Alternative end-joining catalyzes class switch recombination in the absence of both Ku70 and DNA ligase 4

Several independent studies from more than a decade ago reported that Ku70- or Ku80-deficient B cells had very severe CSR defects (Manis et al., 1998; Casellas et al., 1998), leading to the assumption that, as in V(D)J recombination, CSR end-joining is catalyzed strictly by C-NHEJ (Honjo et al., 2004). However, normal CSR requires B cell proliferation, and previously studied Ku70- or Ku80-deficient B cells had severe proliferation defects, with many dying when activated for CSR (Casellas et al., 1998; Manis et al., 1998; Reina San-Martin et al., 2003). Therefore, we have suggested that although Ku70 and Ku80 are required for normal CSR, their overall contribution via C-NHEJ remained to be determined (Manis et al., 2002; Chaudhuri and Alt, 2004; Chaudhuri et al., 2007). Our current protocols for activating B cells stimulate much higher CSR levels in WT B cells than those used in the original Ku70^−/− and Ku80^−/− B cell studies; e.g., in the earlier studies of Ku-deficient cells, only ∼10% of WT B cells underwent CSR to IgG1 (Casellas et al., 1998; Manis et al., 1998), whereas under current stimulation conditions (different purification of cells, sources of cytokines, and activators, etc.) we observe CSR to IgG1 in 60% or more of WT B cells (Cheng et al., 2009; see Fig. 2). Therefore, we have reexamined the requirement for Ku70 and Ku80 in CSR. As Ku70 and Ku80 are required for V(D)J recombination and generation of mature B cells, we generated Ku70^−/− or Ku80^−/− mice that harbored preassembled (“knock-in”) IgH (B1-8-HC) and IgL (3–83k-LC) variable region exons (Pelanda et al., 1997; Sonoda et al., 1997). Various studies have shown that the B1-8-HC V(D)J knock-in allele undergoes normal CSR (Casellas et al., 1998; Manis et al., 1998; Yan et al., 2007). Because of the complex breeding strategies necessary to generate these mice, we generated Ku70^−/− and Ku80^−/− strains that were either homozygous or heterozygous for the B1-8-HC knock-in allele (referred to, respectively, as H/HL and H/+L). The breakdown of results with these genotypes is indicated in each figure or in supplementary figures and tables. Because H/HL and H/+L lines gave similar results for most experiments, we generically refer to both genotypes as HL mice, except where otherwise noted. We verified homozygous disruption of the Ku70 and Ku80 locus and the absence of the Ku70 protein in the Ku70^−/−HL mice and cells analyzed (Fig. 1 a and not depicted).

Ku70^−/−HL and Ku80^−/−HL mice generate substantial numbers of mature splenic B cells, which express surface Ig that derives from their IgH and Igκ knock-in alleles (Fig. 1 b). To test CSR potential to IgG1 and IgE, we purified Ku70^−/−HL and Ku80^−/−HL splenic B cells and stimulated them for 4 d with αCD40 plus IL-4 or, for some experiments, with bacterial LPS plus IL-4. In contrast to earlier findings, neither the Ku70^−/− nor Ku80^−/− B cells showed a dramatic proliferation defect compared with control cells (Fig. 1 c and not depicted). Under these stimulation conditions, up to 50–60% of WT/HL B cells switched to IgG1 as measured by flow cytometry staining for surface IgG1 expression (Fig. 2 a, Fig. S1, and Table S1). Strikingly, however, the surface staining assay also revealed that αCD40 plus IL-4 or LPS plus IL-4–stimulated Ku70^−/−HL B cells and Ku80^−/−HL B cells underwent CSR to IgG1 at levels 20–50% of those of WT/HL B cells (Fig. 2 a, Fig. S1, and Table S1). We confirmed the unexpected CSR ability of Ku70^−/− and Ku80^−/− B cells by using numerous additional types of class switching assays, including measurement of secreted IgG1 by ELISA (Fig. 2 b and Table S2), as well as by the more quantitative clonal B cell ELISPOT (not depicted) or hybridoma IgG1 secretion analyses (Fig. 2 c and Table S3). Southern blotting also allowed us to confirm that CSR in Ku70^−/−HL B cell hybridomas occurred at the DNA level via Sμ to Sγ1 recombination (Fig. S2). IgH class switching to IgE cannot be assayed quantitatively via surface staining because of binding of secreted IgE to Fcε receptors on B cells. However, ELISA demonstrated that Ku70^−/−HL B cells switched to IgE at ∼25% of WT/HL levels (Fig. 2 b). Moreover, hybridoma analyses revealed that ∼50% of WT/HL B cells switched to IgE under our current stimulation conditions and that Ku70^−/−HL B cell hybridomas switched to IgE at ∼30% of WT/HL levels, which is consistent with the ELISA data (Table S3). Note that because the WT hybridomas were derived from the H/HL⁺ genotype, ∼15% were IgE/IgG1 double producers caused by productive CSR to IgG1 on one allele and to IgE on the other (see Table S3 legend). Finally, the hybridoma analyses also allowed us to confirm IgE CSR at the DNA level via Southern blotting (Fig. S2). Thus, these studies implicate a relatively robust Ku-independent form of A-EJ that supports CSR to IgG1 and IgE.

DNA-PKcs–deficient B cells appear more impaired for CSR to other IgH isotypes than for CSR to IgG1 (Manis et al., 2002; Rooney et al., 2005; Callén et al., 2009). Therefore, because Ku functions in part via its role in the DNA–PK complex, we assessed the ability of Ku70^−/−HL or Ku80^−/−HL B cells to undergo CSR to additional IgH isotypes. Stimulation of splenic B cells with LPS alone induces CSR to IgG2b and IgG3. However, we have recently shown that DNA-PKcs– or Artemis-deficient B cells activated for CSR are much more prone to cell death upon LPS stimulation compared with stimulation with αCD40 plus IL-4 because of a p53-dependent response to unrepaired DNA DSBs (Franco et al., 2008). LPS plus anti–IgD-dextran stimulates class switching to IgG2b and IgG3 more robustly than treatment with LPS alone (Cheng et al., 2009). Therefore, to test CSR to these two IgH isotypes, we stimulated purified Ku70^−/−HL splenic B cells for 4 d with LPS plus anti–IgD-dextran. Based on the flow cytometry assay for surface IgG3 expression, this stimulation led to IgG3 switching in up to 40% of the WT/HL B cells (Fig. 3 a; Table S1). Strikingly, Ku70^−/−HL B cells underwent class switching to IgG3 on average at nearly 30% of WT levels under these conditions (Fig. 3 a and Table S1). LPS plus anti–IgD-dextran treatment of Ku70^−/−HL B cells also induced switching to IgG3 at a substantial fraction of WT/HL levels, based on ELISA and ELISPOT analyses (Fig. 3, b and c, and Table S2) and to IgG2b based on ELISA analyses (Fig. 3 b and Table S2). A more limited set of studies also demonstrated that LPS-stimulated Ku80^−/−HL B cells switch to IgG3 or IgG2b, as assayed by flow cytometry (Table S1), ELISPOT (not depicted), or ELISA (Table S2). Finally, we used stimulation with LPS–IL-4–IL-5 and TGFβ for class switching to IgA and found that Ku70^−/−HL B cells also switched to IgA at levels that were a substantial fraction of those observed for WT/HL B cells (Fig. S3).

Although Lig4 deficiency is embryonically lethal (Frank et al., 1998), Ku80^−/−Lig4^−/− mice (Karanjawala et al., 2002) and Ku70^−/−/Lig4^−/− DT40 avian cell lines (Adachi et al., 2001) are viable. In addition, we find that Ku70^−/−Lig4^−/− mice are also viable (see below). Thus, we addressed the question of whether CSR can occur in a fully C-NHEJ–independent fashion by asking whether we observe CSR in cells deficient for both Ku70 and Lig4 (the specific DSB recognition and ligation C-NHEJ components). For this purpose, we generated Ku70^−/−Lig4^−/− mice, which also harbored the preassembled IgH and IgK alleles (HL). Ku70^−/−Lig4^−/−HL mice develop normally, albeit being slightly smaller than their littermates. We confirmed the deletion of Ku70 and Lig4 genes and absence of both proteins in extracts from kidney, brain and thymus by Southern blotting and Western blotting, respectively (Fig. 1 a). Ku70^−/−Lig4^−/−HL mice also generate IgM⁺ splenic B cells (Fig. 1 b), but lack mature thymic or peripheral T cells because of their V(D)J recombination block (not depicted). To test for ability of purified Ku70^−/−Lig4^−/−HL splenic B cells to undergo class switching, we stimulated them for 4 d with αCD40 plus IL-4 and measured class switching to IgG1 by surface staining and ELISA (Fig. 2, a and b, Table S1, and Table S2). We also stimulated Ku70^−/−Lig4^−/−HL splenic B cells for 4 d with LPS plus anti–IgD-dextran and measured class switching to IgG3 and IgG2b by surface staining, ELISA, and ELISPOT (Fig. 3, a–c, Table S1, and Table S2). Based on all of these assays, Ku70^−/−Lig4^−/−HL B cells showed class switching to all tested IgH isotypes at levels comparable to Ku70^−/− B cells. Therefore, we conclude that there is a repair pathway that catalyzes substantial IgH class switching in the complete absence of C-NHEJ.

Most studies of CSR tend to measure this process at time points after stimulation (e.g., day 4) when WT CSR is nearly complete. However, measurements at such late time points might underestimate kinetic differences between mutant and WT B cells, by allowing the mutant B cells to “catch up” (Cheng et al., 2009). Therefore, to further characterize the Ku-independent A-EJ pathway during CSR, we performed time-course kinetics of class switching to IgG1 in αCD40 plus IL-4 stimulated HL, Ku70^−/−HL and Ku70^−/−Lig4^−/−HL purified splenic B cells. For this purpose, we measured surface IgG1expression at days 2.5, 3.5, and 4.5, with the earliest time point corresponding to about the time that AID expression is readily detectable in stimulated B cells (Cheng et al., 2009). These studies showed that Ku70^−/−HL and Ku70^−/−Lig4^−/−HL B cells underwent class switching to IgG1 at a significant fraction of the level observed in WT/HL B cells at all assayed time points (Fig. 4, a and b, Fig. S1, and Table S1). Thus, the Ku70- and Ku70/Lig4-independent pathway of IgH class switching to IgG1 also is relatively robust kinetically as compared with that of WT/HL B cells. However, the relative level of class switching to IgG1 for Ku70^−/−B cells and Ku70^−/−Lig4^−/−HL B cells, as compared with that of WT/HL B cells, appeared slightly lower on average at day 2.5 versus later time points (Fig. 4 b and Table S1), suggesting that CSR in the absence of Ku70 might be slightly less robust kinetically than that of CSR in WT cells (Yan et al., 2007). In this regard, while CSR in B cells in which XRCC4 was conditionally deleted in the periphery appeared kinetically similar to that of WT B cells (Yan et al., 2007), CSR in a Lig4-deficient B cell line appeared kinetically reduced (Han and Yu, 2008).

A substantial fraction of CSR junctions from WT B cells are direct, with most of the remainder displaying short MHs; whereas essentially all CSR junctions from XRCC4-deficient B cells display MHs that are on average longer than those from WT B cells (Yan et al., 2007). To further assess the relationship between the A-EJ pathway that generates CSR joins in Ku-deficient cells versus that which occurs in the absence of XRCC4 or Lig4, we sequenced Sμ/Sγ1 and Sμ/Sε junctions from Ku70^−/−HL, Ku70^−/−Lig4^−/−HL, Lig4^−/−HL, and WT/HL splenic B cells that were activated with αCD40 plus IL-4. As expected, a substantial percentage of Sμ/Sγ1 and Sμ/Sε junctions from WT/HL B cells were direct (27.1 and 33.3% direct, respectively; Fig. 5, b and c, and Table S4). On the other hand, Sμ/Sγ1 and Sμ/Sε junctions from Lig4^−/−HL B cells were almost all MH-mediated (2 and 0% direct, respectively), consistent with the fact that XRCC4 and Lig4 act as a complex in C-NHEJ. Although Sμ/Sγ1 and Sμ/Sε junctions from Ku70^−/−HL B cells also displayed increased usage and mean length of MH-mediated joins compared with those from WT/HL B cells (Fig. 5, b and c; Table S4), a substantial fraction were direct (8.9 and 12.2%, respectively; Fig. 5, b and c; Table S4). Similarly, a substantial fraction of Sμ/Sγ1 and Sμ/Sε junctions from Ku70^−/−Lig4^−/−HL B cells also were direct (18 and 13%, respectively; Fig. 5, b and c, and Table S4). Thus, CSR junctions in Ku-deficient B cells, whether or not Lig4 is present, show a substantial and significant increase in direct joins as compared with junctions generated in the absence of Lig4 (P = 0.04 for Ku70^−/−HL; P = 0.003 for Ku70^−/−Lig4^−/−HL; Fig. 5 C and Table S4). Importantly, the fraction of direct joins is not significantly different between Ku70 versus Ku70 plus Lig4-deficient B cells (P = 0.11 for Sμ/Sγ1 and P = 0.96 for Sμ/Sε junctions; Fig. 5 C), which is consistent with the possibility that another ligase may be predominantly used for CSR end-joining in Ku70^−/−HL B cells.

Recent work has established the role of A-EJ in mammalian DNA repair and made this pathway a major current focus of the DNA repair field. Yet, the nature of A-EJ has remained enigmatic. In particular, other than recent studies that implicate the MRN complex in chromosomal A-EJ (Deng et al., 2009; Dinkelmann et al., 2009; Rass et al., 2009; Xie et al., 2009), little has been forthcoming regarding the identity of chromosomal A-EJ components. In addition, although A-EJ is often discussed as a single pathway, it has remained possible that there are multiple A-EJ pathways that use at least some different components. Our current studies provide several new insights into A-EJ. In particular, we now demonstrate that, in the absence of C-NHEJ, two forms of A-EJ, which appear to be potentially distinct based on both different preferences for junctional MHs and differential reliance on Ku, catalyze the physiological process of CSR. One form of A-EJ, that operates in the absence of XRCC4 or Lig4, utilizes Ku and either Lig1 or Lig3 and, at least in the context of CSR, is almost totally reliant on junctional MHs. Given its apparent dependence on Ku, this form of A-EJ may, as has been suggested (Lieber et al., 2008), reflect the use of the upstream components of C-NHEJ with a different ligase (Lieber et al., 2008); although it remains possible that there are other differences from C-NHEJ. Our current findings further demonstrate a second form of A-EJ during CSR that functions in the simultaneous absence of both the DSB recognition component (Ku70) and the specific ligation component (Lig4) of C-NHEJ. This second form of A-EJ, while also biased toward CSR junctional MHs, generates substantial levels of CSR direct joins. We conclude that this second (Ku plus Lig4-independent) form of A-EJ is totally distinct from C-NHEJ and must use a largely different set of components.

Prior studies found that CSR was abrogated by Ku70- or Ku80-deficiency, leading to the common assumption that Ku70 and Ku80 are essential for end joining during CSR (Honjo et al., 2004). However, Ku70- or Ku80-deficient cells, as assayed at that time, had severe proliferation defects, which also were suggested to potentially contribute to their severe CSR impairment (Manis et al., 1998, 2002; Chaudhuri and Alt, 2004; Chaudhuri et al., 2007). Our current stimulation conditions lead to substantially increased CSR in WT cells and also appear to obviate the very severe proliferation defects previously observed in Ku-deficient B cells. Under these stimulation conditions, we confirm that CSR indeed is impaired in Ku-deficient B cells; but, on the other hand, we also find that Ku-deficient B cells are capable of undergoing substantial CSR. Indeed, Ku70- or Ku80-deficient B cells reach steady-state CSR levels that range from 25–50% those of WT B cells, which are similar in range to those found in XRCC4- or Lig4-deficient B cells. Likewise, our time course experiments confirmed that CSR in the absence of Ku or Ku plus Lig4 is quite robust; although perhaps slightly delayed kinetically as compared with that of WT B cells reminiscent of findings of CSR in Lig4-deficient CH12 cells (Han and Yu, 2008).

The spectrum of CSR junctions differs between Ku70-deficient or Ku70- plus Lig4-deficient versus Lig4- or XRCC4-deficient B cells. Thus, XRCC4- or Lig4-deficient junctions are nearly all MH mediated, whereas the junctions that occur in the absence of Ku70 or Ku70 plus Lig4, although still biased toward MH, have a substantial percentage of direct joins. These findings are consistent with findings of increased, although not to the same degree, use of MH in ISceI-mediated DSB junctions recovered from XRCC4 versus Ku-deficient Chinese hamster organ cells (Guirouilh-Barbat et al., 2004, 2007). Ku was previously thought to function in DNA repair only via C-NHEJ; thus, our finding that Ku actually promotes MH-mediated A-EJ during CSR in the absence of Ligase 4 was unexpected. The Ku heterodimer binds to ends of DSBs and then recruits other C-NHEJ repair factors (for review see Rooney et al., 2004), while also protecting the broken DNA ends from degradation. However, we find that nearly all CSR DSBs in Ku-proficient/Lig4-deficient cells are joined via MHs, which presumably requires end resection; whereas CSR DSBs in Ku plus Lig4-deficient B cells have a higher frequency of direct joins. One possible explanation for these seemingly contradictory findings would be that Ku can preferentially stabilize S region ends held by microhomologies. Alternatively, in the absence of Ku70, end-modifying polymerases, such as TdT, pol µ, or pol λ might gain access to DNA ends and perform nucleotide additions to generate “occult” end-homology (Komori et al., 1996) that cannot be distinguished from “direct” joins. In this context, Pol µ has been shown to have TdT-like activity and to add nontemplate nucleotides even in the absence of Ku, XRCC4, or Lig4 (Gu et al., 2007). A further possibility is that Lig1 and Lig3 may be differentially involved in the Ku70-dependent, MH-mediated A-EJ pathway versus the Ku70-independent pathway and, thereby, provide the end preference. In biochemical assays, Lig3 but not Lig1 performs direct end joining of oligonucleotide substrates (Chen et al., 2000; Cotner-Gohara et al., 2008). Thus, if the end preference observed in biochemical studies holds in Ku70-deficient cells, our findings might be explained by Lig3 preferentially gaining access to DNA ends in Ku70-deficient cells and Lig1 preferentially accessing ends in Ku-70 proficient, Lig4-deficient cells.

In conclusion, our studies demonstrate the existence of an A-EJ pathway that functions to generate substantial levels of chromosomal IgH CSR joins in activated B cells that lack both DSB recognition and joining components of the C-NHEJ pathway. This finding firmly demonstrates the existence of a physiologically functional A-EJ pathway that is distinct from C-NHEJ. In addition, we find that the previously described MH-based A-EJ pathway of CSR that catalyzes CSR in the absence of either XRCC4 or Lig4 is substantially influenced by Ku; thus, this pathway might be a variant of C-NHEJ that uses a different ligase; although other possibilities including a distinct, Ku dependent, A-EJ pathway can also be considered.

Ku70^−/−, Lig4^−/− CD21CreXrcc4^c/c, and Ku80^−/− HL mice have been previously described (Gu et al., 1997, Frank et al., 1998, Yan et al., 2007, Casellas et al., 1998). Ku70^−/−Lig4^−/−HL mice were obtained by breeding double-heterozygous Ku70^+/−Lig4^+/−HL mice. Animal protocols were approved by the Institutional Animal Care and Use Committee of Children’s Hospital (Boston, MA).

Mature B cells were isolated and enriched from the spleen and stimulated with αCD40/IL-4 to induce CSR to IgG1 and IgE as described (Yan et al., 2007). For IgG3 and IgG2b CSR, B cells were stimulated with LPS and dextran-conjugated αIgD as described (Cheng et al., 2009). For IgA CSR, B cells were stimulated with LPS (25 µg/ml), dextran-conjugated αIgD (3 ng/ml), IL-4: 10 ng/ml, IL-5: 1.5 ng/ml and TGFβ: 2 ng/ml. Cytokine sources: αCD40: eBioscience, IL-4: eBioscience, IL-5: BD dextran-conjugated αIgD: Fina BioSolutions LLC. Proliferation studies were conducted on aCD40/IL-4 stimulated splenic cells enriched with B220 magnetic beads (Miltenyi Biotec) and cells were counted every 24 h starting at 1.5 d post-enrichment to ensure that the majority of the cells in culture were B cells. Surface expression was recorded by FACS. ELISPOT and ELISA were performed as described (Yan et al., 2007) at days 4 and 6 of culture, respectively. B cell hybridoma fusions were performed at day 4 of culture and hybridomas were screened by ELISA after 1 wk of selection.

Sμ-Sγ1 and Sμ-Sε junctions were amplified from B cells stimulated in culture with αCD40 plus IL-4 or from hybridomas derived from the above cultures with primers described previously (Yan et al., 2007) and National Center for Biotechnology Information Blast was used for sequence analysis. Calculation of statistical significance was done using the two-tailed Student T test.

Fig. S1 shows a CSR timecourse in Ku70^−/−HL and Ku80^−/−HL mice. Fig. S2 shows CSR-related DNA recombination events in IgG1⁺ and IgE⁺ Ku70^−/−HL and Ku70^−/−Lig4^−/−HL hybridomas. Fig. S3 shows IgA CSR in Ku70^−/−HL B cells. Table S1 is a summary of all FACS experiments. Table S2 is a summary of all ELISA experiments. Table S3 is a summary of all hybridoma fusion data. Table S4 is a summary of all S region function data.

We thank Klaus Rajewsky, Bjoern Schwer, Monica Gostissa, and Maria-Vivienne Boboila for discussions and critical reading of the manuscript.

This work was supported by National Institutes of Health (NIH) grants AI077595 and CA092625 to F.W. Alt, AI037526 to M. Nussenzweig, and T32CA09382-26 to J.H. Wang. C. Boboila is supported by a Cancer Research Institute training grant. F.W. Alt and M. Nussenzweig are investigators of the Howard Hughes Medical Institute.

The authors have no conflicting financial interests.