Characterization of nuclear proteins that bind the EFII enhancer sequence in the Rous sarcoma virus long terminal repeat.

0022-538X/92/116338-15$02.00/0

Copyright © 1992, AmericanSociety for Microbiology

Characterization of Nuclear Proteins That Bind the

EFII

Enhancer

Sequence

in

the

Rous

Sarcoma

Virus

Long

Terminal

Repeat

ROSALIE C. SEARS1ANDLINDA

SEALY',2*

Departmentof Cell

Biology'

andDepartmentof Molecular

Physiology

andBiophysics,2

Vanderbilt

University

School

of

Medicine, Nashville,

Tennessee 37232

Received22April 1992/Accepted 23 July1992

The EFII cis element is a 38-bp sequence at the 5' end of the Rous sarcoma virus long terminal repeat, extending from nucleotides -229to-192(with respecttothe viraltranscriptionstartsite),which isrecognized

bysequence-specific DNA-binding proteinsinavian fibroblast nuclearextracts(L.

Sealy

and R.Chalkley,Mol.

Cell. Biol. 7:787-798, 1987). Wedemonstrate thatmultiple copies of the EFII cis element stronglyactivate transcription ofa reporter gene invivo. We correlate the region of the EFII cis element which activates transcriptioninvivo with thein vitro binding siteforthree nuclearfactors, EFIIa, EFIIb, and EFIIc.The

sequencemotifrecognized byEFlIa,-b, and-cis also found inconsensusbindingsites for members ofa

rapidly

growing

family

of transcription factors related to theCCAAT/enhancer-binding protein (C/EBP).

EFTIa,

-b,

and -care present in fibroblast andepithelial cell lines from various species but aremuch less abundant in differentiated rat liver and kidney cells. The EFlIa bindingactivityis

particularly

abundantin anavianB-cell lymphoma line. Asjudged from molecular weight analysis, cell type distribution,and sequence recognition properties, the EFII factors under study appear todiffer frommostof the

previously

describedC/EBP-related factors and thus may expand the

diversity

oftheC/EBP

family.

The information accumulated about the regulation of gene expression atthe level oftranscription initiation in eukary-otic cells has revealedanimpressive degreeofcomplexity.

Among the cis-acting DNA sequences that regulate

tran-scription, enhancersare nowknown tobe mosaics of

mul-tiplesequencemotifs. Each of these sequence motifs

repre-sents a discrete binding site for one or more trans-acting

factors. All of these cis-acting elements and regulatory

proteinfactorsworkinginconcertspecify appropriatelevels

oftranscription (31, 44, 49, 57). Manyviruses have evolved

to efficiently utilizethese hostcell mechanisms topromote

high-level expression of theirown genomes. For example,

the avian retrovirus Rous sarcoma virus (RSV) contains strong cis-acting enhancer sequences in the long terminal repeat(LTR)regions flanking either end of theproviralDNA

(13, 14, 23, 25,38,42, 52, 69,71).The potenttranscriptional

activity of the RSV LTR enhancer makes it an attractive

model system forstudyingthe moleculareventsinvolved in

theregulationoftranscriptioninitiation.Thus,wehave been

identifyingandcharacterizingnuclear factorswhich bind in

a sequence-specific manner to the RSV LTR enhancer.

Althoughthetypicaltargetcells for transformationbyRSV

and its relative, Rous-associated

virus-i,

are avian

fibro-blasts and lymphoid cells, respectively, the RSV LTR

en-hancer is activeinawidevariety of vertebrate cell types (23, 38,71).This suggests that many of thetrans-actingproteins which mediate the LTR enhanceractivity are common host celltranscription factors.

The major transcriptional control regions for RSV have been localized by deletion mutagenesis (13, 25, 38, 42, 52) and enhancer trap experiments (71) to the U3 region of the RSVLTR, extending from the 5' end of the LTR at position

-229(relativetothetranscription start site) to position -54.

* Correspondingauthor.

We have so far described three nuclear factors, enhancer factors I, II, and III (EFI, EFII, and EFIII, respectively),

which recognize specific nucleotide sequences within the RSV LTR enhancer diagrammed in Fig. 1 (4, 5, 17, 64).

Furtheranalysis of the EFIproteinfactor hasdemonstrated that it specifically recognizestwo inverted CCAAT motifs

(17). The EFIII factor recognizes two sites (4, 5) which

contain a commonsequencemotif knownastheCArGbox

(48, 66). By its

sequence-specific

DNA-binding properties

and antibody recognition, EFIII has been shownto

repre-sentthe avianhomologtotheserumresponsefactor(4, 5).

Wenow reportthe further characterization ofanumber of

protein factors present in chickenembryofibroblast (CEF)

nuclei which bind to the EFII region of the RSV LTR enhancer.

Other investigators have also observed nuclear proteins

which bindspecificallyinvitrotothe EFIIregionof the RSV

LTR(22, 61)andtorelated butnotidentical sequences in the

Rous-associated virus-2 LTR(59, 60). However, when this

workwasbegun, the transcriptional relevance of the EFII

cis element (RSVLTRsequences from -229to -192) had

not been directly addressed. Mutagenesis experiments had

demonstratedthat removal of the EFII sequencesaspartof

larger deletions crippledenhancerfunction(13, 42, 52),and smaller deletions demonstrated thatatleasteightnucleotides

(-201to -208)within theEFII ciselementarerequiredfor

maximal enhancer activity (25). We present evidence here that an isolated EFII cis element, when multimerized,

strongly enhances transcription invivo. This finding is in

agreement with recent experiments performed in GC rat

pituitarytumorcells, which show that the EFIIciselement

(RSV LTR sequences -231 to -193), linked to aminimal

heterologous promoter, is transcriptionally active, and its

activityis increasedbyexpressionofaconstitutive,calcium/

calmodulin-independentmutantof type II

calcium/calmodu-lin-dependent protein kinase (32). We further demonstrate

6338

on November 9, 2019 by guest

http://jvi.asm.org/

I RSVLTR

-229 -54

-229 -192 -175 -150 -146 -121 -112 -6782 -57

_JL ''_JL1__ L_ _J

EFII EFIII EFI EFIII EFI

FIG. 1. Schematic representation of the RSV LTR enhancer. Bindingsites forthree nuclearfactors, designatedEFI, EFII, and EFIII, are shown. The EFI factor has been foundto specifically recognizeinverted CCAAT motifspresentwithin eachof itsbinding

sites (17, 64).The EFIIbindingsite has beenmapped byDNase I footprintingat the 5' end of the RSV LTR(64).TheEFIIIfactoris suggestedto be the avianhomologto theserumresponsefactor,and

it has been foundtospecifically recognizetheCArG motifpresentin

both itsbindingsites(4, 5).

that theupstreammost oftwo near-directrepeatsequences

presentin the EFII cis element isresponsibleformediating strong transcriptional activityinvivo. At least three heat-stable protein complexes,referred to asEFIIa, EFIIb, and

EFIIc, are identified that specifically bind the EFII DNA

sequencein vitro. Thesethreecomplexesbindselectivelyto the upstream repeatwith high affinity. Thus,the

transcrip-tionally active sequences in the EFII cis element in vivo correspond to the

EFIla,

EFIIb, and EFIIc bindingsite in vitro.

All three EFII-binding factors appear to recognize the same nucleotides within the upstream repeat of the EFII DNA sequence. It is now well documented that many

transcription factors occur in families which share DNA sequence recognition properties (30). The nucleotide se-quence recognized by

EFIla,

-b, and -c is also found in

consensusbinding sites for members of arapidly growing family of transcription factors related to the previously

describedCCAAT/enhancer-binding protein (C/EBPa) (1, 7, 10, 15, 27, 29, 34, 39, 56, 58, 61). C/EBP-related family

members belong to the basic region-leucine zipper (bZIP)

class oftranscriptionfactors. bZIPtranscriptionfactorsare

characterized by a conserved DNA-binding domain,

con-taining clusters of basic amino acids, immediately adjacent

to a conserved dimerization domain containing a leucine residue at nearly every seventh position, known as the leucine zipper (36, 40, 41, 62). Three reported members of theC/EBPfamilyoftranscription factors, C/EBPa, IBF,and

Ig/EBP-1(C/EBP-y),have been shownto beable tobind the

EFII region of the RSV LTR in vitro (33, 58, 61). We

demonstrate, based on different sequence-specific DNA recognition properties and apparent-molecular-weight anal-yses, that EFIIa, -b, and -c are probably distinct from

C/EBPa and IBF. Moreover, expression of most of the

identifiedC/EBP-relatednuclearfactors isparticularly abun-dant in differentiatedliver, lung, andadipose cells (1, 3, 7, 10, 11, 15, 18, 56, 73),whileEFIIa, -b,and-caremuch less abundant indifferentiated rat liver andkidneycells than in

numerous fibroblast and epithelial cell lines from various species. However, Ig/EBP-1, EFIIa, andEFIIb havevery similarsequencerecognition characteristics,asjudgedfrom theirabilityto bindboth the EFIIsequenceandthe E site in themurine immunoglobulin heavy-chainenhancer withhigh affinity (58). We also notethat thereisalarge excessofthe EFIIa factor in an avian B-cell lymphoma line. Thus, it is possiblethat the EFIIaprotein playsarole inB-cell-specific gene expression, and further characterization of the three EFII-bindingfactorsmayexpandthediversityofthe C/EBP-relatedfamilyoftranscriptionfactors.

MATERIALS AND METHODS

Cell culture. CEFwere

prepared

from

10-day-old

embryos

(SPAFAS Inc., Preston, Conn.)

and maintained in medium

199

supplemented

with 10% tryptose phosphate broth, 5%

fetal calfserum,and 1% chickenserum.Thefollowingwere

also addedtothe medium: 25 U of

penicillin

G sodiumper

ml,

25 ,ugof

streptomycin

sulfateperml, and 0.14%sodium

bicarbonate. Dulbecco's modified

Eagle's

medium (DMEM)

(high glucose) supplemented

with 10% calfserumwas used

for all othercell

lines, including

A431

epidermal

carcinoma

cells, BALB/c-3T3 fibroblasts,

NIH 3T3

fibroblasts,

baby

hamster

kidney (BHK) cells,

Rat-1

fibroblasts,

avianB-cell

lymphoma

line

Bk3A,

and Cos

cells,

with the following

exceptions.

BHK cellsweremaintained ina 1:1 mixture of

DMEMand Ham's F12

medium;

DMEMforRat-1cellswas

supplemented

with 10% fetal calfserum; DMEM for Bk3A

cellswas

supplemented

with10%tryptose

phosphate

broth,

5% calf serum, and 1% heat-inactivated

(60°C,

30 min)

chickenserum. BHK cellswerea

gift

from

Roger

Chalkley

(Vanderbilt University), BALB/c-3T3

and NIH 3T3 cells

werea

gift

from Jack

Pledger (Vanderbilt

University), Rat-1

cellswere a

gift

from

Mike

Bishop (University

ofCalifornia

atSan

Francisco),

Cos and Bk3A cellswere a

gift

from Steve

Hann

(Vanderbilt University),

and A431 cells were a gift

from Graham

Carpenter (Vanderbilt

University).

Plasmid construction.

pSRA-CAT

was generated in this

laboratory

as described before

(4).

pe- CAT was

con-structedasfollows.Viral sequences in

pRSV-LTR

(64)from

theEcoRIsiteat-54totheBamHIsiteat+524(numbering

with reference tothe start site of

transcription

in the viral

genome)

werecloned into the PvuII siteof

pBR322

tocreate

pRSV-LTRe-.

pRSV-LTRe-

was

digested

withBstEII at

+103 in the viralsequence,treatedwithDNApolymeraseI

at

4°C

tocreateblunt endsasdescribed before

(64),

andthen

digested

withBamHIat

position

+524.

pSV2-CAT (24)

was

digested

with

HindIII,

treatedasabovetocreatebluntends,

and then

digested

withBamHI. The HindIII-BamHI

frag-ment

containing

the

chloramphenicol

acetyltransferase

(CAT)

genewastheninsertedinto

pRSV-LTRe-,

replacing

the

423-bp

BstEII-BamHI

fragment

to generate pe- CAT. The

HindIII

sitefrom theCATgene,now at

position

+103 in pe- CAT

(with

respect tothe start site for transcription), was not

regenerated.

The

p(EFII)nCAT

series of reporterconstructs were

cre-atedasfollows. The

following

double-stranded44-bp

oligo-nucleotide

containing

the

38-bp

EFIIcis element (-229 to

-192 in theRSV LTRare shownin

capital

letters):

5'cogagAATGTAGTCTTATGCAATACTCTTGTAGTCTTGCAACAO

3'cTTACATCATACGTTATGAGAACATCAGAACGTTTggg9ct5'

withAvaI

ends,

was

synthesized

andpreparedasdescribed below. This

44-bp

EFII

oligonucleotide

was treated with

polynucleotide

kinase and 1 mM ATP and

ligated

with T4 DNA

ligase (according

tothemanufacturer's

specifications)

to create head-to-tail multimers. EFII multimers were

treatedasaboveto createbluntends. pe-CATwas

digested

withAcclata

unique

site 235nucleotides upstream from the start of

transcription

and treated as above tocreate blunt ends. EFIImultimersweretheninsertedatthe AccIsiteto create the

p(EFII)nCAT

constructs.

p(EFII3'/5')2+CAT

was created as follows. The double-stranded 50-bp EFII

3'/5'

mutant

oligonucleotide (see

Fig.

3A for sequence),

synthesized

and

prepared

as described

below,

was treated with

polynucleotide

kinase and 1 mMATP and

purified by

polyacrylamide gel electrophoresis, electroelution,

and

on November 9, 2019 by guest

http://jvi.asm.org/

[image:2.612.89.284.82.136.2]

gen chromatography (Qiagen Inc., Chatsworth, Calif.) prior to incubation with T4 DNA ligase at 16°C. The ligated EFII

3'/5' mutant oligonucleotide was then inserted into the

vector pe- CAT, which had been previously digested with AccI and treated as above to generate blunt ends.

The reporter plasmids p(EFII5')4-CAT4 and p(EFII3') 2-CAT4 were created as follows. The following

double-stranded 20-bp and 18-bp EFII oligonucleotides, containing

the 5' EFII sequences from -226 to -213 (20 bp) or the 3' sequences from -205 to -194 (18 bp) in the RSV LTR

(capitalletters):

20bp:5'ccgagGTAGTCTTATGCAAc3'

3'cCATCAGAATACGTTgggct5' 18 bp: 5'cogagGTAGTCTTGCAAc3'

3'cCATCAGAACGTTgggct5'

withAvaIends, weresynthesized and prepared as described

below. The 20-bp and 18-bp oligonucleotides were then

treated withpolynucleotide kinase and 1 mM ATP prior to incubation with T4 DNA ligase (according to the manufac-turer'sspecifications).The ligated 20-bp oligonucleotide was then inserted into the BamHI site in the polylinker of a modified Bluescript M13+ (Stratagene, La Jolla, Calif.) plasmid created in this laboratory which lacks both XhoI and

SmaI sites, after treatment of both the ligated

oligonucleo-tide and themodified Bluescript vector as described above to create blunt ends. A clone containing four copies of the 20-bp oligonucleotide in the modified Bluescript vector,

designated pI120x4, was digested with BamHI and HindIII

to generate afragment of 120 bp containing the four ligated 20-bp oligonucleotides. This fragment was then inserted between the BamHI and HindIII sites of the polylinker at position -105, immediately upstream of the thymidine

ki-nase promoter in the plasmid pBLCAT4 (6), to create

p(EFII5')4-CAT4. The ligated 18-bp oligonucleotide was

treated to create blunt ends and inserted into the plasmid pBLCAT4, which had been previously digested with BamHI

(position -105)andtreated to create blunt ends, to produce

thereporterplasmid p(EFII3')2-CAT4.

TheplasmidpMLS-II, used to make the end-labeled probe

for footprinting, was generated asfollows. The 44-bp EFII

oligonucleotide described above was blunt-ended and

in-serted into the unique SmaI site in the polylinker of a modified Bluescript+ which had been previously digested

withXhoI,blunt-ended asdescribed above, and religated to

destroy the

XhoI

site. The AvaI site on either end of the insert was regenerated.

The inserts in all of the above plasmids were sequenced (45)

toconfirm insert number and orientation and the absence of mutations. We noted that the sequence of both copies of the EFII

3'/5'

mutantoligonucleotide in the p(EFII3'/5')2+CAT clone differed from the sequence given in Fig. 3A in that both

are missing the terminal G of the BamHI linker sequence

insertedto disrupt the 3' repeat.

Transfection and enzymatic assays. Transfections were

performed by the calcium phosphate coprecipitation

tech-nique as described by Graham and Van der Eb (26), with 1 ,ugoftest plasmid DNA, 2.5 ,ug ofpSV40-13-gal DNA (43), and 1.5 ,ug of pUC19 DNA per 10 ml of medium in 100-mm tissue culture dishes. Cells were plated the day before transfection and transfected when they were approximately 80 to 90% confluent. One hour prior to transfection, CEF

were refed with fresh medium 199 (supplemented as de-scribedabove) containing 10 mM HEPES

(N-2-hydroxyeth-ylpiperazine-N'-2-ethanesulfonic acid, pH 7.3). Cells were

exposed tothe CaPO4 DNA precipitate for 6 to 8 h, after

which they were washed once with Tris-glucose buffer containing 137 mM NaCl, 5 mM KCl, 5.6 mM dextrose, 25 mM Tris (pH 7.4), 25 U of penicillin G sodium per ml, 25

,ug

of streptomycin sulfate per ml, and 4.5 ,ug of phenol red sodium salt per ml and then refedwith fresh supplemented medium 199. Transfected cells were allowed to grow for an additional 48 h. Cells were then washed twice in cold (4°C)

phosphate-buffered saline (PBS) and incubated for 5min at 37°C in buffer containing 40 mM Tris (pH 7.5), 1 mM EDTA (pH 8), and 150 mM NaCl to loosen cells from the tissue culture dish.

All subsequent steps involving transfected cells were carried out at4°C. Cells were collected by centrifugation (5 min at 3,700 x g) and lysed by sonication, and cell extracts wereprepared as described by Boulden and Sealy (4). CAT assays with these cell extracts were performed by the procedure of Gorman et al. (24). CAT activity was calculated for each test plasmid by cutting out the acetylated and unreacted forms of [14C]chloramphenicol, quantitating the radioactivity in each by liquid scintillation counting, and determining the percent acetylation. Units of 3-galactosi-dase activity in the same cell extracts were assayed and calculated as described by Norton and Coffin (51). CAT activity was normalized by dividing percent acetylation by units of,B-galactosidase activity for each test plasmid.

Nuclear extracts. Nuclear extracts (0.5 M NaCl) were prepared from 14-day-old chicken embryos (CE) (SPAFAS) aspreviously described by Sealy and Chalkley (64), with the modifications described by Boulden and Sealy (4). The following changes were also made. The phosphatase inhibi-tors 3-glycerolphosphate (10 mM), sodium vanadate (100 FM), and sodium molybdate (10 mM) were added to all buffers. The protease inhibitor benzamidine (10 mM) re-placed leupeptin and pepstatin in buffer G. Homogenized embryos were filtered through cheesecloth prior to passage through Miracloth (Calbiochem Corp., La Jolla, Calif.). Nuclei were prepared and sequentially extracted by one washwith buffer A (5 ml perembryo), one extraction with buffer B (1 ml perembryo), and one extraction with buffer B containing 0.63 M NaCl (3 ml perembryo). This yielded an approximately 0.5 M NaCl extract because of the volume of the nuclear pellet. Following thisextraction, the supernatant wascollected byultracentrifugation in an SW27 rotor at 25 krpm for 60 min. The supernatant was dialyzed against 50 volumes ofdialysis buffer containing 10 mM HEPES (pH 8), 1 mM EDTA (pH 8), 50 mM NaCl, 50% glycerol, 7 mM

,B-mercaptoethanol,

1mM

phenylmethylsulfonyl fluoride,

10 mMbenzamidine, and the phosphatase inhibitors mentioned above.

Nuclear extracts (0.5 MNaCl) from cells grown in mono-layers, including CEF, A431, BALB/c-3T3, NIH 3T3, BHK, Cos, and Rat-1 cells, were prepared as described by Sealy and Chalkley (64), with the following modifications. The phosphatase inhibitors mentioned above were added to all buffers. Nuclei were prepared byonly one wash in buffer A. Thecrude nuclei were then resuspended directly in buffer B containing 0.53 M NaCl (yielding an approximately 0.5 M NaCl extract because of the volume of the nuclear pellet). Thesupernatant was collected at 16,000 x g and dialyzed as described above. Nuclear extracts (0.5 MNaCl) from Bk3A cells were prepared as described for the above monolayer cellsexcept that these suspension cells were pelleted at 750 x g for 10 min, washed twice with cold (4°C) PBS, and resuspended in buffer A to prepare nuclei. The 0.5 M NaCl extraction was performed with 1 ml per 109 cells. Heated nuclearextracts were prepared as described for the

on November 9, 2019 by guest

http://jvi.asm.org/

PROTEINS THAT BIND THE EFII ENHANCER IN THE RSV LTR 6341

layer cells with thefollowingmodifications. After the30-min

0.5 MNaClextraction,the mixturewasheated for5 min at 100°C, placed onice for 10 min, and subjected to

ultracen-trifugation in an SW56 rotor at 30 krpm for 30 min. The

supernatant was dialyzed as described above. Nuclear ex-tracts (0.5 MNaCl) prepared from thefollowing cells were gifts from the indicated individuals and were prepared as describedabove:H4,Tony Ip(VanderbiltUniversity); adult ratliver andkidney, Michelle Cissell (Vanderbilt

Universi-ty). Protein concentrations in extractsweredeterminedby

the method of Schaffner and Weissman (63), with bovine

serumalbumin asthe standard.

Radiolabeled andcompetitor DNAs. Oligonucleotideswere

synthesized(DiabetesResearchTraining CenterDNA core,

VanderbiltUniversity)andpurifiedasdescribed byBoulden

andSealy(4).Equal amountsof eacholigonucleotide,

mea-sured by the

A260,

and its complement were annealed in buffer containing 10 mM Tris (pH 8), 1 mMEDTA(pH8), and from 200 to 400 mM NaCl by placingthe DNAs in a 2-literboiling-waterbath,whichwasallowed to cool slowly

to room temperature overnight. Coding-strand nucleotide

sequences ofthe double-stranded oligonucleotides used for

theDNA-binding studiesaredisplayedinFig.3AandTable

1, except for the oligonucleotides labeled Core, DEI, IBF DNA, and IgH-E, which areshown below:

Core:5'agcttGGGCTGTGGAAAGGAGGGc3' 3'aCCCGACACCTTTCCTCCCgagct5'

DEI:5'togacTATGATTTGTAATGOGGGotga3' 3'agctgATACTAAAACATTACCCCgaget5' IBFDNA:5'CTGTTGGCTGCAATTGCGCCACCGCCACAG3'

3'GACAACCGACGTTAACGCGGTGGCGGTGTC5'

IgH-E:5' TA TGTTGAGTGAGTCAA3'

3'AATTCAAATTTTATAAAAATTTACTTAACTCGTTACAACTCAACTCAGTT5'

TheCoreoligonucleotide representssequencesfrom -48 to -65 from the collagen III promoter, which contains the simian virus 40 (SV40) core enhancer motif (50); linker sequences are indicated by lowercase letters. The DEI oligonucleotide represents sequences from -90 to -107 from the rat albumin promoter (9). IBF DNA represents sequences from the gag gene internal enhancer of RSV, designated OpvuII3' byKarnitz et al. (34). IgH-Erepresents sequences from position301 to position 350 in the murine immunoglobulin heavy-chain enhancer as defined by Ephrussi et al. (16). Since the 44-bp EFII oligonucleotide described above was unable to bind EFII DNA-binding proteins when radiolabeled and used in an electrophoretic mobility shift assay (EMSA) (unpublished results), a 50-bp EFII oligonucleotide containingRSV LTRsequencesfrom -229 to -192 withXhoI andBglIIlinkersonthe5'endand a SmaI linker on the 3' end was synthesized for use in EMSAs(seeFig. 3A).

Radiolabeled probes for EMSAs were prepared as fol-lows.Double-strandedoligonucleotides(200 ng)were radio-labeled with

[y,-32P]ATP

and T4 polynucleotide kinase ac-cording to the manufacturer's specifications. Full-length radiolabeledoligonucleotideswerethenpurifiedby

electro-phoresis on native 12% polyacrylamide gels containing lx TBE (89 mM Trisbase,89 mMboric acid, 2.5 mM EDTA

[pH8.5]),followedbyelectroelutionin lx TBEandethanol precipitation. Nonradiolabeled competitor oligonucleotides were also purified by polyacrylamide gel electrophoresis (PAGE)in1x TBE,andthe amount of DNArecoveredwas quantitated by A260 measurements. All oligonucleotides were resuspended in 10 mM Tris-100 mM NaCl-1 mM EDTA(pH 8).

Radiolabeled probes for footprinting were prepared as

follows.Theplasmid pMLS-II (see above)wasdigestedwith

eitherBamHI (tolabel thecoding strand)orHindIII(tolabel the noncoding strand) and end-labeled with

[a-32P]dATP,

[a-32P]dTTP,

(a-32P]dGTP, [a-32P]dCrP, and DNA

poly-merase I at4°C according to the manufacturer's

specifica-tions. End-labeled, linearized pMLS-II was then digested

with either HindIII (if end-labeled at a BamHI site) or

BamHI (if end-labeledat aHindIIIsite). The resulting

76-bp

end-labeled restrictionfragmentswerepurifiedbyPAGEas

describedabove.

EMSAs.Samplesof0.5 M NaCl nuclear extract,

contain-ingbetween 1 and 10 ,ugof protein, depending onthe cell

type analyzed, were mixed with 0.5 ng of radiolabeled

double-stranded oligonucleotide and 1.3 ,ug of poly(dI):

poly(dC), prepared as described previously (64), in the

presence or absence of nonradiolabeled, gel-purified

com-petitor DNAs. Binding reaction mixes were incubated at

room temperature for 30 min in a final volume of 20 ,ul,

containing7.5 mM HEPES(pH 8), 2.5 mM Tris(pH8), 0.8

mM EDTA (pH 8), 100 mM NaCl, 3.3 mM MgCl2, 5 mM

,-mercaptoethanol, and 20% glycerol. Sampleswere

ana-lyzed on native 4%polyacrylamidegelscontaining 1 xTGE (25 mM Tris, 190 mM glycine, and 1 mM EDTA [pH 8.5]) in

the absence oftracking dye. Gelsweresubjectedto

electro-phoresis inlx TGE for 50minat 200 V andsubsequently

dried for autoradiography. When appropriate, the

protein-DNA complexes were excised from the dried gel, and the

radioactivity in the complexwas quantitated by liquid

scin-tillation counting. Background counts were subtracted by

countingan equivalent section of the driedgel outside the

sample lanes. The relative binding affinityofprotein-DNA

complexes forspecific nonradiolabeled competitor

oligonu-cleotides was determined

by calculating

the percentage of

radiolabeled DNA bound ina specificcomplex in the

pres-ence of competitor DNA relative to that observed in the

absence ofcompetitorDNA, plottedversuslog (competitor

DNA). TheamountofcompetitorDNArequiredtoreduce

the formation ofa specificcomplex by 50%wascalculated

from this plot and divided by the amount of wild-type

competitor DNA

(representing

thesequence of the

radiola-beled DNA used in the

EMSA)

neededtoreducetheamount

of the same

complex by

50% to obtain relative binding

affinities.

OP/Cu footprinting. Samples of 0.5 M NaCl nuclear

ex-tractcontaining either 12 or4 ,ug of protein from CEF or

Bk3Acells,respectively,weremixed with100 to 200kcpm

(approximately

2ng)ofend-labeledprobe(see above)in the presence of 1.3 p.g of

poly(dI):poIy(dC)

and subjected to EMSA as described above withthefollowingmodification:

20-,ulbindingreactionmixeswereanalyzedon6%

polyacryl-amide gels containing lx TG (25 mM Trisbase, 190 mM

glycine [pH 8.5]).Thegelsweresubjectedtoelectrophoresis

in1xTG for 1.5hat200 V.Following

electrophoresis,

gels

werewashedoncewith50 mMTris(pH 8)for15 minatroom

temperature, placedin fresh50 mM

Tris,

and treated in situ

with

orthophenanthroline-copper

(OP/Cu) endonuclease as describedbyKuwabara and Sigman (37)for 6minatroom temperature. Free and bound complexes, visualized by

autoradiography

of thewetgelat4°C,wereexcised from the

gelandelectroeluted in 1xTBE.Cleavedend-labeled DNAs

were then ethanol

precipitated

with 5 ,ug of carrier tRNA.

Samples

containing equal

counts

(defined by

Cerenkov

radi-ation),

and the end-labeled

probes

chemically

cleaved at

purine

residues,

were

separated

on a 20%

polyacrylamide

sequencing gel

asdescribed

by

Maxam andGilbert

(45).

on November 9, 2019 by guest

http://jvi.asm.org/

Apparent-molecular-mass

analysis.

Samplesof 0.5MNaCl nuclear extract

containing

100 or 30 ,g of

protein

from

14-day

CE or Bk3A

cells, respectively,

were subjected to

fractionationon a sodiumdodecyl sulfate (SDS)-12%

poly-acrylamide gel,

transferred to an Immobilon-P membrane

(Millipore Corp., Bedford, Mass.),

and

eluted,

asdescribed

by

Faber and

Sealy

(17),

except that the Immobilon-P

membrane

corresponding

tothe

separating gel

was cutinto 17

slices,

each

approximately

3-mm

wide,

and the

proteins

in each slice were eluted into 40 ,l of cold

(4°C)

elution solution.

Samples (10

RI)

of each SDS-sized fraction were

analyzed by

EMSA. To determine the range of apparent molecularmassesof the

proteins

ineach

fraction,

individual

protein

standards were

separated

on the same SDS-12%

polyacrylamide gel,

transferred to an Immobilon-P

mem-brane,

andvisualized

by

amido black

staining

as described

by

Schaffner and Weissman

(63).

RESULTS

Specific

sequences within the EFII cis element activate

transcription invivo. We have

previously

demonstrated

by

DNase I

footprinting

thata

factor(s) (termed EFII)

in nuclear extracts from avian fibroblasts

specifically

binds to a38-bp

sequenceatthe 5' end of the LTR of the Schmidt

Ruppin-A

strain of

RSV, extending

from -229 to -192 nucleotides

upstreamof thestartsite for

transcription (64).

Toassessthe functional

significance

ofthis

region

of the LTRenhancer for

transcriptional activation,

weinsertedone ormore

copies

of theEFII

binding

site into the reporter

plasmid

pe-CAT.As shown in

Fig. 2A,

pe- CAT is similar to the

plasmid

pSRA-CAT,

whichwasdescribed

previously (4),

exceptthat sequencesin U3

responsible

for LTR enhancer

activity

have been

deleted, leaving just

54 nucleotides upstream of the start site for

transcription.

These 54

nucleotides,

which include a TATA box at

approximately position -30,

are sufficienttodirectaccuratebut very low levels of

transcrip-tion of the CAT gene in vivo

(20). Up

to six

copies

of the

EFII

oligonucleotide

were inserted into pe- CATat a

posi-tion

(-235)

comparable

to that found in the

wild-type

RSV LTR. The

resulting

constructs,

designated

p(EFII)nCAT,

were then tested for their

ability

to activate

transcription

in transient-transfection assays in CEF. As shown in

Fig.

2B,

multiple copies

of theEFIIcis

element,

in eitherthesense or antisense

orientation,

wereableto

strongly

activate

transcrip-tion of the CAT gene

compared

with the enhancer-minus

construct

(pe- CAT).

Six

copies

of the EFIIciselement in the sense orientation

[p(EFII)6+CAT]

were able toactivate

transcription approximately

40-fold.

Inspection

of the nucleotide sequence of the EFII cis

element

(Fig.

3A)

reveals that a

nearly

direct repeat of15

nucleotides is present in the LTR sequence from -227 to -213 and from -206 to

-194,

with the 3' repeat

lacking

an AT present in the 5' repeat. To determine whether the presenceofthese

repeated

sequences isimportantfor

tran-scriptional activation,

wecreateda mutantEFIIciselement in which both repeatsweredisrupted by 10-bpBamHIlinker sequences

(EFII 3'/5'

mutant; see Fig. 3A for sequence). Two

copies

of thismutantoligonucleotidewerealso inserted into the pe- CAT vector at -235. When the resulting construct,

p(EFII3'/5')2+CAT,

was tested in transient-transfection assays, the mutant EFIIcis element failed to

activate

transcription

above the level observed with pe-CAT

(Fig. 2C).

Thus,

it appears that one or both of the

repeated

sequences intheEFIIciselementarenecessary for

transcriptionalactivation, presumablyby actingas abinding

site(s)forone or moretrans-actingfactors.

In ordertodirectlytestthetranscriptional activity of each near-direct repeat, we analyzed two abbreviated EFII cis elements in which eitheronly the upstream 5' repeat, from -226 to -213, oronly the downstream 3' repeat, from -205

to -194,wasreiterated inareporter gene construct. Specif-ically, four copies of the 5' repeat or two copies of the 3' repeat (each separated by a 6-bp AvaI linker sequence) wereinserted immediately upstream of the thymidine kinase promoterand CAT gene in the vector pBLCAT4(6),creating p(EFII5')4-CAT4 and p(EFII3')2-CAT4, respectively.

In-terestingly, in transient-transfection assays,

p(EFII5')4-CAT4 activated transcription 36-fold relative to pBLCAT4, whilep(EFII3')2-CAT4wasunable to activate transcription above the basal activity of pBLCAT4 (Fig. 2C). Althoughwe

have not compared equal copy numbers because such con-structs were notavailable, it is unlikely that more copies of the3' repeat would leadtotransactivationequivalent to that with the 5' repeat when two copies are inactive. Ingeneral, transactivationby multimerized cis elements increases

incre-mentallywithincreasing copy number, as seenfor the

com-plete EFII cis element (Fig. 2B). Our data suggest that

transcriptionalactivationby the EFIIciselement is mediated

primarily throughthe 5' near-direct repeat sequence. Thus,

wewished to furthercharacterize theprotein factors which

specifically bind this element in vitro and, in particular,

determine whetherthey recognizethe upstream repeat.

Nuclearproteins whichbind totheEFIIciselementinvitro specifically recognize the upstream repeat. In our previous

study,weidentifiedtwoEFIIDNA-bindingfactors present

in0.3to0.5 M NaClextractsofquailfibroblast nuclei which

specificallybound the EFII ciselement contained ina286-bp

restriction fragment encompassing the entire RSV LTR

enhancer (64). In order to facilitate further analysis of

proteins recognizingthe EFII ciselement,werepeated this

experiment with a double-stranded 50-bp oligonucleotide,

referred to as EFII DNA (see Fig. 3A for coding-strand

sequence information), containing the 38-bp EFII cis

ele-mentwith linker sequencesonboth ends. As shown in Fig. 3D, lane 1, two major protein-DNA complexes, which we

refer to as EFIla and EFIIb, are present in anEMSA of radiolabeled EFII DNA following incubation with heated

(100°Cfor 5min)0.5 MNaCl nuclearextractpreparedfrom

14-day-old CE. A third minor protein-DNA complex of

slowerelectrophoretic mobility, designatedEFIlc,wasalso detected. Thetwofaster-migrating protein-DNA complexes

likelyarise from some abundant, nonspecific DNA-binding

proteins in the high-NaCl extract, since they were not

inhibited by the addition of a variety of nonradiolabeled

competitor DNAs (data not shown). Results identical to

thosepresentedfor the14-day-oldCE nuclearextractwere observed with both heated and unheated 0.5MNaCl nuclear

extractsfrom CEF(datanotshown).

To assess whetherbindingof EFIIa, EFIIb, or EFIIc in vitro would correlate withtranscriptionalactivation invivo,

weinvestigatedtherole of thetworepeated sequences in the

EFIIciselement in theformation of the three complexes. In addition to the EFII 3'/5' mutant described above, we

obtained two EFII mutantoligonucleotides in which either the 5' repeat (EFII 5' mutant) or the 3' repeat (EFII 3'

mutant) was disrupted by a 10-bpBamHI linker sequence

(Fig. 3A). Representative competition experiments with

thesemutantoligonucleotides and thewild-type EFIIDNA are shown in Fig. 3B. The decrease in relative binding

affinities of the EFIla, -b, and -c factors for each of the

on November 9, 2019 by guest

http://jvi.asm.org/

r-LTREnhancer-TrI

pSRA-CAT: pEBR322 RSVSgenonioe

PVLJII l-229) (489)

pe-CAT: pBR322

(EFII r)

Accl (-235)

) K

EcoRI l+X)

(-54) Lv

CAT

gene

R U5

-Hindlil <- 103)

r raP CAT

gene

U3 1IH|d5

EcoR l- l Hindlil

(-54)

L,>-

(+10O3)

B.

40-30

0

20-LUJ

< 10

-j

LuJ

0

38+4

16+6

d

~~~~1+

.w

go

e

e*

1

'

FIG. 2. Specificsequenceswithin the EFII cis element activatetranscriptioninvivo.(A) pSRA-CATisachimericconstructcontaining

aCATreportergenelinked tothe330-bpRSV LTR and 260bpof upstreamsequencenormally foundatthe 3' end of the RSV provirus

genome.TheRSVLTR enhancer extends from -229to-54 inpSRA-CAT.pe-CAT isanenhancer-minusconstructwhich lacks the RSV genomesequencesfrom the PvuII siteat-489totheEcoRI siteat-54. Thesequencesfrom -54 to +1 in bothplasmidsrepresentaminimal promoter(MP) containingaTATAboxatapproximately -30. (B) Uptosixcopies of the EFII cis elementwereinserted into the plasmid pe- CATatthe indicated AccIsite, creatingthep(EFII)nCATseries,in which1, 2,or6copiesof theEFIIciselementineither thesense

(+)orantisense(-)orientationwereinserted.Samples (1 ,ug)of eitherpSRA-CAT,pe-CAT,orp(EFII)nCATweretransiently transfected

into CEFalongwith 2.5 ptgofpSV40-3-gal, encoding ,-galactosidase, as aninternalcontrol. A representative CATassayis shown for the indicated reporterplasmids. CATactivitywascalculated and normalized asdescribed in Materialsand Methods. RelativeCAT activity,

indicating the foldinduction of normalizedCATactivityfor each reporterplasmid above that attained with the referenceplasmidpe-CAT, isgraphed. Data represent the averages ofat least three separate experiments + standard deviation (SD). (C) The additional reporter constructs p(EFII3'/5')2+CAT, pBLCAT4, p(EFII5')4-CAT4, and p(EFII3')2-CAT4 were also tested in transient transfections as

described forpanelB.p(EFII3'/5')2+CATcontains twocopiesof the EFII3'/5'mutantoligonucleotide (see Fig. 3A for sequence) in thesense

orientation insertedattheAccIsite ofpe-CAT.p(EFII5')4-CAT4contains fourcopiesof the 5' repeat(EFIIsequencesfrom-226to-213;

seeMaterials andMethods),andp(EFII3')2-CAT4containstwocopiesofthe 3'repeat(EFIIsequences -205to -194;seeMaterials and Methods);both inserts areinthe antisenseorientation,and bothwereplaced immediatelyupstreamofthe thymidine kinasepromoterand CATgenein thereporterplasmid pBLCAT4.Relative CATactivity,calculatedasdescribedforpanel B, withpe- CATas areference for p(EFII3'/5')2+CATandpBLCAT4as areference forp(EFII5')4-CAT4 and p(EFII3')2-CAT4, is graphed. Datarepresenttheaveragesof

atleast three separate experiments SD.

mutantoligonucleotides comparedwith theiraffinities for the

wild-typeEFII DNAsequencearepresentedinFig.3C. We observed that theEFIIa, -b,and-ccomplexescould bindto the EFII 3'mutant,which retains theupstream repeat,with

affinityequaltothatseenwith the EFIIDNA. In contrast, all

three EFIIDNA-binding complexesdemonstratedamarked reduction in relativebinding affinitytothe EFII 5' mutant; EFIIa and EFIIb had approximately 9-fold lower relative

binding affinity, and EFIIc binding affinity was reduced to approximately50-fold below thatseenwiththeEFIIDNA. When the EFII 3'/5' mutant was used as the competitor oligonucleotide, all three EFII factors demonstrated greater than 100-fold-lower relativeDNA-bindingaffinity. Thus, at leastoneintactrepeat mustbe present forEFIIa, -b, and-c complex formation, and all three factors have a higher

affinityfor the5' repeatcontainingtheadditionalAT

nucle-otides.

We alsoanalyzedtheabilityof the threeEFII complexes to directly bind each of the EFII mutant oligonucleotides (Fig. 3D). We observed that all three complexes formed equallywell with the radiolabeledEFIIDNA(Fig. 3D,lane 1) orthe radiolabeled EFII 3' mutant (Fig. 3D, lane 2). In contrast, EMSA with the radiolabeled EFII 5' mutant(Fig. 3D, lane 3) resulted in two less-prominent DNA-binding

activities that displayed proportionally slower migrationin the gel than EFIIa and EFIIb, and no EFIIc-likecomplex waspresent. There is evidence that thesameDNA-binding proteincanhave different mobilities inagelmatrix when it occupiesdifferentpositionsonoligonucleotidesof thesame length (35, 55). Therefore,it ispossible that thetwo DNA-A.

on November 9, 2019 by guest

http://jvi.asm.org/

[image:6.612.126.492.76.383.2]

A. EFIIDNA:

EFII 3'mutant: EFII5'mutant:

EFII3a/5'mutant:

B. CompetitorDNA:

-229 -192

tcgagatct AAT GTAG T CTTATG CAATACT CTTGTAG T CTTG CAACAccc

tcgagatctAAT GTAG T C TTATG CAATAC TCTTCGTTC CAccc

tcgagatctAAT G

TCCGGT

TACTCTTGTAG TCT TG

CAACAc

c c

tcgagatctAATGT CGGATCCG ATACTCTCTGTCGCAccc EFIIDNA

Molar Excess: 0 5

EFllc

-40-EFIlb

Ia-EFlla - *12

1 2

10 50 100 0 5 10 50 100 0 5 10 50 100 0

3 4 5 6 7 8 9 10 11 12 13 14 15 16

5'mutant 3'/5' mutant 5 10 50100

M.'

M

tA1

17 18 19 20

3'mutant

C.

D.

100 +*

>

o

Lo

EFIIc-*-

50-FI

_~40

b EFD

Ii

- A

Cr0*303

EFlIa-mu-'-

.4

r

FIG. 3. EFIIa,EFIIb, and EFIIc nuclear factors recognize the upstream repeat in the EFII cis element. (A) Coding-strand nucleotide sequences of double-stranded 50-bp oligonucleotides containing either the wild-type EFII cis element (EFII DNA) or the EFII cis element in which either the 3'repeat (EFII 3' mutant), the 5' repeat (EFII 5' mutant), or both (EFII 3'/5' mutant) have been disrupted by 10-bp BamHI linker sequences, indicated by open boxes. The 38-bp EFII cis element (representing RSV LTR sequences from -229 to -192) is indicated by capital letters. The near-direct repeat sequences present in thewild-typeEFII cis element are underlined. (B) EMSAs were performed as described in Materials and Methods. Briefly, 2,ugof heated 0.5 M NaCl nuclear extract from 14-day-old CE were incubated with 0.5 ng of 32P-labeled EFITDNA in the presence of 1.3 ,ug of poly(dI):poly(dC) and in the absence or presence of a 5- to 100-fold molar excess of nonradiolabeled competitor DNAs. The positions of the retarded EFIIa, EFIIb, and EFIIc protein-DNA complexes are shown. (C) The relative bindingaffinities of EFIIa, -b, and -c for each competitor oligonucleotide were calculated as described in Materials and Methods. The fold decrease in binding affinity relative to wild-type EFII DNA is graphed, and data represent the averages of at least three separate experiments ± SD. (D) EMSAs were performed with 2plgof heated0.5 M NaCl nuclear extract from14-day-oldCE in the presence of 1.3

,ugof poly(dI):poly(dC) and 0.5 ng of the following 32P-labeled EFII wild-type and mutant oligonucleotides: EFII DNA (lane 1), EFII 3' mutant (lane 2), EFII 5' mutant (lane 3), and EFII 3'/5' mutant (lane4). The unbound 32P-labeled EFII DNAs are indicated along with the retarded EFIIa,EFIIb, and EFIIc protein-DNA complexes.

binding activities in Fig. 3D, lane 3, represent EFIIa and

EFIIb, since bothcanbindwith loweraffinitytothe EFII 5'

mutant(Fig. 3B and C). On the other hand, the two

com-plexes inFig. 3D, lane 3, couldrepresentunrelatedproteins,

whicharenormally maskedby the presence of the EFIIa, -b, and -c complexes. EMSA with the radiolabeled EFII 3'/5'

mutant (Fig. 3D, lane4) didnotresult in any specific EFII

DNA-binding activity. The results of both thecompetition

anddirectbinding studies in Fig. 3 are highly consistent with

ouranalysis of thetransactivation properties of the EFIIcis

element invivo. The upstream repeat in the EFII cis element mediates both high-affinity recognition by the EFIIa, -b, and

-cfactors in vitro and strong transcriptional activation in vivo.

OP/CufootprintingofEFIIa,

EFIIb,

and EFlIc.Inorderto

further define the nucleotides within the EFII cis element which EFlIa, -b, and -c recognize for sequence-specific DNA binding, OP/Cu nuclease footprinting of each DNA-protein complex was performed. The EFIIa, -b, and -c

DNA-protein complexes were resolved by EMSA and

treated in situ with the chemical endonuclease OP/Cu.

Cleaved DNA was then isolated from each complex and

analyzed, along with the cleaved unbound DNA, on a

high-resolution sequencing gel (Fig. 4A). Interestingly, all

three complexes exhibited essentially identical footprints.

Strong protection was observed over 15 nucleotides from

-210to -224onthenoncodingstrand, with weaker

protec-tion of 9 nucleotides from -210 to -218 on the coding strand. The footprints on both strands were observed to

coincide closely but not exactly with the 5' repeat, as

illustrated inFig. 4B.Noprotectionwasobserved with any ofthecomplexesoverthe3' repeatoneitherstrand. These

findingsaresimilartoobservations byGoodwin, who found

by footprintinganalysis that erythroid nuclear proteins

pro-tected sequences from -224 to -215 in the RSV LTR enhancer(22).

Since theEFIIa, -b,and -ccomplexesfootprintthesame

on November 9, 2019 by guest

http://jvi.asm.org/

[image:7.612.104.515.78.393.2]

5'

EFII Fa bc

ti

IwF a

-a. m, ..

m;&-

-'W.Z.

81 70 61 52 45 39 34 29 25 22 19 k

EFII Fa bc

-218

/A

G

C

A A T A

-210

CODING

-210 /G

1T T

T

G

C

A T A A G A

C

T\~

-224

EFlic - e ?;

EFIlb

--EFlia

0

jo.MAk

WV

1 2 3 4 5 6 7 8 9 10 11

NON-CODING

B.

-229 5'

3-

3 -192

A GTAGTCTTATGCAATACTCT GTAGTCTTGCAA A TT CATCAGAATACGTTATGAGA CATCAGAACGT T

FIG. 4. EFIIa, EFIIb,and EFIIc protect thesamenucleotides within theEFII cis element.(A)EMSAswereperformedwith 2ng

ofa double-stranded end-labeled76-bprestrictionfragment probe, containingthe38-bpEFII ciselement,and 12,ugof CEF 0.5MNaCl nuclearextract.Following EMSA,thegelwastreated withOP/Cu endonuclease. TheEFIIa, -b,and-cprotein-DNA complexes and the freeprobewereexcised from thegel.Thecleaved,end-labeled DNA was recovered by electroelution and analyzed on a

high-resolutionsequencing gel.The nucleasecleavage patternsofthe free probe (lane F)and theprobeboundbyeitherEFIIa(lane a),EFIIb (lane b), or EFIIc (lane c), for both the coding and noncoding strands of theEFIIciselement,areshown. A schematicof the EFII

cis element with the 5' and 3' near-directrepeats indicatedbyopen

bars is shown beside the lanes. Footprinted nucleotides are indi-cated for each strand.(B)Schematicrepresentationof the EFIIcis element, indicatingtheEFIIa, EFIIb, and EFIIcfootprintsonthe coding (overlined) and noncoding (underlined) strands. 5' and 3' near-direct repeatsequencesareboxed.

sequences in the EFII ciselement, it is possible that they

representthesameDNA-binding proteinand that the differ-ences in electrophoretic mobility are due to the fact that

EFIIa and EFIIbare proteolytic products of EFIIc.

How-ever, no increase is observed in the relative amounts of

EFIIa and EFIIb versus EFIIc with increasing age ofthe extract.Alternatively, thesameDNA-binding proteincould

giverisetotheslower-migrating EFIIbandEFlIc complexes

ifitwasdifferentially posttranslationallymodifiedor associ-ated with othertranscriptionfactors. On the otherhand,the

EFIIa, -b, and -c complexes could contain distinct

DNA-binding proteinswhich haveverysimilar,orperhaps identi-cal,sequencerecognitionproperties.This latterpossibilityis consistent with thegrowingnumber oftranscriptionfactors that have been foundtoexist infamilies which demonstrate

nearlyidentical sequence recognition properties (30). EFHa, EFHb, and EFIIc are composed ofmultiple

DNA-binding proteins. To address whetherthe same ordifferent

DNA-binding proteins arepresent in the EFIIa, -b, and -c

* EFII

DNA FIG. 5. Apparent-molecular-mass determination. Fourteen-day-old CE 0.5 M NaCl nuclear extract (120 pg) was subjected to SDS-PAGE and transferredto an Immobilon-Pmembrane, which wascutinto 17 3-mm-wide slices. Theproteins in each slicewere

eluted

separate!y

and tested inEMSAs. EMSAswere performed with0.5ngof31P-labeledEFII DNAand either2,gof14-day-old CE 0.5 M NaCl nuclear extract in the presence of 1.3 ,ug of poly(dI):poly(dC) (lane 1) or 10 p.l of SDS-PAGE-fractionated 14-day-old CEextract, ranging in size from 81 to 19 kDa, inthe

presence of 100 ng of poly(dI):poly(dC) (lanes 2 to 12). EFIIa, EFIIb,and EFIIc retardedcomplexesintheunfractionatedextract

(lane 1)areindicated. Sizes areshown above the lanes(in kilodal-tons).

complexes, 14-day-old CE nuclearextractwas subjectedto SDS-PAGE and blottedtoImmobilonpaper.Thepaperwas cut into small slices, and the proteins in each slice were

eluted and assayed for EFII DNA-binding activity. As showninFig. 5, three protein-DNA complexes (lanes 4 and 9) migratedtothepositions correspondingtoEFIIa, -b, and -cobserved inunfractionatednuclearextracts(lane 1). The

dominant shifted band in lane 9 (also found in lane 10) is equivalent in mobility to EFIIa and represents a protein complex composedofone or moresubunits of 22to29 kDa. The two slower-mobility complexes in lane 4 migrate to

positionscorresponding toEFIIb and EFIIcand represent complexesmadeupof subunits of52to61 kDa. These three renatured EFIIa, EFIIb, and EFIIc complexes specifically recognizedthe EFIIcis elementincompetitionassayswith thewild-typeandmutantEFIIDNAs (datanot shown).

In additiontotherenaturedEFIIa, -b, and-ccomplexes, anumber of otherprotein-DNA complexeswereobserved, although most did not correspond in mobility to EFIIa, EFIIb, orEFIIc. The faster-migrating complexes in Fig. 5,

lanes 4 to 6, were found tobe nonspecific in competition

assays (data not shown). Although the complex in lane 3

migrates to a position similar to that ofEFIIa, it did not demonstratehigh affinityfortheEFIIcis element in compe-tition assays(datanotshown). For thisanalysis, aminimal amount of nonspecific competitor DNA was used in the EMSAinordertomaximizerecovery.The addition ofmore nonspecific poly(dI):poly(dC) to the EMSA eliminated the nonspecific complexesmentioned above but hadnoeffecton theEFIIa, -b,and-ccomplexesinFig. 5,lanes 4and 9(data not shown). The complexes in Fig. 5, lanes 2 and 7,

A.

A

on November 9, 2019 by guest

http://jvi.asm.org/

[image:8.612.72.308.80.358.2] [image:8.612.332.566.81.275.2]

however, were shown to specifically bind EFII DNA and correspond in apparent molecular mass to one or more subunits of 70to81 kDa and 34to39 kDa, respectively. The significance of the minor complexin lane 2 is not known. However, several of the previously characterized C/EBP family members arein the 35to40 kDa rangeand, through shared sequence recognition properties (see below), would be likelytointeract with theEFII cis element. Interestingly, the complexinFig. 5, lane 7,wasnotobserved when CEF 0.5 M NaCl nuclear extract was analyzed following SDS-PAGE fractionation. Moreover, the complex in lane 7 isnot evident when unfractionated 14-day-old CE extracts are used in EMSAs with 32P-EFII DNA, suggesting that the subunits of 34to39 kDaarenotasabundantasand/or bind theEFIIcis element with lower affinitythantheEFIla, -b, and-cfactors.

Anadditional, slower-migrating complexcanbe detected in Fig. 5, lane 9, which specifically recognizes theEFII cis element by competition analysis (data not shown). This

complex displayed DNA-binding characteristics consistent with the addition ofasecond protein complextoarepeated binding site. For example, we observed that the additional band in lane 9wasonlypresent athigh concentrations of the 25-to29-kDa fraction, and in competition experiments with unlabeled EFII DNA, itwas seento disappear priortoany diminution in the renatured EFIIa band. Moreover, the additional complex in Fig. 5, lane 9,wasabsent when the 25-to29-kDa fractionwas subjectedtoEMSA with the radio-labeledEFII3'mutant,which lacksthelower-affinity down-stream repeat (data not shown). Thus, we interpret the additional shifted complex seen in lane 9 to be a second EFIIa complex, bindingtothe 3'repeatunderconditions of sufficiently highfactor concentration. Thisinterpretation is further substantiated by the footprinting experiments

pre-sented inFig. 7(see below).

Since these molecular mass determinations were

per-formed with the factorspresent ina crudenuclearextract, the potential for proteolysis of the factors is a concern.

However, we performed a similar analysis of the EFII factors afterlysingintact CEF inSDS-containingbuffer and

immediately boiling the total-cell lysate. This approach

should minimizeproteolysisasmuchaspossible. Nonethe-less, results identical to those presented in Fig. 5 were obtained (data not shown). It is entirely possible that

pro-teins in additiontothepolypeptidesdetectedhere afterSDS electrophoretic separation are components of the EFII DNA-binding activitiesin14-day-old CEnuclearextracts.A

polypeptide whose DNA-binding activity cannot be rena-tured after SDS exposure will not be detected, nor will a complex which requires the association of heterologous

subunits for DNA binding, unless all polypeptides in the

native complex have similar molecular masses and are

renatured in thesamefraction. Despitetheselimitations, this analysissuggeststhat, atminimum, twopolypeptides of 22 to29kDa and 52to61 kDa arepresentinthe EFIIa and in

the EFIIb andEFIIccomplexes, respectively.

EFIIa, EFTIb, and EFIHc recognize the same nucleotide sequence. Although EFIIa, -b, and -c protected the same nucleotides withinthe EFII ciselement from OP/Cu cleav-age (Fig. 4), these complexesappear tobecomposed ofat least two polypeptides ofvery different molecular masses which couldberecognizing differentsequencemotifs within the15-nucleotide footprint. Toattempt todiscernany differ-ences inDNA sequence recognition among these polypep-tides,weobtainedaseries ofEFII oligonucleotides harbor-ing small substitution mutations across the 15-nucleotide

TABLE 1. Relative binding affinities of EFIIa, EFIIb, and EFIIc nuclear factors forEFII mutantbinding sites by

competition analysis

Competitor

Footprinted

nucleotide Relative bindingaffinityb

oligoucletide sequence(codingstrand,

-224 to -210) EFIIa

EFIIb

EFIIc

EFII DNA AGTCTTATGCAATAC 1 1 1

EFIIHL mutant AaaaTTATGCAATAC -1.2 -1.2 -1.3 EFII HM mutant AGTCggggGCAATAC -6 -7 -21 EFII HR mutant AGTCTTATGgggTAC -8 -9 -45 EFII5' mutant AcoggatcoggATAC -9 -9 -51 a Sequence information forcompetitor oligonucleotides: EFII DNA and EFII 5'mutant,Fig. 3A; EFII HL, HM, andHR mutants aredouble-stranded 50-bpoligonucleotides identicaltoEFIIDNA exceptfor the mutations shown

inlowercaseletters.

b RelativebindingaffinitieswerecalculatedasdescribedinMaterialsand Methods.Negative values indicateaffinity lower than that of wild-type EFII

DNA.

sequence footprinted by EFIIa, -b, and -c. The relative

binding affinitiesof theEFIIa, -b,and-cfactorsfor each of

these mutant oligonucleotides (obtained from competition

analyses analogous to those shown in Fig. 3) are given in

Table 1. Substitution of the GTC on the left side of the

footprinted sequencewith AAA(EFIIHLmutant) didnot

decrease the relativebindingaffinities of either EFlIa,-b,or

-c. Thus,we can narrowdown the sequence important for

the DNA-binding activities of EFIIa, -b, and -c to the 11

nucleotides TTATGCAATAC (coding strand). Substitution

ofeitherTTAT withGGGG(EFIIHMmutant)orCAAwith

GGG

(EFII

HR

mutant)

reduced thebinding affinitiesofboth

EFIIaand EFIIb 6-to9-fold andthebinding affinityof EFIIc

21- and 45-fold, respectively, relative to wild-type EFII DNA. The decrease in the relative binding affinities of all

three complexes for the EFII HM and HR mutantsis very similartothat observed with the EFII 5'mutant(Table1 and

Fig. 3C),inwhichmostof thefootprintedsequence(10of15

nucleotides) has beendisrupted. The only exception is that

EFIIcbinds tothe EFII HM mutantwith a slightly higher

relativeaffinitythantoeither the EFII 5' mutantorthe EFII HR mutant,possiblybecause ofapartial abilitytorecognize

theremainingGCAATACportionof the

li-nucleotide

bind-ing site. The low-affinity binding of EFIla and EFIIb to

oligonucleotides harboring mutations within the

11-nucleo-tidebindingsite islikelyduetotheabilityof these factorsto

recognizesimilar sequences in the 3' repeat.

Atthis level ofresolution, the EFIIa and EFIIb factors appear to have identical sequence-specific DNA-binding

properties, while EFlIc has a more restricted sequence

requirementforthe 5' repeat.Itisinteresting that the EFII

sequences important for high-affinity binding by all three

EFIIDNA-binding factors contain the sequence motif

TT-NNGCAAT. This motif is found inanumberofhigh-affinity

bindingsites for the previously describedtranscription

fac-torC/EBPa(27, 61, 65, 70)aswellasforagrowingnumber

ofC/EBP-related familymembers(1,7,10, 15, 34,56, 58).In

fact,C/EBPahas been showntobind the EFIIciselement in

the RSVLTRinvitro; Ryden and Beemon demonstrated by DNase I footprinting experiments that a heat-stable factor present in chicken liver nuclear extracts and purified

C/EBPotboth bound theEFIIregionof theRSV LTR(61).

Interestingly, unlike the EFII-binding proteins described

here, footprints appeared concomitantlyoverboth repeats in

the EFIIciselement withincreasing concentrationsofeither

on November 9, 2019 by guest

http://jvi.asm.org/

[image:10.612.70.311.105.184.2]

TABLE 2. Relative bindingaffinities of EFIIa,EFIIb, and EFIIc nuclear factors for various binding sites by competition analysis

Competitor Nucleotide Relativebindingaffinityb oligonucleotidea sequence EFIIa EFIIb EFIIc

EFII DNA TTATGCAATAC 1 1 1

IgH-E TTGAGCAATGT -1.6 -1.5 -4

IBF DNA GGCTGCAATTG -3 -3 -4

DE I TTTTGTAATGG -11 -9 -14

Core TGTGGAAAGGA >100 >100 >100

ISequenceinformation forcompetitor oligonucleotides: EFII DNA, Fig.

3A; IgH-E,IBFDNA,DEI,andCore, Materials and Methods. b See Table 1, footnoteb.

the liver nuclear factororpurifiedC/EBPa(61). The murine C/EBP-related factorimmunoglobulin enhancer-binding pro-tein

(Ig/EBP-1)

hasalso been shownby Roman et al. tobind

the EFII cis element inthe RSVLTRinvitroas a

3-galac-tosidase fusion protein encoded in part by the murine Ig/EBP-1cDNA,and the role ofIg/EBP-1 in regulating RSV

transcription inmurine cells wasdiscussed by these

inves-tigators (58). In addition, a putative C/EBP-related factor,

the internal binding factor (IBF), whichwas characterized

andpurified fromBHKcells,wasreported by Karnitzetal.

to bind the EFII cis element in competition analysis (33).

Given that the EFII factors recognize a sequence motif

which fits the consensus binding site for C/EBP-related

transcription factors, we wished to compare the sequence

recognition requirements of EFIIa, -b, and -c in avian

fibroblasts with those of the members of theC/EBP family

which have been reported tobind the EFII cis element in otherspecies and cell types.

Comparison of the DNA-binding properties of EFIIa,

EFIHb,andEFMc with those ofC/EBPa, IBF, andIJEBP-1.

Initially, a series of competition experiments were

per-formed to ascertain whether the EFIIa, -b, and -c factors could also bind to high-affinity recognition sites used to

characterize C/EBPcx, IBF, and Ig/EBP-1. As shown in

Table 2, the EFII factors showednegligible tolow

recogni-tion oftwohigh-affinity C/EBPa binding sites:theSV40core

enhancer sequence found in thecollagenIII promoter(9, 29, 61)and the DE Isite in theratalbumin promotor(18, 50, 73).

In contrast, oligonucleotide DNAscontaining binding sites for either

Ig/EBP-1

or IBF competed relatively well for

EFIIa, -b, and-cbindingactivities. EFIIa and EFIIb bound

IgH-E, an oligonucleotide which encompasses the E site

from the murine immunoglobulin heavy-chain enhancer,

with only slightly lower affinity than they did EFII DNA.

The IgH-E nucleotide sequence was used to identify the

Ig/EBP-1 expression clone from murine L-cell and B-cell

cDNAlibraries

(58).

Sequence analysisofIg/EBP-1revealed that it hashighhomology toC/EBPa throughout the

DNA-binding domain and leucine zipper region

(58).

EFIIa and

EFIIb also bound to IBF DNA, an oligonucleotide which

containssequencesfromaninternalenhancer located within

the gag gene of RSV (2, 33), with threefold lower affinity

thanto EFII DNA. TheIBFnucleotide sequence wasused

to characterize and purify the IBF DNA-binding activity

from BHK cell nuclear extracts (33, 34). IBF has been

reportedtobe relatedtoC/EBPafrom itsoverlappingDNA

sequencerecognition propertiesand heatstability (8, 34, 39).

EFIIcrecognized both IgH-E and IBF DNA with fourfold lower affinitythan the EFII sequence. These differences in the affinities of EFIIa, -b, and -c for the EFII, IBF, and

IgH-E DNAs, although relatively small, were reproducibly

observed.

Comparisonof the celltypedistribution ofEFMIa, -b, and -c

withIBF,

Ig/EBP-1,

andC/EBPa Although the binding sites for IBF and Ig/EBP-1 were recognized by the three EFII factors present inCEnuclearextractswith moderatetohigh

affinity, similarDNA-binding specificitycouldbe exhibited

by different proteins that may be selectively present in

different cell types. Thus, we nextcharacterized the EFII

DNA-binding activity in cell types in which IBF and Ig/

EBP-1 expression has been documented. As mentioned

above,

IBFwas purified from BHK cell nuclear extracts,

and

Ig/EBP-1

expression, monitored by Northern (RNA

blot) analysis, isobserved in allmurine celltypestested but

isparticularlystronginmurineBlymphocytes.We prepared

nuclear extracts from BHKcells and from anavian B-cell

lymphoma line, Bk3A, to testfor EFIIDNA-binding

activ-ity.AsshowninFig. 6A,lane1,andFig. 6C,lane 5, typical

EFIIa, -b,and-ccomplexesareobserved whenradiolabeled

EFIIDNAisincubated withaBHKcell nuclear extract. To

confirm that the BHK extract contained active IBF, we

incubatedtheextractwith radiolabeledIBF DNA. Astriking

difference between the radiolabeled EFII and IBF

DNA-binding

patternswasobserved

(Fig.

6A, lanes1and 2).The

IBF DNA-binding pattern more

closely

matched that

re-ported by

Karnitz et al.

(34).

We performed competition

analyses with unlabeled EFII DNA in an EMSA with

radiolabeled IBFDNAandBHKextract.As shown in Fig.

6A,

lanes2to7, byouranalysis,IBFhasnegligible affinity

forthe EFII DNA sequence;an80-fold molarexcessof EFII

DNAwasrequired tocompetefor50% of the IBFbinding

activity.

Thus, we conclude that the IBF factor

character-ized inBHKcells isunlikelytobeanintegralpartofeither

the EFIIa, -b,or -cDNA-binding complex. Moreover, this

IBFfactor doesnotappear tobe very abundant in nuclear

extracts from 14-day-old CE since we do not observe its

formation in EMSAs with radiolabeled IBF DNA and

14-day-old

CEextract

(data

notshown).

In contrast, radiolabeled IgH-E and EFII DNAsyielded

nearly identical binding patterns in the Bk3A lymphoma

extractexceptthatnoEFIIccomplexwasobserved to form

with radiolabeledIgH-E

(Fig.

6B). Also, radiolabeledIgH-E

bound EFIIa and EFIIb in EMSAs with 14-day-old CE

extract

(data

not

shown).

Thus,it is possiblethat eitherof theavianEFIIaand EFIIb factors ishighlyrelatedtomurine

Ig/EBP-1.A greatexcessof theEFIIafactor,farexceeding

the level of EFIIb

(or EFIIc),

was detected with either

radiolabeledEFIIorIgH-EDNAin the Bk3Aextract(Fig.

6B). Although

the

proportions

of EFIIa, -b, and -c are

different fromthose observed in CEF, 14-day-old CE, and

BHK extracts, allthree complexesin the Bk3Aextract are

heat stable and demonstrated characteristic DNA-binding

specificities

in

competition

analysis (Fig. 6C,lane9,and data

not

shown).

Wealso observed thatanincrease in theamount

of Bk3A nuclearextractresultedin thedisappearanceof the EFIIb and EFIIcshifts and theformationofa newcomplex

(Fig. 6C,

lane

10),

termedABF

(additional

binding factor).

The ABF complex appeared onlywith the Bk3A extract;

titration of all otherextractsledto auniformincrease in all three factors

(data

not

shown).

As will be discussed later

(Fig. 7),

theABFcomplexappearstorepresenttheaddition of a second EFIIa factor to the downstream near-direct repeatin the EFIIcis

element,

caused

by

the

high

concen-tration of EFIIa in the B-cellextract.

InadditiontoBHKand Bk3A nuclear extracts,wetested anumber offibroblast and

epithelial

cell lines for the

EFIIa,

on November 9, 2019 by guest

http://jvi.asm.org/

A. BHK extract B. Bk3A extract

EFllc_H ^ ^ ^ F

~~~Fllc--_

EFllb _-o

WV,

EFllb

_-p

EFll_IIFIa

DNA ; X _ IBF EFII _ IgHE

DNA DNA DNA

1 2 3 4 5 67 1 2

molar excess: 0 1 5 10 50100I

competitor: EFII DNA

C. Ligprotein: 3 2 3 3 9 7 4 10 1 2

I- ABF

4- EFIIa

6 7 8 9 10

> a) r X OD 3I

[image:11.612.61.286.79.420.2]

wLX

FIG. 6. Comparingthe celltypedistribution ofEFIIa, -b,and-c

with that ofIBF,Ig/EBP-1,andC/EBP.(A)EMSAswereperformed

with 9 ,ug of BHK 0.5 M NaCl nuclear extract and 0.5 ng of 32P-labeled EFII DNA(lane 1) or IBF DNA (lanes 2 to7) in the presenceof 1.3 p,gofpoly(dI):poly(dC)and in the absence(lanes1 and2)orpresence(lanes3 to7)ofa1-to100-fold molarexcessof nonradiolabeled competitor EFII DNA. (B) EMSAs were per-formed with 1,ugofheated Bk3A 0.5 M NaCl nuclear extract and 0.5 ng of32P-labeled EFII DNA(lane 1)orIgH-EDNA(lane 2)in the presenceof 1.3p,gofpoly(dI):poly(dC). (C)EMSAswereperformed

withsamplesof 0.5 MNaCl nuclearextractsfrom the indicatedcell

typesand0.5 ng of32P-labeled EFII DNA in the presence of 1.3,ug

ofpoly(dI):poly(dC). Lanes: 1and2, 14-day-old (14d) CE; 3,A431

epidermal carcinoma cells; 4, BALB/c-3T3 fibroblasts; 5, BHK

cells; 6,rat H4hepatoma cells; 7,adult ratliver;8,adultratkidney;

9 and10, avianB-celllymphomalineBk3A. The amountofprotein

included ineachEMSA is listedabove the correspondinglane.A,

heated extract. EFIIa, EFIIb, EFIIc, and ABF retarded

protein-DNAcomplexesareindicated.

-b, and -c DNA-binding pattern (Fig. 6C). We observed a patternofbinding activitysimilartothat ofEFlIa,-b, and-c

in nuclearextractsfrom human epidermal carcinomaA431 cells (Fig. 6C, lane3) andmurinefibroblasts(BALB/c-3T3; Fig. 6C, lane 4) as well as several other fibroblast lines

(Rat-i, NIH3T3, andCoscells;datanotshown). The EFII

DNA-bindingactivities inthese extracts,asinourstandard CEextract,wereheatstable (Fig. 6C,comparelanes 1 and

2, also lane 3; data not shown). We also obtained nuclear extracts fromadifferentiated ratliverheptomacellline,H4,

and adult rat liver and kidney tissue. Reproducibly less (hepatoma and liver; Fig. 6C, lanes 6 and 7,respectively) or negligible (kidney; Fig. 6C, lane 8) EFII DNA-binding activ-ity was detected in these extracts. Furthermore, retarded complexes corresponding in mobility to EFIIa were not

observed. Although we cannot rule out the possibility that poor extraction or selectiveproteolysis of the EFII factors occurred consistently in the H4, liver, andkidney extracts, all extracts tested were prepared by the same method, and analysis of proteins that bind a hypersensitive site 4,800bp upstream of the phosphoenolpyruvate carboxykinase gene demonstrated that the H4, liver, and kidney extracts we analyzed were not significantly proteolyzed (12). Although the tissues and cell types tested were from variousspecies, itappears fromthis preliminarysurvey thatthe EFIIa, -b, and -c binding activities are not uniformly present in all cell types.

EFIIa binds to the 3' repeat sequenceinthe EFII cis element

athigh concentrations. Wewere intrigued by the unusually large amount of the EFIIa factor present innuclear extracts from Bk3A cells(Fig. 6B and C), as well as the presence of theABF complex when higher concentrations of the extract wereused in theEMSA (Fig. 6C, lane 9). We analyzed Bk3A nuclearextractby SDS gel electrophoresis and Immobilon transferand obtained results identicaltothose shown for the 14-day-old CE extract in Fig. 5 except that in the 25- to

29-kDa fraction, greater amounts of both the EFIIa-like factor and the slower-mobility complex were seen. As shown in Fig. 7A, when we compared the mobility shift

pattern seen in the Bk3A 25-to29-kDa fraction(lane2)with

theEFII DNA-binding pattern of unfractionated Bk3A ex-tract, in which the ABF complex is observed (lane 1), we

found that the ABF factor was identical in mobilityto the slower complexin the 25- to 29-kDa fraction. This slower

complex was previously interpreted to represent an

addi-tional EFIIacomplex binding to the 3' repeat in the EFIIcis

element. We therefore suggest that the ABF complex also represents the binding of a second EFIIa factor to the 3' repeat in the radiolabeled EFII DNA, because of the abun-dance of EFIIa inthe Bk3A cells. In support ofthis,wealso noted that the ABFcomplex did not form inanEMSAwith

the radiolabeled EFII 3' mutant, which contains only one

intact repeat, and that ABF demonstrated thebinding kinet-ics expected for the addition ofasecond factorto arepeated binding site by competition analysis (data not shown) and extracttitration (Fig. 6C, lanes 9 and 10).

In order to directly analyze the possible presence of an

additional proteincomplex bound to the 3' repeat of the EFII cis element in the ABF complex, we performed OP/Cu nuclease footprinting of both the EFIIa protein-DNA com-plex and the ABF protein-DNA comcom-plex present in the Bk3A nuclear extract. Comparison of the EFIIa footprints with the ABFfootprints (Fig. 7B) revealed that proteins in the ABFcomplex protected the same sequences in the 5' repeat onthecoding and noncoding strandsastheproteins in the EFIIa complex, although protection was enhanced in theseregions. However, only proteins in theABFcomplex protected similar nucleotides in the 3' repeat. The additional

ABF footprints in the 3' repeat protected 12 nucleotides from -203 to -192 on the noncoding strand and 10 nucleo-tides from -201to-192 on thecoding strand. A diagram of the ABF footprint along with the EFIIa footprint is pre-sented inFig. 7C.

on November 9, 2019 by guest

http://jvi.asm.org/