Conditional and constitutive expression of a Tbx1-GFP fusion protein in mice

Background Velo-cardio-facial syndrome/DiGeorge syndrome (VCFS/DGS) is caused by a 1.5-3 Mb microdeletion of chromosome 22q11.2, frequently referred to as 22q11.2 deletion syndrome (22q11DS). This region includes TBX1, a T-box transcription factor gene that contributes to the etiology of 22q11DS. The requirement for TBX1 in mammalian development is dosage-sensitive, such that loss-of-function (LOF) and gain-of-function (GOF) of TBX1 in both mice and humans results in disease relevant congenital malformations. Results To further gain insight into the role of Tbx1 in development, we have targeted the Rosa26 locus to generate a new GOF mouse model in which a Tbx1-GFP fusion protein is expressed conditionally using the Cre/LoxP system. Tbx1-GFP expression is driven by the endogenous Rosa26 promoter resulting in ectopic and persistent expression. Tbx1 is pivotal for proper ear and heart development; ectopic activation of Tbx1-GFP in the otic vesicle by Pax2-Cre and Foxg1-Cre represses neurogenesis and produces morphological defects of the inner ear. Overexpression of a single copy of Tbx1-GFP using Tbx1Cre/+ was viable, while overexpression of both copies resulted in neonatal lethality with cardiac outflow tract defects. We have partially rescued inner ear and heart anomalies in Tbx1Cre/- null embryos by expression of Tbx1-GFP. Conclusions We have generated a new mouse model to conditionally overexpress a GFP-tagged Tbx1 protein in vivo. This provides a useful tool to investigate in vivo direct downstream targets and protein binding partners of Tbx1.


Background
Velo-cardio-facial syndrome/DiGeorge syndrome (VCFS/ DGS), also known as 22q11.2 deletion syndrome (22q11DS), is the most common microdeletion syndrome occurring de novo in approximately 1/4,000 live births [1]. TBX1, encoding a T-box transcription factor, is located within the 1.5 Mb critically deleted region and haploinsufficiency of this gene is responsible for the congenital defects associated with 22q11DS [2]. Tbx1 null mutant mice have malformations of the heart, thymus/parathyroid, craniofacial region, and ear that are similar but more severe, to what is typically found in 22q11DS patients [2][3][4]. Previously generated BAC316.23 transgenic mice expressing 8-10 copies of human TBX1 also have developmental defects of the same tissues affected in Tbx1 −/− null mice, resembling clinical features of a recently identified 22q11 duplication syndrome [5][6][7][8][9]. Ablation studies using Tbx1 flox conditional null mutants have proven to be indispensible for understanding the tissue and cell-type specific roles of the Tbx1 gene [10,11]. COET, a transgenic mouse line with conditional overexpression of Tbx1, has previously been reported [12]. However, the COET mouse line does not contain a protein affinity tag on the Tbx1 protein. We have generated a conditional allele targeted at the Rosa26 locus for the purpose of expressing additional copies of Tbx1 fused to a GFP reporter in a tissue-specific manner. We chose a GFP tag in part to detect live GFP as a readout for activation of Tbx1-GFP in vivo and to use it for future biochemical or chromatin immunoprecipitation experiments. Tbx1 mutant rescue experiments demonstrate that Tbx1-GFP fusion protein functions in vivo in a manner similar to that of endogenous Tbx1. The Tbx1-GFP mouse allele provides a new gain-of-function model that is targeted to the constitutively active Rosa26 locus; targeting Tbx1-GFP to the endogenous Tbx1 locus would simultaneously disrupt the normal function of the targeted allele (as is the case for Tbx1 Cre/+ mice, Huynh et al., 2007). As such, our model works to activate 1-2 ectopic copies of Tbx1 while simultaneously tracing cells with GFP, and should therefore be of high value in future experimental studies.

Results and discussion
To generate the Tbx1-GFP allele, we first generated a bacterial plasmid that expresses a Tbx1-GFP fusion protein of approximately 75 kilodaltons (kD) ( Figure 1A), corresponding to the combined mass of Tbx1 (~50 kD) and GFP (~27 kD) ( Figure 1B). The Tbx1-GFP fragment was inserted downstream of a triple polyadenylation (tpA) transcriptional stop sequence that is flanked by LoxP sites in the pBigT vector [13]. The loxP-tpA-loxP-Tbx1-GFP element was cloned into the pROSA26PA vector for targeting to the Rosa26 locus [13,14], and electroporated into WW6 (129/SvJ) mouse embryonic stem cells (ESC). We identified 5/56 correctly targeted ESC clones by Southern blot analysis ( Figure 1C). Clone 45 was chosen for blastocyst injection in C57BL6 females to generate chimeras for germline transmission.
Since inner ear and cardiac defects are two of the most prominent features in the Tbx1 null mutant, we decided to examine those structures in more detail in the Pax2-Cre;Tbx1-GFP flox/+ mutants and Foxg1-Cre;Tbx1-GFP flox/+ mutants, respectively. Foxg1-Cre and Pax2-Cre are both strongly expressed throughout the OV, from which the inner ear forms. Both loss or gain of Tbx1 function in the OV disrupts inner ear morphogenesis [5,18,19]. Paintfilling of the inner ear in Foxg1-Cre;Tbx1-GFP flox/+ Figure 1 Generation of the Tbx1-GFP mouse. (A) Tbx1-GFP was inserted into the multiple cloning site (MCS) of pBigT downstream of a triple polyadenylation sequence (tpA) that is flanked by LoxP sites (triangles). Additional subcloning into the pROSA26PA plasmid resulted in the final targeting construct. 5' and 3' homology arms mediate recombination with the endogenous Rosa26 locus. (B) Whole cell lysates from COS7 cells transfected (T) with Tbx1-GFP plasmid or untransfected (UT). The fusion protein is ≈75 kD and is detected by anti-GFP and anti-Tbx1 on Western blot. (C) Southern blot was used to screen ESC clones. Genomic DNA was linearized with EcoRV and hybridized with a radiolabeled probe that binds upstream of the 5' homology arm. Two of five positive clones are shown. The correctly targeted allele is 4,071 bp and the wildtype allele is 11,517 kb. and Pax2-Cre;Tbx1-GFP flox/+ mutants at E14.5 shows an enlarged endolymphatic duct (ED) and common crus (CC) that joins the anterior and posterior semicircular canals (SCC) ( Figure 2B). The utricle, saccule, and lateral SCC are missing. The cochlea is hypoplastic with varying degree, but enlarged in two instances ( Figure 2B). This phenotype is very similar to that of BAC316.23 transgenic mice [5]; with the exception that BAC316.23 mice still possess the lateral SCC, and in most cases the saccule and utricle are present, albeit malformed ( Figure 2B).
Tbx1 is also known to repress neurogenesis of the VIIIth cranial ganglion that innervates the inner ear [19]. Here we show that in Foxg1-Cre;Tbx1-GFP flox/+ mutants, expression of Neurogenic differentiation factor 1 (NeuroD) is nearly abolished in the VIIIth cranial ganglion, and highly attenuated in cranial ganglia V, VII, IX, and X ( Figure 2C). Immunohistochemistry with anti-Neurofilament confirmed loss of the VIIIth cranial ganglion and abnormal size and misguided projections from other cranial ganglia in Foxg1-Cre;Tbx1-GFP flox/+ mutants ( Figure 2C). In addition, we found that NeuroD expression is completely missing from the olfactory placode of Foxg1-Cre;Tbx1-GFP flox/+ mutants; a region where Foxg1-Cre is active and there is no Tbx1 expression in a wild type context suggesting that gain of function of Tbx1 in the olfactory placode leads to ectopic repression of NeuroD. These findings demonstrate that Tbx1 could act directly or indirectly on neurogenic factors that are common to both otic and olfactory placodes.
Examination of SCC defects in the Pax2-Cre;Tbx1-GFP flox/+ mutants using paintfilling revealed a delay in SCC formation at earlier developmental stages ( Figure 3A). In the wild type, clearing of the fusion plates is visible by E12.25 and complete by E12.5 [20]. In Pax2-Cre;Tbx1-GFP flox/+ mutants, the fusion plates have not yet joined at E12.25 although they appear to partially recover by E14.5 ( Figure 3A). Histological sections at E12.25 confirm that the fusion plates in Pax2-Cre;Tbx1-GFP flox/+ mutants remain attached to the periotic mesenchyme compared to control littermates ( Figure 3A). We examined expression of laminin and netrin-1, a secreted protein that is related to laminins and promotes basal lamina breakdown [21,22]. We observed decreased laminin protein levels along the otic epithelium, especially along the site of lateral SCC formation along with a corresponding decrease in netrin-1 mRNA expression along the fusion plates in Pax2-Cre; Tbx1-GFP flox/+ mutants ( Figure 3B). It is possible that Tbx1 may act to repress netrin-1, thereby modulating cell adhesion properties. Complementation of these results occurred in Tbx1 −/− null mice in which expression of netrin-1 is expanded in the mesenchyme surrounding the OV ( Figure 3C). We did not observe expanded netrin-1 expression in the OV of Tbx1 −/− null mice, possibly due to the severity of the OV phenotype and subsequent absence of vestibular structures.
To test whether Tbx1-GFP could rescue inner ear defects that occur in Tbx1 Cre/null embryos, we crossed Tbx1-GFP flox/flox mice to Tbx1 Cre/+ mice [23]. Tbx1 Cre/mice have been shown to recapitulate the Tbx1 −/− null phenotype [23]. When Cre is expressed, cells from the Tbx1 Cre/+ expressing domain are positive for GFP in a manner that represents the Tbx1 lineage in the OV in embryos ( Figure 4A). Expression of Tbx1-GFP under the Rosa26 promoter alone is not sufficient for direct visualization of natural GFP, therefore Tbx1-GFP protein was detected using an antibody to GFP. Since expression of Tbx1-GFP is persistently driven by the endogenous Rosa26 promoter, we expect the domain of activated Tbx1-GFP to be more extensive than native Tbx1 expression ( Figure 4A). Immunofluorescence with anti-GFP antiserum demonstrated that there is no Tbx1-GFP protein in the absence of Cre expression in the OV ( Figure 4A). Tbx1 Cre/+ ;Tbx1-GFP flox/+ mutants did not exhibit gross morphological defects of the inner ear based on paintfilling ( Figure 4B), perhaps because one allele of Tbx1 is removed by knock-in of Cre. Paintfilling confirmed that the inner ears of Tbx1 −/− null mice lack all discernible vestibular and auditory structures with the exception of what appears to be an enlarged ED ( Figure 4C). When a single allele of Tbx1-GFP is activated in a Tbx1 Cre/background, there is rescue of the anterior and posterior SCCs in all embryos, and partial rescue of the saccule and cochlea (n=6) ( Figure 4C). Complete rescue of the null phenotype is not expected with constitutive activation of Tbx1-GFP because temporal regulation of Tbx1 is required for normal inner ear development.
To do this, we performed western blot on tissue samples ( Figure 5D) and observed that the level of Tbx1-GFP protein expressed from a single allele was higher than that observed for endogenous wildtype Tbx1. However, this may reflect the larger domain of expression of Tbx1-GFP within the cell lineage that is labeled by Tbx1 Cre/+ . Next, we wanted to test whether Tbx1-GFP could also rescue the cardiac defects that occur in Tbx1 Cre/embryos ( Figure 6A), all of which have PTA and VSD (n=3) [12]. Analysis of Tbx1 Cre/-;Tbx1GFP flox/+ embryos at E14.5 showed a complete rescue of outflow tract septation defects (n=8), although 50% still had a misalignment of the outflow tract resulting in a DORV ( Figure 6B). VSD was observed in all Tbx1 Cre/-;Tbx1GFP flox/+ embryos. It is not clear if these defects are due to failure to rescue the Tbx1 null phenotype, or if they are due to overexpression of Tbx1-GFP protein ( Figure 5D), even though protein expression levels as detected by western blot do not provide information about the functional activity levels of Tbx1-GFP protein. It is also possible that failure of complete rescue is due to the fact that Tbx1-GFP is expressed constitutively under the control of the Rosa26 promoter and is therefore not subject to temporal regulation. Activation of both Tbx1-GFP alleles in Tbx1 Cre/-; Tbx1GFP flox/flox embryos resulted in more severe heart phenotypes including PTA in 5/6 embryos in addition to VSD ( Figure 6C) due to further Tbx1 gene dosage imbalance. The expected size of endogenous Tbx1 is ≈50 kD while the Tbx1-GFP fusion protein is ≈75kD. β-actin is detected at ≈40 kD.

Conclusions
In summary, we have generated a new mouse model that conditionally expresses a Tbx1-GFP fusion protein. We show that Tbx1-GFP does not act in a dominantnegative manner and can functionally substitute for endogenous Tbx1 during heart and inner ear development by partial rescue of heart and inner ear morphological defects in a Tbx1 Cre/null background. We cannot completely rule out deleterious effects of the GFP tag protein on the ability of Tbx1 to act as a transcriptional regulator either via direct binding of DNA or via protein: protein interactions as we have not directly tested this. However, ectopic activation of Tbx1-GFP by Foxg1-Cre and Pax2-Cre causes morphological defects of the inner ear that resemble those occurring in BAC316.23 transgenic mice. Taken together with the partial rescue of heart and inner ear defects on a Tbx1 −/− null background, it seems likely that the Tbx1-GFP fusion protein retains functional characteristics of the endogenous Tbx1 protein.
As such, the Tbx1-GFP mouse line provides a genetic tool that is amenable to probing direct downstream targets of Tbx1 and identifying protein-protein interacting partners of Tbx1 in vivo.

Targeting vector
A pTbx1-EGFP-N1 plasmid was generated by cloning Tbx1 cDNA from a Tbx1-TOPO plasmid into pEGFP-N1 (Clontech 6085-1). The following oligonucleotide adaptor sequences were used for an in-frame ligation of the C-terminus of Tbx1 to the N-terminus of EGFP, simultaneously changing the Tbx1 stop codon from TAG to TTG and introducing NotI and BamHI restriction enzyme sites: 5'-GGCCGCGCCGCCCGGTGCCTACGAC TACTGCCCCAGATTG-3'; 5'GATCCAATCTGGGGC AGTAGTCGTAGGCACCGGGCGGCGC-3'. Adaptors were annealed by combining 10 nmol of each oligonucleotide in 200 μL of water and heating to 90°C followed by cooling at room temperature to 40°C. For 20 μL ligation reactions (Roche 1243292, Roche 10799009001), 5 μL of 1:500 diluted annealed adaptors were used. The sequence encoding the Tbx1-GFP fusion protein was inserted into the multiple cloning site of pBigT [13]. The loxP-tpA-loxP-Tbx1-GFP fragment was excised by restriction enzyme digestion with PacI and AscI then cloned into the pROSA26PA plasmid [13] to generate the final targeting vector (Figure 1). The final targeting vector was linearized with BbvCI and purified by phenol chloroform extraction, then electroporated into WW6 ESC and selection performed with G418. Positively selected ESC clones were plated in duplicate for DNA isolation using the DNeasy Blood and Tissue Kit (Qiagen 69506). DNA was digested overnight at 37°C with EcoRV and Southern blot performed with a previously described 5' probe [13].

Mouse models
Tbx1-GFP mice were genotyped with the FastStart High Fidelity PCR System (Roche 03553361001) and the following primers: TB3F (5'-CTGCACCACCATC CCTACAA-3') and GFPR (5'-TGAACTTCAGGGTC AGCTTG-3') for a 421 bp product from the targeted allele, and RO1F (5'-GCAATACCTTTCTGGGAGTT-3') and GFP-wtR (5'-CAATGCTCTGTCTAGGGG TT-3') for a 605 bp product from the wildtype allele. Foxg1-Cre, Pax2-Cre, and Tbx1 Cre/+ mice were kindly provided by Drs. Jean Hebert, Andrew K. Groves, and Antonio Baldini, respectively.. Embryos were dissected according to date of vaginal plug (E0.5). Embryonic stages <E11.5 were confirmed by counting pairs of somites. Animals were maintained in a 12 hr dark/12 hr light cycle in compliance with the Albert Einstein College of Medicine of Yeshiva University Institutional Animal Care and Use Committee (IACUC). Tbx1-GFP mice will become available to the research community upon acceptance of the manuscript.
Primer sequences introduced T7 RNA polymerase binding sites for generating antisense probes and T3 RNA polymerase binding sites for generating sense probes. For wholemount ISH, we rehydrated embryos to PBT and digested them with Proteinase K. This was followed by washes in Glycine solution and PBT. Embryos were then fixed in 4% PFA/0.2% gluteraldehyde for 15 min on ice followed by PBT washes. Embryos were incubated with RNA probes in hybridization buffer overnight at 70°C then washed in a series of SSC, maleic acid buffer, and PBT washes. Then embryos were incubated overnight at 4°C in antibody buffer with 1:10,000 dilution of anti-Digoxigenin-AP antibody (Roche 11093274910) followed by washes in 0.1% BSA/ PBT and AP1 buffer. Staining was performed using BM Purple (Roche 11442074001) followed by fixation in 4% PFA. RNA in situ hybridization to netrin-1 was performed on tissue cryosections according to the David Anderson laboratory protocol. Briefly, tissue sections were fixed in 4% PFA for 20 min at RT and digested with Proteinase K followed by post-fixation in 4% PFA for 15 min. Acetylation was performed in TEA buffer with acetic anhydride followed by washes in PBS and air-drying. Hybridization with RNA probes was done overnight at 68°C in a hyb chamber humidified with 50% formamide/4X SSC. Tissue sections were washed with SSC and PBT then incubated overnight at 4°C in blocking buffer with 1:2,000 dilution of anti-Digoxigenin-AP antibody followed by washes in PBT and AP1 buffer and staining with BM Purple.

Inner ear paintfilling and histology
Embryos were sliced below the forelimbs then fixed in 5% glacial acetic acid, 2% formaldehyde, and 75% ethanol overnight followed by an overnight dehydration in 100% ethanol. Methyl salicylate was then used for clearing of the tissue and long-term storage. Prior to paintfilling, embryos were bisected dorsally and the brain was removed to reveal the inner ear capsule. A micropipette was used to microinject 0.2% correction fluid diluted in methyl salicylate into the inner ear labyrinth. For histological analysis, embryos were fixed in 4% paraformaldehyde overnight at 4°C. They were then dehydrated to 70% ethanol and embedded in paraffin. Tissue sectioning was performed at 10-12 μm thickness. Tissues were cleared in xylene, stained with hematoxylin and eosin (H&E) and then mounted in Permount.