Querkopf is a key marker of self-renewal and multipotency of adult neural stem cells

Adult neural stem cells (NSCs) are present in the subventricular zone (SVZ) and the subgranular layer of the hippocampus and are a source of new neurons throughout life (Zhao et al., 2008). SVZ NSCs give rise to transit amplifying cells, which in turn produce neuroblasts that migrate to the olfactory bulbs and differentiate into olfactory interneurons (Lois et al., 1996). SVZ NSCs have been induced to differentiate into an array of different neural cell types, both in vivo and in vitro (Curtis et al., 2003; Lois and Alvarez-Buylla, 1993). NSCs have therefore aroused great interest, as they hold much promise for the generation of new therapies for neurodegenerative disorders and neural injury. However, two major issues have hindered progress in the field. First, it has been difficult to isolate a pure population of NSCs in their most primitive state. Although some markers are available for the enrichment of NSCs, a definitive marker for NSCs remains to be defined. Second, in order to direct differentiation of NSCs into specific neuronal cell types, which would ideally include cell types not normally formed by NSCs, a greater understanding of factors underlying lineage commitment and hierarchy is required.

NSCs have been shown to express a number of genes including intracellular proteins such as glial fibrillary acidic protein (GFAP), nestin and inhibitor of DNA binding 1 (Id1) (Doetsch et al., 1999; Mignone et al., 2004; Nam and Benezra, 2009), cell surface proteins such as SSEA-1 (also known as FUT4 and LeX) and CD133 (Capela and Temple, 2002; Coskun et al., 2008; Beckervordersandforth et al., 2010), and low levels of heat stable antigen (HSA; HSA is also known as CD24). NSCs also show little affinity for the lectin peanut agglutinin (PNA) (Rietze et al., 2001). Although some of these characteristics have been used to isolate cell fractions enriched for NSCs, most of the markers are expressed strongly in other cells in the brain, whereas others only mark a subset of NSCs. For example, GFAP is expressed by both NSCs and differentiated astrocytes. Of note, none of the markers currently in use to enrich for or to identify NSCs play a non-redundant role in adult NSC biology. We show here that high levels of querkopf (Qkf; also known as Kat6b and Myst4) expression identifies multipotent, self-renewing NSCs.

We have previously shown that a lack of the MYST-family histone acetyltransferase, QKF, leads to defects in the establishment and self-renewal of adult NSCs and consequently to a progressive defect in adult neurogenesis (Merson et al., 2006; Rietze et al., 2001). Remarkably, in humans loss of only one allele of MYST4 (QKF) leads to intellectual disability (Kraft et al., 2011; Clayton-Smith et al., 2011). Unlike other histone acetyltransferases, the expression of Qkf is both spatially and temporally regulated, and strong Qkf expression is localized to neurogenic regions both during development and in the adult (Merson et al., 2006; Thomas et al., 2000). We observed strong expression of Qkf gene in a small subset of neurogenic SVZ cells, raising the question of whether there is a relationship between Qkf gene expression levels and cell identity. Using Qkf–GFP transgenic mice to trace levels of Qkf gene activity, we show that the strongest Qkf–GFP-expressing cells possess all NSC characteristics, namely multipotency and self-renewal. Moreover, we show a remarkable relationship between levels of Qkf–GFP expression and progression in neuronal lineage differentiation.

Qkf–GFP bacterial artificial chromosome (BAC) transgenic mice were generated by microinjection of the BAC vector depicted in Fig. 1A into fertilized oocytes. The BAC vector contained an eGFP reporter inserted at the start codon of Qkf in exon 1. In addition, the BAC contained genomic sequence from 67 kb upstream to 130 kb downstream of the Qkf start codon. Offspring of the BAC transgenic founder mice were screened for expression of the GFP reporter between embryonic day (E)9.0 and E12.5. At E9.5 the endogenous Qkf gene and the Qkf–GFP transgene were expressed ubiquitously (supplementary material Fig. S1A,B). A total of 19 independent founder Qkf–GFP reporter lines were generated. The seven strongest GFP-expressing lines were kept for further analysis.

The presence of GFP reporter expression was confirmed by immunohistochemistry on neurospheres from Qkf–GFP transgenic and wild-type littermates. Qkf–GFP transgenic cells attached to a laminin substrate stained strongly for GFP, which was absent in wild-type cells (supplementary material Fig. S1C–F).

To determine the mRNA products of the Qkf–GFP transgene, a northern blot analysis of RNA isolated from the adult brains of all seven Qkf–GFP transgenic lines was performed. Wild-type adult brain RNA and E12.5 RNA were used as controls. Qkf probe no. 19 was used to determine endogenous Qkf levels. This probe hybridizes to the 5′UTR (untranslated region) and to part of the first coding exon of Qkf, both 5′ and 3′ of the GFP insertion site (see Fig. 1A). Consistent with previous studies, Qkf transcripts of 1 kb and 7.2 kb were present in the wild-type adult brain (supplementary material Fig. S1G) (Thomas et al., 2000). The same transcripts were also present at similar levels in the Qkf–GFP adult brains, suggesting that the endogenous Qkf gene was functionally normal. The E12.5 embryo expressed an additional 7.7 kb message, which has been previously observed (Thomas et al., 2000). Using a 668 bp GFP probe, messages between 1.3 kb and 1.8 kb were detected in the Qkf–GFP transgenic, but not in wild-type samples, with a 1.5 kb mRNA being the most prominent. These RNA species were of the expected size: 1079 bp GFP sequence, 113 bp 5′UTR, 35 bp 3′UTR and 50–250 bp poly(A) tail. The 1.5 kb and 1.8 kb messages also hybridized to Qkf probe no.19 because of the presence of Qkf 5′UTR sequence in the GFP mRNAs. In addition, a 1.1 kb transcript was specifically detected by Qkf probe no. 19 in the Qkf–GFP brain RNA. This transcript was not detected by the GFP probe in any wild-type samples. It is therefore likely to be an aberrant splice product of the Qkf–GFP transgene, which does not contain the GFP coding sequence.

To determine the accuracy of the Qkf–GFP reporter in representing the wild-type Qkf locus, in situ hybridization was carried out probing for GFP and endogenous Qkf in brain sections from all seven transgenic lines (example shown in Fig. 1B–G). To distinguish wild-type endogenous Qkf from Qkf–GFP transgenic transcripts, a Qkf probe (Qkf 1332) that hybridizes to exon 16, which is absent from the transgene, was used. Endogenous wild-type Qkf mRNA was expressed strongly in the SVZ (Fig. 1B) as previously described (Merson et al., 2006). In particular, a small number of cells within the subependymal layer of the SVZ expressed Qkf at high levels (Fig. 1F). Similarly, in an adjacent slide of the same brain, GFP mRNA was strongly expressed in specific cells (Fig. 1C,G). The complex pattern of Qkf gene expression in the mid-gestation embryo (Thomas et al., 2000) was also faithfully reproduced by the Qkf–GFP transgene, with strong expression domains in the developing telencephalon, eyelids, cornea, retina, the lateral aspects of the lens, nasal epithelium, tongue and developing teeth (supplementary material Fig. S1H–I). Taken together, our results suggest that Qkf–GFP transgene expression accurately depicts endogenous Qkf promoter activity pattern in all seven transgenic lines.

To ensure that Qkf–GFP expression levels quantitatively reflect levels of endogenous wild-type Qkf, cells from E12.5 embryos from three independent transgenic lines were sorted according to their level of GFP expression by fluorescence-activated cell sorting (FACS). After selecting for single live cells and excluding debris (Fig. 1H), four cell fractions were collected on the basis of their Qkf–GFP reporter expression, with the highest 10% Qkf–GFP-positive cells designated Qkf–GFP high (Qkf–GFP^Hi). The next three fractions containing 30% of cells each were designated Qkf–GFP medium-high (Qkf–GFP^MedHi), Qkf–GFP medium-low (Qkf–GFP^MedLo), and Qkf–GFP low (Qkf–GFP^Lo; Fig. 1I). Endogenous Qkf mRNA expression levels were assayed. In all instances examined, Qkf–GFP expression levels correlated strongly with endogenous Qkf expression levels (r=0.9619, P<0.0001, n=3 animals; Fig. 1J–L). On average, Qkf–GFP^MedHi cells had 40% less endogenous Qkf expression than Qkf–GFP^Hi cells (P<0.0001). Similarly, Qkf–GFP^Hi cells had more than twice the amount of endogenous Qkf expression than Qkf–GFP^MedLo, and five-times more Qkf expression than the Qkf–GFP^Lo population (P<0.0001). This result and the in situ hybridization data demonstrate that the GFP transgene expression under the Qkf promoter accurately portrays endogenous Qkf expression levels.

To determine whether there is a relationship between Qkf promoter activity and neurogenic cell identity in the SVZ, we carried out in vivo cell cycle studies and imaging, ex vivo cell marker analysis, colony formation, self-renewal and differentiation assays. For ex vivo studies, adult Qkf–GFP SVZ cells were sorted according to their level of GFP expression by FACS. As outlined above, after selecting for single live cells and excluding debris (Fig. 1H), four cell fractions were collected on the basis of their Qkf–GFP reporter expression (Fig. 1I).

Cells sorted from adult Qkf–GFP SVZ of seven transgenic founder lines were assayed for the formation of colonies of proliferating cells (neurospheres) and showed a strong positive correlation between Qkf–GFP expression levels and the ability to form neurospheres (Fig. 2A, data summarized in B; r=0.7800, P<0.0001, two-tailed test). The Qkf–GFP^Hi fraction gave rise to the highest number of neurospheres, which was significantly more than the three other fractions (P<0.0001, n=28 Qkf–GFP sorts using one animal each). By contrast, neurospheres were rare in cell fractions with the lowest Qkf–GFP expression. A 130-fold enrichment for neurosphere forming cells was observed in the Qkf–GFP^Hi fraction over the Qkf–GFP^Lo fraction, a 12-fold enrichment over unsorted cells (P<0.0001, n=28 Qkf–GFP sorts and six unsorted SVZ cultures from one animal each) and a 4.5-fold enrichment over wild-type cells sorted by FACS using the same criteria as those described in Fig. 1H (P<0.0001, n=28 Qkf–GFP sorts, n=3 wild-type sorts). Our data suggest that high Qkf expression correlates with neurogenic activity of SVZ cells. Indeed, the Qkf–GFP^Hi population was so highly enriched for neurosphere-forming cells that sorting of SVZ cells using a combination of stem cell markers, which individually enrich for neurosphere-forming cells (Qkf–GFP^Hi, SSEA-1 positive, CD133 positive, HSA^Lo), did not result in further enrichment (data not shown).

Because plating density can influence the efficiency of neurosphere colony formation (Capela and Temple, 2002), we carried out clonal assays. Cells were sorted as described above, and each sorted event was plated into a well of a 96-well plate. Consistent with bulk culture, a strong, positive correlation between Qkf–GFP expression levels and neurosphere forming ability was observed (r=0.8997, P<0.0001; Fig. 2C). The number of neurosphere forming cells was significantly higher in the Qkf–GFP^Hi fraction than in all other fractions (P<0.0001, n=3). Therefore, plating densities did not affect the relationship between Qkf–GFP expression levels and neurosphere formation.

To ensure that GFP expression on its own did not affect colony formation, SVZ cells from β-actin–GFP transgenic mice (Hadjantonakis et al., 1998) were sorted as described in Fig. 1H,I and assayed for neurosphere colony formation. Although there was some variation in the colony-forming capacity of SVZ cells expressing different levels of β-actin–GFP, this pattern was distinctly different from that observed in Qkf–GFP mice (Fig. 2D compare with B); no significant correlation was found between β-actin–GFP expression and neurosphere formation (r=0.4532, P=0.140, n=4 β-actin–GFP animals). Therefore, the strong relationship between Qkf–GFP expression levels and neurogenic activity is specific to the regulatory elements of the Qkf gene.

In contrast to sorted blood cells, only a minority of the FACS events derived from the enzymatically and mechanically dissociated SVZ represent live cells. Therefore, we evaluated the relationship between FACS events and the number of viable cells in the different Qkf–GFP fractions immediately after sorting. Approximately 35% of sorted events were found to be live cells (Fig. 2E). The relationship between FACS events and live cells was similar in all Qkf–GFP fractions, except for a slight reduction in the Qkf–GFP^Lo fraction (supplementary material Table S5; n=4). This suggests that the correlation between Qkf–GFP expression levels and neurosphere formation is not an artefact due to the proportion of live cells in the different Qkf–GFP fractions at the time of plating.

Enrichment in live cells (rather than sorted events) is commonly used as a baseline in other studies (Capela and Temple, 2002; Coskun et al., 2008; Rietze et al., 2001). When based on the number of sorted cells (Fig. 2E) that are alive 2 hours after plating (supplementary material Fig. S2), one in four live cells in the Qkf–GFP^Hi fraction give rise to neurospheres (Fig. 2F), compared with 1:50 unsorted live cells (P<0.0001, n=28 Qkf–GFP sorts, n=6 unsorted wild-type cultures), showing the high enrichment of neurosphere-forming cells in the Qkf–GFP^Hi fraction.

Self-renewal is one of the defining criteria of stem cells. To determine the relationship between Qkf–GFP expression levels and the self-renewal potential of SVZ neurosphere-forming cells, long-term passaging was carried out as an in vitro readout for self-renewal. Freshly isolated SVZ cells were sorted on the basis of Qkf–GFP expression and plated at equal cell densities. After five passages, only the two highest Qkf–GFP-expressing fractions retained proliferating cells (Fig. 3A; n=6 Qkf–GFP sorts and cultures). Moreover, two Qkf–GFP^Hi and two Qkf–GFP^MedHi cultures were chosen and passaged successfully until passage 12. Importantly, there were nearly 15-fold more cells in the Qkf–GFP^Hi cultures than in the Qkf–GFP^MedHi cultures after five passages, suggesting that a substantial proportion of the Qkf–GFP^MedHi colonies were progenitor colonies with limited self-renewal capacity. The two lowest Qkf–GFP fractions did not maintain any self-renewal activity. Even though live cells were present in these fractions at passage one, by the third passage, no proliferating cells were present in the Qkf–GFP^Lo culture. Similarly, only one out of six Qkf–GFP^MedLo cultures was able to proliferate beyond passage four, and no live cells were present in this culture at passage 6. The strong positive correlation between Qkf–GFP expression and self-renewal ability shows that high Qkf levels are indicative of the self-renewal potential of NSCs (r=0.9045, P<0.0001).

Neurospheres contain stem cells, transit amplifying cells and lineage-restricted proliferating progeny. In the neurosphere assay, stem cells are thought to have an increased proliferation rate and capacity compared with progenitor cells (Louis et al., 2008; Reynolds and Rietze, 2005). We determined the proliferative capacity of SVZ colony-forming cells by assessing the size of neurospheres in the different Qkf–GFP-expressing fractions.

SVZ cells were sorted on the basis of Qkf–GFP expression and plated in a semi-solid medium containing collagen. Colonies were grown for 3 weeks and then scored. Stem cell colonies are designated as colonies that are larger than 2 mm in diameter three weeks after plating (Fig. 3B). The Qkf–GFP^Hi fraction formed the highest proportion of stem cell colonies, more than the Qkf–GFP^MedHi in trend fraction (P=0.092, n=7) and significantly more than the two lowest Qkf–GFP-expressing fractions (P<0.01), which very rarely formed large colonies (Fig. 3C). Interestingly, one in 3.5 Qkf–GFP^Hi colonies were larger than 2 mm, suggesting that 29% of neurosphere-forming cells in the Qkf–GFP^Hi are NSCs. There was a significant, positive correlation between stem cell colony formation and Qkf promoter activity (r=0.8870, P<0.0001).

To determine the proliferation capacity of SVZ cells in bulk neurosphere cultures, the volume of all individual neurospheres in each of the four Qkf–GFP fractions was determined. There was a positive correlation between Qkf–GFP expression levels and the average volume of neurospheres (r=0.3434, P<0.0001; Fig. 3D). Neurospheres in the Qkf–GFP^Hi fraction were on average, more than twice the size of neurospheres in the two low Qkf–GFP-expressing fractions (P<0.0001, n=583 Qkf–GFP^Hi scored neurospheres, 546 Qkf–GFP^MedHi, 233 Qkf–GFP^MedLo, 26 Qkf–GFP^Lo).

Together, our results show that the cells with the highest Qkf levels have the greatest capacity for neurosphere formation (Fig. 2), long-term self-renewal (Fig. 3A) and a high proliferative capacity (Fig. 3C,D), indicative of stem cell activity. The experiments above indicated that Qkf gene activity traced NSC identity as assessed by ex vivo and in vitro experiments. In the next section, we examined the relationship between Qkf gene activity and NSC identity in vivo.

The SVZ neurogenic region contains cell types with different cell cycle characteristics. Transit amplifying cells and neuroblasts are thought to divide approximately every 12–18 hours, whereas NSCs divide once every 14–15 days (Craig et al., 1999; Morshead et al., 1998). Cells can be identified by their cell cycle length using both the thymidine analogue BrdU, which is incorporated into cellular DNA during S phase of the cell cycle, and Ki67, which is an endogenous marker for actively proliferating cells in G1, S, G2 or M phase. To mark fast dividing cells (neuroblasts and transit amplifying cells), four BrdU injections were administered over 4 hours, after which the SVZ cells were sorted, spun onto slides and stained for BrdU. Although there was a positive correlation between Qkf promoter activity and the number of short-term BrdU-labelled cells (r=0.6962, P=0.0027; Fig. 3E), the number of cells showing short-term incorporation of BrdU was not significantly different between the three strongest Qkf–GFP-expressing fractions (P<0.15, n=4). Only the Qkf–GFP^Lo fraction contained significantly fewer cells in S phase than the Qkf–GFP^Hi and Qkf–GFP^MedHi fractions (P=0.0066 and 0.0219). All four fractions contained comparable numbers of Ki67-positive cycling cells (P<0.10, n=3; Fig. 3F). Therefore, fast-cycling cells are present in all four Qkf–GFP fractions.

To assess the number of the infrequently dividing NSCs, we carried out a long-term BrdU incorporation experiment. Mice were injected twice daily with BrdU for 2 weeks, followed by 1 week without treatment. BrdU is distributed to daughter cells through cell division, and thereby becomes more dilute as cells divide. Therefore, any cells positive for BrdU after 1 week without treatment are slow-dividing or terminally differentiated cells. Following the 3-week BrdU labelling procedure, SVZ cells were sorted on the basis of their Qkf–GFP reporter expression, and stained for BrdU. Approximately 10% of Qkf–GFP^Hi cells were positive for BrdU, which was significantly more than in the Qkf–GFP^MedHi and the Qkf–GFP^MedLo fraction (P<0.05, n=3; Fig. 3G). Interestingly, 8% of Qkf–GFP^Lo cells were also positive for BrdU after the 3-week procedure. However, on the basis of the colony-forming and proliferation assays (Figs 2, 3), these cells lacked the ability to give rise to neurosphere colonies or to divide more than five times in vitro. Therefore, it is probable that the long-term BrdU-incorporating cells in the Qkf–GFP^Lo fraction were terminally differentiated cells. Remarkably, this shows that the level of Qkf–GFP expression distinguishes between long-term BrdU-labelling NSCs and terminally differentiated cells. In conclusion, the Qkf–GFP^Hi fraction contains the great majority of slow dividing, self-renewing, neurosphere colony-forming cells, whereas all four Qkf–GFP fractions contain some other type of proliferating cells.

NSCs are widely considered to be astrocyte-like cells in the subependymal layer of the SVZ, which express the mature astrocyte marker GFAP (Doetsch et al., 1999; Garcia et al., 2004; Morshead et al., 2003a). In addition, heat stable antigen (HSA, CD24), which is expressed on the surface of ependymal cells and neuroblasts (Calaora et al., 1996), is expressed at low levels in neurosphere-forming cells (Rietze et al., 2001). This is consistent with the observation that NSCs reside in the subependymal layer of the SVZ (Chiasson et al., 1999; Doetsch et al., 1999; Laywell et al., 2000). SSEA-1 has been used for the enrichment of NSCs from the adult SVZ and embryonic brains. Between 12.5% and 25% of sorted SSEA-1-positive cells form neurospheres (Capela and Temple, 2002; Capela and Temple, 2006). In addition to a GFAP-positive NSC population in the subependymal layer, a smaller, more quiescent population of NSCs expressing CD133 has been reported (Coskun et al., 2008; Beckervordersandforth et al, 2010).

To determine the cellular identity of the four Qkf–GFP fractions, we examined the expression of specific SVZ cell markers either by flow cytometry or by immunofluorescent staining and counting of acutely sorted SVZ cells. There was a strong positive correlation between Qkf–GFP expression and the proportion of GFAP-positive (r=0.7818, P=0.0003, n=4; Fig. 4A), HSA^Lo (r=0.7964, P<0.0001, n=6; Fig. 4B), SSEA-1-positive (r=9079, P<0.0001, n=7; Fig. 4C) and CD133-positive (r=0.7242, P=0.0077, n=3; Fig. 4D) cells. In particular, 65% of sorted events in the Qkf–GFP^Hi population were positive for GFAP, which was threefold higher than the Qkf–GFP^MedHi, and sixfold higher than the Qkf–GFP^MedLo and Qkf–GFP^Lo fractions (P<0.0001; Fig. 4A). Similarly, approximately 80% of sorted events in the Qkf–GFP^Hi and Qkf–GFP^MedHi fractions expressed HSA at low levels, which was significantly more than the two lowest Qkf–GFP fractions (P<0.0001; Fig. 4B). Fifty-five percent of sorted events in the Qkf–GFP^Hi fraction strongly expressed SSEA-1, which was significantly more than the Qkf–GFP^MedHi, Qkf–GFP^MedLo and Qkf–GFP^Lo fractions (Fig. 4C; P<0.0001). We found relatively few CD133-positive cells in the SVZ. However, there was approximately sevenfold enrichment for CD133-positive cells in the Qkf–GFP^Hi fraction, over the Qkf–GFP^MedLo (P=0.0149; Fig. 4D) and Qkf–GFP^Lo (P=0.0099) fractions. This suggests that high Qkf promoter activity enriches for all cells expressing NSC markers rather than a specific sub-population of stem cells. Altogether, our cell marker analyses show that the SVZ fraction with the highest Qkf–GFP expression is highly enriched for the full range of stem cell markers (GFAP positive, HSA^Lo, SSEA-1 positive and CD133 positive).

To determine which fractions are likely to contain neuroblasts, we utilized polysialylated neural cell adhesion molecule (PSA-NCAM), a polysialylated cell surface protein specifically expressed on neuroblasts (Doetsch et al., 1997). The largest proportion of PSA-NCAM-positive cells was found in the Qkf–GFP^MedLo fraction (P<0.005, n=3; Fig. 4E), suggesting that neuroblasts were highly enriched in this Qkf–GFP^MedLo population. Interestingly, we observed a high overlap between PSA-NCAM expression levels and nestin staining (Fig. 4F, n=3), an intermediate filament that is widely expressed in the SVZ (Doetsch et al., 1997). Our results suggest that endogenous nestin is most strongly expressed in neuroblasts and weakly in other SVZ cell types.

Ependymal cells are differentiated cells that line the lateral ventricle and provide nutrients for cells within the SVZ NSC niche. While HSA is weakly expressed on the surface of NSCs and progenitor cells, it is strongly expressed on ependymal cells and neuroblasts (Calaora et al., 1996). To determine whether ependymal cells were enriched in any of the Qkf–GFP fractions, we screened for the presence of HSA expression, and absence of PSA-NCAM staining in the four Qkf–GFP populations (see Fig. 4Bii-v for HSA^Hi population and E for PSA-NCAM^Hi population). Only 9.8% and 11.5% of Qkf–GFP^Hi and Qkf–GFP^MedHi sorted events, respectively, stained strongly for HSA, which was significantly less than the Qkf–GFP^MedLo (53.6%) and Qkf–GFP^Lo (65.4%) fractions (P<0.0001, n=6; Fig. 4B). Migrating neuroblasts coexpress HSA and PSA-NCAM (Calaora et al., 1996). Because 43% of the sorted events in the Qkf–GFP^MedLo fraction were PSA-NCAM positive, it is probable that the majority (43 of 53.6, i.e. 80%) of the HSA-positive cells in this fraction were PSA-NCAM–HSA double positive neuroblasts. By contrast, only a small proportion of FACS sorted events in the Qkf–GFP^Lo fraction were PSA-NCAM positive (16%), whereas strong HSA staining was common (65.4% of all FACS events). Therefore, HSA-positive, PSA-NCAM-negative ependymal cells are likely to be enriched in the Qkf–GFP^Lo population.

Altogether, our cell marker analyses suggest that the Qkf–GFP^Hi fraction contained most of the NSCs, the Qkf–GFP^MedHi fraction was enriched for transit amplifying cells, and the Qkf–GFP^MedLo fraction mostly contained neuroblasts, whereas the Qkf–GFP^Lo population was enriched for ependymal cells.

Our analyses have revealed that the acutely isolated Qkf–GFP^Hi population was enriched for stem cell markers and was highly enriched for self-renewing neurosphere-colony forming cells. To further investigate the overlap between high Qkf expression and cell identity in vivo, whole-mount staining of SVZ tissue dissected from Qkf–GFP animals was carried out (n=16 animals) as described previously (Mirzadeh et al., 2010; Shen et al., 2008). The results are shown in Fig. 5A and all images depict the anterior-dorsal SVZ in the lateral wall of the lateral ventricle. The ependymal surface of the SVZ was marked using a combination of β-catenin and γ-tubulin as previously described (Mirzadeh et al., 2008). The typical honeycomb architecture of the ependymal surface was observed. A small number of GFAP-positive stem cell processes were intercalated in the ependymal layer (Fig. 5D, arrows). Consistent with cell surface marker studies, there was very little Qkf–GFP expression in the ependymal layer (Fig. 5D). Immediately below the ependymal layer, more GFAP-positive cell processes and a small number of cell bodies were evident (Fig. 5E–H). Sparse Qkf–GFP expression was found to overlap with cell bodies of GFAP-positive cells. Deeper in the subependymal layer (Fig. 5I–T), strong overlap was observed between GFAP-positive stem cells and strong Qkf–GFP-expressing cells (Fig. 5I,J,L,O,P,R). Analysis of ten high-magnification confocal images taken in the anterior-dorsal region of the SVZ (Fig. 5A) across four different animals revealed that 88% of GFAP-positive cells in the subependymal zone (i.e. 44 out of 50) overlapped with strong Qkf–GFP expression (e.g. Fig. 5L,R). This is consistent with flow cytometry data, which showed that 87% of GFAP-positive SVZ cells are found in the Qkf–GFP^Hi and Qkf–GFP^MedHi populations. Similarly, strong SSEA-1 expression was found in 66% (42 out of 64) of GFAP-positive cells (Fig. 5I,K,M,O,Q,S). Remarkably, 94% (n=35) of cells that were double positive for SSEA-1 and GFAP were found to express Qkf–GFP strongly (Fig. 5N). This overlap between strong Qkf–GFP, GFAP and SSEA-1 expression was unique, because GFAP and strong Qkf–GFP expression was not found in ependymal cells (Fig. 5B–D) or in PSA-NCAM-positive neuroblasts (Fig. 5U–W). Together, these data reiterate that strong Qkf expression was largely confined to GFAP- and SSEA-1-double-positive stem cells.

Another defining criterion for stem cells is multipotency, i.e. the ability to give rise to a range of progeny. When induced to differentiate, NSCs give rise to all three major neural cell types, namely neurons, astrocytes and oligodendrocytes. To elucidate the correlation between Qkf–GFP transgene expression and multipotency, SVZ cells were sorted and collected on the basis of their levels of Qkf–GFP expression, grown as neurospheres for 7 days and then allowed to differentiate for 7 days before staining for markers of neurons (βIII-tubulin), oligodendrocytes (O4) and astrocytes (GFAP). At total of 25 to 30 neurospheres were differentiated from each Qkf–GFP fraction from each mouse (SVZ cells were sorted from n=3 animals). The exception was the Qkf–GFP^Lo fraction, as it rarely gave rise to neurospheres.

Only the Qkf–GFP^Hi and Qkf–GFP^MedHi fractions generated all three neuronal cell types (Fig. 6A–H). Qkf–GFP^MedLo neurospheres failed to give rise to oligodendrocytes (Fig. 6I–L), and the Qkf–GFP^Lo neurospheres only formed astrocytes (Fig. 6M–P). To determine the proportion of Qkf–GFP^Hi and Qkf–GFP^MedHi colonies that were multipotent, individual neurosphere colonies were randomly picked and differentiated separately. Twenty-five of forty-two (60%) Qkf–GFP^Hi colonies (n=6 animals) were found to be tripotent, eight (19%) were bipotent and nine (21%) only formed astrocytes. This is significantly more than the number of Qkf–GFP^MedHi colonies (six out of eighteen, i.e. 33%; P=0.011) that were tripotent. Three out of eighteen (17%) Qkf–GFP^MedHi colonies were bipotent, and nine out of eighteen (50%) colonies only formed astrocytes. In short, the majority of Qkf–GFP^Hi-derived neurospheres were tripotent, whereas the majority of Qkf–GFP^MedHi-derived neurospheres were not tripotent. These data suggest that multipotency, as determined using the neurosphere assay, is correlated with high Qkf–GFP expression and thus Qkf serves as a marker for multipotent SVZ cells.

Bmi1, a member of the polycomb group repressive complex, is important for maintaining self-renewing NSCs through inhibition of the Ink4a and Arf locus (Molofsky et al., 2003). Owing to a self-renewal defect, the number of NSCs decreases in Bmi1^−/− mice in an age-dependent manner. When Qkf–GFP cell populations were analysed in Bmi1 mutant and wild-type littermates, a reduction in the Qkf–GFP^Hi population was observed in the Bmi1^−/− mice, which declined considerably with age (Fig. 7). This suggests that high Qkf–GFP levels, which mark self-renewing multipotent NSCs, accurately report the time-dependent decrease in NSCs previously described in Bmi1 knockout mice.

In this study, we have utilized Qkf–GFP BAC transgenic mice to examine the expression levels of Qkf in the adult SVZ NSC niche, with respect to stem cell characteristics (summarized in Fig. 8 and supplementary material Table S18).

The definition of a stem cell, as coined by Potten (Potten and Loeffler, 1990), designates stem cells as being (a) undifferentiated, (b) proliferative, (c) self-maintaining (i.e. self-renewing) and (d) able to give rise to a range of functional progeny (i.e. multipotent). Utilizing the neurosphere assay, widely used and accepted as a good test of NSC characteristics (Capela and Temple, 2002; Doetsch et al., 1999; Jensen and Parmar, 2006; Morshead et al., 2003b; Reynolds and Weiss, 1992), we have shown that undifferentiated, proliferative, self-renewing and multipotent cells are largely confined to approximately the 1000–1500 strongest Qkf-expressing cells in the SVZ. In addition, we utilized flow cytometry and confocal microscopy of SVZ whole mounts, to demonstrate that the Qkf–GFP^Hi fraction was highly enriched for the reported stem cell markers GFAP^Pos, HSA^Lo, SSEA-1^Pos and CD133^Pos, and not for markers of ependymal cells or neuroblasts (Capela and Temple, 2002; Coskun et al., 2008; Beckervordersandforth et al, 2010; Garcia et al., 2004; Mirzadeh et al., 2008; Morshead et al., 2003a; Rietze et al., 2001; Shen et al., 2008). Exploiting Bmi1 mutant mice, which have a severe defect in NSC self-renewal (Molofsky et al., 2003), we have shown that there is an age-dependent decline in the Qkf–GFP^Hi population in the Bmi1^−/− mutant mice. Consistent with our conclusion that adult NSCs have the highest requirements for Qkf expression, we have previously shown that QKF is required for normal numbers of adult NSCs (Merson et al., 2006; Rietze et al., 2001), and is essential to promote NSC self-renewal and neuronal differentiation (Merson et al., 2006). We have shown that high level Qkf expression is restricted to the NSCs (and not their cellular neighbours) and that a high-level Qkf expression is essential for long-term self-renewal and multipotency. NSCs are greatly reduced in number in vivo and ex vivo in QKF-deficient animals. Importantly, NSC capacity, which is confined to the strongest QKF-expressing cells, as outlined in this study, declines with age in QKF-deficient animals relative to wild-type controls, consistent with a progressive self-renewal defect in Qkf mutant animals (Merson et al., 2006). Furthermore, neurospheres generated from QKF-deficient animals form fewer neurons than wild-type littermates (Merson et al., 2006), reflecting the lack of neuron production found in Qkf–GFP^Lo neurospheres in the current study. Accordingly, our data suggest that high Qkf promoter activity within NSCs is required to maintain the essential stem cell characteristics of self-renewal and multipotency.

In addition to all NSC characteristics and markers being confined to the strongest Qkf-expressing cells, our data showed a remarkable relationship between Qkf-promoter activity and cell identity within the NSC niche. Although our data suggest that there are also a number of stem cells in the Qkf–GFP^MedHi population, there was enrichment for colony-forming cells with limited self-renewal and proliferative capacity, suggesting that transit amplifying cells are enriched in this Qkf–GFP^MedHi fraction. Until recently, transit amplifying cells have been difficult to isolate and distinguish from NSCs using cell surface markers. Utilizing GFAP–GFP mice, Pastrana and co-workers showed that SVZ cells positive for epidermal growth factor (EGF) receptor, negative for HSA (CD24), with low-level expression of GFAP have transit amplifying cell characteristics (Pastrana et al., 2009). Remarkably, we also obtain a high degree of enrichment of transit amplifying cells solely on the basis of medium-high expression of Qkf. Interestingly, nearly half of the sorted cells in the Qkf–GFP^MedLo fraction were PSA-NCAM positive, which is a marker for neuroblasts (Doetsch et al., 1997). These cells are proliferating, lineage-restricted neural precursors with limited proliferative capacity that migrate from the SVZ to the olfactory bulbs (Doetsch and Alvarez-Buylla, 1996). Our data show that this Qkf–GFP^MedLo fraction also contains the highest percentage of nestin and PSA-NCAM-positive cells, suggesting that migrating, proliferating neuroblasts are enriched in the nestin-positive fraction. While other reports suggest that migrating neuroblasts are indeed nestin positive (Doetsch et al., 1997), NSCs have also been identified as nestin-expressing cells (Reynolds and Weiss, 1992), and various astrocyte subpopulations are reported to express nestin at the highest level within the SVZ neurogenic region (Doetsch et al., 1997). Together with the current study, these data suggest that endogenous nestin is not a unique marker for any specific SVZ subpopulation. Lastly, HSA-positive, PSA-NCAM-negative cells were highly enriched in the Qkf–GFP^Lo fraction and they are likely to represent the ependymal cell population, which has been reported to be PSA-NCAM negative and HSA high (Calaora et al., 1996). Our data document a unique, proportional relationship between Qkf expression levels and cell identity in the SVZ, where the strongest Qkf-expressing cells show all stem cell characteristics, which are gradually lost along with decreasing Qkf promoter activity. Together with our previous reports on the deficiencies in NSC establishment and self-renewal in the QKF-deficient state, we suggest that QKF serves both as a NSC intrinsic regulator and as a marker of ‘stemness’ – the self-renewal potential and multipotency of SVZ cells.

Interestingly, when Qkf–GFP expression was assayed in a Bmi1-null background, a significant decrease in the Qkf–GFP^Hi population was observed. Self-renewing multipotent NSCs decrease in an age-dependent manner in Bmi1 mutant mice because of defects in self-renewal (Molofsky et al., 2003). Given that the Qkf–GFP^Hi population decreased in Bmi1 mutant mice compared with littermate controls as mice aged, this finding supports our hypothesis that Qkf–GFP expression marks and positively correlates with self-renewing and multipotent SVZ cells. In addition, this experiment suggests that Qkf–GFP transgenic mice can be used to screen a large number of mice for potential defects in NSC self-renewal or multipotency in vivo.

QKF is part of the MYST family of histone acetyltransferases. These epigenetic regulators have an essential function in the control of gene expression through their ability to acetylate histones and modify chromatin structure (Thomas and Voss, 2007; Wang et al., 2009). QKF has an identical domain structure to monocytic leukaemia zinc finger protein (MOZ), and all protein domains are highly conserved between these two transcriptional regulators (Voss and Thomas, 2009). Furthermore, the amino acid sequence within the domains is highly similar. We and others have shown that MOZ has an essential function in the maintenance of haematopoietic stem cells (Katsumoto et al., 2006; Thomas et al., 2006). More recently, it has been shown that the role of MOZ in haematopoietic stem cells is dependent on its histone acetyltransferase activity (Perez-Campo et al., 2009), directly demonstrating the importance of chromatin regulation by this subclass of histone acetyltransferases in stem cells. MOZ was originally identified in recurrent translocations leading to acute myeloid leukaemia (Borrow et al., 1996). Several different chromosomal translocations causing leukaemia target the MOZ gene (Aguiar et al., 1997; Carapeti et al., 1998; Chaffanet et al., 2000; Chaffanet et al., 1999; Esteyries et al., 2008; Imamura et al., 2003; Kitabayashi et al., 2001; Rozman et al., 2004), and indeed, translocations targeting the human homologue of Qkf (MYST4/KAT6b) also cause leukaemia (Panagopoulos et al., 2001; Vizmanos et al., 2003). By contrast, deficiency in QKF (MYST4/MORF) leads to intellectual disability in humans (Kraft et al., 2011) and brain developmental defects in mice (Thomas et al., 2000). Taken together, these studies show that QKF is a potent regulator of stem-cell characteristics, where loss of function causes brain developmental defects, and gain of function causes malignancy.

Although we are only beginning to understand the role of chromatin structure in establishing and maintaining stem cells, recent studies have demonstrated that stem cell chromatin is unique and actively modified as stem cells differentiate (Bernstein et al., 2006; Mikkelsen et al., 2007). The role of histone acetylation has been studied in embryonic stem cells. Recently, Ware and co-workers have shown that the histone deacetylase inhibitor sodium butyrate was able to maintain self-renewing mouse and human embryonic stem cells without feeder layers or exogenous factors, highlighting the importance of histone acetylation in maintaining a stem cell state (Ware et al., 2009). Interestingly, we have recently shown that the QKF-related protein MOZ is a key regulator of Hox gene expression in vivo. MOZ is required to maintain the correct levels of histone 3 lysine 9 (H3K9) acetylation at Hox loci during development, and a reduction in MOZ levels leads to aberrant and reduced Hox gene expression and an extensive anterior homeotic transformation of the nervous system and the axial skeleton (Voss et al., 2009). The role of H3K9 acetylation in maintaining chromatin structure extends to NSCs. When oligodendrocyte precursors were redirected to become NSC-like self-renewing multipotent cells, the methylated H3K9 was converted to acetylated H3K9 at the promoter of Sox2 (Kondo and Raff, 2004), a gene essential for maintaining NSCs (Graham et al., 2003). Our data presented here, together with our previous results showing that Qkf mutant mice have severe and progressive defects in adult neurogenesis, suggests that QKF plays a central role in regulating histone acetylation at important stem cell loci within SVZ NSCs.

In conclusion, we have shown that it is possible to isolate the four major populations of the SVZ neurogenic zone, i.e. NSCs, transit amplifying cells, neuroblasts and ependymal cells, on the basis of Qkf promoter activity levels. This system will allow the characterization of neural lineage characteristics including the epigenetic mechanisms that regulate NSC self-renewal and lineage commitment in adult neurogenesis.

The Qkf–GFP BAC construct used for the generation of transgenic mice was generated by homologous recombination in Escherichia coli using the method described by Gong et. al. (Gong et al., 2002). The eGFP coding sequence with its Kozak sequence and poly adenylation signal from pLD53SCA-E-B (a gift from Nathaniel Heintz, Rockefeller University, New York, USA) were inserted into the first coding exon of the Qkf gene on BAC clone RP23-121O15 replacing the Qkf Kozak sequence and start codon. Experiments were performed on male mice aged between 7 and 12 weeks, and conformed to the regulations set by the Australian NHMRC and the Royal Melbourne Hospital Research Foundation. Transgenic mouse lines were maintained as heterozygous for the Qkf–GFP transgene and genotyped by the presence of GFP expression in tail biopsies.

In situ hybridization was carried out with ³⁵S-labelled dCTP (Amersham) using a method described previously (Thomas et al., 2000). Northern blot analysis was performed using standard molecular biology techniques. Qkf probe no. 19 hybridizes to bases 41–1246 of sequence GenBank accession number AF222800, Qkf probe no. 1332 to bases 4019–5633 of AF222800, and the GFP probe to bases 613–1276 of sequence U55763. To determine endogenous Qkf levels in sorted populations, oligonucleotides forward (F) 5′-GCGAATCTCTATGGGTAACG-3′ and reverse (R) 5′-GCACTGCTTCAAGAATCCAC-3′ amplifying from the 5′UTR to exon 1 were used. Expression levels were standardized to 18S (F5′-TCGGAACTGAGGCCATGATT-3′, R5′-CCTCCGACTTTCGTTCTTGATT-3′), Hsp90ab1 (F5′-CCTCCGACTTTCGTTCTTGATT-3′, R5′-AGAATCCGACACCAAACTGC-3′) and Gapdh (F5′-TTCACCACCATGGAGAAGGC-3′, R5′-CCCTTTTGGCTCCACCCT-3′).

SVZ from Qkf–GFP transgenic mice were dissected and enzymatically dissociated using pancreatin–trypsin (2.5% pancreatin, 0.5% trypsin). The tissue was mechanically dissociated using a P200 Gilson pipette. Cells were immediately washed in 0.2% BSA in mouse tonicity (MT-)PBS, passed through a 40 μm sieve (Falcon) and collected by centrifugation. Cells were resuspended in 200–300 μl 0.2% BSA in MT-PBS containing 1 μg/ml propidium iodide (Sigma P-4170) to label dead cells. Cells were sorted into four fractions on the basis of their GFP reporter expression using the FACS DiVA (BD Biosciences) and collected in 2 ml 0.2% BSA in MT-PBS. After collection by brief centrifugation, cells from each of the four Qkf–GFP fractions were suspended in 5 ml NSC proliferation medium [medium described previously (Merson et al., 2006; Voss et al., 2006)] in a 3.5 cm dish. Neurospheres were counted 3–4 days after sorting. For clonal assays, cells were sorted at a density of one event per well of a 96-well plate containing 150 μl NSC proliferation medium, and counted 7 days after plating.

For passaging, neurospheres were collected by brief centrifugation (7 minutes, 120 g), dissociated mechanically using a P1000 Gilson pipette in approximately 900 μl fresh proliferation medium. Dissociated cells were counted using a haemocytometer and plated at a density of 15,000–30,000 cells per cm² depending on the number of cells available, as described previously (Merson et al., 2006; Voss et al., 2006).

To determine lineage potential, Qkf–GFP SVZ cells were sorted and cultured as neurospheres for 7 days. After 1 week in culture, 25–30 neurospheres from each Qkf–GFP fraction were induced to differentiate on laminin-treated coverslips in differentiation medium as previously described (Merson et al., 2006). To determine the lineage potential of individual neurospheres, colonies were picked and plated individually in a 24-well plate. Neurosphere cells were allowed to differentiate for 1 week, washed, fixed in 4% paraformaldehyde, blocked in 10% normal goat serum, and stained for astrocytes (GFAP; Dako Z0334; 1:500), oligodendrocytes (O4; MAB345; 1:500) and neurons (βIII-tubulin; Promega G7121; 1:2000).

For cell surface marker analysis, SVZ cells were prepared for flow cytometry and stained with antibodies raised against GFAP (Chemicon MAB3402; 1:500), HSA (mCD24-PE conjugated; Vector B-1075; 1:200), SSEA-1 (BD Biosciences no. 347420; 1:20), CD133 (APC conjugated; eBioscience no. 17-1331-81; 1:1000) and PSA-NCAM (AbCys AbC0019; 1:400). Alexa Fluor 633 and APC-conjugated anti-mouse IgM and IgG antibodies were used where required. Cells were sorted on the FACS LSR II (BD Biosciences). Sorts were analyzed using FlowJo v8.7 software.

Whole-mount SVZ staining was carried out as described previously (Mirzadeh et al., 2010). Briefly, Qkf–GFP animals were perfusion fixed. The SVZ was dissected and fixed overnight at 4°C in 4% paraformaldehyde. After washing and blocking, the SVZ was incubated with primary antibodies against GFP [using an antibody that was conjugated to biotin (Invitrogen A10263)], GFAP, SSEA-1, PSA-NCAM, β-catenin (Sigma-Aldrich C2206; 1:1000) and γ-tubulin (Sigma-Aldrich T5192; 1:1000) for 48 hours at 4°C. After washing, Alexa Fluor secondary antibodies raised against mouse IgG, IgM and rabbit IgG (Invitrogen Alexa Fluor 633 goat anti-mouse IgG A21052; Alexa Fluor 546 goat anti-rabbit IgG A11035; Alexa Fluor 546 goat anti-mouse IgM A21045; used at 1:500) and FITC–streptavidin (Invitrogen SA1001; 1:500) were applied for 48 hours. Samples were washed, mounted and imaged on the Leica SP2 laser scanning microscope system. Images were processed using the IMARIS v7.2 software.

For BrdU (BioScience Products 010198; 1:10), Ki67 (Novocastra NCL-Ki67p; 1:400) and nestin (Chemicon MAB353; 1:200) staining of SVZ cells collected by FACS, cells were cytospun onto gelatine-coated slides and immunostained as described previously (Voss et al., 2006). Cells were counted and photographed using a compound microscope (Axioplan 2, Zeiss).

Cells were collected by centrifugation and suspended in 40 μl 0.2% BSA in MT-PBS. Trypan Blue (10 μl, Sigma) was added, mixed, and 10 μl of the cell suspension was added to each of four areas of a haemocytometer and counted.

SVZ cells from Qkf–GFP animals were sorted as described above. For the neural colony-forming cell assay, cells were suspended in collagen and grown according to manufacturer's instructions (Stemcell Technologies, Vancouver, Canada). Colonies were scored 3 weeks after plating. To determine neurosphere size in bulk cultures, neurospheres were grown for 4 days, after which the size of each neurosphere in each of the Qkf–GFP fractions was determined using a scale in the eyepiece of an inverted microscope (10×objective, 10 × magnification) converted to metric scale using a haemocytometer.

Statistical analyses were performed using StatView v5.0 and Intercooled Stata v10. Data are presented as means ± s.e.m. and were analyzed using one or two-factorial analysis of variance (ANOVA) with Qkf–GFP expression levels with or without the specific transgenic founder mouse line or the culture passage number as the independent factors followed by Fisher's or Bonferroni's post-hoc test. Pearson's regression analysis was carried out where appropriate, followed by a two-tailed t-test.

We thank C. Gatt, M. Becroft, T. McLennan, N. Downer and the WEHI FACS and Imaging laboratories for excellent technical support. We are grateful for the gift of the eGFP plasmid pLD53SCA-E-B from N. Heintz, for the β-actin–GFP transgenic mice from A. Nagy and the Bmi1 mutant mice from M. van Lohuizen.