A transient subpial neurogenic zone in the dentate gyrus
Pioneering studies by Altman and Bayer (Altman and Bayer, 1990a
; Altman and Bayer, 1990b
) showed that proliferative dentate precursors leave the dentate neuroepithelium and form a migratory stream into the forming dentate gyrus during late embryonic development. To better characterize the spatiotemporal dynamics of the precursors in the dentate migratory stream in mice, we decided to re-examine this framework with new molecular markers. Using Nestin-GFP transgene and Tbr2 to label the dentate precursors and neurogenic transit amplifying cells, respectively (Englund et al., 2005
; Yamaguchi et al., 2000
), we revealed the sequential distribution of these cells during initial dentate gyrus formation.
By E14.5, the domain adjacent to the cortical hem in the ventricular zone (VZ) representing the dentate primordium was marked by strong Nestin-GFP expression (). Concurrently, Tbr2+ cells were abundant in the subventricular zone (SVZ) and a stream of Tbr2+ cells stretched from the SVZ above the forming fimbria to form a line of cells along the pial surface (). A day later at E15.5, Nestin-GFP+ cells emanating from the dentate notch formed a narrow stream oriented toward the subpial region at the junction between the fimbria and forming dentate gyrus (); we have termed this the fimbriodentate junction (FDJ). At high magnification, Tbr2+ cells were also found in the FDJ region () and, in fact, this structure appeared to have formed as a continuation of the subpial collection of Tbr2+ cells seen at E14.5. Interestingly, in the FDJ, the strongly stained Nestin-GFP+ cells did not overlap with the Tbr2 staining, implying that these were largely non-overlapping cell populations (arrowheads in ).
Progressive development of the dentate neurogenic zones is revealed by Nestin-GFP transgene and Tbr2
At E17.5, Nestin-GFP+ cells had fanned out into the hilus () and lined the hippocampal fissure (HF) (). Tbr2+ cells were also distributed across the hilus and along the HF (). At P0, Nestin-GFP+ cells completely covered the dentate side of the HF and the subpial region of the future lower blade, whereas few Nestin-GFP+ cells were located in the hilus (). Tbr2+ cells continued to mirror this distribution pattern ().
By P2, Nestin-GFP+ cells and processes appeared to spread toward the hilus from the marginal zone (MZ) of the granule cell layer (GCL) (), but by contrast, most Tbr2+ cells were still restricted to the emerging molecular layer at this stage (). At the end of the first postnatal week, most Nestin-GFP+ cells populated the hilus and the subgranular zone (SGZ), and had exited the molecular layer (). The Tbr2+ population in the molecular layer was reduced and now found largely in the SGZ (arrowhead in ).
Consistent with work in the adult dentate (Yamaguchi et al., 2000
), perinatal analysis of Nestin-GFP+ cells with acute BrdU labeling showed that the subpial Nestin-GFP+ cells close to the HF were actively dividing with little overlap with Prox1 (see Fig. S1A, A′, A″ in the supplementary material
). Similarly, subpial Tbr2+ cells were proliferating (see Fig. S1B in the supplementary material
) with clear distinction from the reelin+ Cajal-Retzius cells (see Fig. S1C in the supplementary material
). In agreement with their neurogenic nature in the developing cortex (Arnold et al., 2008
; Englund et al., 2005
; Sessa et al., 2008
), the weak Tbr2+ cells also showed low Prox1 expression in both MZ and the newly formed SGZ at P5 (see Fig. S1D-F in the supplementary material
). Taken together, the spatiotemporal distribution of Nestin-GFP+ and Tbr2+ cells reveal two phases of neurogenic zone transitions (): the dentate VZ-to-subpial transition and the subpial-to-subgranular transition.
Formation of the transient neurogenic zone coincides with the appearance of transhilar glial processes
Since previous studies indicate that radial glial cells regulate the morphogenesis of the dentate gyrus (Eckenhoff and Rakic, 1984
; Rickmann et al., 1987
), we further investigated the distribution of the radial glial scaffolding by GFAP staining as the Nestin-GFP+ precursors migrate from the dentate primordium to the subpial neurogenic zone. By E18.5, Prox1+ granule cells already occupied the forming upper blade, whereas GFAP+ glial fibers were enriched at the border of the fimbria (). These GFAP+ fibers appeared to spread out at the entrance of the hilus and project to the pia all around the forming dentate (), whereas Prox1+ granule cells were arranged in parallel to this glial scaffolding in the hilus (). Previous studies assumed that these glial fibers are projecting from radial glial cells with their cell bodies located in the VZ (Eckenhoff and Rakic, 1984
). To determine whether these hilar fibers directly project from the dentate VZ or the new organizing center at the FDJ (, red arrow), we injected DiO solution into the ventricle of E18.5 embryos and allowed these to survive for only 3 hours to physically label all the fibers projecting from the VZ. This labeling prominently marked all the radial glial fibers spanning the whole hippocampal fields except the dentate field (). It suggests that radial glial fibers in the forming dentate during this migratory phase do not directly project to the dentate from the dentate ventricular zone.
Development of the transhilar glial scaffolding proceeds without evidence of VZ connection
To further examine which cells contribute to this scaffolding in the developing dentate, a GFP expression construct was electroporated in utero into the medial wall, including the dentate VZ, at different ages followed by various survival times. In brains of mice electroporated at E13.5 and examined at E14.5, cells in the dentate VZ projected to the pia or cells themselves appeared to be migrating in the same trajectory (, arrowheads). However, when electroporated at E14.5 and examined at E15.5, the dentate VZ showed no direct radial projection into the emerging dentate at all, despite the prominent radial glial scaffolding in the hippocampus proper (). Instead, apparently migrating GFP+ cells were seen at the hilar entrance (, arrow). Therefore, although it appeared that there was a direct VZ to dentate cellular connection at E14.5, this was no longer apparent by E15.5. To further address this at later stages, the medial cortex was electroporated at E17.5. Two days later, extensive labeling of radial fibers was seen in the hippocampal fields, but the dentate was completely devoid of radial fibers from the VZ (). Instead, GFP+ cells formed a distinct stream from the dentate VZ along the fimbria (, arrows) and some of these were again visible at the entry of the hilus ().
Staining with a radial glial marker, brain lipid binding protein (BLBP), labeled a prominent subset of glial processes in the forming dentate at E18.5 (). Strikingly, unlike with GFAP staining, glial cell bodies were clearly identifiable with BLBP in the FDJ (arrow in inset) and some of them had already reached the HF (arrows in ). A collapsed z-projection of the serial sections is shown in , allowing the reconstruction of two whole BLBP+ cells. The somata of the BLBP+ processes spanning the hilus were pinpointed by tracing the overlapping thin sections (). Interestingly, somata were found near the FDJ pia or in the hilus (red and yellow arrows in ), suggesting that some glia were in the process of migration from the FDJ towards the HF.
Reelin is dispensable for the formation of but controls exit from the transient zone
Numerous studies have shown that reelin secreted from Cajal-Retzius cells is essential for the development of the neocortex and hippocampus by controlling the proper lamination of projection neurons (Rice and Curran, 2001
). As the dentate is also quite abnormal in reelin mutants (Forster et al., 2002
), we wondered what role reelin plays in the migration of dentate precursors to the newly identified transient subpial zone. To tackle this issue, we examined the distribution of Nestin-GFP+ cells at birth in reeler mice. In both the controls and reeler mice, Nestin-GFP+ or Tbr2+ cells were properly localized to the subpial zone (arrowheads in and ). Consistent with the known role of reelin in neuronal migration, Prox1+ granule cells were abnormally distributed across the dentate field in the mutants (arrow in ) instead of forming a relatively compact upper blade as in the controls (). Thus, reelin signaling is required for proper granule cell migration but not for the subpial localization of Nestin-GFP+ cells.
Sustained subpial neurogenic zone in the reeler mice
In control mice at P4, most of the Nestin-GFP+ cells were in the SGZ and hilus, whereas the subpial zone was essentially depleted of Nestin-GFP+ cells (). Strikingly, in Reeler mutant mice, the subpial zone was still packed with Nestin-GFP+ cells (). As the upper blade was well-defined at this age, we quantified the distribution of Tbr2+ cells in the dorsal half of the dentate () with the regional schema shown in . Only a small proportion of Tbr2+ cells (28.6±1%) were localized in the control marginal zone, with the majority (71.4±1%) scattered in the GCL, SGZ and hilus (arrow in , inset), whereas in the mutants the Tbr2+ cells (75.4±0.9%) were clustered in the marginal zone (arrow in , inset) with a small number (24.6±0.9%, n
-test, ) in the region outside the marginal zone. Therefore, the dentate neurogenic niche failed to undergo the subpial-to-subgranular transition in the absence of functional reelin. Although previous studies (Forster et al., 2002
) suggest that reelin signaling may cell-autonomously regulate the behavior of radial glial cells, it is also possible that this role for reelin could be due to a non-autonomous role for reelin in organizing other cellular components in the dentate. However, what is clear is that reelin is dispensable for the original positioning of radial glial and transit-amplifying cells in the subpial zone but is indispensable for the later reorganization of this zone.
Formation of the subpial zone requires Cxcl12/Cxcr4 signaling
Previous studies indicate that the chemokine Cxcl12 and its cognate receptor Cxcr4 regulate the morphogenesis of the dentate gyrus (Bagri et al., 2002
; Lu et al., 2002
), but the mechanistic basis of this defect is not well characterized. More recent studies suggest the Cxcl12/Cxcr4 signaling plays a crucial role in regulating the positioning of neurons adjacent to the pia (Borrell and Marin, 2006
; Li et al., 2008a
; Lopez-Bendito et al., 2008
; Paredes et al., 2006
; Tiveron et al., 2006
). Owing to the expression of Cxcr4 in the migratory stream and subpial zone and expression of Cxcl12 by the pial meninges (see Fig. S2 in the supplementary material
) (Berger et al., 2007
), we sought to determine whether Cxcl12/Cxcr4 signaling controls the concentration of Nestin-GFP+ precursors in the subpial region of the developing dentate. At E18.5 in control animals, Nestin-GFP+ cells occupied the subpial region around the entire profile of the forming dentate gyrus from hippocampal fissure superficial to the nascent upper blade and ventrally to the future lower blade (arrows in ). In stark contrast, in Cxcr4−/−
mice the Nestin-GFP+ cells were largely scarce in the subpial region (arrows in ). Consistent with these findings, Tbr2+ cells no longer formed a compact subpial zone along the FDJ and HF in the Cxcr4 mutants as they did in controls (white arrows in ). Instead, Tbr2+ cells were widely dispersed in the dentate (yellow arrow in ). Taken together, these findings indicate that Cxcr4 is required for the proper formation of the subpial neurogenic zone.
Defective subpial neurogenic zone, abnormal radial glial scaffolding and premature granule cell production in Cxcr4 null
Subpial organization of precursors correlates with the transhilar glial scaffolding and neuronal differentiation
The displacement of Nestin-GFP+ progenitors from their subpial location prompted us to ask whether this defect affects the development of the transhilar glial scaffolding. Staining for GFAP showed dense glial processes across the hilus and enriched fiber plexus at the HF in the controls (). By contrast, Cxcr4 mutants showed reduced hilar GFAP+ fibers and decreased fiber plexus in the HF (). In the controls, BLBP staining revealed a subset of transhilar glial fibers emanating from the FDJ and crossing the hilus (). BLBP+ somata were clearly identified in three locations: FDJ, hilus and HF. However, both BLBP+ processes and somata were almost absent in the hilus and HF in mutants (arrows in ).
To determine whether the disruption of the subpial zone and the radial glial progenitors leads to any dynamic consequences for the cellular output of the stem/progenitor cells, we analyzed acute BrdU labeling at E18.5. In agreement with previous studies (Bagri et al., 2002
; Lu et al., 2002
), we found that the number of BrdU+ cells was significantly decreased in the dentate of Cxcr4 mutants compared with the controls (). However, this finding cannot be simply explained by an increase in cell death or a migration defect resulting from the loss of Cxcr4, as the production of granule cells did not seem to have drastically declined despite their abnormal distribution (). We noticed that in association with the decrease in BrdU+ cell numbers in the Cxcr4 mutants at E18.5, there was not only an overall decrease in the number of Nestin-GFP+ cells () but also a corresponding increase in the number of Tbr2+ cells in the dentate field of the Cxcr4 mutants compared with the controls (). This led us to look into the possibility that dentate precursors prematurely differentiate into granule cells when they were displaced from the subpial zone. At E15.5, a robust stream of Nestin-GFP+ cells was present in the controls but it was diminished in the Cxcr4 mutants (arrows in ). Conversely, the mutants had a larger patch of Prox1+ granule cells around the FDJ than the controls. This loss of progenitors and the excess of granule cells at this early developmental stage suggested the premature differentiation of dentate progenitors upon displacement from the subpial zone.
To test this more directly, we birthdated granule cells by injecting BrdU at E15.5 and counted the number of BrdU+/Prox1+ cells at E18.5. Interestingly, the density of double-labeled cells was dramatically higher in the mutants (227±21%) compared with controls (n=4, *P<0.05, Student’s t-test). However, when BrdU was administered at E16.5, the density of BrdU+/Prox1+ cells was significantly lower in mutants Cxcr4 mutants (51.7±7.5%) than controls (n=4, *P<0.01, Student’s t-test) (). This indicates that mutant mice have an early excessive burst of production of granule neurons but fail to produce the appropriate number of granule cells only a day later. Taken together, these data indicate that progenitors displaced from the subpial zone prematurely differentiate and the localization to the subpial transient zone may be required to maintain dentate precursors in an undifferentiated state during late embryonic stages.
SGZ formation is not affected in Emx1-Cxcr4 cKO but severely compromised in Emx1-PTX mice
To examine the SGZ formation, we used a conditional knockout model by crossing the floxed Cxcr4 allele with Emx1ires-cre
(Gorski et al., 2002
) (Emx1-Cxcr4 cKO thereafter) to bypass the prenatal lethality of null Cxcr4. The prenatal development of the dentate gyrus in Emx1-Cxcr4 cKO resembled the null mutants with Tbr2 and Prox1 staining (see Fig. S3 in the supplementary material
By P5, there was distinct reorganization of the radial glial scaffolding in control mice shown by the Nestin-GFP+ cells at the SGZ, whereas in the cKO mice the Nestin-GFP+ cells were scattered throughout the dentate formation (see Fig. S4A, B in the supplementary material
). However, the distribution patterns of Tbr2+ or Ki67+ cells showed subtle differences between controls and cKOs (see Fig. S4C-F in the supplementary material
). Surprisingly, despite the prenatal abnormalities in the early GCL of the Cxcr4−/−
and the Emx1-Cxcr4 cKO, Prox1+ cells formed distinct upper and lower blades of GCLs in the cKO mice (see Fig. S4G, H in the supplementary material
). Our data suggests the granule cells are largely able to adopt appropriate layer positioning in Emx1-Cxcr4 cKO despite the early defects in SPZ.
One possible explanation for the SGZ recovery in the Emx1-Cxcr4 cKOs is that other ligand-receptor systems may compensate for the loss of Cxcr4. To test this, pertussis toxin (PTX) expression (Regard et al., 2007
) was conditionally activated by Emx1ires-cre
(Emx1-PTX thereafter), potently blocking all the trimeric Gi/o
signaling including Cxcr4. As expected, the perinatal subpial neurogenic zone did not properly form in the Emx1-PTX animals (see Fig. S7 in the supplementary material
). By P4, Nestin-GFP+ cells were distributed as patches scattered throughout the dentate () and both Ki67+ and Tbr2+ cells were chaotically dispersed throughout the whole dentate (). Prox1+ granule cells also failed to assume their distinct layered organization and many were ectopically located in the subpial region of FDJ (). Emx1-PTX animals died in the second postnatal week, so we chose P10 animals for further analysis. Compared with the controls (), Emx1-PTX animals almost complete lost organization of the BLBP+ scaffolding in the SGZ (); furthermore, Tbr2+ neurogenic precursors and the BrdU+ or Ki67+ cycling cells in the SGZ were also ectopically localized in the MZ, GCL and hilus (). In addition, the border between the hilus and Prox1+ GCL was obscured owing to ectopic dispersion of granule cells into the hilus and granule cell heterotopias were visible in the remnant of the migratory stream to the dentate (). Therefore, the formation of SGZ appears to rely on a PTX-sensitive pathway.
Compromised organization of the subgranular niche in Emx1-PTX mice
Contribution of subpial progenitors to the formation of the subgranular zone
Previous genetic fate-mapping analysis with the Gli1CreERT2
line revealed that the self-renewing stem cells in the dentate gyrus first appear at the late embryogenesis (Ahn and Joyner, 2005
). In order to test whether the SPZ progenitors may contribute to the neural stem cells settled in the SGZ, we reasoned that when tamoxifen (TM) was injected at E17.5 into the Gli1CreERT2
line in the presence of Rosa-lacZ
reporter, the labeled cells would initially emerge from the subpial zone and then spread toward the GCL from there over time. If the hilar progenitors exclusively contribute to the SGZ, we would expect the opposite. Interestingly, after 24 hours, lacZ
+ cells were first detected in the MZ (inset 1 in ) and the edge between MZ and GCL (inset 2 in ) in the upper blade. After 48 hours, lacZ
+ spread across the GCL and SGZ in the upper blade ( and inset). When tamoxifen was administered at E18.5 and lacZ
expression was analyzed 24 hours later, we found most lacZ
+ cells were restricted in the GCL of the upper blade ( and inset) and others were observed in the future lower blade (). To further analyze the cellular identity of cells produced after recombination induced at E17.5, we turned to the RosaYFP reporter line. The earliest GFP+ cells were detected at P0 () and did not express Prox1 (). In other cases, it appeared that a cell might have just divided and Prox1 could be detected in one of the GFP+ doublet cells (arrows in ). When cell fates were mapped in animals at P14, most recombined cells were Prox1+ granule cells (, inset), and a few of them showed radial glial morphology () and were co-labeled with BLBP () or GFAP (). Taken together, these findings support the idea that perinatal subpial progenitors contribute to the neural stem cells that eventually settle in the SGZ.
Contribution of subpial precursors to the SGZ