|Home | About | Journals | Submit | Contact Us | Français|
Prior studies have shown that oligodendroglial progenitor cells (OPCs) can give rise to neurons in vitro and in perinatal cerebral cortex in vivo. We now report that OPCs in adult murine piriform cortex express low levels of doublecortin (DCX), a marker for migratory and immature neurons. Additionally, these OPCs express Sox2, a neural stem cell marker, and Pax6, a transcription factor characteristic of progenitors for cortical glutamatergic neurons. Genetic fate-mapping by means of an inducible Cre-LoxP recombination system proved that these OPCs differentiate into pyramidal glutamatergic neurons in piriform cortex. Several lines of evidence indicated that these newly formed neurons became functionally integrated into the cortical neuronal network. Our data suggest that NG2+/PDGFRa+ Plp promoter expressing progenitors generate pyramidal glutamatergic neurons within normal adult piriform cortex.
The production of new neurons in the adult mammalian brain is predominantly restricted to two areas, the subventricular zone (SVZ) and the hippocampal subgranular zone (SGZ) (Gage, 2000). Whether neurogenesis continues in the normal adult cerebral cortex, either from SVZ neural stem cells (NSCs) (Gould et al., 1999; Inta et al., 2008), or from intrinsic neural progenitors within cortex (Magavi et al., 2000; Dayer et al., 2005) remains controversial (Gould 2007), though formation of new cortical interneurons from subventricular zone or intrinsic cortical progenitors following cerebral injury has been well documented (Magavi et al., 2000; Arvidsson et al., 2002; Tsai et al., 2006; Ohira et al., 2010). In the present study, we provide genetic fate mapping evidence for the de novo formation of glutamatergic pyramidal neurons from oligodendroglial progenitor cells (OPCs) within the normal adult murine piriform cortex.
OPCs, identifiable by their expression of the proteoglycan NG2 and the platelet-derived growth factor receptor PDGFRα, have been assumed till recently to give rise only to oligodendroglia in intact immature and adult mammals, though these progenitors are known to be capable of generating neurons and NG2+ “type 2 astroglia” in vitro in the presence of appropriate morphogens (Raff et al., 1983, Kondo et al., 2000), and in vivo after transplantation (Belachew et al., 2003; Aguirre et al., 2004). Abundant NG2+ cells persist in CNS white and gray matter after myelination is completed (Dawson et al., 2003). Some extend processes to nodes of Ranvier (Butt et al., 1999), receive glutamatergic synaptic inputs from local neurons, and generate immature action potentials (Bergles et al., 2000; Chittajallu et al., 2004; Karadottir et al., 2008; Mangin et al., 2008; Etxeberria et al., 2010; De Biase et al., 2010). The functional role of this synaptic input to these OPCs remains unknown, but in vitro studies have shown that glutamate inhibits differentiation of OPCs to oligodendrocytes (Gallo et al., 1996; Yuan et al., 1998).
In vivo genetic fate-mapping studies of OPCs by use of oligodendroglial lineage-specific Cre transgenes have not yielded consistent results. Using a platelet-derived growth factor alpha receptor (PDGFRα) promoter-driven Cre, Rivers et al (2008) observed neuronal generation in adult piriform cortex from OPCs, but other investigators, using NG2- or Olig2-promoter-driven Cre transgenes, did not (Dimou et al., 2008; Zhu et al., 2008).
Taking advantage of proteolipid protein (Plp) promoter activity in OPCs to drive expression of a tamoxifen-inducible Cre transgene, we previously reported that NG2+/PDGFRa+ Plp promoter expressing progenitors (“NG2+/PDGFRa+ PPEPs”) give rise to neurons in neonatal mouse forebrain (Guo et al., 2009). We now demonstrate that mature glutamatergic pyramidal neurons are generated in adult piriform cortex from adult NG2+/PDGFRa+ PPEPs that express the PLP promoter and markers for neural stem cells (Sox2) and neuronal progenitors (doublecortin and Pax6), and that these neurons become functionally integrated into CNS circuits.
The Plp-CreERT2 mice (Doerflinger et al., 2003) and Rosa26-STOP-EYFP recombination reporter line (Srinivas et al., 2001) were purchased from The Jackson Laboratory and maintained in C57BL/6 background. The hGFAP-Cre-ERT2 mice were from the Vaccarino colony at Yale University. Plp-CreERT2 and hGFAP-Cre-ERT2 mice were bred to reporter mice Rosa26-STOP-EYFP to yield Plp-CreERT2 / Rosa26-EYFP (PCE/R) and hGFAP-Cre-ERT2 / Rosa26-STOP-EYFP (GCE/R) double transgenic mice. The Rosa26-STOP-EYFP transgene in both PCE/R and GCE/R mice was maintained as homozygous. Both male and female mice were used in our experiments, since we detected no sex differences with respect to Cre induced recombination and neurogenesis. Mice were caged in a 12 h light/dark cycle with free access to food and water. Mouse genotypes were ascertained by Transnetyx Inc. All animal procedures were performed according to guidelines of the Institutional Animal Care and Use Committee, University of California, Davis.
Tamoxifen (TM) (T5648; Sigma-Aldrich) was dissolved in an ethanol/sunflower seed oil (1:9) mixture at a concentration of 30 mg/ml. Early adult mice (postnatal day 45~60, P45~60) were intraperitoneally (i.p.) treated with TM, twice daily for five consecutive days. The dose of injection in the morning was 1.2 mg (40 μl) and that in the afternoon was 1.5 mg (50 μl). With this TM dosage schedule, we obtained highest recombination efficiency with no lethality. No EYFP expression was detected by direct or antibody-amplified fluorescence microscopy of PCE/R mice treated with vehicle only (mixture of ethanol and sunflower seed oil, 1:9) (Fig. S 1 A1-B2 in Guo et al., 2009). EYFP appeared in OPCs and oligodendroglia as early as 12 h after first TM injection, the earliest time-point we assessed.
Eight week postnatal C57Bl mice were used for BrdU cumulative labeling experiment. For 2 h pulse labeling of mitotic cells, mice were i. p. injected with BrdU solution (100mg/kg body weight in sterile PBS at 10 mg/ml). For long term labeling of mitotic cells, BrdU was dissolved in the drinking water (1mg /ml), and mice were given free access to the water for as long as 20 days. BrdU labeling periods of 25 days or more were not used, since we found that after 25 days mice showed evidences of toxicity, for example, loss of hair and body shaking. Brain tissues were analyzed after 2, 4, 6, 8, 10, 14 and 20 days of BrdU labeling to calculate the growth index of OPCs (Nowakowski et al., 1989).
The brain tissues were analyzed on 1, 5, 10, 17, 30, 60, 90, 150 and 182 days after their last TM injection (days post-TM), with 3-8 mice at each time-point. After anesthesia with ketamine (150 mg/kg body weight)/xylazine (16 mg/kg body weight), mice were perfused transcardially with 1x PBS and then with 4% paraformaldehyde (PFA) in PBS. Brains were harvested, postfixed in 4% PFA in PBS for 1~2 h at room temperature (RT), then cryoprotected in 30% sucrose (v/v) prepared in 1x PBS at 4°C until tissues sink down, usually for 1~2 overnight, and then transferred to OCT compound for embedding on ethanol/dry ice.
In this study, we focused on the posterior piriform cortex from Bregma −1.06 mm to Bregma −2.3 (Franklin and Paxinos, 2008), about 1.3 mm spanning the posterior forebrain. For analyzing the subventricular zone (SVZ), we collected sections from Bregma 0.98 mm to Bregma 0.02 mm (Franklin and Paxinos, 2008). Brains were cut coronally on a Leica cryostat (model CM3050 S) to prepare 14-μm-thick sections, which were stored at −80°C until use.
The immunostaining methods have been described previously (Guo et al., 2009), with a few modifications. Briefly, sections were air-dried for at least 30 min at RT, followed by incubation in normal serum blocking solution (dependent on the secondary antibodies used) at RT for at least 1 h (5 % normal serum plus 0.1% Triton X-100 in 1x PBS). When a streptavidin biotin detection system was used, the section was treated with a streptavidin–biotin blocking kit (SP-2002; Vector Laboratories) before normal serum blocking. Primary antibodies were diluted in normal serum blocking solutions and incubated at 4°C overnight or 37°C for 2~3 h, followed by three 20 min washes in PBS plus 0.1% Triton X-100. After incubation with secondary antibodies (all from Jackson ImmunoResearch Inc.) in normal serum blocking solution, sections were washed three times (20 min each) in PBS plus 0.1% Triton X-100 at RT. For streptavidin–biotin detection system, DyLight 488-, DyLight 549- (from Jackson ImmunoResearch Inc.), or Pacific blue-SA (from Invitrogen), diluted in PBS, were incubated for 15 min at RT, followed by three 10 min washes in PBS plus 0.1% Triton X-100 at RT. Finally, DAPI was used to label nuclei, and the sections were mounted with Vectashield mounting medium for fluorescence (Vector Laboratories; H-1000).
In our previous study (Guo et al., 2009), we used a DCX antibody (from Abcam) at 1:250 dilution, which yielded a high background and possibly masked low level OPC DCX expression. In the present study, we optimized immunostaining with this antibody by using 1:400 ~ 1:600 dilutions. This reduced background without affecting overall signals. To evaluate the specificity of DCX immunostaining, peptide absorption experiments were conducted. In brief, DCX antibody (1:500) was incubated with DCX peptide (cat# ab19804, Abcam) or an unrelated peptide (NMDA receptor subunit 1 peptide, Santa Cruz cat# sc-1467P) at 4°C overnight. Tissue sections were then immunostained with these DCX antibody/peptide mixtures, with untreated DCX antibody as a positive control. Results (Fig. S2 A-C) demonstrated that OPC immunostaining by DCX antibody was specific.
For BrdU staining, we used HCl instead of heating to denature double-stranded DNA, since immunoreactivity of EYFP is destroyed by high temperature. After all the immunostaining steps except DAPI staining were completed, sections were post-fixed with 2% PFA in 1x PBS at RT for 15 min, and then denatured in 2N HCl at 37°C for 45 min. After three 5 min washes in PBS, the sections were incubated with anti-BrdU (Rat, Santa Cruz, 1:100) diluted in normal serum blocking solution at 4°C overnight or 37°C for 3 h. Signals from BrdU staining were compared with those obtained with EdU to evaluate the efficacy of our BrdU immunostaining protocol, since EdU signals were visualized via chemical reaction without any DNA denaturing procedures. Eight week postnatal C57Bl/6 mice were given six i.p. doses of BrdU and EdU mixture (same molar concentrations of each) and killed 2 hr after the last injection. Tissue sections were color-developed with EdU first and then immunostained with anti-BrdU. The overlap of BrdU and EdU signals (Fig. S1 J and K) validated the efficiency of our BrdU immunostaining protocols.
Primary antibodies were as described in our previous study (Supplemental Table 1 in Guo et al., 2009) except for the following: Sox2 (Rabbit, 1:200), Pax6 (rabbit, 1:200), SMI312 (Mouse, 1:1000) from Covance; Sox10 (Goat, 1:100), c-Fos (Rabbit, 1:100) from Santa Cruz; NR1 (Rabbit, 1:200), ChAT (Rabbit, 1:500) from Millipore; Iba-1 (Rabbit, 1:1000) from Wako; CD11b (Rabbit, 1:80) from BD Pharmingen; and Tbr1 (Rabbit, 1:50 from Santa Cruz; Rabbit, 1:200 from Abcam). As negative controls for immunohistochemistry, normal IgG (with final concentration 10 μg /ml, usually twice that of the primary antibodies) from corresponding species were incubated in parallel with the primary antibodies to exclude non-specific binding.
NG2+ cultured OPCs were purified from neonate mouse forebrain using methods described in Horiuchi et al., 2010. DCX and PDGFRa double immunostaining was conducted on primary cultured OPCs (Horiuchi et al., 2010).
A Nikon Eclipse C1 confocal laser-scanning microscope was used to image DyLight488 (green), DyLight 549 (red) and DAPI or Pacific Blue (blue). Optical sections were acquired using 20x [numerical aperture (NA), 0.75], 40x (NA, 1.30), or 60x (NA, 1.40) oil objective lens with Nikon EZ-C1 software, version 3.8. The Nikon EZ-C1 3.20 FreeViewer was used to create single-channel views, merged views, and orthogonal views of the images, and Photoshop CS3 was used to combine and label the images, which were exported from EZ-C1 3.20 FreeViewer without any manipulation of contrast. We considered two markers as colocalized only if this colocalization extended from the top to bottom of the z-section images.
For cell counting, we counted the whole area of piriform cortex in 14 μm thick coronal sections. Under 20 x lens, about 5~6 optical fields covered the entire area of the piriform cortex. Six to eight coronal sections located within Bregma −1.06 mm to Bregma −2.3 (about 90 sections collected) from each brain (three to eight animals for each time point) were examined. The number of EYFP+ neurons in layers II and III were quantified; EYFP+ neurons were absent from layer 1. Other quantifications included the whole piriform cortex. For quantification of every defined profile, at least 200 cells were assessed.
We quantified the expression of DCX using Nikon NIS-Elements AR 3.10 software. In brief, DCX signal pixels (range of pixel intensity 0-256) from individual DCX positive cells (considered as a region of interest (ROI)) were measured using the Auto Detection tool, and exported into Excel files. Average DCX expression levels were calculated by dividing “Sum Intensity” by “ROI Area”. This DCX expression level analysis revealed a two-peak distribution pattern, with peak expression levels of 51 and 122, respectively.
To quantify apical dendritic tree length, tissue sections from 30, 90 and 150 days post-TM were immunostained with EYFP in parallel using a biotin-streptavidin detection system to amplify EYFP signals. The distance between the protrusion of the apical dendritic trunk and the end of the apical dendritic tree in piriform cortex layer I, visualized by EYFP signals, was measured using Nikon NIS-Elements AR 3.10 software. At least 30 EYFP+ neurons were examined in this way at each time point.
All data in this study are reported as Mean ± S. D. Statistical analyses were performed by Student's t-test (two-tail) using SPSS 15.0. Statistical significance was set as P = 0.05.
Adult OPCs were identified by their expression of NG2 and PDGFRα. Anti-PDGFRα labeled the cell body and primary processes of OPCs in an asymmetric fashion (Fig. S1 A), whereas NG2 was expressed on the entire OPC surface, including both primary and secondary processes (Fig. S1 B). Blood vessels also expressed NG2 and PDGFRa (for example, Fig. S1 D, arrowhead), but were readily distinguished from OPCs by their morphology and lack of expression of Sox10, an oligodendroglial lineage marker (Fig. S1 D, arrows). In piriform cortex (PC), all PDGFRα+ OPCs co-labeled with NG2, and 96.4 ± 2.3 % (± S.D) NG2+ OPCs were PDGFRα+ (Fig. S1 C, arrows, and E).
It has previously been reported that early postnatal and adult cortex contains a large number of low level DCX-expressing cells which retain properties of multipotential precursors in vitro (Walker T et al., 2007). We observed two populations of DCX-expressing cells in the piriform cortex, high level DCX-expressing (DCX++) (Fig. 1 A1, arrowheads), largely restricted to layer II, and low level expressing (DCX+) cells (Fig. 1 A1, arrows), distributed evenly throughout the piriform cortex. Quantification of DCX expression level revealed an atypical two-peak distribution pattern, with peak expression being 51 ± 12 and 122 ± 18, respectively (p = 7.2E-16, n=150) (Fig. 1 D7). To determine the in vivo identities of these DCX-expressing cells, we double immunostained DCX-expressing cells for PDGFRα, NG2, or HuCD. Virtually none of the DCX++ cells were PDGFRα+ or NG2+, and all DCX++ cells were labeled with HuCD, identifying them as neuronal lineage cells (Fig. 1 D, D1-D3). In contrast, 95.7% of the DCX+ cells were positive for PDGFRα (Fig. 1 A2, arrows, B1-B3, higher magnification) and NG2 (not shown), and most were negative for HuCD (Fig. 1 D4-D5). Interestingly, occasional DCX+ cells were labeled by HuCD; however their morphology was that of OPCs, with primary processes extended from cell bodies and fine distal branched secondary processes (Fig. 1 E1-E3). We further confirmed that the DCX+ cells were members of the oligodendroglial lineage by demonstrating that almost all DCX+/PDGFRα+ cells expressed Sox10, an oligodendroglial lineage-specific maker (Fig. 1 G1-G3). The DCX+ cells could be sub-divided into two populations, perineuronal (Fig. 1 F1-F3, arrows) and non-perineuronal (Fig. 1 D4-D5), according to their relative position to HuCD+ neurons. The cell bodies and some of the processes of perineuronal DCX+ cells were immediately adjacent to neuronal cell bodies. The low level expression of DCX on NG2+/PDGFRa+ OPCs was confirmed by in vitro immunohistology of cultured primary mouse OPCs (Fig. S2 F-H). BrdU cumulative labeling conducted on 8 week postnatal mice showed that only a minority of the NG2+ OPCs (~25%) in piriform cortex were mitotically active, with an average cell cycle duration of 8 days (Fig. S1 F-H), and that 23% (± 3.2%) of the DCX+/PDGFRα+ cells incorporated BrdU (Fig. S2 I-K). None of the DCX++ cells in piriform cortex were labeled with BrdU (Fig. S2 L-N).
The low level DCX expression in OPCs suggested the hypothesis that there is a lineage connection between putative OPCs and neurons. We used a genetic fate mapping procedure to test this hypothesis. Oligodendroglial lineage cells including OPCs express the plp promoter (Mallon et al., 2002; Guo et al, 2009), and we took advantage of this to label OPCs in the adult piriform cortex with the reporter gene EYFP by Cre-LoxP conditional recombination system using PLPcreERT2 /Rosa-STOP-EYFP (PCE/R) double transgenic mice. Early adult mice (P45-P60) were i.p. injected with tamoxifen (TM) twice daily for 5 consecutive days. Consistent with the absence of Cre immunoreactivity in NeuN+ or HuCD+ neurons (Fig. S3 G and H), we confirmed that in adult piriform cortex 10 days after the last tamoxifen injection, 96.7% of EYFP+ cells were Sox10+, and thus members of the oligodendroglial lineage (Fig. 2 A, arrows), and conversely, that about 60% of Sox10+ cells had undergone recombination. Amongst these EYFP+ cells, 16.7% (± 3.2%) were NG2+ OPCs (Fig. 2 B arrows) or PDGFRα+ OPCs (Fig. 2 C, arrows), and 81.4% were PDGFRα-/Sox10+ mature oligodendrocytes. We also found that a very small proportion of EYFP+ cells (~ 2.3%) were subpial GFAP+ cells.
To further assess the effect of tamoxifen on the recombination rate amongst OPCs, we analyzed the tissues on 1, 5, 10, 17 days following the last tamoxifen injection (days post-TM) and quantified the percentages of NG2+ or PDGFRα+ OPCs that were labeled with EYFP. On 1 day post-TM, 12.4% (± 2.3%) of total OPCs were EYFP positive (Fig. 2 D), which was consistent with our previous study (see discussion in Guo et al., 2009). This number increased to 20.8% (± 2.9%) on 5 days post-TM (Fig. 2 D). The recombination rate did not change significantly at later time-points, with the percentage being 21.6% and 20.9% on 10 and 17 days post-TM, respectively (Fig. 2 D), arguing that the TM effect on reporter gene recombination in NG2+ and PDGFRα+ OPCs had been driven to completion within the first 5 days post-TM. Consistent with this observation, Cre immunoreactivity in Sox10+ cells was confined to the nucleus on 1 day post-TM (Fig. S3 I), but, in contrast, restricted to cytoplasm from 5 days post-TM onwards (Fig. S3 J and K). Next, we asked whether the recombination rate decreased with age. P90 and P180 old mice were treated with TM using the same protocol, i.e. 5 consecutive days, twice a day. In contrast to the results in young adults, only 4.9% and 3.6% of NG2+/PDGFRα+ OPCs became EYFP-labeled at P90 and P180, respectively (Fig. 2 E), significantly lower recombination rates than that at P45 (P = 0.0006). However, the recombination rate amongst mature oligodendrocytes and the distribution of EYFP+ mature oligodendrocytes did not change significantly between P45, P90 and P180 using the same tamoxifen paradigm.
Using Cre immunoreactivity as an indicator of Plp transgene expression, we showed that no NeuN+ (Fig. S3 G) or HuCD+ (Fig. S3 H) cells expressed Cre in piriform cortex. To further evaluate whether recombination-induced EYFP expression in piriform cortex was oligodendroglial lineage-specific, we assessed EYFP expression in cells of other lineages during the first 17 days post-TM. During this time interval, there was no neuronal EYFP expression, indicated by NeuN (Fig. 2 F) and HuCD (Fig. 2 G) immunostaining (0 EYFP+/HuCD+ cells out of 1253 EYFP+ cells counted on 5 days post-TM, 0 out of 430 on 10 days post-TM, 2 out of 312 on 17 days post-TM). Occasional cells appeared to be EYFP and HuCD double positive, but we were able to exclude co-localization of these markers by confocal z-stacking, which demonstrated that the EYFP was actually in perineuronal cells (for example Fig. 1 F1-F3). Similarly, as shown in Fig 2H, we did not detect EYFP expression in Iba-1+ microglia at any time post-TM.
Before tracing the progenies of EYFP+ OPCs, we assessed the proportion of EYFP+ OPCs that were mitotically active. NG2+/PDGFRα+, EYFP+/PDGFRα+ and EYFP-/PDGFRα+ cells were designated as overall OPCs, Plp-Cre positive OPCs and Plp-Cre negative OPCs, respectively based on the observation that, shortly after TM treatment, EYFP was expressed only in Cre positive cells (compare Fig. 2 with Fig. S3). After 9 days of BrdU labeling administered in drinking water to 45-55 day postnatal mice, by which time the BrdU labeling index had reached a plateau level, 26.8% (± 12.4%) of EYFP negative OPCs incorporated BrdU, a percentage similar to that in overall OPCs (Fig. 3C), whereas the labeling index of EYFP positive OPCs was 15.3% (± 10.2%). This difference did not reach statistical significance (P = 0.14); hence, EYFP expression did not discriminate adult non-proliferating from proliferating NG2+/PDGFRa+ OPCs, a result consistent with previous data showing that Plp promoter activity does not distinguish proliferating from non-proliferating NG2+ OPCs (Mallon et al., 2002).
Neural stem cells (NSCs) in the subgranular zone (SGZ) of the dentate gyrus, an established adult neurogenic area, express Sox2, an HMG box transcription factor characteristic of NSCs (Suh et al., 2007). Previous studies demonstrated that Sox2 is also expressed in a subpopulation of NG2+/PDGFRα+ progenitor cells not only in the SGZ (Hattiangady et al., 2008) but also, in Sox2-EGFP mice, in the subventricular zone (SVZ) and cerebral cortex (Brazel et al., 2005). To confirm and extend these evidences of Sox2 expression in cortical EYFP+/PDGFRα+ OPCs, we performed triple immuno-fluorescence assays for Sox2, EYFP and PDGFRα. and identified Sox2+ cells that expressed EYFP and PDGFRα (Fig. 3A, arrows). In piriform cortex, 82.3% of EYFP+/PDGFRα+ cells, and 63.6% of total PDGFRα+ cells, expressed Sox2 (Fig. 3D).
In the developing brain, the transcription factor Pax6 is expressed in progenitors that are committed to give rise to glutamatergic neurons (Kallur et al., 2008). Intriguingly, we found that both PDGFRα+ cells (Fig. 3 B1-B4, arrowheads) and EYFP+/PDGFRα+ cells (Fig. 3 B1-B4, arrows) in piriform cortex displayed nuclear Pax6 immuno-reactivity. The EYFP+/PDGFRα+ cells also expressed low level DCX (Fig 3 E, cf Fig. 1 B1-B3). The expression of DCX, Sox2 and Pax6 in OPCs was also consistent with prior gene microarray analyses (Dugas et al., 2006; Cahoy et al., 2008). Thus, EYFP+/PDGFRα+ progenitors in the piriform cortex display some features of neuronal progenitors. In contrast to NG2+ or PDGFRα+ OPCs, no PDGFRα-/Sox10+/EYFP+ mature oligodendrocytes expressed DCX, Sox2 or Pax6.
Interestingly, at time-points later than day 17 post-TM treatment, increasing numbers of EYFP+ neurons became demonstrable in piriform cortex. The piriform cortex is a three-layer structure, layer II of which consists of packed principal neurons that receive sensory input from olfactory bulb (OB) via the lateral olfactory tract (LOT) (Haberly, 2001). The EYFP+ neurons were located principally in layer II (Fig. 4 A, B), and to a much less extent in layer III (Fig. 4 B, wavy arrow), with no EYFP+ neurons in layer I. We employed confocal z-sectioning to verify this co-localization of EYFP and neuronal markers. As shown in Fig. 4 A, the EYFP+ cell bodies were co-labeled with HuCD at all levels of optical sections (Fig. 4 A, boxed area, A1-A3, higher magnification). The presence of fate-mapped piriform neurons was validated and confirmed by EYFP and NeuN double staining (Fig. 4B). In some instances, only one of two closely apposed EYFP+ cell bodies was identified as a NeuN+ neuron (Fig. 4 B1-B2), the other cell being NeuN negative (Fig. 4 B3). To further confirm the presence of fate-mapped EYFP+ neurons, we double stained with SMI312, a pan-neurofilament antibody mixture that strongly labels axons (Soulika et. al., 2009), and EYFP, which is expressed not only in soma, but is also transported from the neuronal perikaryon to distal processes (Guo et al., 2009). As expected, some SMI312+ axons in piriform cortex were co-labeled with EYFP (Fig. 4 C3, box and higher magnification). Some SMI312 negative, but EYFP+ apical dendrites (Fig. 4 C, box and higher magnification), which expressed EYFP+ spines (arrows in Fig 4C1, Fig 8A1-A2), penetrated through layer II and layer Ib into layer Ia (Fig. 4 C2), where they presumably formed synapses with axons extended from OB mitral cells (ul Quraish et al., 2004). The above data clearly demonstrated the presence of EYFP+ neurons in the piriform cortex.
Were these piriform cortical EYFP+ neurons ectopically expressed the PLP promoter (or alternatively Plp-Cre transgene) rather than having been derived from EYFP+ OPCs? If these neurons were generated from EYFP+ OPCs, (a) EYFP+ neurons should be absent from piriform cortex during the first few days or weeks after TM treatment; (b) the number of EYFP+ neurons in piriform cortex should gradually increase with time after TM treatment and (c) EYFP+ neurons were negative for Cre immunoreactivity. All of these predictions proved to be correct. As shown in Fig. 4 D, there were no EYFP+ neurons in piriform cortex before 17 days post-TM. From 17 days post-TM, EYFP+ neurons began to appear in layer II of piriform cortex, and increased 20-fold during the interval between day 17 and day 180 post-TM (Fig. 4 D). Our quantification data showed that 0.4% (± 1.3%), 3.35% (± 1.0%) and 5.8% (± 1.6%) of total EYFP+ cells expressed the neuronal marker NeuN on 17, 90 and 180 days post-TM, respectively. During the same time interval, the percentage of EYFP+/NG2+ cells amongst total NG2+ cells dropped gradually, from 20.9% on day 17 post-TM to 12.1% on 180 days post-TM (Fig. 4 E). We also quantified the frequency (cells/mm2) of piriform cortical NG2+ OPCs using immunohistochemistry and confocal imaging performed on 14 μm sections, and found that this decreased during the six months post tamoxifen treatment (Fig. 4 F). This reciprocal relationship between accumulation of EYFP+ neurons and decrease of EYFP+ OPCs strongly suggests that neurons are continuously being generated from OPCs in young adult mouse piriform cortex, and that the OPCs involved in this process in the normal adult piriform cortex do not self-renew. And again, we did not see any Cre immunoreactivity positive NeuN+ or HuCD+ neurons (Fig. S3 G and H) in piriform cortex at any of the time points we assessed.
Plp promoter activity has been reported in occasional hindbrain neurons (Miller et al., 2009). We evaluated Plp promoter activity distribution in adult forebrain using EYFP reporter and Cre expression in PCER mice, based on the speculation that if a cell expressed Plp promoter activity, it would be Cre immunoreactivity positive and become labeled with EYFP within the first several days after tamoxifen administration into these transgenic mice. At either one or five days post-TM, we detected EYFP+ or Cre+ neurons in the caudate putamen nucleus (CPu) (Fig. S3, A-F), amygdalostriatal transition area (AStr) (Fig. S3, A-D), hypothalamic para- and peri-ventricular nucleus, and posteromedial cortical amygdaloid nucleus (PMCo) (not shown), indicating that Plp promoter activity existed in neurons within these defined areas / nuclei. As expected, the numbers of these immediately post-TM EYFP-labeled neurons did not change over time (compare Fig. S3 A with B), and were independent of age of the mouse (compare Fig. S3 C with D). However, in piriform cortex, in which we observed the generation of EYFP+ neurons, neither Cre positive neurons nor EYFP+ neurons were detected shortly after TM treatment (Fig. S3 G and H) and EYFP+/Cre- neurons progressively increased over time (Fig. 4 D). These PLP promoter transgenic mouse results were consistent with Plp mRNA in situ hybridization data we obtained in adult wild type mice (unpublished data). Taken together, our data indicate that, unlike EYFP+/Cre+ neurons in CPu, EYFP+ neurons in piriform cortex were generated from EYFP+ OPCs.
Prior studies suggested that adult rodent and primate SVZ GFAP+ NSCs generate neurons that migrate to piriform cortex (Shapiro et al., 2007; Bernier et al., 2002). Were the piriform cortical EYFP+ neurons that we visualized also generated from SVZ NSCs? To address this issue, we analyzed EYFP expression in SVZ GFAP+ cells. Ten days after 5 days tamoxifen treatment, when recombination in OPCs had already reached a plateau level (Fig. 2 D), there were no GFAP+ cells labeled with EYFP in the SVZ (at least 4 sections through dorsal and lateral SVZ from each of 3 mice) (Fig. 5 A and higher magnification). Consistent with this observation, no DCX+ SVZ migrating neuroblasts expressed EYFP (Fig. 5 B and higher magnification). As an additional means to exclude GFAP+ SVZ cells as the source for EYFP+ piriform neurons, we bred GFAP-Cre-ERT2/Rosa26-STOP-EYFP (GCE/R) double transgenic mice (Ganat et al., 2006) and P60 GCE/R mice were treated using the same TM paradigm as that for PCE/R mice. No EYFP+ neurons were generated in the piriform cortex of these TM-treated GCE/R mice (Fig. S4, E-E3), although SVZ GFAP+ cells were labeled with EYFP upon TM treatment (Fig. S4, D-D3). Collectively, our data excluded the possibility that the piriform cortical EYFP+ neurons originated from EYFP+ SVZ NSCs, and indicated, instead, that, in young adult mice, EYFP+ neurons are generated in situ from EYFP+ OPCs.
To provide evidence that EYFP+ OPCs first generate immature EYFP+ neurons in piriform cortex, which then become mature EYFP+ neurons, we immunostained for EYFP together with DCX, HuCD, or NeuN. These studies demonstrated three distinct maturation states of EYFP+ cells in piriform cortex (Fig. 6 A and B). EYFP+ progenitor cells expressed low level DCX (Fig. 6 A-A2) and were PDGFRα+ (Fig. 1 B1-B3), but were negative for HuCD (Fig. 6 A3). Some newly generated EYFP+ neurons had few or no fully developed neurites, co-labeled with HuCD (Fig. 6 B-B3, arrows), and expressed high level DCX (Fig. 6 B-B1, arrows), whereas mature neurons had down-regulated DCX expression to undetectable levels and were positive for HuCD (Fig. 6 B-B3, arrowheads and inserts) and NeuN (not shown). To confirm the identify of immature EYFP+ neurons in piriform cortex, we demonstrated that the perikarya (Fig 6. C-C3, arrowheads) and neurites (Fig. 6 C-C3, arrows) of these cells were immunolabeled with another immature neuronal marker, βIII-Tubulin (Tuj1), and that their nuclei were immunostained with Tbr1 (Fig. 6 D-D3, arrows), a marker for postmitotic glutamatergic projection neurons (Englund et al., 2005). To gain additional insight into the time-course of maturation of EYFP+ neurons after TM treatment, we measured the distance between the origin of the apical dendritic trunk and the termination of the apical dendirtic tree, and found that this distance progressively increased between 30 days post-TM (Fig. S5 A, D) 90 days (Fig. S5 B, D), and 150 days (Fig. S5 C, D) post-TM, respectively.
To determine whether mitotically cycling or post-mitotic OPCs were giving rise to EYFP+ neurons in piriform cortex, PCE/R mice were treated with BrdU in drinking water for 15 days, beginning simultaneously with initiation of i. p. TM administration. We validated this BrdU labeling method by showing the presence of HuCD+/BrdU+ cells in neurogenic area SGZ (Fig. S6 A, arrow). Although occasional EYFP+/NG2+ OPCs (Fig. S6 B) and EYFP+/APC+ mature oligodendrocytes (Fig. S6 C) were labeled with BrdU, we found that only two of a total of 320 EYFP+/HuCD+ neurons we assessed were labeled with BrdU (Fig. S6 D), indicating that, with rare exceptions, EYFP+ neurons in the adult piriform cortex were derived from non-proliferating EYFP+ OPCs.
EYFP was expressed not only in the cell bodies of the labeled piriform neurons, but also filled their proximal and distal processes, thus facilitating their morphological characterization. There are two major classes of excitatory principal neurons in piriform layer II and III, pyramidal neurons and semilunar neurons (Suzuki and Bekkers, 2006). Essentially all EYFP+ neurons had the characteristic morphological features of pyramidal neurons. The NeuN+/EYFP+ neuron in Fig. 7 A displayed the typical morphology of a layer II pyramidal neuron: (1) a large cell body (usually >10 μm in diameter); (2) a single apical dendritic trunk with a diameter > 1 μm (Fig. 7 A, arrow) whose branches extend to superficial layer Ia; and (3) multiple basal dendrites extending to layer II and layer III (Fig. 7 A. arrowheads). There was some variability in lengths of apical dendritic trunks of EYFP+ pyramidal neurons (Fig. 7 B, C). Few if any EYFP+ neurons had the morphology of semilunar neurons, i.e. with fork-like apical dendrites and no basal dendrites.
To support our identification of EYFP+ adult OPC-derived piriform neurons as pyramidal neurons, we used immunohistology to examine their neurotransmitter specificities. Using confocal single optical slices, we localized punctate deposits of vesicular glutamate transporter 1 (vGLUT1) immunoreactivity to EYFP+ neurites (Fig. 7 E, arrows). This is consistent with our previous observation that neonatal EYFP+ OPCs generated glutamatergic neurons that expressed vGLUT1 mRNA (Guo et al., 2009). We confirmed the glutamatergic phenotype of the EYFP+ neurons in adult piriform cortex by visualizing immunoreactive glutaminase (an enzyme responsible for converting glutamine to glutamate in glutamatergic neurons) in EYFP+/NeuN+ neurons (Fig. S7 A). Also consistent with their glutamatergic nature, no EYFP+ piriform neurons expressed immunoreactive glutamic acid decarboxylase 67 kD, isoform (GAD67) (Fig. 7 F, arrow), choline acetyltransferase (ChAT) (Fig. 7 G, arrow), or tyrosine hydroxylase (Fig. S5 E), a dopaminergic neuron marker. Furthermore, none expressed interneurons markers, i.e. calretinin (Cal) (Fig. S5 B), somatostatin (SOM) (Fig. S7 C), or parvalbumin (Parv) (Fig. S7 D). Taken together, our data demonstrated that EYFP+ piriform neurons derived from adult OPCs were glutamatergic pyramidal neurons.
The extension of apical dendrites of the EYFP+ pyramidal neurons into superficial layer I (Fig. 7 A-C), their expression of dendritic spines (Fig. 4 C2; Fig. 8 A2, arrows) and their long survival time (up to 300 days post-TM) suggested that these OPC-derived neurons had become integrated into functional circuits. In support of this idea, we found that the EYFP+ spines were often in close proximity to punctate spots of synapsin (Syn), a pre-synaptic protein (Fig. 8 A1, arrows), arguing that these EYFP+ neurons became synaptically integrated into the neuronal network (Jessberger et al., 2008). Since N-methyl D-aspartate receptors (NMDARs) play a role in controlling excitatory synaptic transmission, we explored whether these OPCs-derived neurons expressed NR1, an essential NMDAR subunit. Like their EYFP negative counterparts (Fig. 8 B2-B4, arrows), most piriform cortical EYFP+/NeuN+ neurons were NR1+ (Fig. 8 B1-B4, arrowhead). To lend further support to the hypothesis that these adult OPC-derived glutamatergic pyramidal neurons were functional, we immunostained for c-fos, an immediate early gene known to be expressed by newly generated neurons that have been integrated into neuronal circuits (Magavi et al., 2005, Brill et al., 2009; Ohira et al., 2010). The majority of OPC-derived EYFP+/NeuN+ neurons were co-labeled with c-fos (Fig. 8 C1, arrow and higher magnification).
The present study demonstrates that, in young adult mice, NG2+/PDGFRa+ Plp promoter-expressing progenitors (PPEPs) give rise to neurons in piriform cortex, and provides novel information on both the expression of neurogenic markers by these cortical NG2+/PDGFRa+ PPEPs, and on the phenotypes of the immature and mature neurons derived from this source. Though occasional neurons in other regions of the forebrain (eg, CPu) expressed Plp promoter activity revealed by Cre immunoreactivity (Fig. S3 F), and hence became EYFP+ early after initiation of tamoxifen injection to PCE/R double transgenic mice (Fig. S3 A-F), the lack of piriform cortex neuronal Plp promoter expression (Fig. S3 G, H), and the kinetics of accumulation of EYFP+ neurons in the piriform cortex (Fig. 4 D) were incompatible with the speculation that these EYFP+ neurons were derived from neurons that ectopically express EYFP, since there were virtually no EYFP+ neurons in piriform cortex prior to 17 days post-TM (Fig. 2 F, G), at which time-point the recombination rate of OPCs had already reached a plateau level (Fig. 2 D). While we acknowledge that transgene activity does not necessarily equate to endogenous transcriptional activity, the observations that: (a) all Cre+ cells were Sox10+ in piriform cortex (Fig. S3 I-K); (b) no Cre+/HuCD+ or Cre+/NeuN+ cells were detected in piriform cortex at any of the time points we assessed (Fig. S3 G-H); and (c) no EYFP+/NeuN+ (Fig. 2F) or EYFP+/HuCD+ (Fig. 2G) were detected shortly after TM treatment, all argue for the fidelity with which the Plp-Cre transgene marked oligodendroglial lineage cells, and not neurons, in adult piriform cortex.
It is very unlikely that SVZ GFAP+ NSCs contributed to this adult cortical neurogenesis, since we did not detect any SVZ GFAP+ cells or DCX+ neuroblasts labeled with EYFP 10 days post-TM treatment of PCE/R mice (Fig. 5), nor did we see EYFP+ piriform cortical neurons in GCE/R mice (Fig. S4 D-E3). Furthermore, the EYFP+ neurons in adult piriform cortex in our study were glutamatergic pyramidal neurons rather than GABAergic interneurons, which are the most abundant neuronal type known to be derived from SVZ GFAP+ NSCs (Garcia et al., 2004, Inta et al., 2008). It has been shown that microglia can specifically fuse to apical dendrites of pyramidal neurons in retrovirus-mediated fate-mapping (Ackman, et al., 2006). Bearing this in mind, we carefully assessed the EYFP expression in microglia and found no evidence of EYFP+/Iba1+ (Fig 2 H) or EYFP+/CD11b+ (data not shown) microglia throughout all the time points.
EYFP+ neurons were much less common 60 days post-TM (4 EYFP+/NeuN+ cells out of 1024 EYFP+ cells we counted from 3 mice) in piriform cortex of PCE/R mice in which tamoxifen was first administered to 100 days old PCE/R mice (Fig. S4 A-C). This result was consistent with the much lower number of EYFP+ OPCs in piriform cortex in these age mice (Fig. 2 E), again supporting the conclusion that EYFP+ OPCs were the source of piriform cortical EYFP+ neurons.
The existence of adult neurogenesis in cerebral cortex from OPCs has been controversial. Zhu et al. (2008) reported that no neurons were fate-mapped from NG2+ progenitors in constitutive NG2-Cre mice. Their inability to detect fate-mapped cortical neurons in adult constitutive NG2-Cre mice, despite highly efficient recombination in the oligodendroglial lineage, may have been masked by a background of EYFP+ neurons resulting from ectopic Cre expression in cortex neurons. (see Jackson Laboratory Website, mouse stock number 008533). Fate-mapping in adult Olig2CreERT2 mice also failed to demonstrate labeling of cortical neurons (Dimou et al., 2008), perhaps because they used older mice and a less robust tamoxifen treatment regimen than we employed in the present study, or alternatively, because the Rosa26-EYFP reporter transgene is a better recombination indicator than Rosa26-EGFP or Z/EG transgenes (Young et al., 2010). In accord with our own results, fate-mapped piriform projection neurons were detected after intense tamoxifen treatment of young adult PDGFRαCreERT2 mice (Rivers et al., 2008).
High level DCX is expressed by migrating and immature neurons in SVZ-RMS-OB and SGZ of DG, two well established adult neurogenic areas (Gage 2000). DCX is also expressed in other forebrain areas, for example cerebral cortex including neocortex and piriform cortex (Luzzati et al., 2008, Gomez-Climent et al., 2008, Tamura Y et al., 2007), corpus callosum and striatum (Yang H et al., 2004), and hypothalamus (Fig. S2 E). Although some of cortical DCX expressing cells were identified as belonging to neuronal lineages, the identity of a large number of remnant cortical DCX+ cells is unknown (Yang et al., 2004, Walker et al., 2007). Here we present evidence that there are two populations of DCX+ cells in piriform cortex. As previously reported (Gomez-Climent et al., 2008), we confirmed that the high level DCX+ cells are neuronal lineage (Fig. 1 D-D3), whereas the low level DCX-expressing sub-population were positive for PDGFRα (Fig. 1 A1-B3) and Sox10 (Fig. 1 G1-G4), designating them as putative OPCs instead of neuronal progenitors (Aguirre and Gallo, 2004).
Few EYFP+ neurons incorporated BrdU (Fig. S 4). This result is similar to the observation by Rivers et al (2008) in PDGFRα-Cre-ERT2 mice. There are two possible explanations for the almost complete lack of EYFP+/BrdU+ neurons in adult piriform cortex despite prolonged administration of BrdU in drinking water. First, the EYFP+ neurons may have originated from predominantly post-mitotic EYFP+ OPCs, which comprise ~85% of total EYFP+ OPCs quantified in our study (Fig. 3 C). This explanation would be consistent with our previous report (Guo et al., 2009) of substantial numbers of EYFP+/BrdU+ neurons when fate mapping neonatal PCE/R mice, in which a much greater proportion of EYFP+ OPCs are mitotic than in adults. Alternatively, EYFP+ neurons may, in fact, be derived from mitotically cycling OPC, but these cycling adult OPCs, though possible to label by intracerebroventricular administration of BrdU, are not efficiently labeled by systemically administered BrdU (Kokoeva et al., 2005, 2007).
The piriform cortex we analyzed spans about 1.3 mm from Bregma −2.30 mm to −1.06 mm. Approximately 0.98% of total layer II and III NeuN+ neurons expressed EYFP at 182 days post-TM. By day 182 post-TM, there were approximately 17 EYFP+ neurons per 14 μm section (Fig. 4 D). About 0.1 EYFP+ neuron was added per 14 μm section of piriform cortex per day during the 180 days after tamoxifen treatment was initiated (17/182 = 0.093). Therefore, approximately 10 [0.1 × (1300/14) = 9.3] FP+ neurons were added to adult piriform cortex daily, which would correspond to about 46 total neurons added [17 ÷ 182 × (1300/14) ÷ 0.2 ≈ 46], if the ~20% recombination rate in total NG2+/PDGFRa+ cells is taken into account. The discrepancy between our calculated result and previous data (24 neurons) (Rivers et al., 2008) may have been due to the location along the anterior-posterior axis that was evaluated; our data were gathered from posterior piriform cortex, whereas they focused their study on anterior piriform cortex. Alternately, the rate of neuronogenesis from Pip promoter+/PDGFRa+ OPCs is greater than that in the overall pool of PDGFRa+ OPCs.
Our data strongly suggest that OPC-derived neurons became integrated into the neuronal network in adult piriform cortex. What is their functional role? The piriform cortex is not only the first and largest destination for input from the olfactory bulb, but also acts as an association cortex that integrates sensory information from the environment with behavioral, contextual and cognitive input (Haberly 2001). Given the importance of layer II pyramidal neurons, the role of piriform cortex and the continuous replacement of olfactory bulb interneurons, we hypothesize that the continued addition of OPC-derived neurons to piriform cortex during adulthood is important in preserving the capacity for olfactory discrimination learning and memory, and is driven by continued olfactory bulb input. Consistent with this hypothesis, pyramidal neurons in piriform cortex selectively undergo apoptosis after total bulbectomy (Capurso et al., 1997). Taken together, our findings prove that PLP promoter-expressing OPCs continue to generate glutamatergic pyramidal neurons in the adult piriform cortex.
Supported by funds from the Shriners Hospitals for Children (to FG and DP), the California Institute for Regenerative Medicine (to FG, JM, and DP), and by NIH RO1NS025044 (to DP).