SAMHD1‐mediated dNTP degradation is required for efficient DNA repair during antibody class switch recombination

Abstract

Sterile alpha motif and histidine‐aspartic acid domain‐containing protein 1 (SAMHD1), a dNTP triphosphohydrolase, regulates the levels of cellular dNTPs through their hydrolysis. SAMHD1 protects cells from invading viruses that depend on dNTPs to replicate and is frequently mutated in cancers and Aicardi–Goutières syndrome, a hereditary autoimmune encephalopathy. We discovered that SAMHD1 localizes at the immunoglobulin (Ig) switch region, and serves as a novel DNA repair regulator of Ig class switch recombination (CSR). Depletion of SAMHD1 impaired not only CSR but also IgH/c‐Myc translocation. Consistently, we could inhibit these two processes by elevating the cellular nucleotide pool. A high frequency of nucleotide insertion at the break‐point junctions is a notable feature in SAMHD1 deficiency during activation‐induced cytidine deaminase‐mediated genomic instability. Interestingly, CSR induced by staggered but not blunt, double‐stranded DNA breaks was impaired by SAMHD1 depletion, which was accompanied by enhanced nucleotide insertions at recombination junctions. We propose that SAMHD1‐mediated dNTP balance regulates dNTP‐sensitive DNA end‐processing enzyme and promotes CSR and aberrant genomic rearrangements by suppressing the insertional DNA repair pathway.

Localization of the HIV‐1 restriction factor SAMHD1 to the immunoglobulin switch region on chromosomes regulates AID‐dependent class‐switch recombination and genomic stability in B‐lymphocytes.

Introduction

Activated mature B‐lymphocytes express activation‐induced cytidine deaminase (AID), which diversifies immunoglobulin (Ig) genes by somatic hypermutation (SHM) and class switch recombination (CSR) (Muramatsu et al, 2000). These two B‐cell‐specific events generate the antibody memory that is critical for humoral immunity (Tarlinton & Good‐Jacobson, 2013). SHM introduces random mutations in the Ig variable gene to generate antibodies with high affinity, whereas CSR alters the antibody effector functions by changing the Ig heavy‐chain constant region (CH). Although mechanistically different, both CSR and SHM are initiated by AID‐induced DNA cleavage in the switch (S) or variable (V) regions, respectively (Petersen et al, 2001; Nagaoka et al, 2002). While SHM requires only single‐stranded breaks (SSBs), CSR additionally requires the conversion of SSBs into double‐stranded breaks (DSBs), followed by subsequent recombination of the two‐cleaved S‐regions by classical non‐homologous end‐joining (c‐NHEJ) or by alternative end‐joining (alt‐EJ) (Boboila et al, 2012a). The joining in c‐NHEJ requires little or no homology between the paired DNA ends, whereas alt‐EJ depends on the microhomology (MH) of single‐stranded overhangs of the two‐cleaved DNA ends (Yan et al, 2007; Panchakshari et al, 2018).

Activation‐induced cytidine deaminase‐dependent DNA breaks are, in general, specific to the V‐ or S‐regions of the Ig locus, but the aberrant expression or mistargeting of AID results in oncogenic mutations and chromosomal translocations in both B and non‐B cells (Okazaki et al, 2003; Pasqualucci et al, 2008; Honjo et al, 2012). Recurrent chromosomal translocations involving the Ig locus and proto‐oncogenes such as c‐MYC, BCL‐2, BCL‐6, and FGFR are hallmarks of B‐cell malignancies, which are associated with Burkitt's lymphoma, follicular lymphoma, diffuse large cell lymphoma, and multiple myeloma, respectively (Leder et al, 1983; Kuppers & Dalla‐Favera, 2001; Robbiani & Nussenzweig, 2013).

The expression of SAMHD1, a dNTP triphosphohydrolase, restricts viral infection, particularly in non‐dividing cells, by depleting cellular dNTPs which are essential for viral reverse transcription as well as replication (Hrecka et al, 2011; Laguette et al, 2011; Baldauf et al, 2012; Lahouassa et al, 2012). SAMHD1 is also suspected as a causative agent of Aicardi–Goutières syndrome (AGS), a congenital neurodegenerative autoimmune disorder. SAMHD1 is frequently mutated in AGS and in a variety of human cancers, such as chronic lymphocytic leukemia (CLL) and colorectal cancers (Clifford et al, 2014; Rentoft et al, 2016). In colorectal cancer, heterozygous SAMHD1 mutations that inactivate dNTPase activity lead to an elevated dNTP pool and to increase mutation rate in cancer cells in combination with defective mismatch repair pathway (Rentoft et al, 2016). However, cancer cells with SAMHD1 overexpression are hypersensitive to DNA damage, and exogenously overexpressed SAMHD1 co‐localizes with 53BP1 at DNA damage sites, suggesting that SAMHD1 has a role in DNA damage repair (Clifford et al, 2014). SAMHD1 was also recently reported to have a dNTPase‐independent function in homologous recombination, by recruiting C‐terminal binding protein interacting protein (CtIP) to DNA damage site (Daddacha et al, 2017). In contrast, a recent report attributes the localization of the SAMHD1 to DNA damage sites in human cell lines to the use of a non‐specific antibody (Medeiros et al, 2018). Although SAMHD1 binds to ssDNA/RNA (Goncalves et al, 2012; Beloglazova et al, 2013; Tungler et al, 2013; Seamon et al, 2016) and was proposed to possess a nuclease activity (Beloglazova et al, 2013; Ryoo et al, 2014), subsequent studies suggested the nuclease activity of SAMHD1 was due to the presence of persistent contaminants in the SAMHD1 preparations (Antonucci et al, 2016; Seamon et al, 2016; Welbourn & Strebel, 2016). Many reports indicate ubiquitous expression of SAMHD1, but there is little information regarding the function of SAMHD1 in B cells. A recent study showed that SAMHD1 enhances AID‐induced SHM by promoting transversion but not transition mutations by regulating the MMR/BER pathway (Thientosapol et al, 2018).

Although a number of non‐homologous end‐joining (NHEJ) DNA repair factors have been identified, the underlying mechanism of DNA repair‐pathway choice and its regulation at the context of Ig heavy chain locus (IgH) or non‐IgH rearrangements is far from understood. One of the major limitations is the unknown nature of the repair‐recombination protein complex that forms at the break site induced by AID activation. Here, we attempt to identify proteins that accumulate at AID‐induced DNA‐break sites, and applied Ig locus‐specific insertional chromatin immunoprecipitation (iChIP) (Hoshino & Fujii, 2009; Fujita & Fujii, 2011, 2012; Fujita et al, 2015), which identified several proteins including previously known DNA repair proteins in CSR.

Surprisingly, we identified SAMHD1 in the repair‐recombination complex, which is well known as an HIV1 restriction factor (Goldstone et al, 2011; Powell et al, 2011; Franzolin et al, 2013). Here, we show that AID‐induced CSR and IgH/c‐Myc translocations require SAMHD1 dNTPase activity to promote insertion‐free efficient DNA repair. Our findings revealed a novel role of the cellular dNTP pool in DSB repair and in the maintenance of genomic stability in B cell.

Results

Isolation of S‐region‐binding proteins by iChIP

To identify chromatin components that bind to the Ig S‐region after AID‐induced DNA breaks, we applied the iChIP‐based locus‐specific proteomic approach, which is summarized as follows: (i) An 8X‐repeat of the LexA‐binding element (LexA‐BE) was inserted downstream of the Sα region in CH12F3‐2A cells, a mouse B‐cell line used for studying AID‐induced recombination; (ii) the DNA‐binding domain and dimerization domain of the LexA protein were fused with a 3X‐FLAG tag and a nuclear localization signal (3xFNLDD, hereafter referred as FNLDD) (Fujita & Fujii, 2012; Fujita et al, 2015), and expressed in WT (FNLDD‐alone) and Sα‐engineered (FNLDD‐Sα‐LexA) CH12F3‐2A cells; and (iii) the resultant cells underwent stable isotope labeling with amino acids in cell culture (SILAC), stimulation with CIT (CD40L, IL4, and TGFβ), formaldehyde‐crosslinking, shearing, and immunoprecipitation with an anti‐FLAG antibody (Fig 1A and C; Appendix Fig S1). Collected proteins were then analyzed by mass spectrometry (MS). Prior to the proteomic analysis, Southern blotting analysis was performed to confirm the Sα‐specific insertion of the 8X‐LexA binding elements, and quantitative‐PCR (qPCR) analysis of the immunoprecipitated chromatin was carried out to confirm that it was enriched in the Sα region of the IgH locus (Fig 1B). To reduce non‐specific binding of the FNLDD protein to chromatin, CH12F3‐2A clones with very low FNLDD expression were selected. We also chose clones with similar expression levels of FNLLD for the control (FNLDD‐alone) and Sα‐engineered (FNLDD‐Sα‐LexA) CH12F3‐2A cells (Appendix Fig S1E).

Figure 1. SAMHD1 was identified as a novel regulator of antibody class switch recombination through locus‐specific proteomics.

1. Schematic representation of the IgH locus of WT and of 8X‐LexA binding element (LexA‐BE) knocked‐in CH12F3‐2A cells. The a, b, and c bars indicate the position of primers used for ChIP‐qPCR in B. A protein in which the DNA‐binding domain (DB) of LexA was tagged with 3X‐FLAG, and a nuclear localization signal (FNLDD) was expressed in WT (FNLDD alone) and Sα‐engineered (FNLDD‐ Sα‐LexA) CH12F3‐2A cells.
2. Chromatin immunoprecipitation using α‐FLAG antibody was performed to confirm the pull‐down of Sα chromatin from the Sα‐engineered (FNLDD‐Sα‐LexA) CH12F3‐2A cells. The a, b, and c bars indicate the primers used for ChIP‐qPCR as shown in (A) (n = 3; mean + SD; two‐tailed unpaired Student's t‐test).
3. Schematic representation of the iChIP coupled with SILAC‐based mass spectrometry.
4. Functional classification of the Sα region binding proteins identified by iChIP.
5. Bar graph showing the proteins with ≥ 2‐fold enrichment as calculated by the heavy‐to‐light SILAC ratio.
6. Effect of the siRNA‐mediated KD of the proteins identified by iChIP on IgM to IgA switching in CH12F3‐2A cells. Data are presented as the % IgA switching relative to the IgA switching in cells transfected with control siRNA (siCONT; n = 2; mean + SD).

MS analysis of the proteins obtained by iChIP led to the identification of ~ 39 chromatin‐binding proteins that were enriched ≥ 2‐fold in the FNLDD‐Sα‐LexA cells compared to the control FNLDD‐alone cells (Fig 1D and E; Appendix Table S1). Notably, many proteins with known functions in DNA DSB repair were enriched 1.2‐ to 2.3‐fold, indicating that the iChIP‐based strategy could identify proteins implicated in DNA‐damage repair and recombination at the Sα region of the IgH locus (Appendix Fig S2A). However, AID itself and some proteins with previously known functions in AID‐induced DNA damage and recombination, namely UNG, ATM, KU70/80, MDC1, CtIP, and LIG4, were not detected by the MS analysis. Because many of these proteins presumably bind only transiently to DNA break sites, the amount of these proteins pulled down by iChIP may not have been high enough to be detected by MS analysis. However, these proteins have been found to be associated with the S‐region by standard ChIP assays (Vuong et al, 2009; Stanlie et al, 2014; Yousif et al, 2014).

SAMHD1 is required for CSR and for IgH/c‐Myc chromosomal translocations

To investigate the functional relevance of the proteins identified by iChIP in DNA break repair and recombination, we subjected them to siRNA‐mediated knockdown (KD) in CH12F3‐2A cells and examined the effect on IgM to IgA switching in response to CIT stimulation. We were particularly interested in proteins that were not previously reported to have a role in AID‐dependent recombination. Our screen revealed that the KD of SAMHD1, CPSF6, and DDX21 reduced the IgM to IgA switching to ≤ 50% of the level in cells transfected with control siRNA (siCONT; Fig 1F; Appendix Fig S2B). Since the effects of CPSF6 and DDX21 KD on CSR were likely to be due to their direct effect on the expression of AID, we focused on the role of SAMHD1 in AID‐induced recombination (Appendix Fig S2C). We thus introduced RNAi oligonucleotides recognizing different sequences of the SAMHD1 transcript into CH12F3‐2A cells to KD SAMHD1 and found that the depletion of SAMHD1 significantly reduced the IgA switching without causing significant cell death (Fig 2A–C; Appendix Fig S2D).

Figure 2. SAMHD1 is required for efficient immunoglobulin class switch recombination.

1. Top: Scheme of the IgA‐switching assay in CH12F3‐2A cells. After electroporation of siRNAs, cells were cultured for 24 h, and then stimulated by CIT, cultured for another 24 h, and subjected to FACS analysis. Bottom: Confirmation of the siRNA‐mediated KD of SAMHD1 in CH12F3‐2A cells.
2. FACS profile of the percentage of CH12F3‐2A cells undergoing IgA switching after transfection of the indicated siRNAs into CH12F3‐2A cells cultured with (+) or without (−) CIT stimulation for 24 h. The number in the FACS plot represents the percentage of cells expressing IgA on their surface.
3. Summary of the FACS data obtained from three independent experiments. Data are shown as the % IgA switching relative to the IgA switching in cells transfected with control siRNA (siCONT; n = 3; mean + SD; two‐tailed unpaired Student's t‐test).
4. Quantitative RT–PCR analysis of the μGLT, αGLT, Aid, and Samhd1 expression in the indicated samples. The data were normalized to Gapdh and represent as the mean of the two independent experiments with standard deviations. Statistical significance as determined by two‐tailed unpaired Student's t‐test is shown. ns, not significant.
5. Confirmation of the CRISPR/Cas9‐mediated KO of Samhd1 in CH12F3‐2A cells. Beta‐actin was used as a loading control.
6. CSR time‐course analysis of WT and Samhd1 KO clones. The data are shown as percentage of IgA expressing cells (n = 2; mean ± SD; two‐tailed unpaired Student's t‐test; **P < 0.01).
7. Proliferation of WT and Samhd1 KO clones after stimulation with CIT (n = 2; mean ± SD; two‐tailed unpaired Student's t‐test; ns, not significant).
8. Analysis of the cell proliferation of WT or Samhd1 KO clones of CH12F3‐2A. Histograms show the CFSE dye dilution derived from the indicated samples. Aphidicolin (APH; 2 μg/ml) was also used as a positive control.
9. Top: Schematic representation of the IgG1 or IgG3 switching assay in spleen B cells. Bottom: Western blot analysis of the SAMHD1 expression in spleen B cells derived from WT or Samhd1 KO mice. Tubulin was used as a loading control.
10. Bar graphs show the compilation of IgG1 and IgG3 CSR assays from five independent experiments. The data are shown as percentage of IgG1 (left) or IgG3 (right) expressing cells (n = 5; mean ± SD; two‐tailed unpaired Student's t‐test).

Source data are available online for this figure.

RT–qPCR and immunoblot analyses confirmed that the SAMHD1 RNA and protein were dramatically decreased after introducing specific RNAi oligonucleotides (Fig 2A and D). The RNAi oligonucleotides #69 and #70 resulted in the greatest depletion of SAMHD1 and were therefore selected for subsequent analyses. The absence of SAMHD1 did not negatively affect the levels of other transcripts critical for CSR, such as the germline transcripts (μGLT and αGLT) or AID, confirming that the CSR defect was due to the absence of SAMHD1 (Fig 2D).

To demonstrate unequivocally that SAMHD1 was required for CSR, we used clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR‐associated protein 9 (Cas9) technology and generated Samhd1 KO clones of the CH12F3‐2A cells (Fig 2E, and Appendix Fig S3). To confirm the specificity of the effect of the CRISPR/Cas9‐mediated Samhd1 KO, we introduced GFP‐tagged SAMHD1 into the Samhd1 KO clones and carried out a CSR complementation assay. Complete CSR rescue was observed when the GFP‐tagged SAMHD1 was expressed in clones #17, #28, and #66, whereas the other clones showed partial CSR rescue (Appendix Fig S3A–C). Sequencing of the genomic DNA corresponding to Samhd1 exon 1 of clone #17 revealed the presence of a frameshift mutation that caused the premature termination of translation (Appendix Fig S3D). Time‐course analysis of the IgM to IgA switching in WT and Samhd1 KO clones showed a significant reduction in the IgA‐switched population among the Samhd1 KO cells, confirming the results obtained by the siRNA‐mediated KD of SAMHD1 (Fig 2F). Since CSR is closely linked with cell proliferation, we confirmed that the Samhd1 KO did not obviously compromise the proliferation of CH12F3‐2A cells as measured by cell density and CFSE dilution assay (Fig 2G and H; Appendix Fig S4A and S4B).

To further validate the requirement for SAMHD1 in CSR, we analyzed the efficiency of CSR in spleen B cells derived from WT or Samhd1 KO (Samhd1 ^−/−) C57BL/6 mice upon lipopolysaccharide (LPS)/IL4‐ or LPS alone‐stimulated switching to IgG1 or IgG3, respectively. In agreement with the data obtained from CH12F3‐2A cells, we observed a reduction in both the IgG1 switching and IgG3 switching in Samhd1 ^−/− spleen B cells (Fig 2I and J). Like Samhd1 KO CH12F3‐2A cells, the Samhd1 ^−/− spleen B cells did not show any significant defects in the proliferation (Appendix Fig S4C and D) or the expression of either AID or γ1 (γ1GTL), γ3 (γ3GLT), and μ (μGLT) germline line transcripts (Appendix Fig S4E). We next quantified the concentrations of various isotypes of Ig in the serum derived from WT or Samhd1 KO mice. Although the serum levels of IgA and IgG2b showed a tendency toward decrement (P < 0.1), the levels of other Igs remained largely unchanged (Appendix Fig S5). This is probably because serum Igs can accumulate due to selection and thus may reduce the apparent in vivo defect by SAMHD1 depletion. Taken together, these results clearly indicate that SAMHD1 is required for optimal CSR.

Because AID‐induced recurrent chromosomal translocations involving Ig loci and proto‐oncogenes are hallmarks of B‐cell malignancies, we next analyzed the effect of SAMHD1 depletion on the frequency of chromosomal translocations involving the IgH locus and the c‐Myc oncogene. We found that the siRNA‐mediated KD of SAMHD1 in CH12F3‐2A cells dramatically reduced the frequency of IgH/c‐Myc translocations (Fig 3A–C). We previously reported that the IgH/c‐Myc translocation frequency is increased dramatically by the KD of TOP1 in CH12F3‐2A cells (Husain et al, 2016). Interestingly, SAMHD1 KD in CH12F3‐2A cells stably expressing miRNA directed against Top1 mRNA drastically reduced IgH/c‐Myc chromosomal translocations even with TOP1 depletion (Fig 3D–G). Similarly, Samhd1 KO cells transfected with TOP1 siRNA were severely defective in their ability to induce IgH/c‐Myc translocations, whereas WT cells showed robust translocations (Fig 3H–J). To further confirm the requirement of SAMHD1 for IgH/c‐Myc translocations, we analyzed translocation frequency in spleen B cells derived from WT or Samhd1 KO (Samhd1 ^−/−) mice and found drastic reduction in translocation frequency in Samhd1 KO spleen B cells (Fig 3K–M). Taken together, these results suggest that SAMHD1 facilitates AID‐induced genomic instability. In fact, SAMHD1 depletion resulted in a more severe reduction in the frequency of IgH/c‐Myc translocation than in the frequency of CSR.

Figure 3. SAMHD1 depletion in CH12F3‐2A cells decreases the frequency of IgH/c‐Myc chromosomal translocations.

• A
  Top: Schematic illustration of the experimental design for the translocation assay in CH12F3‐2A cells. Bottom: PCR amplification scheme to detect IgH/c‐Myc chromosomal translocations. Triangles represent the positions of nested PCR primers used to amplify the rearranged regions. The position of the Myc probe used in the Southern blot hybridization is shown as a horizontal black bar.
• B
  Southern blot analysis of PCR‐amplified fragments with a Myc‐specific probe from two independent experiments. CH12F3‐2A cells were transfected with the indicated siRNAs.
• C
  Frequency of IgH/c‐Myc chromosomal translocations derived from two independent experiments (mean ± SD; Fisher's exact test).
• D–G
  Analysis of the IgH/c‐Myc translocation frequency in CH12F3‐2A cells expressing Tet‐inducible miRNA against Top1 mRNA. Schematic illustration of the experimental design for the translocation assay (D). Western blot analysis of the KD of SAMHD1 or TOP1 by the transfection of CH12F3‐2A cells by SAMHD1 siRNA or by tetracycline (50 nM) treatment, respectively. The expression of Tubulin is shown as a loading control (E: left). Relative expression of TOP1 and SAMHD1 as determined by densitometric quantification of Western blot images (E: right). Southern blot analysis of PCR‐amplified fragments with a Myc‐specific probe from two independent experiments (F). Frequency of IgH/c‐Myc chromosomal translocations derived from two independent experiments (mean ± SD; Fisher's exact test) (G).
• H–J
  Analysis of the IgH/c‐Myc translocation frequency in WT and Samhd1 KO CH12F3‐2A cells. (H) Schematic illustration of the experimental design for the translocation assay. (I) Southern blot analysis of PCR‐amplified fragments with a Myc‐specific probe from two independent experiments. (J) Frequency of IgH/c‐Myc chromosomal translocations derived from two independent experiments (mean ± SD; Fisher's exact test).
• K–M
  Analysis of the IgH/c‐Myc translocation frequency in spleen B cell derived from WT (Samhd1 ^+/+ ) or Samhd1 KO (Samhd1 ^−/− ) mice. (K) Schematic illustration of the experimental design for the translocation assay. (L) Southern blot analysis of PCR‐amplified fragments with a Myc‐specific probe from two independent experiments. (M) Frequency of IgH/c‐Myc chromosomal translocations derived from two independent experiments (mean ± SD; Fisher's exact test).

Source data are available online for this figure.

SAMHD1 is dispensable for AID‐induced DSBs and S‐S synapsis

We next asked whether SAMHD1 depletion affected SHM in the 5′ Sμ region, which is an efficient target for SHM, in CH12F3‐2A cells. Although Samhd1 KO CH12F3‐2A cells showed a tendency toward reduced mutation frequency, the mutation base profile remained largely unchanged (Fig 4A–C, and Appendix Fig S6A). The Sμ mutation frequency was reduced to half in primary spleen B cells derived from Samhd1 KO mice with no base bias (Appendix Fig S6B). These results were in agreement with the finding that the SHM in the V region was found to be defective in SAMHD1‐deficient mice (Thientosapol et al, 2018). However, this discrepancy between the effects of SAMHD1 depletion on SHM could be attributed to the use of different translesion polymerases in the CH12F3‐2A and spleen cells.

Figure 4. SAMHD1 depletion does not affect the formation of AID‐induced DNA double‐stranded breaks or synapsis.

1. Top: schematic representation of the Sμ region of CH12F3‐2A cells. Bottom: schematic representation of the Sμ region of CH12F3‐2A cells. The 5′ Sμ region analyzed for SHM is underlined. The genomic DNAs isolated from WT or Samhd1 KO CH12F3‐2A cells from three independent experiments were pooled prior to SHM analysis.
2. Mutation frequency in the 5′ Sμ region of WT or Samhd1 KO CH12F3‐2A cells. Statistical significance as evaluated by Fisher's exact test is shown. ns, not significant.
3. Details of the mutation analysis of the 5′ Sμ region.
4. LM‐PCR analysis of the AID‐induced DNA double‐stranded breaks in the Sμ region of WT or Samhd1 KO CH12F3‐2A cells. Wedges indicate a threefold increase in the DNA amount. Amplification of Gapdh was used as an internal loading control.
5. Gamma‐H2AX‐ChIP signal in the Sμ region is shown as the fraction of immunoprecipitated DNA (% IP) normalized to the total amount of DNA used for the immunoprecipitation (n = 3; mean + SD; two‐tailed unpaired Student's t‐test; ns, not significant).
6. Left: Scheme of long‐range interactions between Sμ–Sα elements in the IgH locus before and after AID activation. Right: Representative gel picture of the 3C assay detecting the Sμ–Sα interaction in WT or Samhd1 KO CH12F3‐2A cells with or without CIT stimulation for 24 h. Gapdh was amplified as a loading control.

To directly examine whether SAMHD1 was dispensable for S‐region cleavage, we assayed the level of DSBs using a linker ligation‐mediated PCR (LM‐PCR) assay for DSB formation at the Sμ region. We did not observe any change in the DSB level between WT and Samhd1 KO cells (Fig 4D). A histone γH2AX ChIP assay further confirmed that AID‐induced DSBs occurred normally at the 5′ Sμ region in the absence of SAMHD1 (Fig 4E). These findings indicate that SAMHD1 is dispensable for the formation of AID‐induced DNA breaks.

We next examined whether SAMHD1 depletion affected synapse formation between the Sμ and downstream Sα region of the IgH locus using a chromosome conformation capture (3C) assay in WT or Samhd1 KO CH12F3‐2 cells. Although the constitutive interaction between the Eμ and 3′Eα regions of the IgH locus was equally detected in unstimulated and CIT‐stimulated cells, the interaction between Sμ–Sα and Eμ‐Sα was significantly increased upon CIT stimulation (Fig 4F). However, the interaction between Sμ–Sα and Eμ‐Sα did not show any significant differences between CIT‐stimulated WT and Samhd1 KO cells. Taken together, these findings indicate that the inhibition of CSR by SAMHD1 depletion was not due to defects in the formation of AID‐induced DNA DSBs or in synapsis of the Sμ and Sα regions.

SAMHD1 promotes end‐joining by suppressing insertions at junctions

Because SAMHD1 depletion did not affect DNA cleavage in the Sμ region, we examined whether the next step after DNA cleavage, namely end‐joining, was affected by SAMHD1 depletion. The DSBs generated by AID during CSR are primarily joined through either the c‐NHEJ or alt‐EJ pathway (Casellas et al, 1998; Chaudhuri & Alt, 2004; Pan‐Hammarstrom et al, 2005; Yan et al, 2007). To analyze the effect of SAMHD1 depletion on the repair pathway choice, we analyzed the junction microhomology of the Sμ–Sα recombination in WT or Samhd1 KO cells after their transfection with either control (siCONT) or DNA ligase 4 siRNA (siLIG4). Junction sequence analysis revealed that the distribution of microhomology lengths of the Sμ–Sα recombination was not dramatically different between the Samhd1 KO and WT CH12F3‐2A cells. However, the Sμ–Sα junctions in the Samhd1 KO cells contained much higher frequencies of junctions with insertions longer than one nucleotide (nt), and lower frequencies of junctions with single nt insertions (Fig 5A–C). When compared with WT cells, the average lengths of the insertions were 2.69‐ and 3.75‐fold higher in Samhd1 KO cells transfected with siCONT or siLIG4, respectively (Fig 5D). To further confirm these findings, we also analyzed the junction sequences at IgH/c‐Myc translocations derived from WT or Samhd1 KO cells. Like the Sμ–Sα junctions, we detected insertions at the IgH/c‐Myc translocation junctions only in the Samhd1 KO cells, indicating an increase in the frequency of IgH/c‐Myc translocations containing insertions (Appendix Fig S7). These findings collectively show that SAMHD1 has a critical role in the end‐joining of AID‐induced breaks by preventing the insertion‐repair.

Figure 5. SAMHD1 suppresses insertions during DNA double‐stranded break repair end‐joining.

• A–D
  Analysis of Sμ–Sα recombination junctions from WT or CH12F3‐2A cells after transfection with control (siCONT) or Ligase 4 (siLIG4) siRNA. The genomic DNAs isolated from WT or Samhd1 KO cells from three independent experiments were pooled prior to junction analysis. A total of 62 (siCONT) and 73 (siLIG4) junctions from WT, and 60 (siCONT) and 60 (siLIG4) junctions from Samhd1 KO cells were sequenced and analyzed. Statistical significance as evaluated by Fisher's exact test (B) or two‐tailed unpaired Student's t‐test (D) or is shown. Schematic representation of the Sμ–Sα junction analysis (A). Summary of the characteristics of Sμ–Sα recombination junctions (B). Sequences at Sμ–Sα recombination junctions with insertions (C). Summary of the average length of insertions at Sμ–Sα recombination junctions (d).
• E–J
  Analysis of the CRISPR/Cas9‐induced chromosomal translocations between the Rosa26 (Chr6) and H3f3b (Chr 11) genes. The genomic DNAs isolated from WT or Samhd1 KO cells from three independent experiments were pooled prior to chromosomal translocation or junction analysis. Schematic representation of the chromosomal translocations (E). Representative images of agarose gels after PCR amplification of the Rosa26/H3f3b translocation from WT or Samhd1 KO CH12F3‐2A cells (F). Relative frequency of Rosa26/H3f3b chromosomal translocations (G). Frequency of junctions containing insertions (H) and average length of insertions (I). Statistical significance as evaluated by two‐tailed unpaired Student's t‐test (G, I) or Fisher's exact test (H) is shown.

We next examined whether SAMHD1 also affected the repair of DSBs induced by events other than AID. To this end, we induced DSBs in the Rosa26 and H3f3B mouse loci using the CRISPR/Cas9 system (Fig 5E). When introduced into WT and Samhd1 KO cells, the Cas9 and gRNA (Rosa26; H3f3b) expression plasmid caused robust chromosomal translocations between these loci (Fig 5F). However, we did not observe any obvious decrease in the frequency of Rosa26/H3f3b translocations in the Samhd1 KO cells (Fig 5F and G). Interestingly, as in CSR and the IgH/c‐Myc translocation junctions, we found that the insertional events were significantly increased in both frequency and average length at the Rosa26/H3f3b translocation junctions in the Samhd1 KO cells (Fig 5H and 5I; Appendix Fig S8). Taken together, our results indicate that SAMHD1 plays a critical role in repair by suppressing the incorporation of longer insertions during end‐joining.

Depletion of dNTP by SAMHD1 promotes AID‐induced recombination

Because SAMHD1 has both dNTPase‐dependent and dNTPase‐independent functions (Goldstone et al, 2011; Powell et al, 2011; Beloglazova et al, 2013; Franzolin et al, 2013; Ryoo et al, 2014), we next asked if the defects in CSR and IgH/c‐Myc translocations upon SAMHD1 depletion were attributed to the accumulation or imbalance of the deoxyribonucleotide pools. To examine this possibility, we performed rescue experiments in Samhd1 KO CH12F3‐2A cells with D311A, an allosteric mutant; D137N, an active site mutant; R451E, a tetramerization defective mutant; or T592A, a phospho‐ablative mutant of human SAMHD1 (Fig 6A). The D311A, D137N, and R451E mutations are known to disrupt SAMHD1's dNTPase activity, whereas the T592A mutation impairs the viral restriction activity of SAMHD1 while preserving its dNTPase activity (Goldstone et al, 2011; White et al, 2013; Zhu et al, 2013). The IgM to IgA switch was rescued in cells expressing WT, but not the dNTPase‐defective mutants (Fig 6B and C). The expression of the phosphorylation‐defective mutant (T592A) caused a significant but incomplete rescue of CSR, which could have been due to the reduced dNTPase activity in the SAMHD1 T592A mutant. These results together indicate that the dNTPase activity of SAMHD1 is required for efficient CSR.

Figure 6. Accumulation of purine deoxyribonucleotides is responsible for the defects in CSR and IgH/c‐Myc translocations.

1. Top: Schematic representation of GFP‐fused WT‐ or mutated‐SAMHD1 proteins. Bottom: Schematic representation of the CSR complementation assays.
2. Confirmation of the expression of GFP‐tagged human SAMHD1 (WT, DN, DA, TA, and RE mutants). Histone H3 expression is shown as a loading control.
3. CSR complementation assay in Samhd1 KO cells. Data are shown as the % IgA switching relative to the IgA switching in WT cells transfected with empty vector and are derived from three independent experiments (mean ± SD; two‐tailed unpaired Student's t‐test).
4. Effects of SAMHD1 depletion or BRD2 depletion on the level of metabolites involved in nucleotide metabolism as analyzed by CE‐TOF‐MS. BRD2 knockdown used as an unrelated control. The deoxyribonucleoside triphosphates or the metabolites derived from them are boxed (n = 2; mean ± SD).
5. Effect of deoxyribonucleoside supplementation individually or in combination on the rate of IgM to IgA switching in CH12F3‐2A cells. Data are presented as % IgA switching relative to the IgA switching in cells treated with an equivalent volume of the solvent in which the deoxyribonucleosides were dissolved (n = 3; mean ± SD; two‐tailed unpaired Student's t‐test; **P ≤ 0.01; ***P ≤ 0.001).
6. Effect of deoxyribonucleoside supplementation individually or in combination on the frequency of IgH/c‐Myc translocations in CH12F3‐2A cells (n = 2; mean ± SD; Fisher's exact test; ns, not significant; ***P ≤ 0.001). The concentration of dNTPs used for supplementation either individually or in combination were as follows: 2′‐deoxyguanosine (dG), 0.15 mM; 2′‐deoxyadenosine (dA), 1.0 mM; 2′‐deoxycytidine (dC), 12 mM; and 2′‐deoxythymidine (dT), 1.6 mM.

Source data are available online for this figure.

A recent study showed that SAMHD1 interacts with C‐terminal binding protein interacting protein (CtIP) and promotes DNA end resection independent of its dNTPase activity (Daddacha et al, 2017). Thus, we next analyzed if K484T, a SAMHD1 mutant with dNTPase activity but defective in the interaction with CtIP, could rescue the CSR in Samhd1 KO cells. We found that expressing K484T in Samhd1 KO rescued CSR as efficiently as WT SAMHD1, whereas D311A, a mutant with impaired dNTPase activity, failed to rescue the CSR (Appendix Fig S9). These findings indicated that the defects in CSR upon SAMHD1 depletion were due to the loss of its dNTPase activity but not its recruitment of CtIP.

Since the dNTPase activity of SAMHD1 was required for efficient CSR, we next compared the dNTP levels in Samhd1 KO or WT CH12F3‐2A cells using a fluorescence‐based PCR assay, in which the incorporation of a limiting dNTP is required for primer extension, and the Taq polymerase‐mediated 5′–3′ exonuclease hydrolysis of a dual‐quenched fluorophore‐labeled probe results in fluorescence (Wilson et al, 2011). This assay revealed that there were substantially higher levels of purine deoxyribonucleotides (dGTP and dATP) in the Samhd1 KO cells than in WT cells, whereas smaller increases in the pyrimidine deoxyribonucleotide levels (dTTP and dCTP) were found (Appendix Fig S10). The greatest increase was observed in the level of dGTP, the preferred substrate of SAMHD1. These findings also suggest that the small pool size of dGTP among the four dNTPs might not have been due to low dGTP synthesis but to its more efficient catabolism by SAMHD1, as previously proposed (Franzolin et al, 2013). To show that the effect of SAMHD1 depletion on CSR and IgH/c‐Myc translocation was not due to the modulation of metabolites other than nucleotides or their derivatives, we analyzed the levels of charged metabolites by capillary electrophoresis time‐of‐flight mass spectrometry (CE‐TOF‐MS). This analysis revealed that Samhd1 KO in CH12F3‐2A cells leads to dramatic accumulations of dATP and dGTP, both of which were below the detection limit of the CE‐TOF‐MS in the WT cells (Fig 6D). In contrast, the level of dTTP was the same in WT and Samhd1 KO cells, and the level of dCTP in both WT and Samhd1 KO cells was below the detection limit. Notably, this metabolome analysis did not reveal any significant differences in the levels of other metabolites (Appendix Table S2). These findings suggested that the accumulation of dGTP and/or dATP might be responsible for the decreased frequencies of CSR and IgH/c‐Myc translocations in the SAMHD1‐depleted cells.

Salvage of deoxyribonucleosides decreases AID‐induced recombination

To further confirm that the accumulation of dNTPs upon SAMHD1 depletion was responsible for the defects in CSR and IgH/c‐Myc translocations, we next analyzed the effect of deoxyribonucleoside supplementation on these events. The uptake of deoxyribonucleosides results in deoxyribonucleotide synthesis by a salvage pathway. We found CSR inhibition was more sensitive to supplementation of purine nucleosides as 0.15 mM 2′‐deoxyguanosine (dG) or 1.0 mM 2′‐deoxyadenosine (dA) reduced the CSR to < 50%, whereas 12 mM 2′‐deoxycytidine (dC) or 1.6 mM 2′‐deoxythymidine (dT) was required to achieve a similar reduction in CSR (Fig 6E; Appendix Fig S11A and B). The combination of purine deoxyribonucleosides (dG + dA) or pyrimidine deoxyribonucleosides (dT + dC), or their mixture (dG + dA + dT + dC), did not show additive effects on CSR. Like CSR, supplementation with purine deoxyribonucleosides (dG or dA) or pyrimidine deoxyribonucleosides (dT or dC) alone, or a mixture of these (dG + dA + dT + dC), decreased the frequency of the IgH/c‐Myc translocations, albeit to a lesser extent than the CSR (Fig 6F). Interestingly, the mixture of purine nucleosides (dG + dA), but not of pyrimidine deoxyribonucleosides (dT + dC), showed an additive effect on the reduced frequency of IgH/c‐Myc translocations in WT, which reached a similar level as that observed in Samhd1 KO CH12F3‐2A cells. Taken together, these findings suggest that the accumulation of purine nucleotides upon SAMHD1 depletion interferes with the more efficient NHEJ pathway resulting in the decrements in both CSR and IgH/c‐Myc translocations.

SAMHD1 deficiency affects Cas9 nickase‐induced CSR

Because DNA cleavage by WT (Cas9) or Cas9 nickase (D10A mutant of Cas9) in the Sμ and Sα regions is known to induce CSR in CH12F3‐2A cells (Ling et al, 2018; Perez‐Riverol et al, 2019), we utilized this system to understand the molecular basis of the repair defects observed in Samhd1 KO cells. In this system, using specific guide RNAs in combination with either WT or Cas9 nickase (Cas9n) well‐defined blunt or staggered DSBs can be generated in the S regions (Fig 7A). When we introduced blunt DSBs in the Sμ and Sα, robust CSR both in WT and Samhd1 KO cells was induced (Fig 7B and C). However, when guide RNAs in combination with Cas9 nickase that generates staggered DSB with 5′ overhangs in Sμ and Sα were used, the CSR was less efficient as previously reported (Ling et al, 2018). Interestingly, the frequency of CSR was further decreased in Samhd1 KO cells (Fig 7B and C). Since AID induces staggered DNA breaks in the IgH locus, efficient end processing is necessary for the repair of DSBs by NHEJ (Rush et al, 2004). These results imply that the accumulation of dNTPs by SAMHD1 deficiency affects the processing and/or end‐joining of the staggered DSBs. In order to verify these possibilities, we performed recombination junction analysis from IgA‐switched cells, following Cas9 or Cas9n‐induced CSR. As expected, Cas9‐induced CSR junctions were predominantly blunt and contained short microhomologies; however, we observed a decrement in the length of resected DNA (P = 0.10) at the junctions in Samhd1 KO cells (Fig 7D and E; Appendix Fig S12). Strikingly, barely any insertion was observed in the recombination junctions, suggesting that sharp blunt or near‐blunt DSBs are not the source of the insertions observed earlier in Samhd1 KO cells. On the contrary, the joining of DSBs with 5′ overhangs showed increments in both the insertional events as well as the average length of insertions without any significant change on the length of resected DNA and the use of microhomology (Fig 7F–I). These data strongly suggest that SAMHD1 is required for the joining of staggered DSBs, which constitute a major portion of AID‐induced breaks (Rush et al, 2004). Thus, SAMHD1‐mediated dNTP balance appears to be crucial for suppressing the insertional DNA repair pathway and promoting appropriate end processing during end‐joining.

Figure 7. SAMHD1 promotes the CSR from staggered DSBs.

• A
  Top: Schematic representation of the IgH locus depicting Sμ and Sα region with the position of guide RNAs (gRNA) that were used to induce blunt end DSBs using WT SpCas9. Bottom: IgH locus depicting Sμ and Sα region with the position of guide RNAs (gRNA) that were used to induce staggered DSBs with 5′ overhangs using D10N mutants of SpCas9 (Cas9n).
• B
  FACS profile of the percentage of WT and Samhd1 KO CH12F3‐2A cells undergoing IgA switching following transfection with the indicated plasmids. The number in the FACS plot represents the percentage of cells expressing IgA on their surface.
• C
  Summary of the FACS data obtained from two independent experiments. The data are shown as percentage of IgA expressing cells upon CSR induced by blunt (left) or staggered (right) DSBs (n = 2; mean + SD; two‐tailed unpaired Student's t‐test); ns, not significant.
• D
  Length of resected DNA at Sμ‐Sα junctions obtained from the joining of DSBs with blunt ends. Statistical significance as evaluated by two‐tailed unpaired Student's t‐test is shown.
• E
  Summary of the characteristics of Sμ–Sα recombination junctions resulting from the joining of DSBs with blunt ends.
• F
  Length of resected DNA at Sμ‐Sα junctions obtained from the joining of staggered DSBs. Left: Length of Sμ and Sα overhangs was 35 and 98 nt, respectively. Right: Length of Sμ and Sα overhangs was 35 and 187 nt, respectively.
• G
  Summary of the characteristics of Sμ–Sα recombination junctions resulting from the joining of staggered DSBs. Statistical significance as evaluated by Fisher's exact test is shown.
• H
  Summary of the average length of insertions at Sμ–Sα recombination junctions. Statistical significance as evaluated by two‐tailed unpaired Student's t‐test is shown.
• I
  Sequences of Sμ–Sα recombination junctions with insertions. The middle column indicates the insertions (in) found in different clones at the junctions.
• J–M
  Analysis of the effect of TdT expression on the frequency of CSR as well as characteristics of Sμ–Sα recombination junctions resulting from the joining of DSBs with blunt ends. (J) Top: Schematic representation of the experimental design to analyze the effect of TdT expression on CSR in WT and Samhd1 KO cells. Bottom: Western blot analysis of the expression of TdT in WT and Samhd1 KO cells. Tubulin was used as a loading control. (K) FACS profile of the percentage IgA switching in WT or Samhd1 KO CH12F3‐2A cells with or without TdT expression. (L) Summary of the FACS data obtained from two independent experiments (n = 2; mean + SD; two‐tailed unpaired Student's t‐test; *P ≤ 0.05). (M) Summary of the characteristics of Sμ–Sα recombination junctions. Statistical significance as evaluated by Fisher's exact test is shown. ns, not significant.

Source data are available online for this figure.

In order to test the possibility that SAMHD1 deficiency‐induced elevated dNTP level may affect the activity of DNA end processing or repair enzymes, we tested the effect of Samhd1 KO on terminal deoxynucleotidyl transferase (TdT), which can add deoxyribonucleotides to the 3′ hydroxyl end of the DSBs in a template‐independent manner (Kato et al, 1967). Since CH12F3‐2A cells do not express TdT, we examined the effect of TdT expression on Cas9‐induced CSR in WT and Samhd1 KO CH12F3‐2A cells (Fig 7J). Interestingly, TdT expression did not affect CSR in WT cells, but reduced it to almost half in Samhd1 KO cells (Fig 7K and L). The decrement in the CSR by TdT expression in Samhd1 KO cells correlates well with decreased blunt end‐joining and increased insertional joining (Fig 7M). These data strongly suggest that elevated levels of dNTPs can modulate the activity of TdT like DNA polymerases (e.g., POLQ) which interfere with the normal ligation program and that the end‐joining by pathways that lead to increments in insertional events does not support optimal CSR. Together, our data strongly implicate the role of SAMHD1 dNTPase activity in DSB repair by NHEJ pathways.

Discussion

In the present study, we applied a locus‐specific proteomics approach to isolate novel DNA repair factors bound to the IgH Sα locus during CSR. In addition to many known CSR factors and NHEJ repair proteins, we found a few novel proteins including SAMHD1, whose function has not been investigated in the context of AID‐induced genomic instability. Although SAMHD1 was isolated by trapping Sα region of the IgH locus, it was also present in the universal donor Sμ, which indicated its possible involvement in Ig‐isotype switching (Appendix Fig S9E). Consistently, not only CH12F3‐2A cells but also splenic B cells derived from Samhd1 ^−/− KO mice showed decreased Ig‐isotype switching in response to in vitro CSR induction. Although the serum levels of the IgA and IgG2b Igs showed a tendency toward a decrease (P < 0.1), the levels of other Igs remained largely unchanged. Since CSR was not completely inhibited in Samhd1 KO mice, a significant reduction in the serum levels of Ig was not expected due to the accumulation of Igs over time. Subsequently, we demonstrated that not only CSR but also associated chromosomal translocations were severely impaired upon SAMHD1 deficiency in CH12F3‐2A as well as spleen B cells. Analysis of the Sμ–Sα and IgH/c‐Myc recombination junctions in the Samhd1 KO B‐cell line revealed an increased frequency of longer insertions, without any significant effect on the utilization of junction microhomology lengths (Fig 5A–D). Similarly, CRISPR/Cas9‐induced Rosa26/H3f3b translocations showed significant increase in the frequency and lengths of the insertions at translocation junctions (Fig 5E–I).

Activation induced cytidine deaminase‐induced DSBs and S‐S synapse, both of which are prerequisites of CSR, remained unchanged upon SAMHD1 depletion (Fig 4). These observations suggested that SAMHD1 is involved in regulating the end‐joining of the DSBs. Although a novel CtIP‐dependent DNA repair function of SAMHD1 has been reported (Daddacha et al, 2017), the K484T mutant, which abrogates SAMHD1's interaction with CtIP and homology‐directed repair, fully complemented the CSR deficiency of Samhd1 KO CH12F3‐2A cells (Appendix Fig S9A–D). However, we did notice decrements in the resection during end‐joining in the absence of SAMHD1, but this was likely independent of the interaction of SAMHD1 with CtIP (Fig 7D). To further explore the function of SAMHD1 in CSR, we took advantage of well‐documented SAMHD1 catalytic mutants and examined their CSR complementation potential in CH12F3‐2A cells. In contrast to K484T mutant, the catalytically defective mutants of SAMHD1 failed to rescue CSR in Samhd1 KO cells, suggesting that the dNTPase activity of SAMHD1 is essential for efficient CSR.

Although SAMHD1 associates with the Ig locus, indicating a local role of its dNTPase activity, a global depletion of the dNTP pool may also promote end‐joining at the Ig locus. The local role of metabolic enzymes such as fumarase in DNA repair has been previously demonstrated (Jiang et al, 2015). Thus, it is possible that the dNTPase activity of SAMHD1 may be more efficient in depleting the dNTP pool locally at the break site. Alternatively, the possibility of an additional dNTP‐independent role at the repair site similar to the loading of CtIP as previously reported cannot be ruled out.

The requirement of SAMHD1 dNTPase activity was in agreement with the accumulation of purine dNTPs in Samhd1 KO CH12F3‐2A cells, and the CSR inhibition upon the addition of deoxyribonucleosides to the culture medium. Consistent with our finding that higher levels of the purine dNTPs (dGTP and dATP) than the pyrimidine dNTPs (dTTP and dCTP) accumulated in Samhd1 KO cells, the inclusion of purine deoxyribonucleosides, particularly dG, in the culture medium inhibited CSR more strongly (Fig 6E). Similarly, a combination of purine deoxyribonucleosides (dG and dA) but not of pyrimidine deoxyribonucleosides (dT and dC) reduced the frequency of IgH/c‐Myc translocations to a level similar to that observed upon SAMHD1 depletion (Fig 6F). Taken together, these findings indicated that the accumulation of purine dNTPs is likely responsible for the inhibition of CSR and IgH/c‐Myc translocations.

Although SAMHD1 is required for the end‐joining of AID‐induced DNA breaks during both CSR and IgH/c‐Myc translocations, it does not affect the frequencies of end‐joining of blunt end DSBs in CRISPR/Cas9‐induced CSR and Rosa26/H3f3b‐induced translocations. However, the frequency of CSR by CRISPR/Cas9‐induced staggered DSBs was significantly reduced by SAMHD1 depletion. Notably, the majority of the AID‐induced DSBs possess staggered ends which require extensive end processing for their effective joining by c‐NHEJ or alt‐EJ pathways (Rush et al, 2004).

It is worth mentioning that the depletion of SAMHD1 results in more drastic reductions in IgH/c‐Myc translocations than CSR. It has been shown that translocations are promoted by DSBs with 5′ overhangs (Ling et al, 2018; Perez‐Riverol et al, 2019), whose joining requires SAMHD1‐mediated optimal dNTP balance. The involvement of SAMHD1's dNTPase activity in CSR and IgH/c‐Myc translocation suggests that the limited dNTP pool favors DNA repair and thereby the recombination efficiency. Remarkably, a number of recent reports indicate that DNA polymerase theta (POLQ) which requires minimal or no DNA template to extend a DNA break end is responsible for nucleotide insertion during DNA break repair, a signature feature of POLQ‐mediated end‐joining (TMEJ) (Chan et al, 2010; Yousefzadeh et al, 2014; Wyatt et al, 2016; Zelensky et al, 2017). Notably, splenic B cells derived from Polq‐null mice not only lack insertions at the CSR junctions but also show an approximately 4‐fold increase in the frequency of IgH/c‐Myc translocations (Yousefzadeh et al, 2014), indicating that POLQ‐mediated repair, which leads to the incorporation of non‐templated insertions, is inhibitory for CSR and IgH/c‐Myc translocations. On the other hand, POLQ can also promote genome instability, as indicated by its requirement for Cas9‐induced translocations and for the fusion of chromosomes at de‐protected telomeres in NHEJ‐deficient cells (Mateos‐Gomez et al, 2015). Consistently, POLQ‐deficient CH12F3‐2A cells showed a drastic reduction in CRISPR/Cas9‐induced Rosa26/H3f3b translocation (Appendix Fig S13). Interestingly, however, when the assay was performed to detect the translocation in the dicentric orientation (type‐3), numerous bands with a range of lengths were detected, suggesting that cells may adopt various modes of DSB repair depending on the nature of the DSB structure, their orientation, and available repair enzymes during ligation.

Therefore, we speculated that the accumulation of purine dNTPs upon SAMHD1 depletion may modulate activity of POLQ‐like enzymes and can lead to the synthesis of longer insertions at the recombination junctions during DSB repair. In fact, the expression of TdT in Samhd1 KO cells increased the frequency of insertional events and reduced CSR activity, confirming that elevated levels of dNTPs may modulate the activities of DNA polymerases. In addition, TdT expression may also convert blunt ends into staggered and further extend existing overhangs and thus reduce their end‐joining in Samhd1 KO cells (Fig 7J–M). Together, these findings suggest that dNTP imbalance can modulate activities of DNA repair enzymes.

We propose that AID‐induced staggered DSBs are efficiently processed and repaired by c‐NHEJ and alt‐EJ when the dNTP pool is limited by SAMHD1, and TMEJ or another pathway that leads to templated sequence insertion (TSI) is restricted (Appendix Fig S14). Elevated dNTP levels may act as a triggering factor for TMEJ or TSI pathways leading to DSB repair with templated or non‐templated insertions (Keskin et al, 2014; Onozawa et al, 2014; Chakraborty et al, 2016). In addition, an elevated dNTP pool may also affect end‐resection required for the efficient end‐joining of the breaks with overhangs. In fact, SAMHD1 has been shown to be required for end‐resection, which, at least in case of AID‐induced breaks, appears to be independent of the interaction of SAMHD1 with CtIP.

In SAMHD1 deficiency, many insertions are derived from different genomic fragments including the transfected Cas9 plasmid at the repair‐junctions of Rosa26/H3f3b translocation. It is likely that they are either copied or captured as a fragment during the repair of the DSBs. Long templated insertion of 507 bp and 732 bp from LIAR1 genomic loci was observed between the V and DJ segments in the human Ig locus, which led to increased antibody diversity to cope with distinct malarial parasites (Pieper et al, 2017). Recently, long multiple insertions at the genomic break points have been reported in Dna2 nuclease‐deficient yeast cells (Yu et al, 2018). Interestingly, impaired V(D)J recombination was also observed due to purine dNTP pool imbalance in adenosine deaminase (ADA) deficiency (Gangi‐Peterson et al, 1999). In the absence of ADA, the elevated dATP pool promoted template‐independent nucleotide insertion at the recombination sites by terminal deoxyribonucleotide transferase (TdT).

Repair of DSBs by the insertion of mobile elements or fragments from mitochondrial or nuclear DNA destabilizes the genome and leads to prevalent genomic aberrations in various cancers including B‐cell tumors (preprint: Li et al, 2017; Pieper et al, 2017; Yi & Ju, 2018). The involvement of dNTPase activity in end‐joining also suggests that dNTP depletion by SAMHD1 not only limits the reverse transcription of viral RNA but may also affect the filling of the DNA gap during the integration of viral DNA into the genome by limiting the activity of DNA polymerases other than POLB (Van Cor‐Hosmer et al, 2013; Goetze et al, 2017). Possible alterations of additional dNTP sensitive DNA/RNA helicases such as DDX1, MTR4, SETX, or BLM may not be mutually exclusive, as they have been shown to modulate CSR activity (Babbe et al, 2009; Lim et al, 2017; Ribeiro de Almeida et al, 2018).

Materials and Methods

Establishment of cell lines for iChIP

The sgRNA sequence recognizing the downstream Sα region was cloned into the pX458 CRISPR/Cas9 plasmid. To construct the targeting vector for LexA sequence insertion, homology arms corresponding to the 5′ and 3′ region of the Cas9‐sgRNA cleavage site were cloned into a plasmid vector containing a loxP‐pGK‐EGFP‐IRES‐Puro‐loxP selection cassette. The 160‐bp 8X‐LexA sequence was subcloned from 8xLexA‐binding elements/pMD20 (Addgene plasmid #48807) (Fujita & Fujii, 2011) and placed between the 5′ homology arm and the selection cassette. The Cas9‐sgRNA plasmid and targeting vector were introduced into CH12F3‐2A cells by nucleofection, and cells were selected by puromycin. One week later, the EGFP‐expressing cells were sorted, and single clones were obtained by limited dilution. Positive clones with 8X‐LexA and the selection cassette inserted at the Sα locus were identified through PCR screening followed by Southern blot validation. The EGFP‐Puro selection cassette was removed by the transient introduction of Cre‐ERT2 and the addition of 4‐OHT in cells, resulting in the Sα‐LexA cell line, which harbored the 8X‐LexA sequence and one loxP site in the Sα locus. Wild‐type CH12F3‐2A cells and Sα‐LexA engineered cells were then transfected with pEF1α‐FNLDD‐IRES‐ZsGreen plasmid expressing FNLDD to obtain the final FNLDD‐alone and FNLDD‐Sα‐LexA cell lines for iChIP assay (Hoshino & Fujii, 2009). For the construction of pEF1α‐FNLDD‐IRES‐ZsGreen, the FNLDD sequence was subcloned from 3xFNLDD/pCMV‐7.1 (Addgene plasmid #48874) (Fujita et al, 2015).

iChIP and MS analysis

The Sα‐engineered (FNLDD‐Sα‐LexA) CH12F3‐2A cells were cultured in RPMI‐1640 and fetal calf serum (FCS) provided in the Pierce SILAC Protein Quantitation Kit (Thermo Fisher Scientific, MA, USA) with 4 mM glutamine, 17 mM HEPES, pH 7.0, 50 μM 2‐mercaptoethanol, ¹³C₆ ¹⁵N_2‐ l‐Lysine‐2HCl (Thermo Fisher Scientific, MA, USA), and ¹³C₆ ¹⁵N₄ l‐Arginine‐HCl (Thermo Fisher Scientific, MA, USA) according to the manufacturer's instructions. The control WT CH12F3‐2A cells expressing FNLDD (FNLDD alone) were cultured in medium containing unlabeled Lysine‐2HCl, and l‐Arginine‐HCl. Fifteen million isotopically labeled (FNLDD‐Sα‐LexA) and non‐labeled (FNLDD alone) cells were mixed and fixed with 1% formaldehyde at 37°C for 10 min. The chromatin fraction from the fixed cells was extracted as described previously (Kustatscher et al, 2014) with minor modifications. Briefly, the cell pellet obtained after formaldehyde fixing and quenching was resuspended in 1 ml of ice‐cold cell lysis buffer (20 mM Tris, pH 8, 85 mM KCl, 0.1% TX‐100) and incubated for 10 min on ice. The cell suspension was centrifuged at 2,000 g for 5 min at 4°C, and the nuclear pellet was resuspended in 330 μl of SDS lysis buffer (50 mM Tris, pH 8, 10 mM EDTA, 4% SDS) and incubated for 10 min at room temperature. After incubation, 1.0 ml of urea buffer (20 mM Tris, pH 8, 1 mM EDTA, 8 M Urea) was added, and the solution was mixed thoroughly and centrifuged at 16,100 g for 30 min at 25°C. The gelatinous chromatin pellet was washed with PBS and resuspended in 800 μl of shearing buffer (20 mM Tris, pH 8, 150 mM NaCl, 1 mM EDTA, 0.1% (v/v) SDS, 0.1% (w/v) Na‐deoxycholate) and fragmented by sonication to an average length of about 2 kbp. The chromatin lysate was cleared by centrifugation at 16,100 g, and used for immunoprecipitation. All the buffers described above were supplemented with protease and phosphatase inhibitor cocktails (Nacalai Tesque, Japan).

The chromatin lysates from two independent preparations were pooled, supplemented with 1% TritonX‐100, incubated with 30 μg of the M2 anti‐FLAG Ab at 4°C overnight, and then incubated with 300 μl of Dynabeads‐Protein G at 4°C for 1 h. The Dynabeads were washed two times each with 1.5 ml of low salt wash buffer, high salt wash buffer, and LiCl wash buffer, and once with 1.5 ml of TBS buffer (50 mM Tris–HCl [pH 7.5], 150 mM NaCl) with 0.1% NP‐40. The immunoprecipitants were eluted twice with 200 μl of elution buffer (500 μg/ml 3x‐FLAG peptide in TBS buffer) at 37°C for 20 min. The chromatin eluates from two such reactions were pooled and precipitated in 1 ml of 2‐propanol with 50 μl of 3 M sodium acetate and 5 μl of 20 mg/ml glycogen at −20°C overnight. After centrifugation (17,400 g) at 4°C for 30 min, the precipitants were washed with 1 ml of 70% ethanol and then incubated in 50 μl of 2× Sample Buffer at 98°C for 30 min, for reverse‐crosslinking and denaturation of proteins.

The reverse‐crosslinked proteins were subjected to SDS–PAGE and analyzed using a nanoLC‐MS/MS system consisting of an LTQ Orbitrap Velos (Thermo Fisher Scientific, MA, USA) coupled with Advance nanoLC (Michrom BioResources, CA, USA) and an HTC‐PAL autosampler (CTC Analytics) at the DNA‐chip Development Center for Infectious Diseases (RIMD, Osaka University, Japan). Data were acquired using Xcalibur™ software (Thermo Fisher Scientific, MA, USA). Quantification was performed using Proteome Discoverer 1.2 (Thermo Fisher Scientific, MA, USA) and the Mascot search engine (Matrix Science, UK) for peptide identification against the SWISS‐PROT database. The initial mass tolerance was set to 10 ppm, and MS/MS mass tolerance was 0.8 Da. The enzyme was set to trypsin/p with two missed cleavages. Carbamidomethylation of cysteine was searched as a fixed modification, whereas N‐acetyl‐protein and oxidation of methionine were searched as variable modifications. A minimum of two peptides were quantified for each protein.

RNAi oligonucleotide transfection

The mouse B‐cell lymphoma line CH12F3‐2A expressing Bcl2 was previously described (Stanlie et al, 2010). Chemically modified Stealth siRNA oligonucleotides (Thermo Fisher Scientific, MA, USA) were introduced into cells using the Nucleofector 96‐well electroporation system (Lonza, Switzerland) to knock down the expression of specific genes. After electroporation, the cells were cultured for 24 h and then were stimulated by CIT (CD40L, IL4, and TGFβ) or OHT (1 mM), and cultured for another 24–48 h before collection and analysis. The lists of antibodies, primers, and Stealth siRNAs used in this study are shown in Appendix Tables S3–S5.

Analysis of CSR and SHM

CH12F3‐2A cells were CIT‐stimulated for 24–48 h to induce CSR, and the surface expression of IgM and IgA was analyzed by staining the cells with FITC‐conjugated anti‐mouse IgM (eBioscience, CA, USA) and PE‐conjugated anti‐mouse IgA (SouthernBiotech, AL, USA) antibodies, respectively. To analyze the CSR efficiency of WT or Samhd1 KO (Samhd1 ^−/−) mice, red blood cell‐depleted spleen B cells derived from these mice were stimulated with LPS and IL4 (for IgG1 switching) or with LPS alone (for IgG3 switching). After 5 days of incubation, the cells were stained with biotinylated anti‐IgG1 or IgG3 coupled with allophycocyanin‐labeled streptavidin and PE‐B220 antibodies. Flow cytometry analyses were performed using a FACSCalibur (BD, NJ, USA) as previously described (Begum et al, 2012).

Analysis of IgH/c‐Myc chromosomal translocations

The IgH/c‐Myc translocation junctions (derivative chromosome 15) were PCR‐amplified from genomic DNA obtained from 48‐h CIT‐stimulated CH12F3‐2A cells using the Expand Long Template PCR System (Boboila et al, 2012b; Stanlie et al, 2014). A total of 16–48 aliquots of genomic DNA were analyzed in separate reactions. The conditions for both the first and second rounds of PCR were as follows: 94°C for 3 min, followed by 25 cycles at 94°C for 15 s, 62°C for 15 s, 68°C for 7.5 min, and a final extension of 5 min at 68°C. The PCR products were separated by electrophoresis on ethidium bromide‐containing 1% agarose gels and subjected to Southern blotting with a Myc‐specific probe. The primer and probe sequences are shown in Appendix Table S4.

Analysis of DNA double‐stranded breaks

Genomic DNAs were prepared from CH12F3‐2A cells stimulated with CIT for 24 h. DSBs were analyzed by ligation‐mediated PCR as previously described (Schrader et al, 2005; Xu et al, 2014). Briefly, live CH12F3‐2A cells were isolated through Percoll density gradient centrifugation and embedded in low‐melting‐point agarose, followed by DNA extraction within low‐melting agarose plugs. A total of 20 μl of genomic DNA was ligated with linker in a final volume of 100 μl. The ligation reaction was stopped by adding 200 μl of H₂O followed by heating at 70°C for 10 min. Threefold dilutions of the input DNA were PCR amplified by KOD‐FX‐Neo polymerase (Toyobo, Japan). The amount of ligated DNA in the input sample was normalized to Gapdh DNA. The PCR products were separated by electrophoresis on 1% agarose gels followed by Southern blotting with a 50 selectable marker (Sμ) probe. The primer and probe sequences are shown in Appendix Table S4.

Author contributions

AH, NAB, JX, MK, and TH designed the research; AH, JX, MN, and NAB performed the research; AH, NAB, JX, and TH analyzed the data; HF provided reagents for iChIP and performed proteomic analysis; JR provided spleens and serum from WT and Samhd1 KO mice; J‐YW provided POLQ‐deficient CH12F3‐2A cells; and AH, NAB, and TH wrote the manuscript.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Source Data for Appendix

Review Process File

Source Data for Figure 2

Source Data for Figure 3

Source Data for Figure 6

Source Data for Figure 7

The EMBO Journal (2020) 39: e102931

Data availability

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (Perez‐Riverol et al, 2019) partner repository with the dataset identifier PXD019042 (https://www.ebi.ac.uk/pride/archive/projects/PXD019042).