A high-throughput genome-wide RNAi screen identifies modifiers of survival motor neuron protein

SUMMARY

Spinal muscular atrophy (SMA) is a debilitating neurological disorder marked by degeneration of spinal motor neurons and muscle atrophy. SMA results from mutations in survival motor neuron 1 (SMN1), leading to deficiency of survival motor neuron (SMN) protein. Current therapies increase SMN protein and improve patient survival but have variable improvements in motor function, making it necessary to identify complementary strategies to further improve disease outcomes. Here, we perform a genome-wide RNAi screen using a luciferase-based activity reporter and identify genes involved in regulating SMN gene expression, RNA processing, and protein stability. We show that reduced expression of Transcription Export complex components increases SMN levels through the regulation of nuclear/cytoplasmic RNA transport. We also show that the E3 ligase, Neurl2, works cooperatively with Mib1 to ubiquitinate and promote SMN degradation. Together, our screen uncovers pathways through which SMN expression is regulated, potentially revealing additional strategies to treat SMA.

Graphical Abstract

In brief

Treatments for spinal muscular atrophy aim to increase survival motor neuron (SMN) protein. Using a genome-wide RNAi screen, McCormack et al. identify modifiers of SMN expression, including genes that are involved in transcription regulation, RNA processing, and protein stability.

INTRODUCTION

Spinal muscular atrophy (SMA) is a neuromuscular disease that is one of the leading genetic causes of infant mortality (Lefebvre et al., 1995). SMA is characterized by the loss of lower motor neurons in the anterior horn of the spinal cord and muscle atrophy leading to muscle weakness (Lefebvre et al., 1995). SMA is caused by the homozygous loss of survival motor neuron 1 (SMN1), the predominant producer of survival motor neuron (SMN) protein (Lefebvre et al., 1995; Lorson et al., 1999). Humans have a second copy of this gene, SMN2, which has a cytosine-to-thymine base pair change leading to exclusion of exon 7 in the transcript to primarily produce a truncated form of SMN known as SMNΔ7 (Lorson et al., 1999; Monani et al., 1999a). SMN2 is unable to fully compensate for the loss of SMN1 in patients with SMA leading to a deficiency of SMN protein. Patients with SMA have variable copy numbers of SMN2, which correlates with disease severity (Mailman et al., 2002; Wirth et al., 2006). Although the critical function of SMN in the context of SMA is not completely known, SMN has been shown to be required for the assembly of small nuclear ribonucleoproteins (snRNPs) (Pellizzoni et al., 2002). It remains unclear how loss of SMN contributes to SMA pathology.

Restoration of SMN protein has been shown to ameliorate the disease phenotype. Strategies to increase SMN in SMA animal models and patients include increasing SMN2 transcription, promoting exon 7 inclusion, blocking SMN protein degradation, and gene delivery (Abera et al., 2016; Avila et al., 2007; Hua et al., 2008; Passini et al., 2010). Currently, there are three recently approved treatments for patients with SMA: nusinersen, an antisense oligonucleotide, which increases SMN protein levels by promoting the inclusion of exon 7 in the SMN2 transcript; gene therapy (Zolgensma), which uses an adeno-associated virus (AAV)-9 to deliver SMN1 complementary DNA (cDNA) to patients; and risdiplam, a small molecule that corrects SMN2 splicing (Foust et al., 2010; Hua et al., 2008; Mendell et al., 2017; Poirier et al., 2018). Although some patients have significant improvements in their ability to sit, walk, and run, other patients see little to no improvement in their motor function (Aragon-Gawinska et al., 2020; Finkel et al., 2017; Mendell et al., 2017; Mercuri et al., 2018). For example, in one study, only 32% of SMA type 1 patients were able to sit unassisted for 30 s following 14 months of nusinersen treatment (Aragon-Gawinska et al., 2020). Therefore, ongoing clinical trials and research focus on identifying alternative ways to further increase SMN protein levels or complementary strategies to improve disease outcomes by targeting non-SMN-modifying genes and pathways.

In addition to SMN2, only two disease modifiers have been identified: plastin-3 and neurocalcin delta (Oprea et al., 2008; Riessland et al., 2017). Plastin-3 is upregulated in unaffected SMN1-deleted females (Oprea et al., 2008). Expression of neurocalcin delta is reduced in asymptomatic, SMN1-deleted individuals with four copies of SMN2 (Riessland et al., 2017). Upregulation of plastin-3, an actin-binding protein, lessens disease severity potentially by stabilizing the growth cone, while downregulation of neurocalcin delta alleviates the disease phenotype by restoring endocytosis in SMA mice (Kaifer et al., 2017; Oprea et al., 2008; Riessland et al., 2017). To date, very few regulators of SMN transcript have been reported. Histone deacetylase (HDAC) inhibitors have been used in cell and animal models of the disease to increase SMN transcript levels (Avila et al., 2007; Sumner et al., 2003) but have been shown to have limited efficacy in clinical trials (Kissel et al., 2011, 2014; Swoboda et al., 2010). Moreover, HDAC inhibitors have been reported to activate expression of ~2% of the genome, which reduces enthusiasm to further develop these drugs for SMA treatment (Kelly et al., 2002; Pazin and Kadonaga, 1997; Van Lint et al., 1996).

Given that current treatment modalities increase SMN levels, it is important to determine all the factors that regulate SMN protein expression. In this study, we sought to identify, using an unbiased approach, modifiers of SMN protein levels utilizing a genome-wide cell-based RNAi screen against >20,000 individual genes. Human embryonic kidney 293 (HEK293) cells stably expressing an SMN2 luciferase minigene cassette were used as the cellular model system for the screen, and subsequent validation analyses were carried out using SMA patient-derived fibroblasts. We have identified and validated regulators of SMN expression that were grouped into transcriptional regulation, RNA processing, and protein stability.

RESULTS

Genome-wide RNAi screen identifies genes that regulate SMN expression

To identify modifiers of SMN expression, we conducted a genome-wide RNAi screen using HEK293 cells that stably express an SMN2 luciferase minigene (Zhang et al., 2001). The luciferase reporter gene assay combines the promoter and splicing-based cassettes in tandem with the major portion of the native SMN2 cDNA with luciferase (Figure 1A). Luciferase is expressed when full-length SMN protein is generated with luminescence signal proportional to SMN protein levels. The screen was conducted in two stages (Figure 1A). To determine baseline signals, standard deviation values, and optimal incubation times post-transfection, we conducted an initial screen using the Ambion Silencer Select Version 4 Human Druggable Genome comprising a non-pooled library of three independent small interfering RNAs (siRNAs) targeting 10,415 genes that are predicted to be druggable. Primary screening data have been deposited into PubChem under the assay ID 1347426. The initial screen was completed successfully with robust SMN2 luciferase expression that could be detected following siRNA transfection. Several genes when knocked down altered luciferase expression, including genes that were previously shown to modulate SMN protein expression (shown in italics) (Table S1) (Hastings et al., 2007; Kashima et al., 2007a; Wee et al., 2014). We next performed a follow-up genome-wide RNAi screen (Figure 1A) using the Ambion genome-wide siRNA library, which is composed of three independent siRNAs targeting 21,556 genes. Candidate gene selection used a robust statistical measure of median absolute deviation (MAD) to standardize siRNA activities from each screen (Chung et al., 2008). Common seed analysis was used to help prioritize clean hits as opposed to those appearing to arise from seed-based off-target effects (Buehler et al., 2012; Marine et al., 2012). Outcomes of screened siRNAs with a positive Z score reflect enhanced full-length SMN expression and a negative Z score showing a reduction in SMN levels (Figure 1B). To identify strong modifiers, we set a Z score cutoff of greater than 1.0 or less than −1.0 (2-fold change). Z score analyses indicated a continuous distribution of hits (Figure 1B); 112 genes gave Z scores smaller than −1.0 (reduced expression) and 89 larger than +1.0 (increased expression), ranging from −5.5 to +3.3 (Tables S2 and S3). Several previously described regulators of SMN expression were re-identified in the screen (Figure S1), thus validating the effectiveness of our assay.

Figure 1. Genome-wide RNAi screen identifies genetic modifiers of SMN expression.

(A) Schematic of SMN2 luciferase reporter cassette stably expressed in HEK293 cells used in an RNAi screen to identify modifiers of SMN. Two independent screens using the Ambion Druggable Genome siRNA library (10,415 genes) and the Ambion Genome-wide siRNA library (21,556 genes) were performed followed by seed correction to eliminate off-target effects.

(B) Distribution of Z scores and distribution of Z scores normalized to median absolution deviation (MAD).

(C) KEGG pathway analysis was used to categorize top hits by predicted molecular function.

See also Figure S1 and Tables S1, S2, and S3.

We divided hits into groups based on predicted molecular functions to gain better insight into pathways that would lead to enhancement of SMN protein levels. We found an enrichment of genes involved in RNA processing and protein metabolism (Figure 1C). Additionally, we sorted the top 100 hits by protein class using Protein Annotation Through Evolutionary Relationship (PANTHER) analysis and found several genes that encode for nucleic acid binding proteins and protein-modifying enzymes (Table 1).

Table 1.

Clustering of top hits by protein class

Protein class	No. of genes
Nucleic acid binding protein	13
Transporter	7
Cytoskeletal protein	4
Metabolite interconversion enzyme	4
Protein-modifying enzyme	4
Translational protein	4
Gene-specific transcriptional regulator	3
Protein-binding activity modulator	3
Membrane traffic protein	2
Scaffold/adaptor protein	2
Chromatin/chromatin-binding or regulatory protein	1
Defense/immunity protein	1
Intercellular signal molecule	1

The top 100 genes from the RNAi screen were sorted by protein class using a Protein Annotation Through Evolutionary Relationship (PANTHER) analysis.

Identification of SMN exon 7 splicing regulators

An overwhelming number of genes (75 out of 201 hits) had well-characterized roles in mRNA splicing, indicating that splicing alterations might be the most robust mechanism to increase SMN expression. Of the SMN-modifying genes, we identified several genes that have previously been shown to enhance SMN exon 7 inclusion. The previously described SMN2 splicing modifiers hnRNP A1, hnRNP A2, U2AF1, U2AF2, PUF60, and SRSF3 were all high-scoring hits in the screen (Table 2) (Hastings et al., 2007; Kashima et al., 2007a; Martins de Araújo et al., 2009; Wee et al., 2014). All these genes were previously shown to promote SMN exon 7 exclusion consistent with an increase in full-length SMN expression when they were inactivated. Interestingly, genes previously described to promote SMN inclusion (ASF/SF2 [SRSF1], SRp30c, Tra2-β, TIA1) were not picked up in the screen (Cartegni and Krainer, 2002; Hofmann et al., 2000; Singh et al., 2011; Young et al., 2002). This may point to the limitations of this assay to detect genes that further suppress SMN expression, as well as genes that have redundant roles in regulating SMN expression.

Table 2.

Reidentified modifiers of SMN expression: Genes identified in the RNAi screen that were previously shown to alter SMN expression

Gene symbol	Gene name	Reference
U2AF1	U2 small nuclear RNA auxiliary factor 1	Hastings et al., 2007; Auslander et al., 2020
PUF60	poly-U binding splicing factor 60 kDa	Hastings et al., 2007
HNRNPA2B1	heterogeneous nuclear ribonucleoprotein A2/B1	(Kashima et al., 2007a)
U2AF2	U2 small nuclear RNA auxiliary factor 2	Hastings et al., 2007
UCHL1	ubiquitin carboxyl-terminal esterase L1	Hsu et al., 2010
MIB1	Mindbomb E3 ubiquitin protein ligase 1	Kwon et al., 2013
SRSF11	serine/arginine-rich splicing factor 11	Wee et al., 2014
SRSF3	serine/arginine-rich splicing factor 3	Wee et al., 2014

Gene cluster and interactome analyses

We next sought to determine whether specific cellular pathways altered SMN expression by grouping hits into functional clusters using Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) analysis (Szklarczyk et al., 2017). The 201 genes that met our initial selection criteria (increased or decreased SMN expression) were selected for further analysis (Figure 2A). We looked for over-representation of biological functions among the candidate genes by gene set enrichment analysis using the Kyoto Encyclopedia of Genes and Genomes (KEGG), Reactome, Gene Ontology (GO), and PANTHER databases.

Figure 2. Interactome analysis of top hits.

(A) STRING analysis of top hits showed four gene clusters: mediator complex, splicing and nucleocytoplasmic trafficking, ubiquitin-proteasome degradation system, and transcription factors.

(B) List of genes and their MAD Z score in each cluster.

Key components of the RNA polymerase II transcription machinery, including the mediator complex, were highly enriched in the genes that when knocked down reduced SMN expression, consistent with their role as transcriptional activators. Indeed, knockdown of CDK8, Med6, Med8, Med14, Med17, Med20, Med24, Med26, Med27, Med30, or Med31 all resulted in decreased SMN expression (Figure 2B). Using STRING analysis, we identified two distinct network clusters that regulated SMN expression: the Transcription-Export (TREX) complex and the ubiquitin-proteasome system (Figure 2B).

The TREX complex regulates SMN levels

Although SMN2 splicing has been well studied, little is known about the factors that regulate SMN mRNA nucleo-cytoplasmic trafficking. The TREX complex couples transcription elongation by RNA polymerase II to mRNA export (Katahira, 2012). TREX coordinates with the elongating RNA polymerase II and facilitates RNP assembly and recruitment of RNA processing factors. Depletion of TREX components reduces expression of some, but not all, genes (Mason and Struhl, 2005). We found from our screen that siRNA targeting THOC1, THOC4 (ALYREF), NFX1, and CPSF6 reduced SMN expression, suggesting they may be indispensable to SMN mRNA nucleocytoplasmic translocation.

Interestingly, we found that knocking down THOC3, another member of the TREX complex, increased SMN expression. This result seemingly conflicts with its predicted role in nuclear export of mRNA (Katahira, 2012). It has been proposed that the TREX and cleavage and polyadenylation specificity factor (CPSF) complexes play sequential roles in mRNA maturation and trafficking (Chan et al., 2011; Katahira, 2012). Unlike other TREX complex subunits, THOC3 contains a WD40 repeat domain, like the CPSF complex protein WDR33, suggesting that both proteins may have redundant or complementary roles in mRNA processing (Chan et al., 2011; Heath et al., 2016). We selected CPSF1, THOC3, and WDR33 to further investigate their roles in SMN RNA processing.

We transfected cells with siRNA targeting CPSF1, WDR33, and THOC3 and isolated protein and mRNA 48 h after transfection. We confirmed that knocking down each gene individually (WDR33 mRNA: 84% decrease, p < 0.0001; CPSF1 mRNA: 87% decrease, p < 0.0001; THOC3 mRNA: 89% decrease, p < 0.0001) increased SMN protein levels (siWDR33: 70% increase, p = 0.0064; siCPSF1: 71% increase, p = 0.0043; siTHOC3: 75% increase, p = 0.0205) (Figures 3A and 3B). Additionally, simultaneously transfecting SMA patient-derived fibroblasts with all three siRNAs significantly increased SMN protein levels over control (171% increase; p = 0.0418) (Figures 3C and 3D). Knockdown of the three genes together resulted in a similar increase in SMN as each of the individual knockdowns (Figure S2). We next examined whether knocking down each gene altered SMN2 transcription, mRNA stability, or cytoplasmic translocation. To determine whether knocking down each gene affected SMN2 transcription, we used RT-PCR primers spanning SMN intron-exon boundaries to quantify SMN2 precursor mRNA (pre-mRNA). We found that SMN2 pre-mRNA levels were not changed by knocking down CPSF1 (p = 0.5059) or THOC3 (p = 0.9635) but are significantly increased when WDR33 is knocked down (51% increase; p = 0.0131) (Figure 4A). In addition, total SMN2 mRNA was unchanged by knocking down each gene (p = 0.5643) (Figure 4B). We next examined whether knocking down CPSF1, WDR33, and THOC3 affects SMN2 mRNA stability. We treated SMA patient cells with actinomycin D to stop gene transcription and collected cells at the designated time points. We found no difference in the rate of degradation of the SMN2 mRNA when CPSF1, WDR33, or THOC3 was knocked down (Figure 4C). To investigate whether knockdown of either CPSF1, WDR33, and THOC3 affected SMN2 splicing, we collected cell pellets from SMA patient-derived fibroblasts 48 h after transfection with siRNA targeting CPSF1, WDR33, and THOC3; isolated RNA; and converted mRNA to cDNA. SMN2 cDNA was amplified by PCR and resolved by agarose gel electrophoresis. We found that the ratio of full-length SMN to SMNΔ7 was not altered when either CPSF1 (p = 0.0537), WDR33 (p = 0.8984), or THOC3 (p = 0.9990) was knocked down (Figure 4D).

Figure 3. Knockdown (KD) of RNA processing machinery components increases SMN protein levels.

SMA patient-derived fibroblasts were transfected with siRNA for 48 h.

(A) Gene expression validates KD of WDR33, CPSF1, and THOC3. Data were analyzed using an unpaired t test (WDR33 mRNA: p < 0.0001; CPSF1 mRNA: p < 0.0001; THOC3 mRNA: p < 0.0001). n = 5.

(B) Western blot and quantification of SMN protein levels 48 h after siRNA transfection. SMN protein levels are increased with KD of each gene. Cells transfected with siRNA targeting either U2AF2 or SRSF3 were run as a control. Data were analyzed using unpaired t tests (control versus siWDR33: p = 0.0064; control versus siCPSF1: p = 0.0043; control versus siTHOC3: p = 0.0205). n = 4–5.

(C) SMA patient-derived fibroblasts were simultaneously transfected with siRNAs targeting WDR33, CPSF1, and THOC3 (triple KD) for 48 h. Gene expression validates KD of WDR33, CPSF1, and THOC3. Data were analyzed using unpaired t tests (WDR33 mRNA: p < 0.0001; CPSF1 mRNA: p < 0.0001; THOC3 mRNA: p < 0.0001). n = 5.

(D) Western blot and quantification of SMN protein levels 48 h after transfection with all three siRNAs. SMN protein levels are increased in fibroblasts with all three genes knocked down. Data were analyzed using an unpaired t test (p = 0.0418). n = 5.

All data are represented as mean ± SEM. *p < 0.05, **p < 0.01, ****p < 0.0001. See also Figure S2.

Figure 4. KD of RNA processing machinery components increases cytoplasmic SMN mRNA.

SMA patient-derived fibroblasts were transfected with siRNA targeting WDR33, CPSF1, or THOC3 for 48 h.

(A and B) SMN2 pre-mRNA (A) and total SMN2 mRNA (B) were examined using qRT-PCR. Total SMN mRNA was unchanged with KD of each gene. SMN pre-mRNA data were analyzed using one-way ANOVA (F(3,16) = 4.233, p = 0.0221) followed by post hoc Dunnett’s multiple comparisons test (control versus siWDR33: p = 0.0131; control versus siCPSF1: p = 0.5059; control versus siTHOC3: p = 0.9635). Total SMN2 mRNA data were analyzed using one-way ANOVA (F(3,16) = 0.7025, p = 0.5643). n = 5.

(C) SMA patient-derived fibroblasts were transfected with siRNA for 48 h and were then treated with actinomycin D to inhibit transcription. SMN2 mRNA expression was examined for 8 h after the addition of actinomycin D. SMN2 mRNA stability was unchanged with reduction of each gene. n = 3–4.

(D) Splicing of the SMN2 transcript and quantification of the ratio of full-length SMN to SMNΔ7. SMN2 splicing was unchanged with reduction of each gene. SMN2 splicing data were analyzed using one-way ANOVA (F(3,12) = 4.071, p = 0.0329) followed by post hoc Dunnett’s multiple comparisons test (control versus siWDR33: p = 0.8984; control versus siCPSF1: p = 0.0537; control versus siTHOC3: p = 0.9990). n = 5.

(E) Cytoplasmic and nuclear SMN2 mRNA levels were examined using qRT-PCR. We found the predominantly nuclear MALAT1 mRNA in the nuclear and not the cytoplasmic RNA fractions, confirming the purity of fractionation. Importantly, cytoplasmic SMN mRNA expression was increased with reduction of WDR33, CPSF1, and THOC3. Data were analyzed using one-way ANOVA (F(3,16) = 19.45, p < 0.0001) followed by post hoc Dunnett’s multiple comparisons test (control versus siWDR33: p < 0.0001; control versus siCPSF1: p < 0.0001; control versus siTHOC3: p = 0.0001). n = 5. Nuclear SMN mRNA expression was decreased with KD of WDR33, CPSF1, and THOC3. Data were analyzed using one-way ANOVA (F (3,16) = 8.846, p = 0.0011) followed by post hoc Dunnett’s multiple comparisons test (control versus siWDR33: p = 0.0040; control versus siCPSF1: p = 0.0014; control versus siTHOC3: p = 0.0014). n = 5.

All data are represented as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. See also Figure S3.

Lastly, we looked at the nucleocytoplasmic distribution of SMN2 mRNA after knocking down each gene. Purity of nuclear and cytoplasmic fractions was verified by examining mRNA levels of MALAT1 and ACTB (Figure S3). Nuclear SMN2 mRNA was reduced following knockdown of WDR33 (p = 0.0040), CPSF1 (p = 0.0014), and THOC3 (p = 0.0014), whereas we found increased cytoplasmic localization of the SMN2 mRNA 48 h after knocking down WDR33 (p < 0.0001), CPSF1 (p < 0.0001), and THOC3 (p = 0.0001), offering one explanation for increased SMN expression (Figure 4E). Together, we show that components of the mRNA trafficking complexes are enriched in our screen and confirm in patient cells that knocking down CPSF1, WDR33, and THOC3 increases SMN expression. Knockdown of WDR33 slightly increased SMN2 pre-mRNA levels, but total SMN2 mRNA levels remained unchanged. Although we found no evidence that each gene altered SMN2 mRNA stability or splicing, knocking down WDR33, CPSF1, and THOC3 increased the cytoplasmic abundance of SMN2 mRNA.

Reducing expression of core components of the ubiquitin-proteasome system increases SMN levels

Functional clustering analysis revealed enrichment of genes regulating SMN expression, including those involved in protein clearance. We and others have shown that SMN protein stability is regulated by the ubiquitin-proteasome pathway (Burnett et al., 2009; Chang et al., 2004; Gray et al., 2018; Han et al., 2012; Hsu et al., 2010; Kwon et al., 2013; Locatelli et al., 2015; Powis et al., 2016). From the screen, we identified core components of the ubiquitin-proteasome degradation system that when knocked down increase SMN levels (Figure 2B). These genes include UBA1, the enzyme that catalyzes the first step in the ubiquitin conjugation cascade, which is responsible for tagging proteins for trafficking and degradation (Amm et al., 2014). Interestingly, mutations in UBA1 have been previously associated with X-linked infantile SMA (Ramser et al., 2008), and increasing UBA1 expression improves the disease phenotype in SMA mice (Powis et al., 2016).

We have shown that the E3 ubiquitin ligase mindbomb-1 (Mib1) directly ubiquitinates and promotes the degradation of SMN (Kwon et al., 2013). Here, we found siRNAs that target the E2-conjugating enzyme UBE2D3, which works in combination with Mib1 to ubiquitinate target proteins, also increase SMN protein levels (Figure 2B). We confirmed in patient-derived cells that knocking down UBE2D3 increases SMN levels (Figure S1B). Although E1 and E2 activating and conjugating enzymes drive the initial stages of protein ubiquitination, the E3 ligase is responsible for the substrate specificity for protein ubiquitination. In addition to Mib1, the screen re-identified the E3 ligase UCHL1 (Figure S1B), which has previously been confirmed to regulate SMN protein levels (Hsu et al., 2010), and two additional E3 ligases, SIAH1 and neuralized-2 (Neurl2), as potential regulators of SMN levels. We confirmed in SMA patient-derived fibroblasts that knocking down Neurl2, but not SIAH1, increased SMN levels (80% increase, p < 0.0001) (Figure 5A). Altogether, we confirmed here that the ubiquitin-proteasome system is a major regulator of SMN protein levels and identified additional genes that may contribute to regulating SMN protein stability, either together or independently.

Figure 5. KD of Neurl2 blocks SMN degradation to increase SMN protein levels.

(A) SMA patient-derived fibroblasts were transfected with siRNA targeting Neurl2. Gene expression and western blot analysis show that Neurl2 KD resulted in increased SMN levels. Data were analyzed using an unpaired t test (Neurl2 mRNA: p < 0.0001; SMN protein: p < 0.0001). n = 5.

(B) SMA patient fibroblasts were transfected with GFP-Neurl2 or Neurl2-GFP. SMN was immunoprecipitated, and Neurl2 was associated, confirming their interaction.

(C) Patient fibroblasts were transfected with siRNA targeting Neurl2, Mib1, or both. Western blot analysis of SMN levels 48 h following transfection. Data were analyzed using a one-way ANOVA (F(5,24) = 65.53, p < 0.0001) followed by post hoc Tukey’s multiple comparisons test (control versus Neurl2: p = 0.0305; control versus Mib1: p < 0.0001; Mib1 versus Mib1 + siNeurl2: p < 0.0001; Mib1 versus Mib1 + siNeurl2 + Neurl2_res: p = 0.7566). n = 5.

(D) Patient cells were transiently transfected with 1 μg HA-ubiquitin and FLAG-Mib1, siRNA targeting Neurl2, or siRNA-resistant Neurl2 (Neurl2_res). SMN was immunoprecipitated, and SMN ubiquitination was determined by western blot using an anti-polyubiquitin (FK1) antibody. n = 3.

(E) Cell-free SMN ubiquitination assay. Recombinant SMN was incubated with E1 and E2 (UBCH5B or UBCH1) enzymes with or without Mib1 or Neurl2 and ubiquitin for 1 h at 37°C. Western blots were probed with an FK1 antibody. Data were analyzed using one-way ANOVA (F(3,16) = 8.874, p = 0.0011) followed by post hoc Tukey’s multiple comparisons test (control versus siNeurl2: p = 0.0060; control versus siMib1: p = 0.0027; control versus siNeurl2 + siMib1: p = 0.0025; siNeurl2 versus siMib1: p = 0.9785; siNeurl2 versus siNeurl2 + siMib1: p = 0.9740; siMib1 versus siNeurl2 + siMib1: p > 0.9999). n = 5.

(F) SMN protein stability following siRNA KD of Neurl2, overexpressing FLAG-Mib1, or both. Data were analyzed using one-way ANOVA (8 h: F(2,9) = 9.920, p = 0.0053; 16 h (F(2,9) = 18.93, p = 0.0006) followed by post hoc Tukey’s multiple comparisons test (8 h: control versus siNeurl2: p = 0.0447; 16 h: control versus siNeurl2: p = 0.0036). n = 4.

All data are represented as mean ± SEM. *p < 0.05, **p < 0.01, ****p < 0.0001.

Neurl2 interacts with SMN and regulates its turnover

We next sought to determine whether Neurl2 regulated SMN expression through direct association and ubiquitination. SMA patient fibroblasts were transfected with GFP-Neurl2 or Neurl2-GFP, and endogenous SMN was immunoprecipitated followed by western blot analysis with an anti-GFP antibody. We found robust association of Neurl2 and SMN, confirming interaction (Figure 5B). We next examined the potential mechanism through which Neurl2 affected SMN protein stability. We investigated if, like Mib1, Neurl2 ubiquitinates SMN. We transfected SMA patient-derived fibroblasts with hemagglutinin (HA)-tagged Neurl2 and collected lysates 48 h after transfection. We immunoprecipitated SMN from these lysates and quantified SMN ubiquitination by western blot analysis. We found that overexpressing Neurl2 and Mib1 individually increased SMN ubiquitination (Neurl2: 89% increase, p = 0.0305; Mib1: 320% increase, p < 0.0001) (Figure 5C). Interestingly, knocking down Neurl2 prevented Mib1-dependent SMN ubiquitination (Mib1 versus Mib1 + siNeurl2: 46% decrease in ubiquitination, p < 0.0001). Co-transfecting siRNA to Neurl2 with an siRNA-resistant Neurl2 (Neurl2_res) restored Mib1-dependent SMN ubiquitination, confirming that Neurl2, and not off-target effects of the siRNA, modulates SMN ubiquitination (Mib1 versus Mib1 + siNeurl2 + Neurl2_res: p = 0.7566) (Figure 5C). To determine whether Neurl2 directly ubiquitinates SMN, we performed an in vitro ubiquitination assay. Because we did not know which E2-conjugating enzyme would facilitate Neurl2-dependent SMN ubiquitination, we performed the assay in the presence of each known E2-conjugating enzyme. We found that Neurl2 directly ubiquitinates SMN only in the presence of the E2 enzyme Ubch1 (Figure 5D). Neurl2 has been reported to, in cooperation with Mib1, ubiquitinate and decrease expression of X-Delta (Song et al., 2006). Therefore, we asked whether Mib1 and Neurl2 cooperatively modulated SMN levels. Knocking down Mib1 and Neurl2 separately increases SMN levels in SMA patient cells (siNeurl2: 73% increase, p = 0.0060; siMib1: 81% increase, p = 0.0027). Knocking down Mib1 and Neurl2 together resulted in similar increases in SMN levels (81% increase compared with control) as knocking down each ligase individually, suggesting both enzymes function together or in series (siNeurl2 versus siNeurl2 + siMib1: p = 0.9740; siMib1 versus siNeurl2 + siMib1: p > 0.9999) (Figure 5E). Finally, we investigated whether Neurl2 cooperates with Mib1 to regulate the rate of SMN turnover. We compared the rate of SMN turnover in cells transiently overexpressing Mib1, alone or when Neurl2 is knocked down. Consistent with our previous studies, SMN is more rapidly degraded in cells overexpressing Mib1 (Figure 5F). However, the rate of SMN turnover is slowed when Neurl2 levels are reduced, suggesting that both genes cooperate to regulate SMN turnover (control versus siNeurl2 8 h: p = 0.0447; 16 h: p = 0.0036).

DISCUSSION

The recently approved therapies to treat SMA have dramatically improved survival and reduced motor deficits of patients with the most severe form of SMA. However, many patients with SMA remain severely debilitated, suggesting additional treatments to improve or complement the current therapies are sorely needed. Based on the current SMA landscape, several SMN-dependent and -independent therapeutic strategies are under clinical consideration, including efforts to provide systemic boosting of SMN levels or enhance skeletal muscle size and function. In this study, we opted to utilize a human genome-wide RNAi screen, robust seed correction, and bioinformatics analysis to help identify genes that could potentiate SMN expression in combination with currently approved treatments. Based on the rigorous nature of our screening strategy, we identified a small list of genes and pathways that are likely to alter SMN gene expression through induction of transcription, RNA processing, and post-translational modifications.

Transcriptional regulation of SMN expression

Very little is known about the regulatory mechanisms regulating SMN1 and SMN2 promoter activity, despite several reports describing differential SMN expression in varying tissues throughout development (Echaniz-Laguna et al., 1999; Monani et al., 1999b; Ramos et al., 2019; Rouget et al., 2005). Putative consensus binding sites for the transcriptional repressor YY1 were previously described in the SMN2 promoter (Echaniz-Laguna et al., 1999). YY1 was a highly ranked hit from our current screen and the highest ranked transcription factor. YY1 is thought to induce transcriptional repression by directing the activity of HDACs at gene promoters. Intriguingly, HDAC inhibitors have been shown to increase SMN expression and were previously under clinical consideration for treating SMA (Kissel et al., 2011, 2014; Krosschell et al., 2018; Renusch et al., 2015; Swoboda et al., 2009, 2010). Although numerous off-target effects precluded further clinical consideration of this class of drug, it is possible that identifying transcription factors could provide another avenue for therapeutic intervention in this pathway.

STRING analysis showed several transcription factors clustered around YY1, including HCFC1, GABP, and SUZ12. HCFC1 gain- and loss-of-function mutations have been described in neurodevelopmental disorders, suggesting it plays critical roles in central nervous system development (Gérard et al., 2015; Koufaris et al., 2016; Yu et al., 2013). HCFC1 associates with DNA-binding transcription factors and chromatin modifiers to either activate or repress transcription. One report estimates that close to 90% of promoters that HCFC1 binds to also bind YY1 or GABP (Michaud et al., 2013). Association of GABP with HCFC1 is thought to be essential for HCFC1 to regulate the transcription of several genes (Michaud et al., 2013). Knocking down YY1, GABP, or HCFC1 individually increases SMN expression, suggesting a critical role in SMN transcriptional induction. Therapies that increase SMN exon 7 inclusion are limited by the level of expression of the SMN2 transcript. Identifying pathways that could be targeted to increase the levels of SMN transcript would provide a mechanism to complement the splicing modifiers to further increase SMN expression.

SMN2 splicing modifiers

Outside of replacing the SMN1 gene, maximizing the production of the SMN protein from the SMN2 gene has been the primary therapeutic focus of SMA research. Indeed, one clinically approved treatment for SMA leveraged the discovery of the intronic splicing silencer N1 (ISS-N1) to design antisense oligonucleotides that target this site to suppress SMN exon 7 skipping (Singh et al., 2006). Subsequently, small-molecule screens have identified compounds that promote SMN exon 7 inclusion, and several of these entities are currently under advanced clinical examination (Cheung et al., 2018; Poirier et al., 2018; Ratni et al., 2018; Sturm et al., 2019). Although promoting SMN exon 7 inclusion has remarkable therapeutic benefits, antisense oligonucleotides targeting ISS-N1 do not help all patients, possibly because the splice sites harbor sequence variants that could render them nonfunctional. This fundamental gap in knowledge of the molecular mechanisms regulating SMN splicing prevents additional mechanism-based drug development to further enhance the efficacy of this class of drug.

To date, only a few proteins have been shown to reduce or increase SMN exon 7 inclusion in knockdown and overexpression studies. The C-to-T transition at position 6 in SMN2 exon 7 is thought to alter the delicate balance between multiple positive and negative splicing modulators (Lorson et al., 1999; Singh and Singh, 2018). SRSF1, a member of the family of serine/arginine (SR)-rich proteins, was shown to increase exon 7 inclusion when knocked down or overexpressed, suggesting a complicated role in SMN splicing (Cartegni and Krainer, 2002; Kashima and Manley, 2003; Wee et al., 2014). hnRNP A1/A2 depletion enhances exon 7 inclusion, suggesting an inhibitory role in SMN exon 7 inclusion (Doktor et al., 2011; Kashima et al., 2007a, 2007b; Singh et al., 2013, 2017). In this study, knocking down SRSF3 or hnRNP A1 increases full-length SMN expression consistent with previous findings. We also re-identified U2AF1 and PUF60 that have been previously reported to suppress SMN exon 7 inclusion (Auslander et al., 2020; Hastings et al., 2007). Together, we found that siRNA targeting almost all the previously confirmed genes that promote SMN exon 7 exclusion is indicative of the robustness of the screening strategy.

Nuclear transport of SMN RNA mediates SMN expression

Prior research on stimulating SMN gene expression through the SMN2-derived mRNA has focused on increasing mRNA synthesis, correcting the splicing to promote exon 7 inclusion, or slowing RNA metabolism (Avila et al., 2007; Farooq et al., 2011, 2013; Hua et al., 2008, 2011; Sumner et al., 2003). Increasing evidence suggests that the export of mRNA is a key regulator of gene expression and selective mRNA export (García-Oliver et al., 2012; Wickramasinghe and Laskey, 2015). Here, we provide evidence that increased export of SMN2 mRNA into the cytoplasm leads to increased SMN protein levels, suggesting an additional mechanism to increase SMN protein.

In mammalian cells, pre-mRNA transcripts are recognized by the CPSF complex, which, through recognition of the hexanucleotide A(A/U)UAAA motif within polyadenylation signals, catalyzes pre-mRNA cleavage and initiates polyadenylation (Schönemann et al., 2014). WDR33, CPSF4, FIP1L1, and CPSF1 are core components of the CPSF complex, which have discrete roles in RNA recognition and maturation. CPSF1 and WDR33 form a heterodimer, independently of the other CPSF subunits, and are required and sufficient for recognition of the polyadenylation motif. In this study, we found that knocking down CPSF1 and WDR33, but not CPSF4 and FIP1L1, increased SMN expression, in line with the indispensable roles of CPSF1 and WDR33 in hexanucleotide recognition and binding. Nevertheless, it is unclear why reducing key components of mRNA biogenesis would result in increased expression of the SMN protein. Further exploration revealed that knocking down WDR33, but not CPSF1 or THOC3, altered the levels of SMN pre-mRNA, whereas none of the knockdowns affected the stability of SMN mRNA. Intriguingly, knocking down WDR33, CPSF1, and THOC3 resulted in increased cytoplasmic SMN transcript, suggesting that cytoplasmic trafficking of SMN mRNA was enhanced when each gene is depleted.

The THO/TREX complex is recruited to maturing RNA and facilitates nuclear export of a subset of mRNAs to the cytoplasm for translation (Guria et al., 2011). In the current study, we found that siRNA targeting core components of the THO/TREX complex reduced SMN expression, as expected. However, transient silencing of THOC3 resulted in increased cytoplasmic localization of SMN mRNA and increased SMN protein expression. Silencing of THOC3 has been previously reported to both increase and decrease expression of genes, suggesting tissue-specific, condition-dependent, and temporal roles in gene expression (Gupta and Senthilkumaran, 2020). This sets up the intriguing possibility that SMN mRNA nucleo-cytoplasmic trafficking could be one factor that drives the tissue-dependent and developmental variability in SMN protein expression in disease-relevant tissues, including motor neurons and muscle.

Regulators of SMN protein turnover

We and others have shown that SMN protein is targeted for degradation by the ubiquitin-proteasome system (Abera et al., 2016; Chang et al., 2004; Kwon et al., 2011, 2013). In this study, we show that siRNA targeting enzymes that catalyze protein ubiquitination or the substrate recognition domains of the proteasome increase SMN levels, consistent with our previous findings. We identified an additional E3 ligase, Neurl2, which promotes SMN ubiquitination and its subsequent degradation. Neurl2 ubiquitinates SMN directly in vitro and in cultured cells. We show that Neurl2 is required for Mib1-dependent SMN ubiquitination and turnover, adding an additional layer of post-translational gene regulation. We have previously shown that a small-molecule inhibitor of SMN ubiquitination by Mib1 increases SMN levels and improves the SMA phenotype in a severe mouse model of SMA (Abera et al., 2016). Although studies to identify a clinical candidate that inhibits Mib1-dependent SMN turnover remain ongoing in our lab, finding additional genes that could potentially be more druggable is prudent.

Conclusions

Many human diseases are caused by toxic accumulation of proteins or the inappropriate clearance of proteins resulting in functional deficits. The genes and pathways that regulate the delicate balance between gene transcription and protein degradation have been difficult to identify because of the number of molecular steps involved and the inherent redundancy involved in maintaining protein homeostasis. High-throughput genome-wide gene suppression assays are ideal to begin identifying critical proteins involved in selectively regulating steady-state levels of disease-associated proteins. Using this strategy, we have identified genes that have not previously been implicated in regulating SMN protein levels. We validated several of the hits in SMA patient-derived cells and provided initial molecular characterization of their modes of action.

STAR★METHODS

Detailed methods are provided in the online version of this paper and include the following:

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Barrington Burnett (barrington.burnett@usuhs.edu).

Materials availability

Plasmids generated in this study are available from the lead contact by request.

Data and code availability

The accession number for the genome-wide RNAi screen dataset reported in this paper is PubChem: 1347426.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Cell culture

3813 fibroblasts derived from an SMA patient were maintained in DMEM supplemented with 10% fetal bovine serum and 1% Pencicillin-Streptomycin Glutamine. Cells were incubated at 37°C with 5% CO₂.

METHOD DETAILS

Plasmids

We generated HA-Neurl2 constructs that were not targets of the Neurl2 siRNA (Thermo Fisher, Assay ID: s44413). Neurl2res was generated by introducing silent mutations in the siRNA-targeting region (T702G, C705T, A708C, and G711A) of HA-tagged Neurl2 constructs.

siRNA screening

SMN2-HEK-luciferase reporter cells were reverse transfected with siRNAs in 384-well optical plates (2 μL per well, 0.8 pmol). Each well contained, 20 μL of DMEM medium and 4 μl/ml RNAiMAX. After 30 min incubation, 20 μL of a 37,500 cells per ml suspension was added to each well for reverse transfection. Plates with cells were incubated at 37°C/5% CO2 for a further 72 h. Media was DMEM supplemented with luciferin (1 μM) and B-27 supplement, and the plates were sealed with an optically clear film. Plates were incubated at 37°C incubator and bioluminescence recorded with a PE ViewLux. Pilot and primary screens were similarly performed with a unique set of siRNAs performed. All screens were performed in temperature-controlled enclosures with automated liquid and plate handling. Luciferase signals were normalized to Ambion Silencer Negative Control #2 siRNA, then log10 transformed to achieve normal distribution.

STRING analysis

Gene Ontology (GO) enrichment analysis of the top genes that increased or decreased SMN expression was performed with the “cellular components enrichment” tool of the STRING database (https://string-db.org/). Statistical significance for enrichment by genome-wide false discovery rate (FDR) was calculated using the STRING database (Szklarczyk et al., 2017).

RNAi screen validation

250,000 fibroblasts were plated and transfected with 30 pmol siRNA using Lipofectamine RNAiMAX in Opti-MEM. Cells were collected 48 hours after transfection for further analysis. siRNAs were all purchased from Thermo Fisher. An siRNA-resistant form of Neurl2 was generated by introducing silent mutations in the siRNA-targeting region (G750A, A753C, G756A, and C759T) of HA-tagged Neurl2 cDNA construct using site-directed mutagenesis. Mib1 was overexpressed using a previously described FLAG-Mib1 plasmid (Itoh et al., 2003).

Protein expression analysis

Cells were lysed in 1% NP-40, 50 mM Tris-HCl (pH 8.0), 150 mM NaCl, and protease inhibitor cocktail on ice for 10 minutes. Protein concentrations were determined using a DC assay according to the manufacturer’s protocol. Protein lysates were resolved by 4%–12% SDS-PAGE and transferred to PVDF membranes. Membranes were blocked in 5% milk in TBST for 1 hour at room temperature. Membranes were incubated with primary antibody (mouse anti-SMN, 1:1000) followed by incubation in secondary antibody. Mouse anti-β-actin peroxidase was used as a loading control. Western blots were quantified using ImageJ.

RNA isolation and gene expression analysis

Total RNA was extracted using the QIAzol-chloroform method as previously described. RNA was purified using the RNeasy kit using the manufacturer’s instructions. cDNA was synthesized from one microgram of RNA using the High-Capacity cDNA Reverse Transcriptase Kit. qRT-PCR was performed using Taqman reagents and TaqMan primers were all purchased from Thermo Fisher.

SMN pre-mRNA

SMN pre-mRNA levels were determined as previously described (d’Ydewalle et al., 2017). Briefly, SMN pre-mRNA was amplified using reverse transcriptase and a strand specific primer. RT-PCR was used to determine relative SMN pre-mRNA levels using TaqMan reagents and a custom TaqMan primer.

SMN2 mRNA splicing

SMN2 transcripts were amplified from cDNA using PCR SuperMix. PCR products were run on a 2% agarose gel at 150 V for 30 minutes. The full-length SMN (421 bp) and the SMNΔ7 (367 bp) products were quantified using ImageJ. The ratio of full-length SMN to SMNΔ7 was calculated.

mRNA stability

Actinomycin D (5 μg/mL) was added to cells 48 hours after transfection with siRNA. Cells were collected 0, 1, 3, 5, and 8 hours after the addition of actinomycin D. RNA was extracted and purified from cells and then transcribed into cDNA. Relative SMN mRNA levels over time were determined using qRT-PCR.

mRNA Fractionation

The SurePrep Nuclear or Cytoplasmic RNA Purification Kit was used for mRNA fractionation. Briefly, cells were lysed using the provided lysis solution and the lysate was spun down for 3 minutes at maximum speed in a benchtop centrifuge to separate the nuclear and cytoplasmic fractions. Cytoplasmic mRNA was isolated from the supernatant and nuclear mRNA was isolated from the pellet. cDNA was synthesized from one microgram of RNA using the High Capacity cDNA Reverse Transcriptase Kit. qRT-PCR was performed using Taqman reagents and TaqMan primers were all purchased from Thermo Fisher. Nuclear SMN2 mRNA expression was normalized to MALAT1 mRNA and cytoplasmic SMN2 mRNA expression was normalized to ACTB mRNA. Purity of nuclear and cytoplasmic fractions was confirmed by examining expression of ACTB and MALAT1 mRNA using qRT-PCR.

Immunoprecipitation

Fibroblasts were transfected with either GFP-Neurl2 or Neurl2-GFP constructs. SMN was immunoprecipitated using an anti-SMN antibody and samples were resolved by 4%–12% SDS-PAGE. Western blots were probed with an anti-GFP antibody.

In vitro ubiquitination

In vitro ubiquitination reactions were performed as previously described (Kwon et al., 2013). Briefly, recombinant SMN protein was incubated with ubiquitin, an E1 ubiquitin-activating enzyme, an E2 ubiquitin-conjugating enzyme, and recombinant Mib1 or Neurl2 in reaction buffer (50 mM Tris-HCl, pH 7.5, 2 mM ATP, 4 mM MgCl₂, 2 mM dithiothreitol (DTT)) for 1 hour at 37°C. Reactions were quenching with 5X Laemmli sample buffer, boiled for 5 minutes, and separated via SDS-PAGE. Western blots were probed with an antibody against polyubiquitin.

Pulse-chase protein labeling

Pulse-chase analysis was performed as previously described (Burnett et al., 2009). Briefly, fibroblasts derived from patients with SMA were incubated in cysteine-methionine-free medium for 2 h followed by incubation in methionine-free medium containing 100 μCi of ³⁵S-labeled cysteine-methionine for 1 h. After labeling, cells were washed with culture medium containing an excess of unlabeled cysteine and methionine and then further incubated with the same medium. Cells were collected at the indicated time points. SMN was immunoprecipitated using an anti-SMN antibody. Samples were separated on SDS-10% PAGE which were then dried and exposed to a phosphoimager screen. The signal was visualized using a Storm phosphorimager system. Densitometry analysis of SMN protein bands was carried out using ImageQuant PhosphorImager software.

QUANTIFICATION AND STATISTICAL ANALYSIS

All experiments were repeated at least three times, and quantitative data are presented as the mean ± SEM. All statistical analyses were performed using GraphPad Prism v7.05. Outliers were determined using Grubbs’ test where α = 0.05. The distribution of data points was assessed using the Shapiro-Wilk normality test and homogeneity of variances was assessed using either an F test or a Brown-Forsythe test. Data were analyzed using either an unpaired t test (for comparisons with one independent variable and 2 groups), or a one-way analysis of variance (ANOVA; for comparisons with one independent variable and > 2 groups). Post hoc analyses were performed using either Dunnett’s or Tukey’s multiple comparisons test. A p value < 0.05 was considered statistically significant.

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Mouse anti-SMN	BD Biosciences	Cat# 610647; RRID: AB_397973
Goat anti-mouse IgG, HRP conjugated	Enzo Life Sciences	Cat# BML-SA204-0100; RRID: AB_2051534
Mouse anti-β-actin-peroxidase	Millipore Sigma	Cat# A3854; RRID: AB_262011
Living Colors GFP Polyclonal Antibody	Takara Bio	Cat# 632592; RRID: AB_2336883
Polyubiquitinylated conjugates monoclonal (FK1)	Enzo Life Sciences	Cat# BML-PW8805-0500; RRID: AB_2052280
Mouse anti-β-tubulin	Millipore Sigma	Cat# T8453; RRID: AB_1841224
Chemicals, peptides, and recombinant proteins
Lipofectamine RNAi MAX	Thermo Fisher	Cat# 13778150
Opti-MEM	Thermo Fisher	Cat# 11058021
Luciferin	Promega	Cat# E1601
B-27 supplement	Thermo Fisher	Cat# 17504044
QIAzol Lysis Reagent	QIAGEN	Cat# 79306
iTaq Universal Probes Supermix	Bio-Rad	Cat# 1725134
Actinomycin D	Millipore Sigma	Cat# A9415
Recombinant SMN Protein	Enzo Life Sciences	Cat# ADI-NBP-201-050
Ubiquitin-activating enzyme (UBE1)	Boston Biochem	Cat# E-305
Human recombinant UBCH5B	Enzo Life Sciences	Cat# BML-UW9060-0100
Human recombinant UBCH1	Enzo Life Sciences	Cat# BML-UW9020-0100
Human recombinant Mindbomb-1 Protein	Origene	Cat# TP321377
Human recombinant Neurl2 Protein	Abcam	Cat# ab171589
Human recombinant ubiquitin	Enzo Life Sciences	Cat# BML-UW8610-0001
35S-labeled cysteine-methionine	Perkin Elmer	Cat# NEG072002MC
PCR Supermix	Thermo Fisher	Cat# 10572014
Critical commercial assays
High Capacity cDNA Reverse Transcriptase Kit	Thermo Fisher	Cat# 4368813
RNeasy Mini Kit	QIAGEN	Cat# 74106
Sure Prep Nuclear or Cytoplasmic RNA Purification Kit	Fisher Scientific	Cat# BP280550
Deposited data
RNAi screen	This manuscript	PubChem: 1347426
Experimental models: Cell lines
SMN2-HEK-luciferase reporter cells	Dr. Elliot Androphy, Indiana University (Zhang et al., 2001)	N/A
GM03813 fibroblasts	Coriell	Cat# GM03813; RRID: CVCL_F172
Oligonucleotides
See Table S4		N/A
Recombinant DNA
GFP-Neurl2	This manuscript	N/A
Neurl2-GFP	This manuscript	N/A
HA-Neurl2	This manuscript	N/A
FLAG-Mib1	Dr. Ajay B. Chitnis, NIH (Itoh et al., 2003)	N/A
HA-Neurl2res	This manuscript	N/A
Software and algorithms
ImageJ	NIH	https://imagej.nih.gov/ij/
GraphPad Prism	Graphpad	https://www.graphpad.com
ImageQuant PhosphorImager	Molecular Dynamics	N/A
Other
PE ViewLux	Perkin Elmer	https://Perkinelmer.com

Highlights.

• Genome-wide RNAi screen identifies genes that regulate SMN expression

• Knockdown of RNA processing genes increases SMN through increased nuclear export

• Knockdown of the E3 ligase Neurl2 increases SMN protein levels

• Neurl2 works with Mib1 to ubiquitinate and promote the degradation of SMN