A secretory cell type develops alongside multiciliated cells, ionocytes and goblet cells, and provides a protective, anti-infective function in the frog embryonic mucociliary epidermis

Epithelia that line internal cavities have several important specialised functions. Such epithelia include the mucosal epithelia of the gut, the mucociliary epithelia of the lung and the secretory epithelia of the kidney. Although their predominant function differs in each case, all of these epithelia have two properties in common: they consist of several different cell types that cooperate to perform the epithelium’s primary function, and they are required to act as a protective barrier for internal tissues.

A number of human diseases arise due to damaged or defective mucosal or mucociliary epithelia and they are usually caused by a defect in one of the many cell types that comprise the epithelial tissue. However, this primary defect can have an impact on adjacent cell types, compromising the function of the epithelium as a whole. This is exemplified by cystic fibrosis in the respiratory epithelium and the gut. The cystic fibrosis transmembrane conductance regulator (CFTR), a chloride/bicarbonate transporter, is principally found in serous acinar cells in the submucosal glands of the lungs (Engelhardt et al., 1992). However, mutations in this channel can have indirect effects on neighbouring cells by affecting ionic homeostasis, making mucus thick and sticky and inhibiting clearance by cilia (Houtmeyers et al., 1999). In the gut, CFTR is present in enterocytes but it has been hypothesised that intercellular communication with adjacent goblet cells can affect secretion of mucins through alterations in levels of bicarbonate (Garcia et al., 2009). Thus, understanding how different cell types interact, and the role that each one plays in the integrity of the epithelium, is of paramount importance for understanding the underlying aetiology of epithelial disease.

Epithelial models that are reconstituted in vitro tend not to recapitulate the complexity that exists in vivo, but experiments in mammalian models are difficult, invasive and expensive. For these reasons, model systems of lower vertebrates, where such epithelia can be readily accessed, have been instrumental in advancing our understanding of mucociliary epithelial development. In recent years, the larval skin of Xenopus has been extensively studied because it is known to have a population of motile multiciliated cells (Drysdale and Elinson, 1992). Indeed, research into the ciliated cells in the epidermis of Xenopus embryos has yielded a number of important insights that are relevant across different biological systems and disease (Deblandre et al., 1999; Mitchell et al., 2007; Park et al., 2008; Stubbs et al., 2012). For example, the PCP protein, Fritz, has been shown to be crucial in controlling the localisation of the cytoskeletal Septin proteins to the base of cilia (Kim et al., 2010). The same study identified mutations in the human Fritz gene in patients suffering from ciliopathies such as Bardet-Biedl syndrome. This illustrates how findings in a model organism such as Xenopus can have direct clinical relevance.

However, if the embryonic Xenopus epidermis is to be a truly powerful model of mucociliary and mucosal epithelia, it is necessary to understand and characterise all cell types that comprise it. With this in mind, we, and others, have recently identified the ionocytes. Ionocytes tend to appear in close proximity to the ciliated cells in the larval epidermis and they have a number of pumps and channels involved in regulating ionic balance and pH (Dubaissi and Papalopulu, 2011; Quigley et al., 2011). Similar cells are found in human mucosal epithelia, such as the serous cells in the respiratory tract (Loffing et al., 2000) and the enterocytes of the gut (Garcia et al., 2009). Depletion of these cells in the Xenopus embryonic epidermis caused a defect in the localisation of basal bodies of neighbouring ciliated cells. This is likely to be due to alteration in pH across the epidermis, which is known to affect the localisation of Dishevelled, a basal body interacting protein (Park et al., 2008; Simons et al., 2009). These findings confirm the power of a multicellular model system in uncovering previously unknown cell-cell interactions that take place in vivo.

The Xenopus larval epidermis also contains ‘goblet cells’ that make up the majority of the epithelium and were originally described at an ultrastructural level in electron micrographs (Billett and Gould, 1971). It is generally assumed that they secrete mucins, like their mammalian counterparts; however, no specific mucins have ever been identified in secretions from these vesicles. They are known to secrete a lectin, Xeel, which is proposed to recognise pathogen-associated glycans (Nagata et al., 2003). The glycosaminoglycan, chondroitin sulphate, has also been identified in the goblet cells but its function is not clear (Nishikawa and Sasaki, 1993).

In this study, we identify and characterise an additional cell type in the larval epidermis of Xenopus tropicalis. This cell type is small and contains an abundance of large apical secretory vesicles. We adopt the term small secretory cells (SSCs), which has been previously used for cells of similar ultrastructural morphology but unknown function (Hayes et al., 2007). Our data suggest that SSCs are specified by Foxa1, secrete a dense, heavily glycosylated material, are responsible for the innate defence mechanisms of the embryo against pathogen invasion, and complete the characterisation of the embryonic epidermis.

The larval Xenopus embryonic epidermis has been shown to contain at least three cell types: ciliated cells, ionocytes and goblet cells (Dubaissi and Papalopulu, 2011). To investigate whether there are additional cell types, we combined double fluorescent in situ hybridisation using a probe for ciliated cells (α-1-tubulin) and a probe for ionocytes (atp6v1a), with antibody staining for goblet cells (anti-Xeel) and the nuclear marker, DAPI, at late tailbud stages (stage 32). As Fig. 1A shows, these three cell types do not account for all of the cells in the epidermis and there is at least one more cell type present (examples in yellow circles).

Foxa1 was previously shown to be expressed in a scattered, spotted epidermal distribution by in situ hybridisation in Xenopus laevis (Hayes et al., 2007) and we confirmed this in X. tropicalis (Fig. 1B). Combining foxa1 with the markers of the other cell types showed that foxa1 is expressed in the remaining cell type (Fig. 1C,D). With foxa1 included, no further gaps were present in the staining (Fig. 1D), which led us to conclude that the foxa1-positive cells complete the composition of the outer epidermal layer at late tailbud stages. But what are these cells and how do they arise?

The epidermal expression of foxa1 begins after the onset of foxj1 and foxi1 expression in ciliated cells and ionocytes, respectively, at about stage 13-14 (supplementary material Fig. S1). Meanwhile, combining markers for ciliated cells, ionocytes and goblet cells at mid-tailbud stages (stage 25) showed that all the outer layer cells are accounted for (data not shown), whereas at late tailbud stages they are not (Fig. 1A). This led us to hypothesise that the foxa1-positive cell type intercalates into the outer layer, like ciliated cells (Deblandre et al., 1999) and ionocytes (Dubaissi and Papalopulu, 2011), but at a later time point.

Indeed, foxa1 expression changes from inner to outer layer between early and late tailbud stages (Fig. 2A). When performing a transplant of a micro-ruby (MR)-labelled piece of outer layer of one embryo onto a fluorescein-dextran (FLDX)-labelled inner layer of another at gastrula stages (see Materials and methods), there is clearly a cell type (marked by yellow arrows) that intercalates into the outer layer in addition to ciliated cells (anti-AcTub antibody, blue) and ionocytes (anti-V1a, white) by stage 32 (Fig. 2B). To see if foxa1 cells do indeed intercalate, we transplanted an FLDX-labelled piece of outer layer of one embryo onto a Cascade Blue (CB)-labelled inner layer of another and allowed to develop until stage 32 before labelling with a foxa1 probe. Foxa1-labelled cells colocalise with CB cells and thus must originate from the inner layer (Fig. 2C). To check for the origin of goblet cells, we transplanted a piece of outer layer from one embryo (FLDX) onto the inner layer (MR) of another and reared until stage 32. The goblet cell marker, anti-Xeel shows colocalisation with FLDX, and hence the goblet cells must differentiate from the original outer layer cells (Fig. 2D).

Using membrane-GFP and co-staining with markers for ciliated cells, ionocytes and goblet cells, we were able to determine the distinct morphology of these cells. This new cell type is small in size compared with its neighbours but contains large apical openings at the surface, evident with mGFP staining and by scanning electron microscopy (SEM; Fig. 3). The SEM image shows ‘pores’ (average diameter of 1.3±0.09 μm) on the surface of this cell type, which are likely to correspond to vesicles that have opened out to release their content (arrow highlights possible secretory material on the surface). Images obtained by transmission electron microscopy (TEM; Fig. 3B), showed an abundance of large vesicles at the apical membrane. Other vesicles, which are presumably immature, were evident deeper within the cell (yellow arrow). Magnification of the apical vesicles showed that they contain material with a dark, presumably dense, core surrounded by lighter material. A comparison with earlier studies of the X. laevis embryonic epidermis showed that these cells have been described by their morphology before (Hayes et al., 2007; Montorzi et al., 2000). Owing to their size and apparent function, these cells were defined as small secretory cells (SSCs), a nomenclature retained here.

The mRNA for inositol trisphosphate kinase b (itpkb), an enzyme that catalyses the conversion of IP₃ to IP₄ and is involved in calcium signalling,_, is known to have a scattered, punctate epidermal expression pattern in Xenopus (Soto et al., 2013) and we confirmed this here (Fig. 4A). To see if itpkb is expressed in SSCs, we tested for colocalisation with foxa1 by double in situ hybridisation and found complete overlap (Fig. 4B). Like foxa1, itpkb did not colocalise with markers for the other cell types (supplementary material Fig. S2). An antibody generated against Itpkb showed localisation in or around the secretory vesicles of the SSCs (Fig. 4C). The material in these vesicles was also recognised by peanut agglutinin (PNA), a lectin that binds the carbohydrate sequence Gal-β(1-3)-GalNAc (Fig. 4D,E). PNA was originally shown to bind to material in the epidermis of X. laevis (Slack, 1985); however, the cell types that secrete it were not defined. Here, we show that they are foxa1-expressing cells (Fig. 4D). PNA also recognised material in the vesicles of goblet cells, but at a much lower level than the SSCs (Fig. 4E, yellow arrow). In addition to the epidermal staining, PNA stained material secreted from the cement gland, which is also positive for a mucin, muc5e (supplementary material Fig. S3).

Another carbohydrate epitope, HNK-1 (HSO₃-3GlcAβ1-3Galβ1-4GlcNAc), also showed staining in a subset of epidermal cells in a previous study, but again the cell type was not properly defined (Somasekhar and Nordlander, 1997). Using a monoclonal antibody we found some vesicles in SSCs were highly positive for HNK-1 (Fig. 4F). PNA and HNK-1 tend to be found in different vesicles (Fig. 4G). However, other cells showed some colocalisation between the two markers (data not shown). Therefore, it is difficult to say at this stage if they represent distinct molecular species or perhaps, different stages of maturation.

By analogy to the regulation of ionocytes by Foxi1 (Dubaissi and Papalopulu, 2011) and ciliated cells by Foxj1 (Stubbs et al., 2008; Yu et al., 2008), we hypothesised that Foxa1 may have a role in the development of SSCs. To test this, we knocked down foxa1 with a morpholino targeting the splice junction of the exon-1-intron 1 boundary (Fig. 5A). Using cell type-specific markers, we found that there was no change in the number of ciliated cells or goblet cells, but there was a large decrease in the number of SSCs (61% reduction; Fig. 5B). There was also a substantial increase in the number of ionocytes (46%), suggesting that these two cell types are somehow linked in their development.

It is possible that SSCs are not properly specified in foxa1 morphants. Alternatively, they may be correctly specified but undergo selective cell death early on. To distinguish these possibilities we examined the number of TUNEL-positive cells, together with an SSC marker (itpkb), at late neurula (stage 21) and late tailbud stage (stage 33) of foxa1 morphant embryos. There was some increase in epidermal cell death at late tailbud stages, but at early stages itpkb expression was significantly reduced, even though there was no increase in cell death (supplementary material Fig. S4). Thus, the decrease in the SSCs upon foxa1 knockdown is not primarily due to cell death.

Overexpression of foxa1 had the opposite effect in that it increased the number of SSCs (supplementary material Fig. S5A). However, this was only evident at a low dose of overexpression (10 pg), whereas higher doses (>50 pg) resulted in the appearance of ‘masses’ of PNA-negative, smaller cells (supplementary material Fig. S5B). This is consistent with previously suggested roles for FoxA1 in tumorigenesis (Nucera et al., 2009; Robinson et al., 2011; Yamaguchi et al., 2008).

When looking for the phenotypic consequences of foxa1 knockdown, we noticed that embryos showed a marked tendency to die shortly after hatching from the vitelline membrane (Fig. 6A). The appearance of the embryo culture was reminiscent of bacterial infection, which occasionally happens, as we do not routinely supplement the medium with antibiotics. We therefore tested whether we can rescue the phenotype with antibiotics.

Embryos injected with control morpholino (MOC) showed excellent survival levels 2 days post-fertilisation (approximately stage 35-37), both with and without application of the broad-spectrum antibiotic, gentamicin (20 μg/ml; Fig. 6A). However, foxa1 morphants, 2 days post-fertilisation, showed much reduced survival levels when cultured in 0.01× Marc’s modified Ringer’s (MMR) solution alone. Upon application of gentamicin, significant rescue of the survival rate was evident (Fig. 6A). Quantification of seven independent experiments (Fig. 6B) reveals that supplementing the media in which foxa1 morphants were reared with antibiotic led to an increase in survival rate from ∼27% to 80% (P<0.01). This compares to over 90% survival for control embryos with or without antibiotic.

To exclude the possibility that the phenotype is due to the increase in ionocytes that accompanies foxa1 MO injections (see Fig. 5B), rather than the decrease in SSCs, we increased specifically the number of ionocytes by overexpression of foxi1 (supplementary material Fig. S6A). This did not significantly affect the survival of the embryos (∼90% survival) in three independent experiments and there was no effect by the addition of antibiotics (supplementary material Fig. S6B).

Loss of SSCs could cause the death of the embryos by having a major effect on the integrity of the epidermis. To answer this question we analysed control and morphants embryos by membrane GFP expression. We have found no obvious difference from the controls at late neurula and tadpole stages, although we did notice an increased tendency of some epidermal cells to round up at the later stages of analysis. However, cell-cell contact appeared normal (supplementary material Fig. S7).

Given the protective action of the SSCs, we sought to identify components of the secretory material that are likely to account for this activity. As PNA recognises the material in the majority of the vesicles in the SSCs (Fig. 4E), we used PNA-HRP as a reagent for further investigation. Two PNA-positive species above 100 kDa were observed when the cement gland was present, but only a single species when the cement gland was removed (Fig. 7A, first four lanes). Comparing cement gland lysate and material aspirated directly from the epidermal surface confirmed that the species running lower in the gel corresponds to epidermal secretions and the material higher in the gel to the cement gland secretions (Fig. 7A, last two lanes). Comparing unreduced (-DTT) and reduced (+DTT) secretions from wild type without cement gland (WT - CG) embryos showed that the reduced samples migrate further in the gel (Fig. 7B). This indicates that the native PNA-positive material can form oligomers (and potentially polymers) through disulphide bonding. A comparison of secretions from foxa1 morphants and controls showed a substantial decrease in the PNA-positive material corresponding to epidermal secretions (Fig. 7C, red arrow). There was little difference in the band corresponding to the cement gland material (Fig. 7C, blue arrow).

The apparent properties of high molecular weight, glycosylation and formation of oligomers through disulphide bonding is reminiscent of mucin glycoproteins. Mucins can be isolated based on their density, so to further enrich the PNA-positive material secreted from the epidermis we separated the samples using caesium chloride (CsCl) density gradient centrifugation. The chart in Fig. 7D shows a peak of PNA reactivity for fractions of density 1.3-1.4 g/ml, which is similar to the density of known mucins (Thornton et al., 1995). PNA-positive fractions were pooled and analysed by mass spectrometry, alongside samples taken before purification from embryos with and without cement gland, to identify any active components.

Table 1 summarises the top protein hits in mass spectrometry for the three sets of samples. One of the top hits in all three samples was a predicted protein, LOC100496483, which matched to a locus in the X. tropicalis genome (Hellsten et al., 2010). This protein appears most closely related (70% identity at the amino acid level) with a glycoprotein called Otogelin described in X. laevis (Hayes et al., 2007), so we called this protein Otogelin-like. As this protein gave the highest hit after purification by density gradient centrifugation and pooling PNA fractions, we postulated that this is the molecule recognised by PNA and secreted from the epidermis. To test this notion we generated an RNA probe to the otogelin-like transcript and observed its expression pattern at different stages of development (Fig. 7E). At stage 20, otogelin-like is expressed principally in the goblet cells, but by stage 32, there is strongest staining in the SSCs (supplementary material Fig. S8A) and a lower level in goblet cells and ionocytes, but not in the ciliated cells (supplementary material Fig. S8B-D). However, by stage 42 the expression of otogelin-like is almost completely confined to the SSCs. Figure 7F shows that the otogelin-like transcript has an overlapping expression pattern with PNA and this is particularly prominent in the SSCs.

To test the effect of otogelin-like depletion, we knocked it down using MO injections (Fig. 8A). Although in morphant embryos, there was a clear reduction in PNA staining (Fig. 8B) and an alteration in the morphology of secretory vesicles of goblet cells and SSCs by TEM (Fig. 8C), the viability of the embryos was normal until at least advanced tadpole stage (stage 42; data not shown). At that point, embryos showed swelling and eventually died, but this was not rescuable by antibiotics (data not shown). Together with earlier data, these results strongly suggest that the dense, oligomeric glycoprotein marked by PNA; secreted from the epidermis and reduced upon SSC depletion is the Otogelin-like species. However, knockdown of otogelin-like alone is not sufficient to phenocopy the foxa1 morphants.

Of the other hits in mass spectrometry, the IgGF_cγ binding protein (Fcgbp) was also found in the purified samples, which is interesting because it has been identified previously in intestinal mucus samples (Johansson et al., 2009). One predicted mucin, Muc5e, was identified but only in the WT+CG sample, indicating that it is secreted from the cement gland, which we confirmed by in situ hybridisation for muc5e (supplementary material Fig. S3B). Other interesting proteins present in the material before purification include two complement factors as well as the glycolipoproteins, vitellogenin and apolipoprotein B, which are implicated in innate immunity (Peterson et al., 2008; Zhang et al., 2011). It remains to be seen whether any of these candidates are secreted from epidermal cells, in particular the SSCs, and whether they contribute to innate immunity.

Using a combination of markers for the known epidermal cell types, new markers and ultrastructural analyses we have characterised a new cell type, the SSCs, in the Xenopus larval epidermis. With the characterisation of these secretory cells, all the cell types in the epidermis up to tadpole stages are accounted for. This is an important step in being able to fully utilise this system as a simple model to study the development and function of mucociliary epithelia and the interactions between cell types.

There has been some confusion in the literature as to the true identity of these cells. The fact that they are nonciliated, undergo an intercalation event and have a scattered epidermal distribution had led many, including ourselves, to assume that these cells are ionocytes (Dubaissi and Papalopulu, 2011; Quigley et al., 2011). Indeed, many of the gene markers for ionocytes and SSCs have been grouped together as being part of one cell type (Hayes et al., 2007). Our current study addresses this issue and shows that the SSCs are indeed a distinct cell type.

The transcription factor Foxa1 was found to be integral for the development of the SSCs. It is of note that this same transcription factor has been shown to be crucial in the development of numerous secretory cell types in mammals, including goblet cells and clara cells in the lung, and enteroendocrine cells in the gut (Besnard et al., 2004; Gao et al., 2010; Ye and Kaestner, 2009). In addition, Foxa1 has also been shown to directly transactivate expression of mucin genes (Jonckheere et al., 2007; van der Sluis et al., 2008). Together with our results, these findings suggest an evolutionarily conserved role of FoxA1 genes in regulating secretory cell fate in epithelial tissues.

Embryos that develop without SSCs, achieved by knocking down Foxa1, die around hatching, at a time that is coincident with SSC intercalation into the outer layer. We were able to rescue embryos from dying by the addition of antibiotics into the media, indicating that these cells provide a first line of defence against infection. We suggest that there are at least three ways by which SSCs may protect the embryo from infection.

First, SSCs may protect the epidermis from invasion by producing a mucus layer, which can trap bacteria and other particulate material and exclude its entry to the underlying epithelium. The major structural components of mucus are large, polymeric gel-forming glycoproteins called mucins (Thornton et al., 2008). Our EM data showed that the vesicles of SSCs contain dense material that visually resemble mucins in mammalian goblet cells (Wang et al., 2009). However, our extensive analysis of secretions did not identify any of the predicted mucins in the X. tropicalis genome, despite the fact that there are at least 25 (Lang et al., 2007). Indeed, as far as we are aware, the only embryonic mucin with known expression in Xenopus is our identification of Muc5e, which is produced by the cement gland. This is in contrast to the adult frog skin, where putative mucins have been identified in subdermal mucous glands and are thought to be secreted onto the surface of the skin (Hauser and Hoffmann, 1992; Schumacher et al., 1994).

Our screen did, however, identify another glycoprotein, Otogelin-like, which has characteristics of mucins. The otogelins are evolutionarily related to mucins, with both containing von Willebrand factor (vWF) domains and cysteine knot domains that facilitate disulphide bonding of monomers to generate large polymers (Lang et al., 2007). Otogelin has been identified in fibre-like structures in the acellular membranes of the inner ear in mammals including humans (Cohen-Salmon et al., 1997; Cohen-Salmon et al., 1999), and mutations in the OTOG gene can result in deafness (Schraders et al., 2012). The fact that Otogelin-like is glycosylated and able to form oligomers suggests that it is capable of forming the matrix of a mucus-like barrier over the surface of the embryo. We suggest that the Otogelin-like protein can undergo massive expansion once it is secreted to form a layer over the surface, just like mucins do (Kesimer et al., 2010). However, our knockdown experiments suggest that this Otogelin-like protein is not sufficient on its own to mediate anti-microbial defence of the epidermis and that there are likely to be other secreted molecules. Indeed, our EM analysis shows that some secretory material (of unknown identity) remains in the SSCs, even once Otogelin-like is knocked down (Fig. 8C).

Identification of the IgGFcγ-binding protein (FCGBP) is also interesting in terms of the generation of a mucus layer because it has been found to be covalently associated with the mucin Muc2 in the intestines of mice (Johansson et al., 2009). It also has many vWF domains, and has been postulated to cross-link mucins, adding stability to the mucus layer. It has also been shown to interact with other molecules in mucus, including the trefoil factors (Albert et al., 2010). Thus, FCGBP could also contribute to a mucus layer on the surface of the tadpole skin, although further tests are needed to confirm this.

A second possible way in which SSCs could provide protection against infection is to secrete innate defence and anti-infective molecules onto the surface. Adult frogs are known to secrete a potent cocktail of antimicrobial peptides (AMPs) onto the skin, including the magainins, from subdermal granular glands (Soravia et al., 1988; Zasloff, 1987). However, these AMPs have never been isolated in tadpoles before metamorphosis despite several attempts to do so (Zasloff, 2009). So how do the tadpoles protect themselves? Candidates in our mass spectrometry screen may give some clues. Some of our top hits were for the glycolipoproteins vitellogenin and apolipoprotein B. Vitellogenin is most commonly known as an egg yolk precursor protein that provides the energy source for developing embryos (Jorgensen et al., 2009). However, it has also been shown to have additional roles in innate immunity (Zhang et al., 2011). It has active antibacterial activity as well as the ability to recognise bacterial structures and promote phagocytosis (Li et al., 2008; Tong et al., 2010). Vitellogenin has also been identified as an antimicrobial agent in fish mucus (Meucci and Arukwe, 2005; Shi et al., 2006). Meanwhile, apolipoprotein B has been shown to antagonise and inhibit expression of invasive genes by Staphylococcus aureus and mice deficient for apolipoprotein B are more susceptible to infection (Peterson et al., 2008). The identification of the complement proteins, C3 and C9, in our mass spectrometry analysis is also of note because the complement cascade is a well-known innate defence system (Dunkelberger and Song, 2010; Janssen et al., 2005). Perhaps some of these molecules are secreted from the SSCs in addition to Otogelin-like. Thus, as summarised in our model in Fig. 9, antimicrobial defence may be due to the combinatorial action of a number of molecules.

Finally, another contributor to the phenotype of enhanced susceptibility to infection could be an indirect effect of depleting SSCs on neighbouring cells. The accompanying paper by Walentek et al. describes how secretion of serotonin from the SSCs enhances the beating of cilia in adjacent ciliated cells (Walentek et al., 2014). In the absence of SSCs, the lack of serotonin could reduce ciliary beating and affect clearance of the mucus-like substance, leading to conditions where pathogens can thrive and infect the epidermis.

Through the initial characterisation of these cells, we have opened up the possibility for future studies into their role in innate immunity as well as other unexplored functions. The resemblance with the mucociliary epithelium of the respiratory tract is striking, and these similarities make the Xenopus larval skin a potentially very powerful model to study development, function and breakdown of these important epithelia.

All morpholinos were purchased from Gene Tools. The sequence of the foxa1 splice morpholino is 5′-GGATTCCTTCTCTTACCTCCTGGGT-3′. The otogelin-like splice morpholino sequence is 5′-TAGAGTCATACATACCTCCATCATC-3′. The control morpholino has the sequence 5′-CCTCTTACCTCAGTTACAATTTATA-3′. Morpholinos were injected with a 15 ng dose at the one-cell stage unless otherwise stated.

Constructs for generating probes were obtained from expressed sequence tag (EST) clones in the Xenopus full-length database (Gilchrist et al., 2004). The clone names were as follows: atp6v1a (TTpA014f09), α-1-tubulin (TNeu122k16), otogelin-like (THdA045k18), foxa1 (TNeu062i17), itpkb (XM_002938844), foxi1 (TNeu069d21), foxj1 (TNeu058m03) and muc5e (TTbA069h06). The constructs were linearised and transcribed with T7 RNA polymerase (Promega) to generate antisense RNA probes. The probes were labelled with fluorescein isothiocyanate (FITC), digoxygenin (Dig) or dinitrophenol (DNP). In vitro transcription was employed to generate GAP43-GFP mRNA from a pCS2-GAP43-GFP construct for injection into embryos and to mark membranes (Kim et al., 1998). The construct was linearised with NotI and transcribed with SP6 RNA polymerase (Promega).

Chromogenic in situ hybridisation was carried out as described (Harland, 1991). Multicolour fluorescence in situ hybridisation was performed using tyramide amplification after addition of probes and antibodies conjugated to horseradish peroxidase (Lea et al., 2012).

Immunofluorescence was largely carried out as described previously (Dubaissi and Papalopulu, 2011; Dubaissi et al., 2012). All primary antibodies for immunofluorescence were used at a dilution of 1:1000. Secondary antibodies (Invitrogen) conjugated to fluorophores were used at a dilution of 1:500. The following primary antibodies were used: mouse anti-acetylated α-tubulin (T7451; Sigma), rabbit anti-V1a (ST170; a gift from Shigeyasu Tanaka, Shizuoka University, Japan), mouse-anti-Xeel (a gift from Saburo Nagata, Japan’s Women’s University, Tokyo, Japan), mouse anti-HNK-1 antibody (C6680; Sigma), rabbit anti-GFP antibody (A-6455; Life Technologies) and a custom-made rabbit anti-Itpkb antibody (Cambridge Research Biochemicals, UK). The lectin, PNA conjugated to Alexa Fluor 568 (Life Technologies), was used at a dilution of 1:1000.

Embryos were injected with the fluorophores MR, FLDX or CB (all Invitrogen), at the eight-cell stage in two ventral blastomeres fated to become epidermal tissue. The embryos were reared in 0.01× MMR until gastrulation and then transferred to transplantation buffer [0.5× MMR, 1% Ficoll and 5 μg/ml gentamicin (Sigma)]. Tungsten wires were used to remove a section of outer layer epidermis from one set of fluorophore-injected embryos. A section of outer layer tissue from a different set of fluorophore-injected embryos was then transplanted onto the inner layer of the first set and the embryos were left for 2 hours to heal. The embryos were then transferred to post-transplantation buffer (0.1× MMR, 5 μg/ml gentamicin) and reared until stage 32, when they were fixed for analysis.

The TUNEL protocol on Xenopus embryos was performed largely as described previously (Hensey and Gautier, 1998). In summary, embryos were fixed in MEMFA and stored in methanol. Embryos were bleached (70% methanol, 10% H₂O₂, 5% formamide) to remove pigment and then rehydrated into PBS. Embryos were incubated in 150 U/ml terminal deoxynucleotidyltransferase (Invitrogen) and digoxygenin-dUTP (1:1000, Roche). Reaction was terminated with 1 mM EDTA in PBS. The normal in situ hybridisation protocol was then followed using anti-digoxygenin antibody conjugated to alkaline phosphatase and chromogenic reaction using nitro-blue tetrazolium chloride/5-bromo-4-chloro-3′-indolyphosphate p-toluidine salt (NBT/BCIP) as a substrate (Harland, 1991).

Embryos were reared in 0.01× MMR for 24 hours until stages 22-24 and then, if required, cement glands were removed with forceps in 0.01× MMR + gentamicin (10 μg/ml). Embryos were then reared until stage 35. To get enough material, 300 embryos were packed into a well containing 600 μl of media. The embryos were reared for 4 hours at room temperature with agitation every 20 minutes. Embryos were then anaesthetised with tricaine and the media removed for analysis of secretions. Material was also aspirated directly from the epidermis using glass Pasteur pipettes. Denaturants such as 6 M urea or 8 M guanidinium chloride were added to denature any proteins. If required, samples were reduced in 10 mM dithiothreitol (37°C for 1 hour) and alkylated in 25 mM iodoacetamide (30 minutes at room temperature in the dark). To concentrate the secretory material, the media samples were centrifuged through Vivaspin columns (Sartorius Stedim Biotech) that trap material above a certain molecular weight. We used filters with a 100,000 MW cut-off so any material above this size was retained.

Agarose gels are used to separate high molecular weight and highly charged molecules such as mucins. Agarose gels were made at 1% (w/v) in tris-acetate-EDTA (TAE). Samples were loaded and electrophoresed at 65 V for 4-6 hours. After separation, proteins were vacuum blotted on to nitrocellulose membranes for 90 minutes in 4× SSC. PNA conjugated to horseradish peroxidase (PNA-HRP, Sigma) was applied to the membranes at a dilution of 1:1000 for two hours. The membranes were then washed four times for 5 minutes each in TBS-Tween, reacted with chemiluminescence reagents (Sigma), and exposed to photographic film (Kodak).

Collected samples were separated by CsCl/4 M guanidinium chloride density gradient centrifugation (Thornton et al., 1999). The starting density was 1.4 g/ml and a Beckman Ti70 rotor was used to generate the gradient in an ultracentrifuge at 164,685 g for 68 hours at 15°C. Fractions (0.5 ml) were taken and assessed for PNA reactivity and density. Positive fractions were pooled, reduced and alkylated, and concentrated on Vivaspin columns before mass spectrometry.

Samples were exchanged into 2 M urea (50 mM ammonium bicarbonate) using Vivaspin columns and digested overnight in trypsin (1 μg, Sigma). The tryptic peptides were purified using ZipTips (Millipore) and solubilised in 0.1% formic acid. Tryptic peptides were separated by reverse phase liquid chromatography (LC) and analysed by tandem mass spectrometry (LC-MS/MS) using a NanoAcquity LC (Waters, Manchester, UK) coupled to a LTQ Velos mass spectrometer (Thermo Fisher Scientific). MS/MS data were searched using Mascot 2.4 (Matrix Science, UK) software against a custom database consisting of the SWISSPROT database with additional predicted Xenopus tropicalis-specific mucin sequences (Lang et al., 2007). The parameters were as follows: Carbamidomethyl (C) as fixed modification, Oxidation (M) as variable modification, peptides mass tolerance of 1.2 Da, fragments mass tolerance 0.6 Da, 1 missed cleavage maximum.

SEM and TEM were performed according to standard protocols. For SEM, embryos were fixed in 2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer (pH 7.2) and then postfixed in osmium tetroxide. Embryos were then critical point dried and sputter-coated with gold. Imaging was performed on an FEI Model Quanta 200 ESEM. For TEM, embryos were fixed overnight in 2% formaldehyde + 2% glutaraldehyde in 0.1 M cacodylate and then post-fixed in reduced osmium. The embryos were then put in 1% tannic acid and then 1% uranyl acetate, for 1 hour each. Samples were then dehydrated in ethanol and embedded in low viscosity resin. Sections of 70 nm were cut on a Reichert-Jung Ultracut E ultramicrotome and observed using an FEI Tecnai 12 electron microscope at 80 kV.

We thank staff in the EM facility in the Faculty of Life Sciences, University of Manchester, for their assistance, and the Wellcome Trust for equipment grant support to the EM facility; Aleksandr Mironov, Leslie Lockey and Patrick Hill for assistance with EM; Raphael Thuret for help with collection of samples and Bethan Powell for assistance with in situ hybridisation; Saburo Nagata and Shigeyasu Tanaka for donation of antibodies; Martin Blum for sharing data and reagents prior to publication; and Kristian Franze for some help in the early stages of analysis of the phenotype.