A Novel Role for Transcription Factor Lmo4 in Thymus Development Through Genetic Interaction with Cited2

Deletion of the transcriptional modulator Cited2 in the mouse results in embryonic lethality, cardiovascular malformations, adrenal agenesis, cranial ganglia fusion, exencephaly, and left–right patterning defects, all seen with a varying degree of penetrance. The phenotypic heterogeneity, observed on different genetic backgrounds, indicates the existence of both genetic and environmental modifiers. Mice lacking the LIM domain-containing protein Lmo4 share specific phenotypes with Cited2 null embryos, such as embryonic lethality, cranial ganglia fusion, and exencephaly. These shared phenotypes suggested that Lmo4 may be a potential genetic modifier of the Cited2 phenotype. Examination of Lmo4-deficient embryos revealed partially penetrant cardiovascular malformations and hypoplastic thymus. Examination of Lmo4;Cited2 compound mutants indicated that there is a genetic interaction between Cited2 and Lmo4 in control of thymus development. Our data suggest that this may occur, in part, through control of expression of a common target gene, Tbx1, which is necessary for normal thymus development. Developmental Dynamics 239:1988–1994, 2010. © 2010 Wiley-Liss, Inc.

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 DNAbinding 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 À/À 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.

Cervical Vertebrae
Abnormalities in Cited2 2/2 Embryos Lmo4 À/À embryos show homeotic-like transformations in the rib cage and cervical vertebrae, although at low penetrance (Hahm et al., 2004). To investigate whether similar defects were present in the Cited2 À/À embryos, skeletal preparations were made at embryonic day (E) 17.5. This revealed that Cited2 À/À embryos (on a mixed genetic background) have fusions of the cervical vertebrae ( Fig.  1). The cervical vertebrae may be completely fused (Fig. 1c,d) from the exoccipital bone through to the atlas (C1), axis (C2), and the third cervical vertebra (C3). Less severe fusions are also observed, which may even be asymmetric (Fig. 1c). The cervical skeleton from one of the less severely affected Cited2 À/À embryos was disarticulated revealing an anterior transformation (Fig. 1g,h). The axis had a structure resembling the anterior arch of the atlas, and the atlas was missing this. The rib cages, however, were unaffected. Similar cervical vertebrae fusions were also observed in embryos on congenic 129Sv and C57Bl/6J genetic backgrounds (data not shown).

Cited2 and Lmo4 Genetically Interact In Vivo
As Cited2 and Lmo4 mutant mice share several specific phenotypes, we tested whether there was a genetic interaction in vivo by intercrossing Lmo4 þ/À mice with Cited2 þ/À mice (Supp . Table S1, which is available online). As Lmo4 þ/À ;Cited2 þ/À mice were viable, we intercrossed them to produce compound and double knockout mice. Mice null for either Cited2 or Lmo4 were not present at weaning, as expected (Supp . Table S2). Embryos from Lmo4 þ/À ;Cited2 þ/À intercrosses were collected at E15.5 and examined by MRI (n ¼ 141). Supplementary Table S3 and Supp. Figures S1 and S2 summarize the defects observed in these embryos, the genotypes of which were all present in normal Mendelian ratios at this point in gestation. We observed no significant differences in the incidence of cardiovascular malformations in the compound heterozygous/knockout or double knockout embryos. Spina bifida was occasionally observed in embryos of the same genotype that exhibited exencephaly (Supp. Fig. 1), a phenotype not described before for either Lmo4 or Cited2 mutant embryos.
The complete lack of thymus in Lmo4 À/À ;Cited2 À/À embryos is a novel phenotype, not previously seen in either individual knockout. This Skeletal defects in embryonic day (E) 17.5 Cited2 À/À embryos. a: Wild-type (Cited2 þ/þ ) embryo skeleton showing the normal appearance of the cervical vertebrae. b-d: Cited2 À/À skeletons showing various degrees of fusion between the cervical vertebrae (indicated). All embryos depicted had exencephaly and are missing the occipital and frontal skull bones. e-h: Disarticulated cervical vertebrae from a Cited À/À skeleton with less severe fusions. e: Wild-type atlas, with anterior arch indicated (arrowhead). f: Wild-type axis. g: Cited À/À atlas which is missing the anterior arch (asterisks). h: Cited2 À/À axis, which has a structure resembling the anterior arch of the atlas (arrowhead). Ex, exoccipital; C1, atlas; C2, axis. Scale bar ¼ 1 mm. Fig. 2. Cardiovascular defects observed in embryonic day (E) 15.5 Lmo4 À/À embryos. a,b: Transverse magnetic resonance imaging sections of wild-type (a) and Lmo4 À/À (b) embryo hearts. A ventricular septal defect (VSD) is visible in the Lmo4 À/À heart. c-f: Three-dimensional reconstructions of embryo hearts and great vessels. c,d: Ventral views of wild-type embryo heart (c) showing normal topology, and Lmo4 À/À embryo heart (d) showing a VSD and doubleoutlet right ventricle (DORV), where both the aorta (Ao) and ductus arteriosus (DA) arise from the right ventricle (RV). e,f: Anterior views of wild-type embryo heart (e) showing normal topology, and Lmo4 À/À embryo heart (f) showing a right-sided aortic arch and ductus arteriosus. RA, right atrium; LA, left atrium; LV, left ventricle; IVS, intra-ventricular septum; PA, pulmonary arteries; Tr, trachea. Scale bar ¼ 500 mm.
indicates that Cited2 and Lmo4 interact to control thymus development. To investigate whether there is a common target for Cited2 and Lmo4, embryos were harvested for quantitative reverse transcriptase-polymerase chain reaction (QRT-PCR) analysis. Pharyngeal arches and hearts were dissected from E10.5 Cited2 À/À and Lmo4 À/À embryos. Expression of Cited2 and Lmo4, as well as potential target candidate genes, Pitx2c and Tbx1, were investigated. Pitx2c was selected as it is a known target of Cited2 , and Tbx1 was selected as it has been implicated in thymus development (Jerome and Papaioannou, 2001).

DISCUSSION
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 singleintervention 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 Fig. 3. Thymus hypoplasia and malpositioning observed in embryonic day (E) 15.5 Lmo4 À/À embryos. a,b: Transverse magnetic resonance imaging (MRI) sections of the thymus of wildtype (Lmo4 þ/þ ; a) and Lmo4 À/À (b) littermate embryos. c,d: Three-dimensional (3D) reconstructions (ventral view) of thymus and trachea showing normal position and equal-sized lobes from wild-type (c) and hypoplasia, abnormal positioning and separation of the lobes in a Lmo4 À/À embryo (d). e: Thymus volumes from Lmo4 þ/þ and Lmo4 À/À embryos, corrected for embryo mass. f: Thymus volumes from Cited2 þ/þ and Cited2 À/À embryos, corrected for whole embryo volume. (Graph values are means 6 SD). Scale bar ¼ 500 mm. Three-dimensional (3D) reconstruction of thymuses from embryonic day (E) 15.5 embryos by MRI. a: Thymus and trachea from a wild-type (Lmo4 þ/þ ) embryo, showing normal position and equal-sized lobes. b: Lmo4 À/À ;Cited2 þ/À thymus illustrating hypoplasia, abnormal positioning, and separation of the lobes. c: Comparison of whole thymus volumes, adjusted for embryo weight, in embryos of all possible genotypes from Lmo4 þ/À ;Cited2 þ/À intercrosses (Mean 6 SEM). Lmo4 À/À ;Cited2 À/À embryos do not have a thymus (*). Scale bar ¼ 500 mm. 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 . 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 . 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 À/À ; Cited2 À/À 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 À/À ;Cited2 À/À 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 . 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. 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 À/À embryos had thymuses 0.46 times the size of wild-type thymuses, and Cited2 À/À thymuses were 0.78 times the size of wild-type thymuses, the expected thymus size of Lmo4 À/À ; Cited2 À/À 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 . 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 Fig. 5. mRNA expression in embryos lacking Cited2 or Lmo4. Quantitative reverse transcriptase-polymerase chain reaction was conducted on pharyngeal arches and hearts harvested from embryonic day (E) 10.5 Cited2 þ/þ , Cited2 À/À , Lmo4 þ/þ and Lmo4 À/À embryos. a: Lmo4 expression is significantly decreased in Cited2 À/À embryos. b: Cited2 expression was not affected in Lmo4 À/À embryos. c,d: Pitx2c expression was significantly decreased in Cited2 À/À embryos (c) but was not affected in Lmo4 À/À embryos (d). e,f: Tbx1 expression was reduced in Cited2 À/À embryos (e) and in Lmo4 À/À embryos (f). Mean 6 SEM are indicated. (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 À/À 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.

Mice
Cited2 þ/À mice (Cited2 tm1Bha ) on a mixed or congenic C57BL/6J background (Bamforth et al., 2001 and Lmo4 þ/À mice (gift from T. Rabbitts, LMB Cambridge; Tse et al., 2004) were used in this study. Embryos were harvested at the indicated time points after detection of a vaginal plug (E0.5), and genotyped using allele-specific PCR (primer details are available on request). All animal experimentation was performed under UK Home Office authorization and regulations.

Imaging
Magnetic resonance imaging was performed on a horizontal 9.4T/21cm VNMRS Direct Drive MR system (Varian Inc., Palo Alto) essentially as described previously  on embryos at E15.5. Skeletons were prepared and stained with alizarin red (bone) and Alcian blue (cartilage) by standard methods.

Statistical Analyses
The chi-squared test was used to calculate the deviation from expected Mendelian ratios. QRT-PCR data was analyzed using a two-tailed, two-sample t-test assuming unequal variance.