The RNA-binding protein MARF1 promotes cortical neurogenesis through its RNase activity domain

Cortical neurogenesis is a fundamental process of brain development that is spatiotemporally regulated by both intrinsic and extrinsic cues. Although recent evidence has highlighted the significance of transcription factors in cortical neurogenesis, little is known regarding the role of RNA-binding proteins (RBPs) in the post-transcriptional regulation of cortical neurogenesis. Here, we report that meiosis arrest female 1 (MARF1) is an RBP that is expressed during neuronal differentiation. Cortical neurons expressed the somatic form of MARF1 (sMARF1) but not the oocyte form (oMARF1). sMARF1 was enriched in embryonic brains, and its expression level decreased as brain development progressed. Overexpression of sMARF1 in E12.5 neuronal progenitor cells promoted neuronal differentiation, whereas sMARF1 knockdown decreased neuronal progenitor differentiation in vitro. We also examined the function of sMARF1 in vivo using an in utero electroporation technique. Overexpression of sMARF1 increased neuronal differentiation, whereas knockdown of sMARF1 inhibited differentiation in vivo. Moreover, using an RNase domain deletion mutant of sMARF1, we showed that the RNase domain is required for the effects of sMARF1 on cortical neurogenesis in vitro. Our results further elucidate the mechanisms of post-transcriptional regulation of cortical neurogenesis by RBPs.

mRNA is expressed in a limited fashion in the developing mouse brain 18 . This finding led us to hypothesize that MARF1 has an essential role in cortical development.
Here, we showed that sMARF1 is expressed in the developing cortex and that its expression increases with neuronal differentiation. Our gain-of-function and loss-of-function experiments revealed that sMARF1 promotes neuronal differentiation both in vivo and in vitro.

Results
Somatic form of Marf1 is expressed in the developing cortex. We previously found that Marf1 mRNA is expressed in the embryonic and postnatal brain 18 . In addition, Marf1 occurs as two isoforms: sMarf1 and oMarf1 16 . To determine which isoform of Marf1 is expressed in cortical neurons, we performed RT-PCR analysis of embryonic primary neurons using ovary tissue as a positive control. E15.5 cortical neurons expressed sMARF1 but not oMARF1 (Fig. 1a). To examine the sMARF1 expression profile during various developmental stages, we performed western blot analysis of cortical brain samples obtained from mice at different stages of development (E10.5, E12.5, E14.5, E17.5, P1, P7, P14, P30, and adult) using a MARF1-specific antibody 16 . We found that sMARF1 expression was prominent from E14.5 to E17.5 and gradually decreased in the adult stages (Fig. 1b). To determine whether sMarf1 expression levels changed depending on the cellular differentiation state, we performed an in vitro differentiation assay using primary E12.5 cortical progenitor cells 18 . We first confirmed the differentiation time frame of progenitor cells to neurons by differentiation assays (Supplementary Fig. 1a and b). We observed that sMarf1 mRNA increased gradually as cortical progenitor cells differentiated to neurons in vitro, but its expression finally declined after 6 days (Fig. 1c). To confirm these findings in vivo, we performed immunostaining of sMARF1 with MAP2 (neuronal marker), Tbr2 (basal progenitor marker) or Pax6 (radial glial marker) in E12.5, E14.5, E16.5 and P0 mouse cortices. The expression of sMARF1 was detected in E12.5, E14.5, E16.5, and P0 cortices ( Fig. 1d and Supplementary Figs 3, 4, 5 and 6). Pax6 + radial glia, Tbr2 + basal progenitors and MAP2 + neurons co-expressed sMARF1 in the E14.5 cortex (Fig. 1e). However, after reaching to the highest expression at approximately E16.5, MAP2 + neuronal expression decreased at P0, as observed in the in vitro expression analysis. These results suggest that sMARF1 is involved in early to mid-term brain development presumably during progenitor differentiation to neurons.

sMARF1 promotes differentiation of cortical progenitor cells in vitro.
To determine whether sMARF1 controls the proliferation and differentiation of cortical progenitors, we performed in vitro gainand loss-of-function experiments using E12.5 cortical progenitor cells, as previously described (Fig. 2a) 19 . For the gain-of-function experiments, we transfected cells with an sMarf1 overexpression construct, and for the loss-of-function experiments, we transfected cells with small-hairpin RNA (shRNA) vectors against sMarf1. We first confirmed the knockdown efficiency of sMarf1 in cortical neurons (Fig. 2d). Overexpression of sMarf1 in E12.5 cortical progenitor cells increased the percentage of Tuj1 + neurons compared with the control cells ( Fig. 2b and c). Next, knockdown of sMarf1 increased the percentage of Ki67 + progenitor cells compared with the control cells ( Fig. 2e and f). In addition, knockdown of sMARF1 increased the percentage of Pax6 + radial glia but decreased Tuj1 + neurons compared with the controls (Fig. 2g,h,i and j). To exclude the possibilities of shRNA vector toxicity and the effect of sMARF1 on the cell survival, we assessed the effect of sMARF1 on the cell death of progenitors and neurons by cell death assays. Transfecting scrambled shRNA vector into progenitors (scr), knockdown of sMarf1 (sMARF1 shRNA) or overexpression of sMARF1 (sMARF1o/e) did not significantly affect progenitor or neuronal survival (Fig. 2k,l and m). These results indicate that sMARF1 promotes the differentiation of cortical progenitor cells but does not affect cell survival. sMARF1 regulates the neuronal differentiation of radial glia in the embryonic cortex in vivo. To confirm the function of sMARF1 in vivo, we conducted in utero electroporation experiments. First, we introduced sMarf1 expression vectors containing an internal ribosome entry site (IRES)-green fluorescent protein (GFP) expression cassette into E13.5 mouse cortices. Then, the cortices were dissected at E16.5, sectioned and stained for sMARF1 to confirm its overexpression ( Supplementary Fig. 2a). To analyze differentiation of radial glia to basal progenitors (BPs), we performed immunostaining for the BP marker Tbr2 as well as GFP (Fig. 3a). sMarf1 overexpression increased the ratio of GFP-labeled Tbr2 + cells within the ventricular zone (VZ) and the subventricular zone (SVZ) (VZ/SVZ) (Fig. 3b). Next, we examined the longer-term effect of sMARF1 expression on cortical progenitor cells. To this end, we performed in utero electroporation at E13.5 and then immunostained for the upper-layer neuronal marker Satb2 as well as GFP at P2 and 7 days after electroporation (Fig. 3c). sMarf1 overexpression increased the percentage of GFP and Satb2 double-positive neurons compared with the control mice (Fig. 3d). These results indicate that sMARF1 promotes cortical radial glial differentiation to BPs and upper-layer neurons. For loss-of-function experiments in vivo, we conducted in utero electroporation of our shRNA vector as well as that of a nuclear GFP expression vector into E13.5 mouse cortices. First, the cortices were immunostained for GFP and MARF1 at E16.5, and we confirmed protein knockdown by shRNA vector in the brains ( Supplementary Fig. 2b). Then, we immunostained brain sections for GFP and Ki67 or Pax6 ( Fig. 3e and g). sMarf1 knockdown increased the ratio of GFP-labeled Ki67 + progenitor cells and Pax6 + radial glia within the VZ/SVZ compared with the control mice ( Fig. 3f and h). To analyze the longer-term effect of sMarf1 knockdown, we electroporated the shRNA vector and nuclear GFP into E14.5 mouse cortices and then immunostained for Satb2 and GFP at P0 (Fig. 3i). sMarf1 knockdown decreased the ratio of GFP-labeled Satb2 + upper-layer neurons compared with the control (Fig. 3j). Moreover, to determine whether the gain-and loss-of-function of sMARF1 alters the neuronal laminar distribution, we analyzed P0 mouse cortices electroporated with the sMARF1 expression vector or shRNA vector at E14.5 by immunostaining for DAPI, GFP and Satb2 ( Fig. 3i and Supplementary  Fig. 7). However, gain-or loss-of-sMARF1 had no effect on laminar distribution compared with the control.  The full-length gel is presented in Supplementary Fig. 8a. (b) Western blotting for sMARF1 in mouse cortical lysates at the embryonic, postnatal, and adult ages. β-Actin was used as a loading control. Lysate of HEK293 cells transfected with a sMarf1 expression vector was used as a positive control. The same full-length image is presented in Supplementary Fig. 8b, because membrane was cut before incubating with antibodies. (c) Relative expression levels of Marf1 mRNA from E12.5 cortical progenitors cultured for 2, 3, 4 and 6 days (2, 3, 4 and 6 days in vitro (div)). *p < 0.05 ***p < 0.001 (n = 4, one-way ANOVA, Tukey-Kramer test).   difference between oMARF1 and sMARF1 occurs outside the NYN/PIN-like domain. The amino acid sequences of the NYN/PIN-like domains are same between both isoforms without frameshifts. These observations suggest that protein function may be conserved between these two isoforms.
To determine whether the NYN/PIN-like domain is necessary for neuronal differentiation, we deleted the NYN/PIN-like domain (ΔNYN) in an sMarf1 expression vector ( Fig. 4a and b). For analysis of the effect of this expression vector on cortical progenitor differentiation, we overexpressed full-length (FL) sMARF1 and ΔNYN sMARF1 ( Fig. 4c and e) in cortical progenitor cells in vitro. Although FL sMARF1 overexpression promoted the differentiation and inhibited the proliferation of progenitor cells, ΔNYN sMARF1 overexpression did not affect either differentiation or proliferation ( Fig. 4d and f). Therefore, these results suggest that sMARF1 may induce neuronal differentiation through its NYN/PIN-like domain in vitro.

Discussion
In the present study, we found that cortical neurons expressed the somatic isoform of Marf1 (sMarf1) but not the oocyte form (oMarf1). The expression of sMARF1 protein peaked between E14.5 and E17.5, after which it was downregulated in later postnatal stages and adult tissues. sMARF1 levels increased and finally declined to control levels as cells differentiated from cortical progenitors to mature neurons in vitro. In the E14.5 brain, not only post-mitotic neurons, but also progenitors, highly expressed sMARF1. Overexpression of sMARF1 promoted the differentiation of cortical progenitor cells to neurons, whereas knockdown of sMARF1 increased the proliferation of progenitor cells without influencing progenitor and neural survival. Transfection of sMARF1 lacking the domain containing RNase activity completely abolished the pro-differentiation activity of sMARF1. These results strongly suggest that sMARF1 acts as a pro-neural protein through its N-terminal RNase activity domain.
The sMARF1 protein expression is prominent from early to mid-term brain development during progenitor differentiation to neurons. Importantly, sMARF1 expression is decreased in the fully differentiated neurons in vitro and in vivo at later stages. This expression change may be consistent with the hypothesis that sMARF1 is specifically required for neuronal differentiation. As shown in Fig. 2k,l and m, Fig. 3i and Supplementary Fig. 7, cell survival or neuronal migration was not affected by sMARF1 knockdown and overexpression. These results also support the idea that sMARF1 specifically regulates progenitor differentiation.
As shown in Fig. 3i, although the Satb2 + neuronal differentiation rate was decreased, laminar distribution was not changed by MARF1 knockdown. There are probably two reasons for this phenomenon. One reason is the timing of in utero electroporation. At E14.5, only Satb2 + upper layer neurons are derived from radial glia and astrocyte production follows a few days later 1 . Therefore, during this transition phase from neurogenesis to gliogenesis, it seems that sMARF1 modulates only Satb2 + upper layer neuronal numbers, not deep layer neurons. Another reason may be that sMARF1 specifically controls neurogenesis but not neural migration. Therefore, even though neuronal differentiation by electroporation of the sMARF1 shRNA vector delays, the laminar distribution of electroporated cells may not be affected. Subsequent glial differentiation may increase from the increased neural stem pool by knockdown of sMARF1 as previously described 20 . The change of Satb2 + cell number in the superficial layer may contribute to the abnormalities of callosal projection to the contralateral hemisphere 1 , although evidence is lacking. Moreover, it is also likely that sMARF1 is involved in the fate specification of cortical neurogenesis. Further experiments such as in utero electroporation at various earlier time points (eg. E10.5, 12.5) and comparison of the rate of neurogenesis of deep layer neurons and superficial neurons would be necessary to clarify the sMARF1 function in brain development.
What molecules could be targeted by sMARF1? Recent advances in sequencing technology, such as the photoactivatable ribonucleoside-enhanced crosslinking and immunoprecipitation (PAR-CLIP) method 21 , enable systematic identification of the targets of RBPs. Conventional RNA immunoprecipitation (RIP) can also identify stably bound RNA targets 22 . Our findings suggest that sMarf1 expression increases as neuronal differentiation progresses (Fig. 1c). Thus, the target molecule of sMARF1 could be a molecule that progenitor cells but not neurons require for development. These molecules could be stem cell maintenance factors, such as Sox2 and Pax6, or proneural genes that only immature neurons require, such as NeuroD and Ascl family 23 . Further examination is required using the abovementioned methods. MARF1 belongs to the NYN/PIN-like domain, RNase domain family, but the function of the NYN/PIN-like domain is still unclear 17 . Therefore, to elucidate the role of MARF1, we compared the NYN/PIN-like domain of MARF1 with other family members.
Another well-characterized NYN/PIN-like domain-containing protein is Zc3h12a (Regnase-1) 24 . Regnase-1 is a PIN domain-harboring RNase that is critical for preventing severe autoimmune inflammatory disease in mice by destabilizing inflammation-related mRNAs, including Il6 and Reg1 itself 24,25 . The PIN domain shares characteristics with the NYN domain 17 , and Zc3h12a is now categorized as a member of the Zc3h12 gene family, which harbors the NYN domain 26 . In addition, RDE-8 in Caenorhabditis elegans encodes a Zc3h12a-like endoribonuclease required for RNA degradation, which functions as a mechanism of RNA interference 27 . Regnase-1 overexpression does not rescue the phenotype associated with the disruption of RDE-8 in Caenorhabditis elegans 27 . These results suggest that NYN/PIN-like domains are diverse. Based on information from the protein database InterPro (http://www.ebi.ac.uk/interpro/), the NYN/PIN-like domain also shows significant similarity with the PIN domain of the N-termini of 5′-3′-exonucleases, such as SMG5/6. In eukaryotes, PIN domains are ribonucleases involved in nonsense-mediated mRNA decay 28 and in the processing of 18 S ribosomal RNA 29 . Taken together, the results suggest that the N-terminal sequence of MARF1 likely displays similar RNase activity or a function in mRNA decay similar to Zc3h12a, RDE-8, and SMG5/6.
Although our results indicate that sMARF1 regulates differentiation through its RNase activity, it is possible that our deletion mutant was inactivated by protein structure changes. Therefore, other experiments, such as introducing point mutations in critical amino acids within the NYN/PIN-like domain, are necessary to confirm our observations. The NYN/PIN-like domain shares a common set of four amino acids with previously characterized nuclease domains 30 . The common amino acids bind to Mg 2+ ions and are essential for nuclease activity 30 . Recent studies have demonstrated a relationship between abnormal cortical neurogenesis and neurodevelopmental disorders. Rodents lacking the neurogenesis-related genes Pax6 or Tbr2 exhibit impaired neurogenesis and behavioral abnormalities associated with an autistic phenotype 31 and hyperactivity 32 . Overproduction of upper-layer neurons induced by drug injection leads to autism-like behaviors in mice 33 . These findings support the hypothesis that normal cortical neurogenesis is critical for the development of the intact brain. In addition, copy number variation (CNV) is a potential risk factor for neurodevelopmental disorders 34 . Chromosome 16p13.11 is known to have CNVs associated with neurodevelopmental disorders [34][35][36][37] . We found that miR-484, which is encoded in the core locus of the chromosome 16p13.11 CNV, regulates cortical neurogenesis 18 . Marf1 is also located on chromosome 16p13.11 contiguous to miR-484 18 . Taken together, these findings suggest that sMARF1 may have synergistic effects with miR-484 on neurogenesis as candidate genes of 16p13.11 CNV. Therefore, future studies of sMARF1 will hopefully uncover not only the detailed mechanisms of sMARF1 functions but also the relationships between sMARF1 and neurodevelopmental disorders.

Materials and Methods
This study was approved by the institutional committee of Osaka University.
Animals. Slc-ICR mice were used in this study, and all mice were purchased from SLC Japan. Mice were euthanized with an overdose of a mixture of 0.5 mg/ml Vetorphale (Meiji Seika Pharma), 0.4 mg/ml Dormicum (Roche), and 0.03 mg/ml Domitor (Orion Pharma) by peritoneal injection. All procedures complied with the Osaka University Medical School Guidelines for the Care and Use of Laboratory Animals.

RT-PCR.
Cells and tissues were homogenized in TRIzol reagent (Thermo Fisher Scientific). Isolated total RNA was purified with an RNeasy Micro kit (QIAGEN). The complementary DNA (cDNA) was synthesized with a High Capacity cDNA Reverse Transcription kit (Applied Biosystems). A no template reaction was used as a negative control. Mouse ovary tissue containing cumulus cells and oocytes was used as a positive control. Primers for Marf1 exon3 were as follows: Forward: 5′-TTCACCAAGATAATGATGCTAAGC-3′, Reverse: 5′-TTTTCCATGCCTTTTGTTCC-3′.

Real-time PCR analysis. TaqMan real-time PCR analysis was conducted using TaqMan Universal Master
Mix II (Thermo Fisher Scientific) with specific probe mixtures for each gene (Marf1: Mm00463593_m1, glyceraldehyde-3-phosphate dehydrogenase (Gapdh): Mm99999915_g1). The expression of each gene was normalized to Gapdh. The reaction and the subsequent quantification were conducted using QuantStudio 7 Flex Real-Time PCR System (Applied Biosystems).

Immunohistochemistry signal amplification and same animal species antigen. The TSA Plus
Fluorescein System (PerkinElmer) was used to amplify the immunohistochemistry signal. After fixing for 15 min in 4% paraformaldehyde in 0.1 M PB, sections were incubated in 0.3% H 2 O 2 in methanol to block endogenous peroxidase activity for 30 min at room temperature. Then, the sections were incubated in sodium citrate buffer (10 mM sodium citrate, 0.05% Tween-20, pH 6.0) for 30 min at 70 °C in an autoclave for antigen retrieval. Thereafter, the samples were immediately washed with PBS(-) and incubated for 1 h in blocking buffer containing 10% goat serum, 5% BSA and 0.3% Triton X-100 at RT. Rabbit anti-MARF1 (1:100, from Dr. Eppig 16 ) was used as the primary antibody. The secondary antibody reaction was performed for 30 min at room temperature with horseradish peroxidase-conjugated antibody (1:2000, Cell Signaling Technology) in blocking buffer. The sections were washed in TNT buffer (0.1 M Tris-HCl pH 7.5, 0.15 M NaCl, 0.05% Tween-20) three times for 5 min each. After the TNT buffer was removed, the sections were incubated in TSA Amplification Reagents for 7 to 10 min. Then, the sections were washed in TNT buffer three times for 5 min each. Finally, the sections were incubated in the sodium citrate buffer for 15 min at 95 °C in the autoclave to inactivate the rabbit anti-MARF1 antibody. Washed samples were incubated in the blocking buffer for 1 h at room temperature. The samples were used for standard immunohistochemistry.
Western blotting. All cells and tissues were lysed in radioimmunoprecipitation assay buffer (50 mM Tris-HCl pH 8, 150 mM NaCl, 1% NP-40, 0.1% sodium dodecyl sulfate (SDS), 0.5% sodium deoxycholate) containing protease inhibitors (Roche). Lysates were loaded onto 8% SDS-polyacrylamide gels and were transferred to polyvinylidene fluoride membranes. Membranes were blocked for 1 h in 5% skim milk in PBS-T (0.05% Tween-20). The primary antibodies (rabbit anti-MARF1 (1:3000, from Dr. Eppig 16 ), mouse anti-V5 (1:2000, Thermo Fisher Scientific), and rabbit anti-β-Actin (1:3000, Cell Signaling Technology)) were diluted in blocking buffer to the indicated concentrations and were incubated with membranes overnight at 4 °C. The secondary antibody reaction was performed for 1 h at room temperature with horseradish peroxidase-conjugated antibody (1:2000, Cell Signaling Technology) in blocking buffer. Enhanced chemiluminescence (GE Healthcare) was used for signal detection. All images were obtained using an LAS-3000 image analyzer (Fujifilm). Lysate obtained from HEK293 cells transfected with an sMARF1 expression vector was used as the positive control for the anti-MARF1 antibody. Empty pCAGIG vector-transfected E15.5 neuron lysates were used as a negative control for the anti-V5 antibody.
Cortical progenitor cell culture and transfection. Cortical progenitor cells were isolated from E12.5 mice. Isolated cortices were transferred to Neurobasal medium (Thermo Fisher Scientific) containing 40 ng/ ml fibroblast growth factor 2 (Promega), 2% B27 (Thermo Fisher Scientific), 120 mg/ml penicillin, 200 mg/ml streptomycin sulfate, and 600 mg/ml glutamine. The cells were plated on poly-D-lysine-coated (Sigma) and laminin-coated (BD) 4-well chamber slides at a density of 4.0 × 10 5 cells/well and were maintained at 37 °C in the presence of 5% CO 2 . For transfections, medium was mixed with 2 μg of plasmid DNA or shRNA vector and 1.5 μl of FuGENE HD (Promega) in 100 μl per well for 15 min at room temperature, after which the transfection mixture was applied to the chamber slides.
In utero electroporation. In utero electroporation was performed using a square-wave electroporator (CUY21SC, NEPAGENE) that delivered five 50-ms pulses of 30 V with 950-ms intervals, as previously described 18,19 . E13.5 or E14.5 mice were injected with 6.0 μg of plasmid DNA in 2 μl of diluted water. For sMARF1 shRNA vector injection, we mixed nuclear EGFP with the solution at a 1:3 ratio. Trypan blue (1%, Gibco) was co-injected as a tracer.
Cell counting and quantification. For the in vitro experiments, 100 GFP + transfected cells/well in 4-well chamber slide were counted as n = 1. Three to four (n = 3-4) independent experiments were performed. In cell death assay, we counted the cells having fragmented or condensed nucleus as apoptotic cells, as described 18 . For the in vivo experiments, three sets of 100 GFP + electroporated cells/brain section were counted as n = 1. n = 3 for individual treated group. Three control embryos and three conditional embryos were selected using similar electroporation conditions and were analyzed for each of the experiments.