|Home | About | Journals | Submit | Contact Us | Français|
The recent discovery of short neural precursors (SNPs) in the murine neocortical ventricular zone (VZ) challenges the widely-held view that radial glial cells (RGCs) are the sole occupants of this germinal compartment and suggests that precursor variety is an important factor of brain development. Here, we use in utero electroporation and genetic fate mapping to show that SNPs and RGCs cohabit the VZ but display different cell cycle kinetics and generate phenotypically different progeny. In addition, we find that RGC progeny undergo additional rounds of cell division as intermediate progenitor cells (IPCs) while SNP progeny generally produce postmitotic neurons directly from the VZ. By clearly defining SNPs as bona fide VZ residents, separate from both RGCs and IPCs, and uncovering their unique proliferative and lineage properties, these results demonstrate how individual neural precursor groups in the embryonic rodent VZ create diversity in the overlying neocortex.
The ventricular zone (VZ) of the dorsal telencephalon contains the progenitor cells which produce all of the various excitatory neurons of the mature neocortex. This process occurs during prenatal mammalian brain development through a precisely regulated series of proliferative and neurogenic divisions. Whether the diversity in neuronal progeny is generated from a similarly heterogeneous pool of precursor cells remains unclear. In fact, numerous studies suggest that the rodent VZ is composed predominantly of a single, multipotent cell type: radial glial cells (RGCs) (E. Hartfuss et al., 2001; T. Miyata et al., 2001; S. C. Noctor et al., 2001; S. C. Noctor et al., 2002; T. E. Anthony et al., 2004; L.Pinto et al., 2008). However, this view contrasts with recent studies in which another cell type, termed short neural precursors (SNPs), was identified in the mouse VZ which differs morphologically and molecularly from RGCs (J. S. Gal et al., 2006; K. Mizutani et al., 2007). Furthermore, primate studies have clearly demonstrated a mix of committed neuronal and glial precursors in the VZ of human and monkey (P. Levitt et al., 1981; N. Zecevic, 2004; B. Howard et al., 2006; Z. Mo et al., 2007). Delineating the composition of the rodent VZ is therefore important both for understanding how neural stem cells and progenitors properly form the cerebral cortex as well as for elucidating possible mechanisms of species-specific diversity.
In addition to the neural precursors in the VZ, a separate class of neuronal progenitors has been described in the overlying subventricular zone (SVZ). These intermediate progenitor cells (IPCs) are generated from RGCs (S. C. Noctor et al., 2004) and are now considered a secondary progenitor population within the rodent and primate neocortex. Although IPCs reside in the SVZ, there is some physical intermixing between IPCs and RGCs in the basal VZ (T. E. Anthony et al., 2004; W. Haubensak et al., 2004; T. Miyata et al., 2004). A small number of IPCs have also been shown to divide at the ventricular surface, where they display morphological characteristics similar to SNPs (S. C. Noctor et al., 2008; T. Kowalczyk et al., 2009). Therefore, it is unclear if SNPs are a distinct population or whether they simply represent a subset of the IPC population.
To resolve these issues, we developed powerful new methods allowing cell cycle and lineage analyses of multiple contiguously cycling populations in vivo. Here we show that the cell cycle kinetics, specifically G1-phase duration, of SNPs and RGCs are markedly different. Moreover, genetic fate-mapping demonstrates that SNPs and RGCs give rise to distinct neuronal lineages due to differential reliance on IPC amplification. These data for the first time uncover the proliferative differences between SNPs and RGCs and demonstrate how these differences lead to diversity of neuronal daughter cells during neurogenesis.
In utero electroporation (IUE) surgeries were performed on ICR mice bred at Children's National Medical Center. Females were checked daily for vaginal plugs; the day of plug was considered embryonic day (E) 0.5. Surgeries were performed at E13.5 or E14.5. Electroporation was carried out as previously described (J. S. Gal et al., 2006). Briefly, dams were anesthetized with ketamine/xylazine cocktail and their uterine horns exposed by midline laparotomy. One microliter of plasmid DNA (3-4 μg/μl) mixed with 0.1% fast green dye in phosphate buffer was injected intracerebrally, via pulled micropipette, through the uterine wall and amniotic sac. The anode of a Tweezertrode (Genetronics) was placed over the dorsal telencephalon outside the uterine muscle. Four, 40 V pulses (50 ms duration separated by 950 ms) were applied with a BTX ECM830 pulse generator (Genetronics). When the desired number of embryos had been electroporated, the uterine horns were placed back inside the abdomen, the cavity was filled with warm physiological saline, and the abdominal muscle and skin incisions were closed with silk sutures. Dams were placed in clean cage to recover and monitored closely. These procedures conform to United States Department of Agriculture regulations, and were approved by the Children's National Medical Center IACUC.
The CAG-RFP plasmid expresses red fluorescent protein under the control of the chicken beta actin promoter (gift from J. LoTurco). The Tα1-hGFP plasmid (gift from S. Goldman) expressed humanized GFP under the control of the Tα1 promoter in short neural precursors (SNPs). The glutamate-aspartate transporter (GLAST) promoter (gift from D.J. Volsky) was subcloned into a promoterless, farnesylated eGFP plasmid (Clontech) to make the GLAST-eGFPf plasmid to label radial glial cells (RGCs). Tα1-Cre plasmid was made by subcloning the Tα1 promoter (from pTα1-hGFP) into the pBS185 Cre vector (Life Technologies); pGLAST-Cre was made by subcloning the GLAST promoter and the Cre fragment from pBS185 into the pBluescript II SK (+/-) vector (Stratagene). pNestin-Cre was a gift from C.Y. Kuan, and the GFP reporter plasmid (pCALNL-GFP), in which the GFP sequence lies downstream of a floxed Neo cassette under the control of a CAG driver, was a gift from T. Matsuda.
MAX Efficiency DH5α Competent Cells (Invitrogen) were transformed with CAG-RFP plasmid. The cells were grown in LB broth containing 50 μg/ml BrdU (Sigma), after which the plasmid DNA was concentrated and purified with the QIAGEN EndoFree Maxi Kit. The resulting BrdU-laced plasmid was injected intracerebrally and drawn into the cortical wall via IUE. Immediately following surgery the brains were removed, fixed overnight in 4% paraformaldehyde (PFA; Sigma) and cryopreserved in 30% sucrose (Sigma) solution for sectioning. To determine the depth of plasmid infiltration after IUE, sectioned brains were processed for BrdU immunofluorescence (see below). It was determined that BrdU reached approximately 160 μm into the tissue.
IUE was performed with native CAG-RFP plasmid, and single intraperitoneal BrdU injections (50mg/kg) were administered 2 hr before (-2 hr), immediately after (0hr), or 2, 4, or 6 hr after the surgery. The 0 hr injection labeled cells in S-phase at the time of electroporation, while the -2 hr injection labeled cells in S-through-M-phase at the time of IUE; other time points labeled cells in S-phase at given intervals after IUE. Animals were allowed to survive 14 hrs after IUE, at which time the mother was sacrificed and the fetuses' heads were removed, fixed overnight in 4% PFA, cryopreserved in 30% sucrose, and sectioned on a cryostat. Sectioned tissue (20 μm) was then processed for BrdU immunofluorescence (see below), using a FITC-conjugated secondary antibody (goat anti-mouse IgG1, 1:200 in 1.5% NGS in PBS; Southern Biotechnology Associates). CAG-RFP+BrdU+ (double positive) cells were counted in a 100 μm × 100 μm area on confocal Z-stacks and divided by the number of RFP+ cells in the same area to determine the percentage of double positive cells.
IUE was performed at E14.5 first using the CAG-RFP plasmid as a control to compare our experimental values to cumulative BrdU values, and subsequently with pTα1-hGFP to label SNPs and pGLAST-eGFPf to label RGCs, once the method had been validated. Single, IP BrdU injections were given at increasing intervals after IUE (one injection per mouse), and mice were sacrificed 2 hr later. After sacrifice, fetuses' heads were removed, fixed, sectioned and processed for BrdU immunofluorescence (see below). Double positive cells were counted in 100 μm × 100 μm area on confocal Z-stacks obtained with a 40× objective. The number of double positive cells was then divided by the number of XFP+ cells in the same area to calculate the labeling indices for the respective populations. The time at which the slope of the labeling index increased from zero was established as the length of G1-phase (TG1).
Tissue from the IUE/BrdU experiments was used for these analyses. The nucleic acid stain SYTO 24 (1:2000; Invitrogen) or propidium iodide (1:1000; Sigma) was used to visualize condensed chromatin on RFP+ or GFP+ sections, respectively. Mitotic electroporated cells at the ventricular surface were counted and divided by the total number of electroporated cells within 70 μm of ventricular surface to calculate the mitotic index. Re-entry of transfected (XFP+) cells into M-phase (i.e. cell cycle duration, TC) was determined to be the time at which the slope of the line increased significantly from zero.
Morphological analyses were performed on 50 μm vibratome sections from postnatal day (P) 10 brains that had been co-electroporated at E14.5 with Tα1-, GLAST- or Nestin-Cre plasmid and a floxed-stop GFP plasmid. The morphology of labeled neurons was assessed as either pyramidal or stellate as compared to Ramon y Cajal's drawings of these subtypes. Zeiss LSM Image Browser software was used to make and rotate 3-D projections of confocal Z-stacks so that entire cell soma and projections could be visualized. Cells that were not easily identifiable as either pyramidal or stellate neurons were not counted.
Immunostaining for BrdU was performed as follows: 20 μm sections were rehydrated in PBS containing 0.1% TritonX-100 (PBS-T; Sigma). After 5 min treatment with 0.1% trypsin (Sigma) in 0.1M Tris (Sigma) buffer containing 0.1% calcium chloride (Sigma), sections were incubated in 2N hydrochloric acid (Fisher Scientific) for 1 hr at room temperature. Sections were then blocked in 10% normal goat serum (NGS; Sigma) in 0.1% PBS-T for 30 min and incubated overnight in mouse anti-BrdU antibody (13 μl in 1 ml PBS-T; Becton Dickinson) at 4°C. Sections were washed 5 × 5 min in 0.1% PBS-T, incubated at room temperature for 1 hr in goat anti-mouse IgG1 antibody, washed 5 × 5 min in 0.1% PBS-T and mounted with Vectashield Mounting Medium for Fluorescence (Vector Laboratories). Sections containing CAG-RFP+ cells were processed for BrdU immunofluorescence using a FITC-conjugated secondary antibody, while GFP+ sections required a TRITC-conjugated secondary antibody (both 1:200; Southern Biotechnology Associtates).
Immunostaining for mouse anti-TUJ1 (1:500, Covance), rabbit anti-Tbr2 (1:500, Abcam), rabbit anti-Pax6 (1:1000, Covance) and mouse anti-Ki67 (1:200, BD Biosciences) was performed on 20 μm coronal sections from embryonic tissue fixed in 4% PFA and cryopreserved in 30% sucrose. Immunostaining for goat anti-Brn1 (1:100, Santa Cruz) was performed on postnatal (P10) tissue from mice perfused intracardially with ice-cold PBS followed by 4% PFA. Tissue was post-fixed overnight in 4% PFA and then sectioned on a vibratome (50 μm, Leica). In all cases, sections were blocked 1 hr at room temperature in 5% serum and 1% bovine serum albumin (BSA, Sigma) in PBS + 0.2% Triton-X100 (Sigma). Primary antibodies were applied overnight at 4°C. Sections were then washed 3 × 5 min in PBS and incubated for 1 hr at room temperature in secondary antibodies (1:200 for all). Secondary antibodies used were: AlexaFluor 546 donkey anti-goat, AlexaFluor 543 goat anti-mouse IgG2a, AlexaFluor 633 goat anti-rabbit (all Invitrogen), and TRITC-conjugated goat anti-mouse IgG1 (Southern Biotechnology Associates). Sections were washed 3 × 5 min in PBS and counterstained with ToPro-3 (1:500, Invitrogen), then coverslipped with Vectashield Mounting Medium for Fluorescence (Vector Labs).
Confocal images were acquired with a Zeiss LSM 510 Meta NLO system equipped with an Axiovert 200M microscope. GFP, SYTO 24 and FITC were excited at 488 nm; RFP, propidium iodide, AlexaFluor 543 and 546 and TRITC were excited at 543 nm; ToPro-3 and AlexaFluor 633 were excited at 633 nm. Z-stacks 5-20 μm thick, composed of 1024 by 1024 pixel, 0.7- to 5-μm-thick optical sections, were collected using 10×, 25× or 40× objectives depending on the experiment.
In order to achieve continuity across experiments, only sections from the same general area of the dorso-lateral neocortex were analyzed. Using anatomical landmarks such as the ganglionic eminences (fetal tissue) and the fimbria and CA1 of the hippocampus (postnatal tissue), sections from as close to the same rostro-caudal level as possible were used throughout.
Image analysis/cell counting was done with the Zeiss LSM Image Browser program. To determine recombined GFP+ cell allocation at E17.5, the depth of the cortical wall was measured and divided by five, yielding five bins corresponding to the VZ, SVZ, IZ and lower and upper CP. GFP+ cells in each bin were counted across a 300 μm-wide column in collapsed confocal Z-stacks, then divided by the total number of GFP+ cells in that column to give the percent of total cells in each bin. For all other experiments, cells were counted in 225 μm2 areas on images taken with the 40× objective. In the Tbr2/TUJ1/Ki67/Pax6 immunostaining images, cells were only counted within 200 μm of the ventricular surface.
Student's T-tests were performed using Microsoft Excel to determine significance between groups in most experiments. However, ANOVA (in SigmaPlot) was used for the IUE/BrdU labeling index, and a nonparametric (Median) regression was performed for the IUE/M-M phase experiment. In all cases, confidence intervals were set at 95% and p-values less than 0.05 were considered significant.
We previously showed that the mouse VZ is heterogeneous with respect to its constituent neural precursors (J. S. Gal et al., 2006). Using time-lapse multiphoton imaging, it was determined that SNPs retract their basal processes during division at the ventricular surface, while RGCs maintain their basal fibers during mitosis (T. Miyata et al., 2001; S. C. Noctor et al., 2001; S. C. Noctor et al., 2002). SNPs and RGCs were also distinguished based on their abilities to express GFP driven by the tubulin alpha-1 (Tα1) or glutamate aspartate transporter (GLAST) promoters, respectively. It was also recently found that VZ cells expressing the Tα1 and GLAST promoters differentially utilize the Notch signaling pathway (K. Mizutani et al., 2007). However, because these were the only known characteristics distinguishing SNPs from RGCs, we sought to identify further differences between these precursor populations, starting with their cell cycle kinetics.
The physical intermixing of SNPs and RGCs in the VZ presented challenges for measuring potential differences in their proliferation in vivo. Cumulative BrdU labeling, a common technique for estimating cell cycle kinetics in vivo which labels all S-phase cells indiscriminately (R. S. Nowakowski et al., 1989), cannot distinguish between SNPs and RGCs. Since SNPs and RGCs were previously distinguished by their preferential promoter expression via in utero electroporation (IUE), this method seemed ideal to adapt for in vivo proliferation kinetics studies. However, despite the increasing use of IUE in developmental neuroscience, very little is known about the mechanics and temporal characteristics of IUE-induced transfection.
In order to use IUE in quantitative measurements of cell proliferation, both the location and the proliferative nature of the starting population of cells needed to be determined. As most IUE is performed with plasmid vectors requiring 10-18hr delays before DNA expression is detectable, it has not been possible to estimate the initial penetration depth of electroporated plasmid since cell division, migration and apoptosis can all occur within this time frame. Thus, to uncover the location of DNA immediately after electroporation, we created a plasmid which did not require DNA transcription and translation to be detectable by physically tagging the plasmid with BrdU (Supp. Fig. 1B; see Methods section). Immediately following IUE with this BrdU-laced plasmid, we found that all VZ cells within the electroporated area of the neocortical wall were physically exposed to the delivered plasmid up to a depth of 160μm from the surface of the ventricle (Supp. Fig. 1C).
The next steps were to determine whether all cells physically exposed to the plasmid also express the exogenous DNA and to uncover which cell cycle phases may be necessary for transfection. To do so, we tagged individual precursor groups by phase location using systemic injections of BrdU at various intervals before or after IUE at E14.5 with the ubiquitously expressed CAG-RFP plasmid. For example, injecting BrdU immediately following the IUE procedure labeled VZ cells in S-phase at the moment of electroporation. In this pulse labeling procedure, quantification of the percent of transfected (RFP+) cells which were also BrdU+ revealed that 70.87 ± 6.60% were co-labeled (Fig. 1Ai, D), strongly suggesting a link between S-phase and plasmid expression. When BrdU was administered 2 hr prior to IUE, which BrdU labeled cells from S-phase into M-phase at the moment of electroporation (Fig. 1Aii), there was almost complete co-labeling of RFP+ cells with BrdU (95.97 ± 4.00%, Fig. 1B, B′, D). Thus, despite the fact that the entire VZ appears to be exposed to plasmid DNA upon IUE, only the cells arrayed through S- and M-phases at the moment of electroporation eventually express the plasmid. To determine the length of time after IUE that transfection and plasmid expression is still possible within the VZ, we allowed longer gap periods between IUE and subsequent BrdU injection. We found that increasing the time between IUE and BrdU administration resulted in steadily decreasing proportions of co-labeled cells (Fig. 1D). Thus, the further away from S-M phases the cells are, the less likely they are to become transfected. From these experiments, we conclude that IUE preferentially transfects cells which transit through S- and M-phases of the cell cycle within 8 hr of the surgery, and that the vanguard of the transfected population is in M-phase at the moment of electroporation.
Subsequent tracking of this temporally-limited cell cohort allowed measurements of cell cycle phase durations. We developed an IUE/BrdU protocol in which IUE with the CAG-RFP plasmid was followed by BrdU injections at different intervals to estimate the duration of G1-phase (TG1) as the time necessary for the electroporated cells to re-enter S-phase (Fig. 2A). A labeling index was created by quantifying (BrdU+RFP+)/RFP+ cells in a 100 × 100 μm2 area in confocal images (Fig. 2B). Our IUE/M-M phase protocol estimated the total cell cycle duration (TC) as the time required for the electroporated VZ cell cohort, the leading edge of which is in M-phase at the moment of electroporation, to re-enter the next mitosis at the surface of the ventricle (Fig. 2A). A mitotic index was created by counting the number of mitotic (visibly condensed chromatin) RFP+ cells at the ventricular surface, divided by the total number of RFP+ cells (Fig. 2C). Estimates of TG1 and TC acquired using the IUE/BrdU and IUE/M-M phase protocols with CAG-RFP (12 hr and 20 hr, respectively) were consistent with previous estimates of the general VZ population (11.8 hr and 17.5-18.4 hr, respectively (T. Takahashi et al., 1995)). Therefore, we conclude that these novel IUE-based methods represent sensitive and robust alternatives to cumulative BrdU labeling for analyses of cell cycle kinetics.
To measure potential differences in cell cycle kinetics between the individual SNP and RGC VZ cell types, we used the IUE/BrdU and IUE/M-M phase methods with cell type-specific pTα1- and pGLAST-GFP reporter plasmids. Quantification of (BrdU+GFP+)/GFP+ cells for both the Tα1 and GLAST time courses resulted in significantly different labeling curves (Fig. 2B′; ANOVA, p<0.0001). Notably, S-phase re-entry for SNPs was delayed by 4 hrs compared to RGCs (16 hrs vs. 12 hrs), indicating that TG1 is 33% longer for SNPs. The IUE/M-M phase analysis uncovered a statistically significant 4 hr delay in M-phase re-entry for SNPs compared to RGCs: 24 hrs vs. 20 hrs (Fig. 2C′; median regression, p<0.009). Together, these data suggest that the increased cell cycle duration in SNPs is due specifically to a lengthened G1-phase. This is the first demonstration of cell cycle differences in the heterogeneous population of mammalian VZ precursor cells. In addition, the biphasic curves for the IUE/BrdU and IUE/M-M phase studies (Fig. 2B′, C′) demonstrate that some SNPs are retained in the VZ, since they returned to the ventricular surface for at least two successive divisions after beginning to express the exogenous plasmid.
To ensure that potential differences in strength between the GLAST and Tα1 promoters were not biasing the results (i.e. driving transcription and translation of GFP more quickly in one cell type than the other), we examined GFP expression in tissue harvested 10 hr post IUE with either pGLAST-eGFPf or pTα1-hGFP (Supp. Fig. 2). Both RGCs (Supp. Fig. 2A) and SNPs (Supp. Fig. 2B) displayed strong, widespread GFP expression. We therefore conclude that there are no temporal differences in promoter expression which could bias the proliferation results. Together, these kinetic experiments confirm that the murine VZ contains multiple types of resident precursors which are distinguished by significant differences in proliferation dynamics.
We next asked how these differences in SNP and RGC cell cycle kinetics affect the growth of the neocortical wall and the allocation of SNP- and RGC-derived cells to the cortical mantle. Previous studies have shown that VZ cells with longer cell cycles tend towards neurogenic, rather than proliferative, divisions (F. Calegari et al., 2005). We therefore hypothesized that SNPs would undergo more direct neurogenic divisions than RGCs, quickly producing neurons that migrate away from the VZ. To test this, we used Cre/Lox-based genetic fate mapping to label the progeny of SNPs and RGCs at mid-neurogenesis and analyzed their distribution across the neocortical wall. Co-electroporation of promoter-specific Cre plasmids and floxed stop GFP reporter plasmid was used at E13.5 to label RGCs (GLAST promoter), SNPs (Tα1 promoter), or all neural stem cells (NSCs, Nestin promoter; Fig. 3A). By E17.5, there were noticeable differences in the allocation of labeled (GFP+) progeny of RGCs and SNPs (Fig. 3B). While less than a third of the total cells produced from each precursor cell type remained in the proliferative areas of the cortical wall (VZ and SVZ; Fig 3C), the percent of RGC progeny in the SVZ was nearly twice that of SNP progeny (18.12 ± 2.66% vs. 9.09 ± 3.43%; Fig. 3C). SNP progeny were primarily localized in the lower half of the cortical plate (40.76 ± 7.05% of GFP+ cells; Fig. 3C) while RGC progeny occupied the upper half (35.23 ± 2.96% of GFP+ cells; Fig. 3C). By comparison, Nestin+ NSCs generated cells that were evenly distributed across the depth of the neocortical wall (Fig. 3B, C), as expected for a general precursor population comprised by both SNPs and RGCs.
Given the inside-out laminar specification of the mammalian neocortex (P. Rakic, 1974), these results suggest either that: 1) SNP-generated neurons are born earlier than RGC-generated neurons and are thus specified to deeper cortical laminae, or 2) SNP-generated neurons are born later and have not finished migrating by E17.5. To test these possibilities, we conducted a longer-term study, performing the IUE-mediated fate mapping at E14.5 and analyzing the final laminar position of the recombined GFP+ neurons on postnatal day 10 (P10; Fig. 4). By measuring the depth of GFP+ neuronal soma from a reference line drawn at the top of layer II/III (determined by immunohistochemistry for Brn1) we found that neurons generated from SNPs reside significantly deeper than neurons from RGCs (Fig. 4B-D). The average distance of GFP+ RGC neuronal progeny, which were found in the lower half of layer II/III, was 236.00 ± 8.58 μm (Fig. 4D), while GFP+ SNP progeny resided more than 100 μm deeper, primarily in layer IV (avg. distance 361.83 ± 21.59 μm; Fig. 4D). By comparison, neurons generated from pNestin+ NSCs displayed a more diffuse arrangement than the tightly packed bands of RGC or SNP progeny (Fig. 4A-A″), with individual neurons spread throughout the depth of layer II/III (avg. distance 163.60 ± 8.76 μm; Fig. 4D). We also classified all fate-mapped progeny based on morphology to determine whether individual VZ precursor types generate specific neuron subtypes. We found that each of the three precursor populations generated both pyramidal and stellate neurons during this stage of neurogenesis (Supp. Fig. 3) and that most of the GFP-labeled pyramidal neurons sent callosum-projecting axons. These data indicate that RGCs and SNPs labeled at E14.5 produce a similar variety of neuron subtypes, but that these neurons are surprisingly specified to different cortical laminae with progeny of SNPs predominantly in layer IV and progeny of RGCs mostly in layer II/III.
To uncover the mechanism underlying the different positioning of SNP and RGC progeny, we expanded our focus to include IPCs, reasoning that the time of specification for SNP- and RGC-derived neurons may be different if one subgroup is sequestered within the IPC pool. We performed immunohistochemistry for markers of proliferation (Ki-67), differentiation (TUJ1) and for the transcription factor T-box brain 2 (Tbr2), which labels IPCs (C. Englund et al., 2005), on tissue electroporated at E14.5 with pTα1- or pGLAST-GFP and sacrificed 24 hr later. We found that 60% more pGLAST+ RGCs co-expressed Tbr2 compared to pTα1+ SNPs (Fig. 5A, A′, D; Table 1). Conversely, two-thirds more SNPs expressed the neuronal marker TUJ1 compared to RGCs (Fig. 5B, B′, D; Table 1). Differences in proliferation (GFP+/Ki67+ cells) between RGCs and SNPs were also apparent, but not significant (Fig. 5D; Table 1). Together with the results from the fate mapping experiments, these data support the hypothesis that SNPs in the VZ serve a direct neurogenic role while RGCs primarily generate neurons indirectly, via IPCs.
Finally, we examined expression of the transcription factor paired box gene 6 (Pax6), which labels VZ precursors (M. Gotz et al., 1998) and is rarely co-expressed in Tbr2+ cells (C. Englund et al., 2005). Previous studies identified Tbr2+ cells dividing at the ventricular surface which are morphologically similar to SNPs (S. C. Noctor et al., 2008; T. Kowalczyk et al., 2009), implying that SNPs may be a misclassified population of IPCs. In tissue electroporated with pTα1- or pGLAST-GFP and stained for Pax6, we found that both pTα1+ SNPs and pGLAST+ RGCs were robustly labeled with Pax6 (Fig. 5C-D), supporting our proliferation study which indicates that SNPs are a resident VZ population (Fig. 2). In addition, in our immunostainings for Tbr2, we did not observe even a single pTα1+/Tbr2+ cell (n=1707 pTα1+ cells, n=75 pTα1/Tbr2+ cells) dividing at the ventricle; most pTα1+/Tbr2+ cells were in the basal VZ or SVZ. Thus, we conclude that SNPs and IPCs are distinct precursor cell types.
In this study we used molecular labeling techniques and novel IUE-based methods to uncover diversity in the proliferative properties and lineage potentials of separate VZ precursor populations. We present three main findings which fundamentally increase our knowledge of the rodent neocortical VZ. First, in contrast to the long-standing view that all VZ cells exhibit similar proliferation parameters (T. Takahashi et al., 1995; L. Cai et al., 1997), we demonstrate that two populations dividing concurrently at the ventricular surface have significantly different cell cycle kinetics. Second, we find that neurons fate-mapped from these two populations migrate to different cortical laminae. Specifically, SNPs and RGCs fate mapped on E14.5 generate neurons that predominantly settle in layer IV and layer II/III, respectively. Finally, we show that this differential laminar allocation is due, at least in part, to an increased reliance of RGCs on intermediate progenitor cells for neuronal production compared to SNPs. From the significant and persistent differences in laminar positioning, we conclude that SNPs and RGCs in the murine VZ represent mutually exclusive populations from E14.5 onwards and generate phenotypically different neuronal progeny via separate mechanisms.
Our data indicate that SNPs labeled at E14.5 transit through one or two cell cycles at the ventricular surface, directly producing TUJ1+ neuron(s) with each division which migrate away to the cortical plate (Figs. 2B′, 2C′, ,5).5). In comparison to SNPs, RGCs labeled at E14.5 display greater perdurance in the VZ, undergoing multiple rounds of asymmetric divisions to generate a proliferative daughter cell and a Tbr2+ IPC with each mitosis (S. C. Noctor et al., 2004). We found that progeny of the labeled RGCs sojourn in the SVZ as IPCs before migrating to the cortical plate, thus arriving later than—and laminating superficially to—neuronal daughters of SNPs (Fig. 6). Though our data do indicate that some SNPs utilize IPCs for neuronal amplification, just as some RGCs are capable of generating neurons directly (T. Miyata et al., 2001; S. C. Noctor et al., 2001), the results suggest that direct and indirect neuronal production are the primary mechanisms employed by SNPs and RGCs, respectively. Not surprisingly, these different modes of neuron generation influence the resulting size of the neuronal progeny: there were considerably more GFP+ RGC-derived neurons than SNP-derived neurons in the brains we examined (see Fig. 4B-C″; avg. number GFP+ cells per volume analyzed: pGLAST = 135.25 ± 20.61, pTα1 = 45.5 ± 8.94). Thus, SNPs augment neuronal output from the RGC population and thereby enhance the overall neuronal production capacity of the VZ. It is therefore tempting to speculate that the SNP pool may have evolved to boost neuronal production in a discrete, region-specific manner. We previously found that SNPs are present in substantial numbers within the murine VZ between E13.5 and E16.5 (J. S. Gal et al., 2006), the birth period of neurons destined to the mid- to superficial layers of the neocortex. Our current fate mapping results indicate that the SNP pool provides focused and temporally limited neurogenesis during this period, most likely generating multiple and successive cohorts of neurons to each developing cortical layer. Whether SNPs play a significant role in neurogenesis before this period (E13.5-16.5) is currently unknown, but it is possible that SNP prevalence may differ across developmental ages and cortical regions and thereby help to finely tune laminar growth across the neocortical mantle.
The results presented here complement and extend work demonstrating a link between the cell cycle of VZ precursors and the laminar position of their neuronal progeny (S. K. McConnell and C. E. Kaznowski, 1991; L. J. Pilaz et al., 2009). McConnell's seminal heterochronic transplantation study established that laminar fate is determined prior to the terminal division of neuronal precursors (S. K. McConnell and C. E. Kaznowski, 1991). Our fate mapping data expand upon this important study, showing that the time and location of terminal division of simultaneously labeled precursors varies among individual precursor groups. Since the terminal division of a VZ precursor's lineage can either occur at the ventricular surface (as in the case of most SNPs) or in the SVZ (as in the case of most RGC descendant cells), laminar specification is therefore critically dependant on the site of terminal division on a cell-by-cell basis. Recently, the Dehay lab found that forced overexpression of cyclin D1 or E1 in E15 mouse VZ precursors can shorten TG1 and result in an accumulation of neurons in cortical layer II/III and the upper half of layer IV at P15. This differed significantly from the neuronal progeny of control precursors, which primarily resided deeper, in layer IV (L. J. Pilaz et al., 2009). Here we demonstrate that the native embryonic mouse VZ contains a heterogeneous pool of endogenous neural precursors which differ both in cell cycle kinetics and whether they undergo secondary proliferation as IPCs. These differences naturally predispose SNPs and RGCs to settle in different laminae.
While numerous studies have identified variations in morphology and antigen and gene expression within the RGC population (P. Malatesta et al., 2000; E. Hartfuss et al., 2001; E. Hartfuss et al., 2003; P. Malatesta et al., 2003; L. Pinto et al., 2008), our results add to a growing body of evidence suggesting that multiple neural precursor cell types contribute to heterogeneity in the rodent VZ. For example, it has been shown that VZ precursors differentially utilize the Notch signaling pathway (A. Kawaguchi et al., 2008). In particular, Notch pathway activation is present in pGLAST+ RGCs but is largely absent in pTα1+ SNPs (K. Mizutani et al., 2007). As Notch activity is known to promote proliferation and inhibit differentiation in both invertebrate and vertebrate systems (N. Gaiano and G. Fishell, 2002), this finding suggested that pGLAST+ RGCs may remain proliferative while pTα1+ SNPs differentiate more quickly. Indeed, the authors demonstrated in vitro that pTα1-EGFP+ cells generated fewer and smaller neurospheres than cells with active Notch signaling (K. Mizutani et al., 2007). Here we present in vivo confirmation of this hypothesis, and identify a functional consequence of these differences in gene expression: namely, SNPs generate neurons on an accelerated time scale but in more limited numbers compared to RGCs.
Diversity of the neural precursor pool is likely a critical component of cortical morphogenesis and function. For example, the mature human brain contains up to 100 billion neurons, roughly one-fifth of which are found in the neocortex (B. Pakkenberg and H. J. Gundersen, 1997). The majority of these are excitatory projection neurons which are further grouped into subtypes based on their morphology, location, connectivity and function. Evidence from prior studies strongly indicates that these major aspects of neuronal fate are specified in the germinal zone (P. Rakic, 1988; E. M. Miyashita-Lin et al., 1999; S. N. Sansom et al., 2005); thus, the full extent of neocortical diversity is thought to require a similarly heterogeneous neuronal precursor population. Data supporting this premise are found in several studies which demonstrate multiple precursor populations co-existing in the human and non-human primate dorsal telencephalic VZ that differ in morphology, antigen expression and regional and temporal prevalence over the course of corticogenesis (P. Levitt et al., 1981, 1983; N. Zecevic, 2004; B. Howard et al., 2006; Z. Mo et al., 2007). Though the number of cortical neurons in rodents is orders of magnitude smaller than in humans (an estimated 4 million in mice (G. Roth and U. Dicke, 2005)), their degree of phenotypic diversity is comparable. Here we provide direct in vivo kinetic evidence of two different proliferative populations intermingled within the murine VZ, demonstrating a fundamental level of similarity between the rodent and primate embryonic neuronal precursor pools. VZ precursor diversity is therefore likely a common mammalian trait necessary to generate the full complement of mature neuronal phenotypes in the neocortex.
We thank J. Corbin and V. Gallo of CNMC for critical reading of the manuscript, and members of the Haydar and Corbin labs for technical help. E.K.S. is a predoctoral student in the Molecular Medicine program of the Institute for Biomedical Sciences at the George Washington University. This work is from dissertation research to be presented to the program in partial fulfillment of the requirements for the Ph.D. degree. This work was supported by NS051852 (T.F.H) and the Cellular Imaging and Statistical Cores of the IDDRC at CNMC (P30HD40677).