Transcriptional Profiling of Wnt3a Mutants Identifies Sp Transcription Factors as Essential Effectors of the Wnt/β-catenin Pathway in Neuromesodermal Stem Cells

Neuromesodermal (NM) stem cells reside in the primitive streak (PS) of gastrulating vertebrate embryos and generate precursors of the spinal cord and musculoskeletal system. Although Wnt3a/β-catenin signaling is crucial for NM stem cell maintenance and differentiation, few key transcriptional effectors have been identified. Through a concerted transcriptional profiling and genetic approach we have determined that two Zn2+-finger transcription factors, Sp5 and Sp8, are regulated by Wnt3a in the PS, and are essential for neural and musculoskeletal patterning. These results identify Sp5 and Sp8 as pivotal downstream effectors of Wnt3a, and suggest that they are essential for the self-renewal and differentiation of NM stem cells.


Introduction
Germ layer specification is a fundamental aspect of embryogenesis that occurs in the PS region during gastrulation. Under the influence of signals emanating from the PS, pluripotent epiblast stem cells differentiate into the multipotent progenitors that will form the germ layers, ie. the ectoderm, mesoderm and endoderm. As development proceeds, the differentiation potential of descendants of these early embryonic stem cells becomes progressively restricted [1]. For example, the descendants of NM stem cells, which originate in the epiblast adjacent to the PS, become restricted to the multipotent neuroectoderm and paraxial mesoderm cell lineages upon differentiation [2][3][4][5][6]. The response of NM stem cells to signals secreted by the PS, such as Wnt3a, determines whether they self-renew or differentiate into these lineages, however the mechanisms that govern NM cell fate decisions are not well understood.
Wnt3a is a member of the 'Wnt' family of secreted glycoproteins that interact with Frizzled/Lrp receptor complexes to prevent the proteolytic degradation of the transcriptional co-activator bcatenin. In the presence of Wnt3a, stabilized b-catenin enters the nucleus where it interacts with members of the Tcf/Lef family of DNA binding transcription factors. Together, b-catenin and Tcf/Lef form the core of a multipartite transcription complex that controls target gene expression [7]. Embryos lacking Wnt3a die at mid-gestation with severe posterior axial truncations that reflect a loss of neuro-musculoskeletal progenitors [4,[8][9][10][11][12]. Cells expressing the stem cell marker and T-box transcription factor, T, including NM stem cells, are transiently established in Wnt3a 2/2 mutants but are not maintained [10]. Interestingly, Wnt3a 2/2 mutants form ectopic neural tubes that are speculated to arise from the transfating of paraxial mesoderm progenitors in the absence of a Wnt signal [10,12]. Similar defects are also observed in Tcf1 2/ 2 ;Lef1 2/2 embryos [13]. Together, these findings strongly suggest that Wnt3a/b-catenin signaling regulates the maintenance and differentiation of NM stem cells. A thorough understanding of the target genes and the gene regulatory network controlled by Wnt3a will provide crucial insights into how the fundamental neural and paraxial mesoderm lineages are established.
Here we report the comprehensive analysis of the Wnt3adependent transcriptome in PS and node stem cells. In silico analysis demonstrates that Wnt3a regulates the expression of many genes associated with the primitive streak, neuroectoderm and paraxial mesoderm. Our analysis identified 2 candidate target genes, Sp5 and Sp8, that when genetically disrupted recapitulate the Wnt3a null phenotype. Thus, Sp5 and Sp8 are redundant factors that function as key effectors of the Wnt3a/b-catenin pathway in NM stem cells.

Materials and Methods
Mice Ctnnb1 tm2Kem , Ctnnb1 lox(ex3) , Wnt3a, T-cre, Sp5 lacZ/lacZ and Sp8 D/+ mice were described previously [10,[14][15][16][17][18]. T-cre mice were used to conditionally delete Ctnnb1 tm2Kem or Ctnnb1 lox(ex3) in NM stem cells and their progeny. Sp5;Sp8 dko embryos were obtained by intercrossing Sp5 lacZ/lacZ ;Sp8 D/+ mice. This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol was approved by the Frederick National Lab Animal Care and Use Committee (Animal Study Proposal #12-408). All efforts were made to minimize suffering. Rodents were euthanized by CO2 inhalation in accordance with the most recent AVMA Guidelines on Euthanasia.

Gene Expression Profiling
Total RNA was isolated from dissected node and PS regions of E7.75-E8 wildtype and Wnt3a mutant embryos as previously described [15]. cDNA and cRNA synthesis, hybridization to GeneChipH Mouse Genome 430 2.0 Arrays, washing, staining, and array scanning were carried out according to manufacturer's protocols (Affymetrix, Santa Clara, CA) as previously described [15,19]. Briefly, 30-50 ng of total RNA from individual embryos was used as template for first cycle cDNA synthesis. For second cycle cDNA synthesis, equivalent amounts of cRNA from three stage-matched embryos per genotype were pooled (600 ng total) to generate labeled-target per array. Array hybridizations were performed in triplicate per genotype.
Statistical analysis was performed on probe-intensity level data (CEL files) using BRB ArrayTools version 4.3.0 or GeneSpringGX software, with similar results. Data presented herein was generated using BRB ArrayTools. Genes were considered statistically significant if their parametric p-value #0.05. For hierarchical clustering of genes in Fig. 1F, centered correlation and average linkage method was used for the maximally expressed probe set measured by average intensity across all arrays. Signal transduction and tissue pathway analysis was performed using Genomatix suite (GePS) software. The microarray data in this study was previously deposited in the GEO database under accession code GSE29995 [19].

In situ Hybridization, Immunofluorescence and Tissue Sectioning
Brachyury, Wnt3a and Uncx4.1 probe synthesis and whole-mount in situ hybridizations (WISH) were performed as previously described [15]. The mouse Sp5 cDNA (cat# EMM1002-5885071) was obtained from Open Biosystems. The mouse Sp8 DNA template was obtained by TA-TOPO cloning (Invitrogen) the Sp8 coding region from ES cell cDNA. Paraffin and cryosectioning was performed at the Pathology and Histology Laboratory (PHL, NCI-Frederick). Cryosections were washed in 16PBT (16PBS +0.1% tween-20), incubated in 16PBS +0.2% Triton-X for 20 minutes, washed in PBT. Sections were incubated in primary antibodies in 0.02% block (Roche Blocking reagent in PBT, filtered) overnight at 4uC. Sections were then washed 3 times in PBT. Secondary antibodies in 0.02% block were applied at room temperature for 2 hours. Sections were washed 3 times, 1 nM DAPI (49,6-diamidino-2-phenylindole) was included in the second wash for 20 minutes. Embryos/sections were photographed on a Leica MZFLIII stereoscope equipped with a Zeiss Axiovision HRc digital camera, and Zeiss Axiovision imaging software (v4.5). Unless indicated otherwise, at least 3 mutant embryos were examined with each probe/antibody, and all yielded similar results.

ES Cell Culture and mRNA Expression Analysis
T-GFP ES cells were maintained on gelatin-coated plates and differentiated as embyoid bodies (EBs) in serum-free, feeder-free conditions, as previously described [20]. Day 2 EBs were dissociated with Trypsin-EDTA and 4610 5 cells were cultured in ultra-low attachment 6-well dishes with 100 ng/ml recombinant Wnt3a (R&D Systems) or 1 mM CHIR99021 (ToCris) for 2 days.
Total RNA was extracted using the RNeasy miniprep kit (Qiagen) and first stand cDNA synthesized using iScript cDNA synthesis kit (Quanta Biosciences). Relative gene expression was analyzed with iTaq universal SYBR Green (BioRad) using the CFX96 Real-Time PCR Detection system (BioRad). Relative fold changes were calculated by normalizing to Gapdh levels using the 2 2DDCt method. qPCR primers are listed in Table S4.

Transcriptional Profiling of PS and Node Stem Cells
Wnt3a is highly expressed in the node and primitive streak during gastrulation, where it signals via b-catenin and Tcf1/Lef1 ( Fig. 1A; [15]) to regulate genes essential for neuromesodermal development. To identify the effectors of this pathway, we used microarrays to generate transcriptional profiles of wild type and Wnt3a 2/2 primitive streak and node tissue. To enrich for early molecular regulators, tissues were excised between E7.75-8.0 when Wnt3a 2/2 mutants are morphologically indistinguishable from wild type controls. Comparisons of wild type and mutant transcriptional profiles identified 729 differentially expressed genes (DEGs, p,0.05) of which 379 were down-regulated and 350 were up-regulated ( Fig. 1B, Table S1 and Figure S1). Included amongst the down-regulated genes are several well-known targets that encode components of the Wnt/b-catenin signaling pathway itself, ex. Axin2, Nkd1, and Lef1 (Fig. 1B, Table S1), thereby providing an initial validation of the screen [21][22][23]. An assessment of DEGs for their association with signaling pathways revealed that the Wnt/bcatenin pathway was most highly represented followed closely by the Notch, Hedgehog, and Fgf pathways (Fig. 1C, Table S2). The DEG data was mined further for associations with literature-based gene regulatory networks using Genomatix Pathway Systems software. This approach identified an interacting network of Wnt/ b-catenin pathway genes that function, in part, to control the expression of Axin2 (Fig. 1D, Fig. S2). Included in this network of Wnt target genes are the transcription factors Msgn1 and Tbx6 that play critical roles in paraxial mesoderm and neural development [19,24,25]. Interestingly, two members of the large and wellcharacterized Sp1 family of DNA-binding Zn 2+ -finger transcription factors, Sp5 and Sp8, were also identified in this network.
We then asked if these DEGs are expressed in the appropriate embryonic tissues to be targets of Wnt3a. Detailed analysis of the DEGs using their associated gene ontology (GO) terms revealed that, as expected, their expression is most commonly associated with stem cells, the PS, neural plate, and mesoderm (Fig. 1E, Table S3). Analysis of genes clustered by their tissue-specific expression patterns revealed that primitive streak and posterior PSM genes were down regulated in Wnt3a 2/2 embryos, while anterior PSM genes were increased (Fig. 1F). These results are consistent with Wnt3a maintaining immature posterior PSM progenitors, while suppressing segmentation genes in the anterior PSM [19,26]. Interestingly, intermediate and lateral plate mesoderm genes showed increased expression levels in the mutants suggesting Wnt3a has a mesodermal patterning role in the PS. It is also notable that anterior neural plate markers increased, but posterior neural markers decreased in the Wnt3a mutants, likely reflecting the established neural posteriorizing role for Wnt signaling [27,28].

Validation of Candidate Wnt3a Target Genes
We focused on the validation of putative Wnt3a target genes that encode transcription factors, as they frequently function as master regulators of stem cell fate. Microarray analysis showed that the Sp5, Sp8, Klf5, Mixl1, and Lmo2 transcription factors were Sp5/8 Act Downstream of Wnt3a/b-catenin Signaling PLOS ONE | www.plosone.org all down-regulated in Wnt3a mutants, and were thus selected for further expression analysis in vivo, along with the secreted Wnt feedback inhibitor Notum. We reasoned that candidate Wnt3a target genes should be spatially and temporally co-expressed with Wnt3a and be responsive to changes in Wnt3a/b-catenin signaling. WISH analysis demonstrated that Sp5, Sp8, Notum, Klf5, and Mixl1 were co-expressed with Wnt3a (Fig. 1a, left) in the wild type node/ primitive steak region at E7.5-8.5 ( Fig. 2A, B, F, G, K, L, P, Q, U, V). Lmo2 was undetectable in the PS at E7.5, but was easily detected at E8.5 (Fig. 2Z, AA). The lack of Lmo2 expression at E7.5, when Wnt3a is easily detected in the PS, suggests that Lmo2 requires additional signals for activation. Reduced Wnt3a/bcatenin activity led to significantly diminished expression of all 6 genes in the PS and posterior mesoderm domains of Wnt3a 2/2 embryos (Fig. 2C, H Sp5 and Sp8 are Wnt3a-responsive and Expressed in the PS Sp5 and Sp8 were identified by microarray as two of the more prominently down regulated genes in Wnt3a mutants. The lack of a phenotype in Sp5 null mice, taken together with the observation that Sp1, Sp5, and Sp8 all bind to the same DNA sequence in vitro [17,29,30], suggests that Sp factors could function redundantly during development. To determine whether additional members of the Sp family are expressed in the PS and are Wnt3aresponsive, we turned to an in vitro mouse embryonic stem cell (ESC) model of PS formation [20]. ESCs were cultured in defined media to allow for the stimulation of the Wnt/bcatenin signaling pathway with purified, recombinant Wnt3a protein in feeder-free and serum-free conditions. ESCs were primed for germ layer differentiation by withdrawing LIF and forming embryoid bodies (EBs) for 2 days. EBs were then treated with Wnt3a or the GSK-3b inhibitor CHIR99021 for an additional 2 days to promote PS and mesoderm gene expression [20,31,32]. Quantitative PCR of EBs revealed that of the 9 Sp genes (Sp1-9), only Sp5 and Sp8 were expressed during the in vitro differentiation process (Fig. 3A, D4 EB untreated). These expression levels were dramatically augmented by treatment with Wnt3a protein or CHIR99021 (Fig. 3A) demonstrating that Sp5 and Sp8 were induced during PS formation and were highly Wnt3a-responsive.
To examine whether Sp5 and Sp8 are expressed in the PS in vivo, a detailed comparative analysis was performed on serial sections of E8.5 embryos processed for WISH (Fig. 3B, C). Section analysis shows definitively that Sp5 and Sp8 are both expressed in the PS and across all three germ layers. In addition to the expression in the PS ectoderm and mesoderm, Sp5 is highly expressed in the dorsal hindgut (Fig. 3D). Sp8 is weakly expressed in the dorsal hindgut but is highly expressed in the PS epiblast/caudal neural plate (Fig. 3E). Together, these results demonstrate that Sp5 and Sp8 respond to Wnt3a/b-catenin activity, and are co-expressed in the PS, including the NM stem cells in the caudal lateral epiblast.

Sp5 and Sp8 Function in the Wnt3a Genetic Pathway
Our bioinformatics analysis and ESC experiments suggest that Sp5 and Sp8 could participate in Wnt signaling. To address this hypothesis we first asked if Sp5 genetically interacts with Wnt3a. Although Sp5 LacZ/LacZ null mice are viable and fertile, both homozygotes and heterozygotes develop ''kinked'' tails at a frequency of ,2-3% (Fig. 4E-F and Fig. S3). This phenotype is observed at a similar frequency in Wnt3a heterozygotes (7.7%, Fig. 4F) and is indicative of defective somite formation leading to fused vertebrae [33]. The frequency of this tail defect rises to 17.5% in Sp5 LacZ/+ ;Wnt3a +/2 mice and 41.5% in Sp5 LacZ/ LacZ ;Wnt3a +/2 animals demonstrating an Sp5 gene dosage-dependent genetic interaction with Wnt3a during tail development (Fig. 4E-F). Similar genetic interactions have been reported for Sp5 and the Wnt3a target gene T [11,17]. Furthermore, Sp5 LacZ/  [15]. Dotted lines demarcate dissection boundaries. B) Volcano plot depicting all statistically significant DEGs (p,0.05, indicated by blue color). C-E) Gene ontology (GO) analysis of DEGs identified a highly significant association with the Wnt/b-catenin pathway (C, D) and with expression in embryonic mesoderm, somites and neural tissue (E). D) A selection of the shortest path network representation of Wnt pathway genes in (C) analyzed in Genomatix GePS. For complete network see Figure S2. F) Analysis of tissue-specific DEGs shows that posteriorly restricted genes expressed in the node/PS, PSM, and NP are down-regulated while anterior PSM, anterior NP, IM, and LPM genes were up-regulated. Green: down-regulated genes; Red: up-regulated genes. doi:10.1371/journal.pone.0087018.g001  ;Wnt3a 2/2 double mutants display a loss of caudal somites and tailbud that is slightly more severe than Wnt3a 2/2 mutants alone (Fig. 4A-D). These results support a role for Sp5 in segmentation and the maintenance of tail progenitors and suggest that Sp5 functions in a genetic network with Wnt3a and T.
Sp8 null embryos are tailless and display spina bifida, reflecting a failure to form caudal somites and maintain a tailbud, and an eversion of the posterior neural folds, respectively [14,34]. Interestingly, the Sp8 D/D axis truncation phenotype bears a striking resemblance to the tail and spina bifida phenotypes observed in the Wnt3a hypomorph, vestigial tail (vt) [33]. The similarities between the hypomorphic Wnt3a and Sp8 mutant phenotypes, together with the demonstration that Wnt3a and Sp5 function in the same genetic pathway and that Sp5 and Sp8 are coexpressed in the PS, strongly suggests that Sp5 and Sp8 could function redundantly in the Wnt3a pathway to regulate NM stem cells. To test this hypothesis, we generated Sp5 LacZ/LacZ ;Sp8 D/D double knockout (hereafter referred to as Sp5;Sp8 dko) embryos. At E9.5, Sp5 LacZ/LacZ and Sp8 D/D single mutants appear similar to wild type controls (Fig. 5A-C). Although Sp5 LacZ/LacZ ;Sp8 D/+ mice are viable and fertile, 33.3% of adults develop ''kinked'' and/or shortened tails (Fig. S3). In contrast, tail defects were observed in only ,2.2% of Sp8 D/+ mice, suggesting a requirement for both Sp5 and Sp8 in tail segmentation and extension. Analysis of Sp5;Sp8 dko revealed embryonic lethality by ,E13.5-14.5. Remarkably, Sp5;Sp8 dko embryos displayed a severe posterior mesoderm deficit at E9.5 (Fig. 5D) that closely resembled the phenotype observed in Wnt3a 2/2 mutants (Fig. 5E). Although Sp5;Sp8 dko formed ,12 somites, compared to the 7-10 somites that typically develop in Wnt3a 2/2 mutants, the T + NM stem cell population is completely depleted in Sp5;Sp8 dko by E9.5, just like in Wnt3a 2/2 mutants (Fig. 5F-H compared to Fig. 4C). It is notable that the open anterior neuropore that was present in all Sp8 nulls (n = 28, Fig. 5C arrowhead) was exacerbated in 17/31 double mutants (arrow, Fig. 5D). These results show that Sp5 and Sp8 are essential for NM cell development and that they may have additional redundant roles in anterior neural development.
In addition to being required for the maintenance of NM stem cells, Wnt3a plays a critical role in directing NM stem cells towards a paraxial mesoderm cell fate. In the absence of Wnt3a activity, presumptive paraxial mesoderm cells are thought to adopt neural fates and form ectopic neural tubes posterior to the last formed somite [10,12]. To determine if Sp5;Sp8 dko mutants also display ectopic neural tissue, we analyzed cross sections of mutant embryos for the expression of the neural marker Sox2. Indeed, a supernumerary Sox2 + neural tube lying ventral to the primary neural tube was observed in Sp5;Sp8 dko similar to Wnt3a 2/ 2 mutants (Fig. 5M-P, Q-T). Fibronectin staining was used to highlight embryonic structure and clearly shows that embryos that display ectopic neural tubes also lack somites (arrowheads, Fig. 5K, O, S). Thus the severe posterior abnormalities resulting from changes in neural and mesodermal fates are nearly identical in both Sp5;Sp8 and Wnt3a mutants. These data show that like Wnt3a, Sp5 and Sp8 play a major role in determining NM stem cell fate.

Discussion
We have characterized the Wnt3a transcriptome in the node and primitive streak (PS) at a developmental stage when the mammalian embryo is undergoing gastrulation, segmentation and axis extension. Our genomic data strongly supports a role for Wnt3a as a major regulator of PS stem cells, including NM stem cells. Many PS-specific genes identified in the microarray were down-regulated implicating Wnt3a as an important regulator of PS fate. The most significant sets of differentially expressed genes were associated with neural or paraxial mesoderm cell identity lending strong support to our hypothesis that Wnt3a directly regulates NM progenitors. Interestingly, neural-specific gene expression was marked by increased anterior and decreased posterior neural gene expression. In parallel, posterior paraxial PSM markers were also down-regulated while the anterior PSM markers, Ripply2 and Mesp2 increased. These findings suggest that Wnt3a suppresses anterior, and activates posterior, gene expression in both the neural plate and the PSM, consistent with a proposed role for Wnt/b-catenin signaling as the major evolutionarily conserved posteriorizing signal during the establishment and maintenance of axial AP polarity [35].
Wnt/b-catenin signaling plays a well-established role in the regulation of embryonic and adult stem cell homeostasis [36]. Somewhat paradoxically, Wnt/b-catenin signaling can promote both the maintenance of pluripotent ESCs as well as their differentiation into germ layer progenitors [32,36]. Although our studies did not focus on pluripotency, Klf5, which encodes a transcription factor redundantly required for ESC pluripotency [37], was down-regulated in Wnt3a mutants suggesting that Klf5 might regulate PS stem cell potency. This screen also led to the identification of the bHLH transcription factor, Msgn1, as a direct target gene of Wnt3a [19]. Msgn1 is a major activator of the  oscillating segmentation clock -an activity that is a defining feature of PSM cells -suggesting that Msgn1 may be the key regulator of PSM differentiation.
Sp5 and Sp8 are expressed in numerous tissues throughout mouse development, [17,38] but are restricted to important signaling centers where Wnts are active and where stem cells reside, such as the primitive streak. The orthologous genes Sp5l and Sp5 have been implicated as important regulators of posterior development in zebrafish [39,40], but a role for mammalian Sp5 has not been clearly established. Although Sp5 is not required for mammalian embryonic development, an Sp5 null allele enhanced the short or kinked tail phenotypes in T [17] and Wnt3a heterozygotes suggesting it may indeed function in NM progenitors. The case for a role for Sp8 in NM progenitors is stronger given that genetic ablation of Sp8 resulted in the loss of all caudal (tail) somites and led to the development of neural anomalies [10,14]. We have now shown, through the generation of Sp5;Sp8 double mutants, that Sp5 and Sp8 have essential, overlapping roles in trunk and tail development. With the exception of the anterior exencephaly defects, the Sp5;Sp8 dko phenotype is nearly identical to the Wnt3a 2/2 phenotype. Double mutants lack posterior paraxial mesoderm and display ectopic neural tubes, strongly suggesting that Sp5 and Sp8 function to regulate the fate of NM stem cells.
We have shown that Wnt3a/b-catenin signaling is both necessary and sufficient for Sp5 and Sp8 expression in the PS, PSM, and in ESCs. The high degree of similarity between Wnt3a, conditional b-catenin, Tcf1;Lef1, and Sp5;Sp8 mutants suggests that Sp5/8 play equally important roles as b-catenin and Tcf1/Lef1 to transduce Wnt3a signals. This raises the intriguing question as to whether Sp5/8 function as terminal effectors of a Wnt3a-initiated transcriptional cascade, or as novel, integral components of the bcatenin/Lef1 transcription complex. Alternatively, Sp5/Sp8 could function in a positive feedback loop to maintain Wnt3a expression. Future studies will address the precise molecular mechanisms of Sp5 and Sp8 activity in stem cells of the early embryo.