Regulation of Emx2 Expression by Antisense Transcripts in Murine Cortico-Cerebral Precursors

Background Emx2 encodes for a transcription factor expressed in the embryonic intermediate mesoderm and central nervous system (CNS). It is implicated in several aspects of cerebral cortex development, including morphogenetic field specification, arealization, precursor proliferation and lamination. Four Emx2-associated antisense transcripts have been found in the urogenital system; one of them, Emx2OS, has been also detected in the adult brain. Until now, however, nothing is known about expression and function of Emx2OS in the developing CNS. Methodology/Principal Findings By quantitative RT-PCR and in situ hybridization, we reconstructed the Emx2OS expression profile in the embryonic CNS, paying special attention to the developing cerebral cortex. Emx2OS was observed in a number of CNS structures expressing also Emx2. Within the cortex, Emx2OS was detectable in periventricular precursors, expressing the sense transcript, and peaked in newly born post-mitotic neurons not expressing such transcript. By integrating lentiviral gene delivery, RNAi, TetON technology, morpholino-mediated gene knock-down, drug-induced perturbation of gene expression, and quantitative RT-PCR, we addressed possible roles of Ex2 antisense RNA in Emx2 regulation, in primary CNS precursor cultures. We found that, in both cortical precursors and their neuronal progenies, Emx2 antisense RNA contributes to post-transcriptional down-regulation of its sense partner, possibly by a Dicer-promoted mechanism. The same RNA, when delivered to rhombo-spinal precursors, stimulates ectopic expression of Emx2, whereas Emx2 knock-out dramatically impairs Emx2OS transcription. This suggests that, within the developing CNS, a reciprocal Emx2/Emx2OS regulatory loop may normally sustain transcription at the Emx2 locus. Conclusions/Significance This study shows that antisense transcripts may contribute to developmental regulation of a key transcription factor gene implicated in CNS patterning, possibly by complex and multilevel mechanisms. The activation of Emx2 by a short antisense transcript may be a prototype of a method for overexpressing single specific genes, without introducing additional copies of them into the genome.


Introduction
Emx2 is a transcription factor gene expressed in the developing urogenital and central nervous systems (CNS) [1,3], crucial for proper morphogenesis of these structures [4,6]. It is implicated in dorso-ventral specification of the rostral neural tube, conferring cortico-cerebral identity to precursors in the dorsal telencephalic vesicle and repressing the activation of striatal morphogenetic programs [7]. Expressed by cortical periventricular precursors along a caudal/medial high -to-rostral/lateral low gradient, it further promotes the specification of hippocampus and occipital cortex, while antagonizing rostral-lateral areal programs [8][9][10]. Within the embryonic brain, it stimulates neural precursors' self-renewal, at expenses of neuronal differentiation; these effects are apparently reverted in post-natal neural stem cells [11][12][13][14]. Finally, Emx2 is crucial to proper inside-out layering of the neocortical primordium. In its absence, pioneer Cajal-Retzius cells orchestrating such process are severely reduced and the neocortical lamination profile is deeply distorted, in a reeler-like way [15,16].
Four main Emx2 antisense transcripts of 6.0, 5.0, 2,3 and 1,8 kb have been detected in the murine urogenital system by Northern blotting and a cDNA corresponding to the second of them, Emx2OS, has been isolated. Emx2OS expression has been studied by in situ hybridisation in the uro-genital system: like Emx2-mRNA, Emx2OS-ncRNA is abundant in normal postmenopausal endometrium, reduced in premenopausal endometrium, absent/ reduced in many primary endometrial tumors [17]. Moreover, its human counterpart has been specifically detected in the adult brain, by Northern blot [17]. However, the Emx2OS expression profile has not been studied in the developing CNS and little is known about possible implication of Emx2OS in fine regulation of Emx2.
Based on various relationships occurring among AS expression levels and those of their sense partners, it has been predicted that the former ones might differentially promote or antagonize the expression of such partners, depending on the gene and on the context. Such prediction has been substantially confirmed in a variety of functional studies, indicating that: (1) AS transcription is not simple transcriptional noise or epiphenomenon of S transcription [36], but often results a prerequisite for proper regulation of ''coding'' sense genes; (2) molecular mechanisms by which AS transcripts act are extremely complex and largely diversified [39,40].
In some cases, poor sequence conservation but similar exon/ intron organisation characterizing antisense transcripts from different species suggests that transcription per se, and not the AS RNA molecule, may be crucial to the function. That has been demonstrated in the case of the silent human provirus HERV-K18, where AS transcription promotes S transcription [41], as well as in a variety of other cases, where AS transcription conversely inhibits S transcription, because of competition between the two transcriptional machineries for shared cofactors, or due to collision between them [42].
However, the AS transcript may also work as such, acting at distinctive regulatory levels. First, it may regulate the epigenetic state of chromatine, modulating the methylation state of DNA and/or histones [27,43,44]. Second, it may facilitate recruitment of transcription factors to enhancers impinging on the partner sense gene, so promoting its transcription. This is the case of the Evf2 ncRNA, cooperating with the homeoprotein Dlx2 in stimulating transcription of its sense partner Dlx6 [25]. Third, the AS ncRNA may modulate the splicing of its partner sense pre-mRNA, as reported for the TRa2 and Zeb2 loci [45,46]. Fourth, AS transcripts may regulate the half-life of their sense partners. Pairing of retrotransposon sense and antisense transcripts paves the way to Dicer-dependent cutting of resulting dsRNAs, followed by siRNA-instructed silencing of sense transcripts, from the same or paralogous loci [47][48][49]. Fifth, antisense RNAs may modulate translation [50]. Last, they may be necessary for proper activity of their sense partners, as shown for the Xist/Tsix pair [51].
In this study, we investigated the expression pattern of Emx2OS-ncRNA in the developing CNS, by quantitative reverse transcription polymerase chain reaction (qRT-PCR) and in situ hybridisation (ISH). Then, by integrating lentiviral cDNA delivery, druginduced perturbation of patterning pathways and gain-or loss-offunction (GOF and LOF, respectively) designs, we addressed possible roles of this transcript on regulation of its partner gene, Emx2, on primary CNS cultures as well as on cell lines. We found that Emx2 antisense transcripts may both stimulate and refine Emx2 expression. Moreover, such effects seem to be exerted at two distinct regulatory levels, transcriptional and post-transcriptional.
These results, beyond their contribution to the comprehension of mechanisms regulating cortico-cerebral development, suggest a possible general exploitation of AS-based methods, as a tool for artificial triggering of endogenous gene expression.
Then, we studied the Emx2OS-ncRNA expression pattern in the developing central nervous system (CNS), by non-radioactive in situ hybridisation (Fig. 1A). At E12.5, within the CNS, the transcript was detectable in the telencephalon (Fig. 3A,B), including both pallium and basal ganglia (Fig. 4A), in the mammillary recess of the hypothalamus (Fig. 3A, arrowhead) and in the midbrain, including tectum and tegmentum (Fig. 3C). No signal was detectable within the rhombo-spinal domain (Fig. 3A,D). Emx2OS was also expressed by the nasal pits (Fig. 3E), the otic vesicle (Fig. 3G), the choroid plexus of II and IV ventricle (Fig. 3A,D), as well as by two clusters of head mesenchyme cells, in the snout region (Fig. 3A, asterisk) and in the surroundings of the hypothalamic optic recess (Fig. 3F, arrowheads).
Focussing our attention on the developing cerebral cortex, we found Emx2OS transcripts within periventricular proliferative layers, from E12.5 to E18.5 ( Fig. 4A-F), as well as in the cortical plate (CP), especially in its more superficial part (Fig. 4E,F). Conversely, no signal was detectable within the preplate (PPL) and its derivatives, marginal zone (MZ) and subplate (SP) (Fig. 4D-F). In particular, no Emx2OS transcripts could be found within Cajal-Retzius (CR) cells, aligned beneath the pia mater and specifically expressing Reln mRNA, in both neocortex and hippocampus ( Fig. 4G-L, empty and solid arrowheads).

Emx2OS Antagonizes Emx2 Expression
The mutually exclusive distribution of Emx2-mRNA and Emx2OS-ncRNA among cortico-cerebral neurons (the former expressed by CR neurons, the latter by CP neurons) suggested to us that reciprocal down-regulation between them might occur and that, in particular, Emx2OS-ncRNA might be implicated in repression of Emx2mRNA.
To assess if upregulation of Emx2 induced by miR-aEmx2OS-774 took place at transcriptional or post-transcriptional level, we then measured Emx2 pre-mRNA levels, following infection of neurospheres with miR-aEmx2OS-774 or negative control lentiviruses. For sake of sensitivity and specificity, measurements were done by two-step quantitative RT-PCR, using two sets of nested primers annealing within the first Emx2 intron (Fig. 1A). Interestingly, no change in Emx2 pre-mRNA levels was found (Fig. 5B, right), suggesting that Emx2OS-dependent regulation of Emx2 took place at post-transcriptional level.
Remarkably, in previous sets of experiments, neural cultures were performed in DMEM/F12/N2 medium, containing the standard growth factors (GFs) mix which promotes the intermitotic/stem state [53]. As during embryonic development Emx2-mRNA disappears in post-mitotic neurons, where Emx2OS-ncRNA is transiently upregulated, the question arises: is Emx2OS-ncRNA able to down-regulate Emx2-mRNA also in nascent neurons? To address this issue, the Emx2OS-LOF tests were repeated culturing the infected cells in the presence of 5% serum, which stimulates neuronal differentiation, in place of GFs. Moreover, in this new experiment an aliquot of miR-C-infected cells was kept under GFs, as a control. Consistently with our expectations, under 5% serum plus miR-C lentivirus, neuronal differentiation was dramatically stimulated, as confirmed by massive activation of neuron-specific b-tubulin (Fig. 5C, top), Emx2-mRNA level decreased, by about 30% (Fig. 5C, bottomright), and Emx2OS-ncRNA was only slightly downregulated (Fig. 5C, bottom-left). Remarkably, knock-down of Emx2OS induced by miR-aEmx2OS-774 rescued the Emx2-mRNA decrease to large extent (Fig. 5C, bottom-right), suggesting that    Emx2OS-ncRNA substantially contributes to Emx2 down-regulation at the time when CP neurons are born.
Then, to confirm this model, we verified it by a complementary GOF approach, i.e. by delivering antisense Emx2 cDNA to embryo-derived neurospheres, via lentiviral vectors, and scoring consequences of that.
Unfortunately, the full-length 5.0 kb Emx2OS cDNA harbors several canonical polyadenylation signals, which make it hardly suitable for our lentiviral expression system. So, in place of it, we overexpressed its 59-most 179 bp fragment, coinciding with the Emx2-mRNA/Emx2OS-ncRNA overlapping region minus a lowcomplexity 74 bp polypyrimidine tract. Assuming -in fact -that Emx2OS-ncRNA works by interacting with Emx2-mRNA via Watson&Crick base-pairing, this short fragment could recapitulate key regulative properties of the full-length Emx2OS-ncRNA. Moreover, in order to assay both consequences of antisense overexpression and their reversibility, we performed Emx2OS-GOF manipulations by a dual, TetON-based, lentiviral delivery system, triggered by doxycycline. Neurospheres were infected by a ''driver virus'', harboring a constitutively expressed rtTA2 S -M2 transactivator gene, plus an ''expressor virus'', harboring the cDNA sequence subject of investigation (Fig. 1C).
As for molecular mechanisms mediating Emx2OS-dependent Emx2-mRNA down-regulation, we hypothesized that the antisense ncRNA might destabilize the sense mRNA, forming a double strand with it and so preparing its degradation by double-strand ribonucleases. Among these enzymes, a reasonable candidate might be Dicer1, involved in dsRNA-mediated destabilization of retrotransposon transcripts [47][48][49] and crucial to generation and survival of cortico-cerebral neurons [54]. To test this hypothesis, we decided to repeat the GOF experiments described above in cells alternatively provided with Dicer1 activity or deprived of it, by morpholino technology [55,56]. To get the best Dicer1 knock-down, NIH/3T3 cells, among the easiest to transfect by established morpholino reagents, were selected as a substrate. The stronger effect elicited by OS1-179(+) in the presence of normal levels of Dicer1 and the more pronounced effect elicited by a-Dicer1 when associated to OS1-179(+) suggested that a specific functional interaction could occur between Dicer1 and the antisense transcript. Remarkably, two-ways ANOVA re-analysis of Emx2 expression data confirmed this suspect, with p,0.025 (Fig. 6B).

Emx2OS Transcripts Promote Both Emx2-mRNA and Emx2OS-ncRNA Expression
The specific co-expression of Emx2OS-ncRNA and Emx2-mRNA by neural precursors belonging to defined regions of the developing CNS suggested that antisense transcripts, in addition to trigger destabilization of Emx2-mRNA, might be also basically implicated in promoting its transcription. This hypothesis was tested by overexpressing Emx2 antisense sequences in neural precursors from the rhombo-spinal tract, a CNS segment which normally does not express Emx2 [1,2] (Fig. 1C). As basal expression levels of Emx2 in rhombo-spinal precursors are very low (near the experimental background) and this might make it difficult to estimate the relative change of transcript level elicited by OS1-179(+), the following device was adopted. The experiment was alternatively run under standard neurosphere medium [15], or upon addition of a specific lithium/cyclopamine drug mix, which, stabilizing the transcription factor (TF) beta-catenin [57] and preventing inhibition of the TF Gli3 [58], promotes Emx2 transcription [59][60][61], so as to increase the baseline of the assay and ameliorate its signal-to-noise ratio (Fig. 7A).
As expected, the Emx2 expression level in rhombo-spinal cells, infected with control virus and kept in the absence of drugs, resulted far lower than in telencephalic cells exposed to lithium/cyclopamine. The Emx2 level slightly rose, upon treatment of rhombospinal cells by either the OS1-179(+) virus or the lithium/ cyclopamine mix (about 2 and 4 times, with p,0.01 and p,0.001, respectively). Interestingly, simultaneous exposure of rhombo-spinal cells to both OS1-179(+) virus and lithium/ cyclopamine elicited a much more dramatic up-regulation effect (almost 30 times, with p%0.001), so confirming the effectiveness of the OS1-179(+) virus and further suggesting a powerful synergy between such virus and the drug cocktail (p,0.001) (Fig. 7A, upper  histograms). In other words, as hypothesized, Emx2OS transcripts may activate Emx2 expression and the amplitude of this phenomenon is much more prominent, upon appropriate pharmacological modulation of the nuclear TF milieu impinging on Emx2 regulation [57,59]. Noticeably, changes of Emx2-mRNA levels elicited by pharmacological and genetic manipulations described above were paralleled by strikingly similar changes of Emx2OS-ncRNA (Fig. 7A, lower histograms). This indicates that both reasonably originated from concerted stimulation of sense and antisense transcription at the Emx2 locus. This also suggests that the upregulation of Emx2-mRNA triggered by artificial OS1-179(+) administration might be further sustained by endogenous antisense transcripts from the same locus, induced by OS1-179(+) itself (Fig. 7A).
As the previous experiment was done by keeping the OS1-179(+) transgene chronically on, it was poorly informative about the kinetics and, in particular, the reversibility of the Emx2 activation triggered by antisense transcripts. To address this issue, we repeated the OS1-179(+) overexpression test over 120 hours, limiting neural precursors exposure to doxycycline to the first 72 hours (Fig. 7B).
Results of this test were as follows. Withdrawal of doxycycline at 72 hours allowed full shutting off of OS1-179(+)-ncRNA at 120 hours, as witnessed by the absence of associated eGFP fluorescence (Fig. 7B, panels to the top). Remarkably, this was accompanied by a collapse of endogenous Emx2 sense and antisense transcripts to levels peculiar to rhombo-spinal cells infected by the control virus (Fig. 7B, histograms).
In other words, even low, but persisting levels of antisense transcripts seem to be sufficient and strictly necessary for proper firing of the Emx2 transcription unit. This might explain why Emx2 sense and antisense transcripts have normally to be co-expressed in proliferative layers of the early developing CNS.

Absence of Emx2 Sense Transcription and/or Its Products Impairs Emx2OS-ncRNA Expression
The association between Emx2OS and Emx2 transcripts might also imply a reciprocal, complementary dependence of Emx2OS-ncRNA expression on Emx2 sense transcription. To address this issue, we compared levels of Emx2OS transcripts in E12.5 cortices from Emx2 +/+ , Emx2 +/2 and Emx2 2/2 mouse embryos [4]. The Emx2 null allele of these mutants -in fact -lacks only a 250bp genomic fragment encoding for the C-term of the homeodomain (not provided of any known cis-regulatory activities), has an intact divergent-promoter region and still harbors the two main enhancers which drive transcription in the developing telencephalon [4,57]. As such, it should in principle be able to normally drive the synthesis of its antisense transcript.
Notwithstanding that, we found that Emx2OS-ncRNA was down-regulated by more than 80% in homozygous null mutants  Ectopic activation of the endogenous Emx2 locus by OS1-179(+). The two graphs show Emx2-mRNA and Emx2OS-ncRNA expression levels in primary neural precursors from distinctive portions of the E10.5 CNS, rhombo-spinal tract (rh/sc) and telencephalon (te), acutely infected with ''driver'' and ''expressor'' lentivectors ( Fig. 1C), kept under doxycycline and added or not with the Li + /cyclopamine mix. Samples are RNA-profiled 72 hours after infection. Data are normalized on telencephalic precursors exposed to Li + /cyclopamine, infected by control lentivirus and kept under doxycycline. Basal Emx2 expression by rhombo-spinal cells, very low as compared to telencephalic precursors, rises about 2 (p,0.01), 4 (p,0.001) and 30 times (p,0.001), upon treatment of these cells with the OS1-179(+) virus, the drug mix, or both, respectively. Two-ways ANOVA indicates that a specific interaction between OS1-179(+) and the drug mix takes place, with p,0.001. A similar course is shown by Emx2OS-ncRNA. (B) Reversibility of OS1-179(+)-dependent activation of the endogenous Emx2 transcription unit. The four panels to the top show time course analysis of eGFP fluorescence in neural precursors, infected by driver and OS1-179(+) lentiviruses and kept under doxycycline for 72 or 120 hours. The absence of fluorescence in the bottom-right panel means that withdrawal of doxycycline at 72 hours is sufficient to reset levels of the eGFP/OS1-179(+) chimaeric transcript to zero by 120 hours. The two graphs to the bottom show Emx2-mRNA and Emx2OS-ncRNA expression levels in primary neural precursors from the E10.5 rhombo-spinal (rh/sc) tract, acutely infected with ''driver'' and ''expressor'' lentivectors ( Fig. 1C), kept under doxycycline for 72 or 120 hours and chronically exposed to Li + /cyclopamine throughout the experiment. Samples are RNA-profiled 120 hours after infection. Data are normalized on rhombo-spinal cells exposed to Li + and by 50% in heterozygous mutants (Fig. 8), suggesting that -as suspected -Emx2 sense transcription and/or its products are in turn necessary for proper expression of Emx2OS-ncRNA.

Discussion
In this study we investigated the expression pattern of Emx2OS-ncRNA, the antisense transcript associated to the transcription factor gene Emx2, in the developing mouse CNS. Then, we preliminarly addressed its involvement in regulation of Emx2 expression.
As shown by qRT-PCR (Fig. 2) and in situ hybridisation (Fig. 3), Emx2OS is specifically expressed by a number of CNS substructures and sense organs also expressing Emx2. Among them: telencephalon, mammillary recess, optic tectum, olfactory placode, otic vesicle [1,3]. Such colocalization of the two transcripts in periventricular layers of defined neural tube domains recalls the identical expression patterns they display in the uterine endometrium [17]. Emx2/Emx2OS colocalization might be an epiphenomenon of shared regulatory mechanisms impinging on both. Alternatively, transcription of one or both units might specifically promote the other and, in particular, antisense transcription might contribute to specifically keep Emx2 on in neural precursors of defined spatial domains.
Remarkably, the two transcripts, both expressed by periventricular neural precursors of the cortical primordium, display a mutually exclusive pattern in post-mitotic progenies of such precursors. Newborn neurons belonging to the cortical plate strongly express Emx2OS-ncRNA, expecially at the end of their radial migration (Fig. 4E,F), but not Emx2-mRNA (while still harboring residual Emx2 immunoreactivity) [1][2][3]. Pioneer Cajal-Retzius neurons lying in the marginal zone conversely express huge amounts of Emx2 mRNA and protein [3], but no antisense transcript at all (Fig. 4G,L). Such mutual distribution recalls the complementary expression patterns displayed by Otx2/Otx2OS and Crx/CrxOS pairs in the adult retina, with the sense transcription factor genes expressed predominantly in photoreceptors and the corresponding OS genes in bipolar interneurons and ganglionic projection neurons [24]. The mutual distribution of Emx2 and Emx2OS RNA products points to a possible negative cross-regulation between them and, in particular, to an involvement of the former in fine control of post-mitotic silencing of the latter.
As for the biological meaning of Emx2/Emx2OS co-expression, we tested the hypothesis that sense and antisense transcription from the Emx2 locus might reciprocally sustain each other.
To assay if Emx2OS-ncRNA is able to promote Emx2-mRNA expression, a 59 fragment of the former was overexpressed in rhombo-spinal precursors, normally expressing none of them. As suspected, delivery of this fragment, OS1-179(+)-ncRNA, into such cells elicits strong Emx2 upregulation. No direct assessment of the level, transcriptional or post-transcriptional, at which this phenomenon takes place, was performed. However two considerations suggests that the former possibility may hold good. First, OS1-179(+) synergizes with a drug mix specifically up-regulating beta-catenin, a key trans-activator binding the two telencephalic enhancers of Emx2 [59]. Second, up-regulation of Emx2 is faithfully paralleled by Emx2OS up-regulation.
Concerning mechanisms mediating transactivating properties of OS1-179(+) (as well of Emx2OS-ncRNA), two hypotheses may be taken into accont. Both molecules might interact with transcription factors impinging on the Emx2 locus, ameliorating their binding to the chromatin and/or their processivity (like the Evf2-ncRNA at the Dlx5/6 locus) [25]. Alternatively, antisense transcripts or their by-products might make the epigenetic state of Emx2 chromatine more permissive to gene expression, by recruiting appropriate modifier enzymes to it. However, modifier enzymes recruited by antisense RNAs, rather than facilitating gene expression, often make chromatine less prone to transcription [27,43,44]. Consistent with these considerations, the need of persistent OS1-179(+) expression to get Emx2 activation and the capability of the only drug cocktail to elicit weak but reproducible Emx2 upregulation induce us to rule out the latter hypothesis and to consider the former more likely. Beyond mechanics of OS1-179(+) action, such activation of Emx2 by a short antisense transcript is nonetheless remarkable, as possible prototype of a general method for overexpressing single specific genes, without any need to introduce additional copies of them into the genome, similarly to classical RNAa [62][63][64][65].
To ascertain if Emx2OS-ncRNA expression reciprocally depends on Emx2 transcription (and/or its products), Emx2OS levels were scored in Emx2 mutant mice, not harboring any gene lesions obviously incompatible with Emx2OS transcription. Remarkably, in such mice, both Emx2 and Emx2OS transcripts are dramatically down-regulated, echoing the collapse of Vax2OS (Vax2 opposite strand) occurring in mice knock-out for the cognate transcription factor gene Vax2 [24]. Multiple are mechanisms possibly underlying such down-regulation. Among the most likely ones, the depression of the Emx2/Wnt mutually-sustaining loop [9,59], leading to a decrease of the nuclear beta-catenin pool which normally stimulates transcription at the Emx2 locus [59]. Such mechanisms -however -were not addressed at all in this study; they will be subject of a further dedicated one.
So, based on Emx2OS down-regulation in Emx2-null mutants as well as on consequences of OS1-179(+) overexpression in rhombospinal precursors, we can speculate now that mutual promotion of sense and antisense transcription at the Emx2 locus may be a crucial pre-requisite for proper activation and expression of Emx2. Maybe that is why Emx2 sense and antisense transcripts have to be co-expressed in defined domains of the early neural tube.
We also tested whether Emx2OS is also implicated in downregulation of Emx2, by artificially modulating Emx2 antisense levels in cortico-cerebral neural precursors. An inverse correlation was found between levels of Emx2OS and Emx2 transcripts, confirming that the former may normally contribute to downregulation of the latter. Changes of Emx2-mRNA levels elicited by Emx2OS-RNAi are not reflected by the course of the corresponding pre-mRNA, that remains stable as found by intronic qRT-PCR. This further indicates that Emx2OS-dependent modulation of Emx2 takes place post-transcriptionally. Antisense-promoted down-regulation of Emx2 is weakened by Dicer1 knock-down, suggesting that Dicer1 might promote such process by ''dicing'' the Emx2-mRNA/Emx2OS-ncRNA hybrid, similarly to retrotransposon silencing [47,49]. Consistently with this model, the concomitant up-regulation of Emx2OS-ncRNA caused by OS1-179(+) might reflect a reduced degradation rate of the endogenous antisense molecule, following the competition by OS1-179(+) for sense transcript binding.
Remarkably, Emx2OS-dependent down-regulation of Emx2 is not limited to proliferating neural precursors. It takes place also in differentiating neuronal progenitors, where a surplus of Emx2OS-ncRNA is available to trigger this process (Fig. 5C, bottom left).
Here, an up-regulation of key enzymes involved in sense RNA degradation, occurring during neuronal differentiation [66], might make the Emx2OS impact on Emx2-mRNA degradation predominant over its positive effects on sense transcription, so accelerating the shutting off of the sense transcript. Actually, the amplitude of Emx2OS-dependent Emx2 downregulation is apparently not huge. Notwithstanding that, such regulation might be instrumental to proper progression of cortico-cerebral morphogenesis. In fact, as Emx2 finely regulates the balance between symmetric, selfrenewing divisions undergone by embryonic neural stem cells and asymmetrical, neuronogenic ones [12], an accurate regulation of its expression levels and, in particular, a finely tuned decay of its mRNA may be crucial to proper growth of the neocortical primordium [14]. A moderate contribution of Emx2OS-ncRNA to Emx2-mRNA destabilization might be just required, for optimal tuning of this process.

Ethics Statement
Animals handling and subsequent procedures were in accordance with European laws [European Communities Council Directive of November 24, 1986 (86/609/EEC)] and with National Institutes of Health guidelines.

Animal Handling
Wt mice (strain CD1, purchased from Harlan-Italy) and Emx2 null mutants [4] used in this study are maintained at the SISSA-CBM mouse facility. Embryos were staged by timed breeding and vaginal plug inspection. Embryos (E10.5-E18.5) were harvested from pregnant dames killed by cervical dislocation.

Preparation of Histological Samples
Dissected embryos were immersion-fixed overnight in 4% paraformaldehyde in 1X phosphate saline buffer (PBS) and then cryoprotected by immersion in 30% sucrose in PBS 1X overnight at 4uC. Tissues were frozen in OCT compound, sectioned at 10 mm in a cryostat, collected on slides SuperFrost Plus (Fischer), air dried for about 30 min, and stored at -80uC until use.

Cell Cultures
Primary cells. Cortical primordia and rhombo-spinal tracts were dissected from E12.5 mouse embryos and mechanically dissociated to single cells, by gentle pipetting. Dissociated neural precursor cells were cultured at 700 cells/ml, in DMEM/F12/ Glutamax medium (Invitrogen TM ), integrated with N2 supplement (Invitrogen TM ), 1 mg/ml BSA, 0.6% w/v glucose, 2 mg/ml heparin, 10 pg/ml Fgf2, 20 pg/ml Egf, 1X Pen/Strept (Gibco), 10 pg/ml fungizone. Cultures were usually blocked and underwent RNA extraction 72 hours after dissection. When a longer culturing time was required, primary neurospheres were dissociated to single cells by trypsin-DNAseI and re-plated at the same initial density. When required, doxycycline, lithium chloride and cyclopamine were added to the culture medium, at 2 mg/ml, 1mM and 1 mM, respectively.

Immunofluorescence
Neural cells were detached form plates by 1X trypsin (Gibco) for 5 minutes at RT, centrifuged at 100 g for 7 min at RT, resuspended in Sato medium and finally left 60 min to attach to SuperfrostH slides, previously kept for 60 min under 20 mg/ml poly-D-lysine at RT. Immunofluorescence was performed as previously described [66], with minor modifications. Briefly, after fixation by 4% paraformaldehyde for 10 min at 4uC and three washes in 1X PBS, 5 min each, cells were incubated for 1 h at RT under blocking mix (1X PBS; 10% FBS; 1 mg/ml BSA; 0.1% Triton-X100) and then incubated at 4uC overnight with primary antibody in blocking mix. Mouse anti-neuron-specific class III b-tubulin-primary antibody (clone Tuj1, Covance), was used, 1:600. Immunoreactivity was revealed after a 2 h incubation with secondary Alexa-594 antibody, 1:600.

In Situ Hybridization
Not radioactive in situ hybridization was performed as previously described, with minor modifications [67]. Two riboprobes were used: Emx2OS, corresponding to mouse chromosome 19: nt 59,500,768-59,501,561, and Reln, corresponding to the 3.3 kb EcoRI fragment from the BS6 clone (a gift by A.Bulfone). Hybridized embryo sections were imaged and analyzed using a fluorescent Nikon (Tokyo, Japan) Eclipse 80i microscope and a DS-2MBWC digital microscope camera. All images were processed by Adobe Photoshop CS3 software.
Quantitative RT-PCR RNA preparation. RNA was extracted from CNS explants and cell cultures by Trizol TM (Invitrogen), according to manifacturer's instructions.
cDNA prepration. At least 1 mg of RNA from each sample was retrotranscribed by SuperScriptIII TM (Invitrogen) in the presence of random hexamers, according to manifacturer's instructions, with minor modifications. In the case of pre-mRNA levels evaluation, at least 4 mg of RNA preparation from each sample, previously treated by DNAseI TM (Promega), were used.
Quantitative PCR. 1/25 of each cDNA sample was analyzed by the SybrGreen TM qPCR platform (Biorad). Each PCR reaction was run at least in triplicate and results averaged. Averages were further normalized against Tbp, except data in Fig. 5B (left and middle graphs), conversely normalized against GAPDH. Specifically in the case of pre-mRNA levels evaluation (reported in Fig. 5B, right graph), to reconcile sensitivity and specificity, 1/2 of each cDNA sample was linearly pre-amplified and 1/100 of the resulting primary reaction product was used as substrate of the subsequent quantitative PCR reaction, driven by nested, internal primers. Finally, supplementary amplifications on not-retrotranscribed samples were run, as negative controls.

Plasmids Construction
All basic DNA manipulations (extraction, purification, ligation) as well as bacterial cultures and transformation, media and buffer preparations were performed according to standard methods. DNAs were transformed in the E.Coli TOP-10 strains (Invitrogen).

Loss-of-Function Constructs and Their Validation
To knock-down Emx2OS by RNAi, we selected the best two siRNAs suggested by the Invitrogen ''BLOCK-iT TM RNAi Designer'' software (miR-aEmx2OS-542 and miR-aEmx2OS-774) plus the BLOCK-iT TM negative control-miR (miR-C) and compared their activities on Hela cells. For this purpose, we built up a set of three bicistronic plasmids, each harboring two distinct transcription units and designed to assay the activity of one specific miR. The former transcription unit, the ''sensor'', included the CMV promoter, the DsRed2 coding sequence, the cDNA corresponding to the Emx2OS-ncRNA fragment targeted in silico by the two miRs above (Genbank AY117414.1: nt796-1289) and the SV40 polyA site. The latter unit, the ''miR expressor'', harbored the CMV promoter, the eGFP coding sequence, a modified 230bp fragment from pri-mmu-miR-155 (where the miR-155 and miR-155* moieties are replaced by those of the artificial miR in order) and the TK polyA site. A fourth plasmid was built up as well, similar to the previous three, but harboring in cis sequences encoding for both miR-aEmx2OS-542 and miR-aEmx2OS-774 (Fig.1B). The pri-miR fragments included in the above mentioned plasmids were prepared as follows. Each of the two sequences 59 TGCTG CATAT TTGCA CTTCT CCGAA GGTTT TGGCC ACTGA CTGAC CTTCG GAGGT GCAAA TATGC AGG 39 and 59 TGCTG CGAAC TTAGA CTCAG ATTCC CGTTT TGGCC ACTGA CTGAC GGGAA TCTGT CTAAG TTCGC AGG 39 was cloned into pcDNA TM 6.2-GW/ EmGFP-miR (Invitrogen), inbetween mmu-miR-155 flanking regions, according to manifacturer's instructions, and the resulting SalI-XhoI chimeric pri-miR cDNA fragments, encoding for miR-aEmx2OS-542 and miR-aEmx2OS-774, respectively, were obtained. Finally, the SalI-XhoI cDNA fragment from ''pcDNA TM 6.2-GW/EmGFP-miR_neg_control_ plasmid'' was used as control pri-miR.
We transfected HeLa cells with these four bicistronic plasmids, counted fluorescent cells and calculated for each plasmid the ratio between red cells and total fluorescent cells, as a comprehensive index of pri-miR processing efficiency and mature miR activity. The DsRed2 + /(total fluorescent cells) ratio varied from 65% (negative control miR), to 39% (miR-aEmx2OS-542), 10% (miR-aEmx2OS-774) and 2% (miR-aEmx2OS-542 & miR-aEmx2OS-774), suggesting that the combination in cis of the two miRs, -542 and -774, would be the best choice (Fig. 5A). However, we noticed that, using the -542/-774 plasmid, the ratio between eGFP + cells and total cells in the plate was consistently lower (data not shown), possibly because of enhanced Drosha-dependent destabilization of the chimaeric eGFP-cds/Pri-miR molecule. As this might pose serious problems in subsequent production of miR-encoding lentiviruses as well as in tracing cells infected by these virions, we prudently opted for simple miR-aEmx2OS-774 for the knocking-down experiments.

Conditional Gain-of-Function Constructs
Emx2OS/Emx2 RNA sequences were conditionally overexpressed by lentiviral vectors and TetON technology.

Morpholino Reagents and Their Validation
Sequences of anti-Dicer1 morpholino and its reverse negative control were obtained from the Gene-Tools free design service (a-Dicer1 and C, respectively) and the former molecule was validated, by testing its capability to inhibit Dicer1-dependent pri-miR-124a maturation [56]. This was assayed by a DsRed2 sensor gene linked to a miR-124a responsive element, cotrasfected into NIH/3T3 cells with pri-miR-expressor plasmids (Fig. S1).
Cell infections were performed without polybrene. In gain-offunction experiments, each virus was used at ''multiplicity of infection'' (m.o.i.) = 10, which resulted in trasduction of almost the totality of cortical precursors [71]. The m.o.i. was raised to 20 in loss-of-function experiments, in order to get a more robust downregulation of the target RNA.