The Negative Regulator of Splicing Element of Rous Sarcoma Virus Promotes Polyadenylation

The Rous sarcoma virus gag gene contains a cis-acting negativeregulator of splicing (NRS) element that is implicated in viralpolyadenylation regulation. To study the mechanism of polyadenylationpromotion at the viral poly(A) site located over 8 kb downstream,we performed in vitro polyadenylation analysis. RNA containingonly the poly(A) site and flanking sequences in the 3' longterminal repeat (LTR) was not polyadenylated detectably in vitro;however, if the transcript contained the NRS upstream of theLTR, polyadenylation was observed. Insertion of the viral env3' splice site sequence between the NRS and the LTR did notalter the level of polyadenylation appreciably. We concludethat the NRS promotes polyadenylation in vitro and can do sowithout formation of a splicing complex with a 3' splice site.We then explored the roles of several cellular factors in NRS-mediatedpolyadenylation. Mutation of the binding sites of U1 and U11snRNPs to the NRS did not affect polyadenylation, whereas hnRNPH strongly inhibited polyadenylation. We propose a model inwhich hnRNP H and SR proteins compete for binding to the NRS.Bound SR proteins may bridge between the NRS and the 3' LTRand aid in the recruitment of the 3'-end processing machinery.

INTRODUCTION

Top

Introduction

The replication of Rous sarcoma virus (RSV), like that of allretroviruses, requires unspliced RNA in the cytoplasm, whereit functions in a dual role as mRNA for gag-pol translationand as packaged genomic RNA (reviewed in reference 8). Becauseretroviruses depend on the cellular machinery for RNA transcription,processing, export, and translation, they have evolved specialmechanisms to aid in the accumulation of unspliced RNA. Somecomplex retroviruses, such as human immunodeficiency virus type1, encode accessory proteins to promote export of incompletelyspliced RNAs; however, simple retroviruses like RSV have notbeen found to encode accessory proteins (6, 8) and instead usecis-acting RNA sequences to interact with cellular factors.Avian retroviruses have been found to contain weak 3' splicesites and a cis-acting negative regulator of splicing (NRS)element within the gag gene, which contribute to the maintenanceof unspliced retroviral RNA (2, 13, 33). In addition, two directrepeat sequences, which flank the src gene of RSV, are involvedin unspliced RNA export (37). The RSV polyadenylation signal(AAUAAA) is in the U3 region of the 3' long terminal repeat(LTR) and is associated with a downstream GU-rich sequence (18).However, polyadenylation of avian retroviruses is relativelyinefficient, and the viral poly(A) site is read through about15% of the time (19).

The NRS is a 230-nucleotide (nt) RNA sequence which has twofunctional domains: (i) SR and hnRNP H protein-binding domainsin its purine-rich 5' portion (12, 30) and (ii) overlappingU1 and U11 small nuclear ribonucleoprotein (snRNP) binding sitesat its 3' end (5, 15, 20, 32, 39). The NRS element has beenproposed to act as a decoy 5' splice site, which generates anonproductively spliced complex with a downstream 3' splicesite (9, 13, 14). Previous evidence suggests that, in additionto its role in splicing inhibition, the NRS may also stimulateRSV polyadenylation. Miller and Stoltzfus (34) showed that largedeletions, which overlapped in a 606-nt region including theNRS, led to a large increase in readthrough transcription ofthe viral RNA. In addition, deletion of a region containingthe env 3' splice site sequence also led to increased readthrough(34).

Recently, more-limited mutations in the NRS have also been shownto increase the level of readthrough transcription. O'Sullivanet al. (38) showed that point mutations in the NRS pseudo-5'splice site sequence lead to increased readthrough of the viralpoly(A) site. A 42-nt deletion in the purine-rich 5' portionof the NRS also causes increased readthrough (40). Avian leukosisviruses (ALVs) bearing NRS mutations induce short-latency B-celllymphomas at a high frequency after infection of chicken embryos(23, 40, 43). In these clonal tumors, readthrough transcriptsfrom ALV into the downstream cellular myb gene have been observed;the presence of these transcripts results in overexpressionof an oncogenic, truncated Myb protein (23, 43). Finally, Fogelet al. (11) carried out extensive mutagenesis of the 5' portionof the NRS and found that the polyadenylation-stimulatory activityof the NRS in vivo does not require hnRNP H but is dependentupon the binding sites for SR proteins and snRNPs. Thus, thereis evidence to show that RSV RNA splicing and polyadenylationregulation are coupled by the NRS, but it is not known how thisoccurs mechanistically or what NRS-bound factors are functionallyimportant.

mRNA 3'-end processing has been shown many times to be closelycoupled with other gene expression events in the nucleus. Recognitionof the cleavage and polyadenylation signal by cleavage factorsis important for transcript termination, and several generalpolyadenylation factors can be found in association with theC-terminal domain of RNA polymerase II (21, 28). In addition,the efficiency of removal of upstream introns is promoted bypolyadenylation (35, 36), just as splicing factors that interactwith the terminal 3' intron reciprocally promote 3'-end processing(28, 29, 44). Polyadenylation itself can be a regulated event;for example, several cellular and viral genes contain alternativepolyadenylation signals that are used differentially duringdevelopment or in different tissues (10). It is of interestto study how polyadenylation signals are regulated and gainfurther insight into the tightly coupled interactions in geneexpression. The mechanism by which the NRS promotes 3'-end processingis novel because most auxiliary elements previously shown toregulate polyadenylation are located within close proximityto the core polyadenylation elements (1, 17, 25), whereas theNRS is located over 8 kilobases upstream of the polyadenylationsite of the primary RSV RNA transcript.

To investigate the mechanism by which the NRS promotes polyadenylationat the 3' LTR from a distance, we used in vitro analysis. Wefound that the NRS is able to promote polyadenylation directlywithout the need for an intervening, downstream 3' splice site.Additionally, we found that hnRNP H strongly inhibits polyadenylation.We suggest that there is a competition between hnRNP H and apolyadenylation-stimulatory factor(s), possibly SR proteins,for binding to the NRS. Further, we propose that SR proteinsare able to form a bridging interaction between the NRS andthe polyadenylation site and use protein-protein interactionsto recruit the 3'-end processing machinery.

MATERIALS AND METHODS

Top

Materials and Methods

Plasmid constructions. The 3' LTR sequence of the PrC strain of RSV from nt 9196 to9392 was synthesized by PCR using forward (5'-GATTCTAGGATCCGAAGGTATCAGACGGGTCTAACATG)and reverse (5'-CTTAGGAATTCAAATGAAGCCTTCTG) oligonucleotideprimers (Operon). This PCR product was inserted into the BamHIand EcoRI sites of pBluescript II KS(+) (Stratagene) to generatethe LTR construct. The NRS sequence from nt 706 to 1007 wassynthesized by PCR using forward (5'-CATGCTTCTAGACCGGAGTGCATCGAG)and reverse (5'-CTAGCAGGATCCCTGGCTCCGCCC) oligonucleotide primersand inserted into the XbaI and BamHI sites of the LTR constructto produce the NRS LTR construct. The triple point mutant NRS.1was described previously (15, 20, 38). The env sequence fromnt 4995 to 5130, including the 3' splice site at nt 5078, wassynthesized by PCR using forward (5'-GATCGTGGATCCGGTACCCTCTCGAAAAG)and reverse (5'-GCTACAGGGATCCTTTCTCCTTGGAGTCC) oligonucleotideprimers and inserted into the BamHI restriction site of theLTR construct to yield the ENV LTR construct. The env 3' splicesite fragment was similarly amplified and inserted into theNRS LTR construct to make the NRS ENV LTR construct.

In vitro polyadenylation assays and poly(A) selections. Capped, ³²P-labeled runoff transcripts were synthesized by invitro transcription using T7 RNA polymerase as previously described(13). Plasmid templates were linearized with HindIII, and transcriptswere purified from 6% polyacrylamide-8 M urea gels prior touse. HeLa nuclear extracts were prepared as previously described(27) from HeLa cells obtained from the National Cell CultureCenter. A typical 13-µl in vitro polyadenylation reactionmixture contained a final concentration of 2.3% polyvinyl alcohol,15 mM creatine phosphate, 1 mM ATP, 1 mM magnesium chloride,and 56% (vol/vol) HeLa nuclear extract. We used adeno-associatedvirus (AAV) RNA made from a template obtained from Jeff Wilusz(7) as a positive control in the polyadenylation assays. Inactivationof U1 and U2 snRNAs using complementary oligonucleotides wascarried out as described by Bruzik and Steitz (4). In reactionswhere hnRNP H protein in the extract was sequestered by GRScompetitor RNA, a synthetic RNA oligomer (5'-GGGGGAGGUGUGGG)was added directly to reaction mixtures, as described previously(1).

Before addition of RNA, the reaction mixtures were preincubatedfor 20 min at 30 degrees. After the RNA was added, polyadenylationreaction mixtures were incubated for 60 min at 30°C. RNAsrecovered from the reactions were poly(A) selected using anoligo(dT)-cellulose column to isolate poly(A)⁺ RNA (42), separatedon a 6% polyacrylamide-8 M urea gel, and quantified on an InstantImager(Packard).

Cross-linking assays. ³²P-labeled NRS RNAs were UV cross-linked to proteins in HeLanuclear extract, digested with RNase A, and electrophoresedon sodium dodecyl sulfate-polyacrylamide gels as described byWilusz and Shenk (45).

RESULTS

Top

Results

The NRS promotes and is required for polyadenylation at the 3' LTR in vitro. In previous studies, it has been shown that mutations in theNRS element of RSV affect splicing regulation and cause readthroughof the cleavage/polyadenylation site in the 3' LTR (11, 34,38, 40). We wanted to use in vitro analysis to determine themechanism by which the NRS promotes viral RNA polyadenylationat a site more than 8 kb downstream. We first wished to determineif a transcript containing the poly(A) signal in the RSV LTR,together with flanking sequences, can be polyadenylated in vitro.

For this purpose, capped, ³²P-labeled runoff transcripts weresynthesized; these transcripts contained a portion of the LTR(nucleotides 9196 to 9392) that flanks the major poly(A) siteat nt 9312 (Fig. 1A). In vitro polyadenylation assays were thenperformed using HeLa nuclear extract. After incubation withthe extract, RNAs were recovered and analyzed directly by polyacrylamidegel electrophoresis. In some cases, the recovered RNAs werealso fractionated on an oligo(dT)-cellulose column to selectfor poly(A)⁺ RNA. As a positive control, we analyzed the polyadenylationof an AAV polyadenylation substrate (7).

View larger version (44K):
[in this window]
[in a new window]
FIG. 1. The NRS promotes polyadenylation in vitro. (A) Schematic representation of the RSV proviral genome with an expanded view of the NRS, the region flanking the env 3' splice site, and the portion of the LTR used in this study. The U3 region of the LTR begins at nucleotide 9058, and the major cleavage/polyadenylation (pA) site is at nt 9312 (arrow). hnRNP H and SR proteins have been shown to bind the 5' end of the NRS, whereas U1 and U11 snRNPs bind at the 3' end. Polyadenylation substrates are schematically represented. (B) In vitro polyadenylation assays. The indicated radiolabeled, capped RNAs were incubated in in vitro polyadenylation reaction mixtures containing HeLa cell nuclear extract in the presence or absence of ATP. The arrowheads indicate the precursor RNA. (C) After in vitro polyadenylation reactions were performed with the indicated transcripts, poly(A)⁺ RNA was isolated using an oligo(dT) column. –, unpolyadenylated RNA that flowed through the column; +, poly(A)⁺ RNA eluted. The arrowhead at the left of each autoradiogram indicates the precursor RNA. Polyadenylated transcripts in the NRS LTR and NRS ENV LTR are denoted reactions.

While the control AAV RNA was efficiently polyadenylated inour in vitro reactions, the RSV LTR substrate was not detectablypolyadenylated (Fig. 1B). To further analyze the LTR RNA afterincubation in the polyadenylation reaction, it was run throughan oligo(dT)-cellulose column to select for poly(A)⁺ RNA. Whilea small amount of LTR RNA was eluted from the oligo(dT) column,it was the same size as the input RNA (Fig. 1C). Thus, the 3'LTR alone showed no detectable polyadenylation in this assay,suggesting that other viral elements may be necessary.

Miller and Stoltzfus (34) previously showed that the efficiencyof RSV 3'-end processing was decreased by deleting either oftwo upstream RSV RNA sequences: (i) a 520-nucleotide sequence(nt 630 to 1149) in the gag gene containing the NRS as wellas flanking sequences and (ii) a 136-nucleotide sequence (nt4995 to 5130) spanning the env 3' splice site (nt 5078), whichincluded exonic splicing enhancer sequences (24). Consequently,we wanted to determine whether either of these sequences promotespolyadenylation in vitro.

First, the region spanning the env 3' splice site from nucleotides4995 to 5130 was cloned in front of the 3' LTR to generate theENV LTR construct (Fig. 1A). Like the 3' LTR alone, this constructwas not polyadenylated (Fig. 1C), suggesting that this env fragmentis not sufficient to promote polyadenylation in vitro. It ispossible that a larger region, encompassing this env gene sequence,may be necessary to play a role in polyadenylation. Alternatively,additional viral elements not included in our ENV LTR constructmay be required for such an effect.

Next, to test whether the NRS promotes polyadenylation in ourin vitro system, a 301-nucleotide fragment containing the NRSwas cloned in front of the 3' LTR to generate the NRS LTR construct(Fig. 1A). Because the NRS has been proposed to act as a decoy5' splice site, which generates a nonproductively spliced complexwith a downstream 3' splice site (9, 13, 14), we also made aconstruct, named NRS ENV LTR. This construct contained the NRS,the env 3' splice site fragment described above, and the 3'LTR, to test whether formation of a splicing complex is necessaryto promote polyadenylation. As shown in Fig. 1B and C, the NRSLTR and NRS ENV LTR transcripts were both polyadenylated, showingthat the NRS promotes polyadenylation in vitro (greater-than-10-foldactivation compared to the LTR alone) and that it does so directlywithout the need to form a complex with a 3' splice site.

It should be noted that the polyadenylation efficiency of theNRS LTR transcript was appreciably less than that of the AAVRNA analyzed in parallel as a positive control (Fig. 1B). Averagingfour experiments, the NRS ENV LTR transcript was polyadenylatedat a level similar to the NRS LTR (6.1% ± 4.5% more efficientlythan NRS LTR transcript, which is not statistically significant).The nuclear extract and conditions used for the in vitro polyadenylationassays are nearly identical to those used in our laboratoryfor in vitro splicing (13, 15). Because the addition of theenv fragment between the NRS and LTR had a minimal effect onpolyadenylation efficiency, we focused our next studies on theNRS LTR construct to determine how the NRS promotes polyadenylation.

A splicing complex containing U1 or U11 snRNP is not necessary for NRS-mediated polyadenylation. After showing that the NRS promotes polyadenylation in vitroin the absence of the env splice site sequence, we investigatedthe role that known NRS-binding factors may play in polyadenylationregulation in vitro. At the 3' end of the NRS, there are overlappingU1 and U11 snRNP binding sites (Fig. 2A) (20). U1 snRNP bindinghas been previously shown to both promote and inhibit polyadenylationdepending on the substrate (17, 25). Our laboratory has shownthat point mutations in the U1 binding site of the NRS in full-lengthviral constructs cause increased readthrough transcription invivo, thus implicating U1 binding in polyadenylation regulationin RSV (38). In particular, the NRS.1 triple point mutation(diagrammed in Fig. 2A), which disrupts both the U1 and U11binding sites, caused the greatest effect in that study by increasingreadthrough transcription from the wild-type level of 15% to23%. We thus chose the NRS.1 mutation to use in this study toinvestigate what effect U1/U11 binding may have on polyadenylationin vitro.

View larger version (40K):
[in this window]
[in a new window]
FIG. 2. Effect of U1/U11 snRNP binding on polyadenylation. (A) Schematic of the NRS LTR construct used in this study and the mutation introduced to disrupt U1/U11 snRNP binding. The consensus major and minor 5' splice site sequences are indicated. (B) In vitro polyadenylation assays. The indicated capped, radiolabeled RNAs were incubated in in vitro polyadenylation reaction mixtures containing HeLa cell nuclear extract, and then poly(A)⁺ RNA was isolated using an oligo(dT) column. –, unpolyadenylated RNA that flowed through the column; +, poly(A)⁺ RNA eluted after washes. (C) U1 and U2 snRNAs in the nuclear extract were inactivated using complementary oligonucleotides. In vitro polyadenylation reactions using the NRS LTR substrate were then performed, followed by isolation of poly(A)⁺ RNA using an oligo(dT) column.

The NRS.1 mutant NRS sequence was cloned next to the 3' LTR,and we performed an in vitro polyadenylation assay followedby a selection for poly(A)⁺ RNA. We found that NRS.1 LTR wasable to be polyadenylated only 2% ± 10% more efficientlythan NRS LTR, which is a statistically negligible difference(Fig. 2B). This result, therefore, suggests that the U1/U11binding sites of the NRS do not play an important role in thepolyadenylation promotion activity of the NRS in vitro. It alsosupports the finding above that formation of a splicing complexis not necessary for the NRS to promote polyadenylation in vitro.

A possible role for U1 snRNP in polyadenylation regulation wasfurther investigated using an oligonucleotide directed againstthe 5' end of U1 snRNA to inactivate the U1 snRNP (4). As acontrol, we used a similar oligonucleotide directed againstU2 snRNP. Polyadenylation of the NRS LTR transcript in vitrowas not detectably affected by preincubation of the nuclearextract with either oligonucleotide (Fig. 2C).

hnRNP H binding to the NRS inhibits polyadenylation. Because intact U1 and U11 snRNP binding sites were not necessaryfor the polyadenylation-stimulatory activity of the NRS, wenext investigated the role that factors that bind the 5' purine-richregion of the NRS may play. It has previously been shown thathnRNP H interacts with the NRS (12, 40). Additionally, hnRNPH is known to promote polyadenylation by binding downstreamof the core polyadenylation signal of some cellular mRNAs (1,3). hnRNP H-specific mutations in the RSV NRS do not affectthe polyadenylation-stimulatory activity of the NRS in vivo;nevertheless, Fogel et al. suggested that hnRNP H may inhibitpolyadenylation (11). We were thus interested in investigatingthe role of hnRNP H binding to the NRS in 3'-end processingin vitro.

Rather than performing extensive mutagenesis of the NRS sequence,we chose to alter the level of available hnRNP H proteins presentin the HeLa nuclear extract. This was carried out by addinga synthetic RNA oligonucleotide, GRS (5'-GGGGGAGGUGUGGG), whichcontains a G-rich binding site for hnRNP H protein, as describedpreviously by Arhin et al. (1). To determine the effect of sequestrationof hnRNP H protein on the processing efficiency of the NRS LTRconstruct, increasing amounts of GRS RNA were added to the invitro polyadenylation reaction mixtures (Fig. 3A). When 230ng of competitor RNA was added to the nuclear extract, NRS LTRpolyadenylation was 1.9-fold ± 0.1-fold greater thanthat observed in an untreated extract (Fig. 3B). As a control,an oligomer containing an unrelated sequence was also incubatedwith nuclear extract and found to have no effect on polyadenylationefficiency (Fig. 3C). We also observed dose-dependent inhibitionof UV cross-linking of hnRNP H to the NRS RNA after incubationof the extract with the GRS competitor RNA but not with thecontrol oligonucleotide (Fig. 3D). Thus, we propose that theincrease in polyadenylation in the presence of the GRS RNA waslikely due to the prevention of hnRNP H binding to the NRS.

View larger version (52K):
[in this window]
[in a new window]
FIG. 3. hnRNP H inhibits polyadenylation. (A) To sequester endogenous hnRNP H, nuclear extract was supplemented with increasing amounts of GRS competitor RNA. In vitro polyadenylation reactions using the NRS LTR substrate were performed as described in Materials and Methods. (B) Poly(A)⁺ RNA was subsequently isolated using an oligo(dT) column. –, unpolyadenylated RNA that flowed through the column; +, poly(A)⁺ RNA eluted. (C) A control oligonucleotide containing an unrelated sequence had no effect on the polyadenylation efficiency of the NRS LTR substrate. In vitro polyadenylation reactions were performed followed by fractionation on an oligo(dT) column to separate poly(A)^– and poly(A)⁺ RNA. Two hundred thirty nanograms of oligonucleotide was used. (D) UV cross-linking of hnRNP H ( $~$ 53 kDa) to the NRS RNA is inhibited after incubation of the extract with increasing amounts of GRS competitor RNA but not with 230 ng of the control oligonucleotide. Protein molecular weight markers are indicated.

To verify that the increase in polyadenylation efficiency wasdue to decreased hnRNP H binding to the NRS rather than to theLTR RNA itself, we measured the processing efficiency of theLTR construct when GRS RNA was added to the polyadenylationreaction mixtures. Sequestering hnRNP H did not activate polyadenylationof the LTR alone (data not shown), suggesting that it is hnRNPH binding to the NRS that is regulating polyadenylation. Wethus can conclude that hnRNP H binding to the NRS greatly inhibitspolyadenylation and define a novel role for hnRNP H as an inhibitorof polyadenylation.

DISCUSSION

Top

Discussion

In this study, we investigated the mechanism by which the avianretroviral NRS element promotes polyadenylation at a site over8 kb downstream in the 3' LTR. While no detectable polyadenylationwas observed in vitro with the LTR RNA substrate alone, additionof the NRS sequence upstream of the LTR enhanced polyadenylation.Surprisingly, the viral env 3' splice site failed to enhancepolyadenylation in vitro. Further, hnRNP H strongly inhibitedthe polyadenylation-stimulatory activity of the NRS. Since SRproteins have been shown previously to compete with hnRNP Hfor binding to the NRS (11, 12, 30), we suggest that SR proteinsare responsible for the polyadenylation promotion activity ofthe NRS. We propose a model in which the NRS directly promotespolyadenylation by binding SR proteins, which form a bridgebetween the NRS and the 3' LTR and help to recruit the polyadenylationmachinery (Fig. 4).

View larger version (19K):
[in this window]
[in a new window]
FIG. 4. Model for direct polyadenylation stimulation by the NRS. hnRNP H and SR proteins both competitively bind to the NRS. hnRNP H inhibits SR protein binding and thus inhibits the polyadenylation-stimulatory effect. We propose that SR proteins are able to bridge between the NRS and the polyadenylation site in the 3' LTR and aid in the recruitment of the 3'-end processing machinery through protein-protein interactions.

Fogel et al. (11) previously suggested an alternative modelin which the NRS promotes polyadenylation by forming a complexwith downstream 3' splice sites, thus stabilizing the bindingof splicing factors at the weak viral 3' splice sites. In contrast,our data show that an interaction with a 3' splice site is notnecessary for the NRS to promote polyadenylation; insertionof the env 3' splice site between the NRS and the poly(A) signaldid not increase levels of polyadenylation in vitro. Furthermore,neither U1 or U11 snRNP binding sites on the NRS nor an activeU1 snRNP in the nuclear extract is required to promote polyadenylationof the NRS LTR transcript. It is possible that the env fragmentwe used was too small to have an effect or that a full-lengthviral RNA transcript is needed. Thus, we cannot rule out thepossibility that additional polyadenylation enhancement is providedin vivo by the interaction of the NRS with the 3' splice sitefragment, which contains ESE sequences (24) that also bind SRproteins.

Sequestration of hnRNP H with an oligonucleotide prevented itscross-linking to the NRS RNA and dramatically increased thepolyadenylation efficiency of the NRS LTR transcript. This suggeststhat binding of hnRNP H to the NRS interferes with the NRS polyadenylation-stimulatoryactivity. The purine-rich 5' end of the NRS contains overlappingbinding sites for hnRNP H and SR proteins (12, 30, 31, 40).SR proteins are essential, multifunctional splicing factorsrequired for different steps of spliceosome assembly and arethought to mediate cross-exon and cross-intron interactionsbetween splicing factors bound at the 5' and 3' splice sites(16). In addition, SR proteins have been demonstrated to beinvolved in RNA nuclear export, RNA stability, mRNA qualitycontrol, and translation (reviewed in reference 22). However,SR proteins have not been shown previously to be involved inpromoting 3'-end processing.

We have previously shown that a 42-nt deletion in the purine-rich5' portion of the NRS causes reduced hnRNP H and SR proteinbinding as well as an increase in viral readthrough transcriptsin vivo (40). This deletion in ALVs is associated with an increasedincidence of rapid-onset lymphomas associated with readthroughinto downstream cellular oncogenes (43). Integrating previouswork with our current data, we propose the following model ofhow the NRS regulates polyadenylation at the 3' LTR. hnRNP Hand SR proteins compete for binding to the 5' half of the NRS.When SR proteins, including SF2/ASF, bind to the NRS, they forma bridge between the NRS and the 3' LTR. The arginine/serine-rich(RS) domains of the SR proteins then make protein-protein interactionswith members of the cleavage and polyadenylation machinery,recruiting it to the cleavage/polyadenylation site in the 3'LTR (Fig. 4).

We were surprised to observe that U1 snRNA inactivation hadno effect in our study, despite a U1 binding site being only170 nt upstream of the polyadenylation site in the NRS LTR construct.Our laboratory has previously shown that U1 snRNP binding sitemutations in full-length viral constructs cause increased readthrough,suggesting that U1 promotes polyadenylation in vivo (38). Itis also known that U1 binding sites near a polyadenylation sitecan strongly inhibit 3'-end processing, even at distances muchlonger than 170 nt (17, 41). However, in our in vitro study,mutating the U1 binding site had no effect on polyadenylationefficiency. We thus can only suggest that U1 may play a minorrole in stimulating polyadenylation in vitro, but the strongereffects of SR proteins, hnRNP H, and other effectors are maskingany observable effects. In addition, a spliceosome-like complex,involving U1 snRNP and 3' splice sites (13), may aid in bringingthe NRS and 3' LTR spatially close together, but we proposethat the SR proteins themselves play the major role in helpingto recruit the 3'-end processing machinery.

It is now becoming clear that polyadenylation signals are verydiverse and that AAUAAA is not ubiquitous at the 3' ends ofall eukaryotic mRNAs as once thought. Data from expressed sequencetag databases have suggested that the actual incidence of AAUAAAis much lower, perhaps as low as 50 to 60% (26). For propergene expression, the correct cleavage and polyadenylation sitemust be identified, just as the proper 5' and 3' splice sitesmust be recognized within vast stretches of intronic RNA. SRproteins are known to bind to exonic splicing enhancers to recruitthe splicing machinery to the proper splice sites. Similarly,SR proteins may bind to specific polyadenylation enhancer sequences(like the NRS) and help to define the proper cleavage/polyadenylationsite and recruit the 3'-end processing machinery. In summary,we propose that SR proteins may be major regulators of all stepsof mRNA processing, recognizing the proper RNA sequences andrecruiting the proper machinery to those sites.

ACKNOWLEDGMENTS

We thank Jeff Wilusz for reagents and much helpful advice, YingyingLi for technical assistance, and Yun-Xing Wang for T7 polymerase.

This work was supported by NIH research grant RO1 CA238796 toK.L.B. and a Pfizer Summer Undergraduate Research fellowshipto J.E.W.

FOOTNOTES

* Corresponding author. Mailing address: Department of Biology, Johns Hopkins University, 3400 North Charles Street, Baltimore, MD 21218. Phone: (410) 516-7289. Fax: (410) 516-7292. E-mail: KLB{at}jhu.edu.

Present address: Watson School of Biological Sciences, ColdSpring Harbor Laboratory, Cold Spring Harbor, NY 11724.

REFERENCES

Top