Recently, the whole genome sequence for Mnemiopsis leidyi was published and became publicly available 17 . Sox genes from non-bilaterian species and human were used in TBLASTN and BLASTP searches of the genome assembly, gene models, and protein models (version 2.2) of the Mnemiopsis leidyi genome, which are available through the Mnemiopsis Genome Project Portal (http://research.nhgri.nih.gov/mnemiopsis). We retrieved seven putative Sox sequences from these searches. After verifying the sequences via RACE-PCR (see ‘Animal collection and in situ hybridization’), the sequences were deposited in GenBank (Accession Numbers KJ173818-KJ173824). In some cases, the final deposited sequence differed from the predicted gene model. Here is a list of how the gene model IDs correspond to the deposited MleSox gene sequences: MleSox1 (KJ173818) = ML047927; MleSox2 (KJ173819) = ML234028; MleSox3 (KJ173820) = ML042722; MleSox4 (KJ173821) = ML06932; MleSox5 (KJ173822) = ML23337; MleSox6 (KJ173823) = ML01787; MleHMG-box (KJ173824) = ML040423.

The dataset was compiled using the available Sox gene complement from all non-bilaterian species plus selected bilaterian species. Anthozoan cnidarians were represented by the set of 14 Sox genes from the sea anemone Nematostella vectensis 33 and six additional published sequences from the coral Acropora millepora 34 . We added a set of 14 Sox genes from the hydrozoan cnidarian Hydra magnipapillata and 10 Sox genes from Clytia hemisphaerica that were previously described 35 . For sponges, we included four sequences from the demosponge Amphimedon queenslandica 11 and three from the demosponge Ephydatia muelleri, plus seven from the calcareous sponge Sycon ciliatum 12 . Sox homologs from non-bilaterian and bilaterian species were used in TBLASTN and BLASTP searches of available genome assemblies and predicted gene models of non-animal eukaryotic phyla, specifically the choanoflagellates Monosiga brevicollis and Salpingoeca rosetta. The filtered protein models for Monosiga v 1.0 36 were downloaded from the Joint Genome Institute genome website. Gene models for S. rosetta were downloaded from the Origins of Multicellularity Sequencing Project at the Broad Institute (https://www.broadinstitute.org/annotation/genome/multicellularity_project/). A set of non-Sox HMG domains from the Tcf/Lef family was used as an outgroup. The 79 amino acid HMG-box domains of the seven putative Mnemiopsis Sox genes, two M. brevicollis Sox-like genes, and two S. rosetta Sox-like genes were aligned to known Sox homologs automatically using MUSCLE 37 . This alignment was used to perform preliminary phylogenetic analyses. Final analyses were done on an alignment that did not include the M. brevicollis Sox-like, S. rosetta Sox-like, or MleHMG-box sequences (Additional file 1). The only missing data were 11 N-terminal amino acids from the HMG-box for the following sequences: CheSox1, EmuSox1-3, and PpiSox2, PpiSox3, and PpiSox12. The tree was based on 136 HMG-box sequences. A second alignment was constructed without the Hydra sequences and was used to examine the effects these sequences had on the overall tree topology.

To choose the best-fit model of protein evolution, we used the program ProtTest v2.4 to apply a variety of possible substitution matrices and rate assumptions 38 . The results from this indicated that the best model for the alignment was LG + Γ, where ‘LG’ indicates the substitution matrix 39 , and ‘Γ’ specifies gamma-distributed rates across sites. Maximum likelihood analyses were performed with the MPI version of RAxML v7.2.8 (RAXML-HPC-MPI) 40 . We conducted four independent searches with a total of 235 randomized maximum parsimony starting trees and then compared the likelihood values among all result trees. For complex datasets, it is often necessary to perform multiple search replicates to find the same best tree multiple times to provide confidence that the tree topology with the best likelihood has been found. We found this to be the case with this dataset. One hundred bootstrapped trees were computed and applied to the best result tree. ML bootstrap values are indicated on the ML tree (Figure 1).

Bayesian analyses were performed with MrBayes3.2 41 . MrBayes does not support the LG model of protein evolution, so we used the second best fit model from ProtTest (RtRev + Γ). Initially, we did two independent five million generation runs of five chains each, with trees sampled every 100 generations. We found that using these parameters, the ‘Average standard deviation of split frequencies’ between the two runs was 0.0148. This diagnostic value should approach zero as the two runs converge and an average standard deviation below 0.01 is a very good indication of convergence, while any value between 0.01 and 0.05 is considered acceptable for convergence. We then did two independent five million generation runs of nine chains each, with trees sampled every 100 generations and a heating parameter of 0.05 (default heating is 0.2) and achieved an average standard deviation of split frequencies of 0.0101. We also ran MrBayes with the ‘mixed’ amino acid model option (prset aamodelpr = mixed) using the same parameters and found no difference in the convergence diagnostic value or in the resulting tree compared to the tree generated with the RtRev + Γ model. Additional convergence diagnostics, examined with the help of AWTY 42 , indicated a conservative burn-in of 0.25. The runs reached stationarity, and adjusting the burn-in did not affect the topology. A majority rule consensus of 37,500 trees was produced and posterior probabilities were calculated from this consensus. Trees were rerooted in FigTree v1.3.1 43 . Bayesian posterior probabilities are shown on the Bayesian tree (Additional file 2: Figure S1).

Mnemiopsis leidyi adults were collected from Eel Pond or the NOAA Rock Jetty, Woods Hole, MA, USA, during the months of June and July and spawned as previously described 44 . RNA was extracted from embryos with TRI Reagent (Molecular Research Center, Cincinnati, OH, USA) and reverse transcribed to generate cDNA (SMART RACE cDNA Amplification Kit, Clontech Laboratories, Inc., Mountain View, CA, USA). This cDNA was used as a template to isolate the genes of interest. Individual RACE-PCR products were cloned and sequenced, and sequences were aligned to the genomic sequence.

For whole-mount in situ hybridization, embryos were fixed at various stages from freshly collected nucleated embryos (0 hpf) to newly hatched cydippids (24 hpf). They were stored in methanol at -20°C until used. Sequences, ranging in length from 650 to 2,000 bp, were used to transcribe digoxigenin-labeled RNA probes (Ambion/Applied Biosystems, Austin, TX, USA). These probes were hybridized for 48 h at 60°C and detected using an alkaline phosphatase-conjugated digoxigenin antibody (Roche Applied Science, Indianapolis, IN, USA), and the substrates nitro-blue tetrazolium (NBT)/5-Bromo-4-chloro-3-indolyl phosphate (BCIP). After detection, specimens were washed with phosphate-buffered saline (PBS) and transferred through a glycerol series up to 70% glycerol. They were then mounted, viewed under a compound microscope (Zeiss AxioSkop 2), and imaged using a digital imaging system (AxioCam HRc with Axiovision software, Zeiss, Thornwood, NY, USA). Color balance and brightness were adjusted using Photoshop software (Adobe Systems Incorporated, San Jose, CA, USA). Additional details of the in situ hybridization protocol for Mnemiopsis have been previously described 44 . All in situ images presented here are available online via the comparative gene expression database, Kahikai (http://www.kahikai.com).

EdU (ethynyl deoxyuridine) is a uridine analog similar to BrdU. To measure cell proliferation, cydippids were fixed and processed for fluorescent detection of incorporated EdU using the Click-iT EdU labeling kit (Invitrogen, Carlsbad, CA, USA), which incorporates EdU in cells that are undergoing the S phase of mitosis. Specifically, cydippids aged 18 to 24 h were incubated in EdU labeling solution for 15 to 20 min and then fixed using 4% paraformaldehyde with 0.02% glutaraldehyde for 30 min. After three washes in PBS, they were stored in PBS at 4°C until subsequent use. Prior to the Click-iT reaction, cydippids were washed for 20 min in PBS plus 0.2% Triton. The Click-iT reaction was performed according to manufacturer instructions, using the Alexa-488 reaction kit. To visualize nuclei, cydippids were also stained with Hoechst 33342 (Invitrogen, Molecular Probes). Cydippids were mounted in PBS, examined, and imaged under a Zeiss Axio Imager or LSM710 confocal microscope.

We identified six members of the Sox family from the Mnemiopsis leidyi genome, all with complete HMG-box domains. A seventh sequence with an HMG-box domain (MleHMG-box) did not fall within the Sox gene family in our preliminary phylogenetic analyses and was excluded from our final alignments and trees. Phylogenetic analyses of the six Mnemiopsis Sox sequences, combined with all previously published non-bilaterian Sox sequences and several representative bilaterian Sox sequences, constructed the metazoan-specific Sox family phylogeny, including the major known groups (B, C, D, E, and F; Figure 1). From this analysis, five Mnemiopsis Sox genes were classified into four groups (B, C, E, and F), with an additional gene (MleSox6) branching at the base of the E and F groups (Figure 1). According to the tree reconstruction, MleSox1 belongs to group B, MleSox2 belongs to group C, MleSox3 and MleSox4 branch within group E, and MleSox5 is found within group F. Each of the Mnemiopsis Sox genes has a clear ortholog in the ctenophore Pleurobrachia, although the two SoxC genes PpiSox2 and PpiSox12 seem to be the result of a lineage-specific duplication within Pleurobrachia and MleSox2 is sister to these two sequences.

As observed in other recent studies 12 35 , a number of the non-bilaterian Sox sequences could not be classified into any of the previously identified major Sox groups (Figure 1), including two ctenophore sequences (MleSox6 and PpiSox4) and two sponge sequences (AquSoxF and EmuSox1) that branch at the base of the E and F groups. Several cnidarian Sox sequences from various species (Acropora millepora, Clytia hemisphaerica, Hydra magnipapillata, and Nematostella vectensis) also could not be classified into the traditional groups, including a group of 14 cnidarian sequences that fall within their own clade in the Sox family (Figure 1). This group includes the nematode CelSox4 gene. To test the possible effects of long-branch attraction due to inclusion of some of the Hydra sequences, we constructed separate trees that did not include any Hydra Sox sequences, but found the same overall tree with only minor rearrangements of branches (data not shown). As noted in other phylogenetic analyses of the Sox HMG-box 13 35 , low statistical support of the major clades likely stems from the short sequence length used for the analyses and the inclusion of a large number of taxa sampled across a wide evolutionary distance.

The trees generated from the maximum likelihood (ML; Figure 1) and Bayesian (Additional file 2: Figure S1) analyses have the same overall topology; there are only a few individual branches that differ between the two trees (specifically, HsaSox30, CheSox2, HmaSox3, TadSox1, SciSox6, and SciSoxE). Overall, our trees (Figure 1; Additional file 2: Figure S1) were in general agreement with other recent surveys of non-bilaterian Sox genes 12 35 , with a few notable exceptions, denoted in bold text in Table 1. A previous analysis of the Sox complement from the calcareous sponge Sycon placed SciSoxE in the SoxE group, SciSoxF1 and SciSoxF2 in the SoxF group, and was unable to classify two other genes (SciSox6 and SciSox7) into any known group 12 . In contrast, our Bayesian analyses consistently place SciSoxE in an unclassified position at the base of the SoxE and SoxF groups, while our ML analyses place it in the SoxE group, calling into question whether sponges have a clear SoxE homolog. Neither of our analyses placed any sponge sequence in the SoxF group. Three Sycon genes (SciSox7, SciSoxF1, and SciSoxF2) branch next to the exclusively bilaterian SoxD group in both of our analyses, albeit with low ML bootstrap support and a low Bayesian posterior probability. The branch uniting the Sycon sequences that places them next to the SoxD clade was unstable in both of our analyses, based on post-tree analysis using PhyUtility 45 , a program that calculates branch attachment frequency and leaf stability metrics. Therefore, it is unclear whether these genes are truly related to SoxD genes, whether this was an artifact of tree reconstruction methods, or whether this may be due to possible sequence convergence. Looking across all result trees from all of our analyses, we see that the Sycon sequences previously classified as SoxF (SciSoxF1 and SciSoxF2) occur together in a position either next to the SoxD group (as seen in Figure 1) or in an unclassified position at the base of the tree in over 90% of trees. Fewer than 10% of our result trees place these two sequences in an unclassified position at the base of the SoxE and SoxF groups together with AquSoxF and EmuSox1. We did not observe the placement of these Sycon sequences in any known group in any of our result trees, regardless of the tree construction method or sequences included.

An asterisk denotes genes with published in situ expression patterns. Gene names in bold text were previously classified differently 12 35 .

^a Not represented in trees because only short partial sequence is available (20 aa missing from HMG-box).

^b Not represented in final trees or counted in the total as preliminary analyses indicated that these are not likely to be true Sox family genes.

^c Classification sensitive to tree search method used in this study, with classification from ML analysis shown.

^d Partial sequence (11 aa missing from 5′ end of HMG-box).

In our trees, a cluster of five paralogous Hydra Sox genes are located in the SoxF group, while previous analyses concluded that the SoxF group had apparently been lost from this lineage 35 . This placement was consistent across all of our result trees, regardless of the tree construction method or the sequences included. The Clytia Sox study 35 placed four hydrozoan Sox sequences (CheSox13, CheSox14, HmaSox1, and HmaSoxBb) and two anthozoan sequences (NveSox3 and AmiSox3) in the SoxB group, while in our trees, these sequences consistently fell in the unclassified group of 14 cnidarian sequences plus the nematode CelSox4. We have summarized our classification of all non-bilaterian Sox genes based on our ML analysis in Table 1.

A previous study identified a putative Sox gene from the choanoflagellate Monosiga brevicollis 36 . We identified two Sox-like sequences from the M. brevicollis genome (Joint Genome Institute ID: 12602, 12133), as well as two Sox-like sequences from the genome of another choanoflagellate, Salpingoeca rosetta (Broad Institute ID: PTSG_01623.1, PTSG_02101.1). In our preliminary analyses, however, these sequences, together with the Mnemiopsis MleHMG-box gene, always clustered together outside the Sox gene family with outgroup sequences, suggesting that they are not true Sox genes. We excluded these sequences from our final alignments and trees but include them in Table 1. Our result is in agreement with a recent in-depth study of transcription factors in the genome of the unicellular holozoan Capsaspora owczarzaki and its close relatives 10 . In that study, the authors found that HMG-box transcription factors arose early in eukaryotic evolution, followed by ‘Sox-like’ HMG-box genes, which arose in the ancestor to choanoflagellates (after the lineage leading to C. owczarzaki diverged), followed by the evolution of Sox and Tcf/Lef HMG-box families at the base of the animals. Further study of the choanoflagellate and ctenophore ‘Sox-like’ sequences will help to clarify the origin and possible functions of these genes.

Two ctenophore Sox sequences (MleSox1 and PpiSox3) fall into group B, within a subclade of exclusively bilaterian Sox sequences that includes three human paralogs (HsaSox15, HsaSry, and HsaSox3). Jager et al. 35 pointed out a similar placement of the PpiSox3 gene in their Sox phylogeny and highlighted the evolutionary implications of this placement, including the possibility that other non-bilaterian orthologs were lost from this subclade or that the placement of the ctenophore Sox group B sequences in this position may be an artifact of tree reconstruction methods or due to possible sequence convergence.

Within group C, there is a non-bilaterian clade consisting of sponge, cnidarian, and placozoan sequences. Three ctenophore SoxC sequences (MleSox2, PpiSox2, and PpiSox12) form a cluster with a sequence from the chordate Branchiostoma floridae (BflSox5) that falls next to a cluster with three human sequences (HsaSox4, HsaSox11, and HsaSox12), one sequence from Ciona intestinalis (CinSoxC), and one sponge sequence (SciSoxC).

A bilaterian SoxF subgroup was recovered in all analyses and included a single non-bilaterian member, CheSox11 from Clytia. Two sister subgroups within the overall SoxF group contain the remaining non-bilaterian sequences. One subgroup has a cluster of five Hydra sequences and a single Nematostella sequence (NveSox7). The other subgroup includes two ctenophore sequences (MleSox5 and PpiSox8), a Nematostella sequence (NveSoxF1), and AmiSoxF from Acropora.

Group E Sox genes include a subgroup of four ctenophore sequences (a set of paralogs from Mnemiopsis, MleSox3 and MleSox4; and another set from Pleurobrachia, PpiSox1 and PpiSox6). This subgroup is found within a larger group of bilaterian and non-bilaterian SoxE sequences. A branch with two unclassified ctenophore Sox sequences (MleSox6 and PpiSox4) falls at the base of Group E and Group F (Figure 1). In a previous study 35 , PpiSox4 was located in the same unclassified position.

Expression of MleSox1, a member of the SoxB group, was not detected by in situ hybridization before or immediately after gastrulation (which occurs around 4 hpf). Light expression is seen in the developing embryo around 7 hpf, around the blastopore, in cells that invaginate to form the pharynx in the cydippid (Figures 2A and B). Expression in a patch of cells in the pharynx can later be seen in the cydippid (Figures 2C). Expression at 7 to 13 hpf is also found in epidermal cells that later form the comb plates (Figures 2A and B); the expression of these epidermal cells expands along the body column as the embryo develops (Figures 2B and E) but then becomes very light and is restricted to the uppermost part of the comb rows in the cydippid (visible in Figure 2F but not in 2C). Under the epidermal expressing cells, expression is found in a small number of cells that later form a part of the upper tentacle bulb in the cydippid (Figures 2C). At 7 to 13 hpf, additional expression is found in three patches of ectodermal cells along the sagittal plane; the innermost patch of these cells contributes to the apical organ. In the cydippid, expression can be seen in the apical organ (Figures 2F). By comparison, PpiSox3 also exhibits expression in the pharynx, tentacle bulb, and apical sensory organ during the juvenile cydippid stage, although comb row expression is not seen 14 .

Expression of MleSox2, the SoxC group member, was detected ubiquitously during early cleavage stages representing maternally deposited expression (Figure 3A). Post-gastrulation (4 to 6 hpf), the expression is split between the oral and aboral halves of the developing embryo, specifically around the blastopore on the oral half, and in mesodermal and ectodermal cells on the aboral half (Figures 3B and E). Expression is ubiquitous in the pharynx and the aboral half of the embryo at 9 to 12 hpf (Figures 3C and F). In the juvenile cydippid, expression is restricted to the pharynx, tentacle bulbs, and the apical sense organ, remaining uniformly expressed in each tissue (Figures 3D and G). In juvenile cydippids from Pleurobrachia, PpiSox2/12 was similarly expressed in the tentacle base and apical sense organ, but not in the pharynx. PpiSox2/12 also exhibited expression in small spots within the comb rows 14 .

MleSox3 is expressed during embryogenesis at 9 to 14 hpf in four groups of mesodermal cells that make up part of the upper tentacle bulb (Figures 4A and C). During the cydippid stage, expression in this region is concentrated in four distinct regions of the upper tentacle bulbs (Figures 4B and D). Additionally, MleSox3 expression is found in groups of cells in the upper pharynx, as well as in the apical sense organ in two main cell groups along the sagittal axis where the base of the apical organ connects to the polar fields (Figures 4B and D). In comparison, PpiSox1, the ortholog to MleSox3, was similarly expressed near the tentacle base, in four small spots around the pharynx, and in five spots in the apical sense organ 14 .

MleSox4 expression is lightly ubiquitously expressed at 9 to 14 hpf in parts of the developing pharynx, in ectodermal and mesodermal cells that make up the tentacle apparatus, and in cells that form the apical organ (Figures 4E and G). During the juvenile cydippid stage, expression encompasses the entirety of the comb rows (Figure 4H). The ubiquitous expression found in earlier stages continues in the pharynx, the tentacle bulbs, and the apical organ of the cydippid (Figures 4F and H). In contrast, PpiSox6 expression during the juvenile cydippid stage was seen exclusively in the comb rows 14 .

MleSox6, which was unclassified by the phylogenetic analysis, is initially expressed around 9 hpf in animals with already developed and functional comb plates. At 9 to 14 hpf, expression is distributed equally throughout the pharynx and stops where the pharynx meets the endoderm; this expression continues throughout the cydippid stage (Figures 5A and B). The aboral expression at 9 to 14 hpf encompasses parts of the mesodermally and ectodermally derived portions of the tentacle bulbs (Figures 5A and C). Expression is also found in cells that later form part of the developing apical sense organ. During the cydippid stage, expression can be found towards the apical ends of the comb rows (Figure 5D). Expression of MleSox6 during the cydippid stage also encompasses the apical organ floor (Figure 5B) and extends out to the polar fields (Figure 5D). The uppermost parts of the tentacle bulbs show expression at this stage (Figure 5B), and light expression continues through mesodermally derived cells connected to the base of the apical sense organ (Figure 5B). There are no expression patterns available for the orthologous gene in Pleurobrachia, PpiSox4.

Despite several attempts, expression patterns were not detected for the Mnemiopsis SoxF group member (MleSox5) during any developmental stage. In support of this, RNA-Seq data generated for the Mnemiopsis genome paper 17 from mixed stage embryos (approximately 15 to 30 hpf) also do not indicate expression of this gene (data available through the Mnemiopsis Genome Web Portal: http://research.nhgri.nih.gov/mnemiopsis/, using the ‘CL2’ track of the Genome Browser). We also made several attempts to generate expression patterns for the Sox-like MleHMG-box gene, but did not detect any expression during development. Similar to MleSox5, the independent RNA-Seq evidence also suggests that this gene is not expressed in the mixed stage embryo sample mentioned above. A comparison of all Mnemiopsis and Pleurobrachia expression patterns is summarized in Table 2.

Orthologs are represented in pairs of columns by Sox group. Expression is indicated by a plus sign (+). A single asterisk (*) shows agreement in expression patterns between orthologous genes from the two ctenophore species. ‘nd’ denotes no data. The double asterisk (**) for MleSox1 expression in the comb row refers to light expression seen only in the uppermost part of the comb row at the cydippid stage.

Mnemiopsis juvenile cydippids (18 to 24 hpf) were labeled with EdU to identify regions of cell proliferation (Figure 6). Results show increased labeling in the tentacle bulbs and the apical sense organ, specifically in the apical organ floor (Figure 6B). There was additional labeling of individual nuclei in the developing comb rows (Figures 6B and C). There were minimal levels of labeling in the epidermis and in the pharynx (not visible in Figure 6). These results are generally consistent with regions of cell proliferation found in adult Pleurobrachia 31 .

A recent extensive set of phylogenetic analyses of animals and closely related non-animals that includes data from the first ctenophore genome (Mnemiopsis) supports a tree with Ctenophora branching before all other animal lineages 17 , which is consistent with some other recent phylogenomic analyses 15 16 . While this new evidence regarding the phylogenetic placement of the ctenophores is compelling, it is worth noting that other phylogenies have been put forward, placing either Porifera 47 or Placozoa 48 as the earliest-branching animal lineage; these alternative phylogenies place Ctenophora in various locations within the non-bilaterian lineages (reviewed in 17 49 ). Regardless of the branching order of the non-bilaterian animal phyla, our analyses of Sox family genes are consistent with the hypothesis that true Sox family genes arose at the base of the animals and that four major groups of Sox genes (B, C, E, and F) were fully diversified in ctenophores (Figure 7). This evolutionary scenario indicates that the Sox family of transcription factors diversified early and remained relatively stable throughout animal evolution. Overall, our trees are largely in agreement with recent studies focused on Sox phylogeny, with a few interesting exceptions. Neither of the sponge sequences that were placed in the SoxF group in a previous study (specifically, SciSoxF1 and SciSoxF2 12 ) were found in the SoxF group in any of our analyses. Although the position of these sequences was ‘unstable’ in our searches, none of our analyses placed them in the SoxF group. The lack of a clear SoxF gene in Porifera in our analyses raises the possibility that this Sox group was lost in this lineage (Figure 7); alternatively, if sponges branch before ctenophores on the animal tree, then the SoxF group may not have been present in the ancestor of all animals, first arising in the lineage leading to ctenophores. Alternately, the two or three sponge Sox genes that branch in an unclassified position outside of the SoxE and SoxF groups (AquSoxF and EmuSox1, also seen in this position in previous studies 13 35 ; and SciSoxE, seen in this position in our Bayesian analysis only) could perhaps be members of the SoxE or F groups that have diverged over time (Figure 1).

In our ML analysis, two Sycon Sox sequences (SciSoxE and SciSox6) are located within the SoxE group (Figure 1); however, in our Bayesian analysis (Additional file 2: Figure S1) these two sequences are found elsewhere (SciSoxE is in an unclassified position outside the SoxE and SoxF groups, and SciSox6 is located in a subgroup next to Group D). Further, in all of our analyses, three Sycon Sox sequences (SciSoxF1, SciSoxF2, and SciSox7) form a poorly supported subgroup next to the bilaterian Group D Sox sequences. The lack of agreement about where these sponge Sox sequences fall may be due to differences in tree reconstruction methods and/or outgroups used. Although Fortunato et al. 12 used the same ‘LG + Γ’ model of evolution for their ML analyses, they did fewer independent ML runs with fewer starting trees for each and used a different set of outgroup sequences. Stable placement of these sponge sequences may be resolved in the future with the addition of more sponge sequences or improvements in tree reconstruction methods. Functional studies of these genes will also help elucidate how closely they align with genes in known Sox groups.

The other notable difference between our trees and previously published trees is the location of several Hydra Sox genes. In our analyses, a cluster of five Hydra Sox genes is clearly nested within the SoxF clade with high support. This placement was consistent across tree construction methods and datasets used, whereas previously, these sequences were found in an unclassified position outside known Sox groups 35 . Our results indicate that SoxF genes were not lost in the lineage leading to Hydra, and are present in all cnidarians surveyed to date. Overall, this suggests that non-bilaterian Group F Sox genes are present in all lineages except Placozoa and Porifera (Figure 7).

We determined the expression patterns of five out of six Mnemiopsis Sox genes in developing embryo and juvenile cydippid stages. Expression patterns determined through in situ hybridization generally revealed spatially restricted Sox expression in somatic cells within zones of cell proliferation, as determined by EdU staining (Table 2, Figure 6). These zones were located in the developing apical sense organ, upper tentacle bulbs, and comb rows in Mnemiopsis, all regions previously shown to be regions of cell proliferation and/or stem cell regions in adult specimens of the ctenophore Pleurobrachia 31 . The experiments here show an overlap between regions of high levels of cell proliferation and regions of Sox gene expression, which supports the hypothesis that cells in these regions are stem cells or progenitor cells. The Sox genes expressed in these regions may be maintaining these cells in an undifferentiated state and/or regulating cell proliferation and renewal. The somatic expression of these genes could also play a role in the regenerative capacity of ctenophores. Mnemiopsis is known to be able to regenerate even when a large portion of its body is removed 50 51 . Overall, our results are consistent with the established role of multiple Sox genes in the maintenance of stem cell pools and as regulators of progenitor and stem cell fate 52 53 .

The Mnemiopsis Sox mRNA expression patterns we generated follow a time course of development from early cleavage stages to the juvenile cydippid stage (approximately 24 hpf). There were no consistent patterns for all Sox genes within a particular developmental stage, with each Sox gene exhibiting its own unique pattern (Table 2).

Sox gene expression has been seen in germline cells in Clytia hemisphaerica (Sox groups B, C, and E), and in adult Pleurobrachia (PpiSox2/12, part of the SoxC group) 35 . With the possible exception of MleSox1 (SoxB group; very light expression in upper comb rows), MleSox4 (SoxE group; expression in comb rows), and MleSox6 (unclassified; light expression in comb rows) which all display some comb row expression in the newly hatched 24 hpf cydippid stage, an interesting finding of this study is the apparent lack of clear Sox gene expression in regions of the developing meridional canals/comb rows (especially by the Mnemiopsis SoxC gene, unlike what is observed in Pleurobrachia), where germ cells presumably arise during the early juvenile stages of Mnemiopsis. Although fate mapping experiments have shown the cellular lineage of many adult structures, it has not yet been possible to follow labeled embryos long enough to determine the precise origin of germ cells 30 . Ctenophores are thought to specify their germline during early cydippid stages via inductive cell signaling, from the meridional canal endoderm, but this has not been confirmed experimentally and other origins of the germline are possible 54 . A small percentage of Mnemiopsis can spawn for a limited period of time as 7- to 13-day-old cydippids that measure 1.8 to 2.8 mm oral-aboral length 30 , which indicates that gonads can be fully developed and functional by this time in some individuals. Adult reproduction begins at 30 mm oral-aboral length, at an age of 13 to 17 days 55 . Repeated attempts to perform in situ hybridizations of Sox genes on slightly older (4 to 7 days) cydippid juveniles were unsuccessful, as their tissues are very fragile and tend to fall apart. Some traditional germline marker genes (such as Piwi) also do not show expression in developing meridional canals/comb rows during the early 24 to 36 hpf cydippid stage in Mnemiopsis 56 . Although the timing of germline specification and origin of germ cells in Mnemiopsis remains an open question, based on the above gene expression evidence, it seems plausible that germ cells may not yet be specified by 24 hpf in Mnemiopsis, when the animals are only 250 to 300 μm in diameter. Further study of MleSox4 (SoxE group) and additional germline marker genes will help to answer open questions regarding germline specification in Mnemiopsis.

Comparisons between Mnemiopsis and Pleurobrachia mRNA expression patterns were possible for the juvenile cydippid stage for four sets of Sox orthologs (Table 2). There was general correspondence in expression for the group B orthologs (MleSox1 and PpiSox3) in the pharynx, tentacle bulb, and apical sense organ, however, there was very light expression in the uppermost part of the comb row of MleSox1 that was not seen in PpiSox3. The group C orthologs had similar expression in the tentacle bulb and apical organ between the two species, but MleSox2 lacked expression in comb rows that was seen in PpiSox2/12 and the Pleurobrachia SoxC genes lacked expression in the pharynx that was seen in MleSox2. Correspondence in expression in the tentacle bulbs, apical sense organ, and pharynx was seen for one set of group E orthologs (MleSox3 and PpiSox1). The second set of group E orthologs (MleSox4 and PpiSox6) exhibited very different expression patterns from one another; MleSox4 had a broad pattern of expression in the tentacle bulbs, apical organ, comb rows, and pharynx at the cydippid stage, while PpiSox6 was expressed exclusively in developing comb rows. Although the expression patterns examined in these two studies each only capture a snapshot in time, the comparisons between the two ctenophore species illustrate how orthologous Sox genes likely share many similar functions, while at the same time, developing some species-specific roles during development. Overall, however, the well-established role of the Sox family in the maintenance of stem cell pools during development 53 seems to be conserved in ctenophores, at least as much as can be indicated by the zones of cell proliferation seen in both Mnemiopsis and Pleurobrachia in the apical sense organ, upper tentacle bulbs, and comb rows, which overlap with regions of Sox gene expression in both species. As discussed by Jager et al. 35 , for most invertebrate Sox genes (including the ctenophore Sox genes), the precise functions of individual Sox genes have not been determined; it remains unknown whether the ancient function of particular Sox groups is primarily associated with cell proliferation (stem cells/progenitor cells) or with differentiating cells. Functional studies of the ctenophore Sox genes in particular cell lineages as they progress would help to address these issues and aid in connecting expression patterns with function.

Sycon SoxB (SciSoxB) is expressed in the ectoderm and in cruciform cells, which are larval sensory cells that may be involved in photoreception 12 . The Mnemiopsis SoxB gene (MleSox1) is similarly expressed in the ectoderm in early developmental stages, and in cells that contribute to the apical sense organ where photoreceptors reside 22 . These results indicate a general pattern of conservation of SoxB gene expression in cells involved in light sensing in early animal lineages. In bilaterians, SoxB genes are broadly involved in neurogenesis and the development and specification of the central nervous system 1 57 , and these roles may have first begun to emerge in ctenophores and possibly sponges.

In invertebrate bilaterians, SoxE genes are often involved in gonad development (mesodermal derivatives), while SoxF genes are commonly involved in endoderm formation 57 . In the anthozoan cnidarians Nematostella and Acropora, SoxE and SoxF genes are similarly associated with endodermal expression 33 34 . Fortunato et al. 12 point out that SciSoxE, SciSox6, SciSoxF1, and SciSoxF2 are expressed in choanocytes and in some mesohyl cells in Sycon, which could be used to support the concept of homology of the choanoderm plus the mesohyl with endomesoderm in sponges. Mnemiopsis SoxE genes (MleSox3 and MleSox4) and the unclassified MleSox6 gene are similarly expressed in areas that overlap with the endomesoderm, and the Pleurobrachia SoxF gene (PpiSox8) is endodermally expressed, indicating the importance of these groups of Sox genes in endomesoderm specification in the earliest animal lineages.