Transcription factor IRF4 controls plasma celldifferentiation and class-switch recombination

Ulf Klein1, Stefano Casola2, Giorgio Cattoretti1, Qiong Shen1, Marie Lia1, Tongwei Mo1, Thomas Ludwig1,Klaus Rajewsky2 & Riccardo Dalla-Favera1

B cells producing high-affinity antibodies are destined to differentiate into memory B cells and plasma cells, but the

mechanisms leading to those differentiation pathways are mostly unknown. Here we report that the transcription factor IRF4 is

required for the generation of plasma cells. Transgenic mice with conditional deletion of Irf4 in germinal center B cells lacked

post–germinal center plasma cells and were unable to differentiate memory B cells into plasma cells. Plasma cell differentiation

required IRF4 as well as the transcriptional repressor Blimp-1, which both acted ‘upstream’ of the transcription factor XBP-1.

In addition, IRF4-deficient B cells had impaired expression of activation-induced deaminase and lacked class-switch

recombination, suggesting an independent function for IRF4 in this process. These results identify IRF4 as a crucial

transcriptional ‘switch’ in the generation of functionally competent plasma cells.

The hallmark of antibody-mediated adaptive immunity is thegeneration of plasma cells producing high titers of antigen-specificantibodies as well as memory B cells destined to become plasmacells after reencounter with antigen. However, the mechanisms thatcontrol the differentiation of germinal center (GC) B cells towardthe plasma cell or memory B cell pathway are not known. Bothcell types derive from antigen-activated B cells that have undergonethe ‘GC reaction’, in which they specifically modify their immunoglo-bulin through somatic hypermutation and class-switch recombination(CSR)1,2. The commitment to plasma or memory cell differentiationis thought to occur in a specific area in the GC structure, the lightzone, where GC centrocytes seem to undergo specific changes intheir transcriptional program. Specifically, centrocytes downregu-late the transcription factor Bcl-6, which is required for GC formationand maintenance3,4, and activate the transcription factors Blimp-1(ref. 5), IRF4 (ref. 6), BMI-1 (ref. 7), NF-kB8, STAT3 (ref. 9)and STAT5 (ref. 10). Although Blimp-1 has been shown to be re-quired for the generation of plasma cells11 and STAT3, for thoseof a specific isotype9, the overall transcriptional programs necessaryfor plasma cell versus memory B cell differentiation have notbeen identified.

Because of its expression in GC centrocytes destined to becomeplasma cells or memory B cells, we have investigated the functionof IRF4 in post-GC B cell differentiation. IRF4 (also called Pip12,LSIRF13, ICSAT14 and MUM1 (ref. 15)) is a member of the inter-feron-regulatory factor family of transcription factors characterizedby a specific DNA-binding domain and the ability to bind toregulatory elements in promoters of interferon-inducible genes16.

Expression of IRF4 is restricted to cells of the immune system,including lymphocytes, dendritic cells and macrophages17,18, inwhich it has been linked to a variety of functions, including prolifera-tion, apoptosis and differentiation17,18. That heterogeneous and often‘opposite’ mode of action may be explained by alternative inter-action with cofactors, including Pu.1 (ref. 12), E47 (ref. 19), STAT-6(ref. 20) and IRF8 (ref. 21), which can influence IRF4 transcriptionalactivity. In the B lineage, IRF4 has a biphasic expression pattern:it is expressed in immature B cells in the bone marrow21 and isabsent from proliferating GC centroblasts6 and then is reexpressedin a subpopulation of centrocytes in the GC and in plasma cells6.Consistent with involvement of IRF4 in early B cell development,mice with germline deletion of IRF4 have a differentiation block atthe transition from the immature to the mature, follicular B cell22

and, as a consequence, lack progeny of the latter cell type, includingGC as well as memory B cells and plasma cells. However, becauseof the developmental block at the immature B cell stage, possibleinvolvement of IRF4 reexpression in post-GC differentiation and inparticular in plasma cell–versus–memory B cell commitmentcannot be investigated in those mice22. Elucidation of the functionof IRF4 in late B cell differentiation is likely to be relevant tounderstanding normal mature B cell development as well as thegenesis of B cell lymphomas. Indeed, IRF4, originally identified alsoas the product of a proto-oncogene involved in chromosomaltranslocations in multiple myeloma, is capable of transforming cellsin vitro15 and is often abnormally expressed in B cell lymphomas6,23,24.To investigate the involvement of IRF4 in post-GC B cell develop-ment, we have generated mice in which Irf4 can be conditionally

Received 15 March; accepted 17 May; published online 11 June 2006; doi:10.1038/ni1357

1Institute for Cancer Genetics, Department of Pathology and Herbert Irving Comprehensive Cancer Center, Columbia University, New York, New York 10032, USA.2CBR Institute for Biomedical Research, Harvard Medical School, Boston, Massachusetts 02115, USA. Correspondence should be to R.D.-F. (rd10@columbia.edu).

NATURE IMMUNOLOGY VOLUME 7 NUMBER 7 JULY 2006 773

A R T I C L E S©

2006

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eim

mun

olog

y

deleted in GC B cells and the fate of the cells carrying the deletion canbe monitored in vivo. We show that IRF4 is critical for plasma celldifferentiation and CSR.

RESULTS

Mice with conditional deletion of Irf4 in GC B cells

To study the involvement of IRF4 in GC and post-GC development,we generated a transgenic mouse strain carrying a loxP-flanked(floxed) Irf4 allele and crossed those mice with mice expressing Crerecombinase specifically in GC B cells. We effected conditionaldeletion of Irf4 by placing loxP sites upstream of the defined promoterregion of Irf4 and downstream of exon 2, containing the translationalstart, respectively (Supplementary Fig. 1 online). To allow tracking ofIRF4-deficient cells after Cre-mediated recombination, we placed agene encoding enhanced green fluorescent protein (eGFP) in anopposite orientation upstream of the Irf4 promoter region, leadingto its expression by the mouse phosphoglycerate kinase promoterplaced in intron 2 (Supplementary Fig. 1). The targeting constructwas flanked by frt sites, which, analogous to the loxP site–Crerecombinase system, are recognized specifically by Flp recombinaseto facilitate the generation of an Irf4-null allele lacking eGFP expres-sion (Supplementary Fig. 1) in mice transgenic for Flp recombinase,

which recombine frt sites early in embryo-genesis25. We identified correctly targetedembryonic stem cell lines by Southern blotanalysis (Supplementary Fig. 1) and testeddeletion of the floxed Irf4 promoter regionwith simultaneous activation of eGFP tran-scription in targeted embryonic stem cells bytransient Cre expression (Fig. 1a). We con-firmed germline transmission of the ‘condi-tional’ Irf4 allele by Southern blot analysis.We crossed the resulting mice heterozygousfor a frt-flanked, floxed Irf4 allele (Irf4frtfl/+

mice; called ‘Irf4fl/+ mice’ here) with Flp-transgenic mice to generate an Irf4-null allele(called ‘Irf4+/– mice’ here; SupplementaryFig. 1). Southern blot analysis of purifiedCD19+ B cells from Irf4fl/–CD19-Cre andIrf4fl/+CD19-Cre mice, in which Cre isunder transcriptional control of the B line-age–restricted gene encoding CD19, showed a7.6–kilobase (kb) band consistent with dele-tion of the floxed region (Fig. 1b), demon-strating the ability of the ‘conditional’ Irf4allele to recombine in vivo. We verified thatthe ‘conditional’ Irf4 allele conferred an Irf4-null phenotype after homozygous deletion(Supplementary Fig. 2 online). We con-firmed the ability of the ‘conditional’ alleleto produce physiological amounts of IRF4,despite the insertion of several gene ‘cassettes’(Supplementary Fig. 1) into the Irf4 locus,by comparing the phenotype and antibodyresponse of Irf4fl/– mice after immunizationwith a T cell–dependent antigen to that ofIrf4+/– and/or Irf4fl/+ mice (SupplementaryFig. 2 and Supplementary Table 1 online).

We then crossed the mice with the ‘condi-tional’ Irf4 allele with Cg1-Cre mice, whichexpress Cre recombinase from a bicistronic

mRNA in which the Cre coding sequence (preceded by an internalribosomal entry site) is downstream of the immunoglobulin heavy-chain g1 constant region (Cg1) coding region, allowing for itsindependent translation26 (Fig. 1c). In these mice, Cre recombinaseis expressed after the induction of germline Cg1 transcription byimmunization with a T cell–dependent antigen, which precedesswitching to IgG1 (ref. 26). In unimmunized mice, less than 2% oftotal splenic B cells show Cre-mediated recombination; this fractionincreases to include 85% of GC B cells (which are rare in unim-munized mice) after immunization26. Cre-mediated recombinationinvolves most IgG1-expressing cells and a small fraction of B cells thathave switched to immunoglobulin classes other than IgG1 (ref. 26). Asthe Cg1 locus is transcribed in a subset of GC B cells, these mice allowanalysis of the function of IRF4 in GC B cells as well as in progenyplasma cells and memory B cells, which are both derived from GC Bcells. Furthermore, cells with deletion of Irf4 can be specificallyidentified in vivo by tracking of eGFP expression (Fig. 1d). Weanalyzed the extent of the Cg1-Cre–mediated deletion of Irf4 in vivoby determining the fraction of eGFP+ versus eGFP– IgG1+ cells in thespleens and peripheral blood of Irf4fl/+Cg1-Cre mice in a T cell–dependent immune response (Fig. 1e). After primary immunization,87.1% ± 2.8% of splenic and 83.6% ± 11.4% of peripheral blood B

a b

c

d e

Mock controlAdeno-Cre

Adeno-Cre:

Floxed15.1 kbWT10.2 kb∆loxP7.6 kb

∆loxP (7.6 kb)WT (10.2 kb)

Floxed (15.1 kb)CD19-enriched: – + – +

∆frt (6.0 kb)

Irf4 fl/+

CD19-Cre

Irf4 mRNA

Irf4

Irf4 Null

in GC andpost-GC B cells

eGFP mRNA

eGFP

All cells

85.6%

IgG1

NP

frt frt

frt

loxP

frt frt

frt

loxP

loxPE1 E2 Cond.

eGFP

Irf4 Null (eGFP+)

Irf4 Null

eGFP P

P

Irf4 fl/–

CD19-Cre

Nhe l, 3′ probe

3′ UTRCreIRES

– +

eGFP

Cγ1-Cre

Cγ1-Cre

Cγ1-Cre

CH1 H CH2 CH3 M1 M2 frt

Cel

l cou

nt

×

×

Figure 1 Functional analysis of the conditionally deleted Irf4 allele. (a) Left, flow cytometry of eGFP

expression by adeno-Cre–treated Irf4fl/+ embryonic stem cells after loxP-mediated deletion of the Irf4

locus. Right, Southern blot analysis of NheI-digested DNA from embryonic stem cell cultures, showing

the expected 7.6-kb band after loxP-mediated deletion (DloxP) along with the 15.1-kb loxP-flanked

allele (Floxed) and the 10.2-kb wild-type allele (WT; Supplementary Fig. 1). (b) Southern blot analysis

of NheI-digested DNA from purified CD19+ and CD19– B cells. CD19-enriched B cells from Irf4fl/+

CD19-Cre mice have the wild-type allele and the allele obtained after loxP-mediated deletion (DloxP)

and those from Irf4fl/–CD19-Cre mice have the allele obtained after loxP-mediated deletion and the

allele obtained after frt-mediated deletion (Dfrt); the faint 7.6-kb band in the CD19– Irf4fl/– sample is

due to cellular contamination of B cells. (c) ‘Knock-in’ of the internal ribosomal entry site (IRES)–Cre

cassette into the Cg1 locus. Black slim rectangles, Cg1 exons. UTR, untranslated region. (d) Status of

Irf4 before (top) and after Cre-mediated recombination in GC B cells of Irf4fl/–Cg1-Cre mice. The

strategy allows identification of Irf4–/– B cells by eGFP expression. Cond., conditional. (e) Flow

cytometry of splenic Thy-1–IgD–IgM–F4/80–Gr-1– cells isolated from NP-KLH–immunized Irf4fl/+Cg1-Cre

mice to assess the fraction of eGFP+ B cells among NP-binding IgG1+ B cells, indicating deletion of

Irf4 (42 d after primary immunization). Right, eGFP expression of all cells (bottom) and of the gated

NP+IgG1+ population (top); number above bracketed line indicates percentage of NP+IgG1+ eGFP+

cells. Representative result of three independent experiments.

774 VOLUME 7 NUMBER 7 JULY 2006 NATURE IMMUNOLOGY

A R T I C L E S©

2006

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eim

mun

olog

y

cells were eGFP+; after secondary immunization, 78.5% ± 4.4% ofsplenic antigen-binding IgG1+ B cells were eGFP+ (Fig. 1e). Theseresults were similar to those obtained after crossing of mice with theCg1-Cre–encoding allele with various reporter mice26 and demon-strate a high efficiency of Cre-mediated deletion of Irf4 in theIrf4–Cg1-Cre double-transgenic mice.

Effect of Irf4 deletion on GC development

We immunized Irf4fl/–Cg1-Cre and Irf4fl/+Cg1-Cre mice with sheep red blood cells andanalyzed GC development on days 10 and14, corresponding to the peak time period ofthe GC reaction, by staining with thelectin peanut agglutinin (PNA) as well asfor CD38 and the B cell marker B220. GCB cells express cell surface antigens with

an as-yet-unknown modified carbohydrate, allowing their recognitionby PNA, which is thus generally used to identify GC B cells27. In mice,CD38 has high expression on circulating naive and memory B cellsand is downregulated at the GC stage28. Irf4fl/–Cg1-Cre and Irf4fl/+Cg1-Cre mice had similar fractions of PNAhiB220+eGFP+ B cells in spleen(35.1% versus 33.5%), which were mostly CD38lo (Fig. 2). In contrast,only approximately 2.0% of PNAlowB220+ cells of both genotypes wereeGFP+. Total numbers of GCs and their histological appearance byimmunostaining analysis were similar in spleens from Irf4fl/–Cg1-Creand Irf4fl/+Cg1-Cre mice (data not shown). These results suggest thatloss of IRF4 expression did not affect GC formation, consistent withthe observation that IRF4+ B cells in GCs are confined to a smallsubset of centrocytes in the light zone in humans6 and mice (data not

35.1%

33.5%

2.2%

2.0%

PNA

Irf4 fl /–

Cγ1-Cre

Irf4 fl /+

Cγ1-Cre

B22

0

eGFP CD38

Figure 2 Development of GC B cells in Irf4fl/–Cg1-Cre mice. Flow cytometry

of spleen cells from Irf4fl/–Cg1-Cre and Irf4fl/+Cg1-Cre mice isolated 14 d

after immunization with sheep red blood cells, to assess B220, PNA,

eGFP and CD38. Histograms show eGFP expression of the gated

PNAhiB220+ or PNAloB220+ populations (middle) and CD38 expression of

the gated eGFP+PNAhi–loB220+ populations (right). Representative result

of experiments using two mice each genotype on days 10, 12 and 14

after immunization.

eGFP

–

eGFP

+

eGFP

–

eGFP

+

Irf4 fl /+

Cγ1-CreIrf4 fl /–

Cγ1-Cre

Irf4

fl/+

Cγ1

-Cre

Irf4

fl/–

Cγ1

-Cre

Irf4 fl /+

Cγ1-Cre

Irf4 fl /–

Cγ1-Cre

Irf4 fl /– Cγ1-Cre Irf4+/– Cγ1-Cre

0.00001

0.0001

0.001

0.01

0.1

1

IgG

1-se

cret

ing

cells

(%

)

0.0001

0.001

0.01

0.1

1

10

Arb

itrar

y un

its

Boost

Primar

y

Preim

mun

e

Boost

Primar

y

Preim

mun

e

Boost

Primar

y

Preim

mun

e

0.001

0.01

0.1

1

10

100

Arb

itrar

y un

itsA

rbitr

ary

units

0.001

0.01

0.1

1

10

100

eGFP

CD

138

Spleen PB BM0.0%0.52%0.0%0.19%0.0%0.58%

0.64% 0.13% 0.14% 0.04% 0.59% 0.06%

a c

NS

NSNS

NP IgM

NS

NS

NS

NSNS

NP IgG2b

NP IgG1<0.01

d

b

Figure 3 Irf4fl/–Cg1-Cre mice lack plasma cells.

(a) Flow cytometry for CD138 expression by

spleen, peripheral blood and bone marrow cells

from Irf4fl/–Cg1-Cre and Irf4fl/+Cg1-Cre mice,isolated 14 d after immunization with sheep

red blood cells (Supplementary Fig. 3, results

obtained with gating on eGFP+ cells). Numbers

above boxes indicate percent CD138+eGFP–

cells (left) or CD138+eGFP+ cells (right). Result

representative of experiments using two mice

each genotype on days 10, 12 and 14 after

immunization. (b) Immunohistochemical analysis

of Bcl-6 (red) and IgG1 (purple; left) and of

Bcl-6 (red) and IgM (purple; right) in spleen

sections from Irf4fl/–Cg1-Cre and Irf4fl/+Cg1-Cre

mice immunized with sheep red blood cells.

Original magnification, �10. (c) Serum response

to NP-KLH immunization of Irf4fl/–Cg1-Cre and

Irf4+/�Cg1-Cre mice. Samples were collected

before immunization (Preimmune) and every

10 d after primary (Primary) and secondary

(Boost) immunization. Sera were tested for thebinding of NP-specific anti-IgG1, anti-IgG2b and

anti-IgM to plates coated with NP23 and BSA;

antibody titers were determined by ELISA as

described3. P values (n ¼ 5) are above bracketed

lines (unpaired Student’s t-test); NS, not

significant. (d) ELISPOT analysis of IgG1

secretion by flow cytometry–isolated eGFP– and

eGFP+ cells of Thy-1–IgM–IgD–Gr-1–F4/80–

splenic cells from Irf4fl/–Cg1-Cre and Irf4fl/+Cg1-

Cre mice 5 d after a boost with NP-KLH. Each

symbol represents the fraction derived from

one mouse.

NATURE IMMUNOLOGY VOLUME 7 NUMBER 7 JULY 2006 775

A R T I C L E S©

2006

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eim

mun

olog

y

shown). Nonetheless, we cannot rule out the possibility that theapparently normal GC formation in Irf4fl/–Cg1-Cre mice was due tothe temporary persistence of IRF4 protein in the GC B cells after Irf4deletion (discussed below). That possibility is unlikely, however, asmost GC B cells are IRF4– (ref. 6), suggesting that IRF4 is dispensablefor the proliferation of centroblasts. Our results here are consistentwith the idea that the lack of GCs in mice with germline Irf4 deletion(Irf4–/– mice22) is most likely secondary to the function of IRF4 inestablishing the naive B cell pool that feeds GC formation.

Plasma cell generation requires IRF4

To investigate whether Irf4–/– GC B cells are able to differentiate intoplasma cells, we immunized Irf4fl/–Cg1-Cre mice and the respectivecontrol mice with sheep red blood cells and analyzed CD138 and B220expression by flow cytometry on days 10 and 14 after immunization(Fig. 3a). The cell surface antigen CD138 (also called syndecan-1) isexpressed by plasma cells and their precursors (plasmablasts) in miceand humans29,30. Although both Irf4fl/–Cg1-Cre and the control micehad similar fractions of eGFP–CD138+ cells (0.58% and 0.64%,respectively; Fig. 3a) that were mostly B220lo–neg (data not shown),Irf4fl/–Cg1-Cre mice completely lacked CD138+eGFP+ cells in thespleen, peripheral blood and bone marrow at all times analyzed(Fig. 3a). Thus, IRF4 deficiency leads to the absence of plasma cellsand plasmablasts.

Immunohistochemical analysis of spleen sections from Irf4fl/–Cg1-Cre and Irf4fl/+Cg1-Cre mice immunized with sheep red blood cells toassess cytoplasmic IgG1 (and Bcl-6, to identify GC location) showedthat in contrast to IgM+ plasma cells, IgG1+ plasma cells were almostcompletely absent from Irf4fl/–Cg1-Cre mice at all time points(Fig. 3b), consistent with a defect in plasma cell formation and,potentially, CSR (discussed below). There were only very rare cellsexpressing IgG1 in the cytoplasm of Irf4fl/–Cg1-Cre mice and thosecells were mostly confined to GCs. Those cells probably representedplasma cells that failed to delete Irf4, given that Cre-mediated deletiondoes not occur in every GC B cell.

In agreement with the flow cytometry and immunostaining results,there was much less IgG1 specific for the hapten NP (4-hydroxy-3-nitrophenylacetyl) in the serum of Irf4fl/–Cg1-Cre mice than in that ofIrf4+/–Cg1-Cre control mice after a ‘boost’ immunization with NP

coupled to keyhole limpet hemocyanin (NP-KLH), whereas these micehad similar titers of other immunoglobulin classes (Fig. 3c). Inaddition, enzyme-linked immunospot (ELISPOT) analysis of spleeneGFP– and eGFP+ cells isolated 5 d after a secondary immunizationwith NP-KLH showed that Irf4fl/–Cg1-Cre mice had almost no IgG1-secreting cells in the sorted eGFP+ fractions (Fig. 3d). However, thesemice had percentages of eGFP– IgG1-secreting cells in the same rangeas that of the Irf4fl/+Cg1-Cre control mice. These cells most likelyrepresent plasma cells that have switched to IgG1 but have failed todelete Irf4. The last observation is consistent with the ability of Irf4fl/–

Cg1-Cre mice to mount a limited IgG1 response after a ‘boost’immunization (Fig. 3c, top). Consistent with the ELISPOT results,immunofluorescence analysis for IgG1 in cells ‘spun down’ onto slidesfrom the sorted eGFP+ and eGFP– fractions demonstrated an absenceof cytoplasmic IgG1+ cells among eGFP+ cells of Irf4fl/–Cg1-Cre mice(data not shown). These results collectively indicate that IRF4 isrequired for plasma cell formation.

IRF4 and Blimp-1 act ‘upstream’ of transcription factor XBP-1

Expression of Blimp-1 and XBP-1, a transcription factor that acts‘downstream’ of Blimp-1 to establish the secretory apparatus of plasmacells31, are essential for plasma cell development11,32. To elucidate therelationship of IRF4 to those factors in plasma cell physiology, we usedlipopolysaccharide (LPS) stimulation to induce plasmablastic differ-entiation of B lymphocytes in vitro. Semiquantitative PCR with twodifferent primer pairs and immunoblot analysis suggested that thegene encoding Blimp-1 was upregulated to a similar extent in wild-type B cells and in B cells with homozygous germline deletion of Irf4(Irf4–/–; Fig. 4a,b). However, differentiation into CD138+B220lo

plasma cells did not occur in Irf4–/– cultures (Fig. 4c). This defectwas not a consequence of the different developmental stages of B cellsin wild-type versus Irf4–/– mice22, as mature B cells in which Irf4 wasdeleted in vitro by TAT-Cre (Cre attached to a hydrophobic peptide(HIV-TAT) to promote cellular uptake)33 were also unable to differ-entiate after LPS treatment (Fig. 4d). Thus, in the absence of IRF4,Blimp-1 expression was not sufficient for plasma cell differentiation.

The establishment of the secretory apparatus requires the processingof Xbp1 mRNA to generate active XBP-1 protein. Immunoblotanalysis for XBP-1 showed that the active form was not produced in

Figure 4 Response of Irf4–/– B cells to LPS

stimulation. (a) Semiquantitative PCR analysis

of transcripts of the genes encoding Blimp-1

(Prdm1) and glyceraldehyde phosphate

dehydrogenase (Gapdh) in B cells enriched from

Irf4–/– and wild-type (Irf4+/+) mice, after isolation

(0 d) and at day 3 of LPS stimulation (3 d).

Wedges indicate sequential 1:4 dilutions. E1–E5and E6–E8, exons of Prdm1 where PCR primers

hybridize. (b) Immunoblot analysis of Blimp-1,

XBP-1 and b-actin expression by ex vivo fractions

(LPS –; day 0) and LPS-stimulated fractions

(LPS +; day 3) from Irf4–/– and wild-type (Irf4+/+)

mice. (c) Flow cytometry of CD138 and B220

expression by LPS-stimulated (day 3) B cell–

enriched fractions from Irf4–/– and wild-type

(Irf4+/+) mice. Numbers above boxes indicate

percent B220loCD138+ cells. (d) Flow cytometry

of CD138 expression by LPS-stimulated (day 3)

B cell–enriched fractions from Irf4fl/– and Irf4fl/+

mice, transduced with TAT-Cre (right) or left untreated (left). TAT-Cre transduction results in the excision of the floxed Irf4 promoter region with simultaneous

activation of eGFP expression (Fig. 1d), thus allowing identification of the resulting eGFP+ Irf4–/– or eGFP+ Irf4+/– cells by flow cytometry. Numbers in

quadrants indicate percent cells in each. The experiment was done twice.

CD138

B22

0 53.6%28.8%0.1%71.2%

7.1%10.5%0.1%28.6%

69.4% 0.2% 57.9% 19.8%

0.8%21.6%0.1%30.3%

eGFP

CD

138

LPS day 3TAT-Cre

LPS day 3No TAT-Cre

Irf4 fl /+

Irf4–/–

Irf4–/–

Irf4–/–Irf4+/+

Irf4+/+

Irf4+/+

Irf4 fl /–

1.9%14.2%

β-actin

Gapdh

XBP-1

Blimp-1LPS – +– +

Prdm1 (E6–E8)Prdm1 (E1–E5)

LPS 0 d 3 d 0 d 3 da d

b

c

776 VOLUME 7 NUMBER 7 JULY 2006 NATURE IMMUNOLOGY

A R T I C L E S©

2006

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eim

mun

olog

y

Irf4–/– B cells after LPS stimulation (Fig. 4b). These observationscollectively indicate that Blimp-1 and IRF4 are nonredundant, criticalcomponents acting ‘upstream’ of XBP-1 in plasma cell differentiation.

IRF4-deficient B cells fail to undergo CSR

In vitro stimulation of B cells by CD40 and/or interleukin 4 (IL-4),which induces Cg1 ‘sterile’ transcripts and CSR to IgG1 (refs. 34,35)but does not induce plasma cell differentiation, strongly upregulatesIRF4 expression in humans8,20 and mice (data not shown). Thatobservation suggests that IRF4 may have a function in CSR indepen-dent of its function in plasma cell differentiation. Initial experimentsinvolving CD40 and IL-4 stimulation of B cells from Irf4fl/–Cg1-Creshowed that these mice had a smaller fraction of class-switched cellsthan did the Irf4fl/+Cg1-Cre control mice (7.4% versus 26.3%, respec-tively; Supplementary Fig. 4 online). However, we did not considerthat result informative, as CSR is a rapid process detectable at themolecular level as early as 3 h after induction36, whereas a longer timeis likely required for complete disappearance of IRF4 protein aftergene deletion and IRF4 protein decay. That idea was supported by thefinding that Cg1-Cre–mediated deletion of Xrcc4, which encodes a

protein required for CSR, also produced a substantial frequency ofclass-switched B cells in vitro (F.W. Alt, personal communication). Wetherefore investigated the function of IRF4 in CSR by analyzing B cellsfrom Irf4fl/– mice in which IRF4 was previously deleted by TAT-Cretransduction. Of those cells, only 0.9% expressed IgG1, compared with10.9% for the Irf4fl/+ control mice (Fig. 5a, right), suggesting that IRF4was required for the production of class-switched B cells. We obtainedsimilar results with B cells from Irf4–/– mice, which acted like TAT-Cre-treated Irf4fl/� B cells (data not shown). Although the slowerproliferation rate of Irf4–/– B cells22 (Supplementary Fig. 4) preventedanalysis of CSR to isotypes that require multiple cell divisions37,38,carboxyfluorescein diacetate succinimidyl diester analysis of cellsstimulated with CD40 plus IL-4 showed that the percentage ofIgG1+ cells was much lower in Irf4–/– than in wild-type culturesafter the same number of cell divisions (Supplementary Fig. 4),excluding the possibility of defective proliferation as a cause ofdefective CSR in Irf4–/– cells.

To investigate at which stage CSR was inhibited, we examined Irf4–/–

B cells stimulated with either CD40 or CD40 plus IL-4 for theoccurrence of I exon (I) g1-Cg1 germline transcripts as an indication

Day 4Day 0Day 0 Day 4

Arb

itrar

y va

lues

β-actinAID

Irf4–/–

Irf4–/–Irf4+/+

Irf4 fl /+

Irf4 fl /–

Irf4+/+

GapdhAicda

PST (Iµ-Cγ1)CT (Iγ1-Cµ)

GLT (Iγ1-Cγ1)

CD40+IL-4CD40+IL-4 CD40CD40

40.1%39.0%59.6%0.1%59.1%0.1%

eGFP

IgG

1

99.0%

0.9% 0.0% 29.6% 0.0%

40.1%

0.2% 0.2% 9.9% 10.9%

32.4%54.5%41.9%0.1%73.1%0.1%

99.6%

0.3% 0.0% 26.8% 0.0%

57.9%

0.2% 0.1% 12.2% 0.9%CD40 + IL-4CD40CD40 + IL-4CD40

No TAT-Cre TAT-Crea b

c

Figure 5 Response of Irf4–/– B cells to stimulation with CD40 plus IL-4. (a) Flow cytometry of IgG1 expression by B cell–enriched fractions from Irf4fl/– and

Irf4fl/+ mice at day 5 of culture; cells were stimulated with CD40 or with CD40 plus IL-4 (IL-4 added at day 1 of culture) and were transduced with TAT-Cre

(right) or left untreated (left). Numbers in quadrants indicate percent cells in each. The experiment was done twice. (b) Semiquantitative PCR analysis of

germline transcripts (GLT), circle transcripts (CT), post-switch transcripts (PST), Aicda and Gapdh in B cells enriched from Irf4–/– and wild-type (Irf4+/+)

mice at day 4 of stimulation with CD40 or CD40 plus IL-4. Wedges indicate sequential 1:4 dilutions. The CD40-stimulated Irf4–/– fraction has less Cg1

GLT than does the comparable wild-type fraction, compensated for by the addition of IL-4. (c) Immunoblot analysis (top) of AID expression by B cells

enriched from Irf4–/– and wild-type mice, at day 0 and day 4 of stimulation with CD40 or CD40 plus IL-4. Bottom, AID protein abundance normalized to

b-actin abundance.

Figure 6 Generation of antigen-specific memory

B cells by Irf4fl/–Cg1-Cre mice. (a) Flow cytometry

of IgG1 and NP binding by peripheral blood

B cells at day 14 from Irf4fl/–Cg1-Cre and

Irf4fl/+Cg1-Cre mice immunized with NP-KLH or

KLH only. Gates were set on IgM–IgD–Gr-1–F4/80–

cells to enrich for class-switched NP-binding

cells. Far right, eGFP expression of the gated

populations. Two mice of each genotype were

analyzed on day 14 and on day 42 (data notshown). (b) Flow cytometry of IgG1 and NP

binding of peripheral blood B cells at day 21

from NP-KLH–immunized Irf4fl/–Cg1-Cre and

Irf4fl/+Cg1-Cre mice (11; left) and 5 d after a

secondary immunization at day 16 (21; right). Cells were isolated and analyzed as described in a. The NP+IgG1+ cells were mostly eGFP+ (data not shown).

Numbers above boxes indicate NP+IgG1+ cells. PE, phycoerythrin; APC, allophycocyanin.

IgG1-APC

NP

-PE

Irf4 fl /+

Cγ1-CreIrf4 fl /+

Cγ1-Cre

Irf4 fl /–

Cγ1-CreIrf4 fl /–

Cγ1-Cre

0.86%

0.44%

0.18%

0.04%

eGFP

eGFP

IgG1-APC

NP

-PE

0.03%0.01%

0.01% 0.03%

Day 5 2° NP-KLHDay 21 1° NP-KLHNP-KLHKLH onlya b

NATURE IMMUNOLOGY VOLUME 7 NUMBER 7 JULY 2006 777

A R T I C L E S©

2006

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eim

mun

olog

y

that the g1 locus is accessible for the CSR machinery as well as forIg1-Cm circle transcripts and Im-Cg1 post-switch transcripts as ‘read-outs’ of active CSR35,36 (Fig. 5b). Although Irf4–/– B cells transcribedthe g1 locus after treatment with CD40 plus IL-4, circle and post-switch transcripts were considerably reduced (Fig. 5b). This resultindicates that IRF4 was required for CSR ‘downstream’ of Cg1germline transcription.

To investigate the mechanism by which IRF4-deficient cells fail toundergo CSR, we assessed whether expression of activation-inducedcytidine deaminase (AID), a molecule critical for CSR35, was affectedin Irf4–/– cells. We therefore analyzed the induction of Aicda mRNA(encoding AID) and AID protein by RT-PCR and immunoblotanalysis, respectively, in B cells stimulated with CD40 plus IL-4 (Fig.5b,c). The results suggested that Irf4–/– B cells were impaired inupregulating AID expression, most likely at the transcriptional level.As AID is required for CSR, this finding suggests that IRF4 maycontrol CSR at least in part by inducing AID expression. As B cellsstimulated with CD40 plus IL-4 do not upregulate Blimp-1 or acquireplasmablast characteristics39,40 (data not shown), the function ofIRF4 in CSR seems to be independent of its function in plasmacell differentiation.

IRF4 in the generation of memory B cells

As most memory B cells in mice express a switched immunoglobulinisotype, the identification of memory B cells seemed problematic inIrf4fl/–Cg1-Cre mice, because IRF4 is required for CSR. However, theearly execution of CSR permitted by the temporary persistence of theIRF4 protein (as described above) allowed us to investigate whetherIRF4 was required for the generation of memory B cells early in theresponse. Indeed, eGFP+ cells with features typical of immunoglobu-lin-switched memory B cells were detectable in Irf4fl/–Cg1-Cre miceearly after immunization. Mononuclear cell fractions derived fromspleens and peripheral blood of Irf4fl/–Cg1-Cre mice had anIgG1+eGFP+ population (Supplementary Fig. 5 online). In peripheralblood, the eGFP+ cells were B220+, CD138–, PNAlo and CD38+

(Supplementary Fig. 5), a phenotype typical of memory B cells30.

We investigated the generation of antigen-specific memory B cells inthe Irf4fl/–Cg1-Cre mice by NP-KLH immunization, which, in contrastto immunization with a complex antigen such as sheep red blood cells,allows analysis of an antigen-specific immune response at the level ofserum (titers of antibody to NP (anti-NP); discussed above), cell (NPbinding) and rearranged antibody gene (somatic hypermutation). At 2weeks after immunization with the T cell–dependent antigen NP-KLH, Irf4fl/–Cg1-Cre mice had the same percentage of NP-bindingIgG1+ B cells in the peripheral blood as did the control Irf4fl/+Cg1-Cremice, and these cells were mostly eGFP+ (Fig. 6a). Sequence analysisof variable heavy-chain region (VH186.2)-Cg1 transcripts amplifiedfrom B220+CD138–eGFP+ peripheral blood B cells showed a slightlylower VH mutation frequency (0.7% versus 1.2%; P o 0.01) andpercentage of rearrangements carrying a tryptophan-to-leucine sub-stitution at position 33, indicative of selection of the corresponding Bcell for improved NP binding41–43 (16% versus 32%), in Irf4fl/–Cg1-Cre mice than in Irf4fl/+Cg1-Cre control mice (Supplementary Table 2and Supplementary Fig. 5). However, compared with that of controlmice, the NP-binding cell population was lower in Irf4fl/–Cg1-Cre miceat day 42 after immunization, as suggested by the much lower fractionof eGFP+ NP-binding cells in spleen and peripheral blood (data notshown). Consistent with that observation, fewer different VH186.2-rearrangements were amplified from B220+CD138–eGFP+ peripheralblood B cells of Irf4fl/–Cg1-Cre mice than for control mice (1–3 versus17–20, with 0.8% versus 2.3% IgV mutation; P o 0.01; Supplemen-tary Table 3 and Supplementary Fig. 5). These findings suggest thatin contrast to the complete lack of plasma cell formation, an initialproduction of switched memory B cells did occur in the Irf4fl/–Cg1-Cre mice, consistent with the idea that IRF4 may not be required formemory B cell generation (discussed below). However, that popula-tion was maintained only partly over time, raising the possibility thatIRF4 is involved in the maintenance of the memory cell pool. IRF4,whose RNA has low expression in memory B cells44, may have aprosurvival function in these cells, as reported for CD4+ T cells45.

The initial production of memory B cells in Irf4fl/–Cg1-Cre miceallowed us to investigate whether IRF4 is required for their prolifera-tion after renewed antigen encounter. At 5 d after a secondaryimmunization with NP-KLH, Irf4fl/–Cg1-Cre and Irf4fl/+Cg1-Cremice had fractions of NP-binding IgG1+ cells in spleen that werefive- to tenfold higher than those of control mice that received onlyprimary immunization (Fig. 6b). This result indicates that thesecondary immunization led to a reactivation and population expan-sion of pre-existing NP-specific memory B cells in the mutant miceand therefore IRF4 was not required for memory B cell proliferation.

IRF4 is required for memory–to–plasma cell differentiation

To investigate whether Irf4–/– memory B cells in the Irf4fl/–Cg1-Cremice were able to differentiate into plasma cells after antigen read-ministration, we determined whether these mice could develop plasmacells deriving from memory B cells 5 d after secondary immunizationwith NP-KLH. Although control mice developed a distinct populationof eGFP+IgG1+CD138+ plasmablasts in the spleen (16.9%; Supple-mentary Fig. 6 online) that bound NP and were CD138+ and B220lo

(Fig. 7), we did not find such cells in the Irf4fl/–Cg1-Cre mice (Fig. 7and Supplementary Fig. 6). These results suggest that the Irf4–/–

eGFP+ B cells present in Irf4fl/–Cg1-Cre mice proliferated after anti-genic restimulation (Fig. 6b) but were unable to differentiate intoplasma cells. The last conclusion was supported by the ELISPOTresults (Fig. 3d) showing a lack of IgG1 secretion by restimulatedIrf4–/– eGFP+ B cells. Thus, IRF4 was required for both ‘de novo’ and‘post–memory plasma cell’ differentiation.

Irf4 fl /–

Cγ1-Cre

Irf4 fl /+

Cγ1-Cre

IgG1-APC IgG1-APC eGFP

B22

0-P

erC

P

CD

138-

Per

CP

NP

-PE

7.7%9.6%

0.0%

38.8%

31.3%4.4%

Figure 7 Lack of memory–to–plasma cell differentiation in Irf4fl/–Cg1-Cre

mice. Flow cytometry of splenic Thy-1–IgD–IgM–F4/80–Gr-1– cells from

Irf4fl/–Cg1-Cre and Irf4fl/+Cg1-Cre mice, isolated by magnetic cell separation

5 d after a secondary immunization at day 16 and assessed for NP binding

(NP-phycoerythrin conjugate (NP-PE)) and for expression of IgG1 and either

CD138 or B220, gated on eGFP+ cells. Non-lymphocytes were excluded

from the analysis by forward- and side-scatter gating. Numbers above ovals

indicate percent CD138+ IgG1+ cells (middle) or B220lo–neg eGFP+ cells(right) in the NP+IgG1+ populations outlined at left (numbers above

rectangles indicate percent NP+IgG1+ cells). PerCP, peridinine chlorophyll

protein complex. Representative result of experiments done using four mice

each genotype.

778 VOLUME 7 NUMBER 7 JULY 2006 NATURE IMMUNOLOGY

A R T I C L E S©

2006

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eim

mun

olog

y

DISCUSSION

Prompted by the biphasic pattern of expression of IRF4 and inparticular its reexpression in a subset of GC centrocytes, we investi-gated the function of IRF4 in GC formation and plasma versusmemory B cell differentiation. Our results indicate that IRF4 maynot be required for GC and memory B cell generation, whereas it hascritical functions in two post-GC processes: the generation of plasmacells both from GC and memory B cells; and CSR.

The high expression of IRF4 in plasma cells and its involvement inchromosomal translocations in plasma cell–derived tumors6,15 havelong suggested its involvement in the physiology of antibody-secretingcells. The findings presented here suggest a critical function for thistranscription factor in the generation of plasma cells. LPS induces theexpression of IRF4, Blimp-1, XBP-1 and surface CD138 and immuno-globulin secretion. In LPS-stimulated Irf4–/– B cells, Blimp-1 wasupregulated at both the transcriptional and protein level. It has beenshown before that Irf4 mRNA is not upregulated in LPS-treatedBlimp-1-deficient B cells46 and that the retroviral introduction ofBlimp-1 into a Blimp-1-deficient cell line leads to IRF4 upregula-tion47. Although those observations suggest that IRF4 is ‘downstream’of Blimp-1 in plasma cell differentiation, we found that despite theupregulation of Blimp-1, LPS-treated Irf4–/– B cells failed to acquire aplasmablastic phenotype or to produce the active form of XBP-1.Thus, our results indicated that Blimp-1 was not sufficient for plasmacell differentiation, but that both Blimp-1 and IRF4 were necessaryfor this process and that both transcription factors acted ‘upstream’of XBP-1.

We were unable to conclusively determine the precise function ofIRF4 in the generation of memory B cells, as the abrogation of CSR inIRF4-deficient B cells eliminates a key molecular marker for theidentification of memory B cells in mice. Because of the ‘architecture’of the Irf4fl/–Cg1-Cre genotype, however, some IRF4 is still present atthe time of the first induction of CSR, thus allowing for the generationof some IRF4-deficient (eGFP+) switched B cells, which have all thephenotypic characteristic of memory B cells, including distribution,morphology (small), immunophenotype (CD38hiCD138–) and anti-gen specificity. These cells may indeed represent early GC emigrants30,as suggested by the fact that the VH rearrangements derived fromIrf4fl/–Cg1-Cre mice consistently had fewer somatic mutations thandid those of the proficient control mice at all time points analyzed.The presence of this IgG1+Irf4–/– cell population and the contrast withthe complete absence of Irf4–/– plasma cells suggest that IRF4 was notrequired for the generation of the cellular phenotype of memoryB cells. Nonetheless, we cannot formally exclude the possibility thatthis cell population may also be generated by the temporary persis-tence of IRF4 at the time when the signals leading to memory B celldifferentiation are delivered. Although conclusive demonstration ofthe function of IRF4 in memory B cell generation will require theengineering of mice capable of a more defined temporal separationbetween Irf4 deletion and CSR, our results are consistent with the ideathat IRF4 is not required for the proliferation of memory B cells, asIrf4–/– (eGFP+) memory B cells undergo population expansion in vivoin response to antigen. Conversely, our results indicate that IRF4 isrequired for the differentiation of memory cells to plasma cells, asdemonstrated by the complete absence of (eGFP+) plasma cells aftersecondary immunization. Thus, IRF4 is critical for the generation ofthe plasma cell phenotype from both GC and memory B cells.

Our results suggest involvement of IRF4 in CSR, as Irf4–/– B cellsstimulated with CD40 and IL-4, a combination of stimuli that stronglyupregulates IRF4 (ref. 20) and induces class switching in wild-typeB cells34,35, developed almost no IgG1+ cells. As Irf4–/– B cells

stimulated with CD40 plus IL-4 expressed ‘sterile’ Cg1 transcriptsbut very few circle or switch transcripts, the developmental block inIRF4-deficient cells is probably after the immunoglobulin locus ismade accessible for CSR but before the recombination mechanism isactually activated. Thus, IRF4 is probably involved as a cofactor in asignaling pathway that induces the expression of genes encoding CSRfactors. One of these factors may be AID, whose expression wassubstantially decreased in Irf4–/– cells after CSR-inducing stimuli. Thereduced AID expression in Irf4–/– cells may partly be mediated by thetranscriptional repressor Id2 (ref. 48), as we noted fivefold highermRNA expression of Id2 in Irf4–/– cells than wild-type cells afterstimulation with CD40 plus IL-4 in a DNA microarray analysis (UKand R.D.-F., unpublished observations). Most GC B cells are IRF4–

(refs. 6,49), including centroblasts in which somatic hypermutation, aprocess that also requires AID expression35, is thought to occur1. Thatwould indicate that the function of AID in somatic hypermutationand in CSR could be independently controlled by distinct transcrip-tional programs, possibly involving IRF8 in somatic hypermutation50

and IRF4 in CSR. Alternatively, IRF4 could have low expressionundetectable by immunohistochemistry throughout the GC, but itsupregulation could be critical for CSR. The finding that the functionof IRF4 in CSR does not require ‘upstream’ regulation of Blimp-1 andcan occur independently of plasma cell differentiation suggests thatits function in CSR is distinct from that in naive B cell22 and plasmacell differentiation.

Finally, our results have implications for the function of IRF4 inlymphomagenesis. Indeed, a substantial fraction of diffuse large celllymphomas show deregulated expression of IRF4 (ref. 6). Thatderegulation may contribute to malignancy by constitutively promo-ting CSR, which in turn may lead to chromosomal translocations.

METHODSGeneration of Irf4fl/–Cc1-Cre mice. The vector to target Irf4 was a derivative of

pEASY-FLOX (M. Alimzhanov (Harvard Medical School, Boston, Massachu-

setts) and K.R.) containing a phosphoglycerate kinase–neomycin-resistance

poly(A) cassette, the herpes simplex virus thymidine kinase gene, two loxP sites,

as well as two frt sites, a promoterless gene encoding eGFP in the opposite

orientation, a triple simian virus 40 poly(A) site in inverse orientation

generated by PCR and joining of the fragments, a phosphoglycerate kinase

promoter in the opposite orientation, and multiple unique restriction sites for

cloning Irf4 segments corresponding to the homology arms (Supplementary

Fig. 1). The vector was constructed so that in the appropriate cell types, Cre-

mediated recombination produced deletion of the Irf4 promoter region with

simultaneous activation of eGFP expression and Flp-mediated recombination

produced deletion of the whole targeted region, including the eGFP gene

(Supplementary Fig. 1). Successively inserted into the cloning sites of the

vector were three DNA fragments of the 129/Sv Irf4 locus comprising the

following: 2.3 kb of the region upstream of the Irf4 promoter region; the Irf4

promoter region up to 300 bp downstream of exon 2 containing the transla-

tional start site; and 4.7 kb of the region downstream of exon 2. The linearized

vector was electroporated into W9.5 embryonic stem cells derived from 129/

SvEvTac, and correctly targeted embryonic stem cell colonies were identified by

Southern blot analysis after selection with gancyclovir and G418 (Supplemen-

tary Fig. 1). Chimeras were obtained after injection of the targeted embryonic

stem cell clones into blastocysts derived from C57BL/6 mice. From two of the

chimeras, derived from independent embryonic stem clones and bred with

C57BL/6 females, we obtained mice with the ‘conditional’ Irf4 allele in germ-

line, as determined by Southern blot analysis. The chimeras were also crossed

with C57BL/6CAGGS-Flpe mice25 to generate a frt-null allele (Supplementary

Fig. 1). Cg1-Cre mice have been described26. All experimental mice were on a

129/Sv–C57BL/6 mixed background. Mice were housed and treated according

to the guidelines of the Institute of Comparative Medicine at Columbia

University (New York).

NATURE IMMUNOLOGY VOLUME 7 NUMBER 7 JULY 2006 779

A R T I C L E S©

2006

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eim

mun

olog

y

Immunization. Mice 8–12 weeks of age were used for immunization experi-

ments. For study of the anti–sheep red blood cell response, mice were

immunized intraperitoneally with 1 � 105 sheep red blood cells in PBS

(Cocalico Biologicals) followed by analysis of spleens, peripheral blood and

bone marrow. For study of the anti–NP-KLH response, mice were immunized

intraperitoneally with NP20-KLH (100 mg per mouse) or KLH only in complete

Freund’s adjuvant. In some experiments, mice were given a secondary immu-

nization with NP20-KLH (100 mg per mouse) in incomplete Freund’s adjuvant.

Then, 14 d and 42 d (and, in the recall immunization experiments, 5 d) after a

secondary immunization at day 16 or 42, peripheral blood and/or spleens were

removed for analysis.

B cell isolation. For spleen cell suspensions, spleens were minced with glass

slides, followed by the removal of erythrocytes by hypotonic lysis. Blood was

withdrawn from the hearts with a heparin-coated syringe. Peripheral blood was

diluted with PBS to 2 ml and was layered on 2 ml Histopaque-1083 (Sigma) in

15-ml tubes. After centrifugation for 20 min at 550g, the interphase was

removed and was washed with 0.5% BSA in PBS. The pellet was resuspended in

1 ml and the gradient centrifugation was repeated. Bone marrow cells were

obtained by flushing of the femurs with 0.5% BSA in PBS.

For enrichment for B cells in the NP immunization experiments, CD19+ B

cells from spleen or peripheral blood were purified by magnetic cell separation

with the MACS system (Miltenyi Biotec) using CD19 microbeads and LS

columns. For experiments with LPS and CD40 stimulation, samples were

enriched for B cells by magnetic depletion of non–B cells (T cells, macrophages,

natural killer cells and plasma cells) with the ‘untouched’ B cell isolation kit

(Miltenyi Biotec) and LS columns. For analysis of the phenotype of antigen-

stimulated memory B cells arising in the recall response, samples were enriched

for class-switched and CD138+ plasma cells by magnetic depletion of T cells,

non–B cells and IgD+ and IgM+ B cells by staining of splenic B cells with

biotinylated antibodies to F4/80 (clone A3-1; Serotec), Thy-1 (clone 53-2.1),

Gr-1 (clone RB6-8C5), IgM (clone R6-60.2; all from BD Pharmingen) and IgD

(clone 11-26; Southern Biotechnology), followed by incubation with strepta-

vidin microbeads (Miltenyi Biotec). Magnetically labeled fractions were passed

over LS columns.

The eGFP+ and eGFP– B cell subpopulations were isolated by flow cytometry

on a FACSAria (Becton Dickinson) after staining for the respective surface

molecules (described below). Stringent gates were set to minimize cellular

contamination. Cells were sorted into 0.5% BSA in PBS and cell pellets were

lysed with TRIzol reagent (Invitrogen Life Technologies) for RNA isolation.

Flow cytometry. The following antibodies and labels were used on single-cell

suspensions from spleen, peripheral blood, bone marrow and cultured B cells:

allophycocyanin-conjugated anti-B220 (clone RA3-6B2); peridinin chlorophyll

protein complex–conjugated anti-B220; allophycocyanin-conjugated anti-IgG1

(clone X56) and anti-IgM; fluorescein isothiocyanate–conjugated anti-CD21

(clone 7G6); phycoerythrin-conjugated anti-CD23 (clone B3B4), anti-CD38

(clone 90) and anti-CD138 (clone 281-2; all from BD Pharmingen); and

phycoerythrin-conjugated anti-IgD (Southern Biotechnology). Biotinylated

PNA (Vector Laboratories) and anti-CD138 were developed with

streptavidin–peridinin chlorophyll protein complex (both BD Pharmingen).

The NP-phycoerythrin conjugate (Biosearch Technologies) was used at a

dilution of 1:200 for staining. Rat monoclonal anti-mouse CD16-CD32 (Mouse

BD Fc Block; BD Pharmingen) was used to block FcgIII/II receptors before the

addition of antibodies. For enrichment for NP-binding B cells in flow

cytometry by ‘gating out’ of non–class-switched cells, CD19+ cell suspensions

enriched by magnetic-activated cell sorting were stained with biotinylated

antibodies to Gr-1, F4/80, IgM and IgD and were developed with streptavi-

din–peridinin chlorophyll protein complex. Data were acquired on a FACSCa-

libur (Becton Dickinson) using CELLQuest software.

Carboxyfluorescein diacetate succinimidyl diester analysis. The proliferation

capacity of Irf4–/– and wild-type B cells purified by magnetic cell separation and

stimulated with CD40 plus IL-4 and LPS was assessed by flow cytometry using

staining with the CellTrace CFSE Cell Proliferation Kit (Molecular Probes;

Invitrogen) according to the manufacturer’s recommendations. Complete

labeling of the fractions was confirmed by flow cytometry at day 0.

B cell culture. Purified B cells from Irf4–/– mice and wild-type littermates, and

from Irf4fl/– and Irf4fl/+ mice for the TAT-Cre experiments, were cultured at a

density of 1 � 106 cells/ml in RPMI medium plus 10% FBS as follows: for 3–5 d

with 20 mg/ml of LPS or for 4–5 d with 1 mg/ml of anti–mouse CD40 (clone

HM40-3; BD Pharmingen) with or without 15 ng/ml of IL-4 (R&D Systems).

Adeno-Cre transduction. Irf4fl/+ embryonic stem cells were incubated for 1 h

with or without adeno-Cre (a gift from J. Kitajewsky, Columbia University,

New York) at a previously determined optimal multiplicity of infection in

DMEM with 2.5% FBS. Cultures were analyzed for eGFP expression by flow

cytometry 48 h after transduction.

TAT-Cre transduction. B cells enriched by magnetic cell separation from Irf4fl/–

and Irf4fl/+ mice were washed three times with HyQADCF-Mab medium

(Hyclone) and 1 � 107 cells/ml were incubated for 45 min at 37 1C with

TAT-Cre33 at a concentration of 35 mg/ml. After being washed twice with RPMI

medium plus 10% FBS, cells were cultured at a density of 2 � 106 cells/ml as

described above with LPS or CD40 with or without IL-4.

Enzyme-linked immunosorbent assay (ELISA) and ELISPOT. ELISA for NP-

specific antibodies to IgG1, IgG2b and IgM using plates coated with NP23 in

BSA was done as described3. Peripheral blood was obtained before immuniza-

tion and every 10 d after primary and secondary immunization. For ELISPOT

assays, B cell fractions were derived from NP-KLH–boosted Irf4fl/–Cg1-Cre and

Irf4fl/+Cg1-Cre mice (secondary immunization 42 d after a primary immuniza-

tion with NP-KLH), eGFP+ and eGFP– B cells were sorted by flow cytometry

and samples were enriched by depletion of Thy-1+IgM+IgD+Gr-1+F4/80+ cells.

Then, IgG1-secreting cells were detected with the ELISPOT Starter Kit (Endo-

gen) according to the manufacturer’s protocol using anti-IgG1 (clone A85-1) as

the ‘catch’ antibody and biotinylated anti-IgG1 (clone A85-3) as the detection

antibody (BD Pharmingen). For each fraction, dilutions of 1:2 (starting from

2.5 � 104 cells) were cultured overnight in 96-well plates coated with anti-IgG1.

Immunohistochemistry. Sections 4 mm in thickness that were formalin-fixed,

paraffin-embedded, dewaxed, ‘antigen-retrieved’ and peroxidase-inhibited were

processed as described3,51. Slides were blocked in 5% defatted milk powder in

Tris-buffered saline–Triton X-100 (Sigma) and were incubated overnight with,

rat anti-mouse immunoglobulin subclasses (BD Pharmingen) or goat anti-

mouse immunoglobulin subclasses (Southern Biotechnology Associates) or

rabbit anti-mouse immunoglobulin subclasses (Sigma). Slides were then

washed, were counterstained with alkaline peroxidase–conjugated species-

specific goat or donkey antibodies (Southern Biotechnology) and were devel-

oped in nitro blue tetrazolium chloride–5-bromo-4-chloro-3-indolyl phos-

phate (NBT/BCIP; Roche). Then slides were stained for Bcl-6 as described3.

Somatic hypermutation analysis. For specific amplification of VH186.2

transcripts from sorted eGFP+ cell fractions (Supplementary Fig. 5 and

Supplementary Tables 2,3 online), a PCR strategy was used that excludes

the recognition of most VH186.2-related gene segments. The cDNA synthesis

with a Cg1-specific cDNA-primer and PCR primers and amplification condi-

tions were as described52 with the following modifications. For the two-round

amplification of certain cDNA fractions, 3 ml of the first-round mixture

(amplification program, 20 cycles) was amplified for another 30 cycles with a

seminested VH186.2 primer (5¢-CATGCTCTTCTTGGCAGCAACAG-3¢)together with the Cg1 PCR primer. Similar amounts of cell equivalents were

used in the PCR. When two rounds of amplification were required, at least

two separate PCR products derived from the cDNA mixture were cloned

and analyzed.

For cDNA synthesis and amplification of Cg1-membrane transcripts (Sup-

plementary Fig. 2 and Supplementary Table 1), 2 ml of a cDNA mixture was

reverse-transcribed with an oligonucleotide hybridizing in the Cg1 membrane–

encoding exon (5¢-TGACAGCAGCGCTGTAGCAC-3¢) and was amplified for

30 cycles with the amplification conditions described above using the VH186.2

primer (ref. 52) and a reverse primer hybridizing in the Cg1 membrane–

encoding exon (5¢-CAGCACAGGTCTCGTCCAGTTG-3¢).

Cloning and sequencing of PCR products corresponding to single colonies

was done as described52 using the ImMunoGeneTics database for sequence

analysis (http://www.genetik.uni-koeln.de/dnaplot/). The unpaired Student’s

780 VOLUME 7 NUMBER 7 JULY 2006 NATURE IMMUNOLOGY

A R T I C L E S©

2006

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eim

mun

olog

y

t-test was used for statistical evaluation of the differences in the VH

somatic mutation frequencies between different genotypes, with n correspond-

ing to the combined number of sequences of the individual mice of the

respective genotypes.

Semiquantitative PCR. RNA was isolated from purified B cell fractions from

Irf4–/– and wild-type mice, before and after stimulation with LPS or with CD40

plus IL-4, and concentrations were determined by spectrophotometry and by

analysis on an ethidium bromide–stained 1% agarose gel. For each fraction,

2 mg RNA was reverse-transcribed into cDNA with Superscript II (Invitrogen).

Semiquantitative PCR was done with serial dilutions of 1:4 in a solution

containing 1.5 mM MgCl2, 100 mM dATP, dGTP, dTTP and dCTP, 0.2 mM of

each primer and 0.4 ml Taq polymerase (Invitrogen). Cycling conditions were as

follows: 94 1C for 2 min, then 30 cycles of 94 1C for 20 s, 60 1C (for

glyceraldehyde phosphate dehydrogenase (Gapdh), 58 1C) for 30 s and 72 1C

for 60 s. Primer sequences were as follows: Blimp-1 (Prdm1), exon 1 forward

(5¢-TGGACTGGGTGGACATGAGAG-3¢), exon 5 reverse (5¢-AAGTGGTG

GAACTCCTCTCTG-3¢), exon 6 forward (5¢-AAGACTCTTCCTTACCCTC

TG-3¢) and exon 8 reverse (5¢-CTGCAGAGATGGATGTAGCTC-3¢); Gapdh

forward (5¢-CGGAGTCAACGGATTTGGTCGTAT-3¢) and reverse (5¢-AGCC

TTCTCCATGGTGGTGAAGAC-3¢); and Aicda forward (5¢-GATATGGACAGC

CTTCTGATG-3¢) and reverse (5¢-TTGCTTTCAAAATCCCAACA-3¢); Ig1, Cg1,

Im and Cm primer sequences were as described35.

Immunoblot analysis. Whole-cell lysates of B cell–enriched fractions corre-

sponding to day 0 and day 3 of LPS stimulation were separated by SDS-PAGE

and were blotted using standard procedures. The following antibodies were

used: rat anti–mouse Blimp-1 (clone 6D2; provided by L.M. Corcoran, Walter

and Eliza Hall Institute of Medical Research, Parkville, Australia); polyclonal

rabbit anti–mouse XBP-1 (Santa Cruz Biotechnology), polyclonal rabbit anti-

AID53 and monoclonal anti–mouse b-actin (Sigma); peroxidase-conjugated

goat anti-rat IgG (Amersham Biosiences); peroxidase-conjugated goat anti-

rabbit IgG (Pierce); and peroxidase-conjugated goat anti-mouse IgG (Pierce).

ECL Western Blotting Detection Reagents (Amersham) were used for detection.

For quantification of AID, band intensities were measured with NIH

Image program 1.62 and AID expression was normalized to that of b-actin

(loading control).

Note: Supplementary information is available on the Nature Immunology website.

ACKNOWLEDGMENTSWe thank P. Smith and M. Benito for help with the immunohistochemistry;M. Deren for genotyping; J. Gao for help with ELISA; K. Gordon andS. Stefanova for cell sorting; V. Miljkovic for sequencing; M. Alizhamov forpEasyFlox; D. Tarlinton (Walter Eliza Hall Institute, Parkville, Australia) foradvice on the detection of NP-binding B cells; and L. Pasqualucci, M. Saito andA. Pernis for discussions. Supported by the National Institutes of Health(CA92625 and CA098285 to K.R. and CA92625 and CA37295 to R.D.-F.)and the Human Frontiers Science Program (U.K.).

AUTHOR CONTRIBUTIONSU.K. designed research, did experiments, analyzed data and wrote themanuscript; R.D.-F. designed research and wrote the manuscript; S.C. andK.R. contributed the Cg1-Cre mouse, which was critical for this analysis, anddesigned research; G.C. did the immunostaining experiments and analyzeddata; Q.S. did the embryonic stem cell injection and ELISA; M.L. did theVH sequencing analysis; T.M. maintained and genotyped the mouse cohort;and T.L. was involved in construction of the transgenic Irf4 mouse andprovided materials.

COMPETING INTERESTS STATEMENTThe authors declare that they have no competing financial interests.

Published online at http://www.nature.com/natureimmunology/

Reprints and permissions information is available online at http://npg.nature.com/

reprintsandpermissions/

1. MacLennan, I.C. Germinal centers. Annu. Rev. Immunol. 12, 117–139 (1994).2. Rajewsky, K. Clonal selection and learning in the antibody system. Nature 381,

751–758 (1996).

3. Ye, B.H. et al. The BCL6 proto-oncogene controls germinal-centre formation and Th2-type inflammation. Nat. Genet. 16, 161–170 (1997).

4. Dent, A.L., Shaffer, A.L., Yu, X., Allman, D. & Staudt, L.M. Control of inflammation,cytokine expression, and germinal center formation by BCL6. Science 276, 589–592(1997).

5. Angelin-Duclos, C., Cattoretti, G., Lin, K.I. & Calame, K. Commitment of B lymphocytesto a plasma cell fate is associated with Blimp-1 expression in vivo. J. Immunol. 165,5462–5471 (2000).

6. Falini, B. et al. A monoclonal antibody (MUM1p) detects expression of the MUM1/IRF4protein in a subset of germinal center B cells, plasma cells, and activated T cells. Blood95, 2084–2092 (2000).

7. Raaphorst, F.M. et al. Cutting edge: polycomb gene expression patterns reflect distinctB cell differentiation stages in human germinal centers. J. Immunol. 164, 1–4(2000).

8. Basso, K. et al. Tracking CD40 signaling during germinal center development. Blood104, 4088–4096 (2004).

9. Fornek, J.L. et al. Critical role for STAT3 in T-dependent terminal differentiation of IgGB cells. Blood 107, 1085–1091 (2006).

10. Scheeren, F.A. et al. STAT5 regulates the self-renewal capacity and differentiation ofhuman memory B cells and controls BCL6 expression. Nat. Immunol. 6, 303–313(2005).

11. Shapiro-Shelef, M. et al. Blimp-1 is required for the formation of immunoglobulinsecreting plasma cells and pre-plasma memory B cells. Immunity 19, 607–620(2003).

12. Eisenbeis, C.F., Singh, H. & Storb, U. Pip, a novel IRF family member, is a lymphoid-specific, PU.1-dependent transcriptional activator. Genes Dev. 9, 1377–1387(1995).

13. Matsuyama, T. et al. Molecular cloning of LSIRF, a lymphoid-specific member of theinterferon regulatory factor family that binds the interferon-stimulated responseelement (ISRE). Nucleic Acids Res. 23, 2127–2136 (1995).

14. Yamagata, T. et al. A novel interferon regulatory factor family transcription factor,ICSAT/Pip/LSIRF, that negatively regulates the activity of interferon-regulated genes.Mol. Cell. Biol. 16, 1283–1294 (1996).

15. Iida, S. et al. Deregulation of MUM1/IRF4 by chromosomal translocation in multiplemyeloma. Nat. Genet. 17, 226–230 (1997).

16. Mamane, Y. et al. Interferon regulatory factors: the next generation. Gene 237, 1–14(1999).

17. Pernis, A.B. The role of IRF4 in B and T cell activation and differentiation. J. InterferonCytokine Res. 22, 111–120 (2002).

18. Marecki, S. & Fenton, M.J. The role of IRF4 in transcriptional regulation. J. InterferonCytokine Res. 22, 121–133 (2002).

19. van der Stoep, N., Quinten, E., Marcondes Rezende, M. & van den Elsen, P.J. E47,IRF4, and PU.1 synergize to induce B-cell-specific activation of the class II transacti-vator promoter III (CIITA-PIII). Blood 104, 2849–2857 (2004).

20. Gupta, S., Jiang, M., Anthony, A. & Pernis, A.B. Lineage-specific modulation ofinterleukin 4 signaling by interferon regulatory factor 4. J. Exp. Med. 190, 1837–1848 (1999).

21. Lu, R., Medina, K.L., Lancki, D.W. & Singh, H. IRF4,8 orchestrate the pre-B-to-Btransition in lymphocyte development. Genes Dev. 17, 1703–1708 (2003).

22. Mittrucker, H.W. et al. Requirement for the transcription factor LSIRF/IRF4 for matureB and T lymphocyte function. Science 275, 540–543 (1997).

23. Rosenwald, A. et al. Molecular diagnosis of primary mediastinal B cell lymphomaidentifies a clinically favorable subgroup of diffuse large B cell lymphoma related toHodgkin lymphoma. J. Exp. Med. 198, 851–862 (2003).

24. Savage, K.J. et al. The molecular signature of mediastinal large B-cell lymphomadiffers from that of other diffuse large B-cell lymphomas and shares features withclassical Hodgkin lymphoma. Blood 102, 3871–3879 (2003).

25. Farley, F.W., Soriano, P., Steffen, L.S. & Dymecki, S.M. Widespread recombinaseexpression using FLPeR (flipper) mice. Genesis 28, 106–110 (2000).

26. Casola, S. et al. Tracking germinal center B cells expressing germline immunoglobuling1 transcripts by conditional gene targeting. Proc. Natl. Acad. Sci. USA 103, 7396–7401 (2006).

27. Kraal, G., Weissman, I.L. & Butcher, E.C. Germinal centre B cells: antigen specificityand changes in heavy chain class expression. Nature 298, 377–379 (1982).

28. Ridderstad, A. & Tarlinton, D.M. Kinetics of establishing the memory B cell populationas revealed by CD38 expression. J. Immunol. 160, 4688–4695 (1998).

29. Sanderson, R.D., Lalor, P. & Bernfield, M. B lymphocytes express and lose syndecan atspecific stages of differentiation. Cell Regul. 1, 27–35 (1989).

30. Blink, E.J. et al. Early appearance of germinal center-derived memory B cells andplasma cells in blood after primary immunization. J. Exp. Med. 201, 545–554(2005).

31. Shaffer, A.L. et al. XBP-1, downstream of Blimp-1, expands the secretory apparatusand other organelles, and increases protein synthesis in plasma cell differentiation.Immunity 21, 81–93 (2004).

32. Reimold, A.M. et al. Plasma cell differentiation requires the transcription factor XBP-1.Nature 412, 300–307 (2001).

33. Peitz, M., Pfannkuche, K., Rajewsky, K. & Edenhofer, F. Ability of the hydrophobic FGFand basic TAT peptides to promote cellular uptake of recombinant Cre recombinase: atool for efficient genetic engineering of mammalian genomes. Proc. Natl. Acad. Sci.USA 99, 4489–4494 (2002).

34. Layton, J.E., Vitetta, E.S., Uhr, J.W. & Krammer, P.H. Clonal analysis of B cells inducedto secrete IgG by T cell-derived lymphokine(s). J. Exp. Med. 160, 1850–1863(1984).

NATURE IMMUNOLOGY VOLUME 7 NUMBER 7 JULY 2006 781

A R T I C L E S©

2006

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eim

mun

olog

y

35. Muramatsu, M. et al. Class switch recombination and hypermutation require activation-induced cytidine deaminase (AID), a potential RNA editing enzyme. Cell 102,553–563 (2000).

36. Kinoshita, K., Harigai, M., Fagarasan, S., Muramatsu, M. & Honjo, T. A hallmark ofactive class switch recombination: transcripts directed by I promoters on looped-outcircular DNAs. Proc. Natl. Acad. Sci. USA 98, 12620–12623 (2001).

37. Hasbold, J., Corcoran, L.M., Tarlinton, D.M., Tangye, S.G. & Hodgkin, P.D.Evidence from the generation of immunoglobulin G–secreting cells thatstochastic mechanisms regulate lymphocyte differentiation. Nat. Immunol. 5, 55–63(2004).

38. Hasbold, J., Lyons, A.B., Kehry, M.R. & Hodgkin, P.D. Cell division number regulatesIgG1 and IgE switching of B cells following stimulation by CD40 ligand and IL-4. Eur.J. Immunol. 28, 1040–1051 (1998).

39. Randall, T.D. et al. Arrest of B lymphocyte terminal differentiation by CD40 signaling:mechanism for lack of antibody-secreting cells in germinal centers. Immunity 8, 733–742 (1998).

40. Knodel, M., Kuss, A.W., Berberich, I. & Schimpl, A. Blimp-1 over-expression abrogatesIL-4- and CD40-mediated suppression of terminal B cell differentiation but arrestsisotype switching. Eur. J. Immunol. 31, 1972–1980 (2001).

41. Weiss, U. & Rajewsky, K. The repertoire of somatic antibody mutants accumulating inthe memory compartment after primary immunization is restricted through affinitymaturation and mirrors that expressed in the secondary response. J. Exp. Med. 172,1681–1689 (1990).

42. McHeyzer-Williams, M.G., McLean, M.J., Lalor, P.A. & Nossal, G.J. Antigen-driven Bcell differentiation in vivo. J. Exp. Med. 178, 295–307 (1993).

43. Smith, K.G., Light, A., Nossal, G.J. & Tarlinton, D.M. The extent of affinity maturationdiffers between the memory and antibody-forming cell compartments in the primaryimmune response. EMBO J. 16, 2996–3006 (1997).

44. Klein, U. et al. Transcriptional analysis of the B cell germinal center reaction. Proc.Natl. Acad. Sci. USA 100, 2639–2644 (2003).

45. Lohoff, M. et al. Enhanced TCR-induced apoptosis in interferon regulatory factor 4-deficient CD4+ Th cells. J. Exp. Med. 200, 247–253 (2004).

46. Kallies, A. et al. Plasma cell ontogeny defined by quantitative changes in Blimp-1expression. J. Exp. Med. 200, 967–977 (2004).

47. Sciammas, R. & Davis, M.M. Modular nature of Blimp-1 in the regulation of geneexpression during B cell maturation. J. Immunol. 172, 5427–5440 (2004).

48. Gonda, H. et al. The balance between Pax5 and Id2 activities is the key to AID geneexpression. J. Exp. Med. 198, 1427–1437 (2003).

49. Cattoretti, G. et al. Nuclear and cytoplasmic AID in extrafollicular and germinal centerB cells. Blood 107, 3967–3975 (2006).

50. Lee, C.H. et al. Regulation of the germinal center gene program by interferon (IFN)regulatory factor 8/IFN consensus sequence-binding protein. J. Exp. Med. 203, 63–72(2006).

51. Cattoretti, G. et al. PRDM1/Blimp-1 is expressed in human B-lymphocytes committedto the plasma cell lineage. J. Pathol. 206, 76–86 (2005).

52. Klein, U., Esposito, G., Baudat, F., Keeney, S. & Jasin, M. Mice deficient for thetype II topoisomerase-like DNA transesterase Spo11 show normal immunoglobulinsomatic hypermutation and class switching. Eur. J. Immunol. 32, 316–321 (2002).

53. Faili, A. et al. AID-dependent somatic hypermutation occurs as a DNA single-strandevent in the BL2 cell line. Nat. Immunol. 3, 815–821 (2002).

782 VOLUME 7 NUMBER 7 JULY 2006 NATURE IMMUNOLOGY

A R T I C L E S©

2006

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eim

mun

olog

y