Identification of Intron and Exon Sequences Involved in Alternative Splicing of Insulin Receptor Pre-mRNA*

The insulin receptor exists as two isoforms, A and B, that result from alternative splicing of exon 11 in the primary transcript. We have shown previously that the alternative splicing is developmentally and hormonally regulated. Consequently, these studies were instigated to identify sequences within the primary RNA transcript that regulate the alternative splicing. Minigenes containing exons 10, 11, and 12 and the intervening introns were constructed and transfected into HepG2 cells, which contain both isoforms of the insulin receptor. The cells were able to splice the minigene transcript to give both A (− exon 11) and B-like (+ exon 11) RNAs. A series of internal deletions within intron 10 were tested for their ability to give A and B RNAs. Intron 10 contained two sequences that modulated exon 11 inclusion; a 48-nucleotide purine-rich sequence at the 5′ end of intron 10 that functions as a splicing enhancer and causes an increase in exon 11 inclusion, and a 43-nucleotide sequence at the 3′ end of intron 10 upstream of the branch point sequence that favors skipping of exon 11. Increasing the length of the polypyrimidine tract at the 3′ end of intron 10 caused exon 11 to be spliced constitutively, indicating that a weak splice site is required for alternative splicing. Finally, point mutations, insertions, and deletions within exon 11 itself were able to regulate inclusion of the exon both positively and negatively.

The human insulin receptor (IR) 1 is encoded by a single gene that is located on chromosome 19 and composed of 22 exons. The mature IR exists as two isoforms, designated A and B, which result from alternative splicing of the primary transcript (1)(2)(3). The A isoform lacks exon 11, is expressed ubiquitously, and is the only isoform in lymphocytes, brain, and spleen; the B isoform contains exon 11 and is expressed predominantly in liver, muscle, adipocytes, and kidney (4 -6). Exon 11 is composed of 36 nucleotides that encode a 12-amino acid segment (residues 717-728) of the carboxyl terminus of the ␣-subunit of IR. A number of investigators have suggested that the isoform ratio could be altered in non-insulin-dependent diabetes mellitus (7)(8)(9)(10), but other studies have produced conflicting results (11)(12)(13)(14). We have found that alterations in isoform ratio in skeletal muscle were associated with hyperinsulinemia rather than diabetes (15). Similar results have been found in the rhesus monkey (16). Along these lines, Sell and co-workers (17) have shown that alternative splicing of the IR gene is regulated by insulin in the Fao hepatoma cell line. Furthermore, we have shown that the alternative splicing is hormonally and developmentally regulated in both the HepG2 hepatoma and 3T3-L1 adipocyte cell lines (18). The changes in splicing were accompanied by increases in insulin sensitivity, as measured by a number of parameters (19). These data indicate that regulation of the alternative splicing of the IR is important for insulin sensitivity and responsiveness.
Splicing of pre-mRNA depends on the presence of relatively short RNA sequence elements, the 3Ј splice site, the 5Ј splice site, the branch point sequence, and the polypyrimidine tract. In alternative splicing, a given splice site may be selected or ignored depending on the cell type or physiological state. This apparent flexibility of the splicing machinery raises the question of molecular mechanisms involved in selection of certain splice sites over others. A number of factors have been implicated in the choice of alternative splice sites, including RNA secondary structure (20 -23), size of the exon (24,25), and relative strengths of the competing splice sites (26). Alterations in the splicing pattern of a number of genes have been demonstrated during cellular differentiation (27); however, the hormonal regulation of alternative splicing is not as common. We have shown that the alternative splicing of exon 11 of the IR gene is modulated by glucocorticoids in HepG2 cells, and, as mentioned above, insulin modulates splicing in Fao cells (17,18). Insulin has also been shown to alter the splicing pattern of the COOH terminus of protein kinase C-␤ in L6 myotubes, and growth factors including epidermal growth factor, platelet-derived growth factor, and basic fibroblast growth factor alter the splicing of the COOH terminus of protein-tyrosine phosphatase 1B in human fibroblasts (28,29). Chew and co-workers (30) have shown that splicing of the insulin-like growth factor-I mRNA is regulated by growth hormone in HepG2 cells. The molecular mechanisms involved in the hormonal regulation of this process are not understood. A prerequisite for mechanistic studies of hormonal regulation is a knowledge of the RNA sequences involved in the alternative splicing event. To that end, we have identified the regions of intron 10 and exon 11 involved in the alternative splicing of the IR gene.

EXPERIMENTAL PROCEDURES
Materials-Cell culture reagents were purchased from Life Technologies, Inc., and fetal calf serum was from Gemini Bioproducts (Calaba-* This work was supported by a merit review award from the Department of Veterans Affairs. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. ‡ Construction of Plasmids and Site-directed Mutagenesis-All recombinant DNA manipulations were carried out according to standard protocols. The minigenes were constructed by amplifying regions of the IR gene from HepG2 genomic DNA by PCR using primers containing unique EcoRI, BamHI, HindIII, and BglII restriction sites to allow replacement of segments of the minigene as cassettes. We were not able to obtain the complete intron 11, as it is Ͼ8 kb. Consequently, we used the known intronic sequence to amplify the two ends of the intron. Thus, all minigenes contain approximately 180 nucleotides at each end of intron 11. In all plasmids, the minigenes were inserted between EcoRI and BglII sites of pSG5 vector (Stratagene, La Jolla, CA), which contains the SV40 early promoter/enhancer, a rabbit ␤-globin intron 2, and an SV40 poly(A) signal after the multiple cloning site. All mutants were generated from the complete minigene B, which was composed of five elements (part of exon 10, entire intron 10, exon 11, a deleted intron 11, and part of exon 12). Fig. 1 shows schematic diagram of the minigene and the sequence of the introns and exons. Deletions and mutations were verified by dideoxy sequencing.
Cell Culture and Transfection-HepG2 cells were maintained routinely in minimum essential medium plus Earle's salts with 10% fetal calf serum at 37°C under 5% CO 2 . The cells were plated at a density of ϳ2 ϫ 10 6 cells/well in six-well plates. Medium was changed every 2 days. Minigene plasmids were transfected into HepG2 cells by the calcium phosphate co-precipitation technique. Cells were harvested 48 h later, and total cellular RNA was prepared using RNAzol B (Tel-Test, Inc., Friendswood, TX) according to the manufacturer's protocol.
Reverse Transcription and Amplification of cDNA-First-strand cDNA was prepared by reverse transcription using 1.0 g of total RNA in a volume of 20 l (250 pmol of random hexamer primers, 1 unit of Inhibit-ACE RNase inhibitor ( To measure exogenous minigene IR transcripts, the primer pair consisted of oligonucleotides spanning nucleotides 1022-1042 (sense primer, 5Ј-TAATACGACTCACTATAGGGC-3Ј) and 1068 -1088 (antisense primer, 5Ј-GCTGCAATAAACAAGTTCTGC-3Ј), which are unique sequences in the pSG5 vector just upstream and downstream of the cloning sites. To measure endogenous IR transcripts, the primer pair consisted of oligonucleotides spanning nucleotides 2140 -2163 (sense primer, 5Ј-AACCAGAGTGAGTATGAGGATTCG-3Ј) and 2424 -2447 (antisense primer, 5Ј-TTCTCAAAAGGCCTGTGCTCCTCC-3Ј) of the IR cDNA (numbering as in Ref. 1), which lie outside the minigene and, thus, are specific for the endogenous gene.
Five l of the cDNA synthesis reaction was used for PCR amplification in a 50-l final reaction volume (0.5 M each oligonucleotide primer, 10 mM Tris-HCl (pH 8.3), 50 mM KCl, 1.5 mM MgCl 2 , 0.1 mM dNTPs, 2 units of Taq DNA polymerase, and 1 Ci of [␣-32 P]dCTP). Twenty-five cycles of amplification were performed using a Perkin-Elmer DNA thermal cycler (System 9600). Each cycle consisted of a 30-s denaturation at 94°C, a 30-s annealing at 55°C, and a 60-s extension at 72°C. The number of cycles was optimized to ensure that the amplification lay within the exponential phase. The products of the PCR amplification were resolved by electrophoresis on 8% polyacrylamide gels. The gels were dried and exposed to film at room temperature. The band densities were quantified using a PhosphorImager (Molecular Dynamics, Sunnyvale, CA).

RESULTS AND DISCUSSION
Intron 10 Contains All the Sequence Information for Alternative Splicing of Exon 11-We have shown previously that the endogenous IR gene in HepG2 cells generates mRNA for both the A and B isoform IR. We used this cell line to investigate which sequences surrounding the alternatively spliced exon 11 are involved in the alternative splicing. A minigene was created from by amplifying each exon and the adjacent intron from the known exon/intron sequences (2). This minigene contained 110 nucleotides of exon 10, 2.3 kb of intron 10, 36 nucleotides of exon 11, 184 nucleotides of the 5Ј end of intron 11, 188 nucleotides of the 3Ј end of intron 11, and finally 103 nucleotides of nucleotides of intron 11, and 103 nucleotides of exon 12. Intron 11 is greater than 8 kb in length; consequently, a large internal deletion was created leaving approximately 180 nucleotides at both the 5Ј and 3Ј ends. Corresponding amino acid residues are indicated below the minigene. Numbers above intron 10 indicate the positions used to create the internal deletions described in this paper. The two As indicate potential BPS upstream of the 3Ј splice site. Minigenes are subcloned into mammalian expression vector pSG5. Panel B, sequence of IR minigene. Sequence of basic minigene is shown divided into exons and introns. The end points of the internal deletions are indicated by numbers below the sequence of intron 10. exon 12 (Fig. 2, minigene B). Transfection of minigene B into HepG2 cells gave RNA corresponding to both the A (Ϫ exon 11) and B (ϩ exon 11) isoform splicing patterns in a 40:60 ratio (Fig. 2, panels B and C). The endogenous gene in these cells expressed both isoform RNAs in a 50:50 ratio using the endogenous gene primer pair (data not shown), suggesting that the minigene contained most of the information for correct splicing. A second minigene (minigene A), containing a large internal deletion of 2.0 kb in intron 10 (between positions 3 and 5 in the minigene), caused an increase in the percent of B splicing isoform (Fig. 2, compare minigenes A and B). A further deletion (to position 7) in minigene C caused a complete loss of B isoform splicing suggesting that sequences in this region (between positions 5 and 7) may be important for skipping of exon 11.
Two potential branch point sequences (BPS) were identified in the intron upstream of exon 11 (indicated by A in Fig. 2A). Removal of the upstream potential BPS by deletion to position 9 (minigene D) caused a large (80%) decrease in the amount of B isoform splicing, suggesting that this BPS may be involved in splice site selection. These deletions suggested that the regulatory sequences might lie in the region upstream of exon 11. Therefore, intron 10 was sequenced from the 3Ј end, and an Alu repetitive sequence was found approximately 300 nucleotides upstream of exon 11. A further minigene was created (minigene E) deleting the central portion of the intron up to, but no further than, the Alu repeat. Transfection of minigene E into HepG2 cells gave a splicing ratio that was identical to minigene A; therefore, the extra sequence up to the Alu repeat did not affect splicing.
The PCR amplification was performed for 40 cycles in an attempt to identify low abundance intermediate splice products for these five minigenes (Fig. 2D). The mature, fully spliced products (mature mRNA) were observed as well as the unspliced precursor RNAs (pre-mRNA) that differed in size due to the internal deletions in the minigenes. No precursor RNA was observed for minigene B as it is predicted to be Ͼ3 kb and is not amplified under standard conditions. A product of ϳ700 base pairs was observed for all minigenes regardless of the size of the internal deletion in intron 10. This PCR product was subcloned, sequenced, and found to be a splicing intermediate lacking intron 10 but still containing intron 11. No intermediates containing intron 10 but lacking intron 11 were isolated. Thus, removal of intron 10 precedes removal of intron 11 in agreement with the processive model of RNA splicing and suggests that the 3Ј splice site of exon 11 is subject to selection. Fig. 2 could be explained by alterations in the spacing of the splice sites. If this were the case, then larger deletions should have an even greater effect. Consequently, minigenes K, L, and M were constructed with larger deletions starting 26 nucleotides downstream of the 5Ј splice site of exon 10 (Fig. 3, position 1) and extending to the end points of minigenes A, C and D (Fig. 3, minigenes K, L, and M, respectively). Surprisingly, transfection of minigene K into HepG2 cells gave Ͻ10% B isoform splicing in contrast to minigene A which gave Ͼ85%.  It is very striking that elimination of intronic sequences in the 5Ј end of intron 10 (between positions 1 and 3) caused a dramatic (75%) decrease in exon 11 inclusion. This suggests that this region contains a sequence that enhances inclusion of exon 11. Deletion of the sequences between positions 5 and 7 (minigene L) caused a large increase in the amount of B isoform splicing consistent with the results from minigenes A and C (Fig. 3). The change in splice site usage is even more dramatic for minigenes K and L than for minigenes A and C, as a result of the absence of the upstream enhancer that favors exon 11 inclusion. These results confirm that sequences between positions 5 and 7 cause exon skipping. Elimination of an additional 26 nucleotides to position 9 (minigene M) caused complete loss of B isoform splicing. A similar loss in B isoform splicing was associated with deletion of this region in minigenes C and D. However, minigene D still showed approximately 20% B isoform splicing, suggesting that this deletion severely weakens the 3Ј splice site but partial recognition is possible in the presence of the upstream enhancer.

Intron 10 Contains a Splicing Enhancer Sequence and an Inhibitory Region That Causes Exon Skipping-The increase in B isoform splicing by the internal deletions in minigenes A, C, and E in
This upstream splicing enhancer was identified by comparing minigenes A and K. Both of these minigenes contain large internal deletions within intron 10, which brings the enhancer closer to the 3Ј splice site. What is the function of this region in the context of the full intron? This region was deleted in the full intron (Fig. 3, minigene W). Deletion of this region caused a 20% decrease in B isoform splicing. Consequently, this region does function as a splicing enhancer but its effect is less pronounced when it is Ͼ2 kb upstream of the splice site (compare minigenes W, A, and K). The enhancer region contains a purine-rich sequence, so two additional deletions were created to assess the contribution of this purine-rich region. Minigene Y lacks the purine-rich sequence, whereas minigene X lacks the remaining sequence in this region. Only minigene Y showed a statistically significant decrease in B isoform splicing. The purine-rich region is 75% GA over a stretch of 60 nucleotides. Repeats of the motif GARGARGAR have been shown to function as splicing enhancers in other genes (31). The sequence in intron 10 does not contain such repeats, but is very GA-rich and may function in a similar manner.
The region between positions 5 and 7 that causes exon skipping was identified, as was the enhancer sequence, using minigenes with large internal deletions in intron 10. What is the role of this inhibitory region in the context of the full intron? A minigene was constructed which lacked only the inhibitory region between positions 5 and 7 (Fig. 3, minigene F). Transfection of minigene F into HepG2 cells gave Ͼ95% B isoform RNA compared with 65% for minigene B with the full intron. Thus, removal of 70 nucleotides from the 2.3-kb intron causes an increase in B isoform splicing. The inhibitory region was further localized to a 43-nucleotide sequence between positions 6 and 7 (minigene N). Deletion of the 27 nucleotides between positions 5 and 6 had no effect on splicing (minigene S). Removal of another 26 nucleotides (minigene G) or 52 nucleotides to the 3Ј splice site of exon 11 (minigene H) dramatically weakens the splice site as would be expected. Similarly, small deletions of 26 or 13 nucleotides (minigenes O and T) caused a dramatic drop in exon 11 inclusion, suggesting that this 13nucleotide sequence is necessary for the efficient use of the 3Ј splice site.
Mutation of the 3Ј Splice Site in Intron 10 -Previous internal deletions in intron 10 had suggested that elimination of the upstream series of adenines between positions 7 and 9 in the minigene impaired the use of the downstream splice site and caused skipping of exon 11 (Fig. 4, minigene O). However, all deletions eliminated other regulatory regions as well. Adenine residues have been identified as the branch point nucleotides in many but not all introns, so elimination of these adenines might explain the alteration in splicing. Mutation of the four adenine residues in this region had no effect on splicing (Fig. 4, minigene I), indicating that this sequence cannot be the functional BPS. However, deletion of the three adenine residues proximal to the splice site (minigene V) gave Ͻ5% exon 11 inclusion. Although this result does not specifically identify the branch point residue, it is likely that one of these adenines is the functional BPS. Alignment of the most distal of the three adenines UCCUCAA with the consensus branch point sequence UNCURAC indicates that this residue could be the branch point, however, accurate identification of the branch point residue will require in vitro branch point mapping. So why does B isoform splicing decrease when the region containing the four upstream adenines is deleted (minigene O)? One possible explanation is that the deletions may have impaired the function of the putative downstream BPS. A deletion from position 8 to 9 between the two stretches of adenines gave a similar reduction in exon inclusion (minigene T). The 3Ј end of this deletion was 2 nucleotides 5Ј to the BPS. All of the previous deletions were constructed by engineering BamHI restriction sites. Therefore, a minigene was constructed with four nucleotide substitutions to create a BamHI site at a position analogous to the deletion mutants (minigene U). This minigene gave Ͻ5% exon inclusion, indicating that mutation of four residues GUCCUCAAAGG to GGAUCCAAAGG could result in exon skipping, presumably by impairing the function of the BPS described above. Conversely, mutation of four nucleotides downstream of the triple adenines had little effect on splicing (minigene AE). The 3Ј splice site in this intron contains a single G residue in the middle of a stretch of nine pyrimidines. Purine residues in the center of the polypyrimidine tract have been shown to have a detrimental effect on RNA splicing (32). Mutation of the guanine residue to thymidine had little effect on exon inclusion (minigene AA). However, increasing the length of the pY tract to 14 pyrimidines caused the exon to be spliced constitutively (minigene AF). Thus, the alternative splicing of exon 11 requires a weak 3Ј splice site and strengthening the site renders the exon constitutive similar to results for other genes (33).
Exon 11 Sequences Are Involved in Splice Site Selection-To investigate whether the alternatively spliced exon itself might be involved in splice site recognition, we introduced mutations into exon 11 in the parental minigene B. Introduction of four point mutations in the middle of the exon caused the exon to be spliced constitutively (Fig. 5, minigene J). An overlapping four point mutations, however, had no effect (minigene AC). A deletion of eight nucleotides or insertion of three thymidine residues caused an almost complete loss of exon inclusion (Fig. 5, minigenes Z and AB). The point mutations that rendered the exon constitutive were tested in combination with deletions of the inhibitory region identified earlier (Fig. 3). There was no additional effect when combined with minigenes F and N that are deleted for the 3Ј inhibitory region (Fig. 5, minigenes P and Q). Interestingly, when these exon mutations were introduced into minigene O, which contained a deletion that impaired function of the putative BPS, there was an increase in the amount of B isoform splicing. Thus, the weakened BPS is able to function, albeit weakly, in the presence of mutations in the exon. These results indicate that the exon 11 sequences play an active role in determining the degree of exon inclusion in both a positive and negative manner.
Models for the Regulation of the Alternative Splicing of Exon 11-The alternative splicing of exon 11 of the IR gene is consistent with the models depicted in Fig. 6. The GA-rich splicing enhancer at the 5Ј end of intron 10 favors inclusion of exon 11. This could be due to a direct effect on the 3Ј splice site. Alternatively, as the enhancer is Ͼ2 kb upstream of the BPS, the effect of the enhancer could be on the adjacent 5Ј splice site. The proximal site (TAG:GUCAGGAC) differs significantly from the consensus (CAG:GUAAGUAU) so the effect of the enhancer may be to strengthen the interaction of the U1 small nuclear ribonucleoprotein particle with the 5Ј splice site (34). How this might affect alternative splicing of the downstream exon is not clear, as this site is used whether or not the exon is included, but it may allow the use of a suboptimal splice acceptor site. SR proteins that recognize purine-rich enhancer sequences are known to favor use of proximal splice sites (35)(36)(37)(38)(39)(40)(41)(42). SF2/ASF is a member of the SR protein family and can bind to GA-rich splicing enhancer sequences similar to that identified in the 5Ј end of intron 10. Overexpression of SF2/ASF has been shown to promote inclusion of alternatively spliced exons in the rat clathrin light chain B and rat ␤-tropomyosin genes (41,42). This is due to the ability of SF2/ASF to promote the use of a proximal splice site, either 5Ј or 3Ј, over a distal site. Interestingly, this activity is antagonized by the hnRNP-A1 splicing FIG. 6. Potential models for splice site selection. Figure depicts different models for splice site recognition. SR proteins binding the purine-rich enhancer sequence at the 5Ј end of the intron could exert effects at either the 3Ј splice site or the proximal 5Ј splice site. Factors recognizing the inhibitory sequences at the 3Ј end of the intron or exon 11 could repress or stimulate recognition of the 3Ј splice site. Alternatively, RNA secondary structure involving both the inhibitory sequence and the exon could sequester the 3Ј splice site to regulate exon skipping.
factor, which favors the use of distal splice sites over proximal. The observed choice of splice sites reflects a balance between SF2/ASF and hnRNP-A1 activities. Both SF2/ASF and hnRNP-A1 are RNA-binding proteins. In contrast to the SR proteins, hnRNP-A1 binds to sequences containing the motif UAGGGA or UAGGGU (43). The 5Ј end of intron 10 also contains the sequence CTTAGGGACC, which includes an hnRNP-A1 binding site (underlined). Whether hnRNP-A1 could regulate the alternative splicing is unknown; however, deletion of the 5Ј end of intron 10 in minigene X eliminates the potential hnRNP-A1 binding site but has no effect on the splicing. Further studies will be required to determine if SR proteins or hnRNP-A1 can recognize the regulatory regions that we have identified in the IR gene and modulate splice site selection.
At the 3Ј end of the intron, there are two regions that influence exon skipping. One is inhibitory and located in the intron upstream of the BPS. Interestingly, secondary structure prediction programs based on the Zuker algorithm indicate that this region has the potential to form a stable stem-loop structure. Whether such a structure is formed in vivo is not known. The role of secondary structure in alternative splicing has been reviewed recently (23). The polypyrimidine tract at the 3Ј splice site is only 9 nucleotides with a central guanine residue. Mutation of the guanine to give a stretch of 10 contiguous pyrimidines did not have an effect on exon recognition; however, increasing the length of the polypyrimidine tract to 14 nucleotides rendered the exon constitutive. Thus, exclusion of the exon requires a suboptimal splice acceptor site. Similar effects have been seen with the growth hormone and other genes (33). The second regulatory region is contained within the alternatively spliced exon itself and appears to have both positive and negative effects. On the one hand, introduction of four point mutations caused the exon to be constitutive, suggesting a negative role for the exon. On the other hand, insertion of three thymidines or deletion of a stretch of eight pyrimidines caused the exon to be skipped indicating a positive role. Interestingly, the constitutive point mutations had little effect when combined with deletions of the inhibitory region upstream. Thus these regulatory regions could function independently to modulate recognition of the 3Ј splice site (Fig. 6) or, alternatively, both sequences could be involved in a larger regulatory region involving secondary structure around the 3Ј splice site. Further studies are planned to test these models.