Developmental Cell

Article

Mechanistic Differences in the TranscriptionalInterpretation of Local and Long-RangeShh Morphogen SignalingTony Oosterveen,1,3,* Sanja Kurdija,1,3 Zhanna Alekseenko,1 Christopher W. Uhde,1 Maria Bergsland,2

Magnus Sandberg,1,2 Elisabet Andersson,1 Jose M. Dias,1 Jonas Muhr,1,2 and Johan Ericson1,*1Department of Cell and Molecular Biology, Karolinska Institutet, 171 77 Stockholm, Sweden2Ludwig Institute for Cancer Research, Box 240, 171 77 Stockholm, Sweden3These authors contributed equally to this work*Correspondence: oosterveen.t@gmail.com (T.O.), johan.ericson@ki.se (J.E.)http://dx.doi.org/10.1016/j.devcel.2012.09.015

SUMMARY

Morphogens orchestrate tissue patterning in aconcentration-dependent fashion during vertebrateembryogenesis, yet little is known of how positionalinformation provided by such signals is translatedinto discrete transcriptional outputs. Here we haveidentified and characterized cis-regulatory modules(CRMs) of genes operating downstream of gradedShh signaling and bifunctional Gli proteins in neuralpatterning. Unexpectedly, we find that Gli activatorshave a noninstructive role in long-range patterningand cooperate with SoxB1 proteins to facilitate alargely concentration-independent mode of geneactivation. Instead, the opposing Gli-repressorgradient is interpreted at transcriptional levels, and,together with CRM-specific repressive input ofhomeodomain proteins, comprises a repressive net-work that translates graded Shh signaling intoregional gene expression patterns. Moreover, localand long-range interpretation of Shh signaling differswith respect to CRM context sensitivity and Gli-activator dependence, and we propose that thesedifferences provide insight into how morphogenfunction may have mechanistically evolved from aninitially binary inductive event.

INTRODUCTION

The secreted protein Sonic hedgehog (Shh) acts in a gradedfashion at long range to establish cell pattern in the ventral neuraltube and developing limb bud and serves as a paradigm ofmorphogen function (Jessell, 2000; Wijgerde et al., 2002).However, studies of morphogens in vertebrates have focusedprimarily on the formation of extracellular gradients and geneticanalyses of signal transduction components (Lander, 2007),whereas little is known of how graded signaling is translated atthe transcriptional level into discrete patterns of gene expression(Dessaud et al., 2008).

Shh signaling is initiated by binding of the Shh ligand to itsreceptor Patched (Ptc), relieving inhibition of Smoothened

(Smo) to result in nuclear translocation of Gli (Gli1–3) transcrip-tion factors (TFs) that bind specific genomic sites to activatetarget genes (Hallikas et al., 2006; Vokes et al., 2007, 2008;Dessaud et al., 2008). While Shh stabilizes full-length Gli2 andGli3 proteins in their activator forms (GliA), in the absence ofligand these bifunctional proteins are processed to transcrip-tional repressors (GliR) (Dessaud et al., 2008). The ratio of GliRto GliA within cells can thereby be modulated in accordancewith the level of Shh pathway activation, establishing opposingintrinsic gradients of GliA and GliR across responding tissues(Ingham and Placzek, 2006).In neural patterning, the Shh gradient emanates from ventral

midline cells and a key role is to regulate spatial expression ofcell fate-determining homeodomain (HD) and basic helix-loop-helix (bHLH) TFs in responding neural progenitors, establishingfive discrete ventral progenitor domains, each of which givesrise to a distinct neuronal subtype (Briscoe et al., 2000; Dessaudet al., 2008). These Shh-regulated TFs are termed class I and IIproteins, depending on whether they are repressed or inducedby Shh, respectively, and their expression has been used asthe primary readout in studies of Shh morphogen activity (Des-saud et al., 2008). However, with the exception of the HD TFNkx2.2 (Lei et al., 2006; Lek et al., 2010), gene regulatoryelements that respond to Shh signaling have not been identifiedfor these TFs. It therefore remains unknownwhether these genesare directly regulated by Gli proteins and the extent to whichother transcriptional regulators influence the regional expressionof these genes. For instance, selective cross-repressive interac-tions between class I and II TFs are important to refine andmaintain ventral progenitor domains (Briscoe et al., 2000; Muhret al., 2001), but whether such interactions are direct and howthey are integrated at the genomic level with input from theShh pathway remain unresolved.Genetic studies have established that the activity of Gli

proteins influences the expression profiles of class I and II TFsand, consequently, ventral pattern formation (Bai et al., 2004;Lei et al., 2004), but partial redundancy between Gli2 and Gli3,together with the bifunctional nature of these proteins, hasmade it difficult to resolve the degree to which GliA and/or GliRprovide positional information at the gene regulatory level(Ingham and Placzek, 2006). Gain-of-function experiments, forexample, suggest a model in which the intrinsic GliA gradientis directly interpreted in the ventral neural tube, with GliR actingprimarily to repress genes at dorsal positions (Stamataki et al.,

1006 Developmental Cell 23, 1006–1019, November 13, 2012 ª2012 Elsevier Inc.

2005). However, certain Shh-regulated class II TFs are activatedin Gli2/Gli3 compound mutants that lack any form of GliA orGliR activity (Bai et al., 2004; Lei et al., 2004), as well as inSmo/Gli3 mutant mice (Wijgerde et al., 2002). These datademonstrate that certain Shh target genes become derepressedwhen all Gli activity is eliminated in neural progenitors, and it hastherefore been alternatively suggested that Shh target geneexpression is regulated by the balance between GliA and GliR(Bai et al., 2004; Ingham and Placzek, 2006). Derepression oftarget genes in these mice implies, moreover, that activatorproteins distinct from GliA could also be integrated in the ventralpatterning process, but the identity and function of these remainelusive.Transcriptional activity of genes is largely determined by the

genomic architecture of cis-regulatory modules (CRMs), andsystematic analyses of CRMs have had significant impact onthe mechanistic understanding of patterning processes in seaurchin and fruit fly development (Stathopoulos and Levine,2005; Davidson, 2010). While a comprehensive genomic anal-ysis of hedgehog-regulated CRMs is lacking, a study of twohedgehog-regulated CRMs active in Drosophila has suggestedthat clustered Ci-binding sites of distinct affinities mediate inter-pretation of a repressor gradient of Ci, the invertebrate homologof Gli, to constrain gene expression differentially (Parker et al.,2011). However, computational modeling of the expressionprofiles of Nkx2.2, Olig2, and Pax6 in various mouse mutantshave argued that these Shh-regulated genes may not directlyinterpret Gli activity levels and suggested instead that thecross-repressive relationship between these TFs underlies theirpatterned expression (Balaskas et al., 2012). These data suggestthat important mechanistic differences may exist betweenthe transcriptional interpretation of Hedgehog gradients inDrosophila and in the vertebrate neural tube. It is notable,however, that vertebrate models of Shh interpretation (Baiet al., 2004; Lei et al., 2004; Dessaud et al., 2007; Balaskaset al., 2012) have yet to be validated directly at gene regulatorylevels. A systematic identification and functional characterization

of Shh-regulated CRMs would likely provide more detailed infor-mation of how Shh is interpreted in the neural tube.

RESULTS

Identification of Shh-Responsive cis-RegulatoryModules Active in the Developing CNSTo examine whether HD or bHLH TFs (Figure 1A; Dessaud et al.,2008) that operate downstream of Shh in neural patterningare directly regulated by the Shh-Gli pathway, we defined aconsensus Gli-binding site (GBS) based on available data invertebrates and Drosophila (Supplemental Information availableonline). This was then used to identify GBSs in evolutionarilyconserved noncoding DNA associated with loci encoding theHD genes Nkx2.2, Nkx2.9, Nkx6.1, Nkx6.2, Dbx1, Dbx2, Pax6,and the bHLH geneOlig2. By this approach, we identified at leastone conserved noncoding element for each locus examined thatcontains one or more conserved GBS, and these conservedsequences varied between!400 and 2,100 bp in length (Figures1B, S1A, and S1B).To test whether the identified elements function as CRMs of

these genes, mouse genomic fragments were isolated andcloned into reporter vectors consisting of a minimal b-globinpromoter and either the eGFP or LacZ reporter genes.Constructs were denoted CRMNkx2.2, CRMNkx2.9, CRMOlig2,CRMNkx6.1, CRMNkx6.2, CRMDbx2, CRMDbx1, and CRMPax6.CRM-LacZ reporter constructs were used to determineenhancer activity in vivo after unilateral electroporation into theneural tube of Hamburger-Hamilton (HH) stage 11 chickembryos (Figure S1C). Embryos were incubated and harvestedat 40 hr postelectroporation (hpe) and analyzed for LacZ expres-sion (shown in white in Figures 1C and S1B). CRM activity wascompared to the expression of endogenous genes as deter-mined by in situ hybridization (Figure 1C), and a hybrid CMV/b-actin promoter-driven eGFP reporter vector was used as aninternal expression control (shown in blue in Figures 1C andS1B). By this approach, one functional element with enhancer

Figure 1. Identification of Neural CRMs Associated with Genes Encoding Class I and II TFs(A) Schematic illustrating relative expression domains of class I and II TFs in the ventral neural tube.

(B) Murine class I and II genomic loci in a 20 kb UCSC genome browser window. Coordinates provided in mm9. Peaks: evolutionary conservation of a 30-Way

Multiz alignment; blocks: evolutionarily conserved regions between mouse and chick/frog; red peaks and blocks contain evolutionarily conserved GBSs; open

rectangles: CRMs; TSS: distance between CRM and transcription start site. Open reading frames and genes depicted in blue.

(C) Expression of class I/II genes and the activities of their respective CRMs (white) and electroporation control (blue) on consecutive sections. +, electroporated

side; ", control side.

See also Figure S1.

Developmental Cell

Transcriptional Interpretation of Shh Signaling

Developmental Cell 23, 1006–1019, November 13, 2012 ª2012 Elsevier Inc. 1007

activity was identified for each of the studied HD and bHLH TFs(Figures 1C and S1B). Importantly, the expression pattern ofeach CRM was similar to its respective endogenous gene (Fig-ure 1C), indicating that they contain sufficient cis-regulatoryinformation to faithfully recapitulate the corresponding expres-sion patterns. In addition to the above stated CRMs, we identi-fied a second element with a conserved GBS for each ofNkx6.1 and Dbx2 (see Supplemental Information), but theseelements were inactive in the neural tube and not studied further.Forced in vivo expression of a constitutively active form of Smo(termed SmoM2; Xie et al., 1998) ectopically activated CRMs ofShh-induced class II genes (Nkx2.2, Olig2, Nkx6.1, and Nkx6.2),whereas CRMDbx1, CRMDbx2, and CRMPax6 were repressed

(Figure 2A). Thus, the identified CRMs respond in vivo to Shhpathway activation in a fashion analogous to endogenous class Iand II genes (Figure 2A).We next examined the activity of GBSs by altering base pairs

known to be essential for binding of Gli proteins to DNA (Pavle-tich and Pabo, 1993). As an internal control, mutations wereintroduced in CRM-LacZ vectors (Figure 2B, shown in white)and the in vivo activity of mutated vectors was compared to cor-responding wild-type CRM-eGFP constructs (Figure 2B, shownin blue). Inactivation of the single GBS in CRMNkx2.2 and the threeGBSs in CRMOlig2 extinguished LacZ expression (Figure 2B; Leiet al., 2006). In CRMNkx6.1, inactivation of a highly conservedGBS (GBS1) completely abolished its activity in vivo (Figures

Figure 2. CRMs of Class I/II Genes Contain Functional GBSs and Are Regulated by Shh Signaling(A) Expression of class I/II genes and CRMs 24 hr after SmoM2 electroporation. Dbx2 is completely repressed by 40 hpe (data not shown).

(B) Mutational analysis of GBS function. Upper panel outlines CRMs and GBSs studied. For each CRM, lower panel shows half the neural tube electroporated

with either two wild-type (WT) CRMs driving the expression of distinct reporter genes (left, white and blue) or mutated (mG) andWT control CRMs (right, white and

blue, respectively).

See also Figure S2.

Developmental Cell

Transcriptional Interpretation of Shh Signaling

1008 Developmental Cell 23, 1006–1019, November 13, 2012 ª2012 Elsevier Inc.

2B and S2A). CRMNkx6.1 also contained a second less conservedGBS (GBS2) (Figure S2A), inactivation of which reduced, but didnot eliminate, the in vivo activity of CRMNkx6.1 (Figures 2B andS2B). Collectively, these data suggest that Gli activators(GliA) are required for activation of CRMNkx2.2, CRMNkx6.1, andCRMOlig2 in vivo; however, this does not resolve whether therequirement of GliA in neural patterning is instructive or permis-sive. Interestingly, whilemutation of the three GBSs in CRMNkx6.2

resulted in reduced ventral expression, it also led to significantderepression in the dorsal neural tube (Figure 2B), as was thecase upon inactivation of the uniqueGBS inCRMDbx1 (Figure 2B).The ventral boundary of CRMDbx1 was not affected by GBSinactivation (Figure 2B), consistent with studies suggestingthat ventral repression of Dbx genes by Shh is indirectlymediated by induction of Nkx6.1 and Nkx6.2 (Vallstedt et al.,2001). These data suggest a direct requirement of GliR to deter-mine the regional expression of CRMDbx1 and CRMNkx6.2 in theneural tube.

Gli-Mediated Repression Provides PositionalInformation via Differential Binding to QualitativelyDistinct GBSsWe next wished to determine whether functional GBSs in CRMshave differential affinity properties and whether this could influ-ence the transcriptional interpretation of the Shh gradient. Wefirst calculated the positional weight matrix (PWM) score ofGBSs (Figure 3A; Hallikas et al., 2006) in CRMs associatedwith neural patterning genes, as well as of a previously identifiedGBS required for floor plate (FP) expression of FoxA2 (GBSFoxA2;Sasaki et al., 1997). Nkx2.2, Nkx2.9, and FoxA2 are locallyinduced in response to high Shh concentrations (Figure 3B),and their CRMs each contain a single high-affinity GBS witha PWM score of 0.9 or higher (Figures 3A and 3C). By contrast,CRMs associated with genes regulated by Shh at long rangewere characterized by PWM scores of GBSs that were lower(PWM< 0.90) than those found in CRMs of locally induced genes(Figures 3A–3C), notwithstanding a high degree of variationregarding the number and predicted affinities of GBSs. Tran-scriptional assays in P19 cells using an obligate activator formof Gli3 (Gli3H; Stamataki et al., 2005) generally supported thePWM score analysis, as multimerized (43) GBSNkx2.2 andGBSFoxA2 each displayed higher transcriptional activity in vitroas compared to selected multimerized GBSs from CRMOlig2,CRMNkx6.1, or CRMDbx1 (Figure 3D). Moreover, in electrophoreticmobility shift assays (EMSAs), excess of neither GBS1Nkx6.1 norGBSDbx1 oligonucleotides could outcompete GBSNkx2.2 forbinding of the zinc finger domain of Gli3 (Figure 3F). We alsoexamined the capacity of isolated GBSs to promote LacZreporter expression in vivo: 43GBSNkx2.2, 43GBSNkx2.9, and43GBSFoxA2 promoted LacZ expression in ventralmostNkx2.2+ p3 progenitors, but not at more dorsal positions (Fig-ure 3E). Meanwhile, 43GBS1Nkx6.1, 43GBSOlig2, or 43GBSDbx1

showed no in vivo activity; nor did 43GBS2Nkx6.1 (Figure 3E),despite a similar PWM score of this last to GBSFoxA2 (Figure 3A).In addition, 43GBS2Nkx6.1 exhibited significantly lower activitythan 43GBSFoxA2 in P19 cells (Figure 3D). These data indicatethat, while PWM scores serve as a useful indicator, they do notnecessarily predict the precise inherent quality of a given GBS.The inability of GliA to activate lower-quality GBSs in vivo

when examined in isolation indicates that CRM context is a crit-ical determinant of GliA-mediated transcriptional activation, andthe evolutionary conservation of GBSs and surroundingsequence may actually be more predictive of transcriptionalactivity as compared to the intrinsic affinity properties ofGBSs. This is supported by the fact that GBS1 is located ina highly conserved region of CRMNkx6.1 (Figure S2A) and is abso-lutely required for transcriptional activation in vivo, despite thefact that this site confers little activity when examined in isolation(Figures 3D and 3E). By contrast, GBS2 in CRMNkx6.1 hasa notably higher PWM score than GBS1 but is located in a lessconserved region (Figure S2A) and was not essential for activa-tion of CRMNkx6.1 when examined in vivo (Figure 2B). Collec-tively, these analyses indicate that locally induced genes, suchas Nkx2.2, Nkx2.9, and FoxA2, are regulated by unique high-quality GBSs, whereas genes regulated by Shh at longerdistance are generally associated with GBSs of lower inherentquality. In the case of the latter, however, there is no predictivecorrelation between gene expression pattern and affinity scoreor number of GBSs, and sequence context appears to be essen-tial for the ability of Gli proteins to induce transcription via GBSsin such CRMs.In models of transcriptional interpretation of the Drosophila

activator protein Dorsal there is no strong correlation betweenbinding site properties and distal CRM activity (Stathopoulosand Levine, 2005). However, in contrast to our findings withrespect to locally induced Shh target genes, genes that arelocally activated by Dorsal are typically associated with low-affinity sites. Our data therefore imply that rather than interpret-ing the ventral-to-dorsal GliA gradient (Stamataki et al., 2005), itmay be the opposing GliR gradient that provides positional infor-mation in neural pattern formation by repressing genes in aconcentration-dependent manner. To explore this possibility,we altered the affinity properties of the critical GBS in each ofCRMNkx2.2 and CRMNkx6.1 by site-specific mutagenesis andexamined the expression of these modified elements in vivo. Incontrast to the complete loss of activity observed upon inactivat-ing GBS1 in CRMNkx6.1, swapping GBS1Nkx6.1 for a high-affinityGBSNkx2.2 or a low-affinity GBSDbx1 left the expression ofCRMNkx6.1 in the ventral neural tube intact (Figure 3G). Expres-sion of CRMNkx2.2 in ventral p3 progenitors was also essentiallyunaffected by substitution of GBSNkx2.2 with a medium-affinityGBS1Nkx6.1 or low-affinity GBSDbx1 (Figure 3G). Thus, alteringthe binding affinities of GBSs had no apparent impact on theGliA-mediated activation of these elements in the ventral neuraltube. However, we also noted a marked dorsal derepression ofCRMNkx2.2 when the endogenous GBSNkx2.2 had been replacedby the low-affinity GBSDbx1 (Figure 3H). Moreover, althoughexpression of CRMNkx6.1 carrying GBSDbx1 was similar tocontrols at 40 hpe, at 24 hpe we observed an ectopic derepres-sion of CRMNkx6.1 when the endogenous GBS1Nkx6.1 had beenreplaced by GBSDbx1 (Figure 3H). These data provide directevidence that the ability of GliR to suppress Shh-regulatedCRMs is sensitive to the inherent quality of the GBS.

GliR Levels Are Limiting in Neural Pattern FormationPrevious studies of Gli genes have proposed that cells interpretthe regional balance between GliA and GliR, and that the GliAgradient could be limiting in ventral pattern formation (Bai

Developmental Cell

Transcriptional Interpretation of Shh Signaling

Developmental Cell 23, 1006–1019, November 13, 2012 ª2012 Elsevier Inc. 1009

Figure 3. Modifications of GBS Affinity Properties Indicate that Shh Signaling Is Transduced by GliR Activity(A) The calculated affinity scores of all identified GBSs. Locally induced genes (black bars); long-range genes (gray bars).

(B and C) Foxa2, Nkx2.2/Nkx2.9, Nkx6.1, and Dbx1 have distinct expression domains (B) associated with CRMs containing a single critical GBS that deviates at

varying positions (C, underlined nucleotides) from the consensus sequence (binding profile from Hallikas et al., 2006).

Developmental Cell

Transcriptional Interpretation of Shh Signaling

1010 Developmental Cell 23, 1006–1019, November 13, 2012 ª2012 Elsevier Inc.

et al., 2004; Stamataki et al., 2005; Dessaud et al., 2008).However, our CRM data argue for a direct interpretation of theopposing GliR gradient and that GliA may provide a morepermissive and nonlimiting activity subordinate to GliR. Consis-tent with such a role for GliR, we found that very low concentra-tions of GliR were sufficient to counteract GliA-mediated geneactivation in P19 cells (Figures 4A and 4B). We therefore de-signed experiments aimed at interfering with the net activity ofGli proteins in vivo without affecting the ratio of GliA to GliR byoverexpressing a myc-tagged protein encompassing the DNA-binding zinc finger domain of Gli3 (GliZnF) (Figure 4C) in theneural tube. If GliR were limiting in neural patterning and GliApermissive, one would predict transcriptional derepression ofShh target genes, whereas a GliA requirement should result inloss or retraction of gene expression. Strict interpretation ofthe balance would leave gene expression largely unaffected.Characterization of GliZnF in vitro confirmed binding to DNA(Figure 4D) and suggested that this protein lacks inherenttrans-activating potential (Figure 4E). DNA binding of GliZnF toDNA was sufficient to interfere with endogenous Gli proteins,as indicated by its ability to block activation of a 43GBSNkx2.2

reporter construct in the ventral neural tube (Figure 4F). Acontrol GliZnF protein harboring a 7 amino acid deletion inzinc finger 5 (GliZnFD5) (Figure 4C) could not bind DNA (Fig-ure 4D; Pavletich and Pabo, 1993) and failed to suppress43GBSNkx2.2-driven reporter activity in vivo (Figure 4F). Expres-sion of GliZnFD5 had no effect on the patterned expression ofendogenous neural Shh target genes in vivo (Figure 4G). Bycontrast, expression of GliZnF resulted in a notable dorsalexpansion (1.5- to 2-fold) of Olig2, Nkx6.1, and Nkx6.2 expres-sion (Figures 4G and S3). The domains of Dbx1 and Dbx2expression were also shifted dorsally in these experiments (Fig-ure 4G) at the expense of more dorsal progenitor identities andwithout any detectable change in Ptc1 expression (data notshown). These data demonstrate that the GliR gradient deter-mines the dorsal boundaries of gene expression for all ventralprogenitor domains regulated by Shh at long distance and argueagainst the possibility that the concentration of GliA is limiting inneural progenitors. GliA was, however, required for expressionof Nkx2.2 in p3 progenitors, and this loss was accompaniedby a ventral expansion of Olig2 (Figure 4H), consistent withdata showing that Nkx2.2 represses Olig2 in p3 progenitors(Novitch et al., 2001). Importantly, however, there was also scat-tered ectopic induction of Nkx2.2 in the Olig2+ pMN domain inthese experiments (Figures 4G and 4H). These data thereforesuggest that the induction of Nkx2.2 is determined by thebalance between GliA and GliR activities (see also Lek et al.,2010) and reveal that local induction of p3 progenitors differsmechanistically from progenitor domains induced by Shh atlong range.

The GliR Gradient Acts in Conjunction withCRM-Specific Repressive InputThe partial dorsal derepression of CRMNkx2.2 and CRMNkx6.1

carrying a low-affinity GBSDbx1 (Figure 3H) implies that repres-sive input in addition to GliR influences the regional expressionof Shh-regulated CRMs. Previous studies have shown thatselective repressive interactions between class I and II proteinsare important to maintain the integrity of ventral progenitordomains (Briscoe et al., 2000; Muhr et al., 2001; Dessaudet al., 2008), and computational analyses using cis-Decoder(Brody et al., 2007) revealed an overrepresentation of conservedHD protein-binding sites (HBS) in all of the identified CRMs (Fig-ure S1A; data not shown). This raised the possibility that repres-sive input by HD proteins is gated through the same CRMs asGliR. Consistent with the cross-repressive relationship betweenNkx6.1 andDbx2 (Muhr et al., 2001; Vallstedt et al., 2001), forcedexpression of Dbx proteins suppressedNkx6.1 and the activity ofCRMNkx6.1 in the ventral neural tube (Figure S4A), and we founda HBS in CRMNkx6.1 required for repression of CRMNkx6.1 in thedomain of Dbx2 expression (denoted HBSDbx; Figures S1A,5A, and 5B). HBSDbx bound Dbx proteins in EMSA (Figure S4B)and was required for the ability of Dbx2 to suppress GliA-medi-ated induction of CRMNkx6.1 in P19 cells (Figure 5C). In addition,we identified a second HBS (denoted HBSMsx) that mediatedrepression of CRMNkx6.1 in the dorsal neural tube as defined byMsx gene expression (Figures S1A, 5A, and 5B). HBSMsx boundMsx proteins in EMSA and was required for repression ofCRMNkx6.1 by any of Msx1–3 in P19 transcriptional assays(Figures 5C, S4B, and S4C). Expression of Msx2 also sup-pressed CRMNkx6.1 activity in the ventral neural tube (Fig-ure S4A), and forced expression of any of Msx1–3 repressedNkx6.1 in vivo (Figure S4D). Thus, although Msx1 and 3 mayhave unique functional properties in the neural tube (Liu et al.,2004), as all three Msx proteins are expressed in the dorsalneural tube (data not shown), our data suggest that they mayact redundantly with respect to the transcriptional repressionof Nkx6.1 expression.The repression of CRMNkx6.1 by Msx and Dbx proteins was

binding site specific, as Msx2 could not repress CRMNkx6.1 viaHBSDbx in P19 cells, and Dbx2 was also unable to act viaHBSMsx (Figure 5C). Upon inactivation of both HBSs, CRMNkx6.1

was uniformly activated along the entire DV axis of the neuraltube at 24 hpe (Figure 5B). Uniform activation was also observedat 16 hpe, but HBSDbx- and HBSMsx-inactivated CRMNkx6.1 wasonly partly dorsally derepressed at 8 hpe (Figure 5D). Consid-ering that CRMNkx6.1 became partly derepressed whenGBS1Nkx6.1 was replaced with the low-affinity GBSDbx1, thesedata indicate that integrated and overlapping activities of GliRand HD proteins are necessary to robustly suppress CRMNkx6.1

activation in the dorsal neural tube (Figure 5G). Accordingly, we

(D) Multimerized GBS-driven luciferase reporter activity in P19 cells transfected with Gli3H (10 ng). Error bars indicate SD (n = 2).

(E) Multimerized high-quality GBSs drive reporter expression in vivo (white) within the Nkx2.2 domain, whereas lower-quality GBSs do not activate reporter

expression. GBS-reporter (white); electroporation control (blue).

(F) EMSA of labeled GBSNkx2.2 oligonucleotides andMycGli3ZnF. Cold competitor of distinct GBSs (in 5- or 100-fold (nx) excess) andMyc antibody are added as

indicated. Upper box: antibody-MycGli3ZnF-DNA complex; central box: specific Gli3ZnF-DNA complex; Lower box: unbound oligonucleotides.

(G) Left image of each panel shows the activity of WT CRMs linked to distinct reporter constructs; other images in each panel show WT (blue) and the indicated

mutated (white) CRM activity.

(H) In vivo activity of WT (red) and GBSDbx1-containing (green) CRMNkx2.2 and CRMNkx6.1.

Developmental Cell

Transcriptional Interpretation of Shh Signaling

Developmental Cell 23, 1006–1019, November 13, 2012 ª2012 Elsevier Inc. 1011

Figure 4. In Vivo Interference with Endogenous Gli Activity Reveals a Dominant Role of GliR over GliA in Delimiting the Boundaries of Expres-sion of Class I and II TFs(A and B) Luciferase assays using 83GBSFoxa2 in presence of Gli2 and uponmutation of GBSs (A) or following addition of Gli2 and GliR (B). Western blot indicates

relative levels of Gli proteins, with GAPDH loading control (B, inset).

(C) Overview of GliZnF constructs. GliZnFDF5 protein lacks indicated amino acids essential to bind GBSs.

(D) EMSA with Mock, MycGliZnF, or MycGliZnFDF5 and labeled GBSNkx2.2 oligonucleotides.

Developmental Cell

Transcriptional Interpretation of Shh Signaling

1012 Developmental Cell 23, 1006–1019, November 13, 2012 ª2012 Elsevier Inc.

found that Gli3R and Msx proteins coordinately repressCRMNkx6.1 in P19 cells (Figure S4F). In addition, forced expres-sion of any of Msx1–3 repressed Dbx1 and Dbx2 expression inthe neural tube (Figure S4D, data not shown) and Msx2 sup-pressed activation of CRMDbx1 in vivo and P19 cell assays(Figures S4E and S4G; data not shown). These data, togetherwith the derepression of the GBS-mutated CRMDbx1, indicate

that cooperative repression by Gli and HD proteins is likely toapply to other Shh-regulated CRMs.

GliA Acts in Synergywith SoxB1Proteins to Activate ShhTarget Genes in the Neural TubeEctopic activation of CRMNkx2.2 and CRMNkx6.1 in the dorsalneural tube implies that GliA is present in dorsal progenitors in

(E) MycGliZnF and MycGliZnFDF5 lack transcriptional activity. Gli3H (50 ng) gives similar protein level to 1 ng MycGliZnF/MycGliZnFDF5 (data not shown).

(F) MycGliZnF, but not MycGliZnFDF5, suppresses 43GBSNkx2.2 activity in vivo (white); electroporation control (blue).

(G) Expression of MycGliZnF, but not of MycGliZnFDF5, abolishes Nkx2.2 expression in the p3 domain but induces ectopic Nkx2.2+ cells in the pMN (arrow);

dorsal boundaries of other class I and II TFs shift dorsally (brackets). Neural tube midline (dashed line), electropated (+), and control (") sides are indicated.

(H) Effects of GliZnF misexpression on Nkx2.2, Olig2, and Shh proteins in vivo.

RLU, relative luciferase units. Error bars indicate SD (n = 2). See also Figure S3.

Figure 5. The Ability of GliA to Activate CRMNkx6.1 in Dorsal Progenitors Is Counteracted Cooperatively by GliR and Region-Specific Repres-sion by HD Proteins(A) In situ hybridization of Nkx6.1, Dbx2, Msx1, and Msx2 delineates three distinct regions of Nkx6.1 regulation.

(B) Schematic illustrating positions of HBSs and GBSs in CRMNkx6.1 (top). Activity in the ventral, intermediate, and dorsal neural tube of CRMNkx6.1 following

mutation of either HBSDbx, HBSMsx, or both 24 hpe.

(C) Luciferase assays in P19 cells using CRMNkx6.1 and HBS-mutated versions activated by Sox3 and Gli3H. Repressive capabilities of Msx2 and Dbx2, tested by

mutation of their respective HBSs. For each construct, luciferase activity in presence of Sox3 and Gli3H was baseline for percentage of repression. Error bars

indicate SD (n = 3).

(D) Activity at 8 and 16 hpe of WT CRMNkx6.1 and a variant with both HBSs inactivated. Graphs indicate the percentage of electroporated GFP+ cells that have

ectopically activated the WT or mutated (mut) CRMNkx6.1. Error bars indicate SD (n = 9).

(E) In vivo activity of HBS-mutated CRMNkx6.1 following mutation of GBS1 or coelectroporation with Ptc1Dloop2, for which expression of which endogenous

Nkx6.1 is also shown.

(F) Gli3 full length (Gli3-190) is expressed in dorsal extremes of stage 16–17 neural tubes, but almost not in the ventral extremes, consistent with the very low levels

of Gli3 mRNA expression in this region. Dashed lines delineate the micro-dissected ventral and dorsal parts of the neural tube, and Gli3-83 indicates the Gli3

repressor form.

(G) Schematic illustrating overlapping temporal requirements of GliR and HD repressive activities in the regulation of CRMNkx6.1.

Mutated sites denoted by X; G1, GBS1; M, HBSMsx ; D, HBSDbx. See also Figure S4.

Developmental Cell

Transcriptional Interpretation of Shh Signaling

Developmental Cell 23, 1006–1019, November 13, 2012 ª2012 Elsevier Inc. 1013

sufficient amounts to trigger gene expression. Consistent withthis, point mutations in the core of GBS1 completely abolishedthe uniform activity of CRMNkx6.1 carrying the HBSDbx andHBSMsx mutations (Figure 5E). To determine whether ongoingShh signaling is required for activity of the derepressed form ofCRMNkx6.1, we cell-autonomously blocked Shh signaling byforced expression of Ptch1Dloop2 (Briscoe et al., 2001). As ex-pected, this abolished expression of Nkx6.1 and CRMNkx6.1

activity in the ventral neural tube, but also resulted in loss ofHBS-mutated CRMNkx6.1 activity along the entire DV axis(Figures 5E and S4H). These data argue that some degree ofShh signaling occurs in the dorsalmost neural tube, and thatthis is capable of stabilizing sufficient amounts of full-lengthGliA to induce CRM activity under non-HD-repressed condi-tions. In accordance with this, western blot analysis of dorsalizedneural stem cells in vitro andmicrodissected dorsal neural tissue

showed that, in addition to proteolyzed Gli3 repressor, the full-length activator form of Gli3 was also present in these cells(Figures 5F, S4I, and S4J).Cross-comparison of CRMs using cis-decoder (Brody et al.,

2007) identified, in addition to the HBSs, a general overrepresen-tation of Sox binding sites (SBSs) in all CRMs (Figure S1A;data not shown; Bailey et al., 2006). Interestingly, while mostClass II CRMs contained several SBSs, at least one of thesewas located in close proximity to a GBS (Figures S1A and 6B).The SoxB1 group of proteins, Sox1–3, are transcriptional activa-tors widely expressed in neural progenitor cells and have animportant role in regulating neural progenitor properties (Fig-ure 6A; Bergsland et al., 2011). Chromatin immunoprecipitation(ChIP) experiments on neural tissue showed that Sox3 bindsthese CRMs (Figure 6C) irrespective of the on-off state of theirassociated genes (Figures 6D and S5A), implying that Gli and

Figure 6. Gli Proteins Activate Transcription in the CNS in Synergy with SoxB1 Proteins(A) Sox3 expression in the HH17 chick neural tube.

(B) Upper panel: schematics of CRMNkx2.2 , CRMNkx6.1, and CRMNkx6.2 with indicated GBS (blue), SBS (green), and TBS (light blue). Lower panel: mutation of

SBSs (indicated by S) in selected CRMs and of two TBSs (T) in CRMNkx2.2. Coelectroporation of two WT CRMs driving expression of distinct reporter genes (left,

white and blue) and the activity of mutated and WT control CRM (right, white and blue, respectively).

(C) Sox3 ChIP on mouse neural tissue for Shh-regulated neural-specific CRMs and a Shh-regulated mesodermal-specific CRMMyf5 (Figure S5B).

(D) Neural stem cells differentiated to dorsal (D), intermediate (I), and ventral (V) identity, as indicated (left) by expression of Nkx6.1 and Sox3. qPCR (right)

following Sox3 ChIP on CRMNkx6.1 and exon 11 of the Sox3 regulated Notch gene (negative control).

(E) Luciferase assays for class II CRMs in P19 cells in presence of Sox3 (10 ng) and Gli3H (10 ng) and following interference by Sox DNA-binding domain (HMG).

(F) Luciferase assays for class II CRMs in P19 cells in presence of Sox3 (10 ng) and variable concentrations of Gli3H. Mutation of either the GBSs (middle) or

specific SBSs (right) abolished activity.

(G) Activity of CRMNkx6.1 carrying mutated HBSMsx and HBSDbx, together with WT (left) or mutated (right) SBS2.

Error bars indicate SD (n = 2). See also Figure S5.

Developmental Cell

Transcriptional Interpretation of Shh Signaling

1014 Developmental Cell 23, 1006–1019, November 13, 2012 ª2012 Elsevier Inc.

Sox proteins could cooperate at the CRM level to activate Shhtarget genes in the neural tube.To examine the regulatory activity of SoxB1 sites, we mutated

the SBSs in selected CRMs. Strikingly, inactivation of the threeSBSs (SBS1–3) in each of CRMNkx6.1 and CRMNkx6.2 abolishedtranscriptional activity of these elements in the neural tube (Fig-ure 6B). The SBSs in CRMNkx2.2 were also important for expres-sion in p3 progenitors, but we noted a functional overlapbetween these sites and two TCF-binding sites (TBS) previouslyimplicated in repression ofNkx2.2 expression in the pMNdomain(Lei et al., 2006). Specifically, CRMNkx2.2 activity in p3 progeni-tors could still be detected after inactivation of either the SBSsor TBSs in CRMNkx2.2, but was abolished when all these siteswere inactivated concurrently (Figures 6B and 7A). By contrast,individual inactivation of SBS1–3 in CRMNkx6.1 showed that theinactivation of a single SBS located at themost proximal positionrelative to GBS1 (termed SBS2) was sufficient to abolish allCRMNkx6.1 activity (Figure 7A, data not shown), suggesting thatthis site has the primary role in mediating Shh responsivenessin CRMNkx6.1. These data show that Shh input alone is insuffi-cient to activate class II CRMs and suggest that GliA coulddirectly cooperate with SoxB1 proteins to induce CRM activity.Accordingly, CRM-luciferase reporter assays in P19 cells re-

vealed synergistic transcriptional activity between Gli3H (Stama-taki et al., 2005) and Sox 2/3 proteins on all class II CRMs(Figures 6E and S5C; data not shown). A synergetic relationshipbetween the Shh-Gli pathway and SoxB1 proteins was alsoobserved when the Shh pathway was activated by SmoM2

and Gli2 (data not shown). Although expression of Gli3H alonehad no effect on CRMNkx6.1, it resulted in some activation ofCRMNkx2.2, CRMNkx2.9, CRMOlig2, and CRMNkx6.2 (Figure 6E).However, P19 cells express Sox2 and Sox3 (data not shown)and endogenously expressed SoxB1 proteins could thereforecooperate with Gli3H in this assay. Indeed, activation of allShh-regulated CRMs examined in P19 cells, induced by Gli3Hor by Gli3H and exogenous Sox3, was abolished by coexpres-sion with the DNA-binding HMG domain (HMG) of Sox3 (Fig-ure 6E). Moreover, activation of CRMNkx2.2 and CRMNkx6.1 inresponse to Gli3H and Sox3 was abolished upon mutation ofeither the critical GBS or SBSs in CRMNkx2.2 or CRMNkx6.1 (Fig-ure 6F). Transcriptional input of SoxB1 proteins was alsorequired for derepression of HBS-mutated CRMNkx6.1 in thedorsal neural tube, as inactivation of SBS2 abolished all activityof this element (Figure 6G). Taken together, these data show thatSoxB1 proteins are essential cofactors that synergistically coop-erate with Gli proteins to enable the largely concentration-inde-pendent mode of gene activation by GliA in neural progenitors.

Interpretation of Local and Long-Range Shh SignalingDiffers with Respect to GliA Dependence and CRMContext SensitivityIn order to achieve a ‘‘GBS-independent’’ loss of CRMNkx6.1

activity in vivo, it is sufficient to inactivate only SBS2, which abutsGBS1 in CRMNkx6.1 (Figures 6B and 7A), whereas in CRMNkx2.2 itwas necessary to inactivate all three SBSs as well as two Tcfbinding sites (Figures 6B and 7A). To test whether high-affinity

Figure 7. The Role of GBSs and Context in the Interpretation of Graded Shh Signaling(A) A high-quality GBS renders CRM activation less context dependent. Mutational analysis of TF function in CRMNkx2.2 and CRMNkx6.1. Top panel outlines

CRM configurations of GBSs (G), SBSs (S), and TBSs (T) For each CRM, mutated binding sites are denoted by an X in the panel at the left, with in vivo expression

of corresponding constructs at the right. WT (blue); mutated (white). TBS-mutated CRMNkx2.2 was occasional derepressed close to the intermediate neural tube

(not shown).

(B and C) Model of the interpretation of graded Shh signaling in tissue patterning.

(B) Extrinsic Shh is translated into opposing intrinsic gradients of GliR and GliA along the neural DV axis. The GliR gradient is interpreted at transcriptional levels,

and determines the positioning of all ventral progenitor domains. This is partly achieved via differential affinity binding of GliR to qualitatively distinct GBSs, and via

cooperation of GliR with CRM-specific input by domain-specific repressors. Local gene activation requires accumulation of GliA to a level sufficient to counteract

GliR, whereas long-range activation simply requires reduction of GliR.

(C) In OFF state: higher cellular concentration of GliR (symbolized by the size of GliR) is required to suppress CRMs regulated by Shh at long range as these

elements are typically regulated by low- ormedium-affinity GBSs, in contrast to locally induced genes that are regulated by high-affinity GBSs. In ON state: locally

induced CRMs require high GliA concentrations, but are less dependent on coactivator (e.g., SoxB1) input (symbolized by relative sizes of GliA and CoA). Long-

range activation of CRMs by GliA is concentration-independent and critically depends on CRM context and input by coactivators, e.g., SoxB1 (symbolized by

relative sizes of GliA and CoA). Size of arrows symbolizes the predominant activity in transcriptional activation.

Developmental Cell

Transcriptional Interpretation of Shh Signaling

Developmental Cell 23, 1006–1019, November 13, 2012 ª2012 Elsevier Inc. 1015

GBSs render gene activation less context dependent, we re-placed GBS1Nkx6.1 with the high-affinity GBSNkx2.2 in CRMNkx6.1

carrying the SBS2 mutation, and observed that its activity wasrestored (Figure 7A). Collectively, these data indicate thatCRMs induced by Shh at long range aremore context dependentthan locally induced CRMs and provide direct evidence thathigh-quality GBSs associated with locally induced genes rendergene activation by GliA less dependent on CRM context andcooperative input by coactivators.

DISCUSSION

ARepressiveGene-RegulatoryModuleDefinesRegionalGene Expression Downstream of Graded Shh Signalingin Neural PatterningThe mechanisms by which graded information provided bymorphogens is interpreted at the genomic level and translatedinto discrete transcriptional outputs have not been resolved invertebrates (Ingham and Placzek, 2006; Dessaud et al., 2008).Through the systematic analysis of Shh-regulated CRMs, weprovide evidence that the relationship between GBS affinityproperties and spatial expression of Gli target genes is theinverse of morphogen interpretation models of transcriptionalactivator proteins in Drosophila (Stathopoulos and Levine,2005). Thus, genes induced by Shh at short range appear tobe regulated by unique high-affinity GBSs, while CRMs associ-ated with genes regulated by Shh at long range contain GBSsthat have variable, but typically lower, affinity binding scores.We show that binding of GliR to GBSs is required to preventectopic activation of CRMDbx1 and CRMNkx6.2, and furthermorethat in CRMNkx2.2 and CRMNkx6.1 this ability is directly influencedby the inherent quality of GBSs. Moreover, the dorsal derepres-sion of long-range target genes upon interference with the DNA-binding ability of endogenous Gli proteins strongly argues thatGliR—but not GliA—levels determine the dorsal boundaries ofthese genes. Gli3 has been suggested to provide the primaryrepressor activity in neural patterning (Bai et al., 2004), but Gli2can be processed to a repressor and has been proposed tosuppress transcription in the neural tube (Lei et al., 2004; Panet al., 2006). Our data indicate indeed that both Gli2R andGli3R act in a partially redundantmanner to constrain the expres-sion domains of Shh-regulated genes. In particular, whereasGli3single mutants exhibit a mild derepression of Dbx1 and Nkx6.2expression (Persson et al., 2002), a more pronounced derepres-sion of CRMDbx1 and CRMNkx6.2 was observed upon GBS inac-tivation in these elements. Likewise, interference with net Gliactivity resulted in a dorsal shift of the Olig2 and Nkx6.1 expres-sion boundaries, which are not affected in either Gli2 or Gli3single mutants (Persson et al., 2002; Bai et al., 2004).

In the fruit fly, it has recently been proposed that cooperativerepression by CiR binding of clustered Ci-binding sites couldbe sufficient to determine the transcriptional output of CRMs inresponse to graded Hh signaling (Parker et al., 2011). However,this model cannot be readily applied to the vertebrate neuraltube, as several Shh-regulated CRMs contain only a single func-tional GBS. We also provide evidence that the GliR gradient isfunctionally integrated at the CRM level with more region-specific repressive input provided by HD proteins. In particular,we identify distinct binding sites for Dbx and Msx proteins in

CRMNkx6.1 that are required for region-specific repression ofthis element in the intermediate and dorsal neural tube, respec-tively. Upon inactivation of both of these sites, CRMNkx6.1 wasderepressed in a progressive ventral-to-dorsal fashion, implyingthat GliR-mediated gene repression rapidly becomes dependenton cooperative input by Msx and Dbx proteins. Conversely, theability of Msx proteins to effectively repress transcriptionappears to be influenced by binding of GliR to CRMs, as replace-ment of GBS1 in CRMNkx6.1 with a low-affinity GBS as well asinactivation of the GBS in CRMDbx1 led to derepression of theseelements in dorsal progenitors defined by Msx gene expression.Accordingly, these findings indicate an overlapping temporalrequirement for and possible synergistic relationship betweenthe GliR gradient and more region-specific repressive input.GliR and class I and II TFs are therefore likely to define corecomponents of a repressive gene regulatory network that trans-lates graded Shh signaling into discrete patterns of gene expres-sion in the ventral neural tube, which is consistent with the factthat genes regulated by Shh at long range become derepressedin a somewhat stochastic fashion in Gli2/Gli3 and Smo/Gli3mutant mice (Wijgerde et al., 2002; Bai et al., 2004), and alsowith data indicating that cross-repression between Dbx1 andNkx6.2 in Gli3 mutants is impaired (Persson et al., 2002). Sucha repressive gene-regulatory network is also in line with theobservation that interfering with the ability of Groucho/TLE core-pressors to interact with class I and II HD repressors results indorsal derepression of Shh-induced genes (Muhr et al., 2001)resembling the derepression observed upon interference withendogenous Gli activity in vivo (Figure 4G).

Low Levels of GliA Are Sufficient to Trigger CRMActivation in Permissive Transcriptional StatesGBSNkx2.2 and GBS1Nkx6.1 are absolutely required for Shh-mediated induction of their respective elements in the ventralneural tube. When they were replaced by a low-affinity GBSDbx1

(mediating a repressive function in its endogenous CRMDbx1

context), not only was ventral activity of these elements restored,but derepression in the dorsal neural tube also occurred. Impor-tantly, these data indicate that this dorsal derepression is in factGliA dependent—a highly unanticipated finding as it has gener-ally been presumed that Shh signaling in dorsal progenitors isnegligible and that these cells therefore predominantly, if notexclusively, express processed repressor forms of Gli2 andGli3. Nevertheless, expression of Ptc1 becomes progressivelyupregulated throughout the neural tube over time (Marigo andTabin, 1996; data not shown), an indication of Shh pathway acti-vation, and our western blot analysis shows that both the full-length activator and repressor forms of Gli3 are present in dorsalneural tissue isolated from HH stage 16–17 chick embryos.Moreover, the facts that uniform derepression of an HDrepressor-insensitive form of CRMNkx6.1 required a functionalGBS1Nkx6.1 and was abrogated upon electroporation withPtcDloop2 (Briscoe et al., 2001) provide direct evidence thatthe ectopic activation of this element is indeed Shh pathwaydependent. Together with the dorsal realignment of ventralprogenitor domain boundaries in response to GliZnF expression,these data collectively suggest a model in which GliA concentra-tions are nonlimiting regarding the transcriptional interpretationof long-range Shh morphogen activity. Instead, GliA functions

Developmental Cell

Transcriptional Interpretation of Shh Signaling

1016 Developmental Cell 23, 1006–1019, November 13, 2012 ª2012 Elsevier Inc.

subordinate to GliR and other CRM-specific repressive input indetermining dorsal limits of Shh target gene expression, and inpermissive nonrepressed states, the levels of GliA in the interme-diate and dorsal neural tube appear sufficient to induce geneexpression (Figures 7B and 7C). This GliR-gradient interpretationmodel differs significantly from prevailing models suggestingthat the GliA gradient is instructive in ventral neural patterning(Stamataki et al., 2005) or alternatively that cells strictly measurethe balance between GliA and GliR (Bai et al., 2004). It is notable,however, that the idea that GliA concentration is directly inter-preted at the transcriptional level is based exclusively on gain-of-function experiments in which obligate GliA proteins wereoverexpressed in the neural tube (Stamataki et al., 2005). It istherefore likely that the inductive responses observed in theseexperiments can be attributed not only to increased GliA levels,but also to transcriptional derepression due to interference withthe endogenous GliR gradient.

SoxB1 Factors Function as Critical Cofactors for GliA inGene Activation in the CNSWe show that GliA-mediated induction of Shh target genes inneural tissue critically requires the cooperative activity ofSoxB1 proteins. This synergistic relationship provides a mech-anistic rationale for how GliA can induce transcription ina largely concentration-independent manner. It is notable thatlow-quality GBSs examined in isolation exhibit little, if any,activity in transcriptional assays in vitro and in vivo, yet mediateactivator functions when operating in the context of endoge-nous CRMs. This indicates that the ability of GliA to stablybind and activate transcription through these sites is CRMcontext dependent. Considering that SoxB1 proteins alonehave low trans-activating potential (Kamachi et al., 2000), onelikely function of these proteins could be to stabilize GliAbinding to DNA, thereby rendering GliA-mediated activationinsensitive to the quality of GBSs and the cellular concentrationof GliA. Hence, rather than being instructive, recruitment of GliAto CRMs could primarily serve as a transcriptional switch thattriggers transcription by potentiating a SoxB1-dependent acti-vator complex once the level of GliR has dropped below a givensuppressive threshold value. Our analysis of CRMOlig2 andCRMNkx6.1 indicates that removal of Gli input, under wild-typeconditions, is not sufficient to achieve transcriptional derepres-sion of these CRMs. However, class II genes regulated by long-range Shh signaling, including Olig2 and Nkx6.1, becomederepressed in Gli2/Gli3 mutant mice (Bai et al., 2004), showingthat GliA is not absolutely required for gene activation and sug-gesting that a SoxB1-containing complex is sufficient to acti-vate certain Shh-induced genes even in the absence of GliA,provided first that GliR is also eliminated. Second, geneticremoval of all Gli activity is likely to result in dysregulatedexpression of many genes (Vokes et al., 2007), which mayfurther facilitate the GliA-independent activation of class IIgenes in these mutants.

The Local and Long-Range Interpretations of ShhSignaling Are Mechanistically DistinctOur study reveals important mechanistic differences betweenlocal and long-range interpretation of Shh signaling, both withrespect to regulation by Gli proteins and CRM context depen-

dence (Figures 7B and 7C). Genes that are locally induced byShh, e.g., Nkx2.2 in p3 progenitors, are associated with high-affinity GBSs that interpret the balance between GliA and GliR.Accordingly, GliA must accumulate to a critical inductivethreshold value necessary to counteract GliR, a mode of induc-tion consistent with the loss of Nkx2.2 expression in Gli2/Gli3mutants and other studies implying that the ratio of GliA toGliR determines the specification of p3 progenitors and FP cells(Bai et al., 2004; Dessaud et al., 2007; Lek et al., 2010). More-over, GliA-dependent induction of local genes is less criticallydependent on CRM architecture and the input of other coactiva-tors than genes regulated by Shh at long range, which containlower-affinity GBSs and interpret the graded activity of GliR.Collectively, these data argue that activation of long-range Shhtarget genes requires the functional integration of Gli activityinto more complex transcriptional networks. It is notable thatthe long-range patterning activities of Bicoid and Dorsal inDrosophila are also highly CRM context dependent (Ochoa-Espinosa et al., 2005; Stathopoulos and Levine, 2005). Consid-ering that the induction of p3 progenitors by Shh resemblesa binary inductive event (Figure 7B), it is feasible that these differ-ences in GliA dependence and CRM context sensitivity reveala mode of mechanistic evolution of morphogen function.

EXPERIMENTAL PROCEDURES

For additional information concerning protein interaction assays, constructs,

probes, mouse neural stem cell cultures, antibodies, and additional reagents,

please refer to the Supplemental Experimental Procedures.

BioinformaticsThe ECR browser (Ovcharenko et al., 2004) was used to screen genomic loci of

the neural class I and II genes of mouse and chick for the presence of evolu-

tionarily conserved YDGGHGGYC motifs, which is a GBS consensus site

extracted from vertebrate and Drosophila literature (Agren et al., 2004, Saitsu

et al., 2005, Gustafsson et al., 2002; Alexandre et al., 1996; Hepker et al., 1997;

ECR browser settings: 300 bp ECR length and 60% ECR similarity). The rela-

tive GBS PWM affinity scores were calculated according to the GBS PWM

given by Hallikas et al. (2006). Shared motifs between CRMs of class I and II

genes were found by using CBS aligner of cis-decoder (6 bp as minimum

length; Brody et al., 2007).

In Ovo ElectroporationDNA was electroporated into the neural tube of chick embryos at HH stage

10–12 as previously described (Briscoe et al., 2000). For each construct, at

least four different embryos from two independent experimentswere analyzed.

Luciferase AssayTranscriptional assays were carried out in P19 cells using Lipofectamine and

Plus Reagent (Invitrogen), according to the manufacturer’s recommendations.

Luciferase and b-galactosidase activity was measured 24 hr posttransfection

using the Luciferase Assay Kit (Biotherma) and Galacto-Light Plus Kit (Applied

Biosystems), respectively. Each figure containing RLU is a representation of

three independent experiments preformed in duplicate. Values calculated for

percentage of repression are the average of three experiments.

Immunofluorescence and In Situ HybridizationImmunofluorescence and in situ hybridization were performed essentially as

described (Briscoe et al., 2000; Schaeren-Wiemers and Gerfin-Moser, 1993).

Protein Interaction AssaysEMSAs and ChIP experiments were performed essentially as previously

described (Agren et al., 2004; Bergsland et al., 2011). Western blots were per-

formed according to standard protocols.

Developmental Cell

Transcriptional Interpretation of Shh Signaling

Developmental Cell 23, 1006–1019, November 13, 2012 ª2012 Elsevier Inc. 1017

SUPPLEMENTAL INFORMATION

Supplemental Information includes five figures and Supplemental Experi-

mental Procedures and can be found with this article online at http://dx.doi.

org/10.1016/j.devcel.2012.09.015.

ACKNOWLEDGMENTS

We thank J. Briscoe, A. Joyner, B. Novitch, S. Mackem, and R. Toftgard for

reagents. We are grateful to A. McMahon for sharing unpublished results

and to T. Jessell for critical reading of the manuscript. This work was

supported by the Swedish Research Council (33X-06555, DBRM), The Royal

Swedish Academy of Sciences by donation from the Wallenberg Foundation,

Swedish Foundation for Strategic research (CEDB; SRL10-0030), The Knut

and Alice Wallenberg Foundation (KAW2011.0161), and research funds of

Karolinska Institutet. T.O. was supported by the Wenner-Gren foundation

and European Union (Marie Curie MEIF-CT-2006-025416).

Received: December 21, 2011

Revised: July 20, 2012

Accepted: September 18, 2012

Published online: November 12, 2012

REFERENCES

Agren, M., Kogerman, P., Kleman, M.I., Wessling, M., and Toftgard, R. (2004).

Expression of the PTCH1 tumor suppressor gene is regulated by alternative

promoters and a single functional Gli-binding site. Gene 330, 101–114.

Alexandre, C., Jacinto, A., and Ingham, P.W. (1996). Transcriptional activation

of hedgehog target genes in Drosophila is mediated directly by the cubitus in-

terruptus protein, a member of the GLI family of zinc finger DNA-binding

proteins. Genes Dev. 10, 2003–2013.

Bai, C.B., Stephen, D., and Joyner, A.L. (2004). All mouse ventral spinal cord

patterning by hedgehog is Gli dependent and involves an activator function

of Gli3. Dev. Cell 6, 103–115.

Bailey, P.J., Klos, J.M., Andersson, E., Karlen, M., Kallstrom, M., Ponjavic, J.,

Muhr, J., Lenhard, B., Sandelin, A., and Ericson, J. (2006). A global genomic

transcriptional code associated with CNS-expressed genes. Exp. Cell Res.

312, 3108–3119.

Balaskas, N., Ribeiro, A., Panovska, J., Dessaud, E., Sasai, N., Page, K.M.,

Briscoe, J., and Ribes, V. (2012). Gene regulatory logic for reading the

Sonic Hedgehog signaling gradient in the vertebrate neural tube. Cell 148,

273–284.

Bergsland, M., Ramskold, D., Zaouter, C., Klum, S., Sandberg, R., and Muhr,

J. (2011). Sequentially acting Sox transcription factors in neural lineage devel-

opment. Genes Dev. 25, 2453–2464.

Briscoe, J., Pierani, A., Jessell, T.M., and Ericson, J. (2000). A homeodomain

protein code specifies progenitor cell identity and neuronal fate in the ventral

neural tube. Cell 101, 435–445.

Briscoe, J., Chen, Y., Jessell, T.M., and Struhl, G. (2001). A hedgehog-insen-

sitive form of patched provides evidence for direct long-range morphogen

activity of sonic hedgehog in the neural tube. Mol. Cell 7, 1279–1291.

Brody, T., Rasband, W., Baler, K., Kuzin, A., Kundu, M., and Odenwald, W.F.

(2007). cis-Decoder discovers constellations of conserved DNA sequences

shared among tissue-specific enhancers. Genome Biol. 8, R75.

Davidson, E.H. (2010). Emerging properties of animal gene regulatory

networks. Nature 468, 911–920.

Dessaud, E., Yang, L.L., Hill, K., Cox, B., Ulloa, F., Ribeiro, A., Mynett, A.,

Novitch, B.G., and Briscoe, J. (2007). Interpretation of the sonic hedgehog

morphogen gradient by a temporal adaptation mechanism. Nature 450,

717–720.

Dessaud, E., McMahon, A.P., and Briscoe, J. (2008). Pattern formation in the

vertebrate neural tube: a sonic hedgehog morphogen-regulated transcrip-

tional network. Development 135, 2489–2503.

Gustafsson, M.K., Pan, H., Pinney, D.F., Liu, Y., Lewandowski, A., Epstein,

D.J., and Emerson, C.P., Jr. (2002). Myf5 is a direct target of long-range

Shh signaling and Gli regulation for muscle specification. Genes Dev. 16,

114–126.

Hallikas, O., Palin, K., Sinjushina, N., Rautiainen, R., Partanen, J., Ukkonen, E.,

and Taipale, J. (2006). Genome-wide prediction of mammalian enhancers

based on analysis of transcription-factor binding affinity. Cell 124, 47–59.

Hepker, J., Wang, Q.T., Motzny, C.K., Holmgren, R., and Orenic, T.V. (1997).

Drosophila cubitus interruptus forms a negative feedback loop with patched

and regulates expression of Hedgehog target genes. Development 124,

549–558.

Ingham, P.W., and Placzek, M. (2006). Orchestrating ontogenesis: variations

on a theme by sonic hedgehog. Nat. Rev. Genet. 7, 841–850.

Jessell, T.M. (2000). Neuronal specification in the spinal cord: inductive signals

and transcriptional codes. Nat. Rev. Genet. 1, 20–29.

Kamachi, Y., Uchikawa, M., and Kondoh, H. (2000). Pairing SOX off: with part-

ners in the regulation of embryonic development. Trends Genet. 16, 182–187.

Lander, A.D. (2007).Morpheus unbound: reimagining themorphogen gradient.

Cell. 26, 245–256.

Lei, Q., Zelman, A.K., Kuang, E., Li, S., and Matise, M.P. (2004). Transduction

of gradedHedgehog signaling by a combination of Gli2 andGli3 activator func-

tions in the developing spinal cord. Development 131, 3593–3604.

Lei, Q., Jeong, Y., Misra, K., Li, S., Zelman, A.K., Epstein, D.J., and Matise,

M.P. (2006). Wnt signaling inhibitors regulate the transcriptional response to

morphogenetic Shh-Gli signaling in the neural tube. Dev. Cell 11, 325–337.

Lek, M., Dias, J.M., Marklund, U., Uhde, C.W., Kurdija, S., Lei, Q., Sussel, L.,

Rubenstein, J.L., Matise, M.P., Arnold, H.H., et al. (2010). A homeodomain

feedback circuit underlies step-function interpretation of a Shh morphogen

gradient during ventral neural patterning. Development 137, 4051–4060.

Liu, Y., Helms, A.W., and Johnson, J.E. (2004). Distinct activities of Msx1 and

Msx3 in dorsal neural tube development. Development 131, 1017–1028.

Marigo, V., and Tabin, C.J. (1996). Regulation of patched by sonic hedgehog in

the developing neural tube. Proc. Natl. Acad. Sci. USA 93, 9346–9351.

Muhr, J., Andersson, E., Persson, M., Jessell, T.M., and Ericson, J. (2001).

Groucho-mediated transcriptional repression establishes progenitor cell

pattern and neuronal fate in the ventral neural tube. Cell 104, 861–873.

Novitch, B.G., Chen, A.I., and Jessell, T.M. (2001). Coordinate regulation of

motor neuron subtype identity and pan-neuronal properties by the bHLH

repressor Olig2. Neuron 31, 773–789.

Ochoa-Espinosa, A., Yucel, G., Kaplan, L., Pare, A., Pura, N., Oberstein, A.,

Papatsenko, D., and Small, S. (2005). The role of binding site cluster strength

in Bicoid-dependent patterning in Drosophila. Proc. Natl. Acad. Sci. USA 102,

4960–4965.

Ovcharenko, I., Nobrega, M.A., Loots, G.G., and Stubbs, L. (2004). ECR

Browser: a tool for visualizing and accessing data from comparisons of

multiple vertebrate genomes. Nucleic Acids Res. 32(Web Server issue),

W280–W286.

Pan, Y., Bai, C.B., Joyner, A.L., andWang, B. (2006). Sonic hedgehog signaling

regulates Gli2 transcriptional activity by suppressing its processing and degra-

dation. Mol. Cell. Biol. 26, 3365–3377.

Parker, D.S., White, M.A., Ramos, A.I., Cohen, B.A., and Barolo, S. (2011). The

cis-regulatory logic of Hedgehog gradient responses: key roles for gli binding

affinity, competition, and cooperativity. Sci. Signal. 4, ra38.

Pavletich, N.P., and Pabo, C.O. (1993). Crystal structure of a five-finger GLI-

DNA complex: new perspectives on zinc fingers. Science 261, 1701–1707.

Persson, M., Stamataki, D., te Welscher, P., Andersson, E., Bose, J., Ruther,

U., Ericson, J., and Briscoe, J. (2002). Dorsal-ventral patterning of the spinal

cord requires Gli3 transcriptional repressor activity. Genes Dev. 16, 2865–

2878.

Saitsu, H., Komada, M., Suzuki, M., Nakayama, R., Motoyama, J., Shiota, K.,

and Ishibashi, M. (2005). Expression of the mouse Fgf15 gene is directly initi-

ated by Sonic hedgehog signaling in the diencephalon andmidbrain. Dev. Dyn.

232, 282–292.

Developmental Cell

Transcriptional Interpretation of Shh Signaling

1018 Developmental Cell 23, 1006–1019, November 13, 2012 ª2012 Elsevier Inc.

Sasaki, H., Hui, C., Nakafuku, M., and Kondoh, H. (1997). A binding site for Gli

proteins is essential for HNF-3beta floor plate enhancer activity in transgenics

and can respond to Shh in vitro. Development 124, 1313–1322.

Schaeren-Wiemers, N., and Gerfin-Moser, A. (1993). A single protocol to

detect transcripts of various types and expression levels in neural tissue and

cultured cells: in situ hybridization using digoxigenin-labelled cRNA probes.

Histochemistry 100, 431–440.

Stamataki, D., Ulloa, F., Tsoni, S.V., Mynett, A., and Briscoe, J. (2005). A

gradient of Gli activity mediates gradedSonic Hedgehog signaling in the neural

tube. Genes Dev. 19, 626–641.

Stathopoulos, A., and Levine, M. (2005). Genomic regulatory networks and

animal development. Dev. Cell 9, 449–462.

Vallstedt, A., Muhr, J., Pattyn, A., Pierani, A., Mendelsohn, M., Sander, M.,

Jessell, T.M., and Ericson, J. (2001). Different levels of repressor activity assign

redundant and specific roles to Nkx6 genes in motor neuron and interneuron

specification. Neuron 31, 743–755.

Vokes, S.A., Ji, H., McCuine, S., Tenzen, T., Giles, S., Zhong, S., Longabaugh,

W.J., Davidson, E.H., Wong, W.H., and McMahon, A.P. (2007). Genomic char-

acterization of Gli-activator targets in sonic hedgehog-mediated neural

patterning. Development 134, 1977–1989.

Vokes, S.A., Ji, H., Wong, W.H., and McMahon, A.P. (2008). A genome-scale

analysis of the cis-regulatory circuitry underlying sonic hedgehog-mediated

patterning of the mammalian limb. Genes Dev. 22, 2651–2663.

Wijgerde, M., McMahon, J.A., Rule, M., and McMahon, A.P. (2002). A direct

requirement for Hedgehog signaling for normal specification of all ventral

progenitor domains in the presumptive mammalian spinal cord. Genes Dev.

16, 2849–2864.

Xie, J., Murone, M., Luoh, S.M., Ryan, A., Gu, Q., Zhang, C., Bonifas, J.M.,

Lam, C.W., Hynes, M., Goddard, A., et al. (1998). Activating Smoothened

mutations in sporadic basal-cell carcinoma. Nature 391, 90–92.

Developmental Cell

Transcriptional Interpretation of Shh Signaling

Developmental Cell 23, 1006–1019, November 13, 2012 ª2012 Elsevier Inc. 1019

1

Developmental Cell, Volume 23

Supplemental Information

Mechanistic Differences in the Transcriptional

Interpretation of Local and Long-Range

Shh Morphogen Signaling Tony Oosterveen, Sanja Kurdija, Zhanna Alekseenko, Christopher W. Uhde, Maria Bergsland, Magnus Sandberg, Elisabet Andersson, José M. Dias, Jonas Muhr, and Johan Ericson

Inventory of Supplemental Information:

Supplemental Figures S1 - S5

Supplemental Figure Legends

Supplemental Experimental Procedures

Supplemental References

2

Figure S1

Figure S2

3

Figure S3

Figure S4

4

Figure S5

5

Supplemental figure legends

Figure S1, related to Figure 1: The relative positions and sequences of relevant Gli, SoxB1 and

Homeodomain binding sites in the class I and II CRMs described in Figure 1.

(A) For each CRM, the genomic localization (above each CRM, given in mm9 assembly coordinates),

sequences and relative positions of Gli (blue, G), Homeodomain (orange/red, H) and SoxB1 (green, S)

binding sites are depicted (distances between sites are indicated in base pairs (bp) below the bars). H1

indicates the Msx binding site and H2 the Dbx binding site.

(B) CRMNkx2.9 activity is ventrally restricted. Schematics showing CRMNkx2.9 coordinates with

indicated binding sites and in vivo expression (white). Electroporation control (blue).

(C) Schematic illustrating in ovo electroporation strategy for assessment of CRM activity in the chick

neural tube.

Figure S2 related to Figure 2: Mutational analysis and vertebrate genomic sequence alignment of

GBS1 and GBS2 in CRMNkx6.1

(A) The sequence and the direct context of GBS1, but not of GBS2, are highly conserved during

vertebrate evolution. Dashed circle indicates dip in mammalian evolutionary conservation of the

GBS2 sequence. Orange *, +, 2 and 1 above alignment indicate gaps of, respectively, 7-12, 11, 2 and 1

bp.

(B) CRMNkx6.1 is inactive when both GBS1 and GBS2 are mutated. Wild type CRMNkx6.1 (blue) was

co-electroporated with the GBS mutated CRMNkx6.1 (white).

(C) Co-electroporation with Ptc1¨loop2 construct that blocks Shh signal transduction inhibits

remaining activity of GBS2 mutated CRMNkx6.1, shown in figure 2. Electroporation control (blue),

CRM expression (white)

Figure S3, related to Figure 4: Quantification of dorsal domain expansion of Nkx2.2, Nkx6.1, Nkx6.2

and Olig2 in the GliZnf experiment. Nkx2.2 expression is generally reduced by a factor 2, while

Nkx6.1, Nkx6.2 and Olig2 expression domains are 1.5-2 fold expanded to more dorsal positions. Error

bars indicate SD (n=6)

6

Figure S4, related to Figure 5: Direct repression by homeodomain proteins on the CRM level, and

Western blot analyzes of ventral and dorsal neural tissue.

(A) Forced expression of Dbx and Msx proteins represses Nkx6.1 (upper panel) and CRMNkx6.1 (lower

panel). Electroporated side (+), control side (-)

(B) EMSA showing binding of Msx1 protein to labeled HBSMsx nucleotides, and Dbx1 binding to the

HBSDbx nucleotides. Mutated HBS nucleotides cannot compete out this binding. Upper boxes: specific

binding to HBS sites, double bands are indicative of homodimerisaton of HD proteins (Zhang et al.,

1997). Lower boxes: unbound oligonucleotides.

(C) CRMNkx6.1 is repressed to a similar degree by each of Msx1-3 proteins in P19 cells (1ng). This

activity is abolished upon mutation of HBSMsx. Repression by Dbx2 is stronger and unaffected by

mutation of HBSMsx. Baseline for % repression is set as full activity of CRMNkx6.1 in the presence of

GliH and Sox3. Error bars indicate SD (n=3)

(D) Forced expression of any of Msx1, 2 or 3 proteins represses Nkx6.1 (upper panel) and Dbx1 (lower

panel). Electroporated side (+), control side (-)

(E) Forced expression of Nkx6 and Msx2 proteins represses Dbx1 (upper panel) and CRMDbx1 (lower

panel). Electroporated side (+), control side (-)

(F) Msx1 (1ng) and GliR (0,5ng) cooperatively repress the Sox3- and Gli3H- induced activation of

CRMNkx6.1 in P19 cells. Baseline for repression % is set as full activity of CRMNkx6.1 when Sox3 and

Gli3H are added. Error bars indicate SD (n=3)

(G) CRMDbx1 is repressed by Msx2 and Nkx6.1 proteins in P19 cells. Baseline for repression % is set

as full activity of CRMDbx1 when Sox3 is added. Error bars indicate SD (n=3)

(H) Ptc1¨loop2 construct represses wild type CRMNkx6.1 expression.

(I) Micro-dissection of ventral and dorsal extremes of the neural tube (delineated by the dashed lines)

results in a clear separation between both tissue types, as shown by a Western blot analysis with

antibodies against the ventral marker Nkx6.1 and dorsal marker Pax7 (brackets indicate their neural

expression domains).

(J) Western blot analysis with a Gli3 antibody reveals a 83 and 190 kD band in extracts from chick

and mouse cells. Mouse neural stem cells (mNSC) specified into a ventral fate (data not shown)

express the full length Gli3 activator form only, whereas the dorsal neural stem cells (data not shown)

7

express both the Gli3 activator and repressor form. Importantly, the mNSCs have a broader ventral

identity than cells isolated from the ventral extreme of the neural tube, and express therefore

presumably higher levels of Gli3 mRNA and, consequently, Gli3 protein.

Figure S5, related to Figure 6: The SoxB1 proteins cooperate with GliA to activate neural specific

CRMs.

A) Quantification of Nkx6.1+ neural stem cells and Nkx6.1 transcript by qPCR. Error bars indicate SD

(n=3)

B) Myf5 CRM (Gustafsson et al., 2002) contains a functional GBS, but does not bind SoxB1 proteins,

and is inactive in the neural tube.

C) A combinatorial activity of Sox2 and Gli3H induces the CRMs of Nkx2.2, Nkx6.1 and Nkx6.2 in a

similar manner as Sox3 and Gli3H. RLU, relative luciferase units; error bars indicate SD (n=2)

8

Supplemental Experimental Procedures

Neural stem cell culture

E14.1 ES cells were propagated on Cell Bind culture flasks (Corning) in KoDMEM (Invitrogen)

supplemented with 2000 U/ml LIF (Chemicon), 10% KSR, 5% FCS, 0.1 mM nonessential amino

acids, 1 mM pyruvate (Invitrogen), and 0.1 ȝM ȕ2-mercaptoethanol (Sigma). For in vitro neural stem

cell differentiations, cells were dissociated with TripLExpress (Invitrogen), washed once with PBS

and plated on gelatinized dishes (10x6 cells per 10 cm2 or 24 well plates with coverslips) in N2B27

differentiation medium (Andersson et al., 2006) supplemented with 10 ng/ml bFGF. An intermediate

neural progenitor identity was induced by 1 ȝM retinoic acid (RA; all-trans, Sigma), and for a ventral

or dorsal progenitor identity a concentration of 100nM RA was used. Furthermore, cells were also

treated for 4 days with 1 ȝM cyclopamine (Sigma) to induce dorsal progenitor fates, 5 nM SHH Ag

1.3 (Curis Inc.) and 0.5ȝM cyclopamine for an intermediate neural tube phenotype and 100nM SHH

Ag1.3 for the induction of ventral progenitors. After 4 days of differentiation the cells were harvested

for qPCR and ChIP or immunocytochemistry.

Primary Antibodies

ß-Gal (goat; Biogenesis cat:4600-1409);

ȕ-actin (Sigma).

c-myc (9E10), Shh, Nkx2.2 (all from Developmental Studies Hybridoma Bank);

c-myc (rabbit; Clontech)

GFP (rabbit; Invitrogen cat:A6455);

Gli2 (Cell Signaling #2585);

Gli3 (rabbit; gift S. Mackem; see Chen, et al., 2004)

Nkx2.2 (rabbit; gift from J. Briscoe);

Olig2 (guinea pig; gift from B. Novitch);

Probes

All probes were prepared from chick cDNAs.

9

Dbx1, Dbx2 (Pierani et al., 1999);

Nkx2.2, Nkx6.1 (Briscoe et al., 2000);

Nkx6.2 (Vallstedt et al., 2001);

Olig2 (Novitch et al., 2001);

Msx1 (Andersson et al., 2006).

Msx2 (chest 410j10, NotI, T3)

Patched1 (Marigo and Tabin, 1996);

Pax6 (chEST 85kp16, NotI, T3)

EMSA

The following double stranded DNA sequences were used to analyze the binding efficiency of

mycGli3ZnF and mycGliZnǻF5 proteins to GBSs:

aaggAAAGCATCTGGTGGTCTTTATT (GBS1Nkx6.1);

aaggAACGCAACAGGTGGTTTTT (GBSDbx1);

aaggCGCGTCCTGGGTGGTCGGA (GBSNkx2.2);

aaggCGCGTCCTGAAAGGTCGGA (mut GBSNkx2.2)

The following double stranded DNA sequences were used to analyze the binding efficiency of Msx1

and Dbx1 proteins to HBSs in the CRMNkx6.1:

aaggCCAGCTAATTATGT (HBSMsx)

aaggCCAGCTAgcTATGT (mut HBSMsx)

aaggAGATGAATTTATAACTT (HBSDbx)

aaggAGAgGccTTTATAACTT (mut HBSDbx)

NIH-3T3 and COS7 cells were transfected with constructs of interest using Lipofectamine and

Plus Reagent (Invitrogen) according to manufacturer’s recommendations and after 24 hours

incubation nuclear extract was prepared as previously described (Ågren et al., 2004). The DNA-

binding ability of Gli3, Gli3ZnF and Gli3ZnFǻF5 proteins was monitored by electrophoretic gel

10

mobility shift assay (EMSA) using a binding buffer containing 1.0 ȝg of poly-dI/C, 0.25 ȝg poly-

dG/dC, 100 mM KCl, 40 mM HEPES (pH 7.9), 10 mM MgCl2, 0.4 mM EDTA, 4 mM DTT and

40% glycerol (modified from: Ågren et al., 2004) and 1xTGE as running buffer. The DNA-binding

ability of Msx1 and Dbx1 was monitored by EMSA using a binding buffer containing 1 ug

poly(dI/dC), 10 mM TRIS-Cl pH 7.5, 75 mM NaCl, 1 mM EDTA, 6% glycerol, 3 mM spermidine,

1 mM DTT, 0.5 mM PMSF (Ferretti, et al 2000) and 0.25x TBE as running buffer. For the EMSA,

10 ȝg of nuclear extract was mixed with appropriate binding buffer (total volume 20 ul) and

incubated for 10 min at room temperature. In indicated competition experiments, a myc-antibody

or a 100-fold excess of unlabeled probe was added as specific competitors and incubated for 15

minutes at room temperature. The DNA/protein-binding reaction was carried out by adding a

double-stranded 32P-labelled oligo and incubation on ice for 30 minutes. DNA-protein complexes

were separated under non-denaturing conditions in a 5% polyacrylamide gel (29:1) in appropriate

running buffer at room temperature. The signals were detected using Phosphoimaging (FLA-7000,

Fujifilm).

DNA constructs

pCAGGS vectors were used to express the following proteins: mycGli3ZnF (a.a. 471-645 of human

Gli3; gift from J. Briscoe), mycGli3ZnFǻ5 (deletion by mutagenesis of a.a. 620-626, corresponding

to the fifth Zinc finger), GliH (Stamataki et al., 2005), SmoM2 (cDNA was subcloned from vector

described in Xie et al., 1998), Ptcǻloop2 (Briscoe et al., 2001), Sox3 (Bylund el al., 2003), Msx1

(Andersson et al., 2006), Msx2 and Msx3 (isolated from cDNA library), and Dbx2 (Muhr et al., 2001).

The GliR expression construct used in cell transfection experiments has been previously described

(Wang et al., 2000). For electroporation: pCAGG-IRES-GFP, BGZA (Yee and Rigby, 1993), BG-

eGFP (modified BGZA vector in which LacZ was replaced with eGFP) and for cell transfection

experiments pTK81luc (Nordeen et al., 1988; with modified multiple cloning site) and HSP68lacZ

(gift from A. Joyner) were used. CRMs were amplified by PCR from BACs (series RP23/24 from

BACPAC Resources Center) and subsequently cloned into reporter constructs. Genomic positions of

isolated functional CRMs are given in supplemental figure 1A (the second inactive element that we

identified for each of Nkx6.1 and Dbx2 were located at chr5:101949539-101950655 and

11

chr15:95469390-95470596, respectively). Mutagenesis was performed with either the QuickChange

single or multiple Site-directed mutagenesis kit (Stratagene). All DNA fragments generated by PCR

amplification and mutagenesis were sequence verified. The 4xGBS-reporter constructs were made by

cloning DNA oligos containing four repeats of the following sequences: (GBS indicated in bold):

CTACTTGGGTGTTCCCT (GBSFoxA2), CGTCCTGGGTGGTCGGA (GBSNkx2.2),

CGCGCTGGGTGGTCGGG (GBSNkx2.9), AAATCTGGGTGGTATGG (GBSOlig2),

AGCATCTGGTGGTCTTT (GBS1Nkx6.1), CAGTTTGGGTGGTATTT (GBS2Nkx6.1),

CCTGACAGGTGGTTAAC (GBSDbx1). For primer sequences and concentration of constructs,

please contact author for correspondence.

Supplemental References

Andersson E., Tryggvason U., Deng Q., Friling S., Alekseenko Z., Robert B., Perlmann T., Ericson J.

(2006). Identification of intrinsic determinants of midbrain dopamine neurons. Cell. 124, 393-405.

�

Briscoe J., Chen Y., Jessell T.M., Struhl G. (2001). A hedgehog-insensitive form of patched provides

evidence for direct long-range morphogen activity of sonic hedgehog in the neural tube. Mol Cell. 7,

1279-91.

Bylund M., Andersson E., Novitch B.G., Muhr J. (2003). Vertebrate neurogenesis is counteracted by

Sox1-3 activity. Nat Neurosci. 6, 1162-8.

Chen Y., Knezevic V., Ervin V., Hutson R., Ward Y., Mackem S. (2004). Direct interaction with

Hoxd proteins reverses Gli3-repressor function to promote digit formation downstream of Shh.

Development. 131, 2339-47.

Nordeen S.K. (1988) Luciferase reporter gene vectors for analysis of promoters and enhancers.

Biotechniques. 6, 454-8.

12

Pierani A., Brenner-Morton S., Chiang C., Jessell T.M. (1999) A sonic hedgehog-independent,

retinoid-activated pathway of neurogenesis in the ventral spinal cord. Cell. 97, 903-15.

Wang B., Fallon J.F., Beachy P.A. (2000). Hedgehog-regulated processing of Gli3 produces an

anterior/posterior repressor gradient in the developing vertebrate limb.

Cell. 100, 423-34.

Yee S.P., Rigby P.W.. (1993). The regulation of myogenin gene expression during the embryonic

development of the mouse. Genes Dev. 7, 1277-89.

Zhang H., Hu G., Wang H., Sciavolino P., Iler N., Shen M.M., Abate-Shen C. (1997).

Heterodimerization of Msx and Dlx homeoproteins results in functional antagonism. Mol Cell Biol.

17, 2920-32.