DiGeorge syndrome (DGS) is a congenital disease causing cardiac outflow tract anomalies, craniofacial dysmorphogenesis, thymus hypoplasia, and mental disorders. It results from defective development of neural crest cells (NCs) that colonize the pharyngeal arches and contribute to lower jaw, neck and heart tissues. Although TBX1 has been identified as the main gene accounting for the defects observed in human patients and mouse models, the molecular mechanisms underlying DGS etiology are poorly identified. The recent demonstrations that the SDF1/CXCR4 axis is implicated in NC chemotactic guidance and impaired in cortical interneurons of mouse DGS models prompted us to search for genetic interactions between Tbx1, Sdf1 (Cxcl12) and Cxcr4 in pharyngeal NCs and to investigate the effect of altering CXCR4 signaling on the ontogeny of their derivatives, which are affected in DGS. Here, we provide evidence that Cxcr4 and Sdf1 are genetically downstream of Tbx1 during pharyngeal NC development and that reduction of CXCR4 signaling causes misrouting of pharyngeal NCs in chick and dramatic morphological alterations in the mandibular skeleton, thymus and cranial sensory ganglia. Our results therefore support the possibility of a pivotal role for the SDF1/CXCR4 axis in DGS etiology.
DiGeorge or velocardiofacial syndrome (DGS) is a congenital disease characterized by cardiovascular defects, craniofacial anomalies, thymus hypoplasia, mental disorders and dysphagia (Shprintzen, 2008). It results primarily from defective development of cranial and cardiac neural crest cells (NCs) that invade the first four pharyngeal arches (PA1-4) and contribute to the formation of the lower jaw, the neck and the heart (Le Douarin and Kalcheim, 1999). In chick, ablation of this pharyngeal NC population phenocopies the cardiocraniofacial anomalies found in DGS patients (Kirby et al., 1983). DGS is caused by hemizygous deletions within chromosome 22q11.2, which have been reproduced on chromosome 16 in the Df1/+ and LgDel/+ mouse models (Puech et al., 1997). The Df1 and LgDel regions carry Tbx1, for which deletion and complementation studies in mouse indicate that it is responsible for many of the abnormalities observed (Lindsay et al., 2001). In human, TBX1 maps to the 22q11 region and its haploinsufficiency or sequence mutations have been found in DGS patients (Scambler, 2010). During mouse development, Tbx1 is first expressed in pharyngeal endoderm and ectoderm and then in the PA mesoderm (Garg et al., 2001; Zhang et al., 2005; Mesbah et al., 2012). However, though not expressed by pharyngeal NCs, loss of Tbx1 impairs their migration and differentiation (Kochilas et al., 2002; Vitelli et al., 2002; Calmont et al., 2009).
Malformations in DGS patients are highly variable, suggesting that complex genetic interactions are involved (Shprintzen, 2008; Scambler, 2010). Despite progress in the identification of Tbx1 targets, the molecular mechanisms involved in pharyngeal NC development that could account for DGS etiology have not been completely elucidated. Interestingly, signaling by CXCR4, a receptor for the chemokine stromal derived factor 1 (SDF1, also named CXCL12), is deficient in brain cortical interneurons of mouse DGS models, resulting in their abnormal migration, probably accounting for the mental disorders observed in DGS patients (Meechan et al., 2012; Toritsuka et al., 2013). Loss of Cxcr4 or Sdf1 in mouse causes cardiac anomalies similar to those observed in DGS (Ma et al., 1998) and disruption of SDF1/CXCR4 signaling in chick demonstrated that these cardiac defects result from abnormal migration of cardiac NCs (Escot et al., 2013). In addition, the Cxcr4 promoter harbors putative TBX1 binding sites (Castellanos et al., 2014). Together, these observations raise the possibility that, in addition to cortical interneurons in the brain, CXCR4 signaling could regulate pharyngeal NC migration and that its misregulation could ultimately cause many of the cardiocraniofacial defects associated with DGS.
RESULTS AND DISCUSSION
Sdf1 and Cxcr4 expression levels are diminished in Tbx1 mutant mice
To investigate whether Sdf1 and Cxcr4 might be Tbx1 targets, we analyzed their expression profiles in mouse embryos at the time of pharyngeal NC migration, by in situ hybridization and using a transgenic Cxcr4-EGFP mouse line (Tran et al., 2007). Analysis of whole embryos (Fig. 1Aa-Ad) and cross-sections through PA2 (Fig. 1Ae-Ah) at embryonic day (E) 9.5 showed Cxcr4 expression in streams of cells along the ectoderm identified as migrating NCs by the marker Sox10. Sdf1 was conspicuous in the ectoderm and pharyngeal endoderm in proximity to the migrating NCs. Interestingly, Tbx1 was expressed in the same regions of pharyngeal endoderm and lateral ectoderm as Sdf1 (Fig. 1Ad,Ah). At E10.5 (Fig. 1Ba-Bd), Cxcr4 expression decreased gradually in NCs occupying the PA and became restricted to some blood vessels, including aortic arches (Fig. S1), whereas Sdf1 and Tbx1 expression shifted to the pharyngeal mesoderm, where they partially overlapped in the mesodermal core. This indicates that Tbx1 and Sdf1 expression patterns overlap transiently in the ectoderm and pharyngeal endoderm during pharyngeal NC migration. To address the question of the Tbx1 regulation of SDF1 signaling, Sdf1 and Cxcr4 expressions were studied in Tbx1 mouse mutants (Fig. 1C). Analysis of whole embryos (Fig. 1Ca-Cd) and cross-sections through the PA (Fig. 1Ce,Cf) at E9.5 showed a strong reduction of both Sdf1 and Cxcr4 expression in the PA of Tbx1 mutants compared with wild-type embryos. This result was confirmed by the quantification of Cxcr4 and Sdf1 messages in Tbx1 mouse mutants, which showed a decrease of nearly 50% in the transcriptional activity of both genes in the Tbx1−/− mouse, compared with their wild-type or heterozygote littermates (Fig. 1Cg). Thus, our data reveal that in mouse, both Cxcr4 and Sdf1 are downstream of Tbx1 in the genetic cascade regulating PA development.
Sdf1 and Cxcr4 expression patterns correlate with pharyngeal NC migration to PAs in chick
We turned to the chick model to determine whether CXCR4 signaling might be involved in pharyngeal NC development. Indeed, NC migration and differentiation have been thoroughly examined in chick, as in mouse, and pharyngeal NCs can be precisely targeted using electroporation of the corresponding neural tube level, in contrast to mouse for which NC promoters available for conditional knockouts do not specifically target the pharyngeal NCs. We showed previously that cardiac NCs migrating towards PA3-4 express Cxcr4 and that Sdf1 exhibits a complementary pattern in the ectoderm along their migratory route (Escot et al., 2013). We therefore investigated the expression patterns of Sdf1 and Cxcr4 in relation to Tbx1 in pharyngeal NCs populating PA1-2 (Fig. 2A; Le Douarin and Kalcheim, 1999). Immunohistology and in situ hybridization analyses revealed that NCs migrating into PA1-2 (Fig. 2B,C; Fig. S2) expressed Cxcr4 in proximity to Sdf1-expressing ectodermal and pharyngeal endodermal cells, exactly as observed for those invading PA3-4. At this level, Sdf1 and Tbx1 expression patterns overlapped in both tissues. At E2.5, once NCs had colonized the PAs, they gradually downregulated Cxcr4 expression, concommitant with Sdf1 and Tbx1 expression being shifted from the ectoderm to the pharyngeal mesoderm (Fig. 2D). From E3.5, Cxcr4 became restricted to endothelia within the aortic arches (Fig. 2E; Fig. S1). Thus, in contrast to other NC populations, e.g. at truncal levels (Fig. S3) or along the digestive tract (Escot et al., 2013), and similarly to mouse, chick pharyngeal NCs share in common a conspicuous Cxcr4 expression throughout their migration along paths lined by Sdf1- and Tbx1-expressing cells, making SDF1 signaling a potential candidate for the NC anomalies observed in DGS patients.
Disruption of CXCR4 signals in chick pharyngeal NCs causes craniofacial, glandular and neural defects
To determine whether defects in CXCR4 signaling might be responsible for the NC anomalies in DGS, CXCR4 activity was inhibited in pharyngeal NCs in chick and the consequences on the formation of their major derivatives analyzed (Fig. 3A). We previously reported that disrupting this signaling pathway in cardiac NCs results in cardiac defects similar to those found in DGS (Escot et al., 2013). We therefore investigated the other tissues known to be affected in this pathology. Using in ovo electroporation, we expressed in pharyngeal NCs either a miRNA against Cxcr4 (miRNA-Cxcr4, Fig. S5) or a dominant-negative form of Cxcr4 (DN-CXCR4) lacking the C-terminus sequences necessary for chemotactic migration (Ahr et al., 2005; Cronshaw et al., 2010). First, we searched for craniofacial dysmorphogenesis due to palate and jaw anomalies, a prominent trait of DGS patients (Fukui et al., 2000; Karpinski et al., 2014). Heads of chick embryos were collected at E9.5 and the jaw anatomy examined. All controls exhibited normal heads (n=18), whereas most embryos in which CXCR4 signaling was affected exhibited anomalies of the lower jaw and hyoid bone, with frequently shortened or absent basibranchial, epibranchial and ceratobranchial bones and Meckel's cartilage, all derived from NCs that populate PA2-3 (Fig. 3A,B; Table S1; n=16). As described previously for the cardiac defects (Escot et al., 2013), jaw anomalies were more frequent in DN-CXCR4-electroporated embryos than in miRNA-Cxcr4 ones (Table S1), owing to the earlier effect of DN-CXCR4 on NC migration (Escot et al., 2013). For this reason, we focused our following phenotypic analyses on DN-CXCR4-electroporated embryos.
One of the NC-related anomalies found in DGS patients is thymus hypoplasia, which is also the origin of immune disorders (Shprintzen, 2008). Indeed, NCs migrating to PA4 contribute to connective tissues and pericytes of the thymus (Fig. 3A), and are essential for the development of the thymic epithelium (Kuratani and Bockman, 1990). Analysis of embryos at E7.5, when the thymus anlage becomes recognizable, revealed smaller thymus primordia in DN-CXCR4-electroporated animals than in controls. The mean surface area of the thymus on transverse sections was reduced from 30% in the milder phenotypes up to 60% in the most severe cases (Fig. 3C; n=9).
Finally, disrupted cranial nerves have been reported in the LgDel mouse, possibly accounting for the dysphagia observed in children diagnosed with DGS (Karpinski et al., 2014). Pharyngeal NCs contribute to some of the major cranial sensory ganglia, i.e. V (trigeminal), VII-VIII (facial and vestibuloacoustic) and IX-X (glossopharyngeal and vagal) (Ayer-Le Lievre and Le Douarin, 1982) (Fig. 3A). In situ hybridization for Sox10 in whole embryos showed considerable mispatterning of cranial sensory ganglia in DN-CXCR4-embryos at E3.5, with some of them partially fused and others atrophied or missing (Fig. 3Da,Db; n=11). These abnormalities resulted at later stages in a variety of phenotypes, characterized by missing, fused or misrouted cranial nerves (Fig. 3Dc-Df; n=8).
Taken together, our results demonstrate that disrupting CXCR4 signals in pharyngeal NCs induces a large spectrum of abnormalities in their skeletal, neuronal and glandular derivatives, which, in addition to the cardiac defects already reported (Escot et al., 2013), recapitulate the defects described in mouse models and DGS patients (Funato et al., 2015; Karpinski et al., 2014; Vitelli et al., 2002).
Craniofacial and neural anomalies after disruption of CXCR4 signaling result from pharyngeal NC misrouting
We then investigated how and when anomalies found in pharyngeal NC derivatives upon disruption of CXCR4 signals were generated. Analysis of NC distribution at E3.5 revealed that in DN-CXCR4-electroporated embryos, NCs initiated migration from the neural tube, but very few reached PA2-4, whereas control embryos showed massive NC colonization (Fig. 4Aa-Ae). At E4.5, in DN-CXCR4-electroporated embryos, PAs remained almost devoid of NCs (Fig. 4Af-Ah; n=8), although apoptosis was not increased during migration (Fig. 4B; n=5), thereby revealing that the colonization defect did not result from cell death or delayed migration. Interestingly, at both at E3.5 (n=6) and E4.5 (n=8) numerous DN-CXCR4-positive NCs were detected in cranial ganglia VII/VIII and IX, which are normally not colonized by CXCR4-expressing NCs (Fig. 4C). This indicates that NC depletion in PAs of DN-CXCR4-expressing embryos resulted from misrouted NC migration, consecutive to the inhibition of CXCR4 chemotactic activity. To validate this interpretation, we analyzed the migratory response of pharyngeal NCs to an ectopic source of Sdf1. When misexpressed in the neural tube, Sdf1 caused misrouting of pharyngeal NCs, resulting in their massive accumulation along the neural tube and their depletion from PAs (Fig. S4). These results also provide an explanation for why cranial sensory ganglia are affected in DN-CXCR4-embryos although they do not rely on SDF1/CXCR4 signaling for their formation.
In conclusion, our study suggests a model in which Tbx1 expressed in the lateral ectoderm and pharyngeal endoderm at early stages of pharyngeal NC migration might regulate Sdf1 expression and consequently the CXCR4-mediated response to SDF1 signals in pharyngeal NCs, thereby controlling their chemotactic guidance towards the PAs. Inhibition of CXCR4 activity results in major defects in the cardiocraniofacial derivatives of pharyngeal NCs that phenocopy most anomalies of the Tbx1 mouse mutant and congenital malformations observed in DGS (Fig. 4D). Taken together with the recent results showing functional defects of CXCR4 signaling in the brain of the LgDel and Df1 mouse models (Meechan et al., 2012; Toritsuka et al., 2013), our study underlines the possibility of a pivotal role for the SDF1/CXCR4 axis in DGS etiology.
MATERIALS AND METHODS
Fertilized chick eggs were from commercial sources and green fluorescent protein positive (GFP+) transgenic chicken eggs (McGrew et al., 2008) from the Roslin Institute (University of Edinburgh, UK). Embryos were staged according to embryonic day of development. The transgenic mouse line [strain Tg(Cxcr4-ECFP)CD73Gsat/Mmucd, developed by the GENSAT project; NINDS contract N01Nso2331 to Rockefeller University, NY, USA] was purchased from Mutant Mouse Resource Research Centers. Mice carrying the Tbx1+/− allele were used and genotyped as described previously (Jerome and Papaioannou, 2001). All mutants were maintained in a predominantly C57BL/6 genetic background. Mouse care and procedures were in accordance with national and European laws.
Chick embryo manipulations
Grafting and in ovo electroporation experiments were performed as described (Escot et al., 2013). For grafting, hindbrains from GFP+ transgenic chick donors at the 5-somite stage (30 h of development) were excised after enzymatic treatment and transplanted isotopically and isochronically into chick hosts in ovo, in which their equivalents have been removed surgically. For electroporations, expression plasmids described previously (Escot et al., 2013; supplementary materials and methods) were mixed with plasmids encoding GFP or RFP and microinjected into the neural tube lumen of 5-somite-stage embryos at the midbrain and hindbrain levels. Electrodes (CUY610 platinum-coated, Sonidel, Ireland) were applied onto the vitelline membrane on either side of the appropriate neural tube region and five square pulses (20 V, 50 ms length, 500 ms gap) were delivered. For the analysis of pharyngeal NC derivatives, pulses were delivered bilaterally to electroporate both sides of the neural tube. Embryos were re-incubated for up to 8 days after electroporation, harvested in PBS and monitored for GFP or RFP fluorescence at the appropriate axial levels. The efficiency of the transfections was verified on whole-mount samples or sections by in situ hybridization for Sdf1 and immunofluorescence labeling for CXCR4.
Immunostaining, in situ hybridization and histological analyses
In situ hybridizations and immunostainings on whole embryos or sections were performed as described (Escot et al., 2013), using the following riboprobes: Sox10 (P. Scotting, Queen's Medical Center, Nottingham), Sdf-1 and Cxcr4 (both produced from EST clones) (Escot et al., 2013), mouse Tbx1 (V. Papaioannou, Columbia University Medical Center, New York), chick Tbx1 (produced by PCR from the mouse Tbox domain conserved in chick) (Garg et al., 2001) and antibodies: anti-cleaved caspase-3 (1:100; 9664, Cell Signaling), anti-NC1/HNK1 (1:10; Tucker et al., 1984), anti-TuJ1 (1:300; 1637, Chemicon), anti-GFP (1:500; 1814460, Roche) and anti-chicken CXCR4 (1:1000; Escot et al., 2013). Nuclei were visualized with DAPI (Sigma). For craniofacial skeleton analyses, heads were fixed in acetic acid and skeletons visualized using Alcian Blue and Alizarin Red.
The pharyngeal region of E9.5 embryos were dissected out from wild-type, Tbx1+/− and Tbx1−/− mutant mice. RT-qPCR was performed as described (Odelin et al., 2014) using the following primers: 5′-CCCCACAACTCTTCCATTCT-3′, 5′-GCAGGAGTGATAGGGGTCAT-3′ for Tbx1; 5′-TTGCCGACTATGCCAGTCAA-3′, 5′-CCGGGATGAAAACGTCCAT-3′ for Cxcr4; and 5′-CACTCCAAACTGTGCCCTTCA-3′, 5′-CACTTTAGCTTCGGGTCAATGC-3′ for Sdf1. RT-qPCR was performed using LightCycler 480 SYBR Green I Master mix (Roche) and cDNA synthesized using first-strand cDNA synthesis kit (Roche) from embryonic RNAs isolated using the NucleoSpin RNA/Protein Kit (Macherey-Nagel). Expression levels between wild-type (n=8), Tbx1+/− (n=6) and Tbx1−/− (n=6) embryos were calculated by the comparative cycle threshold (ΔΔCT) method. Each experiment was performed in triplicate for each genotype and normalized to endogenous Tbp housekeeping gene. Normalized expression levels in the control (Tbx1+/+) were set to 1.0 for each gene. Student's t-test was used to determine statistical significance.
Apoptosis quantitation was performed as described (Escot et al., 2013). The size of thymic primordia was determined using the Nikon NIS Elements imaging software. The outline of the primordium was delineated on each section to calculate its area. The total primordium area in each embryo was calculated by summing the areas of the primordium on each section. Statistical significances were evaluated using Student's t-test.
We thank S. Gournet for illustrations; our colleagues for probes and antibodies; and the GFP+ transgenic egg resource at the Roslin Institute.
S.E. performed experiments and analyzed data. C.B. performed experiments. E.F. and S.Z. performed the analyses on mouse embryos. C.F.-T. and J.-L.D. designed experiments and wrote the manuscript.
This work was supported by the French National Centre for Scientific Research (CNRS); the Université Pierre et Marie Curie; and the Association Française contre les myopathies [grant no. 11,405]. S.E. was a recipient of a doctoral fellowship from the Ministère de l'Enseignement Supérieur et de la Recherche.
The authors declare no competing or financial interests.