Two regions promote U11 small nuclear ribonucleoprotein particle binding to a retroviral splicing inhibitor element (negative regulator of splicing).

The Rous sarcoma virus (RSV) negative regulator of splicing (NRS) is an RNA element that represses splicing and promotes polyadenylation of viral RNA. The NRS acts as a pseudo 5' splice site (ss), and serine-arginine (SR) proteins, U1snRNP, and U6 small nuclear ribonucleoproteins (snRNPs) are implicated in its function. The NRS also efficiently binds U11 snRNP of the U12-dependent splicing pathway, which is interesting, because U11 binds only poorly to authentic substrates that lack a U12-type 3' splice site. It is of considerable interest to understand how the low abundance U11 snRNP binds the NRS so well. Here we show that U11 can bind the NRS as a mono-snRNP in vitro and that a G-rich element located downstream of the U11 site is required for efficient binding. Mutational analyses indicated that two of four G tracts in this region were important for optimal U11 binding and that the G-rich region did not function indirectly by promoting U1 snRNP binding to an overlapping site. Surprisingly, inactivation of U2 snRNP also decreased U11 binding to the NRS. The NRS harbors a branch point-like/pyrimidine tract sequence (BP/Py) just upstream of the U1/U11 site that is characteristic of 3' splice sites. Deletion of this region decreased U2 and U11 binding, and deletion of the G-rich region also reduced U2 binding. The G element, but not the BP/Py sequence, was also required for U11 binding to the NRS in vivo as assessed by minor class splicing from the NRS to a minor class 3'ss from the P120 gene. These results indicate that efficient U11 binding to the isolated NRS involves at least two elements in addition to the U11 consensus sequence and may have implications for U11 binding to authentic splicing substrates.

extreme example: it is estimated that more than 40 viral splice variants are produced in an HIV-infected cell (5)(6)(7). In addition to alternative splicing, retroviruses also exhibit incomplete splicing such that a large fraction of the primary transcripts remain unspliced and, in contrast to unspliced host cell mRNAs, are transported to the cytoplasm where they serve as mRNA and as genomes for progeny virions. These RNA processing events make retroviruses useful tools for studying the cellular RNA processing machinery, and determining how these events are controlled is required for understanding these important aspects of viral replication.
RNA splicing takes place in a large macromolecular complex termed the spliceosome in which five small nuclear ribonucleoprotein particles (snRNPs) and a large number of accessory proteins cooperate to identify splice sites and assemble the complex (8,9). A dynamic network of snRNA-snRNA and snRNA-substrate interactions takes place to further identify the splice sites and to form the catalytic core of the spliceosome (10). Two splicing pathways have been identified (U2-dependent and U12-dependent), which excise introns with distinct cis splicing signals and differ in snRNP utilization (9,11). Although U5 is utilized by both spliceosomes, the more abundant U2-dependent spliceosome contains U1, U2, U4, and U6 snRNPs, whereas the minor, U12-dependent pathway utilizes the low abundance U11, U12, U4atac, and U6atac snRNP counterparts. Non-snRNP splicing factors such as SR proteins are also shared between the two pathways (12,13). The assembly pathways and catalytic properties of the two spliceosomes are quite similar despite the differences in splicing signals and spliceosome composition (11), yet important differences between the two pathways have been observed. Notably, U1 and U2 snRNPs are found as discrete particles in nuclear extracts and can independently bind to the 5Јss and branch point sequence (BPS), respectively, whereas U11 and U12 snRNPs can form a di-snRNP that cooperatively binds the 5Јss and BPS and only poorly associates with isolated splice sites (14 -18).
For retroviruses, levels of unspliced RNA are controlled by the inefficient use of 3Ј splice sites and in avian retroviruses, through the action of a splicing inhibitor element within the gag gene termed the negative regulator of splicing, or NRS, that primarily affects splicing to the src 3Јss (19 -23). Interestingly, the NRS also is required for optimal polyadenylation of viral RNA, because deletions or mutations that disrupt NRS splicing inhibition also lead to read-through transcripts (19,24,25). For splicing inhibition, the NRS is thought to act as a pseudo 5Јss that non-productively associates with 3Ј splice sites to prevent their interaction with the authentic 5Јss (26,27). The NRS can also inhibit splicing of heterologous introns in vivo and in vitro (26). Important splicing regulators called SR proteins bind to high-affinity sites in the upstream portion of the ϳ230-nt NRS and serve to recruit U1 snRNP to a downstream binding site (28 -30). U1 binding appears critical, because mutations that diminish U1 binding inactivate the NRS and abolish interactions with a 3Ј splice site (27,30,31). Other mutations in the U1 binding site activate splicing from the NRS (30 -32), suggesting that U1 binding is a required but not final step in the pathway leading to inhibition. These mutations and the potential for an extended base pairing interaction with U6 snRNA (see Fig. 1) suggest the intriguing possibility that inhibition results from an inappropriate, non-productive U6 interaction. The importance of the NRS was shown with viral constructs harboring U1-site mutations, which produced viruses with mildly delayed replication kinetics (19), and by the increased incidence of lymphomas in infected chickens (33).
Interestingly, the NRS also very efficiently binds U11 snRNP of the U12-dependent splicing pathway, about as well as U1 snRNP despite it being ϳ100-fold less abundant than U1 (26). The NRS contains a consensus U11-type 5Јss sequence that overlaps the U1 site but no corresponding U12 consensus sequence, which is highly conserved in U12-type introns. The NRS U11 site is not used in viral or heterologous constructs containing U2-type 3Ј splice sites, but it is used accurately and efficiently when paired with a U12-type 3Јss (30). However, U11 is dispensable for splicing inhibition, because mutations that selectively eliminate U11 binding actually increase splicing inhibition (30,31). Thus, whereas U1 is required for splicing inhibition, U11 is not but may function to fine-tune NRS activity by controlling the degree of U1 binding to the overlapping site. The U11 site plays a role in the virus life cycle, because viral constructs harboring U11-site mutations produce viruses with mildly delayed replication kinetics (19). 2 We are interested in the cis determinants that underlie the surprisingly efficient binding of U11 snRNP to the NRS in the absence of a corresponding U12-dependent 3Јss. Here, we show that the NRS efficiently binds U11 in vitro, whereas 5Ј splice sites from two different bona fide U12-dependent genes do not, and we show that a downstream G-rich region is required for optimal U11 binding to the NRS. Surprisingly, U11 can still bind the NRS in vitro when U12 snRNP is destroyed. The G element was also required for optimal U11 binding in vivo as assessed by U12-dependent splicing from the NRS to the 3Јss of the P120 gene. Unexpectedly, inactivation of U2 snRNP decreased U11 binding in vitro, and deletion of a U2-type 3Јss-like sequence just upstream of the U11 site also decreased in vitro U2 and U11 binding; this deletion did not affect U12-dependent NRS splicing in vivo, however. Thus, two regions surrounding the U11 site are required for efficient U11 binding to the isolated NRS in vitro, which may distinguish the NRS from authentic U12-dependent 5Ј splice sites that do not associate well with U11 in isolation.

EXPERIMENTAL PROCEDURES
Plasmid Constructs-RSV sequences are from the PrC strain (34) with sequence coordinates as described by Schwartz et al. (35). Plasmids described previously are p3ZBB (to generate the short NRS RNA, nt 701-930) (28), p3ZkxMS (to generate the long NRS RNA, nt 701-1011) (30), and pP120 and pNRS-P120 for expression of the P120 splicing cassette (30). The pSP64-SCN4A plasmid was a gift of Adrian Krainer. p3Z-P120 was made by inserting an HindIII-BglII PCR fragment containing the first nt of exon 6 through nt 180 of exon 7 into the HindIII-BamHI sites of pGEM-3Z (Promega). Point mutations in the G tracts of the NRS were made in a version of p3ZkxMS, p3ZMSxma, in which a T residue at NRS nt 939 was converted by overlap PCR to a G to create an XmaI site: the G tract mutations (pMSmG1, pMSmG2, pMSmG3, and pMSmG4) were also made by overlap PCR and are shown in Fig. 5A (primer sequences available upon request). pMSmG1ϩ2 and pMSmG1ϩ3 had the indicated mutations combined. These mutations were also generated in a background where the U1 site in the NRS was mutated (mU1 (30)), yielding the plasmid series pMSmU1mG1, etc. The 3Ј deletion series pMS⌬G4, pMS⌬G3ϩ4, and pMS⌬G2-4 were made by PCR and have the deletions indicated in Fig.  6. The 5Ј deletion series, pMS⌬G1, pMS⌬G1ϩ2, and pMS⌬1-3, were also made by PCR and are depicted in Fig. 7. The mG1, mG2, and mG1ϩ2 mutations were incorporated into a variant of the chimeric expression plasmid pNRS-P120 (30) in which a unique XhoI site was introduced into the intron to facilitate cloning to generate pmG1-P120, pmG2P120, and pmG1ϩ2-P120.
In Vitro Transcription and Affinity Selection-pGEM-3Z plasmids containing NRS fragments were linearized with XbaI (MS version, 701-1011, 331 nt) except for p3ZBB, which was linearized with BamHI, to generate the 258-nt NRS RNA and with PvuII to generate BBextended (497 nt) RNA and T7 RNA, polymerase was used. p3Z-P120 was linearized with BstNI to make the 5Јss half-substrate. For pSP64-SCN4A, XbaI was used for full-length RNA that included the U1 site, BstXI to exclude the U1 site, and BsaI for the 5Ј-half-substrate (sizes shown in Fig. 3); SP6 RNA polymerase was used for these plasmids. All plasmids were transcribed in vitro in the presence of biotin-11-UTP (20% of total UTP). RNAs were incubated under splicing conditions with ATP in HeLa nuclear extract for 20 min at 30°C (36). Where indicated, 2Ј-O-methyl oligonucleotides to U1 (nt 1-14), U2 (nt 27-49), and U7 (nt 3-20) snRNA were added 15 min prior to addition of biotinylated substrate. For snRNA cleavage, extracts were pretreated for 30 min with the deoxyribonucleotides indicated in the figures at 20 M and 0.08 unit/l RNase H (Promega) to degrade snRNAs. Streptavidin-agarose beads were added and mixed at 300 mM KCl for 1 h at 4°C and then washed extensively at 300 mM KCl. Samples were treated with proteinase K to release bound nucleic acids, which were then phenol-extracted, ethanol-precipitated, and separated on an 8 M urea-8% polyacrylamide (19:1) gel. RNA was then electroblotted to a ZetaProbe membrane (Bio-Rad) and hybridized with a riboprobe to U11, U12, or U2 snRNAs and visualized and quantitated using a PhosphorImager.
Transfection of 293 Cells and Analysis of RNA-293 cells were grown in minimal essential medium supplemented with 10% fetal calf serum. Cells grown to about 40 -60% confluence in 6-cm dishes were transfected with 2-3 g of DNA by the calcium-phosphate method (Amersham Biosciences), and total RNA harvested 40 h later was isolated with Qiagen RNeasy columns according to the manufacturer's instructions. For reverse transcription-PCR, 1-2 g of total RNA was reverse-transcribed with an antisense primer directed to pRSV2 vectors sequences downstream of the transcription unit (GCAGACACTCTATGCCTGTGTGG) and common to all RNAs in 20 l using 200 units of reverse transcriptase (Invitrogen) and the manufacturers recommended reaction conditions. For PCR, 2 l of the reverse transcription reaction was subjected to 18 -25 cycles of PCR as described (30), and product levels in the linear range were quantitated using an AlphaImager. Alternatively, PCR was performed with a radiolabeled 5Ј primer, and products were quantitated on acrylamide gels with a PhosphorImager (with similar results). Fig. 1 depicts the major features of the NRS, including the upstream SR protein binding sites and important sequences for snRNP binding. The consensus U1 and U11 5Јss sequences are shown over the experimentally determined snRNP binding sites within the NRS. U1 binding to its suboptimal site can be substantially improved by mutating the U11 site, whereas only a modest improvement in U11 binding was observed upon mutation of the U1 site (30). Thus, U11 binds the NRS very efficiently and out-competes U1. The efficient U11 binding is surprising, because the NRS does not harbor the highly conserved U12-type BP consensus signal, and it is known that efficient U11 binding to bona fide 5Ј splice sites requires a corresponding U12-type 3Јss (17).

RESULTS
U11 Binding to the NRS Does Not Require U12-Although U11 and U12 snRNPs are thought to exist primarily as a di-snRNP in cells, between 60 and 80% of U11 snRNP exists as a mono-snRNP in extracts while the rest is complexed with U12 (16). Thus, it is possible that U11 snRNP binds the NRS as a free snRNP or as a di-snRNP in vitro. To determine if U12 is required for U11 binding to the NRS, U12 snRNA in nuclear extracts was cleaved with oligonucleotides directed toward nt 2-23 (the di-snRNP remains intact) or nt 53-71 (which destroys U12 snRNP and thus, the di-snRNP). U11 binding to the NRS (nt 701-1011) in these extracts was then assessed by affinity selection on biotinylated NRS RNA and Northern blotting for U11 snRNA. As shown in Fig. 2A and consistent with a previous study (16), treatment of extract with the U12 2-23 oligonucleotide cleaved off the 5Ј end of U12 snRNA, whereas U12 53-71 caused complete degradation (lanes 5 and 6). A control oligonucleotide to U6 snRNA had no effect on U12 snRNA (lane 7). None of the oligonucleotides had an effect on U11 snRNA (data not shown). As shown in Fig. 2B, in untreated extracts NRS RNA selected U11 snRNP well (lane 3), whereas binding to a substrate with mutations at the ϩ6 and ϩ7 positions in the U11 site was very low (lane 2). U11 binding to the NRS was not affected by treatment with the U6 control oligonucleotide (lane 7), nor was there a significant effect upon treatment with either of the U12 oligonucleotides (lanes 5 and 6). These results indicate that, in contrast to authentic splicing substrates, U11 snRNP need not be complexed with U12 to bind the NRS.
A G-rich Downstream Element Promotes U11 Binding to the NRS-To determine the efficiency of U11 snRNP binding to the NRS versus authentic minor class 5Ј splice sites, biotinylated NRS or P120 5Јss and SCN4A RNAs were subjected to affinity selection. P120 and SCN4A represent two U12-dependent substrates that have been well characterized (12,37,38). The NRS RNA contained RSV nt 701-1011, the P120 5Јss substrate contained the entire exon 6 through position 66 of the 99 nt intron, and the SCN4A 5Јss RNA included 119 nt of exon (plus 12 nt from the vector) through nt 66 of the intron. As expected, binding of U11 was well above background for the NRS (Fig. 3, compare lanes 2 and 3) but was low to undetectable with isolated 5Ј splice sites from the two U12-dependent genes (lanes 4 and 5). The difference in U11 binding could not be attributed to the presence of an adenosine rather than a guanosine at position ϩ1 of the 5Јss of the P120 transcript, because conversion to guanosine did not improve U11 binding (data not shown). Furthermore, U11 binding was also low for a complete SCN4A substrate and one that contains a downstream U1-type 5Јss, which has been reported to enhance splicing of this substrate (lanes 6 and 7) (12, 37, 38). The low level of U11 binding likely reflects the short time of incubation (20 min). These results indicated that one or more features of the NRS distinguish it from normal U11-type 5Ј splice sites concerning the efficiency of U11 binding.
Examination of the NRS revealed a G-rich region just downstream of the U11 binding site. The prevalence of G residues in this region is significantly higher than in surrounding regions, and the density of G triplets per kilobase is also non-random and substantially higher (Table I). This suggested a possible role for the G tracts and associated factors in mediating efficient binding of U11 to the NRS. To test this hypothesis, U11 binding to two NRS substrates differing in the presence (MS) or  (35), and shaded areas depict regions important for NRS function, including an upstream SR protein binding region that promotes U1 binding (arrow) and the downstream U1/U11 snRNP interaction sites. Shown below is the sequence of the overlapping snRNP binding sites within RSV nt 912-929; the U1 site is overlined, and the U11 site is underlined. Also shown aligned with the NRS are the U1 and U11 5Јss consensus sequences with cleavage sites denoted by arrows and the potential base pairing interaction with U6 snRNA (nt 32-49).

FIG. 2. U12 snRNP is not required for U11 binding to the NRS.
A, cleavage of U12 snRNP in nuclear extract. Nuclear extract was treated under splicing conditions for 30 min at 30°C with 20 M of the indicated oligonucleotide and 0.08 unit/l RNase H (indicated by ϩ and Ϫ), an aliquot was removed, RNA was extracted, subjected to Northern blotting, and the integrity of U12 snRNP was assessed with a U12 probe. B, U11 binding to the NRS in oligonucleotide-inactivated extracts. The extracts described in A were used for affinity selection using non-biotinylated or biotinylated NRS RNA (nt 701-1011), and U11 binding was monitored by Northern blotting with a U11 probe. A representative blot is shown with percentage of NRS signal (by phosphorimaging) below each lane. non, affinity selection with non-biotinylated RNA; 67, the NRS with a mutation in the U11 binding site. The positions of U12 and U11 snRNA are shown to the right. The data are representative of four independent experiments. absence (BB) of the G tracts was assessed. Both of these constructs are capable of splicing inhibition (39) and bind U1 roughly equivalently in vitro (data not shown). As shown in Fig.  4, U11 binding was less efficient with the substrate lacking the G tracts (compare lanes 3 and 4, and see Figs. 6, 7, and 9 below). The decreased binding to the shorter NRS was not due to an "end effect," because extending this substrate by 239 nt with vector sequences did not restore efficient U11 binding (lane 5). These results indicated that the downstream G-rich region is required for efficient U11 binding to the NRS.
U11 snRNP Binding to G-tract Point Mutants-Closer inspection of the G-rich region downstream of the U11 site revealed four blocks of G residues (Fig. 5A). The G tracts have at least three G residues in a row and a maximum of six and ranged from long sequences of G3CCAG6 (G2) and G5UG3 (G1) to G5CG2 (G3) and G3CG2 (G4). To determine if one or more of the G tracts was involved in promoting U11 binding to the NRS, point mutations were introduced into the G tracts, and their effects on U11 binding were assessed. Because U11 and U1 snRNPs compete for binding to overlapping sites just upstream of the G-rich region (Fig. 1), these substrates also contained a mutation in the U1 site that reduces U1 binding (30) to eliminate the possibility that mutations in the G tracts had an indirect effect on U11 binding, for example by increasing the binding of U1. A 60% decrease in U1 binding was observed with the mU1 substrate in this experiment (Fig. 5B, compare lanes  1 and 3), and small and variable effects on U1 binding were observed with the G tract mutations (lanes 4 -9). There was also little effect of the G tract mutations on U1 binding to substrates containing the wild-type U1 site (data not shown). In contrast, mutations in the first two G tracts substantially reduced U11 binding (compare lanes 4 and 5 to lane 3), whereas mutations in G tracts 3 and 4 had less or little effect, respectively, on U11 binding (lanes 6 and 7). Combining mG1 with mG2 yielded low U11 binding, and the relatively efficient binding of mG3 was eliminated when combined with mG1 (lanes 8  and 9). These results suggested that the first two G tracts are predominantly involved in optimal U11 binding, that neither alone was sufficient, and that the G tracts have little impact on U1 binding in this context. G Tracts 1 and 2 Are Sufficient for Maximal U11 Binding-To investigate the above indication that G1 and G2 are primarily involved in U11 recruitment, the G tracts were sequentially deleted from the 3Ј end and U11 binding was assessed as above. Curiously and in contrast to the above data, deletion of G4 caused a substantial decrease in U11 binding (Fig. 6, lane 4). However, deletion of G4 and G3 (leaving G1 and G2) restored U11 binding to optimal levels (lane 5). Although G1 appears to be important for optimal U11 binding, it is not sufficient because binding was markedly reduced in the G2-4 deletion (lane 6). Taken with the results in Figs. 4 and 5, these data are most easily interpreted as G1 and G2 being required and sufficient for optimal U11 binding and that G3 has a potentially negative effect on U11 binding in the context of the G4 deletion.
G Tracts Are Not Interchangeable-The observations that mutations in G tracts 3 and 4 had little effect (Fig. 5B) on U11 binding suggested that close positioning to the U11 site might be required for function such that G3 and G4 are suboptimally effective due to distance, and thus their mutation shows little effect. Alternatively, the sequence of G3 and G4 may not support efficient U11 recruitment. To determine if G3 and G4 might function when repositioned closer to the U11 site, the G tracts were sequentially deleted from the 5Ј end (Fig. 7). Like the mG1 point mutation, deleting G1 reduced U11 binding to the level of the NRS that lacks the G region (compare lanes 4 and 7), again indicating a requirement for G1 in optimal U11 binding. U11 binding remained inefficient upon deletion of G1 plus G2 and G1 minus G3 (lanes 5 and 6). These results show that placing G3 and G4 at the position of G1 and G2 is not sufficient for efficient U11 binding. We conclude that G3 and G4 do not play a prominent role in efficient U11 binding to the NRS.
U2 snRNP Inactivation Decreases U11 Binding to the NRS in Vitro-As a complement to the data in Fig. 5B that indicated a direct role for the G tracts in U11 binding rather than an indirect effect of competition with U1 binding, we inactivated U1 snRNP in nuclear extract with a 2Ј-O-methyl oligonucleotide to the 5Ј end of U1 snRNA. Consistent with the data in Fig.  5, U11 binding was variably but mildly affected upon inactivating U1 snRNP or by a control oligonucleotide to the nonspliceosomal U7 snRNP (Fig. 8, lanes 4 -7 and data not shown).  Unexpectedly, a control oligonucleotide to U2 snRNP reduced U11 binding in a dose-dependent manner (lanes 8 -11). This oligonucleotide is directed to nt 27-49, which are involved in binding to the branch point of splicing substrates. These data suggested an unexpected role for U2 snRNP in U11 binding to the NRS in vitro. Deletion of the NRS Branchpoint-like/Polypyrimidine Tract Sequence Decreases U2 and U11 Binding to the NRS in Vitro-U2 snRNP was shown by immunoprecipitation to interact with the NRS (26), presumably to a branchpoint-like/pyrimidine tract (BP/Py) region just upstream of the U1/U11 sites. This region resembles splicing signals associated with 3Ј splice sites that are required for U2 binding, but the precise U2 binding site within the NRS has not been determined. To ask if these cis signals might mediate the U2 effect and if they are required for optimal U11 binding in vitro, the BP/Py tract region was deleted and U2 and U11 binding to the NRS was assessed. As shown in Fig. 9, U2 binding as assessed by affinity selection was weak but detectable on the wild-type NRS (lane 3), and the BP/Py deletion reduced U2 binding by ϳ 60% (lane 4). The BP/Py deletion also reduced U11 binding to the level observed for the NRS lacking the G tracts (compare lanes 4 and 5), indicating that this region is also required for optimal U11 binding. Interestingly, U2 binding was also lower on the substrate lacking the G tracts (lane 5). These data suggest that the binding of U2 and U11 are mutually beneficial. In contrast to these results, however, the BP/Py deletion did not appear to affect U11-type splicing (and therefore, binding) from the NRS in vivo (data not shown, see below and "Discussion").
G Tracts 1 and 2 Are Required for U12-dependent Splicing from the NRS in Vivo-To address whether the G element was required for U11 binding to the NRS in vivo, we took advantage of an earlier observation that splicing occurs efficiently and accurately from the NRS U11 site when an authentic U12-dependent 3Јss is provided (30), and thus, NRS minor class splic- ing provides a measure of U11 binding in vivo. We reasoned that, if the mutations that diminished U11 binding in vitro acted similarly in vivo, then U12-dependent splicing from the NRS would be reduced. Human 293 cells were transfected with constructs expressing an authentic P120 splicing cassette, the NRS-P120 chimera, a control construct harboring a mutation in the NRS U11 site, and constructs containing the NRS mG1, mG2, or mG1 plus 2 mutations, which were shown above to compromise U11 binding in vitro (Fig. 10A). As shown in Fig.  10B and quantitated in Fig. 10C, splicing from the NRS-P120 chimera was as efficient as the native P120 substrate, and splicing from the NRS was largely abolished by a mutation in the U11 site that abolishes U11 binding in vitro (lanes 3-5) (30). The latter result demonstrates that the splicing observed is by the U12-dependent pathway. When the single G mutations were introduced, splicing decreased from ϳ70% for wild type to ϳ25%, and combining the mutations inhibited splicing slightly more (lanes 6 -8). This result is consistent with a decrease in U11 binding to the NRS due to the G tract mutations. We conclude that the G tracts are required for optimal U11 binding to the NRS in vivo.

DISCUSSION
The RSV NRS element is distinct from the viral splice sites yet functions to suppress splicing and promote polyadenylation. Considerable evidence suggests that binding of U1 snRNP to the NRS commits RNAs to the inhibition pathway whereby a non-functional spliceosome complex assembles on the NRS and an authentic 3Јss, which sequesters the 3Јss from the normal interaction with the authentic 5Јss (27). It is likely that inhibition takes place at steps subsequent to U1 binding and may involve U6 snRNP. 3 This stalled splicing complex is also proposed to promote 3Ј end processing at the distant RSV polyadenylation site (24). Although U11 snRNP appears to play no direct role in splicing inhibition (30,31), its efficient binding to the NRS may regulate the degree to which the NRS functions and provides an interesting model to investigate novel cis and trans factors that contribute to U11 snRNP binding to 5Ј splice sites. In this report we have identified two regions that are required for optimal binding of U11 to the NRS in vitro: an upstream 3Јss-like sequence and a downstream G-rich element. The G-rich element also promotes U11 binding in vivo as assessed by U12-dependent splicing from the NRS.
The major and minor splicing pathways are remarkably similar, but there are important differences between the two (11). First, the 5Јss and BPS of the U12-dependent introns are highly conserved, which contrasts with U2-dependent introns whose 5Јss and BPS consensus sequences are not strictly adhered to. The minor class introns also lack a polypyrimidine tract between the BPS and 3Јss, which is a hallmark of major class introns (40). Additionally, the minor class snRNPs are of quite low abundance and, unlike U1 and U2 snRNPs of the major pathway, U11 and U12 exist as a di-snRNP and act   9. Analysis of U2 and U11 snRNP binding to ⌬BP/Py and NRS RNA lacking the G-rich region. At the top is the sequence of the branch point-like/ pyrimidine tract region just upstream of the U1 (overlined) and U11 (underlined) binding sites. The region from nt 875 to 901 (between the arrows) was deleted in ⌬BP/ Py. A schematic (not to scale) of the RNAs is at the left, and a representative affinity selection experiment is shown on the right. M, nuclear extract RNA as a marker for U2 and U11 snRNA (positions of U2 and U11 are indicated on the right); non, nonbiotinylated NRS RNA. The percentage of NRS signal for U2 and U11 is shown below each lane. cooperatively in early steps of 5Јss and BPS recognition (16 -18). Thus, U11/U12 binds poorly to isolated splice sites. It is therefore novel that the NRS binds U11 so well in the apparent absence of a standard U12-type BPS (UUCCUUAAC). Consistent with previous reports (17), little U11 associated with authentic U11-type 5Јss substrates in an affinity selection assay where efficient binding is observed with the NRS. Although di-snRNP does interact with the NRS as exemplified by coselection of U12 snRNP (26) (data not shown), surprisingly, U11 binding to the NRS was not affected when U12 snRNA was destroyed, indicating that U11 can bind the NRS efficiently as a mono-snRNP. One possibility is that the high affinity SR protein binding sites in the 5Ј half of the NRS and associated SR proteins account for the differences in U11 binding to the NRS and authentic substrates. It has been shown that SR proteins are required for in vitro splicing of minor class introns (12). However, the activity of a distinct cis element was indicated, because efficient U11 binding is not dependent upon the upstream SR protein binding sites (30). 3 Our results show that a G-rich region just downstream of the U11 site contributes to efficient U11 binding. The ϳ80-nt region exhibits a non-random, increased frequency of G residues and G triplets that roughly segregate into four blocks and whose deletion results in substantially decreased U11 binding. Curiously, the four G blocks are not functionally redundant. The first two G tracts are required for efficient U11 binding, but neither is sufficient, and blocks 3 and 4 cannot substitute even when moved to the position of blocks 1 and 2. Furthermore, the action of G tracts 1 and 2 does not appear to be additive, because mutation or deletion of either one reduces U11 binding to levels near that observed when the region is completely removed. It may be that both G1 and G2 must be occupied by a trans-acting factor, or alternatively, they may comprise a single element. It is not obvious what distinguishes between active and inactive G tracts. Additional mutagenesis studies will be required to understand the subtle differences that differentiate active G1 and G2 from inactive G3 and G4.
The decrease in U11 binding upon mutational inactivation of the G region does not appear to be an indirect effect of increasing the binding of U1 snRNP to the NRS, which competes with U11 for binding to an overlapping site. A more direct role of the G tracts in U11 binding is supported by the finding that U11 binding to the G-tract mutants was not restored by a mutation in the U1 binding site or by 2Ј-O-methyl oligonucleotide inactivation of U1 snRNP. Surprisingly, a control oligonucleotide to U2 snRNP decreased U11 binding in a dose-dependent manner. This is an interesting observation, because it was shown previously by immunoprecipitation that U2 interacts with the NRS (26), and upstream of the U1/U11 site is a sequence that resembles the rat calcitonin-specific BPS followed by a pyrimidine tract that is a strong candidate for the U2 binding region (39,41). Weak binding of U2 to the NRS was eliminated when the BP/Py tract was deleted. This suggests that U2 can associate with this region but additional experiments are required to confirm this possibility. The BP/Py tract also appears important for U11 binding, because its deletion reduced U11 binding to levels seen with substrates lacking the downstream G-rich region. These observations establish a correlation between the BP/Py sequence and U2 and U11 binding. One possibility is that pyrimidine tract-binding splicing factors bind the NRS BP/Py tract and assist downstream binding of U11 in a manner similar to exon definition interactions described for U2AF and U1 snRNP (42,43). However, the observation that the U11 interaction was decreased by the U2 snRNA inactivating 2Ј-Omethyl oligonucleotide suggests a more direct role for U2 in U11 binding. U2 and U11 binding appears to be mutually beneficial, because the substrate lacking the G region and that binds U11 poorly also showed reduced U2 binding. It is possible that U2 and U11 interact through a bridging protein such as hPrp5, as has been proposed for cross-intron and/or exon interactions between U2 and U1 snRNPs (44). Alternatively, the G tracts might also influence U2 binding directly in this context.
The influence of the G tracts on U11 is also not restricted to in vitro binding. The U11 site in the NRS is normally not used, but splicing is activated when the minor class 3Јss from the P120 gene is provided (30). The NRS fragment used in the P120 chimera lacks G tracks 3 and 4, which is consistent with the in vitro observation that these motifs are not required for efficient U11 binding. Importantly, the same mutations that diminished U11 binding in vitro caused a marked decrease in minor class splicing from the NRS in vivo. A decrease was not observed when the upstream SR protein binding sites 3 or the BP/Py-like region was removed (data not shown), indicating a direct affect of the G tracts on U11 binding. It is possible that the BP/Py region is required for optimal U11 binding to the isolated NRS but is dispensable in the presence of a minor class 3Јss.
A large number of cis elements have been identified that influence splicing efficiency in the major splicing system but that are distinct from the splice sites (e.g. intronic and exonic splicing enhancers and silencers) (9). In contrast, aside from the 5Јss and BPS, descriptions of regulatory cis elements have been few for U12-dependent introns, enhancers being the exception (13). Thus, identification of novel elements that promote U11 binding to the NRS may have implications for splicing of authentic minor class introns. Interestingly, like the NRS, intron sequences just downstream of the SCN4A U11 5Јss are rich in G residues and G triplets, although this is not a general feature of all U12-dependent introns (45). This raises the possibility that aspects of efficient U11 binding to the viral NRS element will be applicable to cellular U12-dependent genes and further our understanding of minor class splicing. Identification of factors that associate with the G-rich element to influence U11 binding is an important future goal.