidarians. Ggt was identified as one of the nematocyte-specific genes in Hydra [20] and was reported to be expressed in the ectodermal tissue of the tentacles and body wall of the sea anemone M. senile [60]. This enzyme appears to have evolved before the origin of cnidocytes as many orthologs are present outside of eukaryotes [61], and therefore may have been co-opted to play a role in generating the extreme osmotic pressure that enables nematocysts to fire with such force [12]. Interestingly, the matrix of spirocytes does not generate the same intracapsular pressure as nematocysts and, likewise, does not react with DAPI, suggesting Ggt may not be (as) active during development of spirocytes. It is possible, then, that the lineage of cnidocytes we identified as expressing Ncol3 and Ngal but not Ggt could be developing spirocytes. Alternatively, a study of poly-γ-glutamate in the sea anemone M. senile suggested that the levels of this amino acid are highest in microbasic b-mastigophores in this species [62]. Given that Ggt expression was not detected in the body wall of tentacle bud stage animals but was detected in the tentacles (Fig. 3), an alternative hypothesis is that the cells expressing Ncol3 and Ngal without Ggt may be the small size class of basitrichs in N. vectensis. Chemical inhibition of γ-glutamyl transpeptidase activity and/or knockout of the Ggt gene in vivo would therefore be expected to result in animals with only spirocytes and small basitrichs. This experiment would be a valuable next step in understanding the generation of cnidocyte subtypes.

Conclusions
Expression of minicollagen is the unifying molecular feature of cnidocytes, and cnidocyte diversity has been suggested to result from differences in the spatiotemporal expression of various minicollagen paralogs during cnidogenesis [1,63]. Interestingly, all minicollagens appear to be co-expressed in N. vectensis, but only a subset of cells express minicollagen under the control of PaxA. Indeed, we show that two conserved families of transcription factors are co-expressed in only a subset of developing cnidocytes, suggesting the combined expression of many conserved genes may be sufficient to generate diversity in a novel cell lineage. This system serves as a valuable model for exploring the relationships between the secreted or structural proteins that impart cell identity (e.g., minicollagen and nematogalectin) and the transcription factors responsible for initiating expression of these genes. After the origin of the ancestral cnidocyte, diversity in this cell lineage may have been generated as different classes of conserved transcription factors were recruited to activate the expression of novel structural genes. Examining the diversity of transcription factor families known to activate expression of cnidocyte structural genes across cnidarians will be an important contribution to understanding how diversity evolved in this novel cell lineage.

Animal care and tissue collection
Adult sea anemones were maintained in the dark at 17 °C and spawned as described previously [64]. In vitro fertilization was performed at room temperature (25 °C), and fertilized egg masses were de-jellied by agitation in 1/3X filtered seawater (FSW) containing 4% l-cysteine (C-7352, Sigma, USA). Liberated zygotes were washed three times in clean 1/3X FSW to remove excess cysteine and reared at 17 °C or 25 °C in the dark. At the time of collection, embryos were immobilized in 7% MgCl 2 for up to 10 min at 25 °C, fixed in 4% paraformaldehyde (PFA) with 0.2% glutaraldehyde for 1 min at 25 °C, and then fixed in 4% PFA without glutaraldehyde for 1 h at 4 °C. For in situ hybridization, fixed tissues were washed three times in phosphate-buffered saline containing 0.1% Tween 20 (PTw), washed once in sterile 0.2-μm-filtered autoclaved deionized water (sH 2 O), washed once in 100% methanol, and transferred to clean 100% methanol for storage at −20 °C. For immunohistochemistry, tissues were fixed in a similar manner, washed three times in PTw, and stored at 4 °C in clean PTw before processing. Embryos subjected to ISH followed by IHC analysis were fixed as described for ISH embryos. Staging of embryos is indicated for development at two different temperatures ( Table 2).

Cloning and probe synthesis
To generate a wild-type cDNA library, we extracted total RNA from embryos at multiple stages of development: blastula, gastrula, early and late planula, tentacle bud, and primary polyp stages. Pooled embryos were homogenized in TRI reagent (T9424, Sigma, USA) with a mechanical pestle for 30 s at 25 °C, and total mRNA was extracted following the manufacturer's instructions for TRI reagent. cDNA synthesis was performed following the manufacturer's instructions for the Advantage RTfor-PCR kit (639506, Clontech, USA), and samples were stored at −20 °C until use. To generate RNA probes for in situ hybridization, target sequences (approximately 1000 bp each) were amplified from our cDNA library using a standard PCR protocol, separated by gel electrophoresis on a 1% agarose gel, isolated using the QiaQuick gel extraction kit (28704, Qiagen, USA), and cloned into the PGEM-T vector (A3600, Promega, USA) following the manufacturer's protocols. Targets were PCR-amplified from plasmids using standard SP6 and T7 primers, gel-purified (as above), and used as template for in vitro transcription using the Megascript SP6 or T7 transcription kit (AM1330/AM1334, Ambion, USA). Purified mRNA was solubilized in nuclease-free water and then diluted to a stock concentration of 100 ng/μl in hybridization buffer for storage at −20 °C. At the time of use, probes were further diluted to a working concentration of 1 ng/μl in hybridization buffer and heated to 90 °C for 5 min before application to tissues.

In situ hybridization and immunohistochemistry
In situ hybridization (ISH) was performed as previously described for single-color ISH [65] and for double-fluorescent ISH [66]. Immunohistochemistry (IHC) assays were performed as described previously [64] using Nematostella vectensis-specific antibodies directed against minicollagens 1, 3, and 4 (α-NCOL1, α-NCOL3, and α-NCOL4) which were originally designed and validated by Zenkert et al. [22]. Nuclei were counter-stained for 30 min at 25 °C with 1.43 μM DAPI, and mature cnidocytes were labeled by incubation in 143 μM DAPI (in PTw with 10 mM EDTA) for 30 min at 25 °C. Confocal imaging was performed on a Zeiss LSM 710 microscope, and assessments of cnidocyte abundance were performed in 3D reconstructed images produced from confocal z-stacks. Counting was automated using Imaris software (Bitplane, Concord, MA, USA), and the proportion of embryonic cells comprised of cnidocytes was calculated as the ratio of cells labeled with antibody to nuclei labeled with DAPI. Differences among treatments were evaluated using ANOVA in the R statistical computing environment [67].

Manipulation of gene expression
Splice-and translation-blocking morpholinos were purchased from Gene Tools, LLC (Philomath, OR, USA), reconstituted in nuclease-free water to a final concentration of 1 mM, and stored as directed by the manufacturer at 25 °C in the dark before use. Just prior to microinjection, morpholinos were heated to 60 °C for 5 min and centrifuged for 1 min before being diluted to a final working concentration with 0.2 mg/ml Alexa Flour-555 conjugated dextran (Invitrogen, USA) in nuclease-free water. Morpholinos were microinjected into uncleaved zygotes following the protocol of [68], and embryos were reared to the time of collection at 25 °C in the dark. For splice-blocking morpholinos, disruption of proper splicing was confirmed using PCR and gel electrophoresis with cDNA synthesized from morphant embryos and primers designed to amplify the sequence targeted by the morpholino (see Fig. 9; Table 3 for primer sequences). To assay the effects of ectopic mRNA expression, the complete PaxA coding sequence was cloned upstream of the fluorescent reporter Venus and purified RNA was transcribed from this construct using the mMessage mMachine kit (AM1348, Ambion, USA) following the protocols of Roure et al. [69] and Rottinger et al. [70]. To confirm the specificity of the PaxA::Venus construct in rescuing the morpholino phenotypes, we also injected SoxB2 ATG MO and PaxA sp MO in combination with a publicly available human histone H2B::GFP construct (NCBI AB591038) [71]. Both constructs were microinjected alone, or mixed with morpholino before being injected into fertilized eggs, and embryos were assessed visually for translation of reporter proteins before being reared for additional assays. The effects of in vivo gene manipulation on the expression of putative target genes were assayed using ISH/IHC (above) or by qRT-PCR.
In short, qRT-PCR was performed in morphant cDNA samples collected from multiple independent microinjection experiments per morpholino (N = 5 for control morpholino, N = 2 for PaxA tr MO, and N = 3 for all other morpholinos). Data are reported as mean expression values of target genes in target morpholino-injected embryos relative to mean expression values in control morpholino-injected embryos. The expression of elongation factor 1β (Ef1B) in target and control embryos is set to 1.0, and all other expression values are reported relative to this on a log 2 scale. Differences in relative mRNA expression among target genes were assessed using ANOVA in R [67].