Adaptive correction of craniofacial defects in pre-metamorphic Xenopus laevis tadpoles involves thyroid hormone-independent tissue remodeling

Craniofacial abnormalities, such as cleft palate, fetal alcohol syndrome and microcephaly are seen in over 1/700 births worldwide among humans (Blanco-Davila, 2003; Gregg et al., 1994). Unfortunately, most craniofacial abnormalities affecting humans are untreatable and the few available treatment options involve expensive and invasive surgeries. Thus, research focusing not only on the causes, but also the resolution, of craniofacial defects in vertebrates has never been more crucial.

The spatiotemporally regulated pathways that coordinate cell division, migration, apoptosis and differentiation during craniofacial development are highly conserved across vertebrates (Simoes-Costa and Bronner, 2015; Square et al., 2015). These conserved gene regulatory networks are crucial for the specification and differentiation of key cell populations, such as the cranial neural crest cells: a vertebrate-specific cell population that actively migrates anteriorly during development and differentiates into several craniofacial tissues [e.g. neurons, cartilage, and connective tissue (Simões-Costa and Bronner, 2015; Simões-Costa et al., 2015)]. Environmental factors, including vitamin A levels and pH, also play a role in the specification and differentiation of cells during embryogenesis (Vandenberg et al., 2011; Yelin et al., 2007). Ultimately, the intricate processes involved in creating a head leave organisms vulnerable to the possibility of craniofacial defects during embryogenesis.

Craniofacial abnormalities that are commonly seen in humans can be recapitulated in several other vertebrates (e.g. mice, fish and frogs) through exposure to the same teratogens during development. For example, exposing X. laevis embryos to ethanol can cause eye and brain malformations, similar to those seen in humans with fetal alcohol syndrome (Fainsod and Kot-Leibovich, 2017; Kot-Leibovich and Fainsod, 2009; Yelin et al., 2007, 2005). The similarities among craniofacial phenotypes in vertebrates provide us with the great opportunity to use model organisms to better understand human craniofacial defects.

Although numerous studies have described various causes of craniofacial abnormalities in vertebrates, only one study has reported the natural resolution of craniofacial deformities in vertebrates. This study, carried out by Vandenberg et al., revealed that altering membrane voltage, pH and gene expression patterns in X. laevis embryos by knocking down H⁺-V-ATPase channel function induces a wide range of craniofacial defects in tadpoles, some of which are subsequently normalized over time (Vandenberg et al., 2012). Vandenberg et al. confirmed that the normalization of craniofacial defects occurred prior to metamorphosis by using geometric morphometric analysis to quantify the changes in morphology over time. Interestingly, an unrelated study examining normal X. laevis craniofacial structures noted that variance in craniofacial cartilage morphology among wild-type X. laevis tadpoles decreases prior to metamorphosis (Rose et al., 2015). Combined, these studies suggest that X. laevis tadpoles possess an endogenous mechanism responsible for monitoring and adaptively adjusting craniofacial morphology, prior to metamorphosis.

Thus, we resolved to further investigate the biological mechanisms responsible for the ability of pre-metamorphic X. laevis tadpoles to self-correct abnormal craniofacial phenotypes. This involved qualitative and quantitative analysis of whole-head and craniofacial tissue (i.e. cartilage and eye tissue) morphology, throughout pre-metamorphic stages, in tadpoles with a variety of craniofacial abnormalities; followed by gene expression profile analysis to identify the molecular factors involved in initiating and/or carrying out the self-correction of abnormal craniofacial morphology in these animals. For this research, a set of perturbations that induced consistent abnormal head phenotypes was required; the perturbations used included three different pharmacological exposures (thioridazine, ethanol and ICI 118,551) and one morpholino-based treatment (xMLTK-MO). All of these treatments vary in their known biological effects: thioridazine is a dopamine-receptor antagonist, ICI 118,551 is a β₂ inverse agonist, ethanol is a retinol antagonist and xMLTK-MO hinders chondrogenesis (Table 1) (Suzuki et al., 2012). Despite the differences in biological activity, these perturbations all reliably produce abnormal craniofacial morphologies among early pre-metamorphic X. laevis tadpoles. Once uniform abnormal craniofacial structures were induced, craniofacial morphology was examined at various developmental stages to probe the ability of X. laevis to remodel malformed structures prior to metamorphosis.

By monitoring the craniofacial deformities induced by these four different perturbations throughout pre-metamorphic life stages, we determined that pre-metamorphic X. laevis tadpoles can self-correct some, but not all, craniofacial malformations. We also compared cartilage tissue morphology and gene expression profiles between wild-type tadpoles, tadpoles with correctable craniofacial abnormalities and tadpoles with seemingly permanent craniofacial defects. Using this approach, we found that the improvement of overall craniofacial morphology in tadpoles with correctable craniofacial abnormalities is largely due to the remodeling of cartilage and eye tissue. Furthermore, this adaptive tissue remodeling is independent of thyroid hormone signaling, may depend on prolactin hormone signaling, and involves the upregulation of mmp1 and mmp13.

To examine the potential of tadpoles to self-correct malformed structures, several treatments were used to induce abnormal anterior structures. For example, to reproducibly alter craniofacial morphology (cartilage reduction and smaller eyes), we used a morpholino that has previously been shown to hinder chondrogenesis (xMLTK-MO) by Suzuki et al. (Suzuki et al., 2012) (Fig. 1A, Fig. S1). Additional reproducible craniofacial defects were attained through the temporally regulated exposure of neurula stage embryos to thioridazine, ethanol or ICI 118,551 (Fig. 1, Fig. S1). We found that all three of the pharmacological perturbations altered the expression pattern of the cranial neural crest marker sox9 in tailbud-stage X. laevis (Fig. S2A,B), which suggests that these treatments do not prevent the specification of cranial neural crest cells; instead, they may hinder CNCC migration, differentiation or proliferation. Although xMLTK-MO does not perturb sox9 spatiotemporal expression in developing X. laevis, it hinders chondrogenesis by preventing the upregulation of a chondrogenic transcription factor gene, sox6, by Sox9 (Suzuki et al., 2012).

Each of the pharmacological treatment groups displayed abnormal craniofacial phenotypes. Finite thioridazine exposure during neurulation results in an abnormally rectangular head morphology, as opposed to the more-rounded overall head morphology of stage-matched controls (Fig. 1A). This rectangular head phenotype emerges from the decreased branchial arch size (Fig. 1A) and malformed mouth (Fig. 1B) of the thioridazine-treated tadpole, characteristics that were seen in virtually all specimens exposed to thioridazine (Fig. S1). In addition to the rectangular head phenotype, thioridazine-treated tadpoles frequently have misshaped eyes (Fig. 1A). Similarly, the ICI 118,551-exposed embryos developed into tadpoles with a more-trapezoidal head phenotype, due to a drastically underdeveloped mouth region and reduction in overall head size (Fig. 1A,B). The ethanol exposure (which was also restricted to NF stages 13-26) consistently produced viable tadpoles with reduced overall head size, occasionally with misshaped eyes (Fig. 1A).

Once facial defects had been induced, geometric morphometric analysis was used to quantify the changes to and resolution of malformed or mispositioned craniofacial features (e.g. eyes, nostrils and mouth) throughout pre-metamorphic stages. Focusing on the shape of the eye and relative angles between the midline and the nostrils or mouth edges allowed us to efficiently compare morphology across control and experimental groups, without having to adjust for differences in eye or head size (Fig. 2). Ultimately, this size-independent morphological analysis allowed us to track and investigate significant differences in allometric growth (changes in shape), as opposed to isometric growth.

Following this approach, eye shape was evaluated based on the ratio between the anterior-posterior and lateral axes of each eye. With this measure, we found that ethanol-, thioridazine- and xMLTK-MO-treated tadpoles showed a significant difference in the mean shape of their eye(s) relative to wild-type controls at NF stage 45 (Mann Whitney U-test, P<0.001) (Fig. 2A). Specifically, the eyes of the experimental tadpoles tended to have elongated lateral axes at NF stage 45, correlating with their abnormal eye tissue expansion towards the midline (Figs 1A and 2). The difference between control and experimental groups continuously decreased as the tadpoles progressed to NF stage 49. All three experimental groups clearly demonstrated a large improvement in eye shape between NF stages 45 and 49. In fact, when it came to the mean left eye ratios, the xMLTK-MO-treated group was no longer statistically distinguishable from controls at NF stage 49 (Fig. 2A). We must, however, note that although the xMLTK-MO-treated group showed improvement in abnormal eye shapes, the size of the affected left eye remained smaller than the control right eye (Fig. 1A).

At NF stage 49, the thioridazine-treated group showed the greatest improvement in eye shape (Fig. 2A,B). The mean difference in left and right eye ratios for these experimental tadpoles went from 0.20±0.02 (mean±s.e.m.) and 0.17±0.02 at NF stage 45, to only 0.07±0.02 (mean±s.e.m.) and 0.06±0.02 by NF stage 49. Furthermore, when we tracked the eye shape of individual tadpoles from the thioridazine experimental group, most of the experimental individuals that have abnormal eye shapes early on, match controls by stage 50 (Fig. S4). We also found that it was possible for an individual thioridazine-treated tadpole with two malformed eyes at NF stage 45 to show drastic improvement in the shape of one eye and not the other over time (Fig. 1A). Overall, we did see improvement in the abnormal eye shapes in the thioridazine, ethanol and xMLTK-MO experimental groups; however, not all the malformations were perfectly corrected by stage 50.

For the relative angles between the midline and the nostrils, the ethanol- and xMLTK-MO-treated groups did not differ from the control values by more than 5° at any measured stage (Fig. 2C,D). At NF stage 45, the ethanol- and xMLTK-MO-treated groups have left nostril angle values that do not significantly differ from controls at NF 45 (Kruskal-Wallis U-test, P>0.05). Although they do show statistically significantly different nostril angles at NF stage 49 (Kruskal-Wallis U-test, P<0.001), it is only because the variance observed within control and experimental groups decreased as they progressed from early to late pre-metamorphic stages (Fig. 2C). The thioridazine-treated group again showed the greatest improvement with regards to the nostril-midline angle (Fig. 2C,D). The mean difference in the angle between the nostrils and midline of the thioridazine-treated tadpoles significantly differed from wild-type controls by more than 10° at NF stage 45 (Mann-Witney U-test, P<0.001) (Fig. 2C,D). Yet, by NF stage 49, they only differed from controls by less than 3° (Mann-Witney U-test, P<0.01) (Fig. 2C,D).

A slightly different pattern appeared when we measured the angle between the corner of the mouth and the midline of the experimental groups. At NF stage 45, both the left and right-side mouth measurements for the thioridazine-treated tadpoles were significantly different from controls (Mann-Witney U-test, P<0.001), but by NF stage 49 they had corrected so much that it was no longer statistically distinguishable from wild-type controls (Mann-Witney U-test, P>0.05) (Fig. 2E,F). We also saw that, on average, the xMLTK-MO-treated tadpoles showed irregular acute angles between the mouth corner and midline of their uninjected (right) side than controls at NF stage 45 (Mann-Witney U-test, P<0.001) (Fig. 2F). Although the right mouth corner angle in xMLTK-MO-treated tadpoles more closely matched controls by NF stage 49 (Fig. 2F), it coincided with the left mouth corner angle also becoming significantly more acute than that seen in wild-type control at that stage (Fig. 2E). Although the ethanol-treated tadpoles did not reveal significantly different mean relative mouth corner to midline angles (Fig. 2E,F), when individual ethanol-treated tadpoles were tracked, there were some that, at NF stage 46, had much more acute angles than controls, but by NF stage 50 were not differentiable from controls (Fig. S4).

Given that the chondrocranium is the main determinant of head shape in pre-metamorphic tadpoles, we investigated whether the observed changes in overall head morphology were primarily due to changes in the cartilaginous scaffolding for the head. We characterized and scored changes in cartilage tissue morphology during self-correction of abnormal craniofacial at NF stages 45, 47 and 49 (Fig. 3A-C). The qualitative comparison of representative cartilage stains for control, xMLTK-MO, thioridazine, ethanol and ICI 118,551 experimental groups at NF stage 45 and 49 revealed that there is detectable improvement in the cartilage morphology of xMLTK-MO-, thioridazine- and ethanol-treated tadpoles, but not ICI 118,551-treated tadpoles (Fig. 3A,B).

With the unperturbed control group, we observed a small (3-4%) drop in the mean percentage of severe phenotypes among control tadpoles between stage 45 and 47 (Fig. 3C). However, the craniofacial cartilage scoring distribution of the control group was not significantly different at NF stages 45-47, 47-49 or 45-49 [goodness of fit (GOF) test; χ²<4.21; d.f.=2; P>0.05] (Fig. 3C). On average, with conservative phenotype scoring (Fig. S3), no more than 10% of the control tadpoles had severely or moderately malformed craniofacial cartilage at any given stage (Fig. 3C).

Given that xMLTK-MO specifically hinders chondrogenesis, it is not surprising that ∼80% of xMLTK-MO-injected tadpoles had noticeably malformed craniofacial cartilage at NF stage 45 (Fig. 3C). These malformations were dramatically reduced a few stages later (NF stg. 49) when only 43% of treated animals had malformed craniofacial cartilage (Fig. 3C). Although there was a significant difference between stage 45 and 49 (GOF test; χ²=102.59; d.f.=2; P<0.001), further comparisons between NF stages 45-47 (GOF test; χ²=6.31; d.f.=2; P>0.05) and 47-49 (GOF test; χ²=43.89; d.f.=2; P<0.001) revealed that this was primarily due to changes between the latter pair (Fig. 3C). These data suggest that adaptive cartilage remodeling primarily occurs during and after NF stage 47 in xMLTK-MO-treated tadpoles. Thus, we have confirmed that the malformed cartilage in xMLTK-MO-treated tadpoles improves between stages 45 and 49, although the improvement seen in xMLTK-MO-treated tadpoles was not as drastic as that seen with the thioridazine-treated tadpoles.

For the thioridazine experimental group, the craniofacial cartilage scoring distribution is significantly different between all stages examined (NF stages 45-47, 47-49 and 49-50; GOF test; 196.63≥χ²≥19.99; d.f.=2; P<0.001) (Fig. 3C). At stage 45, more than 60% of the thioridazine-treated tadpoles had severely malformed craniofacial cartilage, but only 13% were severely malformed by stage 49 (Fig. 3C). This suggests that the malformed cartilage is adaptively remodeling its shape during pre-metamorphic stages to reach a target morphology.

Even the ethanol-treated tadpoles, which possessed the mildest cartilage defects of all the experimental groups, still demonstrated a drop from 24% to only 12% severe craniofacial cartilage phenotypes between NF stages 45 and 49 (Fig. 3C). Craniofacial scoring distributions for the ethanol-treated tadpoles were significantly different between NF stage 45 and 47 cartilage (GOF test; χ²=14.23; d.f.=2; P<0.001), but not between stages 47 and 49 (GOF test; χ²=9.80; d.f.=2; P>0.05) (Fig. 3C). Therefore, they do not undergo cartilage remodeling under the same timeline as the xMLTK-MO- or thioridazine-treated tadpoles.

Finally, ICI 118,661 exposure had the most dramatic impact on craniofacial cartilage of all the experiment groups. At any given pre-metamorphic stage, 90-98% of the ICI 118,551-treated tadpoles had severely malformed (or missing) craniofacial cartilage (Fig. 3C). For the ICI 118,551-treated group, the craniofacial cartilage scoring distribution is significantly different between NF stages 45 and 47 (GOF test; χ²=42.37; d.f.=2; P<0.001), but not between stages 47 and 49 (GOF test; χ²=0; d.f.=2; P>0.05) (Fig. 3C). These data suggest that if craniofacial cartilaginous features are missing, they cannot be formed de novo in pre-metamorphic tadpoles, eliminating the possibility that improvement in abnormal cartilage morphology at these stages is a regenerative mechanism.

Given that X. laevis are known to experience major tissue remodeling during metamorphosis, we looked first towards the possible precocious involvement of thyroid hormone-dependent metamorphic pathways in the self-correction of abnormal craniofacial morphology. In wild-type tadpoles, thyroid hormone receptor α (thrα) and retinoid X receptor α (rxrα) gene expression increases at pre-metamorphic stages and their gene products form heterodimers that repress metamorphic genes, such as thyroid hormone receptor β (thrβ) and matrix metalloproteinase 2 (mmp2) (Buchholz et al., 2006; Lee and Privalsky, 2005; Li et al., 2002; Wu et al., 2001). Only when thyroid hormone begins to be produced and binds to the THRα/RXRα heterodimers, do the nuclear receptors become activators of metamorphic gene expression (Das et al., 2006; Fujimoto et al., 2007).

As with truly metamorphic X. laevis tadpoles, tadpoles undergoing precocious metamorphosis show an upregulation of thrβ and mmp2 (Veldhoen et al., 2014). Precocious metamorphosis can be triggered in pre-metamorphic tadpoles through exogenous exposure to the more biologically active form of thyroid hormone, triiodothyronine (T3) (Kerney et al., 2012). However, these precocious metamorphic cases are often extreme in their effect on craniofacial tissue remodeling and are frequently lethal. Consequently, we exposed wild-type pre-metamorphic tadpoles from stage 45 to 50 to a very low dose of T3 to trigger a milder form of precocious metamorphosis (Fig. S5). The gene expression profiles of these +T3 specimens represent a positive control for non-lethal, premature induction of the TH pathway.

We quantified the gene expression levels (relative to wild-type controls) for thrα, rxrα, thrβ and mmp2 among +T3-, xMLTK-MO-, ethanol-, thioridazine- and ICI 118,551-treated pre-metamorphic tadpoles at various pre-metamorphic stages (Fig. 4). We also looked at the expression of retinoic acid receptor α (rarα) (Fig. 4). RARα is a nuclear receptor that also heterodimerized with RXRα; however, it is not directly activated by thyroid hormone (Escriva et al., 2006; Lee and Privalsky, 2005). Although fold changes in mmp2, rxrα and thrα among the experimental groups (particularly the ethanol and xMLTK-MO groups) were found to be significantly different from each other at NF stage 45 (Kruskal-Wallis test, d.f.=4, P<0.05), none of the experimental groups demonstrated statistically significant changes relative to wild-type controls (Dunn's test, d.f.=4, P>0.05) (Fig. 4A). Similarly, at stage 47, neither the experimental groups nor the +T3-positive control group displayed significantly altered gene expression relative to wild-type controls (Kruskal-Wallis test or Dunn's test, d.f.=4, P>0.05) (Fig. 4B). However, for rxrα at this stage, the +T3-positive control and xMLTK-MO-treated groups had opposing effects on gene expression and were found to be significantly different from each other (Dunn's test, d.f.=4, P<0.05) (Fig. 4B). In addition, the +T3-treated group and thioridazine-treated groups demonstrated the opposite effect on thrβ (Dunn's test, d.f.=4, P<0.05) (Fig. 4B).

By stage 49, the +T3 tadpole group began not only to demonstrate noticeable thyroid hormone-dependent morphological differences from wild-type controls (Fig. S5), they also began to show significant differences in mmp2 and thrβ expression relative to wild-type controls (Dunn's test, d.f.=4, P<0.05) (Fig. 4C). Interestingly, the ICI 118,551- and thioridazine-treated groups also showed a biologically significant threefold increase in thrβ expression at NF stage 49; however, this was not statistically significant (Dunn's test, d.f.=4, P>0.05) (Fig. 4C). Furthermore, by the following developmental stage, thrβ expression levels in both the ICI 118,551- and thioridazine-treated groups show no fold change in expression relative to wild-type controls. Meanwhile, the +T3-treated group continued to show elevated mmp2 and thrβ expression (Fig. 4D).

Taken together, the discrepancies among the gene expression profiles of tadpoles with self-correcting craniofacial abnormalities and our +T3-positive control, lead us to conclude that direct activation of the TH signaling pathway by T3 is not occurring in any of the experimental groups. To further explore these observations, we examined the effects of blocking TH signaling during the craniofacial remodeling process using the anti-thyroid drug methimazole (Degitz et al., 2005; Hornung et al., 2010; Tietge et al., 2010). Inhibiting T3 synthesis in pre-metamorphic tadpoles with correctable craniofacial defects does not affect their ability to normalize, providing strong evidence that the self-correction process is not mediated via TH-dependent events (Fig. 5). Furthermore, along with craniofacial morphological changes, the low dose T3 exposure of stage 45-50 tadpoles also caused accelerated limb bud growth (Fig. S5), a phenotype that has not been observed in any of the experimental groups.

Having ruled out the most obvious candidate tissue remodeling pathway, we decided to take a more global RNA-seq-based approach to identifying the molecular pathways involved in the self-correction of abnormal craniofacial morphology in X. laevis larvae. For this experiment, we used whole-head tissue samples from thioridazine-treated tadpoles with malformed craniofacial features and compared them with wild-type controls, at NF stages 45 and 47. We chose thioridazine-treated tadpoles for this experiment because they demonstrated the most consistent abnormal phenotypes and greatest adaptive morphological changes between early and late pre-metamorphic stages. Any tissue-remodeling gene candidates found to be differentially expressed in thioridazine-treated tadpole craniofacial tissue could later be studied via RT-qPCR in other experimental groups.

We found 101 differentially expressed genes between control and thioridazine-treated samples at NF stages 45 and 135 at NF stage 47 (Tables S1, S2). We then performed GO pathway enrichment analysis on the significant gene lists by converting the X. laevis gene symbols to Xenopus tropicalis gene symbols (Table 2, Table 3). When we examined the list of differentially expressed genes, we identified factors that correlated with upregulation (in the thioridazine tadpoles) of endocrine hormone pathways, such as gonadotropin-releasing hormone and prolactin hormone signaling.

The RNA-seq experiment specifically revealed upregulation of the protein-coding sequence for a prolactin 2 protein (prolactin.2.S) (Tables 2 and 3). From previous studies in Xenopus and other animals, we know that prolactin signaling can induce tissue remodeling responses in various tissues (Huang and Brown, 2000; Huang et al., 2009; Jung et al., 2004; Parhar, 2003). Thus, we generated qPCR primers specific to the prolactin.2.S sequence to further investigate this hormone in our treated tadpoles with craniofacial defects. We also generated qPCR primers for the prolactin receptor gene after confirming that the prolactin receptor-binding region was conserved among prolactin.1 and prolactin.2 in X. laevis. With RT-qPCR, we examined the expression of the prolactin.2.S and prolactin receptor gene in the brain tissue of thioridazine-, ethanol- and ICI 118,551-treated tadpoles at NF stages 45 and 47 (Fig. S6A,B). Thioridazine-treated tadpoles exhibited a threefold increase in prolactin.2.S at NF stages 45 and a fivefold increase at stage 47, with no change in the expression of prolactin receptor at either stage (Fig. S6). Thus, the RNA-seq hit for this prolactin gene has been affirmed.

Prolactin gene expression has been shown to be negatively regulated by dopamine signaling in several animals (Fitzgerald and Dinan, 2008; Freeman et al., 2000; Treier and Rosenfeld, 1996). Therefore, we deemed it necessary to investigate the effect that increased dopamine signaling and subsequent downregulation of prolactin has on the correction of thioridazine-induced craniofacial defects in X. laevis. For this, we exposed control and thioridazine-treated tadpoles to a dopamine signaling agonist, pergolide mesylate (PM), during the stages where we previously observed the correction of craniofacial malformations (NF stages 45-49). We found that the +PM thioridazine-treated group was still able to correct abnormal eye shapes (Fig. 6). Thus, dopamine and prolactin signaling do not appear to be necessary for the remodeling of abnormal eye tissue. However, PM exposure did block the ability of thioridazine-treated tadpoles to correct abnormal angles between the nostrils and the midline. Furthermore, PM exposure significantly affected the angles between the midline and the nostrils of tadpoles that originally had wild-type craniofacial phenotypes at NF stage 45, by NF stage 49 (MWU Test, P<0.001) (Fig. 6, Fig S7). A similar but milder effect was also seen with the angle between the midline and the edges of the mouth: at stage 49, both +PM groups had slightly more obtuse mean angle values than −PM controls. Thus, increasing dopamine signaling via PM exposure disrupts the normal maintenance of overall craniofacial morphology in wild-type animals, as well as the correction of abnormal craniofacial morphology in animals with thioridazine-induced craniofacial defects.

Given that matrix metalloproteinases (MMPs) are key factors for tissue remodeling in animals (Jabłońska-Trypuć et al., 2016) and that prolactin signaling regulates mmp1 and mmp13 expression in X. laevis (Jung et al., 2004), we chose to examine the expression of these genes at NF stages 47 and 49 in xMLTK-MO-, ethanol-, thioridazine- and ICI 118, 551-treated craniofacial tissue. We determined that mmp13 was significantly upregulated in xMLTK-MO-treated craniofacial tissues at NF stage 47 (Dunn's test, d.f.=4, P<0.05) (Fig. 7A). Although both mmp1 and mmp13 were upregulated in ethanol-, thioridazine- and ICI 118,551-treated tadpoles at NF stage 49 (Dunn's test, d.f.=4, P<0.05) (Fig. 7B). These data correlate with there being a tissue remodeling response occurring in the craniofacial tissues of these tadpoles that involves various MMPs.

This work demonstrates that X. laevis pre-metamorphic tadpoles are capable of correcting abnormal craniofacial phenotypes resulting from several different embryonic perturbations, in addition to the previously noted knockdown of H⁺-V-ATPase (Vandenberg et al., 2012). These additional examples of perturbation that result in self-correcting abnormalities include ethanol exposure, thioridazine exposure and morpholino-based knockdown of xMLTK (Fig. 1). Furthermore, in accordance with previous studies on the morphological changes that occur in craniofacial cartilaginous features (Rose, 2014; Rose et al., 2015), we found that, as tadpoles progress from early to late pre-metamorphic stages, there is a decrease in the within-group variance in morphological measurements among control and experimental tadpole groups. Therefore, it seems that there is an innate and adaptive mechanism present in all X. laevis larvae with the potential to recognize and address certain malformed or misaligned craniofacial features (Fig. 8).

We know, however, that this adaptive mechanism is not infallible because we have also shown that there are some perturbations that result in permanent craniofacial defects (Fig. 1). For example, in the case of ICI 118,551 exposure, we found that the larvae were unable to correct absent cartilaginous features. Thus, craniofacial cartilage structures cannot be formed de novo at these life stages, similar to tail regeneration in pre-metamorphic tadpoles (Beck et al., 2006; Li et al., 2016). This suggests that the self-correction of abnormal craniofacial morphology requires significant tissue remodeling of existing structures.

As thyroid hormone (TH) signaling mediates craniofacial cartilage and central nervous system remodeling during metamorphosis (Buchholz et al., 2006; Das et al., 2006; Shi et al., 1998; Kerney et al., 2012), we examined whether the tissue remodeling observed in our pre-metamorphic tadpoles could be a mild form of precocious metamorphosis. Ultimately, our TH-dependent gene expression profile analysis and pharmacological inhibition of TH production in these tadpoles revealed that the adaptive craniofacial tissue remodeling observed is not a form of precocious metamorphosis. However, we did see increased expression of rxrα in tadpoles with self-correcting craniofacial deformities. This increase in nuclear receptors is suggestive of increased nuclear receptor signaling, and nuclear receptor-dependent gene regulation is commonly associated with endocrine hormone signaling and the induction of tissue remodeling events (Escriva et al., 2006; Evans and Mangelsdorf, 2014; Laudet, 2011). Thus, alternate endocrine hormone signaling is likely to be playing a crucial role in the correction abnormal craniofacial morphology in pre-metamorphic tissues.

By pursuing an RNA-seq based approach to monitor gene expression changes in pre-metamorphic thioridazine-treated tadpoles with correctable craniofacial malformations, we identified prolactin signaling as the potential pathway used to direct the craniofacial defect self-correction response (Tables 2 and 3). Although thioridazine is a dopamine-receptor antagonist that causes increased prolactin expression, our work (K.P. and K.A.M., unpublished) and previous work (Huang and Brown, 2000) confirmed that increased prolactin signaling during development alone does not cause evident craniofacial defects; the main effect of prolactin overexpression in X. laevis embryos and tadpoles is the retention of their tail into adulthood (Huang and Brown, 2000). Interestingly, prolactin signaling has been previously shown to regulate mmp1 and mmp13 expression in Xenopus cell cultures in a TH-independent manner (Jung et al., 2004) and these Mmp genes are also upregulated in tadpoles with self-correcting craniofacial defects. Thus, it is possible that prolactin signaling is crucial to the remodeling of cartilaginous tissue within the craniofacial region of these pre-metamorphic animals. This is supported by the fact that increased dopamine signaling, which would subsequently decrease prolactin expression and signaling, disrupts the maintenance of normal craniofacial morphology and correction of abnormal craniofacial morphology in pre-metamorphic X. laevis tadpoles (Fig. 6).

Although investigating how malformed craniofacial tissues are naturally corrected in X. laevis tadpoles has led us to valuable insights regarding the remodeling of malformed cartilage and eye tissue, there is much more that needs to be deciphered about this process. For example, it would be beneficial to determine to what extent developmental malformities of muscle and brain tissue can be remodeled in X. laevis pre-metamorphic tadpoles. Additionally, it is well known that tissue remodeling often involves coordinated cell proliferation, apoptosis, migration and differentiation (Bosurgi et al., 2017; Eom and Parichy, 2017; Lorda-Diez et al., 2015; Pellettieri et al., 2010; Shi et al., 2008). Thus, it would be worthwhile to track and study the cell behaviors responsible for the alteration of abnormal tissue and organ morphology in X. laevis larvae. This includes cellular production and release of MMPs into the extracellular matrix (ECM), which may be crucial for ECM remodeling and cell movement.

Finally, we must continue to investigate the impact that various teratogens and developmental perturbations have on cellular mechanisms that are crucial for craniofacial development and morphological regulation in these animals, as there are likely to be distinct regulatory mechanisms involved in monitoring and maintaining the morphology. This will allow us to develop a more precise model for the molecular-, cellular- and tissue-level mechanisms behind the detection and correction of craniofacial defects in Xenopus laevis tadpoles. These discoveries ultimately hold great potential for inspiring and guiding novel tissue-remodeling studies, focused on better understanding the mechanisms responsible for maintaining and altering tissue morphology in vertebrate animals.

All experiments using Xenopus laevis were conducted in accordance with the guide for Care and Use of Laboratory Animals and were approved by the Institutional Animal Care and Use Committee at Tufts University. Embryo culturing was carried out as previously described (Caine and McLaughlin, 2013). Embryos and tadpoles were reared in 0.1× Marc's Modified Ringer's solution [MMR, 10 mM NaCl, 0.2 mM KCl, 0.1 mM CaCl, 0.2 mM MgCl₂, 0.5 MM HEPES, 1 µM EDTA (pH 7.5)] at 14-23°C. The 0.1× MMR rearing media was refreshed at least three times a week. Feeding-stage tadpoles were fed Sera Micron powdered growth food three times a week. Embryos and tadpoles were staged based on Nieuwkoop and Faber (1994).

xMLTK morpholino microinjections were performed at Nieuwkoop and Faber (NF) stage 3, into one of two dorsal blastomeres to primarily restrict the effect of the morpholino to half of the head. Morpholino design and purchasing followed previous work by Suzuki et al. with the xMLTK-MO (Suzuki et al., 2012). Pharmacological exposures to induce craniofacial defects were conducted on neurula stage embryos (NF stage 14-26) that had been reared at 14°C prior to exposure. During the exposure windows, experimental and controls embryos were kept at 18°C; after the exposure they were returned to 14°C until they reached feeding-tadpole stages. Pharmacological treatments included: 1.5% ethanol, 90 µM thioridazine (thioridazine-HCl, Tocris), and 200 µM ICI 118,551 (Tocris). Plate densities during exposures were kept at 50 embryos per 10 ml of media. To generate positive controls for precocious metamorphosis or prevent thyroid hormone production, we applied either a 0.5 nM thyroid hormone (Tocris) or 100 mM methimazole treatment to tadpoles from NF stage 45 to 50, respectively. Prolactin inhibition was carried out through 5 µm pergolide mesylate (Sigma) exposures continuously from stages 45 to 49.

Live specimens were temporarily immobilized for imaging with a mild tricaine exposure, then dorsal view photos for morphological analysis were taken on a Nikon SMZ1500 dissection microscope with a Spot Insight Color digital camera and Spot Image Solutions software. Landmarks were applied to the images with (StereoMorph R package). The primary landmarks used for GMA were: (1) right nostril; (2) right mouth corner; (3) medial edge of right eye; (4) lateral edge of right eye; (5) anterior edge of right eye; (6) posterior edge of left eye; (7) right otolith; (8) left otolith; (9) medial edge of left eye; (10) lateral edge of left eye; (11) anterior edge of left eye; (12) posterior edge of right eye; (13) left mouth corner; (14) left nostril; (15) anterior end of brain; and (16) midpoint between otoliths. Relative angles and distances between landmarks were calculated using a custom R script, while morphometric analyses were carried out using MorphoJ (Klingenberg, 2011).

Euthanized tadpoles were fixed in 1× MEMFA [0.1 M MOPS (pH 7.4), 2 mM EGTA, 1 mM MgSO₄, 3.7% formaldehyde], washed with 1× phosphate-buffered saline (PBS), dehydrated in methanol, bleached overnight at room temperature with a 9% hydrogen peroxide solution under a constant light source, transitioned to 100% methanol and stored at −20°C until used for whole-mount in situ hybridization or Alcian Blue staining. In situ hybridization was performed as described previously (Caine and McLaughlin, 2013), using a sox9 antisense RNA probe (Vandenberg et al., 2011). Alcian Blue staining was carried out over 2 days at 20-37°C, afterwards specimens were placed in a 20% glycerol clearing solution. We characterized changes in cartilage tissue morphology during self-correction of abnormal craniofacial phenotypes by applying Alcian Blue staining and tracking the percent of unperturbed control, thioridazine, ethanol, ICI 118,551 and xMLTK-MO tadpoles with severely malformed, moderately malformed and normal craniofacial cartilage at NF stages 45, 47 and 49; scoring followed a qualitative cartilage morphology scoring key (Fig. S3).

RNA libraries were generated using the Qiagen RNeasy Mini kit and DNAse treated prior to generating cDNA libraries for either RT-qPCR or RNA-seq. RT-qPCR cDNA libraries were generated with M-MuLV Reverse Transcriptase (NEB), following the manufacturer's protocol. RT-qPCR reactions were set up using Applied Biosystems SYBR Green Master Mix and run on Applied Biosystems 7300 Real-Time PCR System. See Table S3 for RT-qPCR primer sequences.

RNA-seq cDNA libraries were constructed using the Illumina Truseq RNA Library Preparation Kit v2, following the manufacturer's protocol. We applied Quibit analysis to quantify the RNA-seq cDNA libraries, which were then pooled and run on Illumina Hi-Seq 2500 at the Tufts University Core Facility to generate 100 bp single-end reads. Subsequent reads were trimmed to remove adapter sequences and filtered to remove any low-quality reads with Trimmomatic, then mapped to the Xenopus laevis genome v9.1 with HISAT2. Transcript count tables were generated with HTSeq-count and differential gene analysis was carried out using the DESeq2 package in R (Love et al., 2014). To perform gene ontology analysis, X. laevis gene hits were BLASTed against the Xenopus tropicalis genome and the corresponding X. tropicalis gene IDs were applied to PANTHER.

To promote suitable statistical power, a minimum total samples size (total n) of 25 tadpoles per control and experimental group was applied; total n is the sum of samples sizes (n) across multiple biological replicates (N). Categorical data across multiple control and experimental groups were statistically analyzed using χ² GOF tests, with expected values set to the average control values at a given developmental stage. Quantitative data across multiple control and experimental groups were statistically analyzed using Kruskal-Wallis non-parametric tests. If significance was detected with the Kruskal-Wallis test, a post-hoc Mann-Whitney U-test or Dunn's test was applied. Bonferroni P-value adjustments were employed to compensate for multiple testing bias. Given that the geometric morphometric data displayed non-normal distributions, standard error bars are presented in these graphs as opposed to confidence intervals.

We appreciate the numerous helpful discussions with members of the McLaughlin and Levin labs, as well as many others in the developmental biology field.