Gene activation of SMN by selective disruption of lncRNA-mediated recruitment of PRC2 for the treatment of spinal muscular atrophy

Analysis of publicly available chromatin immunoprecipitation (ChIP) sequencing data from the ENCODE consortium and the Broad Institute (genome.ucsc.edu/) (14, 15) suggests that PRC2 is associated with SMN2 in different cell types to varying degrees, most notably with the HepG2 cells (Fig. 1A). In addition, ChIP sequencing data from the NIH Roadmap Epigenome Consortium (16) suggests that H3K27me3, the hallmark of PRC2 activity, is associated with SMN2 in human fetal brain (SI Appendix, Fig. S1). To determine whether disruption of PRC2 activity could lead to increases in SMN2 expression, EZH1 and EZH2 mRNAs were knocked down in the SMA fibroblast cell line, GM09677, using antisense oligonucleotides (ASOs) designed for RNaseH-mediated degradation. Two days after transfection of both EZH1 and EZH2 gapmers, their respective mRNA levels were significantly decreased by ∼80% each and were associated with an increase in SMN-FL mRNA as shown by reverse transcription quantitative PCR (RT-qPCR) (Fig.1B). We further analyzed the SMN1 and SMN2 loci (from here on collectively termed “SMN locus”) for chromatin changes upon EZH1/EZH2 knockdown by performing ChIP. Because SMN1 and SMN2 have >99% sequence identity (27,890 of 27,924 base pair match), it is not possible to distinguish between the two chromosomal locations here. We observed decreased association of EZH2 as well as decreased H3K27me3 levels at the locus, without any changes in total H3 (Fig. 1C). This suggests that PRC2 directly regulates the expression of SMN in fibroblasts and potentially other cell types.

Detailed analysis of RNA immunoprecipitation-sequencing (RIP-seq) datasets revealed a previously undescribed PRC2-interacting antisense RNA within the mouse Smn locus (17). Here, we investigated whether the antisense transcript exists in humans and may have a role in PRC2-mediated SMN repression. Next-generation RNA sequencing revealed a lncRNA, which we call SMN-Antisense 1 (SMN-AS1), that is transcribed from the SMN loci (Fig. 2A). Given the high sequence identity between the SMN1 and SMN2 loci, we predict the lncRNA to be transcribed from both loci. As expected, SMN-AS1 was observed in both SMN1- and SMN2-mutated cell lines (Fig. 2B). Furthermore, SMN2 copy number was determined by qPCR for carrier and diseased cell lines (18), and we independently determined the relative expression of SMN-AS1 and observed a direct correlation (Fig. 2B). Northern blot analysis of human fetal brain and adult lung tissues revealed that SMN-AS1 is up to 10 kb long, is heterogeneous in size, and has differential expression between the two tissue types (Fig. 2C). To confirm the specificity of the SMN-AS probe, we turned to a humanized SMA mouse model carrying two copies of the human SMN2 genomic locus (5025 strain) (19). Comparing the brain tissues from wild-type and 5025 mice, we observed a similar set of transcripts in the SMN2-harboring transgenic mice as detected in the human fetal brain (Fig. 2C). By RT-qPCR, we detected SMN-AS1 in patient cell lines, and the level of expression correlated with SMN2 copy number (as determined by ref. 18) (Fig. 2B). In addition, we found that SMN2 mRNA and SMN-AS1 expression is highly correlated, with highest levels in CNS tissues (Fig. 2D). Finally, strand-specific single-molecule RNA-fluorescent in situ hybridization (RNA-FISH) detected expression of SMN-AS1 colocalized with the pre-mRNA transcript at the SMN locus (Fig. 2E) in all SMA fibroblasts (52 out of 52) that were imaged in three independent experiments, suggesting that SMN-AS1 may function in cis. We observed primarily one colocalized foci in the SMA fibroblasts and a wild-type fibroblast line. Although the SMN1 and SMN2 genes are separated linearly by 900 kb, their physical distance is unknown in vivo. Our data suggest that they are close enough in proximity in fibroblasts, such that single molecule RNA-FISH would not be able to discern SMN1 from SMN2-derived signals. Together, these data demonstrate the presence of an antisense transcript in the SMN1 and SMN2 loci.

To investigate the role of SMN-AS1 in the PRC2-mediated epigenetic regulation of the SMN2 locus, we performed native RIP (nRIP) using an antibody against the PRC2 subunit SUZ12, followed by RT-qPCR with two distinct probe sets directed to different regions of SMN-AS1. RIP-qPCR showed that SMN-AS1 is strongly associated with PRC2 in SMA fibroblasts (Fig. 2F). The association was stronger than, or comparable to, that of well-established PRC2-interacting lncRNAs including TUG1 (20) and ANRIL (21). Additionally, PRC2 did not associate with highly expressed negative control transcripts such as GAPDH and RPL19. Similar results were observed with the nRIP for EZH2 (SI Appendix, Fig. S1), further supporting the association of SMN-AS1 with PRC2. Because nRIP identifies both direct and indirect interactions, we next performed RNA electrophoretic mobility shift assays (RNA EMSAs), which specifically detect direct interactions. Using a 441-nt RNA containing the PRC2-interacting region of SMN-AS1 (SMN-AS1, PRC2-binding region), as identified by RIP-seq. (17), we observed that purified recombinant human PRC2 (EED/SUZ12/EZH2) specifically changed its migration (Fig. 2G). Binding was concentration-dependent and was as robust as that of the 434-nt RepA RNA, a conserved domain of XIST RNA that is a well-documented PRC2-interacting lncRNA (17, 22, 23). Dissociation constants (K_d) of both transcripts were estimated to be 350–360 nM, suggesting that the association of this lncRNA with PRC2 at this site is comparable to the RepA domain of XIST. As specificity controls, we observed a low level of background binding to a non–PRC2-interacting 441-nt region of the SMN-AS1 transcript (SMN-AS1, nonbinding region) and to another nonspecific mRNA of similar length, maltose-binding protein from Escherichia coli (22). These data lead us to conclude that SMN-AS1 lncRNA interacts directly and specifically with PRC2.

To investigate the effect of disrupting PRC2:SMN-AS1 interactions, we designed chemically modified ASOs targeting the PRC2-binding region of the lncRNA for hybridization via Watson–Crick complementarity pairing. Depending on the arrangement of DNA and locked nucleic acid (LNA)-modified nucleotides, such base pairing can lead to either RNaseH-mediated degradation of target RNAs or hindering of the interaction between target RNAs and their binding partners. For RNaseH-mediated degradation, a “gapmer”-formatted ASO composed of a central DNA segment greater than 6 nucleotides (i.e., gap) flanked by 2–4 LNA-modified nucleotides is required. Such gapmer ASOs were used for knockdown of EZH1 and EZH2 in earlier experiments (Fig. 1C). In contrast to the gapmer arrangement, a “mixmer”-formatted oligo lacks the central DNA component by the introduction of interspersed chemically modified nucleotides. It does not support the RNaseH-mediated degradation but rather functions as a steric blocker (24). We generated mixmer oligos consisting of LNA interspersed with 2’-O-methyl nucleotides for high-affinity binding to SMN-AS1. Oligos were designed to target regions enriched from EZH2-associated RNAs by RIP followed by next-generation RNA sequencing (17).

Screening of mixmers led us to focus on one efficacious mixmer, RN-0005 (Fig. 3A). Transfecting RN-0005 into SMA fibroblasts significantly increased SMN-FL expression, whereas transfecting the control oligo, RN-0012, which hybridizes to a region outside of a PRC2 interaction domain, did not change SMN-FL expression (Fig. 3B). Consistently, nRIP showed that RN-0005, but not the control oligo, disrupted the binding of PRC2 to SMN-AS1, as shown by RIP-qPCR (Fig. 3C). Furthermore, oligos targeting SMN-AS1 did not affect the interactions between PRC2 and ANRIL, GAPDH, or RPL19 control RNAs. These results were also observed when nRIP was performed using an antibody against EZH2 (SI Appendix, Fig. S2). As expected, single-molecule RNA-FISH for the localization of SMN-AS1 after transfection with RN-0005 showed no change in both the abundance and the localization of SMN-AS1 in ∼90% of cells examined (39 of 42 nuclei) performed in three independent experiments (SI Appendix, Fig. S3). Together, these results demonstrate that selective inhibition of PRC2:SMN-AS1 interaction by a mixmer oligo leads to increased SMN2 expression.

To gain molecular insight into how the active oligo induced SMN expression, we characterized the chromatin changes at the SMN locus in response to the disruption of PRC2:SMN-AS1 interaction using ChIP. When SMA fibroblasts were treated with RN-0005, we observed a loss of EZH2 association as well as decreased H3K27me3 along the SMN gene body (Fig. 3 D and E), suggesting that the oligo blocked the recruitment and activity of PRC2. Furthermore, there was an increase in association of RNA Pol II-phosphoSer2 (RNA polymerase II, phosphorylated at serine 2) and elevated levels of H3K36me3, both of which indicate greater transcriptional elongation (Fig. 3 F and G). By contrast, pan-H3 levels were unaffected by treatment (Fig. 3H). H3K4me3, a mark of transcription initiation, was enriched at the promoter in the lipid control samples. Interestingly, H3K4me3 levels did not change at the promoter with RN-0005 addition (Fig. 3I), suggesting that the increased SMN mRNA levels may be occurring in a setting where basal levels of transcription exist. No changes in PRC2 association were observed at another well-established Polycomb target locus, HOXC13, upon treatment (Fig. 3J). We also performed ChIP on SMA fibroblasts that were treated with a splice-correcting oligo (SCO), which targets the pre-mRNA for exon 7 inclusion but does not alter the transcription rate at the SMN locus (SI Appendix, Fig. S4). No chromatin changes were observed. These data suggest that PRC2 recruitment and histone methyltransferase activity at the SMN locus can be selectively inhibited by an oligo by mechanisms that may include steric blocking of the specific PRC2:SMN-AS1 interaction or disruption of a secondary structure within SMN-AS1 that would be recognized by PRC2.

We further characterized SMN2 mRNA up-regulation, which resulted from the disruption of the PRC2:SMN-AS1 interaction and the subsequent epigenetic changes at the SMN locus. We used the GM09677 fibroblasts, which carry two copies of the SMN2 gene and are homozygous for SMN1 exons 7 and 8 deletion. RT-qPCR analyses with three different primer sets detected a concentration-dependent increase of various SMN mRNA transcripts, including all SMN isoforms (exon 1–2) as well as isoforms including or excluding exon 7, SMN-FL, and SMNΔ7, respectively (Fig. 4A). Furthermore, overall SMN protein levels also increased, as shown by ELISA after 5 d of treatment, which supports the epigenetic evidence of increased transcription (Fig. 4B). Western blot results revealed that this increase could be attributed to the 38-kDa SMN protein (Fig. 4C). Both quantification of ELISA and Western analyses indicated an approximately fourfold protein up-regulation following treatment in SMA fibroblasts with RN-0005. Taken together, blocking the interaction of PRC2 with its recruiting lncRNA resulted in up-regulation of both SMN mRNA and protein.

To determine how targeting the disruption of PRC2:SMN-AS1 interactions might affect PRC2 targets globally, we performed RNA sequencing from samples with specific disruption of PRC2:SMN-AS1 (using RN-0005) or global inactivation of the PRC2 complex (using a SUZ12 gapmer). Treatment with either RN-0005 or the SUZ12 gapmer resulted in significant increases in SMN mRNA levels compared with transfection control samples (Fig. 4D). Globally, there were approximately fourfold more gene expression changes with the SUZ12 gapmer treatment than with RN-0005 treatment that had at least a 1.5-fold change (q < 0.05). This is depicted by a scatterplot of the moderated t statistics of the gene expression changes where for most genes, their individual SUZ12 knockdown (KD) t statistic is usually larger than their respective RN-0005 t statistic. As this is a scatterplot encompassing all genes and we expect that most genes do not change significantly or have a large magnitude of change, the two gene expression profiles correlate overall. Here, a strong linear correlation was observed while simultaneously many more significant changes occurred with the SUZ12 gapmer treatment. Looking closer at expression profiles of genes neighboring SMN1 and SMN2, the nearest neighboring genes that changed significantly in response to RN-0005 treatment were ADAMTS6 (upstream) and BDP1(downstream) at 4.6 Mb and 1.4 Mb away from SMN2, respectively. As the closest changed genes are greater than a megabase away, these data suggest PRC2:SMN-AS1 regulation is localized to the SMN locus. In contrast, the nearest significant neighbor genes that changed after SUZ12 knockdown were TAF9, 0.8 Mb upstream, and BDP1, 1.4 Mb downstream, of SMN2. Furthermore, because both RN-0005 and the SUZ12 KD gapmer do affect SMN expression, we would expect to see potential overlap in downstream changes, which we confirm by identifying 21 pathways overlapping between oligo treatments (only approximately 4–5 overlaps would be expected by chance if there was no relationship between the two treatments). However, we see far fewer pathways modulated significantly with RN-0005 treatment, suggesting fewer changes overall. Taken together, this suggests that RN-0005 has a more localized effect than the more global effects of knocking down PRC2. Pathway gene set analyses identified significant pathways (10% false discovery rate) with each oligo treatment, and although there were overlaps between the datasets, many more pathways changed separately with SUZ12 knockdown (Fig. 4D and SI Appendix, Fig. S5). This suggests that although there is a relationship between the downstream genes effected by modulating SMN expression, there are less overall downstream changes when you up-regulate SMN expression with RN-0005 than when you target SMN through a global knockdown of PRC2.

Although SMN expression is ubiquitous, its expression is highest in the central nervous system (Fig. 2D) (25), particularly in spinal motor neurons where the disease is manifested (5, 6, 26). To assess the activity of RN-0005 in disease-relevant cells, we examined SMN expression in two neuronal cell types. First, we generated induced pluripotent stem cells (iPSCs) derived from SMA patient fibroblasts and differentiated them into SMI32⁺ motor neurons (SI Appendix, Fig. S6). After treating with RN-0005 for 14 d, SMN-FL mRNA displayed a statistically significant twofold increase relative to untreated or control-treated motor neurons (Fig. 5A). Both an inactive oligo that targets SMN-AS1 but does not up-regulate SMN mRNA and an unrelated oligo that does not target SMN-AS1 showed no effect on SMN mRNA levels. As expected, EZH2 knockdown also led to a similar increase in SMN-FL mRNA. We performed a time-course experiment with RN-0005 and observed a delayed increase in SMN-FL mRNA levels in neurons relative to what was seen in fibroblasts. This may be partially due to the mode of delivery (unassisted delivery versus transfection) and/or the nonproliferating state of the neuronal cells versus the highly proliferative fibroblasts. Supporting the latter, the rate of H3K27me3 removal from chromatin of nondividing cells is slower than in proliferating cells (27). Taken together, these data show that disrupting the PRC2:SMN-AS1 interaction leads to SMN up-regulation in disease-relevant and postmitotic motor neuronal cultures.

We also prepared primary cortical neuronal cells from E14 embryos of the 5025 SMA mice and treated them with a chemical variant of RN-0005 that targets the same SMN-AS1 sequence and shows similar blocking of PRC2:SMN-AS1 and mRNA up-regulation (SI Appendix, Fig. S7) but has a more favorable in vivo safety profile. RN-0027 was added to the neurons for 14 d without obvious toxicity or changes in cell morphology (Fig. 5B). We observed a concentration-dependent increase in SMN-FL mRNA with a threefold increase at 10 μM following 14 d of treatment (Fig. 5C) in multiple experiments. Furthermore, the addition of a control oligo did not result in changes in SMN-FL levels. In agreement with the results obtained from patient fibroblasts (Fig. 1B) and motor neuron cultures (Fig. 5A), ex vivo cortical neurons treated with an EZH2 gapmer ASO displayed a concentration-dependent increase in SMN-FL mRNA levels (Fig. 5D). Our findings from ex vivo cortical neurons suggest that there is in vivo relevance of this mechanism in terminally differentiated neuronal cells.

Splice-correcting modifiers have been designed to facilitate the inclusion of exon 7 during SMN2 transcription (28), resulting in the production of SMN-FL mRNA and functional SMN protein. Although steady-state total SMN mRNA levels would not increase with a splice-correcting modifier, the shift to increase SMN-FL mRNA levels has been demonstrated to be beneficial to survival in mice (29, 30) and in humans (31). Because the transcriptional activation approach up-regulates SMN through a distinct mechanism from that of a splice corrector, we reason that combining these two mechanisms will be more effective than with a single approach. To this end, the 5025 mouse cortical neurons, which only harbored human SMN2 and not human SMN1, were treated with either a SCO, our transcriptional activating mixmer, or a combination of the two oligos for 14 d to measure the levels of SMN-FL mRNA (Fig. 5E). Although treatment with the SCO alone resulted in a 2–3-fold increase in SMN-FL mRNA, an additional 1.8-fold increase was observed in combination with the mixmer treatment. This additive effect was also detected with an increase in the human SMN protein levels by a human-specific ELISA (Fig. 5F). Although the SCO up-regulated SMN protein levels ∼2.5-fold, the combination resulted in a fourfold increase of SMN levels. These data further provide evidence that SMN-AS1 inhibition increases SMN-FL mRNA and SMN protein levels by a mechanism that is independent and complementary to that of splice correction. The increased SMN-FL mRNA and protein resulting from the combination approach may provide greater benefit in treating SMA.

RNA from GM09677 fibroblasts that were transfected with RN-0005, SUZ12 gapmer ASO, and lipid controls were sequenced (300 bp paired-end) on the NextSeq500 using Illumina TruSeq stranded total RNA-seq library preparation kits. Refer to SI Appendix, SI Experimental Procedures for a detailed explanation of the differential gene expression analyses.

Probe sets were designed against genomic regions listed in SI Appendix, Table S1. They were labeled with Quasar 570 (SMN1/2 exons), Quasar 670 (SMN1/2 introns), and Cal Fluor Red 610 (SMN1/2-AS1). Stellaris RNA-FISH was performed as described in the Alternative Protocol for Adherent Cells (UI-207267 Rev. 1.0) with the following modifications: 12-mm diameter coverslips were used. We used 25 µL hybridization solution with a final concentration of each probe set of 250 nM. The wash buffer volumes were halved. The FITC, Cy3, Cy3.5, and Cy5.5 channels were used to capture the signals from each probe set, and the FITC channel was used to identify cellular autofluorescence. The filter sets from Chroma were 49001-ET-FITC, SP102v1-Cy3, SP103v2-Cy3.5, and 41023-Cy5.5. The exposure times were 1 s for FITC, Quasar 570, and Cal Fluor Red 610 and 2 s for Quasar 670. GM09677 Human Eye Lens Fibroblast (Coriell) adherent cells were grown in Eagle’s Minimum Essential Medium (EMEM) (ATCC) in a humidified 37 °C incubator at 5% CO₂ in ambient air. F-12K and EMEM media were supplemented with 10% (vol/vol) FBS (Fisher Product number SH30071.03) and 5 mL of Pen/Strep (Life Technologies). F-12 was further supplemented with Normocin (InvivoGen). Cells were grown on 12-mm microscope circular cover glass No. 1 (Fisher #12–545-80) in 24-well flat-bottom cell culture plates (E&K). SMA fibroblasts were transfected at 70% confluence by using oligonucleotides complexed with Invitrogen Lipofectamine 3000 (Pub Part #100022234, Pub #MAN0009872, Rev. B.0) and fixed after 2 d. We used 2 ng DNA and 4 µL P3000 reagent per 50 µL of DNA master mix. We used 0.375 µL Lipofectamine 3000 reagent per 25 µL of Opti-MEM.

Total RNA from 20 human tissues (Clontech) were used for cDNA synthesis using High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Data of RT-qPCR SMN-AS1 levels were normalized to levels from the adrenal gland. GM09677 fibroblasts were plated on a 24-well tissue culture plate at 4 × 10⁴ cells per well in MEM containing 10% (vol/vol) FBS and 1× nonessential amino acids. Fibroblasts were treated with oligos the following day. After 2 d, cells were lysed and mRNA was purified using E-Z 96 Total RNA Kit (Omega Bio-Tek). SMA iPS-derived motor neurons were lysed with TRIzol for RNA isolation per the manufacturer’s protocol. RNA from mouse cortical neurons was extracted using the RNeasy kit (Qiagen) per the manufacturer’s protocol. All cDNAs were synthesized using High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). SMN-FL, SMN Δ7, SMN Exon 1–2 and GUSB mRNA expression was quantified by predesigned TaqMan real-time PCR assays. A list of custom-designed real-time PCR assays is listed in SI Appendix, Table S1.

SMA fibroblasts were transfected at 70% confluence by using oligonucleotides complexed with Lipofectamine 2000 (Invitrogen) following the protocol suggested by the manufacturer in the 96-well and 24-well format. For ChIP, cells were transfected in 15-cm plates and were transfected at 30 nM with Lipofectamine 2000 at a final volume of 20 mL. Cells were harvested 3 d posttransfection.

RIP was performed using the Magna RIP RNA-Binding Protein Immunoprecipitation Kit (EMD Millipore) using ChIP-grade anti-SUZ12 (Abcam), anti-EZH2 (Abcam), and anti-SETD2 (USBiological Life Sciences) antibodies. RNA was extracted with TRIzol (Life Technologies) and transcribed to cDNA using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). qPCR was performed on a StepOnePlus Real Time PCR System (Applied Biosystems) using TaqMan Fast Advanced Master mix (Applied Biosystems).

DNA templates for EMSA probes containing T7 promoter sequences were generated by PCR using Phusion High Fidelity DNA Polymerase (NEB). The specific primer sequences are listed in SI Appendix, Table S1. EMSAs were performed as described previously (Cifuentes-Rojas et al., 2014) (22). RNA probes were transcribed using the AmpliScribe T7 Flash Transcription Kit (Epicentre) and PAGE purified from 6% (vol/vol) TBE urea gel. RNA probes were then dephosphorylated by calf intestinal alkaline phosphatase (NEB), purified by phenol-chloroform extraction, 5′ end-labeled with T4 Polynucleotide Kinase (NEB) and [γ-³²P] ATP (Perkin-Elmer), and purified with Illustra MicroSpin G-50 columns (GE Life Sciences). RNA probes were folded in 10 mM Tris pH 8.0, 1 mM EDTA, and 300 mM NaCl by heating to 95 °C, followed by incubations at 37 °C and at room temperature for 10 min each. MgCl₂ and Hepes pH 7.5 were then added to 10 mM each, and probes were put on ice. We mixed 1 µL of 2,000 cpm/mL (2 nM final concentration) folded RNA with PRC2 (EZH2/SUZ12/EED; BPS Bioscience) at the indicated concentration and 50 ng/mL yeast tRNA (Ambion) in 20 µL final concentration of binding buffer [50 mM Tris⋅HCl pH 8.0, 100 mM NaCl, 5 mM MgCl2, 10 mg/mL BSA, 0.05% Nonidet P-40, 1 mM DTT, 20 U RNaseOUT (Invitrogen), and 5% (vol/vol) glycerol]. Binding reactions were incubated for 20 min at 30 °C and applied on a 0.4% hyperstrength agarose (Sigma) gel in THEM buffer (66 mM Hepes, 34 mM Tris, 0.1 mM disodium EDTA, and 10 mM MgCl₂). Gels were run for 1 h at 130 V with buffer recirculation at 4 °C, dried, and exposed to a phosphorimager screen. Screens were scanned in a Storm 860 phosphorimager (Molecular Dynamics), and data were quantified by Quantity One and normalized as described (37). K_Ds were calculated with Graphpad Prism by fitting the data to a one site-specific binding model.

Cells were lysed 5 d posttransfection using the extraction buffer from the SMN ELISA kit (Enzo) with Protease inhibitor mixture tablets (Roche). Total protein content was determined with the total BCA assay (Promega) for equal loading. Samples and Hi Mark prestained ladder (Invitrogen) were run on a 4% (vol/vol) Bis–Tris gel, and proteins were transferred to nitrocellulose membrane. The membrane was incubated in blocking buffer (Licor) overnight at 4 °C. The SMN antibody (BD catalog no. 610646), Alpha tubulin antibody (Abcam catalog no. ab125267), and secondary anti-mouse and anti-rabbit 800 (Licor) were used, and the membrane was scanned with Odessey (Licor). Band intensities for SMN-FL protein and α-tubulin were quantified using Image Studio software.

GM09677 fibroblasts were plated on a 24-well tissue culture plate at 4 × 10⁴ cells per well in MEM containing 10% (vol/vol) FBS and 1× nonessential amino acids. Fibroblasts were treated with oligonucleotides the following day. After 5 d, cells were lysed and protein was quantified with the SMN ELISA Kit (Enzo Life Sciences, Inc.) and normalized to total protein content as determined by Micro BCA Protein Assay Kit (Thermo Scientific). For the human-specific ELISA used with the cortical neurons, a similar protocol was used. Briefly, cells were washed in cold PBS and lysed in RIPA buffer supplemented with protease inhibitor Complete Tablets, mini EDTA-free EASYpack (Roche). Lysates were quantified by BCA, and ∼20–30 µg were used. A mouse monoclonal anti-SMN antibody was captured on high binding plates (Pierce) at 1 µg/mL; after blocking with BSA in PBS-0.05% Tween-20, lysates were incubated for 2 h at room temperature; a rabbit polyclonal human SMN-specific antibody at 1 µg/mL was used for detection, followed by HRP-goat anti-rabbit (Invitrogen). Signal was measured with SuperSignal ELISA PICO chemiluminescent substrate (Thermo). Total GAPDH in the lysates was also quantified by ELISA (R&D Systems); SMN protein concentration was normalized to total GAPDH content.

Brains were isolated from E14 SMNΔ7 embryos and the cortex was dissected with the MACS neuronal tissue dissociation kit (Miltenyi Biotec). The collected cortical neurons were plated at 0.5 × 10⁶ cells per well in Neurobasal media (ThermoFisher), B-27 supplement (ThermoFisher), and GlutaMax (ThermoFisher) in a 24-well plate coated with poly–d-lysine (Fisher). Cells were incubated at 37 °C, 5% CO₂ for 4 d, allowing the cells to mature and networks to form before unassisted delivery of RN-0005. After 14 d, the cells were harvested for RNA isolation.

SMA patient and control subject dermal fibroblasts or lymphoblastoid cell lines (LCLs) were obtained from the Coriell Institute for Medical Research. All of the cell lines and protocols in the present study were carried out in accordance with the guidelines approved by the Stem Cell Research Oversight Committee (SCRO) and Institutional Review Board (IRB) at the Cedars–Sinai Medical Center under the auspice IRB-SCRO Protocols Pro00032834 (iPSC Core Repository and Stem Cell Program) and Pro00024839 and Pro00036896 (Sareen Stem Cell Program). The iPSCs were grown to near confluence under normal maintenance conditions before the start of the differentiation as per protocols described previously (38). Briefly, IPSCs were then gently lifted by Accutase treatment for 5 min at 37 °C. We subsequently placed 1.5–2.5 × 10⁴ cells in each well of a 384-well plate in defined neural differentiation medium with dual-SMAD inhibition (39). After 2 d, neural aggregates were transferred to low adherence flasks. Subsequently, neural aggregates were plated onto laminin-coated six-well plates to induce rosette formation in media supplemented with 0.1 μM retinoic acid and 1 µM puromorphine along with 20 ng/mL BDNF, 200 ng/mL ascorbic acid, 20 ng/mL GDNF, and 1 mM dbcAMP. Neural rosettes were isolated, and the purified rosettes were subsequently supplemented with 100 ng/mL of EGF and FGF. These neural aggregates, termed iPSC-derived motor neuron precursor spheres (iMPSs), were expanded over a 5-wk period. For terminal differentiation, iMPSs were disassociated with Accutase and then plated onto laminin-coated plates over a 21-d period before harvest using the MN maturation media consisting of Neurobasal supplemented with 1% N2, ascorbic acid (200 ng/mL), dibutyryl cyclic adenosine monophosphate (1 μM), BDNF (10 ng/mL), and GDNF (10 ng/mL). RN-0005 treatments were carried out during this terminal differentiation period. Antibodies used for immunocytochemistry were as follows: SSEA4 and SOX2 (Millipore); TRA-1–60, TRA-1–81, OCT4, and NANOG (Stemgent); TuJ1 (β₃-tubulin) and Map2 a/b (Sigma); ISLET1 (R&D Systems); and SMI32 (Covance).

Cells were cross-linked with 1% formaldehyde for 10 min at room temperature and then quenched with glycine. Chromatin was prepared and sonicated (Covaris S200) to a size range of 300–500 bp. Antibodies for H3, H3K27me3, H3K36me3, EZH2, and RNA polymerase II Serine 2 (Abcam) and H3K4me3 (Millipore) were coupled to Protein G magnetic beads (NEB), washed, and then resuspended in IP blocking buffer. Chromatin lysates were added to the beads and immunoprecipitated overnight at 4 °C. Antibodies against H3, H3K36me3, RNA polymerase II phosphoserine 2, H3K27me3, and EZH2 were obtained from Abcam, and H3K4m3 antibody was obtained from Millipore. We used 10 µg of antibody per IP. IPs were washed, RNase A (Roche) treated, and Proteinase K treated (Roche), and the cross-links were reversed by incubation overnight at 65 °C. DNA was purified, precipitated, and resuspended in nuclease-free water. Custom TaqMan probe sets were used to determine enrichment of DNA. Probes were designed using the custom design tool on the Life Technologies website. Primer sequences are listed in SI Appendix, Table S1.

All RNA-sequencing data were deposited in the Gene Expression Omnibus (GEO) database under accession no. GSE83549.