Altered kinetics of nonhomologous end joining and class switch recombination in ligase IV–deficient B cells

Lig4 is essential for mouse embryo development but dispensable for somatic cell growth. To determine the role of Lig4 in CSR, we disrupted the Lig4 gene on both alleles in CH12F3 cells. Mouse Lig4 has a total of 911 amino acids. The entire coding sequence is located within the second exon. We replaced 78% of the Lig4 coding region with a floxed puromycin marker (Fig. 1 A). The deleted region encodes amino acids 220–911, which contain the catalytic domain and the two BRCT motifs. This created a larger deletion than a previously defined null allele (10).

Six independent clones harboring a correctly targeted allele (designated L4 +/P) were obtained from a total of 192 puromycin-resistant clones. One of these was randomly selected for gene targeting of the second allele. First, the puromycin marker was excised by the Cre-LoxP reaction (designated L4 +/Δ). Then, the same targeting vector was used again to disrupt the second allele. One correctly targeted clone (designated L4 P/Δ) was obtained from a total of 34 puromycin-resistant clones (Fig. 1).

Because NHEJ deficiency is known to result in cellular hypersensitivity to DSB-inducing agents such as ionizing radiation and bleomycin, we tested the Lig4-null CH12F3 cells for this property. As expected, Lig4-null cells exhibited a marked sensitivity to zeocin (a bleomycin analogue; Fig. 1 C), based on a colorimetric assay that measures the metabolic conversion of thiazolyl blue tetrazolium bromide (MTT) in the mitochondria of living cells (20).

To determine the CSR capacity of Lig4-null CH12F3 cells, cells were seeded at a low density (5 × 10⁴ cells/ml) and grown in the presence of cytokines for 72 h, as previously described (see Materials and methods) (16). CSR in Lig4-null cells is reduced to ∼40–50% of that of the Lig4-heterozygous cells (Fig. 2, A and B), which are indistinguishable from wild-type cells (not depicted). This is consistent with the two recent studies (13, 14) showing a similar reduction of CSR in XRCC4-deficient mouse primary B cells and in Lig4^−/− p53^−/− B cells, indicating the existence of alternative end joining that can resolve switch region breaks.

To confirm that the reduction was indeed caused by a Lig4 deficiency rather than a clonal effect, a genetic complementation assay was performed. First, the floxed puromycin marker in L4 P/Δ cells was excised by the Cre-LoxP reaction to generate puromycin-sensitive Lig4-null cells (designated L4 Δ/Δ; Fig. 2 C). Then, L4 Δ/Δ cells were infected with recombinant retroviruses expressing the mouse Lig4 or control virus made from the empty vector. Infected cells were selected by puromycin and subjected to the CSR analysis. As shown in Fig. 2 D, Lig4 virus–transduced cells had restored normal levels of CSR, but the cells transduced with the control virus did not.

Because CSR is regulated by cell division, we analyzed the effect of Lig4 deletion on cell proliferation. Lig4-null CH12F3 cells propagated normally without cytokine stimulation (Fig. 3 A, left). However, slower propagation of Lig4-null cells was detected at 48 and 72 h in the presence of cytokines (Fig. 3 A, right). To distinguish whether this was caused by a reduced proliferation rate or increased cell death, we stained the cells with CFSE and monitored cell divisions through the passive dilution of CFSE. Although the CFSE fluorescence profiles of Lig4-heterozygous and Lig4-null cells completely overlap in the absence of cytokines (Fig. 3 B, left column), a shift toward brighter CFSE was observed for Lig4-null cells in the presence of cytokines at 48 and 72 h (Fig. 3 B, right column), indicating reduced proliferation. We next analyzed cell death using FITC-conjugated annexin V and propidium iodide (PI), which stain early apoptotic and dead cells, respectively. No significant difference in apoptosis or cell death was observed for the two genotypes in the absence of cytokines (Fig. 3 C). However, an increase of apoptosis or cell death was observed in Lig4-null culture in the presence of cytokines at 48 and 72 h (Fig. 3 C). This suggests that diminished repair capacity in Lig4-null cells leads to elevated apoptosis triggered by CSR-associated DSB.

The growth difference of Lig4-heterozygous and Lig4-null cells in the presence of cytokines prompted us to reevaluate the mild effect of Lig4 deletion initially observed on CSR. Two possible mechanisms could account for the reduction of CSR in the absence of Lig4. One possibility is that Lig4 affects cell proliferation in a way unrelated to NHEJ, which in turn affects CSR. The other explanation is that Lig4 is essential for efficient repair of switch region breaks. Without Lig4, accumulation of switch region breaks cause cell-cycle arrest, which is manifested as a cell proliferation defect. If the former were true, CSR level between the two genotypes (L4 +/Δ vs. P/Δ) should be similar at 24 h, as cell proliferation is apparently the same (Fig. 3 A, right). However, if the latter were true, CSR difference between the two genotypes (L4 +/Δ vs. P/Δ) should be more significant at the beginning before the elimination of CSR-failed cells (and the enrichment of switched cells).

To distinguish these two possibilities, a time course of CSR was performed. A representative experiment is shown in Fig. 4 A (left). Although the numerical value of CSR varies slightly in different experiments, the relative ratio of CSR between the two genotypes (L4 +/Δ vs. P/Δ) is consistent. At 24, 48, and 72 h, CSR of the Lig4-null cell was at 13, 29, and 48%, respectively, of that of the Lig4-heterozygous cell (Fig. 4 A, right). Continued subculturing (>7 d) of Lig4-null cells in the presence of cytokines eventually led to ∼80–90% of the wild-type CSR level (unpublished data). CH12F3 cells grow rapidly in culture with an estimated doubling time of ∼8–10 h. Seeding at a very low density is necessary to achieve maximum levels of CSR without subculturing. However, this condition is less optimal for accurately measuring the initial rate of CSR. Therefore, we also seeded the cells at 3 × 10⁵ cells/ml (early log phase) and measured CSR after 24 h. Again, a marked reduction of CSR in Lig4-null cells was observed (15% of wild-type level; Fig. 4 B, right) despite a comparable rate of cell proliferation (Fig. 4 B, left). Thus, the optimal rate of CSR requires Lig4 regardless of stimulation conditions.

These data support the hypothesis that NHEJ is the primary repair mechanism for switch region breaks. The small endpoint reduction observed initially is likely an underestimation of a marked CSR defect, which is probably masked by an enrichment of IgA cells that have managed to repair their switch region breaks through alternative end joining. Nevertheless, in the absence of NHEJ, alternative end joining can effectively repair switch region breaks, albeit at a substantially slower rate.

Numerous studies have shown increased use of microhomology at the junctions in NHEJ-deficient cells (14, 21, 22). To determine whether switch junctions in Lig4-null cells are microhomology mediated, we cloned and sequenced Sμ–Sα junctions from stimulated Lig4-heterozygous and Lig4-null cells. Because Sα is highly homologous to Sμ, we chose to PCR amplify switch junctions from individual cell clones. Cloning individual switch junctions instead of amplifying from a pool of cells helps to eliminate false junctions produced by template switching of stalled PCR intermediates, as well as the bias against large amplicons. In addition, the distinctive size of each PCR product was used to reference the authenticity of the sequenced junction. We found that 27% of the junctions from Lig4-heterozygous cells were direct joins, but there were none for the Lig4-null cells (Fig. 5). The difference is statistically significant (7 out of 26 vs. 0 out of 23; two-tail p-value of 0.01 by Fisher's exact test) and consistent with what was observed for the Sμ–Sγ switch junctions in XRCC4-deficient B cells (14). The distribution of recombination break points is indistinguishable between the two genotypes (unpublished data).

This study highlights the importance of NHEJ as the primary mechanism of DSB repair in CSR but also reveals the flexibility of using alternative end joining, which is in sharp contrast to NHEJ-dependent V(D)J recombination. The mild endpoint reduction of CSR in Lig4-null CH12F3 cells is in agreement with two recent studies using a conditional knockout of XRCC4 in mature B cells (13, 14). However, our study differs from those two regarding cell proliferation under cytokine stimulation. We found a reduced proliferation of Lig4-null cells at the late stage of stimulation, which was not obvious in the two XRCC4 studies. Our observation would argue for a dominant role of NHEJ in CSR, whereas others would argue for the robustness of the alternative end joining. It should be noted that in one of the studies, deletion of XRCC4 is incomplete (13). In the other study, cell proliferation was measured on a p53-deficient background, which lacks the proper cell-cycle checkpoint control (14).

Although alternative NHEJ can efficiently join linearized plasmids, it clearly cannot replace NHEJ for V(D)J recombination or cellular resistance to ionizing radiation. One speculation has been that alternative end joining is not optimized for repair chromosomal breaks, which is not applicable in this case. Another view is that DSBs are repaired in a manner reflecting how they are generated. It is widely believed that the strict dependence of V(D)J recombination on NHEJ stems from the active guidance of the RAG proteins in the postcleavage complex (14, 22). Indeed, fusion of the RAG proteins to the I-Sce I endonuclease can direct I-Sce I breaks solely to NHEJ (23). On the other hand, certain RAG mutations can significantly alleviate the dependence of V(D)J recombination on NHEJ (22). Therefore, the flexibility of using alternative end joining in CSR might originate from the complete accessibility of AID-initiated breaks. Interestingly, a single I-Sce I break engineered at an acceptor switch region can efficiently support CSR (24), strongly suggesting that CSR does not require the type of synaptic complex presumed to exist in V(D)J recombination.

Alternative end joining was found to excessively use microhomology. Because all switch regions are abundant in conserved pentamer sequences (e.g., GAGCT and GGGG/CT; underlining indicates one nucleotide position), these short sequences could potentially serve as homology blocks and promote the utilization of alternative end joining. Indeed, there was a marked shift toward microhomology-mediated joining in Lig4-null cells. The most remarkable observation was that there was no direct join in the absence of Lig4. This is in complete agreement with what has been observed for Sμ–Sγ junctions in XRCC4-deficient mouse B cells (13, 14) and in human patients harboring hypomorphic Lig4 mutations (21).

Mammalian cells have three ATP-dependent DNA ligases (Lig1, 3, and 4). As a dedicated DSB repair enzyme, The Lig4 complex (XLF–XRCC4–Lig4) has a remarkable flexibility of joining blunt, compatible, and incompatible ends, and even ligating across a gap (25). In contrast, Lig1 and Lig3 are considered as nick ligases. However, recent studies have shown that Lig3 is also capable of intermolecular ligation of blunt and short compatible ends (25, 26). Although the exact architecture of alternative end joining remains unclear, biochemical evidence has suggested the XRCC1–Lig3 complex as a primary candidate for the ligation step (27, 28). Disruption of Lig3 or XRCC1 in mice causes embryonic lethality, thereby limiting their genetic characterizations in alternative end joining. However, Lig3 deficiency might not be cellularly lethal given that XRCC1-deficient cells are viable. In this regard, somatic gene targeting of Lig3 might be a suitable way of dissecting the role of alternative end joining in CSR, as well as DSB repair in general in mammalian cells.

CH12F3 cells were obtained from T. Honjo (Kyoto University, Kyoto, Japan). Cells were cultured as previously described (16). Cell propagation was analyzed by counting live cells every 24 h. Unless indicated otherwise in the figures, stimulation of CH12F3 cells was performed by seeding log phase cells at 5 × 10⁴ cells/ml in medium supplemented with 5% NCTC-109 medium, 1 μg/ml anti-CD40 antibody, 5 ng/ml IL-4, and 0.5 ng/ml TGF-β1, and cells were grown for 72 h. Cells were stained with FITC-conjugated anti–mouse IgA antibody and analyzed by flow cytometry. Class switch efficiency was determined as the percentage of IgA-positive cells.

The targeting vector backbone (pTV2DN2) was obtained from J. Mann (Murdoch Children's Research Institute, Parkville, Australia). A 6.1-kb Bgl II fragment (139,231–145,321) and a 1.8-kb Cla I–Bam HI fragment (148,114–149,923; both available from GenBank/EMBL/DDBJ under accession no. AC138397) from bacterial artificial chromosome clone RP23-65O11 were used as the two homology blocks for gene targeting. The relevant part of the targeting vector is shown in Fig. 1 A. 10 million CH12F3 cells were transfected with 10 μg of linear targeting vector and seeded into five 96-well plates. Puromycin was added 48 h later to a final concentration of 1 μg/ml. Puromycin-resistant clones were picked after 7–10 d and screened by PCR and Southern blot analysis.

Cells were seeded at 2 × 10⁵ cells/ml/well in a 24-well plate and grown for 48 h in various concentrations of zeocin. Metabolic conversion of MTT was performed by incubation of harvested cells in 1 mg/ml MTT (in 1× phosphate-based saline) for 2 h. The cell pellet was lysed in 100 μl of acidic isopropanol (1 ml of concentrated hydrochloric acid in 349 ml isopropanol), and the absorbance was measured at 570 nm.

The Lig4 coding region sequence was cloned into plasmid pMSCVpuro (Clontech Laboratories, Inc.). Recombinant retrovirus was made by transfection of pMSCV plasmids into retrovirus packaging cell line Phoenix (provided by the Nolan Laboratory, Stanford University, Palo Alto, CA). The viral supernatants were harvested and used to infect Lig4-null cells (L4 Δ/Δ). Infected cells were selected by puromycin and subjected to CSR assays.

To monitor cell proliferation, cells were stained with 10 nM CFSE (Invitrogen) for 15 min at 37°C and washed three times with ice-cold growth medium. CFSE intensity was measured every 24 h. For detection of early apoptotic and dead cells, cells were stained with FITC-conjugated annexin V and PI and analyzed every 24 h until the culture reached confluence.

Individual IgA-positive clones were isolated by limited dilutions of cytokine-stimulated cultures in 96-well plates. Switch junctions were amplified with primers KY761 (5′-AACTCTCCAGCCACAGTAATGACC-3′) and KY743 (5′-GAGCTCGTGGGAGTGTCAGTG-3′). The PCR product was sequenced at the Genomic Core Facility of Michigan State University.

We thank Dr. T. Honjo for providing CH12F3 cells, Dr. J. Mann for plasmids and advice on gene targeting, and Dr. C.-L. Hsieh for cytogenetic analysis of CH12F3 cells. We are indebted to Dr. K. Meek for discussion and critical review of the manuscript. Christina Brownlee provided technical assistance.

The authors have no conflicting financial interests.