Introduction

The generation of the antibody repertoire in man and mouseoccurs in two stages. The primary repertoire is achieved byintegrative rearrangement of V, D and J gene segments but,following antigen encounter, the immunoglobulin genes inthose B cells triggered by the antigen are now subjected to asecond wave of diversification, this time by somatic hyper-mutation. The hypermutation leads to the introduction ofmultiple single nucleotide substitutions in and around therearranged V gene segment, allowing the production of a secondary repertoire from which those B cells expressingmutated antibodies of improved antigen-binding quality canbe selected.

The mechanism of somatic hypermutation is unknown,but several features suggest a linkage to transcription. Thelocation of the mutation domain is determined by the positionof the transcription start site and recruitment of the muta-tional machinery is affected by the cis-acting transcriptionenhancer elements.1–5 Furthermore, an analysis of the muta-bility of modified immunoglobulin gene substrates has indicated a broad correlation between expression and muta-bility.6,7 However, transcription is not sufficient to ensuremutability and the mutation rates of immunoglobulin trans-genes may be sensitive to integration site.5,6,8,9

We have previously noted that, even with a single trans-genic animal, some germinal centre B cells express the trans-gene well and accumulate multiple nucleotide substitutionswithin the V domain, whereas in other germinal centre Bcells the transgene is poorly expressed and is essentiallyunmutated.6 Because the nucleotide substitutions are likely

incorporated in a stepwise manner with a small number ofsubstitutions being fixed in each cell cycle, we interpretedthese findings in terms of a model in which there is a degreeof clonal stability within the mouse with respect to both theexpression and mutability of the transgene. How might thestable heterogeneity be achieved? Either the mutating andnon-mutating clones differ in the availability of some trans-acting factor (which incidentally can distinguish the endoge-nous and transgenic Igκ loci) or the two types of clones mustdiffer in respect of cis-acting epigenetic modification of thetransgene. This prompted us to compare the methylationstatus of mutated and unmutated copies of the transgenic Vdomain in germinal centre B cells.

Materials and Methods

Transgenic mice

The generation and breeding of the transgenic mice has beendescribed previously.1,5,6,10

Bisulfite modification of DNA

DNA was prepared from tails by proteinase K digestion in the presence of SDS and phenol extraction, and from sorted Peyer’spatch B cells as described previously.11 Methylated dC were detectedby sequencing bisulfite-modified DNA (Fig. 1) as described pre-viously,12 except that DNA was purified from the bisulfite reactionstep using a Wizard DNA clean-up kit (Promega, Madison, WI,USA).

Polymerase chain reaction amplification and sequencing

In all cases, the region amplified spanned Jκ5 and 3′-flanking DNA.For the unrearranged Igκ locus, the forward primer (jol45: TTTTTg-catgcGGTTTATTTAGTTGGATTGGTT (nucleotides yielding arestriction site for subsequent cloning are shown in lower case))

Immunology and Cell Biology (2001) 79, 18–22

Research Article

Somatic hypermutation of immunoglobulin κ transgenes:Association of mutability with demethylation

CHRISTOPHER J JOLLY 1 , 2 and MICHAEL S NEUBERGER 1

1Medical Research Council Laboratory of Molecular Biology, Cambridge, UK and 2Centenary Institute of CancerMedicine and Cell Biology, Sydney, New South Wales, Australia

Summary Following antigen encounter, immunoglobulin genes are diversified by somatic hypermutation. Themechanism by which this mutational process preferentially targets immunoglobulin genes is not known, but islikely linked to transcription. However, transcription is not sufficient to ensure mutability. Here, by polymerasechain reaction amplification of bisulfite-modified DNA, the pattern of demethylation within the Igκ mutationdomain is analysed and transgenes are used to identify an association between demethylation and mutability. Inmice carrying an Igκ transgene that is well transcribed but only poorly targeted for hypermutation, the mutatedtransgene copies have been demethylated within the mutation domain, whereas the methylated copies remain unmu-tated. Thus, the hypermutation mechanism only acts on immunoglobulin gene targets that are demethylated as wellas transcribed, although transcription and demethylation do not themselves guarantee mutability.

Key words: DNA methylation, germinal centre, immunoglobulin mutation, mouse, Mus musculus, transgene.

Correspondence: CJ Jolly, Centenary Institute, Locked Bag 6,Newtown, NSW 2042, Australia. Email: c.jolly@centenary.usyd.edu.au

Received 15 June 2000; accepted 17 July 2000.

primes on a deaminated-version of a sequence immediately upstreamof Jκ5 (positions 1244–1276 in EMBL accession no. MMIGJK15).The forward primer for amplifying the 3′-flank of endogenous (non-transgenic) rearranged Jκ5 alleles (jol65: GGGGTTTTATTaagctt-TAGTGGTAGTGGG) is complementary to the deaminated versionof a consensus FR3 sequence common to many Vκ genes (see Klixet al.5) The forward primer for amplifying the corresponding regionon Lκ-derived transgenes (jol54: ATTgcatgcGTATTTTTAGGGGA-GAAGGTATT) primes in the CDR1 of VκOx1. The reverse primerfor all amplifications was jol42 (AATAtctagaTTCCTATCACTAT-ACCTCAAA, the antisense of a deaminated sequence close to theintronic MAR, positions 2444–2474 in MMIGJK15), with the excep-tion of the Lκ∆[Int

J-C,3′Fl] transgene, for which primer frosch56 was

used (CCATAACggatccTAACTATAAATTTTACCTC, antisense of adeaminated sequence just 3′ to the intronic HindIII site, positions3419–3450 in MMIG25).

Amplification was performed using Amplitaq Gold DNA poly-merase (Applied Biosystems, Foster City, CA, USA), a hot start(15 min 94°C), 15 touchdown cycles (94°C 1 min, annealing temperature 1 min, 72°C 2 min, with the annealing temperaturelowered from 64°C to 50°C by 1°C/cycle), followed by a further30–35 cycles (94°C 30 s, 50°C 30 s, 72°C 2 min). Amplified DNAwas then digested, cloned into M13 and sequenced using the M13forward primer, primer jol55 (GGGATGAGGAATGAAGGAA,MMIGJK15 positions 1676–94), and/or primer jol41 (GAAAG-CATGCTTGTTTGTTATGTAGATATTAT, deaminated MMIGJK15positions 1995–2026) and an Applied Biosystems 377 machine.Methylation sites and mutations were detected using the Gap4 DNAsequence alignment program.13

Results

Previous investigations into the methylation state of the Igloci have focused on the susceptibility to cleavage of the rarerestriction endonuclease target sequences within the loci thatcontain a CpG dinucleotide. Such experiments have revealedthat the relevant restriction sites located close to the J–Cintronic enhancers are methylated in the embryo and most

somatic cells, but are demethylated in B cells. In contrast, theHpaII or AvaI restriction sites in the vicinity of the constantregions remain methylated in mature and activated B cells,although they are usually demethylated in transformed celllines.14–23

Not knowing whether specific individual CpG dinu-cleotides might prove critical for transcription or hypermuta-tion, we used the bisulfite modification technique12 to allowa more detailed analysis of the pattern of CpG methylation inthe mutation domain (Fig. 1). Bisulfite modification deami-nates all unmethylated cytosines to uracils, but does notaffect methylated cytosines. When the modified DNA ispolymerase chain reaction (PCR) amplified with appropriateprimers and sequenced, the methylated cytosines appear as apositive display of C residues. Because methylation of mam-malian DNA nearly always occurs at the C of CpG dinu-cleotides,24 the efficiency of the deamination can be assessedby determining the extent to which the C residues that are notpart of a CpG dinucleotide in the parental sequence are sub-stituted by T in the modified sequences.

We focused on the region downstream of Jκ5, because thisregion has the highest density of CpG dinucleotides in the

Association of hypermutation and demethylation 19

Figure 1 Strategy used to determine methylation of dCnucleotides. (•), Methylated dC.

Figure 2 Methylation of (a) germline Igκ genes in tails and (b)rearranged Igκ genes in germinal centre B cells sorted fromPeyer’s patches (only including unique VJκ5 rearrangements).The maps show potential methylation sites (|, CpG). Deaminationefficiencies (± SEM) were (a) 100 ± 0.0% and (b) 96.2 ± 1.6%.Each row of dots represents one M13 clone. (�), non-methylatedCpG; (�), methylated CpG; ( ), not sequenced.

Igκ locus (Fig. 2) and is efficiently targeted for somaticmutation.5 In tail DNA, all seven CpG dinucleotides exam-ined were heavily (but not invariably) methylated. To examinethe CpG in the equivalent region of rearranged Igκ genes inB cells, we amplified VJκ5 rearrangements using a consen-sus Vκ FR3 primer together with a J–Cκ intron primer. TheVJκ-rearranged sequences were totally demethylated in ger-minal centre B cells (Fig. 2). The high number of PCR cyclesneeded for our analyses, together with the chemical treat-ment, resulted in occasional modifications to the DNA tem-plate (in addition to C demethylation), such that the error ratefrom the bisulfite/PCR was in the order of 1.3 mutations/kbin the final product rather than the 0.7 mutations/kb that weusually observe for PCR amplifications with Taq DNA poly-merase.1 Nevertheless, there was no difficulty in identifyingmutations above this background due to the in vivo somatichypermutation in the sequences obtained from Peyer’s patchgerminal centres (data not shown). Thus, using the bisul-fite/PCR strategy, it is possible to simultaneously monitorCpG methylation and somatic mutation.

Extending the study to Igκ transgenes (Fig. 3), weanalysed the region from Jκ5 into the Jκ–Cκ intron, thereby

spanning 23 CpG dinucleotides. With Lκ (which is a well-expressed transgene and a good hypermutation substrate), allthe CpG were nearly totally demethylated in peripheral Bcells (Fig. 3a). Only a total of eight unmodified CpG weredetected of the 318 CpG analysed, but interestingly half wereat the fifth CpG position. This is unlikely to reflect incom-plete deamination, because the deamination was > 99% efficient as judged by the fact that only three of 3036 non-CpG C in the Lκ database were found to have been leftunmodified. Thus, it appears that this fifth CpG dinucleotidemay be somewhat more refractory than the others todemethylation during B-cell development.

In contrast to what is observed with the complete Lκtransgene, the Jκ 3′ flank in an Lκ derivative that lacks theintronic enhancer/matrix attachment region (Lκ∆[Ei/MAR])remains substantially methylated in peripheral B cells(Fig. 3b). Of the 12 sequences analysed, eight are heavilymethylated, two are sporadically demethylated and none arecompletely demethylated.

Because the Lκ transgene is more efficiently demethy-lated than Lκ∆[Ei/MAR] and also constitutes a better hyper-mutation substrate, we were interested to know whether thedemethylated transgene templates in Lκ∆[Ei/MAR] dis-played an increased accumulation of mutations. However,this was difficult to analyse, because Lκ∆[Ei/MAR] accu-mulates few mutations and is only rarely demethylated and itwould thus be impossible to accumulate a sufficient database.We therefore focused our attention on mice harbouring atransgene that lacks Ei but retains MAR (Lκ∆[Ei]) and inwhich about one-quarter of the transgene copies cloned from germinal centre B cells have been targeted for mutation.6

Following amplification of bisulfite-modified DNA extractedfrom sorted Lκ∆[Ei] germinal centre B cells, 20 of 35 clonesderived from totally demethylated templates with the remain-der being variably methylated (Fig. 4).

CJ Jolly and MS Neuberger20

Figure 3 Methylation of transgenic Igκ sequences in Peyer’spatch B cells. Methylation site 7, as shown in Fig. 2, was deletedfrom the transgenes during their construction10 and methylationsite 6 was not sequenced. Mean (± SEM) demethylation percent-ages are shown in the histogram. The Lκ clones methylated at thefifth CpG derive from independent templates as judged by thepattern of linked C deamination. (�), non-methylated CpG; (�),methylated CpG; ( ), not sequenced.

Figure 4 Relationship between methylation of Lκ∆[Ei] trans-gene copies and their level of mutation in germinal centre B cellssorted from Peyer’s patches. Data were prepared as in Fig. 3. Thebackground mutation rate was determined from analysis ofsequences from sorted B220+ve, low peanut lectin-binding(resting) B cells.

Determination of mutation load among these sequences iscomplicated by two factors. First, the deamination/PCRprocess itself yields a background mutation rate somewhathigher than that of conventional genomic PCR. Second,because deamination causes unmethylated C to be read as T,then C to T transitions that had occurred in vivo on the sensestrand would be missed. However, neither of these consider-ations presents a major problem. With regard to deamina-tion/PCR errors, the contribution of somatic hypermutationis, as discussed earlier, readily discernible above this back-ground. With regard to C deamination, this will lead to aslight underestimate of mutation load, but the effect is mini-mized by the fact that C on the sense strand is one of the lessmutated bases in Lκ transgenes.25

Thus, even with these caveats in mind, analysis of the dis-tribution of somatic mutations among the Lκ∆α[Ei] ampli-fied sequences clearly reveals that the clones harbouring amutation load significantly above background all derive fromunmethylated templates (Fig. 4).

Because the mutated Igκ sequences that we had detected alloccurred on demethylated templates, we wondered if demethy-lation was sufficient for mutability. We therefore analysed themethylation status of an Igκ gene (Lκ∆[Int

J–C,3′Fl]) that we

have previously found to be very well expressed but whichbarely mutates.5 This transgene is well demethylated, as is

another well expressed but relatively poorly mutable Igκtransgene (LκTr5′[Ei/MAR];5 Fig. 5). Thus, while expressionand demethylation of the Igκ transgenes may be a prerequi-site for mutability, they are not sufficient.

Discussion

The results demonstrate that the lymphoid-specific demethy-lation of the Igκ locus that had previously been based on theendonuclease-sensitivity of rare restriction sites19,20,23 corre-lates with a global demethylation of CpG dinucleotidesdownstream of the rearranged Vκ. In the case of an Igκ trans-gene known to undergo only partial demethylation at theHhaI site in the Jκ–Cκ intron (Lκ∆[Ei/MAR]20,26), we havefound that this partial demethylation at a single restrictionsite is not a consequence of all the templates being incom-pletely demethylated (some templates demethylated at someCpG, other templates demethylated at other CpG). Rather, itreflects the existence of a mixed pool containing two majortypes of templates: one type is essentially totally demethy-lated over the Jκ 3′ flank, while the other is wholly methy-lated. Thus, the methylation status of the intronic HhaI site isa reasonable monitor for the overall methylation state of theVJκ mutation domain. Therefore, our finding that regionaldemethylation of the Jκ 3′ flank is compromised both in thetransgenes lacking Ei/MAR and in those lacking Ei alonefully supports the conclusions of earlier work indicating theimportance of these elements in Igκ demethylation.27

The major conclusion of the present work, however, is thatthere is an association between local demethylation andmutability. Most significantly, even within one animal inwhich the transgene is only partially demethylated, mutationis restricted to the demethylated templates.

Thus, there is an intricate interrelationship between tran-scription, demethylation and mutability. With regard todemethylation and transcription, it has long been known thatthere is a correlation between the two processes, althoughdemethylation is not a prerequisite for transcription. So muchis evident from the Lκ∆[Ei/MAR] transgene, which isheavily methylated but quite well expressed; similar conclu-sions have been reached previously with regard to theendogenous Igκ locus.28 With regard to mutability, it appearsthat both transcription (as established in earlier work1,2,6,7) anddemethylation (the present work) are associated with, but notsufficient for, mutability. All of the mutated transgene copiesthat we have detected are demethylated, but the fact thatdemethylation is not sufficient to ensure mutability is wit-nessed by the lack of mutability of Lκ∆[Int

J–C,3′Fl] and the

poor mutability of the LκTr5′[Ei/MAR] transgenes (Fig. 5).Thus, a mutable transgene is one that is both transcribed anddemethylated (rather than merely transcribed), although tran-scription and demethylation are not sufficient to guaranteemutability.

The association between mutation and demethylation doesnot imply that it is demethylation itself that is needed toensure mutability, although that possibility is not excluded.Rather, the data indicate that demethylation correlates withan additional feature that is required over and above tran-scription. Such a feature could be chromatin modification,template localization within the nucleus, transcription rates,or the precise nature of the transcription complexes. Further

Association of hypermutation and demethylation 21

Figure 5 Methylation of transgenic Igκ sequences in Peyer’spatch germinal centre B cells from (a) Lκ∆[Int

J-c, 3′Fl] and (b)

LκTr5′[Ei/MAR] mice. (�), non-methylated CpG; (�), methy-lated CpG.

CJ Jolly and MS Neuberger22

progress in identifying what is needed to ensure mutabilitymay well be facilitated by the availability of lymphoma celllines that perform somatic hypermutation constitutivelyrather than by extending on the analysis of immunoglobulintransgenes in mice as described here.

Acknowledgements

We thank Norman Klix and Andrew Riddell for assistancewith cell sorting and Cristina Rada for providing oligonu-cleotide frosch56 and for sharing unpublished data. We aregrateful to the Howard Hughes Medical Institute for an Inter-national Research Scholar’s award to MSN.

References

1 Betz AG, Milstein C, Gonzalez FA, Pannell R, Larson T, Neuberger MS. Elements regulating somatic hypermutation ofan immunoglobulin κ gene: Critical role for the intronenhancer/matrix attachment region. Cell 1994; 77: 239–48.

2 Peters A, Storb U. Somatic hypermutation of immunoglobulingenes is linked to transcription initiation. Immunity 1996; 4:57–65.

3 Rada C, Yelamos J, Dean W, Milstein C. The 5′ hypermutationboundary of κ chains is independent of local and neighbouringsequences and related to the distance from the initiation of tran-scription. Eur. J. Immunol. 1997; 27: 3115–20.

4 Tumas-Brundage K, Manser T. The transcriptional promoter regulates hypermutation of the antibody heavy chain locus. J.Exp. Med. 1997; 185: 239–50.

5 Klix N, Jolly CJ, Davies SL, Brügemann M, Williams GT, Neuberger MS. Multiple sequences from downstream of the Jκcluster can combine to recruit somatic hypermutation to aheterologous, upstream mutation domain. Eur. J. Immunol.1998; 28: 317–26.

6 Goyenechea B, Klix N, Yélamos J et al. Cells strongly express-ing Igκ transgenes show clonal recruitment of hypermutation: Arole for both MAR and the enhancers. EMBO J. 1997; 16:3987–94.

7 Fukita Y, Jacobs H, Rajewsky K. Somatic hypermutation in theheavy chain locus correlates with transcription. Immunity 1998;9: 105–14.

8 Bertocci B, Quint L, Delbos F, Garcia C, Reynaud C-A, Weill J-C. Probing immunoglobulin gene hypermutation withmicrosatellites suggests a nonreplicative short patch DNA synthesis process. Immunity 1998; 9: 257–65.

9 Tumas-Brundage KM, Vora KA, Manser T. Evaluation of therole of the 3′α heavy chain enhancer [3′α E(hs1,2)] in Vh genesomatic hypermutation. Mol. Immunol. 1997; 34: 367–78.

10 Sharpe MJ, Milstein C, Jarvis JM, Neuberger MS. Somatichypermutation of immunoglobulin κ may depend on sequences3′ of Cκ and occurs on passenger transgenes. EMBO J. 1991; 10:2139–45.

11 González-Fernández A, Milstein C. Analysis of somatic hyper-mutation in mouse Peyer’s patches using immunoglobulin κlight-chain transgenes. Proc. Natl Acad. Sci. USA 1993; 90:9862–6.

12 McDonald LE, Kay GF. Methylation analysis using bisulfitegenomic sequencing: Application to small numbers of intactcells. Biotechniques 1997; 22: 272–4.

13 Bonfield JK, Smith K, Staden R. A new DNA sequence assem-bly program. Nucleic Acids Res. 1995; 23: 4992–9.

14 Akira S, Sugiyama H, Sakaguchi N, Kishimoto T. Immunoglob-ulin gene expression and DNA methylation in murine pre-B celllines. EMBO J. 1984; 3: 677–81.

15 Blackman MA, Koshland ME. Specific 5′ and 3′ regions of theµ-chain gene are undermethylated at distinct stages of B-cell dif-ferentiation. Proc. Natl Acad. Sci. USA 1985; 82: 3809–13.

16 Burger C, Radbruch A. Demethylation of the constant regiongenes of immunoglobulins reflects the differentiation state of theB cell. Mol. Immunol. 1992; 29: 1105–12.

17 Kafri T, Ariel M, Brandeis M et al. Developmental pattern ofgene-specific DNA methylation in the mouse embryo and germline. Genes Dev. 1992; 6: 705–14.

18 Kelley DE, Pollok BA, Atchison ML, Perry RP. The couplingbetween enhancer activity and hypomethylation of κimmunoglobulin genes is developmentally regulated. Mol. Cell.Biol. 1988; 8: 930–7.

19 Mather EL, Perry RP. Methylation status and DNase I sensitiv-ity of immunoglobulin genes: Changes associated withrearrangement. Proc. Natl Acad. Sci. USA 1983; 80: 4689–93.

20 Mostoslavsky R, Singh N, Kirillov A et al. κ chain monoallelicdemethylation and the establishment of allelic exclusion. GenesDev. 1998; 12: 1801–11.

21 Rogers J, Wall R. Immunoglobulin heavy chain genes: demethy-lation accompanies class switching. Proc. Natl Acad. Sci. USA1981; 78: 7497–501.

22 Stavnezer-Nordgren J, Sirlin S. Specificity of immunoglobulinheavy chain switch correlates with activity of germline heavychain genes prior to switching. EMBO J. 1986; 5: 95–102.

23 Storb U, Arp B. Methylation patterns of immunoglobulin genesin lymphoid cells: Correlation of expression and differentiationwith undermethylation. Proc. Natl Acad. Sci. USA 1983; 80:6642–6.

24 Clark SJ, Harrison J, Frommer M. CpNpG methylation in mammalian cells. Nat. Genet. 1995; 10: 20–7.

25 Milstein C, Neuberger M, Staden R. Both DNA strands of anti-body genes are hypermutation targets. Proc. Natl Acad. Sci. USA1998; 95: 8791–4.

26 Neuberger MS, Ehrenstein MR, Klix N et al. Monitoring andinterpreting the intrinsic features of somatic hypermutation.Immunol. Rev. 1998; 162: 107–16.

27 Lichtenstein M, Keini G, Cedar H, Bergman Y. B cell-specificdemethylation: A novel role for the intronic κ chain enhancersequence. Cell 1994; 76: 913–23.

28 Kirillov A, Kistler B, Mostoslavsky R, Cedar H, Wirth T,Bergman Y. A role for nuclear NF-κB in B-cell-specificdemethylation of the Igκ locus. Nat. Genet. 1996; 13: 435–41.