Gene expression profiling in the developing secondary palate in the absence of Tbx1 function

Microdeletion of chromosome 22q11 is associated with significant developmental anomalies, including disruption of the cardiac outflow tract, thymic/parathyroid aplasia and cleft palate. Amongst the genes within this region, TBX1 is a major candidate for many of these developmental defects. Targeted deletion of Tbx1 in the mouse has provided significant insight into the function of this transcription factor during early development of the cardiac and pharyngeal systems. However, less is known about its role during palatogenesis. To assess the influence of Tbx1 function on gene expression profile within the developing palate we performed a microarray screen using total RNA isolated from the secondary palate of E13.5 mouse embryos wild type, heterozygous and mutant for Tbx1. Expression-level filtering and statistical analysis revealed a total of 577 genes differentially expressed across genotypes. Data were clustered into 3 groups based on comparison between genotypes. Group A was composed of differentially expressed genes in mutant compared to wild type (n = 89); Group B included differentially expressed genes in heterozygous compared to wild type (n = 400) and Group C included differentially expressed genes in mutant compared to heterozygous (n = 88). High-throughput quantitative real-time PCR (RT-PCR) confirmed a total of 27 genes significantly changed between wild type and mutant; and 27 genes between heterozygote and mutant. Amongst these, the majority were present in both groups A and C (26 genes). Associations existed with hypertrophic cardiomyopathy, cardiac muscle contraction, dilated cardiomyopathy, focal adhesion, tight junction and calcium signalling pathways. No significant differences in gene expression were found between wild type and heterozygous palatal shelves. Significant differences in gene expression profile within the secondary palate of wild type and mutant embryos is consistent with a primary role for Tbx1 during palatogenesis.

Background 22q11.2 deletion syndrome (22q11.2DS) is the most common human microdeletion [1] occurring with a prevalence of 1:4000 and incidence ranging from 1:2000-6395 [2][3][4]. This microdeletion is associated with several syndromic conditions including DiGeorge (DGS; MIM 188400), velocardiofacial (VCFS; MIM 192430), conotruncal anomaly face (CAFS or Takao syndrome; MIM 217095) and isolated outflow tract (OFT) defects of the heart [5][6][7][8][9]. These conditions are characterized predominantly by the presence of congenital heart defects, thymic and parathyroid hypoplasia, and craniofacial dysmorphism, including oro-facial clefting that predominates as isolated cleft palate, micrognathia and (less commonly) dental defects [10][11][12][13]. The most common deletions are phenotypically indistinguishable from each other and consist of either a 3 Mb segment spanning the low copy repeats (LCR) A-D (around 85% of cases); or a smaller 1.5 Mb deletion that spans LCR A-B seen in around 15% of cases [14][15][16]. A less common LCR C-D deletion of the typical 22q11.2DS region has also been identified, which is associated with a much-reduced prevalence of cardiac malformations and oro-facial clefting [17][18][19]. 22q11.2DS is a contiguous gene and haploinsufficient syndrome with at least 30 different genes potentially contributing to the characteristic clinical features [20,21]. Amongst the genes identified as candidates for the development of 22q11.2DS, T-Box 1 (TBX1), which encodes a T-Box-containing transcription factor is recognised as a major determinant through its location within the 22q11 critical region [21][22][23], expression in organs affected within the clinical spectrum [24][25][26][27] and observations that loss of Tbx1 function in mouse recapitulates the clinical findings seen in many DGS subjects [23,[28][29][30][31]. Supporting this, TBX1 mutation has been identified in a sporadic case of DGS [32] and Tbx1 haploinsufficiency results in the most characteristic phenotypes related to developmental defects in the embryonic pharyngeal apparatus [32,33]. DGS is also referred to as the III-IV pharyngeal pouch syndrome, as the pharyngeal pouches and their associated blood vessels are the structures most commonly affected [23,30]. Apart from the aortic arch, thymus and parathyroid gland defects, Tbx1 murine models also manifest craniofacial anomalies that arise from developmental defects associated with pharyngeal arches I and II [23,34,35]. Indeed, conditional mutant models have revealed a tissue-specific requirement and a dose sensitivity for Tbx1 during murine pharyngeal development [20,[36][37][38].
The majority of 22q11.2DS individuals have a characteristic craniofacial morphology including lateral displacement of the inner canthi, swollen eyelids, small mouth, hypoplastic mandible, flat nasal bridge and square nose [39][40][41]. Cleft palate (including submucous cleft) is also present in approximately 10% of subjects [40]. Morphological studies to assess embryonic malformations in various Tbx1 genotypes also reveal the presence of cleft palate in Tbx1-overexpressing mice [42,43]. Therefore, both loss and gain of Tbx1 function can lead to the development of a cleft phenotype.
The palate is divided anatomically into primary and secondary regions with the secondary palate composed of both hard and soft tissues. Embryologically, the secondary palate is derived from the paired maxillary processes of pharyngeal arch I, which gives rise to the palatal shelves. During palatogenesis, these shelves are initially situated bilaterally adjacent to the developing tongue; however, progressive growth and elevation results in them positioning themselves above the tongue, with further medial growth leading to fusion with their counterpart along the midline to create a single continuous palate. The palatal shelves also fuse with the nasal septum superiorly and primary palate anteriorly, completing separation of the nasal and oral cavities [44][45][46]. In the developing mouse embryo, Tbx1 is expressed in epithelium of the palatal shelves throughout palatogenesis from embryonic day (E)12.5-15.5 [24]. The etiological basis of the cleft palate phenotype in Tbx1 mutants is not fully understood but has been associated with abnormal palatal shelf elevation, possibly due to a combination of increased tongue height, decreased palatal shelf width, perturbed cell proliferation and apoptosis [47]. In addition, inappropriate fusion between the palatal shelf epithelium and tongue has also been described in this mutant, associated with hyper-proliferation and disrupted differentiation [48]. More recently, confocal image analysis has found only subtle differences in levels of proliferation within mesenchyme of the palatal shelves between wild-type and mutant until the later stages of palatogenesis; although significant differences in mesenchymal cell orientation were found in mutant shelves, which might contribute to the cleft phenotype [49].
We are interested in further defining the role of Tbx1 during the process of murine palatogenesis. Specifically, we have investigated regulation of this transcription factor in the secondary palate and carried out a functionally-based microarray using the Tbx1 mouse model. We compared total RNA isolated from dissected secondary palatal shelves derived from E13.5 wild type (WT), Tbx1 +/− (heterozygous) and Tbx1 −/− (mutant) embryos and clustered the data into three groups based on comparison between the three genotypes. Microarray analysis demonstrated that in the absence of functional Tbx1, significant changes occur in the expression profile of numerous genes in mutant versus WT and mutant versus heterozygous groups. The most significant pathways affected in both groups were the hypertrophic cardiomyopathy, cardiac muscle contraction, dilated cardiomyopathy, focal adhesion, calcium signalling and tight junction pathways. High-throughput quantitative RT-PCR validation confirmed significant variation between WT and mutant in the expression of 26 individual genes. We discuss these findings within the context of murine secondary palatogenesis.

Regulation of Tbx1 in the developing secondary palate
Tbx1 transcriptional activity is present in epithelium of the secondary palate shelves throughout the processes of growth, elevation and fusion (Additional file 1) and Tbx1 mutant mice have a fully penetrant cleft palate [23,30,31]. We are interested in further defining the function of this transcription factor during palatogenesis at the molecular level and first sought to understand how Tbx1 transcription might be regulated in the palatal shelf epithelium. We began by investigating the effect of abrogating either Sonic hedgehog (Shh) or Fibroblast growth factor (Fgf) signaling in palatal shelf explants as there are potential associations between these signaling networks and Tbx1 function in the developing palate. Shh is also expressed in the palatal epithelium and lies upstream of Tbx1 in the pharyngeal endoderm [50]; whilst Fgf signaling can maintain Tbx1 expression in early odontogenic epithelium [27]. Specifically, E13.5 secondary palatal shelves were isolated and cultured for 24 h in the presence of either the Shh antagonist cyclopamine or the Fgf receptor inhibitor SU4502. Interestingly, whilst an absence of Shh signaling did not affect Tbx1 transcription, loss of Fgf signaling resulted in a loss of Tbx1 activity in the palatal epithelium after 24 h of culture ( Fig. 1a-g). These results place Tbx1 downstream of Fgf signaling during early palatogenesis and in contrast to the pharyngeal region, loss of Shh does not affect Tbx1.

Altered gene expression in the secondary palate of Tbx1 mutant mice
It is known that Shh, Fgf and Bone morphogenetic protein (Bmp) signaling pathways are important during normal development of the palate [51][52][53]; in particular, reciprocal signaling between epithelial Shh and mesenchymal Fgf10, mediated through fibroblast growth factor receptor 2b (Fgfr2b), regulates cell proliferation in the mesenchyme [54]. Whilst Shh also negatively regulates Bmp4 in the mesenchyme, which is itself upstream of Fgf10 [55]. Tbx1 interacts with a number of these molecules during embryogenesis, being directly upstream of Fgf10 in the early heart field [28,56]; negatively modulating Bmp4 through the binding of Smad1 in cardiomyocytes [36] and being downstream of Shh in endoderm of the early pharynx [50]. Within the palate itself, it has been variously suggested that Tbx1 negatively regulates Fgf10 and Bmp4, whilst positively regulating Fgf8 and Pax9, although there is currently not a consensus on these findings [47,48].
Although we could find no evidence that Tbx1 is downstream of Shh signaling in the palatal epithelium, there is considerable overlap of expression. We therefore investigated known targets of Shh within palatal shelves WT and mutant for Tbx1 using in situ hybridization. Interestingly, we found no significant differences in expression of Shh, Fgf10 and Fgfr2b between WT and mutant ( Fig. 2a-f ). However, whilst Fgf8 expression was also normal in the mutant shelves ( Fig. 2g-h), Bmp4 and paired-box 9 (Pax9) were slightly up and downregulated, respectively in the posterior region of the secondary palate ( Fig. 2i-l). These apparent changes in Bmp4 and Pax9 expression in the mutant might simply be a function of altered numbers of cells expressing these genes in the palate mesenchyme, particularly as the Tbx1 domain within the palatal epithelium does not completely overlie those of Bmp4 or Pax9 in the mesenchyme [48]. However, given the evidence of retarded growth in Tbx1 mutant palatal shelves [47,48] if an alteration in cell number is responsible for any of these changes, it would seem to be more likely for Pax9.

Microarray analysis
To further identify potential transcriptional target genes of Tbx1 implicated in palatogenesis, microarray analysis was carried out using cDNA transcribed from total RNA derived from the dissected secondary palatal shelves of E13.5 Tbx1 +/+ ; Tbx1 +/− and Tbx1 −/− embryos (n = 3 for each genotype).
After normalization and filtering of microarray data, comparison between mutant embryos and WT (Group A), heterozygous and WT (Group B) and mutant versus heterozygous (Group C) were performed (adj. p < 0.1). The WebGestalt database was used to identify biological pathways associated with these differentially expressed transcripts [57]. In Group A, 89 genes were identified to be differentially expressed in mutant compared to WT (adj. p < 0.1, fold change 1.4). From these, 3 genes were upregulated, whereas the majority (n = 86) were downregulated (Table 1). Group B includes differentially expressed genes arising from the comparison of heterozygous and WT palates (n = 400, adj. p > 0.23). This group list was not considered statistically significant (adj. p > 0.1) and therefore was not analysed further (Additional file 2). In Group C, 88 genes were identified to be differentially expressed in the mutant compared to heterozygote palate (adj. p < 0.1, fold change 1.3). Amongst these, 11 genes were upregulated, whereas 77 were downregulated (Table 1). In Group A, from the 89 genes that were searched, 9 Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways were identified (Fig. 3a). The most statistically enriched pathways (adj. p < 0.1) were all associated with cardiac muscle physiology and included hypertrophic cardiomyopathy, cardiac muscle contraction, dilated cardiomyopathy, arrhythmogenic right ventricular cardiomyopathy and vascular smooth muscle contraction. Other pathways included phagosome and focal adhesion, tight junction and calcium signaling pathways and Alzheimer's disease (Additional file 3). In Group C, from the 88 genes that were searched, 10 KEGG pathways were identified (Fig. 3b). The most statistically enriched pathways (adj. p < 0.1) were all also associated with cardiac muscle physiology, including hypertrophic and dilated cardiomyopathy and arrhythmogenic right ventricular cardiomyopathy. Other pathways included tight junction, calcium signalling, focal adhesion, neuroactive ligand-receptor interaction, phagosome and Alzheimer's disease pathways (Additional file 3). We were then interested to further identify the proportion of overlap amongst significantly differentially expressed genes between Groups A and C. (Fig. 4a [58]). The two groups share 58 commonly expressed genes ( Table 2) when compared to WT and heterozygous; whereas 30 genes ( Table 2) were uniquely observed in Group A and 20 in Group C ( Table 2; adj. P < 0.1). The WebGestalt database was used to provide insights into the mechanism of regulation     associated with these 58 common gene transcripts. Adding to the above approach, heat map and dendogram clustering of the commonly expressed genes, as well as uniquely expressed genes in Group A and Group C (n = 99 genes) revealed transcriptional homogenicity between genotypes (Fig. 4b). Genes upregulated in mutants clearly clustered together and were shown to be downregulated in heterozygote and WT samples (red asterisks in Fig. 4b).
In contrast, the downregulated transcriptome of mutant samples was shown to increase its expression in heterozygous and WT palates. Although statistical analysis revealed a non-significant expression pattern of Tbx1 heterozygous samples (adj. p values > 0.1), heat map revealed a similarity in gene expression pattern between heterozygous and WT samples.

Confirmation of microarray data
For validation of the results obtained by microarray, RT-PCR was carried out using gene-specific primers (Applied Biosystems; Additional file 4) and the original RNA samples. In total, 27 genes from Group A and 28 genes from Group C were selected for gene expression verification (Table 3). Changes in gene expression of these transcripts were normalized to that of ß-Actin. In both groups, 27 genes were commonly expressed ( Table 3; Fig. 5a); Alas2 was uniquely present in Group A, whereas Ank1 and Chrna1 were uniquely present in Group C (Table 3; Fig. 5b). All genes tested were confirmed as being significantly changed between WT-mutant and heterozygote-mutant except for Ank1 (Group C; p = 0.102). In Group A, Rapsn,     Sh3bgr, Tnnc2, Tnni2 and Tnnt2 were the most downregulated genes; whereas in Group C, these were Csrp3, Sh3bgr, Sln, Tnnc2, Tnni2, Myh7 and Mylpf.

Discussion
In the present study, functional explant assays and microarray analysis of gene expression was carried out in the palatal shelves of E13.5 mouse embryos WT, heterozygous or mutant for Tbx1. This was prompted by the knowledge that Tbx1 is strongly expressed in epithelium of the palatal shelves throughout palatogenesis, mutant embryos demonstrate cleft palate with complete penetrance [23,24,47,48] and the findings that Tbx1 has multiple potential roles during normal palatal shelf elevation, elongation and adhesion [47,48]. It is known that several regulatory networks underlie signaling between epithelium and mesenchyme during development of the secondary palate and we sought to discover potential genetic pathways disrupted during palatogenesis in the absence of Tbx1. We therefore focused our investigations at E13.5, just prior to the period of rapid growth and elevation [45].
A key finding of this profile is the association between an absence of Tbx1 function and altered expression (primarily downregulation) in a number of muscle-related genes within the shelves of the secondary palate. Developing mononuclear and binucleate myofibril-containing skeletal muscle cells are identifiable within the palatal shelves at E13 [59] and findings of altered gene expression are perhaps not surprising, given the essential role of Tbx1 during the development of branchiomeric musculature and somite-derived tongue muscles [60][61][62] and detectable expression in adult mouse muscle [63,64]. In the embryo, Tbx1 activates the myogenic-determination genes myogenic factor 5 (Myf5) and myogenic differentiation (MyoD) in the mesodermal core of pharyngeal arches I and II [61]. In addition, loss of Tbx1 results in impairment of the onset of myogenic specification [60] and Tbx1 synergizes with the myogenic factor Myf5 for initiation of myogenic cell fate [65]. Our array failed to identify variation in Myf5 and MyoD, but verified downregulation of Myf7 at E13.5 in mutant palatal shelves. This finding suggests that Tbx1 functions upstream of myosin heavy chain 7 (Myh7) during palatal shelf formation and just prior to elevation, possibly as a myogenic factor. The presence of asymmetric expression patterns of myogenic regulatory factors in early first arch-derived muscles of Tbx1 mutant embryos might explain the absence of Myf5 and MyoD gene transcripts [61]. In addition, both skeletal, smooth and non-muscle contractile systems have been identified and implicated in the process of normal palatal shelf elevation [66,67]. A number of the downregulated genes identified have also been implicated in the process of skeletal and cardiac muscle contraction (Tnni2, Tnnt1, Myh3, Myom1, Tnnc2), which might reflect the lack of skeletal myogenic determination. Interestingly, microarray analysis It cannot be discounted that other intrinsic contractile systems might also be disrupted in the secondary palate of Tbx1 mutant mice. Indeed, changes in expression levels were also identified in genes associated with intracellular calcium signaling (Atp2a1, Tnnc2, Cacna1s, Tnnc1), which is known to mediate a number of important physiological processes of relevance to palatogenesis, including skeletal and smooth muscle contraction, apoptosis, cell motility and proliferation [69].
After palatal shelf elevation, periderm cells joined by tight junctions are believed to function as a protective layer, preventing aberrant adhesions and playing an important role in mediating appropriate shelf adherence and epithelial differentiation [70,71]. Loss of periderm is required at the tips of opposing palatal shelves and overall at sites where fusion is required [71]. In Tbx1 mutant mice, aberrant oral adhesions between tongue and palatal shelves have been observed [48]. In the present study, the tight junction genes Myh3, Mylpf, Myh7 and Actn2 were downregulated in mutants at E13.5, suggesting a potential role for Tbx1 in the normal function of tight junctions present within the palatal shelf epithelium.
Comparison between WT-mutant and heterozygous-mutant shelves revealed 58 genes commonly expressed in both groups. From these, 27 genes from Group A and 28 genes from Group C were selected for gene expression verification. Analysis revealed significant downregulation of 26 genes common to both groups (see Fig.  5a) with (Alas2) and (Ank1, Chrna1) individually downregulated in each group, respectively (see Fig.  5b). Statistical analysis revealed significant downregulation of all genes tested through RT-PCR with the exception of Ank1 (p = 0.102; see Fig. 5b). Pathway analysis of these validated genes confirmed the associations between cardiac muscle contraction and calcium signaling, but also suggested links with dilated and hypertrophic cardiomyopathies. Although 22q11.2DS is commonly associated with conotruncal congenital heart defects, hypocalcemic dilated myocardiopathy has also been described in association with this condition [72]. RT-PCR validation of the microarray analysis demonstrated no significant changes in gene expression between WT and heterozygous shelves, consistent with the normal palatogenesis seen in heterozygous embryos [23].
Tbx1 is known to regulate both Fgf8 and Fgf10 expression in the early pharyngeal arches and cardiac outflow tract [64] and influence the spatial distribution of Fgf8 and Bmp4 in the early mandible [73]. It has also been suggested that Fgf8 is significantly downregulated in the palatal shelf epithelium, whilst Fgf10 is upregulated in the mesenchyme at E13.5 in Tbx1 mutant embryos [47]. However, we found no evidence of altered transcript levels associated with these genes in our array. This same report also demonstrated diminished hyaluronic acid (HA) in the palatal shelves of Tbx1 mutant mice and whilst we found no obvious genetic links to this finding within our array, HA has been shown to induce matrix metalloproteinase 9 (MMP9) [74], which was downregulated. However, whilst some members of the MMP family have been directly related to palatogenesis, at least in vitro; this did not include MMP9 [75].
In this microarray experiment, RNA was derived from whole dissected palatal shelves and therefore no formal distinction was made between changes in epithelial and mesenchymal gene activity. Tbx1 is localized to the palatal shelf epithelium at E13.5, but is clearly able to influence signaling activity between epithelium and mesenchyme in the palate (Fig. 6). Indeed, the associations between Tbx1 function and muscle contraction and calcium signaling, both activities that take place in the early mesenchyme, are consistent with this. In addition, Tbx1 seems to act co-operatively with Shh signaling in the palate, through the repression of Bmp4 and induction of Pax9. Interestingly, this co-operative activity would appear to be dependent upon Fgf signaling; Shh in the epithelium is dependent upon reciprocal signaling with Fgf10 in the mesenchyme [54] and our explant studies demonstrate that Tbx1 is also dependent upon Fgf signaling. Although it is currently not known which Fgf ligand is required or whether this is within the epithelium or mesenchyme, maintenance of epithelial Tbx1 transcription is essential for normal palatogenesis. Conditional loss of Tbx1 in either craniofacial mesenchyme [48] or mesoderm [76] does not result in cleft palate, in contrast to loss-of-function in the oral epithelium, which does [48].

Conclusions
We have conducted functional microarray analysis and PCR validation of gene expression in the developing secondary palate at E13.5 in the Tbx1 mutant embryo. Differentially regulated genes were detected in the absence of this transcription factor. In the microarray, a total of 89 genes demonstrated differential expression in Group A and 88 genes in Group C (adj. p < 0.1), whilst high-throughput quantitative RT-PCR confirmed 27 genes significantly changed between WT and mutant and 28 between heterozygote and mutant. Associations existed with cardiac muscle development, hypertrophic and dilated cardiomyopathy, tight junction and calcium signaling. These findings provide further evidence of a primary role for Tbx1 during the process of palatogenesis. Fig. 6 Molecular associations linking Tbx1 with Fgf and Shh signaling in the developing palate. Tbx1 in the palatal shelf epithelium is downstream of Fgf signaling, the ligand/s and source (epithelium/ mesenchyme) are currently unknown. Shh-Fgf10-Fgfr2b epithelial-mesenchymal reciprocal signaling [54] antagonizes Bmp4 [55] and induces Pax9 indirectly through the induction of Osr2 [55,82]. We and others [48] have demonstrated that Tbx1 acts to inhibit Bmp4 and induce Pax9. It has been suggested that Tbx1 activity is required for Fgf8 induction in the epithelium and Fgf10 inhibition in the mesenchyme [47]; however, we and others [48] have found no evidence of this

Mice
Breeding mice were maintained in ventilated cages on an alternating (12:12) light-dark cycle in the Biological Services Unit at King's College London. Time-mated Tbx1 embryos were generated by inter-crossing Tbx1 +/− mice on a C57/Bl6 background [23] such that noon of the day on which vaginal plugs were detected was considered as embryonic day (E) 0.5. Pregnant females were euthanized with cervical dislocation.

Explant culture
Secondary palatal shelves were carefully micro-dissected from E13.5 WT embryos and cultured for 24 h in the presence of cyclopamine or SU4502 as previously described [77]. Briefly, explants were cultured using a modified Trowell technique at 37°C in an atmosphere of 5% CO2 in serum-free Advanced DMEM/F12 (Gib-coBRL) supplemented with 20 U/ml penicillin and streptomycin (GibcoBRL), 10% Fetal Bovine Serum (GibcoBRL), 50 mM transferrin (Sigma) and 150 μg/ml ascorbic acid (Sigma). SU5402 (Calbiochem) was diluted in medium from a 10 mM stock solution in DMSO and cyclopamine (Sigma) was diluted from a 20 mg/ml stock solution in ethanol and added to the culture medium at a final concentration of 75 μM for both inhibitors. A minimum of (n = 6) palatal shelves were used for each experiment.

In situ hybridisation
Wholemount digoxygenin and section 35 S radioactive in situ hybridisation was carried out as previously described [78]. Wholemount (n = 6 palatal shelves) and section (n = 3 embryos) images were photographed using Leica or Zeiss Axioscop microscopes, respectively. For radioactive in situ hybridisation, light and darkfield images were merged in Adobe photoshop CS. Plasmid cDNA was kindly provided by the following investigators: Bmp4 (Brigid Hogan); Fgf8 (Ivor Mason);

Tissue preparation and microarray analysis
Secondary palatal shelves were carefully micro-dissected from E13.5 Tbx1 WT, heterozygous or mutant embryos (3 embryos per genotype), stored as pairs from each embryo in RNAlater (Ambion) and then homogenized using a blunt 20-guage needle to an RNase-free syringe. Total RNA was extracted from homogenate derived from each shelf pair using an RNeasy Isolation Kit (Qiagen). RNA quality was checked using an Agilent Bioanalyzer and quantified with spectrophotometry (NanoDrop ND-1000). In total, 9 sets of RNA were collected, each derived from paired secondary palatal shelves harvested from each embryonic genotype (giving 3 samples from each genotype).

Microarray chip processing and data analysis
The expression profiling analysis was carried out at the Franklin-Wilkins Building Genomics Facility, King's College London. Total RNA was reverse-transcribed and cRNA generated using the MessageAmp II-Biotin Enhanced cRNA Amplification Kit (Ambion). cRNA targets were then hybridized to the Affymetrix Mouse Gene-Chip microarray (MOE430_A_2 GeneChip array), which is a single array containing 22,690 probe sets representing transcripts and variants from over 14,000 well characterized mouse genes. A single chip was used for each pair of palatal shelves per genotype, with hybridization and scanning of array chips carried out according to recommended protocols (www.affymetrix.com).
Microarray data were analysed by the implementation of Bioconductor packages in the programming language R. Intensity values of every chip were imported and evaluated with the packages affy, simpleaffy and affyPLM. Pre-processing, normalization and expression transformations were executed by the function rma of the affy package [79]. Gene expressions were fitted to linear models and moderated t-statistics were calculated for specific comparisons using lmfit and eBayes functions of the limma package [80]. P-values were adjusted for multiple testing with the Benjamini & Hochberg FDR method [81], implemented within the topTable function of the limma package. Venn diagram and heatmap showing hierarchical clustering with complete linkage scaled by genes were constructed using the packages VennDiagram and gplots respectively. Microarray datasets have been submitted to the Gene Expression Omnibus (GEO) at NCBI (GSE37904).

Functional annotation of differentially regulated gene sets
In this study WEB-based GEne SeT AnaLysis Toolkit (WebGestalt, http://www.webgestalt.org/option.php, version 05/20/2014) was utilized to perform functional enrichment analysis on the data sets containing genes from the Tbx1 −/− versus WT shelves comparison (Group A), the Tbx1 −/− versus Tbx1 +/− shelves comparison (Group C) and the commonly expressed gene set of Group A and Group C. For each gene set, WebGestalt used the hypergeometric test to evaluate functional enrichment against predefined categories collected from KEGG. Statistical analysis was performed according to the current default settings.