|Home | About | Journals | Submit | Contact Us | Français|
Autosomal dominant mutations in the bHLH transcription factor TWIST1 are associated with limb and craniofacial defects in humans with Saethre-Chotzen syndrome (SCS). The molecular mechanism underlying these phenotypes is poorly understood. We show that the ectopic expression of the related bHLH factor Hand2 phenocopies Twist1 loss-of-function phenotypes in the limb, and that they display a gene dosage-dependent antagonistic interaction. Twist1 and Hand2 dimerization partner choice can be modulated by PKA and protein phosphatase 2A-regulated phosphorylation of conserved helix I residues. Interestingly, multiple TWIST1 mutations associated with SCS alter PKA-mediated Twist1 phosphorylation, suggesting that misregulation of Twist1 dimerization via either stoichiometric or posttranslational mechanisms underlies SCS phenotypes.
Studies of developing vertebrate limbs have yielded many insights into the process of embryonic pattern formation. Prominent among these are the identification of a growing catalog of transcription factors that orchestrate limb patterning. While the genetic and biochemical interactions of these transcription factors are clearly important for integrating patterning information, these interactions are poorly understood. Twist1 and Hand2 are basic helix-loop-helix (bHLH) transcription factors within the Twist family, and are attractive candidates for investigating such interactions. Each is required for distinct yet subtly related aspects of limb development, and biochemical studies have revealed a complex regulation of their protein-protein interactions1–3.
Early limb bud expression of Twist1 is observed primarily in the peripheral mesenchyme, and Twist1 is required for maintenance of the overlying apical ectodermal ridge (AER)4–7. Twist1 haploinsufficiency in mice and humans is associated with a range of limb abnormalities. Twist1 heterozygous null mice display a partially penetrant preaxial polydactyly8,9. Human TWIST1 nonsense, missense and null mutant alleles8,10–12 are associated with distal limb abnormalities (including polydactyly, brachydactyly and syndactly) as seen in patients with the autosomal dominant craniosynostosis disorder Saethre-Chotzen Syndrome (SCS)12. While some TWIST1 missense alleles display deficient nuclear localization or impaired DNA binding in vitro13,14, a mechanistic understanding underlying many of the TWIST1 SCS mutations is lacking.
Following the onset of limb bud outgrowth, Hand2 is expressed primarily in the posterior limb mesenchyme, where Hand2 expression is both dependent on, and necessary for, normal sonic hedgehog (Shh) expression15–17. Furthermore, ectopically expressed Hand2 within the anterior limb mesenchyme induces preaxial polydactyly and activates the Shh pathway, although it does not always induce ectopic expression of Shh15–17. While Hand2 appears to act through a maintenance loop with Shh, additional functions for Hand2 are possible, as the Hand2 expression domain is broader than that of Shh15–17. Interestingly, both Twist1 and Hand2 are associated with polydactylous limb phenotypes, raising the issue of whether they may interact genetically and/or biochemically.
Although direct interactions between Twist1 and Hand2 have not been described, numerous genetic and functional studies of Twist-related bHLH factors have been reported1,3,18. A number of Twist family proteins, including Twist1 and Hand2, not only dimerize with ubiquitously expressed E proteins but also exhibit promiscuous dimerization with other class B bHLH proteins, and can form stable homodimers19–21. Furthermore, the dimerization partner choice of a related Twist family member, Hand1, can be modulated via the phosphorylation state of specific conserved residues in helix I (ref. 20). Thus, Twist1/Hand2 interactions might depend on partner availability as well as regulatory posttranslational modifications that influence protein-protein interactions18.
Here we investigate the biochemical and genetic interactions between Twist1 and Hand2 both in vitro and during limb development. We show that PKA and B56δ-containing PP2A can regulate Twist1 and Hand2 phosphorylation at the conserved helix I residues, that hypophosphorylation and phosphorylation mimics of these residues alter bHLH dimerization affinities, and that a population of TWIST1 mutations that cause SCS in humans exhibit disregulation of this phosphoregulatory circuit. We also show that ectopic Hand2 expression phenocopies multiple SCS-like limb phenotypes, that the appropriate genetic dosage of Hand2 and Twist1 is critical for proper limb development, and that these interactions require the phosphoregulated helix I residues. These findings support a mechanism where the Twist family dimerization partner choices are modulated by both the relative levels of gene expression and the phosphorylation state of key helix I residues, thereby dictating changes in the regulation of their downstream target genes. We propose that the disruption of this regulatory mechanism underlies the phenotypes observed in a subset of TWIST1 mutants that cause SCS.
Sequence comparisons indicate that the phosphorylation-regulated threonine and serine residues in helix I of Hand1 are highly conserved among the bHLH domains of Twist family proteins, and are conserved among Twist1 homologs as distant as Drosophila Twist (Fig. 1a). By contrast, threonine and serine residues at this position within helix I are not generally found in bHLH proteins outside of the Twist family. This striking conservation raised the possibility that modulating the phosphorylation state of these residues might be a general mechanism that controls the activity not only of Hand1 (ref. 20), but also of Twist1, Hand2 and other Twist-related factors.
To test whether the conserved helix 1 residues of Twist1 and Hand2 could be phosphorylated by PKA, we coexpressed wild type Hand2 and Twist1 or Hand2 T112;S114A and Twist1 T125;S127A double alanine substitution mutants with constitutively active PKA in metabolically-labeled HEK293 cells (Fig. 1b). Immunoprecipitation of radiolabeled proteins shows that PKA can phosphorylate Hand2 and Twist1; moreover, specific mutation of the conserved helix 1 residues reduces the level of 32P incorporated into either Hand2 or Twist1 (Fig. 1b).
We next asked if Twist1 and the PP2A regulatory subunit B56δ can form a complex by coexpressing Twist1 with either B56δ or B56α in HEK293 cells (Fig. 1c). Immunoprecipitation (IP) followed by immunoblot analysis shows that Flag-tagged Twist1 co-immunoprecipitates significant amounts of B56δ, while interaction between Twist1 and the related PP2A subunit B56α, is significantly weaker (Fig. 1c). Interestingly, Twist1 appears to form a stronger interaction with B56δ than Hand1 (Fig. 1b). Thus, similar to observations reported for Hand proteins20, Twist1 can be phosphorylated by PKA and interacts specifically with the PP2A B56δ regulatory subunit.
We next sought to determine if the conserved helix I threonine and serine residues are PKA phosphorylation targets, and if B56δ facilitates specific dephosphorylation of Twist1 and Hand2. We performed phosphopeptide-mapping analyses of both wild type and hypo-phosphorylation mutants of these proteins coexpressed with active PKA and/or B56δ or B56α in metabolically labeled HEK293 cells (Fig. 2). PKA coexpression results in de novo detection of two Hand2 phosphopeptides and increased phosphorylation of an additional phosphopeptide (Fig. 2a,b). Addition of B56α reduces 32P incorporation in a single peptide compared to Hand2 with PKA (Fig. 2c,l). This suggests that B56α can moderately affect Hand2 phosphorylation despite the lack of evidence for direct protein-protein interaction. Coexpression of Hand2, PKA and B56δ results in a more robust reduction (38% of Hand2 plus PKA) in the phosphorylation of this peptide (Fig. 2d,l). Moreover, mutation of the conserved T112 and S114 residues to alanines reduces 32P incorporation into this peptide to undetectable levels (Fig. 2e,l).
In order to further define the phosphorylated residues of Hand2, we raised polyclonal antisera against a Hand protein helix I peptide that contains a phosphoserine at the S114 position, and used it for immunoblotting extracts of HEK293 cells transfected with various Hand2 mutants in combination with PKA (Fig. 2m). This antiserum detects a wild type Hand2 band only when Hand2 is coexpressed with PKA (Fig. 2m). The T112;S114A mutation reduces this phosphospecific band to undetectable levels, while the basic domain T103A mutation has no effect. These data provide additional evidence that the S114 residue of Hand2 is phosphorylated by coexpression of Hand2 with constitutive PKA.
Coexpression of Twist1 with active PKA also results in increased protein phosphorylation (Figs. 1b;2f,g). The addition of B56α has no significant effect on Twist1 phosphorylation (94% of wild type Twist1 phosphorylation levels; Fig. 2h,l). In contrast, coexpression of Twist1 with PKA and B56δ reduces the phosphorylation of a single Twist1 phosphopeptide by 75% (Fig. 2i,l), while mutation of T125 and S127 to alanines reduces this phosphorylation to undetectable levels (Fig. 2j). Thus, Hand2 and Twist1 can be post-translationally modified by PKA and B56δ-containing PP2A, consistent with a possible regulatory circuit that utilizes the helix I phosphorylation state to control the activity of these proteins.
Missense mutations of several basic domain residues in human TWIST1 are associated with SCS8,10–12,22,23. As the mutated residues are adjacent to the phosphorylated helix I residues, this raised the possibility that these mutations might disrupt the Twist1 PKA-PP2A regulatory circuit. We therefore assessed the phosphorylation of Twist1 mutated at each basic domain residue (R118, R120, Q121, R122) individually (Fig. 3a–h). We found that each mutation greatly reduced the phosphorylation of T125;S127 by PKA in metabolically labeled HEK293 cells (Fig. 3).
Three of these mutations replace positively charged residues, suggesting that the degree of PKA-mediated phosphorylation is proportional to the basic character of the basic domain. By contrast, a dominant negative C. elegans Twist E29K mutation24 increases the alkalinity of the basic domain, and thus might enhance PKA-mediated phosphorylation. Phosphopeptide mapping analysis of the analogous Twist1 E117K mutant indicates that this is indeed the case (Fig. 3i,j). Taken together, these data demonstrate that Twist1 helix I phosphorylation can be modified by mutation of nearby residues, including residues with mutations found in SCS patients.
We next tested how differential phosphorylation might impact the activities of Hand2 and Twist1. Previous reports demonstrate that Hand1 and Hand2 can form homo- or heterodimers, and that Hand1 dimerization choices can be phosphoregulated19,20. These observations raise the possibility that interactions between Twist1 and other Twist family members might be similarly regulated. We therefore asked whether Twist1 can heterodimerize with Hand1 or Hand2, and if mimicking phosphorylation at the conserved threonine and serine residues can regulate Twist1 dimerization affinities.
To assess the ability of Twist1 to complex with Hand proteins, we performed IP-immunoblot analysis on extracts of HEK293 cells cotransfected with epitope-tagged forms of Twist1, Hand1 and Hand2 (Fig. 4a). Expression of Flag-tagged Twist1 with myc-tagged Hand1 or Hand2 results in specific co-immunoprecipitation of Twist1 with either Hand protein. As expected, Hand protein heterodimeric and homodimeric complexes are also readily detectable when epitope-tagged Hand1 or Hand2 are coexpressed.
Fluorescence resonance energy transfer (FRET) occurs between donor and acceptor fluorescent molecules when they are in close proximity, and can be used to demonstrate direct protein-protein interactions. To compare interactions between various Twist1 dimerization partners, we used acceptor photobleaching FRET to measure the fraction of donor fluorescence that is absorbed by the acceptor (FRET efficiency)25. We generated both CFP and YFP carboxy-terminal fusion proteins of E47, Hand2, wild-type and T125;S127 mutant forms of Twist1, coexpressed them in HEK293 cells and measured FRET efficiencies (Fig. 4b). Our results show that wild type Twist1 forms dimers with itself, E47, and most strongly with Hand2. The Twist1 phosphorylation mimic T125;S127D displays a general reduction in dimerization, altering E47/Twist1 heterodimers and Twist1 homodimers to a greater extent than the Hand2/Twist1 heterodimer. Altered FRET efficiency is also observed in the hypophosphorylation mutant T125;S127A. While this mutant also shows a general reduction in heterodimerization, the Hand2/Twist1 heterodimers are more severely affected than E47/Twist1. By contrast the Twist1 T125;S127A homodimer is stabilized compared to wildtype Twist1. This results in distinctly different relative FRET efficiencies than those observed with wild type and Twist1 T125;S127D. Taken together, these findings suggest that the alteration of the phosphorylation state of Twist1 at T125 and S127 affects the dimerization affinity of Twist1 for its various potential partners, including Hand2.
Having established that Twist1 and Hand2 can interact in vitro, we next sought to determine if they might interact in vivo during embryonic development. We chose to examine the developing vertebrate limb bud, as previous studies indicated that both genes are expressed in the limb6,7,15–17. There is overlap in the RNA expression of Twist1 and Hand2 in both chick and mouse limb mesenchyme throughout limb morphogenesis (Fig. 5). Twist1 is expressed broadly in the peripheral mesenchyme from Hamburger-Hamilton stage 19 through stage 30 (ref. 26) in chick and from E9.5 through E12.5 in mouse, with elevated levels observed at the anterior and posterior margins. The major Hand2 expression domain is initially restricted to posterior mesenchyme, and expands progressively towards the distal anterior. A minor anterior proximal Hand2 expression domain is also visible. While Twist1 and Hand2 are initially coexpressed within the posterior limb mesenchyme, the region of overlap expands by stage 24 or E11.5 to include much of the distal mesenchyme underlying the AER (Fig. 5a–h,q–t). At the time of digital condensation, extensive interdigital expression of both genes is observed (Fig. 5i–l). Subsequently, overlapping expression is seen in the peridigital mesenchyme at stage 30 of the chick and E12.5 in the mouse (Fig. 5m–p). These results demonstrate that Twist1 and Hand2 are coexpressed within distinct regions of the developing vertebrate limb, consistent with the idea that these two factors may functionally interact.
If Hand2 and Twist1 act together in the limb, then Hand2 misexpression might cause phenotypes related to those seen in Twist1 mutants. To test this, we used a replication-competent retrovirus to misexpress wild type Hand2 in developing chick limbs by general or focal infection at stages 10–20, and examined cartilage, soft tissue and molecular phenotypes through E10 (Fig. 6). We observed preaxial polydactyly in wings stained for cartilage patterning. Affected anterior digit identities varied widely, ranging from cartilage blebs and posterior transformations through induction of ectopic digit 3s (Fig. 6a–c). While these observations are consistent with previous reports, the extent of the digit duplications is more severe than previously described following Hand2 misexpression in chick limbs15,16. We also observed several novel limb phenotypes that are distinct from those previously associated with Hand2 misexpression. These include frequent cases of syndactyly (Fig. 6d,e), shortened cartilage elements in digits and long bones (Fig. 6f,g), and occasional bent digits (Fig. 6h,i). We also detected AER disruptions, as indicated by gaps in the expression of the AER marker Fgf8 (Fig. 6j,k), and regions of reduced distal mesenchyme outgrowth (Fig. 6l,m), that are consistent with disrupted AER maintenance. This combination of limb phenotypes caused by ectopic Hand2 expression resembles those seen in Twist1+/− mice or humans with reduced or altered TWIST1 activity (syndactyly, brachydactyly, hallux valgus, AER disruption and polydactyly)4,12. The similarities between the phenotypes resulting from Twist1 loss of function mutations and Hand2 gain of function experiments are consistent with the idea that these genes have antagonistic functions in limb development.
We next examined genetic interactions between Twist1 and Hand2 to test for functional antagonism. If they do interact antagonistically, then either reducing Hand2 levels to compensate for reduced expression of Twist1, or increasing Twist1 levels when misexpressing Hand2, might change their respective effects on limb development. We first asked whether lowering Hand2 levels would suppress the effects of reduced Twist1. While Hand2 heterozygous null mice are normal, Twist1 heterozygous null mice display a partially penetrant preaxial polydactyly of the hindlimbs (Fig. 7a; refs. 8, 9). We intercrossed Twist1+/− and Hand2+/− mice to determine if lowering Hand2 gene dosage would rescue the Twist1+/− mediated polydactyly (Fig. 7b). 18 of 43 (42%) F1 Twist1+/− animals had hindlimb polydactyly, while their F1 Hand2+/− siblings were all normal (53 of 53). By contrast, polydactyly was never observed (0 of 44) in compound F1 Hand2+/− Twist1+/− offspring (p0.01, χ2). These results reveal a novel dose-dependent requirement for Hand2, and show that the relative levels of Hand2 and Twist1 critically affect limb development.
We then tested whether increased Twist1 expression could compensate for increased Hand2 levels by misexpressing Hand2 either in combination with Twist1, or with GFP as a control. We coinfected the anterior of stage 20 chick wing buds with two viruses that allow simultaneous transduction of two genes. We noted the identities of ectopic digits at E10, scored each limb based on the digit with the most posterior identity, and compared the frequency distribution for each virus combination (Fig. 7c). As the degree of digit posteriorization is proportional to the strength of polarizing signal or Shh pathway activation27,28, this allowed us to quantify the activity of the virally expressed proteins. Hand2 plus GFP virus infection induced a range of duplications, with an ectopic digit 3 noted in 64% of limbs, digit 2 in 21% and no ectopic digit in 15%. The extent of duplications induced by Hand2 plus Twist1 was significantly weaker, with an ectopic digit 3 in only 38% of limbs, a digit 2 in 28% and no ectopic digit in 35% (p0.01, Mann-Whitney). Focal anterior Twist1 virus infection alone had no effect on limb morphology, although Twist1 did cause gross cartilage abnormalities when misexpressed throughout the limb bud, providing independent confirmation of Twist1 virus activity (Fig. 7c, data not shown). Thus increasing Twist1 levels can suppress the effects of increased Hand2.
We also tested whether the conserved helix I threonine and serine residues are required for the ability of Twist1 to suppress the Hand2 misexpression phenotype. We generated a retrovirus carrying the Twist1 helix I double alanine substitution, and tested it in the anterior limb coinfection assay. As observed with wildtype Twist1, anterior misexpression of the mutant Twist1 alone did not cause patterning defects. Unlike the WT gene, however, the mutant Twist1 had no effect on the ability of Hand2 to cause digit duplications (Hand2 plus GFP, 69% ectopic digit 3; Hand2 plus Twist1 mutant, 69% ectopic digit 3; Fig. 7d). Thus the same helix I mutation that prevents Twist1 phosphorylation and that alters Twist1 dimerization partner preference also disrupts Twist1 biological activity in vivo.
In this study, we describe two mechanisms by which the activities of Twist1 and Hand2 can be regulated, either by changes in their relative levels of expression or through phosphorylation. Specific threonine and serine residues in helix I shared by Twist1 and Hand2 are conserved among a majority of Twist family members. We find that these residues can be post-translationally modified via the actions of PKA and B56δ containing PP2A, and provide evidence that dimerization affinity can be altered by such modifications. We also find that ectopic Hand2 expression phenocopies limb phenotypes resulting from mutations in Twist1, including those presented in patients with SCS. We furthermore demonstrate a gene dose-dependent interaction between Twist1 and Hand2 in both combination gain-of-function and combination loss-of-function experiments and show that a mutation that alters Twist1 dimerization and phosphorylation in vitro also alters its genetic interactions with Hand2 in vivo. These data strongly support models in which Twist family bHLH protein dimerization partner choice is crucial for normal development.
Early models of bHLH function posited that competition between transcriptionally active E proteins and transcriptionally inactive HLH proteins for dimerization with tissue specific bHLH proteins was the primary mechanism regulating the formation of bHLH complexes18,29. However, it is now clear that regulation by both phosphorylation and interactions with a wider spectrum of dimerization partners are also significant additions to this model. We previously showed that PKA and B56δ-containing PP2A regulate Hand1 dimerization affinities for other bHLH proteins through phosphorylation of helix I (ref. 20). The current demonstration that Twist1 and Hand2 are similarly modified suggests that other Twist-family proteins that share the phosphorylated helix I residues also share this regulatory mechanism. The embryological significance of the Twist1 post-translational modifications is reflected in the phenotypes observed in SCS patients12, in C. elegans models and in the data presented here (Fig. 7) in which Twist1 point mutations that disrupt the PKA/PP2A circuit alter the developmental activity of Twist1 (ref. 24). These observations provide a potential mechanistic explanation for several mutations found in SCS patients. These include five basic domain mutations and mutation of the conserved phosphorylation-regulated helix I serine10,12,24.
The use of phosphorylation to regulate Twist family dimerization may extend beyond interactions with the Twist family and E proteins. For example Twist1 can bind the myogenic factor MyoD, and negatively regulate its activity30,31. Interestingly, a Twist1 R120A;R122A;R124A triple mutation was reported to disrupt interactions between MyoD and Twist1 (ref. 30). Our results show that mutating these residues individually also disrupts PKA-mediated phosphorylation of Twist1 helix I (Fig. 3). These data are consistent with the idea that the MyoD and Twist1 interaction may be regulated in part by phosphorylation of Twist1.
Although the Twist1 basic domain mutations can affect the ability of PKA to phosphorylate the helix I residues, it is possible that the developmental effects may result solely from changes in DNA binding specificity. We feel this is unlikely for several reasons. First, Twist carrying several of the SCS mutations tested here remains capable of DNA binding as C. elegans Twist-Daughterless heterodimers24. Nevertheless, DNA binding is not required for all Twist1 activity, as a Twist1 mutant with a non-DNA binding basic domain can still inhibit transcriptional activation by MyoD and Mef2 (ref. 31). Similarly, Hand2 that lacks the basic domain remains competent to induce polydactyly when misexpressed in mouse limbs32. Thus multiple lines of evidence point toward Twist1 having DNA binding-independent activities that may be influenced by properties of its basic domain.
While proper Twist1 gene dosage has long been appreciated as critical for normal development, the molecular basis for this is less clear8,9,11,12. Our results indicate that at least in the limb, reduced Twist1 or increased Hand2 dosage disrupts a critical antagonistic balance between Twist1 and Hand2. Hand2 and Twist1 are expressed in the limb in overlapping domains, and these proteins can form heterodimers with each other, form homodimers, or form heterodimers with other partners. Changing the Twist1:Hand2 dosage probably alters the relative amounts of the various possible protein combinations, leading to changes in target gene expression and consequent developmental defects. While we currently do not know which complexes are most critical, there is precedent for specific Twist dimer complexes differentially regulating transcription of target genes21 and thus each unique dimer likely exhibits unique biological activities. Studies in Drosophila using forced dimers show that Twist homodimers specify mesoderm and the somatic myogenic lineage while Twist-daughterless heterodimers repress the transcription of genes required for somatic myogenesis21. A similar experimental strategy may help define the functions of the various Twist1 and Hand2 dimer combinations in the limb.
Some of the ectopic Hand2 limb phenotypes reported here may reflect a normal developmental mechanism where high levels of Hand2 are required to antagonize Twist1 activity. The AER disruption associated with Hand2 misexpression is one example. AER disruption can be observed adjacently to ectopic outgrowths that correspond to sites of Hand2 virus infection, and thus presumably is due to high levels of Hand2 expression (Fig. 6, data not shown). As Twist1 is normally required in the mesenchyme for the function of an FGF signaling loop between mesenchyme and ectoderm that maintains the AER4,5,33, the Hand2 virus phenotype may be caused by interference with this Twist1 activity. Interestingly, Hand2 expression is normally highest in the posterior limb, adjacent to ectoderm just proximal of the AER, and from which the AER has regressed in concert with distal limb outgrowth. Thus, one normal function of high level Hand2 expression may be to limit the posterior-proximal extent of the AER.
Bialek et al.34 recently showed that a novel carboxy-terminal ‘Twist domain’ conserved in Twist1 and Twist2 binds Runx proteins and inhibits Runx transcriptional activity. As Runx2 promotes ossification of cartilaginous and membranous bone, Twist proteins thereby act to delay the onset of ossification. It is proposed that in SCS patients, either mutation of the Runx binding domain or reducing Twist1 levels generates excess free Runx2 protein that promotes premature ossification, resulting in the craniofacial and skeletal defects of SCS. How Hand2 activity fits into this paradigm is unclear. Hand2 levels in the limb may modulate the availability of Twist1 for interaction with Runx proteins. Or, as Runx mutants with polydactyly have not been described35–38, interactions between Hand2 and Twist1 in the early limb may operate independently of Runx pathways. Whether changes in Hand2 gene dosage can alter osteogenic SCS phenotypes is an interesting question. Similarly, the Twist1 double alanine mutation tested here does not suppress the Hand2-induced polydactyly phenotype and also does not cause gross cartilage defects. Whether the loss of this latter activity is due to defective Hand2 interactions or reflects a requirement for these residues for broader aspects of Twist1 functionality remains to be determined. These and future studies will expand our understanding of how combinatorial interactions among broadly expressed transcription factors drive highly context-specific developmental programs.
Twist1 and Hand2 retroviruses were constructed by subcloning cTwist1, cTwist1T109;S111A or mHand2 cDNAs into the RCAS BP(A) and RCAS BP(B) vectors39. The cTwist1 constructs contain an amino-terminal FLAG epitope tag (Sigma). RCAS BP(A) GFP was described previously39. pIRES FLAG-Hand1 and Hand2 encode amino terminal FLAG epitope tag fusions cloned into pIRES NEO (Clontech). B56α and B56δ cDNAs were cloned into the expression vector pCEP4-l. Full length E12 and E47 were a gift from Y. Kee and M. Bronner-Fraser. Myc- and FLAG-tagged mouse Twist1 were a gift from E. Olson. E-protein and Hand YFP and CFP fusion proteins were made as described20. Twist1 YFP and CFP fusion proteins were cloned by inserting the mTwist1 cDNA beginning from codon 3 in frame into pEYFP-N1 and pECFP-N1 (Clontech). Hand2 and Twist1 point mutants were generated using the Quick-Change Mutagenesis kit (Stratagene) following the manufacturer’s protocols.
EK293 cells were grown and transfected as described20 with the indicated constructs. In some experiments a single CaPO4 precipitate was split onto duplicate plates, with one used for metabolic labeling and one for immunoblot analysis. For metabolic labeling 48 hours post transfection cells were incubated with 1mCi of 32P orthophosphoric acid (NEN)/ml of phosphate–free DMEM supplemented with dialyzed FBS for 4 hours. Cells were washed in 20mM Na Hepes pH 7.4 and 150mM NaCl and lysed in 20mMNaPO4, 150mMNaCl, 2mMMgCl2, 0.1%NP40, 10% glycerol, 3ug/ml leupeptin, 3ug/ml pepstatin, 1mM PMSF, 50uM NaVO4, 5mM NaF, 100nM Okadaic acid, and 5mM Beta glycerol phosphate and equal amounts of protein were immunoprecipitated with agarose-conjugated FLAG M2- or myc beads (Sigma) for 2 hours. Samples were resolved through a 12% SDS PAGE gel, and exposed to a phosphoimager screen.
Cell lysates were collected, and immunoprecipitations performed as described20. Cell lysates or immunoprecipitated proteins were run through 12% SDS PAGE gels, electroblotted and incubated with the indicated antibody as described20. Blots were visualized using the Super Signal Luminescent detection protocol (Pierce). Rabbit anti-phosphoHand antisera was commercially prepared by Sigma using the peptide shown in Fig. 2 as antigen.
HEK293 cells were grown, transfected, labeled with 32P, and immunoprecipitated exactly as described20. TLC and gel exposures were visualized by phosphoimager. Quantification of peptide intensities was performed using Image Quant phosphoimager software. Equal-sized volume boxes were placed over variable experimental and invariant control phosphopeptides, and integrated volumes and volume ratios of affected to control peptides were calculated.
HEK293 cells were transfected on glass cover slips with the indicated constructs, grown for 48 hrs, fixed and mounted. Measurement of acceptor photo-bleaching of FRET efficiency was performed exactly as described25. The fluorescence emission from the donor and the acceptor were examined in each cell to ensure that assayed cells expressed equivalent levels of CFP and YFP. The fluorescence emission from the donor and the acceptor were collected sequentially. Average fluorescence intensities of the donor were measured before and after bleaching. A minimum of 30 data points for each dimer pair were collected. The FRET efficiency was calculated by ET = (1 − (IDA/ID))*100, where IDA and ID represent the steady state donor fluorescence in the presence and the absence of the acceptor.
Non-radioactive whole mount and section in situ hybridization were performed as previously described40. Probes for the following genes were used: mHand2 (ref. 41), cHand2 (ref. 42), cTwist1 (ref. 6), mTwist1 (ref. 4).
SPF White Leghorn chick embryos (Charles River) were used for all experiments. Concentrated virus stocks of the indicated viruses were generated as described, and all viruses had titers of at least 5 ×108 infectious particles/ml39. Flag-cTwist1 and Flag-cTwist1T109;S111A protein expression and nuclear localization were confirmed by anti-Flag M2 immunostaining. For single virus experiments, stage 10 through stage 20 embryos were either generally or focally infected in the fore and hind limb fields using standard procedures39. For double virus experiments, (A) and (B) coat viruses were combined to equal titers, and stage 20–21 wing buds were focally infected under the anterior AER. Embryos were harvested through E10 for subsequent analyses. E10 chick limb skeletal and soft tissue morphologies were revealed by alcian green staining followed by clearing in BABB (1:2 v/v benzyl alcohol:benzyl benzoate). Digit identities were scored as described43. Statistical comparisons were performed using the Mann-Whitney test for nonparametric distributions.
Outbred Swiss Webster mice (Taconic) were used for wild type embryo collection. Noon on the day of a mating plug was considered E0.5. Hand2+/− mice41 were obtained on a mixed 129.B6 background and backcrossed to B6 through N5. Twist1tm1Bhr/+ mice4 (Jackson Laboratory) were maintained on a B6 background. We observe 44% penetrance (N=124 Twist1tm1Bhr/+ animals) of hindlimb polydactyly on this background. Hand2+/− B6(N2) to B6(N5) and Twist1tm1Bhr/+ animals were intercrossed, and hindlimbs of F1 progeny scored for the presence of ectopic digits. As the Hand2+/− B6 backcross generation had no effect on the frequency of polydactyly in Twist1tm1Bhr/+;Hand2+/+ animals, data from all generations were pooled. Animals were pcr-genotyped for Twist1 (ref. 9) and Hand2 alleles41 as described. Alizarin red staining of skeletal preparations was performed as described44. Experimental protocols involving vertebrate animal use were approved by the Columbia University College of Physicians and Surgeons Institutional Animal Care and Use Committee.
The following sequences were used for protein comparisons: mouse Twist1, M63649; mouse Twist2, U36384; mouse Hand1 S79216; mouse Hand2, U40039, mouse Paraxis, U18658, mouse Scleraxis, S78079; mouse Ferd3l AF517121; chick Twist1, AF093816; chick Twist2, AJ13110; chick Hand1, U40041; chick Hand2, U40040, chick Paraxis, U72641; human TWIST1, U80998; human TWIST2, G01204; rat Twist1, AF266260; Xenopus Twist1, NM_204084; D. melanogaster Twist, X14569; C. elegans Twist, NP_509367; zebrafish Twist1, AF205258; zebrafish Twist2, AF205259; zebrafish Hand, AF228334; mouse HRT1(hey1) AF172286; mouse Myogenin, M84918.1.
We thank K. Dionne and I. Messina for excellent technical assistance, N. Vargesson for help with the Hand2 experiments and S. Weiner for help with in situ analyses. We also thank L. Field, C. Tabin, T. Jessell, A. Kania, D. Vasiliauskas, L. Zeltser and members of the Laufer lab for comments on the manuscript. This work was supported by the National Institutes of Health (to A.B.F., D.V., D.K.), March of Dimes Birth Defects Foundation (A.B.F.), American Cancer Society (P.C.), American Heart Association (P.C.), and Howard Hughes Medical Institute Research Resources Program for Medical Schools (E.L.).