Double-stranded DNA break polarity skews repair pathway ...

Double-stranded DNA break polarity skews repairpathway choice during intrachromosomal andinterchromosomal recombinationAlexanda K. Linga,1, Clare C. Soa,1, Michael X. Lea, Audrey Y. Chena, Lisa Hunga, and Alberto Martina,2

aDepartment of Immunology, University of Toronto, Toronto, ON, Canada M5S1A8

Edited by Matthew D. Scharff, Albert Einstein College of Medicine of Yeshiva Uni, Bronx, NY, and approved January 29, 2018 (received for review December1, 2017)

Activation-induced cytidine deaminase (AID) inflicts DNA damage atIg genes to initiate class switch recombination (CSR) and chromo-somal translocations. However, the DNA lesions formed during theseprocesses retain an element of randomness, and thus knowledge ofthe relationship between specific DNA lesions and AID-mediatedprocesses remains incomplete. To identify necessary and sufficientDNA lesions in CSR, the Cas9 endonuclease and nickase variants wereused to program DNA lesions at a greater degree of predictabilitythan is achievable with conventional induction of CSR. Herewe showthat Cas9-mediated nicks separated by up to 250 nucleotides onopposite strands can mediate CSR. Staggered double-stranded breaks(DSBs) result in more end resection and junctional microhomologythan blunt DSBs. Moreover, Myc-Igh chromosomal translocations,which are carried out primarily by alternative end joining (A-EJ), werepreferentially induced by 5′ DSBs. These data indicate that DSBs with5′ overhangs skew intrachromosomal and interchromosomal end-joining toward A-EJ. In addition to lending potential insight to AID-mediated phenomena, this work has broader carryover implicationsin DNA repair and lymphomagenesis.

activation-induced cytidine deaminase | class switch recombination |chromosomal translocation | Cas9 | double-stranded DNA repair

During the germinal center reaction, activation-induced cyti-dine deaminase (AID) catalyzes DNA lesions at the Ig heavy

chain (Igh) locus that ultimately become substrates for class switchrecombination (CSR) (1). AID deaminates deoxycytidines toproduce deoxyuridines at repetitive sequences upstream of Igconstant regions known as switch regions (2, 3), which are engagedby base excision repair and mismatch repair pathways to generatenicks or longer gaps (4). These nicks and gaps on opposite DNAstrands give rise to double-stranded DNA breaks (DSBs) in donorand acceptor switch regions (5), which are then synapsed and li-gated together by nonhomologous end-joining (NHEJ) or alter-native end-joining (A-EJ) (6–10), thereby completing CSR andintroducing the expression of a new Ig isotype. Occasionally, AID-induced DSBs at switch regions are joined with AID-induced DSBsin other chromosomes, creating a chromosomal translocation (11).There are two fundamental and longstanding problems associ-

ated with studying the molecular mechanism of AID-dependentrecombination events. First, switch regions are composed of long,repetitive stretches of AID hotspot sequences, making it difficult toprecisely know the cleavage sites and insertional and/or deletionalevents before final ligation. Second, because AID initiates nicks onboth strands of switch regions (12, 13), the resulting DSB inter-mediates represent a spectrum of polarities (i.e., blunt, 5′, or 3′overhang) and varying lengths of overhanging single-strandedDNA. Furthermore, it is unknown how far apart nicks on oppo-site DNA strands can be while still resolving as a DSB, or whichtypes of DSBs are necessary, sufficient, or preferred for CSR andchromosomal translocations. Moreover, it is unknown whetherNHEJ and A-EJ favor repair of a specific subset of these lesionsand what conditions lead to the use of one pathway over the other.

While there are model systems that examine CSR and chro-mosomal translocations by providing transgenic substrates (14,15), these systems are limited in recapitulating the spectrum ofDNA lesions initiated by AID. In this work, we applied CRISPR/Cas9 technology to generate specific DNA lesions at AID targetloci to model AID-induced events in vivo, and characterizedDSB repair factors that promote or inhibit CSR and chromo-somal translocations. We demonstrate that DSB polarity influ-ences the frequency of CSR and Myc-Igh translocations andmicrohomology use at recombination junctions.

ResultsCas9- and Cas9n Nickase-Mediated DSBs Induce CSR in Mouse Cells.To characterize the specific DNA lesions that are sufficient toinduce CSR, we used CRISPR/Cas9 to generate specific DSBs inthe CH12F3-2 mouse B cell line (CH12 cells hereinafter). TheseCas9-generated DNA lesions mimic DSBs that arise downstreamof AID-mediated deamination and base excision. We designedsingle guide RNAs (sgRNAs) targeting upstream of switch regionμ (S′μ) and downstream of switch region α (S′α) (16) to avoid therepetitive elements within the switch regions while still targetingregions of AID-mediated CSR (17, 18). These regions have beenshown to be physiological sites of AID activity during CSR (13, 18,19). To eliminate any contributions by AID to lesions in S′μ, we usedAicda (Aid)−/− CH12 cells (16). Aid−/− CH12 cells transiently

Significance

DNA lesions initiated by activation-induced cytidine deaminase(AID) during class switch recombination (CSR) are diverse, andknowledge of the relationship between specific DNA lesions andDNA repair outcomes remains incomplete. Using the Cas9 endo-nuclease, we show that Cas9 models AID-mediated lesions leadingto CSR and chromosomal translocations, and that double-strandedDNA breaks (DSBs) with 5′ overhangs skew intrachromosomal andinterchromosomal end-joining toward alternative end-joining. Thesedata establish a role for DSB end polarity in choosing the DNA repairpathway. This work provides a greater understanding of how di-verse DNA lesions are differentially repaired, whichwill improve ourunderstanding of themechanismof physiological processes requiredfor adaptive immunity and protective mechanisms that maintaingenomic integrity from diverse threats.

Author contributions: A.K.L., C.C.S., and A.M. designed research; A.K.L., C.C.S., M.X.L.,A.Y.C., and L.H. performed research; A.K.L., C.C.S., and A.M. analyzed data; and A.K.L.,C.C.S., and A.M. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Published under the PNAS license.1A.K.L. and C.C.S. contributed equally to this work.2To whom correspondence should be addressed. Email: [email protected].

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1720962115/-/DCSupplemental.

Published online February 22, 2018.

2800–2805 | PNAS | March 13, 2018 | vol. 115 | no. 11 www.pnas.org/cgi/doi/10.1073/pnas.1720962115

Dow

nloa

ded

by g

uest

on

Dec

embe

r 11

, 202

1

transfected with S′μ and S′α sgRNA undergo Cas9-mediatedswitching from IgM to IgA or to other isotypes by changing theacceptor switch region sgRNA (Fig. 1A and Fig. S1A). Cas9-mediated switching to IgA can be induced in different mouse Bcell lines (Fig. S1 B and C). Multiple DSBs in the S′μ and S′αregions increased Cas9-mediated switching to IgA (Fig. S1D),supporting a report indicating that a multiplicity of breaks withinswitch regions increases CSR (20).To determine whether the Cas9-mediated switching can re-

capitulate other properties of CSR, we assessed two other phe-nomena associated with AID-mediated CSR. First, since ∼7% ofIgA CSR events occur between homologous chromosomes intrans (21, 22), we examined whether Cas9-mediated switchingcan likewise occur in trans. The native KpnI and StuI restrictionsites in the 3′ UTR of Igha on the V(D)J-rearranged chromo-some were “scarred” with Cas9 to generate a cell line in whichonly the unrearranged homologous allele maintained these re-striction sites (Fig. 1B). After inducing Cas9-mediated switchingin these clones and performing reverse transcription PCR of theIgha 3′ UTR followed by restriction digestion, we observed that∼2% of Cas9-mediated switching occurred in trans (Fig. 1B andFig. S1E). Second, although AID-mediated CSR preferentiallyfavors deletional recombination of DSBs in donor and acceptorswitch regions leading to productive Ig expression, recombination

can also instead lead to inversion of the intervening region, resultingin loss of Ig expression (Fig. 1C and Fig. S1F) (23). In contrast toAID-mediated CSR (Fig. 1D), we found that Cas9-mediated switch-ing had a similar proportion of Iglo (i.e., inversional recombination) toIgA+ (i.e., deletional recombination) cells (Fig. 1E), in agreement witha previous report examining I-SceI–mediated switching (23). Weconfirmed that Iglo cells in Cas9-mediated switching truly representedinversional recombination by amplifying a stronger inversion PCRsignal in sorted Iglo compared with IgA+ cells (Fig. S1G). Taken to-gether, these data suggest that Cas9-mediated switching recapitulatessome properties of AID-mediated CSR.

Cas9 Nickase-Mediated Switching Reveals a Preference for 5′ DSBsover 3′ DSBs. AID deaminates dC to dU, which is then engagedby uracil DNA glycosylase (UNG) to generate an abasic site that iscleaved by AP endonuclease to induce a nick in the DNA. Tomodel AID-mediated nicks in vitro, we used the Cas9 RuvC-null(D10A) or the HNH-null (N863A) nickases together with a pair ofproximal sgRNAs to create precise staggered DSBs with a 5′overhang (5′ DSB) or a 3′ overhang (3′ DSB), respectively (24,25). To test whether Cas9-induced 5′DSBs can mediate switching,we transfected two S′μ and two S′α sgRNA-Cas9D10A vectorssimultaneously into Aid−/− CH12 cells to produce 5′ DSBs withvarious overhang lengths (Fig. 1F). S′μ and S′α overhangs do not

Fig. 1. Cas9-mediated DSBs can recapitulate CSR and associated recombination. (A) Percentage of switched Aid−/− CH12 cells when transfected withS’μ1 sgRNA only or in combination with a sgRNA targeting the relevant acceptor switch region. (B, Left) Schematic of two Igh alleles present in CH12 cells. Therearranged allele (VDJ) has been edited using CRISPR-Cas9 to lack KpnI and StuI restriction sites in the 3′ UTR downstream of Cα exon 3 (the “scarred” allele),while the unrearranged allele (WT) retains these restriction sites. (B, Right) To detect trans CSR, KpnI and StuI sites in the 3′ UTR of Igha were scarred withCas9 on the V(D)J-rearranged chromosome. After inducing CSR and amplifying the 3′ UTR of IgA from cDNA, restriction digestion was performed, followed byelectrophoresis and band intensity quantitation (Fig. S1E). (C) Schematics showing the unswitched Igh locus (Top), the deletional CSR that leads to IgA ex-pression (Middle), and the inversional CSR that leads to loss of Ig expression (Bottom). (D) Percentage of WT CH12 cells that have undergone IgA+ CSR, orinversion of the Sμ–Sα intervening region (Iglo) at 48 h after stimulation with CIT (αCD-40, IL-4, and TGF-β) (Fig. S1 F and G). (E) Percentage of Aid−/− CH12 cellsthat have undergone deletional recombination (IgA+) or inversion of the S′μ-S′α region (Iglo). ΔCμ is a sgRNA that targets Cμ exon 1, leading to loss of IgMexpression, and serves as a positive control for Ig loss. (F) Schematic of Igh locus and sgRNA target regions in S′μ and S′α. The distance between theCas9 cleavage site in sense- and antisense-targeting sgRNAs in nucleotides is denoted. (G) Percentage of Aid−/− CH12 cells that have undergone CSR to IgA ontransfection with two pairs of sgRNAs and Cas9D10A to generate 5′ DSBs in S′μ and S′α. Combinations of sgRNAs used in each experiment are denoted bycolored blocks below and correspond to the sgRNAs depicted in A. (H) Same as G, except that one or two sgRNAs have been removed from the transfection.(I) Same as G, except that CasN863A was used to generate 3′ DSBs in S′μ and S′α. (J) Ratio of IgA-switched Aid−/− CH12 cells transfected with Cas9D10A to thosetransfected with Cas9N863A, with the sgRNA combinations used held constant. pX330, empty vector control. Error bars represent SDs.

Ling et al. PNAS | March 13, 2018 | vol. 115 | no. 11 | 2801

IMMUNOLO

GYAND

INFLAMMATION

Dow

nloa

ded

by g

uest

on

Dec

embe

r 11

, 202

1

share significant sequence similarity with each other. The 5′ DSBsinduced switching up to 3% (Fig. 1G), a threefold lower frequencythan that achieved with wild-type Cas9 (Fig. 1A), perhapsreflecting the preference for a blunt DSB over a staggered DSBduring CSR. Surprisingly, a pair of sgRNAs creating a 248-nt 5′overhang was sufficient for Cas9-induced switching (Fig. 1G).Removal of any one sgRNA from transfection greatly diminishedCas9-induced switching (Fig. 1H), suggesting that paired opposingnicks indeed resolve as staggered DSBs to complete the reaction.Similarly, targeting sgRNA to only one strand, in mimicry of theprocessive catalysis of AID in vitro (26), yielded little switching(Fig. S2). The 3′ DSBs created by Cas9N863A can also mediateswitching (Fig. 1I). Although Cas9N863A has similar enzymaticactivity to Cas9D10A (27), we observed a fivefold reduction inCas9-mediated switching using Cas9N863A compared withCas9D10A keeping DSB overhang lengths constant in S′μ and S′α(35 and 24 nt, respectively) (Fig. 1J). As the overhang length isincreased, 5′ DSBs persist as better substrates for Cas9-mediatedswitching to IgA than 3′ DSBs (Fig. 1J). These results suggest therole of a 5′ to 3′ exonuclease or endonuclease that converts distalnicks into staggered DSBs (Fig. S3A; Discussion). We concludethat staggered DSBs of different polarities and with nicks sepa-rated by as much as ∼250 nt can induce switching, although 5′DSBs are preferred over 3′ DSBs.

5′ and 3′ DSBs Promote A-EJ Use During CSR. DSBs created duringCSR are joined by the NHEJ and A-EJ pathways, the latter ofwhich are often characterized by repair junctions with greaterresection and microhomology use compared with junctionsrepaired by NHEJ (7). Since AID initiates staggered DSBs (5),we hypothesized that DSBs with different end polarities anddifferent lengths of ssDNA overhang may be resolved by dif-ferent DNA repair pathways to complete CSR. Thus, we se-quenced individual Cas9 and nickase-mediated S′μ–S′α junctionsfrom Aid−/− CH12 cells to quantify resection and microhomologyuse. Blunt DSBs generated by Cas9 produced junctions with lowlevels of resection (Fig. 2A and Fig. S3B). Using the same sgRNApairs to create a 35-nt overhang in S′μ and a 24-nt overhang in S′α,S′μ–S′α junctions arising from 5′ DSB intermediates exhibited a

greater median resection than 3′ DSBs, although the distribu-tions of resection were not statistically different (Fig. 2A and Fig.S3C). The same trends were observed when the overhang lengthin S′α was increased from 35 nt to 98 nt (Fig. 2A and Fig. S3C).Both 5′ and 3′ DSBs, regardless of overhang length, were resectedto a significantly greater degree than blunt DSBs during Cas9-mediated switching.We next measured the amount of microhomology usage in S′μ–S′α

junctions. 5′ and 3′ DSBs, regardless of overhang length, led tojunctions with increased microhomology compared with thosearising from blunt DSBs, suggesting the predominance of A-EJin the repair of staggered DSBs (Fig. 2B). We observed a positivecorrelation between resection and microhomology use in S′μ–S′αjunctions arising from blunt DSBs in Aid−/− CH12 cells, sug-gesting that blunt DSBs in NHEJ-proficient cells can also pro-duce junctions that are characteristic of A-EJ (Fig. 2C), althoughto a lesser degree than that achieved with 5′ DSBs or 3′ DSBs.Our data thus far suggest that 5′ and 3′ DSBs bias end-joining

toward A-EJ during Cas9-mediated switching compared withblunt DSBs. To determine whether 5′ and 3′ DSBs are indeedpreferential substrates for A-EJ compared with blunt DSBs, weexamined Cas9-mediated switching using blunt, 5′, and 3′ DSBs inAid−/− CH12 cells deficient in ligase IV, a core component ofNHEJ (Fig. S4) (28, 29). We reasoned that if the disparity in μ−αresection and microhomology between junctions from blunt versus5′ or 3′ DSBs in Aid−/− cells were no longer observed in Aid−/−

Lig4−/− cells, then the repair of 5′ and 3′DSBs in NHEJ-proficientcells might truly be driven by A-EJ. Previously, resection could notbe measured from AID-mediated CSR junctions due to the ran-dom targeting of switch region cytidines by AID. Resection at μ−αswitch junctions was increased in ligase IV-deficient cells acrossblunt, 5′, and 3′ DSB intermediates (Fig. 2D and Fig. S3D). Asexpected, microhomology use at μ−α switch junctions was in-creased across blunt, 5′, and 3′ DSB intermediates (Fig. 2E),consistent with increased microhomology use at μ−α switchjunctions in AID-initiated CSR in Lig4−/− CH12 cells (16, 29).Importantly, we no longer observed the disparity in μ−α resectionand microhomology between junctions from blunt versus 5′ or 3′DSBs in Aid−/−Lig4−/− cells. Collectively, these data suggest that 5′

Fig. 2. The μ–α junctions derived from 5′ DSB intermediates have increased resection. (A) Total resection of S′μ and S′α (Top) and microhomology use(Bottom) at μ–α junctions derived from Aid−/− CH12 cells transfected with various sgRNA combinations and Cas9 nickase variants. The median resection lengthis denoted by a black line, and the blue line denotes the sum of overhang lengths generated by Cas9D10A and CasN863A in the S′μ and S′α. (B) Scatterplot ofmicrohomology use from junctions derived from different DSB intermediates, with mean microhomology indicated. (C) Spearman’s correlation between totalresection and microhomology use in μ–α junctions induced by blunt junctions in Aid−/− CH12 cells. (D) Same as A except that μ–α junctions were derived fromAid−/− Lig4−/− CH12 cells. (E) Same as B except that μ–α junctions were derived from Aid−/− Lig4−/− CH12 cells. ***P < 0.001. ns, not significant.

2802 | www.pnas.org/cgi/doi/10.1073/pnas.1720962115 Ling et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 11

, 202

1

and 3′ DSBs skew end-joining toward A-EJ during Cas9-mediatedswitching in contrast to blunt DSBs.

5′ DSBs Skew Repair of Interchromosomal DSBs to A-EJ to CompleteMyc-Igh Translocations. To further investigate whether A-EJ has apreference for joining specific DNA ends, we used CRISPR/Cas9 to model a second AID-mediated genetic transaction, theMyc-Igh chromosomal translocation, that is believed to arisepredominantly from A-EJ (11, 30). We transiently transfectedAid−/− CH12 cells with plasmids encoding wild-type Cas9 andsgRNAs targeting S′μ (mouse chromosome 12) and the 3′ end ofMyc exon 1 (mouse chromosome 15) (Fig. 3A). Myc exon 1 andintron 1–2 are physiological breakpoint sites determined fromMyc-Igh fusions taken from Burkitt’s lymphoma patient samples(31). We detected head-to-head fusion of the Myc coding regionto the Igh locus on mouse chromosome 12, hereinafter denotedas der(12), by nested PCR (Fig. 3B). Through limiting dilution ofgenome equivalents per nested PCR, we determined that themedian frequency of Cas9-mediated der(12) was 0.27 × 10−5

translocations per cell (Fig. 3C), similar to frequencies induced by I-SceI endonuclease, Cre recombinase, or zinc-finger nucleases (15,32). A single DSB in S′μ was not sufficient to induce der(12) (Fig.3B), supporting the notion that DSBs in two partner chromosomesare required for translocation (15). Consistent with previous reports,deficiency in ligase IV increased der(12) frequency by ∼10-fold

(Fig. 3C), supporting the notion that NHEJ suppresses chro-mosomal translocation formation in mouse cells (30, 33).Because AID initiates CSR through staggered DSB intermedi-

ates, we next tested whether 5′ DSBs or 3′ DSBs could triggerder(12) using Cas9D10A and N863A nickases and pairs of proxi-mal sgRNAs targetingMyc and Igh. We found that 5′DSBs inducedder(12) at an equivalent frequency to that induced by blunt DSBs(Fig. 3C), in contrast to the threefold reduction in Cas9-mediatedswitching induced by 5′DSBs compared with blunt DSBs (Fig. 1 AandG). Strikingly, der(12) induced by 3′DSBs in ligase IV-sufficientor -deficient cells were mostly below the limit of detection (<0.031 ×10−5 translocations per cell) (Fig. 3C), indicating that chromosomaltranslocations do not favor 3′ DSBs as intermediates.Der(12) junctions derived from blunt DSBs in Aid−/− CH12 cells

exhibited minimal resection and microhomology use; however, ei-ther the presence of 5′ DSB intermediates or a deficiency in ligaseIV alone was sufficient to increase resection and microhomologyuse at der(12) junctions (Fig. 3 D and E). Most der(12) junctionsarising from 5′ DSBs in Aid−/− CH12 cells were resected to thesummed length of Myc and S′μ 5′ overhangs (Fig. 3D, blue line),while some junctions exhibited filling-in of the overhang or re-section beyond both Myc and S′μ overhangs. As in S′μ–S′α junc-tions derived from Cas9-mediated switching, we also observed apositive correlation between microhomology use and resection ofMyc and S′μ from Cas9-induced der(12) junctions derived from

Fig. 3. Cas9-mediated Myc-Igh translocations are driven primarily by blunt and 5′ DSBs, but not by 3′ DSBs. (A) Schematic of Myc and S′μ genomic loci andpositions of sgRNA target sequences for inducing der(12). sgRNAs are indicated with red bars and nested PCR primers are indicated with blue arrows.(B) Representative gel electrophoresis images of der(12) nested PCR products amplified from Aid−/− and Aid−/−Lig4−/− CH12 cells via Cas9-mediated blunt (Top)and 5′ DSB (Bottom) intermediates. Each lane represents an independent PCR containing the denoted number of genome equivalents as input. (C) Fre-quencies of der(12) translocations in Aid−/− and two independent Aid−/−Lig4−/− CH12 clones induced by blunt, 5′, and 3′ DSBs as determined by nested PCR. (D)Total resection of S′μ and Myc at der(12) junctions arising from blunt and 5′ DSBs in Aid−/− and Aid−/−Lig4−/− CH12 cells. The blue line denotes the sum ofoverhang lengths generated by Cas9D10A in Myc and S′μ. (E) Scatterplot showing the distribution of microhomology use at der(12) junctions arising fromblunt and 5′ DSBs in Aid−/− and Aid−/−Lig4−/− CH12 cells, with mean microhomology indicated. (F) Spearman correlation between total resection andmicrohomology use in der(12) junctions induced by blunt DSBs in Aid−/− CH12 cells. **P < 0.01, ***P < 0.001. ns, not significant.

Ling et al. PNAS | March 13, 2018 | vol. 115 | no. 11 | 2803

IMMUNOLO

GYAND

INFLAMMATION

Dow

nloa

ded

by g

uest

on

Dec

embe

r 11

, 202

1

blunt DSBs in Aid−/− CH12 (Fig. 3F), providing further evidencethat A-EJ is active in NHEJ-proficient cells and can repair DNAends thought to be repaired by NHEJ. Importantly, the resection ofMyc and S′μ overhangs in Aid−/−Lig4−/− CH12 cells was greaterthan the sum of the two overhang lengths (Fig. 3D). We also ob-served an increase in the proportion of der(12) junctions inducedby 5′ DSBs that exhibited templated, complex insertions in AID−/−

CH12 cells (Fig. S5), often composed of sequences homologous tothe S′μ and Myc overhangs. Interestingly, these templated inser-tions were ligase IV-dependent (Fig. S5B). Taken together, ourresults suggest that 5′ DSBs skew interchromosomal end-joiningtoward increased resection and microhomology use, two charac-teristics of A-EJ, in contrast to blunt DSBs.

DiscussionIn this work, we used CRISPR/Cas9 to study the DNA lesionsrequired for CSR and AID-associated chromosomal transloca-tions. By modeling these AID-dependent recombination eventsvia Cas9-induced blunt, 5′, and 3′ DSB intermediates, we foundthat 5′ DSBs induced Cas9-mediated switching and transloca-tions at greater frequencies than 3′ DSBs. 5′ DSBs functioned asbetter substrates for Cas9-mediated switching than 3′ DSBs asoverhang length was increased. Importantly, the Cas9 nickase-induced overhangs used in this study may mimic those generatedin vivo by AID, as suggested by duplications of 9–266 bp of aCSR substrate in CH12 cells, presumably resulting from theformation and filling in of AID-dependent staggered DSBs (34).These results also suggest that a high density of AID-induced

mutations at switch regions is not required for CSR, especially ifdistal nicks are oriented to lead to 5′DSBs. The tolerance for distalnicks as precursors to staggered DSBs are complemented by aprevious finding that nicks 900 nt apart can induce homologousrecombination (HR) (35). The preference for 5′ DSBs over 3′DSBs as substrates for Cas9-mediated switching also implicates a 5′to 3′ exonuclease or endonuclease to facilitate the generation of aDNA end (Fig. S3A), since nicks separated by long distances mightnot necessarily spontaneously “melt” into a staggered DSB. Thispreference for staggered DSBs as substrates for A-EJ is supportedby a previous report demonstrating that knockdown of AID leadingto reduced deamination and a lower density of nicks in the switchregions is correlated with enriched A-EJ activity (18). Mechanisti-cally, we speculate that overhangs would be a poorer substrate forthe Ku complex and therefore predispose to A-EJ.Our results also establish a role for DSB end polarity in de-

termining the choice of DNA repair pathway, in addition to knowncell cycle influences (36, 37). The DNA damage response to dif-ferent prescribed DNA lesions has been investigated in vitro (38,39) and most recently in vivo in yeast (40). We show that 5′ and 3′DSBs undergo increased resection relative to blunt DSBs and leadto intrachromosomal recombination with increased microhomology,while only 5′ DSBs share these characteristics when inducing in-terchromosomal repair. It is not immediately clear why 3′ DSBs arepoorer substrates for chromosomal translocations compared with 5′DSBs. We speculate that in vivo conversion of a 5′ DSB into a 3′DSB by an exonuclease often occurs during the iterative end-joiningprocess. Indeed, creation of a 3′ DSB is required to trigger strandinvasion during HR, yet Cas9 nickase-generated 3′ DSBs induceHR less frequently than 5′ DSBs in a GFP reporter assay (35).These results suggest that the choice of end-joining pathway isinfluenced by DSB polarity, as well as by whether ligation oc-curs within or between chromosomes.The varying degree of resection observed at μ–α and Myc–Igh

junctions derived from 5′ DSBs suggests different simultaneousmodes of resection before ligation. For example, resection equal tothe sum of the partner overhangs suggests the removal of theoverhangs by a flap endonuclease, yielding a blunt DSB for liga-tion. In contrast, resection less than the sum of the partner over-hangs implicates filling in of the ssDNA overhang by a polymerase

before ligation. Resection beyond the sum of the 5′DSB overhangssuggests the conversion of 5′DSBs into 3′DSBs before repair. Ourresults suggest that 5′ DSBs in wild-type cells are repaired byall three of these mechanisms, leading to CSR or chromosomaltranslocations, while complete resection of the overhang and be-yond dominates repair of 5′ DSBs in NHEJ-deficient cells. Whileour model system effectively captures the initial substrates and finalstate of end-joining, future experiments are needed to observe thelikely iterative process of resection and filling before final ligation.In the present study, we have examined molecular characteris-

tics of end-joining from only productive recombination events, i.e.,successful Cas9-mediated switching or chromosomal transloca-tions. Nonproductive repair outcomes at a single target locus, suchas indel formation or faithful repair by NHEJ or HR, are alsolikely outcomes and are the focus of previous studies (27, 41, 42).HR factors, such as BRCA1 and Rad52, have been shown to in-hibit CSR (43, 44). It is possible that HR-mediated repair maycompete for DSB ends at the expense of CSR; indeed, Rad52 hasbeen shown to compete with Ku70/80 for binding to a blunt SμDNA probe in an electrophoretic mobility shift assay. Since ourstudy selectively analyzed DNA repair leading to productive CSR,we did not investigate the role of HR in inhibiting CSR. Exam-ining how specific DSB ends are preferential substrates for CSRvs. HR would be an interesting follow-up study.We acknowledge that Cas9 does not recapitulate all aspects of the

role(s) of AID in CSR. DSB formation in vivo requires AID-mediatedcytidine deamination, followed by uridine excision by UNGand nicking of the phosphodiester backbone by AP endonuclease(12, 45). Therefore, our model examines DNA repair leading toCSR or chromosomal translocation at least two enzymatic stepsdownstream of AID activity. Moreover, it is possible that AIDmight act as a DNA repair factor scaffold, as previously suggested(46), which Cas9 is unlikely to do. Future work along these linescould proceed by substituting the Cas9 nickases used in this studywith catalytically dead Cas9 fused to a cytidine deaminase and/orcatalytically inactivating AID to preserve any scaffolding function.We also recognize that there may be slight differences in enzy-

matic activity or mutagenicity among wild-type Cas9, RuvC-null, andHNH-null Cas9 that could affect our results. Cas9D10A generatedindels with paired sgRNAs at a similar frequency to wild-type Cas9with either individual sgRNA, although indel formation is a poorproxy for cutting efficiency (25). While RuvC-null and HNH-nullCas9 were previously demonstrated to have similar nickase activity(27, 34), Cas9 was shown to release target and nontarget strandsasymmetrically and thus may potentially bias the accessibilityof repair in a strand-dependent fashion (47, 48). Deciphering thepotential idiosyncratic influences of nucleases on modeling AID-mediated phenomena will require additional work, as well as im-proved understanding and implementation of CRISPR technology.In conclusion, our study delineates the changing DNA damage

response to varying DSB structures and highlights the ability ofthe DSB structure to bias end-joining between NHEJ and A-EJpathways. Overall, better knowledge of how diverse DNA lesionsare differentially repaired will improve our understanding of themechanisms of the physiological processes required for adaptiveimmunity and the protective mechanisms that maintain genomicintegrity from diverse threats.

MethodsCell Culture and Transfection. All murine B cell lines were cultured in RPMI1640 medium supplemented with 2 mM L-glutamine, 10% FBS, 5% NCTC-109,0.5 mM β-mercaptoethanol, and penicillin/streptomycin and cultured at 37 °C/5%CO2. CH12 cells were generally electroporated with 4–5 μg of each relevantplasmid unless otherwise indicated, in 4-mm cuvettes with an exponential waveat 325 V, 975 μF, and∞ Ω in a Bio-Rad GenePulser Xcell. Cells were harvested at3 d posttransient transfection for CSR and translocation analysis.

2804 | www.pnas.org/cgi/doi/10.1073/pnas.1720962115 Ling et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 11

, 202

1

CRISPR-Cas9 Vector Construction. The sgRNAs used are summarized in TableS1. sgRNAs were cloned into CRISPR-Cas9 vectors as described previously(40). All sgRNAs used in this study were chosen based on favorable predictedon- and off-target activity (49, 50). SpCas9 bearing an N863A mutation wasobtained from Feng Zhang, MIT, Cambridge, MA, and substituted intopX330 to generate a Cas9N863A CRISPR vector.

CRISPR-Mediated Gene Knockout in CH12 Cells. Aid−/−Lig4−/− CH12 clones1 and 2 were generated by electroporating Aid−/− CH12 cells, provided byKefei Yu, Michigan State University, East Lansing, MI, with 4 μg of Lig4_G2(Table S1) targeting exon 2 of murine Lig4 and expanded in completeRPMI medium. Individual clones were generated by single cell dilution in96-well tissue culture plates. Genomic DNA was harvested from singleclones at 7 d after plating by proteinase K digestion (20 μg/mL) at 55 °Cfor 1 h, followed by heat inactivation at 95 °C for 15 min. Mutant clones

were screened using the mismatch cleavage assay, as described pre-viously (41). Candidate clones exhibiting mismatch cleavage productswere verified by quantitative PCR and sequencing at The Centre forApplied Genomics (TCAG).

Statistics. Unless otherwise indicated in the figure legends, all data wereanalyzed by the Mann–Whitney test using GraphPad Prism 6. In the figures,*P < 0.05, **P < 0.01, ***P < 0.001, and ns indicates not significant. All errorbars represent SDs.

ACKNOWLEDGMENTS. We thank Marc Shulman and members of the A.M.laboratory for fruitful discussions and critical analysis, Kefei Yu for the Aid−/−

CH12 cells, Tania Watts for the M12 and A20 cells, and Chris Paige for theWEHI-279 cells. This research is supported by the Canadian Institutes ofHealth Research (Grant PJT156330, to A.M.).

1. Muramatsu M, et al. (2000) Class switch recombination and hypermutation requireactivation-induced cytidine deaminase (AID), a potential RNA editing enzyme. Cell102:553–563.

2. Petersen-Mahrt SK, Harris RS, Neuberger MS (2002) AID mutates E. coli, suggesting aDNA deamination mechanism for antibody diversification. Nature 418:99–103.

3. Bransteitter R, Pham P, Scharff MD, Goodman MF (2003) Activation-induced cytidinedeaminase deaminates deoxycytidine on single-stranded DNA but requires the actionof RNase. Proc Natl Acad Sci USA 100:4102–4107.

4. Rada C, Di Noia JM, Neuberger MS (2004) Mismatch recognition and uracil excisionprovide complementary paths to both Ig switching and the A/T-focused phase ofsomatic mutation. Mol Cell 16:163–171.

5. Rush JS, Fugmann SD, Schatz DG (2004) Staggered AID-dependent DNA double-strandbreaks are the predominant DNA lesions targeted to S mu in Ig class switch re-combination. Int Immunol 16:549–557.

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

7. Bothmer A, et al. (2010) 53BP1 regulates DNA resection and the choice betweenclassical and alternative end-joining during class switch recombination. J Exp Med207:855–865.

8. Ehrenstein MR, Rada C, Jones AM, Milstein C, Neuberger MS (2001) Switch junctionsequences in PMS2-deficient mice reveal a microhomology-mediated mechanism of Igclass switch recombination. Proc Natl Acad Sci USA 98:14553–14558.

9. Ceccaldi R, et al. (2015) Homologous recombination-deficient tumours are dependenton Polθ-mediated repair. Nature 518:258–262.

10. Pan Q, et al. (2002) Alternative end-joining during switch recombination in patientswith ataxia-telangiectasia. Eur J Immunol 32:1300–1308.

11. Ramiro AR, et al. (2004) AID is required for c-myc/IgH chromosome translocationsin vivo. Cell 118:431–438.

12. 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.

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

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

15. Robbiani DF, et al. (2008) AID is required for the chromosomal breaks in c-myc thatlead to c-myc/IgH translocations. Cell 135:1028–1038.

16. Ramachandran S, et al. (2016) The SAGA deubiquitination module promotes DNArepair and class switch recombination through ATM and DNAPK-mediated γH2AXformation. Cell Reports 15:1554–1565.

17. Luby TM, Schrader CE, Stavnezer J, Selsing E (2001) The μ switch region tandem re-peats are important, but not required, for antibody class switch recombination. J ExpMed 193:159–168.

18. Cortizas EM, et al. (2013) Alternative end-joining and classical nonhomologous end-joining pathways repair different types of double-strand breaks during class-switchrecombination. J Immunol 191:5751–5763.

19. Longerich S, Storb U (2005) The contested role of uracil DNA glycosylase in immu-noglobulin gene diversification. Trends Genet 21:253–256.

20. Gostissa M, et al. (2014) IgH class switching exploits a general property of two DNAbreaks to be joined in cis over long chromosomal distances. Proc Natl Acad Sci USA111:2644–2649.

21. Kingzette M, Spieker-Polet H, Yam PC, Zhai SK, Knight KL (1998) Trans-chromosomalrecombination within the Ig heavy chain switch region in B lymphocytes. Proc NatlAcad Sci USA 95:11840–11845.

22. Reynaud S, et al. (2005) Interallelic class switch recombination contributes signifi-cantly to class switching in mouse B cells. J Immunol 174:6176–6183.

23. Dong J, et al. (2015) Orientation-specific joining of AID-initiated DNA breaks pro-motes antibody class switching. Nature 525:134–139.

24. Shen B, et al. (2014) Efficient genome modification by CRISPR-Cas9 nickase withminimal off-target effects. Nat Methods 11:399–402.

25. Ran FA, et al. (2013) Double nicking by RNA-guided CRISPR Cas9 for enhanced ge-nome editing specificity. Cell 154:1380–1389.

26. Pham P, Bransteitter R, Petruska J, Goodman MF (2003) Processive AID-catalysed cy-tosine deamination on single-stranded DNA simulates somatic hypermutation. Nature424:103–107.

27. Bothmer A, et al. (2017) Characterization of the interplay between DNA repair andCRISPR/Cas9-induced DNA lesions at an endogenous locus. Nat Commun 8:13905.

28. Pan-Hammarström Q, et al. (2005) Impact of DNA ligase IV on nonhomologous end-joining pathways during class switch recombination in human cells. J Exp Med 201:189–194.

29. Han L, Yu K (2008) Altered kinetics of nonhomologous end-joining and class switchrecombination in ligase IV-deficient B cells. J Exp Med 205:2745–2753.

30. Simsek D, Jasin M (2010) Alternative end-joining is suppressed by the canonical NHEJcomponent Xrcc4-ligase IV during chromosomal translocation formation. Nat StructMol Biol 17:410–416.

31. Busch K, et al. (2007) Identification of two distinct MYC breakpoint clusters and theirassociation with various IGH breakpoint regions in the t(8;14) translocations in spo-radic Burkitt lymphoma. Leukemia 21:1739–1751.

32. Simsek D, et al. (2011) DNA ligase III promotes alternative nonhomologous end-joining during chromosomal translocation formation. PLoS Genet 7:e1002080.

33. Lu G, et al. (2016) Ligase I and ligase III mediate the DNA double-strand break ligationin alternative end-joining. Proc Natl Acad Sci USA 113:1256–1260.

34. Gasiunas G, Barrangou R, Horvath P, Siksnys V (2012) Cas9-crRNA ribonucleoproteincomplex mediates specific DNA cleavage for adaptive immunity in bacteria. Proc NatlAcad Sci USA 109:E2579–E2586.

35. Vriend LEM, et al. (2016) Distinct genetic control of homologous recombination repairof Cas9-induced double-strand breaks, nicks and paired nicks. Nucleic Acids Res 44:5204–5217.

36. Escribano-Díaz C, et al. (2013) A cell cycle-dependent regulatory circuit composed of53BP1-RIF1 and BRCA1-CtIP controls DNA repair pathway choice.Mol Cell 49:872–883.

37. Chapman JR, Taylor MRG, Boulton SJ (2012) Playing the end game: DNA double-strand break repair pathway choice. Mol Cell 47:497–510.

38. Reid DA, et al. (2016) Bridging of double-stranded breaks by the nonhomologousend-joining ligation complex is modulated by DNA end chemistry. Nucleic Acids Res45:1872–1878.

39. Chang HHY, et al. (2016) Different DNA end configurations dictate which NHEJcomponents are most important for joining efficiency. J Biol Chem 291:24377–24389.

40. Liang Z, Sunder S, Nallasivam S, Wilson TE (2016) Overhang polarity of chromosomaldouble-strand breaks impacts kinetics and fidelity of yeast non-homologous endjoining. Nucleic Acids Res 44:2769–2781.

41. van Overbeek M, et al. (2016) DNA repair profiling reveals nonrandom outcomes atCas9-mediated breaks. Mol Cell 63:633–646.

42. Daley JM, Wilson TE (2005) Rejoining of DNA double-strand breaks as a function ofoverhang length. Mol Cell Biol 25:896–906.

43. Bunting SF, et al. (2010) 53BP1 inhibits homologous recombination in Brca1-deficientcells by blocking resection of DNA breaks. Cell 141:243–254.

44. Zan H, et al. (2017) Rad52 competes with Ku70/Ku86 for binding to S-region DSB endsto modulate antibody class-switch DNA recombination. Nat Commun 8:14244.

45. Guikema JEJ, et al. (2007) APE1- and APE2-dependent DNA breaks in immunoglobulinclass switch recombination. J Exp Med 204:3017–3026.

46. Larijani M, et al. (2007) AID associates with single-stranded DNA with high affinityand a long complex half-life in a sequence-independent manner. Mol Cell Biol 27:20–30.

47. Richardson CD, Ray GJ, DeWitt MA, Curie GL, Corn JE (2016) Enhancing homology-directed genome editing by catalytically active and inactive CRISPR-Cas9 usingasymmetric donor DNA. Nat Biotechnol 34:339–344.

48. Cong L, et al. (2013) Multiplex genome engineering using CRISPR/Cas systems. Science339:819–823.

49. Hsu PD, et al. (2013) DNA targeting specificity of RNA-guided Cas9 nucleases. NatBiotechnol 31:827–832.

50. Doench JG, et al. (2014) Rational design of highly active sgRNAs for CRISPR-Cas9-mediated gene inactivation. Nat Biotechnol 32:1262–1267.

Ling et al. PNAS | March 13, 2018 | vol. 115 | no. 11 | 2805

IMMUNOLO

GYAND

INFLAMMATION

Dow

nloa

ded

by g

uest

on

Dec

embe

r 11

, 202

1