Striatal astrocytes produce neuroblasts in an excitotoxic model of Huntington's disease

In adult neurogenic niches, astrocytic cells produce neurons throughout life (Fuentealba et al., 2012). Parenchymal astrocytes can become neurogenic in vitro when isolated from the lesioned neocortex (Buffo et al., 2008; Sirko et al., 2013) or in vivo after overexpression of specific transcription factors (Niu et al., 2013). However, whether parenchymal astrocytes can spontaneously generate neurons in vivo is unclear (Dimou and Götz, 2014). The adult brain parenchyma has been generally considered gliogenic and not permissive for the activity of neuronal progenitors (Lim et al., 2000; Ninkovic and Götz, 2013; Shihabuddin et al., 2000). Nonetheless, examples of parenchymal neurogenesis are emerging (Bi et al., 2011; Luzzati et al., 2006,, 2011b; Ohira et al., 2010). For instance, we previously demonstrated that clusters of proliferating cells with features of transient amplifying progenitors (TAPs) produce neurons in the striatum of rabbits (Luzzati et al., 2006) and in a mouse model of striatal degeneration (Luzzati et al., 2011b).

Here, we analysed striatal neurogenesis in the quinolinic acid (QA) lesion mouse model of Huntington’s disease (Fan and Raymond, 2007). We show that such a lesion activates striatal astrocytes to produce neurons.

At 5 weeks post-QA lesion (w.p.l.), numerous DCX⁺ neuroblasts were present in the striatum and organised into clusters or as individual cells (Fig. 1A-E). As in other models of striatal neurogenesis (Liu et al., 2009; Luzzati et al., 2011b), these neuroblasts expressed SP8, a transcription factor typical of lateral/caudal ganglionic eminence-derived interneurons (Ma et al., 2012; Waclaw et al., 2006), and some of them expressed NeuN (RBFOX3 – Mouse Genome Informatics; data not shown) and attained complex morphologies (supplementary material Fig. S1). The clusters of DCX⁺ cells were closely associated to clusters of cells expressing the proliferation marker Ki67 (MKI67 – Mouse Genome Informatics), with numerous cells colabelled for DCX (Fig. 1B,C,E).

Based on clustering and the differential expression of Ki67 and DCX we could define four striatal cell types that were induced by QA: clustered Ki67⁺/DCX⁻ cells (cK), clustered Ki67⁺/DCX⁺ cells (cKD), clustered DCX⁺/Ki67⁻ cells (cD) and individual DCX⁺/Ki67⁻ cells (iD) (Fig. 1C-F). The cK, cKD and cD cells appeared between 2 and 3 w.p.l. (2 versus 3 w.p.l., Tukey's post-hoc test: cK, P=0.010; cKD, P=0.005; cD, P=0.021; Fig. 1F) and, although their number remained constant after 3 weeks (ANOVA: cK, F_2,7=2.464, P=0.155; cKD, F_2,7=0.383, P=0.695; cD, F_2,7=0.419, P=0.673), at all time points a high proportion incorporated BrdU when injected 4 days before sacrifice (supplementary material Fig. S2A,D). This indicates that cK, cKD and cD cells have a high turnover rate. By contrast, iD cells showed a delayed increase that peaked at 4 w.p.l. (Tukey's post-hoc test: 2 versus 3 weeks, P=0.226; 3 versus 4 weeks, P=0.001; Fig. 1F) accompanied by a reduction in the fraction of BrdU⁺ cells over time (Tukey's post-hoc test: 3 versus 4 weeks, P=0.031; supplementary material Fig. S2C,D). Thus, this latter population appears later and has a lower turnover rate.

Interestingly, most cK cells expressed the TAP markers ASCL1 (Parras et al., 2004) and SOX9 (Cheng et al., 2009) (Fig. 1G-I; data not shown). Collectively, these data suggest that, as proposed in other models of striatal neurogenesis (Luzzati et al., 2006,, 2011b), QA stimulates the appearance of TAP-like progenitors (cK cells) that give rise to neuroblasts that initially cluster (cKD, cD cells) and subsequently disperse as individual cells (iD cells). At 6 months after QA, striatal TAPs and neuroblasts were still present and could incorporate BrdU (supplementary material Fig. S3), suggesting that QA results in the long-term establishment of an intrastriatal neurogenic niche.

The induction of neurogenic potential in resident parenchymal cells was further supported by the appearance of self-renewing multipotent neurospherogenic cells in the striatum at 5 w.p.l. (supplementary material Fig. S4; data not shown). Interestingly, clusters of cK, cKD and cD cells were generally closely associated to GFAP⁺ astrocytes, which were occasionally proliferating, as assessed through both Ki67 and BrdU (supplementary material Fig. S5). Using hGFAP-GFP mice (Platel et al., 2009; Zhuo et al., 1997) we could establish that GFP⁺/Ki67⁺ proliferating astrocytes represented 8±3% of all cK cells at 5 w.p.l. (Fig. 1I-L). These observations support the contention that cK cells and their progeny originate from striatal astrocytes.

To explore this possibility further, we first performed cell fate-mapping analysis using transgenic mice expressing the tamoxifen-inducible recombinase CreER^T2 under the control of diverse cell type-specific promoters, namely Glast (Slc1a3 – Mouse Genome Informatics), nestin (Nes) and Ng2 (Cspg4 – Mouse Genome Informatics). GLAST is a pan-astrocytic marker (Dimou and Götz, 2014), whereas nestin more specifically associates with active neurogenic astrocytes (Codega et al., 2014) and some oligodendrocyte progenitors (Boda et al., 2015), while NG2 is specifically expressed by oligodendrocyte progenitors (Zhu et al., 2011). Accordingly, in intact animals 1 week after tamoxifen, recombined YFP⁺ cells represented 44±3% of all S100b⁺ striatal astrocytes in GLAST-CreER^T2 mice and 10±6% in Nestin-CreER^T2 animals. In the NG2-CreER^T2 line, astrocytes were not targeted (supplementary material Fig. S6).

To determine whether cells expressing these genes are the source of intrastriatal TAPs and neuroblasts after lesion, tamoxifen was administered 1 week before QA (bQA). In addition, for each genotype a second group of animals was treated with tamoxifen at 4 w.p.l. (aQA) to identify possible injury-related changes in the phenotype of the neurogenic progenitors after their activation. In all cases animals were analysed at 5 w.p.l. (Fig. 2A). In NG2-CreER^T2 animals, we never observed DCX⁺ neuroblasts expressing the reporter YFP, either in the striatum or in the subventricular zone-olfactory bulb (SVZ-OB) system (data not shown), indicating that NG2⁺ cells are not neurogenic in our model. By contrast, in the SVZ of GLAST-CreER^T2 and Nestin-CreER^T2 animals, YFP⁺ cells included putative TAPs (Ki67⁺/DCX⁻), proliferating (DCX⁺/Ki67⁺) and postmitotic (DCX⁺/Ki67⁻) neuroblasts (Fig. 2A-C). The percentage of YFP⁺ cells did not differ between strains (supplementary material Table S1), indicating similar efficiency of nestin- or GLAST-driven recombination in SVZ neurogenic astrocytes.

In the striatum of GLAST-CreER^T2 bQA and aQA animals, YFP was expressed by cK, cKD, cD and iD cells, indicating that these cells originate from astrocytes that express GLAST both before and after the QA lesion (Fig. 2A,D,E; supplementary material Fig. S7A,B). YFP⁺ postmitotic neuroblasts (cD and iD) were less numerous in the striatum of GLAST-CreER^T2 aQA animals, but increased to levels comparable to those of YFP⁺ proliferative cells (cK and cKD) in the GLAST-CreER^T2 bQA animals (Fig. 2A; supplementary material Table S1A), further supporting that cK and cKD cells are early stages of the striatal neurogenic lineage. Three-dimensional reconstructions in GLAST-CreER^T2 bQA animals indicated that the morphology of the GFP⁺ iD cells was comparable to that of their GFP⁻ counterparts (supplementary material Fig. S8). Thus, GLAST⁺ astrocytes are the source of QA-induced intrastriatal TAPs and neuroblasts.

In Nestin-CreER^T2 bQA mice, YFP⁺ cells corresponding to the striatal cell types (cK, cKD, cD and iD) were very rare (Fig. 2A). However, in Nestin-CreER^T2 aQA animals, the levels of genetic labelling of cK and cKD cells were greatly increased and reached similar levels to those seen in GLAST-CreER^T2 aQA animals (Fig. 2A,F,G; supplementary material Table S1C). Interestingly, although nestin was mostly absent from striatal astrocytes under normal conditions, several YFP⁺ cells with astrocytic morphology appeared in Nestin-CreER^T2 aQA animals (supplementary material Fig. S7C,D). These findings suggest that the resident GLAST⁺ striatal astrocytes upregulate the expression of nestin after lesion and generate cK, cKD, cD and iD cells.

To directly confirm both the striatal origin of the neurogenic progenitors and their astrocytic identity, we performed intrastriatal injections of either a GFP-tagged lentiviral vector (VSVG-GFP; n=3) or an adenoviral vector carrying Cre recombinase under the control of the mouse Gfap promoter (Ad:GFAP-Cre; n=3; Merkle et al., 2007) 1 week before QA lesion. Whereas VSVG-GFP showed broad cellular tropism (data not shown), injection of Ad:GFAP-Cre into the striatum of R26R reporter mice resulted in the expression of YFP almost exclusively in astrocytes (supplementary material Fig. S9). Only animals with no YFP staining in the SVZ-OB system were analysed. In both cases, at 5 w.p.l. serial section 3D reconstructions of the whole striatum revealed multiple examples of YFP⁺ or GFP⁺ cK, cKD, cD and iD cells (Figs 3 and 4). The morphology of these latter cells was consistent with that of GFP^– iD cells (Fig. 4B-E). These data indicate that striatal astrocytes generate TAPs and neuroblasts after lesion.

Notably, in both our genetic and viral fate-mapping analyses about 85% of the striatal Ki67⁺ clusters exhibiting reporter expression were entirely composed of cells expressing YFP or GFP (Fig. 2D-G, Fig. 3C,D and Fig. 4F,G; see Materials and Methods for details), indicating that proliferative clusters originate mostly from the clonal expansion of a single striatal astrocytic progenitor.

SVZ progenitors have been shown to generate neuroblasts for the lesioned striatum (Liu et al., 2009). To examine whether these progenitors can further contribute to the intrastriatal TAPs, we injected Ad:GFAP-Cre or a TAT-Cre to respectively target the dorsolateral and the periventricular SVZ of R26R mice 1 week before QA. In the striatum of these animals at 5 w.p.l. we observed only a few iD cells expressing YFP (supplementary material Fig. S10), suggesting that striatal TAPs originate only from local astrocytes.

Taken together, these results indicate that some striatal astrocytes are quiescent neuronal progenitors that become activated after QA lesion. Like neurogenic astrocytes of other neurogenic niches, these cells upregulate nestin in their active state (Codega et al., 2014) and produce neurons through ASCL1⁺ and SOX9⁺ TAPs (Dimou and Götz, 2014; Fuentealba et al., 2012). However, striatal progenitors and/or their microenvironment may possess unique features that enable them to produce neurons in the brain parenchyma. Unravelling these features might help to unleash the full neurogenic potential of the adult brain, a fundamental prerequisite in order to design cell replacement therapies for brain repair. Interestingly, while this study was under revision the activation of neurogenic potential in striatal astrocytes was also demonstrated in a model of stroke (Magnusson et al., 2014) and under physiological conditions during guinea pig development (Luzzati et al., 2014). Thus, in contrast to SVZ and dentate gyrus neuronal progenitors, which are constitutively active, other populations of neurogenic astrocytes are activated only under specific conditions. The fate potential of these progenitors remains an important issue. Most neuroblasts generated in both the normal and lesioned striatum have a short life-span, but attain complex and specific morphologies (Luzzati et al., 2011a,b, 2014). These transient neurons might sustain a new type of adult brain plasticity that merits further exploration.

All animal experiments were approved by the Italian Ministry of Health and the Bioethical Committee of the University of Turin. Experiments were performed on 8- to 12-week animals. C57BL/6 lesioned mice received two intraperitoneal injections (6 h apart) of 5-bromo-2-deoxyuridine (BrdU, Sigma-Aldrich; 50 mg/kg in 0.1 M Tris pH 7.4) 4 days before sacrifice. Tamoxifen (TAM, Sigma-Aldrich T5648-1G) was dissolved in corn oil (Sigma-Aldrich C8267) and 2.5 mg was administered by forced feeding (oral gavage) twice with a 24 h interval.

Animals were anesthetised with a ketamine (100 mg/kg ketavet, Gellini) and xylazine (33 mg/kg rompun, Bayer) solution and perfused with a solution of 4% paraformaldehyde (PFA) and 2% picric acid (AnalytiCals, Carlo Erba 409302) in 0.1 M sodium phosphate buffer (PB) pH 7.4. Brains were then post-fixed for 3 h, cryoprotected in 30% sucrose (Fluka 84100) in 0.1 M PB pH 7.4, embedded at −80°C in Killik/OCT (Bio-Optica 05-9801), and cryostat sectioned in a series of 50 µm-thick sections.

VSVG-GFP vector stocks were produced by transient transfection of the transfer plasmid expressing eGFP under the control of the CMV promoter, the packaging plasmids pMDLg/pRRE and pRSV.REV, and the VSV envelope plasmid pMD2.VSV-G in HEK293T cells as described (Follenzi et al., 2000). Viral particles were purified and concentrated by ultracentrifugation as described (Dull et al., 1998). Vector titre on HeLa cells was 2×10⁹ TU/ml. The virus was then diluted 1/20 in PBS containing 0.6% glucose and frozen.

Generation of Ad:GFAP-Cre virus was described previously (Merkle et al., 2014). Briefly, HEK293 cells were infected to produce replication-defective adenovirus, which was purified using the Fast-Trap Adenovirus Purification and Concentration Kit (Millipore). The titre was 1×10¹⁰ infectious particles/ml. TAT-Cre recombinant protein was produced as previously described (Peitz et al., 2002).

Mice were anesthetised with 0.3 ml/kg ketamine and 0.2 ml/kg xylazine, positioned in a stereotaxic apparatus (Stoelting) and injected with a pneumatic pressure injection apparatus (Picospritzer II, General Valve Corporation). Injection coordinates: QA (Sigma-Aldrich P6,320-4; 1 µl diluted to 120 mM in 0.1 M PB), +0.1 mm AP, −2.1 mm ML and −2.6 mm DV; VSVG-GFP and Ad:GFAP-Cre, 0.8 mm AP, −2.1 mm ML and −3.2 mm DV (n=3 for each vector). Ad:GFAP-Cre virions driving Cre recombinase expression in GFAP⁺ cells were injected into R26YFP reporter mice. For both vectors, we analysed only animals in which the SVZ and OB were entirely free of reporter-positive cells. To target the SVZ, VSVG-GFP and TAT-Cre were injected respectively at +1.2 mm AP, −1 mm ML and −1.3 mm DV and at +3 mm AP, −0.8 mm ML and −2.9 mm DV.

Sections were incubated for 48 h at 4°C in 0.01 M PBS pH 7.4 containing 2% Triton X-100, 1:100 normal donkey serum and primary antibodies (supplementary material Table S2). For BrdU staining, sections were pre-incubated in 2 M HCl for 30 min at 37°C and then rinsed in 0.1 M borate buffer pH 8.5. Sections were incubated overnight with appropriate secondary antibodies (supplementary material Table S2) and coverslipped with antifade mounting medium Mowiol (4-88 reagent, Calbiochem 475904).

Images were processed using ImageJ (NIH) and Photoshop 7.0 (Adobe Systems). Confocal microscopy serial section 3D reconstructions were performed as described (Luzzati et al., 2011a). Briefly, images from each section were stitched in Fiji (Preibisch et al., 2009), aligned with Reconstruct 1.1 (Fiala, 2005) or with TracKEM2 (Cardona et al., 2010) and neurons were traced in NeuronStudio (Wearne et al., 2005). 3D models were rendered in Blender 2.6. (Blender Foundation) and Vaa3D (Peng et al., 2014).

We thank Claudio Giachino for critical reading of the manuscript; Rosanna Parlato and Ghunther Shutz for the Nestin-CreER^T2 transgenic line; Arturo Alvarez-Buylla for the Ad:GFAP-Cre plasmids; Elisa Vigna for VSVG-GFP plasmids; and Massimiliano Mazzone for TAT-Cre. This work is dedicated to the memory of Aldo Fasolo.