EphA4 preserves postnatal and adult neural stem cells in an undifferentiated state in vivo

Postnatal neurogenesis in mammals persists in two brain regions: the subgranular zone (SGZ) of the dentate gyrus in the hippocampus and the subventricular zone (SVZ) of the lateral ventricles (Lledo et al., 2006; Ninkovic and Gotz, 2007; Zhao et al., 2008). Under pathological conditions, new neuroblasts also migrate into injured cortex and striatum (Kreuzberg et al., 2010; Parent et al., 2002). Neural stem cells (NSCs) in the SGZ and SVZ divide slowly and are preserved throughout life. Little is known about the cellular signaling that guarantees that NSCs are maintained in an undifferentiated state. Postnatal NSCs divide asymmetrically and differentiate into transit-amplifying precursors, which are fast-dividing cells that, in turn, give rise to more differentiated neuroblasts (Abrous et al., 2005; Lledo et al., 2006; Zhao et al., 2008). Neuroblasts in the hippocampal SGZ migrate into the granule cell layer of dentate gyrus and integrate into previously established neural circuits, where they are thought to play a role in certain forms of hippocampal-dependent learning and memory (Aimone et al., 2009; Kempermann et al., 2004). Neuroblasts originating in the SVZ migrate over long distances via the rostral migratory stream (RMS) to the olfactory bulb, where they mature into granule or periglomerular neurons and contribute to olfactory information processing (Alonso et al., 2006; Lledo and Saghatelyan, 2005).

Ephrins and their Eph tyrosine kinase receptors are important regulators of many developmental processes (Klein, 2004; Pasquale, 2005). Receptors and ligands alike are grouped into two subclasses: type A and type B. Although, in general, ephrin A ligands bind to EphA receptors and ephrin B ligands to EphB receptors, cross-specificity has been described for some members (Klein, 2004; Pasquale, 2004). Several ephrins and their Eph receptors regulate different stages of neurogenesis and are differentially expressed on distinct cell types in the neurogenic niches (Chumley et al., 2007; Conover et al., 2000; Depaepe et al., 2005; Furne et al., 2009; Holmberg et al., 2005; Jiao et al., 2008; Qiu et al., 2008; Ricard et al., 2006). For instance, ephrins B2 and B3 and their receptor EphB1 cooperatively regulate proliferation and survival of neural progenitor cells and migration of neuroblasts in the SVZ/RMS and SGZ (Chumley et al., 2007; Conover et al., 2000; Ricard et al., 2006). Two other Eph receptors, namely, EphB2 and EphB3, have opposite function in the adult SVZ. Whereas EphB2 induces proliferation in the SVZ (Katakowski et al., 2005), EphB3 suppresses progenitor cell proliferation via a p53-dependent mechanism (Theus et al., 2010). In the SVZ, the complementary expression of ephrin A2 in transit-amplifying cells and neuroblasts and that of EphA7 on ependymal cells and NSCs controls neural progenitor cell proliferation (Holmberg et al., 2005). In another study, EphA7 was shown to be involved in the modulation of apoptosis of neural progenitors during embryonic development (Depaepe et al., 2005). Also, ephrin B1 has been demonstrated to contribute to the maintenance of NSCs in an undifferentiated state (Qiu et al., 2008).

We recently described the generation of transgenic mice in which EGFP (enhanced green fluorescent protein) is expressed in the entire RMS (Inta et al., 2008), and optimized a procedure for RNA isolation from in vivo fluorescent RMS neuroblasts (Khodosevich et al., 2007). The RMS is a robust and hence an ideal structure in the postnatal brain that allows the harvesting of sufficient cells to perform microarray analysis. Using transgenic 5HT3A-EGFP mice (where EGFP is expressed under the control of the serotonin receptor 3A promoter) with clearly EGFP-labeled RMS, we isolated neuroblasts from two distinct locations: one in the immediate vicinity of the SVZ (posterior RMS, pRMS), and the other more rostral, closer to the bulb (anterior RMS, aRMS) (Khodosevich et al., 2009). It is safe to assume that upregulated genes in cells from the aRMS are most probably involved in migration and differentiation, whereas upregulated genes in cells from the pRMS that is closer to the SVZ might affect cell generation.

We identified Epha4, the gene coding for the Eph receptor A4, as one of the upregulated genes in the pRMS. The upregulation was comparable with that of other genes known to be expressed in NSCs (e.g. GFAP, Nr2e1, Nes). Epha4 mRNA expression levels were below background expression in samples collected from the aRMS or from the periglomerular region of the olfactory bulb.

EphA4 plays an important role in the developing and postnatal brain, including axon guidance (Egea et al., 2005; Gallarda et al., 2008; Wegmeyer et al., 2007), synaptic plasticity (Bourgin et al., 2007; Deininger et al., 2008; Fu et al., 2007; Inoue et al., 2009) and proliferation of precursor cells during cortical neurogenesis (North et al., 2009). Furthermore, expression of Epha4 was found in the adult SVZ (Conover et al., 2000; Liebl et al., 2003). Indeed, there is an increase in the number of neuroblasts in the SVZ and RMS in EphA4 knockout animals (Furne et al., 2009). Interestingly the number of proliferating cells in the SVZ (identified by Ki67 expression) is not changed and there is a decrease in TUNEL-positive (i.e. apoptotic) cells in the adult SVZ in these animals (Furne et al., 2009; Ricard et al., 2006).

In this study we provide evidence that EphA4 is expressed in the adult neurogenic niches exclusively by NSCs, but not by other cell types. In neurosphere cultures and in vivo, knockdown of Epha4 expression caused reduction in cell proliferation and premature differentiation of NSCs.

To establish the cellular expression of EphA4 in the postnatal brain, we employed double- and triple-immunohistochemistry in brain sections from 3-month-old animals. In both adult neurogenic niches, the SVZ and SGZ, EphA4 co-localizes with GFAP (Fig. 1A–C). Furthermore, GFAP–nestin double-positive cells that are bona fide NSCs, at least in the SGZ (Lledo et al., 2006; Zhao et al., 2008), express EphA4 (Fig. 1B for SVZ; Fig. 1C,C′ for SGZ). In SVZ, ependymal cells also exhibit coexpression of GFAP and nestin (Doetsch et al., 1997). However, EphA4 expression does not co-localize with the ependymal cell marker S100β (supplementary material Fig. S1). Because the EphA4 antibody almost exclusively labels cell processes, EphA4 co-labeling with other nuclear or cell body markers was determined by analyzing series of confocal stacks (supplementary material Fig. S2A). To avoid cross-reactivity between secondary antibodies, we used rabbit anti-EphA4, mouse anti-GFAP and chicken anti-nestin antibodies (see Materials and Methods). Furthermore, stainings on separate sections with anti-EphA4 and GFAP antibodies resulted in similar labeling patterns (supplementary material Fig. S2B,C). EphA4-positive cells coexpress Sox2, another neural stem cell marker (Fig. 1D and supplementary material Fig. S3A–D).

EphA4 is not present in transit-amplifying precursors and glial precursors or oligodendrocytes, as revealed by the lack of coexpression with Mash1 and Olig2, respectively (supplementary material Fig. S4A,B). EphA4 is also not co-localized with the neuroblast markers DCX and PSA-NCAM (supplementary material Fig. S4C,D).

We analyzed the proliferating capacity of EphA4-positive cells taking recourse to labeling with 5-bromo-2′-deoxyuridine (BrdU), a marker of dividing cells. After a single injection of BrdU, none of the EphA4-positive cells retained BrdU (Fig. 1E). However, EphA4-positive cells were labeled when BrdU was administered for 1 week followed by 3 weeks with no treatment before sacrificing the animals (Fig. 1F; supplementary material Fig. S3E–H), which indicates that EphA4 is expressed in slow-dividing cells, but not in fast-dividing cells.

The phenotype of EphA4-positive cells in the SVZ is summarized in supplementary material Table S1.

To study the cellular function of EphA4 in NSCs of the postnatal SVZ, we conducted knockdown experiments using two Epha4 short hairpin RNA (shRNA) constructs (shRNA Epha4-1 and shRNA Epha2) (see Fig. 2E; supplementary material Fig. S5) and analyzed NSC proliferation in neurosphere cultures after Epha4 knockdown. We used neurosphere cultures obtained from the SVZ of P4 mice that were infected with retroviruses expressing red fluorescent protein and either control shRNA (scrambled) or one of the two Epha4 shRNAs. After 5 days in culture, BrdU was added to the culture media and cells were fixed after another 12 hours. In the presence of either of the two Epha4 shRNAs, there was a significant decrease in the number of infected proliferating cells in neurospheres (green and red double-positive cells in Fig. 2A,B, quantification in Fig. 2D). Details regarding cell counts in neurospheres are described in Materials and Methods. Knockdown of Epha4 in neurospheres was already pronounced at 4 days in vitro (DIV4) (supplementary material Fig. S6A,B) and by DIV5 was at most 25% of that in control neurospheres (see Fig. 2E for western-blot analysis and supplementary material Fig. S6C,D for immunocytochemistry).

The effect of Epha4 knockdown was completely rescued by coexpression of a modified EphA4, containing three silent mutations that make it resistant to shRNA Epha4-1 (Fig. 2C, see double-infected BrdU-positive cells). To visualize the EphA4 mutant it was linked to copGFP via the T2A peptide sequence (for details on copGFP detection see Materials and Methods).

Next, we investigated the effect of Epha4 knockdown on precursor cell maintenance in neurospheres. We cultured infected neurospheres, and after 6 days we transferred them into media lacking growth factors. After 2 days following growth factor removal, Epha4 knockdown caused a prominent decrease in the percentage of nestin-positive cells relative to the total number of infected cells (Fig. 3A–C). After 4 days of incubation in media without growth factors, the percentage of astrocytes (i.e. number of GFAP-positive infected cells relative to the total number of infected cells) was much higher in neurospheres expressing Epha4 shRNAs than in control neurospheres (Fig. 3D–F). Whereas, in vivo, GFAP labels NSCs and astrocytes, in neurospheres GFAP only labels astrocytes. Nestin- and GFAP-positive cells do not overlap in neurosphere cultures (supplementary material Fig. S7) (see also Campos, 2004; Campos et al., 2004). Although GFAP labels cells in the center of neurospheres (mature astrocytes), nestin labels precursor cells at the periphery. Using another marker for mature astrocytes, namely S100β, we detected a similar increase in infected astrocytes (relative to the total number of infected cells) after Epha4 knockdown as that observed in experiments where GFAP stainings were performed (Fig. 3J–L). Also, in experiments in which differentiation was assessed using the neuronal marker Tuj1, there was an increase in differentiated cells in conditions of Epha4 knockdown (Fig. 3G–I). The Epha4 knockdown phenotype could be rescued when modified EphA4 was coexpressed, whereas a kinase-dead EphA4 mutant was not able to rescue the Epha4 knockdown (Fig. 3C,F,I,L).

To confirm that knockdown of Epha4 indeed affects NSC differentiation, we analyzed the formation of secondary neurospheres derived from primary neurospheres infected by retroviruses expressing control or Epha4 shRNAs. The formation of primary neurospheres was not affected by Epha4 knockdown, except for the slightly smaller average size (data not shown) that was most probably due to a decrease in the number of infected proliferating cells (see Fig. 2A–D). However, the formation of secondary neurospheres was dramatically reduced after Epha4 knockdown in primary neurospheres (Fig. 3M–O). Each secondary neurosphere is a descendant of a single cell from a primary neurosphere. Thus, it can be inferred that, after Epha4 knockdown in primary neurospheres, a decrease in the number of undifferentiated precursor cells would result in fewer secondary neurospheres. Conversely, the number of non-infected secondary neurospheres was similar for control and Epha4 knockdown conditions (Fig. 3P). Thus, in neurosphere cultures, Epha4 knockdown accelerates NSC differentiation, which might account at least in part for the observed decrease in proliferation.

To extend the results obtained in vitro, we silenced EphA4 expression in the postnatal SVZ of wild-type mice using lentiviruses containing shRNAs against Epha4. We injected lentiviruses expressing EGFP and control or Epha4 shRNAs into the SVZ of postnatal day 6 (P6) mice (for details on viral titers see Materials and Methods) that were sacrificed at 4, 7, 15, 25, 40, 60, 80 and 100 days post-injection (Fig. 4). Because the number of infected cells in the SVZ can vary from one injection to another (even when the viral titer is constant), data were normalized and expressed as relative to the number of all infected cells in the SVZ (for further details on the quantification index see Materials and Methods). There was no difference in the number of migrating RMS neuroblasts when comparing control and Epha4 knockdown mice at any time point up to 25 days post-injection. However, at 40 days post-injection and later there was a dramatic decrease of infected neuroblasts in the RMS (Fig. 4A,B,D) of Epha4 knockdown animals. Concomitantly, there was a reduction in the number of infected neurons in the olfactory bulb of Epha4 knockdown animals at 80 days post-injection and later (Fig. 4E,F).

Interestingly, although the effect of Epha4 knockdown could be first detected only as late as 40 days post-injection, the decrease of EphA4 expression levels around the SVZ was already visible 3 days post-injection (supplementary material Fig. S8A,B) and was more pronounced along the whole SVZ at 6 days post-injection (supplementary material Fig. S8C–E).

For rescue experiments in vivo, we used lentivirus expressing a modified EphA4 that contains three silent mutations, making it resistant to shRNA Epha4-1. Cells infected with the ‘rescue’ virus can be visualized by copGFP expression whereas infection with the shRNA Epha4 virus is monitored by EGFP expression. To obtain a significant number of double-infected cells, we used a fivefold higher titer for the ‘rescue’ virus. Co-injection into the SVZ of shRNA Epha4-1 lentivirus (also expressing EGFP, green) together with lentivirus expressing modified EphA4 (also positive for copGFP, magenta) rescued the Epha4 knockdown effect (Fig. 4C,D,F). At 40 days post-injection and later, most RMS neuroblasts expressing shRNA Epha4-1 also contained the ‘rescuing’ modified EphA4 virus (Fig. 4C). Furthermore, at 80 days post-injection and later, there was a significant increase in the number of cells double-infected by Epha4 knockdown and ‘rescue’ lentiviruses in comparison with the control (i.e. gain-of-function effect) (Fig. 4D). Cells infected by the virus expressing the modified EphA4 did not exhibit morphological alterations.

We determined the number of proliferating cells in the SVZ by administering two BrdU pulses to label dividing cells (supplementary material Fig. S9A–C). Although at 4 and 25 days post-injection there was no difference between the numbers of infected proliferating cells in control and Epha4 knockdown conditions, at 40 days post-injection there was a 50% reduction in the number of BrdU-positive infected cells around the SVZ in Epha4 knockdown animals.

The reduction of infected neuroblasts in the RMS after Epha4 knockdown cannot be explained by a change in the number of apoptotic cells in the SVZ, as indicated by the unaltered caspase-3 activation (supplementary material Fig. S9D–F). Thus, both in vitro and in vivo experiments provided evidence that Epha4 knockdown resulted in a decrease of precursor cell proliferation that was not a consequence of altered survival. The observed effect of EphA4 on NSCs is cell autonomous and is mediated by intracellular signaling of EphA4 because it occurs directly in the infected cells. Also, a kinase-dead EphA4 mutant (with a lack of forward signaling) was not able to rescue the Epha4 knockdown phenotype (Fig. 4D).

The immunohistochemistry results indicated that EphA4 is expressed only in NSCs, but not in transit-amplifying progenitors or neuroblasts (Fig. 1). Hence, we expect that the knockdown of EphA4 forward signaling (in the same cell expressing EphA4) decreases proliferation of NSCs but not of other cell types in SVZ. To this end, we analyzed the effect of Epha4 knockdown in the SVZ using retroviruses known to infect only dividing cells. In comparison with neurosphere cultures where NSCs divide fast (Campos, 2004; Qian et al., 1998), in postnatal SVZ the NSCs divide very slowly (Doetsch et al., 1999; Morshead et al., 1994; Morshead and van der Kooy, 2004). Indeed it has been established that the majority of infected cells in the SVZ after retrovirus injection are fast-dividing transit-amplifying precursors or neuroblasts, but not slow-dividing NSCs (Rogelius et al., 2005). Thus, we injected retroviruses expressing tdTomato (a red fluorescent protein) and either control shRNA or shRNA Epha4-1 into the SVZ of P6 mice, and determined the number of infected cells in the RMS or olfactory bulb relative to the total number of infected cells in the SVZ at 5, 8, 12, 18, 25, 40 and 60 days post-injection (Fig. 5A–C). There was no difference in the number of infected cells in the RMS (Fig. 5C) or olfactory bulb (Fig. 5B) between control and Epha4 knockdown mice at any time point. Also, we detected only one or two infected neuroblasts in the RMS per animal at 18 days post-injection and no infected neuroblasts at 40 days post-injection and later (Fig. 5C). The number of BrdU-positive infected cells in the SVZ, labeled by two short BrdU pulses, was not different between Epha4 knockdown and control conditions (Fig. 5D).

Finally, we extended our in vivo injection experiments to adult animals and injected the SVZ of 3-month-old mice with lentiviruses expressing EGFP together with either control shRNA or one of the Epha4 shRNAs. Animals were sacrificed at 4, 25, 40 and 60 days post-injection (Fig. 6). Although there was no difference in the number of infected cells in the RMS of control and Epha4 knockdown animals at 4 and 25 days post-injection, at 40 days post-injection and later the number of neuroblasts in the RMS was significantly decreased after Epha4 knockdown (Fig. 6A–C). Thus, knockdown of Epha4 in adults resulted in the same phenotype as that observed after injections in P6 animals.

We subsequently tested in vivo whether, in addition to proliferation, EphA4 also affected NCS differentiation, as was suggested by the neurosphere experiments. When restricting the analysis to the area around the lateral ventricles, there was a significant decrease in the number of GFAP-positive cells following Epha4 knockdown (Fig. 7A–D), which was already apparent at 15 days post-injection (Fig. 7D; see supplementary material Fig. S10A,B for representative images when using a higher viral titer). In fact, the decrease in NSCs in the knockdown condition might even be an underestimation if one considers that some GFAP-positive cells in the area analyzed might be astrocytes. Similar results were obtained when Sox2 and nestin were used as NSC markers (Fig. 7E–I; for nestin see also supplementary material Fig. S10C,D for representative images when injecting a higher viral titer). Thus, at 4 days post-injection the number of Sox2-positive infected cells was comparable for all conditions, whereas after 15 days we observed a significant decrease in the Sox2-positive cell number in Epha4 knockdown animals in comparison with control or rescue animals.

A decrease in the number of NSCs would eventually result in a decreased number of their progeny cells. Indeed, there was a decrease in the number of transit-amplifying precursors and neuroblasts around the SVZ, as revealed by immunolabeling with Mash1 and DCX, respectively. At 40 days post-injection, the percentage of Mash1-positive infected cells in Epha4 knockdown animals was half that of controls (Fig. 7J–L). The decrease in DCX-positive infected cells around the SVZ corresponded to the decrease in infected neuroblasts in RMS (see Fig. 4D).

If Epha4 knockdown reduces the number of cells expressing stem cell markers and drives direct differentiation, one would expect an increase in neuronal markers in the knockdown cells. Indeed, this was the case. However, to test this, an injection site located slightly more dorso-medially was used to avoid the infection of many striatal neurons that occurred when injections were close to the lateral ventricles. When using these injection coordinates, Epha4 knockdown resulted in an increase of NeuN-positive infected cells around the SVZ in comparison with control animals at 40 days post-injection (Fig. 7M–O).

A similar approach was used to analyze cell differentiation in the SVZ using 3- to 4-month-old mice (supplementary material Fig. S11). Animals were injected into the SVZ with lentiviruses expressing EGFP together with either control shRNA or one of the Epha4 shRNAs and sacrificed at 4, 15 and 40 days post-injection. Similar to early postnatal injections, Epha4 knockdown decreased the number of GFAP- (supplementary material Fig. S11A–C), BrdU- (supplementary material Fig. S11D–F) and Mash1-positive (supplementary material Fig. S11G–I) infected cells around the SVZ. Conversely, there was an increase in NeuN-positive cells (supplementary material Fig. S11J–M).

Finally, we performed BrdU-retaining experiments to directly demonstrate that Epha4 knockdown indeed reduces the number of slow-dividing cells. The SVZ of P6 or adult animals was injected with lentiviruses expressing EGFP together with either control shRNA or one of the Epha4 shRNAs. To label slow-dividing cells, at 40 days post-injection BrdU was administered in the drinking water for 1 week followed by 3 weeks with no treatment before sacrificing the animals. Both early postnatal (Fig. 8A–C) and adult (Fig. 8D–F) Epha4 knockdown animals showed a 50% reduction in the number of infected slow-dividing cells around the SVZ compared with controls. Thus, several lines of evidence indicate that Epha4 expression preserves NSCs in an undifferentiated state.

Ephrins and their Eph receptors regulate many aspects of neural system development during embryogenesis (Klein, 2004; Pasquale, 2005). They also participate in the control of neurogenesis in both postnatal neurogenic regions, the SVZ of the lateral ventricles, and the hippocampal SGZ (Chumley et al., 2007; Conover et al., 2000; Holmberg et al., 2005; Jiao et al., 2008; Qiu et al., 2008; Ricard et al., 2006). In the SVZ, ephrin signaling has been shown to control progenitor cell proliferation (Chumley et al., 2007; Conover et al., 2000; Katakowski et al., 2005; Ricard et al., 2006; Theus et al., 2010) and apoptosis of excessively generated cells (Holmberg et al., 2005; Jiao et al., 2008). In this study, we have shown that Eph receptor A4 is expressed specifically in NSCs, and is necessary to preserve NSCs in an undifferentiated state. Interaction of Eph receptors with ephrins activates both forward signaling (in the cell where the receptor is expressed) and reverse signaling (in the cell expressing the ephrin ligand) (Egea and Klein, 2007). The effects described here following Epha4 knockdown occurred in cells that were infected by viruses expressing shRNA Epha4, and it can be thus inferred that EphA4 preserves NSC fate via forward signaling. This is also supported by the results showing that a ‘kinase-dead’ mutant of EphA4 was not able to rescue the Epha4 knockdown phenotype. EphA4 function is not restricted to the early postnatal SVZ, but continues to be important in the adult SVZ.

The lack of effect on neurogenesis following Epha4 knockdown using retroviruses (that generally infect fast-dividing cells) (Rogelius et al., 2005), together with the observed decrease in RMS neuroblasts only as late as 40 days post-injection using lentiviruses, establish a functional role of EphA4 signaling in NSCs, but not in other proliferative cell types in SVZ. Postnatal NSCs are quiescent for long periods of time and divide rarely, about once in 2 weeks (Doetsch et al., 1999; Morshead et al., 1994; Morshead and van der Kooy, 2004). Thus, a decrease in NSC proliferation will be reflected in the number of their progeny only several weeks later. Also, immunohistochemical data confirm the presence of EphA4 in NSC, but not in precursor cells and neuroblasts. The facts that neuroblast morphology in Epha4 knockdown mice was comparable with controls and that the number of neuroblasts was reduced only at 40 days post-injection, together with unaltered neuroblast number after retroviral injections, speak against a scenario in which EphA4 is involved in migration.

The acceleration of NSC differentiation can most probably account for the observed decrease in their proliferation. This conjecture is supported by several observations. The population of NSCs, indicated by GFAP-, Sox2- and nestin-positive staining, remained relatively constant in control animals (Fig. 7D,F,I). These results are in line with previous reports that postnatal NSCs divide asymmetrically (Morshead et al., 1998; Suh et al., 2007), each NSC generating one NSC and one precursor cell, thus explaining the constancy of the NSC pool. Hence, a decrease in NSC proliferation would not change their number. Therefore, a significant reduction in NSC population after Epha4 knockdown is the result of premature differentiation rather than that of proliferation impairment. Because we did not observe an early increase in the number of transit-amplifying precursors and neuroblasts (as well as an overall proliferation increase), it is possible that NSCs, upon Epha4 knockdown, are enforced to differentiate directly to non-dividing cells (e.g. mature neurons) omitting intermediate phenotypes such as transit-amplifying precursor cells and neuroblasts. This is in line with the in vitro results where we observed an increase in the number of mature astrocytes and neurons together with a slight decrease in the total cell number after Epha4 knockdown. Interestingly, EphA signaling through other two members of EphA family, EphA2 and EphA3, was found to be involved in pre- and postnatal neural precursor cell differentiation towards the neuronal lineage (Aoki et al., 2004).

Potential ligands for EphA4 have been identified in other systems or during embryonic development. Ephrin B1 has been shown to be required for the maintenance of neural progenitor cell state in the ventricular zone of the developing embryonic cortex (Qiu et al., 2008). The interaction of EphA4 with ephrin B1 in embryonic NSCs of the ventricular zone in rodents also lends credence to a proposed function for cortical development (North et al., 2009). Another ligand for EphA4 might be ephrin B3, which was shown to interact with Epha4 in HEK293 cells (Furne et al., 2009; Ricard et al., 2006). After interaction with ephrin B3, EphA4 propagates reverse anti-apoptotic signal to cells expressing ephrin B3, whereas in cells lacking ephrin B3, the reverse signal induces apoptosis (Furne et al., 2009).

Involvement of EphA4 in neurogenesis was also indicated in a study reporting a significant increase in the SVZ and RMS areas in Epha4 knockout animals (Furne et al., 2009). Using knockout mice, however, precludes evaluation of the role of EphA4 in neurogenic cells because EphA4 is expressed in many non-neurogenic cells, such as mature astrocytes, neurons and endothelial cells (Deininger et al., 2008; Goldshmit et al., 2004; Goldshmit et al., 2006; Tremblay et al., 2009). The use of low-titer lentiviruses infecting sparsely distributed cells allows functional studies of EphA4 forward signaling specifically in infected cells. Furthermore, EphA4 plays an important role during embryonic brain development; thus, prenatally derived changes in brain structure and activity of full Epha4 knockout animals can modify postnatal brain function, an effect that is avoided when employing postnatal viral injections.

Postnatal neurogenic niches contain several cell types known to differentially express specific ephrins and/or Eph receptors. Given the redundancy of ephrin–receptor interaction (individual ephrins can interact with several Eph receptors and vice versa), ephrin signaling in neurogenic niches is mediated by complex intracellular networks, most probably involving distinct pathways for certain ligand–receptor interactions and common pathways for others. To better understand the role of ephrin signaling for postnatal neurogenesis, it is thus important to identify the expression of different members in the defined cell types in the neurogenic niches and to link members of the ephrin system with intracellular pathways.

For all our experiments, we used wild-type C57Bl/6 mice. All procedures with animals were performed according to the Heidelberg University Animal Care Committee.

All chemicals and cell culture reagents were purchased from Sigma-Aldrich (Taufkirchen, Germany) and Invitrogen (Karlsruhe, Germany), respectively, unless otherwise specified.

The following antibodies were used: polyclonal rabbit anti-EGFP, 1:10000 (Invitrogen), chicken anti-EGFP, 1:1000 (Abcam, Cambridge, UK), mouse anti-III class β-tubulin, Tuj1, 1:500 (Covance, Princetown, NJ), goat anti-doublecortin, 1:500 (Santa Cruz Biotechnology), rabbit anti-EphA4, 1:250 (Abcam), rabbit anti-EphA4, 1:200 (Zymed, San Francisco, CA), rabbit anti-copGFP, 1:3000 (Evrogen, Moscow, Russia), mouse anti-GFAP, 1:4000 (Sigma-Aldrich), rabbit anti-GFAP, 1:2000 (Dako, Hamburg, Germany), rabbit anti-nestin, 1:500 (Abcam), mouse anti-nestin, 1:500 (BD Biosciences, Franklin Lakes, NJ), chicken anti-nestin, 1:500 (Novus Biologicals, Littleton, CO), rat anti-BrdU, 1:500 (Accurate Chemicals, Westbury, NY), rabbit anti-activated caspase-3, 1:2000 (Chemicon, Temecula, CA), mouse anti-Mash1, 1:500 (BD Biosciences), rat anti-Ki67, 1:100 (Dako), rabbit anti-Sox2 (Santa Cruz Biotechnology), mouse anti-PSA-NCAM, 1:500 (Millipore, Schwalbach/Ts, Germany), goat anti-Olig2, 1:500 (R&D Biosystems), rabbit anti-S100β, 1:1000 (Swant, Bellinzona, Switzerland), mouse anti-S100β, 1:500 (Millipore), Alexa-Fluor-488-conjugated anti-rabbit, anti-mouse, anti-rat, anti-chicken and anti-goat, Alexa-Fluor-350-conjugated anti-rabbit and anti-mouse secondary antibodies (Invitrogen), anti-mouse, anti-rabbit, anti-rat and anti-goat Cy3 coupled, anti-mouse, anti-goat and anti-rabbit Cy5 coupled secondary antibodies (Jackson ImmunoResearch Laboratories, Westgrove, PA), anti-mouse and anti-rabbit HRP-conjugated secondary antibodies (Vector Laboratories, Burlingame, CA).

The target sequences for oligos used to construct shRNA Epha4 expression plasmids were 5′-GCAGCACCATCATCCATTG-3′ for Epha4-1 and 5′-ATCCACCTGGAAGGCGTTG-3′ for Epha4-2. Scrambled shRNA sequences were cloned from pSilencer vector (Ambion, Austin, TX). Complementary pairs of oligonucleotides were cloned into pSUPER vector (Oligoengine, Seattle, WA). To make recombinant lentiviral plasmids for in vivo experiments, we re-cloned the shRNA silencing cassettes from pSUPER vector to pFUGW, a lentiviral vector containing EGFP expressed under the control of the ubiquitin promoter (Lois et al., 2002). To generate retroviral plasmids expressing shRNAs, we re-cloned tdTomato (pRSET-tdTomato; generous gift of Roger Tsien, University of California at San Diego, La Jolla, CA) and the CMV promoter (from pEGFP-C1; Clontech-Takara Bio, Saint-Germain-en-Laye, France) into the pFB retroviral vector (Stratagene, La Jolla, CA), resulting in pFB-CMV-tdT vector. Two shRNA Epha4 cassettes were cloned from the pSUPER shuttle vector into the retroviral vector.

pCMV-SPORT-EphA4 expression plasmid was purchased from Biocat (Heidelberg, Germany). To produce the EphA4 expression plasmid resistant to shRNA Epha4-1, we introduced into the Epha4 open reading frame (ORF) three silent mutations using Quickchange Mutagenesis Kit (Stratagene), and cloned the modified Epha4 ORF into the pCDH-EF1-T2A-copGFP (original plasmid from System Bioscience, Mountain View, CA) or pCDH-EF1-T2A-tdT (pCDH-EF1-T2A-copGFP where we substituted copGFP for tdTomato) lentiviral vector. To obtain retroviral plasmid containing modified EphA4, the whole EF1-modEphA4-T2A-copGFP cassette was re-cloned into pFB retroviral vector.

CopGFP, a green fluorescent protein, was linked to ORF Epha4 through the T2A peptide sequence. We were not able to detect the native fluorescent signal of copGFP when expressed under the control of the EF1 promoter (elongation factor 1), but there was a bright copGFP signal in infected cells when stained with copGFP antibodies (supplementary material Fig. S12A–C). The T2A sequence allows the generation of two proteins expressed at the same level (Osborn et al., 2005). CopGFP antibodies do not cross-react with EGFP (supplementary material Fig. S12C), hence the expression of both fluorescent proteins in the same cell can be detected unambiguously.

The kinase-dead EphA4 mutant was generated as described previously (Kullander et al., 2001), exchanging lysine 653 with a methionine residue. The mutation was introduced into modified EphA4 resistant to Epha4-1 by Quickchange Mutagenesis Kit (Stratagene), and the kinase-dead Epha4 ORF was cloned into pCDH-EF1-T2A-copGFP (System Bioscience) lentiviral vector. To obtain retroviral plasmid containing kinase-dead EphA4, the whole EF1-EphA4KD-T2A-copGFP cassette was re-cloned into pFB retroviral vector.

RNA isolation, cDNA synthesis and quantitative real-time (qRT)-PCR was performed as described previously (Khodosevich et al., 2007). mRNA levels detected by qRT-PCR were normalized to mRNA levels for Gapdh.

The efficiency of shRNA silencing was tested using qRT-PCR and western blot using HEK293 cell culture transfections in triplicates. Two out of five shRNAs that specifically knocked down Epha4 gene expression to less than 25% were chosen for functional experiments described in the study. The efficiency of shRNA silencing was also tested by viral infections of neurospheres and in vivo brain infections.

Recombinant lentiviruses were produced as previously described (Khodosevich et al., 2009). For retrovirus production we used the same conditions as for lentiviruses, except for using GP helper plasmid instead of 8.9 plasmid.

To measure viral titers, a dilution series across five orders of magnitude of viral stock solutions were used for HEK293 cell infection. Each sample was analyzed in triplicate. After 4 days of incubation at 37°C, the number of fluorescent cell plaques at the different viral dilutions was measured and the viral titer was estimated in fluorescent plaque forming units per milliliter (pfu/ml).

SVZ regions were dissected from coronal sections of wild-type P4 mice. All steps of tissue processing were done in dissection media (10× DM: 100 mM MgCl₂, 10 mM kynurenic acid, 100 mM HEPES in 1× Hank's balanced salt solution). Dissected SVZ regions were incubated for 5 minutes with 30 U of papain (Worthington, Lakewood, NJ) and 0.0005% DNase solution, and washed in Neurobasal Media Supplemented [500 ml of Neurobasal media + 10 ml 50× B27-Supplement + 1.25 ml 200 mM l-glutamate + 5 ml penicillin/streptomycin (100 U/ml)] containing also trypsin inhibitor (Sigma-Aldrich) and 0.0005% DNAse. Cells were triturated using a fire-polished Pasteur pipette, counted and plated at a density of 100,000 cells/ml in Neurobasal Media Supplemented containing 20 ng/ml EGF and 20 ng/ml FGF. To obtain secondary neurospheres, the primary neurospheres were dissociated by trypsin incubation and subsequent trituration using a fire-polished Pasteur pipette. Dissociated cells were counted and plated at a density of 20,000 cells per ml in Neurobasal Media Supplemented containing 20 ng/ml EGF and 20 ng/ml FGF.

Proliferation analysis: at DIV1, cells were infected with shRNA-expressing viruses with or without the additional presence of rescue virus. At DIV6, BrdU was added to the culture media to give a final concentration of 10 μM. Neurospheres were incubated with BrdU for 12 hours and subsequently fixed.

Differentiation analysis: at DIV1, cells were infected with shRNA-expressing viruses with or without the additional presence of rescue virus. At DIV7, neurospheres were transferred to Neurobasal Media Supplemented without growth factors, and at DIV9 (for nestin immunostaining) or DIV11 (for GFAP, S100β and Tuj1 immunostainings) cultures were fixed.

The titer of the injected virus was adjusted so as to be equal for all experiments: 2×10⁷ units/ml for lentiviruses and 10⁷ units/ml for retroviruses. For rescue studies we used the same titer for shRNA expressing viruses: 2×10⁷ units/ml, and 10⁸ units/ml of rescue virus. We also confirmed the EphA4 rescue results using a fivefold higher titer (resulting in a higher number of infected cells) for shRNA and rescue viruses (10⁸ units/ml and 5×10⁸ units/ml, respectively). Throughout the text, ‘low-titer’ for shRNA-expressing viruses indicates 2×10⁷ units/ml, and ‘high-titer’ indicates 10⁸ units/ml. An aliquot of 1 μl of recombinant retrovirus or lentivirus (or mix of viruses) was delivered into the SVZ of each hemisphere of P6 C57BL/6 mouse pups with a Hamilton (Hamilton, Bonaduz, Switzerland) syringe using special needles for precise animal injections (reduced needle volume, 20 mm length, 26s gauge and 45° tip angle). The animals were sacrificed 5, 8, 12, 18, 25, 40 or 60 days after injection for retroviruses and 4, 7, 15, 25, 40, 50, 60, 80 or 100 days after injection for lentiviruses, and fluorescent cells in the olfactory bulb, RMS and SVZ were counted.

Adult (3-month-old) animals were injected as previously described (Celikel et al., 2007). The coordinates that were used for injections into the SVZ were: anterior, 0.9; lateral, 1; ventral, 2.3.

Data were analyzed using t-test or ANOVA by GraphPad Prism version 5.00 for Mac OS X (GraphPad Software, San Diego California, www.graphpad.com). All data were normally distributed according to the d'Agostino and Shapiro-Wilk tests. n always defines the number of independent experiments. Absolute cell counts, number of neurospheres and culture coverslips, number of brain sections and animals used for all quantifications are shown in supplementary material Table S2.

Co-labeling in neurospheres was quantified using stacks of confocal images obtained by using 0.3 μm steps acquired with 63× immersion objective (confocal microscope LSM 5 Pascal, Zeiss, Germany). For the markers that label only cell processes (nestin and GFAP) co-localization was determined only after careful examination of tdTomato coexpression in the processes. Low titer retrovirus infection assured scarcity of infected cells in neurospheres. Quantifications in Figs 2 and 3 show mean ± s.d.; number of cells >100, number of neurospheres ≥30, number of culture coverslips for each condition ≥6.

Numbers of fluorescent cells in the RMS or olfactory bulb were evaluated relative to the total number of infected cells around the SVZ. GFAP-, Sox2-, NeuN-, S100β-, nestin-, BrdU-, Mash1, Dcx-, activated caspase-3-positive fluorescent cells in the SVZ were evaluated as percentage of the total number of infected cells around the SVZ. Injections with the same type of virus and titer resulted in approximately the same number of infected cells across animals. In rescue experiments, only double-infected cells were calculated. Mis-injected mice were excluded from further analysis. Quantifications in Figs 4, 5, 6, 7 and 8 and supplementary material Figs S9 and S11 show mean ± s.d., number of infected cells >1000, number of independent experiments ≥6.

This way of estimating the quantification index should not be affected by altered proliferation in the SVZ after Epha4 knockdown for the following reasons. Although cell proliferation after Epha4 knockdown decreased by 0.5% at 40 days post-injection (BrdU-labeling, 1 day after BrdU injection), these cells do not enter the equation as newly born cells in the SVZ and eventually migrate to the olfactory bulb; hence they are not considered to be ‘infected cells around the SVZ’ once they reach the RMS. Most importantly, all normalized quantifications (except NeuN co-localization) showed a decrease in the number of specifically labeled cells (i.e. GFAP+, Sox2+, BrdU+, Mash1+, nestin+) after Epha4 knockdown. Because the index is obtained by dividing specifically labeled infected cells by all infected cells in the SVZ, reduced proliferation in the SVZ would result in a decreased number of total infected cells in the SVZ, and hence in a higher index in Epha4 knockdown experiments than controls, which was not the case.

To label fast-proliferating cells, mice were injected once with 50 mg BrdU/kg body weight. To label slow-proliferating cells, 1 mg/ml BrdU was administered in the drinking water for 1 week followed by 3 weeks of normal water consumption to obtain BrdU dilution in fast-proliferating cells. For the analysis of dividing cells after lentiviral infection, mice were injected with BrdU twice, the two time-points being 6 hours apart, and sacrificed 24 hours later. Postnatal animals were injected with 20 mg BrdU/kg body weight and adult animals with 50 mg/kg.

Sagittal brain sections (50–75 μm) were cut with a vibratome (Leica VT1000S, Leica, Germany). Immunostainings were carried out on free-floating sections. Slices were blocked in 0.2–1% Triton and 2–5% normal goat serum or 5% bovine serum albumin and incubated overnight with primary antibodies at 4°C followed by incubation with secondary antibodies at room temperature. Primary and secondary antibodies were listed above. For nuclear staining we used DAPI. For BrdU stainings, slices were preincubated with 1 N HCl at room temperature for 1 hour, then neutralized with 10 mM Tris (pH 8.5) at room temperature for 15 minutes and further processed as for other stainings. Sections were mounted onto slides with Mowiol (Carl Roth, Karlsruhe, Germany) or Immu-mount (ThermoScientific, Bonn, Germany) and subsequently analyzed on a fluorescent microscope (Axioplan 2; Zeiss, Germany) or confocal microscope (LSM 5 Pascal; Zeiss, Germany).

For western-blot analysis, protein samples were boiled in SDS gel sample buffer. Denatured proteins were separated by SDS-PAGE, transferred onto PVDF membranes and probed with antibodies. For statistical analysis, antibody signals were quantified using NIH ImageJ software and values normalized to the corresponding β-actin signals. Data were analyzed using the Student's t-test.

We thank Ulla Amtmann, Regina Hinz-Hernkommer and Irmgard Preugschat-Gumbrecht for technical assistance. This work was supported in part by the Schilling Foundation and DFG (SFB488 grant).