Six1 proteins with human branchio-oto-renal mutations differentially affect cranial gene expression and otic development

Branchio-otic (BOS) and branchio-oto-renal (BOR) syndromes together comprise the second most prevalent birth defect involving hearing loss (Hilgert et al., 2009; Smith, 2018). These syndromes, which are inherited in an autosomal dominant manner with 100% penetrance (Chang et al., 2004; Fraser et al., 1980), are characterized by deformities of the external, middle and inner ear structures, as well as second branchial arch defects that include fistulas and cysts; renal malformations also characterize BOR (Kochhar et al., 2007; Smith, 2018). External ear malformations include pre-auricular pits, tags and cupped/lop ears; those of the middle ear include ossicle deformities, and of the inner ear include auditory and vestibular canal hypoplasia (Smith, 2018). There is considerable variability in the severity, presence and type of these abnormalities among patients, including family members carrying the same mutation. How these varied defects arise in the embryo has not yet been addressed.

Mutations in two genes have been identified in about half of patients with BOS/BOR (reviewed in Smith, 2018; Moody et al., 2015). The gene encoding the SIX1 transcription factor is mutated in ∼4% of patients (BOS3/BOR3), and the gene encoding EYA1, which binds to SIX1 to modify its transcriptional activity, is mutated in ∼40% of the patients (BOS1/BOR1). SIX1 is highly related to Drosophila Sine oculis (SO) (Cheyette et al., 1994; Serikaku and O'Tousa, 1994; Kawakami et al., 1996). Like other proteins of the SIX family, SIX1 possesses two highly conserved domains: a SIX-type homeodomain (HD) that binds to DNA and an N-terminal SIX domain (SD) that binds to cofactor proteins including EYA1 (Pignoni et al., 1997; Kawakami et al., 2000; Kobayashi et al., 2001; Kenyon et al., 2005a). Several SIX1 single-nucleotide mutations that result mostly in amino acid substitutions have been reported in BOS/BOR patients (Fig. 1A), but only a few have been functionally analyzed and only in cell culture. The V17E mutation in the first α-helix of the SD is reported to abolish the SIX1-EYA interaction and subsequent translocation of EYA to the nucleus (Patrick et al., 2009). The R110W mutation in the sixth α-helix of the SD is reported to decrease the SIX1-EYA interaction (Kochhar et al., 2008; Ruf et al., 2004) and be deficient in transcriptional activation (Patrick et al., 2009). The W122R mutation, adjacent to the HD (Sanggaard et al., 2007), is predicted to affect DNA binding (Patrick et al., 2009), and the Y129C mutation in the HD results in significantly reduced DNA binding (Ruf et al., 2004; Patrick et al., 2009). To our knowledge, the gene expression consequences of these mutations on craniofacial development are unknown.

The craniofacial structures affected in BOS/BOR are derived from non-neural ectoderm (external ear epidermis, external auditory canal), cranial neural crest (second branchial arch, middle ear ossicles, external ear cartilage) and otic placode [inner ear, auditory-vestibular ganglion (VIIIg)] (Groves and Fekete, 2012; Moody et al., 2015). The otic placode is specified at neural plate stages, invaginates from the surface ectoderm to form a flattened otic vesicle, and then progresses through complex morphogenesis to form an inner ear comprised of both ventral auditory and dorsal vestibular subdomains (Alsina and Whitfield, 2017). In general, the morphological changes that transform the otic vesicle from an epithelial sac into an elaborate fluid-filled labyrinth are conserved across vertebrates (Bever et al., 2003; Barald and Kelley, 2004; Quick and Serrano, 2005; Groves and Fekete, 2012; Alsina and Whitfield, 2017). Studies in several vertebrates indicate that Six1 is a crucial regulator of cranial placode development (Brugmann and Moody, 2005; Grocott et al., 2012; Moody and LaMantia, 2015; Moody and Saint-Jeannet, 2014; Saint-Jeannet and Moody, 2014; Schlosser, 2006, 2010; Streit, 2007). Six1 loss-of-function studies demonstrate reduced expression of several placode genes and defects in otic development, including loss of hair cells and defective patterning of the sensory epithelium in which they reside (Laclef et al., 2003; Zheng et al., 2003; Brugmann et al., 2004; Ozaki et al., 2004; Schlosser and Ahrens, 2004; Zou et al., 2004; Bricaud and Collazo, 2006; Konishi et al., 2006; Ikeda et al., 2007; Chen et al., 2009; Christophorou et al., 2009; Ikeda et al., 2010; Bricaud and Collazo, 2011; Zhang et al., 2017; Sullivan et al., 2019). Six1 gain of function expands placode domains at the expense of the adjacent epidermal and neural crest regions (Brugmann et al., 2004; Schlosser et al., 2008; Christophorou et al., 2009; Maharana and Schlosser, 2018), and promotes utricular hair cell over VIIIg neuron formation (Bricaud and Collazo, 2006).

Given the key role of Six1 during early development of the inner ear and that known human mutations in this gene result in BOS/BOR, our goal was to determine whether these mutations have differential effects on the development of the cranial neural crest and placodes, the precursor populations that contribute extensively to the dysmorphic tissues. Using Xenopus, which evolved for terrestrial hearing and therefore extensively shares otic development, morphology and gene expression with mammals, we expressed Xenopus Six1 proteins carrying the human mutations in a wild-type (WT) background and assessed their effects on early gene expression and otic morphology. Each mutant caused specific and often different disruptions in the size of the domains of several neural border, neural crest and pre-placodal ectoderm genes. Each mutant also had differential effects on otic vesicle genes and tadpole inner ear morphology. This detailed description of early gene expression changes caused by SIX1 mutations suggests that subtle changes in their patterns in the embryonic ectoderm and otocyst contribute to the variable BOS/BOR patient phenotypes.

Previous studies in human cell lines demonstrated that V17E, R110W and Y129C mutations render SIX1 unable to activate transcription in the presence of mouse Eya1 or human EYA2 (Ruf et al., 2004; Patrick et al., 2009). We observed the same with the Xenopus mutants that we generated (Fig. 1B). Co-transfection of HEK293T cells with WT Xenopus Six1 (Six1WT) plus Eya1 caused a ∼7-fold increase (P<0.0001) in luciferase activity relative to levels after transfection of control vector, Six1WT alone or Eya1 alone. All four Xenopus Six1 mutants failed to significantly induce luciferase activity over control levels either alone (V17E, P=0.99988; R110W, P=0.99999; W122R, P=0.99999; Y129C, P=0.99999) or in the presence of Eya1 (V17E, P=0.27122; R110W, P=0.99999; W122R, P=0.99764; Y129C, P=0.99947). To exclude the possibility that the lack of transcriptional activation was caused by the mutations blocking Six1 from either localizing to the cell nucleus or translocating Eya1 to the nucleus, cells were co-transfected with FLAG-tagged versions of Six1WT or each Six1 mutant plus Myc-tagged Eya1 followed by confocal immunofluorescence microscopy (Fig. 1C-T). Consistent with the biochemical results of Patrick et al. (2009), we observed that Six1WT and all four Six1 mutants were located almost exclusively in the nucleus. Thus, the nuclear localization of Xenopus Six1 is not disrupted by these mutations, even though the nuclear localization signal is likely to be located in the SD (Brodbeck et al., 2004). This assay also showed that R110W, W122R and Y129C efficiently translocate all of the transfected Eya1 into the nucleus; in no cell was Eya1 detected in the cytoplasm. However, whereas Patrick et al. (2009) showed that V17E is unable to translocate EYA2 to the nucleus, we observed Eya1 immunofluorescence in the cytoplasm and the nucleus (Fig. 1E,Q). This discrepancy is likely because instead of the mammary cancer cells used by Patrick et al. (2009), we used HEK293T cells, which are known to express low levels of SIX1 (see Materials and Methods); thus, some Eya1 is translocated to the nucleus even in the absence of Six1 transfection. In agreement with Patrick et al. (2009), we found V17E to be the only mutant that had impaired transport of Eya1 into the nucleus. Overall, these results demonstrate that the Xenopus constructs we created to model the BOS/BOR mutations are translated, and even though they access the nucleus, they are transcriptionally deficient in the presence of Xenopus Eya1. These results match the biochemical analyses of the mouse and human mutated proteins.

BOS/BOR is an autosomal dominant syndrome in which patients express one WT allele and one mutated allele. Therefore, to provide a combination of WT:mutant protein we expressed Six1 mutant mRNAs in a WT background. However, mRNAs were injected on only one side of the embryo so that the uninjected side could serve as an internal control in each animal. We also analyzed embryos injected with two different doses of Six1 WT mRNAs (150 pg, 400 pg) to assess whether the same amount of mutant protein had similar or different effects compared to Six1WT. Because we found that V17E mRNA caused early lethality at >150 pg/blastomere, its effects were compared to 150 pg/blastomere of Six1WT mRNA (Six1WT-150). The other Six1 mutants, which did not cause lethality, were expressed at 400 pg/blastomere and compared to 400 pg/blastomere of Six1WT mRNA (Six1WT-400), a dose used in numerous previous publications to determine the early developmental function of Six1 (Brugmann et al., 2004; Schlosser et al., 2008; Sullivan et al., 2019). Expression of either mutant or WT Six1 caused variable changes in gene expression (reduced, broader or no change) in nearly every batch of embryos; we report the frequency of each phenotype for every gene assayed in the graphs in Figs 2-4. For simplicity, in the text we note which phenotype is the most frequent. Statistical comparison of the frequencies of phenotypes caused by expression of Six1WT versus Six1 mutants allows one to determine whether the mutation has changed the activity of the protein.

Neural border genes are required for the subsequent formation of the neural crest and pre-placodal ectoderm (PPE) (Groves and LaBonne, 2014; Saint-Jeannet and Moody, 2014; Grocott et al., 2012). Therefore, we first evaluated the effect of Six1 mutants on msx1, a neural border-specifying gene that controls the expression of neural crest-specifying genes (Meulemans and Bronner-Fraser, 2004; Monsoro-Burq et al., 2005). Both Six1WT-150 and V17E caused the msx1 domain to be smaller or reduced in intensity on the injected side of the majority of embryos (Fig. 2A,D); the frequencies were not significantly different (P>0.05), indicating that V17E had the same activity as Six1WT on this target. Six1WT-400 also caused a decrease in the msx1 domain (Fig. 2B,D). However, both R110W and W122R showed this decrease significantly less frequently than WT, and also caused the domain to be broader more frequently (Fig. 2C,D); here, and in descriptions below, broader domains were either similar in intensity or reduced in intensity compared to the control side of the same embryo. In contrast, Y129C showed a decrease in msx1 expression at the same frequency as Six1WT-400 (Fig. 2D). These results indicate that, in the embryo, V17E and Y129C have about the same effect on a neural border gene as Six1WT, whereas R110W and W122R decrease the msx1 domain less frequently and broaden it more frequently compared to Six1WT.

Since neural crest and PPE expression domains are influenced by the previous expression of neural border genes (Meulemans and Bronner-Fraser, 2004; Monsoro-Burq et al., 2005), we expected the Six1 alteration of the msx1 expression domain to affect both. Both Six1WT-150 and V17E could either broaden or reduce the domain of foxd3, a neural crest specifier (Meulemans and Bronner-Fraser, 2004), and the frequencies of these effects were similar (Fig. 2E,F,J). Six1WT-400 also caused both effects (Fig. 2G,J). R110W effects were not different from those of WT (Fig. 2J), whereas W122R and Y129C caused a significantly higher frequency of foxd3 domain expansion (Fig. 2H,I,J). zic2 is also expressed in the neural crest domain and is required for neural crest development (Elms et al., 2003; Warner et al., 2003; Teslaa et al., 2013). In most cases, Six1WT-150 and V17E reduced zic2 and at similar frequencies (Fig. 2M). Six1WT-400 reduced neural crest zic2 in every embryo (Fig. 2K,M), whereas R110W, W122R and Y129C each did so at a significantly lower frequency, and in many cases made the domain broader (Fig. 2L,M). In Xenopus, sox9 is expressed in both the neural crest and otic placodes (Park and Saint-Jeannet, 2010; Lee and Saint-Jeannet, 2011). In most cases, Six1WT-150 caused reduced neural crest expression of sox9, as did V17E, albeit at a significantly lower frequency (Fig. 2P). In most cases, Six1WT-400 also caused reduced neural crest expression of sox9 (Fig. 2N,P); in contrast, R110W, W122R and Y129C caused reduced sox9 at significantly lower frequencies and expanded it at significantly higher frequencies (Fig. 2O,P). Thus, compared to WT, each Six1 mutant has different effects on three genes expressed in the early neural crest: V17E was less effective than WT at reducing sox9; W122R and Y129C were less effective than WT at reducing foxd3 and tended to broaden it; and R110W, W122R and Y129C were less effective than WT at reducing zic2 and sox9 and tended to broaden them.

As previously reported, in most cases, Six1WT-150 expanded the PPE expression of sox11 (Sullivan et al., 2019) (Fig. 3A,H). In contrast, V17E reduced it significantly more frequently (Fig. 3B,H). In most cases, Six1WT-400 reduced sox11, as did R110W and Y129C, but W122R more frequently caused expansion (Fig. 3C,D,H). As previously reported (Sullivan et al., 2019), Six1WT-150 caused reduction of irx1 in the PPE, as did V17E, albeit at a significantly higher frequency (Fig. 3I). Six1WT-400 also reduced irx1 in the PPE, as did R110W and Y129C (Fig. 3E,I). In contrast, W122R caused a reduced irx1 PPE domain at a significantly lower frequency and often caused expansion (Fig. 3F,I). Six1WT-150 reduced otic placode expression of sox9, whereas V17E reduced sox9 at a significantly lower frequency and often broadened it (Fig. 3J). In most cases, Six1WT-400 also reduced sox9 otic placode expression (Figs 2N and 3J), whereas R110W, W122R and Y129C each reduced sox9 otic placode expression significantly less frequently than WT and in some cases broadened its domain (Figs 2O and 3G,J). Thus, compared to WT, each Six1 mutant protein has different effects on three placode genes: they were less effective than Six1WT (V17E broadened sox11 less frequently; W122R reduced irx1 less frequently), more effective than Six1WT (V17E reduced irx1 more frequently) or caused the opposite phenotype in a subset of embryos (V17E, R110W, W122R and Y129C for sox9).

To assess whether these changes in early gene expression impact the initial formation of the inner ear, we raised similarly injected embryos to otic vesicle stages and assessed the expression of several transcription factors required for inner ear development (Groves and Fekete, 2012; Saint-Germain et al., 2004). Six1WT-150 reduced sox9 otic expression in about half of the embryos (Figs 4A and 5A), whereas V17E caused a significantly greater frequency of the reduced phenotype (Figs 4A and 5B). Six1WT-400 also reduced sox9 otic expression (Figs 4A and 5C); Y129C caused the same at a similar frequency (Fig. 4A), whereas R110W and W122R reduced expression significantly less frequently and more frequently caused expansion of the otic vesicle (Figs 4A and 5D). irx1 was reduced by Six1WT and at similar frequencies by each mutant (Figs 4B and 5E). tbx1 was both reduced and enlarged by Six1WT-150, whereas in most cases V17E caused reduction (Fig. 4C). In most cases, Six1WT-400 reduced tbx1, as did W122R and Y129C at a similar frequency (Figs 4C and 5F); in contrast, R110W reduced tbx1 at a significantly lower frequency and expanded it more frequently (Fig. 4C). In most cases, dlx5 was reduced by Six1WT-150 (Figs 4D and 5G); the frequency of dlx5 reduction was significantly greater with V17E (Figs 4D and 5H). In most cases, dlx5 was reduced by Six1WT-400, W122R and Y129C at a similar frequency; dlx5 reduction was significantly less frequent with R110W and more embryos showed expansion (Fig. 4D). otx2 was reduced by both levels of Six1WT, and each mutant reduced it at a similar frequency to Six1WT (Figs 4E and 5I,J). Whereas pax2 expression was dramatically reduced by both levels of Six1WT, all four mutants caused reduction at a significantly lower frequency as well as a low frequency of expansion (Figs 4F and 5K).

Because the reduction in gene expression assayed in whole-mount in situ hybridization (ISH) could result from either reduced gene expression or loss of otic vesicle tissue, we sectioned a subset of the larval heads and calculated the volume of control and injected otic vesicles in the same embryo (Fig. 5K; Table S1; two-tailed paired Student's t-test). Six1WT-150 otic vesicle volumes were significantly larger than the otic vesicle volumes on the control side of the same embryo, whereas V17E otic vesicle volumes were significantly smaller than those on the control side (Fig. 5L). Six1WT-400 otic vesicle volumes were significantly smaller than those on the control side of the same embryo, whereas R110W, W122R and Y129C otic vesicle volumes were slightly larger, although this only reached significance for R110W (Fig. 5L; P<0.05). Interestingly, using the two-tailed, unpaired Student's t-test, we found that the V17E otic vesicle volume (5.78×10⁶ µm³) was significantly smaller than that of Six1WT-150 (1.92×10⁶ µm³; P<0.05), but indistinguishable from that of Six1WT-400 (1.52×10⁶ µm³; P>0.05); this comparison suggests that a low level of V17E (150 pg) has an activity comparable to that of a higher level of Six1WT (400 pg). The otic vesicle volumes of R110W, W122R and Y129C (5.24, 5.12 and 4.48×10⁶ µm³, respectively) each were significantly larger than that of Six1WT-400 (1.52×10⁶ µm³; P<0.05), but indistinguishable from that of Six1WT-150 (5.78×10⁶ µm³; P>0.05); this comparison suggests that a 400 pg dose of these mutant proteins has a weaker activity than the comparable level of Six1WT.

Together, these analyses demonstrate that all four Six1 mutants affect both otic vesicle volume and the expression of a subset of genes that are required for otic specification and patterning in ways that are not identical to Six1WT. Their effects are likely due to the mutations rather than overexpression because each is different from the comparable level of Six1WT.

To determine whether the changes reported above affect the morphology of the inner ear and surrounding, neural crest-derived otic cartilage, we performed injections as above and fixed tadpoles at stages when the auditory and vestibular regions, including hair cell sensory patches are recognizable (Bever et al., 2003; Quick and Serrano, 2005). We then used several techniques to assess aspects of ear morphology. To reveal the cartilaginous otic capsule that is derived from cranial neural crest, heads were stained with Alcian Blue. For V17E, R110W and W122R, there were subtle asymmetries between the otic capsule on the control versus injected sides (Fig. 6A and data not shown); the Y129C otic capsule showed more obvious asymmetries than the other mutants (Fig. 6B). To quantitate these changes, heads were sectioned, and the volume of the otic cartilage on each side was calculated (Table S2) and tested for whether the experimental volume was different from that on the control side (two-tailed paired Student's t-test). V17E, W122R and Y129C resulted in a significantly smaller otic cartilage volume compared to the control sides (Fig. 6C,D).

In bright-field examination of live tadpoles, in which the heads are transparent, we noticed that in some specimens the otoliths of the saccule and utricle appeared reduced. Because otolith volumes are difficult to assess in fixed specimens, we used optical coherence tomography (OCT) imaging, which uses differential light interference to detect morphology in living specimens, to test whether, like otic cartilage, otolith volumes were reduced by mutant Six1 proteins (one-tailed paired Student's t-test). The fluid-filled lumen of the living inner ear appears dark whereas the mineralized otoliths, which reflect light, appear bright (Fig. 6E). Upon calculating otolith volumes (Table S3), we found that the mean otolith volumes were significantly smaller in Six1WT-150, V17E and Y129C inner ears compared to those on the control sides of the same tadpoles (Fig. 6F). We compared the mean volumes of the mutant otoliths to those of Six1WT otoliths to determine whether the mutant protein reduced this structure (one-tailed unpaired Student's t-test). Otoliths of V17E (7.2×10⁵ µm³) were similar in volume to those of Six1WT-150 (1.6×10⁶ µm³; P>0.05). R110W (1.8×10⁶ µm³) otoliths were significantly larger compared to Six1WT-400 otoliths (6.6×10⁵ µm³; P<0.05), and Y129C otoliths (1.0×10⁵ µm³; P<0.05) were significantly smaller than Six1WT-400 otoliths; W122R otoliths (1.2×10⁶ µm³; P>0.05) were not different. Thus, two of the four mutants have effects that are significantly different from those of Six1WT.

In some BOS/BOR patient reports, the vestibular and/or cochlear ducts are reported to be hypoplastic. To determine whether the Six1 mutants also caused a reduced luminal volume, the lumen of the inner ear labyrinth was traced from OCT images of living tadpoles (Fig. 6E). Six1WT or Six1-mutant-expressing labyrinths contained all tadpole stage vestibular and auditory structures in the appropriate position/orientation, although loss of some interior cartilage structure was noted in some Y129C inner ears stained with Alcian Blue (Fig. 6B). Luminal volumes obtained from live tadpoles by OCT were calculated (Table S4) and compared (one-tailed paired Student's t-test); those for V17E and Y129C were significantly smaller compared to control sides (Fig. 7E). Using a one-tailed unpaired Student's t-test, we found that the mean luminal volumes of V17E (7.4×10⁶ µm³) were significantly smaller than those of Six1WT-150 (2.1×10⁷ µm³; P<0.05). Compared to the lumens of Six1WT-400 (6.0×10⁶ µm³), R110W lumens were significantly larger (1.8×10⁷ µm³; P<0.05), Y129C lumens were significantly smaller (6.6×10⁵ µm³; P<0.001) and W122R lumens were not different (1.4×10⁷ µm³; P>0.05). Thus, three of the four mutants have effects that are significantly different from those of Six1WT.

To obtain a higher-resolution measurement of luminal volume, fixed tadpole heads were stained with phalloidin to reveal the luminal border. Confocal z-stacks were generated through the entire depth of the inner ear on both the injected and control side (Fig. 7A), the luminal surface traced in each stack (Fig. 7B), and the luminal volume calculated, compared (Table S5; one-tailed paired Student's t-test) and displayed in 3D (Fig. 7C). Concordant with OCT imaging of live tadpoles, fixed specimens of Six1WT and Six1 mutant labyrinths contained all structural elements (Fig. 7C). Luminal volumes of Six1WT-400 were significantly increased compared to the control sides (Fig. 7F), whereas three of the Six1 mutant proteins (V17E, R110W, Y129C) caused a significant decrease in luminal volume (Fig. 7F). In these fixed specimens, we did not observe significant differences between Six1WT effects and Six1 mutant effects on mean luminal volume (one-tailed unpaired Student's t-test; P>0.05).

Each of the compartments of the inner ear contain a sensory end organ containing hair cells, which have stereocilia that stain intensely with phalloidin, making them readily recognizable in the z-stacked images (Fig. 7A). Because loss of Six1 affects the differentiation of the hair cells (Bricaud and Collazo, 2006; Ahmed et al., 2012; Zhang et al., 2017), we also measured the volumes of five of the sensory patches that are well developed in the stage 47 tadpole (anterior, posterior, horizontal, saccule, utricle; Fig. 7D). For Six1WT-150 inner ears, only the posterior canal sensory patches were significantly different from those on the control side (larger) (P<0.05; Fig. S1), and for Six1WT-400 inner ears, only the utricle sensory patches were significantly different from those on the control side (larger) (P<0.05; Fig. S1). For the Six1 mutant inner ears, only Y129C caused significantly different volumes: the posterior canal and horizontal canal sensory patches were significantly smaller than those on the control side (P<0.05; Fig. S1). V17E sensory patch volumes were not significantly different from those of Six1WT-150 (Fig. S2A). Compared to Six1WT-400 (Fig. S2B-F), only Y129C caused significantly smaller posterior canal (WT 1.8×10⁵ vs 7.4×10⁴ µm³; P<0.05) and utricular (WT 1.6×10⁵ vs 1.0×10⁵ µm³; P<0.05) sensory patches. These latter comparisons indicate that two of the four mutants have effects on some sensory patches that are different from Six1WT effects.

Together, these morphological measurements show that otic capsule and several aspects of inner ear morphology are disrupted by each BOS/BOR Six1 mutation in ways that differ from overexpression of the WT protein, and that those caused by Y129C were the most consistently significant.

It is important to decipher the consequences of mutated SIX1 on the development of the involved tissues to understand the function of the altered proteins as well as the underlying causes of the BOS/BOR phenotypes. These include the external ear (pinna deformities, preauricular pits), middle ear (ossicle defects), inner ears (vestibular and cochlear aqueducts), second branchial arch (fistulas/cysts) and kidney (Smith, 2018). Many of these phenotypes contribute to the characteristic hearing loss. Several studies have attempted to categorize patient phenotypes according to the mutation in EYA1 or SIX1, but phenotypes are so variable, even within the same family, that no significant correlations were found (Ceruti et al., 2002; Kemperman et al., 2002; Ruf et al., 2003, 2004; Ito et al., 2006; Sanggaard et al., 2007; Kochhar et al., 2008; Krug et al., 2011). One contributing factor to the inability to sort phenotypes according to genotypes is that the clinical reports often are not sufficiently detailed; they rarely use magnetic resonance imaging to describe inner and middle ear defects or ultrasound to detect renal defects. Another potential contributor to variability is that early gene expression changes may subtly alter the regulatory networks controlling the differentiation of the affected tissues. To address this, we assessed whether four different mutations found in SIX1 affect different aspects of craniofacial embryogenesis: early ectodermal gene expression (neural border, neural crest, PPE); otic vesicle gene expression; and the morphology of the otic capsule and inner ear. We chose to study whether mutations in different regions of the protein, two in the SD (V17E, R110W) and two near or in the HD (W122R, Y129C), have different effects on one or more of these early developmental processes, and whether these effects differ from those caused by expressing Six1WT.

A study of BOS/BOR mutations in EYA1 concluded that the mutated proteins act in a dominant-negative fashion rather than as hypomorphs (Li et al., 2010). Likewise, it has been proposed that mutant SIX1 proteins act in a dominant-negative manner by either competing with WT protein for DNA binding sites or competing for the EYA1 co-activator (Kochhar et al., 2007; Hoskins et al., 2007). However, since patients carry one normal SIX1 allele and one mutated allele that is transcriptionally deficient, SIX1 mutants could simply be hypomorphic. To address this, the interactions of mutant SIX1 proteins with EYA proteins and DNA were assessed by a variety of biochemical assays in various mammalian cell lines. Patrick et al. (2009) showed that V17E does not form a complex with the protein-protein interaction domain (ED) of EYA2, does not translocate EYA2 to the nucleus and does not activate the transcription of the same MEF3 reporter we used in this study (Spitz et al., 1998). This could indicate that V17E acts as a dominant negative by binding to DNA without an Eya cofactor. However, since the presence of EYA2 was shown to stabilize all mutants tested except V17E (Patrick et al., 2009), it remains possible that V17E acts as a hypomorph. In a yeast two-hybrid assay, R110W showed a reduced interaction with the ED of Eya1 (Ruf et al., 2004), but in a different assay it bound the ED of EYA2 and transported it into the nucleus (Patrick et al., 2009). Both studies showed that although R110W can bind to DNA, it does not activate MEF3 reporter transcription either alone or in combination with Eya1/EYA2. Another consideration is that mutations in the SD could result in changing the specificity of cofactor binding, as demonstrated in the fly (Kenyon et al., 2005b). The functional deficits of W122R have not been directly studied in SIX1. However, it has been proposed that the linker region between the sixth α-helix in the SD and the HD may directly contact the minor groove upon DNA binding (Patrick et al., 2009), and therefore mutations in this region could prevent or reduce DNA binding efficiency. Y129C can interact with the ED of Eya1/EYA2 and translocate EYA2 to the nucleus, but it does not bind DNA or activate MEF3 reporter transcription (Ruf et al., 2004; Patrick et al., 2009). The Y129C mutation in zebrafish six1a likewise fails to drive transcription via an ARE reporter (Bricaud and Collazo, 2011). Our MEF3 reporter luciferase results for the four mutations in Xenopus Six1 corroborate the human, mouse and fish assays, demonstrating that there is similar biochemical activity across species.

Because biochemical assays indicate that the different mutant SIX1 proteins can have different binding affinities for either Eya proteins or DNA, we asked whether they would have different effects on embryonic gene expression and whether their effects differ from those of Six1WT. To accomplish this, we made four of the human mutations in Xenopus Six1 and expressed the mutant proteins on one side of embryos to assess their effects on the expression of a large number of genes that are required for the development of precursors of the tissues affected in BOS/BOR: the neural crest, PPE and otic placode. Several studies have shown that expressing Six1WT in embryos causes changes in early ectodermal gene expression (Brugmann et al., 2004; Schlosser and Ahrens, 2004; Christophorou et al., 2009; Schlosser et al., 2008; Sullivan et al., 2019). To determine whether the mutant Six1 proteins have a different effect on gene expression from Six1WT, we compared the frequency of gene expression changes (determined by comparing to the control side of the same embryo) caused by each manipulation. We categorised the Six1 mutant protein effects as follows: (1) stronger than those of a comparable level of Six1WT, i.e. the phenotype occurred significantly more frequently; (2) weaker than those of a comparable level of Six1WT, i.e. the phenotype occurred significantly less frequently; or (3) opposite to those of Six1WT, e.g. most frequently causing a broader expression domain when Six1WT most frequently causes a smaller expression domain (Table 1).

There were a few changes that were common to all four Six1 mutants compared to comparable levels of Six1 WT: (1) a weaker effect on the neural crest and otic placode expression of sox9; (2) a weaker effect on pax2 otic vesicle expression; (3) an opposite effect on sox9 neural crest expression in a subset of embryos; and (4) no significant effects on irx1 or otx2 otic vesicle expression. However, for the most part, the four Six1 mutants had effects that were distinct from each other. For example, V17E showed the largest number of stronger effects, particularly on otic vesicle genes, and the fewest number of opposite effects. Two effects (sox9 neural crest, sox11 PPE) were categorized as opposite because a higher frequency of embryos showed the opposite effect compared to Six1WT-150. However, the frequency of V17E effects on early gene expression were not significantly different from those of Six1WT-400 (Chi-square, P>0.05), except for sox9 otic placode expression, indicating that this mutation causes the protein to act more repressive, just like high levels of the WT protein. It will be interesting to test in future experiments whether this effect is due to the lack of interaction with Eya1 and/or limited amounts of endogenous Eya1. Interestingly, overall, R110W and W122R, which are in close proximity in the protein, had many similar effects. For example, both affected msx1, the neural border gene, differently from Six1WT. Both affected otic vesicle genes in a dual fashion: they were repressed less than Six1WT (weaker) and expanded more (opposite). However, R110W and W122R do not show identical effects, perhaps because R110W is likely to affect partner interactions only, whereas W122R is predicted to affect both partner interactions and DNA binding. Finally, neither V17E nor R110W differentially affected foxd3 expression, whereas both W122R and Y129C had a weaker effect on this neural crest gene.

By assessing several neural crest, PPE/cranial placode and otic vesicle genes, we revealed that each mutant showed a combination of the effects depending upon which gene was assessed, which did not segregate in strict accordance with tissue progenitor type (Table 1). For example, V17E had a stronger effect on irx1 PPE expression but a weaker and opposite effect on sox11 PPE expression. V17E also had a stronger effect on some otic vesicle genes and a weaker effect on others. W122R and Y129C had stronger effects on foxd3 neural crest expression, yet weaker and opposite effects on zic2 and sox9 neural crest expression. One potential explanation for these different effects is tissue-specific and temporal-specific availability of cofactors that affect Six1 transcriptional activity, such as Eya1 and a large number of other potential interactors that are expressed in these tissues (Neilson et al., 2010). For example, a single-cell RNA-sequencing dataset indicates that neural crest cells do not express Eya1, but do express Eya3 and Eya4 (Briggs et al., 2018). Interestingly, Mutsuddi et al. (2005) predicted that the effects of BOS/BOR mutations in EYA1 could differ from target gene to target gene by being deficient in recruiting different sets of interacting proteins to promoters and thus leading to abnormal transcriptional output. Since cofactors other than Eya1 could affect whether Six1 functions as a transcriptional repressor versus activator (Silver et al., 2003; Brugmann et al., 2004), identifying the interaction of Six1 mutant proteins with the cofactors available in each progenitor domain is an important next step. One also needs to consider whether mutant Six1 proteins alter the developmental influence of Six4, which also is expressed in placodes and binds to similar elements on the DNA (Kawakami et al., 1996, 2000). Another consideration is that the genes we monitored are likely a mixture of direct and indirect targets of Six1; Six1 could directly repress a repressor leading to activation of an indirect target. Since several large-scale screens in multiple animals have revealed many hundreds of Six1 transcriptional targets in a variety of tissues (e.g. Ando et al., 2005; Jusiak et al., 2014; Yan et al., 2015; Riddiford and Schlosser, 2016), sorting out the details of the Six1 gene regulatory network involved in craniofacial tissue development is an important next step.

Several studies have described severe otic defects in Six1-null mice that lead to hearing loss (Laclef et al., 2003; Zheng et al., 2003; Zhang et al., 2017). While the otic vesicle forms at the appropriate developmental time, it does not undergo proper morphogenesis. In contrast, Six1-heterozygous inner ears, which are similar to BOS/BOR in having one normal copy of SIX1, appear to develop normally at the histological level (Laclef et al., 2003), although one mouse line was characterized by a low frequency of truncated inner ear morphology and conductive hearing loss due to middle ear defects (Zheng et al., 2003). Thus, Six1 haploinsufficiency is mostly tolerated in inner ear development. Interestingly, an N-ethyl-N-nitrosourea (ENU)-induced mutation in mouse (E135G) that changes the second residue in the N-terminal region of the HD (residue E125 in human; Fig. 1A), showed that the presence of one copy of Six1WT influences the effects of the mutation (Bosman et al., 2009). Whereas the homozygous mutants are missing semicircular canals, and have fewer hair cells, VIIIg defects and malformed middle ear ossicles, the heterozygotes have only very subtle inner ear defects. This indicates that the availability of endogenous Six1WT in the presence of mutant protein abrogates the severe homozygous mutant inner ear defects. Similarly, we found that Six1 mutant protein in the presence of endogenous Six1WT disrupted gene expression in the otic placode and otic vesicle, but these changes did not severely perturb the morphology of the otic capsule or structures in the inner ear. The otic capsule, otoconia and hair cells differentiate and the structural elements were all recognizable. However, there were demonstrable effects: the volumes of the otic cartilage, otoliths, inner ear lumen and sensory end organs are variably reduced by each mutant. Interestingly, the changes in otic vesicle volume did not predict the changes in otic cartilage volume. For example, although Y129C caused enlarged otic vesicles, otic cartilage volume was significantly reduced. Similarly, although V17E caused a large reduction in otic vesicle size, the otic cartilage volume was only modestly reduced. These apparent discrepancies may be due to the otic cartilage being derived from cranial neural crest rather than the otic vesicle, and the mutations having differential effects on these two precursor populations. Another puzzling difference between early and later effects is that Y129C showed the fewest disruptions in otic placode and otic vesicle gene expression, yet it showed the greatest disruption of tadpole ear morphology, and V17E showed the most effects at otic vesicle stages but the mildest effects in the tadpole ear (Table 1). While the latter might be due to only mildly affected V17E embryos surviving to tadpole stages, alternatively, the two types of disruptions – loss of co-factor binding (V17E) versus loss of DNA binding (Y129C) – may lead to very different transcriptional roles during PPE/placode formation versus tadpole ear morphogenesis.

The expression of each of the mutant Six1 proteins causes early and distinctly different gene expression disruptions that affect the progenitors of the neural crest and otic cells that ultimately give rise to the inner, middle and external ears and branchial tissues that are affected in BOS/BOR. The effects we report are highly variable within each experimental group, perhaps because we sampled a genetically diverse population derived from different outbred parents. However, similar variability in phenotypes is a hallmark of BOS/BOR (Smith, 2018). An important next step will be to discover how these varied defects that arise during embryonic stages ultimately lead to the variable patient phenotypes.

Many of the methods were supported by Xenbase (http://www.xenbase.org/, RRID: SCR_003280) and the National Xenopus Resource (http://mbl.edu/xenopus/, RRID: SCR_013731).

Four different mutations associated with BOS/BOR (Fig. 1A) were introduced into the Xenopus laevis pCS2-Six1 plasmid (Brugmann et al., 2004) with the QuikChange Lightning Site-Directed Mutagenesis kit (Agilent). Each mutant was constructed by a single-nucleotide base change: Six1-V17E (GTG to GAG), Six1-R110W (AGG to TGG), Six1-W122R (TGG to AGG) and Six1-Y129C (TAC to TGC). All constructs were fully sequenced in both directions.

mRNAs encoding Six1WT (Brugmann et al., 2004), each of the Six1 mutants and a nuclear-localized β-galactosidase (nβgal) lineage tracer were synthesized in vitro according to the manufacturer's protocols (mMessage mMachine kit, Ambion). Antisense RNA probes for ISH were synthesized in vitro (MEGAscript kit; Ambion) as previously described (Yan et al., 2009).

Fertilized embryos were obtained by natural mating of outbred, WT adult Xenopus laevis males and females as described (Moody, 2000, 2018). The breeding colony is comprised of adult, outbred Xenopus laevis of unknown age obtained from Nasco (Fort Atkinson, WI, USA) and maintained in compliance with The George Washington University Institutional Animal Care and Use Committee protocol A233. Embryos were chosen at the two-cell stage to accurately identify the dorsal and ventral animal blastomeres (Klein, 1987; Miyata et al., 1987; Moody, 2018). When selected embryos reached eight cells, the dorsal-animal and ventral-animal blastomeres that predominantly give rise to the neural crest and cranial placodes (Moody and Kline, 1990) were microinjected with 1 nl mRNA according to standard methods (Moody, 2000, 2018). Embryos were cultured in diluted Steinberg's solution until fixation.

HEK293T cells (American Type Culture Collection, CRL-3216), cultured with Dulbecco's modified Eagle medium (HyClone)+10% fetal bovine serum (Gibco)+penicillin-streptomycin (Gibco), were seeded onto 24-well plates at 200,000 cells/well; 24 h later, cells were transfected using X-tremeGENE 9 DNA (Sigma-Aldrich) transfection reagent with 200 ng/well pGL3-6xMEF3-Firefly luciferase (reporter plasmid, Spitz et al., 1998), 20 ng/well TK-Renilla luciferase, 400 ng/well of either pCS2, pCS2-Six1WT, pCS2-5′Myc-Eya1, pCS2-Six1WT plus pCS2-5′Myc-Eya1, pCS2-Six1-V17E, pCS2-Six1-V17E plus pCS2-5′Myc-Eya1, pCS2-Six1-R110W, pCS2-Six1-R110W plus pCS2-5′Myc-Eya1, pCS2-Six1-W122R, pCS2-Six1-W122R plus pCS2-5′Myc-Eya1, pCS2-Six1-Y129C or pCS2-Six1-Y129C plus pCS2-5′Myc-Eya1. The pGL3-6xMEF3 luciferase reporter was previously shown to bind Xenopus Six1 (Ford et al., 2000). Forty-eight hours after transfection, cells were lysed with passive lysis buffer (Promega), and the resulting extracts analyzed for Firefly and Renilla luciferase activities using the Dual-luciferase reporter assay system (Promega). Experiments were repeated five times; after testing the data for normality (Kolmogorov–Smirnov test), a one-way ANOVA with Tukey post hoc multiple comparisons test was performed using GraphPad Prism 8 software. Expression of exogenous proteins from the transfected plasmids was confirmed by standard quantitative western blotting using primary antibodies against Six1 (D5S2S, #16960; 1:1000), Myc-tag (9B11, #2276, 1:1000) and β-actin (13E5, #4970. 1:1000) (Cell Signaling Technology, Danvers, MA, USA) and secondary antibodies (IRDye 680RD donkey anti-rabbit IgG #92568073, 1:5000; IRDye 800CW donkey anti-rabbit IgG #92532213, 1:5000; IRDye 800CW goat anti-mouse #92532219, 1:5000; LI-COR Biosciences, Lincoln, NE, USA) (A.L.P.T., unpublished).

HEK293T cells, obtained and cultured as above, were seeded onto Nunc Lab-Tek II chamber slides (Nalge) at 400,000 cells/chamber; 24 h after plating, cells were transfected with a total of 2000 ng/chamber of pCS2-5′Myc-Eya1 plus one of the following: pCS2-Six1-3′FLAG, pCS2-Six1-V17E-3′FLAG, pCS2-Six1-R110W-3′FLAG, pCS2-Six1-W112R-3′FLAG or pCS2-Six1-Y129C-3′FLAG. Forty-eight hours after transfection, cells were fixed in 4% paraformaldehyde and processed for immunostaining by standard methods using mouse anti-FLAG (9A3, #8146, 1:1600) and rabbit anti-Myc (71D10, #2278, 1:200) monoclonal antibodies (Cell Signaling Technology, Danvers, MA, USA) followed by Alexa Fluor 488-conjugated anti-rabbit (#4412, 1:1000) and Alexa Fluor 568-conjugated anti-mouse (#A1104, 1:1000) secondary antibodies, and NucBlue nuclear counterstain (#R37605) (Thermo Fisher Scientific, Waltham MA, USA). Experiments were repeated three times, and at least five fields per slide analyzed using a Zeiss LSM 710 confocal microscope. It should be noted that low levels of endogenous SIX1 mRNA and protein can be detected in the HEK293T cell line (https://www.proteinatlas.org/ENSG00000126778-SIX1/cell#rna and confirmed by A.L.P.T., unpublished).

Embryos were cultured to neural plate border (stage 13), neural plate (stages 16-18) and otocyst (stages 28-32) stages (Nieuwkoop and Faber, 1994), fixed in 4% paraformaldehyde, stained for β-Gal histochemistry, and processed for ISH as described previously (Yan et al., 2009). In embryos in which the nβgal lineage tracer was located in the appropriate tissue domains, the position, intensity and size of the expression domains of dlx5, foxd3, irx1, msx1, otx2, pax2, sox9, sox11, tbx1 and zic2 were compared between the injected, lineage-labeled side and the control, uninjected side of the same embryo, thus controlling for inter-embryo variation. Embryos for each assay were derived from a minimum of three different sets of outbred parents. Embryos were scored for gene expression changes independently by at least two of the authors, and the values reported are means of their independent scores. Gene expression changes were scored in three categories: (1) decreased expression, which could be either a smaller domain or same sized domain with reduced intensity compared to the control side of the same embryo; (2) broader expression domain, which was of the same intensity or reduced intensity compared to the control side of the same embryo; or (3) no change compared to the control side of the same embryo. The Chi-square test was used to determine if the frequency of gene expression changes were significantly different between Six1-mutant-injected embryos and Six1WT-injected embryos (P<0.05).

A subset of larvae (stages 30-34) that were processed by ISH expression were embedded in a gelatin-based medium [0.5% gelatin, 30% bovine serum albumin, 20% sucrose, hardened with glutaraldehyde (75 µl/ml)], and vibratome sectioned at 40 µm in the transverse plane. Serial section images through the otic vesicle of a minimum of six embryos per mRNA injection were collected using a Leica Tiling microscope. The area of both control and injected otic vesicles from every section per embryo was measured using Zen Blue software (Zeiss Zen 2.0) to calculate the volume. Because the larvae were of different sizes, the volume of each mutant otic vesicle was expressed as a percentage of the control vesicle of the same larva and plotted as percent change from control: [(experimental−control)/control]×100. A two-tailed paired Student's t-test was used to determine if the volume of the Six1-mutant-injected otic vesicle was significantly different from that on the control side (P<0.05). A two-tailed unpaired Student's t-test was used to determine if the volumes of Six1-mutant-injected otic vesicles were significantly different from those of Six1WT-injected otic vesicles (P<0.05).

To assess the development of the neural-crest-derived cartilage that forms a capsule surrounding the inner ear, embryos were microinjected as above, grown to tadpole stages, fixed in 4% paraformaldehyde and stained in 0.1% Alcian Blue in acidic alcohol (70% EtOH, 0.37% HCl) according to Young et al. (2017). They were vibratome sectioned at 40 µm as above. Serial section images were collected, and the area of otic cartilage on each side of the head (injected and control) traced in every section and converted to volume using Olympus cellSens Standard software (version 1.17). Because the tadpoles were of different sizes, the otic cartilage volume of each mutant ear was plotted as percentage change, as above. A two-tailed paired Student's t-test was used to determine if the volume of the Six1-mutant-injected otic capsule was significantly different from that on the control side of the same embryo (P<0.05).

Tadpoles (stages 46-47) were anesthetized in 0.01% benzocaine solution, aligned in an agar-coated dish with dorsal sides facing the OCT imaging field and live imaged with a Thorlabs Telesto series OCT Imaging System with a 1300 nm spectral domain. Images collected from living tadpoles were processed with the ThorImageOCT software (version 5.1.1). Sectional image stacks, collected every 2 µm, were exported as tiff images into Free-D 3D image reconstruction and modeling software (version 1.15; Andrey and Maurin, 2005). To measure volumes in the living tadpole, which avoids fixation artefacts, the outer boundary of each saccular and utricular otoliths and the luminal surface were traced in every other section of the injected and control inner ears of each embryo. Because the tadpoles were of different sizes, the otolith and luminal volumes of each mutant ear were expressed as percentage changes from controls, as above. A one-tailed paired Student's t-test was used to determine if the otolith and luminal volumes on the Six1-mutant-injected side were significantly smaller than those on the control side (P<0.05). A one-tailed unpaired Student's t-test was used to determine if Six1 mutant otolith and luminal volumes were significantly smaller than those of Six1WT specimens (P<0.05).

To measure the volumes of the inner ear lumens and sensory hair cell end organs, embryos were microinjected as above, grown to the same tadpole stages, fixed in 4% paraformaldehyde/PBS and stained with Alexa Fluor 568 Phalloidin (Thermo Fisher Scientific, 1:250 dilution) to detect F-actin. Z-stacks through the entire inner ear were acquired every 1 µm with a Zeiss LSM710 confocal microscope, and the stacked confocal images imported into IMARIS software (Bitplane version 9.2) to generate a 3D reconstruction of the inner ear. The luminal surface and the boundary of each sensory hair cell patch was traced in each optical section of the injected and control inner ears of each embryo, setting the same threshold parameters for both ears. After segmenting the image data, IMARIS calculated the surface area and volumes for each sample. Because the tadpoles were of different sizes, the luminal volume of each mutant ear was expressed as percentage change from control, as above. A one-tailed paired Student's t-test was used to determine if sensory patch and luminal volumes on the Six1-mutant-injected side were significantly different from those on the control side (P<0.05). A one-tailed unpaired Student's t-test was used to determine if they were significantly different between Six1WT-injected and Six1-mutant-injected specimens (P<0.05).

This work was made possible with the support of Xenbase (http://www.xenbase.org/; RRID: SCR_003280) and the National Xenopus Resource (http://mbl.edu/xenopus/; RRID: SCR_013731).