Mapping the SF2/ASF Binding Sites in the Bovine Growth Hormone Exonic Splicing Enhancer

Splicing of the last intron (intron D) of the bovine growth hormone pre-mRNA requires the presence of a downstream exonic splicing enhancer (ESE). This enhancer is contained within a 115-nucleotide FspI-PvuII (FP) fragment located in the middle of the last exon (exon 5). Previous work showed that the splicing factor SF2/ASF binds to this FP region and stimulates splicing of intron D in vitro. However, the precise sequences recognized by SF2/ASF within the FP region had not been determined. Here we used multiple strategies to map the SF2/ASF binding sites and determine their importance for ESE function. Taking advantage of the fact that SF2/ASF ultraviolet (UV) cross-links specifically to RNA containing the FP sequence, we first mapped a major SF2/ASF binding site by UV cross-linking and reverse transcription. This strategy identified a 29-nucleotide SF2/ASF binding region in the middle of the FP sequence containing the 7-nucleotide purine-rich motif described previously. Interestingly, this binding region is neither sufficient, nor absolutely required for SF2/ASF-mediated splicing, suggesting that additional SF2/ASF binding sites are present. The location of these additional sites was determined by electrophoretic mobility shift analysis of various subfragments of the FP sequence. Antisense 2'-O-methyl oligoribonucleotides complementary to selected SF2/ASF binding sites block bovine growth hormone intron D splicing. Thus, multiple SF2/ASF binding sites within the exonic splicing enhancer contribute to maximal enhancer activity.

However, little is known about the SR protein binding sites within t h e s e enhancer sequences, although it appears that SR p r o t e i n -d e p e n d e n t splicing specificity is mainly due to the RRM domains (40,41).
Splicing of the last intron (intron D) of the bovine growth h o r m o n e (bGH) pre-mRNA requires the presence of a downstream splicing e n h a n c e r in the last exon (17,42). This ESE is contained within a 115-nucleotide Fsp I-P v u II (FP) fragment in the middle of exon 5 (42). The SR protein SF2/ASF binds specifically to the FP region and, presumably as a result of this binding, it enhances splicing of bGH intron D in vitro (19). Although there is evidence that a purine-rich sequence in the middle of the FP sequence is part of the bGH ESE sequence (17), it was not clear w h e t h e r SF2/ASF binds to this purine-rich sequence and/or to adjacent regions. I n this study, we take advantage of the fact that SF2/ASF binds to the FP sequence in the absence of other proteins, and utilize this simplified recognition event to study one of the earliest steps in splicing, ESE recognition. Mapping of SF2/ASF cross-links to the FP sequence and EMSA showed that there are multiple SF2/ASF binding sites within the FP sequence, including a site that is centered around the previously described purine motif (17). The functional significance of these SF2/ASF binding
As expected, the ∆FP substrate did not splice after supplementation w i t h recombinant SF2/ASF (lane 12). Interestingly, the UV-XL substrate spliced upon supplementation of the dilute nuclear extract with recombinant SF2/ASF (lane 8), but at lower levels than the wild-type and ∆UV-XL substrates. This difference was consistent over a range of SF2/ASF concentrations (data not shown).
The question remains as to why UV-XL spliced so poorly i n comparison to wild-type and ∆UV-XL substrates, given that cross-links t o SF2/ASF were only observed within the 29-nucleotide (UV-XL) sequence.
Perhaps the 29-nucleotide sequence is indeed a strong SF2/ASF binding   Figure 1A.
The third group consists of the FP subfragments (nos. 2, 4, 5 and 9) t h a t show the highest levels of binding to SF2/ASF. This group includes subfragment no. 4 which contains most of the cross-links mapped in Figure   1A, pre-incubating the oligoribonucleotides with the splicing substrate for 1 5 -30 minutes at 0°C prior to addition of the nuclear extract.
As can be seen in Figure 6, oligoribonucleotides B7 and B8 do n o t significantly inhibit bGH pre-mRNA splicing (lanes B7 and B8) which is consistent with the fact that E5/∆FP RNA does not contain strong SF2/ASF binding sites (19). Oligoribonucleotide B6 also does not significantly affect splicing, consistent with the EMSA data in Figure 5 showing little or n o Additionally, there is a preference for pyrimidines over purines in crosslinking (58). Thus, it is possible that SF2/ASF does in fact bind to t h e purine motif identified earlier (17,30), but that this interaction does n o t lead to formation of a cross-link.
Deletion of a 29-nucleotide sequence (UV-XL) containing this putative SF2/ASF binding site(s) led to only a 50% reduction in t h e SF2/ASF-mediated stimulation of intron D splicing as compared to wildtype bGH pre-mRNA ( Figure 3). In addition, this sequence by itself functioned only as a moderate splicing enhancer, even though it clearly bound SF2/ASF (Figure 4). Although it is possible that these effects w e r e caused by the deletion mutants affecting the secondary structure of bGH pre-mRNA, we do not believe this to be the case. Extensive analyses of bGH mRNA secondary structure, using Zuker's computer algorithm (60-62) i n combination with previous experimental data obtained with s t r u c t u r esensitive probes (for example, RNAses T1 and V1) has not suggested a n y obvious alterations. Instead, we favor the hypothesis that additional SF2/ASF binding sites are required for maximal enhancer activity. T h e presence of additional SF2/ASF binding sites in the FP sequence w a s confirmed by the fact that ∆UV-XL RNA was significantly shifted b y SF2/ASF in the EMSA (Figure 4). In fact, ∆UV-XL RNA shifted even b e t t e r than wild-type FP RNA. This observation raises the possibility that deletion of the UV-XL sequence activated a new/cryptic SF2/ASF binding site which, together with the normal SF2/ASF binding sites in the FP sequence, compensated for the loss of the UV-XL binding site. Apparently this compensation results in even better SF2/ASF binding. However, if this assumption is true, this new/cryptic SF2/ASF binding site does not a p p e a r to be functional, since the ∆UV-XL pre-mRNA never splices better than t h e wild-type bGH pre-mRNA ( Figure 3). Thus, the precise location of SF2/ASF binding sites within the enhancer and the interrelationship between t h e various SF2/ASF binding sites appear to be important factors i n determining enhancer function.
In an attempt to locate the additional SF2/ASF binding sites, the FP sequence was divided into similar sized subfragments of 1 4 -1 7 nucleotides. These subfragments were inserted into bGH exon 5 in place of the entire FP sequence. EMSA analysis of these FP subfragments s h o w e d that they could be divided into three groups: those that showed little or n o binding to SF2/ASF (subfragment nos. 6 and 8), those that s h o w e d moderate binding to SF2/ASF (subfragment nos. 1, 3 and 7), and those t h a t showed strong binding to SF2/ASF (subfragment nos. 2, 4, 5 and 9). T h e SF2/ASF cross-links that were mapped in Figure 1A