How RNA structure dictates the usage of a critical exon of spinal muscular atrophy gene

Abstract

Role of RNA structure in pre-mRNA splicing has been implicated for several critical exons associated with genetic disorders. However, much of the structural studies linked to pre-mRNA splicing regulation are limited to terminal stem-loop structures (hairpins) sequestering splice sites. In few instances, role of long-distance interactions is implicated as the major determinant of splicing regulation. With the recent surge of reports of circular RNA (circRNAs) generated by backsplicing, role of Alu-associated RNA structures formed by long-range interactions are taking central stage. Humans contain two nearly identical copies of Survival Motor Neuron (SMN) genes, SMN1 and SMN2. Deletion or mutation of SMN1 coupled with inability of SMN2 to compensate for the loss of SMN1 due to exon 7 causes spinal muscular atrophy (SMA), one of the leading genetic diseases of children. In this review, we describe how structural elements formed by both local and long-distance interactions are being exploited to modulate SMN2 exon 7 splicing as a potential therapy for SMA. We also discuss how Alu-associated secondary structure modulate generation of a vast repertoire of SMN circRNAs.

1. Introduction

Alternative pre-mRNA splicing enhances the coding potential of eukaryotic genome by generating multiple transcripts from a single gene [1–4]. All human genes with two or more exons have potential to be alternatively spliced. Complexity of alternative splicing multiplies with the increasing number of exons within a gene. Pre-mRNA splicing can occur in both directions, forward and backward [5]. While forward splicing generates linear transcripts, including messenger RNAs (mRNAs) and long non-coding RNAs (lncRNAs), backsplicing produces circular RNAs (circRNAs) [5–7]. Based on the experimental validations in case of Survival Motor Neuron (SMN) gene, a 30-kb long human gene has capability to generate hundreds of transcripts if forward and back splicing are appropriately factored [8–11]. Not all transcripts generated by alternative splicing are stable due to RNA surveillance mechanisms, which preferentially degrade transcripts with premature termination codons (PTCs) and/or other suboptimal features [12,13]. Hence, the abundance of transcripts generated by a gene cannot be accurately captured. Arrangement of cis-elements on a pre-mRNA coupled with the relative expression of transacting factors are the drivers of alternative splicing [14,15]. Depending upon the context, similar cis-element may have positive or negative effect on alternative splicing of an exon [16,17].

Context of an exon is defined by multiple factors, including size of the exon, strength of the splice sites, positioning of cis-elements within the exon and the flanking intronic sequences [14–21]. RNA structure adds an important dimension to the context of an exon as it modulates accessibility of the cis-elements and shortens the distance between cis-elements located far apart [22–28]. In case of backsplicing that generates circRNAs, RNA structure plays a pivotal role in bringing the downstream 5′ splice site (5′ss) in close proximity to the upstream 3′ss [29]. RNA structure-mediated alternative splicing events are sensitive to the surrounding environment that affects structure, including temperature, ionic strength and the interacting molecules [25,30–32]. Consequently, RNA structure may serve as a trigger for differential splicing of exons under changed environment such as thermal, osmotic and oxidative stress. In living cells splicing is coupled to transcription and vice versa [33]. Chromatin structure can influence the context by modulating the rate of transcription and consequently determining the outcome of alternative splicing [34,35]. Protein factors and small molecules that modulate the rate of transcription through modification of chromatin structure can have effect on the context of an exon as well [33]. There is also evidence to suggest that the nascent pre-mRNA structure influences transcription and consequently modulates co-transcriptional splicing regulation [36].

Humans have two nearly identical copies of Survival Motor Neuron (SMN) gene: SMN1 and SMN2 [37]. Both SMN genes code for SMN, an essential protein linked to diverse functions, including snRNP assembly, pre-mRNA splicing, transcription, translation, stress granule formation, macromolecular trafficking and maintenance of the cytoskeletal dynamics [38–40]. A critical C-to-T mutation at the 6^th position (C6U substitution in RNA) of exon 7 distinguishes SMN2 from SMN1 [41,42]. Although C6U substitution alone is sufficient to trigger SMN2 exon 7 skipping [41], more than fifty cis-elements and transacting factors have been implicated in regulation of SMN exon 7 splicing (Fig.1A). While a complete description of all regulatory elements of SMN is beyond the scope of this report, reviews published on this topic over the recent years capture how our understanding of SMN exon 7 splicing regulation is continuing to evolve [43–46]. The overriding evidence support that the effect of a disease-causing exonic mutation (such as C6U substitution within SMN2 exon 7) could be overcome by compensatory mutations within both exonic and intronic sequences. In this review, we focus on a structure as a modulator of SMN exon 7 splicing.

Figure 1.

Regulation of SMN exon 7 splicing. (A) Diagrammatic representation of intronic and exonic cis-elements as well as trans-acting factors that modulate SMN exon 7 splicing. Upper-case letters signify exonic sequences, small-case letters, intronic sequences. Exons and introns are also shown as colored boxes and lines, respectively. Numbering of nucleotides, neutral and positive, starts from the first exonic and intronic position, respectively. The 5′ and 3′ splice sites (5′ss and 3′ss) are indicated by arrows. Cr1 and Cr2 represent cryptic 5′ splice sites recently reported [59]. Negative and positive regulators of exon 7 splicing are indicated by (−) and (+), respectively. For details see [28,46]. (B) Effect of Spinraza™ on SMN2 exon 7 splicing. Exons are represented by colored boxes, introns by lines. First 27 nts of intron 7 with highlighted ISS-N1 element are given. Annealing positions of Spinraza™ to ISS-N1 target are shown. SMN2 pre-mRNA splicing results in predominant skipping of exon 7 in mRNA (SMN2Δ7) and production of truncated SMN protein (SMNΔ7). Sequestration of ISS-N1 by Spinraza™ promote SMN2 exon 7 inclusion (SMN2FL) and as a consequence increases levels of full-length SMN protein (SMN).

Due to predominant skipping of SMN2 exon 7, SMN2 contributes very little towards the generation of the overall pool of SMN in human cells (Fig. 1B). Consequently, reduction in SMN levels due to the loss of SMN1 causes spinal muscular atrophy (SMA), a leading genetic disease of children and infants [47]. While degeneration of motor neurons is the first manifestation of the disease, peripheral tissues are also affected in SMA [48,49]. In addition, based on the severe impairment of male reproductive organ development and male infertility observed in mild SMA mouse models, the disease appears to be gender-specific [50–52]. Incidentally, testis produces high levels of SMN due to an adult-specific splicing switch from exclusion to mostly inclusion of SMN2 exon 7 [50]. The mechanism of this splicing switch remains unknown. In general, SMN2 is dispensable for survival of humans. However, testis-specific switch of SMN2 exon 7 splicing to fulfill the demand for an elevated levels of SMN during testicular development offers solid rationale to why humans have retained copies of SMN2 during evolution. Considering SMN2 is almost universally present in SMA patients, correction of SMN2 exon 7 splicing provides one of the best avenues for SMA therapy.

Nusinersen (Spinraza™), the first drug approved by FDA in 2016 for the treatment of SMA, acts through correction of SMN2 exon 7 splicing [53,54]. Nusinersen is an antisense oligonucleotide (ASO) that sequesters Intronic Splicing Silencer N1 (ISS-N1) located within intron 7 (Fig. 1B) [55,56]. Positioned within a complex structural context, ISS-N1 harbors multiple cis-elements, including a GC-rich sequence and two hnRNP motifs and a cryptic 5′ss, Cr1 (Fig. 1A) [57–59]. Two putative TIA1 binding sites are located immediately downstream of ISS-N1 (Fig. 1A) [60]. Another cryptic 5′ss, Cr2, located 28 nucleotides (nts) downstream of the Cr1 falls within the TIA1 binding site (Fig. 1A) [59]. Similar to ISS-N1 targeting ASOs, engineered U1 snRNPs (eU1s) annealing to ISS-N1 or sequences in its vicinity also promote SMN2 exon 7 inclusion [59,61–63]. It is proposed that the stimulatory effect of ISS-N1 targeting ASO is at least in part due to the structural rearrangement at the 5′ss of exon 7 [56]. Similar mechanism could hold true for the stimulatory effects of eU1s that target ISS-N1 or nearby sequences. In case when the authentic 5′ss of exon 7 is mutated, eU1s annealing to ISS-N1 or sequences in the vicinity preferentially activate Cr1 [59,63]. Here again, eU1-induced structural arrangement could play an important role in bringing Cr1 and the 3′ss of intron 7 in close proximity. A number of small molecules that modulate SMN2 exon 7 splicing with high specificity are at different stages of the pre-clinical and/or clinical development [64–67]. At least two of these small molecules are shown to interact with RNA structures within SMN pre-mRNA [66,67].

2. Role of local structures formed by exon 7

As indicated by the results of in vitro structure probing, most of the 54 nts of human SMN exon 7 is engaged in the formation of RNA secondary structures comprised of two terminal stem loops (TSLs) and an internal stem (IS) (Fig. 2) [68,69]. TSL1 and TSL2 sequester the 3′ss and 5′ss, respectively (Fig. 2) [23]. The IS1 is formed by the central region of exon 7 and intron 6 sequence (Fig. 2) [23]. Here we discuss the role of all probed structures in modulation of SMN exon 7 splicing. Where relevant, we also discuss potential role of the predicted secondary structures involved in SMN exon 7 splicing regulation.

Figure 2.

Local structures formed by SMN exon 7 and adjacent upstream/downstream intronic sequences. Interactions of small exon 7-splicing modifying compounds with a local structure. (A) Secondary structure of SMN2 exon 7 and its surrounding intronic sequences derived from enzymatic structure probing [23]. Existence of TSL2 and its effect on exon 7 splicing was also confirmed by mutational analysis. Intron 6 and intron 7 sequences are shown in small-case green and blue letters, respectively. Exon 7 sequence is shown in upper-case black letters. Numbering of nucleotides, neutral, positive and negative, starts from the first position in exon 7, first position of intron 7 and the last position in intron 6, respectively. The splice sites of exon 7 are indicated by the red arrows. IS1 structure is boxed. Nucleotides in magenta indicate site-specific mutations, single and compensatory, used to confirm the existence of TSL2 and its effect on exon 7 splicing [23]. Abbreviation: IL, internal loop; IS, internal stem; TSL, terminal stem loop. (B) Interactions of a small exon 7-splicing modifying compound, SMN-C3, with a local structure. Secondary structure of SMN2 exon 7 and its surrounding intronic sequences was derived from in-cell SHAPE-mutation profiling [66]. Markings and abbreviations are the same as in (A). Chemical structure of the small compound, SMN-C3, that increases inclusion of exon 7 is shown. Binding site of SMN-C3 (in the context of a double strand) is highlighted with red box. Residues, whose sensitivity to a structure probing agent, NAI, is increased and decreased upon SMN-C3 binding are highlighted in a pink and a blue circle, respectively [66]. (C) Interactions of a small exon 7-splicing modifying compound, PK4C9, with TSL2 structure. Upper-case letters signify exonic sequence, small-case letters in blue, intron 7 sequence. Nucleotides in red indicate site-specific mutations, single and double, used to confirm the effect of destabilization or stabilization TSL2 on exon 7 splicing. The 5′ss of exon 7 is indicated by the red arrows. Chemical structure of PK4C9 is shown. In the lower panel PK4C9 is represented as a red star. Upon PK4C9 binding to TSL2, several residues at the base of the stem highlighted with pink circles form hydrogen bonds, hydrophobic interactions or stack with the compound [67]. These interactions destabilize TSL2 and increase the accessibility of the 5′ss. In addition, binding of PK4C9 to TSL2 stabilizes the triloop of this RNA structure.

2.1. Role of TSL1

The TSL1 is formed by the first sixteen nucleotides of exon 7 (Fig. 2A) [68,69]. Six of these sixteen nucleotides are located in the loop of TSL1; whereas, remaining ten residues form the stem. C6U substitution that distinguishes SMN2 from SMN1 is located at the loop-closing position and is predicted to increase the stem size of TSL1 from 5-bp to 6-bp duplex (Fig. 2) [68,69]. However, the results of enzymatic structure probing showed unpaired nature of 6U, suggesting that the stem size of TSL1 is not changed in SMN2 [23]. In this scenario, the UAGACA motif created by C6U substitution falls within the loop of TSL1. It has been proposed that UAGACA motif serves as a strong binding site for hnRNP A1, a negative regulator of SMN2 exon 7 splicing [70]. Hence, UAGACA location within the loop makes this motif highly accessible for interactions with hnRNP A1. In case of SMN1, the loop of TSL1 is represented by the CAGACA motif that overlaps with CAGACAA motif, which is proposed to interact with ASF/SF2, a positive regulator of SMN1 exon 7 splicing [71]. Aside from the loop harboring protein-binding motifs, mutations within the stem region of TSL1 promote SMN2 exon 7 inclusion, suggesting that the entire TSL1 structure is inhibitory [69]. Based on the results of extensive point mutations, we concluded that TSL1 is located within an extended inhibitory context (Exinct) spanning from the 3^rd to 15^th positions of exon 7 [69]. An antisense microwalk independently validated the presence of Exinct [72]. In particular, ASOs disrupting TSL1 promoted SMN2 exon 7 inclusion [72]. In vivo selection serves as one of the unbiased methods to interrogate the position-specific significance of nucleotide residues in alternative splicing [73]. Incidentally, SMN1 exon 7 was the first entire exon subjected to in vivo selection [74]. The results of in vivo selection of SMN1 exon 7 supported the presence of Exinct [74]. Of note, unlike SMN1 exon 7, SMN2 exon 7 has a strengthened Exinct due to C6U substitution. More recently, a machine-learning-based computational analysis validated the findings of in vivo selection and suggested that the Exinct-like motifs created by genetic mutations are associated with genetic diseases [75].

Employing in vitro Selective 2′-Hydroxyl Acylation analyzed by Primer Extension (SHAPE)-based RNA structure probing method, an alternative TSL structure was proposed at the location previously designated as TSL1 (Fig. 2A) [23, 66]. Disparity between these two structures at the beginning of exon 7 could be due to two different in vitro methods used for structure probing [23,66]. Also, differences in structures could be due to different lengths of sequences analyzed for in vitro structure probing. Interestingly, only a part of the alternative structure at the beginning of exon 7 was found to be preserved when sequences were analyzed by in-cell SHAPE structure probing method [66]. Future studies will determine if RNA structures adopt different confirmations based on the presence of the exon 7-interacting factors in the cell.

2.2. Role of TSL2

The TSL2 is the first experimentally validated structure of SMN pre-mRNA (Fig. 2) [23]. TSL2 is formed by a 19-nt long sequence encompassing the last seventeen and the first two residues of exon 7 and intron 7, respectively. Based on the results of in vivo selection of the entire exon 7, TSL2 encompasses an inhibitory region called the “3′-Cluster” [73,74]. Antisense microwalk as well as the machine-learning-based computational analysis validate the inhibitory nature of the 3′-Cluster [72,75]. Supporting the negative effect of TSL2, point mutations that abrogate TSL2 stimulate SMN2 exon 7 inclusion (Fig. 2C) [23]. Conversely, compensatory mutations that restore TSL2 promote SMN2 exon 7 skipping (Fig. 2C) [23]. Located within TSL2, wild-type A residue at the last position (54A) of exon 7 has been determined to be highly inhibitory [74]. Consistently, an A-to-G substitution (A54G) at the last position of exon 7 fully restores SMN2 exon 7 inclusion. The stimulatory effect of A54G substitution happens to be so pronounced that absence of most known positive cis-elements within exon 7 is tolerated [74]. A54G substitution is predicted to destabilize TSL2 as it changes the canonical U-A base pair into wobble U-G base pair. In addition, 54G increases the base pairing between endogenous U1 snRNA and the 5′ss of exon 7. An eU1 with extended base-pairing with 5′ss of exon 7 fully restored SMN2 exon 7 inclusion [23]. These findings support that the poor recruitment of U1 snRNP is one of the limiting factors for SMN2 exon 7 splicing. Recently, a small compound PK4C9 that promotes SMN2 exon 7 inclusion, has been shown to interact with TSL2 [67]. Based on this interaction, authors propose that PK4C9 favors recruitment of U1 snRNP through remodeling of the 5′ss of exon 7.

2.3. Role of IS1

Based on the enzymatic structure probing, IS1 is a 10-bp long duplex in the middle of SMN exon 7 (Fig. 2A) [23]. While the purine-rich 3′-strand of IS1 is provided by exon 7, the pyrimidine-rich 5′-strand of IS1 comes from intron 6. IL1, a 4-nt (ACAG) internal loop, bridges the 5′-strand of IS1 with the 5′-strand of TSL1 (Fig. 2A). IL2, a 6-nt purine-rich (AAAAG) internal loop, bridges the 3′-strand of TSL1 with the 3′-strand of IS1 (Fig. 2A). IL3, another 5-nt internal loop, bridges the 3′-strand of IS1 with the 5′-strand of TSL2 (Fig. 2A). Based on the results of in vivo selection, sequences involved in the formation of IL2, IS1 and IL3 appear to be stimulatory [74]. Incidentally, IL2 overlaps with the binding site of Tra2-β1, which recruits other factors to exon 7 (Fig. 1A). List of factors interacting with exon 7 is continuing to grow, although no effort has yet been made to associate these interactions with specific structures. It is likely that the single stranded regions, including IL2 and IL3 provide primary binding sites for positive regulators. Supporting this argument, mutations within either IL2 or IL3 have been found to promote SMN1 exon 7 skipping [74].

Unlike enzymatic structure probing, results of a recent in vitro chemical structure probing by SHAPE revealed a shorter IS1 (Fig. 2B) [23,66]. Further, findings of in vitro SHAPE showed different structural context upstream and downstream of IS1. However, results of in cell SHAPE only partially validated the structural context revealed by in vitro SHAPE. Interestingly, a small compound, SMN-C3, that stimulates SMN2 exon 7 inclusion has been proposed to interact with IS1. However, this interaction is proposed to be modulated by protein factor(s) [66].

3. Role of structural context of intron 7

Secondary structure of SMN2 intron 7 deduced from the chemical structure probing was reported in 2013 (Fig. 3) [27]. The key findings of the probed secondary structure are the presence of internal stems formed by unique long-distance interactions (LDIs). Here we describe functional significance of some of these LDIs and as well as other structures formed by intron 7.

Figure 3.

Secondary structure of SMN intron 7. (A) SHAPE-derived structure of intron 7 [27]. The structure is based on combined probing results obtained with multiple extension primers. An exon 7/intron 7 junction as well as the 5′ss are indicated. Exon 8 is represented by a green box. Numbering of nucleotides, neutral and negative, starts from the first position of intron 7 and the last position of exon 7, respectively. Binding sites of hnRNP A1/A2 and TIA1 are highlighted in pink and green, respectively. Residues with normalized 1M7 reactivity > 0.5 are marked with large circles, while residues with normalized 1M7 reactivity between 0.3 and 0.5 are indicated with small circles. Broken lines indicate potential base-pairing. Positions corresponding to Reverse Transcriptase (RTase) falloffs are marked with “F”. Residues in red constitute a region whose structure was not determined due to RTase fall off. Abbreviations: ISTL, internal stem formed by a long-distance interaction; LS1, long-distance interaction site 1; TSL, terminal stem-loop; ISS-N2, intronic splicing silencer 2. (B) Site-specific mutations in ISTL1 and adjacent regions that effect this structure and cause changes in a pattern in exon 7 splicing with/without antisense oligonucleotides [27]. Arrows and letters in red indicate positions and identities of mutations, single and compensatory, that break (when single) and restore (when combined) ISTL1 structure. Arrows and letters in green indicate positions and identities of mutations deep within intron 7, which increase the size of the stem and cause increase in exon 7 skipping.

3.1. Role of ISTL1

Internal stem formed by the long-distance interaction 1 (ISTL1) is a unique intra-intronic structure located within intron 7 [27]. The two strands of the 8-bp ISTL1 are separated from each other by 279 nts. Together with TSL2, ISTL1 sequesters the entire 5′ss of exon 7. In addition, the 5′-strand of ISTL1 encompasses half of the GC-rich motif that has been confirmed to be an effective target for a short (8-mer) ASO [57,76]. The last base of the 5′-strand of ISTL1 is a C residue located at the 10^th intronic position (¹⁰C), which also happens to be the first residue of the 15-nt long ISS-N1 [27]. The significance of ¹⁰C first emerged when two 14-nt long ASOs, F14 and L14, targeting ISS-N1 produced opposite effects on SMN2 exon 7 splicing [18]. While F14 sequestered the first 14 residues (including ¹⁰C) of ISS-N1 and promoted exon 7 inclusion, L14 sequestered the last 14 residues (excluding ¹⁰C) of ISS-N1 and promoted exon 7 skipping [18]. Results were validated using ASOs with two different chemistries (locked nucleic acid and 2′O-methyl modifications), ruling out any chemistry-specific effect [18]. Subsequent experiments confirmed that F14 destabilizes ISTL1 due to sequestration of ¹⁰C, whereas, L14 stabilizes ISTL1 due to formation of a duplex with the remainder of the ISS-N1 sequence. Supporting this argument, deletion of deep intronic sequences encompassing the 3′-strand of ISTL1 eliminated the negative effect associated with L14 [27]. Also, increasing the length of ISTL1 duplex substantially enhanced the skipping of SMN2 exon 7. These results provided one of the rarest examples of how a single nucleotide shift in the annealing position of an ASO could flip the effect on splicing. Findings also suggested that the effect of an ISS-N1-annealing ASO may not be due to the displacement of the negative factor(s) (hnRNP A1/A2) only, it could also be due to the change in structural context upstream and/or downstream of the ASO annealing position. While ISTL1 is the first validated regulatory structure formed by an LDI in the context of a human genetic disease, a similar structure in intron 3 of Proteolipid Protein 1 (PLP1) gene has been implicated in Pelizaeus-Merzbacher disease (PMD) [77].

3.2. Role of other ISTLs

In addition to ISTL1, additional structures formed by LDIs within intron 7 include ISTL2, ISTL3 and ISTL4 (Fig. 3) [27]. Two strands of ISTL2, ISTL3 and ISTL4 are separated from each other by 226, 85 and 55 nts, respectively. Similar to ISTL1, ISTL2 and ISTL3 exert negative effect on SMN2 splicing, although the effect is not as pronounced as in case of ISLT1. The negative effect of ISTL2 could at least in part be attributed to the sequestration of TIA1 binding site located immediately downstream of the ISS-N1. Interestingly, a continuous intronic sequence, termed ISS-N2, provides the 3′-strands of ISTL1, ISTL2 and ISTL3. Deletion or an ASO-mediated sequestration of ISS-N2 confers stimulatory effect on SMN2 exon 7 inclusion [27]. Unlike ISTL1, ISTL2 and ISTL3, the significance of ISTL4 in exon 7 splicing regulation has not yet been established.

3.3. Other structures within intron 7

Results of structure probing supported the formation of several TSLs, ILs and bulges (Fig. 3) [27]. Significance of some of these structures could be inferred from the location of the critical cis-elements within these structures. For instance, one of the putative hnRNP A1 binding sides within ISS-N1 is located within an IL. This loop also harbors four of the eight residues of the GC-rich motif that serves as the target for the shortest ASO. Presence of this loop makes ISS-N1 and GC-rich sequence highly accessible for targeting by ASO. Other putative hnRNP A1 motif associated with ISS-N1 is sequestered within TSL3, which also locks one of the TIA1 binding sites located downstream of the ISS-N1. It is likely that the stimulatory effect of the ISS-N1-targeting ASO is linked at least in part to freeing of the TIA1 motif. Sequences downstream of ISTL4 form two TSLs, TSL6 and TSL7. Deletion of intron 7 sequences encompassing TSL6 and TSL7 moderately stimulated SMN2 exon 7 inclusion [27]. Future studies will determine if these and other structures within intron 7 play critical role in regulation of exon 7 splicing.

4. Role of RNA structure in backsplicing

Primate-specific Alu elements are the most abundant short interspersed nuclear elements (SINEs) totaling more than one million copies and accounting for ~11% of the human genome [78,79]. Surprisingly, a whopping 39% sequence of human SMN genes are represented by Alu elements as inverted repeats in the intronic regions [80]. Presence of inverted Alu repeats results in formation of double stranded structures that are conducive for the generation of circRNAs through backsplicing [81,82]. Indeed, a recent report confirmed the existence of a large repertoire of circRNAs generated from SMN genes (Fig. 4) [11]. Based on the nature of exons, SMN circRNAs are broadly categorized into four types [11]. While type 1 circRNAs harbor one or more of the early exons (exons 1 through 4), type 2 circRNAs harbor middle exons in addition to early exons (exons 1 through 7). Some of the type 1 circRNAs contain novel exons derived from sequences within intron 1. Type 3 circRNAs harbor the 3′-terminal exons with or without upstream exons. Some of the type 3 circRNAs also contain one or more of the four novel exons derived from intergenic region located downstream of the annotated SMN gene. Type 4 circRNAs are generated by a combination of backsplicing and trans-splicing, such that one or more exons in these circRNAs originate from other genes. Overall, more than fifty SMN circRNAs are universally produced in all cell types examined, with type 1 circRNAs being the most predominant [11]. All SMN circRNAs are generated by backsplicing from canonical splice sites. Both SMN1 and SMN2 generate circRNAs with near equal frequencies. Most common type 1 circRNAs produced by SMN are C2A-2B-3–4, C2B-3–4 and C3–4 (Fig. 4). These three circRNAs share the same 5′ss of exon 4 suggesting a random pairing between the 5′ss of exon 4 with the 3′ss of any of the upstream exons 2A, 2B and 3. Predicted secondary structures support such pairing through RNA:RNA duplex formation between sequences at the 5′ss of exon 4 and the 3′ss of the upstream exons (Fig. 4A) [11]. C5–6 is the predominant circRNA among type 2 SMN circRNAs (Fig. 4) [11]. Formation of C5–6 is likely to be facilitated by inverted Alu repeats that are predicted form secondary structures bringing the 5′ss of exon 6 and the 3′ss of the exon 5 in close proximity (Fig. 4A). Among type 3 circRNAs of SMN, C6–7-8A and C6–7-8A-9tr1–10 are the most common. Here again, generation of type 3 circRNAs is likely facilitated by Alu-associated secondary structures that bring the 3′ss of exon 6 in close proximity with 5′ss of any of the downstream exons (Fig. 4B). Type 4 circRNAs of SMN are generated by incorporation exonic sequences from the neighboring Alu-rich genes, including ERBIN and SERF1 [11].

Figure 4.

Repertoire of circRNAs generated from the SMN genes. (A) Diagrammatic representation of circRNAs identified using primers that anneal to exons 2a, 2b, 3, 4 and 5. A genomic overview of the SMN gene layout is given. Exons are indicated by colored shapes, introns by lines. Exon sizes are given using numbers in black located below exons, intron sizes are shown using numbers in grey located above introns. A newly identified 5′ss within exon 8 is indicated with black arrow. Colored arrows show canonical backsplicing events. Arrow thickness reflects the estimated prevalence of each event. Schematic of the likely role of the Alu-associated secondary structures in the generation of the most abundant types 1 and 2 circRNAs are given (diagram not to the scale). (B) Diagrammatic representation of circRNAs identified using primers that anneal to exons 6, 6b, 7 and 8. A genomic overview of the 3′ portion of the SMN gene layout is given. Labeling and color coding are the same as in (A). Schematic of the likely role of the Alu-associated secondary structures in the generation of the most abundant type 3 circRNAs are given (diagram not to the scale).

Generation of SMN circRNAs appears to be a posttranscriptional event, as suppression of transcription elongation had no appreciable effect on the relative abundance of SMN circRNAs. At the same time, depletion of DHX9, an RNA helicase that specifically unwinds Alu-associated secondary structures, enhanced expression of several SMN circRNAs, including C3–4, one of the most predominant circRNAs of SMN [11]. Interestingly, depletion of DHX9 also promoted skipping of exons 3 and 4 in various combinations, suggesting that the C3–4 is generated from intronic lariats harboring these exons. Considering the 5′ss of exon 4 is used to generate several circRNAs, including C2A-2B-3–4, C2B-3–4 and C3–4, an enhanced expression of C3–4 under the DHX9-depleted conditions consequently led to decreased expression of C2A-2B-3–4 and C2B-3–4. These findings also suggested that the base pairing between inverted Alu elements located upstream and downstream of exons 3 and 4, respectively, is under the control of DHX9. Many of the types 2 and 3 circRNAs showed elevated expressions under the conditions of depleted DHX9, suggesting that the base pairing between Alu elements located toward the 3′-terminal region of SMN pre-mRNA are also under the control of DHX9 [11]. Future studies will determine the role of specific secondary structures that are primary drivers for the generation of SMN circRNAs.

6. Conclusions

Generation of mature RNAs, including mRNAs, lncRNAs and circRNAs requires an accurate removal of introns during pre-mRNA splicing. Information to remove an intron is distributed among cis-elements located within both intronic and exonic sequences. Studies conducted for SMN pre-mRNA splicing regulation are instructive in suggesting that a large part of essential information for intron removal is locked into structural elements. The inhibitory structural context is one of the primary causes of SMN2 exon 7 skipping linked to SMA. The currently approved drug, Spinraza, works at least in part through abrogation of the inhibitory structural context at the 5′ss of exon 7. Many of the small compounds that have recently shown promise for SMA therapy appear to also work through modulation of RNA structure at the 5′ss of exon 7. A vast repertoire of the newly reported SMN circRNAs are generated through structural elements abundantly present within Alu-rich SMN. While functions of SMN circRNAs are not yet clear, generation of these circRNAs come at the expense of linear transcripts required to produce SMN, an essential protein involved in multiple aspects of RNA metabolism. Future studies will determine how transacting factors and structural elements collaborate to maintain a fine balance between linear transcripts and circRNAs generated by SMN.