Available online at www.sciencedirect.com

Developmental Biology 317 (2008) 686–694www.elsevier.com/developmentalbiology

Genomes & Developmental Control

Prediction and characterisation of a highly conserved, remote and cAMPresponsive enhancer that regulates Msx1 gene expression in cardiac neural

crest and outflow tract

Kerry Ann Miller, Scott Davidson, Angela Liaros, John Barrow, Marissa Lear,Danielle Heine, Stefan Hoppler, Alasdair MacKenzie⁎

School of Medical Sciences, Foresterhill, University of Aberdeen, Aberdeen AB25 2ZD, UK

Received for publication 5 July 2007; revised 25 January 2008; accepted 9 February 2008Available online 21 February 2008

Abstract

Double knockouts of the Msx1 and Msx2 genes in the mouse result in severe cardiac outflow tract malformations similar to those frequentlyfound in newborn infants. Despite the known role of the Msx genes in cardiac formation little is known of the regulatory systems (ligand receptor,signal transduction and protein–DNA interactions) that regulate the tissue-specific expression of the Msx genes in mammals during the formationof the outflow tract. In the present study we have used a combination of multi-species comparative genomics, mouse transgenic analysis andin-situ hybridisation to predict and validate the existence of a remote ultra-conserved enhancer that supports the expression of the Msx1 gene inmigrating mouse cardiac neural crest and the outflow tract primordia. Furthermore, culturing of embryonic explants derived from transgeniclines with agonists of the PKC and PKA signal transduction systems demonstrates that this remote enhancer is influenced by PKA but not PKCdependent gene regulatory systems. These studies demonstrate the efficacy of combining comparative genomics and transgenic analyses andprovide a platform for the study of the possible roles of Msx gene mis-regulation in the aetiology of congenital heart malformation.© 2008 Elsevier Inc. All rights reserved.

Keywords: Msx1 gene; Comparative genomics; Transfac; Enhancer element; Transgenic mouse; Embryo; Cardiac neural crest; Outflow tract; Transcription; PKA

Introduction

Proper morphogenesis of the outflow region relies on theventral migration of a population of cells called the cardiacneural crest (cNC) (Jiang et al., 2000). Ablation of cNC cellsderived from embryonic hindbrain rhombomeres 6, 7, and 8 inchick embryos resulted in a series of outflow tract defects thatare frequently found in newborn human infants (Waldo et al.,1999, 2005). These cNC cells contribute to the formation of theendocardial cushions of the outflow tract that subsequentlyforms the conotruncal septum that, in turn, divides the aortic andpulmonary channels and contributes to the heart valves and theupper portions of the intraventricular septum (Kirby and Waldo,1995). A class of homeobox genes called the muscle segmenthomeobox (Msx) genes are expressed in migrating cNC (Chan-

⁎ Corresponding author.E-mail address: mbi167@abdn.ac.uk (A. MacKenzie).

0012-1606/$ - see front matter © 2008 Elsevier Inc. All rights reserved.doi:10.1016/j.ydbio.2008.02.016

Thomas et al., 1993; Houzelstein et al., 1997) and have recentlybeen shown to be critical in normal morphogenesis of theoutflow tract as deletions of the Msx1 and Msx2 genes causeextensive outflow tract defects that reflect those frequently seenin newborn human infants (Ishii et al., 2005; Lallemand et al.,2005; Ogi et al., 2005). Although these genes have been shownto be critical to formation of the outflow tract virtually nothingis known of the regulatory systems that coordinate and supportthe tissue-specific expression of these genes in cardiac neuralcrest or outflow tract. The reasons for this lack of knowledgestems from a previous inability to accurately predict thelocation, functional linkage and tissue-specific characteristicsof key regulatory sequences that may possibly be located atsome distance from the genes they regulate.

The current study describes the use of multi-speciescomparative genomics and Transfac bioinformatics to predictthe location and tissue-specific characteristics of a novel enhancer.The status of this enhancer as a cNC specific enhancer of the

687K.A. Miller et al. / Developmental Biology 317 (2008) 686–694

Msx1 gene was subsequently validated using a combination oftransgenic embryo studies and in-situ hybridisation. In addition topredicting and validating many of its characteristics we alsodemonstrate, using transgenic embryonic explant culture, that thisenhancer is responsive to PKA but not PKC signalling stronglysuggesting the role of a PKA mediated signal transductionpathway in the regulation of Msx1 in cNC cells.

The prediction and validation of the tissue-specific proper-ties of this highly conserved cNC specific enhancer demon-strates the power of using predictive multi-species comparativegenomics in facilitating an understanding of the regulatorysystems controlling the expression of clinically importantgenes. Furthermore, the discovery of this ultra-conservedfunctional enhancer element will serve as a platform forunderstanding the possible effects of gene mis-regulation in theproduction of cardiac outflow tract defects and lends greaterweight to the assertion that, in addition to mutations of codingregions, regulatory regions should be examined when attempt-ing to determine the roles of key genes in the production ofdevelopmental malformations and disease.

Methods

Bioinformatic analysis

Comparative genomic analysis of non-coding DNA surrounding the Msx1gene was carried out using the ECR browser (http://ecrbrowser.dcode.org)(Ovcharenko et al., 2004). Verification of transcription factor binding matriceswas carried out using TRANSFAC professional (BIOBASE BiologicalDatabases) (Matys et al., 2003, 2006).

Plasmid constructs

p1230/KE— Primers KE.for; TAT GTT TAG CCC ACC CTG GA and KE.rev; TGA GGC TGG CCTATC TGA CT were used to amplify the Human KEelement from human placental DNA (Sigma) using a high fidelity polymerase(Expand HiFi system, Roche) and annealing temperature of 57 °C. The PCRproduct was digested with enzymes EcoRI and SpeI, and ligated into compatibleends of plasmid pGEM-7Zf (+) (Promega) to produce pGEMKE. The KE insertfrom pGEMKE was removed by digestion with SphI (made blunt end usingKlenow enzyme) and then with KpnI. This fragment was then ligated into theHindIII (made blunt end using Klenow enzyme) and KpnI restriction sites ofthe pBGZ40–1230 LacZ reporter plasmid (Yee and Rigby, 1993) to form p1230/KE. p1230/KE was linearised and released from the plasmid backbone usingKpnI and NotI prior to pronuclear injection.

Pronuclear injection

Pronuclear microinjection of 1-cell mouse embryos was performed aspreviously described (Nagy et al., 2003). Briefly, linearised p1230/KE DNA at aconcentration of 2 ng/μl was injected into the pronucleii of 1 cell embryosderived from superovulated and mated (CBA×c57BL/6)F1 females. Survivingeggs were transferred into the oviducts of CD1 pseudo pregnant mice.

Detection of transgene activity

After dissection from extra-embryonic tissue in room temperature PBS,embryos were washed briefly in standard wash (2 mMMgCl2 in PBS) and fixedin 4% paraformaldehyde (PFA) at 4 °C for 1 h. Embryos were washed usingdetergent wash (2 mMMgCl2, 0.1% sodium deoxycholate, 0.02% Nonidet P-40and 0.05% BSA in 0.1 M phosphate buffer (pH 7.3) for 3×20 min at roomtemperature. Embryos were transferred to X-gal stain (0.085% NaCl2, 5 mMK3Fe (CN)6, 5 mM K4Fe (CN)6 and 0.1% 5-bromo-4-chloro-3-indolyl-beta-D-

galactopyranoside (X-Gal); Melford Laboratories) and incubated at 37 °C for 2 hto overnight. After staining, embryos were transferred to 4% PFA at 4 °Covernight and stored in 100% methanol at −20 °C.

In-situ hybridisation

Plasmid pGEM7/Msx1 was linearised with NsiI for production of antisenseprobe. Digoxygenin (DIG) labelled antisense probes were transcribed usingcomponents of the T7 RNA polymerase Maxiscript In Vitro Transcription Kit(Ambion) as described in the manufacturers instructions with the followingmodifications. DIG labelled antisense RNA probes were transcribed in thefollowing reaction; 1× transcription buffer, 0.01 M DTT, 2 mM rNTP mix(0.5 mM ATP, 0.5 mM CTP, 0.5 mM GTP, 0.32 mM UTP and 0.18 mM dig-UTP; (Roche)), 5 μg linearised template plasmid DNA, 40 units RNAaseinhibitor, RNAasin (Roche) and 90 units of the T7 RNA polymerase innuclease-free water to a final volume of 50 μl. Whole-mount in-situhybridisation with fresh embryos was carried out as previously described(Lettice et al., 2002).

Immunohistochemistry

Embryonic tissue was fixed in 4%PFA for 1 h and left overnight in 20%OCT medium. Embryos were then orientated within pure OCT medium andfrozen. Sections were cut at −30 °C using in a Hacker Bright Clinicut cryostatand sections were mounted on Polysine-coated slides (VWR). Sections werewashed in Tris Buffered Saline (TBS) for 5 min then treated with 1% sodiumdodecyl sulfate (SDS) solution in TBS for 5 min to facilitate antigen recovery.Slides were then washed four times for 5 min in TBS with gentle rocking. Theslides were then transferred to a humidified chamber and samples were pre-incubated with TBST (TBS plus 0.1% TritonX-100 at pH 7.5) containing 10%Fetal Calf Serum (FCS) for 10 min, at room temperature, to suppress non-specific binding of the antibody. The samples were washed in TBST 3 times for5 min with gentle rocking. Consecutive sections were separately incubated withprimary antibodies Anti-Msx1/2 (Hybridoma Bank) and 40-1a (anti-β-galactosidase; Hybridoma Bank) at a dilution of 1:20 in TBST/10% FCS.After overnight incubation in a humidified chamber, unbound antibody wasremoved from samples by washing 3×5 min in TBST with gentle rocking. Allsamples, including control, were then incubated in the dark with secondaryantibody (1:200, Alexa Fluor Rabbit anti-mouse IgG 594; Molecular Probes/Invitrogen) for 1 h at room temperature. The slides were then washed in TBST3×5 mins with gentle shaking. Slides were then mounted in VECTASHIELD®mounting medium with DAPI (Vector labs) and visualised immediately with anAxioskop 2 plus microscope (Zeiss) with HBO100 FluoArc mercury lamp andGFP filters. Images were analysed using Axiovision Viewer 3.0 imagingsoftware.

Vibratome sectioning

For vibratome sectioning embryos were fixed in 4% paraformaldehyde(PFA)/PBS from 1 h to overnight at 4 °C. Embryos were equilibrated through4% and 20% sucrose/PBS solutions at 4 °C for several hours or until theembryos had sunk. Embryos were then transferred to BSA/gelatin mix (0.5%gelatin, 1% BSA and 5% sucrose in PBS) overnight or longer. Prior toembedding, embryos were removed from BSA/gelatin and excess mix removedby blotting. Embryos were fixed in 25% glutaraldehyde for no longer than 1 minand embedded in BSA/gelatin containing 0.25% glutaraldehyde. Blockscontaining embryos were then prepared for sectioning by gluing the block tothe cutting dish using cyanoacrylate based adhesives. Sections of 50 μm weretaken at speed 2–10, amplitude 8 on a Lancer Vibratome series 1000. Sectionswere floated onto a SuperFrost microscope slide (VWR) and mounted inaqueous mountant (7% gelatin, 50% glycerol and 0.1% phenol in water).

Transgenic explant agonist studies

Embryonic day 10.5 (E10.5) transgenic embryos were recovered anddivided in half transversely at a level posterior to the heart primordium in ice-cold PBS. Transgenic anterior halves were then divided sagitally using a sterile

688 K.A. Miller et al. / Developmental Biology 317 (2008) 686–694

scalpel. Transgenic explants were identified by 30 min X-gal staining of theirposterior halves. AG-1X2-formate beads (a gift from Cheryl Tickle) weresoaked for 1 h in either the adenylyle cyclase agonist forskolin (100 μM) or thePKC agonist phorbol dibutyrate (PDBu; 1 μg/ml) in dimethyl sulfoxide(DMSO). These beads were then implanted within the otic vesicles of transgenicexplants. Head explants were cultured on Millicell Culture Plate Inserts(Millipore) with D-MEM medium supplemented with 2 mM GlutaMAX(Invitrogen), 100 U/ml penicillin, 100 μg/ml streptomycin (Invitrogen) and 10%Fetal Bovine Berum (FBS; Biosera) for 24 h at 37 °C/5% CO2. Alternatively,either forskolin or PDBu were added to the medium at concentrations of10 μM and 100 ng/ml respectively. In each case similar quantities of DMSOwithout the addition of agonists were added to control halves of the same head.Following incubation for 24 h agonist effects were assessed by X-gal stainingfor 6 h. X-gal staining was stopped by multiple washes in PBS and 1 h fixationin 4% paraformaldehyde (PFA)/PBS at 4 °C.

Results

Comparative genomic analysis reveals a highly conservednon-coding sequence preserved in-cis with Msx1 for 450million years

In order to determine the location and identity of a possiblecNC specific MSX1 enhancer we used the ECR browser toundertake a comparative genomic analysis of theMSX1 flankingregions covering 100 kb in both a 5′ and a 3′ direction withreference to the human MSX1 locus. We compared the flankingsequences of the MSX1 locus to 8 other vertebrate speciesspanning 450 million years of vertebrate evolution (Blair Hedgesand Kumar, 2003) (Fig. 1A). Although comparison of human

Fig. 1. (A) A multispecies rVISTA plot from the ECR browser comparing 50 kb of hutop to bottom) zebrafish (Zeb.), puffer fish (Fugu), Xenopus (Xen.), chicken (Chick),reference to the human genome sequence. The scale bar represents 4 kb compared torepresents levels of sequence conservation between 50% and 100%. Red, blue, pink100 base pairs in intergenic non-coding, exonic, intronic and 3′ and 5′ untranslated resequence including LINEs, SINES and Alu sequences. Previously described enhancedistal enhancer (DE; MacKenzie et al., 1997; Miller et al., 2007). KE is also highligmarsupial (monodom), chicken, frog, fugu and tetradon (tetra) KE enhancer sequencefor vertebrates). Putative transcription factor binding sites as predicted by TRANSFareas of coloured sequence above each of the lines of sequences. A key for each of tS8 is also known as Prx2.

sequence with more diverged non-mammalian genomes such asthat of chicken significantly reduced the numbers of theseconserved sequences, it had the effect of greatly increasing theaccuracy of prediction. This was highlighted by the accurateverification of previously characterised enhancer elements (PEandDE, see Fig. 1A) (MacKenzie et al., 1997;Miller et al., 2007).Furthermore, accurate prediction of an active and Wnt inducibleTCF4 binding site within PE was previously possible using acombination of comparative genomics and transfac analysis(Miller et al., 2007). In addition to the PE and DE elements thepresence of a significant area of non-coding conservation over40 kb 5′ of the Msx1 transcriptional start site was highlighted.This highly conserved sequence, that we called KE, continued tobe highly conserved even in amphibians and fishes suggesting itscritical role in some biological process common to all vertebrates.A clue to its possible purpose came from the observation that in allof the species analysed, KEwas always maintained in ciswith theMsx1 gene. Considering the known high rate of genomicrearrangement events known to occur during evolution, thecontinued retention of KE in synteny with the Msx1 gene over450 million years of evolution strongly suggests that KE isfunctionally linked to the Msx1 gene (Mackenzie et al., 2004).Furthermore, recent studies have shown that over 75% of non-coding sequences demonstrating similar degrees of conservationconsist of tissue-specific cis-regulatory elements such asenhancers (Pennacchio et al., 2006). We therefore explored thehypothesis that this sequence was a tissue-specific enhancerrequired for some aspect of the regulation of the MSX1 gene.

man genomic flanking sequence 5′ of theMsx1 transcriptional start site to (fromMarsupial (Mars.), rat, mouse and dog. The x-axis represents linear distance withthe human genome sequence that has been used as the base sequence. The y-axisand yellow peaks represent areas of sequence conserved for more that 75% overgions (UTR) respectively. Green bars and grey boxes represent areas of repetitivers are highlighted in black boxes and include the proximal enhancer (PE) and thehted within a black box. (B) A linear alignment of the human, dog, mouse, rat,s demonstrating levels of conservation (highlighted in pink for tetrapods and redAC professional (matrix similarity≥0.90, core similarity=1) are highlighted ashe transcription factor binding sites is displayed in the bottom right hand corner.

689K.A. Miller et al. / Developmental Biology 317 (2008) 686–694

The highly conserved KE contained a number of perfectlyconserved binding sites for transcription factors specific toneural crest and heart cell lineages

In order to predict the possible tissue-specific properties ofKE we carried out a Transfac analysis on the sequence usinghigh search stringencies (core=1, matrix N0.9). We were able todetect the nearly perfectly conserved putative binding consensiiof a number of neural crest (Sox5 and Sox9) (Perez-Alcala et al.,2004; Sakai et al., 2006) and heart specific (GATA, NKX2.5, S8(Prx2)) (Durocher et al., 1997; Leussink et al., 1995) trans-cription factors (see Fig. 1B). This observation was consistentwith the hypothesis that KE was an MSX1 cardiac neural crestspecific enhancer.

KE is a tissue-specific enhancer and supports transgeneexpression within Msx1 expressing cNC

In order to validate our hypothesis that KE represented acNC specific enhancer sequence we amplified and isolated thehuman KE from human placental DNA. The KE sequence wasthen cloned into a lacZ reporter construct containing a LacZmarker gene (pBGZ-1230) fused to a human β-globin minimalpromoter to form the p30KE plasmid. pBGZ-1230 has beenused as a reporter plasmid several times in the past and isincapable of supporting any degree of consistent tissue-specific

Fig. 2. (A–F) Whole mount analysis of E8 (A and B), E9.5 (C and D) and E10.5 (E) pp30KE transgene construct. hb; hindbrain, hf; head fold, dl; dorsal lip, ov; otic vesicl4–6, pf; pontine flexure. White arrowheads (E) highlight the positions of cells betweentransgene.

expression (Yee and Rigby, 1993). We linearised the p30KEplasmid and produced 3 different transgenic mouse lines bypronuclear injection of 1-cell mouse embryos. In laterdevelopment, several differences were observed betweenthese lines in terms of their staining within limb buds and thedeveloping nervous system (not shown). These differences wereattributed to integration effects. However, all lines demonstratedstrong and reproducible expression in presumptive neural crestcells migrating between the embryonic hindbrain region and thedeveloping heart field in cells surrounding the otic vesicle.Because of its reproducibility this expression was directlyattributable to the activity of the KE element. Using Xgalhistological staining we analysed the expression of thetransgene throughout the early stages of embryonic develop-ment from embryonic day 8 (E8) to E12.5. At E8 expression ofthe p30KE transgene was strongest in an area of the neural headfold lip destined to form the hindbrain. Expression of thetransgene also extended ventrally and laterally within theepiblast (Figs. 2A and B). At E9.5 expression of the transgenepersisted within the dorsal lip of rhombomeres 4 and 6 of thehindbrain (Figs. 2C and D). In addition LacZ expression wasobvious in 2 different streams of presumptive neural crest cellsmigrating from rhombomeres 4 and 6 migrating either side ofthe otic vesicle (Fig. 2D). These observations are entirelyconsistent with the reported expression of a LacZ based markergene used to target the second exon of the Msx1 gene in mice

30KE transgenic mouse embryos stained with X-Gal to detect expression of thee, 1,2,3; brachial arches 1, 2 and 3, ey; optic vesicle, r4, r5 and r6; rhombomeresthe epithelium and neuroepithlium and caudal of the otic vesicle that express the

690 K.A. Miller et al. / Developmental Biology 317 (2008) 686–694

(Houzelstein et al., 1997). In the study, expression of the lacZmarker can clearly be seen in cells migrating from rhombo-meres 4 and 6 either side of the otic vesicle (Houzelstein et al.,1997). By E10.5 expression of the transgene was still obviouswithin migrating neural crest cells either side of the oticvesicle (Fig. 2E) that also expressed Msx1 transcripts (Figs.3A and B). These observations were further verified usingmonoclonal antibodies specific to the Msx1 protein (Figs. 3Dand G) and LacZ protein (Figs. 3E and H) that detected bothMsx1 and LacZ within presumptive neural crest derivedmesenchyme cells immediately caudal of the otic vesicle (Figs.

Fig. 3. (A) Vibratome section of an E10.5 mouse embryo sectioned coronally throughin-situ hybridisation using a DIG-labelled Msx1 specific anti-sense probe (purple stadetect expression from the transgene and sectioned coronally through the hindbrain.clarify the positions of each cell layer. In both panels A and B the rostro-caudal axistaken at the same level it should be noted that morphological differences betweenhybridisation. DAPI stained (C and F), and fluorescent immunohistochemical analysibacterial β-galactosidase protein (LacZ; E and H) in adjacent 8 μm cryostat sectioexpression of Msx1 and LacZ is highlighted using white arrow heads and can be seen3rd branchial artery (3ba; G and H) around which neural crest are known to migrate inexpression (purple stain) within the outflow tract (oft) of an E10.5 wild type embryo hestained whole mount preparation of an E11.5 embryo heart transgenic for the p1230/KEusing white arrow heads. (K) sectional analysis of X-Gal stained E11.5 embryonic heartlining the outflow tract (oft). The extent of transgene expression is highlighted using

3D and E) and in cells migrating next to the 3rd branchial archartery (Figs. 3G and H), a known conduit of cardiac neuralcrest migration (Sieber-Blum, 2004). We also examined theexpression of the transgene in the developing heart and wereable to detect LacZ expression within cells lining the outflowtract previously shown to express Msx1 (Chan-Thomas et al.,1993) (Figs. 3J and K). However, the expression of thetransgene in the outflow tract was detected at least a day laterthan the expression of the Msx1 gene (Figs. 3I–K) and wasnot as widespread throughout the outflow tract as previouslyreported for Msx1 transcripts.

the hindbrain region at the level of the otic vesicle and analysed by whole mountining). (B) Vibrotome section of an E10.5 p30KE embryo stained using X-gal toIn both panels A and B tissue layers are delimited using broken white lines toruns from left (rostral; Rost.) to right (caudal; Caud.). Although the sections areA and B reflect the different methods used for LacZ staining and for in-situs using monoclonal antibodies raised against the Msx1 protein (D and G) and thens of an E10.5 mouse embryo transgenic for the p1230/KE plasmid. Cellularin mesenchyme cells caudal to (D and E) the otic vesicle (ov), and adjacent to theto the heart field. (I) Whole mount in-situ hybridisation analysis ofMsx1mRNAart where the extent of expression is highlighted using black arrow heads. (J) X-galconstruct demonstrating transgene expression within the outflow tract, highlightedtransgenic for the p1230/KE construct demonstrating LacZ expression within cellswhite arrow heads. at, atrium; NT, neural tube.

691K.A. Miller et al. / Developmental Biology 317 (2008) 686–694

Tissue-specific activity of the KE element is responsive tocAMP mediated but not PKC mediated signal transductionpathways

Transfac analysis of the KE element demonstrated thepresence of several highly conserved putative binding sites ofthe Sox9 transcription factor that is known to be expressedduring, and required for, neural crest differentiation andmigration (Kordes et al., 2005). Furthermore, previous studieshave shown that the expression of Sox9 in these neural crestcells is responsive to PKA signalling (Sakai et al., 2006). Inorder to explore the hypothesis that the KE enhancer wasresponsive to cAMP mediated PKA signalling, we recoveredE10.5 embryos transgenic for the p30KE transgene and dividedtheir heads sagittally. These halves were then cultured in thepresence or absence of the adenylyl cyclase agonist forskolin. Inparallel, we also cultured sagittally divided transgenic embryoheads in the presence of the PKC agonist phorbol 12,13-dibutyrate (PDBu). Both agonists were applied either to themedium or were soaked in AG-1X2-formate beads that werethen placed within the otic vesicle of the transgenic embryonicexplant. Following overnight culture the embryo head explantswere fixed and stained for equal periods of time in equalconcentrations of X-gal. Comparisons were made betweenPDBu exposed transgenic explants and those not exposed toPDBu. No discernable difference could be detected in thestaining intensity of either transgenic explant half exposed toPDBu placed either in the medium or applied using formatebeads (data not shown). However, we were able to discern aclear and significant change in both in the intensity anddistribution of LacZ staining cells in head explants cultured withforskolin (see Fig. 4). In the case of the implanted beadselevated expression of the transgene was clearly evident in cells

Fig. 4. (A–H) X-gal stained embryo head explant cultures derived from E10.5 mouexplants that have been implanted with AG-1X2-formate beads soaked in DMSO (A a(B and D). White arrows highlight position of implanted beads within the otic vesiccontaining 100 ng/ml of forskolin in DMSO (F and H) or DMSO only (E and G; m

surrounding the bead implantation site to the extent thatdiscerning the location of the bead was difficult following LacZstaining (Figs. 4A–D). Furthermore, expression of the trans-gene was greatly enhanced throughout the explant and extendedinto regions surrounding the developing eyes and pharyngealarches by addition of forskolin to the medium. These are areasof expression not previously associated with the activity of thetransgene in the untreated explants (Figs. 4E–H) and clearlydemonstrate that the KE element is responsive to cAMP/PKAmediated pathways.

Discussion

The high degrees of conservation of regulatory regions thatmodulate the expression of developmentally important genessuggests that control of the expression of these genes has beenat least as important as the function of the proteins theyproduced during evolution (Davidson et al., 2006a,b; Mac-Kenzie and Quinn, 2004; Mackenzie et al., 2004). It has beenrealised for a number of years that comparative genomics is animportant tool in the search for sequences within the genomecritical to normal human development and health (Loots andOvcharenko, 2005; Loots et al., 2002; Ovcharenko and Loots,2003; Wasserman et al., 2000; Wei et al., 2002). Although theprimary reason for the sequencing of the human genome, aswell as a number of other vertebrate genomes, was to findnovel genes, it is now obvious that these genome sequences arealso extremely useful in detecting the regulatory elementsresponsible for controlling the expression of these genes. In thepresent study we have used comparative genomics to locate,isolate and characterise a highly conserved enhancer elementand provide evidence for the regulatory systems that modulateits activity.

se embryos transgenic for the p1230/KE construct. Panels A–D represent headnd C; bead −ve) and in DMSO containing a 10 micromolar solution of forskolinle (ov). Panels E–H represent head explants that have been cultured in mediumedium –ve). v; ventral aspect, d; dorsal aspect, e; eye, 1; first pharyngeal arch.

692 K.A. Miller et al. / Developmental Biology 317 (2008) 686–694

Clues to the regulatory systems supporting Msx1 expression indorsal neural tube lip and migrating cNC

The Msx genes have been shown to be critical components ofthe mechanisms required for normal formation of the outflowtract of the heart (Ishii et al., 2005; Lallemand et al., 2005; Ogi etal., 2005). The current study describes the discovery of a highlyconserved cardiac neural crest specific enhancer, KE, that hasbeen conserved in cis with the Msx1 gene for over 450 millionyears and supports the activity of a marker gene in dorsal neuraltube lip and presumptive cNC cells migrating from the fourth andsixth hindbrain rhombomeres that also expressMSX1 transcripts(Houzelstein et al., 1997). In addition, the regulatory systems thatsupport the tissue-specific properties of this enhancer are cAMPresponsive, suggesting a role for a PKA dependent signaltransduction pathway that may have been conserved from ourfish ancestors. The role of cAMPmediated PKA pathways in thespecification, migration and differentiation of neural crest is welldocumented (Benjanirut et al., 2006; Chen et al., 2005; Huber etal., 2003; Sakai et al., 2006). Furthermore, many of these PKAmediated regulatory systems include the involvement oftranscription factors such as Sox 5 and 9 (Perez-Alcala et al.,2004; Sakai et al., 2006). Therefore, our observation that the KEelement is not only cAMP inducible but also contains anumber of perfectly conserved putative Sox transcription factorbinding sites is consistent with our current understanding withregard to the SoxE/PKA dependent regulatory systems knownto modulate the specification of cardiac neural crest (Perez-Alcala et al., 2004; Sakai et al., 2006). Thus, our data, incombination with the published literature, suggests a role forthe KE element in the proper morphogenesis of the heart.

A role for KE in heart morphogenesis

Understanding where KE is important in heart developmentnecessitates an understanding of the origins and destination of thecNC. The role of the cardiac neural crest in morphogenesis of theoutflow tract of the heart can be broken down into 3 main steps;specification, migration and differentiation (Hutson and Kirby,2003). Defects in the molecular pathways controlling each ofthese steps have been shown to result in outflow tract defects.Specification of the neural crest from the dorsal lip of the neuraltube involves a morphological change from an epithelial cell typeto that of a migratory cell type (Firulli and Conway, 2004; Jung etal., 2005). A number of different studies have implicated theMsx1 gene as being amajor player in neural crest cell specification(Burstyn-Cohen et al., 2004; Monsoro-Burq et al., 2005; Tribuloet al., 2003) along with members of the SoxE family (Sox 8, 9 and10 (Kordes et al., 2005)), the Snail family, FoxD3, Lbx1, c-Mycand Pax3 (Barembaum and Bronner-Fraser, 2005). In the currentstudy we demonstrate that KE is active in the neural tube dorsallip region at E8 which is whereMsx1would need to be expressedin order to play a role in neural crest specification. Thus, the highdegrees of evolutionary conservation of the KE element, theputative transcription factor binding sites contained within it andits activity in the early dorsal neural tube lip supports a role forMsx1 in the specification of cardiac neural crest.

In addition to being expressed in the neural plate dorsal lipduring neural crest specification,Msx1 continues to be expressedin neural crest cells undergoing migration in chick embryos(Taneyhill and Bronner-Fraser, 2005) an observation confirmedin mammalian embryos in the current study. Further evidencecomes from previous examinations of the expression of a LacZmarker gene used to target the Msx1 locus where it was shownthat LacZ activity was present within migrating neural crest cellsderived from rhombomeres 4 and 6 (Houzelstein et al., 1997).Based on this existing literature and the findings described herewe propose that the KE element is required to support theexpression of the Msx1 gene within cardiac neural crest cellsduring their specification and migration where Msx1, in its roleas a potent antagonist of cell differentiation (Odelberg et al.,2000; Thompson-Jaeger and Raghow, 2000), may be required toact as a brake to differentiation thus supporting cell migration.The mechanism through which Msx1 achieves this mightinvolve a direct interaction with the Pax3 protein that is alsoexpressed in dorsal neural tube lip and migrating cNC and is alsoknown to be required for cNC specification (Epstein et al.,2000). Msx1 and Pax3 have been shown to cooperate in thespecification of neural crest (Monsoro-Burq et al., 2005). Inaddition, previous studies have suggested that within other organsystems the Msx1 protein prevents cellular differentiation bydirectly binding to the Pax3 protein (Bendall et al., 1999;Bendall et al., 1998). Such a relationship between Pax3 andMsx1 may also exist within migrating cardiac neural crest cellswhere Pax3 may be required to specify cNC (Epstein et al.,2000) but whereMsx1 is required to modulate this activity of thePax3 protein thus allowing cell migration.

We have succeeded in detecting the activity of the KE elementin presumptive dorsal neural tube lip cells and in migratingcardiac neural crest cells, that we show also express Msx1transcripts at E.10.5, consistent with a role for KE in supportingMsx1 expression during cNC specification and migration.However, the observed activity of KE in the developingendocardiac cushions during formation of the outflow tract wasless conclusive. For example, Msx1 is strongly expressed in theendocardiac cushions from E10.5 onwards but we were unable todetect activity of KE within the outflow tract until E11.5. Theseobservations suggest that, although important for Msx1 expres-sion in the specification and migration of cNC, KE may not playsuch an important role in the expression ofMsx1 in differentiatingendocardiac cushions. A number of studies have demonstratedthat the expression of theMsx1 gene during the morphogenesis ofthe limb buds and neural tube roof plate is associated with, and issupported by, the expression of bone morphogenic proteins(BMPs) (Lallemand et al., 2005). In addition, there have been anumber of studies which suggest that the expression of the Msxgenes in the endocardial cushions of the differentiating heartoutflow tract is supported by BMP mediated regulatory systems(Abdelwahid et al., 2001; Angello et al., 2006; Eisenberg andMarkwald, 1995). Thus, we hypothesise that at least twoenhancers are required for the expression of Msx1 during thedevelopment of the OFT. We believe that we have identified oneof these enhancers in the form of the highly conserved and cAMPdependent KE element that may be required for expression of

693K.A. Miller et al. / Developmental Biology 317 (2008) 686–694

Msx1 during early specification and migration of cNC cells.Thus, the requirement for expression of Msx1 in migrating cNCis an ancient one and is reflected by the exceptional degrees ofKE conservation. However, we believe that a second enhancer,that supports Msx1 in the endocardiac cushions during theirdifferentiation, remains to be identified. We hypothesise that thissecond enhancer is likely to be BMP dependent and is requiredfor Msx1 expression in the endocardiac cushions. We wouldfurther venture that this second enhancer will not be so easy toidentify as it will not be as highly conserved as KE. This lack ofconservation will reflect its role in the later stages of themorphogenesis of the mammalian outflow tract that is acomparatively recent adaptation within the hearts of birds andmammals from an evolutionary perspective. We are currentlyextending our analysis to determine the possible location of thiselement.

Although Msx1 has been shown to be expressed in cardiacneural crest during their specification, migration and differentia-tion and is, together with Msx2, critical for normal morphogen-esis of OFT, the precise stage at which these genes are mostcrucial to OFT development remains to be determined. Therequirement forMsx1 in the specification and migration of cNCcells is suggested by the extreme conservation of the KE elementwhich our studies suggest supports expression of Msx1 in cNCcells during these stages. This observation is not withoutprecedence as although it has been shown that the Pax3 gene isessential for the normal formation of the OFT migrating cNCcells stop expressing Pax3 before their arrival in the heart(Epstein et al., 2000). Thus it is entirely feasible that because therequirement for the expression ofMsx1 in the dorsal neural tubelip and in migrating cardiac neural crest cells precludes theexpression of Msx1 in cardiac outflow tract that expression ofMsx1 in dorsal lip and migrating cNC represent the mostimportant stages of Msx1 expression required for normal heartdevelopment. Therefore, concentrating exclusively on the regu-latory mechanisms governing the expression of Msx1 in theendocardial cushions may be a mistake as many outflow tractdefects might result from problems in the specification and mig-ration of cNC rather than the formation of endocardial cushions.

It has been largely accepted that theMsx1 protein is a classictranscription factor that influences gene activity through DNAbinding. The two mouse Msx1 gene deletion models previouslyproduced reflected this belief and, in both cases, only sequence3′ and including the third helix of the homeodomain weredeleted leaving the rest of the homeodomain intact (Houzelsteinet al., 1997; Satokata and Maas, 1994). More recent evidencenow suggests that another mode of action of the Msx1 proteinincludes protein–protein interactions with other transcriptionfactor proteins such as the Pax, Dlx and Lhx genes via the N-terminus and the first helix of the homoedomain (Bendall et al.,1998, 1999; Bryan and Morasso, 2000; Zhang et al., 1997).Thus, existing Msx1 targeted deletion models may not fullyreflect the developmental significance of the Msx1 gene in theinhibition of gene expression through protein–protein interac-tion with known cNC specification factors such as Lhx1 andPax3. This possibility warrants further analysis through theproduction of a complete Msx1 null mouse model.

Conclusions

OFT defects make up one of the most frequent class of birthdefects affecting newborn infants. Understanding the causes ofsusceptibility to OFT defects through genetic analysis of patientDNA through linkage and association studies of DNApolymorphisms is one of the most important directions wecan take towards understanding how outflow tract defects occur.Until recently, however, much of the effort into understandingthe causes of OFT defects has been directed against the codingregions of genes. The current study highlights the extremeconservation of a regulatory element responsible for supportingthe expression of a gene known to be involved in OFTformation and underlines the need to look at regulatory regions,as well as coding regions when looking for the polymorphismsthat affect susceptibility to OFT defects.

Acknowledgments

KAM and SD were supported by BBSRC studentships. Thisproject was funded by Tenovus Scotland and the BBSRC.

References

Abdelwahid, E., Rice, D., Pelliniemi, L.J., Jokinen, E., 2001. Overlapping anddifferential localization of Bmp-2, Bmp-4, Msx-2 and apoptosis in theendocardial cushion and adjacent tissues of the developing mouse heart. CellTissue Res. 305, 67–78.

Angello, J.C., Kaestner, S., Welikson, R.E., Buskin, J.N., Hauschka, S.D., 2006.BMP induction of cardiogenesis in P19 cells requires prior cell–cellinteraction(s). Dev. Dyn. 235, 2122–2133.

Barembaum, M., Bronner-Fraser, M., 2005. Early steps in neural crestspecification. Semin. Cell Dev. Biol. 16, 642–646.

Bendall, A.J., Rincon-Limas, D.E., Botas, J., Abate-Shen, C., 1998. Proteincomplex formation between Msx1 and Lhx2 homeoproteins is incompatiblewith DNA binding activity. Differentiation 63, 151–157.

Bendall, A.J., Ding, J., Hu, G., Shen, M.M., Abate-Shen, C., 1999. Msx1antagonizes the myogenic activity of Pax3 in migrating limb muscleprecursors. Development 126, 4965–4976.

Benjanirut, C., Paris, M., Wang, W.H., Hong, S.J., Kim, K.S., Hullinger,R.L., Andrisani, O.M., 2006. The cAMP pathway in combination withBMP2 regulates Phox2a transcription via cAMP response elementbinding sites. J. Biol. Chem. 281, 2969–2981.

Blair Hedges, S., Kumar, S., 2003. Genomic clocks and evolutionary timescales.Trends Genet. 19, 200–206.

Bryan, J.T., Morasso, M.I., 2000. The Dlx3 protein harbors basic residuesrequired for nuclear localization, transcriptional activity and binding toMsx1. J. Cell Sci. 113, 4013–4023.

Burstyn-Cohen, T., Stanleigh, J., Sela-Donenfeld, D., Kalcheim, C., 2004.Canonical Wnt activity regulates trunk neural crest delamination linkingBMP/noggin signaling with G1/S transition. Development 131, 5327–5339.

Chan-Thomas, P.S., Thompson, R.P., Robert, B., Yacoub, M.H., Barton, P.J.,1993. Expression of homeobox genes Msx-1 (Hox-7) and Msx-2 (Hox-8)during cardiac development in the chick. Dev. Dyn. 197, 203–216.

Chen, S., Ji, M., Paris, M., Hullinger, R.L., Andrisani, O.M., 2005. The cAMPpathway regulates both transcription and activity of the paired homeoboxtranscription factor Phox2a required for development of neural crest-derivedand central nervous system-derived catecholaminergic neurons. J. Biol.Chem. 280, 41025–41036.

Davidson, S.,Miller, K.A., Dowell, A., Gildea, A.,MacKenzie, A., 2006a. Cellularco-expression of a LacZ marker gene driven by the amygdala-specific ECR1enhancer with the substance P neuropeptide. Mol. Psychiatry 11, 323.

Davidson, S., Miller, K.A., Dowell, A., Gildea, A., MacKenzie, A., 2006b.

694 K.A. Miller et al. / Developmental Biology 317 (2008) 686–694

A remote and highly conserved enhancer supports amygdala specificexpression of the gene encoding the anxiogenic neuropeptide substance-P.Mol. Psychiatry 11, 410–421.

Durocher, D., Charron, F., Warren, R., Schwartz, R.J., Nemer, M., 1997. Thecardiac transcription factors Nkx2–5 and GATA-4 are mutual cofactors.EMBO J. 16, 5687–5696.

Eisenberg, L.M., Markwald, R.R., 1995. Molecular regulation of atrioventri-cular valvuloseptal morphogenesis. Circ. Res. 77, 1–6.

Epstein, J.A., Li, J., Lang, D., Chen, F., Brown, C.B., Jin, F., Lu, M.M., Thomas,M., Liu, E., Wessels, A., Lo, C.W., 2000. Migration of cardiac neural crestcells in Splotch embryos. Development 127, 1869–1878.

Firulli, A.B., Conway, S.J., 2004. Combinatorial transcriptional interactionwithin the cardiac neural crest: a pair of HANDs in heart formation. BirthDefects Res. C Embryo Today 72, 151–161.

Houzelstein, D., Cohen, A., Buckingham, M.E., Robert, B., 1997. Insertionalmutation of the mouse Msx1 homeobox gene by an nlacZ reporter gene.Mech. Dev. 65, 123–133.

Huber, W.E., Price, E.R., Widlund, H.R., Du, J., Davis, I.J., Wegner, M., Fisher,D.E., 2003. A tissue-restricted cAMP transcriptional response: SOX10modulates alpha-melanocyte-stimulating hormone-triggered expression ofmicrophthalmia-associated transcription factor in melanocytes. J. Biol.Chem. 278, 45224–45230.

Hutson, M.R., Kirby, M.L., 2003. Neural crest and cardiovascular development:a 20-year perspective. Birth Defects Res. C Embryo Today 69, 2–13.

Ishii, M., Han, J., Yen, H.Y., Sucov, H.M., Chai, Y., Maxson Jr., R.E., 2005.Combined deficiencies of Msx1 and Msx2 cause impaired patterning andsurvival of the cranial neural crest. Development 132, 4937–4950.

Jiang, X., Rowitch, D.H., Soriano, P., McMahon, A.P., Sucov, H.M., 2000. Fateof the mammalian cardiac neural crest. Development 127, 1607–1616.

Jung, J., Mysliwiec, M.R., Lee, Y., 2005. Roles of JUMONJI in mouseembryonic development. Dev. Dyn. 232, 21–32.

Kirby, M.L., Waldo, K.L., 1995. Neural crest and cardiovascular patterning.Circ. Res. 77, 211–215.

Kordes, U., Cheng, Y.C., Scotting, P.J., 2005. Sox group E gene expressiondistinguishes different types and maturational stages of glial cells indeveloping chick and mouse. Brain Res. Dev. Brain Res. 157, 209–213.

Lallemand, Y., Nicola, M.A., Ramos, C., Bach, A., Cloment, C.S., Robert, B.,2005. Analysis of Msx1; Msx2 double mutants reveals multiple roles forMsx genes in limb development. Development 132, 3003–3014.

Lettice, L.A., Horikoshi, T., Heaney, S.J., van Baren, M.J., van der Linde, H.C.,Breedveld, G.J., Joosse, M., Akarsu, N., Oostra, B.A., Endo, N., Shibata,M., Suzuki, M., Takahashi, E., Shinka, T., Nakahori, Y., Ayusawa, D.,Nakabayashi, K., Scherer, S.W., Heutink, P., Hill, R.E., Noji, S., 2002.Disruption of a long-range cis-acting regulator for Shh causes preaxialpolydactyly. Proc. Natl. Acad. Sci. U. S. A. 99, 7548–7553.

Leussink, B., Brouwer, A., el Khattabi, M., Poelmann, R.E., Gittenberger-deGroot, A.C., Meijlink, F., 1995. Expression patterns of the paired-relatedhomeobox genes MHox/Prx1 and S8/Prx2 suggest roles in development ofthe heart and the forebrain. Mech. Dev. 52, 51–64.

Loots, G.G., Ovcharenko, I., 2005. Dcode.org anthology of comparativegenomic tools. Nucleic Acids Res. 33, W56–W64.

Loots, G.G., Ovcharenko, I., Pachter, L., Dubchak, I., Rubin, E.M., 2002. rVistafor comparative sequence-based discovery of functional transcription factorbinding sites. Genome Res. 12, 832–839 [doi].

MacKenzie, A., Quinn, J.P., 2004. Post-genomic approaches to exploringneuropeptide gene mis-expression in disease. Neuropeptides 38, 1–15.

MacKenzie, A., Purdie, L., Davidson, D., Collinson, M., Hill, R.E., 1997. Twoenhancer domains control early aspects of the complex expression pattern ofMsx1. Mech. Dev. 62, 29–40.

MacKenzie, A., Miller, K.A., Collinson, J.M., 2004. Is there a functional linkbetween gene interdigitation and multi-species conservation of syntenyblocks? BioEssays 26, 1217–1224.

Matys, V., Fricke, E., Geffers, R., Gossling, E., Haubrock, M., Hehl, R.,Hornischer, K., Karas, D., Kel, A.E., Kel-Margoulis, O.V., Kloos, D.U., Land,S., Lewicki-Potapov, B., Michael, H., Munch, R., Reuter, I., Rotert, S., Saxel,H., Scheer, M., Thiele, S., Wingender, E., 2003. TRANSFAC: transcriptionalregulation, from patterns to profiles. Nucleic Acids Res. 31, 374–378.

Matys, V., Kel-Margoulis, O.V., Fricke, E., Liebich, I., Land, S., Barre-Dirrie,

A., Reuter, I., Chekmenev, D., Krull, M., Hornischer, K., Voss, N.,Stegmaier, P., Lewicki-Potapov, B., Saxel, H., Kel, A.E., Wingender, E.,2006. TRANSFAC and its module TRANSCompel: transcriptional generegulation in eukaryotes. Nucleic Acids Res. 34, D108–D110.

Miller, K.A., Barrow, J., Collinson, J.M., Davidson, S., Lear, M., Hill, R.E.,MacKenzie, A., 2007. A highly conserved Wnt-dependent TCF4 bindingsite within the proximal enhancer of the anti-myogenic Msx1 gene supportsexpression within Pax3-expressing limb bud muscle precursor cells. Dev.Biol. 311, 665–678.

Monsoro-Burq, A.H., Wang, E., Harland, R., 2005. Msx1 and Pax3 cooperate tomediate FGF8 and WNT signals during Xenopus neural crest induction.Dev. Cell 8, 167–178.

Nagy, A.G., M., Vinterstein, K., Behringer, R., 2003. Manipulating the MouseEmbryo. Cold Spring Harbor laboratory Press, Cold Spring Harbor.

Odelberg, S.J., Kollhoff, A., Keating, M.T., 2000. Dedifferentiation ofmammalian myotubes induced by msx1. Cell 103, 1099–1109.

Ogi, H., Suzuki, K., Ogino, Y., Kamimura, M., Miyado, M., Ying, X., Zhang, Z.,Shinohara, M., Chen, Y., Yamada, G., 2005. Ventral abdominal walldysmorphogenesis of Msx1/Msx2 double-mutant mice. Anat. Rec. ADiscov. Mol. Cell Evol. Biol. 284, 424–430.

Ovcharenko, I., Loots, G.G., 2003. Comparative genomic tools for exploring thehuman genome. Cold Spring Harbor Symp. Quant. Biol. 68, 283–291.

Ovcharenko, I., Nobrega, M.A., Loots, G.G., Stubbs, L., 2004. ECR Browser: atool for visualizing and accessing data from comparisons of multiplevertebrate genomes. Nucleic Acids Res. 32, W280–W286.

Pennacchio, L.A., Ahituv, N., Moses, A.M., Prabhakar, S., Nobrega, M.A.,Shoukry, M., Minovitsky, S., Dubchak, I., Holt, A., Lewis, K.D., Plajzer-Frick, I., Akiyama, J., De Val, S., Afzal, V., Black, B.L., Couronne, O.,Eisen, M.B., Visel, A., Rubin, E.M., 2006. In vivo enhancer analysis ofhuman conserved non-coding sequences. Nature 444, 499–502.

Perez-Alcala, S., Nieto, M.A., Barbas, J.A., 2004. LSox5 regulates RhoBexpression in the neural tube and promotes generation of the neural crest.Development 131, 4455–4465.

Sakai, D., Suzuki, T., Osumi, N., Wakamatsu, Y., 2006. Cooperative action ofSox9, Snail2 and PKA signaling in early neural crest development.Development 133, 1323–1333.

Satokata, I., Maas, R., 1994. Msx1 deficient mice exhibit cleft palate andabnormalities of craniofacial and tooth development. Nat. Genet. 6, 348–356.

Sieber-Blum, M., 2004. Cardiac neural crest stem cells. Anat. Rec. A Discov.Mol. Cell Evol. Biol. 276, 34–42.

Taneyhill, L.A., Bronner-Fraser, M., 2005. Dynamic alterations in geneexpression after Wnt-mediated induction of avian neural crest. Mol. Biol.Cell 16, 5283–5293.

Thompson-Jaeger, S., Raghow, R., 2000. Exogenous expression ofMsx1 rendersmyoblasts refractory to differentiation into myotubes and elicits enhancedbiosynthesis of four unique mRNAs. Mol. Cell. Biochem. 208, 63–69.

Tribulo, C., Aybar, M.J., Nguyen, V.H., Mullins, M.C., Mayor, R., 2003.Regulation of Msx genes by a Bmp gradient is essential for neural crestspecification. Development 130, 6441–6452.

Waldo, K., Zdanowicz, M., Burch, J., Kumiski, D.H., Stadt, H.A., Godt, R.E.,Creazzo, T.L., Kirby, M.L., 1999. A novel role for cardiac neural crest inheart development. J. Clin. Invest. 103, 1499–1507.

Waldo, K.L., Hutson, M.R., Stadt, H.A., Zdanowicz, M., Zdanowicz, J., Kirby,M.L., 2005. Cardiac neural crest is necessary for normal addition of themyocardium to the arterial pole from the secondary heart field. Dev. Biol.281, 66–77.

Wasserman, W.W., Palumbo, M., Thompson, W., Fickett, J.W., Lawrence, C.E.,2000. Human–mouse genome comparisons to locate regulatory sites. Nat.Genet. 26, 225–228.

Wei, L., Liu, Y., Dubchak, I., Shon, J., Park, J., 2002. Comparative genomicsapproaches to study organism similarities and differences. J. Biomed.Inform. 35, 142–150.

Yee, S.P., Rigby, P.W., 1993. The regulation of myogenin gene expressionduring the embryonic development of the mouse. Genes Dev. 7,1277–1289.

Zhang, H., Hu, G., Wang, H., Sciavolino, P., Iler, N., Shen, M.M., Abate-Shen,C., 1997. Heterodimerization of Msx and Dlx homeoproteins results infunctional antagonism. Mol. Cell. Biol. 17, 2920–2932.