A latent neurogenic program in astrocytes regulated by Notch signaling in the mouse

See allHide authors and affiliations

Science  10 Oct 2014:
Vol. 346, Issue 6206, pp. 237-241
DOI: 10.1126/science.346.6206.237


Neurogenesis is restricted in the adult mammalian brain; most neurons are neither exchanged during normal life nor replaced in pathological situations. We report that stroke elicits a latent neurogenic program in striatal astrocytes in mice. Notch1 signaling is reduced in astrocytes after stroke, and attenuated Notch1 signaling is necessary for neurogenesis by striatal astrocytes. Blocking Notch signaling triggers astrocytes in the striatum and the medial cortex to enter a neurogenic program, even in the absence of stroke, resulting in 850 ± 210 (mean ± SEM) new neurons in a mouse striatum. Thus, under Notch signaling regulation, astrocytes in the adult mouse brain parenchyma carry a latent neurogenic program that may potentially be useful for neuronal replacement strategies.

Astrocytes' hidden ability to generate neurons

New neurons could be useful in the brain to respond to damage done by disease, trauma, or just plain age. But the adult brain does not produce many new neurons. Working in mice, Magnusson et al. hunted for intrinsic genetic programs that can help adult brains produce replacement neurons. They found that astrocytes, spidery cells that are interspersed between neurons, carry a latent neurogenic program. The ability to entice astrocytes to replace neurons could obviate the need for neuronal replacement strategies.

Science, this issue p. 237

Neurogenesis in the adult brain is largely restricted to the dentate gyrus and the subventricular zone lining the lateral ventricles. However, astrocytes close to a lesion can display neural stem cell properties when assayed in vitro (13), and astrocytes can be forced to either convert into (4, 5) or produce neurons (6) when reprogrammed by ectopic expression of transcription factors in vivo.

To explore the in vivo neurogenic potential of astrocytes, we used Connexin-30–CreER (Cx30-CreER) transgenic mice (7) carrying a R26R–yellow fluorescent protein (YFP) reporter allele (8) for genetic fate mapping after experimental stroke. Stroke was induced by transient occlusion of the middle cerebral artery, a procedure that generated an ischemic lesion primarily in the striatum. We waited 1 week between giving tamoxifen and inducing the stroke, to allow for the tamoxifen to be eliminated and exclude the possibility of recombination in other cell types after stroke (Fig. 1A). The Cx30-CreER transgene allows specific recombination in a large subset of parenchymal astrocytes throughout the brain (37 ± 10% of glutamine synthetase+, S100+ cells in the striatum, mean ± SEM) (fig. S1), as well as in subventricular zone astrocyte-like neural stem cells (9).

Fig. 1 Striatal astrocytes produce neuroblasts after stroke.

(A) Experimental time line. In uninjured Cx30-CreER mice, recombined DCX+ neuroblasts are restricted to the subventricular zone (B) but appear also in the striatum after stroke (C and D). Arrowheads highlight examples of neuroblasts. LV, lateral ventricle. Error bars in (D) show SEM. w, weeks. (E to G) Many recombined striatal astrocytes up-regulate Ascl1 [(D) and (E)], and clusters of recombined, proliferating (Ki67+) cells expressing Ascl1 (F) or DCX (G) appear in the striatum. (H) Some recombined cells express the mature neuronal marker NeuN. (I to N) Injection of Adeno-GFAP-Cre virus with astrocyte-specific Cre expression into the striatum does not trigger a neurogenic reaction [(I) and (K)], but after stroke recombined astrocytes up-regulate Ascl1 and produce neuroblasts [(J) and (L) to (N)]. Scale bars: (C) and (J), 500 μm; (E) to (H) and (K) to (N), 10 μm.

As previously described, stroke triggered the appearance of many doublecortin-positive (DCX+) recombined neuroblasts in the affected striatum (10, 11) (Fig. 1, B to D) (n = 12 mice). We found signs that some of these neuroblasts might have been generated locally within the striatum, rather than having migrated from the subventricular zone. Beginning at 2 days after stroke, scattered astrocytes in the medial striatum expressed Ascl1 (Fig. 1E and fig. S2), a proneural transcription factor (12). At this time point, all S100+,Ascl1+ astrocytes proliferated (fig. S2, C to F) and subsequently formed clusters of recombined, proliferating Ascl1+ cells (Fig. 1F), no longer expressing the astrocyte marker S100 (fig. S2, G and H). Up to ~70% of S100+ astrocytes expressed Ascl1 in the most dense patches along the border of the lesion (fig. S3). Two weeks after the stroke, tightly packed clusters of recombined DCX+,Ki67+ neuroblasts appeared in the medial striatum; these neuroblasts were small and round and lacked the bipolar processes seen on migrating neuroblasts (Fig. 1G). All DCX+ cells coexpressed polysialylated neural cell adhesion molecule (PSA-NCAM) (fig. S2I). The number of cells in the striatum expressing Ascl1 and DCX increased from 2 to 7 weeks (3900 ± 990 DCX+ cells per lesion at 7 weeks, mean ± SEM) after the stroke (Fig. 1D) but maintained their general distribution (fig. S3). All described intermediate stages were continuously present, indicating that new astrocytes were being recruited for at least 7 weeks. Some of these neuroblasts developed into mature NeuN+ neurons (340 ± 160 cells per striatum at 7 weeks after stroke, mean ± SEM) (Fig. 1, D and H)—of which several expressed neuronal nitric oxide synthase (nNOS), a marker for GABAergic medium-sized striatal interneurons—and formed synaptic connections (fig. S4).

Because Cx30-CreER mice also target cells in the subventricular zone, it was not possible to exclude a periventricular origin of these new cells. We specifically fate-mapped striatal astrocytes by injecting adenovirus expressing Cre under either the general cytomegalovirus (CMV) promoter or the astrocyte-specific human glial fibrillary acidic protein (GFAP) promoter in the striatum of R26R-YFP mice (94 to 99% specificity) (fig. S5). We used an oblique injection route, allowing local recombination of striatal astrocytes without transducing cells in the subventricular zone or rostral migratory stream (Fig. 1, I and J, and fig. S6, A and B). The virus injection itself did not cause any ectopic Ascl1+ or DCX+ cells to appear (Fig. 1, I and K, and fig. S6, A and C). Animals were subjected to stroke and analyzed 7 weeks later. Many Ascl1+ cells and DCX+ neuroblasts, some of which expressed NeuN, expressed YFP in the virus-injected area of the striatum (Fig. 1, J and L to N, and fig. S6, B and D to G) (n = 6 mice), demonstrating that striatal astrocytes had generated neuroblasts after the stroke.

Seven weeks after stroke, 31 ± 4% (mean ± SEM) of recombined neuroblasts were found in clusters in Cx30-CreER mice, indicative of a local striatal origin. A fate-mapping study demonstrated that 33% of striatal neuroblasts derive from nestin-expressing subventricular zone stem cells after stroke (13), which suggests that approximately between one-third and two-thirds of the new neurons derive from striatal astrocytes after stroke.

Cx30-CreER–recombined striatal astrocytes in uninjured animals were negative for the neural stem cell marker nestin, whereas 95% were Sox2-positive (fig. S1, B, F, and H). In transgenic reporter mice where activation of the Notch1 receptor leads to expression of β-galactosidase (β-Gal) (14), β-Gal mRNA was abundant in the uninjured brain parenchyma, and half of all S100+ striatal astrocytes were positive for the β-Gal protein (Fig. 2, A, B, and E) (n = 5 mice). The nuclei of these cells also contained the cleaved Notch1 intracellular domain (NICD), which is present in cells with active Notch signaling (Fig. 2, A and E).

Fig. 2 Reduced Notch signaling in astrocytes that enter a neurogenic program.

Many astrocytes in the uninjured striatum have active Notch1 signaling, as revealed by Notch1 activity–induced β-Gal expression and nuclear NICD protein (A and B). Two weeks after stroke in Notch1 reporter mice, cells in a large area of the injured striatum lack NICD protein (C) and β-Gal mRNA (D), but many still contain the long-lived β-Gal protein (C). (E) NICD and β-Gal in striatal S100+ astrocytes. Most Ascl1-expressing astrocytes lack nuclear NICD in the injured striatum (F) but retain β-Gal protein (G). Arrowheads highlight individual cells. Scale bars: (A), (C), (F), and (G), 10 μm; (B) and (D), 100 μm. DAPI, 4′,6-diamidino-2-phenylindole.

Neural stem cell quiescence is regulated in part by Notch1 signaling (12, 15). We found that stroke caused a reduction in the protein levels of Notch receptors 1, 2, and 3, as well as the Notch ligands Dll1, Jagged1, and Jagged2, whereas Dll3 levels were increased, in the affected striatum 3 days after stroke (fig. S7). Two weeks after stroke, both NICD protein and β-Gal mRNA in Notch1 reporter mice had become undetectable in a large area in the striatum, demonstrating that cells no longer had active Notch1 signaling (Fig. 2, C and D) (n = 6 mice). However, many NICD-negative astrocytes still contained detectable β-Gal protein (Fig. 2, C and E), probably due to a relatively high stability of this protein (16). Most Ascl1+ astrocytes were located on the very border of the area devoid of NICD immunoreactivity and were negative for NICD (Fig. 2F). They did, however, still contain low levels of β-Gal protein 3 days and 2 weeks after stroke (Fig. 2G). The absence of NICD and presence of β-Gal protein suggested that Notch1 signaling was reduced as astrocytes entered the neurogenic program and that the astrocytes that up-regulated Ascl1 and produced neuroblasts after stroke were the ones that had been identifiable by having active Notch1 signaling before the injury (fig. S8).

To assess whether reduction of Notch1 signaling is required for triggering striatal astrocytes to enter the neurogenic program in response to stroke, we artificially maintained Notch1 signaling after stroke. We used Cx30-CreER mice homozygous for conditional NICD alleles, in which tamoxifen administration results in ectopic expression of NICD and green fluorescent protein (17). Seven weeks after stroke, Cx30-CreER;R26R-YFP control mice had recombined Ascl1+ and DCX+ cells in the striatum, as expected (6 and 16 cells/mm2, respectively; n = 4 mice). However, virtually no recombined Ascl1+ cells (0.04 cells/mm2) or DCX+ cells (0.02 cells/mm2) were found in the striatum when NICD was ectopically expressed in astrocytes (n = 4 mice) (Fig. 3 and fig. S9). Thus, reduced Notch1 signaling is necessary for activation of the latent neurogenic program in the striatum.

Fig. 3 Ectopic Notch1 activation in astrocytes inhibits stroke-induced neuroblast production.

Ectopic Notch1 activation is sufficient to inhibit Ascl1 expression and neuroblast production by astrocytes after stroke. Error bars show SEM. GFP, green fluorescent protein.

We next asked whether experimental reduction of Notch signaling in striatal astrocytes could activate the latent neurogenic program, even in the absence of stroke. We deleted RBP-Jκ (CSL), an obligatory transcription factor for canonical Notch signaling that is present in all NICD+ striatal astrocytes (fig. S10A), using Cx30-CreER; R26R-YFP mice homozygous for conditional RBP-Jκ null alleles (18). Abolishing Notch signaling in astrocytes largely phenocopied the effect of stroke (n = 20 mice) (Fig. 4, A and B, and fig. S10, B to F). Ascl1+ astrocytes first appeared in the medial striatum 2 weeks after tamoxifen administration and 1 week later across the entire striatum. Proliferation was restricted to the S100+,Ascl1+ cells in the medial striatum (fig. S11). From 3 weeks after recombination, tightly packed clusters of recombined, proliferating cells expressing Ascl1 and DCX, or containing a mix of both, appeared in the same region (figs. S10, C to F; S11F; and S12). The vast majority of neuroblasts (83 ± 2%, mean ± SEM) were found in clusters in the striatum 3 weeks after tamoxifen administration, suggesting that striatal astrocytes represent the dominating source of neuroblasts after blocking Notch signaling. Whereas the number of striatal Ascl1+ cells was highest 4 weeks after tamoxifen injection, neuroblast numbers increased with time, peaking at up to 17,000 cells per striatum 8 weeks after the administration of tamoxifen (6400 ± 3200 DCX+ cells per striatum, mean ± SEM) (Fig. 4C). Even 34 weeks after tamoxifen administration, Ascl1+ astrocytes and some DCX+ cells were present in the striatum. If each cluster originated from a single astrocyte, we estimated that each Ascl1+ astrocyte initiates a series of four to five cell divisions before differentiating into DCX+ cells that divide once, resulting in ~40 neuroblasts. Over time, DCX+ cells developed elaborate processes and neuronal morphology, and up to 1500 recombined cells per striatum expressed NeuN (850 ± 210 cells per striatum at 34 weeks, mean ± SEM) (Fig. 4, D and E), of which the majority were positive for nNOS and some had formed synaptic connections (fig. S13).

Fig. 4 Deletion of RBP-Jκ triggers neuroblast production by astrocytes in the striatum and medial cortex.

(A to C) Deletion of RBP-Jκ in astrocytes of adult Cx30-CreER mice leads to the appearance of ectopic Ascl1+ cells and DCX+ neuroblasts in the striatum in the absence of injury. Arrowheads point to examples of neuroblasts. cKO, conditional knockout. (D and E) With time, some striatal neuroblasts mature into neurons. (F to L) Injection of Adeno-GFAP-Cre virus into the striatum of animals heterozygous for the mutated RBP-Jκ allele does not cause Ascl1+ or DCX+ cells to appear [(F) and (H)] but in homozygous animals induces astrocytes to up-regulate Ascl1 and produce neuroblasts [(G) and (I) to (L)]. Arrowheads in (J) to (L) highlight individual cells. (B) and (M to R) Cx30-CreER–mediated RBP-Jκ deletion triggers the appearance of Ascl1+ cells and neuroblasts also in the superficial medial cortex [red areas in (M), arrows in (B)]. Arrowheads in (R) show cell clusters. Error bars in (C), (D), and (N) show SEM. Scale bars: (B) and (G), 500 μm; (E) and (H) to (L), 10 μm; (O) and (Q), 200 μm; (P) and (R), 25 μm.

Deletion of RBP-Jκ exclusively in striatal astrocytes, using either the Ad-GFAP-Cre or Ad-CMV-Cre virus, caused up-regulation of Ascl1 in 43 ± 4% (mean ± SEM) of the recombined astrocytes and the generation of recombined neuroblasts (13 ± 5% of YFP+ cells at 8 weeks after virus injection) in mice homozygous (n = 9 mice), but not heterozygous (n = 21 mice), for the conditional RBP-Jκ mutation (Fig. 4, F to L, and fig. S14, A to E). Some neuroblasts developed branched processes and up-regulated NeuN (fig. S14F). No migration of recombined neuroblasts was observed from the subventricular zone in these experiments, demonstrating the specificity of the recombination strategy to striatal astrocytes and corroborating on-site neurogenesis within the striatum.

We finally asked whether the Notch-regulated latent neurogenic program was restricted to astrocytes in the striatum. Analysis of the brains of Cx30-CreER;R26R-YFP mice in which RBP-Jκ had been ablated in astrocytes revealed a neurogenic potential of astrocytes also in the medial cortex (Fig. 4M). Ascl1-expressing astrocytes and cell clusters, as well as up to 2900 DCX+ neuroblasts (1600 ± 530 DCX+ cells 8 weeks after tamoxifen administration, mean ± SEM), were seen in the superficial cortex in all animals (n = 20 mice), in a band <200 μm on either side of the medial longitudinal fissure (Fig. 4, B and M to R, and fig. S15). DCX+ cells in this area never acquired bipolar processes but were always small, round, and tightly clustered together. No ectopic DCX+ cells were seen in other parts of the central nervous system. In mice in which astrocytes were targeted instead using a GLAST-CreER transgene (7), recombined ectopic Ascl1+ and DCX+ cells were generated in the same regions as in the Cx30-CreER mice, both after stroke (striatum; n = 9 mice) and after RBP-Jκ deletion (striatum and medial cortex; n = 6 mice) (fig. S16). Neither stroke nor deletion of RBP-Jκ resulted in generation of neuroblasts from oligodendrocyte progenitors (fig. S17).

Our results reveal a latent neurogenic potential in a population of astrocytes outside the neurogenic niches and delineate its molecular regulation. In contrast to other mammals, adult humans exhibit substantial neurogenesis in the healthy striatum (19). Our results raise the possibility that some new striatal neurons may derive from local astrocytes in the adult human brain, rather than from the adjacent subventricular zone. Neuronal loss in the striatum characterizes several neurological conditions. Blocking canonical Notch signaling resulted in a comparable density of new neurons (42 neurons/mm3), as has been reported to have beneficial effects after interneuron transplantation (39 neurons/mm3; see supplementary materials and methods) in an animal model of Parkinson’s disease (20), suggesting that inducing endogenous generation of interneurons can be an attractive alternative to transplantation. Our findings of a latent neurogenic program within the striatum and the characterization of its molecular regulation point to a route for developing neuronal replacement therapies.

Supplementary Materials

Materials and Methods

Figs. S1 to S17

References (2125)

References and Notes

  1. Acknowledgments: We thank E. Andersson, E.-B. Braune, J. Dias, S.-B. Jin, and U. Lendahl for valuable discussions. This study was supported by grants from the Swedish Research Council, the Swedish Cancer Society, the Karolinska Institute, Tobias Stiftelsen, The Swedish Heart and Lung Foundation, StratRegen, StemTherapy, European Union project TargetBraIn (279017), Torsten Söderbergs Stiftelse, and Knut och Alice Wallenbergs Stiftelse. D.O.D. was supported by the Portuguese government (SFRH/BD/63164/2009). C.G. is a Hållsten Academy and Wallenberg Academy fellow. We thank F. W. Pfrieger for providing Cx30 and GLAST transgenic mice through a material transfer agreement with the Institut Génétique Biologie Moléculaire Cellulaire (Strasbourg, France). Requests for mice should be directed to F. W. Pfrieger (CNRS, France). We declare no conflicts of interest. The supplementary materials contain additional data.
View Abstract

Navigate This Article