Corrections

APPLIED BIOLOGICAL SCIENCESCorrection for “An algorithm-based topographical biomaterialslibrary to instruct cell fate,” by Hemant V. Unadkat, MarcHulsman, Kamiel Cornelissen, Bernke J. Papenburg, Roman K.Truckenmüller, Anne E. Carpenter, Matthias Wessling, GerhardF. Post, Marc Uetz, Marcel J. T. Reinders, Dimitrios Stamatialis,Clemens A. van Blitterswijk, and Jan de Boer, which appearedin issue 40, October 4, 2011, of Proc Natl Acad Sci USA (108:16565–16570; first published September 26, 2011; 10.1073/pnas.1109861108).The authors note that the following statement should be added

to the Acknowledgments: “This work was supported in part byNIH Grant R01 GM089652 (A.E.C.).”

www.pnas.org/cgi/doi/10.1073/pnas.1302919110

EVOLUTIONCorrection for “Diversification of rhacophorid frogs providesevidence for accelerated faunal exchange between Indiaand Eurasia during the Oligocene,” by Jia-Tang Li, Yang Li,Sebastian Klaus, Ding-Qi Rao, David M. Hillis, and Ya-PingZhang, which appeared in issue 9, February 26, 2013, of ProcNatl Acad Sci USA (110:3441–3446; first published February11, 2013; 10.1073/pnas.1300881110).The authors note that, within the author line, “Yang Lia,c”

should instead appear as “Yang Lib,c”. The corrected author lineappears below. The online version has been corrected.

Jia-Tang Lia,b, Yang Lib,c, Sebastian Klausd, Ding-Qi Raoa,David M. Hillise, and Ya-Ping Zhanga,f

www.pnas.org/cgi/doi/10.1073/pnas.1304031110

IMMUNOLOGYCorrection for “Integrity of the AID serine-38 phosphorylationsite is critical for class switch recombination and somatic hy-permutation in mice,” by Hwei-Ling Cheng, Bao Q. Vuong,Uttiya Basu, Andrew Franklin, Bjoern Schwer, Jillian Astarita,Ryan T. Phan, Abhishek Datta, John Manis, Frederick W. Alt,and Jayanta Chaudhuri, which appeared in issue 8, February 24,2009, of Proc Natl Acad Sci USA (106:2717–2722; first publishedFebruary 5, 2009; 10.1073/pnas.0812304106).The authors note that the National Institutes of Health Grant

AI31541 should instead appear as AI077595.

www.pnas.org/cgi/doi/10.1073/pnas.1303069110

IMMUNOLOGYCorrection for “Alternative end-joining catalyzes robust IgHlocus deletions and translocations in the combined absence ofligase 4 and Ku70,” by Cristian Boboila, Mila Jankovic, CatherineT. Yan, Jing H.Wang, Duane R.Wesemann, Tingting Zhang, AlexFazeli, Lauren Feldman, Andre Nussenzweig, Michel Nussenzweig,and FrederickW.Alt, which appeared in issue 7, February 16, 2010,of Proc Natl Acad Sci USA (107:3034–3039; first published January25, 2010; 10.1073/pnas.0915067107).The authors note that the National Institutes of Health Grant

AI031541 should instead appear as AI077595.

www.pnas.org/cgi/doi/10.1073/pnas.1303073110

IMMUNOLOGYCorrection for “Downstream class switching leads to IgE anti-body production by B lymphocytes lacking IgM switch regions,”by Tingting Zhang, Andrew Franklin, Cristian Boboila, AmyMcQuay, Michael P. Gallagher, John P. Manis, Ahmed AmineKhamlichi, and Frederick W. Alt, which appeared in issue 7,February 16, 2010, of Proc Natl Acad Sci USA (107:3040–3045;first published February 1, 2010; 10.1073/pnas.0915072107).The authors note that the National Institutes of Health Grant

AI031541 should instead appear as AI077595.

www.pnas.org/cgi/doi/10.1073/pnas.1303075110

IMMUNOLOGYCorrection for “Robust chromosomal DNA repair via alternativeend-joining in the absence of X-ray repair cross-complementingprotein 1 (XRCC1),” by Cristian Boboila, Valentyn Oksenych,Monica Gostissa, Jing H. Wang, Shan Zha, Yu Zhang, Hua Chai,Cheng-Sheng Lee, Mila Jankovic, Liz-Marie Albertorio Saez,Michel C. Nussenzweig, Peter J.McKinnon, FrederickW.Alt, andBjoern Schwer, which appeared in issue 7, February 14, 2012, ofProc Natl Acad Sci USA (109:2473–2478; first published January30, 2012; 10.1073/pnas.1121470109).The authors note that the National Institutes of Health Grant

AI031541 should instead appear as AI077595.

www.pnas.org/cgi/doi/10.1073/pnas.1303078110

www.pnas.org PNAS | April 2, 2013 | vol. 110 | no. 14 | 5731

CORR

ECTIONS

Downstream class switching leads to IgE antibodyproduction by B lymphocytes lacking IgMswitch regionsTingting Zhanga,b,c,d, Andrew Franklina,b,c,d, Cristian Boboilaa,b,c,d, Amy McQuaya,b,c,d, Michael P. Gallaghera,b,c,d,John P. Manisb,e, Ahmed Amine Khamlichif, and Frederick W. Alta,b,c,d,1

aDepartment of Genetics, Harvard Medical School, Boston, MA 02115; bthe Children’s Hospital; cImmune Disease Institute; dHoward Hughes Medical Institute;eDepartment of Pathology, Harvard Medical School, Boston, MA 02115; and fCentre National de la Recherche Scientifique UMR 5089-IPBS, Université deToulouse, F-31077 Toulouse, France

Contributed by Frederick W. Alt, December 30, 2009 (sent for review December 22, 2009)

Ig heavy chain (IgH) class-switch recombination (CSR) replaces the IgHCμ constant regionexonswithoneof several sets ofdownstream IgHconstant region exons (e.g., Cγ, Cε, or Cα), which affects switchingfrom IgM to another IgH class (e.g., IgG, IgE, or IgA). Activation-induced cytidine deaminase (AID) initiates CSR by promoting DNAdouble-strand breaks (DSBs) within switch (S) regions flanking thedonor Cμ (Sμ) and a downstream acceptor CH (e.g., Sγ, Sε, Sα) thatare then joined to complete CSR. DSBs generated in Sμ frequently arejoined within Sμ to form internal switch region deletions (ISD). AID-induced ISDandmutationshavebeenconsidered rare indownstreamSregions, suggestingthatAID targetingto theseS regions requires itsprior recruitment to Sμ. We have now assayed for CSR and ISD in Bcells lackingSμ (Sμ−/−B cells). In Sμ−/−B cells activated forCSR to IgG1and IgE, CSR to IgG1wasgreatly reduced; but, surprisingly, CSR to IgEoccurred at nearly normal levels. Moreover, normal B cells had sub-stantial Sγ1 ISDand increasedmutations in andnearSγ1, and levels ofboth were greatly increased in Sμ−/− B cells. Finally, Sμ−/− B cellsunderwent downstream CSR between Sγ1 and Sε on alleles thatlacked SμCSR to these sequences.Ourfindings showthatAID targetsdownstreamSregions independentlyofCSRwithSμand implicateanalternative pathway for IgE class switching that involves generationand joining of DSBs within two different downstream S regionsbefore Sμ joining.

Antibodies arecomprisedof Igheavy (IgH)and light (IgL)chains.The N-terminal portion of IgH and IgL chains, termed the var-

iable region, binds antigens, whereas the C-terminal portion of theIgHchain, termed the constant region, determines antibody class andeffector functions. The IgH variable region is encoded in a distinctexon from those that encode the constant region. The variable regionexon is assembled early in B-cell development from VH, D, and JHsegments through V(D)J recombination (1). The IgH V(D)J exon isfirst expressed with proximal downstream exons that encode the Cμconstant region, leading to expression of μ IgH chains and IgMantibody.Newly generatedBcells express surface IgMandmigrate tothe periphery where they can be induced to express a different IgHclass (e.g., IgG, IgE, or IgA). Themouse IgH locus contains eight setsofCHexons, referred to as “CHgenes,” arranged as 5′-V(D)J-Cμ-Cδ-Cγ3-Cγ1-Cγ2b-Cγ2a-Cε-Cα-3′ over a 200-kb region (2). IgH classswitching involves a recombination/deletion event, termed class-switch recombination (CSR), in which the Cμ gene is replaced with adownstream CH gene (2, 3). Both CSR and the related somatichypermutation (SHM) process, which introduces mutations intovariable regionexons toallowproductionofhigheraffinityantibodies,are initiated by activation-induced cytidine deaminase (AID) (4, 5).Each CH gene that undergoes CSR is organized into a unit from

5′ to 3′ that includes a transcriptional promoter followed by anoncoding exon (termed I exon), a switch (S) region, and a set ofCH exons (6). CSR involves the introduction of a DNA double-strand break (DSB) in thedonorS regionflanking theCμ gene (Sμ)and into an acceptor S region flanking a downstreamCH gene; thisis followed by the joining of the breaks by general DSB end-joining

pathways (7). S regions are long (1–10 kb) intronic sequencescontaining characteristic repeated motifs that include AID targetmotifs, and the Sμ is the most repetitive and harbors the highestnumbers of AID targets (6, 8). Although there is some homologybetween Sμ, Sε, and Sα, there is little or no homology between Sμand Sγ regions (8). Transcription through S regions targets AIDactivity, which generates primary lesions that are processed intoDSBs in Sμ and the downstream acceptor S region required toinitiate CSR (7, 9). Thereby, specific induction of transcriptionfrom the I-region promoter flanking a particular acceptor S regiontargets that region for CSR. AID also introduces lesions into IgHand IgL variable region exons through a transcription dependentprocess, and they are converted into SHMs (9, 10). During CSR,AID also generates SHMs in S regions and immediate flankingsequences (11).S regions serve primarily as specializedDNA structures that target

the AID DSB-inducing activity. Thus, most CSR junctions occurwithin or occasionally just outside of S regions (12). In addition, Bcells that lack the donor Sμ are greatly impaired for CSR to all testedCH genes (13, 14). Correspondingly, deletion of Sγ1 abrogates CSRonly to Cγ1 (15). Finally, recombinational IgH CSR from IgM toIgG1 can be achieved in cells lacking those S regions whenDSBs areintroduced into their former sites by a yeast endonuclease (16).Thereare several mechanisms by which transcribed S regions may becomeAID targets. First, mammalian S regions form R loops that providesubstrates for the single-stranded DNA-specific activity of AID (17–19). Second, AID seems capable of gaining access to sequences, suchas S regions rich in AID target motifs, in a phosphorylation andReplication Protein A (RPA)-dependent fashion (20).AIDactivity cangeneratemultipleDSBswithinSμ (21, 22). Such

DSBs within a given S region can be religated, joined to anotherDSB in the same S region, or joined to a DSB in a downstreamS region to affect CSR.Religation of a resectedDSBwithin a givenS region or ligation of two intra-S region DSBs will generateinternal S-region deletions (ISD). ISD are usually assayed byactivating B cells for CSR and then, generating IgM-producinghybridomas from them; theses are assayed for ISD large enough tobe viewed by Southern blotting. These large ISD occur frequentlywithin the Sμ region inB cells orB-cell lines activated forCSR (23–26). They areAID-initiated (24) andmostly seem to occur throughend joining (27). However, prior studies have found ISD indownstream S regions to be quite rare (24–26, 28), leading tosuggestions that Sμ drives CSRby providing excess DSBs to ensure

Author contributions: T.Z., J.M., and F.W.A. designed research; T.Z., A.F., C.B., A.M., andM.G. performed research; A.A.K. contributed new reagents/analytic tools; T.Z., A.F., C.B.,A.M., J.M., and F.W.A. analyzed data; and T.Z. and F.W.A. wrote the paper.

The authors declare no conflict of interest.1To whom correspondence should be addressed. E-mail: alt@enders.tch.harvard.edu.

This article contains supporting information online at www.pnas.org/cgi/content/full/0915072107/DCSupplemental.

3040–3045 | PNAS | February 16, 2010 | vol. 107 | no. 7 www.pnas.org/cgi/doi/10.1073/pnas.0915072107

joining to less frequently occurring DSBs in downstream S regions(6, 16, 29).AIDactivity duringCSRalso generates SHMs in Sμ andadjacent sequences, including the upstream I region (29, 30).However, such mutations again were rarely found in downstreamCHgenes inwild-type (WT)B cells (29).The lack of SHMor ISD indownstream S regions also led to suggestions that AID activity onthese sequencesmayonly occur afterAIDhas actedonSμ, perhapsbecause of induction of factors or formation of complexes requiredfor AID access to downstream S regions (25, 29). However, AID-initiated mutations in downstream S regions were observed in thecontext of a DNA-repair-deficient background (30).IgH class switching to certain downstream CH genes (e.g., Cε)

often occurs in humans and mice through sequential CSRinvolving two downstream acceptor S regions (31–34). Amongthe lines of evidence for sequential CSR are findings of remnantsof an additional intervening S region in CSR junctions betweenSμ and a downstream S region. In particular, such findings sug-gest that most switching to IgE may involve a pathway in whichSμ first joins to Sγ1 to form a Sμ/Sγ1 fusion S region that sub-sequently switches to Sε (31–34). Although it is assumed thatsuch junctions occurred by the pathway outlined above, it wasnot ruled out that some might occur through a pathway in whichtwo downstream S regions are joined before joining with Sμ.However, such a pathway would require simultaneous DSBs intwo downstream S regions, which might be considered unlikelygiven the apparently low frequency of DSBs in downstream Sregions found in prior studies.Deletion of sequences upstream of Sμ, including the IgH JH

segments and the intronic enhancer element, inactivated CSRand led to increased Sγ1 ISD in appropriately activated B cells(35). Although there may be several explanations for this finding,we considered the possibility that inactivation of Sμmight lead toincreased AID targeting of downstream S regions. To test thispossibility and also better define potential CSR pathways, weassayed for IgH class switching, ISD and SHM of downstream Sregions, and downstream CSR between the Sγ1 and Sε in B cellsstimulated for CSR to IgG1 and IgE.

ResultsIgH CSR to IgE Occurs at Nearly Normal Levels in B Cells Lacking Sμ.To further elucidate mechanisms that target AID during IgHCSR, we have studied mice generated previously by gene-targetedmutation that harbor an essentially complete deletion of theentire 4.6-kb Sμ region in their germline (13). Mice homozygousfor this Sμ deletion mutation (Sμ−/− mice) show a great reductionin IgH CSR to all tested antibody classes, including IgG3, IgG2b,IgG1, IgG2a, and IgA (13). However, CSR to IgE was not pre-viously studied in these mice. To further characterize potentialIgH CSR defects in Sμ−/− mice, we purified splenic B cells fromWT and Sμ−/− spleens and stimulated them in vitro with anti-CD40 and IL-4 to stimulate CSR from IgM to IgG1 and IgE. Weassayed for CSR to IgG1 by surface staining and flow cytometry atdays 2, 3, and 4 of stimulation. Consistent with previous studies(13), we observed greatly decreased (between 5- and 10-fold) IgHclass switching to IgG1 at all analyzed time points (Fig. 1A andB).We confirmed this finding by hybridoma analyses that showedincreased levels of IgM-expressing hybridomas and greatlydecreased levels of IgG1-expressing hybridomas from Sμ−/− ver-sus WT B cells (Fig. 1C and Table S1).IgE class switching cannot be readily analyzed by flow cytom-

etry, because anti-CD40 treatment induces up-regulation ofCD23 (FcεRII), the Fc receptor for IgE, on activated B cells (36),which leads to binding of soluble IgE antibodies to non–IgE-producing B cells and causes false-positive staining. Therefore, weanalyzed IgE switching by assaying the frequency of IgE-pro-ducing hybridomas generated from day 4 anti-CD40 plus IL-4stimulated B cells. Under our current stimulation conditions,activated WT B cells switch to IgE at quite substantial levels;

indeed, ∼40% of the recovered hybridomas were IgE producers(Fig. 1C and Table S1) (27). In this regard, our stimulation con-ditions also seem to give very high levels of IgG1 class switching(Fig. 1C) (27, 37). Unexpectedly, Sμ−/− B cells also displayed veryhigh levels of IgE switching; we recovered IgE-expressing Sμ−/−hybrdomas at nearly 70% the frequency of WT IgE-producinghybridomas. Indeed, the frequency of IgE-expressing Sμ−/− Bhybridomas was more than 3-fold greater than that of IgG1-expressing Sμ−/− hybridomas (Fig. 1C and Table S1).To further elucidate the nature of the apparently high level of

IgE class switching in Sμ−/− B cells and to determine if it trulyrepresented CSR, we employed a PCR approach, using primersfrom Iμ and just downstream of Sε, to isolate potential CSRjunctions from a series of WT and Sμ−/− IgE-expressing hybrid-

A

B

C

IgM IgG1 IgE0

10

20

30

40

50

60

70

80WT

S -/-

Antibody class

1.5 2.0 2.5 3.0 3.5 4.0 4.50

10

20

30

40

50

60

70Wildtype

S -/-AID-/-

Day

100 101 102 103 104

100

101

102

103

104

9.2e-3 0.41

98.90.72

100 101 102 103 104

100

101

102

103

104

0.15 26.9

71.51.49

100 101 102 103 104

100

101

102

103

104

0 1.46

98.20.29

100 101 102 103 104

100

101

102

103

104

0 0.16

99.20.64

100 101 102 103 104

100

101

102

103

104

0.86 46.8

49.92.36

100 101 102 103 104

100

101

102

103

104

0 3.59

95.50.96

100 101 102 103 104

100

101

102

103

104

0 0.2

99.40.4

100 101 102 103 104

100

101

102

103

104

0 2.34

97.20.41

100 101 102 103 104

100

101

102

103

104

0 0.37

98.80.8

WT S -/- AID-/-

IgG

1

B220

Day 2

Day 3

Day 4

Fig. 1. IgG1 and IgE CSR in Sμ−/− B cells. (A) FACS analysis on days 2, 3, and 4 ofanti-CD40/IL-4–stimulated splenic B cells from WT, Sμ−/−, and AID−/− mice bysurface staining with anti–IgG1-FITC and anti-B220-PE-Cy5 antibodies. The per-centage of IgG1+B220+ cells represents CSR level to IgG1 at each time point. (B)Statistical analysis of three independent FACS analysis experiments for IgG1switching of anti-CD40/IL-4–stimulated B cells fromWT, Sμ−/−, andAID−/−mousespleens at day2, 3, and 4. (C) Hybridomaanalysis by supernatant ELISAon clonesgenerated from day 4 anti-CD40/IL-4–stimulated B-cell fusion with NS-1 fusionpartner cell line. IgM, IgG1, and IgE single-positive clones were counted andanalyzed for percentage in total single positive clones for each IgH isotype inthree independent fusionsonWTandSμ−/−B cells. SDs calculated fromthe threeexperiments are shown. Detailed numbers are listed in Table S1.

Zhang et al. PNAS | February 16, 2010 | vol. 107 | no. 7 | 3041

IMMUNOLO

GY

omas (Fig. S1Upper).Wewereable to isolate junctions from∼80%of assayed WT and Sμ−/− B-cell hybridomas, confirming that IgEswitching had occurred through a recombination/deletion event.The recovered WT junctions were similar to those previouslyreported (38). Thus, they frequently contained microhomology(MH) at the junctions and less frequently were direct with nooverlapping nucleotides, or they contained insertions. These WTSμ to Sε CSR junctions, similar to CSR junctions (31, 38), linkedsequences in the more 5′ portion of Sμ with sequences in the more3′ portion of Sε (Fig. S1), perhaps reflecting PCR biases of theassay, AID targeting hotspots, or continued AID activity and ISDwithin fused Sμ/Sε regions. CSR junctions from the Sμ−/− B-cellhybridomas occurred in a similar region of Sε as those recoveredfrom WT. As expected (13), the junctions mostly occurred in theregion just upstream of Sμ, but one junction occurred in a plasmidsequence inserted through gene targeting and the other occurredin the 5′ portion of Cμ (Fig. S1). Based on these findings, weconclude that B cells lacking the donor Sμ region utilize sequencesoutside of Sμ to undergo substantial levels of class switching to IgE.

Frequent Rearrangement of Sγ1 in Normal and Sμ-Deleted B Cells inthe Absence of CSR to Sμ. To further analyze factors that mayregulate AID targeting on downstream S regions, we assayed forISD in Sγ1 of IgM-secreting hybridomas generated from day 4anti-CD40 plus IL-4 stimulatedWT and Sμ−/−B cells. To assay forpotential rearrangements of Sγ1 that did not involve bona fideCSR, we digested genomic hybridoma DNA with EcoRI and thenused Southern blotting to assay for non-germline fragments thathybridized to an Iγ1 probe. The Iγ1 probe derives from sequencesjust upstreamof the Sγ1 region and hybridizes to an∼17-kbEcoR1fragment that encompasses the Iγ1 exon andSγ1 region; therefore,any rearrangement within this region (except for CSR, whichdeletes the Iγ1 exon) will alter the size of the EcoRI fragment thathybridizes to the Iγ1 probe (Fig. 2A Upper; Fig. S2). In markedcontrast to other studies, we found, in three independent sets ofIgM-secretingWT hybridomas, that nearly 20% of the IgH alleleshadundergone non–CSR-associated Sγ1 rearrangements (Fig. 2Aand B; Table S2). Thus, our findings indicate that, under our anti-CD40plus IL-4 activation conditions,AID targets the downstreamSγ1 region independently of Sμ to Sγ1 CSR on that IgH allele.Strikingly, we found that nearly 80% of IgM-producing hybrid-omas isolated from three sets of anti-CD40 plus IL-4–stimulatedSμ−/−B cells contained Sγ1 rearrangements on alleles in which Sγ1had not undergone bona fide CSR with Sμ (Fig. 2 A and B; TableS2). Thus, there is a substantial accumulation of apparent AIDtargeting events at Sγ1 in the absence of an upstream donor Sμ.

Frequent Mutation of the Iγ1/Sγ1 Region in the Absence ofRearrangement to Sμ. To further test whether or not AID cantarget downstream S regions independently of initiating a bonafide CSR event, we assayed for SHM of sequences within theregion covering the 3′ portion of Iγ1 through the 5′ portion of Sγ1after anti-CD40 plus IL-4 stimulation of WT and Sμ−/− B cells for7 days. This ∼1.6-kb region, which contains about 500 bp from the3′ end of the Iγ1 exon and then extends ∼1.1 kb into the 5′ portionof Sγ1, has been assayed similarly for AID-initiated hyper-mutations by other groups (29, 30). Analyses of mutations withinthe Iγ1/Sγ1 region from three sets of activated WT and controlAID−/−B cells revealed a significantly higher mutation level in theWT B cells (∼4 × 10−4/bp versus 3 × 10−5/bp) (Fig. 3A; Table S3).Moreover, activated Sμ−/− B cells had an even higher level ofmutations within the Iγ1/Sγ1 region (∼1.4 × 10−3/bp) (Fig. 3A). Inaddition, there was a higher level of mutations within this givensequence from the Sμ−/− samples than in WT samples (Fig. 3B).Among the scored mutations, which include point mutations,insertions, and deletions, there seemed to be an increased fre-quency of deletions in the Iγ1/Sγ1 sequences from the Sμ−/−samples (Table S3), potentially reflecting increasedAID targeting.

However, the overall spectrum of point mutations was similarbetweenWT and Sμ−/− sequences, consistent with the same intactmutation mechanism in both (Fig. S3). These results provide fur-ther evidence that AID targets downstream S regions independ-ently of Sμ andmoreover, thatAID-initiatedmutations are greatlyincreased in downstream S regions on IgH alleles that lack Sμ.

Frequent Recombination Events Between Sγ1 and Sε in the Absenceof Recombination with Sμ. To investigate AID targeting efficiencyat downstream S regions, we assayed for Sε rearrangements in thesame sets of IgM-producing hybridomas in which we assayed forSγ1 rearrangements.We first employed Southern blotting to assaygenomic DNA from these hybridomas for rearrangements withinSε. For this purpose, we assayed BamHI- or EcoRI-digested DNAfor hybridization to an Iεprobe and separately, for hybridization toa Cε probe to detect potential Sε ISD (Fig. 4AUpper; Fig. S4). Wenoted thatmost of these hybridomas clearly arose fromactivatedBcells that had not undergone CSR to Sγ1 involving Sμ but that hadundergone some type of Sγ1 rearrangement, because they con-tained two distinct, rearranged Iγ1-hybridizing alleles (Fig. 2A).Despite the apparently high level of IgE switching and particularlyhigh frequency of Sγ1 rearrangements that did not involve CSR

A

WT IgM+ hybridomas129/S

vJ

NS-1

GL* * * *- - - -

B

S -/- IgM+ hybridomas

GL** ** * * ** * ** *** * * ** * *** *

129/SvJ

*

Un

sw

itc

he

dre

arr

an

ge

d

S1 a

llele

s/T

ota

l

un

sw

itch

ed

allele

s (

%)

0

10

20

30

40

50

60

70

80

90

C 1S 1I 1

I 1 probe

RI RI RI

WT S -/-

Fig. 2. Sγ1rearrangementsonunswitchedalleles inWTandSμ−/− IgM-producinghybridomas. (A Top) Map of the Cγ1 gene. The EcoRI sites are indicated (RI), andthe Iγ1 probe is indicated. (Middle and Bottom) Southern blotting analysis ofEcoRI-digested genomic DNA extracted from IgM-producing hybridomas gen-erated from day 4 anti-CD40/IL-4–stimulated WT (Middle) and Sμ−/− (Bottom) Bcells for hybridization to an Iγ1 probe (18). Iγ1-hybridizing germline bands ofEcoRI-digested genomic DNA from WT (129/SvJ background) mouse kidneys areindicated with an arrow head and labeled as GL. Lanes with – on top show nobands hybridizing to the Iγ1probeor a JH4orCμprobe (used to showthat assayedhybridomas had rearranged JH alleles and to confirm their genotype as WT ormutant, respectively) (Fig. S2). These lanes were not counted in the calculationsshown in Fig. S2B, whereas lanes with * on top contain one or two unswitchedalleles that have undergone rearrangements of the Iγ1-hybridizing fragment. (B)Percentage of total unswitched Sγ1 alleles in WT and Sμ−/− IgM-producinghybridomas that contain rearrangements of the Iγ1-hybridzing fragments (pre-sumed Sγ1 rearrangements). SDs calculated from the three experiments areshown. Detailed numbers are listed in Table S1.

3042 | www.pnas.org/cgi/doi/10.1073/pnas.0915072107 Zhang et al.

with Sμ in Sμ−/−B cells, we detected only one potential Sε ISDwiththe Iε probe in 20 Sμ−/− IgM-producing hybridoma clones (Fig.S4). However, when DNA from the same set of Sμ−/− IgM-pro-ducing hybridomas was assayed by hybridization to a Cε probe, weobserved a substantial number of rearrangements (Fig. 4BUpper).Because many of these Sμ−/− IgM-producing hybridomas con-tained two rearranged Iγ1-hybridizing alleles (Fig. 2A), the rear-rangements detected with the Cε probe cannot all be explained bythe recombination events on the Sμ-deleted allele that lead to classswitching (Fig. S1). Additionally, as outlined above, they are not SεISD. Therefore, we considered the possibility that they representdownstream CSR events between Sγ1 and Sε. In support of thisnotion, stripping and reprobing these blots with an Iγ1 proberevealed that a substantial proportion of the Sε rearrangementsdetected with the Cε probe are exactly the same size as the Iγ1-hybridizing rearrangements (Fig. 4 A and B; common bands areindicated with asterisks).To directly test for downstream CSR between Sγ1 and Sε in

activated Sμ−/− cells, we assayed for Sγ1–Sε junctions in genomicDNA from a set of IgM-producing Sμ−/−-activated B-cell hybrid-omas through PCR with a forward primer derived from the 3′Iγ1region and reverse primer derived from the region immediatelydownstream of Sε (Fig. 4A Lower). In unrearranged genomicDNA, the two PCRprimers are derived from regionsmore than 60kb apart, and therefore, they would only be amplified if theyunderwent a rearrangement, such as an Sγ1 to Sε recombination,that placed them more proximal to each other. Although we didnot amplify bands by this approach from the WT IgM-secretingB-cell hybridoma DNA, we amplified bands of distinct sizes from15 of 42 Sμ−/− IgM-secreting B-cell hybridomas (Fig. 4C). DNA ofcloned Iγ1/downtream Sε PCR fragments from six Sμ−/− IgM-producing hybridomas confirmed that each harbored a unique Sγ1to Sε junction (Fig. S5). These findings show that Sγ1 and Sε canharbor AID-initiated breaks simultaneously and undergo a formof downstream CSR in the absence of bona fide CSR with Sμ.

DiscussionMechanisms that coordinate action of AID on donor and acceptorS regions are not fully understood. One hypothesis, based on priorfindings thatAIDseemed tohave little or noactivity ondownstreamS regions other than in the context ofCSR,was thatAIDmight haveto be recruited to Sμ and/or introduce lesions into Sμ to gaindownstream access (21, 25, 26). However, we now show that WTIgM-producing B cells activated for CSR to IgG1 accumulate ISDevents within Sγ1 and mutations within Iγ1/Sγ1, both hallmarks ofAID activity, on alleles that have not undergone CSR with Sμ. Onedifference between our studies and most earlier studies is that ourcurrent anti-CD40 plus IL-4–activation conditions allow muchhigher levels of IgG1and IgE switching (e.g., refs. 27 and 37 and thisstudy). Thus, these potentially increased-activation conditions mayfacilitate observation of AID targeting events in downstream S

A

B

0.0

2.5

5.0

7.5

10.0

12.5

15.0

17.5

0 1

2

3 4 6

73

0

1 2

3

4

5 6 7 8

65

WT S -/ -

Mu

tati

on

fre

qu

en

c y

(x10 -4

bp

)

I 1/S 1 mutation

WT S -/ -

Fig. 3. Mutations accumulate in the Iγ1/Sγ1 region in WT and Sμ−/− B cells.(A) Mutation frequency in the Iγ1/Sγ1 region of WT and Sμ−/− B cells. SDs arecalculated from three independent experiments. Detailed numbers are listedin Table S3. (B) The total number of Iγ1/Sγ1 sequences from WT and Sμ−/− Bcells analyzed is indicated in the center of each pie chart. The percentage ofindividual sequences containing different numbers of mutations is listed indifferent segments of the pie charts and is proportional to the segment sizesin the pie charts.

A

Iγ1

Sμ-/- IgM+ hybridomas

* * * * * *10 kb

8 kb

6 kb

C

- --B

GL

Cε* * * * * *

10 kb

8 kb

6 kb

GL/ NS-1

- --

-

PCR of Sγ1-Sε junctions

WT IgM+ hybridomas Sμ-/- IgM+ hybridomas (a158)

+

-

+

Sε.1-RvIγ1-Fw

Iγ1-Fw Sε.1-Rv

Cεprobe

Iγ1 probe

RI RI RI RI

RI RICεprobe

Iγ1 probe

RI Iεprobe

Iγ1 Sγ1 Cγ1 Iε Sε Cε

Iγ1 Sγ1/Sε Sε

*

*

Fig. 4. Sμ−/− IgM-producing hybridomas contain Sγ1–Sε junctions on IgHalleles that have not undergone CSR with Sμ. (A) Diagram of the Southernblotting and PCR strategies to detect Sγ1–Sε junctions. (A) Upper diagramshows germline Cγ1 and Cε genes, which are separated by 60 kb, and Lowerdiagram shows putative CSR product between the two genes. RI, EcoRI sites.The Cμ and Iγ1 Southern blot probes and the PCR primers used in the assayare also indicated. The PCR primers would only generate a band in therearranged allele (Lower) because of the long distance of their separation inthe germline configuration (Upper). (B Upper and Lower) Southern blot withCε and Iγ1 probes, respectively. Rearranged bands that cohybridize with theCε and Iγ1 probe are indicated with an asterisk. Digested NS-1 fusion-partnergenomic DNA hybridizes to the Cε probe at the same size as germlinegenomic DNA, which is shown as GL/NS-1 (NS-1). (C) PCR analyses of genomicDNA from a set of WT (Left) and Sμ−/− IgM+ hybridomas (Right). The same-sized bands observed in the Sμ−/− lines were observed in repeat analyses. Inanalyses of three separate sets of WT hybridomas no PCR bands wereobserved in 24 samples analyzed, whereas in analyses of three separate setsof Sμ−/− hybridomas, we observed bands in 15 of 42 samples analyzed, with10 representative samples illustrated here.

Zhang et al. PNAS | February 16, 2010 | vol. 107 | no. 7 | 3043

IMMUNOLO

GY

regions. Moreover, we find that Sμ−/− B cells activated for CSR toIgG1 and IgE accumulate greatly increased levels of Sγ1 ISD,downstreamCSRevents involving Sγ1 and Sε, andmutations in Iγ1/Sγ1. These findings clearly show that AID can access downstream Sregions without first interacting with Sμ. Finally, the downstreamCSR observed between Sγ1 and Sε, in the absence of CSR with Sμ,implies thatAIDactivity can lead to simultaneousDSBs in these twodownstream S regions, thereby suggesting an alternative pathwayfor sequential CSR (see below). In this regard, whereas thedependence of I-region promoter transcription and CSR to mostdownstreamS regions on the 3′ IgH regulatory regionmay precludeactivating more than one of these promoters at a time (39, 40), theIγ1 promoter functions relatively independently fromknown IgH 3′regulatory region elements and is activated by the same pathways asthe Iε promoter, which could allow simultaneous activation andfacilitate downstream Sγ1 to Sε CSR.We now show that downstream S regions can be targeted for

AID-induced mutations and DSBs in the absence of AID activityon Sμ and in the absence of CSR with Sμ. In this context, inde-pendent recent studies from our lab have confirmed this findingand have also confirmed previous observations that normal B cellsactivated for CSR to IgG1, under our current conditions, showmore ISD within Sμ compared with Sγ1 (24–28). Factors that leadto large ISD detectable by Southern blotting likely include DSBfrequency within an S region and aspects of S-region sequencethat might influence end-joining events with respect to rejoiningDSBs within an S region or to a DSB within a separate S region—potentially by using different end-joining pathways (27). Assum-ing that DSB frequency is a major factor, the higher level ofdetectable ISD at Sμ may reflect greater AID activity at Sμ thanother S regions because of its transcription, sequence, or otherfactors (29). In this context, a high frequency of DSBs in Sμ maywell drive CSR to ensure that DSBs in downstream S regions,which we now show can be independently introduced, find apartner Sμ DSB for CSR (7). This general model also has beensupported by our findings that independently introduced ISceI-endonuclease DSBs within the IgH locus can be joined at longrange by general cellular repair processes to promote recombi-national IgH class switching (16). Finally, the finding of greatlyincreased levels of AID-initiated events on downstream S regionsin activated Sμ−/− B cells is intriguing. One explanation for thisfinding is that such events simply accumulate in cells that cannotundergo efficient CSR in the absence of Sμ. A more interestingpossibility is that deletion of Sμ, which is such an efficient AIDtarget, allows AID to act more effectively on the downstream Sregions. Although only speculative, such a model could suggestfunctional roles for ISD in CSR by promoting increased AIDtargeting to downstream S regions.The Sμ deletion we have studied removes most Sμ AID target

sites (13), removes the major region of Sμ involved in R-loop ini-tiation (18), and greatly decreases the size of the CSR targetregion. Therefore, this deletion should vastly diminish the level ofDSBs that could serve as upstreamCSRdonors, and residualDSBspresumably should occur through low-level AID targeting by R-loop (41, 42) or non–R-loop mechanisms (7). Thus, given theapparent role for Sμ DSBs in driving CSR, the dramatic decreasein CSR to most IgH isotypes in Sμ−/− B cells was expected (13).However, our finding that IgE CSR in Sμ−/− B cells is relativelyunimpaired was indeed unexpected, especially considering thatCSR to IgG1 in the same cell population was severely decreased.One can conceive of several possible explanations for this finding,but we propose a general model that we consider attractive (Fig.S6). Sε CSR junctions in Sμ−/− cells must involve direct or indirectjoining of Sε to a presumably rare DSB in the remaining regionupstream of Cμ, suggesting that increased Sε DSBs might com-pensate for loss of SμDSBs in Sμ−/−Bcells and thereby, driveCSR.In this regard, Sμ−/−B cells showed increased ISD and SHM in Sγ1along with substantial levels of downstreamCSR between Sγ1 and

Sε. Thus, in the absence of Sμ, frequent Sγ1 breaks may drivedownstreamCSRwith Sε, potentially then leaving the fusedSγ1/Sεregion as a major driver for CSR with infrequent breaks upstreamof Cμ. If this model is correct, we must explain why Sε CSR junc-tions isolated from Sμ−/− cells lacked intervening Sγ1 regionsequences. Although such intermediate sequences might beeliminated by ongoing ISD in the hybrid Sγ1/Sε region, a moreintriguing possibility is that the Iγ1 promoter ismore effective thanthe Iε promoter in driving AID access to Sε and that, based on itsSμ homology, Sεmight serve as a better AID target than Sγ1 (Fig.S6). Finally, our current work raises the possibility that someobserved sequential CSR (31–34) might involve an alternativepathway in which the downstream CSR provides the firstintermediate sequence.

Materials and MethodsMice. Sμ−/− mice were generated previously (13) and were maintained with ahomozygous 129/SvJ IgH locus background. The genotype of each exper-imental mouse was confirmed by Southern blotting analysis. Sμ−/− mice wereanalyzed at 8–16 weeks of age with age-matched controls that included WTmice on 129/SvJ background and AID−/− mice (provided by Dr. TasukuHonjo). All experiments with mice followed the protocols approved byBoston Animal Care Facility of the Children’s Hospital, Boston, MA 02115.

CSR Assay and FACS Analysis. B cells were harvested from spleens of age-matched mutant mice and controls and purified by selection of the CD43-negative population through a mouse B-cell enrichment kit (STEMCELLTECHNOLOGIES). Purified B cells were cultured in RPMI-based mediumcontaining 50 μM β-ME and 15% FCS (as described in ref. 37) in the presenceof anti-CD40 antibodies (eBioscience) and IL-4 (20 ng/mL; eBioscience) tostimulate CSR to IgG1 and IgE. Cells were kept under 0.5 million/mL andassayed by surface staining and flow cytometry on days 2, 3, and 4 with anti-IgG1-FITC and anti-B220-PE-Cy5 antibodies as previously described (38).

Hybridoma Analysis for CSR. Five to ten million anti-CD40/IL-4–stimulated Bcells from each mouse were fused with NS-1 fusion partner myeloma cells onday 4 and recovered after 7 days selection with 1× Hypoxanthine Amino-pterin Thymidine (HAT) medium. Single clones from each well were pickedand screened by ELISA on their supernatants with IgM, IgG1, and IgE cap-turing and revealing antibodies (Southern Biotech). Only clones that weresingle positive for one of the three antibody classes were counted. Clonesthat were negative for all IgH isotypes or positive for more than one IgHisotype were found at very low levels (below 5% of total clones) and nottaken into the calculations.

Somatic Hypermutation Assay. Five to 10 million anti-CD40/IL-4–stimulated Bcells were harvested on day 7 for genomic DNA extraction with DNeasy Bloodand Tissue Kit (QIAGEN). iProof High-Fidelity DNA polymerase (Bio-Rad LifeSciences) was used to amplify the Iγ1/Sγ1 region with forward primer Sg1.2F(TGTCAATCCTGTTCTTAGTCAATCA) and reverse primer Sg1.2R (CCATCAGCTC-TAGCCATGTAGTATT). PCR products were purified and cloned into vectors aspreviously described (37). Plasmids with proper inserts were sequenced, and a1,640-bp region from the 3′ end of Ig1 into 5′ Sg1 was analyzed.

Southern Blotting Analysis. At least 5 μg genomic DNA isolated from eachhybridoma clone was digested with EcoRI or BamHI overnight and run on a0.7% agarose gel. DNA was transferred from the gels to membranes thatwere hybridized with corresponding probes (Figs. 2 and 4 and Figs. S2 andS4), washed, and put on XAR film (Kodak Biomax) for exposure.

Switch-Region Junction Cloning. Sμ–Sε and Sγ1–Sε junctions were amplified byAdvantagecDNApolymeraseMixandPCRKits (Clontech) following theirprotocols.For Sm–Se junction cloning, we used forward primer ImF1 (ACTCAGTCAGT-CAGTGGCGTGAAGGGCT)andreverseprimerSε-1.Rv (CATCAGGCTTTGCTCACTCA).For Sγ1–Sε junction cloning, we used forward primer Sg1.2F (TGTCAATCCTGTTCT-TAGTCAATCA) and reverse primer Sε-1.Rv (CATCAGGCTTTGCTCACTCA).

ACKNOWLEDGMENTS. We thank Jing H. Wang, Duane Wesemann, Yu NeeLee, Feilong Meng, and Michael G. Kharas for stimulating discussions andMichael Lieber and Barry Sleckman for critical review of the manuscript. Thiswork was supported by National Institutes of Health Grants AI031541 andCA092625 (to F.W.A.). C.B. is supported by a Cancer Research Institute traininggrant. F.W.A. is an investigator of the Howard Hughes Medical Institute.

3044 | www.pnas.org/cgi/doi/10.1073/pnas.0915072107 Zhang et al.

1. Dudley DD, Chaudhuri J, Bassing CH, Alt FW (2005) Mechanism and control of V(D)Jrecombination versus class switch recombination: Similarities and differences. AdvImmunol 86:43–112.

2. Honjo T, et al. (1981) Rearrangements of immunoglobulin genes during differentiationand evolution. Immunol Rev 59:33–67.

3. Honjo T (1983) Immunoglobulin genes. Annu Rev Immunol 1:499–528.4. Muramatsu M, et al. (2000) Class switch recombination and hypermutation require

activation-induced cytidine deaminase (AID), a potential RNA editing enzyme. Cell102:553–563.

5. Revy P, et al. (2000) Activation-induced cytidine deaminase (AID) deficiency causes theautosomal recessive form of the Hyper-IgM syndrome (HIGM2). Cell 102:565–575.

6. Chaudhuri J, Alt FW (2004) Class-switch recombination: Interplay of transcription,DNA deamination and DNA repair. Nat Rev Immunol 4:541–552.

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

8. Stavnezer J (1996) Immunoglobulin class switching. Curr Opin Immunol 8:199–205.9. Di Noia JM, Neuberger MS (2007) Molecular mechanisms of antibody somatic

hypermutation. Annu Rev Biochem 76:1–22.10. Li Z, Woo CJ, Iglesias-Ussel MD, Ronai D, Scharff MD (2004) The generation of

antibody diversity through somatic hypermutation and class switch recombination.Genes Dev 18:1–11.

11. Liu M, Schatz DG (2009) Balancing AID and DNA repair during somatichypermutation. Trends Immunol 30:173–181.

12. Dunnick W, Hertz GZ, Scappino L, Gritzmacher C (1993) DNA sequences atimmunoglobulin switch region recombination sites. Nucleic Acids Res 21:365–372.

13. Khamlichi AA, et al. (2004) Immunoglobulin class-switch recombination in micedevoid of any S mu tandem repeat. Blood 103:3828–3836.

14. Luby TM, Schrader CE, Stavnezer J, Selsing E (2001) The mu switch region tandemrepeats are important, but not required, for antibody class switch recombination. JExp Med 193:159–168.

15. Shinkura R, et al. (2003) The influence of transcriptional orientation on endogenousswitch region function. Nat Immunol 4:435–441.

16. Zarrin AA, et al. (2007) Antibody class switching mediated by yeast endonuclease-generated DNA breaks. Science 315:377–381.

17. Chaudhuri J, et al. (2003) Transcription-targeted DNA deamination by the AIDantibody diversification enzyme. Nature 422:726–730.

18. Chaudhuri J, Khuong C, Alt FW (2004) Replication protein A interacts with AID topromote deamination of somatic hypermutation targets. Nature 430:992–998.

19. Huang FT, et al. (2007) Sequence dependence of chromosomal R-loops at theimmunoglobulin heavy-chain Smu class switch region. Mol Cell Biol 27:5921–5932.

20. Yu K, Chedin F, Hsieh CL, Wilson TE, Lieber MR (2003) R-loops at immunoglobulin classswitch regions in the chromosomes of stimulated B cells. Nat Immunol 4:442–451.

21. Schrader CE, Linehan EK, Mochegova SN, Woodland RT, Stavnezer J (2005) InducibleDNA breaks in Ig S regions are dependent on AID and UNG. J Exp Med 202:561–568.

22. Stavnezer J, et al. (1999) Switch recombination in a transfected plasmid occurspreferentially in a B cell line that undergoes switch recombination of its chromosomalIg heavy chain genes. J Immunol 163:2028–2040.

23. Alt FW, Rosenberg N, Casanova RJ, Thomas E, Baltimore D (1982) Immunoglobulinheavy-chain expression and class switching in a murine leukaemia cell line. Nature296:325–331.

24. Dudley DD, et al. (2002) Internal IgH class switch region deletions are position-independent and enhanced by AID expression. Proc Natl Acad Sci USA 99:9984–9989.

25. Reina-San-Martin B, Chen J, Nussenzweig A, Nussenzweig MC (2007) Enhanced intra-switch region recombination during immunoglobulin class switch recombination in53BP1-/- B cells. Eur J Immunol 37:235–239.

26. Reina-San-Martin B, et al. (2003) H2AX is required for recombination betweenimmunoglobulin switch regions but not for intra-switch region recombination orsomatic hypermutation. J Exp Med 197:1767–1778.

27. Boboila C, et al. (2009) Alternative End-joining Catalyzes Robust IgH Locus Deletionsand Translocations in the Combined Absence of Ligase 4 and Ku. Proc Natl Acad SciUSA, in press.

28. Reina-San-Martin B, Chen HT, Nussenzweig A, Nussenzweig MC (2004) ATM isrequired for efficient recombination between immunoglobulin switch regions. J ExpMed 200:1103–1110.

29. Schrader CE, et al. (2003) Mutations occur in the Ig Smu region but rarely in Sgammaregions prior to class switch recombination. EMBO J 22:5893–5903.

30. Xue K, Rada C, Neuberger MS (2006) The in vivo pattern of AID targeting toimmunoglobulin switch regions deduced from mutation spectra in msh2-/- ung-/-mice. J Exp Med 203:2085–2094.

31. Du L, et al. (2008) Involvement of Artemis in nonhomologous end-joining duringimmunoglobulin class switch recombination. J Exp Med 205:3031–3040.

32. Mandler R, Finkelman FD, Levine AD, Snapper CM (1993) IL-4 induction of IgE classswitching by lipopolysaccharide-activated murine B cells occurs predominantlythrough sequential switching. J Immunol 150:407–418.

33. Mills FC, Mitchell MP, Harindranath N, Max EE (1995) Human Ig S gamma regions andtheir participation in sequential switching to IgE. J Immunol 155:3021–3036.

34. Snapper CM, Finkelman FD, Stefany D, Conrad DH, Paul WE (1988) IL-4 induces co-expression of intrinsic membrane IgG1 and IgE by murine B cells stimulated withlipopolysaccharide. J Immunol 141:489–498.

35. Gu H, Zou YR, Rajewsky K (1993) Independent control of immunoglobulin switchrecombination at individual switch regions evidenced through Cre-loxP-mediatedgene targeting. Cell 73:1155–1164.

36. Challa A, Pound JD, Armitage RJ, Gordon J (1999) Epitope-dependent synergism andantagonism between CD40 antibodies and soluble CD40 ligand for the regulation ofCD23 expression and IgE synthesis in human B cells. Allergy 54:576–583.

37. Cheng HL, et al. (2009) Integrity of the AID serine-38 phosphorylation site is critical forclass switch recombination and somatic hypermutation in mice. Proc Natl Acad SciUSA 106:2717–2722.

38. Yan CT, et al. (2007) IgH class switching and translocations use a robust non-classicalend-joining pathway. Nature 449:478–482.

39. Cogné M, et al. (1994) A class switch control region at the 3′ end of theimmunoglobulin heavy chain locus. Cell 77:737–747.

40. Seidl KJ, et al. (1999) Position-dependent inhibition of class-switch recombination byPGK-neor cassettes inserted into the immunoglobulin heavy chain constant regionlocus. Proc Natl Acad Sci USA 96:3000–3005.

41. Roy D, Lieber MR (2009) G clustering is important for the initiation of transcription-induced R-loops in vitro, whereas high G density without clustering is sufficientthereafter. Mol Cell Biol 29:3124–3133.

42. Roy D, Zhang Z, Lu Z, Hsieh CL, Lieber MR (2010) Competition between the RNAtranscript and the nontemplate DNA strand during R-loop formation in vitro: A nickcan serve as a strong R-loop initiation site. Mol Cell Biol 30:146–159.

Zhang et al. PNAS | February 16, 2010 | vol. 107 | no. 7 | 3045

IMMUNOLO

GY