SMUG1 is able to excise uracil from immunoglobulin genes: insight into mutation versus repair

Transfected SMUG1 affects IgV mutation spectrum but reduces mutation load in DT40 cells

SMUG1 is able to excise uracil from both single- and doubled-stranded DNA in vitro as judged by biochemical assays (Kavli et al, 2002). Furthermore, siRNA-mediated knockdown of SMUG1 expression in mouse embryo fibroblasts leads to increased C:G → T:A transition mutations, suggesting that endogenous SMUG1 can access and repair uracil in chromosomal DNA in mouse cells in vivo (An et al, 2005). We were therefore surprised by our recent finding (Rada et al, 2004) that enforced overexpression of SMUG1 does not restore efficient in vitro switching by UNG-deficient mouse splenic B cells. The observation suggested to us that either (i) SMUG1 is, for some unidentified reason, unable to gain access to the AID-generated U:G lesion in the immunoglobulin switch region, (ii) that is able to gain access but cannot do so in such a way as to potentiate switch recombination or (iii) that it can potentiate switch recombination but does so at an efficiency that scored below our sensitivity of detection.

To gain some insight into the possible explanation, we extended our studies to a different system and asked whether transfected SMUG1 could affect the pattern of IgV gene diversification in chicken DT40 cells. Parental DT40 cells diversify their IgV genes constitutively during in vitro culture through a gene conversion process that is triggered by AID-dependent cytidine deamination (Buerstedde et al, 1990; Arakawa et al, 2002; Harris et al, 2002). The DT40X2U derivative cell line carries targeted inactivation of the xrcc2 locus causing it instead to perform IgV gene somatic hypermutation at high frequency (Sale et al, 2001). DT40X2U cells also express a transfected Ugi cassette (which encodes a protein inhibitor of UNG; Friedberg et al, 1975; Wang and Mosbaugh, 1988) rendering them deficient in uracil excision and thereby causing the IgV gene mutations to be largely restricted to nucleotide transitions at C:G pairs, as the abasic site needed for the production of transversions is no longer generated (Di Noia and Neuberger, 2002).

Previous studies monitoring the ability of cell extracts to excise uracil from oligonucleotide substrates have revealed that, unlike the extracts from mouse and human cells, which have been examined where endogenous SMUG1 provides a backup for UNG activity (Nilsen et al, 2001; Kavli et al, 2005), DT40 cells exhibit little if any UNG-backup uracil excision activity (Di Noia and Neuberger, 2002, 2004). We have observed the same apparent lack of backup activity in other chicken cell lines and in chicken primary lymphoid tissue (not shown), suggesting that this lack of UNG-backup activity is not peculiar to DT40 but likely a characteristic of the species. Backup activity can, however, be provided by transfection of chicken DTX2U cells with a human SMUG1 expression cassette, with these transfectants exhibiting a ratio of SMUG1 to UNG activity similar to that observed in mammalian cell lines (Figure 1A).

SMUG1 from all species so far examined (but not UNG) is able to excise 5-hydroxy-me-U from DNA both in vitro and in vivo (Boorstein et al, 1992, 2001; Kavli et al, 2002; Ulbert et al, 2002). Consistent with their lack of SMUG1, DT40 cells lack 5-hydroxy-me-U-excising activity. However, introduction of the human SMUG1 expression vector confers 5-hydroxy-me-U excision activity, with the transfected SMUG1 being localised to the nucleus, as judged by use of an hSMUG1-GFP fusion gene (Figure 1B and C).

Although DT40X2U [hSMUG1] cells showed indistinguishable growth curves from the mock-transfected controls, we noticed that they consistently (in four independent transfectants) gave rise to a lower proportion of surface IgM-loss variants following clonal expansion (Figure 1D and Table I). Such loss variants are caused by mutation of the IgV genes, suggesting that hSMUG1 expression had reduced the mutation frequency. Sequencing the IgVλ genes amplified from sorted IgM-loss variants supported this interpretation. Thus, the hSMUG1 transfectants gave a higher proportion of unmutated sequences (despite comparable efficiencies of sorting for sIgM loss) and a lower mutation load than the DTX2U control clones (averaging 1.4 versus 2.6 mutations per mutated sequence; Figure 1E and F and Table I). It is also striking that the hSMUG1 transfectants exhibited a major shift in the mutation pattern. Although both hSMUG1 and control transfectants gave rise to mutations that were nearly exclusively at C:G pairs, 89% of these mutations in the controls were nucleotide transitions (consistent with previous work; Di Noia and Neuberger, 2002), but this dropped to only 45% in the DTX2U [hSMUG1] transfectants (Figure 1F and Table I).

The large increase in transversion mutations at C:G pairs provides good evidence that transfected hSMUG1 is able to excise the uracil produced by AID-mediated deamination at the IgVλ locus in DT40 cells. Nonetheless, the diminished mutation accumulation suggests that hSMUG1 not only acts to alter the pattern of IgV diversification but also to diminish the degree of mutation, possibly by stimulating repair.

SMUG1 overexpression partially restores isotype switching in msh2^−/−ung^−/− mice

The evidence from transfected DT40 cells that SMUG1 is able to access the AID-generated U:G lesion obviously led us to ask why UNG-deficient mouse B cells had exhibited a class-switching defect as judged by in vitro assays with neither endogenous mouse SMUG1 nor transgenically overexpressed SMUG1 apparently acting to compensate for the UNG deficiency (Rada et al, 2004).

A possible explanation relates to the sensitivity of the in vitro assay. We previously noted that, in UNG-deficient mice, there is an MSH2-dependent pathway of isotype switching, which is readily evident from analysis of serum immunoglobulin titres but is less apparent from in vitro switching assays (Rada et al, 2002, 2004). However, in order to use serum analysis as a test for the ability of overexpressed SMUG1 to potentiate switching, it was necessary to cross the SMUG1 transgene into an msh2^−/−ung^−/− double knockout background, as the high background of serum IgG in ung^−/− single knockouts would otherwise render the assay insensitive.

The results were dramatic. Serum ELISA as well as SDS–PAGE analysis of total serum immunoglobulin that had been purified by use of Protein L revealed that the hSMUG1 transgene triggered the production of considerable quantities of IgG1 and IgG3 in msh2^−/−ung^−/− double knockout mice (Figure 2A and B). Consistent with previous findings (Rada et al, 2004), very little IgG is detectable in the sera of non-transgenic msh2^−/−ung^−/− double knockout controls (with the somewhat higher backgrounds detected in younger mice being likely largely attributable to maternal immunoglobulin). However, in the presence of the SMUG1 transgene, greatly elevated IgG1 and IgG3 levels are detected, although these levels are still several fold below those present in control animals.

These findings led us to revisit the in vitro switching assay and ask whether it would be possible to detect UNG-independent switching by analysing the cultures after a longer period of culture with LPS+IL4. We found that by extending the period of in vitro culture from 5 to 8 days, significant switching could indeed be detected in the cultures from the ung^−/− single knockout mice (Figure 2C) as well as in the culture from the hSMUG1-transgenic msh2^−/−ung^−/− double knockout mice, but not from the non-transgenic msh2^−/−ung^−/− double knockout controls (Figure 2D). These results indicate that the in vivo switching (as judged by serum IgG) that can be achieved in UNG-deficient mice through either the MSH2-dependent backup pathway or through enforced SMUG1 expression correlates with in vitro switching (as judged by the production of sIgG1⁺ cells during culture of spleen cells with LPS+IL4). It is, however, notable that a relatively weak restoration of in vitro switching correlates with a more substantial restoration of serum IgG titres. Analogous observations have been made with other mutant mouse lines and, as discussed previously (Rada et al, 2004), may well reflect in vivo selection for switched cells.

SMUG1 is downregulated during B-cell activation in normal mice

The observation that the hSMUG1 transgene restores significant amounts of serum IgG to msh2^−/−ung^−/− mice clearly reveals a distinction between endogenous and transgenic SMUG1. A likely explanation seemed that endogenous SMUG1 is expressed at too low a level to potentiate switching. Indeed, we have previously found that SMUG1 mRNA is poorly expressed in mouse lymphoid tissues as compared to heart, skeletal muscle, kidney and liver (Rada et al, 2004). However, activated B cells do display detectable Ugi-resistant uracil excision activity (Rada et al, 2004). This backup activity would appear to be largely due to SMUG1, as a combination of Ugi (to inhibit UNG) and the PSM1 anti-SMUG1 neutralising antibody (Kavli et al, 2002) eliminated nearly all detectable uracil excision activity in B-cell extracts (Figure 3A). Interestingly, though, while UNG activity increased following activation of resting B cells with LPS+IL4 (consistent with the association between UNG and DNA replication), SMUG1 activity actually diminished (Figure 3B). Thus, as judged by these types of assay, the ratio of UNG:SMUG1 activity increased from 4:1 to >50:1 following B-cell activation (Figure 3C). We have not discriminated nuclear from mitochondrial UNG in our assays. Given that nuclear UNG is practically absent from resting cells (Otterlei et al, 1999; Caradonna and Muller-Weeks, 2001; Imai et al, 2003; Fischer et al, 2004), the actual increase in the ratio of nuclear UNG:SMUG1 during B-cell activation is likely to be even more pronounced.

SMUG1 overexpression reduces the extent of switching and mutation in normal mice

The results indicate that SMUG1 is poorly expressed in activated mouse B cells but can, if overexpressed, access the AID-generated U:G lesion to potentiate class switching although it does not fully restore switching. However, our results with DT40 cells had raised the possibility that while SMUG1 can assist antibody diversification (as witnessed by the increase in transversion mutations in UNG-deficient DT40), it might often access the AID-generated U:G lesion to channel it into a repair pathway rather than potentiate antibody diversification.

We therefore wondered whether overexpressed SMUG1 might act similarly in mouse B cells—while able to assist diversification, acting more frequently to potentiate conventional repair. Thus, in an animal that is unable to switch (owing to deficiency in both UNG and MSH2), one would be well placed to detect the (albeit imperfect) switch-potentiating effect of SMUG1 overexpression. However, in a UNG-proficient wild-type background where switching and mutation occur normally, one might expect the repair activity of overexpressed SMUG1 to become apparent and lead to a consequent reduction in antibody gene diversification. This is indeed what is observed. Following in vitro incubation with LPS+IL4, splenic B cells from hSMUG1-transgenic UNG-proficient animals consistently exhibited an approximate three-fold reduction in switching to IgG1 as compared to B cells from non-transgenic littermate controls (Figure 4A). The B cells from the transgenic mice were indistinguishable from those of their non-transgenic littermates with respect to surface markers (not shown) as well as in their proliferative and blasting response to LPS (Figure 4B and Supplementary Figure 3A–C).

SMUG1 overexpression affects the IgV mutation spectrum in msh2^−/−ung^−/− mice

Sequence analysis of the IgV_H 3′-intronic flank in germinal centre B cells from Peyer's patches also revealed that the presence of the SMUG1 transgene resulted in a substantial depression of mutation accumulation (Figure 4C). Thus, in all littermate pairs of UNG-proficient mice analysed, the hSMUG1-transgenic animal revealed a diminished mutation accumulation compared to its littermate control whether judged by the percentage of sequences that carried mutations (averaging 43% in SMUG1 transgenics versus 62% in controls) or by the mean number of mutations per mutated sequence (averaging 7.2 in hSMUG1 transgenics versus 12.7 in controls) (Table II). A similar effect of the hSMUG1 transgene on the depression of mutation accumulation was seen when analysed in both ung^−/− and msh2^−/−ung^−/− knockout backgrounds (Table II and Supplementary Figure 1A). Consistent with the lack of impairment of B-cell proliferation (Figure 4B), the depression in mutation accumulation did not correlate with any alteration in the size or appearance of germinal centres (Supplementary Figure 3D). The presence of the hSMUG1 transgene also had no detectable effect on the distribution of mutations along the target sequence (Supplementary Figure 1B) and, despite the lower mutational load, the major hotspots were still conserved, with a major proportion of the mutations conforming to the hypermutation consensuses (75% of G:C mutations at WRC; 48% of A:T mutations at WA in hSMUG1-trasngenic msh2^−/−ung^−/− mice).

Although these results are consistent with the proposal that overexpression of SMUG1 in normal mice leads to a depression of mutation accumulation by increasing faithful repair of the initiating U:G lesion, one would expect in the light of our previous results that SMUG1 overexpression might restore some transversion mutations at C:G pairs in msh2^−/−ung^−/− mice. This is the case. Indeed, not only is there a restoration of transversions at C:G pairs in [hSMUG1] msh2^−/−ung^−/− transgenics, there is also some restoration of mutations at A:T pairs (Table II). Furthermore, while somatic mutation is msh2^−/−ung^−/− mice is not accompanied by the occasional deletions/insertions seen in normal mice (presumably reflecting the lack of induction of error-prone repair pathways), SMUG1 overexpression leads to a restoration of such aberrant events as witnessed by the identification of six insertion/deletion mutations.

Thus, overexpression of SMUG1 in msh2^−/−ung^−/− mice, while causing an overall depression of mutation (presumably through increased faithful repair), also leads to an alteration in the mutation pattern. This likely reflects that the SMUG1-generated abasic site can (like the UNG-generated abasic site; Rada et al, 2002, 2004) act as a template for the production of transversions at C:G pairs as well as a lesion that triggers (albeit somewhat poorly) mutation creation at A:T pairs. Interestingly, the presence of the SMUG1 transgene in ung^−/− MSH2-proficient mice, while leading to the expected depression of mutation accumulation, gave only a mild increase in transversions at C:G pairs (Table II). We suspect that, in this background, MSH2 may compete with SMUG1 for the processing of the U:G lesion. In any case, the results provide clear evidence that overexpressed SMUG1 can access and excise the uracil from the immunoglobulin locus.

Mice and cell lines

Transgenic mice expressing human SMUG1 under the control of the mouse H2-K^k promoter and IgH intronic enhancer (which show a 25-fold increase over controls in SMUG1 enzymatic activity in splenocytes) as well as msh2^−/−ung^−/− double knockout mouse line have been previously described (Rada et al, 2004). The expression of hSMUG1 in lymphoid tissues produced no alterations in the abundance of B and T cells, as judged by flow cytometry with anti-CD45R(B220) and anti-CD3, nor in the relative proportions of CD4⁺ and CD8⁺ T cells (not shown).

The chicken B-cell line DT40CL18CL4 (an sIgM⁺ subclone of DT40CL18) is described by Sale et al (2001) as is its XRCC2-deficient derivative DT40X2. The Ugi-expressing transfectant of DT40X2 (designated DT40X2U here) is described by Di Noia and Neuberger (2002).

Expression of human SMUG1 in DT40

The open reading frame encoding human SMUG1 was PCR-amplified from IMAGE cDNA clone 726197 (MRC Geneservice, Cambridge, UK) using oligonucleotides TGGACATATGCCCCAGGCTTT and CCCCAAGGGCACTCATTTCAA, cloned into pCR-Blunt II TOPO (Invitrogen) and then subcloned as an EcoRI fragment into pIRES-Neo3 (BD Clontech). Chicken DT40 cells were transfected with the PvuI-linearised DNA of the resultant plasmid and G418-resistant clones were screened for Ugi-resistant uracil excision activity. Four clones showing such activity were selected and expanded.

For the subcellular localisation experiment, SMUG1 was PCR-amplified using the oligonucleotides GGAATTCTGGACATATGCCCCAGGCTT and CGGGATCCTTTCAACAGCAGTGGCAGCA, cloned into the EcoRI/BamHI sites of pEGFP-N3 (Clontech) and transfected into DT40. Cells from two GFP⁺ G418^R clones (that exhibited SMUG1 activity) were incubated on slides coated with poly-L-lysine, fixed with 4% paraformaldehyde and analysed by confocal microscopy using propidium iodide to stain DNA.

Uracil excision assays

Uracil excision activity in whole cell extracts of DT40 and mouse B cells was measured essentially as described (Di Noia and Neuberger, 2002) except that the oligonucleotide CATAAAGTG(HOmU)AAAGCCTGG-FITC with a single 5′-hydroxymethyl deoxyuridine (HOmU) was annealed to CCAGGCTTTACACTTTATG or CCAGGCTTTGCACTTTATG and used as a substrate for the measurement of HOmU excision activity. Where indicated, extracts were preincubated on ice for 20 min with recombinant Ugi (NE Biolabs) and/or the PSM1 anti-SMUG1 antibody (a kind gift from Professor G Slupphaug, Institute of Cancer Research and Molecular Biology, Trondheim, Norway). UDG activity in resting versus activated mouse B cells was measured in extracts from normal mice splenocytes fractionated by CD43⁺ depletion using an AutoMacs device and anti-CD43 magnetic beads (Miltenyi Biotec). CD43-negative cells were >93% B cells (CD45R(B220)⁺) and >90% were arrested in G1 phase of the cell cycle as judged by propidium iodide staining. A fraction of these cells was stimulated to proliferate with LPS+IL4 for 4 days before processing for activity measurements, as described by Rada et al (2004).

Analysis of class switching

Serum immunoglobulin isotypes were quantified by ELISA (Rada et al, 2002) and immunoglobulin profiles were evaluated by Protein L-agarose precipitation followed by SDS–PAGE and staining with Coomassie blue (Rada et al, 2004). In vitro switch recombination to IgG1 was evaluated in samples of purified B cells that had been cultured for 4 days with LPS+IL4 as previously described (Rada et al, 2004). Cell blasting and proliferation following incubation with LPS was assayed by flow cytometric analysis of cell scatter and by monitoring incorporation of [³H]thymidine as previously described (Rada et al, 2002). The colorimetric assay for cell metabolic activity was carried out by measuring reduction of MTS tetrazolium CellTiter96 substrate (Promega) according to the manufacturer's instructions.

Analysis of mutation

The IgVλ region from DT40 cell populations was amplified by PCR and sequenced as described previously (Di Noia and Neuberger, 2002). Sequence determination of the J_H4-proximal intronic region from sorted germinal centre B cells from Peyer's patches is described by Jolly et al (1997).