Mouse Tbx3 Mutants Suggest Novel Molecular Mechanisms for Ulnar-Mammary Syndrome

The transcription factor TBX3 plays critical roles in development and TBX3 mutations in humans cause Ulnar-mammary syndrome. Efforts to understand how altered TBX3 dosage and function disrupt the development of numerous structures have been hampered by embryonic lethality of mice bearing presumed null alleles. We generated a novel conditional null allele of Tbx3: after Cre-mediated recombination, no mRNA or protein is detectable. In contrast, a putative null allele in which exons 1-3 are deleted produces a truncated protein that is abnormally located in the cytoplasm. Heterozygotes and homozygotes for this allele have different phenotypes than their counterparts bearing a true null allele. Our observations with these alleles in mice, and the different types of TBX3 mutations observed in human ulnar-mammary syndrome, suggest that not all mutations observed in humans generate functionally null alleles. The possibility that mechanisms in addition to TBX3 haploinsufficiency may cause UMS or other malformations merits investigation in the human UMS population.


Introduction
The transcription factor TBX3 is critical for human development: heterozygotes bearing point, deletion and insertion mutations in TBX3 have ulnar-mammary syndrome (UMS) consisting of congenital limb defects, apocrine and mammary gland hypoplasia, and dental and genital abnormalities [1]. More recently, heart and conduction system defects have been described in mice (Tbx3) and humans (TBX3) [2,3,4,5,6,7]. Abnormal TBX3 expression occurs in multiple cancers [8]. It is becoming apparent that T-box proteins have functions in addition to transcriptional regulation [9] and that their activities are highly dosage sensitive [5].
We generated novel Tbx3 gene targeted alleles in mice and found that ablation of different regions of the N-terminal genomic sequence has different molecular and phenotypic consequences. We show that deletion of the T-box encoding region does not inevitably generate a null allele, as has been presumed [2,3,4]: splicing of residual 59 untranslated and 39 coding sequences can generate an abnormal mRNA that is translated into an aberrant protein predominantly localized in the cytoplasm. Homozygotes for this allele have different phenotypes than those observed with a true null allele that produces no mRNA or protein. These findings have important implications for interpreting the phenotypes of other presumed null alleles of Tbx3 in mice, and when considering the molecular mechanisms of congenital defects in humans with different types of TBX3 mutations..

Results
Three gene targeted Tbx3 alleles have been previously reported by other laboratories. These alleles were presumed null because they either delete exons encoding the T-box and DNA binding domain (Tbx3 tm1Pa [2]), or have insertions that disrupt the normal translational start codon Tbx3 Cre [3] and Tbx3 neo [4]. Homozygotes for these Tbx3 alleles die in broad windows from embryonic day (e) 10.5-e16.5. Heterozygotes are reportedly normal with the exception of mildly abnormal external genitalia in the Tbx3 tm1Pa/ tm1Pa mutant females [2]. The bases of the variable phenotypes observed with these alleles have not been determined: different genetic backgrounds likely play a role however, the possibility that these mutant alleles produce aberrant forms of Tbx3 mRNA and protein has not been tested.
We examined the genomic targeted sequence of Tbx3 D1-3 and found that it has the potential to encode a transcript restoring the normal Tbx3 reading frame in the terminal portion of Exon 3 Exons are depicted in thick black bars and numbered 1-7. Black arrowheads beneath locus indicate location of primers used for PCR analyses in panels C and D. Bidirectional arrow above wild type locus indicates region deleted between two endogenous EcoRI sites to generate the Tbx3 D1-3 targeted allele shown in B. Blue and red regions indicate portions of Exon1/59UTR and exon 3 that remain post targeted deletion. B) The Tbx3 D1-3 allele. Blue and red regions indicate portions of Exon1/59UTR and exon 3 that remain post targeted deletion. The red and blue arrowheads indicate primers (F, R) used for PCR analyses to assay for novel mRNA spanning the deletion. B') Genomic sequence resulting from deletion between endogenous EcoRI sites and insertion of polylinker to create the Tbx3 D1-3 allele. Note preservation of reading frame from exon 1 through polylinker into exon 3, which would produce a protein with wild type sequence beginning in exon 3. C) Quantitative real-time PCR analysis of cDNA from reverse transcribed Tbx3 mRNA in e10.5 mouse embryos. Bar graphs compare levels of transcripts detected using primer sets depicted in panels A and B. In Tbx3 +/+ embryos, no transcripts reflecting splicing from the 59UTR into exon 3 are detected (none). In Tbx3 +/D1-3 heterozygotes, 50% of wild type levels of exon 1/2 containing transcripts are present as expected from the single wild type allele. 100% of wild type levels of exon 5/6 containing transcripts are present (black bar). This reflects the contributions from the wild type allele (gray bar) and from the abnormal mRNA containing Cterminal sequences as detected with the F/R primer set (blue bar). In Tbx3 D1-3/D1-3 homozygotes, no exon 1/2 containing transcripts are present (none) and aberrant transcripts from the D1-3 allele (blue bar) are present at levels comparable to that of wild type transcripts in wild type embryos.
( Figure 1B). We used primers to assay Tbx3 transcripts containing the exon 1/2 junction by qRT-PCR of Tbx3 +/D1-3 embryos and, as expected, detected 50% of the wild type transcript level. However, primers that assay the exon 5/6 junction ( Figure 1C) detected 100% of the wild type transcript level in Tbx3 +/D1-3 mutants. We then designed primers to assay for the presence of an aberrant transcript that would be the product of the residual 59UT and exon 3, as predicted in Figure 1B. We detected this transcript by qRT-PCR, and by standard PCR followed by visualization of an amplicon of the predicted size only in embryos bearing the Tbx3 D1-3 allele. These findings revealed that an mRNA is produced from the remaining C-terminal exons 3-7 of the Tbx3 D1-3 allele. More importantly, when we employed a custom antibody to a Cterminal peptide unique to Tbx3, we detected Tbx3 protein in wild type embryos as expected ( Figure 1E), but also in Tbx3 D1-3/D1-3 embryos ( Figure 1F). Tbx3 protein is normally detected predominantly in the nucleus with variable amounts detected in the cytoplasm, depending on cell type. For example in the limb bud, mesenchymal cells contain mostly nuclear Tbx3 protein ( Figure 1E4) while ectodermal cells have cytoplasmic protein ( Figure 1E4). In contrast, the Tbx3 D1-3 mutant protein is cytoplasmic, even in mesenchymal cells ( Figure 1F4). This is consistent with deletion of a previously described nuclear localization signal [10].
The need for a Tbx3 conditional allele that is a true null in the recombined state is clear: it will allow us to bypass embryonic lethality seen in homozygotes of all previously reported mutant alleles (null or otherwise); it will permit conditional gene ablation approaches to study the role of Tbx3 in specific tissues, at later stages of development and in adult animals; it circumvents hypomorphic phenotypes resulting from insertion of exogenous sequences at other regions of the Tbx3 locus ( [5]; Moon, unpublished). The Tbx3 flox allele ( Figure 2A) was designed such that activity of Cre recombinase deletes 4.6kb of genomic DNA encompassing the promoter, 59UT, transcriptional start site and first exon of Tbx3 (Tbx3 Dflox ). This design required insertion of two loxP elements at the Tbx3 locus: one located 3.3 kb 59 of the translation initiating ATG and the second within the first intron. In addition to mutagenic capacity, site selection was influenced by the absence of any immediately proximate evolutionarily conserved sequences, reducing the chance that the presence of loxP elements would alter gene regulation in cis.
The value of Tbx3 flox as a conditional allele requires that it fulfill three requirements: it must behave as wild type in the unrecombined state; it must be a null allele in the recombined state (Tbx3 Dflox ); finally, it must be reliably competent for tissuespecific recombination by Cre recombinase. We addressed each of these requirements.
Heterozygous Tbx3 +/flox and homozygous Tbx3 flox/flox mice were indistinguishable anatomically and behaviorally from their wild type siblings, confirming the neutrality of the loxP sites. We generated Tbx3 +/Dflox heterozygotes and unlike Tbx3 +/D1-3 animals, both sexes were fertile and mothers had normal ability to nurture their litters (N = 28 females, 7 males). Furthermore, no abnormalities of the genitalia were detected.
In contrast to functional mRNA produced in Tbx3 D1-3/D1-3 embryos, Tbx3 Dflox/Dflox embryos contained no Tbx3 mRNA when assayed with primers to detect either the 59 or 39 ends of the message ( Figure 2B). Also in contrast to Tbx3 D1-3/D1-3 embryos (95% which were dead by e12.5), 30% of Tbx3 Dflox/Dflox embryos survived to e13.5 (Table 1). Despite their longer survival, it is notable that the limb and structural heart phenotypes of Tbx3 Dflox/ Dflox mutants are more severe than those reported for other presumed null mutants ( Figure 3). In E13.5 Tbx3 Dflox/Dflox mutants, we observe truncation of the hindlimbs beyond the tibia, absence of the fibula and an abnormal pelvis in 100% of mutants (N = 6, D) Visualization PCR products of reverse transcribed Tbx3 mRNA. Amplicons obtained using the F/R primer set shown in panel A, and with primers to detect HPRT as a control, were run on an agarose gel and visualized with ethidium bromide. Tbx3 D1-3 transcripts are not detected in wild type embryos. PD, primer dimer. E, F) The Tbx D1-3 mRNA is translated into protein that is predominantly localized to the cytoplasm. Confocal micrographs of sectioned E10.0 limb buds after fluorescent immunohistochemical detection of Tbx3 using an antibody to the C-terminus. E1-4) Tbx3 +/+ limb bud.    Figure 3H). In the forelimbs, there is some variability in severity of digit loss such that either digits 4 and 5 or digits 3-5 are absent and the left side is usually more severely affected ( Figure 3D,E). There is duplication of the condensations and soft tissue of the first digit. Tbx3 Dflox/Dflox mutant survivor have heart defects similar to those reported in Mesbah et al. [7], although the genetic backgrounds are not the same which makes objective comparison difficult. However, 18/25 Tbx3 tm1Pa/tm1Pa mutants had anterior displacement of the atria and 4/25 had more severe looping defects with thin walled ventricles; most survived to e13.5 with malformed outflow tracts [7]. In contrast, 50% our conditional null homozygotes have severe early looping defects with thin heart tubes and only 3 survived to e13.5; most died prior to the initial stages in outflow tract remodeling. At E9.0 we noted that 50% of Tbx3 Dflox/Dflox mutants have hypoplastic, thin-walled abnormally looped heart tubes (Figure 3 I, J); this phenotype is more severe than reported in Tbx3 tm1Pa/tm1Pa [7], Tbx3 Cre/Cre [3] or Tbx3 neo/neo [4] homozygotes. It would be ideal to examine other presumed null alleles for production of aberrant mRNA/protein and to perform side-by side phenotypic comparison in the same genetic background.
We tested the Tbx3 flox allele for efficient tissue-specific recombination in developing embryos using Prx1Cre [11] to recombine the allele in the forelimb mesenchyme. No mesenchymal Tbx3 protein is detectable In Tbx3 flox/flox ;Prx1Cre mutants ( Figure 2D1, D4), whereas there is robust signal in the control littermate in this tissue ( Figure 2C1, C4). Note also that as predicted with Prx1Cre, Tbx3 function and protein production is preserved in the apical ectodermal ridge of the limb bud ( Figure 2D1).

Discussion
UMS phenotypes vary within and between families, and all families evaluated to date have different mutations [6,12,13,14]. This variability, and the absence of malformations in several structures/organs that express TBX3 during development, suggest that dosage sensitivity to TBX3 is present in human embryonic development as it is in mice: normal murine cardiac structure and function require tight regulation of the dosage of several Tbx genes [5,15,16,17].
It has been hypothesized that the mechanism of UMS is TBX3 haploinsufficiency: transcripts from C-terminal deletion or missense mutant alleles were thought to be functionally null due to degradation by nonsense mediated decay or loss of critical functional domains [12,14]. Carlson and colleagues subsequently demonstrated that C-terminal truncated proteins could be produced in vitro and that a dominant repressor domain in the C-terminus is required for TBX3 to immortalize primary fibroblasts [10]. A correlation between mutations upstream of, or in the T-domain and more severe UMS phenotypes has been suggested more recently [6]. We have now shown that removal of the T-box encoding region allows production of a variant transcript and the resulting protein is aberrantly localized to the cytoplasm. Our findings indicate that the genital, fertility and nurturing phenotypes seen in Tbx3 +/D1-3 heterozyogotes are not attributable to loss of wild type Tbx3 protein, but due to negative effects of the aberrant protein produced from the Tbx3 D1-3 transcript. The fact that Tbx3 D1-3/D1-3 homozygotes die earlier than Tbx3 Dflox/Dflox null mutants indicates that deleterious effects of the Tbx3 D1-3 protein exacerbate those due to of loss of the wild type protein. A similar situation occurs in mice with respect to the T gene. Defects in the Brachyury mutant (a 200-kb deletion removes the entire gene) are thought to result from haploinsufficiency of the T protein [18]. Other T mutations (T wis , T c ,T c-2H ) which encode frameshifts and truncated proteins are thought to generate dominant-negative proteins and cause more severe developmental defects [19]. The heart and skeletal phenotypes of Tbx3 Dflox/Dflox true null mutants are more severe than those of other presumed null mutants previously reported.
Many human UMS mutations have the potential to generate abnormal proteins containing either N-or C-terminal regions. Our observations raise the possibility that some nonsense and missense human TBX3 mutations previously thought to function as null alleles may also produce aberrant transcripts and proteins with unexpected activity. Since it has been shown that Tbx3 protein has both activator and repressor domains [10], it is likely that a mutant protein with preservation of one domain in the absence of another will have markedly different activity in vivo.

Gene Targeting
The Tbx3 D1-3 allele has a 4.6kb deletion between 2 EcoRI sites located 920 bp 5' of the translational start site and 3.7 kb 39 of the ATG ( Figure 1A). This deletes the a portion of the 59UTR, all coding regions of exons1, 2 and 2a and all but the terminal portion of exon 3 encoding the final 12 amino acids of this exon. An FRTflanked neomycin (neo r ) selection cassette at the BglII site 3.2 kb 59 of the ATG was removed with Flip recombinase with the B6.SJL-Tg(ACTFLPe)9205Dym/J strain. These mice were maintained in a mixed Bl6/SV129/FVB background.
The Tbx3 flox conditional allele was generated by inserting the 59 loxP sequence in an AgeI site 4.2 kb upstream of the ATG; the 39 LoxP sequence was inserted in a SpeI site 1100 bp 39 of the ATG (midway between the first and second exons, Figure 2A). Adjacent to the 39 loxP is an FRT-flanked neo r cassette used as a positive selectable marker; a thymidine kinase negative selectable marker was included outside the region of genomic homology. Following electroporation ES cells were selected for G418 r , Ganc r , and 182 cells lines were isolated for further characterization. The initial allele was designated Tbx3 floxneo and ES cells carrying Tbx3 floxneo were injected into blastocysts to generate chimeric mice that successfully transmitted the mutant allele. Heterozygous Tbx3 +/ floxneo progeny of the chimeras were bred with mice expressing FLPe recombinase (Gt(ROSA)26Sor tm1(FLP1)Dym ) to remove the neo r cassette, creating the Tbx3 flox allele. These mice were maintained in a mixed Bl6/SV129 background.
Tbx3 +/Dflox animals were generated by breeding Tbx3 flox/flox males to hprtCre females which causes recombination in the egg; Tbx3 +/Dflox males and females were then intercrossed to obtain.

Ethics Statement
All mouse work was performed under a protocol in Dr. Moon's name approved by the University of Utah IACUC and euthanasia was performed in accordance with AVMA requirements. Preparation of RNA from Embryos for Reverse Transcription and qRT-PCR Tissues were dissected in ice cold PBS and stored in RLT buffer (Qiagen) at 280uC. Total RNA was extracted from samples (RNeasy Micro Kit, Qiagen).One hundred micrograms of total RNA was transcribed to cDNA using the Superscript III First-Strand Synthesis System (Invitrogen). Quantitative PCR was performed with iQ SYBR Green Supermix on the iCycler system (Bio-Rad) and normalization was to hprt, gapdh, and b-actin. qPCR data is presented using the DDC(t) method [20].
Primer Sequences for Real-time Quantitative PCR Exon 1 forward: 59 TGAGGCCTCTGAAGACCATG 39. Exon 2 reverse: 59 TCAGCAGCTATAATGTCCATC 39. The pubis is malformed and lacking one element, likely the pubic bone. s, scapula; h, humerus; r, radius; u, ulna; digits 1-5; fe, femur; fi, fibula; t, tibia; il, ilium; is, ischium; p, pubic bone. G-J) Ventral (G, I) and left lateral (H,J) views of whole mount E9.0 control (G, H) and mutant (I, J) embryos centered on the heart. The mutant heart is short, bowed ventrally rather than the normal loop to the right. Only a single, thin walled ventricular chamber (V) is present whereas in the lateral view of the control, the nascent right ventricle (RV) is present. doi:10.1371/journal.pone.0067841.g003

Generation of Anti-Tbx3 C-terminal Antibody
We synthesized a KLH conjugated peptide unique to the Cterminus of Tbx3: GLEAK PDRSCSGSP. The antiserum was generated in rabbits by Covance and the polyclonal antibody was affinity purified, and validated by western blot, immunoprecipitation and immunohistochemistry. Immunohistochemistry E10.5 embryos were harvested in 1XPBS, fixed overnight at 4 degrees in 4% paraformaldehyde. Limb buds were dissected and processed for paraffin sectioning. Immunohistochemistry was carried out on 10 micron paraffin sections using Anti-Tbx3 Cterminal antibody. Citrate antigen retrieval was performed and sections incubated with primary antibody (1:200) overnight at 4 degrees and detected using donkey anti-rabbit conjugated to Alexa fluor 488(1:500) from Invitrogen. Nuclei were stained with Hoescht. Slides were imaged with a Nikon ARI inverted confocal microscope at the University of Utah Imaging Core.

Statistical Analysis
Comparison of the recovered versus expected ratios of genotypes from mating male and female Tbx3 +/D1-3 and Tbx3 +/ Dflox heterozyogtes was done using Pearson's chi-squared test. Comparison of the numbers of Tbx3 +/D1-3 and Tbx3 +/Dflox adult females demonstrating abnormal pheontypes was done using Fisher's exact test. Statistical test were performed using GraphPad software (www.graphpad.com).