Pa2G4 is a novel Six1 co-factor that is required for neural crest and otic development

Abstract

Mutations in SIX1 and in its co-factor, EYA1, underlie Branchiootorenal Spectrum disorder (BOS), which is characterized by variable branchial arch, otic and kidney malformations. However, mutations in these two genes are identified in only half of patients. We screened for other potential co-factors, and herein characterize one of them, Pa2G4 (aka Ebp1/Plfap). In human embryonic kidney cells, Pa2G4 binds to Six1 and interferes with the Six1-Eya1 complex. In Xenopus embryos, knock-down of Pa2G4 leads to down-regulation of neural border zone, neural crest and cranial placode genes, and concomitant expansion of neural plate genes. Gain-of-function leads to a broader neural border zone, expanded neural crest and altered cranial placode domains. In loss-of-function assays, the later developing otocyst is reduced in size, which impacts gene expression. In contrast, the size of the otocyst in gain-of-function assays is not changed but the expression domains of several otocyst genes are reduced. Together these findings establish an interaction between Pa2G4 and Six1, and demonstrate that it has an important role in the development of tissues affected in BOS. Thereby, we suggest that pa2g4 is a potential candidate gene for BOS.

1. Introduction

One of the most prevalent birth defects in developed countries is congenital hearing loss (Hilgert et al., 2009). Branchiootorenal Spectrum disorder (BOS), comprised of Branchiootic syndrome and Branchiootorenal syndrome, is the second most common type of autosomal dominant syndromic hearing loss (Smith, 2014). Both syndromes are characterized by malformations of the external ear (derived from the surface ectoderm and branchial arches 1 and 2), hyoid region (derived from branchial arch 2), middle ear and ossicles (derived from branchial arches 1 and 2), and inner ear (derived from the otic placode); BOR is diagnosed when there also are kidney malformations. Although this is a fully penetrant syndrome, patients present with highly variable degrees of craniofacial malformations and hearing loss, the latter of which can be conductive, sensorineural or mixed in type. While the craniofacial abnormalities are not life threatening, affected infants need to be identified as early as possible to intervene for hearing impairments and potential renal dysfunction.

Mutations in two genes have been identified in about half of patients with BOS (reviewed in Moody and Saint-Jeannet, 2014; Smith, 2014; Moody et al., 2015). The gene encoding SIX1, a home-odomain-containing transcription factor, is mutated in about 4% of patients (diagnosed as BOS3 or BOR3); the gene encoding EYA1, a co-factor protein that binds to SIX1 and modifies its transcriptional activity, is mutated in about 40% of patients (diagnosed as BOS1 or BOR1). Genetic screening for mutations in these two genes is considered a best practice to identify at-risk BOS children and begin early treatment (Smith, 2014), but over half of BOS cases cannot be identified with this approach. Thus, it is imperative to identify additional genes that underlie this birth defect.

SIX1 is one of 6 vertebrate transcription factors highly related to Drosophila Sine oculis (SO). SO has been studied in great detail because it is a major player in the formation of the Drosophila visual system (Cheyette et al., 1994; Blanco et al., 2010; Piñeiro et al., 2014; Serikaku and O'Tousa, 1994; Hayashi et al., 2008; Pauli et al., 2005; Pignoni et al., 1997; Yan et al., 2003; Zhang et al., 2006; Zhou et al., 2014). All SO/Six proteins contain a highly conserved Six-type home-odomain, which binds DNA, and an N-terminal domain called the Six domain (SD) (Fig. 1) that interacts with co-factor proteins (Kawakami et al., 2000; Kobayashi et al., 2001; Pignoni et al., 1997). In vertebrates, Six1, Six2, and Six4 are expressed in the cranial placodes, which give rise to sensory organs including the otic placode precursor of the inner ear, and in the developing kidney (reviewed in Brugmann and Moody, 2005; Saint-Jeannet and Moody, 2014). Six1 loss-of-function in Xenopus and chick results in reduced expression of several placode genes and defects in otic development (Brugmann et al., 2004; Christophorou et al., 2009; Schlosser et al., 2008). Six1 knock-down in zebrafish results in loss of inner ear hair cells (Bricaud and Collazo, 2006, 2011), and Six1-null mice show defects in the olfactory placode, inner ear and cranial sensory ganglia (Chen et al., 2009; Ikeda et al., 2007, 2010; Konishi et al., 2006; Laclef et al., 2003; Ozaki et al., 2004; Zheng et al., 2003; Zou et al., 2004). Mutations in SIX1 associated with BOS are not truncations leading to loss-of-function (reviewed in Moody et al., 2015). Instead, most of them occur in the SD (Fig. 1), and there is experimental evidence that some of these disrupt EYA binding (Patrick et al., 2009, 2013; Ruf et al., 2004; Sanggaard et al., 2007). Therefore, we have proposed that a key to identifying additional genes that underlie BOS is to discover novel co-factors that bind to the SD of SIX1 (Moody et al., 2015).

Fig. 1.

The protein binding domain in human and frog Six1 are identical. A comparison of the amino acid sequence of the N-terminal Six Domain (SD, turquoise bar) and homeodomain (HD, black bar) of human and Xenopus Six1 compared to fly SO. Sequences in human versus Xenopus are identical. Sequences that are identical in all three animals are blocked in red, whereas those that differ in fly are blocked in white.

The focus of SO/Six co-factor research has been primarily on Eya (Eyes absent) and Gro (Groucho)/Grg (Groucho-related) family members because they are known to bind to Six proteins, affect their transcriptional activity and are expressed in the same tissues as many Six genes (Bajoghli et al., 2005; Bane et al., 2005; Brugmann and Moody, 2005; Brugmann et al., 2004; Kobayashi et al., 2001; Li et al., 2010; Neilson et al., 2010; Ohto et al., 1999; Ozaki et al., 2002; Zhu et al., 2002). But, SO/Six1 proteins also can interact with other proteins. For example, yeast two-hybrid assays in fly identified more than 25 proteins that interact with SO (Anderson et al., 2014; Giot et al., 2003; Kenyon et al., 2005; Pignoni et al., 1997). As expected, Eya and Gro have the highest interaction scores, but several other proteins were identified as potential co-factors. We took advantage of this data set to screen for potential vertebrate Six1 co-factors using Xenopus, a tetrapod that has highly homologous middle/inner ear gene expression, morphology and hearing characteristics compared to humans (Elepfandt et al., 2000; Mason et al., 2009; Powers et al., 2012; Quick and Serrano, 2005; Schoffelen et al., 2008; van Dijk et al., 2011). Because the amino acid sequence of the SD of fly SO is highly conserved in vertebrate Six1 and the frog SD amino acid sequence is identical to human (Fig. 1; Neilson et al., 2010; Moody et al., 2015), we predicted that proteins that bind to fly SO are likely to bind to vertebrate Six1, and thereby be relevant to BOS. One of the proteins we identified as a potential co-factor is Proliferation associated-2G4 (Pa2G4, also known as Ebp1 or Plfap). It is homologous to the fly SO interactor CG10576, and is expressed in the otic placode, cranial neural crest and nephric mesoderm of frog and mouse (Gray et al., 2004; Neilson et al., 2010).

Previous studies indicate that Pa2G4 contains domains that are thought to interact with other proteins enabling its ability to regulate cell proliferation and differentiation (Kowalinski et al., 2007; Monie et al., 2007; Figeac et al., 2014). Pa2G4 also contains distinct motifs in the C-terminus that are involved in RNA binding (Kowalinski et al., 2007; Monie et al., 2007; Squatrito et al., 2004). In mammals, Pa2G4 can interact with a diverse range of proteins including several receptors, kinases and cell cycle regulators (Xia et al., 2001a; Yoo et al., 2000; Zhang et al., 2008). In mammalian cell culture, it associates with the inactive form of the ErbB-3 receptor, and translocates to the nucleus following receptor activation (Yoo et al., 2000). Several types of cancers demonstrate atypical expression and/or activity of Pa2G4 (Kim et al., 2010; Lu et al., 2011; Zhou et al., 2010; Ko et al., 2016). Loss-of-expression studies in mice and chick indicate a role in regulating proliferation and differentiation during developmental and adult stages of myogenesis (Figeac et al., 2014). Consistent with a role in proliferation, Pa2g4-deficient mice are 30% smaller compared to wild type littermates and show cellular hallmarks associated with growth retardation (Zhang et al., 2008). Nonetheless, to our knowledge there are no studies of its function in the craniofacial development of any vertebrate.

Herein we show that Xenopus Pa2G4, which has high similarity in structure to the human protein, binds to the Six1 protein and affects the formation of the Six1-Eya1 complex in human embryonic kidney cells. In these cells, which have endogenous levels of Six1, Pa2G4 represses transcription of two different luciferase reporters, whereas in Xenopus fibroblast-like cells Pa2G4 activates the same reporters. In the embryo, pa2g4 is expressed in the embryonic precursors of the tissues affected in BOS: the cranial neural crest (branchial arch 1 and 2 precursors of the middle ear, ossicles and mammalian external ear), otic placode/otocyst (precursor of the inner ear and auditory nerve ganglion), and kidney. Loss of Pa2G4 function leads to down-regulation of neural border zone, neural crest and cranial placode genes, and to a concomitant expansion of neural plate genes. Gain-of-function leads to a broader neural border zone, expanded neural crest and altered cranial placode domains. In both loss- and gain-of-function experiments, the later expression domains of branchial arch and otocyst genes are significantly impacted. Together these findings establish an interaction between Pa2G4 and Six1, and demonstrate that it has an important role in the development of tissues affected in BOS.

2. Material and methods

2.1. Cloning

Full-length Xenopus laevis pa2g4 was purchased (Open Biosystems; BC084760). The pa2g4 open reading frame (pCS2⁺-pa2g4) and a 5′HA tagged version (pCS2⁺-5′HA-pa2g4) were generated and introduced into the Xba1 site of pCS2⁺(Genscript, Clone EZ PCR cloning kit). A 3′HA tagged version that contains the native 5′UTR (pCS2⁺-5′UTR-pa2g4-3′HA) was constructed (Agilent, Quik Change Lightning Site-directed Mutagenesis kit) by first introducing the 3′-HA tag and then adding the native 5′-UTR. 5′ and 3′ FLAG-tagged versions of Xenopus laevis six1 (Pandur and Moody, 2000) were generated and inserted into StuI and XbaI cut pCS2⁺ vector (pCS2⁺-5′FLAG-six1; pCS2⁺-3′FLAG-six1). All plasmids were confirmed by full-length sequencing and used as templates for in vitro synthesis of mRNAs (Ambion, mMessage mMachine kit).

2.2. Embryo microinjection

Wild type, outbred Xenopus laevis embryos were obtained by natural matings, cultured and microinjected as previously described (Moody, 1999, 2000). pa2g4 mRNA (100 or 200 pg/nl) was mixed with nuclear localized βgal mRNA (nβgal, 50 pg/nl) as a lineage tracer, and 1 nl of the mixture was microinjected into the two lateral-animal blastomeres of the 8- or 16-cell embryo that are the major progenitors of the neural crest and cranial placodes (Moody, 1987; Moody and Kline, 1990). The uninjected side of the embryo was used as an internal control.

2.3. Morpholino knock-down of expression

To knock-down endogenous levels of Pa2G4 protein in the embryo, two translation-blocking antisense morpholino oligonucleotides (MOs) that target both homeologues were purchased (Gene-Tools, LLC) (Supplemental Fig. 1). An equimolar mixture of pa2g4 MOs (1.125, 2.25, 4.5, or 9.0 ng per blastomere) was microinjected into two blastomeres of the 8-cell embryo, as above. Both MOs are lissamine labeled so that cells in the embryo in which knock-down was achieved could be identified. To verify the ability of the MOs to block pa2g4 translation, Xenopus oocytes were first injected with 9 ng of the MO cocktail and then injected with either 2 ng of 5′UTR-pa2g4-3′HA mRNA (MO sensitive) or 2 ng of 5′HA-pa2g4 mRNA (MO insensitive, i.e. rescue mRNA). The oocytes were cultured overnight at 18 °C, lysates prepared and Western blotting performed with an HA antibody as previously described (Neilson et al., 2012) (Supplemental Fig. 2A, B). In addition, the reversal of the MO knock-down phenotype in whole embryos was demonstrated by injecting 400 pg of rescue mRNA (5′HA-pa2g4) immediately after embryos were injected with 2.25 ng of the MO cocktail (Supplemental Fig. 2A and B).

2.4. Whole embryo in situ hybridization

Embryos were cultured to stages: 12.5–13 (neural border zone stages); 16–18 (neural plate, pre-migratory neural crest, pre-placodal ectoderm stages); 22–24 (neural crest migration, otic placode stages); and 28–36 (branchial arch, otocyst stages) (Nieuwkoop and Faber, 1994). They were processed for in situ hybridization (ISH) using antisense Dig-labeled RNA probes of genes expressed in various ectodermal domains that were synthesized as previously described (Yan et al., 2009). The expression patterns of neural border, neural crest, pre-placodal ectoderm (PPE), neurogenic placodes, neural plate, branchial arch and otocyst genes were compared on the experimental (injected) and control (uninjected) sides of the same embryo; samples were derived from at least three different clutches of eggs from different sets of outbred, wild-type parents. In some cases, the widths of the expression domains were measured with an eyepiece micrometer and control and experimental sides were compared by the paired t-test.

2.5. Vibratome sectioning

Embryos processed for otocyst gene expression by ISH at larval stages were embedded in a mixture of gelatin: bovine serum albumin: saccharose (2.2 g:135 g:90 g/450mls PBS) cured with glutaraldehyde. Specimens were sectioned with a vibratome at 25–50 µm, mounted on glass slides and coverslipped with 10% glycerol in PBS. The dorsal-ventral diameter of the otocyst on control and manipulated sides of embryos injected with either pa2g4 MOs or mRNA was measured with an eyepiece micrometer at 65X. At least three sections per embryo were measured. Measurements of control versus experimental sides of the same embryo were compared by the paired t-test.

2.6. Proliferation assay

Embryos were injected on one side with pa2g4 mRNA (mixed with 50 pg/nl cytoplasm-localized βgal mRNA [cβgal]) or pa2g4 MOs, as described above. Upon reaching stages 16 (MO-injected) or 18 (mRNA-injected), the vitelline membranes were manually removed and embryos were fixed in 4% paraformaldehyde in PBTw, processed for βGal histochemistry, refixed and processed for whole mount immunostaining with an antibody that recognizes phosphorylated Histone 3 (pH3; #9701, Cell Signaling), which marks cells in M-phase of mitosis (Hendzel et al., 1997), as previously described (Zaghloul and Moody, 2007). Control and experimental sides of each embryo were photographed and an equivalent 100 µm X 200 µm area drawn over the lineage labeled neural plate border zone on both control and experimental sides. Each image was assigned a randomized code to blind the experimental manipulation of the sample to the authors performing the cell counts. Counts were made independently by two of the authors and their results averaged. Control and experimental cell counts were compared by the paired t-test.

2.7. Cell death assay

Embryos were injected on one side with pa2g4 MOs, fixed at stage 16 and prepared for staining as above. They were processed for whole mount TUNEL labeling according to the manufacturer's instructions (Roche, In Situ Cell Death Detection kit) as adapted for Xenopus embryos (Zaghloul and Moody, 2007). Two authors blinded to the experiment counted the numbers of TUNEL-labeled cells on control versus experimental sides, as above. Positive and negative controls were simultaneously processed according to manufacturer's instruction to verify the efficacy of each assay.

2.8. Co-IP assays

Human embryonic kidney cells (HEK 293T cell line), which endogenously express genes in the SIX1 pathway (Chai et al., 2006), were transfected with the various plasmids using X-tremeGENE-HP DNA transfection reagent (Sigma) following manufacturer's instruction (1 µg of DNA per well of a 6 well plate). Cells were extracted 48 h after transfection in 1X MBS, 1% TritonX-100, and a protease/phosphatase inhibitor cocktail (Pierce) with 5 mM EDTA. Lysates were spun at 16,000 g for 30 min at 4 C, and the supernatants were incubated with 2 µg of anti-Flag antibody (Sigma, M2) bound to protein-AG-agarose beads (Pierce) overnight at 4 C. Beads were washed 3 times with 1 ml of extraction buffer, eluted using reducing Laemmli buffer, proteins separated by SDS-PAGE, and transferred to PVDF (Amersham) using a semi-dry transfer apparatus. Membranes were blocked using 5% non-fat dry milk in TBS with 0.1% Tween20. Western blots were performed using anti-HA (G036, ABM), anti-Myc (9E10, DSHB) and anti-Flag (M2, Sigma) antibodies at 0.2 µg/ml followed by incubation using the Clean-Blot IP detection kit (Pierce). Signal detection was performed using chemiluminescence exposed to X-OMAT film (Kodak). Both 5′-FLAG-Six1 and 3′FLAG-Six1 were used without significant differences. Unless noted otherwise, the results with the 3′FLAG-Six1 are presented.

2.9. Luciferase assays

HEK 293T or Xenopus fibroblast-like cells (XTC-2 cell line; Pudney et al., 1993) were transfected as above. Each well of a 6-well plate was transfected with 100 ng of a reporter of Six1/SO transcriptional activity, ARE-luciferase (Silver et al., 2003; Bricaud and Collazo, 2011) and 10 pg of a normalizer plasmid containing Renilla luciferase under the control of a CMV promoter (pRL-CMV). The same assay was performed with a Tfap2α-luciferase reporter that we constructed; 2500 bp of Xenopus laevis genomic sequence upstream of the Tfap2α transcription start site (Karpinka et al., 2015; Vize and Zorn, 2016) was cloned by PCR upstream of the luciferase sequence in the pGl3 plasmid (Promega) using J-Strain Xenopus laevis genomic DNA (National Xenopus Resource; Pearl et al., 2012) and the In-Fusion Kit (Clontech). Tfap2α was chosen because in the embryo it is down-regulated by six1 mRNA injection (13/20 embryos) and in an in vitro assay the promoter binds Six1 (data not shown). 300 ng of all other plasmids were co-transfected with the reporter plasmids. 24 h after transfection, cells were rinsed with 1X PBS and luciferase levels were assayed using the Dual-Luciferase Reporter Assay System (Promega) according to manufacturer's instructions. In brief, proteins were extracted from the cells in 500 µl of Passive Lysis Buffer. 10 µl of lysate was used for the assay and incubated with 25 µl of firefly luciferase substrate followed by 25 µl of Renilla luciferase substrate. Luciferase measurements were read in a luminometer. For each sample, the firefly luciferase measurement was normalized to the Renilla luciferase to give a luciferase ratio. This ratio was then divided by the luciferase ratio of the reporter plasmids co-transfected with empty vector in order to determine the activation of the reporter over the background levels. Experiments were repeated at least 3 times on different days; the histogram represents the average from these three experiments.

3. Results

We hypothesized that new candidate genes that underlie BOS will be found amongst the proteins that associate with the Six1 transcription factor (Moody et al., 2015). Using the Drosophila interactome as our guide, we identified new Six1 candidate co-factors, and verified that they are expressed in the embryonic precursors of tissues that are most relevant to BOS: the neural crest of branchial arch (BA) 1 and BA2 (middle ear, ossicles and mammalian external ear), otic placode (inner ear) and kidney (Neilson et al., 2010). Pa2G4, also known as Ebp1 or Plfap, is one of these factors. To verify that Pa2G4 is a Six1 interactor and is involved in the development of BOS-affected tissues, we performed experiments to determine if it binds to Six1, affects Six1 transcriptional activity, and/or is required for normal gene expression in the precursors of the middle and inner ear.

3.1. Does Pa2G4 interact with Six1?

In fly, CG10576 interacts with SO in a yeast-two-hybrid screen (Giot et al., 2003), but this association has not been determined with vertebrate proteins. When Xenopus six1-flag was transfected into HEK 293T cells and the protein immunoprecipitated using the Flag antibody, endogenous human PA2G4 was found to be associated (shot-gun mass spectrometry data not shown). To confirm these results, we co-transfected Xenopus six1-flag plus Xenopus pa2g4-HA and immuno-precipitated with either the anti-Flag or the anti-HA antibodies; in these experiments the two proteins also associated (Fig. 2A for FLAG IP). To test whether Pa2G4 modifies Eya1 binding to Six1, we transfected cells with equimolar amounts of six1-flag and eya1-myc plasmids, and varying amounts of pa2g4-HA plasmid (Fig. 2B). When Pa2G4 levels were at a 0.125 or 0.25 M ratio to Six1+Eya1, there was no detectable effect on Eya1 co-immunoprecipitation by Six1. However, when Pa2G4 levels were increased to a 0.5 or 1.0 M ratio to Six1+Eya1, the amount of Eya1 that co-immunoprecipitated with Six1 dramatically decreased. These results indicate that Pa2G4 can compete with Eya1 for Six1 binding.

Fig. 2.

Six1 binds Pa2G4. (A) HEK 293T cells were transfected with six1-Flag and pa2g4-HA plasmids. Pa2G4-HA was detected by Western blot (WB) when Six1-Flag was immunoprecipitated with an anti-Flag antibody (IP). (B) HEK 293T cells were transfected with equimolar amounts of six1-Flag and eya1-myc plasmids. In addition they were transfected with varying relative amounts of pa2g4-HA plasmids: equal (1.0), half (0.5), quarter (0.25), eighth (0.125) or none (0). Eya1-Myc was detected by Western blot (WB) when Six1-Flag was immunoprecipitated (IP). Eya1 bound to Six1 when Pa2G4 levels were low (0.125 or 0.25 M ratios), but did not bind when Pa2G4 levels were increased (0.5 or 1.0 M ratios). Left-most lane shows no transfection.

3.2. Does Pa2G4 alter Six1 transcriptional activity?

Since Pa2G4 can bind to Six1, we used luciferase reporter assays to determine whether it affects Six1 transcriptional activity. In one set of assays we used the ARE-luciferase reporter plasmid, which consists of 7 Six binding sites and has been used previously in Drosophila S2 cells and in primate COS-7 cells to assay SO/Six1 activity (Silver et al., 2003; Bricaud and Collazo, 2011). When transfected into HEK 293T cells, Six1 alone and Pa2G4 alone repress reporter activity whereas Eya1 alone activates it (Fig. 3A). Transfecting Six1+Eya1 activates transcription by more than 2-fold; adding Pa2G4 to this transcription mix reproducibly decreased the activation, albeit not to a significant level. In a second set of assays, we cloned the endogenous enhancer-promoter of Tfap2a upstream of luciferase and used this plasmid (AP2-luciferase) as a reporter. In HEK 293T cells, Six1 alone induced the reporter expression significantly whereas Eya1 alone had no significant effect. Again, Pa2G4 alone significantly represses transcription of the AP2-luciferase reporter. Comparable to the ARE-luciferase reporter, Six1+Eya1 significantly activates transcription and addition of Pa2G4 dampened this effect (Fig. 3B). These results indicate that in HEK 293T cells, which contain endogenous Six1 and Eya1, Pa2G4 alone can repress both an artificial and a natural reporter. In addition, Pa2G4 can modestly modulate the transcriptional activity of the Six1-Eya1 complex in these cells. We then tested the ability of Six1 or Pa2G4 alone to induce the two reporters in a Xenopus fibroblast-like cell line (XTC-2). We found that Pa2G4 increased the transcription of both reporters, whereas Six1 had no effect on its own (Fig. 3C). Together these experiments show that Pa2G4 can affect gene expression both positively (XTC-2) or negatively (HEK 293T) depending on the cell context. Because we do not know the complete catalogue of transcription factors endogenous to each of these cell lines, it is unclear at present whether Pa2G4's transcriptional activity is entirely dependent on its interaction with Six1.

Fig. 3.

Pa2G4 affects the transcriptional activity of two different reporters. Histograms representing the relative luciferase activity in cells transfected with the various constructs and luciferase reporters. The luciferase activity was normalized to Renilla expressed with a constitutive promoter. All values are represented as a ratio to the empty vector (CS2), error bars represent the standard deviation of 3 independent experiments performed on 3 different days. (A) In HEK 293T cells, Six1 and 2G4 significantly repress whereas Eya1 and Eya1+Six1 significantly activate expression of the ARE-luciferase reporter. Addition of Pa2G4 dampens this activation. (B) In HEK 293T cells, Six1 and Six1+Eya1 significantly activate and 2G4 significantly represses expression of the AP2-luciferase reporter. Addition of Pa2G4 dampens this activation. (C) In XTC cells, Pa2G4 significantly activates the ARE-luciferase reporter; AP2-luciferase expression also is enhanced but did not reach significance. *=p < 0.05; **=p < 0.01; ***=p < 0.001.

3.3. Is Pa2G4required for craniofacial development?

If Pa2G4 is relevant to BOS, it should be involved in the development of craniofacial tissues that are affected. Therefore, we tested whether it has a role in the development of: 1) the cranial neural crest, subsets of which (BA1 and BA2) contribute to the middle ear, ossicles, and mammalian external ear (Fekete and Noden, 2013; Thompson and Tucker, 2013; Cox et al., 2014; Sandell, 2014); and 2) the cranial placodes, a subset of which give rise to the inner ear. This was accomplished by assessing the expression of a large number of relevant genes after Pa2G4 loss-of-function and gain-of-function approaches.

For loss-of-function analyses, we designed MOs that bind to the 5′UTR or to the 5′ORF (including the ATG start site) to block the translation of endogenous pa2g4 transcripts (Supplemental Fig. 1). An equimolar mixture of the two MOs, which were lissamine-labeled to lineage trace the affected cells, was injected into sites in both the dorsal and the ventral animal blastomeres on one side of the 8-cell or 16-cell embryo. Detailed fate maps show that this targets the delivery of the MOs to the embryonic progenitors of the neural crest and cranial placodes (Fig. 4A; Moody, 1987; Moody and Kline, 1990), thus reducing non-specific effects and also providing the uninjected side of the embryo as an internal control. We confirmed the efficacy of the MO mixture in reducing pa2g4 translation by expressing 5′UTR-pa2g4-3′HA mRNA, which contains both MO recognition sites, in the presence and absence of the MOs (Supplemental Figs. 1 and 2). We confirmed the specificity of the MO effect by showing that the 5′HA-pa2g4 mRNA, which is resistant to MO knock-down, can rescue the MO-induced foxd3 phenotype (Supplemental Figs. 1 and 2).

Fig. 4.

Knock-down of endogenous Pa2G4 disrupts the formation of the neural border zone and its derivatives. (A) Xenopus embryos showing “typical” cleavage patterns at the 8-cell stage (8CS) and 16-cell stage (16CS). Diagram on right shows the 16CS fate map of which blastomeres are the major progenitors of the neural plate (NP), neural crest (NC), pre-placodal ectoderm (PPE) or epidermis (EPI) (based on Moody, 1987). Green asterisks indicate the sites of microinjection that target knock-down to the precursors of the neural crest and cranial placodes. (B) In situ hybridization assays of gene expression at neural plate or neural tube stages monitored after unilateral knock-down of endogenous Pa2G4. Right side of each image (except zic2-b) is the MO-injected side, and the left is the uninjected, control side that demonstrates the normal gene expression pattern. Genes expressed in the neural border zone are

Since pa2g4 mRNA is expressed in the neural border zone, cranial neural crest and pre-placodal ectoderm (PPE) (Neilson et al., 2010), we analyzed the expression of genes that mark these embryonic precursors of the branchial arches and inner ears. Reducing endogenous Pa2G4 significantly reduced the size and/or staining intensity of the expression domains msx1, pax3 and tfap2a in a high proportion of the embryos (Table 1; Fig. 4B). Since border zone genes are required for the subsequent formation of the neural crest and PPE (reviewed in Groves and LaBonne, 2014; Saint-Jeannet and Moody, 2014), we next evaluated the effect of reducing Pa2G4 on genes expressed at later stages when neural crest and PPE have segregated. Neural crest (foxd3, sox9), PPE (six1, sox11, irx1), epibranchial placode (sox3) and otic placode (sox9) genes were down-regulated in a high percentage of embryos (Table 1; Fig. 4B); note that in Xenopus, sox9 is expressed in both the neural crest and the otic placode (Park and Saint-Jeannet, 2010; Lee and Saint-Jeannet, 2011). There was a concomitant expansion of the neural plate domains of several genes (sox2, zic2, sox11, irx1) (Table 1; Fig. 4B) and maintenance of a neural ectodermal gene (foxd4l1) several stages after its expression is normally extinguished (69.7% of embryos, n=33; Fig. 4B). These results suggest that reducing Pa2G4 expands the neural plate at the expense of neural border zone derivatives. For sox2 and zic2, the expansion of the width of the neural plate domain also resulted in fainter staining intensity, which for zic2 obscured its distinct neural crest domain (Fig. 4B). Interestingly, when the pa2g4 MO containing cells extended ventrally into the epidermis, zic2 was ectopically expressed (66.7%, n=15; Fig. 4B, zic2-b).

Table 1.

Percentage of embryos (n) that show a change in gene expression after Pa2G4 knock-down

Neural border	msx1	pax3	tfap2a
reduced	95.3 (43)	84.2 (38)	89.2 (28)
Neural crest	foxd3	sox9
reduced	90.2 (92)	66.7 (39)
PPE/Placodes	six1	sox11	irx1	sox9	sox3
reduced	85.9 (71)	54.5 (44)	66.7 (57)	87.2 (39)	71.4 (14)
Neural plate	sox2	zic2	sox11	irx1
broader	83.3 (48)	85.4 (41)	77.3 (44)	64.9 (57)
Branchial Arches	sox9	dlx5	tbx1	irx1
reduced	73.7 (19)	84.2 (19)	75.0 (4)	50.0 (20)
Otocyst	six1	pax2	bmp4	irx1	sox9	dlx5	tbx1	otx2
reduced	65.4 (26)	64.7 (17)	16.7 (12)	85.0 (20)	78.9 (19)	85.7 (21)	100 (4)	82.4 (34)

To assess the later consequences of the early effects on neural crest and PPE genes, we analyzed a number of genes expressed in the branchial arches and otocyst in pa2g4 morphants raised to larval stages (st 28–36). Four genes that are expressed in the branchial arches (irx1, sox9, dlx5, tbx1) were notably reduced (Table 1; Fig. 4C). For sox9 and dlx5, this included BA1 and BA2 that across vertebrates contribute subsets of cells to the middle ear, ossicles and mammalian external ear (Fekete and Noden, 2013; Thompson and Tucker, 2013; Cox et al., 2014; Sandell, 2014). In the otocyst, six1 and pax2 expression was somewhat reduced compared to the contralateral control side, whereas bmp4 expression was not significantly altered (Table 1; Fig. 4C). However, the levels of expression of several genes expressed downstream of Six1 and Pax2 (irx1, sox9, dlx5, tbx1, otx2) were so dramatically reduced that the otocyst was barely visible (Table 1; Fig. 4C). For otx2 there was no staining in 41.2% of the embryos (n=34), and in the remainder the ventral expression domain was significantly reduced in size (Fig. 4D). Gene expression data from early tadpole stages (st. 36–39) were similar (not shown), indicating that the phenotypes were not simply due to a developmental delay. To determine if the reduced staining was due to loss of otic tissue, embryos at stages 34–36 were sectioned and the dorsal-ventral diameters of the control and morphant otocysts in the same embryo compared. In addition to reduced gene expression, Pa2G4 knock-down also significantly (p < 0.00001, paired t-test) reduced the size of the otocyst (Fig. 5A and C), a phenotype that is similar to that seen in Six1-null mouse embryos (Zheng et al., 2003).

Fig. 5.

Loss and gain of Pa2G4 have different effects on the otocyst. (A) Two examples of pa2g4 morphants. In both, the dorsal-ventral diameter of the otocyst is smaller on the MO-injected side and gene expression levels are reduced. (B) In contrast, when Pa2G4 levels are increased by mRNA injection, the otocyst diameter is not significantly different from the control side of the same embryo, but gene expression levels also are reduced. (C) The dorsal-ventral diameters of pa2g4 morphants, measured in relative units by an eyepiece micrometer, were significantly smaller compared to the control side of the same embryo (n=13; p < 0.00001, paired t-test), whereas they were not significantly different in embryos injected with pa2g4 mRNA (n=20; p > 0.05).

3.4. Increasing Pa2G4 alters the sizes of neural crest, placode and neural plate domains

We next increased Pa2G4 levels by microinjecting mRNA (100 pg or 200 pg) into the neural crest and PPE precursor blastomeres (Fig. 4A). Complimentary to the loss-of-function phenotypes, the domains of three neural border zone genes were expanded (Table 2; Fig. 6A); measuring the widths of msx1 and pax3 expression domains showed that they both were significantly broader compared to control sides (p < 0.0001, paired t-test). For msx1 and tfap2a, the frequency of the broader domain phenotype was comparable between the 200 pg and 100 pg doses, whereas for pax3, the 200 pg dose was significantly more effective. At subsequent stages, the expression domains of neural crest genes (foxd3, sox9, zic2) were expanded (Fig. 6B); measuring the width of the foxd3 expression domain confirmed it was significantly broader (p < 0.00001, paired t-test). Interestingly, the neural crest domain of foxd3 was more frequently broader at the 200 pg dose, whereas that of sox9 was more frequently broader at the 100 pg dose; dosage did not significantly alter the effect on zic2 (Table 2). Six1 expression in the PPE was sensitive to the level of Pa2G4. Injection of 200 pg of pa2g4 mRNA repressed six1 expression, whereas 100 pg broadened its domain (Fig. 6C; Table 2). In contrast, irx1, which is initially expressed in the neural border zone and then in the PPE (Glavic et al., 2004), was expressed at lower levels on the injected side in some embryos and at higher levels in others by both doses of pa2g4 mRNA (Fig. 6C; Table 2). Likewise, the otic placode expression of sox9 was repressed in some embryos and expanded in others at both doses (Table 2). In contrast, the domains of genes associated with the neurogenic precursors in the cranial ganglion placodes (sox2, sox11) were expanded by both doses of pa2g4 (Fig. 6C; Table 2); for sox2, this effect was significantly greater at the 100 pg dose. Expressing Pa2G4 in the neural plate surprisingly showed phenotypes similar to Pa2G4 knockdown. The neural plate domains of zic2, sox11 and irx1 were obviously enhanced and/or broader (Fig. 6B and C; Table 2). Although the sox2 neural plate domain was infrequently scored broader at the 100 pg dose (Fig. 6C; Table 2), measuring its width showed a significant difference compared to control sides (p < 0.01; paired t-test). In addition, foxd4l1 expression was maintained several stages after it normally is extinguished (85.8% of embryos, n=127; Fig. 6C). Interestingly, when Pa2G4 was ectopically expressed in the major neural plate precursor blastomere, it ectopically induced msx1 (81.1%, n=42), foxd3 (95.6%, n=45) and zic2 (79.3%, n=29) in the midline of the neural plate (Fig. 6D). Since six1 is expressed at a negligible level in the neural plate (Pandur and Moody, 2000; Ghanbari et al., 2001), this phenotype appears to result from a Six1-independent action of Pa2G4. Consistent with this, we found that the ectopic neural plate induction of foxd3 expression induced by Pa2G4 was nearly eliminated by co-expression of Six1 (5.5%, n=72) (Fig. 6E). In addition, increasing Six1 above endogenous levels ameliorated the Pa2G4 expansion of the foxd3 domain in the neural border; the domain was still slightly broader than control but was reduced in intensity (81.6%, n=76; Fig. 6E). Finally, in Six1 morphants, Pa2G4 induced ectopic neural plate expression of msx1 (80%, n=15) and foxd3 (100%, n=20) at frequencies similar to expression in wild type embryos. In addition, Pa2G4 expression in Six1 morphants expanded msx1 (100%, n=6) and foxd3 (96.0%, n=25) domains beyond the border zone into the epidermis (e.g., Fig. 6E). These results in Six1 morphants indicate that Pa2G4 has different effects on craniofacial ectodermal domain sizes that vary depending on the presence or absence of Six1.

Table 2.

Percentage of embryos in which gene expression domains changed after increasing Pa2G4 levels

Neural border	msx1-200	msx1-100	pax3-200*	pax3-100	tfap2α-200	tfap2α-100
broader	100 (27)	100 (20)	83.3 (48)	58.1 (43)	70.7 (41)	83.9 (31)
Neural crest	foxd3-200*	foxd3-100	sox9-200*	sox9-100	zic2-200	zic2-100
broader	88.0 (125)	76.2 (42)	59.7 (67)	83.9 (31)	96.6 (58)	97.1 (34)
PPE	six1-200*	six1-100	irx1-200	irx1-100
reduced	53.1 (81)	15.3 (72)	55.3 (47)	57.7 (52)
broader	16.0 (81)	73.6 (72)	40.4 (47)	32.7 (52)
Placodes	sox11-200	sox11-100	sox2-200*	sox2-100	sox9-200	sox9-100
reduced	19.5 (77)	19.6 (56)	17.6 (68)	1.9 (53)	43.1 (51)	51.4 (37)
broader	70.1 (77)	73.2 (56)	67.6 (68)	88.7 (53)	51.0 (51)	43.2 (37)
Neural plate	sox11-200	sox11-100	sox2-200*	sox2-100	zic2-200	zic2-100	irx1-200	irx1-100
broader	89.7 (78)	100 (44)	8.0 (100)	29.5 (78)	89.1 (64)	83.3 (36)	90.9 (22)	6.7 (43)
Branchial Arches	sox9-200	dlx5-200	tbx1-200	irx1-200
reduced	50.9 (51)	33.3 (33)	37.0 (27)	39.4 (33)
Otocyst	six1-200	pax2-200	bmp4-200	irx1-200	sox9-200	dlx5-200	tbx1-200	otx2-200
reduced	53.8 (39)	50.0 (32)	47.8 (23)	83.9 (31)	61.0 (41)	42.9 (21)	51.4% (37)	64.0 (25)

Legend: Number after gene name indicates picograms of pa2g4 mRNA injected into border zone progenitor blastomeres. Number in parentheses indicates the number of embryos analyzed. All experiments were performed at least three independent times.

^*indicates a significant difference between 100pg and 200pg data (Chi-squared analysis, p<0.05)

Fig. 6.

Pa2G4 gain-of-function enlarges the neural border zone and neural crest at the expense of placodes. (A) Three genes required for neural border zone formation are expanded when the level of Pa2G4 is elevated. Black bars indicate the width of the expression domain on the control side of an embryo, and red bars indicate the width in the same embryo on the side injected with 200 pg of pa2g4 mRNA, indicated by the pink lineage tracer. (B) Three genes required for neural crest formation are expanded (red bars for foxd3, sox9; red arrow for zic2) on the pa2g4 side of the embryos, compared to control sides (black) of the same embryo. For zic2, the neural plate also is expanded (red bar). For sox9, the otic placode domain (red arrow) is reduced. (C) Genes required for cranial placode formation are differentially affected by increasing pa2g4. For six1, a low dose (100 pg) broadens the PPE domain (red bar), whereas 200 pg reduces the intensity of expression (red arrow). The placode domains of sox11 and sox2 are enlarged at both doses (also see Table 2). The placode domain of irx1 is reduced at 100 pg, but either reduced (irx1-200-a) or expanded (irx1-200-b) with 200 pg. The neural plate expression of sox11 and irx1 are expanded (red bars) at either dose. The sox2 neural plate domain infrequently appeared broader, but quantitation verified that it is (see text). foxd4l1 expression is maintained in the neural plate long after it should normally be extinguished (red arrows). (D) Expression of pa2g4 in the precursor of the neural plate (see Fig. 4A) causes the ectopic expression of neural border (msx1) and neural crest (foxd3, zic2) genes in the neural plate midline (red arrows). (E) Co-expression of Six1 with Pa2G4 (left image) eliminates the ectopic neural plate expression of foxd3 (red arrow; cf. 5D) and ameliorates the significant expansion of foxd3 in the border zone (cf. 5B), suggesting that these phenotypes are due to interaction with another factor. Expression of Pa2G4 in the absence of Six1 (right image) preserves the ectopic neural plate (np) expression of foxd3 and broadly expands it past the border zone (bz) into the lateral epidermis (epi).

Consistent with the early reduction of otic placode genes (six1, irx1, sox9) increasing Pa2G4 ultimately resulted in reduced expression of a number of otocyst marker genes (Table 2; Fig. 7), regardless of the mRNA dose (data not shown). In contrast to the Pa2G4 morphant phenotype, the otocysts in these larvae were not significantly reduced in size (p > 0.05), but otic gene expression was nonetheless reduced (Fig. 5B and C). Interestingly, despite broadening of the pre-migratory neural crest gene domains, abnormalities in branchial arch gene expression domains were noted only at low frequencies at these later stages (Table 2; Fig. 7); this is consistent with the luciferase reporter assays that show that Pa2G4 effects are cell context dependent.

Fig. 7.

Gain of Pa2G4 disrupts gene expression in the otocyst. Otic genes are either reduced in intensity or undetected in the otocyst on the injected side (right image, red arrows, pink lineage tracer). Left image shows gene expression in the otocyst (black arrows) on the control side of the same embryo. Green arrows denote reduction in gene expression in the neural crest-derived branchial arches. Note that the otocysts are present in all cases. For all phenotypes in this figure, please see Table 2 for frequencies and the number of embryos analyzed.

3.5. Are the effects of altering levels of Pa2G4 due to changes in cell death or proliferation?

In mammalian cell lines, Pa2G4 has been implicated in cell cycle regulation, growth control and cancers. To determine if Pa2G4 affects proliferation in the developing embryo, we stained embryos for the expression of phosphorylated Histone3 (PH3), which marks cells in M phase (Hendzel et al., 1997). This was performed in both pa2g4 morphants and in pa2g4 mRNA-injected embryos. We found a significant reduction in the number of PH3-positive cells in Pa2G4 morphants (Fig. 8A, p < 0.05, paired t-test), indicating that it is required for normal levels of proliferation. However, increased levels of Pa2G4 do not significantly affect proliferation (Fig. 8B, p > 0.05, paired t-test). In many developing tissues, cells that are prevented from entering the cell cycle undergo apoptotic cell death. To test this possibility, we stained pa2g4 morphants for DNA strand breaks using the TUNEL assay. We found that when Pa2G4 was reduced there were significantly fewer apoptotic cells (Fig. 8A; p < 0.05, paired t-test). These assays demonstrate that reducing Pa2G4 in the precursors of the cranial neural crest and otic placode results in fewer proliferative cells, congruent with the smaller size of mouse mutants (Zhang et al., 2008), but this effect is not due to increased apoptosis in Xenopus embryos.

Fig. 8.

Pa2G4 levels affect cell death and cell proliferation. (A) Knock-down of endogenous Pa2G4 by MO injection significantly reduces the number of apoptotic cells in the neural crest/PPE domain, as indicated by the TUNEL assay. Loss of Pa2G4 also significantly reduces the number of mitotic cells, as indicated by phosphorylated H3 (PH3) immunostaining. Embryos evaluated at stage 16. (B) Increasing levels of Pa2G4 by mRNA injection (100 pg, 200 pg, or 400 pg) has no significant effect on proliferation in the neural crest/PPE domain when embryos are evaluated at stage 18. Ctrl, control side; KD, knock-down side; inj, mRNA-injected side. *, p < 0.05, paired t-test.

4. Discussion

The genetic cause of BOS has only been identified in about half of patients. We previously proposed that novel Six1 binding proteins are strong candidates (Moody et al., 2015). We suggested pa2g4 as a potential BOS candidate gene based on its high similarity to a Drosophila Six1/SO-binding partner (CG10576; Giot et al., 2003), and its embryonic expression in tissues that express Six1 in Xenopus, mouse and chick (Gray et al., 2004; Neilson et al., 2010; Figeac et al., 2014) and are affected in BOS. While there is substantial information about the multiple functions of Pa2G4/Ebp1 in cancer cells (Ko et al., 2016), there is a paucity of detail regarding its developmental functions. The Drosophila homologue CG105756 is highly expressed in all cerebral neuroblast populations of the developing brain (Yang et al., 2016). CG10576 over-expression by P-element insertion resulted in embryos with neurogenic-like patches within developing muscle populations (Bidet et al., 2003). Targeted embryonic loss of CG10576 by RNAi resulted in viable and fertile animals (Neely et al., 2010), but no specific developmental phenotypes have yet been reported. Postnatal Pa2g4-null mice show decreased litter sizes, decreased body weight and kinked tails (http://www.informatics.jax.org/allele/genoview/MGI:4353886?counter=1#limbs_digits_tail_id). siRNA-mediated knock-down of Pa2g4/Ebp1 in mouse myoblast cell lines and in chick embryo myoblasts prevents muscle differentiation (Figeac et al., 2014). In this report, we provide novel evidence that Pa2G4 has a critical role in controlling the sizes of the neural plate, neural crest and cranial placode domains, in part by interacting with Six1 and modulating the interaction between Six1 and Eya1.

4.1. Pa2G4 interacts with numerous proteins to regulate cell proliferation and gene transcription

Structural studies indicate that Pa2G4 protein is homologous to type II methionine aminopeptidases based on its “pita bread” fold structure (Kowalinski et al., 2007; Monie et al., 2007). However, it lacks enzymatic activity, and instead this fold is thought to participate in protein interactions important for the regulation of cell proliferation and differentiation. The C-terminal region of Pa2G4 contains other distinct motifs that are reported to be involved in RNA and protein binding in numerous cell lines (Yoo et al., 2000; Xia et al., 2001b; Squatrito et al., 2004; Kowalinski et al., 2007; Monie et al., 2007; Zhang et al., 2008; Zhou et al., 2010; Ko et al., 2016). However, there is no indication from the protein structure that Pa2G4 interacts directly with DNA.

The ability of Pa2G4 to influence cell proliferation appears to depend upon isoform expression. Mammalian Pa2g4 genes are alternatively spliced to produce p42 and p48 proteins (Liu et al., 2006; Ko et al., 2016). The p48 isoform is more abundantly expressed, can localize to either the nucleus or the cytoplasm, and inhibits apoptosis when over-expressed in cell lines. In contrast, the shorter isoform is predominantly cytoplasmic and over-expression inhibits cell proliferation, leading to the suggestion that it acts as a tumor suppressor (Ko et al., 2016). Consistent with these results, knock-down of Pa2G4/Ebp1 in mouse satellite cells and myoblasts reduces proliferation, and in chick dermatomyoblasts prevents differentiation (Figeac et al., 2014). We found in intact Xenopus embryos that additional full-length Pa2G4 (e.g., equivalent to p48) has no effect on proliferation, whereas its reduction results in reduced proliferation and reduced apoptosis. These cellular studies are supported by the observation that Pa2g4-null mice are smaller than wild type littermates (Zhang et al., 2008). Since isoform specificity may be relevant to the phenotype differences observed in this report, and the Xenopus mRNA contains the same alternative translational start sites as mouse and human, it will be important to determine if the two isoforms are differentially expressed and/or altered by binding partners during development of the relevant craniofacial tissues.

There are several lines of evidence that Pa2G4 can modulate transcriptional activity in different types of cancer cells (Ko et al., 2016). The p42 isoform translocates to the nucleus in response to growth factor stimulation, and the p48 isoform is located in both the nucleus and cytoplasm. p42 can complex with the tumor suppressor RB protein to repress transcription (Xia et al., 2001a; Zhang et al., 2002), and p48 is part of a complex, which includes Sin3A, Rb, and HDAC2, that functions to repress transcription in cancer cell lines (Zhang et al., 2002, 2005a, 2005b; Zhang and Hamburger, 2004). Consistent with these observations, we found that the p48 form of Pa2G4 can repress the expression of two different reporters in HEK 293T cells. In contrast, the p48 isoform activates the same reporters in XTC-2 cells. This indicates that the transcriptional effect of Pa2G4 is dependent upon the cellular context, which is not surprising considering the number of proteins and RNAs with which it can interact. While it is clear the HEK 293T cells express endogenous Six1, the published work on XTC-2 protein expression only reports that these cells express several members of the TGFβ family (Smith and Tata, 1991). It is likely that differences in endogenous transcription factors between the cell lines will explain the differences in reporter activity. This is consistent with the observation that Pa2G4 affects embryo gene expression in the neural plate and epidermis differently depending on the presence or absence of Six1. While more detailed biochemical assays are needed to determine the constituents of the Pa2G4 complexes that result in the activation and repression of Six1 reporters in different cell types, our study is the first to indicate that Pa2G4 can modulate the expression of genes relevant to craniofacial development. Since there is no evidence to suggest that Pa2G4 directly binds to DNA, we must surmise that it interacts with other transcription factors as well as Six1.

Our Co-IP assays clearly demonstrate that Pa2G4 can disrupt the interaction between Six1 and Eya1. Therefore, it was surprising that adding Pa2G4 to HEK 293T cells co-transfected with Six1+Eya1 did not significantly disrupt reporter activation. However, similar results are reported for the ability of Gro to repress Six-Eya mediated transcriptional activation. In both Drosophila S2 cells and monkey COS7 cells, Gro can repress SO/Six1 reporter activity, but in the presence of Eya/Eya1 it has a small to undetected effect (Silver et al., 2003; Bricaud and Collazo, 2011). These authors suggest that once Six1 and Eya1 form a complex on the DNA, Gro is not able to interfere. It will be important to determine the exact complex that Pa2G4 forms with these proteins to fully understand these results and their implications on craniofacial development.

4.2. A role for Pa2G4 in craniofacial development

Previous studies in fly, mouse and chick embryos and myogenic cell lines demonstrate that Pa2G4 plays an important role in muscle differentiation. In fly, over-expression disrupts muscle progenitors, leading to neurogenic-like states (Neely et al., 2010). In mouse and chick, Pa2G4 regulates muscle progenitor proliferation and differentiation independent of the ErbB3 signaling pathway (Figeac et al., 2014). Interestingly, Six1, which also is expressed during myogenesis (Grifone et al., 2005; Yajima et al., 2010), is increased upon Pa2G4 knock down in a mouse myoblast cell line (Figeac et al., 2014). In contrast, we observed a decrease in Six1 expression in the PPE in Pa2G4 morphant embryos, and a dose-dependent increase when Pa2G4 was over-expressed specifically in the PPE. Thus, its role in myogenesis is different from its role in craniofacial development. This may not be surprising considering the number of different types of proteins and RNAs with which Pa2G4 can interact in different cellular contexts.

The results we present demonstrate a unique and specific role for Pa2G4 in craniofacial development. It is required for proper formation of the neural border zone and both of its cranial derivatives: the neural crest and placodes. The expansion of the neural plate into the border zone territory in pa2g4 morphants is not surprising since border zone genes repress neural plate genes (Moody and Saint-Jeannet, 2014; Moody and LaMantia, 2015). Release from this territorial repression can result in neural plate expansion. Complimentary to the loss of Pa2G4, gain of Pa2G4 results in a larger border zone and enlarged neural crest domains. However, the PPE domain is either broader or reduced, depending upon the level of Pa2G4. This likely occurs because neural crest genes repress PPE genes (Bugmann et al., 2004; Moody and LaMantia, 2015). Thus, the levels and precise domains of over-expression may result in the variable placode phenotypes. It was surprising that gain of Pa2G4 also results in an enlargement of the neural plate domains. This might result from the maintenance of foxd4L1 expression and the ectopic expression foxd3 and zic2, each of which would prolong the period during which the neural ectoderm is in a proliferative stem-like state (Yan et al., 2009; Buitrago-Delgado et al., 2015; Gaur et al., 2016). The effects of Pa2G4 expression in the neural ectoderm and in the neural crest domain are most likely due to interactions with proteins other than Six1, which is barely detectable in these domains. This conclusion is supported by the observation that the neural plate and neural crest phenotypes can be reversed by Six1 co-expression, presumably by competition for binding. One possible candidate pathway is via the activated ErbB3 receptor, which is expressed in neural crest and to which Pa2G4 binds in cultured cells. Both this competition and the mutual repression between neural crest genes and PPE genes (Brugmann et al., 2004) would be lost in Six1 morphants, resulting in the expansion of msx1 and foxd3 expression into both the neural plate and the PPE/ectoderm.

Loss of Pa2G4 ultimately results in perturbed gene expression in embryonic structures derived from two precursor populations: the neural crest-derived branchial arches and the otocyst. Since the adult structures derived from these tissues are affected in BOS, and Pa2G4 binds to and affects the activity of the two other genes known to be mutated in BOS, we suggest that it is a reasonable candidate for this syndrome. However, considering all of the various activities of Pa2G4 that have been revealed in cancer and myogenic cells, it will be very important to demonstrate whether the developmental phenotypes that we observed in the intact embryo result from the ability of Pa2G4 to affect gene transcription, or from one of its many cytoplasmic functions. In cancer cells, these include binding to nucleolar and cytoplasmic RNAs, signaling pathway components, as well as ubiquitination pathway components (Ko et al., 2016). We predict that if Pa2G4 is mutated in BOS patients, it will harbor amino acid changes that alter binding efficiencies, as seen in the majority of SIX1 and EYA1 mutations. Our next challenge will be to identify the interaction between Pa2G4, Six1, Eya1 and potentially other interacting proteins in developing craniofacial tissues, particularly the cranial neural crest and otic placode.