MOLECULAR AND CELLULAR BIOLOGY, Jan. 1989, p. 312-320 Vol. 9, No. 10270-7306/89/010312-09$02.00/0Copyright X3 1989, American Society for Microbiology

New B-Lymphocyte-Specific Enhancer-Binding ProteinA. DORN, C. BENOIST, AND D. MATHIS*

Laboratoire de Genetique MoMculaire des Eucaryotes du Centre National de La Recherche Scientifique and Unite 184 deBiologie Moleculaire et de Genie Genetique de l'Institut National de la Sante et de la Recherche Medicale, Institut de

Chimie Biologique, Faculte de Medecine, Strasbourg, France

Received 23 June 1988/Accepted 19 October 1988

We report the discovery of a new B-lymphocyte-specific enhancer-binding protein. A series of gel retardationassays using fragments that scan the -2172 to -1180 region of the major histocompatibility complex class IIgene E. reveal a site (W) that serves as the recognition sequence for two nuclear proteins, one B-cell restrictedand the other ubiquitously occurring. Certain characteristics of the NF-W1 and NF-W2 pair recall the OTF-2/NF-A2 and OTF-1/NF-Al pair that binds to the immunoglobulin octamer, but we demonstrate that the twoprotein pairs are distinguishable by several criteria. NF-W1 and NF-W2 interact differentially with theircommon GTTGCATC binding site, display a different affinity for it, and have molecular weights that differ byabout 20,000. Yet, proteolysis experiments and cross-linking analyses indicate that the two W complexes showstructural relatedness.

B lymphocytes are a favored system for studying cell-type-specific regulation of gene transcription. To a greatextent, this may be ascribed to the availability of an array ofcultured cell lines representing discrete stages in the B-celllineage. It may also be attributed to the existence of well-characterized gene families coordinately regulated through-out B-cell differentiation: the immunoglobulin heavy- andlight-chain genes and the major histocompatibility complex(MHC) class II genes.The B-cell specificity of immunoglobulin gene expression

is dictated by multiple regulatory elements (23). Reportshave appeared of a tissue-specific enhancer (4, 27, 30-32), atissue-specific promoter (10, 12, 13, 26, 32), tissue-specificsilencer elements (17, 18, 20, 43, 45), and perhaps sequencesresiding in the body of the gene (14).Much less is known about the regulation of MHC class II

gene expression in B cells. The 5'-flanking region of themurine Ea gene shows B-cell-specific enhancer activity thatis due to the interplay of two regulatory elements, onelocated on the -2172 to -1148 fragment and another in the-215 to +12 region (21, 21a). Additional evidence thatdiscrete B-cell-specific elements operate to control EBa geneexpression has been provided in transgenic mouse studies;mice carrying an Ea transgene 5' truncated to position -1180do not transcribe it in essentially all B cells, even thoughexpression is more or less normal in other cell types (9a, 42,46). Yet, no convincing evidence has ever been presented fora B-cell-specific protein interacting with the Ea promoter.Two proteins (NF-X and NF-Y) are known to bind to highlyconserved motifs (the X and Y boxes) of Ea and other MHCclass II genes, and these proteins have been extensivelycharacterized (8, 9, 15, 21a, 29, 40). But they are ubiqui-tously occurring proteins that bind to sequences in thepromoter-proximal region whose influence is felt in all celltypes that express class II molecules (5, 9, 40). In short, wehave very little understanding of how MHC class II genesare regulated in B cells.

* Corresponding author.

In this report, we describe a B-cell-restricted and a ubiq-uitously occurring protein that share a common binding sitealong the Ea 5'-flanking region. Certain parallels are notedbetween the NF-Wi and NF-W2 pair binding to the E. Wsequence and the OTF-2INFA-2 and OTF-1/NF-Al pairrecognizing the immunoglobulin octamer sequence, but wedemonstrate that these two protein pairs are not the same.Evidence is provided that NF-Wi and NF-W2 bind differ-ently to the GTTGCATC recognition sequence, that theyhave distinguishable affinities for it, and that they havedistinct molecular weights, but that by two independentcriteria, NF-Wi and NF-W2 are structurally related.

MATERIALS AND METHODS

Cell lines. All the cell lines used in this study were ofmurine origin and were M12, CH27, and WEHI 231 (B-lymphoma lines that constitutively express MHC class IImolecules); 70Z/3 (a pre-B-lymphoma line that is class IInegative); BW5147 (a T hybridoma, also class II negative);LMTK (a fibroblast); MCA (an epithelial cell line); T-113 (ateratocarcinoma); and P388D1 (a macrophagelike line thatcan be induced to express MHC class II molecules bygamma interferon treatment).

70Z/3 cells were induced to differentiate by treatment with20 ,ug of lipopolysaccharide (LPS) per ml for up to 30 h.P388D1 cells were activated by treatment with 50 ,ug ofgamma interferon (Boehringer Ingelheim, Vienna, Austria)per ml for 24 h; induction of cell surface MHC class IImolecules was verified by cytofluorimetric analysis.

Extracts and NF-W1 and NF-W2 purification. Nuclearextracts were prepared by the technique of Dignam et al. (7),except for the modifications already detailed (8). In someexperiments, whole cell extracts were used, and they wereprepared by the method of Manley et al. (24).NF-Wi and NF-W2 were purified as follows. Extracts

from 20 liters of CH27 B-lymphoma cells (-2 x 1010 cells;500 mg of protein) were made 60% in ammonium sulfate andwere incubated overnight on ice. The precipitate was recov-ered by centrifugation at 30,000 rpm (R45 Ti rotor) for 60 min

312

B-CELL-RESTRICTED DNA-BINDING PROTEIN 313

W 5' GCATGCACCACAGAATTGTTGCATCAGCCTCTTGACCC3' CGTACGTGGTGTCTTAACMC6TAGTCGGAGMCTGGG

W-MUT1 5' GCATGCACCACAGMTTGTTGMATCAGCCTCTTGACCC3' C6TACGTGGTGTCTTAACMCTTAGTCGGAGAACTGGG

W-MUT2 5' GCATGCACCACAGAATTGTT6CATXAGCCTCTTGACCC3' CGTACGTGGTGTCTTAACAACGTATTCGGAGAACTGGG

3'5'

3'5'

3'

5'

OCTAMER 5' AATAA1MmGCATACCCTCACTG 3'3' TTATTAAACGTATGGGAGTGAC 5'

B 5' TCTCAACAGAGG6GACTmTCCGAGAGGCCATCTGGCAGTT 3'3' AGAGTTGTCTCCCCTGAAAMGCTCTCCGGTAGACCGTCM 5'

X 5' AAGGAACCCTmTCCTAGCAACAGATGTGTCAGTCTGAA 3'3' TTCCTTGGGAAAGGATCGTTGTCTACACAGTCAGACTT 5'

Xi 5' TACACAGTGCTACTTAGCMCTTATGATGCTGCCGAGT 3'3' ATGTGTCACGATGAATCGTTGAATACTACGACGGCTCA 5'

W-STICKY 5' TGAAACGCATGCACCACAGMTT6TTGCATCAGCCTCTTGACCC 3'3' CGTACGTGGTGTCTTAACAACGTAGTCGGAGMCTGGGACTTTG 5'

FIG. 1. Oligonucleotides used in this study.

at 4°C and was suspended in 50 ml of AS buffer (50 mM Trishydrochloride [pH 7.9], 6 mM MgCl2, 40 mM (NH4)2SO4,0.2 mM EDTA, 1 mM dithiothreitol, 15% glycerol). Thesolubilized material was applied to a heparin-agarose column(column volume, -70 ml; Heparin Ultrogel A4R; IBF); theflowthrough was collected, and the column washed with 4bed volumes of He-40 buffer (25 mM N-2-hydroxyethylpi-perazine-N'-2-ethanesulfonic acid [HEPES, pH 7.9], 1 mMdithiothreitol, 1 mM phenylmethylsulfonyl flouride, 2 mMMgCl2, 25% glycerol, 40 mM KCl). Bound protein waseluted by 70-ml steps of He buffer augmented with progres-sively higher KCI concentrations (100 to 1,000 mM KCl).Fractions were assessed for NF-W1 and NF-W2 activity bythe gel retardation assay; both activities generally peaked inthe 0.5 M KCl fraction.Peak fractions from the heparin-agarose column were

pooled, diluted to 100 mM KCI, and applied to a W-oligonucleotide (oligo) affinity column. The oligo used formaking the column had 6-base protruding ends (Fig. 1),which permitted the formation of oligomers upon ligation.The column was made and run essentially as described byKadonaga and Tjian (19). In short, after application of theheparin-agarose fraction, collection of the flowthrough, andwashing with 10 bed volumes of Z buffer (25 mM HEPES[pH 7.9], 12.5 mM MgCl2, 1 mM dithiothreitol, 1 mMphenylmethylsulfonyl flouride, 0.1% Nonidet P-40, 20%glycerol, 100 mM KCI), we eluted the bound material with500-pJl steps of Z buffer supplemented with progressivelyhigher concentrations of KCl (100 to 1,500 mM). NF-W1 andNF-W2 activities were evaluated by the gel retardationassay, peak fractions were pooled, and the material wasdiluted to 100 mM KCl. This material was applied to a

second W-oligo affinity column which was washed andeluted as above. NF-W1 and NF-W2 activity were againmeasured by the gel retardation assay, and positive fractionswere stored as beads at -90°. Due to their partial separationon the affinity column, it was possible to obtain fractionswith various ratios of NF-W1 to NF-W2.

Binding studies: gel retardation assays and methylationinterference mapping. Gel retardation assays were per-formed as published previously (8). For the competition

experiments, various quantities (0 to 300 fmol) of unlabeledcompetitor oligo were added at the beginning of the bindingreaction. Residual binding was calculated after scanning theNF-W1 or NF-W2 band on autoradiograms; the plottedvalues are averages of at least three experiments.

Methylation interference mapping was conducted accord-ing to a published protocol (8).

Structural studies: pore gradient gel electrophoresis, prote-ase treatment, and formaldehyde cross-linking. Pore gradientgel electrophoresis was performed as described by Hooftvan Huijsduijnen et al. (15), except that a 10 to 15% gradientof polyacrylamide was used.

Protease treatment after the binding reaction has also beendescribed (15). In some experiments, crude nuclear extractwas digested with protease (10 to 200 ng of proteinase K or1 to 100 ng of trypsin) for 15 min before the 30-min bindingreaction of a methylation interference assay.For the cross-linking experiments, a typical gel retardation

assay was run, except that the reaction volume was in-creased 3-fold, the labeled DNA 30-fold, the poly[d(I-C)]3-fold, and the extract 3-fold. NF-W1 and NF-W2 bandswere located by autoradiography of the wet gel, and theexcised bands were incubated in 1 ml of 0.5x TBE-1%formaldehyde at 4°C for increasing times from 5 min to 12 h.Cross-linked complexes were extracted from the gel byelectroelution, precipitated with 5 volumes of acetone, andelectrophoresed on a 12.5% sodium dodecyl sulfate-poly-acrylamide gel. Radioactive bands were located by autora-diography.

RESULTS

Search for a B-cell-specific DNA-binding protein. Twotypes of experiments have established that the -2172 to-1180 region at the 5' end of the E,,, gene is critical for itsexpression in B cells. First, this fragment shows potentB-cell-specific enhancer activity. That is, it can very effi-ciently stimulate transcription when cloned one kilobaseupstream of a heterologous promoter, but only in mature Bcells that normally express class II genes (21). Second,sequences in this region are required for expression ofan E., transgene in the B cells of transgenic mice (9a, 42,46).With these results in mind, we sought to localize se-

quences on the -2172 to -1180 fragment that might serve asthe recognition site for a B-cell-specific DNA-binding pro-tein. We launched an extensive series of gel retardationexperiments to scan the -2172 to -1180 fragment for thebinding site of a protein limited to B cells. Nineteen over-lapping 30- to 650-base-pair fragments were incubated withnuclear extract from either the B-lymphoma line M12 or thefibroblast line LMTK, and the resulting DNA-protein com-plexes were analyzed by the gel retardation assay. A believ-able and reproducible difference between the M12 andLMTK extracts was observed only with fragments abuttingposition 1180 (data not shown).To facilitate a more precise definition of the binding site of

the putative B-cell-specific protein, we prepared a 38-base-pair double-stranded oligonucleotide spanning positions-1219 to -1182 (Fig. 1, W oligo). This oligo corresponds tothe smallest restriction enzyme fragment that engendered aband with M12 but not LMTK extracts. When the 38-merwas incubated with an M12 or LMTK extract, severalcomplexes were formed, as illustrated in Fig. 2. The patternsare quite similar with the two extracts, the most striking

VOL. 9, 1989

314 DORN ET AL.

M12 LMTK A.mm

WEHI 7OZ/3LMTK M12 CH27 231 LPSr---7I m--- m--

a.4 i. ...NF-W2 2-a -t M6aNF-W1- Uw-a q

...

2.1_

.o

B. z

;2 n- + ,.

0 a

FIG. 2. Identification of a B-cell-restricted and a ubiquitousprotein binding to the TGTTGCATC sequence at positions -1203 to-1195 of the Ea gene. One molar NaCl extracts of nuclei from M12(a B cell) or LMTK (a fibroblast) were incubated with 32P-labeled Woligo in the presence of poly[d(I-C)] competitor. Nucleoproteincomplexes were resolved from free DNA by nondenaturing poly-acrylamide gel electrophoresis and were revealed by autoradiogra-phy. For each extract, two amounts of poly[d(I-C)] were used (100and 400 ng). The arrows indicate a B-cell-restricted and a ubiqui-tously occurring band, which represent NF-W1 and NF-W2, respec-tively.

difference being that M12 but not LMTK cells have theprotein represented by band 1.

It is evident that other complexes form on the 38-mer.Their specificity was investigated in a series of methylationinterference and competition experiments (data not shown;see below). Band 1 and band 2 proved to correspond toproteins making specific contacts on the DNA. They willhereafter be referred to as NF-W1 and NF-W2, and theircommon binding site (see below) will be termed W. Theother complexes will not be discussed in this report; we havejudged each one less interesting than the NF-W1 and NF-W2complexes for one or more of the following reasons. (i)Appearance of the retarded band is not reproducible withdifferent extracts prepared from the same cell line. (ii)Appearance of the band is inhibited by all unlabeled com-petitor fragments, signifying that the complex is nonspecific.(iii) No unambiguous and repeatable methylation interfer-ence footprint can be observed with DNA extracted from theband. Nonetheless, we are not ready at this time to state thatall of these complexes are irrelevant to the control of E.transcription, but for the sake of simplicity, we will concen-trate on NF-W1 and NF-W2.

Cell-type distribution of NF-W1 and NF-W2. We haveconfirmed the B-cell specificity of NF-W1 and the ubiquityof NF-W2 by a series of gel retardation assays. We foundNF-W1 at various levels in extracts from mature B cells that

NF-W24.3 £NF-W1-.

FIG. 3. Cell-type distribution of NF-W1 and NF-W2. (A) MatureB cells. Gel retardation assays were run as in Fig. 2 except that allextracts had been partially purified by running them over a heparin-agarose column (Materials and Methods). In all cases, the 0.5 M KClfraction was used; no other fraction showed significant NF-W1activity. Two poly[d(I-C)] amounts of 240 to 1,200 ng were used foreach extract, depending on the extract. LMTK is a fibroblast line;M12, CH27, and WEHI 231 are B-lymphoma lines that constitu-tively express MHC class II molecules; 70Z/3 is a pre-B-lymphomaline that has been induced to differentiate (e.g., to express immu-noglobulin K) by treatment with LPS for 24 h. (B) Other cell types.Gel retardation assays with unfractionated nuclear extracts fromM12, LMTK, BW5147 (a T hybridoma), P388D1 (a macrophagelikeline induced to express MHC class II molecules by gamma inter-feron treatment), MCA (an epithelial cell line), and Terato-113 (anundifferentiated teratocarcinoma). The gels were purposely over-loaded to emphasize the lack of NF-W1 in these extracts.

express MHC class II molecules: M12, CH27, WEHI 231,TA3, A20, and isolated splenocytes (Fig. 3). This proteinalso occurred in extracts from 70Z/3 pre-B cells that hadbeen provoked to differentiate by treatment with LPS. TheNF-W1 band varied somewhat with the different extracts,sometimes showing up as a doublet. NF-W1 occurred atvery low levels or was absent from cells that do not expressclass II molecules: undifferentiated 70Z/3 cells, the T lym-phoma BW5147, the fibroblasts LMTK and NIH 3T3, theepithelial cell line MCA, and the macrophagelike lineP388D1 with or without gamma interferon treatment. NF-

MOL. CELL. BIOL.

B-CELL-RESTRICTED DNA-BINDING PROTEIN 315

TIME OF LPS STIMULATION

- 15' 2h 8h 2Qh 3Oh M12 LMTmmi

A

NF-W1-- 00-4u-:, ? W

B -A.

NF-A2/OTF2-10TWF2-W

NF-KcB I

FIG. 4. Kinetics of induction of NF-W1 and otherproteins in 70Z/3 pre-B cells treated with LPS. 70:treated or untreated with 20 j±g of LPS per ml forbefore preparation of nuclear extracts. Gel retardaticconducted with the W oligo (A), the immunoglobuoligo (B), and the immunoglobulin K B oligo (C). IdentNF-KB band was by several criteria, including comethylation interference experiments; its appearancpublished reports (e.g., 2, 38, 39) because we use an

both the B motif and an adjacent Ephrussi sequence

press). The sequences of the oligos are listed in Fig. 1.for each time point correspond to different con

poly[d(I-C)] competitor (100 and 400 ng).

W2, on the other hand, was present in extractslines tested so far. This protein has also beetseveral mouse tissue extracts, e.g., those ofand teratocarcinoma.

Since at least two immunoglobulin gene

factors (NF-A2/OTF-2 and NF-KB) are induccells by LPS treatment (2, 39, 41), we were

exploring further the NF-W1 induction pherdemonstrated by the kinetic experiment depictNF-W1 is induced between 2 and 8 h of LPS tr(is the same window during which NF-A2/OTF(Fig. 4B). However, NF-KB induction is detearlier, even after 15 min of LPS treatment (Idichotomy of NF-KB versus NF-A2/OTF-2 ai

further emphasized by the fact that NF-KB is in3 cells treated with LPS plus cycloheximide, Kand NF-A2/OTF-2 are not (data not shown).Comparison of the binding properties of NF

W2. Preliminary methylation interference exp

cated that NF-W1 and NF-W2 had slightly diffcharacteristics. We were interested in furthethis point but experienced difficulties becausconcentrations of these two proteins in crudebecause of the presence of extraneous proteithe W oligo. To resolve these problems, w

enriched NF-W1 and NF-W2 by passage ov

agarose column followed by two passes ov

affinity column (see Materials and Methods).

the WT lanes (Fig. 5A), this procedure substantially im-K proves visualization of both proteins.

That NF-W1 and NF-W2 make distinct contacts on theircommon binding site was established by methylation inter-ference experiments with the purified material (Fig. 5B). Itappears that NF-W1 is more dependent on one of the G

0 0

| w -" residues on each strand; i.e., the contacts are GTTGCATC for

NF-W1, and CAACGTAG for NF-W2.S

A greater dependence of NF-W1 on the external G ofthe antisense strand is also suggested by gel retardationassays with mutated W oligos (Fig. 5A). Substitution of theinternal G on the antisense strand by a T residue (Mut 1,

- Octamer- *GrGAATC ) results in greatly reduced binding of bothNF-W1 and NF-W2. Substitution of a T for the external G(Mut 2, CAACTATA) has markedly less influence on thebinding of both proteins, but NF-W1 is more affected than

- B - NF-W2. This is most clearly seen by comparing lanes 4 ofWT and Mut 2 (Fig. 5A).NF-W1 and NF-W2 binding to the W oligo can also be

differentiated on the basis of self-competition experiments.r DNA-binding When the binding reaction to 32P-labeled W oligonucleotideZ/3 cells were is conducted in the presence of increasing amounts of thevarious times same oligo unlabeled, the NF-W1 band disappears more

)n assays were quickly than the NF-W2 band. The shapes of the curvesilin H octamer depicted in Fig. 5C thus indicate that NF-W1 has the highertification of the affinity of the two proteins for the W sequence.)mpetition and To summarize, three properties distinguish the binding ofwe differs from NF-W1 and NF-W2 to their common recognition site: sen-oligo carrying sitivity to A+G methylation, sensitivity to mutation, andThe two lanes strength of attachment.centrations of Comparison of the structures of NF-W1 and NF-W2. As a

first step in elucidating the basis of these binding differences,we compared several structural parameters of NF-W1 andNF-W2. These were: (i) molecular mass, (ii) proteolysisproducts, and (iii) cross-linking products.

i from all cell (i) Molecular mass. The sizes of NF-W1 and NF-W2 weren detected in initially compared by pore gradient gel electrophoresis. Thisspleen, liver, method entails running the protein of interest on a nondena-

turing polyacrylamide gel of graded porosity; the proteintranscription migrates until it reaches its pore-size limit, which is molec-ible in 70Z/3 ular mass dependent (1). DNA-protein complexes wereinterested in formed on the W 38-mer as usual, by using a purified fractioniomenon. As that contains approximately five times as much NF-W2 asLed (Fig. 4A), NF-W1. The incubated material was loaded onto a gradienteatment. This polyacrylamide gel along with a series of protein size mark--2 is induced ers. By comparing the migration of the two complexes withectable much the migration of the marker proteins, we arrived at aFig. 4C). The molecular mass estimate for the NF-W1/W-oligo complex ofnd NF-W1 is 73 kDa and for the NF-W2/W-oligo complex of 93 kDa (Fig.iduced in 70Z/ 6).while NF-W1 The contribution of the W oligo to the molecular mass of

the complex is rather difficult to assess; its calculated'-Wi and NF- molecular mass is 24 kDa but DNA and protein of the sameeriments indi- mass will usually migrate differently on a polyacrylamide'erent binding gel. In any case, the oligo contribution is the same for ther elaborating two complexes, so we can probably say that NF-W1 is about,e of the low 20 kilodaltons (kDa) smaller than NF-W2.extracts and That the oligo does not cause totally aberrant molecular

Lns binding to mass estimates is evidenced by the fact that NF-W1 activitye extensively runs with a molecular mass of 50 to 75 kDa when a crudeer a heparin- nuclear extract is applied to a gel filtration high-performanceer a W-oligo liquid chromatograph and that NF-W1 sediments with anAs shown in apparent molecular mass of 60 kDa on glycerol gradients

M.F

VOL. 9, 1989 ;3X0 w ? 1gF

316 DORN ET AL.

WT Mutl Mut2I I 'I

123412341234

W2m- wmumWl * "

FM

'SW04

S AS

F 21 F 21~~~~~~~~~._T<~~

C.

3 NF-W2a NF-Wl

510 20 30 50 100

Cold Competitor (Femtomoles)

V Nt--S.V: =N%7A=O

N- N CC'..AC-

FIG. 5. Comparison of binding properties of NF-W1 and NF-W2. (A) Gel retardation assay with wild-type and mutated templates. NF-W1and NF-W2 were extensively purified from CH27 whole-cell extracts by passage over a heparin-agarose column and two W-oligo affinitycolumns. WT is the wild-type W oligo. Mut 1 has a G- T transversion at the internal G on the antisense strand contacted by both NF-W1and NF-W2; Mut 2 has a G--T transversion at the external G on the antisense strand contacted more distinctly by NF-W1. The sequencesof the oligos are presented in Fig. 1. Four amounts of poly[d(I-C)] were used: 120 (lane 1), 240 (lane 2), 480 (lane 3), and 570 (lane 4) ng. Theasterisked band probably represents a degradation product of NF-W2 (Discussion). (B) Methylation interference mapping. As in panel A, an

extensively purified CH27 fraction containing NF-W1 and NF-W2 was used. Binding reactions were conducted with methylated DNA 5' endlabeled with 32p on the sense (S) or antisense (AS) strand. After electrophoresis on a standard nondenaturing polyacrylamide gel, free DNA(F) or DNA from the NF-W2 (2), NF-W1 (1), or degraded NF-W2 (*) bands was electroblotted, eluted, cleaved, and displayed on a sequencinggel. Arrowheads indicate bases whose methylation interferes with protein-DNA complex formation. (C) Self-competition experiments. Gelretardation assays were conducted with a fixed amount of extensively purified CH27 extract, poly[d(I-C)], and 32P-labeled W oligo. Increasingamounts of unlabeled W oligo (5 to 100 fmol) were added at the initiation of the binding reaction. After electrophoresis of the incubatedsamples and autoradiography, residual binding was quantitated by densitometry. The vertical axis expresses residual binding of NF-W1 or

NF-W2 in the presence of competitor W oligo as a percentage of binding in the absence of competitor W oligo. The values represent averagesof three independent experiments.

(data not shown). NF-W2 activity was not recovered in theseexperiments, as it seems rather sensitive to dilution.

(ii) Proteolysis products. Useful structural informationoften derives from an analysis of the proteolysis products of

PORE-GRADIENTRETARDATION

kD

-136

NF-W2-NF-Wi -

6

-46

FIG. 6. Comparison of the molecular masses of the W complexesby pore gradient gel electrophoresis. A binding reaction was set upwith the W oligo, 120 ng of poly[d(I-C)], and an extensively purifiedNF-W1-NF-W2 fraction that had approximately 5x more NF-W2than NF-W1. After the standard incubation, the material was loadedonto a 5 to 15% gradient polyacrylamide gel together with proteinmolecular mass markers.

a DNA-binding protein. For example, such studies haveprovided evidence for a structurally distinct DNA-bindingcore (6, 15, 16, 22, 25, 36) and for DNA-binding fingers(28).The protease sensitivity of the W complexes was exam-

ined by using purified fractions that contain essentially onlyNF-W1 or essentially only NF-W2. The purified fractionswere incubated with the W oligo and then treated withincreasing amounts of proteases before electrophoresis on a

neutral polyacrylamide gel (Fig. 7). Two important pointsemerge. First, both NF-W1 and NF-W2 have a protease-resistant DNA-bound core detectable with all three enzymesand with subtilisin B (data not shown). Second, the NF-W1and NF-W2 cores appear very similar in size-charge afterdigestion with each of the three proteases.

It was clearly of interest to determine whether the DNA-binding domains ofNF-W1 and NF-W2 retain the differentialbinding properties of the intact proteins. To answer thisquestion, we needed to produce a protease-resistant core

with uncomplexed protein and to show that it is still capableof binding to the W oligo. This was possible after proteinaseK but not after trypsin digestion. Therefore, we performed a

methylation interference mapping experiment, comparingthe patterns obtained with purified, proteinase K-digestedNF-W1 and NF-W2 fractions. As is evident from a compar-ison of the rightmost panel of Fig. 7 with the righthand panelof Fig. 5B, the proteolyzed NF-W1 core, just like intactNF-W1, has an extra contact site on the external G of the

A.

MOL. CELL. BIOL.

B-CELL-RESTRICTED DNA-BINDING PROTEIN 317

TRYPSIN

NF-WlI0 1 2 34 5 6 7

PAN-PROTEASE

NF-W2

0 1 2 3 4 5 6 7

NF-Wl NF-W2

0 1 2 3 4 5 6 0 1 2 3 4 5 6

PROTEINASE K

NF-W1 NF-W2012365O012 34 5 67 0 1 2 3 4 6 5 7

6.wt| t§n :tw

FIG. 7. Protease treatment of NF-W1/W-oligo and NF-W2/W-oligo complexes. Binding reactions were set up with the W oligo, 120 ng ofpoly[d(I-C)], and purified CH27 fractions that contained essentially only NF-W1 (Wl tracks) or essentially only NF-W2 (W2 tracks). Aftera 30-min incubation, various amounts of the different proteases were added. Trypsin lanes: 0, no protease; 1 to 7, 1, 10, 20, 30, 50, 80, 100ng. Pan protease lanes: 0, no protease; 1 to 6, 1, 10, 25, 50, 80, 100 ng. Proteinase K lanes: 0, no protease; 1 to 7, 10, 20, 50, 150, 100, 200ng. After a 15-min incubation, the digested samples were electrophoresed on the standard nondenaturing polycarylamide gel. For themethylation interference experiment depicted in the rightmost panel, the purified CH27 extracts were first treated with 20 ng of proteinaseK for 15 min. A 30-min binding reaction was then conducted by using the digested extracts, 120 ng of poly[d(I-C)], and methylated 32P-labeledW oligo. The incubated samples were electrophoresed on a nondenaturing gel, and bands were located by autoradiography. Afterelectroblotting, elution, and cleavage, the DNAs were displayed on a sequencing gel. F, free DNA. B, DNA from the more slowly migratingband of the core doublet produced by proteinase K. The lower band of the doublet gives exactly the same pattern as the upper band, and sois not shown. The arrow indicates the external G on the antisense strand whose methylation inhibits the binding of NF-W1 but not of NF-W2(cf. Fig. 5B).

antisense strand (arrow, Fig. 7). The DNA-binding cores ofNF-W1 and NF-W2 thus show the same differential bindingcharacteristics as the entire proteins.

(iii) Cross-linking products. A potentially powerful tool forstudying the structure of a protein-DNA complex is cross-linking analysis. We have used formaldehyde to cross-linkthe NF-W1/W-oligo and NF-W2IW-oligo complexes in gelslices derived from a typical gel retardation experiment.After progressively longer times of formaldehyde treatment,the cross-linked complexes were electroeluted and thenelectrophoresed on a sodium dodecyl sulfate-polyacrylamidegel together with protein size markers. Fig. 8 illustrates atypical experiment. The cross-linked NF-W1/W-oligo com-plex migrates as a single, sharp band to a position indicativeof a molecular mass of 64 kDa. The band is faintly visibleafter 5 min of formaldehyde treatment and increases inintensity with longer fixation times. In contrast, the NF-W2/W-oligo complex gives rise to two cross-linked products.One appears to be the same 64-kDa band seen with NF-W1;the other migrates as a diffuse band to a position indicative ofa molecular mass of about 82 kDa. Both bands are evidentafter only 5 min of formaldehyde fixation; the ratio of thelower to the upper band is maximum at the shortest times.The molecular mass value obtained for the cross-linked

NF-W1/W-oligo complex (64 kDa) is in fairly good agree-ment with that calculated after pore gradient gel electropho-resis of the native complex (74 kDa). Likewise, the molec-ular mass of the largest cross-linked NF-W2/W-oligocomplex (82 kDa) is similar to that estimated for the nativecomplex (94 kDa). The range probably reflects differences inthe techniques employed, in particular, variation in thecontribution of the oligo tag when measured by pore gradientversus sodium dodecyl sulfate-polyacrylamide gel electro-phoresis. It is striking, though, that both techniques predictthat NF-W2 is about 20 kDa larger than NF-W1.

DISCUSSIONNF-W1 and NF-W2 display distinct binding properties but

related structural features. We report herein the discovery

and characterization of a pair of proteins that bind to the Wmotif of the murine class II gene E.. One of these proteins(NF-W1) is essentially restricted to mature B cells thatnormally express MHC class II molecules, while the other(NF-W2) occurs in all cell types so far examined.

This is not the first report of a differentially distributed pairof proteins recognizing the same regulatory sequence; theB-cell-specific OTF-2/NF-A2 and the ubiquitously occurring

A. B.NATIVE FORMALDEHYDE CROSS-

COMPLEXES LINKED COMPLEXES

5' 15 30'

W2WlW2W1W2W1

-94

NF-W2 --_ - 8NF-Wl- * -68

-43

-30

-20.1

0*-FIG. 8. Formaldehyde cross-linking of the W complexes. (A)

Native complexes. A standard gel retardation assay was set up usinga purified CH27 NF-W1-NF-W2 fraction containing about 5 x asmuch NF-W1 and NF-W2. The most rapidly migrating complex(unlabeled) is probably a degraded derivative of NF-W2 (Discus-sion). (B) Cross-linked complexes. Bands from a gel retardationassay like that depicted in panel A were excised and treated withformaldehyde for 5, 15, or 30 min. The fixed complexes were thenelectroeluted, precipitated, and electrophoresed on a 12.5% SDSsodium dodecyl sulfate-polyacrylamide gel along with protein mo-lecular mass markers (see list in Materials and Methods).

NF-W1 NF-W2

PK PKF B B¶I 1=

q5 _

VOL. 9, 1989

318 DORN ET AL.

OTF-1/NF-A1 both interact with the immunoglobulin oc-tamer motif, for example (11, 34, 37, 41). But as far as weknow, this is the first report of such a pair of proteinsdisplaying distinct binding properties. NF-W1 and NF-W2interaction with the common GTTGCATC recognition se-quence can be distinguished by at least three criteria: sensi-tivity to A+G methylation, sensitivity to G transversions,and degree of affinity. Apparently, the B-cell-restrictedprotein interacts more strongly with the binding site, perhapsbecause it makes additional contacts.

Despite these distinct binding properties, NF-W1 andNF-W2 show an interesting structural relatedness. Twoindependent methods of size determination lead us to esti-mate that NF-W2 is about 20 kDa larger than NF-W1. Thisdisparity does not hold, however, when one focuses on theDNA-binding domains; digestion by four different enzymesreveals very similar protease-resistant cores for the twoproteins. These cores exhibit (to at least some extent) theaforementioned characteristic binding properties. Furtherevidence of a structural relatedness comes from the formal-dehyde cross-linking experiments: both NF-W1 and NF-W2give rise to a 64-kDa cross-linked product; only NF-W2gives an additional (82-kDa) product. Considering all of thisstructural information, we are prompted to suggest thatNF-W2 is composed of two domains, a DNA-binding seg-ment resembling NF-W1 plus an additional 20-kDa segment.The appendage is most likely a noncovalently bound sub-unit, because NF-W2 gives rise to two easily distinguishablecross-linked products. If so, the two subunits must be rathertightly associated because we never detect the entity similarto NF-W1 in non-B cells and because the kinetics of cross-linking suggest that the two subunits are linked to each otheras fast as or faster than the binding subunit is linked to theDNA (i.e., there is not a clearly progressive conversion ofthe 64-kDa cross-linked product into the 82-kDa product).Our interpretation of the formaldehyde cross-linking studiesis substantiated by UV cross-linking experiments; when thecross-linking agent permits only DNA-protein liaisons, theNF-W2 band gives rise to a single 64-kDa cross-linkedproduct (data not shown).An alternative interpretation of the data is that NF-W1 is

a degradation product of NF-W2 generated in a cell-type-specific fashion. We do not favor this explanation for severalreasons. First, mixing experiments argue against it. Weprepared nuclear extracts simultaneously from LMTK cells,M12 cells, and an equal mixture of LMTK and M12 cells.With the LMTK extract, gel retardation assays showed aratio of NF-W1 to NF-W2 of 0:10; with the M12 extract, theratio was about 7:3; and with extracts from the mixture,there was an intermediate ratio. If M12 cells contain aprotease that converts NF-W2 to NF-W1, we would haveexpected a higher ratio with the extracts from mixed cellsapproaching that with pure M12 extracts. Second, NF-W1and NF-W2 display distinct patterns of contact sites on theW sequence. While it is certainly possible that a degradationproduct can show an altered binding specificity, this has notbeen reported to our knowledge. Conversely, there areseveral reports of degradation products showing the samepattern of contact sites as the intact DNA-binding protein(15). This argument is rendered more convincing by thedemonstration that purposely proteolysed NF-W1 andNF-W2 DNA-binding cores also display distinct patterns ofcontact sites. Third, proteolysis experiments with four en-zymes (proteinase K, pan protease, trypsin, and subtilisin B)show no evidence of a conversion of NF-W2 into NF-W1. Infact, a major proteolysis intermediate of NF-W2 gives rise to

a band in gel retardation assays just below the NF-W1 band,and DNA from this band has a methylation interferencefootprint indistinguishable from that of NF-W2 (Fig. 7, W2panels and Fig. 5, panels A and B). Thus, if proteolysis isinvolved in the conversion of NF-W2 to NF-W1, it must behighly specific to yield a product with a higher affinity for itsbinding site than the native protein.

It is also conceivable that NF-W1 and NF-W2 actuallyderive from the same gene. Alternative splicing or posttrans-lational modifications could then give rise to related butdistinct DNA-binding proteins. Precedents certainly existfor such posttranscriptional events influencing the expres-sion of transcription factors (2, 3, 35, 39).

Relationship between NF-W1, NF-W2, and previously de-scribed DNA-binding proteins. Are NF-W1 and NF-W2 re-lated to previously described DNA-binding proteins specificfor B lymphocytes? The most likely relatives of NF-W1 andNF-W2 are NF-A2/OTF-2 and NF-A1/OTF-1 (11, 34, 37,41). The sites recognized by these two protein pairs arehighly related, GTTGCATC versus ATTTGCAT, and the factthat both sites attract a ubiquitous and a B-cell-specificprotein is quite suggestive. In addition, NF-W1 and NF-A2/OTF-2 have similar kinetics of induction by LPS in 70Z/3cells (Fig. 4A and B). They also have similar molecularmasses (around 60 kDa [37]). Nonetheless, the two proteinpairs are clearly distinguishable. An unlabeled immunoglob-ulin octamer oligo does not compete with a labeled W oligofor the binding of either NF-W1 or NF-W2 (data not shown).In addition, the bands detected in gel retardation assays withthe two oligonucleotides are clearly different (Fig. 4A andB). Finally, some extracts (e.g., M12) contain NF-W1 butnot NF-A2/OTF-2 (data not shown).NF-W1 and NF-W2 are also distinguishable from NF-KB

by a number of criteria. First, an unlabeled B oligonucleotideis unable to compete with a labeled W oligonucleotide for thebinding of either NF-W1 or NF-W2 (data not shown).Second, gel retardation assays show distinctly differentpatterns with the two oligo probes (Fig. 4A and C). Third,the kinetics of induction of NF-W1 and NF-KB by LPS in70Z/3 cells are not the same (Fig. 4A and C). Competitionexperiments also established that NF-W1 and NF-W2 arenot the Pu box-binding protein, the protein which bindsspecifically to the Ephrussi sequences ,u-E3 and K-E3 or theCCAAT-box-binding protein NF-Y (data not shown).

In summary, we describe a pair of proteins which showspecificity for the same DNA sequence, but interact with itin a distinguishable fashion. We have yet to establish exactlywhat role the two proteins play in the regulation of Ea genetranscription. We do know that the W motif is part of acomplex enhancer (21). Can the striking correlation betweenthe cell-type distribution of NF-W1 and NF-W2 and theactivity of the enhancer, which according to its contextexhibits ubiquitous or B-lymphocyte-specific activity, beoverlooked?

ACKNOWLEDGMENTSWe are grateful to P. Gerber for skillful technical assistance, to A.

Staub and F. Ruffenach for oligonucleotides, to the secretarial andphotography services for preparing the manuscript, to R. Hooft vanHuijsduijnen and Xiao-Yan Li for advice, and to R. Accolla, G.Haughton, M. Pierres, and C. Paige for cell lines.A.D. received fellowships from the Deutsche Forschungsge-

meinschaft and the Deutscher Akademischer Austauschdienst. Thiswork was supported by institutional funds from the Institut Nationalde La Santd et de La Recherche Medicale and the Centre Nationalde La Recherche Scientifique and by a grant from the Associationpour la Recherche sur le Cancer to D.M. and C.B.

MOL. CELL. BIOL.

B-CELL-RESTRICTED DNA-BINDING PROTEIN 319

LITERATURE CITED1. Andersson, L.-O., H. Borg, and M. Mikaelsson. 1972. Molecular

weight estimations of proteins by electrophoresis in polyacryl-amide gels of graded porosity. FEBS Lett. 20:199-202.

2. Atchison, M. L., and R. P. Perry. 1987. The role of the K

enhancer and its binding factor NF-KB in developmental regu-lation of K gene transcription. Cell 48:121-128.

3. Baeuerle, P. A., and D. Baltimore. 1988. Activation of DNA-binding activity in an apparently cytoplasmic precursor of theNF-KB transcription factor. Cell 53:211-217.

4. Banerji, J., L. Olson, and W. Schaffner. 1983. A lymphocyte-specific cellular enhancer is located downstream of the joiningregion in immunoglobulin heavy chain genes. Cell 33:729-740.

5. Boss, J. M., and J. L. Strominger. 1986. Regulation of a

transfected human class II major histocompatibility complexgene in human fibroblasts. Proc. Natl. Acad. Sci. USA 83:9139-9143.

6. Briggs, M. R., J. T. Kadonaga, S. P. Bell, and R. Tjian. 1986.Purification and biochemical characterization of the promoter-specific transcription factor, Spl. Science 234:47-52.

7. Dignam, J. D., M. R. Lebovitz, and R. G. Roeder. 1983.Accurate transcription initiation by RNA polymerase II in a

soluble extract from isolated mammalian nuclei. Nucleic AcidsRes. 11:1475-1489.

8. Dorn, A., J. Bollekens, A. Staub, C. Benoist, and D. Mathis.1987. A multiplicity of CCAAT box-binding proteins. Cell 50:863-872.

9. Dorn, A., B. Durand, C. Marfing, M. Le Meur, C. Benoist, andD. Mathis. 1987. Conserved MHC class II boxes -X and Y-are transcriptional control elements and specifically bind nu-

clear proteins. Proc. Natl. Acad. Sci. USA 84:6249-6253.9a.Dorn, A., H. Fehling, W. Koch, M. Le Meur, P. Gerlinger, C.

Benoist, and D. Mathis. 1988. B-cell control region at the 5' endof a major histocompatibility class II gene: sequences andfactors. Mol. Cell. Biol. 8:3975-3987.

10. Falkner, G. G., and H. G. Zachau. 1984. Correct transcription ofan immunoglobulin K gene requires an upstream fragment con-taining conserved sequence elements. Nature (London) 310:71-74.

11. Fletcher, C., N. Heintz, and R. G. Roeder. 1987. Purification andcharacterization of OTF-1, a transcription factor regulating cellcycle expression of a human histone H2b gene. Cell 51:773-781.

12. Foster, J., J. Stafford, and C. Queen. 1985. An immunoglobulinpromoter displays cell-type specificity independently of theenhancer. Nature (London) 315:423-425.

13. Gopal, V. T., T. Shimada, A. W. Baur, and A. W. Nienhuis.1985. Contribution of promoter to tissue-specific expression ofthe mouse immunoglobulin kappa gene. Science 229:1102-1104.

14. Grosschedl, R., and D. Baltimore. 1985. Cell-type specificity ofimmunoglobulin gene expression is regulated by at least threeDNA sequence elements. Cell 41:885-897.

15. Hooft van Huijsduijnen, R., J. Bollekens, A. Dorn, C. Benoist,and D. Mathis. 1987. Properties of a CCAAT box-bindingprotein. Nucleic Acids Res. 15:7265-7281.

16. Huet, J., and A. Sentenac. 1987. TUF, the yeast DNA-bindingfactor specific for UAS,p upstream activating sequences: iden-tification of the protein and its DNA-binding domain. Proc.Natl. Acad. Sci. USA 84:3648-3652.

17. Imler, J. L., C. Lemaire, C. Wasylyk, and B. Wasylyk. 1987.Negative regulation contributes to tissue specificity of theimmunoglobulin heavy-chain enhancer. Mol. Cell. Biol. 7:2558-2567.

18. Ishihara, T., A. Kudo, and T. Watanabe. 1984. Induction ofimmunoglobulin gene expression in mouse fibroblasts by cyclo-heximide treatment. J. Exp. Med. 160:937-942.

19. Kadonaga, J. T., and R. Tjian. 1986. Affinity purification ofsequence specific DNA binding proteins. Proc. Natl. Acad. Sci.USA 83:5889-5893.

20. Kitamura, D., H. Maeda, K. Araki, A. Kudo, and T. Watanabe.1987. Regulation of immunoglobulin gene transcription by labilerepressor factor(s). Eur. J. Immunol. 17:1249-1256.

21. Koch, W., C. Benoist, and D. Mathis. 1989. Anatomy of a new

B-cell-specific enhancer. Mol. Cell. Biol. 9:303-311.

21a.Koch, W., S. Camdeias, J. Guardiola, R. Accolla, C. Benoist,and D. Mathis. 1988. An enhancer factor defect in a mutantBurkitt lymphoma cell line. J. Exp. Med. 167:1781-1790.

22. Little, J. W., and S. A. Hill. 1985. Deletions within a hingeregion of a specific DNA-binding protein. Proc. Natl. Acad. Sci.USA 82:2301-2305.

23. Maniatis, T., S. Goodbourn, and J. A. Fischer. 1987. Regulationof inducible and tissue-specific gene expression. Science 236:1237-1245.

24. Manley, J. L., A. Fire, A. Cano, P. A. Sharp, and M. L. Gefter.1980. DNA-dependent transcription of adenovirus genes in asoluble whole-cell extract. Proc. Natl. Acad. Sci. USA 77:3855-3859.

25. Marzouki, N., S. Camier, A. Ruet, A. Moenne, and A. Sentenac.1986. Selective proteolysis defines two DNA binding domains inyeast transcription factor . Nature (London) 323:176-178.

26. Mason, J. O., G. T. Williams, and M. S. Neuberger. 1985.Transcription cell type specificity is conferred by an immuno-globulin VH gene promoter that includes a functional consensussequence. Cell 41:479-487.

27. Mercola, M., X. F. Wang, J. Olsen, and K. Calame. 1983.Transcription enhancer elements in the mouse immunoglobulinheavy chain locus. Science 221:663-665.

28. Miller, J., A. D. McLachlan, and A. Klug. 1985. Repetitivezinc-binding domains in the protein transcription factor IIIAfrom Xenopus oocytes. EMBO J. 4:1609-1614.

29. Miwa, K., C. Doyle, and J. L. Strominger. 1987. Sequence-specific interactions of nuclear factors with conserved se-quences of human class II major histocompatibility complexgenes. Proc. Natl. Acad. Sci. USA 84:4939-4943.

30. Neuberger, M. S. 1983. Expression and regulation of immuno-globulin heavy chain gene transfected into lymphoid cells.EMBO J. 2:1373-1378.

31. Picard, D., and W. Schaffner. 1984. A lymphocyte-specificenhancer in the mouse immunoglobulin K gene. Nature(London) 307:80-82.

32. Picard, D., and W. Schaffner. 1985. Cell-type preference ofimmunoglobulin K and X gene promoters. EMBO J. 4:2831-2838.

33. Queen, C., and D. Baltimore. 1983. Immunoglobulin gene tran-scription is activated by downstream sequence elements. Cell33:741-748.

34. Rosales, R., M. Vigneron, M. Macchi, I. Davidson, J. H. Xiao,and P. Chambon. 1987. In vitro binding of cell-specific andubiquitous nuclear proteins to the octamer motif of the SV40enhancer and related motifs present in other promoters andenhancers. EMBO J. 6:3015-3025.

35. Santoro, C., N. Mermod, P. C. Andrews, and R. Tjian. 1988. Afamily of human CCAAT-box-binding proteins active in tran-scription and DNA replication: cloning and expression of mul-tiple cDNAs. Nature (London) 332:218-220.

36. Sauer, R. T., C. 0. Pabo, B. J. Meyer, M. Ptashne, and K. C.Backman. 1979. Regulatory functions of the lambda repressorreside in the amino-terminal domain. Nature (London) 279:396-400.

37. Scheidereit, C., A. Heguy, and R. G. Roeder. 1987. Identificationand purification of a human lymphoid-specific octamer-bindingprotein (OTF-2) that activates transcription of an immunoglob-ulin promoter in vitro. Cell 51:783-793.

38. Sen, R., and D. Baltimore. 1986. Multiple nuclear factorsinteract with the immunoglobulin enhancer sequences. Cell 46:705-716.

39. Sen, R., and D. Baltimore. 1986. Inducibility of K immunoglob-ulin enhancer-binding protein NF-KB by a posttranslationalmechanism. Cell 47:921-928.

40. Sherman, P. A., P. V. Basta, and J. P. Y. Ting. 1987. UpstreamDNA sequences required for tissue-specific expression of theHLA-DRa gene. Proc. Natl. Acad. Sci. USA 84:4254-4258.

41. Staudt, L. M., H. Singh, R. Sen, T. Wirth, P. A. Sharp, and D.Baltimore. 1986. A lymphoid-specific protein binding to theoctamer motif of immunoglobulin genes. Nature (London) 323:640-642.

VOL. 9, 1989

320 DORN ET AL. MOL. CELL. BIOL.

42. van Ewlk, W., Y. Ron, J. Monaco, J. Kappler, P. Marrack, M.Le Meur, P. Gerlinger, B. Durand, C. Benoist, and D. Mathis.1988. Compartmentalization ofMHC class II gene expression intransgenic mice. Cell 53:357-370.

43. Wall, R., M. Briskin, C. Carter, H. Govan, A. Taylor, and P.Kincade. 1986. A labile inhibitor blocks immunoglobulin K-light-chain-gene transcription in a pre-B leukemic cell line. Proc.Natl. Acad. Sci. USA 83:295-298.

44. Weinberger, J., D. Baltimore, and P. A. Sharp. 1986. Distinctfactors bind to apparently homologous sequences in the immu-

noglobulin heavy-chain enhancer. Nature (London) 322:846-848.45. Weinberger, J., P. S. Jat, and P. A. Sharp. 1988. Localization of

a repressive sequence contributing to B-cell specificity in theimmunoglobulin heavy-chain enhancer. Mol. Cell. Biol. 8:988-992.

46. Widera, G., L. C. Burkly, C. A. Pinkert, E. C. Bottger, C.Cowing, R. D. Palmiter, R. L. Brinster, and R. A. Flaveli. 1987.Transgenic mice selectively lacking MHC class II (I-E) antigenexpression on B cells: an in vivo approach to investigate Ia genefunction. Cell 51:175-187.