Polyadenylation Sites - Molecular and Cellular Biology - American

0 downloads 0 Views 3MB Size Report
Sequence Elements Upstream of the 3' Cleavage Site Confer. Substrate .... poly(A) site is the binding of cleavage and polyadenylation ...... Montell, C., E. F. Fisher, M. H. Caruthers, and A. J. Berk. 1983. ... Schek, N., C. Cook, and J. C. Alwine.
Vol. 14, No. 7

MOLECULAR AND CELLULAR BIOLOGY, July 1994, p. 4682-4693

0270-7306/94/$04.00+0

Copyright X 1994, American Society for Microbiology

Sequence Elements Upstream of the 3' Cleavage Site Confer Substrate Strength to the Adenovirus Li and L3 Polyadenylation Sites JOHN PRESCOTTt AND ERIK FALCK-PEDERSEN* Department of Microbiology, W R. Hearst Research Foundation, Cornell University Medical College, New York New York 10021 Received 1 February 1994/Returned for modification 24 February 1994/Accepted 28 April 1994

The adenovirus major late transcription unit is a well-characterized transcription unit which relies heavily on alternative pre-mRNA processing to generate distinct populations of mRNA during the early and late stages of viral infection. In the early stage of infection, two major late transcription unit mRNA transcripts are generated through use of the first (Li) of five available poly(A) sites (LI through L5). This contrasts with the late stage of infection when as many as 45 distinct mRNAs are generated, with each of the five poly(A) sites being used. In previous work characterizing elements involved in alternative poly(A) site use, we showed that the LI poly(A) site is processed less efficiently than the U3 poly(A) site both in vitro and in vivo. Because of the dramatic difference in processing efficiency and the role processing efficiency plays in production of steady-state levels of mRNA, we have identified the sequence elements that account for the differences in Li and U poly(A) site processing efficiency. We have found that the element most likely to be responsible for poly(A) site strength, the GU/U-rich downstream element, plays a minor role in the different processing efficiencies observed for the Li and L3 poly(A) sites. The sequence element most responsible for inefficient processing of the Li poly(A) site includes the Li AAUAAA consensus sequence and those sequences which immediately surround the consensus hexanucleotide. This region of the Li poly(A) site contributes to an inability to form a stable processing complex with the downstream GU/U-rich element. In contrast to the Li element, the L3 poly(A) site has a consensus hexanucleotide and surrounding sequences which can form a stable processing complex in cooperation with the downstream GU/U-rich element. The L3 poly(A) site is also aided by the presence of sequences upstream of the hexanucleotide which facilitate processing efficiency. The sequence UUCUUUUU, present in the L3 upstream region, is shown to enhance processing efficiency as well as stable complex formation (shown by increased binding of the 64-kDa cleavage stimulatory factor subunit) and acts as a binding site for heterogeneous nuclear ribonucleoprotein C proteins.

and a less conserved GU/U-rich region located within 50 nucleotides downstream of the cleavage site (21, 28, 29, 33, 38, 39, 52, 53, 71). For poly(A) sites associated with retroviral long terminal repeats, additional cis elements are found upstream of the AAUAAA (5, 16, 49, 50, 62). The simian virus 40 (SV40) late poly(A) site is an example of a nonretroviral poly(A) site which contains a very well characterized upstream 3' processing efficiency element (4, 55). Relatively little is known about how these upstream elements are involved in mRNA 3'-end formation. Studies by Gilmartin et al. (22) have recently demonstrated that the human immunodeficiency virus (HIV) upstream element helps to stabilize the preprocessing complex on the HIV pre-mRNA substrate in vitro. Purification of the factors involved in 3' processing (1, 9, 10, 23, 24, 36, 59-61) and their use in ordered addition studies have demonstrated the sequence of events leading to successful 3'-end processing. The first step in the recognition of a poly(A) site is the binding of cleavage and polyadenylation specificity factor (CPSF) (formerly known as PF2, CPF, and SF) to the AAUAAA hexanucleotide (24, 25, 31, 64). The CPSF-pre-mRNA complex is relatively unstable. Addition of cleavage stimulatory factor (CStF) (formerly known as either CStF or CF1) to the CPSF-pre-mRNA complex results in formation of a stable preprocessing complex (24, 25, 64). Formation of a CStF-stabilized complex is facilitated by the presence of a GU/U-rich sequence located downstream of AAUAAA (58, 70). CStF does not form a stable complex with the substrate pre-mRNA independent of CPSF. Once the

Based on the work of several laboratories, the biochemistry of mRNA 3'-end formation (for reviews, see references 35, 63, and 65) demonstrates a level of complexity and sophistication strikingly similar to that described for initiation of transcription by RNA polymerase II (for a review, see reference 72). The requirement of the consensus AAUAAA in the first step of building a 3' processing complex is reminiscent of the role of the TATA sequence at the 5' end of the transcription unit; the subsequent steps required to stabilize the preprocessing complex are also reminiscent of the events which lead to formation of a stable transcription initiation complex. As is the case with polymerase II promoters, which vary in their ability to form stable initiation complexes, characterization of 3' processing sites for many transcription units indicates that poly(A) sites vary in their capacity to act as a substrate for binding of the 3' processing factors. Since successful 3'-end formation is a mandatory step in mRNA production, this step can be used to regulate mRNA expression (2, 13). In transcription units transcribed by polymerase II in mammalian cells, two RNA sequence elements have been identified as basic prerequisites for processing at a poly(A) site, the highly conserved AAUAAA hexanucleotide, usually located 10 to 35 nucleotides upstream of the cleavage site (20, 33, 40, 66), * Corresponding author. Phone: (212) 746-6514. Fax: (212) 7468587. t Present address: Dept. of Microbiology, University of California at San Francisco, San Francisco, Calif.

4682

DETERMINANTS OF POLY(A) SITE STRENGTH

VOL. 14, 1994

stable CPSF-CStF complex has formed on the pre-mRNA, the pre-mRNA is considered to be committed to cleavage and polyadenylation (24). While cleavage at the poly(A) site requires the addition of at least two more protein components [cleavage factor and poly(A) polymerase], their role in stable complex formation and the cleavage reaction is not well characterized (9, 23, 24, 60, 61). We have been using 3' processing elements from the adenovirus major late transcription unit (MLTU) to study the mechanisms involved in controlling 3'-end formation in complex transcription units. The MLTU contains five poly(A) sites (L1 through L5). During the course of an adenoviral infection, the predominant MLTU mRNA generated during the early stage of infection (before viral DNA replication) is processed at the Li poly(A) site. After DNA replication and the onset of late gene expression, all five poly(A) sites are used to generate stable mRNA (reviewed in reference 45). Previous studies using minigene adenovirus recombinants have shown that the poly(A) sites Li and L3 contain all of the cis information necessary to elicit a temporal switch in poly(A) site use (14, 18). Studies have also shown that the Li poly(A) site is a weak substrate RNA for 3' processing machinery compared with L3, with L3 being 20- to 50-fold more efficient than Li in competing for processing factors in vitro (48). Characterization of these poly(A) sites in vivo supports the notion that the Li poly(A) site is weak in comparison to the L3 poly(A) site (17). This difference in processing efficiency is likely to be an important part of the early to late MLTU poly(A) site processing switch mechanism. On the basis of previous characterizations of the cis elements required for efficient polyadenylation, it is not clear why these sites differ so greatly in their ability to support the formation of a stable cleavage complex. To determine the basis of the difference in substrate processing efficiency, we have divided each poly(A) site into three segments, an upstream element which includes the sequence 5' to AAUAAA, a core element which includes the sequence from 10 nucleotides 5' to the hexanucleotide to the site of cleavage (including AAUAAA), and a downstream element including all of the sequence downstream of the cleavage site. A series of chimeric poly(A) sites in which upstream, core, or downstream elements were exchanged between the weak Li and the strong L3 poly(A) sites was constructed. Using a combination of in vivo and in vitro assays, we have determined the relative contribution of each element to processing efficiency. The inefficiency of the Li poly(A) site is in large part the result of the sequence present in the Li core element, presumably by causing weak interaction between CPSF and CStF during the formation of the cleavage complex. In the L3 poly(A) site, where the core AAUAAA functions well in complex formation, we have shown that sequences upstream of the core element also function to increase processing efficiency, both in L3 and in Li-L3 chimeric constructs. A U-rich sequence located approximately 50 nucleotides upstream of the L3 AAUAAA is an important part of the upstream element. The presence of this element is associated with enhanced cleavage efficiency and more efficient or more stable cleavage complex formation.

MATERIALS AND METHODS Plasmids. The parent plasmids pL1 and pL3, which contain -200 bp of adenovirus sequence flanking the Li and L3 poly(A) sites, respectively, have been described previously (48). ApaI and NcoI restriction endonuclease sites were introduced 10 bp upstream of the AATAAA and at the cleavage

4683

site, respectively, to generate pL1.1.1. and pL3.3.3., as described previously (48). The ApaI and NcoI restriction endonuclease sites in pL1.1.1. and pL3.3.3. were used in conjunction with XbaI and EcoRI polylinker sites as breakpoints for the exchange of the upstream sequences, the AATAAA-containing core sequences, and the downstream GT/T-rich sequences between the two plasmids. Using these restriction endonuclease sites and standard molecular cloning techniques, the family of chimeric poly(A) site-containing plasmids (pL1.1.3., pL1.3.1., pL3.1.1., pL3.3.1., pL3.1.3., and pL1.3.3.) was generated. Substitution of upstream sequences with nonspecific spacer sequences was done by digesting the parent plasmids (pL1.1.1. and pL3.3.3.) with HincIl and ApaI to remove all adenovirus sequences greater than 10 bp upstream of the AATAAA. The 3' overhang generated byApaI digestion was removed with T4 DNA polymerase, and a 100-bp ThaI fragment of pGEM2 (positions 2057 to 2157) was inserted into the vector in both orientations to generate pLsp.1.1.(+), pLsp.3.3.(+), pLsp. i.i.(-), and pLsp.3.3.(-). The chloramphenicol acetyltransferase (CAT) constructs were made by first removing the SV40 poly(A) site from pML.SIS.CAT (30) by digesting with KpnI and NarI. The respective 3' and 5' overhangs were blunted with T4 DNA polymerase and deoxynucleoside triphosphates (dNTPs), and the vector was recircularized with T4 DNA ligase to generate pML.SIS.CAT(-PA). All poly(A) sites were first removed from the appropriate pGEM2 constructs by XbaI and EcoRI digestion, and the resulting 5' overhangs were filled in with dNTPs and the Klenow fragment of DNA polymerase I. These isolated poly(A) sites were cloned either directly into pML.SIS.CAT(-PA) that had been linearized with BamHI and filled in with dNTPs and the Klenow fragment or into the BamHI-linearized pML.SIS.CAT(-PA) with the use of BamHI synthetic linkers (5'-CGGGATCCCG-3'). Spinner cells and extract preparation. HeLa cells were grown in suspension in Joklik's modified minimal essential medium containing 5% horse serum and used for nuclear extract preparation as described previously (48). Preparation of 32P-labeled pre-mRNA. High-specific-activity [a-32P]UTP-labeled pre-mRNAs, used in all in vitro cleavage assays, were transcribed and purified as described previously (48). In the UV cross-linking experiments, either [c_-32P]UTPlabeled pre-mRNA or uniformly labeled pre-mRNA was prepared with 250 ,uM 7meG(5')ppp(5')G (Pharmacia LKB Biotechnologies, Inc.); 25 pRM GTP, ATP, UTP, and CTP (Pharmacia); and 0.4 ,uM a-32P-labeled GTP, ATP, UTP, and CTP (3,000 Ci/mmol; 10 mCi/ml) (Du Pont, New England

Nuclear). In vitro cleavage and polyadenylation assays. In vitro cleavage and polyadenylation assays (34, 42-44) were conducted as described previously (48). Monolayer cells, transfections, and CAT assays. HeLa cells were grown as monolayers in Dulbecco's modified Eagle's medium containing 5% calf serum. For transfection, cells were plated onto 6-well plates containing 35-mm wells (Becton Dickinson-Falcon) 24 to 48 h prior to transfection. At 80% confluency, cells were transfected with 1 pRg of pRSVLUC, 3.5 jig of pGEM2, and 1 ,ug of the specific pML.SIS.CAT construct per well by the CaPO4 precipitation procedure (54). Forty-eight hours after transfection, cells were washed three times with phosphate-buffered saline (PBS) (without Ca2+ or Mg2+), incubated for 5 min at room temperature in 1 ml of scrape buffer (0.04 M Tris-HCI [pH 7.4], 1 mM EDTA, 0.15 M NaCl), gently scraped off the plates, pelleted by a 1-min microcentrifuge spin, and resuspended in 100 RI of 0.25 M Tris

4684

PRESCOTT AND FALCK-PEDERSEN

(pH 7.8). Cells were lysed by freezing on dry ice and thawing at 30'C three times. Cellular debris was removed by a 5-min microcentrifuge spin, and 30 [L1 of the supernatant was assayed for luciferase activity with an Analytical Luminescence Laboratory Monolight 2010 according to the method of Braiser et al. (3). The remaining 60 p.1 of extract was heated at 650C for 5 min to inactivate deacetylases, and any precipitate was removed by a 5-min microcentrifuge spin. After correcting for transfection efficiency, an appropriate volume (5 to 50 RI1) of the heat-inactivated extract was assayed for CAT activity (54). Reaction products were separated on a silica gel thin-layer chromatography plate in chloroform-methanol (95:5) and visualized by exposure to XAR-2 Kodak film. Reaction products were quantitated with a Molecular Dynamics Phosphorlmager. Preparation of 32P-end-labeled SI DNA probes. Plasmids pML.SIS.CAT Li and pML.SIS.CAT L3 were digested with the restriction endonucleases EaeI and NaeI, and the respective 723- and 720-bp fragments were purified from an agarose gel after electrophoretic separation. The 3' end of the complementary strand was labeled with the Klenow fragment of DNA polymerase I in 0.7 mM [at-32P]dGTP and 400 mM (each) dTTP and dCTP. Random primer labeling of DNA probes. Double-stranded DNAs containing either luciferase cDNA (1,650-bp HindIIIEcoRI fragment from pRSVLUC) or CAT cDNA (1,080-bp XbaI-BamHI fragment from pML.SIS.CAT) were purified from an agarose gel after electrophoretic separation. Heatdenatured DNAs were primed with random hexamers and labeled with the Klenow fragment of DNA polymerase I in 0.3 p.M [cX-32P]dCTP and 400 p.M (each) dGTP, dATP, and dTTP as described by Feinberg and Vogelstein (19). Si nuclease protection assays. Forty-eight hours after transfection with either pML.SIS.CAT Li or pML.SIS.CAT L3, HeLa cells were washed three times with PBS and RNA was isolated as described by Chomczynski and Sacchi (8). Twenty micrograms of RNA and 6 fmol of 32P-end-labeled cDNA (-30,000 cpm/fmol) were heat denatured in 40 p.l of hybridization buffer (80% formamide, 1 mM EDTA, 400 mM NaCl, 40 mM PIPES [piperazine-NN'-bis(2-ethanesulfonic acid]). After hybridization overnight at 52°C, 500 p.l of ice-cold Si buffer (400 mM NaCl, 30 mM sodium acetate, 1 mM ZnSO4, 20 mg of denatured herring sperm DNA per ml) and 250 U of Si nuclease (Bethesda Research Laboratories, Inc.) were added, and reactions were incubated for 1 h at 37°C. Following Si digestion, reactions were extracted with 500 p.1 of phenolCHCl3 (1:1) and precipitated with 2 volumes of ethanol. Reaction products were separated on a denaturing polyacrylamide gel and visualized by autoradiography. Reaction products were quantitated with a Molecular Dynamics PhosphorImager. Measurement of CAT mRNA half-life. In our experience, transfected HeLa cells have reached steady-state levels of RNA by 36 h posttransfection and maintain near linear expression for up to 72 h posttransfection. We measured RNA turnover in transfected HeLa cells by adding 6 pRg of actinomycin D (Sigma) per ml to cells 40, 42, 44, 46, and 47 h after transfection. Cells were harvested at 48 h after transfection, with samples representing exposure to actinomycin D for 0, 1, 2, 4, 6, and 8 h. We also carried out mRNA half-life determinations by adding actinomycin D at 40 h posttransfection and harvesting cells at 0, 1, 2, 4, 6, and 8 h after adding actinomycin D (data not shown). The half-lives of LI calculated by these two technical variations were in close agreement at 2.5 and 2.8 h, respectively. In all cases, RNA purification was as follows. Cells were washed three times with PBS, and RNA was isolated as described by Chomczynski and Sacchi (8). RNA

MOL. CELL. BIOL.

from two 35-mm wells was resuspended in 400 [L1 of DNase I digestion buffer (40 mM Tris-HCl [pH 8], 10 mM NaCl, 6 mM MgCl2) and incubated for 30 min at 370C with 3 U of RNase-free DNase I (RQ1) (Promega). Reactions were extracted with phenol-CHCl3 (1:1), and the aqueous phase was precipitated with 0.25 M NaCl and 2 volumes of ethanol. The RNA was resuspended in water and quantitated by UV absorption. Aliquots (5 pRg) were immobilized on a nitrocellulose membrane with the Minifold II Slot-blotter (Schleicher & Schuell) in duplicate. The nitrocellulose filters were hybridized to either a random primer-labeled luciferase cDNA or a random primer-labeled CAT cDNA. Hybridizations were carried out in 50% formamide-5 X Denhardt's solution-150 mg of denatured herring sperm DNA per ml-0.25% sodium dodecyl sulfate (SDS)-5x SSPE (lx SSPE is 0.18 M NaCl, 10 mM NaH2PO4, and 1 mM EDTA [pH 7.7]) for 24 h at 420C. Filters were washed with 0.1 X SSPE-0.1% SDS at 650C and exposed to Kodak XAR-2 film. Counts per minute were quantitated with a Molecular Dynamics PhosphorImager. UV cross-linking and immunoprecipitations. UV crosslinking reactions contained 0.5 mM MgCl2, 2 mM EDTA, 1% polyvinyl alcohol, 12 mM HEPES (N-2-hydroxyethylpiperazine-N'-2-ethanesulfonic acid [pH 7.9]), 90 mM KCI, 6% glycerol, 0.3 mM dithiothreitol, 0.06 mM phenylmethylsulfonyl fluoride, - 1 mg of protein from a HeLa nuclear extract per ml, and 0.44 mg of Escherichia coli tRNA per ml. Aliquots (20 pl) were added to 20 fmol of 32P-labeled pre-mRNA, and reactions were incubated for 1 min at 30°C. Reactions were then exposed to UV irradiation (MR-4 254-nm UV light; George W. Gates & Co., Inc.) (5 cm from the source) for 3 min on ice. After irradiation, samples were digested with 7 pug of pancreatic RNase A (Boehringer Mannheim) for 20 min at 37°C. Reaction products were separated on SDS-denaturing polyacrylamide gels and visualized by autoradiography. For immunoprecipitation, UV cross-linking reactions were done in quadruplicate. After RNase A digestion, one reaction was used as a control while the other three were used for immunoprecipitation reactions. Two hundred microliters of IP-2 buffer (59) (50 mM Tris [pH 7.6], 50 mM NaCl, 0.05% Nonidet P-40) was added, along with either 50 p.l of hybridoma supernatant containing the 3A7 monoclonal antibody directed against the 64-kDa subunit of CStF (59) (kindly provided by J. Manley), 1 p.1 of mouse ascites fluid containing the 4F4 monoclonal antibody directed against the heterogeneous nuclear ribonucleoprotein (hnRNP) C1 and C2 proteins (7) (kindly provided by G. Dreyfuss), or 50 p.1 of the M73 hybridoma supernatant containing anti-ElA antibody (kindly provided by S. Shiff). Reactions were incubated on ice for 8 h with occasional gentle shaking. One hundred microliters of goat anti-mouse immunoglobulin G (IgG) (whole molecule)-agarose (Sigma), washed and diluted 1:10 in IP-2 buffer, was added, and the reactions were rotated gently at 4°C for an additional 8 h. Reaction mixtures were pelleted in a microcentrifuge and washed two times in IP-2 buffer before being separated on an SDSdenaturing polyacrylamide gel and visualized by autoradiography. RESULTS In vitro cleavage studies of the weak Li and the strong L3 poly(A) sites demonstrated a 4- to 10-fold difference in processing efficiency and a 20- to 40-fold difference in the ability of the two poly(A) site substrates to compete for 3' processing factors in a HeLa nuclear extract (48). We have recently shown that these poly(A) sites function in vivo in a manner which resembles their use in vitro (17), with Li as a weak poly(A) site

DETERMINANTS OF POLY(A) SITE STRENGTH

VOL. 14, 1994

A

B

mRNA

SV40 E/P

2

1

3

5

6

CAT

pML.SIS.CAT Li |

pML.SIS.CAT L3

Ll

I Li

D

L3

mLI

C

m

I

615 492 -

VW

369-

, cleaved and

poyadenylated

2 3 4 5 6 7

1 374nt ** **

L3

-PA

8 hr 4 hr 2 hr pML.SIS.CAT L1 1 hr o hr-

unprocessed

a

-

397nt

4

4685

-~ m 4M

8 hr 4 hr 2 hr 1 hr

pML.SIS.CAT L3

Ohrl - / +control

unprocessed Li processed LI unprocessed 13 processed L3

FIG. 1. Characterization of CAT gene products and mRNAs resulting from transfections of pML.SIS.CAT constructs using the adenovirus Li and L3 poly(A) sites. (A) Schematic of the plasmid used to transfect HeLa monolayer cells. DNA fragments (-220 bp) containing either the Li or the 12 poly(A) sites were subcloned into the 3' untranslated region in place of the SV40 late poly(A) site present in the parent plasmid. EaeI and NaeI restriction endonuclease cleavage sites, used for the production of SI endonuclease protection probes, are indicated. Intron sequences are represented by the narrow portion of the CAT transcribed sequences. (B) CAT assays were carried out with extracts from cells transfected with LI (lanes 1 and 2), 12 (lanes 3 and 4), or no poly(A) site (-PA) (lanes 5 and 6)-containing plasmids. Reaction products were separated on a silica gel thin-layer chromatography plate in CHCl3-MeOH (95:5). (C) SI nuclease protection assays were carried out with RNA isolated from untransfected cells (lanes 2 and 6) or cells transfected with CAT plasmids containing either the Li (lane 3) or the L3 (lane 7) poly(A) site. Hybridizations contained 3' 32P-end-labeled DNA probes complementary to the Li-containing (lanes 2 and 3) or the 12-containing (lanes 6 and 7) CAT mRNAs. Fiftyfold dilutions of the Li (lane 1)- and 12 (lane 5)-containing end-labeled DNA probes indicate the size and integrity of each undigested probe. DNA molecular weight markers (lane 4) indicate approximate sizes of SI nuclease digestion products. Reaction products were electrophoresed on a 3.5% polyacrylamide-8 M urea gel. End-labeled (*) DNA probes and SI nuclease digestion products are shown schematically at the bottom. (D) Slot blot analysis of RNA isolated from transfected cells at 0, 1, 2, 4, and 8 h after actinomycin D addition, immobilized in duplicates to a nitrocellulose filter, and hybridized to 32P-labeled CAT cDNA. The plasmids used to transfect the cells, along with the length of exposure to actinomycin D, are indicated. Negative and positive hybridization controls, consisting of 0.5 ng of pGEM4 and pML.SIS.CAT, respectively, are indicated.

and L3 as a strong site. We have determined the in vivo processing efficiency of each of these sites by a transient transfection assay. This assay employs an adapted CAT expression system, pML.SIS.CAT (30), to measure the CAT gene product as a function of poly(A) site efficiency (Fig. 1A and B) (17). Transfection of pML.SIS.CAT(-PA) [which lacks an authentic poly(A) site] into HeLa cells, followed by standard extract preparation and CAT assay protocols, results in CAT gene product levels which are undetectable by our standard assay. This is in contrast to the abundant gene product levels found in cells transfected with the plasmid pML.SIS.CAT 23, which contains the L3 poly(A) site used in the in vitro cleavage poly(A) site is shown to be 10- to 20-fold assay (48). The more efficient than the Li poly(A) site when parallel pML. SIS.CAT Li and pML.SIS.CAT 13 transfections are assayed (all assays have been normalized with the cotransfection control pRSVLUC, as indicated in Materials and Methods).

The in vivo assay consistently reveals a greater difference between Li and L3 processing efficiency than that observed with the in vitro cleavage assay (17, 47a). Possible explanations for this observation are that the in vivo assay is a more sensitive indicator of processing efficiency than the in vitro assay or that secondary events affecting mRNA stability or translatability in vivo influences the observed levels of CAT gene product. Total RNA was isolated from cells transfected with each of the poly(A) site plasmid constructs used in the CAT assay, and mRNA was characterized by an S1 nuclease hybrid protection assay (Fig. 1C). The Ll-generated mRNA was much less abundant than mRNA generated by processing at the L3 poly(A) site. This result is consistent with the level of translated CAT gene product generated from transfections using the same respective plasmids. To eliminate the possibility that differences in mRNA stability contributed to the differences noted between Li and L3 mRNA in terms of gene product

4686

MOL. CELL. BIOL.

PRESCOTT AND FALCK-PEDERSEN

A

Apa

Nco

Ll wt L1.1.1.

--.CACCGA r.CGTTGGTTT-TCT-TGTATT CCT'TCGCCCCAGGCTGGGGAGAATGrMAAAAAAAAAAAAAAAfAGCAT(ikTC.CAAAATAAAAAACTCACCAAGCFCATG

L3 wt L3.3.3.

AC TCGGAGATGATTATTTA CCCC >ACTT A^TA^^GATATTAT.TT CCAWCTTCTTTTTGTCACTTG\A^ACATGA^ACATGTACT

NcoI

Apal CCACTTCTrTTTGTCACTTGAAAAAACATGTAtAAATAATGTACTi OGC

CTTTCA^T^AAAGGCAAAT GT-TTTTATTTGT CG

TGATTATTTACCCC

25 nt GU/Ul

IAAUAAAI

L1.1.1. Li .1 .3.

1AAU AAA I W

L1.3.1. L3.1.1. L3.3.3. [S . @ Sf:.>$: L3.3.1. L3.1 .3. L 1 .3.3.

.

GU/UI

X_~~~GU

He

.B:f~~i......... .:' .:':''GUU/U....... :.&f.:.:.. . . 2'S. . :BB :N:: B~~ ~ ~g.

B tfu 0,

10

3.I60 C)

= 40 0 a

4C 20 C) A

0.~ -I

Y,

.

-

i

(W .

.r

Y-

0:

a-

Y-

X

JJJ1

0-

0-

,-:

0-

.10i

1r

-i

FIG. 2. In vivo determination of 3' processing site efficiency for Li and L3 chimeric poly(A) sites. (A) DNA sequences encoding the wild-type adenovirus Li (L1 wt) and L3 (L3 wt) poly(A) sites, aligned with altered parental constructs (L1.1.1. and L3.3.3.) used to construct the chimeric poly(A) sites. Sequence alterations are indicated by vertical bars between the pairs, and the resulting restriction endonuclease sites are boxed and labeled. The family of chimeric precursor mRNAs are shown at the bottom. The known cis processing elements, AAUAAA, the cleavage site (contained within the NcoI recognition site), and the GU/U-rich element, are indicated on each precursor mRNA. All chimeric pre-mRNAs are drawn to scale, as indicated by the 25-nucleotide bar above them. (B) The average normalized CAT activity of each chimera compared with those of the wild type and the altered parent. Data represent a minimum of three transfections (each done in duplicate) for each construct. Reaction products were quantitated with a Molecular Dynamics Phosphorlmager as described in Materials and Methods.

accumulation,

we

performed

an

actinomycin D RNA half-life

measurement of transcripts generated from both the pML.SIS.

CAT L3 and pML.SIS.CAT LI plasmids (Fig. 1D). The degradation rate for these two mRNAs are slightly different, with the weaker Li poly(A) site having the longer half-life (2.3 h for Li and 1.1 h for L3). This indicates that the steady-state measurements of mRNA and CAT protein slightly overrepresent the products from pML.SIS.CAT Li compared with those from pML.SIS.CAT L3. We conclude from these experiments that the in vivo CAT assay is a slightly more sensitive assay than the in vitro cleavage assay and that the in vivo processing efficiency of each poly(A) site is consistent with its ability to compete for processing factors in vitro (17, 48). Assuming that the transcription rate from these constructs is the same, data also indicate that only a small fraction of available Li premRNA transcript is converted into stable cytoplasmic mRNA. Our studies do not reveal the fate of the remainder of the unprocessed Li pre-mRNA, but an efficient and rapid nuclear RNA degradation pathway is known to exist (11).

Sequences upstream of the LI and L3 cleavage sites have the greatest impact on poly(A) site strength. To identify the sequence elements responsible for the differences in processing efficiency between Li and L3, we divided each poly(A) site into three comparable segments: sequences upstream of AAUAAA, sequences immediately flanking and including AAUAAA, and sequences downstream of the cleavage site (Fig. 2A). This was done by first introducing point mutations upstream of AATAAA in both poly(A) site plasmids, generating a unique ApaI restriction endonuclease site 7 to 12 bp upstream of AATAAA. Then, since the Li poly(A) sitecontaining plasmid naturally contains an Ncol site at the cleavage site, a unique Ncol restriction endonuclease site was similarly introduced at the cleavage site in the L3-containing plasmid. These base substitutions are in areas not expected to influence poly(A) site efficiency (12,56). Comparing constructs containing the parent Li or L3 poly(A) sites to the modified

versions, L1.1.1. and L3.3.3., indicates the substitutions had a minimal effect on processing as determined by the CAT assay

(Fig. 2B; compare Li with L1.1.1. and L3 with L3.3.3.). The resulting constructs, pL1.1.1. and pL3.3.3., were used to create the family of chimeric constructs in which the downstream GT/T, the core AATAAA, or the upstream sequence elements were exchanged. Using this family of chimeric poly(A) sites, we were able to assess the relative contributions of each of these sequence elements to poly(A) site strength. Given the sequence heterogeneity seen among GU/U-rich elements and the role the downstream region plays in the formation of a stable cleavage complex on the pre-mRNA, it seemed likely that the processing efficiency of the Li and L3 poly(A) sites would in large part be determined by differences in their respective GU/U-rich elements. The chimeric constructs L1.1.3. and L3.3.1. represent a reciprocal exchange of downstream elements. When expression from these constructs was compared with that from their respective parent constructs (Fig. 2B; compare LM.1. 1. with LM.1.3. and L3.3.3. with L3.3.1.), we found that the exchange of the downstream region between the strong L3 and the weak Ll poly(A) sites had only a modest effect on poly(A) site strength. Thus, while the downstream GU/U-rich elements of both Li and L3 are required for poly(A) site processing (27, 51), they are functionally interchangeable elements and not responsible for the difference in poly(A) site strength. The constructs which showed an unexpectedly large change in processing efficiency were those which exchanged the core AAUAAA-containing elements, L1.3.1. and L3.1.3. The L1.3.1. construct demonstrated a 300% increase in the amount of CAT gene product compared with that of the L1.1.1. construct but was considerably less efficient than L3.3.3. The L3.1.3. construct resulted in a 70% reduction in the amount of CAT gene product compared with that of L3.3.3. (but this amount was still greater than that seen with L1.1.1.). Thus, while the required consensus hexanucleotide AAUAAA is present in both the Li and L3 poly(A) sites, the sequences flanking the core hexamers are quite different and contribute significantly to the strength of these poly(A) sites. The results seen with the exchange of the AAUAAAcontaining elements between Li and L3, along with the known roles of AAUAAA and GU/U-rich sequences in poly(A) site processing, suggest that the Li AAUAAA core element is deficient in its ability to cooperate with a GU/U-rich element in the formation of a stable cleavage complex. One possible explanation for this proposed lack of cooperation is that differences in spacing between AAUAAA and the GU/U-rich element in the poly(A) sites affect the stability of complex formation. In the L1.1.3. and L3.1.3. poly(A) sites, the L3 GU/U-rich downstream element is 6 nucleotides closer to the Li AAUAAA than it is in the L3 poly(A) site. To decrease the effect of altered spacing on processing efficiency, we extended the spacing between the Li AAUAAA and the L3 GU/U-rich region in both L1.1.3. and L3.1.3. by 4 nucleotides. In both cases, extending the spacing between the AAUAAA and the GU/U-rich elements did not have a statistically significant effect on processing efficiency (data not shown). Data suggest that the context of the Li AAUAAA is such that it is inherently limited in its capacity to form a CStF-stabilized preprocessing complex in collaboration with a downstream GU/U-rich element in comparison to the L3 AAUAAA core element. The final and most dramatic observation made with the chimeric L1-L3 constructs was seen when the upstream regions of Li and L3 were exchanged. Replacing the upstream region of L1.1.1. with that of L3 to generate L3.1.1. increased CAT product formation by approximately fivefold (Fig. 2B). Replacing the upstream region of L3.3.3. with that of Li to generate

4687

DETERMINANTS OF POLY(A) SITE STRENGTH

VOL. 14, 1994

B

A

.-I

a

60t

Z-

-

a 0

401

Z%

a-a 201

I

AI v

20l ~ I--

Ir: V: W:i

-4, 'If

W:

Ir:

V-

CL

j

x W:

T-

W:

3

3

CL

j

coi v At n .

CL

co C

X fi

3 3 FIG. 3. Effect of replacing sequences upstream of the L1 and L3 poly(A) sites with nonspecific sequences on poly(A) site strength. (A) Nonspecific sequence was used to replace the upstream sequence present in the Li poly(A) site. Lsp.1.1.(+) and Lsp.1.1.(-) contain opposite orientations of the nonspecific sequence element. Quantification of CAT activity was performed after transfection of HeLa cells with pML.SIS.CAT plasmids containing the indicated poly(A) site as previously described. (B) Replacement of the L3 upstream element with nonspecific elements as previously described.

L1.3.3. reduced the production of CAT by 75%. The exchange of these elements indicates either that the Li upstream region inhibits processing in the L1.3.3. construct or that the L3 upstream sequence is acting to facilitate processing in the L3.1.1. construct. The mRNA half-life measurements presented in Fig. 1 indicate that the Li transcripts were more stable than their L3 counterparts. Because the effect of the L3 elements is to increase steady-state levels of mRNA, we feel that the observed differences in steady-state mRNA levels of the chimeric constructs are not the result of changes in RNA stability. The U upstream region facilitates processing at a poly(A) site. To determine whether the L3 upstream sequence element acts to enhance processing or the Li upstream element acts to inhibit processing, we replaced these elements with two different random sequences (see Materials and Methods). If the sequences upstream of the Li poly(A) site constitute an inhibitory element, then replacing them with random sequences will significantly increase Li poly(A) site strength. If, however, the sequences upstream of the L3 poly(A) site constitute an enhancing element, then replacing them with random sequences will significantly decrease L3 poly(A) site strength. The experiment depicted in Fig. 3 shows the effect on poly(A) site strength that replacing sequences upstream of the Li and the L3 poly(A) sites with two different random sequences had. Replacing sequences upstream of the Li poly(A) site with nonspecific sequences did not affect L1 processing efficiency (Fig. 3A). Replacing sequences upstream of the L3 poly(A) site with random sequences resulted in a decline in CAT expression to a level very similar to that seen with L1.3.3. (Fig. 3B). This experiment indicates that the sequence upstream of the Li poly(A) site does not repress processing; instead, data confirm the proposal that sequences upstream of the L3 poly(A) site act to enhance processing efficiency. An upstream U-rich sequence contributes to the function of the U upstream element. The 75-nucleotide sequence element present upstream of the L3 poly(A) site facilitates processing efficiency in vivo. Within this element is the 8-nucleotide sequence UUCUUUUU. Because of the similarity between

4688

PRESCOTT AND FALCK-PEDERSEN

A

MOL. CELL. BioL.

B rI

C

60

'S

401

= 60

I

J

m~I m

J

unprocessed

-

I-0

4

0 20

201 0

0 V.:

.I

-J

I

5' cleavage product

-J

FIG. 4. Effect of deleting the sequence UUCUUUUU from the L3 upstream efficiency element on enhancer activity. In vivo effect on CAT activity when the UUCUUUUU sequence is deleted from the L3 upstream element. (A) Sequence deleted from L3.1.1. as described in Materials and Methods. (B) Effect of mutating the UUCUUUUU sequence in the intact L3 poly(A) site. Quantification of CAT activity was performed after transfection of HeLa cells with pML.SIS.CAT plasmids containing the indicated poly(A) site. Reaction products were quantitated with a Molecular Dynamics PhosphorImager as previously described.

this sequence and those known to function downstream of poly(A) cleavage sites, we wanted to determine if it was involved in the function of the L3 upstream region. Using a PCR-directed method, we deleted the sequence TTC lTTIT (UUCUUUUU in the pre-mRNA) from the L3 upstream enhancing element of pL3.1.1. (replacing it with the NotI restriction endonuclease recognition sequence GCGGCCGC) to generate pL3.1..LAT. When the L3.I..AT poly(A) site was assayed for in vivo processing efficiency (using pML.SIS.CAT L3.1.I.AT), a loss in processing efficiency occurred (Fig. 4A). A similar observation was made when the U-rich element was deleted from the parent L3 poly(A) site (Fig. 4B). At the level of the in vivo CAT assays, the U-rich region is an important component of the L3 upstream sequence. Having established that the upstream U-rich element contributes to processing efficiency in vivo, we used the in vitro cleavage assay system (described in Materials and Methods and reference 48) to further characterize the function of this sequence in processing complex formation. We determined that the upstream L3 element could facilitate cleavage of the L3.I.1. pre-mRNA compared with cleavage of the L1.1.1. pre-mRNA in vitro (Fig. 5; compare lanes 2 and 4). We also demonstrated that when the T-rich region is deleted from the upstream element, a decrease in cleavage product is observed. By comparing the processing efficiency of L3..I.LAT with those of L3.1.1. and L1.1.1., we show that deleting this 8-nucleotide sequence element from the pre-mRNA eliminates the in vitro function of the 75-bp L3 upstream element. The upstream U-rich region quantitatively influences the proteins which bind to L3.i.1. pre-mRNA. Having identified a sequence which functionally impacts on 3' processing efficiency, we next wanted to assess how this sequence element affects the association of proteins on the pre-mRNA substrate. UV light cross-linking of a labeled precursor RNA to the available subset of substrate-bound proteins was one of the initial assays used to characterize specific proteins in association with a 3' processing substrate pre-mRNA (41, 68-70). The 64-kDa subunit of CStF was one of the polypeptides which

1 2 3 4 5 6 FIG. 5. Effect of deleting the sequence UUCUUUUU from the U3 upstream efficiency element on in vitro poly(A) site processing efficiency. RNA processing reactions were carried out under cleavage conditions with the LM.1.1. (lanes 1 and 2), L.1.1. (lanes 3 and 4), and L.1.1.AT (lanes 5 and 6) 32P-labeled pre-mRNAs either in the presence (even-numbered lanes) or the absence (odd-numbered lanes) of HeLa nuclear extract. Reaction products were electrophoresed on a 3.5% polyacrylamide-8 M urea gel. 5' cleavage products and unprocessed pre-mRNA are indicated.

consistently bound to a pre-mRNA substrate in an AAUAAAdependent manner. Binding of the 64-kDa polypeptide is also dependent on the presence of the downstream GU/U-rich region. A second set of proteins, the hnRNP C polypeptides, has been shown to bind to downstream GU/U-rich elements independent of AAUAAA. Presumably, both of these proteinRNA interactions occur through direct interaction with the downstream GU/U-rich region. We used the UV cross-linking assay to determine if the upstream L3 U-rich element has a noticeable effect on the association of proteins on a premRNA substrate, particularly L3.1.1. In our assays, since the downstream region as well as the core hexanucleotide element originates from the Li poly(A) site, we included L1.1.1. pre-mRNA for comparison with L3.1.1. and L3.1.1.AT premRNAs (Fig. 6A, lanes 1, 4, and 6, respectively). An Li precursor containing a mutation in the AAUAAA hexanucleotide (to AAGAGA) (Fig. 6A, lane 2) functioned as a negative control for poly(A) site-specific protein cross-linking. A second type of precursor mRNA corresponding to precleaved L1.i.1., L3.1.1., and L3.1.1.AT (Fig. 6A, lanes 3, 5, and 7, respectively) was included to identify proteins which do not require the downstream GU/U-rich element for interaction with each of the indicated substrates. Using binding and UV cross-linking conditions which were slightly modified cleavage conditions (described in Materials and Methods), we identified a number of similarities and differences in the population of proteins labeled by each of the substrates (Fig. 6B). The pre-mRNAs used for Fig. 6A correspond to the input pre-mRNAs used for Fig. 6B. The most striking result from the UV cross-linking experiment was the increased quantity of each polypeptide labeled with the L3.1.1. precursor compared with L1.1.1. and L3.1.1.AT (Fig. 6B; compare lane 4 with lanes 1 and 6). A prominent doublet of approximately 41/43 kDa was labeled by each pre-mRNA substrate except the precleaved L3.1.1.AT.

DETERMINANTS OF POLY(A) SITE STRENGTH

VOL. 14, 1994

A

1 2 3

V.-._Li pmLl LI PC

B

1 2 3

4 5 6

C

7

pcL3.1.1

L3.1.1AT

5 6 7 8

9 10 11 12

U

0:.

_

L3.1.1 L3.1.1AT PC

pC

64kD{ 4 5 6 7

1---75-BOkD

CStF 64kD-[

L3.1.1 1 2 3 4

4689

p

IS

3, S% 43kDI hnRNP-C 41kD [

FIG. 6. UV-induced cross-linking of HeLa nuclear extract proteins to precursor mRNAs. (A) [32P]UMP-labeled pre-mRNAs of L1.1.1. (1) (lane 1), Li with an AAGAGA hexanucleotide mutation (pmLl) (lane 2), NcoI-precleaved L1.1.1. (Llpc) (lane 3), L3.1.1. (lane 4), NcoI-precleaved L3.1.1. (L3.1.1 pc) (lane 5), L3.1.1.AT (lane 6), and NcoI-precleaved L3.1.1.AT (lane 7). (B) Each of the pre-mRNA substrates in panel A was incubated under RNA processing cleavage conditions (described in Materials and Methods) and then underwent UV light irradiation (as described in Materials and Methods). After digestion with pancreatic RNase A, reaction products were electrophoresed on a 10.5% polyacrylamide-0.1% SDS Laemmli gel. The apparent molecular masses of the most prominently labeled proteins are indicated. The lane designations correspond to those in panel A. (C) UV cross-linking assays performed with L3.1.1. (lanes 1 to 4), precleaved L3.1.1. (lanes 5 to 8), and L3.1.1.AT (lanes 9 to 12) were processed as described before (lanes 1, 5, and 9) or immunoprecipitated (as described in Materials and Methods) with anti-64-kDa monoclonal hybridoma supernatant (lanes 2, 6, and 10), anti-RNP C monoclonal antibody (lanes 3, 7, and 11), or anti-ElA monoclonal antibody (lanes 4, 8, and 12) and electrophoresed on an SDS-9% polyacrylamide gel to demonstrate that the putative 64-kDa CStF subunit is recognized by the appropriate monoclonal antibody in a specific manner.

The apparent molecular weights of these proteins, and the observation that deleting the sequence UUCUUUUU from the pre-mRNA (Fig. 6B; compare lanes 4 and 6) disrupts their binding, suggested that they were the hnRNP C1 and C2 proteins (41, 57, 69). We have used the anti-RNP C monoclonal antibody 4F4 (7) in an immune precipitation assay to directly demonstrate that the 41/43-kDa polypeptides are the RNP C1 and C2 polypeptides (data not shown). Not surprisingly, the upstream U-rich region can directly interact with the RNP C proteins (Fig. 6B; compare lane 4 with lane 6 and lane 5 with lane 7). On the basis of these studies, we cannot assign a direct function in 3' processing to the association of the RNP C proteins with the upstream L3 element. In addition to the 41/43-kDa proteins, a protein of approximately 64 kDa was also radiolabeled by UV cross-linking to the LM.1.1., L3.1.1., and L3.1.1.AT pre-mRNAs (Fig. 6B, lanes 1, 4, and 6, respectively). The magnitude of the signal corresponding to this protein was higher with the L3.1.1. pre-mRNA substrate than with either the L1.1.1. pre-mRNA or the L3.1.1.AT pre-mRNA. The molecular weight of this protein suggested that it may be the 64-kDa subunit of CStF (41, 57, 69). Immunoprecipitation assays using the anti-64-kDa monoclonal antibody (59) (Fig. 6C, lanes 2, 6, and 10) indicate that this predominant signal is attributable to the CStF 64-kDa polypeptide. The binding of CStF to a pre-mRNA is indicative of stable cleavage complex formation on a pre-mRNA (22, 24, 25, 64); it is dependent on the presence of an intact AAUAAA

and is influenced by the presence of the downstream GU/Urich region. Each of these characterizations of poly(A) site cleavage complex formation is consistent with the data we have obtained which indicates the L3 U-rich element functions to stabilize the formation of the poly(A) site cleavage complex. As seen in Fig. 6B and C with the precleaved substrate RNAs, the absence of the downstream GU/U element abolishes labeling of the 64-kDa polypeptide. This holds true for the L3.1.1. precleaved pre-mRNA, which contains the upstream U-rich element. In the context of these substrates, the upstream U-rich element is not able to functionally replace the downstream GU/U-rich sequence with respect to 64-kDa protein binding. We can also see in Fig. 6B, lane 2, that the mutant Li pre-mRNA does not bind the 64-kDa protein. In fact, with the exception of the RNP C polypeptides, the mutation of the AAUAAA hexanucleotide had a dramatic effect on all proteins which were cross-linked to the Li pre-mRNA. These proteins, which are between 43 and 64 kDa in molecular mass, are clearly influenced by both an intact hexanucleotide (Fig. 6B; compare lanes 1 and 2) and the presence of the upstream U-rich element. Overexposure of the anti-64-kDa immunoprecipitation (Fig. 6C, lanes 2, 6, and 7) indicates that some of these proteins were coprecipitated by the anti-64-kDa monoclonal antibody. This is in contrast to immune precipitation with monoclonal antibody 4F4 against RNP C (Fig. 6C, lanes 3, 7, and 11) or with a completely irrelevant monoclonal

4690

PRESCOTT AND FALCK-PEDERSEN

antibody against the ElA polypeptides of adenovirus (Fig. 6C, lanes 4, 8, and 12). These monoclonal antibodies precipitated a small amount of protein immediately below the 64-kDa protein but none of the other subunits. We are in the process of determining if the association of these proteins in an AAUAAA-dependent manner is unique to Li hexanucleotide core substrates. The presence of the upstream U-rich region in L3.1.1. also resulted in an enhanced signal corresponding to a third class of protein(s) which migrates to a position corresponding to a polypeptide with a molecular mass of 75 to 80 kDa. To our knowledge, this protein has not been identified in previous studies of this type. Because there was nearly a quantitative loss of this polypeptide in the U-deletion substrates and because this protein is labeled at relatively high levels with the L3.1.1. precleaved substrate (Fig. 6B and C), we feel that this protein directly interacts with the U-rich sequence in a specific manner. Results from the UV cross-linking experiment have allowed us to make several observations about the effect of the upstream U-rich element on proteins associating with a poly(A) site pre-mRNA substrate. The U-rich element directly interacts with the RNP C polypeptides but is not the only element that interacts with these proteins. The U-rich element is directly associated with increased binding of the 75- to 80-kDa protein, which is independent of a requirement for the downstream GU/U-rich element. Finally, in the full-length substrate L3.1.1., the U-rich element quantitatively increases the binding of several proteins to this precursor RNA, including the 64-kDa subunit of CStF. These observations complement the data in Fig. 5, where we observed enhanced processing efficiency, and are in agreement with observations made by Gilmartin et al. characterizing the HIV upstream element (22). In their characterization of the HIV upstream efficiency element, they demonstrated a correlation between the stability of the preprocessing complex, increased in vitro processing efficiency, and the presence of the upstream element of HIV. DISCUSSION Poly(A) site use in the adenovirus MLTU presents examples both of alternative poly(A) site use in a complex transcription unit and of poly(A) sites which demonstrate striking differences in their function as substrates for the 3' processing machinery. In this study, we focused on identifying the sequence elements which make the Li poly(A) site a poor substrate for 3' processing complex formation in comparison to the L3 poly(A) site. We have shown that poly(A) site strength is the result of interplay between several elements, each of which can contribute to the overall efficiency of processing at a given poly(A) site, but no single element is sufficient to produce either maximal or minimal processing efficiency. Again, we can make a comparison between the 5' and 3' ends of a polymerase II transcription unit. The strategy of using multiple elements to create a strong or a weak poly(A) site is similar to the strategy employed to vary promoter strength at the 5' end of the transcription unit. Multiple sequence elements regulate poly(A) site strength. While much is known about the factors involved in the cleavage and polyadenylation reaction, there is still very little known about the determinants which regulate poly(A) site strength. Although the AAUAAA hexanucleotide is invariant in the vast majority of poly(A) sites, the sequences flanking this element, including the GU/U-rich element, are highly variable. Variability in the primary sequence of the GU/U-rich element, along with its known role in the formation of a stable cleavage

MOL. CELL. BIOL.

complex, has led to the hypothesis that poly(A) site strength may be determined through variations in the sequence of the GUIU-rich element. In support of this hypothesis, mutations in the downstream GU/U-rich elements of two poly(A) sites, SV40 early and adenovirus E2A, have been shown to affect poly(A) site processing efficiency (26, 37). Studies using the L3 core element with different downstream elements show that poly(A) strength can vary in a manner dependent on the downstream GU/U-rich region (17, 64). While the downstream GU/U-rich elements are certainly required for both Li and L3 poly(A) site utilization (27, 51), we have shown that the Li and L3 downstream elements are functionally equivalent, indicating that alterations in the GU/U-rich region are not responsible for the observed 20-fold difference in processing efficiency between these two sites. This study has identified two distinct elements which contribute to the difference in Li and L3 processing efficiency. (i) The core element of the Li poly(A) site is ineffective in comparison to the core element of L3, and (ii) the sequences upstream of the L3 poly(A) site act to facilitate 3' processing. Both core elements contain the AAUAAA consensus hexanucleotide but differ in their respective flanking sequences (Fig. 2A). The result of distinct sequences flanking the AAUAAA hexanucleotide is a significant difference in processing efficiency, with the Li core element behaving as a weak element compared with the L3 core element. The simplest explanation for this difference is that the Li core element is unable to bind CPSF as efficiently as the L3 core element. This explanation seems unlikely, since previous work has shown that a precleaved Li pre-mRNA can be adenylated in vitro as efficiently as a precleaved L3 pre-mRNA (48). The adenylation of a precleaved substrate pre-mRNA requires both CPSF and the AAUAAA hexanucleotide (36, 67). These studies indicate that the initial binding of CPSF to the Li AAUAAA sequence is not responsible for the weak function of the Li core element. A second explanation for this inhibitory effect is that the Li core element is unable to cooperate with a functional GU/Urich element to direct the formation of a stable cleavage complex on pre-mRNA. Data presented here and in our previous study support the latter explanation, in which poly(A) site strength is determined by the ability of individual elements to interact with the core element to stabilize the CPSF-CStF interaction. According to this model, differences in either the core element or the downstream GU/U-rich element would have the potential to impact on processing efficiency. While sequences upstream of these two poly(A) sites are not required for basal poly(A) site utilization, exchanging these sequence elements had a dramatic effect on Li poly(A) site processing efficiency. The sequences upstream of the Li AAUAAA do not contribute to poly(A) site strength in either a positive or a negative manner. This observation is in agreement with observations made by DeZazzo and Imperiale (15) that deletion of upstream elements had no apparent effect on processing of the Li poly(A) site when it was assayed as the only functional poly(A) site in a transcription unit. In contrast to the upstream Li sequences, the sequences upstream of the L3 AAUAAA facilitate processing of the L3 poly(A) site as well as the Li poly(A) site in the chimeric construct. Unlike the AAUAAA hexanucleotide or the downstream GU/U-rich region, the upstream element is not required for basal poly(A) site use. Instead, it is able to stimulate basal site use as defined by the individual core and GU/U-rich elements. Our observation that the L3 poly(A) site contains an upstream efficiency element adds L3 to a growing list of poly(A) sites functioning in mammalian cells that contain upstream elements. These include poly(A) sites from SV40 (4), hepatitis B viruses (46,

VOL. 14, 1994

47), and HIV (5, 6, 16, 62). Functional sequences identified in these upstream enhancing elements are as follows: SV40 late, UGUGAAAUUUGUGAUGCUAU; hepatitis B virus PS1B, AAAUUAUUUGUAUUA; HIV, CAGCUGCUULUUMGC CUGU; and adenovirus L3, UUCUUUUU. While there is no clear consensus sequence seen in the upstream enhancing elements, they all contain a U (or AU)rich sequence. From our experiments, the U-rich element in the context of the Li core element causes a general enhancement of protein binding to the pre-mRNA substrate. The binding proteins correspond to the same population of proteins which lose their affinity for the pre-mRNA when a mutation in the AAUAAA hexanucleotide is present along with the hnRNP C1 and C2 proteins. In spite of our observation that precleaved L3.1.1. does not demonstrate cross-linking to the 64-kDa subunit of CStF, it is tempting to suggest that upstream U-rich elements are functionally redundant with downstream GU/U-rich elements. In a simplistic attempt to test this possibility, we placed the downstream GU-rich element of the SV40 late poly(A) site upstream of the weak Li poly(A) site. In this position, the downstream SV40 downstream element was not able to enhance Li processing efficiency either in vitro or in vivo (unpublished data). Since the SV40 late downstream element contains only a GU-rich sequence element and not a U-rich sequence element, the possibility still remains that upstream and downstream U-rich elements are functionally redundant. On the basis of existing information and our observation that the U-rich element can associate strongly with the hnRNP C proteins, one could invoke a role for RNP C proteins in 3' processing. However, hnRNP C proteins are clearly not required for efficient poly(A) site processing in vitro when purified processing factors are used. Since we have worked only with nuclear extracts, we are not in a position to define a role (either positive or negative) for hnRNP C proteins in 3' processing. The association of the 75- to 80-kDa polypeptides with the upstream U-rich element is, in our opinion, the most unique effect of the upstream L3 U-rich element. Further investigation, particularly using purified 3' processing factors in reconstitution assays, should reveal a stronger functional relationship between the proteins interacting directly with the premRNA substrate and the efficiency with which a given substrate is cleaved. In fact, it should be mentioned that the enhanced levels of protein labeling observed in the presence of the U-rich element correlates with enhanced cleavage efficiency, but direct proof of a functional correlation between these proteins and cleavage efficiency will come from reconstitution assays using purified factors. Implications for the regulation of poly(A) site use in complex transcription units. The adenovirus MLTU is not the only transcription unit that has a dominant proximal weak poly(A) site. The Igju heavy chain transcription unit contains two poly(A) sites, a promoter proximal (Rs) site and a downstream (Im) site. Use of the upstream ,s site produces an mRNA encoding the secreted form of the IgM heavy chain, whereas use of the downstream Rm site leads to the production of a membrane-bound Ig. Plasma cells, much like cells in the early stages of an adenovirus infection, use the promoter proximal site exclusively. B cells, like cells in the later stages of an adenovirus infection, use both available poly(A) sites (reviewed in reference 32). In addition, the promoter proximal Rs site is approximately 10-fold weaker than the downstream m site, both in vivo (47) and in vitro (unpublished data). These similarities in poly(A) site strength and utilization seen with the IgM heavy chain and adenovirus MLTUs suggest a common mechanism of regulation.

DETERMINANTS OF POLY(A) SITE STRENGTH

4691

Any model used to explain how these transcription units regulate their poly(A) site usage must explain the observed preference for promoter proximal poly(A) site processing. This is particularly true with a poly(A) site such as Li which is clearly an inefficient substrate for 3' processing. All other things being equal, promoter proximal poly(A) sites have a processing advantage over more distal sites simply because of the 5' to 3' directionality of the transcription complex: promoter proximal sites are exposed to processing factors before distal sites are. In addition, one can imagine models in which the poly(A) site recognition machinery itself possesses 5' to 3' directionality. Such a scanning model has been proposed for both splice site recognition and for poly(A) site recognition (46). Work from Imperiale's laboratory has shown that sequence elements surrounding the Li poly(A) site impact on downstream processing efficiency both in vivo and in vitro (14, 15, 67a), giving support to a transcription-coupled scanning mechanism. If poly(A) sites are recognized in a directional manner, giving the promoter proximal poly(A) site a processing advantage over more distal sites, then variations in poly(A) site strength may be an important aspect of properly regulated use. In the adenovirus MLTU, which contains five poly(A) sites whose use is tightly regulated during the course of infection, such variations in poly(A) site strength appear to be crucial to properly regulated gene expression. ACKNOWLEDGMENTS We thank Gideon Dreyfuss for the generous gift of the anti-hnRNP C1 and C2 antibody, James Manley for the generous gift of the anti-64-kDa CStF subunit antibody, and Steven Shiff for the anti-ElA M73 antibody. We thank Monika Lusky and Yoshio Takagaki for helpful advice on immunoprecipitations. REFERENCES 1. Bienroth, S., E. Wahle, C. Suter-Crazzolara, and W. Keller. 1991. Purification of the cleavage and polyadenylation factor involved in the 3'-processing of messenger RNA precursors. J. Biol. Chem.

266:19768-19776. 2. Birnstiel, M. L., M. Busslinger, and K. Strub. 1985. Transcription termination and 3' processing: the end is in site! Cell 41:349-359. 3. Brasier, A. R., J. E. Tate, and J. F. Habener. 1989. Optimized use of the firefly luciferase assay as a reporter gene in mammalian cell lines. BioTechniques 7:1116-1122. 4. Carswell, S., and J. C. Alwine. 1989. Efficiency of utilization of the simian virus 40 late polyadenylation site: effects of upstream sequences. Mol. Cell. Biol. 9:4248-4258. 5. Cherrington, J., and D. Ganem. 1992. Regulation of polyadenylation in human immunodeficiency virus (HIV): contributions of promoter proximity and upstream sequences. EMBO J. 11:1513-1524. 6. Chemngton, J., R Russnak, and D. Ganem. 1992. Upstream sequences and cap proximity in the regulation of polyadenylation in ground squirrel hepatitis virus. J. Virol. 66:7589-7596. 7. Choi, Y. D., and G. Dreyfuss. 1984. Monoclonal antibody characterization of the C proteins of heterogeneous nuclear ribonucleoprotein complexes in vertebrate cells. J. Cell Biol. 99:1997-2004. 8. Chomczynski, P., and N. Sacchi. 1987. Single-step method of RNA isolation by acid guanidinium thiocyanate-phenol-chloroform extraction. Anal. Biochem. 162:156-159. 9. Christofori, G., and W. Keller. 1988. 3' cleavage and polyadenylation of mRNA precursors in vitro requires a poly(A) polymerase, a cleavage factor, and a snRNP. Cell 54:875-889. 10. Christofori, G., and W. Keller. 1989. Poly(A) polymerase purified from HeLa cell nuclear extract is required for both cleavage and polyadenylation of pre-mRNA in vitro. Mol. Cell. Biol. 9:193-203. 11. Citron, B., E. Falck-Pedersen, M. Salditt-Georgieff, and J. E. Darnell, Jr. 1985. Transcription termination occurs within a 1000 base pair region downstream of the mouse j3 maj-globin gene. Nucleic Acids Res. 12:8723-8731. 12. Conway, L., and M. Wickens. 1987. Analysis of mRNA 3'end

4692

13. 14.

15. 16.

17.

18. 19.

20. 21. 22. 23. 24. 25.

26. 27. 28.

29. 30.

31.

32. 33. 34.

35.

PRESCOTT AND FALCK-PEDERSEN

formation by modification interference: the only modifications which prevent processing lie in AAUAAA and the poly(A) site. EMBO J. 6:4177-4184. Darnell, J. E., Jr. 1982. Variety in the level of gene control in eukaryotic cells. Nature (London) 297:365-371. DeZazzo, J. D., E. Falck-Pedersen, and M. J. Imperiale. 1991. Sequences regulating temporal poly(A) site switching in the adenovirus major late transcription unit. Mol. Cell. Biol. 11:59775984. DeZazzo, J. D., and M. J. Imperiale. 1989. Sequences upstream of AAUAAA influence poly(A) site selection in a complex transcription unit. Mol. Cell. Biol. 9:4951-4961. DeZazzo, J. D., J. E. Kilpatrick, and M. J. Imperiale. 1991. Involvement of long terminal repeat U3 sequences overlapping the transcription control region in human immunodeficiency virus type 1 mRNA 3' end formation. Mol. Cell. Biol. 11:1624-1630. Edwalds-Gilbert, G., J. Prescott, and E. Falck-Pedersen. 1993. 3' RNA processing efficiency plays a primary role in generating termination-competent RNA polymerase II elongation complexes. Mol. Cell. Biol. 13:3472-3480. Falck-Pedersen, E., and J. Logan. 1989. Regulation of poly(A) site selection in adenovirus. J. Virol. 63:532-541. Feinberg, A. P., and B. Vogelstein. 1983. A technique for radiolabeling DNA restriction endonuclease fragments to high specific activity. Anal. Biochem. 132:6-13. Fitzgerald, M., and T. Shenk. 1981. The sequence 5'-AAUAAA-3' forms part of the recognition site for polyadenylation of late SV40 mRNAs. Cell 24:251-260. Gil, A., and N. J. Proudfoot. 1984. A sequence downstream of AAUAAA is required for rabbit P-globin mRNA 3'-end formation. Nature (London) 312:473-474. Gilmartin, G. M., E. S. Fleming, and J. Oetjen. 1993. Activation of HIV pre-mRNA 3' processing in vitro requires both an upstream element and TAR. EMBO J. 12:4419-4428. Gilmartin, G. M., M. A. McDevitt, and J. R. Nevins. 1988. Multiple factors are required for specific RNA cleavage at a poly(A) addition site. Genes Dev. 2:578-587. Gilmartin, G. M., and J. R. Nevins. 1989. An ordered pathway of assembly of components required for polyadenylation site recognition and processing. Genes & Dev. 3:2180-2189. Gilmartin, G. M., and J. R. Nevins. 1991. Molecular analyses of two poly(A) site-processing factors that determine the recognition and efficiency of cleavage of the pre-mRNA. Mol. Cell. Biol. 11:2432-2438. Green, T. L., and R. P. Hart. 1988. Mutations in poly(A) site downstream elements affect in vitro cleavage activity. Mol. Cell. Biol. 8:1839-1841. Hales, K. H., J. M. Birk, and M. J. Imperiale. 1988. Analysis of adenovirus type 2 Li RNA 3'-end formation in vivo and in vitro. J. Virol. 62:1464-1468. Hart, R. P., M. A. McDevitt, H. Ali, and J. R. Nevins. 1985. Definition of essential sequences and functional equivalence of elements downstream of the adenovirus E2A and the early simian virus 40 polyadenylation sites. Mol. Cell. Biol. 5:2975-2983. Hart, R. P., M. A. McDevitt, and J. R. Nevins. 1985. Poly(A) site cleavage in HeLa nuclear extract is dependent on downstream sequences. Cell 43:677-683. Huang, M. T. F., and C. M. Gorman. 1990. Intervening sequences increase efficiency of RNA 3' processing and accumulation of cytoplasmic RNA. Nucleic Acids Res. 18:937-947. Keller, W., S. Bienroth, K. M. Lang, and G. Christofori. 1991. Cleavage and polyadenylation factor CPF specifically interacts with the pre-mRNA 3' processing signal AAUAAA. EMBO J. 10:4241-4249. Leff, S. E., and M. G. Rosenfeld. 1986. Complex transcriptional units: diversity in gene expression by alternative RNA processing. Annu. Rev. Biochem. 55:1091-1117. Levitt, N., D. Briggs, A. Gill, and N. J. Proudfoot. 1989. Definition of an efficient synthetic poly(A) site. Genes Dev. 3:1019-1025. Manley, J. L. 1983. Accurate and specific polyadenylation of mRNA precursors in a soluble whole-cell lysate. Cell 33:595-605. Manley, J. L. 1988. Polyadenylation of mRNA precursors. Biochim. Biophys. Acta 950:1-12.

MOL. CELL. BIOL. 36. McDevitt, M. A., G. M. Gilmartin, W. H. Reeves, and J. R. Nevins. 1988. Multiple factors are required for poly(A) addition to a mRNA 3' end. Genes Dev. 2:588-597. 37. McDevitt, M. A., R. P. Hart, W. W. Wong, and J. R. Nevins. 1986. Sequences capable of restoring poly(A) site function define two distinct downstream elements. EMBO J. 5:2907-2913. 38. McDevitt, M. A., M. J. Imperiale, H. Ali, and J. R. Nevins. 1984. Requirement of a downstream sequence for generation of a poly(A) addition site. Cell 37:993-999. 39. McLauchlan, J., D. Gaffney, J. L. Whitton, and J. B. Clements. 1985. The consensus sequence YGTGTTYY located downstream from the AATAAA signal is required for efficient formation of mRNA 3' termini. Nucleic Acids Res. 13:1347-1365. 40. Montell, C., E. F. Fisher, M. H. Caruthers, and A. J. Berk. 1983. Inhibition of RNA cleavage but not polyadenylation by a point mutation in mRNA 3' consensus sequence AAUAAA. Nature (London) 305:600-605. 41. Moore, C. L., J. Chen, and J. Whoriskey. 1988. Two proteins crosslinked to RNA containing the adenovirus poly(A) site require the AAUAAA sequence for binding. EMBO J. 7:3159-3169. 42. Moore, C. L., and P. A. Sharp. 1984. Site-specific polyadenylation in a cell-free reaction. Cell 36:581-591. 43. Moore, C. L., and P. A. Sharp. 1985. Accurate cleavage and polyadenylation of exogenous RNA substrate. Cell 41:845-855. 44. Moore, C. L., H. Skolnick-David, and P. A. Sharp. 1986. Analysis of RNA cleavage at the adenovirus-2 L3 polyadenylation site. EMBO J. 5:1929-1938. 45. Nevins, J. R., and S. Chen-Kiang. 1981. Processing of adenovirus nuclear RNA to mRNA. Adv. Virus Res. 26:1-35. 46. Niwa, M., C. C. MacDonald, and S. M. Berget. 1992. Are vertebrate exons scanned during splice-site selection? Nature (London) 360:277-280. 47. Peterson, M. L., and R. P. Perry. 1989. The regulated production of p., and p,. mRNA is dependent on the relative efficiencies of Rs poly(A) site usage and the Cp.4-to-M1 splice. Mol. Cell. Biol. 9:726-738. 47a.Prescott, J. Unpublished data. 48. Prescott, J. C., and E. Falck-Pedersen. 1992. Varied poly(A) site efficiency in the adenovirus major late transcription unit. J. Biol. Chem. 267:8175-8181. 49. Russnak, R., and D. Ganem. 1990. Sequences 5' to the polyadenylation signal mediate differential poly(A) site use in hepatitis B viruses. Genes Dev. 4:764-776. 50. Russnak, R. H. 1991. Regulation of polyadenylation in hepatitis B viruses: stimulation by the upstream activating signal PS1 is orientation-dependent, distance-dependent, and additive. Nucleic Acids Res. 19:6449-6456. 51. Ryner, L. C., Y. Takagaki, and J. L. Manley. 1989. Sequences downstream of AAUAAA signals affect pre-mRNA cleavage and polyadenylation in vitro both directly and indirectly. Mol. Cell. Biol. 9:1759-1771. 52. Sadofsky, M., and J. C. Alwine. 1984. Sequences on the 3' side of hexanucleotide AAUAAA affect efficiency of cleavage at the polyadenylation site. Mol. Cell. Biol. 4:1460-1468. 53. Sadofsky, M., S. Connelly, J. L. Manley, and J. C. Alwine. 1985. Identification of a sequence element on the 3' side of AAUAAA which is necessary for simian virus 40 late mRNA 3'-end processing. Mol. Cell. Biol. 5:2713-2719. 54. Sambrook, J., E. F. Fritsch, and T. Maniatis. 1989. Molecular cloning: a laboratory manual, 2nd ed. Cold Spring Harbor Laboratory, Cold Spring Harbor, N.Y. 55. Schek, N., C. Cook, and J. C. Alwine. 1992. Definition of the upstream efficiency element of the simian virus 40 late polyadenylation signal by using in vitro analysis. Mol. Cell. Biol. 12:53865393. 56. Sheets, M. D., S. C. Ogg, and M. P. Wickens. 1990. Point mutations in AAUAAA and the poly(A) addition site: effects on the accuracy and efficiency of cleavage and polyadenylation in vitro. Nucleic Acids Res. 18:5799-5805. 57. Swanson, M. S., and G. Dreyfuss. 1988. Classification and purification of proteins of heterogeneous nuclear ribonucleoprotein particles by RNA-binding specificities. Mol. Cell. Biol. 8:22372241.

VOL. 14, 1994 58. Takagaki, Y., C. C. MacDonald, T. Shenk, and J. L. Manley. 1992. The human 64-kDa polyadenylation factor contains a ribonucleoprotein-type RNA binding domain and unusual auxiliary motifs. Proc. Natl. Acad. Sci. USA 89:1403-1407. 59. Takagaki, Y., J. L. Manley, C. C. MacDonald, J. Wilusz, and T. Shenk. 1990. A multisubunit factor, CstF, is required for polyadenylation of mammalian pre-mRNAs. Genes Dev. 4:2112-2120. 60. Takagaki, Y., L. C. Ryner, and J. L. Manley. 1988. Separation and characterization of a poly(A) polymerase and a cleavage/specificity factor required for pre-mRNA polyadenylation. Cell 52:731-742. 61. Takagaki, Y., L. C. Ryner, and J. L. Manley. 1989. Four factors are required for 3'-end cleavage of pre-mRNAs. Genes Dev. 3:17111724. 62. Valsamakis, A., S. Zeichner, S. Carswell, and J. C. Alwine. 1991. The human immunodeficiency virus type 1 polyadenylation signal: a 3' long terminal repeat element upstream of the AAUAAA necessary for efficient polyadenylation. Proc. Natl. Acad. Sci. USA 88:2108-2112. 63. Wahle, E., and W. Keller. 1992. The biochemistry of 3'-end cleavage and polyadenylation of messenger RNA precursors. Annu. Rev. Biochem. 61:419-440. 64. Weiss, E. A., G. M. Gilmartin, and J. R. Nevins. 1991. Poly(A) site efficiency reflects the stability of complex formation involving the downstream element. EMBO J. 10:215-219. 65. Wickens, M. 1990. How the messenger got its tail: addition of poly(A) in the nucleus. Trends Biochem. Sci. 15:277-281. 66. Wickens, M., and P. Stephenson. 1984. Role of the conserved AAUAAA sequence: four AAUAAA point mutants prevent

DETERMINANTS OF POLY(A) SITE STRENGTH

4693

messenger RNA 3' end formation. Science 226:1045-1051. 67. Wigley, P. L., M. D. Sheets, D. A. Zarkower, M. E. Whitmer, and M. Wickens. 1990. Polyadenylation of mRNA: minimal substrates and a requirement for the 2' hydroxyl of the U in AAUAAA. Mol. Cell. Biol. 10:1705-1713. 67a.Wilson-Gunn, S. I., J. E. Kilpatrick, and M. J. Imperiale. 1992. Regulated adenovirus mRNA 3'-end formation in a coupled in vitro transcription-processing system. J. Virol. 66:5418-5424. 68. Wilusz, J., D. I. Feig, and T. Shenk. 1988. The C proteins of heterogeneous nuclear ribonucleoprotein complexes interact with RNA sequences downstream of polyadenylation cleavage sites. Mol. Cell. Biol. 8:4477-4483. 69. Wilusz, J., and T. Shenk. 1990. A uridylate tract mediates efficient heterogeneous nuclear ribonucleoprotein C protein-RNA crosslinking and functionally substitutes for the downstream element of the polyadenylation signal. Mol. Cell. Biol. 10:6397-6407. 70. Wilusz, J., T. Shenk, Y. Takagaki, and J. L. Manley. 1990. A multicomponent complex is required for the AAUAAA-dependent cross-linking of a 64-kilodalton protein to polyadenylation substrates. Mol. Cell. Biol. 10:1244-1248. 71. Woychik, R P., R H. Lyons, L. Post, and F. M. Rottman. 1984. Requirement for the 3' flanking region of the bovine growth hormone for accurate polyadenylation. Proc. Natl. Acad. Sci. USA 81:3944-3948. 72. Zawel, L., and D. Reinberg. 1993. Initiation of transcription by RNA polymerase II: a multistep process. Prog. Nucleic Acid Res. Mol. Biol. 44:67-108.