The transcription factor CITED2, which binds CREBBP with high affinity (Bhattacharya et al., ¹⁹⁹⁹), acts as a co-factor for transcription factors such as TFAP2, LHX2, PPARA, and SMAD2/3 (Glenn and Maurer, ¹⁹⁹⁹; Braganca et al., ²⁰⁰³; Tien et al., ²⁰⁰⁴; Chou and Yang, ²⁰⁰⁶), as well as acting as a repressor of hypoxia-activated transcription (Bhattacharya et al., ¹⁹⁹⁹). Mice null for Cited2 die in utero from phenotypically heterogeneous cardiac malformations including atrial, ventricular, and atrioventricular septal defects (ASD, VSD, AVSD); outflow tract defects (double-outlet right ventricle [DORV], common arterial trunk [CAT], tetralogy of Fallot [TOF], transposition of great arteries [TGA]); and interrupted and right-sided aortic arch (Bamforth et al., ²⁰⁰¹, ²⁰⁰⁴; Yin et al., ²⁰⁰²; Weninger et al., ²⁰⁰⁵). Loss of Cited2 also gives rise to exencephaly, cranial ganglia fusion, abnormal migration of cardiac neural crest cells, absent adrenal glands, and placental abnormalities (Bamforth et al., ²⁰⁰¹, ²⁰⁰⁴; Barbera et al., ²⁰⁰²; Yin et al., ²⁰⁰²; Weninger et al., ²⁰⁰⁵; Withington et al., ²⁰⁰⁶; Val et al., ²⁰⁰⁷). On a congenic C57Bl/6J background, Cited2-null mice show left–right patterning defects characterized by right atrial and pulmonary isomerism, and abnormal ventricular topology (Bamforth et al., ²⁰⁰⁴; Weninger et al., ²⁰⁰⁵). We have recently shown that the phenotypically heterogenous and penetrant cardiac malformations in Cited2 deficiency arise from a primary requirement in epiblast derivatives for left–right patterning, with a secondary cell-autonomous role in the mesoderm (MacDonald et al., ²⁰⁰⁸).

Cited2 binds the LIM homeodomain transcription factor, Lhx2, by means of LIM domains (Glenn and Maurer, ¹⁹⁹⁹). Proteins containing LIM domains function as transcriptional regulators, and this domain may act as a molecular adaptor for multiprotein complexes. The subfamily of LIM-only (LMO) proteins comprises four members (Lmo1–Lmo4). These non-DNA binding transcription factors may affect transcription by protein–protein interaction with DNA-binding transcription factors or chromatin modeling proteins (Hahm et al., ²⁰⁰⁴). Lmo4 comprises two zinc-binding LIM domains and is widely expressed in the developing embryo, including the pharyngeal arches, thymus, and neural crest (Grutz et al., ¹⁹⁹⁸; Kenny et al., ¹⁹⁹⁸; Sugihara et al., ¹⁹⁹⁸). Mouse embryos lacking Lmo4 display a phenotype that overlaps with Cited2^−/− embryos. This includes exencephaly, in utero or perinatal lethality, and cranial nerve fusions (Hahm et al., ²⁰⁰⁴; Tse et al., ²⁰⁰⁴; Lee et al., ²⁰⁰⁵).

The shared specific phenotypes observed in Cited2 and Lmo4 mutant mice suggest that there may be a genetic interaction. In this study, we show that embryos lacking Lmo4 have partially penetrant cardiac malformations including VSD, DORV, and right-sided aortic arch, which have not previously been reported. The shared specific phenotypes are further extended by demonstrating cervical fusions and abnormal thymus in Cited2 and in Lmo4 mutant embryos. The thymus is significantly smaller in compound mutant embryos, and completely absent in embryos lacking both Lmo4 and Cited2, indicating that there is a genetic interaction. Expression of Tbx1, which is important for thymus development, was significantly reduced in Lmo4 and Cited2 mutant embryos. This suggests a novel role for Cited2 and Lmo4 acting, likely in part through Tbx1, to control development of the thymus.

Genetic interactions can be identified through screens testing pairwise combinations of mutant genes that either suppress or enhance the original mutant phenotype (Hartman et al., ²⁰⁰¹). In pairwise intervention experiments, the expected outcome of a double-intervention is the product of the two single-intervention effects; departure from this model indicates a functional relationship (i.e., either an aggravating or an alleviating effect, indicating compensatory function or gene-action in series, respectively) between the two target genes (St Onge et al., ²⁰⁰⁷). This approach has been used in Drosophila melanogaster and Caenorhabditis elegans to identify genes acting in the same pathway (Lehner et al., ²⁰⁰⁶; Sambandan et al., ²⁰⁰⁸). Also, genes that share similar or identical GO term annotations are more likely to interact genetically, as shown by the analysis of thousands of mutant Saccharomyces cerevisiae yeast strains (Wong et al., ²⁰⁰⁴). This is also true for genes that encode proteins found in the same subcellular location, or that act in the same protein complex. Another excellent predictor of a genetic interaction is a shared specific mutant phenotype (Tong et al., ²⁰⁰⁴). Mouse embryos lacking Cited2 or Lmo4 share specific phenotypes, both displaying exencephaly and cardiovascular, thymus, skeletal, and cranial ganglia defects. Experiments to identify a genetic interaction revealed that Lmo4^−/−;Cited2^−/− embryos completely lacked a thymus, a novel phenotype, not observed in either parental line. However, there was no significant increase in penetrance of cardiac malformation in Lmo4^−/−;Cited2^−/− embryos.

As discussed by Wong et al., a genetic interaction between two genes is indicated when the phenotype observed is more pronounced than the phenotype produced by the genes individually (Wong et al., ²⁰⁰⁴). Having re-analyzed previously published studies, Mani et al. (²⁰⁰⁸) concluded that the expected phenotype in offspring should be the product of the parental fitness, and not the average of parental fitness or that of the least fit parent as previously thought (Tong et al., ²⁰⁰¹, ²⁰⁰⁴). This allows for greater expected deviation from parental fitness in offspring. For instance, if one parent had a fitness of 0.6 and one a fitness of 0.8, the expected fitness of the offspring is not 0.7 (average) or 0.6 (least fit parent), but 0.48 (product). Thus, applying this principle to these results, where Lmo4^−/− embryos had thymuses 0.46 times the size of wild-type thymuses, and Cited2^−/− thymuses were 0.78 times the size of wild-type thymuses, the expected thymus size of Lmo4^−/−;Cited2^−/− embryos should be 0.36 times the size of those in wild-type embryos. However, a total lack of thymus was discovered, dramatically exceeding the expected decrease. This indicates that a genetic interaction between Lmo4 and Cited2 is necessary for thymus development, suggesting that the genes act together in a common developmental process or molecular target.

Pitx2c is a known target of Cited2, but QRT-PCR data showed that it is not a target of Lmo4. Nor is Cited2 a target of Lmo4, although it appears that Lmo4 could be a target of Cited2. The data also revealed that Tbx1 expression levels are reduced in response to loss of both Cited2 and Lmo4 expression. This could indicate that Tbx1 is a common target of Cited2 and Lmo4, all of which are necessary for normal thymus formation. Indeed, Ivins et al. identified a reduction in Lmo4 expression in embryos lacking Tbx1 expression at E9.5 (Ivins et al., ²⁰⁰⁵). Tbx1 is expressed in the developing pharyngeal endoderm, which contributes to thymus formation. Absent thymus is a feature of DiGeorge / 22q11 deletion syndrome (Lischner, ¹⁹⁷²; de la Chapelle et al., ¹⁹⁸¹), for which the majority of phenotypes observed are largely attributed to haploinsufficiency of TBX1 during development (Chieffo et al., ¹⁹⁹⁷; Jerome and Papaioannou, ²⁰⁰¹; Lindsay and Baldini, ²⁰⁰¹). Moreover, in the mouse, morphology of the thymus is sensitive to Tbx1 dosage (Zhang and Baldini, ²⁰⁰⁷). Lmo4 is strongly expressed in the pharyngeal arches during mouse embryonic development (Hahm et al., ²⁰⁰⁴), specifically in the migratory cranial neural crest, and highly expressed in proliferating T lymphocytes in adult thymus (Kenny et al., ¹⁹⁹⁸). Cited2 is also strongly expressed in the pharyngeal arches during development, in endoderm, mesenchyme and ectoderm (Weninger et al., ²⁰⁰⁵; MacDonald et al., ²⁰⁰⁸), and it is the third pharyngeal pouch from which the thymus is derived. The expression of Tbx1 in the fourth pharyngeal arch is also important for normal formation of the fourth pharyngeal arch arteries (Lischner, ¹⁹⁷²; Jerome and Papaioannou, ²⁰⁰¹), the development of which are affected in embryos lacking Cited2 and/or Lmo4.

Our data show that Lmo4^−/− embryos have incomplete penetrance of cardiovascular defects in addition to those phenotypes already known, which has not been previously reported. Thymus defects have also not been previously reported and were detected by MRI, despite Tse et al. (²⁰⁰⁴) noting a normal cellular architecture of the thymus from histological sections, and no hypoplasia was recorded (Tse et al., ²⁰⁰⁴). Adding cardiovascular and thymus defects to the list of similarities between the recorded phenotypes of both Cited2 and Lmo4 knockout mice gave strong grounds for a physical and genetic interaction between Cited2 and Lmo4. Indeed, using GST pull-down assays, we were able to show that LMO4 physically interacts with residues 123–161 of CITED2, although we were unable to co-immunoprecipitate Lmo4 and Cited2 (data not shown).

To fully understand the genetic network in which Cited2, Lmo4, and Tbx1 are potentially interacting to control thymus development, further experiments will need to be performed. Although QRT-PCR data gives a quantitative measure of altered expression, qualitative assays will be needed to complement this, for example, by using in situ hybridization techniques to investigate whether the expression of Tbx1 is reduced specifically in the endoderm as may be expected. A more detailed investigation of candidate genes in wild-type and double heterozygote embryos and double homozygous null embryos will also need to be carried out, looking at the expression profile of genes specifically expressed in thymus such as Foxn1 (Nehls et al., ¹⁹⁹⁶), and those expressed in the pharyngeal arches and known to affect thymus development, for example, Pax9, Eya1, and Six1 (Peters et al., ¹⁹⁹⁸; Xu et al., ²⁰⁰²; Zou et al., ²⁰⁰⁶).

To summarize, in this study we have shown that mice lacking Lmo4 exhibit cardiovascular and thymus defects, which are novel phenotypes and add Lmo4 to the genetic network involved in cardiovascular development. Lmo4 interacts genetically with Cited2 in vivo to control thymus development, possibly through a common target gene in Tbx1.

We thank T. Rabbitts for Lmo4 mice. A.C.M is the recipient of a British Heart Foundation DPhil Studentship. S.B. is supported by a British Heart Foundation Chair Award [CH/09/003]. This work was funded by Wellcome Trust Program Grant Award [083228] to S.B., and by a Wellcome Trust Core Grant Award [075491/Z/04].

Additional Supporting Information may be found in the online version of this article.

Supp. Fig. S1. Neural tube, cleft palate, and cardiac defects observed in mutant embryos. Embryos at embryonic day (E) 15.5 generated by intercrossing Lmo4+/−;Cited2+/− mice were examined. a: Normal wild-type embryo. b: Lmo4−/−;Cited2+/− embryo with exencephaly (Ex) and spina bifida (SB). Scale bar = 1 mm. c–l: Two-dimensional images generated by magnetic resonance imaging. c,d: Sagittal sections through wild-type and Lmo4−/−;Cited2+/− embryos. c: Wild-type embryo with complete palate (P) over the tongue (T). d: Lmo4−/−;Cited2+/− embryo with cleft palate (cP) and edema (E). e,f: Coronal sections of a normal wild-type (e) and abnormal Lmo4−/−;Cited2−/− (f) heart. g,h: Coronal sections of organs from normal wild-type (g) and abnormal Lmo4+/−;Cited2−/− (h) embryo. The mutant displays right pulmonary isomerism and right-sided stomach. i–l: Transverse sections of hearts. i: Normal wild-type heart. j: Lmo4+/−;Cited2−/− heart with a common atrium (A) and a primum atrial septal defect (ASDP). A common atrioventricular valve (AVV) opens into both ventricles. k: Lmo4+/−;Cited2−/− heart with a common atrium (A) and heart apex pointing to the right. l: Lmo4−/−;Cited2+/+ heart positioned severely to the left. IVS, interventricular septum; LV, left ventricle; MV, mitral valve; PAS, primary atrial septum; RV, right ventricle; TV, tricuspid valve; WT, wild-type. Scale bar = 500 μm.

Supp. Fig. S2. Cardiovascular defects observed in mutant embryos. Embryos at embryonic day (E) 15.5, generated by intercrossing Lmo4+/−;Cited2+/− mice, were examined by magnetic resonance imaging (MRI), and three-dimensional (3D) reconstructions were made. a–f: Ventral views. a: Wild-type (WT) heart showing normal topology. b: Lmo4−/−;Cited2+/− heart showing a ventricular septal defect (VSD) and double outlet right ventricle (DORV). The ductus arteriosus is right-sided. c: Lmo4−/−;Cited2+/− heart showing right-sided aortic arch (RAA) and ductus arteriosus (RDA), and a mirror image arrangement of the pharyngeal arch arteries. The heart is twisted severely to the left. d: Lmo4+/−;Cited2−/− heart showing an interrupted aortic arch (IAA) and DORV, with an aberrant right subclavian artery (ARSA). e: Lmo4−/−;Cited2−/− heart with a common arterial trunk (CAT). f: Lmo4+/−;Cited2−/− heart with CAT, RAA, and VSD. g,h: Dorsal views. g: Wild-type (WT) heart showing normal topology of venous return to the heart. h: Lmo4+/−;Cited2−/− heart showing a common atrium (A) with bilateral systemic venous sinuses (LSVS, RSVS) into which drain bilateral superior and inferior caval veins. Scale bar = 500 μm.

Supporting Tables.