BMP and WNT signalling cooperate through LEF1 in the neuronal specification of adult hippocampal neural stem and progenitor cells

Neuronal production from neural stem cells persists during adulthood in the subgranular zone of the hippocampal dentate gyrus. Extracellular signals provided by the hippocampal microenvironment regulate the neuronal fate commitment of the stem cell progeny. To date, the identity of those signals and their crosstalk has been only partially resolved. Here we show that adult rat hippocampal neural stem and progenitor cells (AH-NSPCs) express receptors for bone morphogenetic proteins (BMPs) and that the BMP/P-Smad pathway is active in AH-NSPCs undergoing differentiation towards the neuronal lineage. In vitro, exposure to the BMP2 and BMP4 ligands is sufficient to increase neurogenesis from AH-NSPCs in a WNT dependent manner while decreasing oligodendrogenesis. Moreover, BMP2/4 and WNT3A, a key regulator of adult hippocampal neurogenesis, cooperate to further enhance neuronal production. Our data point to a mechanistic convergence of the BMP and WNT pathways at the level of the T-cell factor/lymphoid enhancer factor gene Lef1. Altogether, we provide evidence that BMP signalling is an important regulator for the neuronal fate specification of AH-NSPCs cultures and we show that it significantly cooperates with the previously described master regulator of adult hippocampal neurogenesis, the WNT signalling pathway.

Immunostaining. Cultured cells were fixed with 2% paraformaldehyde (Panreac). Samples were incu- Gene Expression analysis. RNA was extracted from cells or 2-month old wild type (CD-1) mice using Illustra RNAspin Mini Kit (GE Healthcare) and measured with a Microplate Reader (Infinite M200 TECAN). cDNA was obtained by reverse-transcription (RT) employing PrimeScrip RT Reagent kit (Takara). Gene expression was determined by quantitative polymerase chain reaction (qPCR) in a LightCycler480 (Roche) using SYBR PremixEX Taq (2×) (Takara) and the corresponding forward and reverse primer for each gene, following the manufacturer protocol. Data were analysed according to the 2 −ΔΔCt method 24 . Primer sequences are available upon request.
Statistical Analysis. The statistical significance of the difference between means for the kinetics experiments was assessed by one-way ANOVA, using the Tukey test as post-hoc comparison. To determine the significance of the BMP2/4 and WNT3A synergic effect we used two-way ANOVA analysis of the percentage of neurons. The significance between means of the remaining experiments was calculated using paired 2-tailed Student t test. All the values correspond to average ± sem, and those with a P value < 0.05 were considered significant ( * P < 0.05; ** P < 0.01; *** P < 0.001). Arcsen transformation was carried out for the statistical analysis of percentages.

Results
BMP ligands and BMP receptors are expressed in the adult hippocampal dentate gyrus and are differentially regulated during AH-NSPC differentiation. Based on sequence similarity and function, BMP ligands can be subdivided into several subgroups that share a number of BMP type 1 and type 2 receptors (Fig. 1A). To gain a deeper understanding on their role in adult neurogenesis, we first examined their expression in the adult mouse hippocampal dentate gyrus (DG) and in cultured adult hippocampal neural stem and progenitor cells (AH-NSPCs) from rat by quantitative RT-PCR. On the basis of the C t (cycle threshold) values, we found that most BMP ligands and receptors are expressed at the mRNA level in the adult DG ( Supplementary Fig. S1).
We next evaluated the timing of BMP receptor expression in AH-NSPC cultures undergoing differentiation. AH-NSPCs were induced to differentiate for up to 14 days in vitro (DIV) using a combination of Retinoic Acid (RA) and Forskolin (FSK), a well-established treatment that facilitates the acquisition of the neuronal fate ( Fig. 1B,C; Hsieh et al., 2004). After RNA extraction and cDNA synthesis, the gene expression pattern of the BMP type 1 receptors (Bmpr1a, Bmpr1b and Acvr1) and the BMP type 2 receptors (Bmpr2, Acvr2a and Acvr2b) was determined by quantitative RT-PCR and referred to the housekeeping genes Sdha (Fig. 1) and 18 S (Supplementary Fig. S2). Increased expression of the neuronal cytoskeleton gene βIII-tubulin (Tubb3) and decreased expression of the immature intermediate filament gene Nestin (Nes) were used to confirm the differentiation of the cultures (Fig. 1D,E). Our results showed a decrease in the expression of Bmpr1a and an increase in the expression of Bmpr1b and Acvr1 as the cells differentiate ( Fig. 1F-H). Expression of Bmpr2 at the mRNA level remained relatively constant during the process whereas Acvr2a and Acvr2b expression progressively decreased ( Fig. 1I-K). This regulated gene expression profile of the type 1 and type 2 receptors suggests an early role for BMPR1A, BMPR2, ACVR2A, ACVR2B and a late role for BMPR1B, ACVR1, BMPR2 during adult hippocampal neurogenesis. BMP2 and BMP4 are pro-neurogenic in AH-NSPCs. Since AH-NSPCs express different receptor combinations throughout differentiation, we next analysed whether BMP ligands with different affinities for those receptors influence neurogenesis. To this end, we cultured AH-NSPCs in the absence of mitogenic stimulation but in the presence of an increasing concentration of several prototypic BMP ligands: BMP2 and BMP4 recombinant proteins, belonging to the Dpp subgroup that preferentially bind to BMPR1A; and BMP7 recombinant protein, belonging to the 60 A family that is best bound by BMPR1B and ACVR1 14,15 . After 4 DIV, neurogenesis was measured by immunostaining against the neuronal marker βIII-tubulin (TuJ1). Our data show that BMP2 and BMP4 induce a marked dose-dependent increase in the percentage of TuJ1 + neurons, while BMP7 has a very mild effect ( Fig. 2A,B). BMP2/4 also increased the number of neurons expressing the immature neuronal marker doublecortin (DCX) and the more mature neuronal marker MAP2 (25 ng/ml BMP4: 6.7 ± 1.1% MAP2 + cells, avergage ± sem, n = 3) compared to the control condition (N2 medium: 0.4 ± 0.2% MAP2 + cells, avergage ± sem, n = 3, P < 0.01, two-tailed T-test; Supplementary Fig. S3). Western blot analysis showed increased protein levels of βIII-tubulin and doublecortin (DCX) after BMP2/4 exposure ( Supplementary Fig. S4), indicating altogether that ligands of the BMPR1A receptor such as BMP2/4 may be involved in instructing a neuronal cell fate in vitro. In support of this view, transduction of AH-NSPCs with a retroviral vector overexpressing a constitutive active form of BMPR1A mimicked the pro-neurogenic effect triggered by BMP2/4 (Fig. 2C). In addition, we found a reduction in oligodendrocytes in cultures treated with BMP2/4 but no change in the number of astrocytes ( Fig. 2D and Supplementary Fig. S3), suggesting that BMP2/4 induce neurogenesis while decreasing oligodendrogenesis. Of note, a large fraction of the cells in the culture remained undifferentiated at 4 DIV (28.4 ± 2.4% of the cells were SOX2 + stem/progenitors in 25 ng/ml BMP4 vs. 62.6 ± 5.2% in control N2 medium, avergage ± sem, n = 3, P < 0.01, two-tailed T-test). Out of the SOX2 + cells, only 16.4 ± 3.7% were cycling (Ki67 + ) in the presence of BMP4, suggesting that most of the AH-NSPCs that did not engage in the differentiation programme stayed quiescent 19 .
Since Bmpr1a is markedly expressed in undifferentiated AH-NSPCs and decreases throughout the neuronal differentiation process 19 (Fig. 1F), we hypothesized that BMP2/4 would be acting very early in the neurogenic lineage, influencing the cell fate choice decision of the AH-NSPC progeny. To test this idea, we transiently exposed AH-NSPCs to BMP2 or BMP4 for 24 hours (1 DIV) and then cultured the cells for 3 additional days in the absence of BMP stimulation (Fig. 3A). We found that AH-NSPCs exposed to BMP2 or BMP4 for 1 DIV reached similar neuronal differentiation levels than AH-NSPCs cultured for 4 DIV in the continuous presence of BMPs ( Fig. 3B-D). These data show that a brief BMP2/4 treatment during the initial stages of differentiation is sufficient to promote neurogenesis and indicate that BMP2/4 can specify the neuronal fate of AH-NSPCs within the first 24 hours of the differentiation time course.
BMP2 and BMP4 induce neurogenesis through the activation of the BMP canonical pathway. We next wanted to check whether BMP2/4 signal through the SMAD-dependent canonical pathway.
AH-NSPCs were treated with BMP2 or BMP4 and were then analysed at different time points by immunofluorescence, Western blot and quantitative RT-PCR. As shown in Fig. 4, BMP2 and BMP4 rapidly triggered the phosphorylation and nuclear translocation of SMAD1/5/8 ( Fig. 4A-C) and upregulated the expression of the SMAD target gene Id1 ( Fig. 4D and Supplementary Fig. S3) confirming the activation of the canonical pathway. In agreement with this finding, inhibition of non-canonical BMP signalling employing the P38MAPK inhibitor SB203580 had no effect on the increase in neurogenesis induced by BMP2/4 exposure (Fig. 4E). Phosphorylation of SMAD1/5/8 was transient and decreased after the initial 24 hours of stimulation with the BMP ligands, indicating that AH-NSPCs did not maintain sustained activation of the BMP canonical pathway during the 4 DIV differentiation period ( Fig. 4C and Supplementary Fig. S6). This further reinforces the view that the signalling inducing the neuronal fate occurs in the first 24 hours (1 DIV) of the differentiation process (Fig. 3). Together, the data indicate that BMP2 and BMP4 enhance neuronal production through the transient activation of the P-SMAD1/5/8-dependent canonical signalling pathway. BMP2/4 canonical signalling synergizes with WNT canonical signalling to increase neurogenesis. It has been previously reported that adult hippocampal neurogenesis is tightly regulated by the canonical WNT/β-catenin signalling pathway, both in vitro and in vivo [reviewed in 25 ], and that WNT3/WNT3A ligands markedly enhance the neuronal differentiation of AH-NSPCs 10,26 . Given the phenotypic similarity resulting from the activation of the BMP and WNT pathways, we searched for a possible crosstalk between the two.
To this end, we first exposed AH-NSPCs to increasing amounts of the WNT3A ligand and confirmed that recombinant WNT3A increased the number of neurons at 4DIV in a dose-dependent manner (Fig. 5A,B). Interestingly, as with BMP2/4, the acquisition of the neuronal fate in the presence of WNT3A was also accompanied by a reduction in the oligodendroglial fate (Fig. 5C). In addition, we confirmed that WNT3A triggered the canonical pathway in AH-NSPCs, since it increased the phosphorylation of LRP6 (Fig. 5D) and the expression of the Axin2 gene (Fig. 5E).
Having verified the pro-neurogenic effect of WNT3A, we next exposed AH-NSPCs to low doses of the BMP2/4 and WNT3A ligands separately or in combination to test for a possible interaction between the two pathways ( Fig. 6A-C). Statistical analysis by 2-way ANOVA revealed a significant synergistic effect in neurogenesis of the BMP and WNT factors (F 1,28 = 14.642, P < 0.001). At BMP2 and WNT3A concentrations of 25 ng/ml, a marked raise in TuJ1 + cells was elicited with a 19-fold increase in the number of neurons when both ligands were present, relative to the 5-and 3.5-fold increases elicited by BMP2 or WNT3A treatments alone, respectively (Fig. 6A). A similar profile was obtained for BMP4 and WNT3A (F 1,30 = 5.851, P < 0.05, Fig. 6B).
The pro-neurogenic BMP2/4 activity requires endogenous canonical WNT signalling and is linked to LEF1 expression. Given that WNT signalling is absolutely required for neurogenesis to proceed   in the adult hippocampus 25 , we speculated that, not only the BMP and WNT pathways synergize when both ligands are added exogenously to the AH-NSPC culture, but that BMP signalling requires a basal level of WNT signalling to exert its pro-neurogenic effect. In other words, we hypothesized that the BMP treatment alone could be potentiating endogenous WNT signalling. To test for this hypothesis, we differentiated AH-NSPCs in the presence of BMP2/4 (50 ng/ml) concomitant with increasing concentrations of the canonical WNT signalling inhibitor XAV939 27 to prevent endogenous canonical WNT signalling (Fig. 6D). Blockade of the signalling elicited by exogenously added WNT3A (50 ng/ml) was used as a control for the XAV939 treatment. Interestingly, we found that endogenous WNT signalling is required at least in part for the neurogenic action of the BMP2/4 ligands.
Canonical WNT signalling triggers the expression of pro-neurogenic genes through T-cell factor/lymphoid enhancer factor (TCF/LEF)-binding sites in their promoters 26 . It has been reported that the expression of Lef1 can be activated by BMP4 in BMP-mediated inductive tissue interactions 28 , raising the possibility that the convergence between BMP2/4 and WNT3A signalling is located nearby LEF1. Thus, we next explored whether BMPs could increase Lef1 expression in AH-NSPCs. As shown in Fig. 7A, following a 6-hour or 24-hour treatment with BMP4, Lef1 mRNA levels were increased. This resulted in the accumulation of LEF1 protein (Fig. 7B).
We next analysed the Lef1 gene promoter region to check for the presence of conserved BMP regulatory elements. We searched the rat genome databases, retrieved the 5 kb sequence upstream of the initial Lef1 ATG start codon located in Chr2 and performed a theoretical analysis of the sequence using promoter inspector tools. We found a validated TATA-less promoter sequence spanning from −501 to +100 relative to the translational start site ATG codon at +1 (601 bases in length, P1) and a second theoretical promoter region (684 bases in length, P2) within the first intron of the Lef1 gene (Fig. 7C). The structure of the rat Lef1 gene resembles that of the human LEF1 gene, a multipromoter gene characterized by a first 5′ promoter (P1) lacking a consensus TATA box and a second intronic promoter (P2) located downstream of exon 1 29,30 . In human, the P1 promoter drives the expression of a full-length LEF1 polypeptide; the P2 promoter instead produces a truncated protein that lacks the beta-catenin binding domain and supresses WNT signalling 30 . From these two promoter sequences, we selected the upstream −501/+100 P1 promoter for further analysis and scanned the sequence for transcription factor binding sites (TFBS). Out of the 154 total TFBS sites mapped to the proximal upstream promoter region, two corresponded to canonical SMAD protein binding sites. The SMAD binding sites were located at position −302 ("5′-GTCT-3′ SBE" or SMAD4 binding element) and −261 ("GCCG-like motif " or GC-rich SMAD1/5 binding element) from the start codon, and were recognized as putative binding sites for P-SMAD/SMAD4 dimers (Fig. 7D,E). To investigate the direct interaction between SMAD proteins and the Lef1 promoter, we performed a Chromatin Immunoprecipitation (ChIP) assay with anti-SMAD4 antibodies, followed by PCR analysis with primer pairs amplifying the region containing the putative SMAD binding sites (Fig. 7D). As shown in Fig. 7F, the ChIP-PCR results revealed the presence of SMAD4 on the Lef1 gene promoter region comprised between −361 and −625 nucleotides in AH-NSPCs exposed to BMP4. Thus, our results show that in response to BMP stimulation SMAD4 associates with the DNA at the Lef1 proximal promoter region, which includes the predicted SMAD binding elements. The binding correlates with an increase in Lef1 mRNA levels (Fig. 7A). Finally, we found that transduction of AH-NSPCs with a retroviral vector overexpressing Lef1 is sufficient to enhance neuronal production even in the absence of exogenous WNT stimulation (Fig. 7G), thereby mimicking the BMP2/4 effect.
Altogether, the data indicate that Lef1 is a direct target gene of SMAD4 and suggest that the increase in the expression of the transcription factor LEF1 downstream of BMP signalling may be potentiating adult hippocampal neurogenesis.

Discussion
It has been previously reported that neurogenesis in the adult mammalian brain is dynamically regulated by a number of local signals from the neural stem cell microenvironment, including BMPs and WNTs. High levels of canonical BMP signalling have been detected in the hippocampal niche, the majority of the stem and progenitor cells showing phosphorylated SMAD1/5/8 proteins in the nucleus 19,20 . Moreover, ligands from the BMP family  19,20 and in the control of progenitor maturation at multiple stages along the neurogenic lineage 18 . However, the contribution of the BMP pathway to the fate specification of the hippocampal neural stem cell progeny during adulthood, in concert with other key niche signals, has remained largely unexplored. The in vitro data included herein give further insight into the extensive role of BMP molecules as regulators of stem cell differentiation in the adult hippocampus, and support a model whereby BMPs would have WNT-dependent instructive effects, favouring the acquisition of the neuronal fate possibly at the expense of the oligodendroglial fate (Fig. 8). This lines up with the early pro-neurogenic role of BMP signalling during the neurogenic phases of forebrain development 31,32 . BMP signalling has been previously shown to promote an astrocytic fate in embryonic subventricular zone multipotent NSPC mouse cultures grown in EGF 33,34 . In our in vitro setup, however, BMP2/4 did not change the astroglial fate of adult hippocampal NSPCs isolated from rat and expanded in FGF2. This inconsistency may be due to a difference in the origin of the cells, in the cell culture conditions or perhaps in their basal level of WNT signalling. Interestingly, in vivo it has been shown that increasing BMP4 levels in the adult mouse hippocampus, by means of injecting a lentivirus overexpressing BMP4, decreases neural stem/progenitor cell proliferation, delays maturation, but it does not increase the proportion of cells that differentiate into astroglia 18 . This suggests that under physiological conditions BMP may not be pro-gliogenic in the adult hippocampal niche. However, we cannot exclude pro-gliogenic effects of the BMP pathway upon sustained up-regulation of BMP signalling or under pathological conditions that lead to an increase in gliosis 33,35 .
As for the contribution of the specific subtypes of BMP receptors, the gene expression patterns obtained suggest that BMPR1A receptor has an early role in the differentiation process, while BMPR1B and ACVR1 may possibly have a late role. The early role of BMPR1A is consistent with the results obtained in the differentiation assays employing the preferred BMPR1A ligands (BMP2/4) and with the retroviral experiments in which we overexpressed a constitutive active form of BMPR1A. Moreover, we found that the differentiation levels attained under an initial and transient exposure to BMP2/4 were similar to those obtained with a continuous exposure to the ligands, further indicating that the signalling inducing the neuronal fate occurs in the first stages of the differentiation process and thus is likely transduced by the most highly expressed type 1 receptor in those initial stages, BMPR1A. Our results partly complement a previous study 36 , in which a dominant negative form of BMPR1A (dnBMPR1A) was overexpressed in AH-NSPCs. The dnBMPR1A, with a non-functional intracellular domain but a functional extracellular domain that traps the endogenously produced BMP ligands, increased astroglial differentiation of the cultures and this was likely mediated by the concomitant increase in the expression of BMPR1B in the cultures 36 .
To gain insight into potential mechanisms underlying the BMP pro-neurogenic action uncovered in our assay, we examined its dependency on WNT signalling, a master regulator of adult hippocampal neurogenesis 10 . We found that the BMP pro-neurogenic effect requires basal levels of endogenous WNT signalling and that the BMP and WNT pathway show a synergistic interaction. When combining the BMP and WNT ligands at non-saturating concentrations, the effect of the two taken together was greater than the sum of their separate effects at the same doses. In the light of these data, it seemed likely that the two signalling pathways would share components of the same mechanistic route.
The BMP and WNT pathways have been shown to interact with each other at multiple levels, for instance, via the binding of both SMAD4 and TCF/LEF transcription factors to the promoter and enhancer regions of certain target genes 37,38 . The TCF/LEF family of transcription factors mediate the downstream effects of canonical Wnt/β-catenin cascade and has four members in mammals, LEF1 being the most well characterized 39 . Given LEF1 is expressed in adult hippocampal neural stem cells and is required for neurogenesis 3,26,40 , and since the Schematic of our proposed model summarizing the main results. BMP2/4 induce neurogenesis through the activation of the P-SMAD canonical pathway downstream of the BMPR1A type 1 receptor. The pro-neurogenic effect of BMP signalling is partly dependent on endogenous WNT signalling, and the mechanism relies on the up-regulation of the Lef1 gene, a direct SMAD target. WNT-independent effects of the BMPs are not excluded (dashed line, red). BMP2/4 also decrease the number of oligodendrocytes. expression of Lef1 can be activated by BMP4 in BMP-mediated inductive tissue interactions 28 , we focused on the Lef1 gene as a possible convergence point between the BMP and the WNT routes. We identified functional BMP regulatory elements on the Lef1 promoter, we found that Lef1 expression was increased shortly after BMP exposure and we demonstrated that overexpressing Lef1 is sufficient to enhance neurogenesis, uncovering a novel mechanistic link between the BMP/WNT signalling pathways operating in adult hippocampal neural stem and progenitor cells. A similar mechanism could be active as well during the development of the hippocampal dentate gyrus, since it has been recently reported that Lef1 expression at embryonic stages is regulated by the BMP type 1 receptor ACVR1 41 .
In summary, signalling through the BMP type 1 receptor BMPR1A plays a relevant role in the modulation of adult hippocampal neurogenesis, regulating not only NSC quiescence [18][19][20] but possibly also the early stages of the differentiation cascade. At least in vitro, the pro-neurogenic canonical WNT signalling interplays with, and is subjected to regulation by, BMP signalling. This may be achieved by increasing the expression of the transcription factor LEF1. Although future in vivo studies are needed to further evaluate the effect of BMP signalling in fate specification, our in vitro model already shows that the role of BMPs/BMPRs during adult neurogenesis is much more complex than was previously thought.