IgG1+ ovalbumin-specific B-cell transnuclear miceshow class switch recombination in rare allelicallyincluded B cellsStephanie K. Dougana, Souichi Ogataa,b, Chih-Chi Andrew Hua,c, Gijsbert M. Grotenbrega,d, Eduardo Guillena,Rudolf Jaenischa,1, and Hidde L. Ploegha,1

aWhitehead Institute for Biomedical Research, Cambridge, MA 02142; bJanssen Oncology Research and Development, a division of Janssen Pharmaceutica NV,Beerse B2340, Belgium; cDepartment of Immunology, H. Lee Moffitt Cancer Center and Research Institute, Tampa, FL 33612; and dImmunology Programmeand Department of Microbiology, National University of Singapore, Singapore 117456

Contributed by Rudolf Jaenisch, June 22, 2012 (sent for review May 7, 2012)

We used somatic cell nuclear transfer (SCNT) to generate a mousefrom the nucleus of an IgG1+ ovalbumin-specific B cell. The result-ing OBI mice show generally normal B-cell development, withelevated percentages of marginal zone B cells and a reduction inB-1 B cells. Whereas OBI RAG1−/− mice have exclusively IgG1 anti-ovalbumin in their serum, OBI mice show elevated levels of anti-ovalbumin of nearly all isotypes 3′ of the γ1 constant region in theIgH locus, indicating that class switch recombination (CSR) occursin the absence of immunization with ovalbumin. This CSR is asso-ciated with the presence of IgM+IgG1+ double producer B cells thatrepresent <1% of total B cells, accumulate in the peritoneal cavity,and account for near-normal levels of serum IgM and IgG3.

allelic exclusion | natural antibodies | TN mice

Bcells exist as a polyclonal pool such that an antibody responsemay be mounted against any possible pathogen. When a B-cell

recognizes its cognate antigen through its B-cell receptor (BCR),the B cell proliferates and differentiates into antibody-secretingplasma cells and a smaller population of memory B cells. Clonalselection theory rests on the idea that a B cell expresses a BCR ofa single specificity; harmful consequences could ensue if a B cellactivated in the normal course of an immune response also pro-duced a second antibody that reacted with self-antigen. Severalmechanisms ensure that a B cell produces only a single specificityBCR, including monoallelic initiation of recombination, restrictedaccess of the RAG proteins, rapid entry into the cell cycle, chro-matin remodeling, and subunit pairing constraints (1, 2). Thus,nearly all B cells express a BCR encoded by single alleles at the IgHand Igκ or λ loci. Allelic exclusion, however, is not perfect, and∼0.01% of B cells express two rearranged IgH genes (3), whereas1–7% of B cells express two rearranged Igκ genes (4, 5).B-cell development begins in the bone marrow when B-cell

progenitors express RAG1/2 and rearrange the Ig heavy chainlocus (6). D to J rearrangement occurs first, often on both chro-mosomes, followed by V to DJ rearrangements. A productive, in-frame VDJ results in cell surface expression of the Ig heavy chainpaired with VpreB and λ5 surrogate light chains. Pre-BCR sig-naling induces proliferation and prevents further rearrangementson the other chromosome. After several rounds of cell division,pre-B cells re-express the RAG genes and engage in V to J rear-rangement of the Igκ light chain locus. Surface expression of theBCR marks transition to the immature B-cell stage.Once in the periphery, B cells engage a wider array of anti-

gens, including those from food and commensal microbes.Transitional B cells further differentiate into one of three majorB-cell populations: long-lived marginal zone (MZ) B cells thatreside in the marginal zone sinus of the spleen; follicular B cellsthat form the B-cell zones of spleen and lymph nodes; and B-1 Bcells that reside mainly in the peritoneal cavity and are a majorsource of natural antibodies (7, 8). The fate decision made bytransitional B cells is linked to the signaling capacity of the BCR,and it has been suggested that BCR affinity for peripheral self-antigens directs B cells into particular lineages, with higher

affinity B cells becoming B-1 B cells, whereas lower affinity Bcells develop into MZ B cells (7, 8).When naïve follicular B cells encounter antigen, the BCR-

bound antigen complex is internalized into endolysosomes whereantigen is degraded and peptide fragments are presented on classII MHC (9). Toll-like receptor (TLR) ligands such as LPS or CpGcan further activate a B cell to express the costimulatory ligandsCD80 and CD86. Surface expression of peptide-loaded class IIMHC with costimulatory ligands engages CD4 T cells of the ap-propriate specificity. These CD4 T cells provide CD40L stimula-tion and release cytokines such as IL-4 and IL-6 that induce theformation of a germinal center, where B cells undergo affinitymaturation and isotype switching, both of which are mediated byactivation-induced deaminase (AID). Rare B cells with mutationsthat increase their affinity for antigen are selected for expansionand development into plasma cells and memory B cells (9).All B cells initially express IgM, but may switch to other iso-

types upon AID-induced double-strand breaks in the GC-richswitch region that precedes each constant region (10). Resolu-tion of these breaks causes looping out and deletion of the in-tervening DNA from the genome. Thus, an IgG1+ B cell hasdeleted the μ, δ, and γ3 constant regions, at least on the allelecontaining the productively rearranged VDJ.Chicken ovalbumin (OVA), the major protein component of

egg whites, has been a favorite of immunologists for years. T-cellreceptors from CD8 and CD4 cells specific for immunodominantepitopes of ovalbumin were cloned and used to generate OT-Iand OT-II transgenic mice as a source of monoclonal lympho-cytes of defined specificity (11, 12). The availability of ovalbu-min-specific CD8 and CD4 T cells has inspired the generationof dozens of engineered pathogens, tumor cell lines, and trans-genic mice that express ovalbumin to study the many aspects ofadaptive immunity. Notwithstanding the widespread use of ov-albumin as a model antigen, no ovalbumin-specific BCR trans-genic mice have been reported.Somatic cell nuclear transfer (SCNT) from antigen-specific

lymphocytes allows the generation of transnuclear (TN) micewith lymphocytes of a single, defined specificity (13). Productionof TN mice is rapid, requiring ∼6 wk from lymphocyte harvest toobtaining chimeric animals, and requires no DNA vector con-struction or genetic manipulation of embryonic stem cells. Earlierwe reported a panel of TN mice derived from CD8 T cells specificfor Toxoplasma gondii (13) and we now applied the same tech-nique to B cells specific for OVA to obtain antigen-specific B-celltransnuclear mice. The resulting OBI mice contain B cells thatare ovalbumin specific, have no genetic alterations other than the

Author contributions: S.K.D., S.O., R.J., and H.L.P. designed research; S.K.D. and S.O.performed research; G.M.G. and E.G. contributed new reagents/analytic tools; S.K.D.,C.-C.A.H., R.J., and H.L.P. analyzed data; and S.K.D. and H.L.P. wrote the paper.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.1To whom correspondence may be addressed. E-mail: ploegh@wi.mit.edu or jaenisch@wi.mit.edu.

www.pnas.org/cgi/doi/10.1073/pnas.1210273109 PNAS | August 21, 2012 | vol. 109 | no. 34 | 13739–13744

IMMUNOLO

GY

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 18

, 202

0

physiological BCR rearrangements, and are the closest approx-imation of the high-affinity B cells that result in primary andmemory antibody responses in vivo. Our data show that the Bcell that served as nucleus donor for SCNT had already classswitched to IgG1. Whereas the use of IgG1 is fully compatiblewith B-cell development, allelic exclusion is imperfect and allowsthe emergence of B cells that rearrange the remaining wild-typeIgH locus to yield a presumably diverse repertoire of IgM. TheseIgM+ IgG1+ cells express productively rearranged BCRs of twodifferent specificities and can initiate class-switch recombinationin animals not deliberately exposed to ovalbumin, resulting in theproduction of isotypes other than IgM from the wild-type allele,and ovalbumin-specific class-switched immunoglobulins from thetransnuclear allele.

ResultsGeneration of OBI Mice. Somatic cell nuclear transfer is most effi-cient when using F1 hybrid mice as a source of donor nuclei (13–

15). Accordingly we used B6xBALB/c F1 males as a source ofB cells. To identify antigen-specific B cells, we mixed biotinylatedovalbumin with streptavidin-phycoerythrin (PE) to generate tet-rameric phycoerythrin-labeled ovalbumin (tOVA-PE). Splenocytesfrom control mice showed ∼0.03% of B cells binding to tOVA-PE,a frequency too low to proceed with isolation of antigen-specific Bcells and SCNT. We therefore immunized mice intraperitoneallywith 100 μg of ovalbumin in complete Freund’s adjuvant (CFA),followed by two doses of 100 μg ovalbumin in incomplete Freund’sadjuvant (IFA), which allowed us to identify a rare population(∼0.1%) of B cells that stained with tOVA-PE (Fig. 1A). Sevendays after the final immunization, we isolated isotype-switchedCD19+, IgM−, tOVA-PE+ B cells by fluorescence activated cellsorting (FACS) and used them as a source of donor nuclei forSCNT. A total of 154 nuclear transfers yielded three ES cell lines,one of which showed tOVA-PE+ cells in peripheral blood of chi-meric mice and gave germline transmission (Fig. 1B). B cells fromthe resultant OBI TN mice readily stained with OVA-Alexa 488and anti-IgG1 (Fig. 1C). The OBI TN IgH and Igκ loci werebackcrossed to B6 and placed onto a RAG1−/− background toprevent endogenous Ig rearrangements. Subsequent experimentswere performed on mice that were backcrossed for 8–10 gen-erations onto the B6 or B6;RAG1−/− backgrounds.B cells sorted from OBI RAG1−/− mice were used as a source

of cDNA for 5′ RACE to determine the sequence of the BCRheavy- and light-chain loci (Fig. 1D), which showed somaticmutations in both the IgH and Igκ variable regions, evidence thatthe original donor B cell had undergone affinity maturation ina germinal center. The heavy-chain (HC) VDJ was joined to γ1(IgG1), whereas the light-chain VJ was connected to the κ con-stant region. Thus, the original donor nucleus came from a high-affinity IgG1+Igκ+ B cell.To define the epitope recognized by the OBI BCR, we syn-

thesized overlapping 10-mer peptides from chicken ovalbuminand spotted them onto nitrocellulose. OBI serum recognizes anepitope centered on the sequence DKLPGFGDSI, contained ina surface-exposed loop of ovalbumin (Fig. 1E). The OBI epitopeis located in the N-terminal portion of ovalbumin and is distinctfrom the more C-terminally located OT-I and OT-II epitopes.

OBI Heavy Chain Alone Can Confer Binding to Ovalbumin. To in-vestigate the role of the OBI heavy chain in antigen binding, weisolated B cells from OBI HC mice that inherited the rearrangedOBI heavy chain in the absence of the OBI light chain. OBI HCor wild-type B cells were cultured with CpG for 3 d, labeled tosteady state with [35S]methionine/cysteine, and supernatantswere immunoprecipitated with ovalbumin-conjugated sepharose

Fig. 1. OBI mice generated by somatic cell nuclear transfer. B6xBALB/c F1male mice were immunized three times with ovalbumin in CFA/IFA adjuvant.Splenocytes were harvested 7 d after the final immunization and stained withanti-IgM and ovalbumin-PE tetramers (tOVA-PE). IgM−, ovalbumin+ cells weresorted by FACS and used as donor nuclei for SCNT to generate OBI TN mice. (A)Splenocytes from nonimmunized and triply immunized mice were stained withanti-IgM and tOVA-PE. Numbers indicate number of cells per 100,000. (B) Pe-ripheral blood from the chimeric founder shows ovalbumin reactive IgM−

B cells. Population shown is gated on CD19+ cells. (C) Peripheral blood B cellsfrom germline transmitted OBI mice stain brightly with monomeric ovalbumin-Alexa 488 (Invitrogen) and with anti-IgG1. (D) cDNA with synthesized fromOBI RAG1−/− B cells and subjected to 5′ RACE analysis. The sequence of the IgHand Igκ chains is shown. Nucleotides that differ from germline are highlighted inbold. (E) Overlapping 10-mer peptides from chicken ovalbumin were synthesizedand spotted onto nitrocellulose. The membrane was probed with serum froman OBI mouse. Observed reactivity to the peptides is indicated in bold.

totalIg

postOvaIgOva1 Ova2 Ova3

totalIg

postOvaIgOva1 Ova2 Ova3

totalIg

postOvaIgOva1 Ova2 Ova3

totalIg

postOvaIgOva1 Ova2 Ova3

wild type cell lysate OBI heavy chain cell lysate wild type media OBI heavy chain media

75

100

150

50

37

25

IgM

IgG1

Ig

Fig. 2. OBI heavy chain alone can confer binding to ovalbumin. B cells wereharvested from spleens of B6 mice or mice bearing only the IgH OBI TN locusand cultured for 4 d in media containing LPS. Cells were then labeled with [35S]methionine/cysteine for 4 h. Supernatants were collected and serially immu-noprecipitated with ovalbumin-conjugated sepharose beads. After four se-quential precipitations, the ova-depleted supernatants were precipitatedwith protein G to isolate the remaining nonovalbumin reactive antibodies.

13740 | www.pnas.org/cgi/doi/10.1073/pnas.1210273109 Dougan et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 18

, 202

0

beads. Sequential precipitation with ovalbumin-beads removedall anti-ovalbumin antibodies, and the remaining ovalbumin-depleted supernatants were immunoprecipitated with anti-IgMand anti-IgG1 to discern the amount of nonovalbumin-reactiveantibodies produced (Fig. 2). Although most of the Ig produced inOBI HC mice is not reactive with ovalbumin, the OBI heavy chainalone is sufficient to confer binding to ovalbumin, when pairedwith ∼10% (as assessed biochemically) of available light chains.

OBI Mice Have an Increase in MZ B Cells and a Decrease in B-1 B Cells.We found no obvious impairment of B-cell development in theOBI mice (Fig. 3). Compared with wild type, OBI and OBIRAG1−/− bone marrow populations showed a higher percentageof surface BCR+ cells, as expected in cells that have alreadyundergone V(D)J recombination (16, 17). The presence of ma-ture B cells in mice that received the OBI heavy chain alone(OBI HC) suggests that IgG1 already pairs adequately with theλ5 surrogate light chain in the course of B-cell development (18).Transgenic light chains have been reported to substitute for λ5(19); thus in OBI mice with both VDJ and VJ rearrangements,we cannot exclude the possibility that IgG1 pairs with Igκ at thepre–B-cell stage. OBI mice have normal or slightly elevatedpercentages of B cells in peripheral lymphoid organs, and theabsolute numbers of B and T cells in OBI mice are normal.Experiments exploring the role of BCR affinity in fate decisions

have thus far relied either on manipulated expression levels oftransgenic BCRs and target antigens or on alteration of the sig-naling capacity of the BCR (7, 8). Here we find a subtle increasein MZ B cells in OBI and OBI RAG1−/− mice and a decrease inB-1 B cells in OBI mice, which became more pronounced on theRAG1−/− background (Fig. 3). This skewing of the MZ to B-1 Bcell ratio suggests that the particular specificity of the OBI BCRdirects cell fate decisions in the transitional B-cell pool.

OBI Serum Antibody Titers Show Extensive Isotype Switching. Inaddition to a loss of peritoneal cavity B-1 cells, OBI mice showedfewer CD5+ B cells in spleen (Fig. 2). This spleen-resident B-1B-cell population has been described as composed of short-termantibody-secreting cells and—together with B-1 B cells in theperitoneal cavity—is largely responsible for maintaining serumantibody titers (20, 21). We therefore analyzed serum from 3-mo-old mice of the following genotypes: wild-type B6, OBI HC,

OBI, and OBI RAG1−/− (Fig. 4A). Serum from OBI RAG1−/−

mice contained anti-ovalbumin antibodies composed exclusivelyof IgG1 and Igκ isotypes, as expected, because OBI RAG1−/−

mice lack the CD4 T cells that provide the CD40L signal toinitiate isotype switching (9). However, OBI and even OBI HCmice showed anti-ovalbumin IgG2a, IgG2b, and IgA (Fig. 4A) inthe absence of deliberate immunization. To confirm that thesecondary antibodies used for ELISA did not cross-react withIgG1, we generated a hybridoma from OBI spleen cells and in-cluded hybridoma supernatant in each assay as a positive controlfor anti-ovalbumin IgG1 and Igκ and a negative control for allother isotypes. The high concentration of anti-ovalbumin IgG1 inOBI serum most likely outcompetes other anti-ovalbumin iso-types for binding to the ovalbumin-coated ELISA plate. Indeed,when serum samples from OB1 mice were first depleted withprotein G sepharose to remove IgG and then assayed for oval-bumin-specific IgA, higher titers were detected, which did notplateau over the serum dilutions tested (Fig. 4A, last panel).The B6 mouse genome encodes IgG2c, but not IgG2a (22);

therefore, the presence of anti-ovalbumin IgG2a in the OBImouse strain indicates that it was the BALB/c allele in theoriginal B6xBALB/c F1 donor B cell that gave rise to the pro-ductive VDJ recombination. We could not detect anti-ovalbuminIgG2c in OBI serum, although our detection threshold was

CD23

CD

21

B220

CD

5A

A4.

1

CD23

GL7

B22

0

B220

CD

3

B220

CD

19

CD19

CD

43

IgD

IgM

or I

gG1

wild type OBI RAG1+/- OBI RAG1-/- wild type OBI RAG1+/- OBI RAG1-/-

sple

en

Lymph nodes

Peritoneal cavity

Bone m

arrow

1845

32

1645

32

1067

15

72

11

39 36

3.0 1.2 1.3

16 16 11 21 9 22

1.4 0.12 17 4.2 2.2

3226 68

8.7

68

10

75

18

65

5937

4551

198

2.6

Fig. 3. OBI mice have near-normal B-cell development with increased MZand decreased B-1 B-cell populations. Cells were harvested from spleen,mesenteric lymph nodes, peritoneal cavity, and bone marrow and stainedwith the indicated antibodies. Spleen populations (Left column) were gatedon CD19+ cells. Bone marrow populations (Lower Right) were gated onB220+ cells. Results are representative of three to six mice per group.

A

B

-510 -410 -310 -210

OVA

-Ale

xa55

5

wild type OBI OBI RAG1-/-

IgA

Bone m

arrow(gated on C

D138+)

Peyer’s patches

(gated on CD

19+)

6.2%0.6% 0.0%

0.01%0.69%0.28%

22%36%

0.0%

0.0%

4.2%2.6%

00.20.40.60.8

11.21.41.61.8

anti-

ova

O.D

.

IgG2a

00.20.40.60.8

11.21.41.61.8

anti-

ova

O.D

.

IgG2b

00.20.40.60.8

11.21.41.61.8

anti-

ova

O.D

.

IgA

00.20.40.60.8

11.21.41.61.8

anti-

ova

O.D

.

IgA (IgG-depleted)

00.20.40.60.8

11.21.41.61.8

anti-

ova

O.D

.

IgG1

00.20.40.60.8

11.21.41.61.8

anti-

ova

O.D

.

Igκκ

OBI HCOBIOBI RAG1-/-wild typehybridoma

00.20.40.60.8

11.21.41.61.8

anti-

ova

O.D

.

Igλλ

-510 -410 -310 -210 -510 -410 -310 -210 -510 -410 -310 -210

-510 -410 -310 -210-610 -510 -410 -310-610 -510 -410 -310

Fig. 4. Serum levels of antiovalbumin antibodies. (A) Serum was harvestedfrom 3-mo-old B6, OBI TN heavy chain only (OBI HC), OBI RAG-proficient(OBI), or OB-I RAG1−/− mice and analyzed by ELISA for ovalbumin-specificantibodies of the isotypes shown. All mice were born to non-OBI mothers toeliminate the possibility of antiovalbumin antibody transfer via breast milk.Serum was used at a starting concentration of either 100-fold dilution or1,000-fold dilution (IgG1 and Igκ) and titrated at 10-fold serial dilutions.Error bars indicate SDs of six individual mice per group. Supernatant from anOBI hybridoma (IgG1+Igκ+) was included on each ELISA plate to control fornonspecific binding of the secondary ELISA reagents to the OBI IgG1. Hy-bridoma supernatant was used neat and at 10-fold serial dilutions. Resultsare representative of three independent experiments. Last panel: serumsamples were preincubated with protein G sepharose to deplete IgG beforeanalysis by ELISA. (B) Cells were isolated from bone marrow and Peyer’spatches of wild-type, OBI, and OBI RAG1−/− mice and stained with anti-IgAand fluorescent ovalbumin. Results are representative of two mice pergroup. Bone marrow populations are gated on CD138+ cells. Peyer’s patchpopulations are gated on CD19+ cells.

Dougan et al. PNAS | August 21, 2012 | vol. 109 | no. 34 | 13741

IMMUNOLO

GY

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 18

, 202

0

higher than for other isotypes, because the anti-IgG2c secondaryantibody used for ELISA showed slight reactivity with IgG1 fromhybridoma supernatant.Class-switched serum anti-ovalbumin derives from class-switched

plasma cells, so we examined bonemarrow fromOBImice. Amongthe CD138+ plasma cells, 6% were IgA+ and bound fluorescentovalbumin (Fig. 4B). The IgA+ ovalbumin− plasma cells likelyderived from OBI B cells in the germinal center that had sustainedAID-induced point mutations that reduced binding to ovalbumin.Because these mice were never immunized, ovalbumin would notbe present on follicular dendritic cells, and selection for increasedaffinity for ovalbumin would not occur. Although the frequency ofovalbumin-reactive IgA+ cells was low, none were detected in OBIRAG1−/− bone marrow. Likewise, Peyer’s patches from OBI miceshowed a significant population of IgA+ ovalbumin+ B cells absentfrom wild-type and OBI RAG1−/− mice (Fig. 4B).

Rare Double Producer Cells Account for Isotype Switching in OBI Mice.How does CSR occur in OBI mice? The lack of CSR in OBIRAG1−/− mice confirms that CD4 T-cell help is required.However, the frequency of anti-ovalbumin CD4 cells in a naïvepolyclonal T-cell pool must be exceedingly small (23). In light ofthis conundrum, we hypothesized that rare OBI cells mightrearrange the wild-type BCR allele and emerge as IgM+IgG1+

cells. These double producers would have BCRs of at least twodifferent specificities, assuming that allelic exclusion of the lightchain locus would be efficient. In such a situation, the non-OBIBCRs might bind to environmental antigens, interact with ap-propriately specific CD4 T cells, and undergo class switching.To investigate this possibility, we looked for IgM+ cells in OBI

mice. Total serum IgM and IgG3 are near-normal in OBI andOBI HC mice (Fig. 5A). However, the number of IgM+ cellsdetectable in spleen was ∼1% (Fig. 5B). Nonetheless, we sorted byFACS IgM+IgG1+ cells from OBI mice, cultured these cells for3 d with LPS and CpG, and labeled them with [35S]methionine/cysteine. Immunoprecipitation with anti-κ antibodies recoveredequivalent amounts of IgM and IgG1 from the culture media ofthe double producer cells, demonstrating that they synthesize andsecrete both isotypes at comparable efficiencies. These cells aretherefore the likely source of serum IgM (Fig. 5C).OBI mice have reduced populations of B-1 B cells (Fig. 2).

Peritoneal cavity cells from OBI mice were enriched inIgM+IgG1+ cells, and also IgM+IgG1− cells (Fig. 6A). TheseIgM-only cells, found exclusively in the peritoneal cavity, areeither rare cells that failed to express the OBI transnuclear IgG1and were preferentially directed to and expanded in the B-1 B-cell compartment or arose from IgM+IgG1+ cells that selectivelylost IgG1 expression during affinity maturation (24, 25).Total peritoneal cavity cells were cultured for 3 d with LPS

and CpG and labeled with [35S]methionine/cysteine. Immu-noprecipitation with anti-κ antibodies recovered a substan-tial amount of IgM, confirming that the peritoneal cavity isenriched in double producers (Fig. 6B). Of the IgG1+ B cellsin the peritoneal cavity, 7% also express IgM (Fig. 6C). Of theIgG1+CD5+ B-1 B-cell population, 81% expressed IgM, sug-gesting that within the B-1 B-cell compartment, IgM+IgG1+ cellsare the predominant population.

DiscussionAllelic exclusion prevents expression of two different BCRs ona single B cell. Allelic exclusion is mediated in part throughsuppression of further VDJ recombination once a productivelyrearranged Ig heavy chain locus has been generated; however,other factors must contribute because two productive VDJrearrangements can be found in allelically excluded cells (1, 26).We estimate that, similar to IgM+ B cells in wild-type mice (3),allelic exclusion applies to 99% of B cells in the OBI mouse asmeasured by cytofluorimetry. In wild-type mice, cytofluorimetryoverestimates the frequency of IgH double expressing cells dueto the inclusion of two-cell doublets mis-scored as double pro-ducers (3). The actual rate of allelic exclusion in OBI mice is

likely to be higher than 99%, although a determination of theexact frequency would require analysis by limiting dilution ofsorted putative allelically included cells (3). Despite this tightregulation, a small population of OBI B cells manages to expresssurface IgM, derived from rearrangement of the wild-type IgHallele. Although such double producers account for <1% of themature B-cell pool in spleen as estimated by cytofluorimetry,these double-positive B cells, or rather their differentiatedprogeny, produce enough IgM to maintain near-normal serumIgM levels (27). We cannot exclude the possibility that an envi-ronmental antigen cross-reacts with the OBI BCR; however, weview this as unlikely because the vast majority of OBI B cellshave a naïve phenotype and OBI mice housed at separate lo-cations have similar levels of the class-switched serum isotypes.Thus, in the absence of deliberate exposure to ovalbumin, doubleproducers are the most likely B cells capable of engaging CD4 Tcells. Once AID is activated in a B cell, it acts on transcribed IgHand IgL loci (28), resulting in class switching not only of the wild-type IgH locus (which is the only possible source of IgG3 in OBImice) but also of the transnuclear OBI IgG1 locus. Thus, OBImice on a RAG-proficient background express near-normal

A

IgM(+), IgG1(+)- EndoH EndoF - -

Supernatant/medium

IgM(-), IgG1(+) UnsortedEndoH EndoF EndoH EndoF

75

100

150

50

37

25

IgM

IgG1**

Igκ

IgG1

IgM*

IgM**

B

0

50

100

150

200

wildtype

OBI HC OBI OBIRAG1-/-

seru

mIg

G3

(μg/

mL)

0

50

100

150

200

wildtype

OBI HC OBI OBIRAG1-/-

seru

mIg

M(μ

g/m

L)

C

1.0%

IgM

IgG1

Fig. 5. A rare population of OBI B cells expresses IgM from the endogenousallele and is responsible for maintaining serum IgM levels. (A) Serum washarvested from 3-mo-old male B6, OBI TN heavy chain only (OBI HC), OB-IRAG-proficient (OBI) or OBI RAG1−/− mice, diluted 100-fold and analyzed byELISA for total IgM and total IgG3. Three individual mice per group areshown. Error bars show SD of triplicate samples. (B) Splenocytes from anOBI mouse were stained with antibodies to IgM and IgG1. Results arerepresentative of three independent experiments. The IgG1+IgM+ cellsand a comparable number of IgM−IgG1+ cells were FACS sorted and used inC. (C) IgM+IgG1+ or IgM−IgG1+ cells were sorted from OBI spleens by FACS,cultured for 3 d in media containing LPS and CpG. Cells were then labeledwith [35S]methionine/cysteine for 4 h. Supernatants were immunoprecipi-tated with anti-κ antibody to retrieve both IgM and IgG1, and immuno-precipates were digested with endoglycosidase H or endoglycosidase F. *,partially deglycosylated IgM; **, fully deglycosylated IgM or IgG1.

13742 | www.pnas.org/cgi/doi/10.1073/pnas.1210273109 Dougan et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 18

, 202

0

levels of total serum IgG3 and detectable serum titers of anti-ovalbumin IgG2a, IgG2b, and IgA.IgE, the major isotype responsible for asthma and allergies, is

generally present at low levels (29). We could not reliably detectantiovalbumin IgE in naïve OBI mice from 6 wk to 6 mo of age.Allergic responses in mice are generally induced by prior sensi-tization with the antigen. Given that the OBI mice already havea preexpanded pool of antigen-specific B cells and high serumtiters of specific antibodies, we were curious as to whether OBImice would phenocopy ovalbumin-sensitized mice with respectto food allergies. Administration of ovalbumin via the drinkingwater or by oral gavage of OBI mice did not induce an acutedrop in body temperature characteristic of anaphylaxis (30).These results are consistent with the lack of detectable IgE.Nevertheless, if properly sensitized or if induced to class switchto IgE, OBI B cells could be useful for studying allergic responsesversus oral tolerance to egg whites, a food that causes significantmorbidity in allergic humans.OBI mice show B-cell subset skewing toward marginal zone B

cells with a paucity of B-1 B cells, suggesting either no in-teraction with self-antigen or a weak signal strength of the OBIBCR (7, 8). This skewed MZ to B-1 B-cell ratio in OBI micecould not be reversed by 4 wk of continuous ovalbumin in thedrinking water. Peritoneal cavity (PC) B-1 B cells are capable ofT-cell–independent antibody secretion and tend to secrete IgMthat reacts weakly with common bacterial cell wall componentsor with viral constituents (20, 21, 31, 32). The PC B-1 cellcompartment, as well as CD5+ cells in spleen and lymph nodes,was reduced in OBI mice and absent in OBI RAG1−/− mice. Thefew B-1 B cells present in the peritoneal cavity of OBI mice wereenriched in IgM+IgG1+ double producers, with nearly all of the

CD5+ cells being IgM+, suggesting that direction of OBI B cellsinto the B-1 B-cell lineage is driven largely by the BCR specif-icities of the allelically included receptor.IgG1, with its long cytoplasmic tail, differs from IgM in terms of

its signaling properties (33, 34). Still, the OBI IgG1 supports B-celldevelopment and mediates nearly complete allelic exclusionin vivo. The OBI heavy chain confers much of the specificity forovalbumin, and mice bearing only the OBI heavy chain haveroughly 10% of the output of their B cells as immunoglobulinsspecific for ovalbumin, measured by immunoprecipitation. OBIHC mice, which possess polyclonal anti-ovalbumin antibodies inthe context of many other B-cell specificities, may more accuratelyreflect normal physiology and serve as a platform for vaccine de-velopment or therapies aimed at activating B-cell memory, espe-cially given that many memory B cells are IgG1+ (34).Here we report a B-cell transnuclear mouse specific for oval-

bumin. We defined the epitope, sequenced the rearranged VDJand VJ loci, and determined that B-cell development and subsetformation are, for the most part, normal. OBI B cells secreteanti-ovalbumin antibodies of nearly all isotypes, although isotypeswitching can be eliminated by crossing to a RAG-deficientbackground. The OBI mouse will be a valuable resource forstudies of antigen-specific B-cell responses.

Materials and MethodsAnimal Care. Animals were housed at the Whitehead Institute for BiomedicalResearch and maintained according to protocols approved by the Massa-chusetts Institute of Technology Committee on Animal Care. C57BL/6 andRAG1−/− mice were purchased from Jackson Labs. TN mice were generatedas previously described (13–15).

Sequencing of the BCR Genes. OBI RAG1−/− B cells were purified by negativeselection using CD43 magnetic beads (Miltenyi Biotec) and used as a sourceof RNA; 5′RACE was performed according to the manufacturer’s protocol(GeneRacer, L1502-01; Invitrogen).

Flow Cytometry. Cells from the indicated organs were subjected to hypotoniclysis, stained, and analyzed using a FACSCalibur (BD Pharmingen). Alexa 488and Alexa 588 ovalbumin were from Invitrogen. All antibodies were fromBD Pharmingen.

Serum ELISAs. Serum was collected from age-matched mice. High-binding 96-well plates (Costar) were coated overnight with 10 mg/mL ovalbumin (Sigma)or 2 μg/mL antimouse H+L (Southern Biotech) diluted in PBS. Plates werewashed three times with PBS, 0.05% Tween-20, blocked with 10% (vol/vol)FCS, washed three times, and incubated with serum. Serum samples wereused at the indicated dilutions. Hybridoma supernatant was used neat andat 10-fold serial dilutions. Plates were washed five times, and HRP-coupledsecondary antibodies recognizing IgM, IgG1, IgG2a, IgG2b, IgG2c, IgG3, IgA,Igκ, Igλ, or IgE [1 μg/mL in 10% (vol/vol) FCS; Southern Biotech] were added.Plates were washed seven times and bound antibody was detected using3,3′,5,5′-tetramethylbenzidine substrate.

Production of OBI Hybridoma. Hybridomas were generated and screened aspreviously described (35). Spleen cells from OBI HC mice were fused with NS-1cells. The resulting hybridomas were screened for ovalbumin reactivity by ELISA,and the Igκ genes were amplified by RT-PCR and sequenced. A single OVA-reactive hybridoma bearing the OBI IgG1 and IgVK135 genes was cultured for3 d in RPMI with low Ig FCS (Gibco), and supernatants were harvested.

Metabolic Labeling and Immunoprecipitation. B cells were cultured in RPMIwith 10% FBS. In some cases, LPS (20 μg/mL) and CpG (1 μM) were added tothe culture medium. For metabolic labeling, plasmablasts were starved for1 h in methionine- and cysteine-free medium, then labeled for 4–6 h with[35S]methionine/cysteine (PerkinElmer). Supernatants were harvested, andcells were lysed in Nonidet P-40 buffer. Supernantants and/or lysates wereanalyzed by immunoprecipitation, SDS/PAGE, and fluorography. Sequentialimmunopreciptitations were performed as described (36). Enzymatic degly-cosylation was performed using endoglycosidase H (Endo H) or PNGase F(New England Biolabs).

ACKNOWLEDGMENTS. We thank Patti Wisniewski and Chad Araneo forcell sorting and John Jackson for mouse husbandry. Oktay Kirak performed

IgG1

IgM

IgG1

CD

5

81

19

1.4%

IgG1

IgM

wild type OBI OBI RAG1-/-

0.028 2.11.5 0.0 0.0

75

100

150

50

37

25

IgM

IgG1

Igκ

IgG1

IgM

OBI

7

93

A

B C

Fig. 6. Peritoneal cavity is enriched in IgM+IgG1+ double producers. (A)Peritoneal cavity cells were isolated from wild-type, OBI, and OBI RAG1−/−

mice, stained with antibodies to IgM and IgG1 and analyzed by flowcytometry. Results are representative of three mice per group. (B) Totalperitoneal cavity cells were cultured for 3 d in media containing LPS andCpG. Cells were then labeled with [35S]methionine/cysteine for 4 h. Super-natants were immunoprecipitated with anti-κ antibody to retrieve both IgMand IgG1. (C) OBI peritoneal cavity cells were stained with the indicatedantibodies and analyzed by flow cytometry. Upper panel is gated on IgG1+

cells. Lower Right is gated on CD5+IgG1+ cells. Results are representative offour individual mice.

Dougan et al. PNAS | August 21, 2012 | vol. 109 | no. 34 | 13743

IMMUNOLO

GY

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 18

, 202

0

SCNT and the epitope mapping experiments. S.K.D. received a fellowshipfrom the Cancer Research Institute. S.O. was funded by Janssen Pharma-

ceutica NV. H.L.P. and R.J. are funded by grants from the NationalInstitutes of Health.

1. Brady BL, Steinel NC, Bassing CH (2010) Antigen receptor allelic exclusion: An updateand reappraisal. J Immunol 185:3801–3808.

2. Mostoslavsky R, Alt FW, Rajewsky K (2004) The lingering enigma of the allelic ex-clusion mechanism. Cell 118:539–544.

3. Barreto V, Cumano A (2000) Frequency and characterization of phenotypicIg heavy chain allelically included IgM-expressing B cells in mice. J Immunol 164:893–899.

4. Casellas R, et al. (2007) Igkappa allelic inclusion is a consequence of receptor editing.J Exp Med 204:153–160.

5. Velez MG, et al. (2007) Ig allotypic inclusion does not prevent B cell development orresponse. J Immunol 179:1049–1057.

6. Jung D, Giallourakis C, Mostoslavsky R, Alt FW (2006) Mechanism and control of V(D)Jrecombination at the immunoglobulin heavy chain locus. Annu Rev Immunol 24:541–570.

7. Pillai S, Cariappa A (2009) The follicular versus marginal zone B lymphocyte cell fatedecision. Nat Rev Immunol 9:767–777.

8. Pillai S, Cariappa A, Moran ST (2005) Marginal zone B cells. Annu Rev Immunol 23:161–196.

9. Shlomchik MJ, Weisel F (2012) Germinal center selection and the development ofmemory B and plasma cells. Immunol Rev 247:52–63.

10. Chaudhuri J, et al. (2007) Evolution of the immunoglobulin heavy chain class switchrecombination mechanism. Adv Immunol 94:157–214.

11. Hogquist KA, et al. (1994) T cell receptor antagonist peptides induce positive selec-tion. Cell 76:17–27.

12. Barnden MJ, Allison J, Heath WR, Carbone FR (1998) Defective TCR expression intransgenic mice constructed using cDNA-based alpha- and beta-chain genes underthe control of heterologous regulatory elements. Immunol Cell Biol 76:34–40.

13. Kirak O, et al. (2010) Transnuclear mice with predefined T cell receptor specificitiesagainst Toxoplasma gondii obtained via SCNT. Science 328:243–248.

14. Hochedlinger K, Jaenisch R (2002) Monoclonal mice generated by nuclear transferfrom mature B and T donor cells. Nature 415:1035–1038.

15. Kirak O, et al. (2010) Transnuclear mice with pre-defined T cell receptor specificitiesagainst Toxoplasma gondii obtained via SCNT. J Vis Exp 43:2168.

16. Pelanda R, et al. (1997) Receptor editing in a transgenic mouse model: Site, efficiency,and role in B cell tolerance and antibody diversification. Immunity 7:765–775.

17. Sonoda E, et al. (1997) B cell development under the condition of allelic inclusion.Immunity 6:225–233.

18. Kitamura D, et al. (1992) A critical role of lambda 5 protein in B cell development. Cell69:823–831.

19. Pelanda R, Schaal S, Torres RM, Rajewsky K (1996) A prematurely expressed Ig(kappa)transgene, but not V(kappa)J(kappa) gene segment targeted into the Ig(kappa) locus,can rescue B cell development in lambda5-deficient mice. Immunity 5:229–239.

20. Choi YS, Dieter JA, Rothaeusler K, Luo Z, Baumgarth N (2012) B-1 cells in the bone

marrow are a significant source of natural IgM. Eur J Immunol 42:120–129.21. Hastings WD, Gurdak SM, Tumang JR, Rothstein TL (2006) CD5+/Mac-1- peritoneal B

cells: A novel B cell subset that exhibits characteristics of B-1 cells. Immunol Lett 105:

90–96.22. Martin RM, Brady JL, Lew AM (1998) The need for IgG2c specific antiserum when

isotyping antibodies from C57BL/6 and NOD mice. J Immunol Methods 212:187–192.23. Moon JJ, et al. (2007) Naive CD4(+) T cell frequency varies for different epitopes and

predicts repertoire diversity and response magnitude. Immunity 27:203–213.24. Lozano F, Rada C, Jarvis JM, Milstein C (1993) Affinity maturation leads to differential

expression of multiple copies of a kappa light-chain transgene. Nature 363:271–273.25. Rajewsky K (1993) Immunology. The power of clonal selection. Nature 363:208.26. ten Boekel E, Melchers F, Rolink AG (1998) Precursor B cells showing H chain allelic

inclusion display allelic exclusion at the level of pre-B cell receptor surface expression.

Immunity 8:199–207.27. Hasan M, Polic B, Bralic M, Jonjic S, Rajewsky K (2002) Incomplete block of B cell

development and immunoglobulin production in mice carrying the muMT mutation

on the BALB/c background. Eur J Immunol 32:3463–3471.28. Dudley DD, Chaudhuri J, Bassing CH, Alt FW (2005) Mechanism and control of V(D)J

recombination versus class switch recombination: similarities and differences. Adv

Immunol 86:43–112.29. Hamilton RG (2010) Clinical laboratory assessment of immediate-type hypersensitiv-

ity. J Allergy Clin Immunol 125(2, Suppl 2):S284–S296.30. Proust B, et al. (2008) A single oral sensitization to peanut without adjuvant leads to

anaphylaxis in mice. Int Arch Allergy Immunol 146:212–218.31. Lam KP, Rajewsky K (1999) B cell antigen receptor specificity and surface density

together determine B-1 versus B-2 cell development. J Exp Med 190:471–477.32. Jayasekera JP, Moseman EA, Carroll MC (2007) Natural antibody and complement

mediate neutralization of influenza virus in the absence of prior immunity. J Virol 81:

3487–3494.33. Waisman A, et al. (2007) IgG1 B cell receptor signaling is inhibited by CD22 and

promotes the development of B cells whose survival is less dependent on Ig alpha/

beta. J Exp Med 204:747–758.34. Horikawa K, et al. (2007) Enhancement and suppression of signaling by the conserved

tail of IgG memory-type B cell antigen receptors. J Exp Med 204:759–769.35. Love JC, Ronan JL, Grotenbreg GM, van der Veen AG, Ploegh HL (2006) A micro-

engraving method for rapid selection of single cells producing antigen-specific anti-

bodies. Nat Biotechnol 24:703–707.36. McGehee AM, et al. (2009) XBP-1-deficient plasmablasts show normal protein folding

but altered glycosylation and lipid synthesis. J Immunol 183:3690–3699.

13744 | www.pnas.org/cgi/doi/10.1073/pnas.1210273109 Dougan et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 18

, 202

0