Enhancer SINEs Link Pol III to Pol II Transcription in Neurons.

Spatiotemporal regulation of gene expression depends on the cooperation of multiple mechanisms, including the functional interaction of promoters with distally located enhancers. Here, we show that, in cortical neurons, a subset of short interspersed nuclear elements (SINEs) located in the proximity of activity-regulated genes bears features of enhancers. Enhancer SINEs (eSINEs) recruit the Pol III cofactor complex TFIIIC in a stimulus-dependent manner and are transcribed by Pol III in response to neuronal depolarization. Characterization of an eSINE located in proximity to the Fos gene (FosRSINE1) indicated that the FosRSINE1-encoded transcript interacts with Pol II at the Fos promoter and mediates Fos relocation to Pol II factories, providing an unprecedented molecular link between Pol III and Pol II transcription. Strikingly, knockdown of the FosRSINE1 transcript induces defects of both cortical radial migration in vivo and activity-dependent dendritogenesis in vitro, demonstrating that FosRSINE1 acts as a strong enhancer of Fos expression in diverse physiological contexts.

Introduction

All organisms respond to environmental conditions by modifying gene expression in a manner that is strictly regulated both temporally and spatially. Although this adaptive response is of fundamental importance for any cell type, it is particularly relevant to neurons. Failure to rapidly adapt the transcriptional output to ever-changing conditions compromises most brain tasks, including learning and memory formation (Flavell and Greenberg, 2008, Sweatt, 2016, West and Greenberg, 2011).

In eukaryotic cells, three RNA polymerases regulate the transcription of largely non-overlapping sets of genes. RNA polymerase II (Pol II) transcribes protein-coding genes, and Pol I and Pol III transcribe rRNA and tRNA genes, respectively (Roeder, 1996). Pol III also transcribes the 5S rRNA, small RNAs, microRNAs, and RNAs derived from DNA-repetitive elements such as short interspersed nuclear elements (SINEs) (Dieci et al., 2007). Despite the fact that most genes transcribed by Pol III are highly conserved across species, the genomic occupancy of the Pol III complex varies greatly between organisms and between cell types within the organism (Barski et al., 2010, Moqtaderi and Struhl, 2004, Moqtaderi et al., 2010, Raha et al., 2010), implying that it may have additional, cell-specific functions. In human cells, Pol III complexes bind preferentially to genomic regions adjacent to Pol II transcriptional start sites (TSSs) (Moqtaderi et al., 2010, Oler et al., 2010), and expressed tRNA genes are predominantly located in the vicinity of active Pol II promoters (Oler et al., 2010). Thus, Pol III and Pol II transcription may be functionally linked.

Gene expression is regulated by multiple mechanisms, including the interaction of promoters with distal enhancers, which are short genomic elements (typically < 200 bp) often positioned several kilobases away from their target genes (Kolovos et al., 2012). Enhancers function in an orientation-independent manner and are characterized by distinctive features, such as an “open” chromatin and the presence of the histone modifications H3 lysine 4 monomethylation (H3K4me1) and lysine 27 acetylation (H3K27ac) (Heintzman et al., 2009, Kolovos et al., 2012, Zentner et al., 2011). Enhancers are often transcribed into non-coding RNAs known as enhancer RNAs (eRNAs), which stabilize the formation of DNA loops, possibly facilitating the interaction of enhancers with gene promoters (Kolovos et al., 2012). Recently, a distinct class of enhancers has been shown to provide spatiotemporal specificity to gene expression during neuronal development (Frank et al., 2015) and in mature neurons (Kim et al., 2010, Malik et al., 2014, Schaukowitch et al., 2014, Telese et al., 2015). However, the mechanisms by which they regulate transcription remain poorly understood.

We previously showed that, in neurons, a group of SINEs undergoes de novo histone acetylation and recruits the Pol III general transcription factor TFIIIC (Crepaldi et al., 2013). SINEs are an abundant class of retrotransposons often considered as non-functional DNA because of their non-coding, repetitive nature. They are short modular sequences that, similarly to other Pol III-transcribed genes, possess an internal Pol III promoter containing A and B boxes (Muotri et al., 2007, White, 2011). These elements bind the multi-subunit complex TFIIIC, which, in turn, recruits TFIIIB and Pol III. Most SINEs carry mutations that disrupt the promoter region, leaving only a few subtypes with transcriptional potential (Ichiyanagi, 2013). Recent studies indicated that SINEs often acquire novel functions in the host genome in a phenomenon known as exaptation (Huda et al., 2010, Rebollo et al., 2012). In mammalian fibroblasts exposed to heat shock, for example, SINE transcripts inhibited Pol II-dependent transcription by interacting with the Pol II enzyme (Allen et al., 2004, Mariner et al., 2008). Two members of the ancient SINE family Amniota SINE1 (AmnSINE1) were shown to act as enhancers for the fibroblast growth factor 8 (Fgf8) and special AT-rich sequence-binding protein 2 (Satb2) genes in the developing mouse brain (Sasaki et al., 2008, Tashiro et al., 2011). Similarly, in humans, members of the Alu family are enriched for enhancer-like histone modifications in a tissue-specific manner and preferentially engage in long-distance interactions with gene promoters and other Alu elements (Su et al., 2014).

Here, we identified a class of SINEs that function as enhancers of activity-regulated neuronal genes. Genome-wide analyses revealed that enhancer SINEs (eSINEs) bear the epigenetic marks H3K4me1 and H3K27ac, recruit TFIIIC, and are transcribed by Pol III in response to depolarization. We discovered that an eSINE located in the proximity of the activity-dependent gene Fos (which we named FosRSINE1) functions as a Fos enhancer and is transcribed in depolarized neurons. FosRSINE1 eRNA interacts with Pol II at the Fos promoter and is necessary for the relocation of Fos to Pol II transcription factories upon depolarization. Strikingly, this mechanism is required for activity-dependent dendritogenesis and for cortical radial migration and neuronal differentiation of neural progenitors during embryonic development. Together, our findings demonstrate the profound effect of FosRSINE1 on Fos gene expression and reveal a functional link between Pol III and Pol II transcription.

Results

Genome-wide Occupancy of the Pol III Machinery Identifies eSINEs

In response to neuronal depolarization, a group of SINEs located near activity-dependent genes undergoes de novo acetylation at H3K9/K14 (Crepaldi et al., 2013). Because acetylated SINEs possess an internal Pol III promoter (Muotri et al., 2007, White, 2011), we reasoned that they may represent a class of Pol III-transcribed neuronal enhancers. To investigate this hypothesis, we first employed chromatin immunoprecipitation sequencing (ChIP-seq) to map the Pol III machinery genome-wide in resting or depolarized mouse primary cortical neurons. We employed paired-end sequencing to obtain reads that were significantly longer than SINEs and, therefore, easily mappable (Figure S1A). For each experimental condition, we produced over 32 million high-quality mapped reads, providing excellent depth (Figure S1B).

ChIP-seq was performed using antibodies that recognize the catalytic subunit of Pol III (Rpc155) or the DNA binding subunit of TFIIIC (Gtf3c1). As expected, both Rpc155 and Gtf3c1 were highly enriched at tRNA genes (Figures S1C and S1D). Gtf3c1 was widely distributed across the genome, with 66,662 and 58,090 peaks in control (Ctrl) and depolarized conditions, respectively (Table S1). Similar to other cell types (Barski et al., 2010, Oler et al., 2010), only a small fraction of Gtf3c1 binding sites recruited Rpc155 (14.8% and 6.1% in Ctrl and depolarized conditions, respectively; Table S1). Interestingly, a considerable fraction of Rpc155 and Gtf3c1 peaks overlapped at least one SINE (17.4% and 20.0% in resting and stimulated neurons, respectively, for Gtf3c1 and 20.7% and 21.1% for Rpc155; Table S1). Comparative analysis in untreated and depolarized neurons revealed that 7,700 genomic regions showed significant recruitment of Gtf3c1 in response to depolarization, of which 1,151 overlapped with SINEs (Table S2). We named this group of 1,151 SINEs characterized by activity-dependent recruitment of TFIIIC eSINEs. The levels of Gtf3c1 binding to these elements were higher than for randomly selected SINEs (rndSINEs, p < 2.2e−16, Mann-Whitney test for both Ctrl and KCl conditions; Figure 1A). Importantly, Gtf3c1 binding to eSINEs increased in response to stimulation (p < 2.2e−16, Mann-Whitney test), whereas it was unchanged for rndSINEs. Similarly, Pol III binding was higher on eSINEs compared with rndSINEs (p < 2.2e−16, Mann-Whitney test for both Ctrl and KCl conditions; Figure 1B), and recruitment to eSINEs increased in KCl-treated neurons (p = 2.9e−11, Mann-Whitney test). ChIP-seq tracks of representative eSINEs are shown in Figure S1E.

Figure 1

Genome-wide Identification of eSINEs

(A and B) Box and whisker plots of Gtf3c1 (A) and Rpc155 (B) distribution. Shown is ChIP-seq tag density at eSINEs or at randomly selected SINEs (rndSINEs) of comparable size. The solid line denotes the median. Lower and upper box limits indicate the 25th and 75th percentiles, respectively. Whiskers indicate 1.5 times the interquartile distance, measured from the median.

(C) Distribution of Gtf3c1-bound eSINEs under control (dashed line) or depolarized (solid line) conditions relative to the TSS of inducible (left) and housekeeping genes (right).

(D and E) Box and whisker plots summarizing the distribution of Gtf3c1 (D) and Rpc155 (E). Shown is ChIP-seq tag density at 120 eSINEs or at 120 randomly selected SINEs of comparable size. The solid line denotes the median. Lower and upper box limits indicate the 25th and 75th percentiles, respectively. Whiskers indicate 1.5 times the interquartile distance, measured from the median.

(F and G) Binding density profiles of Gtf3c1 and Rpc155 at gene bodies of the indicated sets of genes under control (dashed line) or depolarized (KCl, 50 mM for 45 min, solid line) neurons. Inducible genes (IGs, F), housekeeping genes (HGs, G), a randomly selected set of genes (RGs, G), and silent genes (SGs, G) were analyzed.

(H) Transcription of activity-regulated genes correlates with Gtf3c1 recruitment to proximal eSINEs. mRNA fold induction for each gene is plotted versus Gtf3c1 induction at the closest eSINE. R2 (Pearson correlation coefficient) and p value are shown.

See also Figure S1.

To investigate whether Gtf3c1 recruitment to eSINEs correlated with activity-dependent transcription genome-wide, we analyzed RNA sequencing (RNA-seq) data performed under similar experimental conditions (Kim et al., 2010, Malik et al., 2014). We observed that eSINEs were enriched in the proximity of inducible, but not housekeeping, genes (Figure 1C). Although alternative roles cannot be excluded, this genomic localization suggests that eSINEs may represent a new class of regulatory elements that coordinate activity-dependent transcription in neurons. Next, 120 genes that were induced at least 2-fold in response to depolarization were paired with the closest eSINE (distances ranging from < 1 kb to several hundred kilobases from the TSS; Table S3). This group of eSINEs showed recruitment of the Pol III machinery in response to depolarization (Figures 1D and 1E, right plots; p < 2.2e−16 and p = 2.1e−7 for Gtf3c1 and Rpc155 respectively, Mann-Whitney test) whereas 120 rndSINEs did not (Figures 1D and 1E, left plots).

In accordance with previous studies indicating that Pol II and Pol III co-localize on the proximity of expressed genes, the enrichment of Rpc155 and Gtf3c1 at the TSSs and gene bodies of the paired 120 genes increased in depolarized neurons (Figure 1F; Moqtaderi et al., 2010, Oler et al., 2010, Raha et al., 2010). As expected, no significant recruitment of the Pol III machinery was observed on housekeeping genes (HGs; Figure 1G); lower occupancy of Pol III and Gtf3c1 was observed at randomly selected genes (RGs) and silent genes (SGs), with no changes upon neuronal activity (Figure 1G). A positive correlation (r2 = 0.5, p = 9.7e−8; Figure 1H) was found between the recruitment of Gtf3c1 at eSINEs following depolarization and the transcription of genes for which the closest eSINE was located at a distance of 100 bp or less from the TSS (45 of 120 genes). This subset of genes included Fos, Gadd45b, and other well-characterized activity-dependent genes, such as FosB, JunB, Egr4, Crem, Npas4, Nr4a1, and Nr4a3 (Figure 1H; Table S3).

eSINEs Bear the Epigenetic Hallmarks of Enhancers and Are Transcribed

Epigenetic marks commonly used to identify putative enhancers include H3K27ac and H3K4me1 (Kolovos et al., 2012, Zentner et al., 2011). ChIP-seq experiments performed on Ctrl and depolarized neurons revealed that both histone modifications were higher at the 1,151 eSINEs than at rndSINEs (p < 2.2e−16, Mann-Whitney test for both Ctrl and KCl conditions; Figures 2A and 2B). H3K4me1 and H3K27ac were present at eSINEs in resting neurons, indicating that they may be epigenetically primed prior to neuronal activation. H3K4me1 and H3K27ac were also enriched at the 120 eSINEs paired with activity-dependent genes (p = 2.5e−15 and p = 6.2e−12, respectively, under Ctrl conditions; Figures S2A and S2B). The recruitment of Gtf3c1 to a panel of eSINEs in response to depolarization and the presence of H3K4me1 and H3K27ac were confirmed using ChIP-qPCR (Figure S2C). It should be noted that eSINEs displayed levels of H3K4me1 and H3K27ac similar to previously identified enhancers for the activity-dependent genes Arc and Fos (Kim et al., 2010, Schaukowitch et al., 2014; Figure S2C). In contrast, SINEs located in the proximity of housekeeping (Actb and Hprt) and repressed (Prrt1 and Ndufb5) genes did not recruit Gtf3c1 and were devoid of the enhancer marks H3K4me1 and H3K27ac (Figure S2C).

Figure 2

eSINEs Show the Hallmarks of Enhancers and Are Transcribed

(A and B) Box and whisker plots summarizing the distribution of H3K4me1 (A) and H3K27ac (B). Shown is ChIP-seq tag density at eSINEs or at randomly selected SINEs of comparable size. The solid line denotes the median. Lower and upper box limits indicate the 25th and 75th percentiles, respectively. Whiskers indicate 1.5 times the interquartile distance, measured from the median.

(C–E) Cortical neurons were treated with DNaseI (3 units for 20 min, 3) or vehicle (0) and depolarized with KCl for 45 min. Histograms show DNaseI digestion efficiency at eSINEs (C), non-enhancer SINEs (D), and the TSS of Actb and Fsh (E) expressed as fold over undigested. Data are represented as mean ± SEM (n = 3). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, two-way ANOVA.

(F and G) Expression profile of a panel of eSINEs (F) and non-enhancer SINEs (G) in response to KCl stimulation, assessed by qRT-PCR. Data are represented as mean ± SEM and normalized to 0 min KCl (n = 3). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, one-way ANOVA.

See also Figures S2 and S3.

Enhancers are active genomic regions that, similarly to promoters, are associated with an open, easily accessible state of the chromatin (Boyle et al., 2008, Frank et al., 2015). When the chromatin accessibility of eSINEs was tested using the DNaseI hypersensitivity assay, we observed increased sensitivity to DNaseI digestion after depolarization (Figure 2C), indicating that the chromatin surrounding eSINEs becomes depleted of nucleosomes and primed for transcription. In contrast, SINEs located in the proximity of either housekeeping (Actb and Hprt) or repressed (Prrt1 and Ndufb5) genes were not sensitive to DNaseI (Figure 2D). Notably, the hypersensitivity of eSINEs to DNaseI after neuronal stimulation was comparable with that observed at the TSS of the housekeeping gene Actb (actin beta; Figure 2E), whereas Fsh (follicle-stimulating hormone), a gene that is not expressed in neurons, did not show DNaseI hypersensitivity (Figure 2E).

Enhancers are transcribed in a number of organisms and cell types in response to external stimuli (Kaikkonen et al., 2013, Koch et al., 2011, Wang et al., 2011), and the eRNAs produced are critical for enhancer function (Lam et al., 2013, Li et al., 2013, Melo et al., 2013, Schaukowitch et al., 2014, Telese et al., 2015). We therefore investigated whether eSINEs were expressed in response to neuronal depolarization using qRT-PCR and found that they were transcribed in activated neurons (Figure 2F). Importantly, eSINE RNAs terminated in the proximity of canonical Pol III transcriptional terminators (Figure S3), indicating that they are bona fide Pol III transcripts. As expected, SINEs located proximal either to housekeeping (Actb and Hprt) or repressed (Prrt1 and Ndufb5) genes were not transcribed (Figure 2G). Thus, eSINEs represent a new class of neuronal enhancers that recruit TFIIIC and are transcribed by Pol III.

Functional Characterization of FosRSINE and Gadd45bB1F eSINEs

To gain further insights into the biochemical and functional properties of eSINEs, two elements located in the proximity of Fos and Gadd45b (named FosRSINE1 and Gadd45bB1F) were selected for further analysis. Fos and Gadd45b are activity-dependent genes with well known functions in the nervous system (Greenberg et al., 1986, Ma et al., 2009) that are transcribed in vivo in cortical neurons upon exposure to a novel enriched environment (NEE) and in response to neuronal depolarization in vitro (Crepaldi et al., 2013, Sheng et al., 1990, Sultan et al., 2012). Moreover, both eSINEs are de novo acetylated on H3K9K14 in the somatosensory cortex of mice exposed to NEE (Crepaldi et al., 2013). Despite the fact that SINE sequences are highly repetitive, FosRSINE1 and Gadd45bB1F have distinctive features that allow unequivocal identification (Figure S4A). We first confirmed that FosRSINE1 and Gadd45bB1F recruited Gtf3c1 and Rpc155 in response to depolarization (Figures 3A and 3B) and displayed the H3K4me1 and H3K27ac marks (Figures 3C and 3D). FosRSINE1 and Gadd45bB1F also showed activity-dependent hypersensitivity to DNaseI comparable to Fos and Gadd45b TSSs (Figure 3E). qRT-PCR of FosRSINE1 and Gadd45bB1F RNA at different times after neuronal depolarization revealed that both eSINEs are robustly transcribed (Figure 3F). Importantly, transcription was not detected at genomic regions located immediately upstream (USr) or downstream (DSr) of FosRSINE1 (−170 and +750 bp) and Gadd45bB1F (−400 and +200 bp) (Figure 3F), indicating that the eSINEs are not part of larger transcriptional units. Interestingly, expression of both FosRSINE1 and Gadd45bB1F preceded Fos and Gadd45b transcription (Figure 3F), suggesting that FosRSINE1 and Gadd45bB1F eRNA may control the expression of these genes.

Figure 3

FosRSINE1 and Gadd45bB1F Are Bona Fide eSINEs

(A–D) Mouse cortical neurons were exposed to KCl (50 mM, 45 min) or left untreated and subjected to Gtf3c1 (A), Rpc155 (B), H3K4me1 (C), and H3K27ac (D) ChIP-qPCR. Histograms show ChIP efficiency expressed as percentage of chromatin input; immunoglobulin G (IgG) ChIP was used as a negative control. Data are represented as mean ± SEM (n = 3). ∗p < 0.05, ∗∗∗p < 0.001, Student’s t test.

(E) Cortical neurons were treated with DNaseI (3 units for 20 min, 3) or vehicle (0) and depolarized with KCl. Histograms show DNaseI digestion efficiency expressed as fold over undigested. Data are represented as mean ± SEM (n = 3). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, two-way ANOVA.

(F) KCl-dependent expression levels of FosRSINE1, Gadd45bB1F, Fos, Gadd45b, and the genomic regions located immediately upstream (USr) or downstream (DSr) of the two SINEs were analyzed by qRT-PCR. Fos (top) and Gadd45b (bottom). Data are represented as mean ± SEM (n ≥ 5). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, one-way ANOVA.

(G) Luciferase assay of mouse cortical neurons transfected with the indicated vectors and stimulated with KCl for 6 hr or left untreated. Data are represented as mean ± SEM (n ≥ 3). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, two-way ANOVA.

See also Figure S4.

We next investigated whether FosRSINE1 enhanced activity-dependent transcription using luciferase reporter assays. Neurons were transfected with plasmids expressing the luciferase coding region under the control of three activity-response gene promoters (Fos, Gadd45b, and Arc) and either depolarized or left untreated. Under stimulated conditions, inclusion of FosRSINE1 at the 3′ end of the luciferase cassette markedly increased luciferase expression for all tested promoters (Figure 3G). FosRSINE1 did not enhance luciferase expression when cloned 3′ of either the SV40 promoter of the pGL3-promoter vector or the Gapdh promoter (Figures S4B and S4C). Importantly, in neurons transfected with equimolar amounts of a vector carrying FosRSINE1 and a separate luciferase construct encoding the Fos promoter, FosRSINE1 failed to enhance luciferase expression (Figure S4D), indicating that genomic contiguity is required for enhancer function.

To establish whether transcription of FosRSINE1 is necessary for enhancer function, we generated a mutated version of the eSINE bearing a six-nucleotide mismatch that disrupts Gtf3c1 and Pol III recruitment (Orioli et al., 2012). Mouse neuroblastoma cells (Neuro-2a) were transfected with a vector carrying either wild-type (pBS_RSINE1WT) or mutant FosRSINE1 (pBS_RSINE1mut) and stimulated with forskolin. FosRSINE1 expression levels were significantly lower in cells transfected with the mutant compared with the wild-type vector (Figure S4E). Strikingly, this mutated version of FosRSINE1 failed to enhance activity driven by Fos, Gadd45b, and Arc promoters (Figure 3G). Thus, Pol III binding to FosRSINE1 is required for enhancer activity.

FosRSINE eRNA Is Required for Enhancer Function

To investigate the mechanisms by which eSINEs regulate activity-dependent transcription, the FosRSINE1 genomic sequence was disrupted using a high-fidelity variant of Cas9 (SpCas9-HF1; Kleinstiver et al., 2016). A guide RNA targeting FosRSINE1 was expressed in Neuro-2a cells together with a vector encoding SpCas9-HF1. As a control, we transfected the SpCas9-HF1-encoding plasmid and an empty vector (no guide). Efficient editing of the FosRSINE1 locus was assessed by Surveyor assay (Figure S5A) and confirmed by DNA sequencing of individual clones (Figure S5B). In cells transfected with FosRSINE1 guide RNA, the induction of FosRSINE1 eRNA upon forskolin stimulation was drastically reduced compared with empty vector (Figure 4A), whereas induction of a different eSINE (Dusp1MIRc) was not affected by (Figure 4A). Strikingly, disruption of the FosRSINE1 genomic locus was sufficient to abolish forskolin-dependent transcription of Fos (Figure 4A), indicating that the genomic integrity of FosRSINE1 is necessary for Fos expression upon neuronal stimulation.

Figure 4

FosRSINE1 eRNAs Is Required for Enhancer Functions

(A) Neuro-2a cells were transfected with a vector encoding SpCas9-HF1 protein and either an empty RNA vector (BPK1520, no guide) or an RNA vector containing a small guide RNA targeting FosRSINE1 (guide RNA). Cells were either left untreated or treated with forskolin (50 μM, 45 min). Shown are mean ± SEM of fold induction of FosRSINE1, Fos, and Dusp1MIRc over control no guide (n = 4). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, two-way ANOVA.

(B and C) qRT-PCR of cortical neurons treated with ML60218 (50 μM, 15 min) or DRB (25 μM, 15 min) and exposed to KCl for 45 min. Shown are mean ± SEM of fold induction over DMSO control of FosRSINE1 eRNA (B) and Fos mRNA (C) (n = 3). ∗p < 0.05, ∗∗p < 0.01, two-way ANOVA.

(D) RNA FISH analysis of Fos mRNA in cortical neurons transfected with GFP and either LNACTR or LNARSINE1 and exposed to KCl for 45 min or left untreated. Shown are maximal projections of confocal scans; reconstructed cell edges are shown in green, DAPI-counterstained nuclei are shown in blue, and Fos mRNA ribonucleoparticles are shown in red.

(E) Quantitative analysis of RNA FISH experiments. Shown are means ± SEM of at least 42 cells per condition (n = 3). ∗p < 0.05, ∗∗∗p < 0.001, two-way ANOVA.

(F) Luciferase assay of cortical neurons transfected with the indicated vectors in combination with either LNACTR or LNARSINE1. Data shown as mean ± SEM (n = 4). ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, two-way ANOVA.

See also Figures S5–S7.

Experiments that knockdown eRNAs have recently shown that enhancer transcripts regulate the expression of nearby genes (Lam et al., 2013, Li et al., 2013, Melo et al., 2013, Mousavi et al., 2013, Schaukowitch et al., 2014). To investigate whether FosRSINE1 enhancer activity is mediated by Pol III-dependent eRNA synthesis, we employed several complementary approaches. First, cortical neurons were treated with the specific Pol III inhibitor ML60218 (50 μM, 15 min; Wu et al., 2003) and depolarized. ML60218 treatment decreased transcription of the Pol III-transcribed gene tRNALeu (Figure S6A), indicating efficient inhibition of Pol III. Importantly, inhibition of Pol III markedly reduced transcription of both FosRSINE1 and the Fos gene in response to neuronal activation (Figures 4B and 4C). The effect of ML60218 on Fos transcription was not due to non-specific inhibition of Pol II because the expression levels of the Pol II-transcribed β-Actin were unchanged (Figure S6B). Conversely, treatment of cortical neurons with the specific Pol II inhibitor 5,6-dichloro-1-β-D-ribofuranosylbenzimidazole (DRB) (Yankulov et al., 1995) led to a strong reduction in Fos induction and β-Actin transcription but had no effect on either FosRSINE1 or tRNALeu transcription (Figures 4B and 4C; Figures S6A and S6B). Similar results were obtained for two additional eSINEs (Gadd45bB1F and Dusp1MIRc) and the paired activity-dependent genes (Gadd45b and Dusp1) (Figures S6C–S6F).

Next we used locked nucleic acid (LNA) oligonucleotides to specifically target FosRSINE1 eRNA for degradation. In Neuro-2a cells, an LNA targeting FosRSINE1 eRNA (LNARSINE1) reduced FosRSINE1 levels without affecting Dusp1MIRc expression (Figures S6G and S6H). Cortical neurons were transfected with either LNARSINE1 or a control LNA (LNACTR) and depolarized, and the Fos transcript was measured using RNA fluorescence in situ hybridization (FISH). As expected, in LNACTR-treated neurons depolarization induced the appearance of clearly detectable ribonucleoparticles containing Fos mRNA (Figure 4D, top right, white arrowheads). Remarkably, LNARSINE1 caused a 70% reduction of Fos mRNA expression compared with LNACTR (Figures 4D and 4E). Similarly, an LNA targeting Dusp1MIRc eRNA (LNAMIRc) strongly reduced Dusp1 mRNA levels following KCl stimulation (Figures S7A and S7B). The specificity of LNARSINE1 was tested by performing luciferase assays on neurons expressing vectors encoding luciferase under the control of either the Fos promoter and FosRSINE1 (prom_RSINE1) or the Fos promoter alone (prom). In neurons transfected with prom_RSINE1, LNARSINE1 completely abolished FosRSINE1 enhancer activity but was ineffective in decreasing luciferase expression in neurons transfected with a vector containing Fos promoter only (Figure 4F). Together, these data show that Pol III-dependent transcription of FosRSINE1 is necessary for enhancing Fos expression in depolarized neurons.

FosRSINE1 eRNA Controls Fos Gene Relocation to Transcription Factories

Transcription factories (TFs) are discrete nuclear foci formed by clusters of Pol II and transcription factors that create a permissive environment for transcription (Rieder et al., 2012) where enhancers and promoters may physically interact (Cook, 2010). Because activity-dependent genes relocate to TFs upon depolarization (Crepaldi et al., 2013), we investigated whether eSINE eRNA regulates this process by performing immuno-DNA FISH on cortical neurons transfected with LNAs (Figure 5A). TFs were labeled using a Pol II antibody, and the distance between the center of the DNA-FISH signal and the nearest TF was measured. In neurons transfected with LNACTR, neuronal stimulation decreased the distance between the Fos locus and the nearest TF (Figure 5B) and led to higher co-localization of FISH signals with TFs (Figure 5C). Strikingly, transfection with LNARSINE1 inhibited depolarization-dependent relocation of Fos to TFs (Figures 5A–5C), whereas the nuclear position of Csn2, a gene that is not expressed in neurons, was not affected (Figures 5A–5C). Our findings indicate that FosRSINE1 eRNA promotes Fos relocation to TFs in response to stimulation and may contribute to the spatial organization of chromatin within neuronal nuclei.

Figure 5

FosRSINE1 eRNA Mediates Fos Relocation to Transcription Factories

(A) Representative images of confocal sections of immuno-DNA FISH showing nuclear localization of Fos and Csn2 loci (green) relative to TFs in neurons transfected with GFP and either LNACTR or LNARSINE1 and exposed to KCl for 45 min or left untreated. TFs were detected by Pol II immunostaining (4H8, red); nuclei were stained with DAPI (blue). For each image series, the distance between the center of the FISH signal and the nearest TF is indicated (top right inset). Scale bars, 2 μm (images on the left) and 0.5 μm (magnified images).

(B) Box and whisker plot of the distribution of the distance between Fos or Csn2 loci and the nearest TF. Whiskers denote the 90th and 10th percentiles, box edges denote the 75th and 25th percentiles, solid lines denote medians, and dashed lines denote averages. n = 33–48 foci across 3 biological replicates.

(C) Percentage of co-localization of the Fos and Csn2 loci with TFs, both under basal conditions and in response to KCl, in neurons transfected with either LNACTR or LNARSINE1. ∗p < 0.05, Fisher’s exact test.

FosRSINE1 eRNA Interacts with Pol II to Mediate Transcriptional Initiation and Elongation

It has been proposed that eRNAs may contribute to Pol II loading onto target promoters (Mousavi et al., 2013). Moreover, transcripts originating from SINEs have been shown to bind Pol II in mouse fibroblasts (Allen et al., 2004, Mariner et al., 2008). To investigate whether FosRSINE1 eRNA interacted with Pol II, neurons were subjected to RNA immunoprecipitation (RIP) using a Pol II antibody. We observed that, upon depolarization, the Pol II antibody immunoprecipitated FosRSINE1 eRNA at high levels (Figure 6A), indicating binding. Additional evidence of the interaction was obtained by performing electrophoretic mobility shift assays (EMSAs) using in vitro-transcribed α-32P-labeled FosRSINE1 eRNA. The radioactive probe was loaded on a non-denaturing polyacrylamide gel either alone (probe only) or in the presence of neuronal nuclear proteins. At least three distinct protein complexes were bound to FosRSINE1 eRNA (Figure 6B, lane 3, white arrowheads). Antibodies that recognize the pre-initiating, initiating, or elongating forms of Pol II (8WG16, 4H8, and H5, respectively; Sutherland and Bickmore, 2009) displaced or super-shifted at least one band, confirming that FosRSINE1 eRNA was associated with Pol II and indicating that the interaction can occur at different stages of transcription (Figure 6B, lanes 4–6).

Figure 6

FosRSINE1 eRNA Interacts with Pol II and Affects Pol II Initiation and Elongation

(A) RNA immunoprecipitation (RIP) of cortical neuron nuclear extracts incubated with Pol II 8WG16 antibody and subjected to qRT-PCR. RIP efficiency was expressed as increment over control IgG. Data are represented as mean ± SEM (n = 3). ∗∗p < 0.005, Student’s t test.

(B) EMSA of FosRSINE1 eRNA. A FosRSINE1 radiolabeled probe was incubated with cortical neuron nuclear extracts. An excess of either sense (S) or antisense (A) unlabeled probe was used as control. Three bands were detected (lane 3, white arrowheads) that were super-shifted or displaced by Pol II antibodies (lanes 4–6).

(C–F) Neuro-2a cells were transfected with the indicated LNAs, stimulated with forskolin (Fsk), and subjected to ChIP using 8WG16 (C), 4H8 (D and E), and H5 (F) antibodies. Histograms show ChIP efficiency at the Fos promoter and exon 4, expressed as fold over control. Data are shown as mean ± SEM (n = 3). ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, two-way ANOVA.

We next asked whether FosRSINE1 eRNA influenced Pol II recruitment to the endogenous Fos promoter using Pol II ChIP. In untransfected or LNACTR-transfected Neuro-2a cells, Pol II binding to the Fos gene significantly increased upon stimulation at both the promoter (8WG16 and 4H8 antibodies) and at the coding region (4H8 and H5 antibodies) (Figures 6C–6F). In cells transfected with LNARSINE1, stimulus-dependent binding of pre-initiating Pol II to the Fos promoter was conserved (Figure 6C), whereas recruitment of both the initiating and elongating forms of Pol II at the Fos promoter and coding region was prevented (Figures 6D–6F). These results indicate that, in the absence of the Fos RSINE1 eRNA, Pol II is unable to progress into active transcription and provide the first evidence of a functional link between a Pol III-dependent transcript and Pol II.

FosRSINE1 eRNA Regulates Neuronal Differentiation and Cortical Development

The biological significance of FosRSINE1-dependent regulation of Fos transcription was investigated by studying dendritogenesis of primary cortical neurons. Dendritic growth in response to external cues requires the transcription of activity-regulated genes (West and Greenberg, 2011), many of which are regulated by Fos (Greenberg et al., 1986, Malik et al., 2014). Thus, early and robust activation of Fos may be necessary for dendritic arborization. Neurons were transfected with a GFP-expressing vector and either LNACTR or LNARSINE1 and maintained under basal or depolarized conditions for 48 hr. As expected, neurons transfected with LNACTR showed an increase in both dendritic length and complexity (Figures 7A–7C). Strikingly, activity-dependent dendritogenesis was completely abrogated in neurons transfected with LNARSINE1 (Figures 7A–7C).

Figure 7

FosRSINE1 Enhances Fos Expression in Diverse Physiological Contexts

(A) Representative images of cortical neurons transfected with a GFP-expressing vector in combination with either LNACTR or LNARSINE1 and maintained under basal or depolarizing (KCl, 50 mM) conditions for 48 hr, followed by GFP immunostaining. Scale bar, 50 μm.

(B) Quantification of the total length of the dendritic processes of 29–31 neurons (n = 3). Shown are means ± SEM. ∗∗p < 0.01, two-way ANOVA.

(C) Sholl analysis of neurons analyzed in (B). For each distance point, the average number of intersections ± SEM is shown. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗∗p < 0.0001, two-way ANOVA.

(D) qRT-PCR of mouse neural progenitor cells (NPCs) and post mitotic-neurons (PMNs). Data show mean ± SEM of fold induction of Hprt SINE RNA, FosRSINE1 eRNA, and Fos mRNA (n = 6) normalized to expression levels in NPCs. ∗∗p < 0.005, ∗∗∗∗p < 0.0001, paired t test.

(E) E13.5 embryos electroporated in utero with LNAs and a CAG-Fos-expressing construct as indicated and analyzed at E15.5. Shown are representative images of coronal sections immunolabeled for GFP (green) and Tbr1 (red). Scale bar, 100 μm.

(F) Quantification of the distribution of cells electroporated as in (E) between the ventricular-subventricular zone (VZ-SVZ), intermediate zone (IZ), and cortical plate (CP). Data are from 3 independent experiments and 16–22 embryos per condition.

(G) Quantification of neurons expressing Tbr1. Data are from 3 independent experiments and 16–22 embryos per condition.

(H) Coronal sections of mouse embryonic brains immunolabeled with GFP (green) and Ki67 (red) antibodies. Scale bar, 100 μm.

(I) Quantification of cells expressing Ki67. Data are from 3 independent experiments and 10–21 embryos per condition.

(J) Coronal sections of mouse brains immunolabeled with GFP (green) or Sox2 (red) antibodies. Scale bar, 100 μm.

(K) Histograms showing quantification of the percentage of cells expressing Sox2. Data are from 3 independent experiments and 10–21 embryos per condition.

Data are represented as mean ± SEM; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, two-way ANOVA with Tukey’s post test (F) and one-way ANOVA with Tukey’s post test (G, I, and K).

Because Fos has been shown to regulate mouse brain development (Velazquez et al., 2015), we reasoned that FosRSINE1 may also enhance Fos expression in this physiological context. First, we assessed whether FosRSINE1 was expressed in embryonic cortical neurons. Mouse neural progenitor cells (NPCs) were dissected at embryonic day 12.5 (E12.5) and differentiated in vitro into post-mitotic neurons (PMNs) (Nitarska et al., 2016). A remarkable increase in both FosRSINE1 and Fos expression levels was observed following differentiation of NPCs to PMNs (Figure 7D). Conversely, transcription of the non-enhancer SINE HprtSINE was unchanged (Figure 7D). Next, we asked whether FosRSINE1 regulated the expression of Fos during neuronal development in vivo by performing in utero electroporation experiments. LNACTR or LNARSINE1 were electroporated together with a GFP expression plasmid into E13.5 mouse brains, and, after 2 days, neural migration and expression of differentiation markers were assessed. As observed previously (Nitarska et al., 2016) after 48 hr, 33% of the progeny of neural progenitors electroporated with LNACTR in the ventricular zone (VZ) had reached the cortical plate (CP) (Figures 7E and 7F). Inhibition of FosRSINE1 expression resulted in an accumulation of cells within the VZ and a reduction of neurons migrating toward the CP (Figures 7E and 7F). Neural progenitors electroporated with LNARSINE1 failed to differentiate and express T-Box Brain 1 (Tbr1) (Figures 7E and 7G), a transcription factor found in early-born neurons (Hevner et al., 2001). Consistent with the phenotype observed in Fos mice (Velazquez et al., 2015), depletion of FosRSINE1 eRNA also resulted in increased proliferation of neural progenitors, as indicated by the elevated number of Ki67- and Sox2-positive cells (Figures 7H–7K). Co-electroporation of neuronal progenitors with a vector expressing Fos under the control of a cytomegalovirus (CMV) early enhancer/chicken β-actin (CAG) promoter fully rescued the defects induced by the inhibition of FosRSINE1 (Figures 7E–7K), ruling out potential off-target effects of LNARSINE1.

Thus, during the early stages of cortical development, FosRSINE1 regulates radial migration and neuronal differentiation of NPCs, whereas at the onset of neuronal maturation, FosRSINE1 eRNA is required for activity-dependent dendritogenesis.

Discussion

eSINEs Are a Class of Neuronal Enhancers

Repetitive elements are extremely abundant in the mammalian genome, however their function remains largely obscure. In recent years, SINE transcripts have been linked to protein synthesis, mRNA turnover, and RNA editing in both human and mouse cell lines (Athanasiadis et al., 2004, Kim et al., 2004, Levanon et al., 2004, Ohnishi et al., 2012). SINEs may also have regulatory functions independent of their transcription, providing for example, sites of alternative splicing (Levanon et al., 2004), generating new TSSs and polyadenylation signals (Chen et al., 2009, Faulkner et al., 2009), and offering an accessible chromatin environment for transcription factor binding (Gomez et al., 2016). SINEs also act as insulators, transcriptional repressors, or developmentally regulated enhancers (Allen et al., 2004, Lunyak et al., 2007, Nakanishi et al., 2012, Tashiro et al., 2011). We have identified a new class of enhancer SINEs that are transcribed by Pol III and likely regulate the expression of many Pol II-dependent genes in response to neuronal depolarization. eSINEs share biochemical and functional properties with previously identified neuronal enhancers, including the epigenetic marks H3K4me1 and H3K27ac (Figures 2A and 2B, and 3C and 3D) and hypersensitivity to DNaseI (Figures 2C and 3E), which is associated with open, nucleosome-depleted chromatin. Importantly, although we were not able to perform a genome-wide analysis of eSINE transcripts because of technical issues related to their repetitive nature, we showed that many eSINEs are transcribed in response to neuronal depolarization (Figures 2F and 3F) and that transcription is critical for target gene induction (Figure 4; Figure S7).

A Functional Link between Pol III and Pol II Transcription

The Pol II and Pol III transcriptional machineries have been traditionally considered independent entities that regulate the expression of largely non-overlapping genes. However, recent studies have demonstrated that they often co-localize on genomic loci in both murine and human cells and that co-localization of the two machineries correlates with transcription (Barski et al., 2010, Moqtaderi et al., 2010, Oler et al., 2010, Raha et al., 2010). Whether the two machineries are functionally linked and the mechanisms of such regulation is still unknown.

Transcription of enhancer elements was first observed more than 20 years ago (Collis et al., 1990), and, more recently, RNA synthesis has been demonstrated at active enhancers of most cell types across many species (Kaikkonen et al., 2013, Koch et al., 2011, Wang et al., 2011). All neuronal enhancers identified to date are transcribed by Pol II and bind transcription factors that are also enriched on the promoter of their target genes (Kim et al., 2010, Malik et al., 2014, Telese et al., 2015). It has been postulated that recruitment of transcription factors to both enhancers and promoters helps to create a molecular bridge that physically connects the two elements, conferring specificity to enhancer-promoter pairings (Lam et al., 2014). In contrast, eSINEs are transcribed by Pol III and bind the general transcription factor complex TFIIIC (Figures 1A and 1B and 3A and 3B). Interestingly, eSINEs do not possess consensus sequences for any transcription factor associated with Pol II-transcribed neuronal enhancers (L.C. and A.R., unpublished data). Because TFIIIC and Pol III are recruited to both eSINEs and TSSs of activity-regulated genes (Figures 1A, 1B, and 1F), they may play a role in pairing eSINEs with their gene promoters in response to depolarization.

Binding of the Pol III machinery to eSINEs correlates with transcription of the closest genes in response to neuronal activation (Figure 1H), suggesting that contiguity of eSINEs with their target genes is necessary for enhancer functions. Consistent with this finding, FosRSINE1 eRNA failed to enhance luciferase expression driven by the minimal Fos promoter when encoded by a separate vector (Figure S4D). However, the higher-order structure of chromatin within the nucleus enables enhancers to contact promoters even when they are located far apart on the chromosome. We discovered that FosRSINE1 eRNA binds Pol II (Figures 6A and 6B) and is necessary for relocation of the Fos genomic region to transcription factories (Figure 5), perhaps inducing changes of chromatin structure that favor enhancer-promoter interactions. The finding that both Pol III and TFIIIC are recruited to Pol II-regulated genes (Figure 1F) further indicates that the two machineries may come into close proximity within transcription factories.

FosRSINE1 Enhances Fos Expression in Diverse Physiological Contexts

The biological relevance of FosRSINE1 was demonstrated by the fact that FosRSINE1 eRNA is required for activity-dependent dendritic growth and branching (Figures 7A–7C). We previously showed that silencing of TFIIIC in cortical neurons increased dendritic length (Crepaldi et al., 2013). These findings may be explained by the fact that more than 80% of Gtf3c1 binding identified by ChIP-seq is located in genomic regions other than eSINEs and independently of Pol III (< 20% of co-localization on SINEs and < 15% genome-wide). Therefore, global silencing of TFIIIC is likely to affect genome organization and transcription in ways distinct from the inhibition of FosRSINE1 and Fos expression alone. Strikingly, the role of FosRSINE1 was not limited to regulating Fos expression in response to neuronal activity but extended to early brain development. Both Fos and FosRSINE1 levels increased during neuronal differentiation in vitro (Figure 7D), and FosRSINE1 eRNA was required for embryonic cortical development in vivo (Figures 7E–7K). Thus, in very different physiological contexts, Pol II-dependent transcription of the Fos gene is regulated by the Pol III-mediated transcription of its nearby eSINE.

In 1969 Britten and Davidson, formulated the hypothesis that transposable elements such as SINEs participate in gene-regulatory networks, contributing to speciation novelty (Britten and Davidson, 1969). More recently, repetitive elements have been shown to represent key determinants of macroevolution and clade-specific phenotypes (Tashiro et al., 2011). The considerable number of eSINEs identified in neurons suggests that they may have an unexpected and complex role that goes beyond controlling the expression of isolated genes and extend to globally linking Pol III with Pol II transcription.

Experimental Procedures

CRISPR-Cas9

Single guide RNAs were designed toward the Fos eSINE using http://crispr.mit.edu/. The backbones for the vectors expressing the guide RNA (U6-BsmBIcassette-Sp-sgRNA, Addgene 65777) and the high-fidelity Cas9 (CMV-T7-humanSpCas9-HF1(N497A, R661A, Q695A, Q926A)-NLS-3xFLAG, Addgene 72247; Kleinstiver et al., 2016) were obtained from Addgene. Annealed oligos composing the guide RNAs were cloned into the BsmBI site of U6-BsmBIcassette-Sp-sgRNA. Neuro-2a cells were plated into a 12-well plate and transfected 48 hr later with 250 ng U6-BsmBIcassette-Sp-sgRNA (either empty vector or containing FosRSINE guide RNA) and 750 ng CMV-T7-humanSpCas9- HF1 using Lipofectamine 2000. The medium was changed 2–3 hr after transfection, and cells were harvested 48 hr later. For RNA analysis, cells were serum-starved 16 hr before adding forskolin (50 μM, 45 min). The guide RNA sequence used was GGTCATGCACTTGAGGTCATGGG (the last 3 letters are protospacer adjacent motif [PAM]).

Immuno-DNA FISH

Immuno-DNA FISH experiments were performed as described previously (Crepaldi et al., 2013) with some modifications. Briefly, cells were fixed for 10 min in 3% paraformaldehyde (PFA) in PBS, followed by permeabilization for 10 min in 0.5% Triton X-100 in PBS. After blocking with PBS+ (PBS plus 0.1% casein, 1% BSA, 0.2% fish skin gelatin) for 1 hr, coverslips were incubated overnight with Pol II-Ser5p (1:500, 4H8, Millipore 06-623) and GFP (1:2,000, Abcam ab13970) antibodies in PBS+. For detection, coverslips were incubated with anti-mouse Alexa Fluor 568 (1:1,000, Life Technologies) and anti-chicken Alexa Fluor 488 (1:1,000, Life Technologies) for 1 hr in PBS+. Post-fixation in 3% PFA in PBS (10 min) was followed by permeabilization in 0.1 M HCl and 0.7% Triton X-100 (10 min on ice) and by denaturation in 70% formamide in 2× saline sodium citrate (SSC) (80°C, 30 min). FISH hybridization with digoxigenin-labeled probes was carried out overnight at 42°C. The probes (bacterial artificial chromosome/clones [BACs] Fos RP24-233K8 and Csn2 RP23-110B6) were labeled with digoxigenin-deoxyuridine triphosphate (dUTP) using a nick translation kit (Roche), denatured (95°C, 5 min), and pre-annealed (37°C, 45 min) with Cot-1 DNA in hybridization buffer (50% formamide, 20% dextran sulfate, 2× SSC, and 1 mg/mL BSA) immediately before hybridization. FISH signals were amplified using sheep anti-digoxigenin fluorescein Fab fragments (1:50, Roche 11207741910) and fluorescein rabbit anti-sheep antibodies (1:100, Vector Laboratories FI-6000); DNA was counterstained with DAPI. Confocal images of neuronal nuclei were acquired using a Leica SPE3 confocal microscope. The co-localization threshold was set to 225 nm, corresponding to the distance at which the two smallest detectable objects overlap.

In Utero Electroporation

E13.5 pregnant mice were anesthetized with isoflurane in oxygen carrier (Abbot Laboratories), and the uterine horns were exposed through a small incision in the ventral peritoneum. Plasmid DNA solution (0.5–1.5 μg/μL), prepared using the EndoFree plasmid purification kit (QIAGEN), was mixed with 50 μM antisense LNA GapmeR (in vivo-ready, Exiqon) and 0.05% Fast Green (Sigma) and injected through the uterine wall into the lateral ventricles of the embryos using pulled borosilicate needles and a Femtojet microinjector (Eppendorf). Five electrical pulses were applied at 35 V (50-ms duration) across the uterine wall at 950-ms intervals using 5-mm platinum Tweezertrodes (Harvard Apparatus) and an ECM-830 BTX square wave electroporator (Harvard Apparatus). The uterine horns were replaced in the abdominal cavity and the abdomen wall, and the skin was sutured. 48 hr after surgery, pregnant mice were sacrificed, and embryos were subjected to immunofluorescence to assess radial migration and expression of proliferation and laminar markers.

Radial Neural Migration Analysis

Embryos were electroporated in utero with the indicated GFP vectors, and analysis of radial migration was performed as described using ImageJ and an Excel macro (Nitarska et al., 2016). Images were acquired on an SP8 confocal microscope with Leica Application Suite Advanced Fluorescence (LAS AF) software using a 20× objective at 1,024 × 1,024 pixel resolution. Confocal images were run through a band-pass filter to segment and isolate cell-sized shapes, thresholded and segmented into 10 radial regions between the ventricle and the pial surface. Individual cell position along the radial axis was recorded and imported into Excel along with the coordinates of top (pial) and bottom (ventricle) boundaries obtained using ImageJ’s Path Writer plugin. The distance and percentage of migrating cells in each area were calculated using an Excel macro.

Author Contributions

C.P., L.C., and A.R. conceived the project and designed the experiments. C.P., L.C., E.B., J.N., S.M.F., and A.C. performed the biochemical, cellular, and molecular biology experiments. L.C. performed the bioinformatics analyses. C.P., L.C., E.B., and A.R. wrote the paper, and all authors commented on the manuscript.