Department of Oral Health Sciences, Life Sciences Institute, University of British Columbia, Vancouver, Canada
Institute of Animal Physiology and Genetics, Academy of Sciences of the Czech Republic and Department of Anatomy, Histology and Embryology, University of Veterinary and Pharmaceutical Sciences, Brno, Czech Republic
Facial morphogenesis is a result of orchestrated signals originating from the endoderm, ectoderm, and brain, and affecting the patterning of neural crest–derived mesenchyme. Each component of the embryonic face, frontonasal mass, lateral nasal, maxillary and mandibular prominences is dependent to a greater or lesser extent on certain classes of signals. The frontonasal mass gives rise to the facial midline skeleton and its development is partially dependent on retinoids produced by the brain (Niederreither et al.,1999; Schneider et al.,2001; Halilagic et al.,2007). The actions of sonic hedgehog produced by the foregut endoderm are an important inductive signal for the frontonasal mass (Benouaiche et al.,2008) and for midline specification (Cordero et al.,2004; Marcucio et al.,2005). Blocking FGF receptor signaling in the forebrain impedes growth of the frontonasal mass (Hu and Marcucio,2009). Antagonizing bone morphogenetic protein (BMP) signaling with Noggin also leads to decreased proliferation and ultimately to inhibition of skeletal formation (Foppiano et al.,2007; Hu et al.,2008). Therefore, at least four different signals impact development of the upper face.
The frontonasal mass not only is the major contributor to the midline of the face but also has a major role in lip fusion. Fusion of the primary palate or lip occurs when the frontonasal mass corners (globular processes) meet the cranial-medial edges of the maxillary prominences to form a bilayered epithelial seam. Following breakdown of the epithelium (Sun et al.,2000), a mesenchymal bridge forms, which then proliferates to form a continuous upper lip (Wang et al.,1995; Jiang et al.,2006). Since non-syndromic cleft lip with or without cleft palate (CL/P) is the most common craniofacial birth defect (Jugessur and Murray,2005), much effort has been focused on identifying some of the predisposing genetic alterations (Vieira et al.,2005,2008). In addition, due to the complex interaction of genes and environment, it is necessary to use a variety of approaches and model organisms to understand the basis of this malformation. We have previously shown that interference with either BMP (Ashique et al.,2002a) or FGF signaling (Szabo-Rogers et al.,2008) in the primary palate reduces outgrowth of the prominences resulting in cleft lip. Mutations in ligands or receptors in the FGF and BMP pathways have been shown to cause cleft lip in humans (Riley et al.,2007; Suzuki et al.,2009). In addition, it is unlikely that decreased gene function can explain all of CL/P. We need to consider also that ectopic activation of genes by environmental factors or mutations in repressors or antagonists may lead to a net gain-of-function accompanied by abnormal morphogenesis (Gritli-Linde,2008). The chicken model is an excellent organism in which to simulate a gain-of-function mutation in the facial prominences.
Not only are secreted signals important in orchestrating the steps involved in lip fusion but so are the downstream transcription factors. Some of the more characterized genes in the chicken embryo include AP2α, MSX1, MSX2, DLX5, TBX2, TBX3 (Brown et al.,1993; Shen et al.,1997; Gibson-Brown et al.,1998; Barlow et al.,1999; Ashique et al.,2002a; Szabo-Rogers et al.,2008,2009). Our recent expression profile study has also identified many genes differentially expressed in the frontonasal mass and one of the most abundant was the T-box transcription factor, TBX22 (Buchtová et al.,2009). In situ hybridization studies in amniotes have documented expression of TBX22 in the frontonasal mass and maxillary prominences of chickens (Haenig et al.,2002; Handrigan et al.,2007), mice (Braybrook et al.,2002; Bush et al.,2002; Herr et al.,2003), and humans (Braybrook et al.,2002). The high degree of evolutionary conservation was recently confirmed in zebrafish, where expression is also restricted to the perioral mesenchyme (Jezewski et al.,2009). Interestingly, despite the expression of TBX22 in the frontonasal mass, mutations in the human gene primarily affect fusion of the maxillary-derived secondary palate (CPX, OMIM 303400; Braybrook et al.,2001; Marcano et al.,2004; Stanier and Moore,2004; Suphapeetiporn et al.,2007). Recently, a null mutation of Tbx22 was reported in the mouse and the main phenotypes included submucous cleft palates and posteriorized tongues (Pauws et al.,2009). A minority of null embryos had overt cleft palate. These mouse and human studies do not reveal functions of Tbx22 in the frontonasal mass possibly due to functional redundancy with other T-box genes also expressed in this region (Chapman et al.,1996; Kraus et al.,2001; Singh et al.,2005; Farin et al.,2007; Zirzow et al.,2009). In addition, Tbx22 has a rather late onset of expression in mouse, first appearing between E9.5 (Farin et al.,2008) and E10.5 (Bush et al.,2002). Thus, the main aspects of neural crest and frontonasal mass patterning would have taken place prior to the onset of Tbx22 expression. It is possible that precocious or ectopic expression of Tbx22 might affect development of the face and even lead to cleft lip as has been shown in the Dancer mice that have a gain in function of Tbx10 (Bush et al.,2003).
TBX22 is grouped within the TBX1 subfamily (Naiche et al.,2005) and is most closely related to TBX15 and TBX18. While some T-box genes act mainly as transcriptional activators (e.g., TBX4, 5), the TBX15, 18, and 22 subfamily are all repressors (Andreou et al.,2007; Farin et al.,2007). The regulation of TBX22 is not well characterized, although it has recently been described that the transcription factor meningioma 1 (Mn1) regulates Tbx22 expression in mouse (Liu et al.,2008). The signaling pathways controlling Tbx22 expression have not yet been studied.
In this study, we show for the first time that FGFs and BMPs regulate TBX22 expression in the frontonasal mass, whereas we exclude roles for SHH. In order to investigate whether TBX22 was capable of mediating the effects of these growth factors on lip fusion, we performed gain-of-function experiments using viral misexpression. Increased levels of TBX22 caused a failure of outgrowth due to decreased proliferation of mesenchyme. These early defects led to clefts affecting the primary palate or lip. In addition, we found that TBX22 selectively inhibited intramembranous bone differentiation. Furthermore, we identified two downstream targets of TBX22, DLX5 and MSX2, the loss of which could have contributed to the clefts in the primary palate and inhibition of osteogenesis. Finally, we investigate the role of the endogenous gTBX22 gene in mediating the FGF-induced and viral-induced phenotypes.
FGFs Are Required for the Induction of TBX22 in the Frontonasal Mass
Previous chicken studies (Haenig et al.,2002) had shown there was no TBX22 expression in migratory neural crest cells. Thus, de novo expression occurred sometime after the cessation of neural crest cell migration. A similar result has recently been reported in zebrafish, where expression around the mouth is not detected until relatively late in organogenesis (Jezewski et al.,2009). TBX22 is a mesenchymally restricted transcription factor and the upregulation in the facial mesenchyme is presumably dependent on local interactions with epithelium or perhaps with the forebrain in the case of the frontonasal mass (Marcucio et al.,2005; Hu and Marcucio,2009). We examined expression of TBX22 in more detail and found that no expression could be seen in the face prior to stage 15 (25 body somites or less), but at 26 somite pairs, a slight signal medial to the nasal pits was detected (Fig. 1A, A1). The intensity and area increased up to 29 somites (Fig. 1B-C1) and was high until approximately stage 22 when expression began to decrease in the frontonasal mass (data not shown). Other areas of the face acquired expression after stage 20 including the lateral nasal prominences, medial edge of the maxillary prominences, and lateral, cranial mesenchyme of the mandibular prominence (Haenig et al.,2002; Handrigan et al.,2007). While a different member of the TBX family, TBX2, is induced by FGF8 (Firnberg and Neubuser,2002), the regulation of TBX22 by secreted signals has not been studied previously.
There is very strong expression of FGF8 across the frontonasal ectoderm from stage 15 onwards (Hu et al.,2003; Hu and Marcucio,2009; Szabo-Rogers et al.,2009), preceding the onset of TBX22 expression by about 12 hr. There is also FGF8 expressed in the rostral/anterior telencephalon (Creuzet et al.,2004; Hu and Marcucio,2009). Thus, the thin layer of frontonasal mass mesenchyme is sandwiched between two FGF8-expressing tissues. To test whether FGF receptor signaling from either the forebrain or epithelium was required for induction of TBX22 expression, we injected the antagonist, SU5402 (Mohammadi et al.,1997), into the frontonasal mass at stage 15 when TBX22 is not yet expressed. TBX22 failed to upregulate on the injected side in these embryos (n = 6/7, Fig. 1E, Table 1) whereas normal induction was observed in DMSO-injected embryos (Fig. 1D, Table 1). To test whether epithelial-derived FGF was sufficient to maintain expression, we implanted FGF2 beads at stage 20 and examined TBX22 expression at stage 24. We found that FGF2 could sustain expression of TBX22 beyond stage 20 (Fig. 1F, Table 1). The SU5402 and FGF2 bead data suggest that FGFs are one class of endogenous signal responsible for regulating TBX22 during frontonasal mass growth.
Table 1. Regulation of TBX22 by Altering Secreted Signals in the Frontonasal Mass
Time point (hr); n
FGF2 1 mg/ml
6; n = 7
16; n = 5
16; n = 4
SU5402 1 mg/ml
6; n = 6
BMP4 0.1 mg/ml
6; n = 5
16; n = 5
BMP7 0.1 mg/ml
6; n = 5
16; n = 4
Noggin 0.65 mg/ml
6; n = 6
16; n = 5
SHH 5 mg/ml
6; n = 10
16; n = 10
SU5402 1 mg/ml
16; n = 7
Next, we addressed whether FGF signaling can regulate the expression of TBX22 during stages when lip fusion is taking place. FGF8, 9, 18 are expressed in the epithelium lining the nasal slit in the stage-24 frontonasal mass (McGonnell et al.,1998; Karabagli et al.,2002; Havens et al.,2006; Szabo-Rogers et al.,2008) whereas FGF2 is expressed more generally throughout the epithelium and mesenchyme (Richman et al.,1997). We have shown that certain regions of the frontonasal mass are dependent on FGF signaling for normal outgrowth and lip fusion (Szabo-Rogers et al.,2008) and that FGFs can replace epithelium and promote outgrowth of the frontonasal mass skeleton (Richman et al.,1997). There is also evidence from human genetic studies that mutations in FGF receptors cause cleft lip (Dode et al.,2003; Riley et al.,2007). We found that SU5402 reduced TBX22 expression in the globular process and caudal edge of the frontonasal mass (Fig. 1H, Table 1; control beads had no effect, Fig. 1G, n = 3). We next tested whether FGF-soaked beads would expand the expression of TBX22 and the temporal characteristics of this induction. Either FGF2 or FGF8 beads upregulated TBX22 in as little as 6 hr (Fig. 1J, K, Supp. Fig. S1A, B, which is available online; Table 1) whereas control beads did not affect expression (Fig. 1I). Therefore, FGFs are positive regulators of this transcription factor. Our studies demonstrate that FGFs are both required and sufficient to maintain TBX22 expression in the frontonasal mass. The rapid induction of TBX22 by FGF2 could also mean that TBX22 was mediating the FGF2 stimulation of proliferation described in our other studies (Szabo-Rogers et al.,2008).
Regulation of TBX22 by BMPs
Several BMPs (BMP2, BMP4, BMP7) and BMP antagonists (NOGGIN, BAMBI) are expressed in the epithelium and mesenchyme of the frontonasal mass, in some cases overlapping with TBX22 and other times in complementary domains (Francis-West et al.,1994; Ashique et al.,2002a; Foppiano et al.,2007; Higashihori et al.,2008). From these data, it is not possible to predict whether increased or decreased levels of BMP signaling affect TBX22 expression. To test whether BMPs positively or negatively regulated TBX22, we implanted BMP- or Noggin-soaked beads into the corner of the frontonasal mass at stage 24. BMP4 and BMP7 recognize different BMP receptors (ALK3/6 and ALK2, respectively; Onichtchouk et al.,1999). We, therefore, tested both ligands to see which was more likely to act as an endogenous regulator of TBX22. We found both treated embryos showed decreased expression at both 6- and 16-hr treatment (Fig. 2A–D, Table 1). These results indicate BMP signaling via both receptors could repress TBX22 expression. In contrast, blocking endogenous BMP signaling with Noggin strongly induced TBX22 after both 6 and 16 hr (Table 1; Fig. 2E, F, S1C, D).
We also considered other classes of epithelial signals that might regulate TBX22 expression. Extensive mapping of WNT ligands by our group (Geetha-Loganathan et al.,2009) did not reveal any obvious temporal and spatial correlations between TBX22 expression and WNT expression. Therefore, these were not examined. SHH expression in the stomodeal epithelium (Hu et al.,2003) abuts TBX22 mesenchymal expression along the caudal edge of the frontonasal mass. Moreover, in our previous work we had noted that Noggin beads led to an increase not only in FGF8 but also in SHH expression (Ashique et al.,2002a). Since we have already shown that FGF8 is capable of inducing TBX22, we also tested whether SHH beads would induce TBX22. Only a minority of specimens have very slightly decreased expression (n = 5/20, data not shown; Table 1). Therefore, SHH is not an upstream regulator of TBX22 under the conditions of these experiments and the most likely endogenous signaling pathways that regulate TBX22 are FGFs and BMPs.
The Effect of RCAS::hTBX22 Is to Reduce Proliferation and Outgrowth of the Frontonasal Mass
Interestingly, Noggin treatment reduces proliferation (Ashique et al.,2002a; Foppiano et al.,2007) whereas FGF2 dramatically increases proliferation (Szabo-Rogers et al.,2008) and yet both proteins induce TBX22 expression. In order to determine whether TBX22 has any effects on cell proliferation, we overexpressed the gene in the frontonasal mass and then labeled embryos with BrdU prior to fixation. The targeting of the virus to the right side of the frontonasal mass was first confirmed with the 3C2 antibody (Fig. 3A, B). In RCAS::hTBX22-infected embryos, there was a significant decrease in proliferation in the region of high viral expression (Fig. 3A1, A2, B1, B2). The percentage of BrdU-positive cells in hTBX22-injected samples was significantly lower (n = 3, 27.9%, P = 0.0046, SD ± 4.1%) than in the GFP controls (n = 3, 41.4%, SD ± 2.81%). We also examined PCNA RNA expression to gain a three-dimensional view of TBX22 effects on proliferation. PCNA is expressed in a V-shape in the frontonasal mass corresponding to areas of increased BrdU labeling in sections (Fig. 3A1, B1, C). Several regions with low BrdU labeling also have lower expression of PCNA including the globular process and anterior medial edge of the maxillary prominences (Fig. 3C). We acknowledge, however, that PCNA expression was much less abundant in the frontonasal mass than elsewhere in the face. A shorter detection period would likely have revealed more spatial differences in expression in the maxillary and mandibular prominences. With these caveats in mind, PCNA is a good representation of proliferation in the frontonasal mass. The effect of TBX22 virus was to decrease PCNA expression mainly in the cranial and lateral edge of the treated frontonasal mass (Fig. 3C). This is consistent with the sections stained with BrdU antibody.
While analyzing the effects on cellular dynamics, we noted there was a qualitative reduction in outgrowth of the frontonasal mass. To quantify the extent of this decrease, we measured the area of the lateral third of the frontonasal mass on the treated and control sides (Fig. 3D). In normal embryos, the ratio between the two sides is 1:1 indicating a high degree of symmetry. In treated embryos there was a 12% reduction in the size of the right side of the frontonasal mass (Fig. 3E, P < 0.001). The size deficits could be attributed to an increase in apoptosis. Therefore, we performed TUNEL reaction on serial sections of the BrdU, virus-injected embryos. It was clear that there was no increase in the number of apoptotic cells in treated versus control virus-infected embryos (Fig. 3A3, B3).
Another possibility for the loss of tissue in the frontonasal mass could be changes in cell expansion. Small groups of cells labeled with DiI expand in a characteristic manner in different regions of the face due to a combination of cell proliferation and cell movement. Moreover, the shape of the expansion is altered in treatments that truncate frontonasal mass outgrowth (McGonnell et al.,1998). We, therefore, injected DiI into TBX22 or GFP viral-infected embryos and did not detect any difference in the shape of the DiI-labeled cell group (n = 6, data not shown). Therefore, the reduction of the frontonasal mass in the RCAS::hTBX22 viral misexpression model is mainly caused by a decrease in cell proliferation in the cranial and lateral frontonasal mass. These data also suggest that the main role of TBX22 during normal development is to inhibit rather than to stimulate proliferation. In bead implant experiments, it is more likely that TBX22 mediates the effects of Noggin protein rather than those of FGF.
Viral Misexpression of TBX22 in the Frontonasal Mass Induces Cleft Lip
We hypothesized that overexpression of TBX22 would cause cleft lip in our chicken model, based on the early growth defects we had characterized. Control embryos injected with GFP expressing virus at stage 15 developed normally and showed high expression in the frontonasal mass-derived midline elements (Fig. 4A, A1, and inset, Table 2). In contrast, embryos injected with RCAS:hTBX22 developed clefts on the treated side of the beak (82% skeletal clefts and 64% visible externally, Fig. 4B–C3, Table 2). Furthermore in severely affected embryos, the maxillary bone was displaced proximally and the maxillary process of the palatine bone was noticeably shorter (Fig. 4C2, C3). These defects resemble closely those produced by SU5402 (Szabo-Rogers et al.,2008), which is also caused by decreased proliferation. Other intramembranous bones, such as the jugal bone and maxillary process of premaxillary bone, were also reduced (Fig. 4B1, C3, Table 2). This suggests that TBX22 could also be affecting intramembranous ossification.
Table 2. Effects of RCAS::hTBX22 on Morphology of the Beak
Viral suspension injected posterior to the optic stalk.
DF1 cells injected into the maxillary region.
Primary palate cleft
Soft tissue cleft or notch
Gap between premaxillary and maxillary bones
Effects on frontonasal mass derivatives
Maxillary process of premaxilla reduced
Effects on maxillary prominence derivatives
Jugal % missing or reduced
Maxillary bone – displaced or reduced
Palatine bone – reduced nasal process or maxillary process
Secondary palate ‘cleft’
Palatine bone – increased gap between bones
Effects on first arch derivatives
Mandibular bones – missing or reduced
Entoglossum – missing or reduced
In order to determine if there were stage-specific effects of RCAS::hTBX22 misexpression, we treated another set of embryos at stage 10. This earlier manipulation leads to widespread infection of the upper and lower beaks, including bilateral spread in some embryos (3/23 had bilateral clefts). Once again, we observed cleft lip (Fig. 5B, B1, B3, B4). However, the frequency was lower than in stage-15-injected embryos (44 vs. 82% with skeletal clefts; Table 2). Embryos injected at stage 10 displayed additional defects in the skeleton compared to those injected at stage 15 (Fig. 5A–A2, B2–B5). The intramembranous bones derived from the maxilla, frontonasal mass, and mandibular prominence were reduced in size (Table 2; Fig. 5A1, A2, B2–5). In addition, the entoglossum or cartilaginous tongue skeleton was reduced (Fig. 5B2). Interestingly, RCAS::hTBX22 does not inhibit differentiation and outgrowth of the chondrocranium (i.e., prenasal cartilage, interorbital septum, Meckel's cartilage). Thus, the tongue cartilage, which is derived from the mesencephalic neural crest (Couly et al.,1996), seems to be particularly susceptible to increased levels of TBX22, compared to other first arch or frontonasal derived cartilages.
Since RCAS::hTBX22 misexpression in the frontonasal mass affected lip fusion, we hypothesized that the some of the genes expressed in the globular process might be targets of TBX22. We considered DLX5 and MSX2 as prime candidates, first because they are restricted to the globular process and second because upon closer examination, they are expressed in a complementary pattern to TBX22 (Fig. 6A–C). The mutually exclusive expression patterns of TBX22 in relation to MSX2 and DLX5 was consistent with TBX22 acting upstream to repress their expression. Control embryos injected with DF1 cells expressing GFP virus had no changes in expression (Fig. 7A, A1, C, C1; n = 10). However, DF1 cells expressing hTBX22 reduced expression of both DLX5 and MSX2 (DLX5, Fig. 7B, B1, n = 5/6; MSX2, Fig. 7D, D1, n = 7/8). Therefore, during normal development, TBX22 may act as a repressor to restrict the expression of DLX5 and MSX2. In addition, the phenotypes caused by RCAS::hTBX22 may be mediated by decreased DLX5 and MSX2 expression. In particular, since RCAS::TBX22 decreases both proliferation and MSX2 expression, two effects elicited by Noggin beads, it is possible that TBX22 is mediating the effects of Noggin on facial morphogenesis. In contrast, the TBX22 virus has opposite effects to those of FGF2 on proliferation and MSX2 expression. Therefore, it seems unlikely that TBX22 is mediating the effects of FGF2 beads.
FGF Is Partly Able to Rescue the Phenotype Induced by RCAS::hTBX22
The induction of TBX22 by FGF was unexpected, especially in view of the dramatic induction of proliferation by FGF reported previously by our group (Szabo-Rogers et al.,2008) and the antiproliferative effects of exogenous hTBX22 shown here. Was the induction of gTBX22 by FGF related to the proliferation increase or was the induction of gTBX22 serving a different function, unrelated to proliferation. To distinguish between these possibilities, we designed a dual treatment where embryos were first infected with RCAS::TBX22 at stage 15 and then implanted with FGF2 beads at 48 hr when expression of the virus is high. Following these treatments, we hybridized embryos to either gTBX22 or PCNA.
The first question we addressed was whether FGF2 could still induce gTBX22 even in the presence of RCAS::hTBX22. Under the stringent conditions used, the probe we generated against gTBX22 does not hybridize to the human viral transcript. We found that instead of the upregulation observed with FGF2 beads alone, there was clear downregulation of gTBX22 (Fig. 8A, A1, n = 11/11). This suggested that not only was the human gene interfering with the induction of gTBX22 but the virus was also blocking the normal expression of the gene. We confirmed this result in specimens treated only with the retrovirus (Fig. 8B, B1; n = 9/9). Thus, under the dual treatment conditions, there was a net knock-down of gTBX22 gene expression.
This knockdown of gTBX22 could be underlying the phenotypes of RCAS::hTBX22, but additional experiments were required to determine whether this was true. We, therefore, examined the ability of FGF2 to rescue proliferation and morphogenesis in the dual-treated embryos. If rescue was not elicited, this would support the view that repression of gTBX22 was essential to the phenotypes. However, if measurable rescue was observed, then it would show that low levels of gTBX22 did not impede outgrowth. This second scenario would also diminish the importance of gTBX22 levels in generating the outgrowth phenotypes caused by RCAS::hTBX22.
First we showed that our marker of cell proliferation, PCNA expression, could be induced by FGF2 in the frontonasal mass (Fig. 8C, n = 7/7). In contrast, the findings from the doubly manipulated embryos (virus plus FGF2 bead) were striking. Even though very high levels of viral-derived hTBX22 were present in the mesenchyme and, therefore, low levels of gTBX22, FGF2 beads were able to rescue and expand the expression of PCNA (Fig. 8D, n = 6/8). Finally, to quantify the effects of FGF2 on the viral-infected frontonasal mass, we measured the treated and non-treated contralateral sides of the frontonasal mass as before (Fig. 3D). An FGF2 bead on its own can greatly increase the size of the frontonasal mass (Fig. 8C; Szabo-Rogers et al.,2008). Analysis with ANOVA and Tukey's Post-hoc testing revealed that in the RCAS::hTBX22-infected embryos (n = 15), the treated side was significantly smaller than non-treated side compared to normal (n = 11, P < 0.001) and FGF2-treated embryos (n = 7; P < 0.001, Fig. 8E). The FGF2-treated embryos were on average slightly larger than non-treated embryos (Fig. 8E; P < 0.05, with ANOVA). The virus plus FGF2 bead-treated embryos were indistinguishable from the non-treated controls (n = 8, no significant difference), showing rescue of frontonasal morphogenesis. We conclude, therefore, that at least in the short term, an FGF2 bead is capable of overcoming the proliferation and growth deficits induced by hTBX22 and these effects are not dependent on induction of gTBX22 nor are they affected by the presence of exogenous hTBX22. We conclude that FGF2 can work independently of TBX22 to stimulate proliferation and that in non-viral infected embryos the induction of TBX22 by FGF serves a different function.
We showed for the first time in a series of bead implant experiments that TBX22 is downstream of both the FGF and BMP pathways. The evidence for TBX22 mediating the effects of BMPs is strongest and is based on the similarity in tissue and gene responses to exogenous Noggin or TBX22. In contrast, FGFs may be important for the initial expression of TBX22 in the frontonasal mass but TBX22 is not a likely mediator of mitogenic effects of FGF. Instead, the main function of TBX22 as shown in misexpression experiments is to decrease proliferation. We subsequently uncovered three targets of TBX22, DLX5, MSX2, and TBX22 itself. All are downregulated in the presence of viral-derived hTBX22. In order to determine whether auto-downregulation of gTBX22 was mediating the effects of RCAS::hTBX22, we implanted FGF-soaked beads into viral-infected embryos. From this experiment, we were able to conclude that the loss of the gTBX22 gene caused by the virus is not the main cause of the cleft lip phenotype and that the positive effects of FGF on gTBX22 expression are related to a function other than proliferation.
Phenotypes Are Caused by Overexpression of hTBX22 and Not By the Repression of the Endogenous Gene
We designed our study using the human gene in order to observe the effects of exogenous viral hTBX22 on the endogenous gene and, consequently, we were able to show clearly the downregulation of gTBX22. Furthermore, with the targeted viral injections, especially those into the frontonasal mass, we were able to determine the effects on other facially expressed TBX genes. There are three possible explanations to account for the phenotypes: (1) Either the human gene is able to function just like the chicken gene and expression in ectopic places is the basis for the phenotypes; (2) it is loss of the chicken gene that causes the phenotypes; or (3) it is a combination of both mechanisms.
There are two pieces of data that are consistent with the ectopic expression of hTBX22 being the main cause of the phenotypes. Ectopic expression of hTBX22 is consistently created in the stage-15 injections into the frontonasal mass. At the time the virus is first expressed, gTBX22 has already downregulated the majority of the frontonasal mass so autoregulation is not a major factor. We determined with careful BrdU analysis that the virus decreases proliferation in the cranial, lateral edge of the frontonasal mass, overlapping precisely with the major area of viral infection but low gTBX22 expression. In contrast, the globular process, where gTBX22 is repressed does not exhibit measurable effects on BrdU labeling. Thus, the proliferation phenotype overlaps with the region of greatest expression of RCAS::hTBX22 and not the location where autoregulation occurs.
The second experiment that refutes the idea that decreased gTBX22 is responsible for the phenotype is the dual treatment with virus and FGF2 beads. The data convincingly show that FGF2 can rescue the phenotype even though there is no reinduction of gTBX22. Therefore, the RCAS::hTBX22 results in stable and effective repression of the gTBX22 gene but this repression has no impact on the function of FGF. The observation that RCAS::hTBX22 can duplicate in the chicken, the same type of autoregulation as seen between hTBX22 and the human TBX22 promoter (Andreou et al.,2007), suggests that human TBX22 protein is functionally similar to the chicken protein. We acknowledge that the specificity of hTBX22 protein for chicken T-box response elements needs to be confirmed biochemically.
An additional reason that we feel misexpression is at the root of the phenotypes described here, is that the features of CPX syndrome are distinct from those produced in our chicken experiments. Although the tongue is abnormal in our study, the defects are not homologous to those observed in human or mouse Tbx22 loss-of-function. In human and mouse, the mesodermally-derived muscle attachments to the floor of the mouth are affected, whereas in chicken the neural crest–derived skeleton (Couly et al.,1996) is deleted. Finally, as we have mentioned earlier, the lip is not affected in CPX syndrome while it is susceptible to ectopic expression of hTBX22 in our study.
Thus, the arguments in favor of a role for knockdown of gTBX22 in generating the chicken phenotypes are not as strong at the present as those against. However, since we do not have a true loss-of-function experiment in the chicken embryo, we must leave open the possibility that some aspect of the phenotype is caused by the decrease in gTBX22.
The Roles of TBX22 in Regulating Cell Proliferation and Apoptosis
Our retroviral studies suggest that TBX22 may negatively regulate proliferation during normal facial development in a stage- and spatially-specific manner. In the immediate post-migratory neural crest stage when facial mesenchyme is interacting with the epithelium, neuroepithelium, and endoderm, there is initially no expression of TBX22. Subsequently, there is upregulation between stage 15 and 18 that is dependent on FGF signaling. Based on the negative effects of the virus, this early expression of TBX22 is unlikely to promote cell proliferation but instead could fill other, as yet undetermined, functions. Between stage 20 and 22, the coordinated loss of the mitogenic signal FGF8 and the antiproliferative transcription factor TBX22 may also help to maintain an even level of proliferation. Indeed, proliferation for the lateral frontonasal mass has been measured and found to be constant until stage 25 (Minkoff and Kuntz,1977). It is the maintenance of proliferation relative to a drop in the basal parts of the facial mesenchyme that results in outgrowth of the facial prominences (Minkoff,1991).
After stage 22, the expression of TBX22 is restricted to the globular processes and caudal edge. Both of these TBX22-enriched areas have distinctively low proliferation (Fig. 9A; this report and Szabo-Rogers et al.,2008). Extending beyond the frontonasal mass, correlations are also good in the maxillary and mandibular prominences. The lateral, cranial mesenchyme of the mandibular prominence has relatively lower proliferation (Barlow and Francis-West,1997; McGonnell et al.,1998) overlapping TBX22 expression. In the maxillary prominence, there is abundant TBX22 expression along the medial edge where the presumptive palatal shelves will form. We have shown partial overlap of TBX22 signal with the lowest proliferating region located at the cranial, medial edge close to the fusion zone. In addition to our data, other very detailed studies from the Minkoff lab report statistically significant lower proliferation on the medial side of the maxillary prominence compared to the lateral side (Bailey et al.,1988). Thus, taken together, the correlation of TBX22 with lower proliferation in the stage 24–26 facial prominences and the negative effects of TBX22 virus on cell proliferation suggest that at these stages TBX22 provides regional control of proliferation.
We have proposed a role for TBX22 in regulating proliferation in the face. Therefore, it would be interesting to know whether there was increased proliferation in the Tbx22 knockout mice. Thus far, no differences in proliferation were reported (Pauws et al.,2009). However, Tbx2, which is expressed in a similar fashion to Tbx22 (Gibson-Brown et al.,1998; Firnberg and Neubuser,2002; Szabo-Rogers et al.,2009) and is a transcriptional repressor (Jacobs et al.,2000; Lingbeek,2002; Harrelson et al.,2004; Naiche et al.,2005; Teng et al.,2007), causes increased proliferation when targeted (Zirzow et al.,2009). This increase in proliferation is stage specific and restricted to the anterior palatal shelves. Thus it remains possible that loss of Tbx22 would cause subtle increases in proliferation if examined in detail.
In contrast to proliferation, we have less evidence for TBX22 playing a role in regulating apoptosis. We base this conclusion in part because there are no temporally regulated increases in apoptosis following the downregulation of FGF8 and TBX22 at stage 22 (McGonnell et al.,1998; Ashique et al.,2002a; Song et al.,2004). We could not detect a decrease in apoptosis in viral injected embryos. However, since the basal level of cell death is low in this region, any differences might have been below the level of detection. It is possible that further experimentation using viral misexpression in areas of naturally high apoptosis such as the mandibular midline might still reveal a role for TBX22 in this pathway.
TBX22 Is Likely to Be a Mediator of BMP Signaling
We have provided compelling evidence that TBX22 should be added to the list of transcriptional targets of the BMP pathway although the precise position of TBX22 in relation to other genes still needs to be determined biochemically (Fig. 9B, D). For example, since MSX2 is a repressor (Catron et al.,1996) and is induced by exogenous BMPs (Barlow and Francis-West,1997; Mina et al.,2002), the down-regulation of TBX22 caused by BMPs or the upregulation induced by Noggin is explained most simply by the model that MSX2, or other similar repressor, acts upstream of TBX22 (Fig. 9B). However, we have also shown that RCAS::hTBX22 can decrease MSX2 expression (Fig. 9D). Thus, we propose that MSX2 could lie both upstream and downstream of TBX22, depending on the context.
In mouse studies, the regulation of Tbx22 by BMPs was examined in a genetic model where Bmpr1a was conditionally deleted. However, no effect on Tbx22 expression was observed in either the primary or secondary palate (Liu et al.,2005b). Since BMP signaling can also operate via the Type Ib (ALK6) and ALK2 receptors and we have shown that stimulating either receptor with BMP2 or BMP7 activates TBX22, we cannot completely rule out a role for compensation by these other BMP pathways in the Bmpr1a-targeted embryos. It would be interesting to see whether there is increased expression of Tbx22 in conditional knockouts of Bmp4 (Liu et al.,2005a). In addition, we have only studied one BMP antagonist and there are many others. Our data suggest that Tbx22 expression may also be affected by genetic manipulations to a variety of BMP antagonists (Gazzerro and Canalis,2006).
Possible Roles for TBX22 in Mediating FGF Functions Other Than Proliferation
Although we have shown that endogenous FGF signaling is important for the induction of TBX22 and maintenance of expression, the effects of FGF on growth are not mediated via TBX22 (Fig. 9C). We base these conclusions, first, on the completely opposite effects on proliferation caused by FGF and overexpression of hTBX22 and, second, on the ability of FGF2 to rescue hTBX22-induced growth defects independent of the levels of exogenous or endogenous TBX22. Nonetheless, we have shown that TBX22 is downstream of FGFs (Fig. 9C). It is possible that TBX22 is mediating other effects of this protein, such as the promotion of cell survival. This idea is supported by the minimal apoptosis observed in RCAS::hTBX22 misexpression embryos in the present study. In addition, we previously showed that the effects of FGF2 beads are to decrease BMP4 expression (Szabo-Rogers et al.,2008), which likely improves cell survival. Similarly, Noggin beads induce FGF8 expression and this is accompanied by increased thickness of the epithelium, a direct result of decreased apoptosis (Ashique et al.,2002a). Thus, the induction of TBX22 by not only FGF but also BMP antagonists could be promoting cell survival.
A second possible function of the induction of gTBX22 is to act as a negative feedback loop (Fig. 9C) to control the dramatic increase in proliferation caused by FGF (Szabo-Rogers et al.,2008). The induction of such negative feedback loops is well documented in both the FGF and BMP pathways in bead experiments (Minowada et al.,1999; Lee et al.,2001; Ashique et al.,2002a; Szabo-Rogers et al.,2008) and in other systems (Pizette and Niswander,1999; Foppiano et al.,2007).
TBX22 Affects Some Aspects of Intramembranous Bone Formation
The major effect of RCAS::hTBX22 other than an anti-proliferation effect of ectopic and sustained hTBX22 expression was on intramembranous bones (Fig. 9D). Frontonasal mass, maxillary and mandibular bones were reduced in size with little effect on the adjacent cartilage elements, even when virus was injected very early in development. The usual effect of altering chondrogenesis is to see large beak deviations due to deletion of cartilage (Ashique et al.,2002b; Havens et al.,2008) and these were rarely observed in the present study. It appears that viral-derived hTBX22 reduces the size of ossification centers, since the bones are generally present but smaller. These observations may be tied to the inhibition of proliferation, which would decrease the number of preosteoblastic cells. The bone effects may be mediated by the two transcription factors that were identified as targets of DLX5 and MSX2. DLX5 has been shown to promote intramembranous bone formation when overexpressed in chicken calvaria (Holleville et al.,2007). Activating mutations in human MSX2 lead to craniosynostosis or bone overgrowth (Jabs et al.,1993), and homozygous null Msx2 mouse embryos have deficits in their calvaria (Satokata et al.,2000). Thus, the loss of both MSX2 and DLX5 could be mediating the effects of hTBX22 on intramembranous bone differentiation. We propose that TBX22 could be a novel transcription factor involved in the expansion of intramembranous bone condensations.
The loss of function of mouse Tbx22 also reduces the size of two membranous bones, the vomer and palatine, with the greatest effects being seen on the vomer (Pauws et al.,2009). In the majority of the skull, however, there are no differences in osteogenesis. This is likely due to the restricted temporal and spatial expression of the gene. Interestingly, the loss of Tbx22 results in increased Msx2 expression, which is complementary to the results in our study. However, the increased Msx2 expression was limited to the mesodermally derived tongue and did not extend to the skeletogenic neural crest–derived mesenchyme. Our data suggest that if Tbx22 levels were to be increased as, for example, in the situation with Tbx10 in the Dancer mouse, there would be more widespread effects on bone formation and these would be accompanied by a decrease in Msx2 as well as Dlx5. Interestingly, the effects of Tbx10 overexpression on Tbx genes or other transcription factors were not reported in the study by Bush et al. (2004).
We have learned from the present study that any genetic or environmental insult where TBX22 is induced could lead to smaller facial prominences and predispose an individual to having cleft lip. In the Dancer mouse mutant, the effect of the mutation is to cause ubiquitous upregulation of Tbx10 throughout the embryo. However, despite generalized expression, the phenotypes are very restricted, affecting only the inner ear and face. Dancer mice (Deol and Lane,1966) along with Twirler (Gong et al.,2000) and A strain mice (Wang et al.,1995) display true cleft lip where the mesenchymal bridge between the medial nasal and maxillary prominences fails to form. Based on our data and the highly penetrant cleft lip observed in Dancer mice, we propose that genetic or environmental insults such as those accompanied by decreased BMP signalling, leading to a gain-of-function of TBX22 could contribute to human clefting. Furthermore, our work highlights that a deeper understanding of how facial restricted genes act in both gain- and loss-of-function experiments in several animal models (Gritli-Linde,2008) is necessary before we can hope to prevent or ameliorate congenital facial defects.
Embryos and Bead Implantation
Fertilized White leghorn eggs were obtained from the University of Alberta and incubated to the appropriate stage (Hamburger and Hamilton,1951). Stage-24 embryos were treated with several different signaling molecules or proteins using microscopic beads. In some cases, the compounds were applied locally to the lateral corner of the frontonasal mass (globular process), and in other embryos beads were implanted into the medial maxillary prominence.
AG1X-2 beads (BioRad, Hercules, CA; format form, 200-μm diameter) were soaked in 1 mg/ml SU5402 (EMD Biosciences, Czech Republic) dissolved in dimethyl sulfoxide (DMSO) for 1 hr at room temperature. Control beads were soaked in DMSO only. Affi-Gel blue agarose beads (BioRad) were soaked in 0.1 mg/ml BMP4 and BMP7, 0.65 mg/ml Noggin (Regeneron, Tarrytown, NY), 1 mg/ml FGF2 and FGF8 (Peprotech, Rocky Hill, NJ). SHH protein (5 mg/ml) was expressed in the Richman lab and proven to induce ectopic digits in other studies (Towers et al.,2008).
In order to block FGF signaling coming from the brain and/or epithelium, injections rather than beads were used. Stage-15 embryos were injected medial to the right nasal placode with either SU5402 (1 mg/ml) or DMSO. Injections were performed using a Picospritzer (General Valve Corp.) with electrically pulled needles.
Whole-Mount In Situ Hybridization
Embryos were fixed in 4% PFA and processed into 100% methanol. Whole-mount in situ hybridization using digoxigenin-labelled RNA probes was performed as described using an Intavis in situ hybridization robot (Song et al.,2004). The following individuals generously provided gallus cDNAs for this study: A. Kispert, TBX22; M. Kessel, DLX5; S. Wedden, MSX2; M. Towers, PCNA.
For double whole mount in situ hybridization, the probe for the gene of interest was synthesized with digoxigenin-tagged UTP, and the probe for hTBX22 or pol was synthesized with fluorescein-tagged UTP. Hybridization was carried out with both labeled probes at the same time and detection carried out sequentially with BCIP (5-Bromo-4-chloro-3-indolyl phosphate p-toluidine salt) alone, or BCIP + NBT (Nitrotetrazolium Blue chloride) combined.
RCAS::hTBX22 Retroviral Construction and Injection of Virus
The full-length clone for human TBX22 was purchased from Invitrogen (Carlsbad, CA; Clone ID IOH46151). The human protein is 69% overall identical to gallus and in the T-box domain the homology is 90%. The Gateway recombinase reaction (Invitrogen) was performed using the RCASBP-DV Gateway compatible destination vector (Loftus et al.,2001). RCASBPA containing GFP was generously provided by S. Gaunt. Virus was grown up using methods described by Logan et al. (Logan and Tabin,1998). Microinjections were used to deliver virus particles or virus-infected DF1 chicken fibroblast cells (ATCC). DF1 cells were transfected with proviral DNA and cultured for a minimum of 2 weeks prior to injection into embryos. Stage-10 injections were directed into the mesenchyme adjacent to the mesencephalon, just caudal to the optic stalk. Stage-15 injections were directed to the lateral edge of the frontonasal mass (medial to the nasal placode).
BrdU and TUNEL Analysis
Forty-eight hours after frontonasal viral injection into the frontonasal mass (virus injected at stage 15, BrdU label injected at stage 24), 50 μl of 10 mM BrdU (bromodeoxyuridine) was injected in the amniotic sac. Embryos were fixed after 2 hr, embedded in wax, and sectioned at a thickness of 7 μm. Sections were steamed with 10 mM sodium citrate (pH 6.0) for 10 min and treated with exonucleases. Primary antibody for BrdU (GE Healthcare) was applied for 1 hr at room temperature. Alexa Fluor 488 (Invitrogen) labeled goat anti-mouse antibody (1:200) was incubated at room temperature for 20 min. Slides were coverslipped using Prolong Gold Antifade with DAPI (Invitrogen). Total (DAPI) and proliferating (BrdU) cell counts were made using Northern Eclipse software and the percentage of labeled cells was determined. An area of 250 × 450 μm was counted in the lateral frontonasal mass mesenchyme. Three technical replicates were analyzed for each sample. Sections were 14 μm away from each other and average values were obtained for the technical replicates. These mean values were used for the Student's t-tests that compared proliferating cells in the control and experimental groups. TUNEL reaction was performed on adjacent sections to those used for BrdU staining (Szabo-Rogers et al.,2008).
The 3C2 antibody (Developmental Studies Hybridoma Bank, concentrated supernatant made by our lab, diluted 1:3) was used to detect viral Gag protein. Adjacent sections to those used for BrdU and TUNEL were pretreated the same way as BrdU. However, exonucleases were not used. Blocking was performed using 10% Goat serum in PBS for 1 hr at room temperature. Supernatant was added to slides and incubated for 1 hr at room temperature. The same secondary antibody and mountant was used as for BrdU.
Virus Injections Combined With Bead Implants
Embryos were injected into the lateral edge of the frontonasal mass at stage 15 with RCAS::hTBX22 expressing DF1 cells. Following 48 hr of incubation, embryos had reached stage 24–25. An FGF2-soaked bead (1 mg/ml) was implanted into the right side of the frontonasal mass and embryos were reincubated for 16 hr. Embryos were fixed, processed for in situ hybridization, and photographed for frontonasal measurements.
Frontonasal Mass Measurements
Embryos treated with either FGF2-soaked beads, RCAS::hTBX22 DF1 cells, or dually treated with FGF-soaked beads and RCAS::hTBX22 were photographed frontally following in situ hybridization. Control stage-matched embryos were treated with Tris soaked beads, hybridized to a PCNA probe, and then photographed. Image J was used to measure the total width of the frontonasal mass, between the medial edges of the nasal slits. One third of this distance from the nasal slit was used as the maximum width of the area measured. In other words, the area of the lateral third of the frontonasal mass was measured on the treated and control sides (Fig. 3D). The centre was omitted because this region was variable in terms of vial infection. The virus delivered with DF1 cells did not spread to the untreated lateral third. The relative difference between the sides was calculated as a ratio of treated/untreated area. One-way ANOVA and Tukey's post-hoc testing (Statistica) was used to determine whether any of the treatments altered the size of the treated frontonasal mass.
To study bone and cartilage morphology, late stage embryos (stage 38) were first fixed in 100% ethanol, then processed through 100% acetone and stained with Alizarin Red and Alcian Blue as described (Plant et al.,2000).
The authors thank Cheryl Whiting and Kathy Fu for technical support and Emma Kim for cell proliferation counts. N.H. was funded by a Department of Trade and Industry Canada Post-doctoral Fellowship. This work was funded by CIHR grants and an Alpha Omega Foundation award to J.M.R.