The Dlx5 and Foxg1 transcription factors, linked via miRNA-9 and -200, are required for the development of the olfactory and GnRH system

During neuronal development and maturation, microRNAs (miRs) play diverse functions ranging from early patterning, proliferation and commitment to differentiation, survival, homeostasis, activity and plasticity of more mature and adult neurons. The role of miRs in the differentiation of olfactory receptor neurons (ORNs) is emerging from the conditional inactivation of Dicer in immature ORN, and the depletion of all mature miRs in this system. Here, we identify specific miRs involved in olfactory development, by focusing on mice null for Dlx5, a homeogene essential for both ORN differentiation and axon guidance and connectivity. Analysis of miR expression in Dlx5−/− olfactory epithelium pointed to reduced levels of miR-9, miR-376a and four miRs of the -200 class in the absence of Dlx5. To functionally examine the role of these miRs, we depleted miR-9 and miR-200 class in reporter zebrafish embryos and observed delayed ORN differentiation, altered axonal trajectory/targeting, and altered genesis and position of olfactory-associated GnRH neurons, i.e. a phenotype known as Kallmann syndrome in humans. miR-9 and miR-200-class negatively control Foxg1 mRNA, a fork-head transcription factor essential for development of the olfactory epithelium and of the forebrain, known to maintain progenitors in a stem state. Increased levels of z-foxg1 mRNA resulted in delayed ORN differentiation and altered axon trajectory, in zebrafish embryos. This work describes for the first time the role of specific miR (-9 and -200) in olfactory/GnRH development, and uncovers a Dlx5–Foxg1 regulation whose alteration affects receptor neuron differentiation, axonal targeting, GnRH neuron development, the hallmarks of the Kallmann syndrome.


Introduction
During mammalian embryonic development, olfactory receptor neurons (ORNs) are specified and differentiated within two distinct neuroepithelial regions: the main olfactory epithelium (OE) and the Vomero-Nasal (VN) epithelium. Although similar, OE-and VN-derived mature neurons express distinct classes of odorant receptors and carry out different odour-transducing functions. Such molecular and physiological distinctions are apparently maintained in all vertebrates.
It is increasingly being recognized that biological processes are governed by complex regulatory modules and networks of molecular interactors, rather than simplistically by individual genes with individual functions. In these networks, non-coding RNAs (miR, lncRNAs, linc-RNAs, anti-sense RNAs and pseudogenes) play an important role (Arora et al., 2013;Esteller, 2011;Konopka, 2011;Mayanil, 2013;Ng et al., 2013;O'Brien et al., 2012;Salmena et al., 2011;Satoh, 2012;Schonrock et al., 2012). Thus, it is conceivable that mutations or misexpression of non-coding RNAs could participate in the molecular pathogenesis of KS/nCHH. Gaining knowledge on the RNA networks and regulations underlying olfactory differentiation, neuronal connectivity and guidance would be of great importance.
MicroRNAs (miRs) represent a class of short non-coding RNAs that act as negative post-translational regulators on longer coding and non-coding RNAs (Bartel, 2004). Annealing of complementary sequences enables miR to induce degradation or inhibit translation of target mRNAs (Plasterk, 2006). The neuronal functions of miR range from patterning and cell differentiation during embryonic development to physiology of more mature and adult neurons, including their survival, homeostasis, activity and plasticity (Agostini et al., 2011;Aranha et al., 2011;Bian and Sun, 2011;Brett et al., 2011;Fiore et al., 2011;Gao, 2010;Gaughwin et al., 2011;Li et al., 2011;Luikart et al., 2011;Olde Loohuis et al., 2012;Shi et al., 2010). More specifically, a role of miRs in the development of sensory neurons, including olfactory sensory neurons, is beginning to emerge. In Drosophila, miR-7 has been implicated in the differentiation of photoreceptor cells via regulation of the EGF receptor signalling (Li and Carthew, 2005). In Caenorhabditis elegans miR-273 and lsy-6 have been shown to be required for asymmetric expression of taste receptors in chemosensory neurons (Chang et al., 2004;Johnston and Hobert, 2003). In Danio rerio (zebrafish) the miR-200-class is required for the proliferation, differentiation and survival of ORNs (Choi et al., 2008). In Xenopus laevis miR-124 regulates changes in the sensitivity of retinal ganglion cells' growth cones to the guidance signal SEMA3A (Baudet et al., 2011), implicated in the pathogenesis of KS (Cariboni et al., 2011;Hanchate et al., 2012). In the mouse, the conditional disruption of Dicer in the developing olfactory system results in impaired ORN differentiation and reduced survival (Choi et al., 2008), indicating that mature miRs are required for these processes; however, without revealing their identity.
Since the activity of single miR is context-and time-specific, their functions should be examined within these contexts. With this in mind we generated high-throughput data from the developing olfactory system, focusing on the Dlx5 homeogene: its targeted inactivation leads to a fully penetrant KS-related defects consisting in delayed ORN differentiation, impaired axonal connectivity and failure of GnRH neurons to reach the forebrain (Levi et al., 2003;Long et al., 2003;Merlo et al., 2007). We screened for miR expression in ORNs, comparing wild-type vs Dlx5 mutant tissues, and identified miR-9 and miR 200-class as the molecular link between Dlx5 and Foxg1. Using reporter zebrafish strains to visualize the embryonic olfactory axons Sato et al., 2005;Yoshida et al., 2002) or the GnRH + neurons (Abraham et al., 2008(Abraham et al., , 2010, we show that miR-9 and miR-200-class play a role in ORN differentiation and axonal organization. We also show that these miRs are required for early GnRH neuron genesis and position. Thus we have identified a novel miR-based regulation that participates in the control of olfactory development, axon connectivity and GnRH neuron development.

Materials and method
2.1. Dlx5 mutant mouse strain All procedures using mice were approved by the Ethical Committee of the University of Torino, and by the Italian Ministry of Health. Mice with targeted disruption of Dlx5 have been previously reported (Acampora et al., 1999). The null allele, denominated Dlx5 lacZ , allows for detection of the Dlx5-expressing cells by staining for βgalactosidase (β-gal) expression. The olfactory phenotype has been previously characterized (Levi et al., 2003;Long et al., 2003;Merlo et al., 2007). Dlx5 +/− (heterozygous) males and females were crossed, and generated the expected Mendelian ratios of embryos with genotype WT, Dlx5 +/− and Dlx5 −/− . Pregnant females were sacrificed at the chosen embryonic age by cervical dislocation.

Collection of embryonic olfactory epithelia and RNA extraction
Embryos were collected clean of extra-embryonic tissues (used for genotyping) by manual dissection, transferred in RNAse-free PBS, and further dissected to separate the head. This was then included in 3% low-melting agarose (Sigma-Aldrich) in PBS, let harden and vibratome-sectioned at 250 μm thickness. Sections were transferred in RNase-free PBS, and manually dissected with fine pins to collect the OE or the VNO epithelia. For the Dlx5 mutant tissues, the entire epithelial lining of the nasal cavity was collected, since it is not possible to discriminate the olfactory vs. the respiratory epithelium. The excised tissues were individually collected in RNA-later (Life Technologies AM7020) and stored at -20 until extraction. The collected tissues with the same genotype were pooled into three samples, used to extract total RNA with the TRIzol reagent (Life Technologies), following the manufacturer's instructions. Correct pooling was further verified by RT-PCR for the Dlx5 mRNA. RNA samples were quantified using a NanoDrop1000 spectrophotometer (Nanodrop Technologies, Inc.), and analysed by capillary electrophoresis on an Agilent Bioanalyzer. Only samples showing an RNA Integrity Number N 5 were further processed. The quality of small RNAs was further assayed by using a small RNA Chip Bioanalyzer (Agilent).

miR profiling and data analysis
RNA samples were labelled using the one-colour method, with the miRNA Complete Labelling and Hybridization kit (Agilent). For the profiling, the mouse 8x15K arrays were used (Agilent) on a mouse V2 microarray platform. These arrays comprise probes for 627 mature mouse miRs and 39 viral miRs. Data were extracted using conventional spot-recognition and significantly-above-background tools; the signal intensity was normalized across samples using the Lowess cyclic normalization algorithm. Differentially expressed miRs (DEM) were detected using the SAM two-class unpaired statistical tool, using a FDR = 5%. Of the 627 mouse miR probes present on the arrays, 118 miRNA were found to be expressed significantly above the background.

Softwares and databases used
For preliminary Gene Ontology (GO) analyses we used DAVID (http://david.abcc.ncifcrf.gov/) and KEGG (http://www.genome.jp/ kegg/pathway.html). For improved categorization and visualization, we used ClueGO (Bindea et al., 2009). To examine the embryonic expression of individual RefSeq we used the publicly available transcriptome-wide in situ expression databases Genepaint (www. genepaint.org) and Eurexpress (www.eurexpress.org). For the prediction of miR targets we used TargetScan6.2 (www.targetscan.org) and Hoctar, a miR target prediction tool based on a combination of miR-mRNA co-expression and seed sequence recognition (Gennarino et al., 2011). For photo-documentation, digital micrographic images were contrast balanced and colour matched using PhotoShop 7 (Adobe), cropped, rotated and assembled into figures with QuarkXPress (Pantone).

Real-Time qPCR analysis for coding mRNAs
Relative mRNA abundance was determined by Real-Time qPCR. Total RNA (500 ng) was reverse-transcribed at 42°C with Reverse Transcriptase (Promega), amplified with the GoTaq qPCR Master Mix (Promega), and detected with the Universal Primer System (UPS, Roche) on the AB7900 equipment (Applied Biosystem). The abundance of TATA-binding protein (TBP) and of GAPDH mRNAs were used for normalization. Primers were designed with the dedicated UPS on-line tool (Roche) and provided in Suppl. Table I. Experiments were repeated twice on independent samples, every point was done in triplicate. Data analysis was performed using the ΔΔCt formula, with ABI software, version 2.1 (Applied Biosystems).

Identification of Dlx5 binding sites and chromatin immunoprecipitation
Dlx5 binding sites and putative target genes were predicted essentially as described (Vieux-Rochas et al., 2013). Briefly, the PWM provided by JASPAR (PH0024.1) was used to score sites using standard loglikelihood ratios. Only those sites conserved in at least two vertebrate species were considered. For chromatin immunoprecipitation (ChIP) we used the human SHSY-5Y neuroblastoma cells, which express low endogenous levels of Dlx5, miR-9 and miR-200, transfected with 5 μg of DLX5-myc-tag expression vector (from Open-Biosystem) or with the same vector in which the Q178P mutation (Shamseldin et al., 2012) was introduced (BioFab, Rome, sequence verified). Chromatin was crosslinked, sonicated, precipitated with anti-myc TAG antibody (SantaCruz, sc-40) and de-crosslinked according to instructions (EZ Magna ChipG, Millipore). Genomic fragments comprising the predicted Dlx5 putative binding sites were PCR-amplified with primer flanking the site (sequences in Suppl. Table II). Chromatin from cells transfected with an empty vector, or with a vector expressing the Q178P DLX5, was used for comparison. ChIP with an irrelevant IgG was also used as negative control. Total chromatin was used as positive control (input).

Transfection of pre-miRs and anti-miRs
To downmodulate endogenously expressed miR-9 and miR-200 we used the commercially available Ambion anti-miR inhibitors (Life Technologies). To overexpress miR-9 and miR-200 exogenously we used commercially available Ambion pre-miR precursors (Life Technologies). The human neuroblastoma cells SH-SY5Y were plated in 6-well plates at 50% confluency and 24 h later were transfected with 100 nM of the anti-miR or 75 nM of the pre-miR vector, using the Lipofectamine 2000 reagent (Life Technologies), according to the manufacturer's instructions. Cells were examined for miR over-expression or knock-down 48 hrs after transfection, by single-assay miR Real-Time qPCR (Applied Biosystem).

Antibodies and immunoblotting
For Western immunoblotting, the following primary antibodies were used: anti-Foxg1 (1:1000, Abcam) and anti-vinculin. Secondary antibody was horseradish peroxydase-conjugated anti-mouse and anti-rabbit (Sigma, 1:250). Total proteins were extracted either from transfected SH-SY5Y cells (following trypsinization and centrifugation) or from the forebrains of E16.5 embryos (following dissection and quickly freezing in liquid N 2 ) using the same procedure. Essentially, cells were lysed with an electrical tissue homogenizer (Turax) in modified RIPA buffer composed as follows: 150 mM NaCl, 50 mM Tris, pH 7.5 with 1% Triton X-100, 0.5% Na deoxycholate, 0.1% SDS, 2 mM EDTA, 1 mM sodium orthovanadate, 1 mM sodium fluoride and protease inhibitors. Homogenates were centrifuged at 13,000 rpm for 10 min at 4°C, extracts were quantified with the Bio-Rad Assay, then separated in 10% acrylamide gel and transferred to a PVDF membrane (Millipore). Membranes were saturated with BSA 5% in TBS-Tween-20 0.3%, washed in TBS-Tween, incubated with the primary antibodies ON at 4°C, then washed in TBS-Tween, incubated with HRPconjugated secondary antibodies (diluted 1:5000), washed and developed with chemo-luminescence reagent Luminata TM Forte Western HRP Substrate (Millipore). Images were acquired with the ChemiDoc system (Bio-Rad), exported and analysed with Quantity One software. Signals were normalized against vinculin.

miR in situ hybridization in tissue sections
WT and Dlx5 −/− mouse embryos were collected at 13.5 days postcoitum (E13.5), fixed in 4% PFA/0.1 M phosphate buffer (PB, pH 7.4) for 12-16 hrs, washed in PBS, dehydratated in methanol, processed for paraffin embedding and sectioned at 6 μm. Hybridization was carried out with DIG-labelled riboprobes that specifically detect the mature form of mouse miR-9 and miR-141 (Exiqon) in according with manufactory instruction. The sections were hybridized with the probe for 16 hrs, washed, incubated with an anti-DIG-AP antibody (Roche) and developed with NBT-BCIP (Sigma). To control the efficiency of the procedure and RNA preservation, we adjacent sections were hybridized with a probe that detects the snU6 small ncRNA.

Zebrafish strains and treatments
All procedures using zebrafishes are authorized by the Ethical Committee of the University of Torino and the Italian Ministry of Health. The following two strains were used for visualization of the olfactory axons: OMP 2k :gap-CFP rw034 and TRPC2 4.5k :gap-Venus rw037 Sato et al., 2005;Yoshida et al., 2002). These strains were obtained from Drs. Nobuhiko Miyasaka and Yoshihiro Yoshihara (RIKEN Brain Science Institute, Japan). The OMP::CFP transgene labels the OEtype neurons of the fish olfactory placode, while the Trpc2::Venus transgene labels the VNO-type neurons . The fish strain GnRH3::GFP (Abraham et al., 2008(Abraham et al., , 2010 was obtained from Prof. Y. Zohar (Univ. Maryland Biotechnology Institute, Baltimore, USA).
Adult fishes were maintained, bred and genotyped according to standard procedures, kept under a 14 hrs-light and 10 hrs-dark photoperiod at approximately 28°C. Allelic transmission followed the expected Mendelian ratios. Following fertilization, eggs were collected and embryos (WT or micro-injected) were grown in the presence of 0.003% 1-phenyl-2-thiourea (PTU) to prevent formation of melanin pigment. To down-modulate specific miRs we utilized a conventional antisense morpholino oligo (MO)-mediated strategy (from GeneTools, LLC, Philomat OR, USA) (Flynt et al., 2007;Kloosterman and Plasterk, 2006). For z-dlx5a two MOs were designed: one that prevented exon1-intron1 splicing, and consequently led a premature Stop codon upstream of the homeodomain, and a second one annealing on the ATG start codon and preventing translation. Anti-z-miR-9 MO was designed with the on-line dedicated tool https://oligodesign.gene-tools.com/request/, while the anti-z-miR-200 MO mix was as previously published (Choi et al., 2008). Sequences are provided in Suppl. Table III. The complete zebrafish foxg1a cDNA (gene ID 30274), including the CDS, the 5′ and the 3′ UTRs, was cloned in the pCS2 + plasmid, and the mRNA was in vitro transcribed, polyadenylated and capped with the mMESSAGE mMACHINE SP6 kit (Life Technologies).
Fertilized eggs were collected at one cell stage and injected under stereological examination with 4 ng of MO, in the presence of Phenol Red for further selection of the injected embryos. 48-72 h post fertilization (hpf) embryos were fixed with 4% PFA at 4°C ON, washed in PBS and embedded in 4% low melting agarose, 0.1% Tween-20. The apical portion of the head was manually dissected from the rest of the embryo. The OMP::CFP+ and the Trpc2::Venus + (YFP+) axons were viewed in a frontal plane. Confocal microscopy analysis was performed using Leica TCS SP5 (Leica Microsystems). Images were acquired as Z-stacks of 1 μm-thick optical sections. Digital micrograph images were contrast balanced and colour matched using PhotoShop 7 (Adobe), cropped, rotated and assembled into figures with QuarkXPress (Pantone).
GnRH3::GFP+ neurons were examined on fixed 72 hpf embryos, at 20× and 40× magnification, by confocal microscopy. For counting, we followed the procedure detailed in Y. Zhao et al. (2013). Briefly, stacks of Z-slices at 40 × were used to manually count the GFP + neurons with the assistance of ImageJ image-processing software (from NIH), after adjustment of contrast and brightness. To avoid double-counting of the same neuron, each optical slice (1 μm-thick) was also individually reexamined. Merged images of all Z-slices confirmed the estimated number.
For time-lapse video recording, embryos were collected at 34 and 60 hpf, anaesthetized with tricain (standard dose) embedded in 0.8% low-melting agarose and examined by confocal microscopy using the following parameters: 20× magnification, 10 Z-slices (5 µm each), acquisition every 20 min. for a total of 16 hrs. Laser intensity was adjusted to avoid photo-damage. Row images were adjusted for contrast and brightness, oriented and registered with ImageJ, videos were mounted with the same software.

Results
3.1. Differentially expressed miRs in Dlx5 −/− OE RNA samples from WT and Dlx5 −/− OE (age E12.5) were used to hybridize on Agilent arrays for the detection of mature mouse miRs. Of the 627 miR probes present on the arrays, 118 miRs were found to be expressed significantly above the background. We compared normal vs. Dlx5 −/− OE using both SAM two-class unpaired and the Rank Product statistical analyses, with a FDR = 5 %. We found eight miRs differentially expressed, six down-regulated (miR-9, miR-141, miR-200a, miR-200b, miR-429 and miR-376a) and two up-regulated (miR-450a-5p and miR130b*) in the Dlx5 −/− OE (Fig. 1a). The most altered miR was miR-9, with a fold change of -2, while the other miRs showed a fold-change between −1.9 and +1.3.
To confirm the modulation of these miRs in the absence of Dlx5, we quantified their expression by Real-Time qPCR, using newly collected RNA samples from WT and Dlx5 −/− embryonic OE. The results confirmed a reduced level of all the down-regulated miRs, while did not confirm the up-regulated ones (Fig. 1b); hence, these were not further considered. As a further confirmation, we carried out in situ hybridization on sections of WT and Dlx5 −/− embryonic OE, at the age E12.5, to detect miR-9, miR-141 and miR-429, using specific mouse DIG-labelled probes. We observed a reduction of miR-9, miR-141 and miR-429 signal in the Dlx5 −/− OE, compared to the WT (Fig. 1c), while hybridization with two positive controls, Sp8 (expressed in the OE) and Sox5 (expressed in chondrogenic condensations), yielded an equivalent positive signal in both genotypes, indicating adequate RNA preservation. miR200a and miR200b could not be tested by in situ hybridization due to high sequence conservation between all members of the miR-200class.

Genomic regulation of miR-9 and miR-200 by Dlx5
The three loci miR-9.1, -9.2 and -9.3, located on chromosomes 3, 13 and 17 respectively, generate identical mature miR when transcribed, referred to as "miR-9". The miR-200a, -200b and -429 loci are closely located on chromosome 4, while miR-141 and -200c are closely located on chromosome 6. miR-376a is clustered with 16 other miRs on chromosome 12. Suppl. Fig. 1 provides a summary of the sequence and chromosomal location of the mentioned miRs. We sought to establish if any of the downmodulated miRs might be a direct Dlx5 transcriptional target. First we searched for the presence of Dlx5 binding sites near the identified miRs loci in genomic regions conserved between human, mouse and other mammalian species. We used a previously generated genome-wide prediction of Dlx5 binding sites, based on its Position-Weight Matrix present in the Jaspar database (Portales-Casamar et al., 2010;Vieux-Rochas et al., 2013). We predicted one Dlx5 binding site near the miR-9.2 locus, located about 1.5 kb downstream, three sites near the miR-9.3 locus, located about 4, 5 and 6 kb downstream, and two sites near the miR-200a-200b-429 locus, located about 5 kb upstream (Fig. 2a). No Dlx5 binding site was predicted within a 50 kb range from the miR-9.1, miR-141, miR-200c and miR-376a loci.
To test whether the DLX5 protein physically occupies the Dlx5 sites near the miR-9.3 and miR-200a/b/miR-429 loci, Chromatin Immuno-Precipitation (ChIP) analysis on these sites was performed. myc-tagged version of either the WT or the Q178P mutant DLX5 were expressed in the SH-SY5Y human neuroblastoma cells, which express DLX5, miR-9 and miR-200 endogenously. The Q178P mutation falls in the DNAbinding domain of the protein, and has been found to co-segregate with ectrodactyly in a human family (Shamseldin et al., 2012). For both locations, an enrichment in the myc-immunoprecipitated chromatin in cells transfected with WT DLX5 was observed, compared to cells transfected with the empty vector, while the transfection with DLX5-Q178P failed to show binding (Fig. 2b, c). The identity of the amplified ChIP fragments was sequence-verified.
To determine whether the forced expression of DLX5 may result in an upregulation of miR-9 and miR-200-class RNAs, SH-SY5Y cells were transfected with myc-tagged wild-type DLX5 or Q178P mutant DLX5 expression vectors, and the relative abundance of miR-9 and miR-200 was quantified by Real-Time qPCR. The over-expression of DLX5 induced a 2.5-3 fold increase in the abundance of miR-9 in this system, while the Q178P mutant DLX5 did not (Fig. 2d). On the contrary, DLX5 overexpression did not induce changes in miR-200 expression, either in SH-SY5Y (Fig. 2d) or in GN11 (neuroendocrine) or in U2OS (osteosarcoma) cells (data not shown). Thus, Dlx5 is likely to regulate the expression of miR-9.3 directly, and the expression of miR-200a/b/miR-429 indirectly. Alternatively the expression of miR-200a/b/miR-429 could require additional transcription (co)factors not present in these cells.
3.3. Searching for functionally relevant targets of miR-9 and miR-200 clsss in the OE miR-9 is widely expressed in the forebrain and olfactory sensory system of the mouse embryo and has been implicated in neural development (La Torre et al., 2013;Shibata et al., 2011;C. Zhao et al., 2013). The miRs of the -200 class are also expressed in the developing OE and are known to play a role in ORN differentiation (Choi et al., 2008). Since miR-200a, miR-200b, miR-141 and miR-429 share very similar seed sequences (Suppl. Fig. 1), they are predicted to bind to, and negatively regulate, common sets of RNA targets.
To compile lists of best predictable targets, we initially used TargetScan-6.2, a tool that uses sequence base-pairing and conservation criteria. We integrated the results obtained with TargetScan with Hoctar (Gennarino et al., 2011), a prediction tool that combines sequence basepairing with co-expression and anti-correlated co-regulation criteria. With these two tools, we predicted the most reliable miR-9 targets, and functionally classified the top scoring ones, to search for significantly enriched categories. The results were organized as "Biological Process", "Molecular Function" and "Cellular Component" using the GeneOntology (GO)-based tool ClueGO (Bindea et al., 2009). For miR-9 we detected only three enriched categories: regulation of cell differentiation, cell junction assembly and neuron development (Suppl . Tables IV  A, B, C).
We carried out the same prediction and categorization analyses for the miR of the -200 class; in this case the two subfamilies (miR-141/200a and miR-200b/c/429/548a) were examined separately, and indeed yielded lists which were similar but not identical. For miR-141/200a, the enriched categories included regulation of cell differentiation, neurogenesis, development and regulation of transcription (Suppl. We then searched for predicted targets that have seed sequences for two or three of the miRs downregulated in the absence of Dlx5. The analysis yielded, respectively, 54 and 30 mRNAs, whose functional annotation indicated an enrichment of the following categories: cell differentiation, neurogenesis, neuron differentiation, projection development and regulation of transcription (Suppl. Tables VIIA, B, C and VIIIA, B). Amongst these we noted Foxp1 and Foxg1, which however were not differentially expressed in the Dlx5 −/− OE with the adopted cut-off value and statistical parameters . Two possible explanations: either changes in the abundance of miR-9 and miR-200-class cause changes in the abundance of target RNAs that are too modest to Asterisks indicate statistical significance. n.s., not significant. c. in situ hybridization on coronal sections of WT (top panels) and Dlx5 −/− (bottom panels) embryos, at the age E12.5, to detect endogenous miR-9, miR-141 and miR-429. Hybridization with probes detecting SP8 (expressed in the OE) or Sox5 (expressed in chondrogenic condensations) was carried out as control for RNA preservation (panels on the right). The olfactory epithelium (OE) and the septum are indicated (dotted lines). Note the reduced expression of all the tested miRs in the mutant specimens.
pass the imposed cut-off value, or these miRs preferentially affect translation and not stability of the target mRNAs.
Next we intersected the predicted miR-9 and miR-200-class targets with the coding mRNAs found to be differentially expressed in the Dlx5 −/− OE compared to the WT . A significant enrichment of miR-9 and miR-200-class target sequences was detected in the 3′ UTR of genes up-regulated in the Dlx5 −/− OE (Table 1A, B). Some of the miR-9 and miR-200-class targets upregulated in the mutant OE (Qk, Foxf2) are mesenchymally-expressed rather than OEexpressed, while other targets were actually downregulated in the absence of Dlx5 (Akap6, Elmod1, Snap25) ( Table 1C). The overall poor anti-correlation between DEMs and DEGs may be explained considering that changes in the miR abundance not necessarily cause changes in the abundance of the target mRNA. These genes were not further considered.
To determine whether miR-9 and miR-200-class may modulate Foxg1 protein level, the effect of introduction of pre-miR-9 or depletion of endogenous miR-9 on Foxg1 protein level was assayed by Western blot analysis in SH-SY5Y cells, which express DLX5, miR-9, miR-200-class and Foxg1 endogenously. These cells appear relatively more similar to the olfactory neurons (from a molecular point of view, and limited to these target genes) than other cellular models that we considered. The expression of pre-miR-9 induced a 6-fold reduction in Foxg1 protein level, while expression of anti-miR-9 induced a 2-fold increase in Foxg1 level (Fig. 3a,b). We also determined the level of endogenous Foxg1 mRNA, by Real-Time qPCR, upon expression of pre-miR-9 or anti-miR-9, and observed, respectively, a 2-fold decrease and a 2.5-fold increase in the relative Foxg1 mRNA abundance (data not shown). In the same cells, the expression of pre-miR-200 led to a 3.9-fold decrease in Foxg1 proteins level (Fig. 3c). Thus, both miR-9 and miR-200 negatively regulate Foxg1 protein level.
Next, the level of Foxg1 protein in extracts from the forebrain of Dlx5 +/− vs. Dlx5 −/− embryos, at the age E13.5, were compared by Western blot analysis. In the Dlx5 −/− samples we detected a 1.9-fold increase of Foxg1 protein, relative to the Dlx5 +/− sample (Fig. 3d). Finally, we examined the expression of Foxg1 protein on coronal sections of the nasal region of WT and Dlx5 −/− embryos, at the age E13.5, by immunostaining with anti-Foxg1 antibody. We observed an increase in Foxg1 signal in the Dlx5 −/− OE, compared to the WT (Fig. 3e). To confirm that Foxg1 staining corresponds to the OE in both genotypes, and to verify the delayed differentiation in the absence of Dlx5, we stained adjacent sections with the anti-βIII-tubulin antibody TuJ1. We detected a reduced βIII-tubulin staining in the Dlx5 −/− sections compared to the WT specimen. Thus, the level of Foxg1 protein is increased in the absence of Dlx5.
In summary, since miR-9 and miR-200-class are down-modulated in the absence of Dlx5, while Foxg1 protein level is up-regulated, and since the 3′ UTR of the Foxg1 mRNA is a predicted target of these miRs, we can infer that the Dlx5-miR-Foxg1 regulation is most likely a direct one.

Dlx5 is required for normal olfactory development in zebrafish embryos
Olfactory development and GnRH ontogenesis are highly conserved amongst vertebrates including zebrafish (Lohr and Hammerschmidt, 2011;Roa, 2013;Whitlock et al., 2006;Wierman et al., 2011). D. rerio (zebrafish) embryos are increasingly being used in developmental genetics, thus we opted to gain functional in vivo knowledge on the identified miRs, by using two lines of transgenic fishes that allow a direct visualization of the olfactory axons Yoshida et al., 2002). In these lines, two distinct fluorescent reporters are expressed in the two main populations of peripheral ORNs, and the reporter proteins are efficiently transported in their axons (Suppl. Fig. 3). In one strain CFP is expressed under the control of OMP promoter, and marks the majority of ORNs lying in the basal portion of the embryonic OE, projecting their axons to the dorsal OB. In the other strain, the Venus (YFP) reporter is expressed under the control of the Trpc2 promoter and marks a sub-population of sensory neurons that occupies the apical part of the OE and projects to the ventro-lateral OB. The fish CFP+ and the YFP+ neurons are thought to correspond, respectively, to the OE and VNO receptors in the mammalian system ; however, in the fish embryos these two sensory neuron populations originate from the olfactory placode but then remain intermingled and occupy the same neuroepithelium, while in the mouse they occupy distinct nasal structures.
To deplete z-dlx5a, the fish ortholog of mammalian Dlx5 (Ellies et al., 1997;MacDonald et al., 2010;Quint et al., 2000), we used a combination of two morpholino oligos (MO), as previously reported : the first designed to anneal with the exon1-intron1 splice junction, expected to interfere with the splicing of the z-dlx5a primary transcript and to result in a premature stop codon upstream of the homeodomain; the second designed to anneal with the ATG start codon and expected to inhibit the translation of the z-dlx5a mRNA (sequences in Suppl. Table III). The depletion of z-dlx5a was previously shown to cause axon misorientation and altered glomeruli formation Fig. 2. Dlx5 regulates the expression of miR-9 and miR-200-class loci. a. Position of predicted Dlx5 binding sites near the miR-9.2, the miR-9.3 and the miR-200a/b/miR-429 loci (black arrowheads), annotated on the UCSC mouse genome browser. The position of the miR loci and the mammalian conservation are reported. Note that the predicted Dlx5 sites fall in conserved regions. b, c. ChIP analyses with anti-myc TAG antibody, of human genomic fragments encompassing the Dlx5 sites predicted for miR-9.3 and miR-200a/b/miR-429 (shown in a), upon transfection of human SH-SY5Y cells with an empty vector (left lanes), with a DLX5-myc (middle lanes) or with a Q178P mutant DLX5-myc (right lanes) expression vector. Enrichment is observed only when wild-type DLX5 (red arrows), and not Q178P mutant DLX5 (red asterisks), is transfected (red arrows). d. Quantitative determination of endogenous level of miR-9 and miR-200 in SH-SY5Y cells, upon exogenous expression of DLX5-myc or Q178P DLX5-myc. Expression of WT DLX5, but not Q178P mutant DLX5 results in increased levels of miR-9, but not miR-200. in the MO-injected zebrafishes . Here we focused on OE differentiation at 72 hpf. First we verified that the splice-variant MO caused efficient intron 1 retention of the endogenous z-dlx5a mRNA (Fig. 4d). Injection of the control MO in the eggs from OMP::CFP × Trpc2::YFP strains did not cause significant reduction in the number of embryos that were CFP + (observed frequency 72%, expected frequency 67%, total analysed N = 40). The expected frequency represents the fact that we examine single-positives from CFP + × YFP + breeds, and the double-negative are then eliminated. Upon injection of the z-dlx5a MO, about 70% of the CFP + embryos showed a clear reduction of the CFP signal intensity (Fig. 4a-c), while injection of the same in the Trpc2::Venus strain yielded 72% of YFP+ embryo (78% in the control injected), in which the YFP+ fluorescence was seldom reduced (b10% of YFP+) (Fig. 4a-c). The reduced fluorescence intensity of the OMP::CFP reporter and the reduced number of CFP + embryos counted after injection suggest a delayed differentiation of the OMP-type olfactory neuron. In contrast, the differentiation of the Trpc2-type neurons seemed minimally unaffected. To better document this, we measured the abundance of endogenous z-OMPa, z-OMPb, z-Trpc2, z-ngn1 and z-S100β mRNAs in the head-piece of injected embryos, by Real-Time qPCR. As Trpc2 and OMPa/b are early markers for the differentiation of VNO-type and OE-type neurons, respectively, and since these mRNAs are not expressed in other embryonic territories, they are well suited to monitor overall olfactory epithelium differentiation. With the exception of z-S100β, all these differentiation markers showed a 40-60% reduced expression in the injected embryos, indicating a true differentiation delay (Fig. 4e). Overall, embryonic development and morphology were not significantly changed, revealed by visual observation. As a further control for the general developmental progression, the abundance of endogenous z-hoxA7a and z-hoxA10b did not significantly change upon MO injection (Fig. 4e). Thus, reduced z-dlx5a level does not affect general embryonic development.
In conclusion, the depletion of z-dlx5a in the fish embryos causes a differentiation delay on the OMP-type ORN and a general alteration of trajectory and connectivity of the olfactory axons , phenotypes fully recapitulating all main aspects of the Dlx5 −/− phenotype in the mouse (Levi et al., 2003;Long et al., 2003;Merlo et al., 2007). precursor (pre-miR-9) (a) or treatment of the cells with and anti-miR-9 (b). Quantification was done using vinculin as house-keeping control, and is shown on the right of each blot. A down-modulation and an up-modulation of Foxg1 is observed, respectively, with the pre-miR and with the anti-miR. c. Western blot analyses of endogenous Foxg1 protein level in SH-SY5Y cells, upon transfection of a pre-miR-200 vector. Relative quantification is shown on the right. A down-modulation of Foxg1 is observed. d. Western blot analyses of Foxg1 protein level in extracts of the forebrain of WT, Dlx5+/− and Dlx5 −/− embryos (E15.5), normalized against signal for vinculin. Relative quantification is shown on the right. Foxg1 protein level is increased 1.5-folds and 2.5-folds, respectively, in the heterozygous and the homozygous Dlx5 mutant samples. e. Immunofluorescent staining of coronal sections of WT (top panels) or Dlx5 −/− (bottom panels) embryos, at the age E13.5, with anti-Foxg1 antibody (left panels, red staining) and anti-βIII-tubulin (right panels, green fluorescence) antibodies. The Foxg1 signal is increased n the OE of the mutant specimen, compared to the WT control.

Depletion of miR-9 and miR-200-class in zebrafish results in delayed ORN differentiation
To functionally demonstrate a role of miR-9 and miR-200-class for olfactory development, and the involvement of Foxg1 in this regulation in vivo, the zebrafish model was again used. The sequence of miR-9 and mi-200-class shows a high degree of identity between mouse and zebrafish (95% to 100%), as well as high similarity in their expression territories in early embryos ((Choi et al., 2008;Wienholds et al., 2005) and public databases). The knock-down of miR-9 in zebrafish embryos, via injection of a MO previously shown to be specific and effective (Leucht et al., 2008) (sequence in Suppl. Table III), led to a significant and dose-dependent reduction of the endogenous miR-9, relative to control-injected ones, accompanied by a 3.5-fold increase of the . Normalization is carried out relative to control samples, made = 1. (on the right) Relative abundance of z-hoxA7a and relative abundance of z-hoxA10b mRNAs in whole embryos injected with z-dlx5a MO (grey bars), to monitor developmental progression and exclude a generalized delay. endogenous z-foxg1 mRNA (Fig. 5d, e). Upon injection of the anti-miR-9 MO, only approximately 45% of the embryos were found to be CFP + (72% in the control injected), and in these we observed a clear reduction of the CFP+ signal. On the contrary, we counted a nearly normal percentage of YFP + embryos (63%), and these seldom showed reduced fluorescence (Fig. 5a-c). The majority of anti-miR-9 injected embryos displayed a normal placode organization, a normal pattern of olfactory axon fasciculation, extension and connectivity, and normal glomeruli formation. These data indicate that the depletion of miR-9 results in a delayed or absent differentiation of the OMP+ type ORN, with only a minimal effect of the Trpc2+ type neurons, and minimal consequences on axon/glomeruli organization.
To further confirm the role of miR-9 in ORN differentiation, we carried out quantitative measurements of the mRNA levels of the OE Fig. 5. Depletion of z-miR-9 in zebrafish embryos causes delayed olfactory differentiation. a. Micrographs of Trcp2::Venus (left, yellow fluorescence) and OMP::CFP (right, cyan fluorescence) zebrafish embryos injected with control (top panels) or anti-z-miR-9 (bottom panels) MOs. The control MO did not cause any significant alteration. Arrows indicate the normal axonal pathway in the control embryos. Asterisks indicate the regions of reduced fluorescence intensity. b. Whole-mount bright field micrographs of injected embryo, showing a normal embryonic morphology and growth rate. c. Proportion of embryos showing either YFP or CFP fluorescence, upon injection of control or z-dlx5a MO. A significant loss of CFP+ embryos is detected. d. Quantification of endogenous miR-9 in zebrafish embryos injected with control or anti-z-miR-9 MOs, by Real-Time qPCR. Results show efficient depletion. e. Quantification of endogenous z-foxg1 mRNAs in zebrafish embryos injected with anti-z-miR-9 MO, by Real-Time qPCR. Results show that depletion of miR-9 causes a significant increase of the z-foxg1 mRNA. f. Quantification of developmental markers, by Real-Time pPCR, in zebrafish embryos injected with anti-miR-9 MOs, relative to control MO injected embryos. Samples were collected 72 hpf. Results are shown relative to the control injected samples, made = 1. The relative abundance of z-hoxA7a and relative abundance of z-hoxA10b mRNAs were determined, to monitor progression of development and exclude a generalized delay. differentiation genes, which include z-foxg1, z-OMP-a, z-OMP-b, S100β and z-ngn1, in the dissected frontonasal piece of fish embryos injected with MOs anti-miR-9. While the z-foxg1 mRNA increased, the level of other differentiation markers decreased, with the exception of S100β (Fig. 5f). To exclude a generalized effect on development, we detected the mRNA abundance of z-hoxa-7a and z-hoxa-10b, two genes controlling general patterning and body plan, whose expression is strongly dependent upon embryonic age, and found that their abundance remains nearly constant (Fig. 5f). This provides an indication that the differentiation delay observed upon depletion of miR-9 is specific for the olfactory and anterior brain regions. miR-200a, miR-200b, miR-141 and miR-429 share the same seed sequence and likely target the same mRNAs; for this reason they are grouped in a single miR class (named miR-200-class). We depleted the miR-200 class in fish zygotes, by injecting a mix of anti-miR-200 MO previously described and found to efficiently down-modulate several miR Results show that depletion of miR-200 causes a significant increase of z-foxg1 mRNA abundance. g. Quantification of developmental markers, by Real-Time pPCR, in zebrafish embryos injected with anti-miR-200 class MOs, relative to control MO injected embryos, set = 1. The abundance of z-hoxA7a and abundance of z-hoxA10b were also determined and used as in Fig. 5. of the class-200 and to affect ORN differentiation (Choi et al., 2008). Upon injection of the anti-miR-200 MO mix, only about 24% of examined embryos turned out CFP+ (vs. 70% in the control injected), and in these we observed a clear reduction/absence of the CFP signal. The trajectory and glomeruli of CFP+ fibres could not be examined. In the same experiment, about 45% of the examined embryos were found to be YFP+ (vs. 60% in the control injected), and in these we often observed a clear reduction in the YFP signal intensity (Fig. 6a-c). In addition, we also observed defects of olfactory placode organization, alterated trajectory and reduced glomeruli formation of the YFP + fibres (Fig. 6d). Real-Time qPCR on anti-miR injected embryos demonstrated a dose dependent depletion of endogenous miR-200 class, compared to control injected ones, accompanied by a 3.5-fold increase of z-foxg1 mRNA in the same embryos (Fig. 6e, f).
Quantitative measurements of the mRNA levels of the OE differentiation genes z-foxg1, z-OMPa, z-OMPb, S100β and z-ngn1, in the frontonasal piece of fish embryos injected with MOs anti-miR-200class revealed increased levels of z-foxg1 mRNA, while the other differentiation markers decreased, with the exception of S100β (Fig. 6g). The abundance of z-hoxa-7a and z-hoxa-10b mRNAs did not greatly change, indicating that the differentiation delay observed upon depletion of miR-200-class is specific.

Depletion of miR-9 and miR-200-class in zebrafish results in altered GnRH neuron genesis and position
To determine whether miR-9 and miR-200-class play a role in GnRH neuronal differentiation and migration, we used the GnRH3:GFP transgenic zebrafish strain, in which the GFP reporter is expressed under the transcriptional control of a fragment of the z-GnRH3 promoter. The GnRH3-GFP + neurons have been widely characterized, and they consist in a population of terminal-nerve associated neurons (see Fig. 7a), thought to correspond to the mammalian hypothalamic neurons of olfactory origin (Abraham et al., 2008(Abraham et al., , 2010Wang et al., 2010). These neurons begin expression of GnRH around 24-30 hpf, and no other marker is known to specifically identify them, prior to GnRH. Their migration velocity has been estimated to be around 1 µm/hr (Y. Zhao et al., 2013), e.g. quite slow, while neurite extension is relatively faster.
We used the same MOs indicated above to deplete miR-9 and miR-200 class in GnRH3::GFP zygotes, and examined the effect on the number and position of the GFP+ neurons associated to the terminal nerves, between 36 and 72 hpf. Counting was done on higher magnification (40×) Z-stack images on fixed 72 hpf embryos, as previously indicated (Y. Zhao et al., 2013) (see Materials & Methods section, and Suppl. Fig. 5); the position was determined by confocal imaging of fixed embryos at 60 and 72 hpf; migration was assessed by time-lapse video recording on live embryos, between 36 and 52 hpf. In control embryos, we counted an average of 13 (+/− 2) GnRH3::GFP+ neurons/embryo at 72 hpf, while in miR-9 and miR-200 MO injected embryos the average number was, respectively, 5 (+/− 1) and 6 (+/− 1) (Suppl. Fig. 5). We arbitrarily established 8 neurons (−35% with respect to the control) as threshold-value to score for the presence of the phenotype "reduced number". Compared with the control injected embryos (N = 15), the depletion of miR-9 (N = 29) resulted in a higher fraction of GFP+ embryos with this phenotype (42% of cases. Fig. 7). In addition, in several cases the GFP + neurons appeared mispositioned or scattered; we termed this phenotype "scattered position" and detected this condition in 54% of the cases (Fig. 7). Similarly, the depletion of miR-200-class (N = 23) resulted in a reduced number of GFP + neurons in 22% of GFP+ embryos with the phenotype "reduced number" and 50% of the cases showing the phenotype "scattered position" (Fig. 7).
To complement the previous (static) data with live images of the migrating GnRH3 neurons, we carried out few time-lapse video recordings on untreated (4) and z-miR-200-class MO injected (4) embryos at earlier ages (36-52 hpf), in order to observe the first appearance of these neurons. During this time-window, migration of GFP + cells turned out to be quite slow (1 µm/hr), in agreement with published data (Abraham et al., 2008), thus this experiment would require a recording time incompatible with maintaining the embryos alive. Furthermore, with such a slow migration, differences are unlikely to be detected. Instead, we could easily monitor the number and position of early GFPexpressing neurons, and noted that upon depletion of miR-200 class they appear reduced in number but normally clustered. This could be taken as an indication that the "reduced number" phenotype is due to reduced neurogenesis of these cells, while the "scattered position" phenotype is acquired later and therefore possibly due to altered migration. Representative videos are provided as Supplementary Material.
Overall, these phenotypes recapitulate that described upon depletion of the Kallmann disease-genes ortholog z-kal1a/b (Whitlock et al., 2005;Yanicostas et al., 2009), nelf ) and z-fgfr1a , and provide the first evidence of a role of specific miRs in GnRH neuron development.

Exogenous expression of z-foxg1 mRNA results in reduced ORN differentiation
The forkhead transcription factor gene Foxg1 is expressed in the forebrain and OE of early mouse (Suppl. Fig. 4) and zebrafish embryos, and plays a role in maintaining olfactory progenitor cells in a "stem state" and prevent their premature differentiation (Duggan et al., 2008;Kawauchi et al., 2009b;Manuel et al., 2011). We raised the hypothesis that, in the absence of Dlx5 and reduced levels of miR-9 and -200-class, Foxg1 protein level is increased due to higher stability/translation of the Foxg1 mRNA. In turn, an excess of Foxg1 could maintain ORNs in a progenitor-like state and prevent their timely differentiation. We set forth to test this hypothesis using zebrafish embryos.
We previously verified that the depletion of miR-9 and miR-200-class in zebrafish embryos leads to higher level of z-foxg1 mRNA (no Ab efficiently recognizes the z-foxg1 protein). We injected anti-miR-9 and anti-miR200 (or control) MOs in WT zygotes, then at 48 hpf we extracted total-RNA from these and carried out Real-Time qPCR analyses. z-foxg1 mRNA level increased by three-folds when either miR-9 or miR-200-class were depleted (Figs. 5e and 6f).
Next, we in vitro transcribed the z-foxg1 mRNA, injected this in OMP::CFP + and Trpc2::Venus+ 1-cell zygotes and 72 hpf examined ORN differentiation and axon extension/trajectory. Higher doses (20 ng) of z-foxg1 mRNA caused severe developmental anomalies, including small size, microcephaly, ocular dysmorphology, etc. … (data not shown). Nevertheless, lower doses (5-7 ng) affected general embryonic development less severely, and resulted in a reduced frequency of reporter-positive embryos (22% for Venus, 28% for CFP, N = 20) and a strongly reduced expression of both CFP and Venus reporters (Fig. 8a,b). In those few cases in which fluorescence was detected, we observe only mild effects of the injected mRNA on olfactory axon trajectory and fasciculation (Fig. 8a). To confirm that reduced fluorescence intensity truly reflects altered gene expression and delayed differentiation, and to rule out generalized effects on embryonic development, we examined the relative abundance of the olfactory differentiation-related mRNAs z-OMPa, z-OMPb, z-Trpc2, z-ngn1 and z-S100β upon injection of z-foxg1 mRNA in WT zygotes, by Real-Time qPCR. The results show that all the differentiation-related mRNAs (examined) were significantly down-regulated (about 70-80% reduction), compared to the control, with the exception of z-S100β. As control for general developmental progression, the abundance of z-hoxa7a and z-hoxa10b mRNAs was mildly increased (Fig. 8d). These results indicate that higher expression of foxg1 has similar effects as Dlx5, miR-9 and -200 depletions on olfactory differentiation, in vivo.

Discussion
The results presented here indicate that loss of Dlx5 causes a downmodulation of miR-9 and of miR-200-class, which results in the overexpression of the Foxg1 protein. Starting from profile data obtained from a mouse model of Kallmann syndrome, we functionally examined this pathway in zebrafish showing that miR-9 and miR-200-class are required for normal differentiation of the ORNs, for the extension and connectivity of the olfactory axons, and for the migration of the GnRH neurons from the nasal primordium to the forebrain. Then we determined that altered levels of Foxg1 are also associated with delayed/ reduced differentiation of the ORNs, and altered olfactory axons trajectory.

The role of miR in ORN differentiation and axonal trajectory
Olfactory development is often used in studies on the cellular and molecular bases of neuronal differentiation, migration and axon targeting. Research has focused on the identification of protein-coding genes that control ORN differentiation and axon targeting, and several mutant mouse strains show specific phenotypes affecting these processes. Conversely, little information is available on the role of specific miRs and other non-coding RNAs, in the same processes. An indication that miR are needed for olfactory development comes from the results of the conditional deletion of Dicer using the OMP-Cre or the Foxg1-Cre deleter mice (Choi et al., 2008). In these models, all mature miRs are depleted, in the Cre + cells. Mice in which Dicer was depleted using OMPcre show delayed ORN differentiation and survival, while showing little or no effect on axon guidance, glomerulus formation and topography. Accordingly, GnRH neuron migration was found to be nearly normal. It is important to notice that he OMP promoter drives Cre expression in OE and VNO neurons from the age E14.5 onward, i.e. when olfactory connections are formed and GnRH neuron have reached the forebrain. Thus, this model fails to reveal the role of miRs in olfactory axon guidance and connectivity, as well as in GnRH neuron migration, events that occur at a much earlier age. Conversely, mice in which Dicer is removed with Foxg1-Cre show a severely impaired ORN differentiation and reduced cell survival (Choi et al., 2008). Since Foxg1 is expressed much earlier than OMP in placodal cells, it can be concluded that miRs are required at an early stage of development, and not late. However, later aspects of olfactory development, such as full ORN differentiation, axonal connection and GnRH migration, were not examined. Furthermore, since the disruption of Dicer prevents the maturation of all miRs, the Dicer conditional mutant models lack specificity.
The sequence-based specificity of miR function is the key to begin decipher the molecular pathway underlying olfactory development. The most abundant miRs expressed in the developing mouse OE are: the miR-200-class (-200a, -200b, -200c, -141 and -429), miR-199, miR-152, miR-214, miR-205, miR-183, miR-182 and miR-96 (Choi et al., 2008). However, when Dlx5 is absent we do not see a generalized reduction of miR expression, rather the downmodulation of specific ones. This might explain the fact that in the OE of Dlx5 −/− embryos we see a relatively mild differentiation phenotype compared to that of the Dicer flox ; Foxg1::Cre mice or the anti-miR-200-class morphants (Choi et al., 2008). In this work we define the role of miR-9 and miR-200-class in the development of the olfactory system, with functions ranging from ORN differentiation to axon guidance, glomerulus formation and GnRH neuron migration. We also show that miR-9 and miR-200-class target (amongst others) the foxg1 mRNA, through which they likely exert their functions. Previous results in which zebrafish embryos were injected with anti-miR-200 class MOs found a delayed ORN differentiation, but axonal organization and GnRH neuron migration was not assessed (Choi et al., 2008). Examining olfactory development more thoroughly we now can implicate the miR-9 and miR-200-class networks in a more complex phenotype reminiscent of the Kallmann syndrome (see below).

miR-9 and miR-200 mediate the Dlx5-Foxg1 cascade
Foxg1 is a winged helix transcription factor, member of the fork-head family, expressed in the early developing forebrain and peripheral olfactory system. In the forebrain, the loss of Foxg1 causes premature lengthening of progenitor cell cycles and increased neurogenic divisions, leading to a severe brain hypoplasia (Manuel et al., 2011). It has been proposed that these proliferation defects could be a secondarily caused by altered expression of FGF8 and BMPs in the forebrain of Foxg1 null mutants. By the generation and analysis of Foxg1 +/+ ↔ Foxg1 −/− chimaeras, Foxg1 appears to control forebrain progenitor proliferation cell autonomously (Manuel et al., 2011).
Foxg1 is also expressed in early progenitor cells of the OP, and later the expression becomes restricted to the ventro-lateral OE and VNO, in cells within the basal compartment of the OE, the location where OE stem and progenitor cells are known to reside (Kawauchi et al., 2009b). In early Foxg1 −/− embryos a small number of progenitors are initially specified but show reduced proliferation and differentiation. Older Foxg1 −/− embryos show no recognizable OE, VNO and OB (Duggan et al., 2008). BrdU pulse-chase labelling of Sox2-expressing stem cells indicated that these cells are delayed or halted in their development, in the absence of Foxg1 (Kawauchi et al., 2009a,b), suggesting that the proliferation and/or subsequent differentiation of Sox2+ stem cells in the OE is regulated by Foxg1. Thus Foxg1 is required for the development of the central and peripheral olfactory system, to maintain progenitors in a proliferative state (Duggan et al., 2008).
The 3′ UTR of tetrapod and zebrafish Foxg1 mRNAs hosts miR-9 and miR-200 target sequences. Indeed Foxg1 has been experimentally shown to be negatively regulated by miR-9. The mouse miR-9 targets Foxg1 mRNAs for proper generation of Cajal-Retzius neurons in the medial pallium (Shibata et al., 2008). miR-9 expression is medio-laterally graded, being most intense in the cortical hem; it contrasts with the Foxg1 expression in a reciprocal gradient. miR-9 over-expression in developing forebrain at E11.5 resulted in ectopic Reelin + cells over the cortex beyond the marginal zone, while conversely the inhibition of endogenous miR-9 function caused the regression of Wnt3a positive cortical hem and reduction of Reelin+, p73+ and NeuroD1+ cells (Shibata . A significant loss of both YFP and CFP+ embryos is detected. Asterisks indicate p b 0.005. c. Whole-mount bright field micrographs of injected embryo, showing slightly altered embryonic morphology and nearly normal growth rate. d. (left) Quantification of endogenous z-OMPa, z-Trcp2, z-ngn1 and z-S100β mRNAs in embryos injected with z-foxg1 mRNA, by qPCR. Values are normalized against the control injection, set = 1. The over-expression of foxg1 causes a significant decrease of these mRNAs, with the exception of z-S100β. (right) Quantifications of z-hoxA7a and z-hoxA10b mRNAs in WT embryos injected with anti-z-foxg1 mRNA, relative to control injected embryos, used as in Fig. 5. Quantification of z-foxg1 mRNA was used to verify the injection. et al., 2008). Thus, a fine modulation over the level of Foxg1 protein is required for efficient and timely differentiation of specific subset of cortical neurons.
Here we show that mouse and fish foxg1 mRNA is a target of miR-9 and miR-200 class, both of which are down-modulated in the Dlx5 null embryonic OE. We also show that Dlx5 promotes expression of miR-9 and miR-200 class, thereby tends to repress Foxg1 protein translation. Hence, the most likely scenario is that in the absence of Dlx5 the Foxg1 protein persists for a longer time or at increased level, preventing progenitor cells from efficiently exiting the progenitor state and initiate differentiation. This possibility is clearly consistent with the results reported by Shibata et al. (2008), in which they show that the depletion of miR-9 resulted in abnormally high levels of Foxg1 proteins, and this caused a delayed differentiation of the Cajal-Retzius neurons in the cortex. It has also been shown that miR-200 represses neural induction of human embryonic stem cells, via modulation of Pax6 and Zeb transcription factors (Du et al., 2013). We may propose that in neural stem cells, this regulation is carried out also via Foxg1. Our results reinforce the notion that the expression level of master transcription factors (such as Dlx5 and Foxg1 for ORN) must be precisely regulated in either directions to assure normal development.
Finally, Foxg1 and Emx2 have been shown to be key transcription factors that control the switch between gliogenesis and neuronogenesis, in stem cells (Brancaccio et al., 2010). Since Emx2 and Foxg1 are both expressed in the early olfactory progenitors, this observation is intriguing, as it is conceivable that an increased Foxg1 expression may imbalance this choice and lead to a retarded neuronogenesis, at the advantage of the generation of supporting cells, or other non-neuronal cell types present in the OE. We will address this issue in future works.

miRs, coding RNAs and the Kallmann syndrome
Part and current research on KS/nCHH has strongly focused on the identification of protein-coding genes. Conversely, there is little or no information as to weather miRs (and other non-coding RNAs) might be involved in the molecular pathogenesis of this disorders. One indications comes from the SEMA3A gene, which causes KS in human and a KS-like phenotype in mice (Cariboni et al., 2011;Hanchate et al., 2012); in retinal neurons the SEMA3A mRNA is regulated by miR-124 and this regulation modulates the guidance response of these cells to the Sema3a signal (Baudet et al., 2011). Another indication comes from a study in zebrafish, showing a role of miR-200-class for olfactory development (Choi et al., 2008). Recently, miR-200b and miR-429 have been linked to pituitary endocrine functions controlling ovulation and fertility in female mice (Hasuwa et al., 2013). Interestingly, we show here that miRs of the 200-class are involved in olfactory axons extension and GnRH neuron migration, thus suggesting that these miRs are indeed essential for male and female sexual maturation. Finally, the lin28/let-7 miR system has been implicated in the maturation of the post-natal hypothalamus and induction of puberty, in a mouse model of CHH (Gaytan et al., 2013;Sangiao-Alvarellos et al., 2013).
In spite of this scattered knowledge, studies specifically addressing the expression and function of miR in GnRH neurons are needed to dissect their role in the genesis, migration and maturation of this specific cell type. At present, a systematic study to identify specific miRs relevant for early GnRH neuron development, in vitro or in vivo, has not been reported, as yet. Thus, our results provide the first evidence of the participation of miR-9 and miR-200-class in these early events. We further link two transcription factors with the action of these miRs.
A pathway is emerging, yet to be fully unravelled. The possibility that Dlx5 may regulate aspects of the development of GnRH neurons was suggested by the finding that Dlx and Msx binding sites are present in the mammalian GnRH promoter (Givens et al., 2005). Furthermore, Dlx5 is expressed in murine embryonic GnRH neurons, and in Dlx5 −/− mice the GnRH neurons fail to properly migrate (Levi et al., 2003;Merlo et al., 2007). These and the present results, however, do not fully demonstrate that Dlx5 is needed for GnRH neuron migration cell autonomously; rather the observed defects could be consequent of impaired olfactory axon elongation and/or connectivity, hence non-cell autonomous. These possibilities will have examined in further studies.