Direct and efficient transfection of mouse neural stem cells and mature neurons by in vivo mRNA electroporation

Manipulation of gene expression in the brain of living animals represents a key approach in neurobiological research today. Over the past decades, mouse genetic approaches have been developed and widely used to study biological processes either via ectopic/overexpression or gene knockout in transgenic mice. However, the generation of genetically modified mice is a long-term project that requires considerable resources and infrastructure. Moreover, overexpression or inactivation of a given gene in a specific tissue/cell type or at a defined moment implicates complex strategies involving the concomitant presence of several transgenes in a given animal, rendering this approach expensive and time-consuming.

Several methods were developed to express transgenes in specific cell populations of living animals, including neurons. Viral vectors have been widely used for in vivo gene function studies or even therapeutic approaches (Kotterman and Schaffer, 2014; Papale et al., 2009). Although these tools allow the relatively rapid manipulation of defined brain structures, safety issues and immune reactions in the host tissue represent considerable limitations for their use (Lowenstein et al., 2007; Sack and Herzog, 2009).

Electroporation of plasmid DNA expression constructs to induce overexpression or knockdown of protein-coding genes or microRNAs (miRNAs) is nowadays a broadly used approach in studies of the developing and postnatal mouse brain (De Vry et al., 2010). However, DNA electroporation is of limited efficiency and allows generally only the targeting of embryonic (Saito and Nakatsuji, 2001) and early postnatal neural stem cell compartments (Boutin et al., 2008, 2010). Postmitotic neurons or adult neural stem cells in the subventricular compartment are highly refractory to DNA electroporation (Barnabé-Heider et al., 2008; De la Rossa and Jabaudon, 2014), even though some invasive techniques were developed to locally electroporate adult mature neurons (Molotkov et al., 2010; Ohmura et al., 2015; Pagès et al., 2015; Porrero et al., 2016).

In specific experimental models, such as Ciona or Xenopus Laevis oocytes, mRNA injection was extensively and efficiently used for ectopic gene expression or knockdown (Bertrand et al., 2003; Mimoto and Christian, 2011). Recently, mRNA transfection of combination of different transcription factors has been used in vitro in human cells to reprogram somatic cells into induced pluripotent stem cells (Luni et al., 2016) and to drive the neuronal differentiation of progenitor cells into specific neuronal cell types (Faedo et al., 2016). Based on these findings, we investigated the potential of in vitro-generated mRNAs for functional studies in the mouse brain. We show that mRNA electroporation is considerably more efficient than DNA for targeting the postnatal neural stem cell pool and leads to more homogeneous expression levels. mRNA-based protein expression is stable over days. Importantly, mRNA electroporation allows efficient targeting of neural stem cells in the adult ventricular/subventricular zone (VZ-SVZ) and even fully mature neurons integrated in the cortex and other brain structures. We demonstrate the usefulness of mRNA electroporation for functional analyses in the adult brain. First, we reveal that transient expression of the neuronal gene Myt1l can drive the transformation of neural stem cells into proliferative progenitors. Second, we show that stable recombination of the Dicer locus through CRE-mRNA electroporation in adult neural stem cells and layer 5 mature cortical neurons interferes with dendritic spine morphology and density, providing new in vivo evidence for the involvement of the microRNA pathway in synapse regulation.

Postnatal in vivo brain electroporation allows the transfection of DNA constructs into the stem cell compartment surrounding the walls of the lateral ventricles (LV). Position of the electrodes around the head of the animal, and therefore the orientation of the electric field, allows specific targeting of the lateral or dorsal aspects of the ventricular wall (Boutin et al., 2008; de Chevigny et al., 2012; Fernández et al., 2011) (Fig. 1A). Electroporation of the lateral wall of the LV with an expression plasmid coding enhanced green fluorescent protein (EGFP) (pCAGGS-EGFP, 2 µl of a 5 µg/µl solution) produced at 1 day post-electroporation (dpe) a typical sparse salt-and-pepper labeling with clearly distinguishable individual cells that showed radial glia morphology (Fig. 1A,C). DNA electroporation of the dorsal wall led generally to higher EGFP⁺ cell densities. In both, lateral and dorsal compartments, EGFP intensity was highly variable between individual cells (Fig. 1C).

We followed the same experimental protocol using in vitro-synthesized mRNA encoding EGFP (StemMACS, EGFP mRNA, Miltenyi Biotec; 2 µl at 0.5 µg/µl, approximately equimolar to the used DNA concentration). As for DNA, orientation of the electroporation electrodes allowed the specific targeting of the lateral or dorsal VZ-SVZ domains (Fig. 1B). Both, lateral and dorsal electroporation of mRNA induced considerably stronger EGFP labeling of the ventricular walls at 1 dpe (Fig. 1D) compared with DNA-based transfection.

To quantitatively evaluate the targeting efficiency of neural stem cells by DNA versus mRNA electroporation, we analyzed ventricular Lateral Wall Whole Mount (LWWM) (Mirzadeh et al., 2008) preparations at 1 dpe. Low-magnification views of the lateral walls after DNA electroporation showed scattered EGFP staining with varying fluorescence levels among positive cells (Fig. 1E,G). In contrast, mRNA efficiently and homogeneously labeled the almost entire lateral wall of the lateral ventricle (Fig. 1F,G).

High-magnification images of whole-mount preparations (Fig. 1G) allowed us to precisely quantify the percentage of EGFP-positive cells in the ventricular wall. Overall, there was a 17-fold increase in the amount of transfected cells after mRNA electroporation compared with DNA (Fig. 1G,H).

We then quantified the mean EGFP fluorescence intensity in individual electroporated cells on LV coronal sections. The vast majority of DNA electroporated cells were poorly fluorescent. Only a small fraction showed very high intensity levels. In contrast, after mRNA electroporation almost all cells were intensely fluorescent (Fig. 1J). Moreover, the mean EGFP fluorescence intensity among individual electroporated cells (Fig. 1I,J) showed a twofold lower standard deviation after mRNA electroporation compared with DNA (Fig. 1J). Thus, mRNA electroporation, in addition to targeting more cells, led to more homogenous protein levels in electroporated neural stem cells than using DNA plasmids (Fig. 1I,J). Moreover, EGFP fluorescence level per cell varies with the mRNA concentration, permitting control of protein concentration in gain-of-function situations (Fig. S1). The impact of transcript size on electroporation efficiency and protein expression has not been studied in the context of this work. In conclusion, in vivo mRNA electroporation provides an efficient means of labeling postnatal neural stem cells, improving both the rate and homogeneity of electroporation compared with the use of plasmid DNA.

We investigated the dynamics of protein appearance after mRNA versus DNA electroporation. Two hours after lateral DNA electroporation, fluorescence was undetectable. After 6 h sparsely distributed individual cells were EGFP positive (Fig. 1K). In contrast, mRNA electroporation induced homogeneous EGFP fluorescence in large amounts of cells as early as 2 h after electroporation, which increased in intensity at 6 h (Fig. 1K). Thus, in agreement with the absence of need for a transcription step, protein production was considerably faster after mRNA electroporation compared with DNA.

Next we investigated the stability of mRNA over time after in vivo electroporation. mRNA electroporation followed by microdissection of VZ-SVZ tissue and qRT-PCR (quantitative reverse transcription polymerase chain reaction) for the presence of EGFP mRNA demonstrated that, at 2 h post electroporation, high quantities of transcript could be detected. However, at 1, 2 and 7 dpe very low EGFP mRNA levels were present, indicating rapid degradation of transcripts (Fig. 1L). Nevertheless, EGFP fluorescence was detectable until 7 dpe, pointing towards the high stability of EGFP protein in cells (Fig. 1M) (Corish and Tyler-Smith, 1999). Thus, although electroporated mRNAs generate high quantities of protein during a short period, active protein can be identified for prolonged periods.

The above results show that mRNA electroporation allows the highly efficient transient expression of exogenous proteins in vivo. In the next steps, we aimed to combine the efficiency of mRNA electroporation with the permanent genetic modifications induced by CRE-loxP-induced recombination to allow long-term cell-tracing and gene-function analysis.

First, we asked whether CRE/loxP mRNA-mediated recombination could be used for the efficient long-term labeling of defined stem cell and neuron populations in vivo. During postnatal neurogenesis, the vast majority of newborn olfactory bulb (OB) neurons is generated from the lateral ventricle walls and populates mainly deep positions of the OB after migration. Smaller subsets are generated from the dorsal and medial aspects, and integrate mainly into the superficial layers of the granule cell layer (GCL) and the glomerular layer (GL) (Merkle et al., 2007).

We electroporated mRNA encoding CRE recombinase (StemMACS CRE Recombinase mRNA, Miltenyi Biotec; 0.5 µg/µl) into the lateral and medial aspects of the forebrain ventricles of newborn R26-Ai14 mouse pups, carrying a CRE-inducible tomato allele in the ROSA26 locus (Madisen et al., 2010). Twenty-one days later large parts of the lateral and medial ventricular walls showed strong tomato fluorescence, demonstrating efficient recombination in the stem cell compartment (Fig. 2A). Consistent with this high transfection level, we observed large amounts of cells in the rostral migratory stream (Fig. 2C) as well as many postnatally generated tomato-positive neurons in the OB (Fig. 2D-F). After lateral electroporation, these fluorescent neurons were predominantly localized in deep positions in the GCL (Fig. 2D,E) whereas after medial electroporation they settled mainly in the GL (Fig. 2D,F), in agreement with the localization of the respectively targeted stem cell pools (Fernández et al., 2011). Moreover lateral electroporation labeled numerous fluorescent astrocytes in the striatum (Fig. 2B). Together, these results show that combining the efficiency of mRNA electroporation with the permanence of CRE recombination is a powerful approach with which to label and manipulate large amounts of newborn cells in the postnatal mouse brain. We also found that CRE mRNA is suited to efficiently induce recombination in the embryonic forebrain (Fig. S2).

Although DNA electroporation is routinely used to study embryonic and postnatal neurogenesis, adult neurogenesis of the olfactory bulb is only poorly targeted by this approach. Here, current in vivo transfection strategies rely on the use of virus infection, which does not allow the targeting of defined subregions of the ventricular walls. We asked whether mRNA electroporation could be used to overcome these limitations.

Electroporation of a pCAGGS-CRE plasmid (10 µg/µl) in the lateral wall of 8-week-old R26-Ai14 mice resulted in the appearance of individualized cells in the ventricular wall at 5 dpe. These cells had generally the typical cuboid morphology of ependymocytes (Fig. 3A). When CRE mRNA (0.5 µg/µl) was used under the same conditions, 5 days later the entire lateral wall was covered with tdTomato-positive cells (Fig. 3A). High magnification images and S100β immunostaining validated ependymal targeting (Fig. 3A,B, yellow arrowheads). Moreover, individual cells in subependymal positions and extending basal processes showed tdTomato fluorescence (Fig. 3B, white arrowheads). Position, morphology and the observation that many of these cells expressed the glial marker GFAP (glial fibrillary acidic protein) indicated that they were adult B1-type neural stem cells (Fig. 3A,B). Supporting this observation, ventricular lateral wall whole-mount preparations of CRE mRNA-electroporated animals at 5 dpe showed characteristic pinwheel structures (Mirzadeh et al., 2008) with recombined multi-ciliated ependymocytes (white dotted line) surrounding recombined mono-ciliated B1 stem cells (white arrowhead in Fig. 3C).

In accordance with the previous observation that adult neural-stem cells could be targeted by mRNA electroporation, 28 days after CRE mRNA electroporation animals displayed a substantial amount of adult-born neurons in the olfactory bulb (Fig. 3D,E). This was strikingly different from DNA electroporations in which OB neurons were observed only sporadically (Fig. 3D-F). Thus, mRNA electroporation allowed the efficient and reliable genetic manipulation of adult neural stem cells and their derivatives.

Next, we asked whether it was possible to target mRNA to specific adult VZ-SVZ subregions, and subsequently label neurons integrated in specific layers of the OB (Fig. 4A), comparable with the situation in postnatal animals. Both the dorsal and lateral walls could be labeled by CRE mRNA electroporation into Ai14 mice. Both targeted compartments gave rise to new olfactory interneurons at 28 dpe, with lower amounts of cells generated from dorsal electroporations (Fig. 4B), as expected. We analyzed the distribution of the adult generated cells in the OB. Similar to what we observed with postnatal electroporation, lateral electroporation strongly biased the neuronal production towards deep granule cells, whereas dorsal electroporation mostly generated superficial granule cells and periglomerular cells (Fig. 4C). Thus, mRNA electroporation represents an efficient new tool with which to manipulate forebrain neurogenesis and neural stem cell regionalization in the adult mouse brain.

Postmitotic neurons in the adult brain are refractory to transfection with DNA (De la Rossa and Jabaudon, 2014). We investigated whether the use of mRNA could overcome this limitation. Intra-parenchymal injection of CRE mRNA in adult R26-Ai14 mice was followed by electroporation (Fig. 5A). This approach reliably targeted mature neurons in different regions of adult brain, including layer 5 pyramidal neurons in S1 somatosensory cortex (Fig. 5B), medium spiny neurons of dorsal striatum (Fig. 5E) and medial septum neurons (Fig. 5G), allowing the study of neuronal morphology at synaptic resolution (Fig. 5C,F,H). Injected but non-electroporated control brains never showed tdTomato-positive cells.

Electroporated cells were mostly located close to the injection site, likely reflecting low diffusion of mRNA in the tissue. In addition to mature neurons, about 20% of the targeted cells were NeuN negative. These cells showed typical astrocytic morphologies and over 90% expressed the glial marker GFAP, altogether showing that mRNA electroporation targets preferentially, but not exclusively, neurons (Fig. 5D). Overall, our results show that CRE mRNA electroporation efficiently transfects adult postmitotic and fully integrated neurons in different brain areas.

Next, we provided proof of principle that mRNA electroporation can be used for functional analyses using both gain- and loss-of-function approaches. As a prerequisite, we investigated the degree of co-transfection when two mRNAs were used in parallel. Co-electroporation of equal amounts of EGFP and CRE mRNA at P1 showed that almost 70% of all fluorescent cells were double positive at 1 dpe (Fig. 6A,B). Thus co-expression of two proteins by mRNA electroporation is efficient.

On the basis of this, we first performed a gain-of-function approach. The pan-neuronal basic helix-loop-helix transcription factor Myt1l has been implicated in the acquisition of neuronal identity by repressing non-neuronal fates (Mall et al., 2017). Moreover, expression of Myt1l, together with Brn2 and Ascl1, has been used for the induction of highly proliferative neuronal precursors from fibroblasts (Lim et al., 2015).

We investigated the consequences of Myt1l overexpression in postnatal neural stem cells. Co-electroporation of equal amounts of EGFP and Myt1l mRNAs at P1 led to increased numbers of BrdU-positive cells in the SVZ at 2 dpe (Fig. 6C,D), in agreement with the induction of additional proliferative interneuron precursors.

In the next step we performed a loss-of-function approach, thereby focusing on the microRNA pathway. Dicer is a cytoplasmic nuclease that is essential for the biogenesis of functional microRNAs. Dicer mutant mice have been widely used to study the microRNA pathway in vivo. However, constitutive or conditional loss of function for Dicer in mice induces severe developmental phenotypes, rendering the analyses of late and/or cell-autonomous roles of the microRNA pathway difficult (Bernstein et al., 2003). Nevertheless, specific microRNAs have been implicated in the control of synapse formation and plasticity in the developing and adult brain (Pathania et al., 2012; Siegel et al., 2011).

Based on these observations, we asked whether CRE-mRNA electroporation could be used to study the consequences of Dicer deficiency on synaptic spine density and structure in adult-generated OB neurons and fully mature cortical projection neurons. We generated a double transgenic mouse line harboring a conditional Dicer allele (Dicer^flox) (Harfe et al., 2005) and the Ai14 allele (see above) to allow detection of recombination events and morphological analysis at synaptic resolution. First, we investigated Dicer function during adult OB neurogenesis. We performed intraventricular injection of CRE-mRNA in 2-month-old Dicer^wt/flox and Dicer^flox/flox mice, and electroporated into adult neural stem cells of the lateral wall. Analyses of adult-born granule cells at 28 dpe (Fig. 6E) revealed a 30% decrease in the density of apical tuft dendritic spines (Fig. 6F), without any significant difference regarding the length of these spines (Fig. 6G).

Next, we asked whether Dicer expression is essential for the maintenance of a correct spine density and structure in integrated and fully mature cortical neurons that were generated during embryogenesis. We performed intra-parenchymal injection of CRE-mRNA into the adult neocortex of Dicer^wt/flox and Dicer^flox/flox mice and investigated spine density and morphology of recombined layer 5 pyramidal neurons of S1 somatosensory cortex (Fig. 6H). Quantitative analyses at 28 dpe revealed no change in spine density (Fig. 6I). However, in this situation dendritic spine length was significantly decreased (Fig. 6J).

Thus, we show that mRNA electroporation is an efficient and alternative method to analyze gene function in the brain. We show that overexpression of a transcription factor by this approach interferes with proliferation in postnatal neural stem cells. When applied to Dicer-conditional mutant mice, we demonstrate a function of the microRNA pathway in synaptic spine density and architecture in the adult brain.

We show here that mRNA electroporation holds considerable advantages over classical approaches based on DNA electroporation. First, transfection efficiency in vivo using mRNA is greatly increased over DNA. Electroporation of equimolar amounts of mRNA and DNA led to an over 17-fold increase in the amount of EGFP-expressing cells along the ventricular wall of the forebrain. Transfection efficiency in electroporation is a function of the size of the used nucleic acid molecules (Ribeiro et al., 2012). Given that the EGFP mRNA is about ten times smaller than the corresponding DNA plasmid, its transfer into the cell might be facilitated. Moreover, certain cell types in the ventricular wall might be refractory to DNA-based expression but permissive for RNA. For example, in our hands postmitotic ependymocytes are difficult to electroporate with DNA constructs, while both EGFP and CRE mRNAs are efficiently expressed in this cell type. It is possible that the breakdown of the nuclear membrane during mitosis allows the introduction of episomal DNA into the nucleus, a key prerequisite for the generation of functional mRNA and subsequently protein. Once the nuclear membrane is stable and permanent, nuclear traffic becomes highly regulated and DNA transfer into the nucleus is not efficient anymore. In contrast, RNA does not need to enter the nucleus of target cells to be expressed.

It would have been interesting to investigate whether electroporation efficiency is similar between mRNA of significantly different sizes. However, analyzing the influence of mRNA size on electroporation efficiency has several drawbacks. For example, addressing this point based on the expression of a protein translated from the electroporated mRNA is biased by the detection threshold of each protein. On the other hand, direct measurement of the abundance of electroporated transcripts in electroporated cells is hampered by the short half-life of mRNAs in the cell and the potential impact of the diverging sequences on transcript stability.

Second, mRNA electroporation leads to faster and more homogeneous expression of transgenes. Once introduced into a cell, DNA needs to enter the nucleus in order to be transcribed into mRNA, which then needs to be re-exported to the cytoplasm for translation. In contrast, mRNA can be directly used for translation in the cytoplasm, leading to the very rapid accumulation of a wanted protein. Only 2 h after mRNA electroporation, EGFP fluorescence was detectable in the targeted cells. This rapid and homogeneous expression might be of major impact for functional studies. For example, major transcriptional regulators that control neural stem cell identity and fate might act during short and early time windows. mRNAs appear well suited to provide better quantitative and temporal control. To exemplify this point, we were able to significantly induce the generation of proliferative progenitors by electroporating Myt1l mRNA into neural stem cells.

Third, mRNA electroporation can be used to conveniently manipulate both stem cells and postmitotic neurons in the adult brain. Currently, acute transgenesis in adult animals is mainly performed by infection with recombinant viruses. However, this approach has some important limitations. For example the production and use of active viral particles demand particular safety measures, permissions and know-how, which are not always available. Moreover, virus infection can induce immunological reactions of the host tissue that can interfere with the analysis of a given phenotype (Lowenstein et al., 2007; Sack and Herzog, 2009).

The use of mRNA electroporation is limited by the shorter persistence of the electroporated RNA compared with DNA. To overcome this limitation, we combined RNA electroporation with CRE-LoxP-induced genetic manipulation to induce long-term changes in gene expression. Using this approach, we were able to efficiently and durably change gene expression in adult neurons of the olfactory bulb, after targeting adult neural stem cells of the VZ/SVZ region. Importantly, we were also able to alter gene expression in post-mitotic cortical projection neurons by simple intra-parenchymal injections of very small liquid volumes, thus avoiding severe disruption of the surrounding brain tissue. Previous approaches that demonstrated electroporation of DNA into neurons were of low efficiency and highly invasive, requiring the implantation of electrodes into the brain tissue (Molotkov et al., 2010; Ohmura et al., 2015; Pagès et al., 2015; Porrero et al., 2016). We provide the first non-viral method for transfecting adult neurons in different brain regions based on non-invasive surface electrodes.

Finally, we validated the mRNA electroporation approach to study gene function in vivo by investigating the role of the microRNA pathway in synapse formation and stability in adult brain. Modified expression of specific microRNAs was shown to impact dendritogenesis and spine formation in developing neurons (Bicker et al., 2014). Moreover, microRNAs were implicated in activity-dependent translation at the synapse (Cohen et al., 2011; Gu et al., 2015; Weiss et al., 2015). Alterations in such processes may lead to psychiatric disorders. For example, deregulation of miR-137 is associated with schizophrenia and mental disability in humans (Siegert et al., 2015).

So far, studies aiming at deciphering the role of microRNAs in synaptic functions have been performed in vitro and/or in a context where the change in microRNA expression was constitutive. The mRNA electroporation approach we present here reveals the specific involvement of the microRNA pathway in adult animals in the establishment of synapses during neurogenesis, as well as the maintenance of spine morphology in neurons that were already functionally integrated.

CD1 (Charles-River, Lyon, France), Ai14 transgenic reporter (Jackson Laboratories, stock number 007914) and Dicer-floxed (Jackson Laboratories, stock number 006001) male and female mice were used. Animal experiments were carried out in accordance with European Communities Council Directive and approved by French ethical committees (Comité d'Ethique pour l'expérimentation animale no. 14; permission number: 62-12112012).

The pCAGGS-CRE and pCAGGS-EGFP vectors were derived from pCX-MCS2 (Morin et al., 2007). Both plasmids were used at a concentration of 5 µg/µl (0.1% Fast Green), except for adult electroporation (Fig. 3), where pCAGGS-CRE was used at 10 µg/µl.

mRNAs were provided by Miltenyi Biotec (Miltenyi Biotec): EGFP mRNA (130-101-114), CRE Recombinase mRNA (130-101-113) and Myt1l mRNA (130-104-379) (a kind gift from A. Bosio, Miltenyi Biotec). mRNAs were synthesized in an animal component-free production process using T7-based transcription from a cDNA vector (optimized with regard to human codon usage) and subsequent 5′ capping and 3′ polyadenylation. Optionally, Ψ-UTP- and 5-mCTP-modified nucleotides were introduced to reduce innate immune responses against mRNA in downstream applications. mRNAs were DNase treated, sterile filtered and lyophilized. mRNAs were then diluted in RNAse-free PBS at a concentration of 0.5 µg/µl (same range of molarity compared with 5 µg/µl DNA plasmid), except for Fig. S1, where mRNA concentrations are indicated.

Postnatal electroporation was performed as described previously (Boutin et al., 2008). Briefly, P1 pups were anesthetized by hypothermia. A glass micropipette was inserted into the lateral ventricle and 2 μl of plasmid or RNA solution was injected by expiratory pressure using an aspirator tube assembly (Drummond). Successfully injected animals were subjected to five 95 V electrical pulses (50 ms, separated by 950 ms intervals) using the CUY21 edit device (Nepagene) and 10 mm tweezer electrodes (CUY650P10, Nepagene) coated with conductive gel (Control Graphique Medical). Electroporated animals were reanimated in a 37°C incubator before returning to the mother.

Timed pregnant E15.5 CD1 mice were anesthetized and the uterine horns were exposed. CRE mRNA (at 0.5 µg/µl) was then injected into the lateral ventricle of the embryos with a glass micropipette by expiratory pressure using an aspirator tube assembly (Drummond). Fast Green was added at 0.01% for visualization of injection. Successfully injected animals were subjected to five 35 V electrical pulses (50 ms, separated by 950 ms intervals) using the CUY21 edit device (Nepagene) and 3 mm tweezer electrodes (CUY650P3, Nepagene).

Eight- to ten-week-old animals were deeply anesthetized intraperitoneally [3.0 mg of ketamine (Imalgene); 0.25 mg of xylazine (Rompun; Bayer)] and stereotaxically injected with 2 µl (injection in the lateral ventricle) or 200 nl (intraparenchymal injections) of nucleic acid solution using a Nanoject II (Drummond) equipped with a glass micropipette, at a rate of 230 or 23 nl/s, respectively. The coordinates for injections in the lateral ventricle were AP 0.0; L 1.0; V −1.5 mm, AP 0.2; L 2.0; V −0.8 mm for somato-sensory cortex; AP 0.2; L 0.5; V −2 mm for septum; and AP 0.2; L 2.0; V −2.5 mm for dorsal striatum, relative to Bregma. After injection, animals were subjected to five 200 V electrical pulses.

Immunostaining of ventricular lateral-wall whole-mount (LWWM) preparations were carried out as described previously (Boutin et al., 2014). Briefly, dissected brains were subjected to a 12 min fixation in 4% paraformaldehyde+0.1% Triton X-100, blocked 1 h in PBS+3% BSA, incubated overnight with primary antibodies diluted in PBS, 0.1% Triton X-100 and 3% BSA, and incubated for 1 h with adapted secondary antibodies at room temperature. LWs were dissected further and mounted with Mowiol before imaging using confocal microscope (Leica SP8). Primary antibodies were rabbit anti-ZO1 (1:400; Thermo Fisher Scientific, 61-7300), mouse IgG1 anti-NeuN (1:100; Millipore, Mab 377) and chicken anti-EGFP (1:1000; Aves Labs, GFP-1010). Secondary antibodies were Alexa Fluor 647 goat anti-rabbit (1:800; Thermo Fisher Scientific, A-21244), Alexa Fluor 568 goat anti-mouse IgG2b (1:800; Thermo Fisher Scientific, A-21144), Alexa Fluor 488 goat anti-mouse IgG2b (1:800; Thermo Fisher Scientific, A-21141) and Alexa Fluor 488 donkey anti-chicken (1:500; Jackson ImmunoResearch, 703-545-155).

BrdU (Sigma) was injected intraperitoneally 1 time at 50 mg/g body weight, 2 days after electroporation. Mice were sacrificed 2 h after BrdU injection and stained with a rat anti-BrdU antibody (Bio-Rad). BrdU-positive cells among EGFP cells were then counted in the SVZ at 2 dpe using ImageJ software.

For histological analysis, pups were deeply xylazine/ketamine anaesthetized. Perfusion was performed intracardiacally with a solution of 4% paraformaldehyde in PBS. The brain was dissected out and incubated overnight in the same fixative at 4°C. Sections were cut at 50 µm using a microtome (Microm). Floating sections were first incubated overnight at 4°C with the following primary antibodies: rabbit anti-EGFP (1:500; Life Technologies), mouse anti-NeuN IgG1 (1:100; Merck Millipore, MAB377), rabbit anti-GFAP (1:2000, DAKO, Z0334), mouse anti-S100B (1:1000, Sigma), mouse anti-CRE recombinase (1:4000, Millipore, MAB3120), rat IgG2a anti-BrdU (1:1000, Bio-Rad, OBT0030) before incubation 2 h at room temperature with the corresponding fluorescent labelled secondary antibody. Before mounting, cell nuclei were stained with Hoechst 33258. Optical images were taken either using a fluorescence microscope (Axioplan2, ApoTome system, Zeiss) or a laser confocal scanning microscope (LSM510 or LSM780, Zeiss).

Cytoplasmic fluorescence intensity was measured using ImageJ software on 12-bit z-stack images. Each cell was circled, and mean pixel intensity was extracted and normalized to an arbitrary unit (a.u.) corresponding to 4095 bit depth.

Mice were killed by decapitation and brains were dissected out. SVZ was micro-dissected from 300 μm vibratome brain sections. RNA was then extracted using the miRNAeasy kit (Qiagen) and cDNA were prepared using super-script III reverse transcriptase (Invitrogen). Quantitative PCR was performed on a Bio-Rad CFX system using SYBR GreenER qPCR SuperMix (Invitrogen). β-Actin was used as a reference gene. The oligonucleotides used for amplification were: β-actin FORseq, CTAAGGCCAACCGTGAAAAG; β-actin REVseq, ACCAGAGGCATACAGGGACA; EGFP FORseq, AGAACGGCATCAAGGTGAAC; EGFP REVseq, GAACTCCAGCAGGACCATGT.

Images were taken using a laser confocal scanning microscope (LSM780, Zeiss) with a 63× objective. Lateral and z-axis resolutions were 100 nm and 380 nm, respectively. Spine density and spine length quantifications were performed on z-stack projections using the NeuronJ (Meijering et al., 2004) plug-in of ImageJ software. All dendritic protrusions showing a clearly visible head were quantified as dendritic spines. Imaging and quantifications were all performed blindly to experimental groups.

Statistical analyses were performed using R software and R Commander package (https://CRAN.R-project.org/package=Rcmdr). Data are presented as mean±s.e.m. except for Fig. 1J where data are presented as mean±s.d. Non-parametrical Wilcoxon rank sum test was used to assess mean differences between two groups and Kolgomorov–Smirnov test was used to assess differences between distributions (Fig. 6G,J). Differences were considered statistically significant when P<0.05 (ns, P>0.05; *P≤0.05; **P≤0.01; ***P≤0.001). Levene's test was used to assess equality of variances (Fig. 1J) and variances were considered statistically different if P<0.05.

Imaging was performed using the PiCSL-FBI core facility (IBDM, AMU-Marseille, France) supported by the Agence Nationale de la Recherche through the ‘Investments for the Future’ program (France-BioImaging, ANR-10-INBS-04). We acknowledge the IBDM animal facilities. We thank Jean-Claude Platel for fruitful discussions and critical reading of the manuscript.