Pbx1 is required for adult subventricular zone neurogenesis

The subventricular zone (SVZ) of the lateral ventricle walls is one of the neurogenic niches in the adult rodent brain capable of generating neurons in a non-pathological context. Multipotent neural stem cells of the SVZ produce transient amplifying progenitors (TAPs), which predominantly generate neuroblasts in vivo. Adult neural stem cells are subsets of SVZ astrocytes, defined by their characteristics of self-renewal and multipotency, while TAPs and neuroblasts can be easily distinguished on the basis of specific marker gene expression. TAPs, for instance, express the basic helix-loop-helix (bHLH) transcription factor achaete-scute family transcription factor 1 (Ascl1, or Mash1), whereas neuroblasts can be recognized by their expression of doublecortin (Dcx) and neuronal βIII-tubulin (Tubb3, or Tuj1) or production of polysialylated NCAM (PSA-NCAM). Neuroblasts migrate along the rostral migratory stream (RMS) into the olfactory bulb (OB), where they differentiate into a wide variety of inhibitory neuron phenotypes, including GABA-ergic granule cells (GCs) or dopaminergic, calbindin- or calretinin-expressing periglomerular interneurons (PGNs), and thus contribute to a continuous neuronal turnover in the adult OB (Lim and Alvarez-Buylla, 2014). Defective replacement of OB interneurons has been implicated in impaired olfaction-related behavior, whereas excessive proliferation in the SVZ has been linked to the formation of glial tumors (Sakamoto et al., 2014; Barami et al., 2009). Adult SVZ neurogenesis therefore needs to be tightly regulated in order to maintain a proper balance between self-renewal and differentiation towards neuronal or glial lineages.

Pre B-cell leukemia homeodomain (PBX) transcription factors constitute the PBC subgroup of TALE (three amino acid loop extension) homeodomain (HD)-containing proteins. PBX transcription factors are essential regulators of embryonic development. They contribute to the correct patterning of the anterior-posterior body axis, confer regional identity and are involved in the regulation of proliferation, apoptosis and differentiation (Berkes et al., 2004; Ferretti et al., 2011; Gordon et al., 2011; Koss et al., 2012; Yao et al., 2013). Pbx1 participates in many developmental processes, as demonstrated by the complex phenotypes associated with Pbx1 loss-of-function in mice (Brendolan et al., 2005; DiMartino et al., 2001; Ferretti et al., 2011; Golonzhka et al., 2015; Koss et al., 2012; Manley et al., 2004; Selleri et al., 2001; Stankunas et al., 2008; Vitobello et al., 2011). Genes encoding PBC class HD proteins share a high degree of sequence homology and have overlapping functions in domains of co-expression in vivo (Capellini et al., 2011). In fact, select developmental defects associated with Pbx1 loss-of-function were only uncovered when the Pbx1 deficiency was combined with homozygous or heterozygous loss of Pbx2 or Pbx3 (Capellini et al., 2011; Ferretti et al., 2011; Koss et al., 2012). Mechanistically, PBX1 associates with members of the MEIS/PREP subclass of TALE-HD proteins, but can also form heteromeric complexes with HOX proteins, non-HOX HD-containing proteins, bHLH or PAX proteins (Ladam and Sagerström, 2014; Longobardi et al., 2014; Schulte, 2014). As we have recently shown, MEIS2 is an essential co-factor of the neurogenic transcription factor PAX6 and as such is required for the acquisition of a general neuronal fate in the SVZ and the subsequent differentiation of a subpopulation of these cells towards dopaminergic periglomerular neurons (Agoston et al., 2014; Brill et al., 2008; Hack et al., 2005; Kohwi et al., 2005, 2007). Pbx1 expression in structures associated with adult forebrain neurogenesis in rodents has been reported, but its functional relevance has remained unresolved (Redmond et al., 1996). Here, we now define a function for Pbx1 as an early regulator of neurogenesis in the mouse SVZ.

We first characterized Pbx1 mRNA expression and protein localization in the brain of 7- to 11-week-old mice. Groups of cells staining positive for Pbx1 transcripts and protein were located directly underneath the ependymal cell layer (EpCL) at the dorsal and lateral walls of the SVZ (Fig. 1A-F). Cells exhibiting nuclear immunoreactivity for PBX1 contribute to the Ki67⁺ rapidly proliferating cell population in the SVZ, with 65.4±5.5% of the Ki67⁺ cells also labeling for PBX1 (Fig. 1D,H, Table S3). Consistent with expression in TAPs, 92.5±5.4% of the Ascl1-expressing cells in the adult SVZ label for PBX1 (Fig. 1E,H) (Kim et al., 2007). The gene encoding the intermediate filament protein nestin is expressed in ependymal cells, TAPs and slowly proliferating bona fide neural stem cells (Cattaneo and McKay, 1990; Lendahl et al., 1990). Nestin⁺ cells in the subependyma, but not the EpCL itself, were also immunopositive for PBX1, further strengthening the notion that PBX1 protein is present in neural progenitor cells of the adult SVZ (Fig. 1F). Of the Tuj1-expressing neuroblasts in the SVZ or RMS, 81.75±13.2% were immunoreactive for PBX1 and virtually all of these co-labeled for MEIS2 (Fig. 1G-I) (Agoston et al., 2014). Notably, the dentate gyrus of the hippocampus, which is the second major stem cell niche in the adult mouse brain, is devoid of Pbx1 transcripts or protein (Fig. 1J,K). Similar to Meis2, Pbx1 expression thus specifically marks the SVZ neurogenic niche (Agoston et al., 2014). By contrast, almost all cells in the adult SVZ, RMS, corpus callosum, cortex and striatum stained positively for PBX2, consistent with the widespread expression of Pbx2 in the embryo (Fig. 1L-M, Fig. S1) (Selleri et al., 2004). Only a few cells in the SVZ and RMS expressed Pbx3 and these were mostly immunonegative for PBX1 or MEIS2 (Fig. 1N-O″). Double labeling for each of the three PBX-encoding genes together with MEIS2 established that virtually all MEIS2-expressing cells stained positively for PBX1 and PBX2, whereas only 13.9% of the MEIS2⁺ cells were immunoreactive for PBX3 (Fig. 1P).

In the OB, PBX1-immunoreactive cells contribute to GCs and PGNs, with virtually all GCs and, on average, 22.8% of the calbindin (calbindin 1)⁺, 23.65% of the calretinin (calbindin 2)⁺ and 94.46% of the dopaminergic tyrosine hydroxylase⁺ (TH)⁺ PGNs labeling for PBX1 (Fig. 2A-G, Fig. S2, Table S3). PBX3 was absent from the TH⁺ PGN subtype (Fig. 2H, Fig. S2).

Collectively, the Pbx1 expression profile suggests an early role for PBX1 in neuronal lineage specification in the SVZ and a later contribution to the adult generation of OB interneurons.

The neurosphere assay allows the propagation of adult neural stem cells and TAPs in the presence of EGF and FGF2 in vitro while maintaining their multipotency, at least during early passages in culture (Reynolds and Weiss, 1992, 1996; Reynolds and Rietze, 2006). Free-floating adult neurospheres (aNSs) mostly consist of rapidly proliferating TAPs as well as a smaller number of presumably activated and EGF-responsive stem cells, both of which express nestin (Pastrana et al., 2009, 2011). Consistent with the prominent expression of Pbx1 in TAPs in vivo, PBX1 was detected in most nestin⁺ aNS cells (Fig. 3A). To distinguish between TAPs and neural stem cells, we labeled aNSs with the green fluorescent cytoplasmic dye 5-carboxyfluorescein diacetate, acetoxymethyl ester (CFDA AM; Fig. 3B-E). When applied to free-floating aNSs, the CFDA label is quickly diluted out by cell division in TAPs, but retained in the more quiescent stem cells, which are visible as brightly labeled cells positioned in the center of large aNSs (Fig. 3C, arrow). Intense nuclear staining for PBX1 was seen in 98.3% of the CFDA-negative, nestin⁺ TAPs, but in only 4% of the CFDA-retaining, nestin⁺ putative SVZ stem cells (Fig. 3D,E). Upon growth factor withdrawal and plating on appropriate substrates, aNS cells will differentiate into neurons, astroglia or oligodendrocytes (Doetsch et al., 2002; Reynolds and Weiss, 1992). When we induced SVZ-derived aNS cells to differentiate on laminin, PBX1 protein was retained in newborn Tuj1-expressing neurons and GFAP⁺ astrocytes but lost from oligodendrocytes (immunoreactive for the O4 antigen) (Fig. 3F-I, Table S3).

To investigate whether Pbx1 has a role in neural cell fate specification, we used a conditional allele of Pbx1 (Pbx1^fl/fl) that allows Cre-mediated deletion of exon 3 of the Pbx1 gene (Koss et al., 2012). PBX1 immunoreactivity was lost within 48 h when aNSs derived from Pbx1^fl/fl mice were infected with retroviruses carrying Cre recombinase together with an IRES-GFP cassette (Cre-GFP; Fig. 4A,B). Upon differentiation, Cre-GFP-transduced, Pbx1-deficient aNS cultures generated neurons and astrocytes at frequencies comparable to cultures transduced with GFP alone (Fig. S3). We therefore considered the possibility of functional compensation. Because PBX2 colocalizes with PBX1 in the SVZ (Fig. 1) and is known to compensate for loss of Pbx1 in embryonic domains of co-expression, we crossed the conditional Pbx1 mutant strain to a mouse line homozygous mutant for Pbx2 to obtain Pbx1^fl/fl;Pbx2^−/− compound mutant animals (Capellini et al., 2011; Ferretti et al., 2011; Koss et al., 2012; Selleri et al., 2004). Mice with a constitutive knockout of Pbx2 are viable and show no obvious phenotype (Selleri et al., 2004). Absence of PBX2 protein in Pbx2 homozygous mutant animals was confirmed in vivo and in aNS cultures derived from these animals in vitro by immunohistochemical staining and western blot analysis (Fig. S4). Similar to aNS cultures lacking Pbx1, cultures obtained from Pbx2 single-mutant animals did not show defective neurogenesis or gliogenesis, nor did cultures obtained from animals of different heterozygous mutant genotypes (Fig. S3). We therefore transduced aNSs from homozygous Pbx1^fl/fl;Pbx2^−/− animals with Cre-GFP-expressing virus to induce loss of Pbx1 on a Pbx2-deficient background [hereafter ‘double knockout’ (dKO)]. Pbx1^fl/fl;Pbx2^−/− cells infected with viruses that express only GFP served as control (single knockout for Pbx2, hereafter ‘sKO’). Upon differentiation, the generation of neurons from dKO progenitor cells was significantly reduced compared with sKO control cells (Fig. 4C,E). This was paralleled by an increased production of cells of the oligodendroglial lineage, as evident in a sharp rise in the number of cells expressing the oligodendrogliogenic transcription factor OLIG2 or the O4 antigen, a marker for mature oligodendrocytes (Fig. 4D,E). By contrast, the number of astrocytes generated from dKO and sKO progenitors did not differ (Fig. 4E, Table S4).

We next injected Cre-GFP- or GFP-expressing retroviruses into the SVZ of Pbx1^fl/fl;Pbx2^−/− mice in vivo. We first confirmed efficient loss of PBX1 protein in Cre-GFP-transduced cells, observing successful elimination in 97.25% of the Cre-expressing cells 3 days after virus injection (Fig. 5A-C). As in vitro, retroviral transduction of Cre recombinase reduced the relative proportion of DCX⁺ or PSA-NCAM⁺ neurons and enhanced the proportion of OLIG2-immunoreactive oligodendrocyte precursor cells (Fig. 5D-L, Table S4). Because oligodendrocyte precursor cells originating from the dorsal SVZ migrate into the overlying corpus callosum or the laterally located white matter tracks of the striatum, instead of entering the RMS, we mapped the distribution of sKO and dKO cells in brain sections obtained 3 days after injection of GFP- or Cre-GFP-expressing retroviruses into the SVZ (Menn et al., 2006). We observed that there was a tendency for dKO cells to populate the corpus callosum more frequently than sKO cells (Fig. 5M-O). Loss of Pbx1 expression in adult neural progenitor cells thus alters the gene expression profile and migratory behavior of the cells. Collectively, these results implicate Pbx1 in the regulation of neurogenic versus oligodendrogliogenic cell fate decisions of adult SVZ progenitor cells.

PBX family proteins can form stable heterodimers with MEIS proteins (Ladam and Sagerström, 2014). In the SVZ-OB neurogenic system, MEIS2 together with PAX6 (and the Distal-less homolog DLX2), is necessary for the acquisition of a dopaminergic PGN fate (Agoston et al., 2014; Brill et al., 2008; Hack et al., 2005; Kohwi et al., 2005). Therefore, we first examined whether PBX1 participates in the formation of higher order MEIS2/PAX6-containing transcriptional complexes. Indeed, PBX1-specific antibodies successfully precipitated both proteins from OB extracts (Fig. S5). To test whether Pbx1 also has a role in dopaminergic PGN differentiation, we stereotactically injected GFP- or Cre-GFP-expressing retroviruses into the RMS, where dopaminergic neurons of the OB undergo their final mitosis (Hack et al., 2005).

Because dopaminergic PGNs are characterized by a particularly slow maturation, the mice were analyzed up to 60 days post injection. It was previously reported that genetic ablation of Pax6 by a similar experimental approach, or forced expression of function-blocking forms of PAX6 or MEIS2 in neuroblasts of the RMS, elicited a cell fate change from dopaminergic to calretinin-expressing PGNs (Agoston et al., 2014; Hack et al., 2005; Kohwi et al., 2005). To our surprise, only a few dKO cells could be recovered from Cre-GFP-infected Pbx1^fl/fl;Pbx2^−/− brains, whereas sKO cells were abundant in the granule cell layer (GCL) and glomerular layer (GL) of GFP-infected littermates (Fig. 6A-C). To more closely compare the survival of sKO and dKO cells, we performed simultaneous lineage tracing of both genotypes in the same animal (Fig. 6D-G). Equal-titer retroviral stocks expressing either tdTomato or Cre-GFP were mixed and injected into the RMS of Pbx1^fl/fl;Pbx2^−/− animals. Ten days post injection, many red fluorescent (sKO) and green fluorescent (dKO) cells were seen in the core of the OB, where the migratory stream of SVZ-born neuroblasts enters the OB (Fig. 6D). At 21 days post injection, both sKO and dKO cells were found dispersed throughout the GCL, and some sKO cells had settled in the GL (Fig. 6E). We then followed the survival and migration of sKO and dKO cells by assessing their frequency and distribution at 19, 21, 28, 35 and 50 days post injection. Previous BrdU birthdating experiments had shown that only a fraction of the neuroblasts that leave the SVZ can successfully integrate into the OB circuitry and survive for longer than 1 month (Petreanu and Alvarez-Buylla, 2002; Winner et al., 2002). In agreement with these studies, we observed a continuous, gradual decline in the total number of tdTomato-labeled sKO cells during the 31 days of our analysis (Fig. 6F, left). Interestingly, the total number of dKO cells decreased even more sharply, with only very few cells still present in the OB at 50 days post injection (Fig. 6F, right). When we grouped the cells according to their location in the OB, the relative distribution of sKO cells in the GCL, plexiform layer and GL suggested that equal proportions of the surviving cells entered the GL or remained in the GCL, with a transient population of cells found in the plexiform layers, probably entering them on their way to the GL (Fig. 6F,G, left panels). By contrast, dKO cells were rarely observed in the GL (Fig. 6F,G, right panels). Because apoptotic cell death is a common feature of regions of ongoing neurogenesis in the brain, including the adult OB, we quantified the proportion of tdTomato⁺ or Cre-GFP⁺ cells that co-stained with an antibody against cleaved (activated) caspase 3 (ac-caspase3) (Biebl et al., 2000; Winner et al., 2002). We found that between 1.2% and 1.5% of the dKO cells were ac-caspase3⁺ in the RMS at 3, 7 and 10 days post injection. Notably, 5.8% of the dKO cells in the GCL were ac-caspase3⁺ at 28 and 35 days post injection. By contrast, apoptosis never occurred in sKO cells within the RMS and reached only 3.5% in the GCL and GL 35 days post injection. Pbx1 is thus required for the long-term survival of adult-born OB neurons. Considering that dopaminergic PGNs are PBX1 immunoreactive (Fig. 2), the detection of TH⁺ dKO cells 60 days post injection was unexpected (Fig. 6H). These cells are likely to have survived as a result of ectopic upregulation of PBX3 expression, since sKO cells in the OB were mostly PBX3 negative, whereas 75% of the dKO cells expressed Pbx3 (Fig. 6I-K). Given the high degree of homology between PBX1, PBX2 and PBX3, PBX3 might indeed functionally replace PBX1 in the regulation of genes required for dopaminergic differentiation in dKO cells.

Dcx and Th are regulated jointly by MEIS2 and PAX6. As we previously showed by chromatin immunoprecipitation followed by quantitative PCR (ChIP-qPCR), a MEIS/PBX consensus site located 2725 bp upstream of the start codon of the murine Dcx gene (DCXI) is bound by MEIS2, PAX6 and the PAX6-interacting transcription factor DLX2 in neuroblasts (Agoston et al., 2014). Moreover, the endogenous Dcx promoter/proximal enhancer (Karl et al., 2005; Piens et al., 2010) (NCBI AY590498, BX530055) is transcriptionally activated by MEIS2 and PAX6 (Agoston et al., 2014). In in vitro generated neurons, a PBX1-specific antibody efficiently precipitated the DCXI chromatin fragment (Fig. 7B). Reflecting the robust expression of Dcx in these cells, the DCXI site carried the H3K4me3 epigenetic mark, which is associated with active promoters (Fig. 7C) (Barski et al., 2007). PBX1 also induced the expression of a Dcx promoter/enhancer-driven luciferase reporter in HEK293T cells, both alone (probably owing to low-level endogenous MEIS/PREP expression in these cells) and together with ectopically expressed MEIS2 (Fig. 7D). By contrast, a reporter in which the MEIS/PBX consensus site was deleted could not be stimulated by transfection of Pbx1 and Meis2 (Fig. 7E). Interestingly, we found that PBX1 was already bound to the DCXI site in aNSs, even though these cells do not yet express Dcx (Fig. 7F). In aNS cells, the DCXI genomic region lacked H3K4me3 and H3K27me3 epigenetic marks, indicative of a transcriptionally non-restricted chromatin state (Fig. 7G). The PBX1-specific antibody was ineffective in ChIP assays on the DCXI site in in vitro differentiated astroglia or on a validated PBX binding site within the myogenin promoter in aNS cells (Fig. 7H) (Berkes et al., 2004). In addition, siRNA-mediated knockdown of Pbx1 effectively reduced PBX1 binding to the DCXI site in aNSs, further confirming the specificity of the PBX1 antibody that was used in the ChIP experiments (Fig. S6). Together, these results identify PBX1 as transcriptional activator of the Dcx promoter/enhancer and show that PBX1, in contrast to MEIS2, already associates with the DCXI site prior to Dcx expression.

We also examined whether PBX1 binds a known TALE-HD binding site, THI, located 3479 bp upstream of the start codon of the mouse Th gene within a known promoter/proximal enhancer. In a previous study, ChIP analysis with a MEIS2-specific antibody and chromatin prepared from adult mouse OB showed enrichment of MEIS2 at this motif (Fig. 8A) (NCBI AF415235) (Agoston et al., 2014). ChIP-qPCR experiments with the PBX1-specific antibody precipitated the THI genomic fragment from freshly isolated OB or SN4741 cells, a dopaminergic neural progenitor cell line derived from mouse embryonic substantia nigra (Fig. 8B-D) (Son et al., 1999). PBX1 thus binds the THI site in Th-expressing cells. Interestingly, ChIP-qPCR experiments from primary aNSs or from isolated SVZ tissue also showed enriched PBX1 at the THI site, despite the fact that Th is not expressed in either cell population and H3K27me3 epigenetic marks prevail at the THI site in aNS cells (Fig. 8E-H, Fig. S7) (Cave et al., 2014).

We also compared by qPCR the transcript levels of Th and Dcx in isolated SVZ, OB tissue and aNS cultures. Dcx-expressing neuroblasts are present in the SVZ and OB, whereas Th-expressing PGNs are only found in the OB. Consistent with this, Th expression in the SVZ and in aNSs was barely detectable (Fig. 8I). Particularly in aNS cells, Dcx transcript levels, despite being very low overall, still exceeded those of Th by three orders of magnitude (Fig. 8I′). These findings reflect the bivalent configuration of the Dcx promoter/enhancer in aNSs compared with the transcriptionally repressed state of the Th gene promoter/enhancer in these cells. Moreover, our results are consistent with the notion that low levels of Dcx transcripts are already detectable in GFAP/prominin double-positive, bona fide neural stem cells in the SVZ (Beckervordersandforth et al., 2010).

Here, we describe sequential functions of Pbx1 in the SVZ adult neural stem cell system. Using conditional ablation of Pbx1 in adult neural progenitors or neuronally committed neuroblasts, together with ChIP-qPCR on the promoter/enhancers of selected downstream genes, we show that PBX1 is required for the acquisition of a neuronal instead of oligodendroglial cell fate, is necessary for the long-term survival of adult-generated young neurons, and can bind its consensus sites in the regulatory regions of selected target genes well before these are transcriptionally activated.

Pbx1 is expressed in rapidly proliferating ASCL1⁺/nestin⁺ progenitor cells in the SVZ. Its expression thus precedes that of MEIS2, which exhibits strong nuclear immunoreactivity only at the neuroblast stage and hence in a developmentally more restricted cell population of the SVZ (Agoston et al., 2014). Neural progenitor cells still possess the capacity to give rise to neurons, astrocytes and oligodendrocytes, whereas neuroblasts will mature exclusively into neurons. Notably, when PBX1⁺ SVZ progenitor cells were differentiated in vitro, PBX1 immunoreactivity was retained in neurons and astrocytes but rapidly lost in oligodendrocytes. Neurogenesis and oligodendrogliogenesis are competing cell fates (Hack et al., 2005; Ortega et al., 2013). Specifically, by continuous live imaging, Ortega and colleagues found that the oligodendroglial lineage tree separates early from cells committed towards a neuro-astroglial fate. Given that targeted manipulation of MEIS2 with dominant-negative constructs in the SVZ caused a neurogenic-to-astrogliogenic fate switch (Agoston et al., 2014), our results argue that neuronal cell fate specification and differentiation in the adult SVZ involves the sequential activity of two TALE-HD proteins: PBX1 acts in an early progenitor cell population, directing it towards a neuronal as opposed to oligodendroglial fate, while MEIS2 is necessary for neuronal differentiation from more specified, neuro-astroglial progenitor cells. Our results might therefore have implications for the understanding of demyelinating pathologies. The number of SVZ-derived oligodendrocytes increases fourfold after a demyelinating lesion in the corpus callosum, indicating that stem or early progenitor cells of the SVZ can respond to the insult by specifically upregulating the production of those cells that are needed for repair (Menn et al., 2006). According to the data presented above, this process should require downregulation of Pbx1 as a direct response to the demyelinating stimulus.

PBX proteins bind DNA cooperatively with other transcription factors (Longobardi et al., 2014; Penkov et al., 2013). Additional proteins thus need to participate in PBX1-dependent lineage decisions. Likely candidates are PAX6 and DLX2, which have known functions in SVZ neurogenesis and bind to TALE-HD proteins with high affinity in solution, or even additional proteins of the MEIS/PREP family (Agoston et al., 2014; Brill et al., 2008; Hack et al., 2005; Kohwi et al., 2005, 2007). However, no member of this protein family other than MEIS2 has thus far been implicated in SVZ neurogenesis. The precise composition of PBX1-containing transcriptional complexes in distinct cell populations in the SVZ requires further investigation.

Genetic deletion of Pax6 or forced expression of a function-blocking form of MEIS2 in neuroblasts of the RMS elicits a cell fate change from dopaminergic to calretinin-expressing PGNs (Agoston et al., 2014; Kohwi et al., 2005; Hack et al., 2005). Because both PAX6 and MEIS2 form higher order protein complexes with PBX1, a decrease in the number of TH⁺ cells with a concomitant increase in calretinin⁺ cells was expected in dKO compared with sKO cells. However, instead of maturing towards an alternative PGN subtype fate, Pbx1/Pbx2-deficient neuroblasts were gradually eliminated from the OB, with the first apoptotic dKO cells appearing in the RMS. In addition, dKO cells were infrequently found in the GL and the majority of these exhibited ectopic upregulation of PBX3. Owing to the high structural homology among PBX proteins, PBX3 is likely to compensate for the loss of PBX1 and PBX2 during dopaminergic differentiation in dKO cells. Two scenarios might account for the predominant loss of dKO cells from the GL: dKO cells might be unable to reach the GL due to defective cell migration, or they might be eliminated by cell death before they can reach the GL. The fact that we did not observe accumulating dKO cells in the RMS, or anywhere else, during their radial migration in the GCL, together with the higher rate of programmed cell death in dKO compared with sKO cells, argues for compromised cell survival as an underlying cause. However, we cannot formally rule out a contributing migratory defect. It is worth pointing out that the high-affinity netrin receptor deleted in colorectal cancer (DCC) is downregulated in mesencephalic dopaminergic neurons in Pbx1-deficient animals (Sgadò et al., 2012). However, we found no correlation between the localization of DCC and PBX1 in the OB (data not shown), leaving an open question as to whether PBX1 also has a role in guiding neuroblasts during their migration in the OB.

Interestingly, the survival of dopaminergic PGNs requires the TALE-HD-interacting transcription factor PAX6 (Ninkovic et al., 2010). Moreover, dopaminergic genes in C. elegans, including the nematode homolog of Th, are cooperatively activated by the Pbx gene homologs ceh-20, ceh-40 and ceh-60, together with ceh-43, a Dlx homolog (Doitsidou et al., 2013). Intriguingly, a DLX consensus binding site is present in close proximity to the THI TALE-HD binding site, which is bound by PBX1 and MEIS2 (Agoston et al., 2014) (Fig. 8). Together with the results presented above, these findings argue for a phylogenetically conserved role for PBX family proteins in dopaminergic neuron differentiation (Doitsidou et al., 2013).

We observed PBX1 binding to the regulatory regions of the downstream targets Dcx and Th prior to their transcriptional activation. The presence of PBX1 at the Th promoter/enhancer in chromatin isolated from the SVZ or aNSs is particularly intriguing, as Th transcripts are not detectable and the Th promoter carries repressive histone modifications in both cell populations, strongly suggesting that the Th gene is transcriptionally silent in these cells. In fact, full maturation of dopaminergic OB neurons, including Th expression, takes several weeks and requires odor-mediated afferent synaptic activity (Akiba et al., 2009; Baker and Farbman, 1993; Baker et al., 1988; Winner et al., 2002). Moreover, aNSs in culture will not express Th, even with specialized differentiation protocols, unless they are genetically modified (Cave et al., 2014; Deleidi et al., 2011).

PBX1 chromatin binding can thus anticipate de novo transcriptional activation of the Th gene by weeks if not months, which is suggestive of a priming function for PBX1 in this context. Priming transcription factors are a special class of transcriptional regulators that can penetrate silent chromatin and bind regulatory regions at times when the overall chromatin structure still prevents access of other transcription factors (Iwafuchi-Doi and Zaret, 2014). Notably, observations made in the context of skeletal muscle development, hindbrain patterning or breast cancer had already indicated a role for PBX1 in transcriptional priming in these systems (Choe et al., 2014; Berkes et al., 2004; Magnani et al., 2011). Specifically, during skeletal muscle differentiation, PBX1 is constitutively bound to the promoter of the myogenin (Myog) gene and transcriptional activation of this promoter requires recruitment of the pro-myogenic transcription factor MYOD by PBX1 (Berkes et al., 2004). Likewise, Pbx4 primes the hoxb1a promoter during zebrafish hindbrain development, whereas in the breast cancer cell line MCF7, PBX1 acts as pioneer factor for the estrogen receptor alpha (ERα)-dependent transcriptional program following estrogen stimulation (Magnani et al., 2011).

Thus, in addition to its role in the regulation of neurogenic cell fate decisions and the survival of newly generated neurons, our results suggest that PBX1 might act as initial ‘mark’ that endows the Dcx and Th genes with the competence for later activation in the SVZ adult neurogenic system. Collectively, the present study establishes a role for PBX1 in adult SVZ neurogenesis that goes beyond that of its heterodimerization partner MEIS2.

Stereotactic injections in 8- to 12-week-old Pbx1^fl/fl;Pbx2^−/− mice of mixed gender (Koss et al., 2012; Selleri et al., 2004) were performed as described and with published coordinates (Brill et al., 2008; Agoston et al., 2014; Hack et al., 2005). Animals received independent injections into both hemispheres. The number of animals examined per experimental setting is given in Table S4. All procedures involving animals were approved by the local animal care committee and are in accordance with the law for animal experiments issued by the Hesse state government. Statistical analysis was performed with two-tailed, unpaired Student's t-test (Graph Pad Prism 5.01); the s.e.m. represents variance between different injections.

For lineage tracing with tdTomato- and Cre-GFP-expressing viruses, a mixture of both viral stocks, diluted to equal titer, was injected into Pbx1^fl/fl;Pbx2^−/− littermates. In each series of experiments, all animals received injections from the same premixed viral stock. For analysis, serial 75 µm thick vibratome sections of the OB were cut and every third section was stained with antibodies specific for tdTomato or GFP. Fig. 6F shows the absolute number of cells that could be recovered for each genotype by this approach.

In situ hybridization was performed on 75 µm thick vibratome section as described (Heine et al., 2008). The Pbx1-specific probe comprised nucleotides 1510 to 2398 of NCBI AF202197 and the Pbx2-specific probe comprised nucleotides 1045 to 1916 of NCBI NM_017463.

For immunocytochemical and immunohistochemical staining, cells were fixed with 2% paraformaldehyde (PFA) in PBS (pH 7.4). Free-floating aNSs were allowed to attach to poly-D-lysine-coated coverslips for 30 min prior to fixation. Immunohistochemical analysis on PFA-perfused frozen or vibratome brain sections was performed as described (Agoston et al., 2014). Primary and secondary antibodies are listed in Table S1.

Images were taken with a Nikon 80i or a Nikon Eclipse TE2000-E confocal microscope. The number of specimens and sample sizes analyzed are given in Table S3. Cells were counted blind. Standard deviation was calculated between technical replicates and statistical significance assessed by unpaired Student's t-test. Chromogen staining was performed with a Ventana DISCOVERY XT automated staining system, with antigen retrieval protocol Conditioner #1, OmniMap HRP detection and counterstaining with Hematoxylin.

ChIP analysis from tissues or cells was performed as described (Agoston et al., 2014). Antibodies and primers used for ChIP are specified in Table S2. Experiments were conducted at least in triplicate and plotted as s.e.m. Statistical analysis was performed with two-tailed, unpaired Student's t-test between experimental samples and the control sample precipitated with normal IgG antibodies.

Sphere-forming cells were isolated from the lateral walls of the lateral ventricle of 8- to 12-week-old C57BL/6 mice and propagated under non-adherent conditions in DMEM/F12 and GlutaMAX (Life Technologies) containing 1× B27 supplement (Life Technologies), 10 mM HEPES pH 8.0 (Sigma Aldrich), 2 mM L-glutamine (Sigma Aldrich), 20 ng/ml EGF (human recombinant, Peprotech), 20 ng/ml bFGF (FGF2; human recombinant, Peprotech) and 1× penicillin/streptomycin (Sigma Aldrich) as described (Agoston et al., 2014). Unless otherwise noted, passage 1 aNSs cultured for no more than 5 days were used for infection with GFP- or Cre-GFP-expressing viruses. For quantification, experiments were counted blind; statistical analysis was performed with two-tailed, paired Student's t-test between GFP- or Cre-GFP-transduced cell cohorts. To obtain cultures enriched for neurons for ChIP, aNSs were transduced with Pax6 followed by differentiation on laminin-coated dishes (1 µg/cm²; Roche) in medium lacking EGF but containing 2 ng/ml FGF2 and 20 ng/ml brain-derived neurotrophic factor (BDNF; human recombinant, Peprotech). To obtain cultures enriched for astrocytes, aNSs were differentiated on poly-D-lysine-coated dishes (140 µg/ml; Sigma Aldrich) in medium lacking EGF, FGF2, but containing 10 ng/ml ciliary neurotrophic factor (CNTF; human recombinant, Peprotech) and 0.5% FCS (SeraPlus, Life Technologies).

For the label-retention assay, primary aNSs were passaged, grown for 24 h and then pulsed for 1 h with 2.5 µM 5-carboxyfluorescein diacetate, acetoxymethyl ester (C1354, Thermo Fisher Scientific) for 5 min at room temperature, washed and cultured at 37°C. After 4 days, the spheres were dissociated and grown for 24 h on laminin-coated dishes in the presence of EGF and bFGF prior to immunohistochemical analysis. Because, under these conditions, cells occasionally spontaneously exit the cell cycle and differentiate, cultures were stained for the stem/progenitor cell marker nestin. Only nestin⁺/CFDA⁺ cells were counted as quiescent neural stem cells in vitro. The number of technical replicates and number of cells counted for each experiment are given in Table S4.

Retroviral constructs were pCLIG-GFP or corresponding vectors carrying a Cre-IRES-GFP cassette or tdTomato (Brill et al., 2008; Hildinger et al., 1999; Hojo et al., 2000). Virus production was performed as detailed (Agoston et al., 2014). Viral titers were between 4.7×10⁶ and 1.7×10⁷ p.f.u. Viral stocks were diluted to equal titer before use. Reporter assays using a luciferase reporter under the control of the 2073 bp endogenous promoter/enhancer of the Dcx gene (corresponding to NT 143935144 to 143933071 of NC_000086.7, Mus musculus C57BL/6J chr.X, GRCm38.p4; Piens et al., 2010) cloned in pGL3-basic and the Pbx1a-pCS2⁺ and Pbx1b-pCS2⁺ mammalian expression plasmids were performed as described (Agoston et al., 2014). For deletion of the DCXI site, site-directed mutagenesis was carried out with primers 5′-GTGATTAAAATCATGCATATATCTTGCATT-3′ and 5′-TGAAAATAGAAACAGCCCAGATGTCTGT-3′ on a Dcx promoter/enhancer comprising nucleotides 1-1870 of the above reporter construct (corresponding to NT 143935144 to 143933274 of NC_000086.7; Piens et al., 2010) using the Phusion Site-Directed Mutagenesis Kit (New England Biolabs). Transcription factor binding site prediction was performed with MatInspector [Genomatix Software Suite (Cartharius et al., 2005)]. siRNA-mediated knockdown of Pbx1 was carried out with siRNAs (5′-GGUUGGCAGGAUGCUACUA-3′), 1.6 nmol transfected per 2×10⁶ cells; Eurogentech, Belgium). Non-targeting control siRNAs were purchased from Eurogentech (SR-CL000-005). RNA duplexes were transfected with Metafectene Pro (Biontex). After siRNA transfection, cells were grown for 48 h as free-floating spheres before they were used for ChIP.

We thank Magdalena Götz for the Cre-IRES-GFP and Duran Sürün for the tdTomato viral vectors; Stefan Momma for the SN4741 cell line; and Michael Cleary, Arthur Buchberg, Hermann Rohrer and Jane Johnson for antibodies.