Unfolding the mystery of alternative splicing through a unique method of in vivo selection

Abstract

Alternative splicing of pre-messenger RNA (pre-mRNA) is a fundamental mechanism of gene regulation in higher eukaryotes. In addition to creating protein diversity, alternative splicing provides the safest mode of gene evolution. Of late, more and more forms of alternatively spliced transcripts (mRNAs) are being discovered for key genes. Some of the alternatively spliced transcripts are also associated with major human diseases. This has created a sense of urgency to find the methods by which regulation of alternative splicing of specific exons could be best understood. Here I review a powerful in vivo selection method that uses a combinatorial library of partially random sequences. Several advantages of this method include in vivo analysis of large sequences, identification of unique sequence motifs, determination of relative strength of splice sites and identification of long-distance interactions including role of RNA structures. This unique method could be applied to identify tissue-specific cis elements. Similarly, the method is suitable to find cis elements that become active in response to specific treatments of cells. Considering this unbiased method uses in vivo conditions, it has potential to identify critical regulatory elements as therapeutic targets for a growing number of splicing-associated diseases.

2. INTRODUCTION

Nature has embraced a well-thought-out plan in which organisms have continued to evolve by mingling relatively less-conserved non-coding sequences (introns) with the increasingly conserved coding sequences (exons). Thanks to this arrangement, alternative splicing creates more than 200,000 proteins from about 30,000 genes in humans. A similar number of genes in mouse produce equally high number of proteins. However, several differences between human and mouse genome is responsible for producing species-specific alternative splicing. Spliceosome, a macromolecular complex, catalyzes the dynamic process of pre-mRNA splicing (1). However, non-spliceosomal factors are key to the process of alternative splicing (2–7). Some of these non-spliceosomal factors are expressed only in certain tissue types providing the tissue-specific splicing (8–10). Unique arrangement of cis-elements (or regulatory sequences) on pre-messenger RNA (pre-mRNA) assists the splicing proteins in defining the exon-intron boundaries or the 5′ and 3′ splice site (5′ ss and 3′ ss) located towards the 5′ and 3′ ends of intron, respectively. Traditionally, cis-elements have been referred to as exonic or intronic splicing enhancers (ESEs or ISEs) and silencers (ESSs or ISSs). Enhancers and silencers promote or suppress splice-site selection, respectively (4). Most of our knowledge of cis-elements is limited to small sequence motifs that could be identified using various computational programs (11–13). Currently, there is no method to identify large regulatory motifs including RNA structures formed by local and long-distance intra-molecular RNA-RNA interactions. An increasing number of reports have unequivocally confirmed the very significant role of RNA structures in alternative splicing (14–16). Further adding to complexity, alternative splicing has been linked to transcription (17–19) and polyadenylation (20–22). But the most intriguing part of alternative splicing is the production of several unknown intermediates that are either degraded or not translated due to lack of initiation codon or presence of a termination codon (23–30). Considering there is also a proofreading at the second step of splicing (31–33), some of the transcripts may be eliminated after the very first step of splicing (34). Thus, it is hard to predict the fraction of pre-mRNAs that eventually make up to a translatable mRNA. Most commonly reported alternatively-spliced transcripts include exon skipping, intron retention, insertion of Alu-like sequences, and alternate usage of the 3′ and/or 5′ ss (35–37; Figure 1). In some cases, majority of the above-mentioned variants are present in mRNAs derived from a single gene (38, 39). Alternative splicing can be induced by inherited or acquired mutations. In fact, it is hard to find a single major human disease (genetic or acquired) that is not associated with alternative splicing. These include anxiety (40), Alzheimer’s disease (41), cancer (42), cardiovascular disease (43), diabetes (44), neuromuscular disease (45, 46), obesity (47–49), and Parkinson’s disease (41, 50). As the number of patients affected by alternative splicing rises, there will be pressing need to find novel therapies for alternative splicing-associated human diseases.

Figure 1.

Types of alternative splicing of a hypothetical gene containing four exons. Exons are shown as colored boxes, whereas introns are shown as solid lines. Dotted lines indicate the joining of exonic sequences by removal of intronic sequences. A. In the absence of alternative splicing all exons are joined by removal of intronic sequences. B. Exon 3 is skipped due to absence of recognition of both, the 3´ and 5´ ss of exon 3. C. Only a portion of exon 4 is included due to activation of an alternative 3´ ss within exon 4. D. Only a portion of exon 2 is included due to activation of an alternative 5´ ss within exon 2. E. An Alu-like sequence is included due to activation of the additional 5´ and 3´ ss within intron 1. F. Intron 2 is retained due to absence of recognition of both, the 5´ ss of exon 2 and the 3´ ss of exon 3. Abbreviations: Int stands for intron, whereas ss stands for splice site.

As for now, we are confronted with the challenge of understanding the very mechanism by which exon-intron boundaries are accurately defined. One thing is clear; every single exon is regulated by a distinct rule. The only common principle is the ability of the cellular machinery to adopt a rule of combinatorial control in which decision to include or exclude an exon is made by multiple cis-elements and their cognate transacting factors. In this process, some cis-elements may play a dominant role because of their location vis-à-vis location of other cis-elements. In a recent review, Baralle and coworkers have wisely termed the dominant cis-elements as “master checkpoints” (51). Such checkpoints may serve as the effective targets for correcting aberrant splicing in human diseases. Here I describe the lessons learned from the very first in vivo selection experiment performed on the entire exon 7 of human Survival Motor Neuron 1 (SMN1) gene (52, 53). This method offers one of best possible alternatives to identify novel cis-elements that are currently beyond the scope of being predicted by available computational programs or otherwise.

3. IN VIVO SELECTION OF THE ENTIRE SMN1 EXON 7

Humans have two copies of SMN: SMN1 and SMN2. Because of a critical C to U mutation at position 6 (designated as C6U) in exon 7 of SMN2, majority of mRNAs derived from SMN2 lack exon 7 (reviewed in 54). Hence, inability of SMN2 to compensate for the loss of SMN1 results into spinal muscular atrophy (SMA), a leading genetic cause of infant mortality. SMA has become a model disease in which a defective copy of the gene (SMN2) is almost universally present and holds the potential to be corrected by merely altering the aberrant splicing of one exon (i.e. exon 7). There have been several studies that have implicated a number of cis-elements and their cognate proteins in regulation of alternative splicing of SMN exon 7 (reviewed in 55). Here I describe the results of a unique experimental design of in vivo selection that revealed novel cis-elements without their cognate protein factors being identified (52, 53).

3.1. Experimental design

In vivo selection in this review refers to iterative selection or cycled selection of exonic sequences that promote their own inclusion during the process of pre-mRNA splicing in an intact cell. In contrast to in vitro selection, in vivo selection recapitulates a true process of molecular evolution in which selected sequences undergo through several quality controls in addition to their ability to be included in the processed transcript. For example, sequences that undergo through nonsense-mediated decay are not selected. Similarly, sequences that interact with the insoluble aggregates of the cell are eliminated.

In vivo selection method used for SMN1 exon 7 analysis is an advanced version of iterative in vitro selection method “SELEX” (Systematic Evolution of Ligands through EXponential Enrichment) that was originally developed to isolate tight binding nucleic acid ligands called “aptamers” (56–58). Most in vitro SELEX experiments have been performed using ~1015 molecules generated by complete randomization of small sequences. Because complete randomization creates artificial binding motifs, aptamer-protein interactions show no resemblance with the RNA-protein interactions prevalent in nature. Hence, not surprisingly aptamers have been generated against proteins lacking nucleic acid binding domains (56). To suppress the creation of artificial binding motifs, many modified SELEX experiments have been performed using partially random sequences. These experiments have revealed valuable and rather critical features of RNA-protein interactions prevalent in nature (59–63). In contrast to complete randomization, partial randomization allows examination of larger sequences, while maintaining the wild-type characteristics of the molecule. Except for the limitations of oligonucleotide synthesis, there is no limit on the size of molecules that could be analyzed using partial randomization.

Since SMN exon 7 is only 54-nucleotide long, it presented an ideal size to test the very first in vivo selection of an entire exon. Partial randomization was used to create the initial pool of ~1011 unique molecules (52). Limitations of in vivo experiment do not allow analysis of a larger pool. However, a pool size of ~10¹⁰ unique molecules is enough to test the position-specific role of every nucleotide within the entire exon 7. We used 30% randomization by doping 10% each of the non-wild-type residues at every exonic position. Traditionally, this degree of randomization has been successfully used to obtain valuable information (59–63). The commercially synthesized partially random exon 7 sequences were inserted in SMN1 minigene and the percentage of randomization was confirmed by sequencing of 51 individual clones from the initial pool. It should be mentioned that the sequencing of the initial pool is necessary to normalize the skewed randomization during commercial synthesis (52). Of note that the commercial synthesis of the partially random oligonucleotides may be the only expensive step of the selection protocol, however cost of oligonucleotide synthesis is continuously going down.

For iterative selection, cells were first transfected with exon-7-randomized SMN1 minigene and specific primers were used to amplify exon-7-included transcripts. The amplified exon-7-included products were purified and reamplified with specific primers that allowed restoration of the minigene (Figure 2). The success of in vivo selection experiment was apparent from the very first cycle. For example, there was no smear during the amplification of the spliced products (52). A complete randomization would have created cryptic-splice sites resulting into non-amplifiable bands. After only four cycles substantial enrichment was achieved. To further confirm the enrichment, a total of 59 clones from the 4th cycle were analyzed. As expected, none of the selected clones showed exon-7-excluded products in the highly sensitive radioactive RT-PCR (52).

Figure 2.

Strategy for the iterative in vivo selection of the entire SMN1 exon 7. Exons are shown as colored boxes, whereas introns are shown as solid lines (diagram not to the scale). SMN1 minigene containing exon 6, short intron 6, partially-random exon 7 (Exon 7R), intron 7 and exon 8. Ns represent all positions of exon 7 that were partially randomized. Details of minigene construction are described elsewhere (52). Cells were first transfected with minigene containing Exon 7R. After in vivo splicing, exon-7 included products were amplified by RT-PCR, separated on gel and purified. A second amplification step was used to amplify exon 7 sequences with flanks containing special restriction sites. Upon digestion at the special restriction site, intact E7R was released and inserted in the original vector to restore the minigene (52). Repeated steps of selection procedure amplified exon 7 sequences that promote their own inclusion (52).

3.2. Analysis of results and validation

Having obtained the required pool of sequences that promote exon 7 inclusion, a very strict criteria was applied to determine the position-specific role of residues. For this purpose, all sequences were aligned end-to-end. This approach is distinct from those in which sequences are aligned to obtain small motifs (64, 65). Upon end-to-end alignment of sequences, exonic positions that were found to retain the wild-type residues were considered conserved. Similarly, exonic positions that were found to discard the wild-type residues were considered mutable. A conserved position is supposed to have positive impact on exon 7 inclusion, whereas a mutable position is likely to have the negative impact on exon 7 inclusion. The concept of mutability was originally applied in 1995 to find the important residues within a protein (66). Since then, this approach has been successfully tested in other systems (62, 67–69).

3.2.1. Mutability profile reveals three novel cis-elements

Based on the mutability of residues, exon 7 was found to contain two negative and one positive element (Figure 3). Negative elements were termed as Exinct (stands for “Extended inhibitory context”) and 3´-Cluster located towards the 5´ and 3´ ends of exon 7, respectively. The lone positive element was termed as Conserved tract located in the middle of exon 7. Nature of above elements was validated by site-specific mutations (52, 70). Both Exinct and Conserved tract are large sequence motifs. Such motifs are currently not predicted by computational programs. It is possible that a large sequence motif provides a higher degree of specificity due to interaction with a larger protein complex. In fact most splicing proteins (particularly SR and SR-like proteins) are capable of forming large complexes by interacting among each other before interaction with their cognate cis-elements on a given pre-mRNA. It is well known that a multi-component protein complex may have different specificity to a RNA sequence than the individual components of the complex. Consistently, Gold and colleagues have demonstrated that even a slight change in stoichiometry of two proteins in a complex is sufficient to produce different sequence motifs in an in vitro SELEX experiment (71). Extending this principle to pre-mRNA splicing, Singh and Vacarcel have recently reasoned that alternative splicing is regulated by specific arrangement of proteins that are generally nonspecific on their own (6). Hence, it is rather reassuring that results of in vivo selection revealed larger sequence motifs that may be unique to a particular exon.

Figure 3.

Mutability of residues based on the results of in vivo selection. The negative bars (in green) represent conserved positions, whereas the positive bars (in red) represent the mutable position. Mutable and conserved positions play negative and positive roles in exon inclusion, respectively. The values of −1 and +17.6 represent the absolutely conserved and the least conserved residues, respectively. The dotted horizontal lines show the cutoff points with the mutability values of +0.2 and −0.2, corresponding to the mutable and the conserved residues, respectively. Based on the stretches of the mutable and the conserved residues, two negative elements (Exinct and 3′-Cluster) and one positive element (Conserved tract) have been shown. Nature of these elements was confirmed by site-specific mutagenesis (52, 70). The exceptionally high mutability of position 54 is consistent with the dominant effect of A54G substitution on exon 7 inclusion (52). Interestingly, 3′-Cluster overlaps with TSL2, a terminal stem-loop structure that was recently found to play an inhibitory role (16; Figure 4).

Another valuable outcome of in vivo selection was the identification of the variable impact of residues within a given cis-element. For example, not all residues within (Exinct) were inhibitory, whereas, not all residues within the conserved tract were stimulatory. Such residues are referred to as incompatible residues. Due to several compelling reasons, presence of an incompatible residue within a cis-element could have been preferred. For example, incompatible residues may have avoided creation of tight binding sites for protein complexes. It is not in the interest of a dynamic and efficient process of cellular regulation to have a tight-bound complex in which protein components are trapped. On the other hand a loose complex would allow protein components to be recycled multiple times. Further, role of an incompatible residue could be associated with providing secondary contacts for proteins that interact with other motifs. It is also possible that incompatible residues are important for providing the stability to the transcripts, as well as may help in transport of transcripts across the nuclear membrane.

3.2.2. Role of terminal exonic positions on exon skipping

Terminal gaunosine residues play an important role in exon definition and are predominantly represented in internal exons (72). Yet, plethora of exons without terminal gaunosines is efficiently included in the processed transcripts. Hence it is difficult to predict the exact contribution of terminal residues in defining an exon. In vivo selection of entire SMN1 exon 7 provided the first opportunity in which relative significance of both terminal positions were tested. Underscoring the significance of terminal guanosines, the wild-type guanosine at the first position (1G) was conserved, and a non-wild-type guanosine at the last position (54G) was overwhelming selected. Additional experiments confirmed that 1G is absolutely required to include exon 7 in SMN1. However, effect of 54G was found to be more dominating than 1G. For instance, inhibitory effect of G1H mutation (H stands for A or C or U) was easily compensated by 54G. Further experiments showed that 54G also compensates for the lossof an ESE associated with Tra2-β1. In fact 54G was able to compensate for the simultaneous loss of two cis-elements, Tra2-β1 ESE and 1G (52). These results are in full agreement with the results of in vivo selection that showed that the last position of exon 7 is the-most-inhibitory position (or highly mutable position) within the entire exon 7. It must be mentioned that finding of 54G as the most powerful position within exon 7 is a major discovery thanks to in vivo selection of entire exon 7. This very fact has been confirmed in a recent report, which compared the results of in vivo selection with available computational programs (51). Among various possible effects, it is likely that 54G breaks an inhibitory context created by a RNA structure at the 5′ ss of exon 7 (16). However, the exact mechanism by which 54G has such a dominant impact is beyond the scope of this review.

3.2.3. Relative impacts of weak splice sites on exon skipping

As mentioned above, C6U is the critical difference between exon 7 of SMN1 and SMN2 (also see references in 54). Prior to in vivo selection of the entire exon 7, all published reports focused exclusively on the weak 3´ ss as the major cause of exon 7 skipping in human SMN2 (cited in 16, 54). Only after realizing the fact that 54G compensates for the loss of a number of positive elements, it became obvious that the 5´ ss of exon 7 is extremely weak. Surprisingly, existing computational programs show their strong limitation as they fail to detect any significant difference between 54G and 54A on the strength of splice sites (51). Thus in vivo selection offers a strong alternative to the existing computational programs for predicting the relative strength of splice sites of a skipping exon. In addition to the presence 54A, other cis-elements also contribute towards the weak 5´ ss of exon 7. The first reported example is the 3´-Cluster, a major cis-element revealed by in vivo selection of entire exon 7. Most recently, an intronic element, ISS-N1, was also found to contribute towards the weak 5´ ss of exon 7 (73). Noticeably, 3´-Cluster and ISS-N1 are not conserved between human and mouse, suggesting an evolutionary significance. It will not be surprising if factors interacting with 3´-Cluster also communicate with ISS-N1 or vice versa. Of note that discovery of ISS-N1 is a major step forward for our understanding of SMN exon 7 splicing. Antisense oligonucleotides that block ISS-N1 were found to restore exon 7 inclusion in SMN2 mRNA (73). Hence, ISS-N1 provides the very first gene-specific target for correction of SMN2 exon 7 splicing. ISS-N1 also serves as one of the classic examples of correcting exonic defects through use of unique intronic sequences. It should be mentioned that in vivo selection of entire exon 7 was one of the crucial steps that led to an early discovery of ISS-N1.

3.2.4. Role of RNA structure on exon skipping

The very first application of the partially random large sequence in an iterative selection experiment gained significance due to identification of a unique structure involving non-canonical interaction (59). Further use of the partially random large sequences in the iterative selection experiments confirmed the role of other critical structural motifs (60, 61, 63). RNA structure remains one of the most challenging aspects of alternative splicing. Single mutations that create/abrogate a cis-element could also bring a drastic change in RNA structures. But, delineating the role of RNA structure using site-specific mutations is a daunting task. Remarkably, mutability values obtained after in vivo selection provided a strong clue about the inhibitory roles played by local RNA structures. For example, two terminal stem-loop structures, TSL1 and TSL2, formed by exon 7 were found to harbor the majority of highly mutable residues (Figure 4). Most of these mutations are capable of breaking TSL1 and TSL2. Indeed, a recent structural analysis by enzymatic probing confirmed the existence ofTSL1 and TSL2 (16). Further, a systematic analysis of residues within TSL2 confirmed the inhibitory nature of TSL2 (16). Most convincing among them were compensatory mutations that restored exon 7 exclusion by reinstating the inhibitory structure, TSL2. Having proved its use in revealing critical role of RNA structure, in vivo selection approach could be applied to large exonic sequences. With proper modifications, method could also be applied to structures that involve base pairing between exonic and intronic sequences.

Figure 4.

RNA secondary structure of human SMN1 exon 7. The same structure holds true for human SMN2 exon 7 (except for C6U transition, a C to U mutation at position 6 in exon 7 of SMN2; also see reference 54). Numbering starts from exon 7 (the entire exon 7 is boxed). Intron 7 starts from position 55. Negative numbers represent intron 6. The exonic sequences are shown in large-case letters, whereas intronic sequences are shown in small-case letters. The local secondary structure of exon 7 RNA consists of two terminal-stem loops (TSL1 and TSL2) and one internal-loop (IL2). TSL1 is exclusively formed by exonic sequences. TSL2 is mostly formed by exon 7 sequences, although the last two base pairs of the TSL2 are formed by sequences of exon 7 and intron 7. Shaded residues represent mutable positions determined by in vivo selection of entire exon 7. A recent report validated the above structure by enzymatic probing (16). This report also confirmed the inhibitory nature of TSL2. Positions of the 3´ ss, 5´ ss and ISS-N1 (a negative element) are shown. Details of ISS-N1 are described elsewhere (73).

4. CONCLUSIONS AND PERSPECTIVES

Alternative splicing is an essential process of gene regulation in higher eukaryotes. This process is also associated with a growing number of human and animal diseases. Methods are being developed to identify critical cis-elements that regulate alternative splicing of a given exon. A proper understanding of splicing cis-elements has multiple ramifications. Most important among them is the use of cis-elements for correction of the aberrantly spliced genes. In vivo selection of the entire exon 7 of SMN1 is the first successful in vivo application of a very promising and unbiased methodology. Not only this method was capable of identifying novel regulatory elements, but was able to reveal critical structural motifs, identify prominent role of certain key positions within the entire exon and disclose the relative strength of splice sites. There is no report of any in vivo selection study in which such a wealth of information was generated in a single experiment.

Success of in vivo selection method could be largely attributed to the partial randomization. This approach maintained the wild-type character of the molecule, as well as tested the relative significance of every exonic position. Hence, identification of cis-elements was made in the context of the entire exon. Method provides flexibility in choosing the degree of randomization for examination of specific region of an exon. For example, a higher degree of randomization could be used to check the more conserved regions, whereas a lesser degree of randomization could be used to check the highly mutable regions. This way, significance of every single position within an exon could be tested. In any case, information generated from in vivo selection of large sequences would be of great help for future generations of computational programs.

Now that the concept of in vivo selection of an entire exon has been validated, experiments could be designed to analyze complex exons. Such exons are generated by incorporation of different portions of the adjoining intronic sequences. Actually, number of complex exons that contain multiple splice sites within a single exon is continuing to grow. Alternative splicing of mu opioid receptors is one of the classic examples in which addition of adjoining intronic sequences could produce up to five different variants of exon 5 (38). Incorporation of different portions of the last intron creates alpha and beta isoforms of glucocorticoid receptors (GRs) (74). Since more than 10% genes in humans are affected by GRs (74), it becomes rather essential to understand the critical role of cis-elements that determine the very ratio of alpha and beta isoforms of GRs. A recent study confirmed that different antidepressants have variable effects on the expression of alpha and beta isoforms of GRs (75). It is likely that certain cis-elements become more responsive to treatment of a particular drug. In vivo selection of large sequences has a great potential to address these issues. Another example of alternative splicing involving incorporation of different portions of the last intron is myelin oligodendrocyte glycoprotein (MOG) (39). Interestingly, one of the MOG isoforms is specific to primates (39). Conditions under which these isoforms are generated are not known. Here again, in vivo selection of large exonic sequences will be able to reveal critical cis-elements that promote inclusion of the last exon.

Several genes contain more than one alternatively spliced exon. Examples of these include but not limited to, B cell receptors, cystic fibrosis transmembrane receptor (CFTR), dopamine receptors, iodothyronine deiodinase (DIO1), LDL receptors, metabotropic glutamate receptors, MOG, mu opioid receptor, neurofibromatosis protein (NF1), phosphodiesterase (PDE), SECIS binding protein 2 (SBP2), Tau, Toll-like receptors, Tra2 and vascular endothelial growth factor (VEGF). There is a strong likelihood that alternative splicing of one exon affects the alternative splicing of another exon. However, there is no systematic study to this effect. In vivo selection method coupled with partial randomization of exonic sequences would undoubtedly provide an unparallel approach to analyze alternative splicing of multiple exons in a single experiment. In fact such experiments would revolutionize the process of target discovery for correction of a number of splicing-associated diseases.

Pre-mRNA splicing is coupled with other cellular processes in vivo. These include transcription, 5´-capping, polyadenylation, exon-junction complex (EJC) formation, mRNA transport, nonsense-medicated decay (NMD), nonsense-altered splicing (NAS) and translation (see references in 52). Additionally, it is likely that RNA structure affects most of the above processes. Unfortunately, there is no method that could address the interrelationship between pre-mRNA splicing and these processes. In vivo selection method could be tailored to ask specific questions. Since method using partial randomization allows analysis of large sequences without creating artificial motifs, it is highly feasible to test the role of multiple overlapping wild type motifs (e.g. associated with alternative splicing, NMD, NAS, EJC formation and transport) in the context of a particular gene.

In summary, there is a paucity of methods that reveal complex biological motifs of fundamental significance. In vivo selection method described here offers several advantages over existing methodologies and has a huge potential to advance our understanding of posttranscriptional regulation. The method will help find key regulatory elements that have profound impact on alternative splicing of genes associated with a growing number of human diseases. Once the key regulatory elements are known, different strategies could be adopted to target these elements for producing the needed protein isoforms. For example, the negative cis-elements could be blocked by antisense oligonucleotides to promote exon inclusion in a given mRNA. Similarly, to promote exon exclusion, positive cis-elements could be targeted. Structure-specific aptamers could be used to target critical structural motifs. Aptamers are emerging as the powerful tool of target validation and therapy (57, 58). To an equal significance, in vivo selection method described here will address several fascinating questions of biological evolution. After all, in vivo selection is an unbiased process of biological evolution with the advantage of receiving very specific answers in a time and manner of our own choice.