Hedgehog participates in the establishment of left-right asymmetry during amphioxus development by controlling Cerberus expression

Left-right (LR) asymmetry is an essential feature of embryonic development in most bilateral animals. Correct positioning of asymmetric organs along the LR axis is important for their proper function and abnormality results in heterotaxy and situs inversus (Sutherland and Ware, 2009). LR asymmetry patterning in vertebrates requires asymmetric expression of Cerberus (Cer), Nodal, Lefty and Pitx2 in either the left or right side of embryos (Yoshiba and Hamada, 2014). During the development of embryos, Cer [including Cerl2 (Dand5) in mice, Coco in frog and Charon in medaka] first becomes asymmetrically expressed in a L<R manner around the node (Hojo et al., 2007; Marques et al., 2004; Schweickert et al., 2010). Antagonized by Cer protein, Nodal activity then becomes biased to the left side of the node, from which a Nodal cue is transmitted unilaterally to the left lateral plate mesoderm (LPM), where it in turn induces asymmetric expression of Nodal, Lefty and Pitx2 (Nakamura and Hamada, 2012). The mechanism responsible for asymmetric expression of Nodal, Lefty and Pitx2 has been analyzed in mice (Brennan et al., 2002; Marques et al., 2004; Saijoh et al., 2003). However, the molecular mechanism regulating Cer expression remains elusive (Yoshiba and Hamada, 2014). Interestingly, work in our lab revealed that Cer is also first asymmetrically expressed in developing amphioxus embryos and acts upstream of Nodal, Lefty and Pitx (Li et al., 2017). This finding led us to further investigate the regulatory mechanism of Cer in amphioxus.

In addition to aforementioned molecules, Hedgehog (Hh) signaling is also implicated in LR asymmetric development. In chick, Shh is asymmetrically expressed in the left side of Hensen's node, and blocking Hh activity at the node disrupts cNR-1 (chick Nodal gene) expression in the LPM and LR patterning of heart laterality (Dathe et al., 2002; Levin et al., 1995; Pagán-Westphal and Tabin, 1998). The loss of Hh signaling by Smo gene knockout in mouse led to multiple LR defects in visceral organ patterning and in Nodal, Lefty2 and Pitx2 expression (Zhang et al., 2001). Similarly, Hh knockdown in sea urchin disrupted right-side Nodal expression and right-side characteristics, but did not affect development of left-side structures (Hojo et al., 2007; Warner et al., 2016). However, the function of Hh signaling in LR patterning remains to be elucidated in zebrafish and Xenopus, although left-side ectopic expression of Shh perturbs their LR asymmetry development (Chen et al., 1997; Sampath et al., 1997; Schilling et al., 1999). Thus, to fully understand the function of Hh signaling in LR patterning and to evaluate the conservation of this function during evolution, studies of species from other taxonomic groups are evidently needed.

Amphioxus belongs to subphylum Cephalochordata and represents the most ancient of living chordates. Owing to its vertebrate-like body plan, compact genome and similar embryonic development to that of vertebrates, amphioxus has been regarded as an valuable animal model for understanding the mechanisms of vertebrate development (Bertrand and Escriva, 2011; Koop and Holland, 2008). Amphioxus exhibits consistent LR asymmetry in pharyngeal structures at larval stages, with the mouth and preoral pit formed on the left and club-shaped gland and endostyle positioned largely on the right (Soukup et al., 2015). In addition, amphioxus embryos develop somites asymmetrically along the LR axis, with the left somites positioned more anteriorly than those on the right from the five-somite stage (Minguillon and Garcia-Fernandez, 2002). Consistent with its compact genome, amphioxus encodes only a single Hh homolog, representing the archetype of vertebrate Shh, Ihh and Dhh (Shimeld, 1999). Interestingly, Hh is asymmetrically expressed on the left side in amphioxus, like Shh in chick (Levin et al., 1995; Shimeld, 1999). Our recent study found that Hh knockout in amphioxus disrupts the normal placement of several organs, such as the preoral pit and gill slits, along the LR axis (Wang et al., 2015). This implicated a role of Hh in amphioxus LR asymmetry establishment.

In this study, we show that Hh mutation in amphioxus results in morphologically symmetrical embryos, in which the left-side organs are duplicated and present on both sides, whereas the right-side structures are lost. Moreover, we examine the expression patterns of Cer, Nodal, Lefty and Pitx, and find that Cer expression disappears completely whereas expression of Nodal, Pitx and Lefty becomes bilaterally symmetric in Hh mutants. These results reveal a novel mechanism for Hh in LR patterning by regulating Cer expression.

As described in our previous study, we obtained amphioxus Hh^−/− mutants by inbreeding of Hh^+/− individuals (Wang et al., 2015). When the offspring developed to mouth-open stage, ∼25% exhibited a bilaterally symmetric phenotype. To establish whether this phenotype was caused by loss of Hh function, we performed a genotyping analysis using a PCR product cleavage assay. The results showed that the larvae with normal morphological features were Hh^+/− or wild type (WT, i.e. Hh^+/+) (10/10), whereas those with bilaterally symmetrical features were always Hh^−/− (10/10) (Fig. 1B). DNA sequencing of amplicons from the Hh^−/− individuals revealed an 8 bp deletion at the TALEN target site (Fig. 1C).

Next, we carefully examined the morphological features of the offspring and found no differences between WT and heterozygous individuals until adult. The WT/heterozygous larvae develop the mouth, preoral pit, left part of the endostyle and club-shaped gland (CSG) duct on the left side of the body, with the right part of the endostyle, glandular region of CSG and the first gill slit on the right side of the body (Fig. 2A-D). Their muscle segments are arranged asymmetrically in an interwoven pattern along the anteroposterior axis (Fig. 2E). However, H&E section staining showed that all Hh^−/− larvae had gill slits ventrally but no mouth opening on either side (Fig. 2C′,D′), and SEM results indicated that the Hh^−/− larvae developed preoral pits on both sides without mouth opening (Fig. 2F′,G′). The left part of the endostyle and the duct of CSG were also duplicated, like the preoral pit, but the right part of the endostyle and glandular region of the CSG had disappeared (Fig. 2A′-D′). In addition, the muscle segments of these homozygous mutants became symmetrically aligned on both sides of the notochord (Fig. 2E′).

We also examined the expression pattern of Ptch, a direct target of Hedgehog signal (Hooper and Scott, 2005), in the Hh^−/− embryos to confirm whether the phenotype was caused specifically by the Hh mutation as opposed to off-target effects on other genes. As expected, Ptch expression was dramatically decreased in Hh^−/− as compared with WT/Hh^+/− embryos (Fig. 2H).

We next examined the pattern of several asymmetrically expressed organ-specific genes in Hh^−/− mutants and their WT/Hh^+/− siblings. Early larvae were chosen for this analysis since major organogenesis starts at this stage and these genes display an asymmetric expression pattern in the pharyngeal region. In agreement with the above morphological observations, the expression patterns of all examined genes were changed in Hh^−/− mutants (Fig. 3). In WT/Hh^+/− larvae, FoxE4 was specifically expressed in the CSG; Krox was strictly limited to within the CSG glandular region; Nkx2.1 was predominantly expressed in the right side of the endostyle; and Dkk1/2/4 in the left side of the pharynx where the mouth will form (Fig. 3A-H). However, in Hh^−/− mutants, expression of FoxE4 and Krox in the CSG glandular region was lost, and that of FoxE4 in the CSG duct was symmetrical on both sides (Fig. 3B′,D′). Likewise, Nkx2.1 expression faded out in the right-side endostyle of Hh^−/− mutants, whereas that in the left side was duplicated and became symmetric (Fig. 3F′). Interestingly, the expression domain of Dkk1/2/4 in the preordained mouth opening region appeared on both sides in the mutants (Fig. 3H′). This indicated that, like the left-side organs described above, the mouth primordium is also ectopically formed on the right side in Hh^−/− mutants. In addition, we detected another gene, Err, which was asymmetrically expressed in a staggered pattern of somites and dorsal compartment motor neurons of WT larvae (Bardet et al., 2005). Err expression became bilaterally symmetric in the somites and dorsal compartment motor neurons of Hh^−/− mutants (Fig. 3I,I′). These results confirmed the above morphological observations and demonstrated the requirement of Hh in amphioxus asymmetric organ development.

Considering the fact that Hh signaling is involved in dorsoventral patterning of the central nervous system (CNS) and somites in vertebrates (Hammerschmidt et al., 1997), we examined eight CNS/notochord/somite marker genes, namely Wnt3, Pax2/5/8, Nkx2.1, m-actin, MRF2, Zic, Netrin and Brachyury, to determine whether they are affected in the Hh knockout. In WT neurula, expression of Wnt3 and Pax2/5/8 is restricted to the dorsal part of the neural tube and that of Nkx2.1 is confined to the ventral area of the neural tube (Kozmik et al., 1999; Schubert et al., 2001; Venkatesh et al., 1999), and these patterns appeared unaffected in Hh^−/− embryos (Fig. 4A′-F′, Fig. S1). Similarly, the expression of Netrin and Brachyury in notochord (Holland et al., 1995; Shimeld, 2000), m-actin and MRF2 in myotomes and Zic in the somite dorsolateral wall (Gostling and Shimeld, 2003; Kusakabe et al., 1997; Schubert et al., 2003) were unaffected during Hh^−/− embryo development, except for the change in somite staggered pattern (Fig. 4G′-P′, Fig. S1).

Cer, Nodal, Lefty and Pitx2 are key players in vertebrate LR patterning. All these genes exhibit an asymmetric expression pattern during vertebrate embryonic development, with Cer in a right-biased manner and the other three in a left-biased fashion (Nakamura and Hamada, 2012). Orthologs of these genes in amphioxus are also asymmetrically expressed in a similar pattern to that in vertebrates (Boorman and Shimeld, 2002; Le Petillon et al., 2013; Yu et al., 2002). Recent reports further indicated that both Cer and Nodal signaling are required for LR patterning in amphioxus and that Cer acts upstream of Nodal signaling by inhibiting its activity on the right side (Soukup et al., 2015; Li et al., 2017). To dissect the molecular mechanism of Hh signaling in amphioxus LR patterning, we compared the expression patterns of these genes in Hh^−/− mutants and their WT/Hh^+/− siblings at the mid-gastrula (G4), late gastrula (G5), neurula with 3-4 somites (N1) and neurula with 8 somites (N2) stages. In WT/Hh^+/− embryos, the four genes were symmetrically expressed at G4 and G5 (Fig. 5A,B,E,F,I,J,M,N) and asymmetrically expressed at N1 stage, with Cer in an L<R manner and the other three in an L>R manner (Fig. 5C,G,K,O). At N2 stage, expression of Nodal, Lefty and Pitx was restricted to the left side (Fig. 5H,L,P) and that of Cer had moved to the midline (Fig. 5D). However, Cer expression was completely abolished at all four stages examined in the Hh^−/− mutants (Fig. 5A′-D′), and Nodal, Lefty and Pitx were apparently unaffected at G4 and G5 (Fig. 5E′,F′,I′,J′,M′,N′) but became bilaterally symmetric at N1 and N2 (Fig. 5G′,H′,K′,L′,O′,P′).

The above experiments highlighted an essential role of Hh in asymmetry development in the amphioxus embryo by regulating Cer expression. To validate this finding further, we injected synthesized Hh mRNA into unfertilized eggs from Hh^+/− females and then fertilized them with Hh^+/− sperm to see whether the re-expression of the Hh gene could rescue Hh^−/− phenotypes. Based on morphological observation and genotype diagnosis, we found that the Hh^−/− mutants recovered their LR asymmetry and developed almost normal morphological features after Hh mRNA injection (Fig. 6B′,B″). We further compared Cer expression between the injected and uninjected embryos. About 75% (26/35) of uninjected embryos obtained from the inbreeding of Hh^+/− showed a normal Cer expression pattern and 25% (9/35) lost Cer expression, which is consistent with a Mendelian segregation ratio (3:1). However, after Hh mRNA injection, all of the embryos (40/40) displayed a normal Cer expression pattern (L<R) (Fig. 6C-D″).

After demonstrating that Hh signal is necessary for Cer expression in amphioxus, we then asked whether the signal also provides a cue for generating asymmetric Cer transcription in the embryo. We first compared the expression patterns of Cer, Hh and Ptch (a receptor and also a target of the Hh signaling pathway) in carefully staged embryos spanning from early gastrula to around the three-somite neurula (150 min after late gastrula) stages according to our previous report (Li et al., 2017). Cer transcripts were first detected in the dorsoanterior endomesoderm at late gastrula stage (0 min) in an L=R manner and then shifted to the first pair of somites in an L<R fashion from 30 min after late gastrula (Fig. 7Aa-g). Hh expression was turned on earlier than Cer expression and first detected at early gastrula stage. After late gastrula, Hh expression was conspicuous in the dorsolateral endomesoderm of the embryo, which covered the Cer expression area. However, in contrast to Cer, expression of Hh in the dorsal endomesoderm was symmetric (Fig. 7Ba-g). As a receptor and direct target gene of Hh signaling, Ptch transcription started shortly after Hh expression. In all stages examined, Ptch expression domains essentially followed that of Hh and no signs of asymmetric expression were apparent along the LR axis (Fig. 7Ca-g). These results suggested that Hh signal might not be a direct regulator for the establishment of Cer asymmetric expression in amphioxus embryos.

To confirm this proposal, we then examined Cer expression in Hh mRNA-injected embryos and in Ptch^−/− mutants (our unpublished data) at N1 stage, when Cer is expressed on the right in WT, both of which should lead to ectopic activation of Hh signal. If Hh signal participates in the LR asymmetric expression of Cer, such global activation of Hh signaling would result in Cer expression on both sides in Hh mRNA-injected embryos and Ptch^−/− mutants. Contrary to this, Cer expression was still confined to the right side and the embryos developed with a normal asymmetric morphology (Fig. 7D-K, Fig. S2). Furthermore, to detect whether Hh signaling activity exists on both sides at N1 stage, we examined Cer expression in Hh knockout embryos after SB505124 Nodal inhibitor treatment, which should result in a bilaterally symmetric Cer expression pattern in amphioxus (Soukup et al., 2015; Li et al., 2017). The results indicated that Cer expression became bilaterally asymmetric after SB505124 treatment in WT/Hh^+/− embryos, but no expression was observed in either side of Hh^−/− mutants (Fig. 8). Taken together, these results demonstrated that Hh regulates Cer expression without LR bias.

We identified a canonical Gli binding sequence (5′-GACCACCCA-3′) 66 bp upstream of the Cer transcription initiation site in the amphioxus genome, like that in zebrafish (Wang et al., 2013). Comparative genomic analysis showed that this sequence was completely conserved among three amphioxus species: Branchiostoma floridae, B. belcheri and Epigonichtys cultellus. To test whether this Gli binding sequence is necessary for initiating downstream gene expression, we cloned a 1.1 kb DNA segment upstream of Cer including the Gli binding sequence and then inserted it into a luciferase reporter vector. The luciferase assay showed that the DNA segment including the Gli binding sequence could promote reporter gene expression in amphioxus embryos. However, the segment containing a mutated Gli binding site also promoted reporter gene expression (Fig. S3), indicating that this canonical Gli binding sequence is not required for Cer expression.

In the present study, we reveal that Hh knockout in amphioxus disrupts the placement of asymmetric organs along the LR axis and changes the expression pattern of asymmetric genes, highlighting a novel mode of Hh participation in early LR asymmetry development of amphioxus embryos. Combining the current data with the previous finding that Cer acts as an upstream antagonist of Nodal signaling (Li et al., 2017), we propose a model to explain how Hh functions in the establishment of amphioxus LR asymmetry.

That Hh signaling is involved in LR patterning was first demonstrated in chick. During chick embryo development, Shh expression is confined to the left side of Hensen's node at stage 4⁺. Blocking or ectopic activation of Shh function disrupts the LR patterning of heart (Levin et al., 1995). In addition, blocking Shh activity at the left side of the node inhibits cNR-1 expression in the left LPM, and ectopic activation of Shh at the right side of the node leads to ectopic activation of cNR-1 in the right LPM (Levin et al., 1995; Pagán-Westphal and Tabin, 1998). These findings indicated that Shh-Nodal signal is important for normal LR asymmetry development in chick. In mouse embryos, Hh and Ptch1 are expressed bilaterally; nevertheless, Shh^−/− or Shh-Ihh compound mutants or Smo mutants exhibit an absence of Nodal expression in the LPM and of heart looping (Tsukui et al., 1999; Zhang et al., 2001). Further study showed that the initiation and propagation of asymmetric Nodal expression in the LPM did not require Hh signal at the node, but depended on the Hh activity in the LPM (Tsiairis and McMahon, 2009). Hh signaling also functions in a similar way in the LR asymmetry development of sea urchin embryos, and Hh knockdown by morpholino resulted in the loss of Nodal expression in the right lateral ectoderm but did not affect the initial asymmetric Nodal expression in the right coelomic pouch (Warner et al., 2016). Together, these results indicate that Hh is necessary for Nodal expression or, in other words, that Hh signaling affects asymmetry development via controlling Nodal expression.

In amphioxus, preliminary studies on Hh expression suggested that it might not be essential for LR patterning since the onset of Hh asymmetric expression occurs long after that of Nodal (Shimeld, 1999; Yu et al., 2002). However, our observations support a crucial role of Hh signal in amphioxus LR asymmetry. We demonstrated that amphioxus larvae develop an abnormal symmetric phenotype after Hh is knocked out (Wang et al., 2015). All Hh mutants duplicated their left-side organs and lost right-side organs (Fig. 2). Moreover, loss of Hh function disrupted the expression pattern of asymmetric genes such as Cer, Nodal, Lefty and Pitx. We conclude that Hh indeed participates in LR pattern formation during amphioxus embryo development.

Cer is the earliest asymmetrically expressed gene in developing embryos of amphioxus and vertebrates (Marques et al., 2004; Schweickert et al., 2010). The mechanism regulating Cer expression remains largely unknown, although a recent study in mouse indicated that the LR asymmetry of Cerl2 expression is determined post-transcriptionally by unilateral decay of Cerl2 mRNA on the left side (Nakamura et al., 2012). In our study, we showed that Hh knockout abolished Cer expression in developing amphioxus embryos from gastrula stage, suggesting that Hh signal is required for Cer gene expression. However, the mechanism by which Hh signaling regulates Cer expression remains to be elucidated, but appears not to involve a canonical Gli binding sequence that we identified in the Cer promoter region.

Additionally, WISH revealed that Cer is expressed asymmetrically in amphioxus embryos 30 min after late gastrula stage, when Hh and Ptch exhibit a symmetric expression pattern (Fig. 7Aa-Cg). Moreover, Cer expression was still confined to the right side at N1 stage when Hh signaling was constitutively activated by Ptch knockout or Hh mRNA injection. We also showed that in embryos treated with the Nodal inhibitor SB505124, bilateral Cer expression appeared in WT/Hh^+/− embryos but was not present in Hh^−/− embryos (Fig. 8), demonstrating that Hh signal was activated on both sides. These results suggested that Hh is necessary for Cer expression but that it might not provide the asymmetric cue.

As discussed above, Hh has been shown to participate in the development of asymmetry in chick, mouse or sea urchin embryos via controlling Nodal expression. However, our results indicate that a different situation exists in amphioxus as Hh^−/− mutants express Nodal, Lefty and Pitx but with a bilaterally symmetric expression pattern (Fig. 5), demonstrating that these genes do not need Hh for activation in amphioxus embryos. Similar abnormal expression of these genes was also observed in Cer^−/− amphioxus (Li et al., 2017). Moreover, Hh mRNA injection revealed that Cer expression could be recovered in Hh^−/− mutants and that the morphological asymmetry of these larvae was also rescued (Fig. 6). Cer encodes an antagonist of Nodal signaling and inhibits its action on the right side, which results in the expression of Nodal and of its targets Lefty and Pitx on the left side only (Nakamura and Hamada, 2012). Therefore, the abnormal LR patterning observed in Hh^−/− mutants can be explained by the absence of Cer expression.

Based on these results together with previous reports, we propose a new model for Hh participation in LR asymmetry development of amphioxus embryos (Fig. 9). In this model, Hh controls LR asymmetric morphogenesis by regulating the expression of Cer rather than of Nodal. In WT/Hh^+/− embryos, Hh signaling continually induces Cer expression from late gastrula stage (G5). When embryos develop into early neurula stage, Cer expression shifts to the right side due to some unknown factor inhibiting Nodal expression on the right, which results in normal LR patterning (Li et al., 2017). When Hh is knocked out, Cer expression is abolished. Hence, Nodal expression becomes bilaterally symmetric in the absence of inhibition by Cer on both sides. Thus, abnormal LR patterning occurs, with left-side organ duplication and loss of right-side structures.

The model tells us that Hh participates in the asymmetric development of amphioxus embryos via activation of Cer but that it might not provide the LR signal for Cer expression. In some vertebrates, such as mouse (Marques et al., 2004), frog (Schweickert et al., 2010), zebrafish and medaka (Hojo et al., 2007; Lopes et al., 2010), cilia-generated flow in the LR organizer provides the signal for asymmetric Cer expression. In amphioxus, cilia movement might play some roles in symmetry breaking, because ciliogenesis and embryonic rotation appear just prior to the onset of Cer asymmetric expression (Soukup et al., 2015). However, there is as yet no direct evidence to support that Cer symmetry breaking is related to cilia movement. SB505124 treatment alters the asymmetric expression pattern of Cer in amphioxus (Soukup et al., 2015; Li et al., 2017). Therefore, whether Nodal signaling plays roles in symmetry breaking is worthy of further study. Knockout of genes required for ciliogenesis or in the Nodal signaling pathway using gene editing (Li et al., 2014) should provide insight into this issue.

In contrast to vertebrates, the amphioxus mouth opens on the left at the larval stage. This has led to controversy over whether the amphioxus mouth is a homolog of the vertebrate mouth (Kaji et al., 2016; Soukup et al., 2015). In vertebrates, Hh signaling is necessary and sufficient to initiate mouth formation, and its activity is required in a dose-dependent way to determine mouth size (Tabler et al., 2014). A recent study in amphioxus reported that the formation of the mouth was controlled by Nodal signaling (Kaji et al., 2016). However, in Hh^−/− amphioxus embryos, Nodal, Lefty and Pitx were symmetrically expressed on both sides due to the loss of Cer expression, and the larvae had no mouth opening on either side. However, Cer^−/− amphioxus embryos also expressed Nodal, Lefty and Pitx symmetrically but two mouths appeared, one on each side (Li et al., 2017). The different mouth phenotypes of Cer^−/− and Hh^−/− mutants indicated that Hh signal is necessary for amphioxus mouth formation. Notably, WISH revealed that the oral mesovesicle (the definitive source of a larval mouth in amphioxus) marker genes Dkk1/2/4 and Pou4 were still expressed normally in Hh^−/− larvae (Fig. 3G-H′, Fig. S4), similarly to Cer^−/− mutants (Li et al., 2017). This implies that Hh is required for mouth opening, but not oral mesovesicle formation, in amphioxus. Moreover, upregulating Hh signaling activity by Hh mRNA injection or Ptch knockout led to an increase in the size of the larval mouth (Fig. S2). These results provide molecular evidence of an orthologous relationship of the mouth between amphioxus and vertebrates.

Amphioxus Branchiostoma floridae originally introduced from Dr Jr-Kai Yu's laboratory (Institute of Cellular and Organismic Biology, Academia Sinica, Taiwan) was bred in our aquarium as described previously (Li et al., 2012). Thermal induction spawning, egg fertilization and embryo culture were carried out according to our previous reports (Li et al., 2013, 2015; Zhang et al., 2007). Embryos and larvae at required developmental stages were fixed with 4% paraformaldehyde (PFA) in MOPS buffer (pH 7.4) at 4°C overnight. Stages were defined as mid-gastrula (G4), late gastrula (G5), pre-hatching neurula (N0), early neurula (N1), mid-neurula (N2), late neurula (N3), early larva (L1) and mouth-open larva (L2) in accordance with previous reports (Hirakow and Kajita, 1991; Hirakow and Kajita, 1994; Lu et al., 2012).

Hh gene knockout amphioxus were generated previously using the TALEN method (Wang et al., 2015). Briefly, a pair of TALENs recognizing exon 1 of the amphioxus Hh gene (Fig. 1A) were designed and assembled as described (Li et al., 2014). The resulting plasmids were linearized with SacI and the TALEN mRNA synthesized using the mMESSAGE mMACHINE T3 Transcription Kit (Ambion).

TALEN mRNA was microinjected into unfertilized eggs, which were then fertilized. One day after injection, genomic DNA was isolated from the injected embryos and used as template for PCR. The PCR products were digested with BsrGI to estimate the somatic mutation ratio. To obtain germline mutants, the TALEN-injected embryos (F₀) were raised to adulthood and crossed with WT amphioxus. The progeny (F₁) were genotyped using a PCR/sequencing assay. In this study, homozygous mutants (Hh^−/−) were obtained from the inbreeding of heterozygotes, as Hh^−/− mutants cannot survive to adult. Genotyping of offspring was carried out using the PCR product cleavage assay. Primers AmphiHhF2/R2 used for this amplification respectively match the upstream/downstream sequences of the TALEN target site (Table S1). Amplicons were digested with BsrGI and then subjected to agarose gel electrophoresis to distinguish genotypes: those containing the mutation were indigestible owing to loss of this restriction site, whereas those from WT were cleaved into short fragments.

Samples were prepared using the t-butyl alcohol freeze-drying method (Inoué and Osatake, 1988). First, embryos at L2 stage were fixed in 2.5% glutaraldehyde in filtered PBS (pH 7.4) at 4°C overnight. Next day, the fixed samples were washed three times in 1 M PBS for 10 min each, and then dehydrated in a graded ethanol series. Specimens were then transferred into 100% t-butyl alcohol, and the liquid was changed five times at 10 min intervals. Finally, specimens were immersed in t-butyl alcohol at 4°C overnight. Next day, the specimens were dried in a vacuum freeze dryer, mounted on an aluminium stub, coated with platinum and observed under a JSM-6390 field emission scanning electron microscope (JEOL) at 20 kV.

RNA probes were amplified using the primers listed in Table S1. cDNA templates for PCR were derived from total RNA extracted from mixed embryos and larvae. PCR products were cloned into the pGEM-T Easy vector (Promega) and transformed into E. coli. After sequencing verification, digoxigenin (DIG)-labeled antisense probes were synthesized for target genes using SP6 or T7 RNA polymerase (depending on insert orientation). WISH was performed according to Holland et al. (1996) with slight modifications as follows: the duration of proteinase K treatment varied from 3 to 10 min depending on embryonic stage, and probe incubation was performed at 65°C overnight.

After imaging as whole mounts, embryos were embedded in 1% agarose (in 0.5 M PBS), then dehydrated in an ethanol series (30%, 50%, 60%, 70%, 80%, 95% and 100%, 10 min each). Finally, the embryos were embedded in paraffin, sectioned at 5 µm, mounted on a polylysine-treated slide and stained with Eosin according to standard histological methods.

Full-length amphioxus Hh cDNA was amplified using primers AmphiHhF1/R1 (Table S1) and the purified cDNA fragment subcloned into the Addgene plasmid pXT7. After sequencing verification, we synthesized capped Hh mRNA using the mMESSAGE mMACHINE T7 Transcription Kit (Ambion) and stored the mRNA at −80°C until use.

Unfertilized eggs and sperm were obtained separately using the heat-shock method (Li et al., 2013) and Hh mRNA mixed with red fluorescent dye was injected into unfertilized eggs as described (Liu et al., 2013). The final concentration of Hh mRNA was 1.5 μg/μl, and the injection dose was 2 pl for each egg. After injection, the eggs were fertilized with preserved sperm, and any unfertilized eggs or those without red fluorescence sorted out under a fluorescence microscope. The remaining eggs were cultured in an incubator at 25°C and 95% humidity. Samples were collected for in situ hybridization at the desired developmental stages.

We thank Dr Luming Yao for technical assistance with scanning electron microscopy observation.