Soxb1-Driven Transcriptional Network Underlies Neural-Specific Interpretation of Morphogen Signals

SoxB1-driven transcriptional network underlies neural-speciﬁc interpretation of morphogen signals

Tony Oosterveena,1, Sanja Kurdijaa,1, Mats Ensteröa,b, Christopher W. Uhdea, Maria Bergslandb, Magnus Sandberga,b, Rickard Sandberga,b, Jonas Muhra,b, and Johan Ericsona,2

aDepartment of Cell and Molecular Biology, Karolinska Institutet, 171 77 Stockholm, Sweden; and bLudwig Institute for Cancer Research, 171 77, Stockholm, Sweden

Edited* by Thomas M. Jessell, Columbia University College of Physicians and Surgeons, New York, NY, and approved March 20, 2013 (received for review November 28, 2012)

The reiterative deployment of a small cadre of morphogen signals and critically requires the direct synergistic input of SoxB1 pro- underlies patterning and growth of most tissues during embyo- teins at the transcriptional level (9). The SoxB1 group of TFs, Sox1 genesis, but how such inductive events result in tissue-specific to -3, are broadly expressed in the developing CNS and play an responses remains poorly understood. By characterizing cis-regula- important role in maintaining neural progenitor properties (10, tory modules (CRMs) associated with genes regulated by Sonic 11). Importantly, these TFs are expressed in a neural-specific hedgehog (Shh), retinoids, or bone morphogenetic proteins in the manner (Fig. 1B) (12), raising the possibility that SoxB1 proteins fi CNS, we provide evidence that the neural-specific interpretation of contribute to the tissue-speci c selection of target gene activation morphogen signaling reflects a direct integration of these pathways upon Shh pathway activation. with SoxB1 proteins at the CRM level. Moreover, expression of Graded Shh signaling underlies anteroposterior patterning of SoxB1 proteins in the limb bud confers on mesodermal cells the the developing limb bud, and Gli proteins are bound to the CRMs of class II genes Nkx2.2 and Nkx6.1, despite the fact that potential to activate neural-specific target genes upon Shh, retinoid, they are transcriptionally silent in this tissue (8, 9). We therefore or bone morphogenetic protein signaling, and the collocation of wished to determine whether expression of SoxB1 proteins in binding sites for SoxB1 and morphogen-mediatory transcription fac- mesodermal cells of the limb bud is sufficient to activate neural-

fi BIOLOGY tors in CRMs faithfully predicts neural-speci c gene activity. Thus, an specific Shh-target genes in response to Shh signaling. To test unexpectedly simple transcriptional paradigm appears to conceptu- this theory, we electroporated expression constructs into the DEVELOPMENTAL fi ally explain the neural-speci c interpretation of pleiotropic signal- base of the presumptive limb bud at Hamburger–Hamilton ing during vertebrate development. Importantly, genes induced in (HH) stage 15 embryos (Fig. 1A) and collected embryos for anal- a SoxB1-dependent manner appear to constitute repressive gene ysis 6 or 40 h postelectroporation (hpe). The ambient level of regulatory networks that are directly interlinked at the CRM level Shh signaling was low at the base of the limb at 40 hpe, as indi- to constrain the regional expression of patterning genes. Accord- cated by low expression of the Shh receptor Patched 1 (Fig. S1A), ingly, not only does the topology of SoxB1-driven gene regulatory which is a primary response gene of Shh signaling (13). Activa- networks provide a tissue-specific mode of gene activation, but it tion of the Shh pathway by forced expression of a constitutively also determines the spatial expression pattern of target genes active form of Smoothened (SmoM2) induced Ptc1,aswellas within the developing neural tube. a Shh-responsive CRM associated with Ptc1 (Fig. 1 B and E) (14). However, neither SmoM2 nor Sox2 nor Sox3 expressed Sox3 | Gli | Sox2 | positional information individually in the limb bud were sufficient to activate expression of class II genes (Fig. 1B, and Fig. S1 B–E). In contrast, misexpression of SmoM2 and SoxB1 in combination resulted in Results a striking induction of Nkx2.2, Nkx6.1, Nkx6.2,andOlig2 in the limb bud (Fig. 1B,andFig. S1 D and E). Similarly, CRMNkx2.2-, rowth and patterning of the developing embryo is ac- CRMNkx6.1-, and CRMNkx6.2-driven reporter constructs (Fig. 1C), complished by redeploying a remarkably small number of G E signaling pathways, but how these activities are interpreted in which are active in neural tissue but not mesodermal cells (Fig. 1 ) a tissue-specific manner is poorly understood (1, 2). During CNS (9), were activated in the limb bud by Shh signaling in a SoxB1- development, the morphogens Sonic hedgehog (Shh), retinoids, dependent manner (Fig. 1E). Direct binding of SoxB1 to CRMs was required for transcriptional activation, as indicated by the and bone morphogenetic proteins (BMPs) impart positional Nkx2.2 Nkx6.1 identity to progenitor cells at ventral, intermediate, and dorsal abolished activity of CRM and CRM following in- positions of the neural tube, respectively (3–5). These signals act activation of the Sox-binding sites (SBSs) in these elements by regulating the spatial expression patterns of homeodomain and (Fig. 1 C and F). Nkx2.2 is a known repressor of Olig2 in the basic-helix–loop–helix transcription factors (TFs) that specify cell neural tube (15), and we noted that expression of Olig2 was – lower than that of the other class II genes at 40 hpe, but identity along the dorsoventral (DV) axis of the neural tube (3 7), Olig2 and many of these TFs are activated by these pathways specifically CRM exhibited no activity, implying that Nkx2.2 represses Olig2 in the limb bud. Consistent with this finding, at 6 hpe we in neural progenitors but not other developing tissues. For ex- Olig2 ample, graded Shh signaling underlies patterning of the ventral observed that the class II genes, including Olig2 and CRM , neural tube by repressing or inducing expression of class I and were induced. Moreover, Nkx2.2 and Olig2 were expressed in class II TFs, respectively, but this set of genes is not induced during Shh-mediated patterning of the developing limb bud (8). We have recently identified cis-regulatory modules (CRMs) for Author contributions: T.O., S.K., M.E., M.B., M.S., R.S., J.M., and J.E. designed research; the Shh-regulated class I and II TFs Nkx2.2, Nkx2.9, Olig2, Nkx6.1, T.O., S.K., M.E., C.W.U., M.B., and M.S. performed research; M.E. contributed new reagents/analytic tools; T.O., S.K., M.E., C.W.U., M.B., R.S., J.M., and J.E. analyzed data; Nkx6.2, Dbx1, Dbx2, and Pax6 (9). These CRMs recapitulate the and T.O., C.W.U., R.S., and J.E. wrote the paper. expression patterns of their respective endogenous genes in the fl neural tube, and are directly regulated by bifunctional Gli tran- The authors declare no con ict of interest. scription factors of the Shh pathway (5) via conserved Gli-binding *This Direct Submission article had a prearranged editor. sites (GBS) (Fig. 1C and Fig. S1F) (9). Importantly, functional 1T.O. and S.K. contributed equally to this work. analyses of these elements indicate that activation of Shh-induced 2To whom correspondence should be addressed. E-mail: [email protected]. Nkx2.2 Olig2 Nkx6.1 Nkx6.2 CRMs (CRM ,CRM ,CRM ,andCRM )byac- This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. tivator forms of Gli proteins (GliA) is sequence context-dependent 1073/pnas.1220010110/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1220010110 PNAS Early Edition | 1of6 Downloaded by guest on September 30, 2021 Fig. 1. SoxB1 proteins confer competence to activate neural-specific Shh target genes in the limb bud. (A) Schematic illustrating experimental setup for electroporation of the proximal limb bud. (B) Expression of Sox3, Ptc1, and class II genes in the neural tube and limb bud following constitutive activation of Shh signaling by forced expression of SmoM2 alone (Left) or in combination with Sox3 (Right) in the proximal limb bud. Electroporated region of limb bud indicated by Ptc1 expression. Dashed line delineates proximal border of limb bud. NT, neural tube; LB, limb bud. (C) Schematics illustrate the position of SBSs and the GBS in each class II CRM. (D) Schematic of transverse section from chicken embryo. Blue boxes indicate regions of neural or limb bud tissue shown in E. (E) Activities of CRMs associated with Ptc1 and class II genes (white) in the ventral neural tube (Left, magnification 80×) and limb bud following electroporation with SmoM2 alone (Center) or in combination with Sox3 (Right). GFP electroporation control (blue). CRM activity in limb is shown after 40 hpe (magnification 70×), except for CRMOlig2 that is 6 hpe (magnification 100×). (F, Upper) Schematics indicate the relative positions of SBSs (red) and GBSs (blue) in CRMs. X indicates mutationally inactive sites. Activity of wild-type or mutated CRMs (white) in the limb bud; GFP electroporation control (blue) (magnification 50×). (G) Expression of the neural-specific class II genes in the proximal limb bud 6 hpe in response to forced expression of Sox3 and SmoM2 (magnification 100×). (H and I) Expression of Sox3, dHand,andSox1 in the limb bud following forced expression of Sox3 (magnification 30×). Dotted red oval indicates electroporated area, as determined by Sox3 expression.

a mutually exclusive fashion (Fig. S1D), indicating that cross- GBS and the nearest conserved SBS in CRMs associated with repressive interactions between neural patterning transcription Shh-induced class II genes was 36 bp (9), implying that collocation factors are recapitulated in this assay (see also Fig. S1E) (5). of conserved SBSs and GBSs in noncoding genomic sequences The rapid induction of class II genes in the limb bud (within 6 could be sufficient to identify genes regulated by Shh signaling hpe), together with the requirement for direct binding by SoxB1 in neural tissue. Binding of the transcriptional coactivator p300 and the fact that Gli is bound to CRMNkx2.2 and CRMNkx6.1 ele- has been shown to accurately predict tissue specific enhancer ments in the limb bud (8, 9), strongly indicates that the tissue- activity during embryogenesis (16), and we therefore devised specific expression of Shh-regulated genes is primarily de- a computational method to identify TF-binding sites (TFBSs) that termined by the combinatorial activity of Sox and Gli at the were enriched within 50 bp of conserved GBSs in p300-bound transcriptional level. It is notable that expression of the meso- elements from limb bud (n = 25) and neural tissue (n = 72) (SI dermal marker dHand was maintained in Sox3-expressing cells 40 Experimental Procedures and Fig. S2A). This TFBSs analysis hpe (Fig. 1 H and I) and that a negligible induction of the neural revealed a significant overrepresentation of SBSs around GBSs marker Sox1 was detected at this stage. Importantly, these data in elements bound by p300 in neural compared with limb tissue argue that the induction of neural-specific class II genes in the (Fig. 2A and Table S1). limb cannot be explained by reprogramming of mesodermal cells To directly test the predictive value of a GBS-SBS signature, we into bona fide neural progenitors, although we do not exclude the performed a genome-wide search for consensus GBS and SBS- possibility that long-term expression of SoxB1 proteins in the limb containing noncoding elements conserved between mouse, hu- could eventually result in reprogramming of mesodermal cells. man, and opossum, and examined the expression patterns and Our data show that SoxB1 proteins account for the neural- regulation of nearby genes (Table S2). Eighty-three presumptive specific activation of class II genes in response to Shh signaling, CRMs were identified when the maximum distance between the suggesting that other genes may be regulated according to a sim- GBS and SBS was set at 36 bp (Table S3), and the number of ilar transcriptional logic. The farthest distance between a functional positive regions decreased as a function of increased distance

2of6 | www.pnas.org/cgi/doi/10.1073/pnas.1220010110 Oosterveen et al. Downloaded by guest on September 30, 2021 BIOLOGY DEVELOPMENTAL

Fig. 2. The collocation of a SBS and GBS in CRMs defines a genome-wide transcriptional code for Shh-regulated genes active in the CNS. (A) In silico screen of p300-bound mouse elements from limb or CNS (as determined by ref. 16) for association strength between GBSs and TFBSs that are conserved between mouse, human, and opossum. Significant values for association of Sox-binding sites with GBSs are highlighted in red. Error bars indicate SD (n = 10). (B) Statistics for database and literature survey of the tissue- and region-specific expression patterns of transcription factors associated with collocated SBS and GBS sites. (C) Neural expression of genes identified following a genome-wide screen for a SBS and GBS located at a maximum distance of 36 bp from one another. GBS affinity scores are shown in Table S10.(Upper) SBS and GBS sequence logos. (D) Expression of SoxB1-Gli target genes in the unelectroporated limb bud (control) or following forced expression of SoxB1 and SmoM2. (E) Transcriptional assays in P19 cells on CRMs containing a SBS-GBS signature, indicating a synergistic re- lationship between SoxB1 and the Shh pathway for activation of these CRMs. RLU: relative luciferase units. Error bars indicate SD (n = 2).

between conserved Gli and SoxB1 binding sites (Fig. S2 B and C). in the limb bud assay, isolation of the CRM associated with A survey of gene-expression databases and the literature indicated phosphatidic acid phosphatase type 2B (Ppap2b, CRMPpap2b) that genes encoding TFs that could be linked to mammalian GBS- showed that this element could be synergistically activated by SBS elements have a statistically significant higher probability Sox3 and an activator form of Gli3 (Gli3-High, Gli3H) in P19 cells (Fisher’s exact test, P < 0.01) of being expressed in the ventral (Fig. 2E), and similar results were obtained for three CRMs as- neural tube (57%) compared with the posterior limb bud (7%), sociated with Ebf3 (CRMEbf3), C20orf19 (CRMC20orf19), and where Shh signaling levels are high (Fig. 2B and Table S4) (8, 17). Zfp238 (CRMZfp238) (Fig. 2E). Taken together, these data dem- A similar search for GBS-SBS signatures between chick and onstrate that enrichment and proximity of GBSs and SBSs in mouse identified 45 presumptive CRMs enriched for conserved conserved noncoding sequences can faithfully predict genes GBSs and SBSs (Table S5). This relatively low number of regions regulated by Shh signaling in the developing CNS, providing probably reflects the incomplete sequence information of the evidence that a common transcriptional strategy underlies tissue- chick genome at the time of analysis. Nevertheless, we used this specific activation of Shh target genes in the developing CNS. dataset to evaluate whether GBS/SBS elements promote neural We next wished to determine whether SoxB1 proteins contrib- gene expression in a Shh-dependent manner. Of the closest ute to the neural-specific interpretation of other pleiotropic signals neighboring genes, we selected those located within 150 kb of active in the developing CNS. The class II genes Dbx1 and Dbx2 a Sox-Gli signature (except for one located ∼1 × 106 bp away). are expressed in the intermediate neural tube, but whereas re- After excluding from our dataset the class I and II genes for which pressor forms of Gli proteins (GliR) are important to suppress Shh-regulated CRMs have been identified previously (9), we expression of these genes in the dorsal neural tube (9, 18), Shh obtained 15 functional riboprobes from the remaining 20 neigh- signaling is not required to activate them (3, 9). Instead, retinoic boring genes identified. Expression analysis of this set of genes acid (RA) signaling by paraxial mesoderm has been implicated revealed that 87% (13 of 15) of genes showed a clear ventral in the induction of these genes in the neural tube (3). RA binds bias of expression in the neural tube (Fig. 2C). Strikingly, most nuclear receptors directly to activate transcription (19) and of these genes could be ectopically activated in the limb bud in examination of CRMDbx1 and CRMDbx2 revealed that each con- response to Sox3 and SmoM2 expression (Fig. 2D). Moreover, tains two conserved nuclear receptor binding sites (NRBS) (Fig. although Ppap2b was one of two genes that were not induced 3A and Fig. S1F) resembling RARE (retinoic acid response

Oosterveen et al. PNAS Early Edition | 3of6 Downloaded by guest on September 30, 2021 element) half sites. Mutational inactivation of these sites abol- P19 cells overexpressing Sox3 were also exposed to increasing ished the activity of these CRMs in vivo (Fig. 3A), providing ev- concentrations of RA (Fig. 3B). Furthermore, this synergistic idence that RA directly regulates transcription of Dbx1 and Dbx2. activation was abolished in CRMDbx1 carrying either the inacti- Like CRMs for Shh-induced class II genes (9), CRMDbx1 and vated RARE or the closely located SBS4 (Fig. 3B). Thus, SoxB1 CRMDbx2 contain four and three conserved SBSs, respectively, proteins are required for and appear to act synergistically with and ChIP experiments showed that SoxB1 proteins bound these RA to activate transcription of Dbx genes in the developing elements in vivo (Fig. 3C). Inactivation of the three SBSs in neural tube. CRMDbx2 led to a complete loss of CRM activity (Fig. 3B), Analysis of conserved RAREs in p300-bound elements from whereas mutation in CRMDbx1 of SBS4, which lies nearby a typical limb bud and neural tissue revealed an overrepresentation of RARE (DR2) (Fig. 3A), eliminated CRM-driven reporter activity SBSs located in proximity to RAREs in elements active in neural in the neural tube (Fig. 3A). Analyses of CRMDbx1 in transcrip- tissue (Fig. 3F and Table S6), suggesting that an SBS-RARE tional assays in P19 cells indicated that exposure of cells to RA transcriptional code may be of predictive value in determining was not sufficient to induce CRMDbx1 in vitro, whereas forced neural-specific gene expression. Indeed, a complementary ge- expression of Sox3 alone resulted in a moderate up-regulation of nome-wide analysis identified 545 RAREs conserved between transcriptional activity (Fig. 3B). Importantly, however, there was mouse and chick (Table S7), 52 of which are located within 50 bp a synergistic and dose-dependent activation of CRMDbx1 when of a conserved SBS, and these collocated sites lie nearby genes

Fig. 3. SoxB1 proteins are required for activation of retinoid and BMP target genes in the CNS. (A)(Upper) Schematics indicating positions of SBSs and NRBSs in CRMDbx1 and CRMDbx2.(Lower) Neural activity of CRMDbx1 (24 hpe, magnification 100×) and CRMDbx2 (40 hpe, magnification 90×). For each CRM (Left) two wild-type CRMs driving expression of distinct reporter genes (white and blue); (Center) NRBS- inactivated and wt control CRM (white and blue, respectively); (Right) SBS-inactivated and wild-type control CRM (white and blue, respectively). (B) Transcriptional assays in P19 cells on wild-type CRMDbx1 or elements carrying either an inactivated SBS4 or inactivated NRBS, as indicated, following addition of RA alone (Left) and in combination with Sox3 (Center, Right). (C) Sox3 ChIP on mouse tissue for CRMDbx1 and CRMDbx2.(D) Expression of Sox3 and class I genes in the neural tube and limb bud following forced expression of Sox3 in the proximal limb bud. Electroporated region of limb bud indicated bySox3 expression. Dashed line delineates proximal border of limb bud. NT, neural tube; LB, limb bud. (E) Activity in the limb of a wild-type (Top, Middle) or NRBS-inactivated (Bottom) CRMDbx1 (red) and GFP electroporation control (green), electroporated alone or in combination with Sox3 (magnification 70×). (F) In silico screen of p300-bound mouse elements from limb or CNS (as determined by ref. 16) for association strength between RARE and TFBSs that are conserved between mouse, human, and opossum. Significant values for association of Sox- and Fox-binding sites with RAREs are highlighted in red and blue, respectively. Error bars indicate SD (n = 10). (G) Enrichment of functional annotation as CNS or limb of genes associated with an SBS-RARE signature following a genome-wide screen for these sites that are conserved between mouse and chick and either ≤50 bp or >50 bp of one another. (H) Expression of Msx1 and Gsh1 following forced expression of a constitutively active BMP type 1 receptor (Alk-2CA) alone (Left)orin combination with Sox3 (Right) in the proximal limb bud. Only the electroporated regions of limb bud are shown (magnification 70×). (I, Upper) Schematics illustrate the position of the isolated CRM in the Msx1 locus and the location of its SBS and SmBSs. Expression of Msx1 (Far Left) and activity of its associated CRMMsx1 in the dorsal neural tube (magnification 90×). (Center Left) Two wild-type CRMs driving expression of distinct reporter genes (white and blue); (Center Right) SBS- inactivated and wild-type CRM (white and blue, respectively); (Far Right) SmBS-inactivated and wt CRM (white and blue, respectively).

4of6 | www.pnas.org/cgi/doi/10.1073/pnas.1220010110 Oosterveen et al. Downloaded by guest on September 30, 2021 significantly enriched for functions in neural development but not limb development (Fig. 3G and SI Experimental Procedures), among them Pax6, which has been implicated in the regulation of DV patterning in the neural tube (5). In contrast, genes lying nearby RAREs located >50 bp away from an SBS showed essentially no difference in the level of enrichment for functional annotation of neural versus limb tissue (Fig. 3G). Retinoids have also been shown to critically regulate limb bud development and to be present at relative high concentrations in the proximal limb bud (20), and consistent with this, we found that expression of Sox3 alone in the base of the presumptive limb was sufficient to activate Dbx1, Dbx2, and Pax6 in this tissue (Fig. 3D). CRMDbx1 was also activated in the limb bud in response to Sox3 expression, and, importantly, this activation required the RARE in this element (Fig. 3E). Taken together, these analyses imply that SoxB1 proteins and retinoids may cooperatively regulate a broad array of genes. BMP signaling is mediated by nuclear translocation of Smad TFs (Smad 1/5/8) (21), and studies of differentiating hematopoi- etic cells have indicated that the lineage-specific TFs GATA1 and C/EBPα function to recruit SMAD1 to cell type-specific enhancers active in erythroid and myeloid cells, respectively (22). BMP signaling induces members of the Msx, Pax, and Gsh families of TFs in the dorsal neural tube (4, 7), and we found that Msx1 and -2, Fig. 4. Recapitulation of dorsoventral neural patterning in the limb bud and Pax3 and -7,andGsh1 and -2 were associated with presumptive fi CRMs containing conserved SBS and SMAD binding sites a model of the neural-speci c responses to morphogen signaling. (A) Schematic of transverse section from chicken embryo. Blue boxes indicate regions of (SmBSs) (Table S8). Msx proteins act as repressors of Nkx6.1 and neural or limb bud tissue shown in B.(B) Forced expression of Sox3 confers

Dbx1 expression in the neural tube (9), and we isolated a non- on mesodermal cells the potential to recapitulate the DV patterning of the BIOLOGY coding sequence associated with Msx1 (CRMMsx1) containing two

neural tube (see Left), as indicated by the induction of the Shh-regulated DEVELOPMENTAL conserved SmBSs and one SBS (Fig. 3I and Fig. S1F) and assessed Nkx2.2, retinoid-regulated Dbx2, and BMP-regulated Gsh1/2 genes in the the transcriptional activity of this element in the neural tube. chick limb bud (Right). Shh*, Shh pathway activated by SmoM2; RA*, en- Msx1 CRM precisely recapitulated the endogenous expression pat- dogenous levels of RA signaling; Bmp*, Bmp pathway activated by Alk-2CA. tern of Msx1 in the dorsal neural tube (Fig. S3A), and this activity (C) Neural-specific gene activation: Shh, RA, and BMPs emanate from local was lost when either the two SmBSs or the unique consensus SBS sources and act in a graded fashion to pattern cells along the DV axis of the in CRMMsx1 was inactivated (Fig. 3I). Msx genes are also induced neural tube. Many cell fate-determining TFs operating downstream of these in nonneural tissues in response to BMP signaling (23, 24), and in morphogens are specifically induced in neural tissue via the cooperative ac- gain-of-function experiments, activation of BMP signaling by tivity of SoxB1 proteins and signal-transducing TFs at the CRM level, enabling forced expression of a constitutively active form of BMP receptor the neural-specific selection of target gene activation in response to these type I (Alk-2CA) alone induced expression of the endogenous Msx1 pleiotropic signals. Regional restriction of gene expression: Genes induced in gene, as well as both wild-type and SBS-mutated forms of a SoxB1-dependent manner constitute functional, repressive GRN that act CRMMsx1 in both neural tissue and limb bud (Fig. 3H and Fig. locally to constrain the regional expression domains of cell fate-determining S3B) (23, 25). Taken together, these data argue that SoxB1 pro- TFs operating downstream of morphogen signals in neural patterning, and the repressive inputs of these Shh-, RA-, and BMP-driven networks are directly motes Msx1 expression but is not absolutely required for Smad- interconnected at the CRM level to control the patterning of target genes mediated induction of Msx1 in the neural tube. Compared with Msx fi along the DV axis of the neural tube. Accordingly, not only does the topology genes, Gsh genes are expressed in a more neural-speci cfashion of SoxB1-driven GRNs provide a tissue-specific mode of gene activation, but it (26, 27), and we found that induction of Gsh1 in limb bud meso- also determines the spatial expression pattern of target genes within that derm by BMPs required the cooperative activity of SoxB1 proteins tissue. G, Gli; NR, nuclear receptor; S, SoxB1; Sm, Smad. (Fig. 3H), implying that expression of at least a subset of genes induced by BMP signaling in neural tissue critically requires the cooperative activity of SoxB1 proteins. of gene regulatory networks (GRNs) that interpret positional information in neural patterning (6, 15, 28–32),butithas Discussion remained elusive how these repressive inputs operate together Collectively, our data provide evidence for a functional integration with morphogen signals at the genomic level to regulate patterns of SoxB1 proteins and nuclear mediators of Shh, retinoid, and of gene expression. Class I and II TFs determine the positional BMP signaling at the CRM level, offering an unexpectedly simple identity of progenitors downstream of the Shh gradient (5), and transcriptional paradigm for the neural-specific interpretation of our recent analyses indicated that these genes are induced by pleiotropic morphogen signals during embryonic development. a synergistic interaction between SoxB1 and Gli proteins at the Accordingly, SoxB1 is sufficient to confer on mesodermal cells the CRM level; positional information is primarily imparted at the potential to up-regulate neural-specific target genes induced by CRM level by cooperative transcriptional repression mediated by morphogen signaling in the limb bud in a manner that recapit- Gli-repressors and cross-repressive interactions between class I ulates the DV patterning activity of Shh, RA, and BMPs in the and II proteins themselves (9). In addition to the induction of neural tube (Fig. 4B). The neural-specific interpretation of Shh these repressive factors, it is notable that many in silico-identified and retinoid signaling appears to be contingent on genome-wide SBS-GBS–associated genes have been implicated previously in signatures consisting of a collocated SBS-GBS or SBS-RARE in Shh signaling and ventral pattern formation, including Tcf4, Tle4, genomic sequences, respectively. In the case of Shh signaling, such Nkx2.1, Six3, Nf1a, hedgehog interacting protein (Hhip), and motifs can faithfully predict expression of associated genes in the beta-transducin repeat containing protein (Btrc) (Table S9). ventral neural tube. Similarly, genes associated with presumptive Furthermore, we observed a functional recapitulation of tissue SBS-RARE–regulated CRMs have a higher probability of being patterning and cross-repressive interactions between class I active in neural tissue compared with limb bud. andIIproteinsinthelimbbud(Fig.4B and Fig. S1E). SoxB1 It has long been established that cross-repressive interactions and Gli proteins therefore appear to define the central node of between Shh-regulated TFs are themselves an integral feature a neural-specific GRN required to translate graded Shh

Oosterveen et al. PNAS Early Edition | 5of6 Downloaded by guest on September 30, 2021 signaling into regional gene-expression patterns in the ventral Experimental Procedures neural tube (Fig. 4C). Although SoxB1 and Gli proteins are For additional information regarding cell transfection assays, ChIP, bioin- sufficient to trigger activation of this network in the limb, this formatics, DNA-constructs, RNA-probes, antibodies, and additional reagents, does not mean that they are the only activators involved, as please refer to SI Experimental Procedures. many genes cooperatively activated by SoxB1 and Gli proteins In Ovo Electroporation. in neural tissue (and in the limb) encode transcriptional activa- pCAGGS expression vectors and reporter constructs tors or repressors that themselves are integral components of the (BGZA or BG-eGFP) containing wild-type or mutated CRMs were electroporated either alone or in combination according to the text, into HH stage 10–12 (neural network. Such proteins are likely to act in a more CRM context- tube) (6) or HH stage 15 (limb bud) (34) chicken embryos. After 6, 24, or 40 h of dependent manner to influence the regional expression pattern of incubation, embryos were isolated and processed for immunohistochemistry Shh-regulated genes within the neural tube, as exemplified by Tcf and in situ hybridization. Reporter constructs containing mutated CRMs were proteins, which act cooperatively with SoxB1 and Gli to activate electroporated at concentrations that would result in clear detectable activities transcription of Nkx2.2 in the ventral-most neural tube (9) and for their wild-type versions. For each experimental condition, at least three which are induced upon SoxB1 and SmoM2 expression in the limb different embryos from two independent experiments were analyzed. bud. Such a node-driven process provides a mechanistic rationale Immunofluorescence and in Situ Hybridization. Immunofluorescence and in for how established GRNs can be coopted and redeployed by cells situ hybridization were performed essentially as previously described (6, 35). simply by up-regulating core activator components of the GRN (33). Positional information provided by retinoids and BMP sig- P19 Cell Transfection Assays. P19 cells were transfected by using Lipofect- naling in neural tissue is likely to be similarly interpreted by amine and Plus Reagent (Invitrogen). SoxB1-driven repressive GRNs (Fig. 4C). Importantly, these GRNs are directly interlinked at the CRM level, as illustrated by ChIP. ChIP experiments were performed as previously described (9). Nkx6.1, which is induced by SoxB1 and Gli proteins but posi- tionally restricted to ventral progenitors by RA- and BMP-in- ACKNOWLEDGMENTS. We thank J. Briscoe, M. Goulding, B. Novitch, M. Ros, and R. Toftgård for reagents. This work was supported by the Swedish Re- duced genes (9), and Dbx1, which is regulated by both RA and search Council (33X-06555; Developmental Biology for Regenerative Medi- Gli-repressors (present study) (9). Accordingly, the SoxB1 tran- cine); The Royal Swedish Academy of Sciences by donation from the scriptional code provides not only a strategy for the tissue-specific Wallenberg Foundation; the Swedish Foundation for Strategic Research selection of target genes, but also for the induction of GRNs that (Center of Excellence in Developmental Biology, SRL10-0030); The Knut and Alice Wallenberg Foundation (KAW2011.0161); research funds of Kar- drive morphogen interpretation and determine the positional olinska Institutet; and Wenner-Gren Foundation and European Union Marie identity of cells (Fig. 4C). Curie MEIF-CT-2006-025416 (to T.O.).

1. Wolpert L (1996) One hundred years of positional information. Trends Genet 12(9): 19. Chambon P (1996) A decade of molecular biology of retinoic acid receptors. FASEB J 359–364. 10(9):940–954. 2. Rogers KW, Schier AF (2011) Morphogen gradients: From generation to in- 20. Niederreither K, Vermot J, Schuhbaur B, Chambon P, Dollé P (2002) Embryonic reti- terpretation. Annu Rev Cell Dev Biol 27:377–407. noic acid synthesis is required for forelimb growth and anteroposterior patterning in 3. Pierani A, Brenner-Morton S, Chiang C, Jessell TM (1999) A sonic hedgehog- the mouse. Development 129(15):3563–3574. independent, retinoid-activated pathway of neurogenesis in the ventral spinal cord. 21. Massagué J, Seoane J, Wotton D (2005) Smad transcription factors. Genes Dev 19(23): Cell 97(7):903–915. 2783–2810. 4. Liu A, Niswander LA (2005) Bone morphogenetic protein signalling and vertebrate 22. Trompouki E, et al. (2011) Lineage regulators direct BMP and Wnt pathways to cell- nervous system development. Nat Rev Neurosci 6(12):945–954. specific programs during differentiation and regeneration. Cell 147(3):577–589. 5. Dessaud E, McMahon AP, Briscoe J (2008) Pattern formation in the vertebrate neural 23. Pizette S, Abate-Shen C, Niswander L (2001) BMP controls proximodistal outgrowth, tube: A sonic hedgehog morphogen-regulated transcriptional network. Development via induction of the apical ectodermal ridge, and dorsoventral patterning in the 135(15):2489–2503. vertebrate limb. Development 128(22):4463–4474. 6. Briscoe J, Pierani A, Jessell TM, Ericson J (2000) A homeodomain protein code specifies 24. Vainio S, Karavanova I, Jowett A, Thesleff I (1993) Identification of BMP-4 as a signal progenitor cell identity and neuronal fate in the ventral neural tube. Cell 101(4):435–445. mediating secondary induction between epithelial and mesenchymal tissues during 7. Liu Y, Helms AW, Johnson JE (2004) Distinct activities of Msx1 and Msx3 in dorsal early tooth development. Cell 75(1):45–58. neural tube development. Development 131(5):1017–1028. 25. Timmer JR, Wang C, Niswander L (2002) BMP signaling patterns the dorsal and in- 8. Vokes SA, Ji H, Wong WH, McMahon AP (2008) A genome-scale analysis of the cis- termediate neural tube via regulation of homeobox and helix-loop-helix transcription regulatory circuitry underlying sonic hedgehog-mediated patterning of the mam- factors. Development 129(10):2459–2472. malian limb. Genes Dev 22(19):2651–2663. 26. Valerius MT, et al. (1995) Gsh-1: A novel murine homeobox gene expressed in the 9. Oosterveen T, et al. (2012) Mechanistic differences in the transcriptional interpretation central nervous system. Dev Dyn 203(3):337–351. of local and long-range Shh morphogen signaling. Dev Cell 23(5):1006–1019. 27. Szucsik JC, et al. (1997) Altered forebrain and hindbrain development in mice mutant 10. Bylund M, Andersson E, Novitch BG, Muhr J (2003) Vertebrate neurogenesis is for the Gsh-2 homeobox gene. Dev Biol 191(2):230–242. counteracted by Sox1-3 activity. Nat Neurosci 6(11):1162–1168. 28. Ericson J, et al. (1997) Pax6 controls progenitor cell identity and neuronal fate in 11. Bergsland M, et al. (2011) Sequentially acting Sox transcription factors in neural lin- response to graded Shh signaling. Cell 90(1):169–180. eage development. Genes Dev 25(23):2453–2464. 29. Muhr J, Andersson E, Persson M, Jessell TM, Ericson J (2001) Groucho-mediated 12. Guth SI, Wegner M (2008) Having it both ways: Sox protein function between con- transcriptional repression establishes progenitor cell pattern and neuronal fate in the servation and innovation. Cell Mol Life Sci 65(19):3000–3018. ventral neural tube. Cell 104(6):861–873. 13. Marigo V, Tabin CJ (1996) Regulation of patched by sonic hedgehog in the de- 30. Zhou Q, Anderson DJ (2002) The bHLH transcription factors OLIG2 and OLIG1 couple veloping neural tube. Proc Natl Acad Sci USA 93(18):9346–9351. neuronal and glial subtype specification. Cell 109(1):61–73. 14. Agren M, Kogerman P, Kleman MI, Wessling M, Toftgård R (2004) Expression of the 31. Lek M, et al. (2010) A homeodomain feedback circuit underlies step-function in- PTCH1 tumor suppressor gene is regulated by alternative promoters and a single terpretation of a Shh morphogen gradient during ventral neural patterning. De- functional Gli-binding site. Gene 330:101–114. velopment 137(23):4051–4060. 15. Novitch BG, Chen AI, Jessell TM (2001) Coordinate regulation of motor neuron sub- 32. Balaskas N, et al. (2012) Gene regulatory logic for reading the Sonic Hedgehog sig- type identity and pan-neuronal properties by the bHLH repressor Olig2. Neuron 31(5): naling gradient in the vertebrate neural tube. Cell 148(1-2):273–284. 773–789. 33. Davidson EH (2010) Emerging properties of animal gene regulatory networks. Nature 16. Visel A, et al. (2009) ChIP-seq accurately predicts tissue-specific activity of enhancers. 468(7326):911–920. Nature 457(7231):854–858. 34. Suzuki T, Ogura T (2008) Congenic method in the chick limb buds by electroporation. 17. McGlinn E, Tabin CJ (2006) Mechanistic insight into how Shh patterns the vertebrate Dev Growth Differ 50(6):459–465. limb. Curr Opin Genet Dev 16(4):426–432. 35. Schaeren-Wiemers N, Gerfin-Moser A (1993) A single protocol to detect transcripts of 18. Persson M, et al. (2002) Dorsal-ventral patterning of the spinal cord requires Gli3 various types and expression levels in neural tissue and cultured cells: In situ hybrid- transcriptional repressor activity. Genes Dev 16(22):2865–2878. ization using digoxigenin-labelled cRNA probes. Histochemistry 100(6):431–440.

6of6 | www.pnas.org/cgi/doi/10.1073/pnas.1220010110 Oosterveen et al. Downloaded by guest on September 30, 2021