GTF2IRD1 in Craniofacial Development of Humans and Mice

See allHide authors and affiliations

Science  18 Nov 2005:
Vol. 310, Issue 5751, pp. 1184-1187
DOI: 10.1126/science.1116142


Craniofacial abnormalities account for about one-third of all human congenital defects, but our understanding of the genetic mechanisms governing craniofacial development is incomplete. We show that GTF2IRD1 is a genetic determinant of mammalian craniofacial and cognitive development, and we implicate another member of the TFII-I transcription factor family, GTF2I, in both aspects. Gtf2ird1-null mice exhibit phenotypic abnormalities reminiscent of the human microdeletion disorder Williams-Beuren syndrome (WBS); craniofacial imaging reveals abnormalities in both skull and jaws that may arise through misregulation of goosecoid, a downstream target of Gtf2ird1. In humans, a rare WBS individual with an atypical deletion, including GTF2IRD1, shows facial dysmorphism and cognitive deficits that differ from those of classic WBS cases. We propose a mechanism of cumulative dosage effects of duplicated and diverged genes applicable to other human chromosomal disorders.

The vertebrate head is a complex structure whose assembly is coordinated by a hierarchy of regulatory genes, many of which are members of transcription factor families (e.g., Hox, Msx, Dlx) (1). The TFII-I gene family encodes related transcription factors that all contain a leucine zipper and multiple I-repeat motifs, whose complex and incompletely defined biology includes regulation of genes important in vertebrate development (24). The three family members are GTF2I (or TFII-I) (5), GTF2IRD1 (or GTF3, MusTRD1, BEN, WBSCR11) (6, 7), and GTF2IRD2 (8). GTF2IRD1 and GTF2I are invariably deleted in WBS (9) as part of a larger deletion. GTF2I, the first member identified, acts as a basal transcription factor that binds to initiator elements of various promoters and also regulates transcription through E-box elements at enhancers in response to upstream signaling events (5). GTF2IRD1 can bind regulatory elements upstream of genes involved in tissue development and differentiation (3, 4). In Xenopus, GTF2IRD1 (or XWBSCR11) acts as a positive regulator of goosecoid in response to the transforming growth factor–β family member activin (3); during mouse embryonic development, Gtf2ird1 interacts with the enhancer that controls early-phase Hoxc8 expression (4). In humans, heterozygous deletions on chromosome 7q11.23, including these genes, are the genetic basis of the multisystem developmental disorder WBS (9).

Individuals with WBS have characteristic dysmorphic facies alongside other developmental abnormalities, including vascular problems (especially supravalvular aortic stenosis, SVAS), short stature, and a unique cognitive profile (WBSCP), with relatively proficient language and face-processing skills but serious impairments in spatial and numerical ability (10, 11). Personality traits include overfriendliness and charismatic speech. Many experience anxiety and simple phobias. The roots of the mental and cognitive aspects probably lie in disruption of normal neurodevelopment, because brain morphology and neural organization are abnormal (12). At the molecular level, WBS is a contiguous gene disorder that involves a heterozygous deletion of ∼1.5 megabase (Mb) encompassing some 28 genes (13, 14) (Fig. 1A). The only firm genotype-phenotype correlation is between vascular problems caused by haploinsufficiency for the elastin gene (9).

Fig. 1.

(A) Transcript map of the WBS region on human chromosome 7q11.23 showing key individuals with partial deletions (1518). WBS18 has classic WBS (∼1.5 Mb deletion); HR has atypical WBS (∼1 Mb deletion); CS has SVAS (∼0.94 Mb deletion). Gray boxes denote areas where WBS deletion breakpoints cluster. CEN, centromere; TEL, telomere; HR, CS, PM, and TM are UK patients. The syntenic mouse region on chromosome 5G1 retains gene order but is inverted (not to scale). (B) Gene expression analysis in HR by RT-PCR. Left to right: control, HR, and WBS18. Numbers represent percentage of gene expression relative to the normal individual and normalized against the HPRT1 control. HR and WBS18 show reduced CYLN2 and GTF2IRD1 expression. (C) Gene expression analysis in mice by quantitative real-time RT-PCR. The histogram shows the relative quantities of Gtf2ird1 normalized against the control gene Lamc1 in adult wild-type (WT) and mutant (Gtf2ird1-null) mice (-ve RT, negative RT blank control). Gtf2ird1 expression is absent in the mutant mice.

Attempts to identify the genes underlying the craniofacial, cognitive, and behavioral features of WBS rely on the identification of rare individuals with smaller deletions in the WBS region, supplemented by studies of mouse models. Early links made between haploinsufficiency for LIMK1 and the characteristic cognitive profile (15) are ambiguous and are contradicted by reports of individuals who have this gene deleted but lack the WBSCP (16). An SVAS patient (16), CS, with 23 of the 28 defined genes in the critical region deleted (including LIMK1) (Fig. 1A), has none of the craniofacial, cognitive, or behavioral abnormalities that are seen in individuals with the full deletion. CS retains genes at the telomeric end of the critical region, which suggests that the decisive genes responsible for craniofacial and neurological abnormalities are located in this region (which harbors the TFII-I gene family), a claim supported by other patient studies (1618). Here, we show that both GTF2IRD1 and GTF2I are responsible for the main aspects of WBS.

We have identified an atypical WBS individual, HR, with a smaller genetic deletion relative to classic WBS cases but a larger deletion than in SVAS patient CS, including two extra telomeric genes, CYLN2 and GTF2IRD1. HR is a 4.5-year-old girl with surgically corrected pulmonary artery stenosis. Her birth-weight (3380 g) and growth appear normal, and at 4.5 years her height is just above the 50th centile. Her facial features are suggestive of, but not classical for, WBS (Fig. 2A). Early developmental milestones such as sitting and walking were within normal limits. However, at 18 months (and unlike her older sibling at the same age) she had a vocabulary of only a few single words, and by age 4 she continued to show a delay in language acquisition as well as serious deficits in spatial cognition [not to the degree seen in WBS (19)]. She does not exhibit the overly friendly personality.

Fig. 2.

Craniofacial characterization of HR. (A) 2D photographs of (i) WBS child (4 years) (WBS18) with characteristic facial dysmorphism; (ii) atypical individual HR (3.5 years) with mild dysmorphic features. Note periorbital fullness, long philtrum, upturned nose. (B) 3D face surfaces of HR and a typical female WBS child alongside averages of control and WBS subgroups. Each is synthesized in a DSM made up of all 270 images (excluding HR's). (C) Scatterplot display of the unseen classification of HR against the control and WBS DSMs. Horizontal broken lines represent the position of the mean control and WBS faces at –1 and +1, respectively. Vertical axis: position of each face relative to the means normalized to [–1, +1]; horizontal axis: age in years. Background scatterplot represents the classification of the unseen test sets used to determine the average discrimination performance of the 20 DSMs built with the randomly generated training sets. Overlaid is the average position of the unseen closest mean classification of HR's image estimated using the 20 DSMs. HR is classified as borderline (i.e., mildly dysmorphic).

The WBS-like dysmorphic features of HR were quantified using three-dimensional (3D) face surface images captured with stereo photogrammetric devices. Images were compared unseen to dense surface models (DSMs) of face shape constructed from a collection of 185 control and 85 WBS individuals aged 2 weeks to 20 years (Fig. 2B). Such comparisons can reliably distinguish between dysmorphic and normal faces (92% discriminating accuracy) (20, 21). HR is on the periphery of both the WBS and control groups and is classified as having mild, but not classic, WBS features (Fig. 2C). Because some members (adults and children) of two families with a smaller deletion (ELN + LIMK1) have been reported with some dysmorphic features (15), we compared 3D face images captured from two adults with a similar deletion (PM and TM, Fig. 1A), but no phenotype other than SVAS, to DSMs constructed from 132 control and 45 WBS adults. No WBS-like features were present (fig. S1). Their childhood photographs also showed no facial dysmorphology.

Chromosome analysis of HR with an ELN/LIMK1 probe identified a heterozygous deletion at 7q11.23. The extent of the deletion was determined by polymerase chain reaction (PCR) analysis of DNA isolated from somatic cell hybrids containing either the deleted or the normal copy of chromosome 7. These data indicate that the heterozygous deletion in HR spans ∼1 Mb encompassing the interval between genes NOL1R and GTF2IRD1 (Fig. 1A) (fig. S2). The proximal breakpoint lies within the centromeric low-copy repeat (C-mid), comprising pseudogenes, repeats, and partial genes (14). The distal breakpoint is located ∼7 kb downstream of exon 1 and within intron 1 of GTF2IRD1. Intron 1 is large (53,895 base pairs) and rich in repeats (∼53% content of short and long interspersed nuclear elements and long terminal repeats) that appear to make the region unstable and prone to microdeletions. Semiquantitative reverse transcription PCR (RT-PCR) of lymphoblastoid cell line RNA from HR revealed reduced expression of GTF2IRD1 and normal levels of GTF2I expression (Fig. 1B). HR is therefore haploinsufficient for GTF2IRD1, even though the gene is only partially deleted and the translation start codon in exon 2 is still present.

To further study the effects of lack of GTF2IRD1, we used the transgenic mouse strain Tg(Alb1-Myc)166.8 (22) in which integration of a c-myc transgene on distal chromosome 5 has induced a deletion of ∼40 kb in the mouse WBS syntenic region. The deletion starts downstream of Cyln2 and includes the upstream transcription start site and exon 1 of Gtf2ird1 while retaining the translation start site in exon 2. It therefore closely resembles the GTF2IRD1 deletion breakpoint in HR. Gtf2ird1 expression was abolished in the transgenic line, whereas expression of Cyln2 was unaffected (Fig. 1C) (fig. S3). These mice are prone to liver-specific hepatocellular carcinomas due to overexpression of the c-myc transgene (22), but our analysis has revealed additional phenotypes relevant to WBS.

In mice, loss of Gtf2ird1 leads to growth retardation and craniofacial abnormalities. Gtf2ird1-null mutants are viable and fertile but smaller and lighter than wild-type controls (Fig. 3A). All Gtf2ird1-null mice display a characteristic facial appearance that includes periorbital fullness and a short snout (Fig. 3, B and C). About 20% of the homozygous mutants have a more severe craniofacial abnormality involving a misaligned jaw, which leads to chronic overgrowth of teeth, and a twisted snout (Fig. 3, B and C). This does not appear to affect neonatal suckling; however, once weaned, these mutants are unable to manipulate hard feed pellets. With a modified diet of soft mash and regular teeth clipping, their life-span is normal but they are smaller than their homozygous littermates from birth to adulthood (Fig. 3A).

Fig. 3.

(A) Growth deficiency in Gtf2ird1-null mice. Histogram showing body weights of wild-type (males, n = 13 to 24; females, n = 8), Gtf2ird1-null (males, n = 26 to 27; females, n = 10 to 14), and Gtf2ird1-null mice with misaligned jaws (Mutant*) (males, n = 8; females, n = 9) at ages 1 and 2 months. All mutants have a lower body weight relative to the wild type (9.4 to 13.9% decrease, depending on age); mutants with misaligned jaws (Gtf2ird1-null*) have the lowest body weights (24 to 42% decrease, depending on age). Pairwise comparisons show that differences in body weight between the genotypes are statistically significant. (Two-tailed t tests; P values of wild-type relative to mutant mice at 41 and 2 months, respectively: females, 4.618 × 10–3, 3.72 × 10–4; males, 2.79549 × 10–6, 3.64363 × 10–5. P values of wild-type relative to mutant* mice at 41 and 2 months, respectively: females, 1.25987 × 10–7, 1.15 × 10–4; males, 6.41166 × 10–14, 1.62929 × 10–7.) (B) Craniofacial abnormalities. Misaligned jaw and shorter snout of Gtf2ird1-null mice at ages 12 days and 9 months (top and bottom panels, respectively). (C) Adult Gtf2ird1-null mice display a smaller body size, a dysmorphic face with “periorbital fullness,” and a shorter snout. Lower left panel shows a Gtf2ird1-null mutant (*) with a twisted snout. (D) 2D images of 3D surface scans of two age-matched mice comparing face shape. Top panels show a wild-type and a Gtf2ird1-null male with a twisted snout (mutant*). Bottom panels show the average 3D surfaces of wild-type (n = 10) and mutant* (n = 7) mice.

Quantitative 3D facial scans of seven Gtf2ird1-null mice with the more severe phenotype were compared with scans from 10 wild-type mice. The craniofacial abnormality consistently involves both skull and jaws, in particular the periorbital area. Gtf2ird1 mutants have a shorter snout than wild-type controls, and the face shape is typically asymmetric, with the snout twisted to the left or right by 7° to 12° (Fig. 3D) (fig. S4). Like Gtf2ird1 mutant mice, WBS individuals also exhibit differing degrees of severity in their craniofacial dysmorphology, no doubt due to the influence of genetic background and modifier genes. Gtf2ird1-null mice also display a distinctive hind leg clasping and “kicking” reflex not seen in wild-type controls and suggestive of neurological dysfunction. Growth and craniofacial development appear normal in heterozygous Gtf2ird1+/– mice, which suggests that humans are more sensitive to the craniofacial effects (although apparently not the growth effects) of GTF2IRD1 gene dosage.

Our findings show that homozygous loss of Gtf2ird1 in mice results in craniofacial abnormalities reminiscent of those seen in WBS, together with growth retardation and neurological abnormalities. This is consistent with the expression pattern of Gtf2ird1 in the developing brain and craniofacial areas (23) and its ability to regulate expression of goosecoid (Gsc) and Hoxc8, genes that control craniofacial and skeletal development (3, 4). Gsc-null mice die soon after birth with craniofacial defects, rib fusions, and sternum abnormalities (24); Hoxc8 EE–/– mice, where temporal regulation of Hoxc8 is altered but not abolished, have skeletal pathologies and signs of neurological dysfunction (25). We used a short interfering RNA (siRNA) that knocks down levels of endogenous GTF2IRD1 by ∼60% to show regulation of GSC by GTF2IRD1 in human embryonic kidney (HEK) 293T cells. In cotransfection experiments, the GTF2IRD1 siRNA reduced expression of a pGL3-GSCprom luciferase reporter construct (Fig. 4), suggesting that GTF2IRD1 influences gene transcription at endogenous levels. To date, few upstream regulators of Gsc have been defined in vivo, even though its developmental importance in Drosophila and humans is well established. Misregulation of GSC expression due to absence or lower levels of GTF2IRD1 probably contributes to the craniofacial pathologies seen in WBS and Gtf2ird1-null mice.

Fig. 4.

Gene regulation by GTF2IRD1 through the GSC promoter in vivo. (Top) Western blots of siRNA-transfected HEK-293T cell lysates. Both endogenous GTF2IRD1 and TFII-I are knocked down. (Bottom) RNA interference assays. HEK-293T cells were transfected with a GSC promoter–luciferase gene reporter construct (PGL3-GSCprom) and with either GTF2IRD1 or TFII-I siRNAs. Knockdown of GTF2IRD1 induced ∼60% down-regulation of luciferase activity (mean ± SEM of triplicates). TFII-I control siRNA has no effect under these conditions.

Of the other genes deleted in HR, CYLN2 probably contributes to, but is not solely responsible for, the neurological phenotype in WBS. Although no human case of isolated haploinsufficiency for CYLN2 has been reported, Cyln2–/– and Cyln2+/– mice present with mild structural brain abnormalities, hippocampal dysfunction, and deficits in motor coordination, alongside mild growth deficiency but no craniofacial defects (26). However, haploinsufficiency for CYLN2 and GTF2IRD1 alone cannot explain all the clinical features of WBS, because HR displays a milder clinical profile. The more pronounced facial and cognitive/behavioral phenotypes associated with the larger deletion implicate other telomeric genes in these features, specifically GTF2I, which lies distal to GTF2IRD1 and is deleted in classic WBS cases. GTF2I and GTF2IRD1 share many structural properties and are likely to have overlapping functions; indeed, both have been shown to regulate gene activity through a common DNA element, DICE (27). This work introduces members of the TFII-I gene family as critical regulators of craniofacial and neurological development.

Our findings allow us to compile a picture of the molecular pathology of WBS, a classical human microdeletion syndrome. No single gene is responsible for the craniofacial or cognitive features of WBS. We suggest that cumulative dosage of TFII-I–family genes explains the main phenotypes. Gtf2ird1-null mice and classic WBS individuals have two functioning copies (in trans and cis, respectively), whereas HR has three functioning genes of the GTF2IRD1/GTF2I cluster and shows milder WBS phenotypes. Whether one gene has more effect than the other or whether the effects are additive or multiplicative remains to be determined. Haploinsufficiency for other genes in the WBS critical region explains the vascular features and may contribute to the full syndrome, but it is not the main cause. Because adjacent duplicated but diverged genes are common in the human genome, such cumulative dosage effects may underlie the pathology of other chromosomal syndromes where members of different gene families are deleted or duplicated.

Supporting Online Material

Materials and Methods

Figs. S1 to S4


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article