Background

The discovery that three homeobox genes, Gsx, Pdx and Cdx, are organised in a small gene cluster raised the possibility that these genes might act in a coordinated way. This ParaHox cluster was first described in the amphioxus Branchiostoma floridae [1], and later in human, several other vertebrates, an echinoderm, a hemichordate and two molluscs [2,3,4,5,6,7,8]. The cluster has broken in many taxa through genomic rearrangements or gene losses [1, 3, 9, 10]. The Pdx gene, also called Pdx1, Xlox, Lox, Ipf1, Idx1 or Stf1, is expressed in developing gut endoderm in vertebrates, with sharp anterior and posterior limits, and later in the pancreas and where the duodenum meets the stomach [11,12,13]. In amphioxus embryos, Pdx is also expressed strongly in the midgut endoderm and, puzzlingly, in two putative receptor cells in the neural tube [1, 14]. The latter lie adjacent to a large pigment cell [1, 15] and comprise the first photosensory organ of Hesse (dorsal ocellus) to develop [16]. The amphioxus Cdx gene is also expressed in the gut, more posteriorly where the anus will form, and in other posterior tissues at early developmental stages [1]. Similarly, vertebrate Cdx genes are expressed caudally, including the posterior gut, although multiple paralogues have subtly different patterns and additional expression sites [17,18,19]. Gsx is expressed primarily in neural tissue in amphioxus, Drosophila and mouse [1].

Together with data from echinoderms, annelids and molluscs, these patterns led to the hypothesis that ParaHox genes were components of an ancient system for patterning the bilaterian gut [1, 20,21,22,23]. The origin of a ‘through-gut’, with the mouth, digestive regions and anus, may have facilitated the evolution of predation and active burrowing, key drivers of animal diversification in the Cambrian [1, 20, 24] (but see [25]). To test this hypothesis, we need to know if ParaHox genes consistently specify region-specific fate, instructing cells to differentiate according to their head to tail position. Such a role has been suggested for vertebrate Pdx. Mutation of both Pdx alleles results in the absence of a pancreas in mice and humans [12, 26,27,28], conditional deletion in mouse causes homeotic-like transformation of endoderm cells [13], and ectopic expression in chick causes endoderm cells to change molecular identity and behaviour [29]. Pdx has a similar role in sea urchin larvae, where morpholino knockdown inhibits the formation of the sphincter between the midgut and hindgut [22, 30]. In adult mammals, the Pdx protein is a transcriptional activator of insulin and other genes in β cells of the endocrine pancreas [31]. No insights are possible from Drosophila or nematodes as they have lost Pdx, while functional data are difficult to obtain in other invertebrates.

For Cdx genes, there is evidence for a role in gut patterning in several taxa. For example, heterozygous mutation of mouse Cdx2 causes homeotic-like transformation of posterior gut cells [32] and inhibiting Cdx function in sea urchin development allows posterior gut cells to express a more anterior marker [22]. In Drosophila, Cdx is necessary for the invagination of the hindgut [33], and Caenorhabditis elegans Cdx has roles in the development of the rectum [34]. A second role for Cdx genes in body axis or tail extension is also seen in some taxa, including mice, Xenopus and zebrafish [18, 35,36,37], an ascidian [38], and short-germ arthropods Artemia, Tribolium and Gryllus [39, 40]. There is no similar function in Drosophila, however, which may be because Cdx was recruited into a novel gene network patterning the rapidly developing long-germ band embryo [41]. While roles in the gut and tail may be shared between distant phyla, there are also taxon-specific roles. Thus, mutations in mouse Cdx1 or Cdx2 cause homeotic transformations of mesodermally derived vertebrae, at least in part by altering Hox gene expression [18, 42], and interfering with Cdx function in Xenopus causes disruption to neural patterning [37, 43]. In Drosophila, Cdx also specifies fate of the cells forming the anal plates on the exterior of the most posterior segment [44]. Similarly, many cell types are affected by mutation of Cdx (pal-1) in the nematode Caenorhabditis elegans, including posterior epidermal rays [45].

Expression of Gsx is variable between taxa, making inference of its ancestral role in bilaterians difficult. The larvae of some annelids and a mollusc show clear Gsx expression around the forming mouth, consistent with a general link between ParaHox genes and the through-gut [9, 23, 46]. This pattern, however, is not seen in all molluscs nor in ecdysozoans [47]. Wollesen et al. [47] have therefore argued that Gsx expression in the anterior nervous system was more likely an ancestral bilaterian trait, with recruitment into foregut a derived character of some taxa. Either way, deuterostomes such as amphioxus are probably not suitable for inferring ancestral roles of Gsx as the deuterostome mouth may not be homologous to the mouth of ancestral bilaterians [20].

Experimental data from a diversity of invertebrates are necessary to evaluate which roles of Pdx and Cdx are conserved between taxa and to evaluate their ancestry. Key questions are whether these genes specify region-specific cell fate in the gut and formation of the anus. Since there is no living outgroup to bilaterians that possesses a homologous through-gut, it is necessary to address this question using ingroups from each bilaterian clade. The recent development of TALEN mutagenesis methods applicable to amphioxus has opened up an opportunity to investigate these functions in this animal occupying a pivotal phylogenetic position [48, 49].

Results

Generation of stable Pdx and Cdx mutant lines in amphioxus

We deduced gene structures for amphioxus Pdx and Cdx and designed constructs encoding TALEN pairs targeted to the first coding exon of each (Fig. 1; Additional file 1: Figs. S1-S6). For each gene, two TALEN mRNAs were co-injected into unfertilised eggs of B. floridae and successful mutagenesis detected by PCR on pools of neurula stage embryos (Additional file 1: Table S1, Fig. S7). Remaining embryos were reared to maturity to generate mosaic F0 animals; founders were spawned and mutations typed by PCR on sperm or pools of neurulae generated by outcrossing. Embryos from F0 × wild type crosses were reared to maturity to generate F1 heterozygous mutants; these lines were expanded by crossing with wild type animals, rearing embryos to maturity and intercrossing offspring.

Fig. 1
figure 1

Gene structure, sequences targeted by TALENs and mutations generated. E1, E2, E3, coding regions of exons; grey, homeobox; WT, wild type sequence

Full size image

We generated amphioxus lines with 4-bp, 11-bp and 13-bp deletions (4Δ, 11Δ, 13Δ) in Pdx (Fig. 1). Each deletion causes a frameshift predicted to give a protein comprising 31–33 amino acids of wild type protein followed by peptide sequence with no strong similarity to any known protein (Additional file 1: Fig. S8). We generated a Cdx mutant line with a 7-bp deletion (7Δ) predicted to give a short peptide comprising the first 5 amino acids of Cdx plus 4 additional amino acids before a stop codon (Fig. 1; Additional file 1: Fig. S9). None of the predicted products contains a homeodomain.

Pdx is necessary for correct spatial differentiation of the midgut

Embryos derived from crosses between heterozygous Pdx 4Δ mutants, and crosses between Pdx mutant lines, generated pools of morphologically identical embryos; genotyping revealed these include homozygotes, heterozygotes and wild type with no significant reduction in frequency of homozygous mutants (18/96; chi square p = 0.16). Since Pdx is strongly expressed in the gut and the first Hesse organ receptor cells in 13-h neurulae, we tested whether marker genes for these tissues were affected in Pdx mutants. In situ hybridisation revealed no difference in expression of Mop (melanopsin, expressed in the Hesse organ receptors; Fig. 2a, b) or Mitf (expressed in the adjacent pigment cell; Fig. 2c, d). Dorsal views show that the two receptors cells (Fig. 2a’, b’) and single pigment cell (Fig. 2c’, d’) are present in mutants. We also detected no difference in expression pattern in Pdx itself, Cdx or Ilp1 (insulin-like peptide gene) in the endoderm of neurulae (Fig. 2e–j). Later in development, however, a morphological phenotype was visible in mutants when viewed under fluorescent illumination. In wild type and heterozygote larvae, endogenous green fluorescence is detected in cirri around the mouth and in a clearly delineated patch of the midgut endoderm, noted here for the first time. In contrast, in 4Δ/4Δ, 4Δ/11Δ and 11Δ/13Δ mutant larvae, clear fluorescence is seen only in buccal cirri. The difference is first detectable at the 5 gill slit stage (6 to 16 days development depending on batch) and persists to larvae with many gill slits (around 46 days; Fig. 2k, l, Additional file 1: Fig. S10, S11).

Fig. 2
figure 2

Pdx mutant phenotype. In situ hybridisation to wild type and Pdx 4Δ homozygous mutant neurulae stage (13 h) amphioxus embryos for Mop (a, b), Mitf (c, d), Pdx (e, f), Cdx (g, h) and Ilp1 (i, j). Embryos were genotyped by PCR after hybridisation. Green fluorescence in 7 gill slit amphioxus larvae showing endodermal patch in wild type (k, k’) but not homozygous Pdx 4Δ mutants (l, l’). Inset k” shows confocal imaging of sibling larva showing fluorescence in endoderm cells. Anterior to the left in all images; dorsal to the top in all cases except a’, b’, c’ and d’ which are dorsal views with the right side to the top of the image. Scale bar in a, 50 μm, refers to a to j; scale bar in k, 200 μm, refers to k to l’

Full size image

Cdx is necessary for the formation of the anus and posterior axial extension

Homozygous Cdx−/− mutants are identical to wild type amphioxus in external appearance through gastrula and neurula stages, and until the mouth opens around 24 h post-fertilisation (Fig. 3; Additional file 1: Fig. S12). Starting at the 2 to 3 gill slit stage (30 to 40 h post-fertilisation), a morphological difference gradually begins to be evident as truncation of the body axis in homozygous mutants (Fig. 3a, b). As development proceeds, the posterior of the animal continues to extend only in wild type and heterozygous larvae, enhancing the phenotypic difference. By the 4 gill slit stage, homozygous mutants are approximately half the length of wild type larvae (Additional file 1: Fig. S13).

Fig. 3
figure 3

Cdx mutant phenotype. a Wild type and b homozygous mutant 3 gill slit (40 h) larvae showing the anus or lack of the anus (arrow) and distance from gut restriction to the tip of the tail (double-headed arrow). Scanning electron micrographs showing the anus and tail fin in 40-h wild type larva (c, c’) and closed anus and reduced tail fin in mutant larva (d, d’). In situ hybridisation to Rootletin RNA in e late neurula (16 h) wild type embryo, f late neurula (16 h) homozygous mutant embryo, g early larva (21 h) wild type embryo and h early larva (21 h) homozygous mutant embryo. In situ hybridisation to Cyp26-3 RNA in i late neurula (16 h) wild type embryo, j late neurula (16 h) homozygous mutant embryo, k early larva (21 h) wild type embryo and l early larva (21 h) homozygous mutant embryo. Anterior to the right, dorsal to the top of the image in a–c. Anterior to the left, ventral to the top in d. Anterior to the left, dorsal to the top in e to l. Scale bar in a, 50 μm, refers to a and b; scale bar in e, 50 μm, refers to e to l.

Full size image

A second morphological difference becomes evident around the 3 gill slit stage in the gut. In homozygous mutants, the gut remains closed by the epithelial cell layer of the endoderm; in wild type animals, the anus perforates to generate the through-gut (Fig. 3c, d). When cultured algal cells are provided to 2-day and 3-day larvae, mutant larvae take up food material and in some animals the closed anus ruptures leaving a ragged terminus. The tail fin is also markedly reduced in mutants and develops from a smaller zone of the posterior ectoderm; this effect is most marked ventrally (Fig. 3a, b). The development of the amphioxus tail fin is driven by the extension of specialised epidermal cells containing long intracellular ciliary rootlets containing a coiled-coil protein Rootletin [50,51,52]. We find the Rootletin gene is expressed by fewer cells in mutant than in control sibling larvae (Fig. 3e–h). This alteration of Rootletin gene expression is consistent with a smaller number of epidermal cells being specified to differentiate into tail fin in Cdx mutant animals. The expression of a posteriorly expressed gene Cyp26-3, encoding a retinoic acid-degrading enzyme, is also greatly reduced in Cdx mutant embryos (Fig. 3i–l). We infer that Cdx is necessary for fate specification of posterior ectoderm cells, for formation of the anus and for tail extension.

Cdx acts via the retinoic acid pathway in amphioxus posterior development

The imperforate anus, tail truncation, reduced Rootletin expression and tail fin phenotype observed in Cdx mutant larvae are similar to a suite of changes caused by exogenous treatment of amphioxus larvae with retinoic acid (RA) [51, 53]. We therefore asked whether Cdx mutation caused these phenotypes through an increase in RA in the posterior region.

To test if dampening RA signalling partially rescues the mutant phenotype, we treated embryos derived from Cdx heterozygote crosses with BMS493, an inverse agonist of the pan-retinoic acid receptor (RAR; Fig. 4). Homozygous mutant larvae treated with BMS493 had enlarged tail fins compared to untreated mutants, especially dorsally, with the fin developing from more epidermal cells (Fig. 4b, d). Posterior growth of the body increased marginally but the anus did not open. In wild type and heterozygous embryos, BMS493 treatment gave the opposite phenotype to Cdx mutation (Fig. 4a, c): an enlarged larval tail fin derived from a larger region of posterior ectoderm (see also [53]). These phenotypes were presaged by subtle changes to expression of the Rootletin gene at earlier developmental stages (Fig. 4e–j). We infer, therefore, that the role of Cdx in tail fin formation and tail extension acts, at least in part, through the RA pathway.

Fig. 4
figure 4

Inhibiting RA pathway partially rescues Cdx mutant phenotype. The large tail fin evident in wild type and Cdx+/− heterozygous (‘normal’) amphioxus larvae (a, 30 h) is severely truncated in Cdx−/− homozygous mutants and develops from fewer epidermal cells (b). Inhibition of the RA pathway using BMS493 enlarges the tail fin of normal larvae (c) and partially rescues the truncation in Cdx−/− mutants (d). Tail fin development is presaged by the expression of the Rootletin gene in 18-h late neurula embryos, visualised by in situ hybridisation (e), which is also downregulated in Cdx−/− mutants (f). Inhibition of the RA pathway results in the upregulation of Rootletin RNA to varying degrees in normal (g, i) and mutant (h, j) embryos. In contrast, posterior expression of Cyp26-3 in 16-h embryos (k) is downregulated in Cdx−/− mutants (l) and by inhibition of RA action (m, n). These data are consistent with a model (o) involving inhibition of RA signalling by Cdx and a feedback loop; interactions in dotted lines are either not confirmed (Cdx effect on RA independent of Cyp26-3) or deduced from previous work (RA inhibition of Cdx [14]). Anterior to the left, dorsal to the top in all images. Scale bar in a, 50 μm, refers to a–d; scale bar in e, 50 μm, refers to e–n

Full size image

We examined a key player in the RA pathway, a cytochrome P450 family 26 (Cyp26) gene encoding an enzyme that degrades and clears excess RA [54]. Amphioxus has three Cyp26 genes, derived from cephalochordate-specific duplication [53, 55, 56] (Additional file 1: Fig. S14, S15). In B. floridae, we find that Cyp26-3 is expressed posteriorly at the neurula stage (Fig. 4k), unlike the pattern previously reported in B. lanceolatum [53]. In homozygote Cdx mutants, posterior Cyp26-3 expression is downregulated but not abolished (Fig. 4l); treatment with BMS493 also downregulates Cyp26-3 expression, to a more extreme degree (Fig. 4m, n). To investigate if Cyp26 genes are direct transcriptional targets of Cdx in amphioxus (as in mouse [57]), we injected unfertilised amphioxus eggs with a luciferase reporter gene construct driven by 3.1 kb of Cyp26-3 upstream sequence and assayed luciferase at the late neurula stage. The sequence includes several putative Cdx-binding sites (Additional file 1: Fig. S16) of which two have a close match to a consensus [58]; mutation of either or both decreases luciferase activity consistent with Cdx-binding positively regulating Cyp26-3 expression (Fig. 5; Additional file 1: Tables S3, S4).

Fig. 5
figure 5

Reporter gene analysis of Cyp26-3. a Region 5′ of Cyp26-3 gene showing two putative Cdx-binding sites (BS1, BS2). Numbers are distance before start codon. Four reporter gene constructs were tested in amphioxus embryos with 0, 1 or 2 Cdx sites mutated. b Violin plot showing relative levels of luciferase expression for each construct, a control luciferase vector pGL3 without Cyp26-3 sequence, and uninjected control (values in Additional file 1: Table S4)

Full size image

These results suggest that a major role of Cdx expression is to ensure RA activity is kept at a low level in the posterior part of the amphioxus embryo, at least in part via positive regulation of Cyp26-3. Downregulation of Cyp26-3 expression by BMS493 is consistent with a feedback loop between RA and Cyp26 activity as previously shown in zebrafish [59,60,61] (Fig. 4o).

Effects of Cdx and Pdx mutation on amphioxus transcriptome

To examine the downstream molecular consequences of Pdx and Cdx mutation, we sequenced transcriptomes of mutant larvae soon after morphological phenotypes are evident: 5 gill slits for Pdx−/− and 2–3 gill slits (34 h and 42 h) for Cdx−/− (42 h described below; mixed analysis in Additional file 2: Supplementary Data). Comparisons were made to transcriptome data from mixed wild type and heterozygous siblings to identify direct and indirect targets of Cdx and Pdx that are up- or downregulated after mutation (Fig. 6a, b; Additional file 1: Fig. S17, S18).

Fig. 6
figure 6

Transcriptome analysis of Pdx and Cdx mutant amphioxus embryos and larvae. a MA plot showing fold change in expression (log 2 scale) between wild type and Pdx mutant larvae in relation to mean expression level (read counts). Coloured dots are contigs meeting significance criteria. b MA plot showing fold change in expression (log 2 scale) between wild type and Cdx mutant embryos in relation to mean expression level (read counts). Coloured dots are contigs meeting significance criteria. c Violin plot comparing expression level (FPKM) between wild type (WT) and Pdx mutant larvae (KO) for selected GFP, ILP1 and ILPR contigs, showing downregulation of GFP and upregulation of ILPR in mutants. d Violin plot comparing expression level between wild type (WT) and Cdx mutant embryos (KO) for selected iLBP contigs, showing downregulation in mutants. e Linear regression showing colinear-like response in expression level fold change of Hox genes to Cdx mutation. f Proportional Venn diagram showing overlap between contigs classed as gut-enriched (blue) and those expressed differentially after Pdx mutation (orange) or Cdx mutation (green)

Full size image

We first asked which of the multiple amphioxus GFP-encoding genes [62, 63] are affected in Pdx mutants. We found 11 contigs encoding GFP that were significantly downregulated (1.8 to 8.3 fold); 10 derive from tandemly duplicated amphioxus GFP genes that cannot be distinguished using short read data: GFP-8, GFP-10, GFP-12 and GFP-13 [63] (Additional file 1: Table S5). We conclude that one or more of these closely related GFP genes were affected when regionalisation of the gut was disrupted by Pdx mutation (Fig. 6c).

Second, we asked if signalling pathways were disrupted by ParaHox gene mutation. In Pdx−/− mutants, we detect significant changes to expression of genes encoding components of the insulin-signalling pathway (~ 1.6 fold downregulation of two IGFBP contigs; 1.4 fold upregulation of ILPR [64]), but not the gut-expressed Ilp1 gene itself (Fig. 6c; Additional file 1: Table S6). In Cdx−/− mutants, we found downregulation of intracellular lipid-binding protein (iLBP) genes, which in vertebrates includes CRABP (cellular retinoic acid-binding protein), CRBP (cellular retinol-binding proteins) and FABP (fatty acid-binding proteins). In amphioxus, these have undergone extensive independent duplications [56, 65]. The contigs affected by Cdx mutation correspond to amphioxus iLBP4 (formerly called CRABP [66]; 4.2 to 13.5 fold downregulation) and iLBP6 (1.6 fold downregulation; Fig. 6d; Additional file 1: Table S7). The biochemical activities of these genes are unclear; it is possible that one or both encode proteins that bind RA, inhibiting the RA pathway [56]. Downregulation of iLBP genes in Cdx mutants suggests a second possible mechanism through which Cdx suppresses RA activity in the posterior of amphioxus: through positive regulation of iLBPs. Consistent with disrupted tail formation, which persists through development, we find the Rootletin gene is downregulated in the transcriptome of Cdx mutants (Additional file 1: Table S7). We detect no significant change to expression levels of FGF or Wnt genes, RAR, RXR or RALDH (Additional file 1; Additional file 2: Supplementary Data), but an increase in expression of Brachyury-2 in Cdx mutants (Additional file 1: Table S7).

Third, we examined the effect of amphioxus Cdx mutation on Hox genes because in vertebrates the role of Cdx genes in body axis elongation is mediated, at least in part, through activation of central and some posterior Hox genes [18, 43, 67, 68]. Vertebrate Cdx genes may also have repressive effects on the most anterior Hox genes giving a colinear-like response: in Xenopus tropicalis, perturbation of Cdx gene activity caused the upregulation of paralogy groups 1 and 2 Hox genes and downregulation of paralogy groups 5 to 10/11 [19]. In amphioxus, Hox genes differ greatly in expression intensity in normal embryos [69,70,71] (Additional file 1: Table S8, Fig. S19, S20). Despite these quantitative differences, we detect a significant colinear-like response of Hox genes to amphioxus Cdx gene activity. In amphioxus Cdx−/− mutants, Hox-1 expression is upregulated (1.4 fold); Hox2, Hox3 and Hox4 are unaffected; Hox5 and Hox6 are mildly downregulated; and Hox7 is strongly downregulated, although not each change is significant when genes are analysed individually (Fig. 6e; Additional file 1: Table S8, Fig. S21). This colinear-like response is consistent with amphioxus Cdx in normal development activating central and posterior Hox genes and repressing anterior Hox genes as part of an axis patterning system.

Pdx and Cdx mutation affects gut-associated gene expression

To investigate further the association between Pdx and Cdx genes and the development of the gut, we tested whether gene sets affected by mutations include a high proportion of ‘gut-enriched’ genes. We used published B. lanceolatum transcriptome data [69] to define 2083 gut-enriched genes, represented by 4705 contigs in our study. These are genes with far higher expression in level in the gut than in most other adult tissues. We find 482 are differentially expressed in Pdx−/− mutants and 218 in Cdx−/− mutants (37 in common; Additional file 2: Supplementary Data). This equates to 8.3% of the Pdx differentially expressed contigs and 15.3% of the Cdx differentially expressed contigs (Fig. 6f), a significant enrichment (Additional file 1: Fig. S22). Hence, mutation of Pdx or Cdx has a disproportionate and significant effect on the expression of gut-enriched genes.

We also used this dataset to investigate the nature of changes that occurred in the gut of Pdx−/− mutants. In the set of 482 gut-enriched contigs with altered expression in mutants, we identified 218 annotated protein-coding genes (Additional file 2: Supplementary Data). These include homologues of vertebrate proteins with well-characterised roles in gut function, including the digestive enzyme chymotrypsin (1.5 fold downregulation), a mucin that coats epithelia (1.9 fold downregulation), a glycosidase enzyme used for starch digestion (1.5 fold upregulation) and a proton-coupled transporter for intestinal absorption of folates (1.5 fold upregulation); Additional file 1: Table S9.

Discussion

The development of technologies for generating targeted mutations has great potential in comparative and evolutionary developmental biology. The most widely used technology is based on CRISPR/Cas9, but this has not yet been applied successfully to cephalochordates. In contrast, TALENs have been used to generate inherited mutations in B. floridae [48, 49]. The method is efficient and reproducible, although not straightforward in amphioxus because of the practical difficulties of rearing from egg to adult and long generation times (4 to 6 months for B. floridae; much longer for B. lanceolatum). Here we report the introduction of germline mutations into two ParaHox genes, Pdx and Cdx. Each produces a clear phenotype. We do not know if all are complete null mutations or if alternative start codon usage, perhaps at low frequency, generates a hypomorphic allele. This is a possibility for Pdx since the mutant phenotype is subtle and not evident at embryonic stages, even though Pdx has no detectable maternal RNA [1, 14, 69]. Despite this caveat, the mutations allow robust conclusions to be drawn about important roles of amphioxus Pdx and Cdx genes, although they may not reveal every function.

Descriptive work has previously suggested that Pdx and Cdx genes might have functions in specific regions of the gut, notably the midgut for Pdx and the posterior gut or anus for Cdx. One of the earliest regional markers in amphioxus gut is Ilp1, expressed from the pharyngeal region to the midgut from gastrula to neurula, overlapping but larger than the Pdx domain [72, 73]. We found that mutation of amphioxus Pdx did not alter Ilp1 transcript abundance in neurulae or larvae. Even if Pdx does not specify the Ilp1 domain in early embryos, this does not exclude the possibility of Pdx being a transcriptional regulator of the Ilp1 gene at later stages or in adults, comparable to Pdx regulation of insulin in the vertebrate pancreas [74]. A later regional marker in amphioxus endoderm is a stripe of endogenous green fluorescence, evident in the midgut from the 5 gill slit stage to larvae with many gill slits (around 46 days of development). Confocal imaging confirmed this is not fluorescence from ingested algae as it is present inside endoderm cells. Previous studies involving UV excitation of amphioxus larvae and adults have described green fluorescence primarily in ovaries and buccal cirri [62, 75] attributed to the presence of multiple GFP genes [62, 63]. To our knowledge, the stripe in larval gut has not been described previously, and it proved to be a fortuitous marker of endoderm regionalisation. The most striking phenotype we observed in Pdx−/− mutant amphioxus was the absence of the GFP stripe. Transcriptome analysis reveals the endogenous gut GFP is encoded by one or more of an array of very similar genes (GFP-8, GFP-10, GFP-12, GFP-13). We interpret loss of the GFP stripe, and lower abundance of transcripts, as the consequence of incorrect cell fate specification in a region of endoderm. An alternative hypothesis, that the phenotype reflects an effect on GFP gene expression only, is not compatible with the large effect on the transcriptome we detect in Pdx−/− mutants, including altered expression of genes encoding digestive enzymes and other gut markers. Since overall shape and size of the gut seems normal, we suggest that Pdx mutation has caused cells in one region of the gut to be transformed to another endodermal cell type, invoking Pdx in cell determination functions.

A gut phenotype is also evident in amphioxus Cdx mutants, specifically the failure of the anus to open. Opening of the anus in amphioxus development involves rearrangement of cell junctions within epithelial cell layers; cells at the extreme posterior of the gut must lessen connections with neighbours and form new junctions with outer epidermal cells. In this way, two nested epithelial cell layers (gut and epidermis) become contiguous, with the edges of the anus opening marking the boundary where the two layers fused. In terms of geometry, this marks a topological transition from an indented sphere to a torus [76]. Biologically, the propensity to break and reform epithelial cell junctions must be a property specific to the extreme posterior cells of the gut (and/or the epidermis). Hence, loss of this property in Cdx mutant larvae likely reflects a change in cell fate specification, with terminal gut cells losing a region-specific character.

Transcriptome analysis allowed further insight into gut-related functions. We found that a set of ‘gut-enriched’ genes was disproportionately represented among the (up and down) differentially regulated genes in both Pdx and Cdx mutant animals; for Pdx, these include genes encoding homologues of vertebrate digestive proteins. The true number of gut-associated genes affected is likely to be even higher, since our definition of ‘gut-enriched’ is relatively strict. This enrichment further indicates that correct development of the gut, with functional regionalisation, is severely perturbed in Pdx and Cdx mutants.

We also asked if amphioxus Pdx and Cdx genes have roles in tissues outside the gut, because Pdx is also expressed in the first Hesse organ receptor cells to form and Cdx is expressed in all germ layers at the posterior of the embryo [1]. Hesse organ receptors are primary rhabdomeric photoreceptors and express the amphioxus melanopsin (Mop) gene encoding the microvillar light transducing protein [77, 78]. We find that mutation of the Pdx gene did not remove Mop expression from Hesse organ receptor cells and the associated Mitf-positive pigment cell also formed. As in Xenopus [19], we find that disruption of Cdx function in amphioxus has a colinear-like effect on Hox genes. Morphological impacts of such change, such as homeotic transformations, are difficult to detect because of the lack of overt anatomical differences along the segmented mesoderm. We do detect cell fate change in the posterior epidermis in Cdx mutants, however, manifest as a smaller region of the epidermis expressing the Rootletin gene. The protein product of this gene drives cell shape changes underpinning tail fin development [50,51,52]; hence, this change in epidermal cell fate has a direct effect on larval morphology.

The other major developmental process in which Cdx genes are implicated in a range of taxa is posterior growth of the body or axial elongation. This role for Cdx is well characterised in vertebrates, and a similar function is evident in some arthropods, although homology of the process between distant taxa is not proven. We found that mutation of amphioxus Cdx dramatically disrupts growth of the body axis, as in vertebrates and some arthropods. We also detect some mechanistic similarities to vertebrates. The most important effectors through which Cdx genes control tail extension in vertebrates are central Hox genes and the retinoic acid (RA), FGF and Wnt signalling pathways [57, 58, 67, 68]. The role of vertebrate Hox genes in axial extension is complex. Although mouse Hox gene mutants do not generally have a tail phenotype (apart from Hoxb13 [79]), ectopic expression of central genes, such as Hoxa5 or Hoxb8, can partially rescue the tail truncation phenotype of Cdx mutants [68]. An opposite effect was found for Hoxa13, Hoxb13 and Hoxc13 suggesting a role for the extreme posterior Hox genes in tail growth termination [68, 80]. Furthermore, there is evidence that Cdx2 is a direct activator of posterior and central Hox genes and these likely stimulate axial growth [41,42,43]. In amphioxus, we do not have definitive evidence for Hox gene involvement in tail development, but our finding of activation of central Hox genes by Cdx is consistent with this possibility and comparable to vertebrates.

There are clear similarities between amphioxus and vertebrate Cdx genes with respect to RA signalling in the tail. Low posterior levels of RA are necessary for tail growth in vertebrates [81,82,83] and amphioxus [51]. We show that Cdx is upstream of RA activity in amphioxus, as it is in vertebrates, since inhibition of RA action partially rescues the axial truncation and reduced tail fin phenotype of Cdx mutants. Young et al. [68] found that mutation of mouse Cdx2 and Cdx4 genes causes a downregulation of the gene encoding an RA-clearing enzyme Cyp26A1. Savory et al. [57] found this is a direct transcriptional effect and that it impacts RA signalling. Similarly, we find downregulation of expression of a Cyp26 gene in amphioxus Cdx mutants and show this is also likely to be a direct transcriptional effect. The clear inference is that in both mice and amphioxus, Cdx genes suppress RA signalling in the tail through the same mechanism, activation of Cyp26 expression, thereby permitting axis extension. There may be a second route by which amphioxus Cdx dampens posterior RA signalling, through positive regulation of genes encoding putative RA-binding proteins (iLBP4 and iLBP6), although currently it is unclear if these proteins bind RA or another molecule. We also note differences to vertebrates. In particular, we do not detect altered expression of genes encoding FGF or Wnt signalling components in amphioxus Cdx mutants. This is consistent with the findings of Bertrand et al. [84], who showed that in amphioxus (unlike vertebrates) FGF does not act as an antagonist to RA signalling during posterior axial extension.

Conclusions

We conclude that Pdx and Cdx genes are essential for regional cell fate specification in the endoderm of the developing amphioxus gut, including specification of midgut and formation of an anus. Combining with data from other taxa, specification of cell fates in the middle and posterior of the gut has probably been a function of Pdx and Cdx genes since the origin of Bilateria. This conclusion is consistent with the proposal that these two ParaHox genes played a role in the evolution of a through-gut, an innovation of the bilaterians and a possible contributor to sedimentary mixing and animal diversification in the Cambrian [1, 20]. We also conclude that a role for Cdx in posterior axial extension is homologous between amphioxus and vertebrates, with similar mechanistic basis, and thus dates back at least to the base of Chordata. This second role for Cdx may have facilitated the evolution of the extended post-anal tail and active locomotion in chordates.

Methods

Amphioxus culture and targeted mutation

Amphioxus (Branchiostoma floridae) were obtained from a stock maintained by Jr-Kai Yu originating from Tampa, Florida. Cultures were maintained in Xiamen University under previously described conditions [85]. Gametes were obtained using thermal-shock (20 to 26 °C) [86]; fertilisation and culturing of embryos at 26 °C was carried out as described [87]. The TALEN method was used to generate Pdx and Cdx mutants with TALEN pairs designed to target coding sequence (Additional file 1: Fig. S1-S6). TALEN construct assembly, mRNA sythesis and mutation efficacy assays were conducted following published methods [48]. Mosaic founder animals were spawned to generate F1 heterozygotes, using PCR and sequencing to detect, characterise and follow mutant alleles [49] (Additional file 1: Table S1). Homozygous mutants were generated by crossing heterozygous animals.

Whole-mount in situ hybridisation (WMISH)

Coding sequences of Cdx, Pdx, Rootletin, Ilp1, Mop, Mitf and Cyp26-3 genes were amplified from cDNA libraries from amphioxus embryos using primers given in Additional file 1: Table S2. PCR products were cloned into the pGEM-T Easy vector (Promega, USA) and confirmed by DNA sequencing; the resultant plasmids were linearised, cleaned using phenol-chloroform and used for templated synthesis of digoxigenin-labelled antisense RNA probes using T7 or SP6 RNA polymerase (Promega, USA). Embryos and larvae were fixed in 4% paraformaldehyde in MOPS buffer at 4 °C for 24 h and stored in 70% ethanol at − 20 °C. Hybridisation and detection was performed as described [88]; genotypes were checked by PCR after hybridisation and imaging.

Scanning electron microscopy (SEM) and GFP detection

SEM used an adaptation of published methods [89]. In brief, embryos were fixed in 2.5% glutaraldehyde in PBS at 4 °C overnight, washed 3 × 10 min in 0.1 M PBS (pH 7.4) and transferred to 100% ethanol through a graded ethanol series. Specimens were then washed 5 × 10 min in 100% tertiary butyl alcohol and stored at 4 °C overnight. Samples were dried in a vacuum freeze dryer, placed on conductive tape, sprayed with platinum and observed under JSM-6390 scanning electron microscope. For GFP detection, amphioxus larvae were mounted in 1% methylcellulose in seawater and photographed under an SZX10 fluorescent stereoscope (Olympus, Japan) or an LSM 780 confocal microscope (Zeiss, Germany).

BMS493 treatment

The retinoic acid antagonist BMS493 (Sigma Aldrich) was dissolved in DMSO as a 1 mM stock solution. Stock solution was diluted in filtered seawater to 1 μM and applied to amphioxus embryos from 5 h post-fertilisation onwards (early gastrula stage). Control embryos were treated with filtered seawater containing an equal amount of DMSO. Most embryos were washed and fixed at 16 or 18 h post-fertilisation for in situ hybridisation; others were continuously cultured in BMS493 until 30 h post-fertilisation for morphological observation.

Reporter gene assays

A 3.1-kb region 5′ of Cyp26-3 was cloned upstream of the firefly luciferase gene in pGL3 (Promega) and modified by PCR to generate three mutant versions with altered Cdx-binding sites (Additional file 1: Table S3). Solutions containing 3 ng/μL Renilla luciferase vector pRL-TK (Promega), 20% glycerol, 5 mg/ml Texas Red Dextran, with or without 30 ng/μL of one of the Cyp26-3 constructs, were microinjected into unfertilised B. floridae eggs as previously described [87]. For each experiment, ~ 60 embryos were collected and assayed at 16 h post-fertilisation; uninjected embryos from the same batch were used as negative control. Levels of firefly luciferase from pGL3 and Renilla luciferase from pRL-TK were detected with the Dual Luciferase Kit (Promega) using a GloMax luminometer with an integration of 10 s; the level of firefly luciferase was normalised to Renilla luciferase activity for each experiment. All experiments were repeated three times (Additional file 1: Table S4).

Embryo genotyping

Genotyping of live embryos was performed as previously described [49], and genotyping of embryos fixed with 4% PFA-MOPS-EGTA or analysed by whole-mount in situ hybridisation was conducted with the same protocol except that an extra 30-min wash in 500 ml filtered seawater or PBS was added before lysis, to remove the fixative.

RNA sequencing

Embryos and larvae obtained from crosses between Pdx+/− 4Δ × Pdx+/− 11Δ and Cdx+/− 7Δ × Cdx+/− 7Δ were sorted by visual phenotype at the 5 gill slit stage for Pdx and 2–3 gill slits (34 h or 42 h post-fertilisation) for Cdx. Samples for RNA analysis were pools of ~ 100 (Pdx) or ~ 300 (Cdx) embryos or larvae. Each cross gave two matched sibling pools: ‘mutant’ individuals with a morphological phenotype (inferred Pdx 4Δ/11Δ and Cdx 7Δ/7Δ) and ‘control’ individuals without clear phenotype (mixture of heterozygotes and wild type). RNA sequencing was performed by BGI (Shenzhen, China) on the BGISEQ platform, > 60 million 100 nt paired-end reads per sample, using mRNA enrichment, random priming reverse transcription and low PCR cycle number; initial total RNA ranged from 1.11 to 2.17 μg (RIN 10.0) for Pdx and 0.69 to 1.89 μg (RIN 8.7 to 10) for Cdx. Reads were mapped using STAR v2.7.0 [90] using --twopassmode Basic and --outSJfilterOverhangMin 12 12 12 12 options [91] to B. floridae superTranscriptome BfsuperV5 (215,490 superTranscripts) downloaded from the IncDNA-BF database [92]; genes may be represented by more than one contig. Hox genes were assembled manually to remove and correct an artefactual fusion between 6 distinct Hox genes and a lncRNA (giving 215,495 superTranscripts [93];). Between 45.2 and 52.4 million reads per sample were mapped. Since superTranscripts can join exons not normally spliced together, STAR aligner settings were adjusted to permit mapping near non-canonical splice junctions (STAR --quantMode GeneCounts --twopassMode Basic --outSJfilterOverhangMin 12 12 12 12). STAR mapped 91% and 90% of reads from the Pdx and Cdx experiments. Reads were quantified using featureCounts [94] in the Subread package v1.6.3 allowing multiple mapping, thereby permitting analysis of duplicate genes (featureCounts -O –M -p –B –fraction).

Differential gene expression analysis

Analysis of differentially expressed genes (DEGs) used DESeq2 v3.8 [95] including principal component analysis to test for outlier samples and batch effects. In the Cdx experiment, two control and two mutant 42-h samples grouped distinctly from one control and one mutant 34 h sample, likely due to a batch or developmental age effect. Two DEG analyses were therefore performed, 42 h alone and 34 h/42 h combined (Additional file 1: Fig. S17). In the Pdx experiment, one control sample (WT2) grouped distinctly and was excluded from DEG analysis which used 3 mutant and 2 control samples (Additional file 1: Fig. S18). For a contig to be considered differentially expressed, we required expression change of > 0.5 log2 (fold change) and adjusted p value < 0.05, plus absolute expression level of > 2 fpkm in at least one condition [19]. To assess the accuracy of embryo sorting, raw reads matching mutant or wild type allele sequences were counted. Embryo pools classed as Cdx−/− had 4.3 to 5.3% of Cdx ‘wild type’ reads, suggesting that 1 or 2 embryos in each mutant pool of 20 were heterozygous; the control pool had 38 to 70% wild type reads, consistent with a mix of heterozygous and genetic wild type embryos. The same method could not be applied accurately to Pdx mutants because control pools had predominantly wild type reads (91 to 100%), suggesting downregulation of mutant allele expression in heterozygotes. This obviates the applicability of read counts for assessing heterozygote number. Gut-enriched genes were identified from published B. lanceolatum data [69] (NCBI GEO GSE106430) as genes with higher mean expression level in the gut than any other adult tissue (eggs and embryos excluded) and expression level in the gut at least double that in 7 out of 8 other adult tissues; the 2083 genes were matched to contigs in the current study using blastn with an e value cut off of 1e−70 giving 4705 gut-enriched contigs.