552 | Nature | Vol 582 | 25 June 2020
Article
ionotropic glutamatergic receptors (NBQX and APV) and an antagonist
of GABAA receptors (picrotoxin (PiTX)) blocked the signals, indicat-
ing that the converted neurons respond to synaptic inputs from both
glutamatergic and GABAergic neurons. No neuronal electrophysiologi-
cal properties were detectable in astrocytes transduced with control
virus (Extended Data Fig. 3e–h). These results demonstrate a one-step
conversion to functional neurons by depletion of PTB.
Generating new neurons in mouse brain
We next attempted to directly reprogram astrocytes into neurons in the
mouse brain. We designed an adeno-associated virus (AAV; serotype
2) vector to express shPTB (AAV-shPTB) (Fig. 2c) and a corresponding
empty vector lacking shPTB (AAV-empty) as control. To enable lineage
tracing, 5′ to the shPTB hairpin, we placed a red fluorescent protein
(RFP)-coding sequence that was initially silenced (by a loxP–Stop–loxP
cassette) but activated in cells expressing Cre recombinase. Focusing on
the substantia nigra of the midbrain where DA neurons reside (Fig. 2d),
we found that RFP+ cells were almost completely absent 10 weeks after
injecting either AAV-empty or AAV-shPTB in wild-type mice at 1 to 2
months of age, a developmental stage when astrocytes have already lost
their neurosphere-generating potential in the midbrain^16. By contrast,
RFP was expressed in response to both AAVs in Gfap-cre transgenic
mice expressing Cre recombinase from the astrocyte-specific Gfap
promoter^17 (Extended Data Fig. 4a, b).
Ten weeks after injection of AAV-empty into substantia nigra, most
RFP+ cells were astrocytes, as indicated by typical astrocytic mor-
phology and expression of the astrocyte markers S100b and ALDH1L1
(Extended Data Fig. 4c), with no evidence for viral transduction in NG2
cells (Extended Data Fig. 4d). We detected RFP in about 1% of NeuN+ neu-
rons (Extended Data Fig. 4e), demonstrating minimal Cre expression
in endogenous neurons in young adult mice. By contrast, 3 weeks after
AAV-shPTB injection, around 20% of RFP+ cells expressed NeuN; the
percentage of RFP+NeuN+ cells more than tripled by 5 weeks; and by 10
weeks around 80% of RFP+ cells were NeuN+GFAP– (Fig. 2e, f, Extended
Data Fig. 4e). At this time point, most converted neurons also expressed
multiple mature neuron markers (for example, MAP2, NSE and PSD95)
(Extended Data Fig. 4f ) and markers for glutamatergic (VGlut2) or
GABAergic (GAD65) neurons (Extended Data Fig. 4g). These results
demonstrate shPTB-mediated, time-dependent astrocyte-to-neuron
conversion in the mouse midbrain.
Progressive maturation of new DA neurons
We next monitored the gradual appearance of DA neurons among
RFP-labelled cells from 3 to 12 weeks after AAV-shPTB injection in the
midbrain (Fig. 2e, f). On the basis of staining with the DA neuron markers
DOPA decarboxylase (DDC) and TH, we detected a progressive increase
in the number of converted DA neurons, which reached 30–35% of
RFP+ cells 12 weeks after injection (Fig. 2f, Extended Data Fig. 5a, b).
All RFP+TH+ DA neurons were detected proximal to, but not distal
from, the site of injection where endogenous TH+RPF− DA neurons
reside (Extended Data Fig. 5c–e), indicating restricted astrocyte-to-DA
neuron conversion within the dopamine domain. Converted neurons
also expressed multiple DA neuron markers, such as DAT, VMAT2, EN1,
LMX1A and PITX3 (Extended Data Fig. 5f ), with morphology similar to
that of endogenous DA neurons (Extended Data Fig. 5g). A substantial
population of RFP+ cells (about 22% of RFP+ cells) expressed TH and
GIRK2 (a marker of A9 DA neurons and a subpopulation of A10 neurons),
whereas a minor population (about 7% of RFP+ cells) expressed TH
and calbindin-D28k (a marker of A10 DA neurons) (Fig. 2g), indicating
that different subtypes of DA neurons were generated. Furthermore,
SOX6-marked RFP+ DA neurons were confined to the substantia nigra
and OTX2-marked RFP+ DA neurons were confined to the ventral teg-
mental area (VTA); both types expressed a common DA neuron marker,
ALDH1A1 (Extended Data Fig. 5h–j). No RFP+TH+ cells were detected
following injection with AAV-empty (Extended Data Fig. 5k).
Patch clamp recordings of these converted neurons (illustrated in
Extended Data Fig. 6a, b) showed typical voltage-dependent currents of
sodium and potassium channels, repetitive action potential firing and
spontaneous postsynaptic currents. We also recorded the activity of
hyperpolarization-active and cyclic nucleotide-gated (HCN) channels
that could be specifically blocked with CsCl (Fig. 2h) and relatively wider
action potentials compared to those of GABAergic neurons (Extended
Data Fig. 6c, d)—both characteristics of mature DA neurons^18 ,^19. We
recorded no HCN channel activities and rather infrequent firing of
spontaneous action potentials at 6 weeks, and found HCN activities
and increased firing of spontaneous action potentials in a fraction of
RFP+TH+ DA neurons 12 weeks after injection (Extended Data Fig. 6e–g).
These results demonstrate progressive functional maturation of new
DA neurons within the dopaminergic neuron-containing domain of
the midbrain.NeuN+ cells
(per cent of RFP+ cells)TH+ cells
(per cent of RFP+ cells)
Midbrain Striatum CortexMidbrain Striatum Cortex 0.33 (NS)02550751000102030403.4 × 10 –5
3.9 × 10 –5NeuN/RFPTH/RFPcRFP/DAPI3 weeks 8 weeks 12 weeks dDensity of RFP+ bresRFP/DAPI
VMAT2/PSD95/RFPDay 90–100:
injected
retrobeadsDay 0:
injected
AAV-shPTBDensity of RFP+ bresDensity of
RFP+TH+ bres048120204060 0.01
0.6 (NS)e f(^03812)
5
10
15
20
25
Time (weeks)
0.17
(NS)
Retrobeads RFP TH Merge
g CPuNAcSeptOT
a b
CPuNAcSeptOT
2.2 × 10 –3
0.03
Fig. 3 | Regional specificity in astrocyte-to-neuron conversion and axonal
targeting. a, b, Induction of NeuN+ neurons in three brain regions, showing
that TH+ neurons were detected only in the midbrain. Scale bar, 40 μm. n = 3
mice. c, d, Progressive targeting of RFP+ fibres to striatum over the course of 12
weeks. Scale bar, 10 μm. RFP+ fibre density was determined by the sphere
method and quantified (d) from images (c) at each time point. n = 3 biological
repeats. e, Targeting of RFP+ fibres to multiple subregions around the
striatum, particularly septal nuclei (sept) (left), but RFP+TH+ fibres are
targeted mainly to CPu and NAc (right). Quantification was performed on
images collected at week 12. n = 3 mice. f, Evidence for synaptic connection in
CPu, as indicated in the magnified inset by colocalization of the presynaptic
marker VMAT2 (arrowheads) and the postsynaptic marker PSD95 (arrows) on
RFP+ fibres. Scale bar, 10 μm; magnified inset, 2 μm. g, Labelling of RFP+TH+
cells in substantia nigra with retrograde beads injected into striatum 90–100
days after reprogramming (left). Arrowheads indicate converted cells, one of
which was labelled by the beads; arrows show endogenous (TH+RFP−) DA
neurons labelled by the beads. Scale bar, 20 μm. In b, d, e, ANOVA with post hoc
Tukey test; data are mean ± s.e.m. NS, not significant. P-values are indicated. In
f, g, three independently repeated experiments with similar results.