Sobp modulates the transcriptional activation of Six1 target genes and is required during craniofacial development

Hearing is an important component of human communication and hearing loss disorders negatively affect quality of life. Around 32 million children worldwide have disabling hearing loss, 40% of which have a genetic cause (Krug, 2016). Although hundreds of genes are associated with syndromic and non-syndromic congenital hearing loss, many cases still have an unknown cause (Shearer et al., 2017). Branchio-otic and branchio-oto-renal syndromes (BOR) are autosomal dominant disorders in which affected individuals present with variable degrees of hearing loss due to defects in inner, middle and outer ears, as well as branchial arch-associated dysmorphologies, including branchial fistulas and cysts (Moody et al., 2015; Smith, 2018). Variants in EYA1 or SIX1 have been identified in ∼50% of affected individuals, but the underlying genetic causes of remaining cases are unknown (Ruf et al., 2003, 2004; Sanggaard et al., 2007; Lee et al., 2007; Klingbeil et al., 2017). Thus, identifying genes that interact with SIX1 may uncover novel genes linked to BOR or other deafness syndromes.

The Six1 transcription factor, which is homologous to Drosophila Sine oculis (So) (Cheyette et al., 1994), plays a role in many cellular processes and in the embryo regulates the development of craniofacial tissues affected in BOR (Kawakami et al., 2000; Kumar, 2009; Xu, 2013; Moody and LaMantia, 2015). Loss of Six1 in mouse and Xenopus causes several craniofacial defects, including disruption of inner, middle and outer ear development (Zheng et al., 2003; Li et al., 2003; Laclef et al., 2003; Ozaki et al., 2004; Brugmann et al., 2004; Schlosser et al., 2008; Guo et al., 2011; Tavares et al., 2017; Sullivan et al., 2019). A hallmark of Six1 transcriptional activity is that it can be modulated by co-factors that influence it to either activate or repress target genes (Ohto et al., 1999; Li et al., 2003; Silver et al., 2003; Brugmann et al., 2004). Transcriptional activation is mediated by the Eya family of co-activators. Upon binding to Six1, Eya factors are translocated into the nucleus, where Eya phosphatase activity triggers the recruitment of co-activators that switch Six1 function from repression to activation (Ohto et al., 1999; Li et al., 2003).

A screen to identify proteins that interact with Drosophila So identified So-binding protein (Sobp), which is co-expressed with so in the anterior region of the eye disc (Kenyon et al., 2005a). In mouse, Sobp was identified as the spontaneous recessive variant causing deafness and vestibular-mediated circling behavior in the Jackson circler (jc) mouse (Calderon et al., 2006; Chen et al., 2008). In humans, SOBP was linked to mental retardation, anterior maxillary protrusion, strabismus and mild hearing loss (MRAMS; OMIM #613671) in seven individuals of the same family. The homozygous variant inserts an early stop codon at arginine 661 (p.R661X), causing a 212 amino acid truncation in the C terminus (Basel-Vanagaite et al., 2007; Birk et al., 2010).

As Sobp variants cause otic phenotypes in mouse and humans, and are expressed in the pre-placodal ectoderm (PPE) and otic vesicle in Xenopus embryos (Neilson et al., 2010), we speculated that Sobp might function by associating with Six1 and thereby contributing to BOR (Moody et al., 2015). Here, we address the developmental function of Sobp. We find that, although sobp is expressed in the same tissues as six1, higher levels of Sobp are detected where Six1 expression is lower, suggesting that it may repress Six1 activity. We show that Sobp binds to Six1, competes with Eya1 binding to the complex and significantly interferes with the transcriptional activation of Six1+Eya1 target genes. Structural and functional analyses show that Sobp constructs either lacking a conserved C-terminal nuclear localization signal (NLS) or containing a variant similar to human p.R661X (Xenopus Sobp p.R651X) can still interact with Six1, indicating that the domain that interacts with Six1 is not in the C terminus. In vivo studies establish Sobp as a crucial factor for PPE and neural crest (NC) gene expression, and for later otic vesicle development. These findings also show that, while the p.R651X variant disrupts gene expression in the PPE in the same manner as full-length Sobp, changes in other domains and otic vesicle patterning are distinct. Together, these results demonstrate that Sobp plays a crucial role in Six1 transcriptional function during several aspects of vertebrate craniofacial development, and suggest it may be a candidate gene for BOR and other deafness syndromes.

We reported previously that sobp is expressed in several of the embryonic tissues that also express six1, including the PPE and otic vesicle (Neilson et al., 2010). A closer comparison at the neural plate stage showed that sobp expression was stronger in the anterior PPE and weaker in the posterior PPE, whereas six1 expression was the reverse (Fig. 1A,B). At larval stages, both six1 and sobp were expressed in the medial wall of the otic vesicle, but sobp expression appeared more intense dorsally compared with its ventral domain (Fig. 1C,D; Fig. S1A,B). Co-transfection assays in HEK293T cells showed that Sobp protein was mainly nuclear (Fig. 1E-G), as previously described in mouse (Chen et al., 2008), where it colocalized with Six1 in most cells expressing both constructs (Fig. 1H-K). These findings are the first direct evidence in vertebrates of Sobp and Six1 co-expression in tissues and cellular compartments that would allow their interaction.

Drosophila Sobp has been shown to bind to So by yeast two hybrid and GST-pulldown assays (Kenyon et al., 2005a). To test whether this interaction occurred with vertebrate proteins, HEK293T cells were co-transfected with constructs driving expression of Six1-Flag and/or HA-Sobp followed by immunoprecipitation (IP). Western blot analysis detected Six1 after HA-Sobp IP (Fig. 1L); reverse IP confirmed this finding (data not shown).

To assess whether Sobp modulates Six1 transcriptional activity, HEK293T cells were co-transfected with a Six1-inducible reporter (Ford et al., 2000) and with different combinations of Six1, Eya1 and/or Sobp (Fig. 1M; Fig. S2). As demonstrated previously (Patrick et al., 2009), in the presence of Eya1, Six1 induced a significant increase in luciferase activity over control (∼5 fold increase, P<0.0001). In contrast, Sobp by itself (P=0.8899) or in the presence of Six1 (P=0.9184) did not cause a significant change in luciferase activity. However, co-transfection of Sobp with both Six1 and Eya1 significantly repressed luciferase activity compared with Six1+Eya1 levels (P<0.0001), returning luciferase activity to levels statistically indistinguishable from control (P=0.2226). These data show that Sobp can be classified as a bona fide Six1 co-factor because it binds to Six1 and is able to significantly interfere with the transcriptional activation of the Six1+Eya1 reporter.

Pa2g4 is a recently identified Six1 co-factor that reduces the transcriptional activation of Six1+Eya1 target genes by competing with the ability of Eya1 to bind to Six1 (Neilson et al., 2017). To determine whether Sobp also competes for binding to Six1, HEK293T cells were co-transfected with Six1-Flag, Myc-Eya1 and increasing amounts of HA-Sobp. Co-IP experiments (Fig. 2A) showed that equimolar amounts (1.0×) of Six1, Eya1 and Sobp did not disrupt Six1 binding to Eya1, whereas the amount of Six1 bound to Eya1 was diminished at a five- or 10-fold increase in Sobp. Interestingly, low (0.5×) and equimolar (1.0×) levels of Sobp resulted in higher levels of Six1 bound to Eya1 relative to when Six1 and Eya1 were co-transfected without Sobp. Because Eya factors are localized in the cytosol (Fig. 2C-F) and become nuclear when bound to Six factors (Fig. 2G-J) (Ohto et al., 1999; Li et al., 2003), we tested whether the competition for binding to Six1 would disturb the ability of Six1 to translocate Eya1 to the nucleus (Fig. 2O-C′). Cells co-transfected with combinations of Six1-Flag, Myc-Eya1 and increasing concentration of HA-Sobp showed that Eya1 was translocated to the nucleus by Six1 (Fig. 2O,S,X,Z) at equimolar amounts of Six1, Eya1 and Sobp, with most transfected cells also expressing Sobp in the nucleus (Fig. 2P,T,W,AA). Remarkably, Sobp was able to bind to (Fig. 2B) and partially translocate Eya1 to the nucleus (Fig. 2K-N) without Six1 co-expression. However, and in agreement with the competition co-IP results, a fivefold increase in Sobp led to detection of Eya1 in the cytosol (Fig. 2Q-R,BB,CC) even though both Six1 (Fig. 2U,BB) and Sobp (Fig. 2V,CC) were nuclear. Together, these data indicate that Sobp interference with the transcriptional activation of Six1+Eya1 target genes is achieved through a dose-dependent competition mechanism.

Because Sobp is located in the nucleus in the absence of Six1 co-transfection (Fig. 1E-G) (Chen et al., 2008), it must contain a NLS. A bioinformatic analysis (Kosugi et al., 2009) identified in Xenopus Sobp N-terminal and C-terminal putative NLSs, similar to that detected in the mouse (Chen et al., 2008) and human (Birk et al., 2010) sequences. However, only the putative C-terminal NLS had a cutoff score above 8, which indicates protein localization exclusively in the nucleus (Fig. 3A-B) (Kosugi et al., 2009). Deletion of this C-terminal domain caused HA-Sobp-NLSdel to be detected mainly in the cytosol (Fig. 3C-E). Interestingly, when cells additionally expressed Flag-tagged Six1, Sobp-NLSdel was translocated to the nucleus (Fig. 3F-I). Removal of the NLS did not disrupt binding of Sobp to Six1 (Fig. 3J) or diminish the ability of Sobp to reduce the transcriptional activation of the Six1+Eya1 reporter (P<0.0001; Fig. 3K). These findings identify a single C-terminal NLS in Sobp and demonstrate that, in its absence, Sobp can bind to Six1 in the cytosol and be translocated to the nucleus via this interaction.

We introduced the human p.R661X variant found in a MRAMS family (Birk et al., 2010) in Xenopus Sobp by replacing the arginine located at amino acid 651 with a stop codon (c.1951 A>T; p.R651X) generating a truncated Sobp lacking part of the C-terminus, including the NLS (Fig. 3A). As expected, HA-p.R651X is located primarily in the cytosol (Fig. 3L-N) as the NLS was missing. Cytosolic HA-p.R651X was translocated to the nucleus in most cells also expressing Six1-Flag (Fig. 3O-R). Despite the large deletion, p.R651X still bound to Six1 (Fig. 3S) and significantly reduced the transcriptional activation of the Six1+Eya1 reporter (P<0.0001; Fig. 3T). These data demonstrate that, although the p.R651X variant of sobp lacks a large region of the C terminus, the domain for Six1 interaction remains intact.

To determine whether Sobp has a role in NC and PPE development, we performed two types of knockdown (KD) experiments: a translation-blocking antisense morpholino oligonucleotide (MO) against sobp; and an F0 analysis after CRISPR/Cas9-mediated genome editing (Fig. 4; Fig. S3). First, we assessed by qPCR whether KD in the whole embryo caused changes in the expression of genes characteristic of neural plate (sox2), NC (foxd3), PPE (six1) and epidermis (krt12.4) at neural plate stages (Fig. 4A). KD using either method showed decreased mRNA levels for foxd3 (NC) (MO, P<0.001/q=0.000602; CRISPR, P<0.01/q=0.002520) and six1 (PPE) (MO, P<0.0001/q=0.000007; CRISPR, P<0.001/q=0.000274), but not for sox2 (MO, P=0.033437/q=0.033771; CRISPR, P=0.071790/q=0.036254) or krt12.4 (MO, P=0.114167/q=0.086482; CRISPR, P=0.104590/q=0.042254). We also monitored levels of sobp mRNA (Fig. 4A) and confirmed that, as expected, it was decreased by CRISPR/Cas9 editing likely followed by nonsense-mediated mRNA decay (P<0.0001/q=0.000162), but not by a translation-blocking MO (P=0.359946/q=0.218127). To assess whether these whole-embryo gene expression changes were also observed as changes in the sizes of specific expression domains, we monitored similar KD on one side of embryos by in situ hybridization (ISH). Sobp KD by MO or by CRISPR each caused a broader sox2 domain that was fainter in intensity (MO, 77.0%, Fig. 4D,T; CRISPR, 64.9%, Fig. 4H,U) and reduction of the foxd3 (MO, 90.2%, Fig. 4E,T; CRISPR, 66.6%, Fig. 4I,U), six1 (MO, 73.9%, Fig. 4F,T; CRISPR, 76%, Fig. 4J,U) and krt12.4 (MO, 95.8%, Fig. 4G,T; CRISPR, 78.6%, Fig. 4K,U) domains on the injected side of the embryos. The results from the two different assays are summarized in Table S1. The qPCR and ISH approaches both showed reductions in foxd3 and six1 expression, but the ISH assay was more sensitive to changes in the spatial domains of sox2 and krt12.4. Whereas qPCR did not detect a change in sox2 levels in the whole embryo, the sox2 domain was broader and fainter, which would likely lead to no significant overall change as assessed by qPCR. Whereas qPCR did not detect a change in krt12.4 levels in the whole embryo, the krt12.4 domain was obviously reduced along the anterior neural plate (Fig. 4G,K); this may be too subtle a change to be detected by qPCR of a whole embryo covered by epidermis.

To determine whether increasing Sobp protein above endogenous levels altered gene expression, we increased Sobp expression by mRNA injections. First, we assessed by qPCR whether increasing Sobp throughout the embryo caused overall changes in gene expression (Fig. 4B). Increasing Sobp significantly reduced whole-embryo mRNA levels of foxd3 (P<0.01/q=0.002780) and six1 (P<0.01/q=0.002780), and increased levels of krt12.4 (P<0.05/q=0.009927); sox2 levels were not significantly altered (P=0.397516/q=0.200746). To assess whether these whole embryo gene expression changes also were observed as changes in the sizes of specific expression domains, we targeted sobp mRNA to the dorsal-animal and ventral-animal blastomeres of eight-cell embryos that are the major precursors of NC and PPE (Moody and Kline, 1990) and assessed changes in expression domains by ISH. Similar to the qPCR data, the majority of embryos showed no changes in the sox2 domain on the injected side (67.5%, Fig. 4L,V). However, we also detected a decrease in the sox2 domain in some embryos (25.0%). The effects on the foxd3 domain were pleiotropic: it was smaller in 39.0% (Fig. 4M,V) and broader in 39.0% (Fig. 4V) of embryos. Similar to the qPCR data, the six1 domain was smaller in the majority of the embryos (86.9%, Fig. 4N,V). Even though the krt12.4 domain was increased in some embryos with ectopic expression of this gene overlapping the lineage tracer (33.3%, Fig. 4O,V), consistent with the qPCR results, for most embryos the domain of this gene was not detectably changed (54.2%, Fig. 4V). Overall, these data, which are summarized in Table S1, demonstrate that increasing the levels of Sobp disrupts the relative sizes of the gene expression domains of foxd3 and six1 and can induce ectopic expression of krt12.4.

Because the p.R651X variant of sobp does not disrupt the interaction of this protein with Six1, we assessed whether p.R651X would cause the same changes in gene expression as full-length Sobp. First, qPCR analysis of whole embryos (Fig. 4C) showed a significant decrease in six1 (P<0.01/q=0.002134) and significant increase in krt12.4 (P<0.01/q=0.004706), but no significant changes in foxd3 (P=0.211730/q=0.142565) or sox2 (P=0.307810/q=0.155444) mRNA levels. The only difference in comparison with full-length Sobp was a lack of effect on foxd3 levels (Table S1). By ISH, we found that the sox2 domain was not affected in the majority (71.4%, Fig. 4P,X) of embryos, similar to the qPCR results and similar to, but slightly more frequently than, the full-length Sobp ISH data (P<0.05). The foxd3 domain was reduced in about half (47.4%, Fig. 4Q,X) and increased in a small number (7.9%, Fig. 4X) of embryos, even though qPCR did not detect a change. In comparison with full-length Sobp, the foxd3 domain showed a significantly different distribution in phenotypes (P<0.01). Consistent with the qPCR results and indistinguishable from full-length Sobp (P=0.3772), the six1 domain was decreased at a high frequency (95.0%, Fig. 4R,X). Although whole embryo levels of krt12.4 were increased, its domain was not detectably changed in most embryos (77.3%, Fig. 4X). However, a low percentage of embryos showed ectopic expression of krt12.4 in the area where the lineage tracer was present (19.7%, Fig. 4S,X), albeit at a lower frequency than full-length sobp (P<0.05). These data demonstrate that introduction of the p.R651X variant in Sobp caused similar changes in six1 expression compared with full-length Sobp, but its effects on sox2, foxd3 and krt12.4 expression were subtly different. Importantly, in spite of its inability to translocate to the nucleus in the absence of Six1 (Fig. 3L,N), p.R651X is sufficiently functional to alter NC and PPE gene expression in whole embryos.

As both KD and increased expression of Sobp altered PPE gene expression, we analyzed whether these changes led to disruptions in the otic vesicle at larval stages (Fig. 5, Table S2). KD of sobp by CRISPR (Fig. 5A,D,G,J) revealed a decrease in the intensity and domain size of the otic expression of dlx5 (42.9%, Fig. 5D,J) and pax2 (54.0%, Fig. 5G,J); interestingly, six1 was not altered in most sobp crispants (68.6%, Fig. 5A,J) as only 25.7% of embryos showed a slight decrease in expression (Fig. 5J). Similar findings were shown by qPCR (Fig. 5M): decreased dlx5 (P<0.01/q=0.002700) and pax2 (P<0.01/q=0.002668), and no significant change in six1 (P=0.450307/q=0.227405).

Increased Sobp had variable effects on otic vesicle genes (Fig. 5B,E,H,K; Table S2). six1 expression was decreased in 50.0%, increased in 6.7% and unchanged in 43.3% (Fig. 5B,K) of larvae. dlx5 showed increased expression with a variable domain size in 46.4% (Fig. 5E,K) and decreased expression and domain size in 39.3% (Fig. 5K) of larvae. pax2 showed increased expression with a variable domain size in 59.5% (Fig. 5H,K) and decreased expression in 16.2% (Fig. 5K) of larvae. qPCR of whole heads (Fig. 5N) confirmed the effects were variable because there were no significant changes in the mRNA levels for six1 (P=0.106946/q=0.079244) or dlx5 (P=0.117690/q=0.079244). Only pax2 showed a significant increase in mRNA levels (P<0.01/q=0.005414) after increased Sobp expression. These findings show that altering Sobp levels that change PPE gene expression are followed by disruptions of otic vesicle development.

We tested whether the p.R651X variant of sobp also altered otic gene expression (Fig. 5C,F,I,L; Table S2). Effects on otic gene expression were variable with most of the analyzed larvae showing no changes in six1 (51.3%), dlx5 (42.1%) or pax2 (56.7%; Fig. 5L). Less frequently, we observed a decrease in the expression and domain size for six1 (35.9%, Fig. 5C,L) and dlx5 (36.8%, Fig. 5F,L), whereas pax2 showed increased expression (30.0%, Fig. 5I,L). Comparing the frequencies of gene expression changes between embryos injected with wild-type sobp versus p.R651X mRNAs showed effects on six1 were similar (P=0.742), whereas dlx5 and pax2 were affected by p.R651X significantly less frequently (dlx5, P<0.0001; pax2, P<0.05). qPCR of whole heads (Fig. 5O) confirmed the effects were variable because there were no significant changes in six1 (P=0.671981/q=0.678701), dlx5 (P=0.621603/q=0.678701) or pax2 (P=0.435624/q=0.678701) mRNA levels. These data demonstrate that the p.R651X variant of sobp mildly disrupts six1 otic vesicle expression similar to full-length Sobp, but has diminished effects on dlx5 and pax2.

Because altered Sobp levels affected NC gene expression (Fig. 4), we examined whether this leads to defects in cranial cartilages at tadpole stages (Fig. 6). Gross analysis of surviving tadpole heads revealed that sobp KD by CRISPR (Fig. 6A) caused mild to severe cranial cartilage hypoplasia on the injected side in a subset (5 nl, 18.6%; 10 nl, 46.7%) of F0 tadpoles, whereas injection of sobp mRNA (Fig. 6B) caused only mild hypoplasia. Tadpoles injected with the p.R651X variant of sobp (Fig. 6C) did not have apparent cranial cartilage defects. Staining tadpoles with Alcian Blue demonstrated that the observed hypoplasia in Sobp F0 crispants was the consequence of deformed Meckel's and ceratohyal cartilages, hypoplastic branchial arch cartilages, and absent quadrate and otic cartilages (Fig. 6D,G,J). Most of the cartilaginous elements of the sobp mRNA-injected tadpole heads were normal; only the otic capsule was mildly hypoplastic in 56.6% of tadpoles (Fig. 6E,H,J). p.R651X mRNA-injected tadpole cranial cartilages, including the otic capsule, were minimally affected; only 3.8% of tadpoles showed a mildly hypoplastic otic capsule (Fig. 6F,I,J). These findings indicate that early changes in gene expression caused by loss or increased expression of Sobp lead to craniofacial skeletal defects, particularly disruption of the otic capsule.

Development of the cranial sensory placodes is a complex process that requires different signaling pathways and expression of transcription factors that induce and/or repress gene expression (Saint-Jeannet and Moody, 2014; Moody and LaMantia, 2015). Previous work from our lab and others showed that Six1 functions at early stages of placode formation (Christophorou et al., 2009; Brugmann et al., 2004), during otic vesicle patterning (Zheng et al., 2003; Ozaki et al., 2004; Laclef et al., 2003) and hair cell formation in the inner ear (Zhang et al., 2017; Li et al., 2020; Ahmed et al., 2012). One important characteristic of Six1 factors that allows them to act in different contexts is their interaction with co-factors that modulate their transcriptional activities (Neilson et al., 2020, 2017; Li et al., 2003, 2020; Kenyon et al., 2005a,b; Heanue et al., 1999; Guo et al., 2011; Brugmann et al., 2004; Ahmed et al., 2012). We identified Sobp as a potential Six1 co-factor based on findings in Drosophila (Neilson et al., 2010; Kenyon et al., 2005a), but little is known about its function in vertebrate development or the mechanisms by which it interacts with Six1. Here, we demonstrate that Sobp is a bona fide Six1 co-factor that can repress the transcriptional activation of Six1+Eya1 target genes. In addition, we show that altering the levels of Sobp disrupts the patterning of embryonic ectodermal domains and otic vesicle genes (Fig. 7).

Sobp was first characterized in Drosophila as part of the retinal gene determination network based on its co-expression in the eye disc with So (Silver et al., 2003; Kenyon et al., 2005a). Our work is the first demonstration that in vertebrates Sobp interacts with Six1 and Eya1. Binding to Six1 was expected as Drosophila Sobp binds to So (Kenyon et al., 2005a), but binding to Eya1 was not expected; this interaction is reminiscent of fly Eya binding to the So co-repressor Dachshund (Chen et al., 1997). Similar to Eya factors, Sobp does not have a DNA-binding domain (Birk et al., 2010) and therefore relies on interaction with transcription factors to modulate transcription. However, unlike Eya1, a cytosolic protein that requires binding to Six factors for nuclear translocation (Ohto et al., 1999; Buller et al., 2001), Sobp is a nuclear protein (Chen et al., 2008; Birk et al., 2010). Our findings demonstrate that Sobp has a functional C-terminal NLS and after deletion of this crucial domain or a larger C-terminal domain containing the NLS, Sobp is located almost exclusively in the cytosol. Interestingly, variants of Sobp containing a deletion of this NLS or of the 220 C-terminal amino acids, including the NLS (p.R651X) still interfere with the transcriptional activation of Six1+Eya1 target genes after binding to Six1 and translocation to the nucleus. These data indicate that an interaction between Six1 and cytosolic proteins is not exclusive to Eya factors (Ohto et al., 1999).

Upon binding to the Six1+ Eya1 complex, Sobp reduces the ability of this complex to activate the transcription of target genes. Our findings indicate that this repressive function occurs through a dose-dependent competition mechanism: low levels of Sobp increase the interaction between Six1 and Eya1, whereas high levels of Sobp disrupt this interaction, leading to at least some Eya1 remaining in the cytosol. As Six1 protein stability is higher in the presence of Eya factors (Patrick et al., 2009), it is possible that low levels of Sobp similarly increase Six1 protein stability without inhibiting its transcriptional function. In addition, as Sobp can partially translocate Eya1 into the nucleus by itself, the increased amount of Six1 bound to Eya1 at lower Sobp concentrations may be a consequence of a higher level of Eya1 in the nucleus.

Because deletion of the C-terminal region of Sobp did not affect its ability to interfere with the transcriptional activation of Six1+Eya1 target genes, we speculate that other remaining regions of the protein are responsible for its activity. For example, the FCS zinc-finger domains (ZF1 and ZF2) (Chen et al., 2008) are similar to those in constituents of the polycomb repressive complex 1 (PRC1), which modifies histones in chromatin leading to transcriptional repression (Chen et al., 2008; Blackledge and Klose, 2021). The SUMO-interacting motifs (SIMs) may also be necessary as SUMOylation also causes transcriptional repression by modifying transcription factors and the assembly of nuclear protein complexes (Sun and Hunter, 2012), and SUMOylation of PRC1 is required for repression (Gill, 2010). It will be important to test the function of these domains in future experiments. It is interesting that individuals with BOR and some with MRAMS present with hearing loss but there are no reports of defects in balance caused by vestibular perturbations (Smith, 2018; Birk et al., 2010; Basel-Vanagaite et al., 2007), whereas the jc mouse presents with both hearing and balance defects (Chen et al., 2008; Calderon et al., 2006). A comparison between the MRAMS p.R661X and the jc S449fsX490 shows that the latter loses additional domains, including both SIMs and a region of the proline rich domain (PRD). As PRDs can be involved in protein/protein interactions (Yu et al., 1994), the jc variant could interrupt binding to Six1 or other factors. A full understanding of the Sobp functional domains and sites of protein-protein interactions is an important next goal.

The PPE gives rise to all cranial sensory placodes and is characterized by the expression of six1 and eya1 (Schlosser and Ahrens, 2004; Pandur and Moody, 2000). sobp also is expressed in the PPE (Neilson et al., 2010), but we noticed by ISH that its expression is weaker in the posterior placodal region that will give rise to the otic and epibranchial placodes. This regional difference in expression is supported by a single cell RNA-seq (scRNA-seq) study that allowed us to compare the anterior and posterior placode regions of neural plate stage embryos (Briggs et al., 2018; Peshkin and Kirschner, 2020). sobp enrichment occurs in the anterior placodal cells that co-express tle1 and tle2, two members of the Groucho family of known Six1 co-repressors (Roth et al., 2010). Together, these findings suggest that at PPE stages, Six1 may induce gene expression in the posterior PPE by transcriptional activation and its activity may be limited in the anterior PPE by the presence of multiple co-repressors, including Sobp (Fig. 7A). This is consistent with the proposed role of Sobp in the fly eye field (Kenyon et al., 2005a).

The changes in gene expression after loss or increased Sobp expression that we observed also show that Sobp is required for the proper balance in size of the other embryonic ectodermal domains: neural plate, neural crest and epidermis. Our data indicate that Sobp appears to have a role in inducing directly or indirectly epidermal genes and repressing neural ectodermal genes (Fig. 7A). It is likely that NC and PPE were disrupted after loss of Sobp because the neural plate domain expanded, and that they were disrupted after increasing Sobp levels because the epidermis expanded. As Sobp and Six1 are co-expressed in the PPE, which is intercalated between the neural plate and epidermis, it is possible that, within the PPE, Sobp in combination with Six1 inhibits a neural plate fate while promoting an epidermal one. Alternatively, as Six1 is exclusively expressed in the PPE (Pandur and Moody, 2000; Briggs et al., 2018; Peshkin and Kirschner, 2020), Sobp function in the other ectodermal domains could be independent of its ability to modulate Six1+Eya1-mediated transcriptional activity. scRNA-seq analyses show that, at neural plate stages, sobp is not expressed in cells expressing sox2 but is co-expressed in krt19+ cells (Briggs et al., 2018). Because increased expression of the truncated p.R651X caused similar changes to full-length Sobp in the PPE, but not to genes in the other domains, we posit that the truncation in Xenopus p.R651X/human p.R661X does not disrupt the interaction with Six1. It is, however, possible that the truncation hampers interactions with other nuclear factors because p.R651X cannot translocate into the nucleus in domains where Six1 is not present. It will be important to determine with which other proteins Sobp interacts in different embryonic ectodermal domains.

Our in vivo KD studies show that Sobp does not appear to be required for six1 expression in the otic vesicle. The decrease in otic six1 after increasing either Sobp or p.R651X is likely due to a reduction in the interaction between Six1 and Eya1 because, in mouse, Eya1 is required for otic vesicle expression of Six1 (Zheng et al., 2003). Although changes in gene expression in the otic vesicle may be influenced by earlier changes in the PPE, because six1 is still expressed in the majority of Sobp crispants, changes in PPE gene expression do not completely account for otic vesicle changes. Instead, we propose that the Sobp-Six1 relationship in the otic vesicle likely impacts dorsal-ventral (D-V) patterning.

D-V patterning of the otic vesicle is crucial for the development of ventral auditory structures and dorsal vestibular structures (Ohta and Schoenwolf, 2018; Nakajima, 2015; Bever et al., 2003). Published data and data presented here show that Six1, Eya1 and Sobp are co-expressed in the medial wall of the otic vesicle (Durruthy-Durruthy et al., 2014; David et al., 2001). Unlike the apparent uniform expression of six1, sobp expression is more intense dorsally, a region that, in mouse, express two known Six1 co-repressors: Dach1 and Dach2 (Ozaki et al., 2004; Li et al., 2002, 2003; Durruthy-Durruthy et al., 2014). These expression data suggest that the transcriptional activation of Six1+Eya1 target genes is repressed in the dorsal-medial domain of the otic vesicle, in agreement with previous findings in mouse that Six1 and Eya1 induce expression of ventral genes, such as Otx1, while repressing dorsal genes, such as Dlx5 (Zheng et al., 2003; Ozaki et al., 2004). We suggest that, in the otic vesicle, there is likely a dose-dependent mechanism of activation and repression of Six1+Eya1-mediated transcriptional activity involving co-factors such as Eya1, Sobp and Dach that contributes to D-V patterning (Fig. 7B). Dorso-medially, higher levels of Sobp and Dach would repress this transcriptional activity, allowing localized expression of dorsal genes such as Dlx5. Ventro-medially, Sobp induces Pax2 expression independently of Six1 expression, as Pax2 expression does not depend on Six1 in mouse (Ozaki et al., 2004); concurrently, Six1 and Eya1 are transcriptionally active in this domain because of lower levels of Sobp and the absence of Dach. Sobp likely interacts with other factors to accomplish otic D-V patterning because, compared with full-length Sobp, changes in gene expression were less frequent after increased p.R651X, which required increased expression of Six1 for nuclear translocation.

Although we have focused our analyses on early stages of otic development, the changes after altering Sobp levels in the embryonic ectodermal domains ultimately led to severe cranial cartilage defects. The differences in severity between loss and increased Sobp expression might be explained by the targeting approach: for CRISPR/Cas9-mediated KD, embryos were injected at the two-cell stage, whereas mRNA injections were performed at the eight-cell stage only in precursors of the PPE and NC. Because Sobp appears to not be expressed in Xenopus branchial arches (Neilson et al., 2010), the defects in the cranial cartilages likely are secondary to changes in the early ectodermal domains rather than disruption of NC patterning in the branchial arches. Although additional work is required to understand the function of Sobp during formation of these cartilages, it is notable that the otic cartilage was consistently hypoplastic when levels of Sobp were altered. Perhaps this defect contributes to the hearing and vestibular deficits observed in jc mice.

In summary, we show that Sobp is a Six1 co-repressor that interacts with Six1 and most likely other factors during otic and craniofacial development. Although Sobp does not have an identified DNA-binding domain (Birk et al., 2010), it functionally modulates the Six1+Eya1 transcriptional complex. The variants found in individuals with MRAMS (Birk et al., 2010) and jc mice (Chen et al., 2008) that have overlapping but also very distinct phenotypes suggest that disruption of different domains in Sobp may perturb interactions with different partners. To better understand the function of Sobp during normal development and in congenital syndromes, it will be essential to assess the role of its various domains and identify any additional binding partners.

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).

A full-length Xenopus tropicalis sobp plasmid was purchased from Open Biosystems (BC154687); subsequent sequence analysis identified it as having ∼95% homology to the predicted sequences for Xenopus laevis L-homeolog sobp (XM_018263336.1; XM_018263335.1) and S-homeolog sobp (XM_018265405.1). The ORF was subcloned into the BamHI site of pCS2⁺ (pCS2⁺-sobp) using the Clone EZ PCR cloning kit (GenScript). pCS2⁺-5′HA-sobp and pCS2⁺-sobp-3′HA were generated using the QuikChange lightning Site-directed mutagenesis kit (Agilent). The same kit was used to sequentially remove three nucleotides located at the 5′ end of the pCS2⁺-3′HA-sobp ORF to generate a construct whose transcribed mRNA does not bind to the designed translation-blocking antisense morpholino oligonucleotide (pCS2⁺-sobpMOI-3′HA; morpholino insensitive). To generate a plasmid containing the human p.R661X variant, the mutagenesis kit was used to introduce into the ORF of pCS2⁺-sobp and of pCS2⁺-5′HA-sobp a stop codon at amino acid 651 (c.1951A<T; p.R651X; pCS2⁺-p.R651X and pCS2⁺-5′HA-p.R651X). Finally, a conserved putative C-terminal nuclear localization signal (NLS) was identified by analyzing the amino acid sequences for Xenopus tropicalis Sobp (NP_001096678.1), Xenopus laevis Sopb (XP_018118825.1; XP_018120894.1), Homo sapiens SOBP (NP_060483.3) and Mus musculus Sobp (NP_780616.4) using cNLS mapper (http://nls-mapper.iab.keio.ac.jp/cgi-bin/NLS_Mapper_form.cgi). The putative NLS was deleted using the mutagenesis kit (pCS2⁺-5′HA-sobp-NLSdel). All plasmids were confirmed by full-length sequencing in both directions.

HEK293T/17 cells (ATCC CRL-11268) were grown in Dulbecco's modified Eagle medium (Hyclone) supplemented with 10% fetal bovine serum (Lonza) and penicillin-streptomycin (Gibco). Cells were plated into 24-well plates (Thermo Fisher Scientific) for luciferase assays; into six-well plates (Fisher) for co-immunoprecipitation experiments (co-IP); and into one-well Nunc Lab-Tek Chamber slides (Thermo Fisher Scientific) for immunofluorescence (IF). Cells were transfected according to the manufacturer's protocol using X-tremeGENE 9 DNA (Sigma) for luciferase assays and IF; and lipofectamine 3000 (Thermo Fisher Scientific) for co-IP. Cells were processed for each assay 48 h after transfection.

HEK293T cells were transfected with combinations of pCS2⁺-3′Flag-Six1, pCS2⁺-5′HA-Sobp, pCS2⁺-5′HA-sobp-NLSdel, pCS2⁺-5′HA-p.R651X and/or pCS2⁺-5′Myc-Eya1. 1 µg of each plasmid was used unless noted in the figure legend. Cells were extracted after 48 h using 500 µl of ice-cold Pierce IP lysis buffer (Thermo Fisher Scientific) plus Halt Protease Inhibitor Cocktail with EDTA (Thermo Fisher Scientific). Cell debris was pelleted at 13,000 g for 10 min at 4°C. The supernatant was subjected to immunoprecipitation using Pierce anti-HA magnetic beads (Thermo Fisher Scientific), Pierce anti-c-Myc magnetic beads (Thermo Fisher Scientific) or Pierce anti-DYKDDDDK magnetic agarose (Thermo Fisher Scientific). Proteins were washed five times and eluted using the Pierce lane marker non-reducing sample buffer (Thermo Fisher Scientific). The eluted proteins were reduced using 100 mM DTT at 100°C for 5 min prior to SDS-PAGE and Western blot. For control experiments, 10 µl of each sample in IP lysis buffer was diluted with Laemmli sample buffer with 2% BME, incubated at 100°C for 5 min prior to SDS-PAGE and Western blot. Immobilon-FL PVDF membranes (Fisher) were probed with mouse anti-HA (6E2, Cell Signaling, 1:1000) to detect Sobp, mouse anti-Myc (9B11, Cell Signaling, 1:1000) to detect Eya1, rabbit anti-Six1 (D5S2S, Cell Signaling, 1:1000) and rabbit anti-β-actin (13E5, Cell Signaling, 1:1000). Secondary antibodies were IRDye 680RD donkey anti-rabbit IgG (92568073, Licor, 1:5000) and IRDye 800CW goat anti-mouse IgG (92532210, Licor, 1:5000). Experiments were repeated at least three independent times. Blots were scanned using the Licor Odyssey infrared imaging system.

Transfected HEK293T cells were harvested and analyzed using the Dual Luciferase Assay kit (Promega) following the manufacturer's directions. Each transfection included 200 ng of pGL3-6XMEF3-Firefly luciferase reporter (Ford et al., 2000) and 20 ng of Renilla luciferase reporter (pRL-TK), in addition to different combinations of pCS2⁺ (control), pCS2+-3′Flag-Six1, pCS2+-5′Myc-Eya1, pCS2+-5′HA-Sobp, pCS2+-5′HA-sobp-NLSdel and/or pCS2+-5′HA-p.R651X (400 ng each). At 48 h post-transfection, cells were resuspended directly in 100 µl of passive lysis buffer (PLB) and 20 µl of lysate was used in the analysis. Experiments were repeated at least five times. ANOVA with Tukey's post-hoc multiple comparisons test was performed using GraphPad Prism 9 software. Expression of exogenous proteins from the transfected plasmids was confirmed by standard western blotting using antibodies described for co-IP (Fig. S2).

HEK293T cells were transfected with different combinations of pCS2⁺-3′Flag-Six1, pCS2⁺-5′HA-Sobp, pCS2⁺-5′HA-sobp-NLSdel, pCS2⁺-5′HA-p.R651X and/or pCS2⁺-5′Myc-Eya1. 2 µg of each plasmid was used unless noted in the figure legend. Cells were processed as described previously (Shah et al., 2020). Briefly, 48 h after transfection, cells were fixed in 4% paraformaldehyde and processed for immunostaining by standard methods using mouse anti-Flag (9A3, Cell Signaling, 1:400), rabbit anti-Myc (71D10, Cell Signaling, 1:250), rabbit anti-HA (C29F4, Cell Signaling) or mouse anti-HA (6E2, Cell Signaling, 1:800) followed by Alexa Fluor 488-conjugated anti-rabbit (4412, 1:1000) and Alexa Fluor 568-conjugated anti-mouse (A1104, 1:1000) secondary antibodies, and DAPI nuclear counterstain (R37605, Thermo Fisher Scientific). Experiments were repeated at least three independent times, and at least five fields per slide were analyzed using a Zeiss LSM 710 confocal microscope.

mRNAs encoding Xenopus tropicalis sobp, Xenopus tropicalis 5′HA-sobp, Xenopus tropicalis sobpMOI-3′HA, Xenopus tropicalis p.R651X and a nuclear-localized β-galactosidase (nβgal) lineage tracer were synthesized in vitro according to manufacturer's protocols (mMessage mMachine kit, Ambion). Antisense RNA probes for in situ hybridization (ISH) were synthesized in vitro (MEGAscript kit; Ambion), as previously described (Yan et al., 2009b).

To knock down endogenous levels of Sobp protein in embryos, a 3′-carboxyfluorescein-labelled translation-blocking antisense morpholino oligonucleotide (MO) was purchased (GeneTools): TCCCCTCTTTTTCCATTTCTGCCAT. The MO binds at the ATG start site for both the predicted L- and S-homeologs for Xenopus laevis (Fig. S3A) and to Xenopus tropicalis sobp. To verify the ability of the MO to block sobp translation (Fig. S3B), Xenopus stage VI oocytes were injected with 9 ng of MO and then injected with either 2 ng of 5′HA-Sobp (MO-insensitive), Sobp-3′HA (MO-sensitive) or sobpMOI-3′HA (MO-insensitive) mRNA. Oocytes were cultured overnight at 18°C. Lysates were prepared and western blotting performed with rabbit anti-HA antibody (C29F4, Cell Signaling) as previously described (Neilson et al., 2017).The specificity of the MO was tested by injecting embryos with 9 ng of MO followed by the immediate microinjection of 50 pg of the MO-insensitive sobp mRNA (sobpMOI-3′HA), and processing embryos for foxd3 expression by ISH (Fig. S3C,D), as described below. Only 29.4% of these embryos had reduced foxd3 expression (Fig. S3C,D) compared with 90.2% for MO alone (Fig. 4E,T), demonstrating significant rescue of the foxd3 phenotype.

Based on the Xenopus laevis sobp genomic sequence obtained from Xenbase (http:/xenbase.org), a 20 bp target was designed by CRISPRscan (http://www.crisprscan.org) during the Xenopus Genome Editing Workshop at the National Xenopus Resource at the Marine Biological Laboratory (Woods Hole, MA, USA). Potential off-targets were predicted by GGGgenome (https://gggenome.dbcls.jp) using the most up-to-date version of the Xenopus laevis genome. The 5′ dinucleotides were converted to GG (Gagnon et al., 2014). The sequence is GGTTCTTGGATGGTACGGTA and targets the L- and S-homeologs for sobp in its second exon that encodes the first conserved region (Fig. S3A) (Chen et al., 2008; Kenyon et al., 2005a). A DNA template was produced by a PCR-based method using a universal reverse primer with a gene-specific forward primer containing a T7 polymerase promoter. The MEGAshortscript T7 transcription kit (Thermo Fisher Scientific) was used for sgRNA synthesis. sgRNA was mixed with Cas9 Protein (PNA Bio) and Texas Red Dextran, Lysine Fixable (Thermo Fisher Scientific) prior to injections. Injected embryos were incubated at 21°C and at least eight embryos were processed for sequencing to confirm DNA editing for the L- and S-homeologs for sobp each time injections were performed. Genotyping primer sequences are listed in Table S3. Insertion/deletion frequencies were calculated with the TIDE software package (http://shinyapps.datacurators.nl/tide/; Fig. S3E,F).

Fertilized Xenopus laevis embryos were obtained by natural mating and in vitro fertilization (Moody, 2000). Injections for ISH and cartilage staining were performed unilaterally with 1 nl of MO (9 ng/nl) or 1nl of mRNA (100 pg mixed with 100 pg nβgal mRNA) in the dorsal-animal and ventral-animal blastomeres of eight-cell stage embryos; these blastomeres predominantly give rise to the neural crest and cranial placodes (Moody and Kline, 1990); or with 5 nl or 10 nl of 75 pg/nl of sgRNA mixed with 0.2 ng/nl of Cas9 protein (PNA Bio) and 0.1% Texas Red Dextran in one cell of two-cell stage embryos. Injections for quantitative real time PCR (qPCR) were performed with 1 nl of MO (2 ng/nl) or 2 nl of mRNA (100 pg mixed with 100 pg nβgal) in each cell of two-cell stage embryos; or with 10 nl of sgRNA/Cas9 solution in one-cell stage embryos. Microinjections were performed according to standard methods (Moody, 2018). Embryos were cultured in diluted Steinberg's solution until fixation or harvest.

Embryos were cultured to neural plate (stage 16-18) and larval (stage 28-34) stages (Nieuwkoop and Faber, 1994), fixed in 4% paraformaldehyde (PFA), stained for β-Gal histochemistry and processed for in situ hybridization (ISH) as described previously (Yan et al., 2009a). In embryos in which the fluorescent label/dextran or nβgal lineage tracer were located in the appropriate tissue domains, the position, intensity and size of the expression domains of sox2, foxd3, six1, krt12.4, dlx5 and pax2 were compared between the injected, lineage-labeled side to the control, uninjected side of the same embryo, thus accounting for inter-embryo variation. Embryos for each assay were derived from a minimum of three different sets of outbred, wild-type parents. Differences in the frequency of gene expression changes were assessed for significance (P<0.05) by the Chi-square test using GraphPad Prism 9. A set of control larvae that was processed for six1 and sobp ISH was embedded in a gelatin-based medium [0.5% gelatin, 30% bovine serum albumin and 20% sucrose, hardened with glutaraldehyde (75 μl/ml)], and vibratome sectioned at 50 µm in the transverse plane. Whole-mount and serial section ISH images were collected with an Olympus SZX16 stereomicroscope coupled to an Olympus UC90 camera and cellSens Entry software.

Three embryos at neural plate (stage 16-18) or three dissected heads at larval (stage 28-34) stages were collected in TRI-reagent (Zymo) and processed for RNA extraction with DNAse I treatment using the Direct-zol RNA Miniprep kit (Zymo). cDNA was synthesized using the iScript Advanced cDNA Synthesis kit (Bio-Rad). qPCR was performed using 5 ng cDNA with the SsoAdvanced Universal SYBR Green Mix (Bio-Rad). Primer sequences are listed in Table S4. qPCR of four biological replicates was performed in duplicate. PCR and data analysis were performed using a CFX Connect thermocycler (Bio-Rad). Statistical analysis was performed with GraphPad Prism 9, with significance calculated by unpaired t-tests followed by a False Discovery Rate approach using the two-stage step-up method (Benjamini et al., 2006) (FDR=1%).

Embryos were grown to tadpole stages and the ones that remained alive were photographed and counted for quantification of survival rates. They were subsequently processed as described previously (Young et al., 2017): fixed in 4% PFA for 1 h at room temperature then incubated in a solution of acid/alcohol containing 0.1% Alcian Blue. When staining was complete, tadpoles were washed in the acid/alcohol solution without Alcian Blue, bleached with a solution containing 1.2% hydrogen peroxide and 5% formamide, and cleared in 2% KOH with increasing concentrations of glycerol.

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). We thank the Xenopus Genome Editing Workshop at the National Xenopus Resource that helped with sgRNA design and training on CRISPR/Cas technology. We also thank Francesca Pignoni, Kristy Kenyon and Dominique Alfandari for helpful discussions during the execution of this work. Finally, we thank many members of the Xenopus community for providing plasmids.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.199684