Somatic hypermutation of immunoglobulin variable genes, which increases antibody diversity, is initiated by the activation‐induced cytosine deaminase (AID) protein. The current DNA‐deamination model posits that AID deaminates cytosine to uracil in DNA, and that mutations are generated by DNA polymerases during replication or repair of the uracil residue. Mutations could arise as follows: by DNA replicating past the uracil; by removing the uracil with a uracil glycosylase and replicating past the resulting abasic site with a low‐fidelity polymerase; or by repairing the uracil and synthesizing a DNA‐repair patch downstream using a low‐fidelity polymerase. In this review, we summarize the biochemical properties of specialized DNA polymerases in mammalian cells and discuss their participation in the mechanisms of hypermutation. Many recent studies have examined mice deficient in the genes that encode various DNA polymerases, and have shown that DNA polymerase H (POLH) contributes to hypermutation, whereas POLI, POLK and several other enzymes do not have major roles. The low‐fidelity enzyme POLQ has been proposed as another candidate polymerase because it can efficiently bypass abasic sites and recent evidence indicates that it might participate in hypermutation.
Antibody variable genes are initially formed by assembling immunoglobulin variable (V), diversity (D) and joining (J) gene segments by non‐homologous end joining. To produce antibodies with higher specificity, a second cycle of diversification takes place by somatic hypermutation (SHM). During this process, base substitutions are introduced into a region of ∼1 kb surrounding the antibody‐coding sequence (Lebecque & Gearhart, 1990). The process of SHM seems to increase the antibody repertoire enough to limit the frequency of environmentally acquired infections and thereby increases the individual fitness of many vertebrate species, including humans and mice. In one of his lectures on this subject, Myron Goodman proposed SHM as the answer to the question “Why don't you die when I sneeze?” Indeed, it might be argued that the demise of the invaders in the War of the Worlds could be attributed to their lack of an adequate mechanism for diversifying their antibodies.
The base‐substitution mechanism is initiated by the activation‐induced cytidine deaminase (AID) protein, which is encoded by the AICDA gene (Honjo et al, 2005; Neuberger et al, 2005) and is expressed only in B lymphocytes. AID deaminates cytosine to uracil in DNA and has a strong preference for WRC (W is A or T and R is A or G) hotspot motifs in single‐stranded DNA (ssDNA; Pham et al, 2003). Such ssDNA might arise transiently during gene transcription and SHM is, in fact, dependent on the active transcription of antibody genes to create the target for cytosine deamination by AID (Barreto et al, 2005).
Subsequent steps of SHM have long been postulated to involve a low‐fidelity DNA polymerase, and the history of this concept has been reviewed (Gearhart & Wood, 2001). As it is now known that SHM is initiated by AID, it is possible to formulate models involving not only DNA polymerase errors, but also mutations produced during replication or repair of the lesion. U:G mispairs are formed in immunoglobulin DNA by the action of AID, which acts preferentially on a cytosine in the non‐transcribed strand (Martomo et al, 2005a). As first envisaged by Neuberger and colleagues, there are several possible pathways to produce mutations from uracil (Fig 1; Neuberger et al, 2005). One pathway (or ‘phase’ as it is sometimes called) mainly influences events at C:G base pairs, and a second pathway affects substitutions at A:T pairs. After cytosine deamination by AID to generate uracil, copying of this base by a high‐fidelity polymerase during chromosomal replication will cause a C:G to T:A transition (C:G mutagenesis type I). In human cells, uracil in DNA is normally removed after incision by a uracil‐DNA glycosylase and repaired by base‐excision repair (Fig 1). A uracil‐DNA glycosylase, such as the enzyme UNG, eliminates the uracil base to produce an apyrimidinic (AP) site. Such sites can be cleaved by several mechanisms, either by an AP endonuclease in duplex DNA or by an AP lyase activity in ssDNA. During error‐free base‐excision repair in cells, the gap of one or a few nucleotides is usually filled by DNA polymerase β. In SHM, error‐prone base‐excision repair can occur if other low‐fidelity DNA polymerases are used in this step. Alternatively, mutations at C:G base pairs might arise (C:G mutagenesis type II; Fig 1) if a bypass DNA polymerase incorporates a base opposite an AP site before repair is initiated.
By contrast, mutations of A:T base pairs seem to be generated by a genetically distinct process. This pathway involves the MSH2–MSH6 mismatch‐recognition complex, which can bind to a U:G mismatch (Wilson et al, 2005). A current hypothesis is that the low‐fidelity DNA polymerase H (POLH; also known as pol η) generates mutations at A:T pairs during a mismatch‐repair event that is initiated at a U:G base pair, but involves a larger gap‐filling reaction that might use exonuclease I (Bardwell et al, 2004; Wilson et al, 2005). This process could also generate mutations at C:G base pairs.
Bypass polymerases and translesion synthesis
Damage to DNA alters its structure and so DNA polymerases responsible for chromosome replication often stall at sites of lesions. Fortunately, in addition to DNA‐repair pathways, cells can tolerate damage by bypassing it. Specialized bypass DNA polymerases can insert a nucleotide opposite a lesion and, in some cases, continue DNA synthesis a few nucleotides further. In mammalian cells, six enzymes—pol ζ, POLH, POLI (pol ι), POLK (pol κ), REV1 and POLQ (pol θ)—have been shown to bypass DNA damage. POLH, POLI, POLK and REV1 belong to the Y‐family polymerase group and share some protein sequence similarity. The catalytic subunit of pol ζ, REV3L, belongs to the B‐family of polymerases, which includes DNA polymerases α, δ and ϵ. POLQ belongs to the A‐family of enzymes, which are similar to Escherichia coli DNA pol I. Translesion DNA‐synthesis reactions consist of at least two steps (Fig 2A; Wang, 2001; Prakash et al, 2005): the first is insertion of a base opposite the lesion by a bypass polymerase and the second is extension from this inserted base by another polymerase. In yeast, pol ζ has some ability to extend past misincorporations (Prakash et al, 2005). There are human counterparts of the pol ζ subunits (Lin et al, 1999a; Gibbs et al, 2000), so the enzyme might also extend mispairs in higher eukaryotes. After extension, a switch back to a high‐fidelity replicative DNA polymerase (for example, pol δ) must take place so that normal DNA synthesis can continue.
Bypass DNA polymerases and SHM
POLH. There is good genetic evidence that mammalian POLH (Table 1) is involved in SHM. POLH efficiently bypasses cyclobutane pyrimidine dimers induced by ultraviolet radiation, and its absence in humans causes the variant form of xeroderma pigmentosum (XP‐V; Johnson et al, 1999; Masutani et al, 1999). POLH is a low‐fidelity enzyme that generates replication errors at an average frequency of ∼3 × 10−2 mutations per base pair (Table 1). In XP‐V patients, the frequency of SHM is normal but the mutation pattern changes; the frequency of mutations at A:T bases is decreased and there is a concomitant increase in C:G base mutations (Zeng et al, 2001; Faili et al, 2004; Zeng et al, 2004; Mayorov et al, 2005). In vitro experiments show that POLH often makes errors at A:T base pairs, particularly at WA sequences (W is A or T), which are potential hotspots for SHM in immunoglobulin genes (Rogozin et al, 2001). In the mouse, SHM hotspots in vivo and in vitro are strongly correlated (Pavlov et al, 2002). Recent analyses of Polh‐deficient mice have established that POLH is involved in A:T mutagenesis during SHM (Delbos et al, 2005; Martomo et al, 2005b). POLH physically interacts with the MSH2–MSH6 complex, which is also implicated in mutagenesis at A:T base pairs (Wilson et al, 2005). However, POLH might not be involved in C:G mutagenesis type II, as mice deficient in this polymerase still have a high frequency of mutations of C:G. POLH does not efficiently insert a base opposite AP sites, and extension from a primer opposite an AP site is limited (Table 1; Wang, 2001; Kusumoto et al, 2002; Prakash et al, 2005).
Pol ζ and REV3L. The role of pol ζ in SHM has also been considered. In the yeast Saccharomyces cerevisiae, the enzyme consists of the catalytic subunit Rev3 (a 173‐kDa protein) and the accessory subunit Rev7. In mammalian cells, a 350‐kDa protein homologous to Rev3 is known as REV3L (Murakumo et al, 2001). REV3L is expressed in germinal centre B cells, which suggests that it might be involved in SHM (Winter et al, 2000; Poltoratsky et al, 2001; Zan et al, 2001; Zeng et al, 2001). When Rev3L expression was inhibited using antisense oligonucleotides in the mutating CL‐01 cell line, there was a moderate reduction of both A:T and G:C mutations in an immunoglobulin variable gene (Zan et al, 2001). A similar effect was reported in transgenic mice expressing Rev3l antisense RNA (Diaz et al, 2001). The modest decrease in SHM in these mice is difficult to explore further because the complete disruption of Rev3l causes lethality during embryonic development (Bemark et al, 2000; Esposito et al, 2000a; Wittschieben et al, 2000; O‐Wang et al, 2002; Van Sloun et al, 2002). Presumably, the antisense inhibition is incomplete and enough pol ζ remains in the mice to allow viability. Biochemical studies using yeast pol ζ suggest that it might extend the mismatched primers generated by POLH in A:T mutagenesis and complete the gap‐filling reaction, or extend a primer opposite AP sites generated by bypass in C:G mutagenesis type II (Wang, 2001; Prakash et al, 2005).
POLI and other Y‐family and X‐family DNA polymerases. POLI is a paralogue of POLH that also has low fidelity (Table 1). POLI can efficiently insert a base opposite an AP site (Wang, 2001; Vaisman et al, 2002; Prakash et al, 2005). When the POLI gene was knocked out in the Burkitt's lymphoma cell line BL2, most SHM (which in this cell line is largely at C:G base pairs) was dependent on POLI (Faili et al, 2002). However, mice from the 129 strain, which have an inactivating nonsense mutation in Poli, had normal SHM (McDonald et al, 2003). Addition of the 129 strain Poli mutation to a Polh‐ or Polk‐knockout mouse did not change the phenotype (Delbos et al, 2005; Shimizu et al, 2005). Currently, the reason for the different results in Poli‐deficient 129 mice and the POLI‐knockout cell line is not understood; this issue might be clarified by experiments with mice that have a complete or inactivating knockout of the Poli gene in another mouse strain.
Speculation that POLK might participate in SHM arose because mutation‐spectrum analysis indicates that POLK preferentially mutates the SHM hotspot sequence RGYW (R is a purine, G is a guanine, Y is a pyrimidine and W is A or T; Rogozin et al, 2001). However, disruption of Polk has no significant effects on SHM in the mouse (Schenten et al, 2002; Shimizu et al, 2003). Purified POLK can bypass an AP site by a primer/template‐slippage mechanism (Ohashi et al, 2000a), but such frameshift mutations are uncommon events in SHM.
Mammalian cells also have several DNA polymerases in the X family: POLB (pol β), POLL (pol λ) and POLM (pol μ). POLB functions in error‐free base‐excision repair of DNA, whereas POLL and POLM have roles in non‐homologous end‐joining of DNA during double‐strand‐break repair (Bertocci et al, 2003; Ma et al, 2004; Nick McElhinny & Ramsden, 2004). No alterations in the SHM process have been found in mice with knockouts of Polb (Esposito et al, 2000b), Poll (Bertocci et al, 2002) or Polm (Bertocci et al, 2002; Lucas et al, 2005). However, POLL and POLM do have a role in generating antibody diversity during VDJ joining (Ma et al, 2004; Nick McElhinny & Ramsden, 2004). Mice carrying a specific defect in the proofreading 3′ to 5′ exonuclease function of the replicative DNA polymerase δ (a B‐family polymerase) also lack defects in SHM (Longacre et al, 2003).
REV1. REV1 has deoxycytidine monophosphate (dCMP) transferase activity that can insert a cytosine opposite an AP site (Table 1). Experiments with budding yeast show that REV1 is involved in the bypass of AP sites and other lesions, but its exact role remains to be determined (Nelson et al, 2000; Prakash et al, 2005). The dCMP transferase activity is dispensable for its in vivo role in translesion synthesis opposite an AP site (Nelson et al, 2000). REV1 interacts with other Y‐family member polymerases as well as REV7, which is a subunit of pol ζ (Murakumo et al, 2001; Guo et al, 2003; Ohashi et al, 2004; Tissier et al, 2004), and might be a docking protein for translesion DNA‐synthesis proteins in SHM. Disruption of the REV1 gene in chicken DT40 B cells indicates that REV1 might be involved in SHM by producing C:G to G:C transversions (Simpson & Sale, 2003). No significant effects on SHM were found in a mouse lacking the BRCT domain of REV1, which is important for protein–protein interactions (Jansen et al, 2005). The SHM spectrum in hypermutating DT40 cells could be explained by the insertion of cytosine opposite abasic sites, but this explanation cannot account for the SHM spectrum in mice. Specific analysis of the role of the dCMP transferase activity in SHM needs to be performed using a dCMP transferase mutant Rev1 mouse or cell line.
POLQ. The POLH, POLI, POLK and REV1 enzymes belong to the Y family, and are characterized by their ability to manipulate highly distorted templates. However, POLQ is a low‐fidelity bypass polymerase that belongs to the A family (Table 1; Seki et al, 2003; Seki et al, 2004). A‐family polymerases are generally high‐fidelity enzymes that operate during DNA repair, recombination and replication, and do not normally engage in the bypass of lesions (Seki et al, 2004). POLQ is unusual in this respect, as it can efficiently insert an adenine opposite an AP site and subsequently extend the DNA (Fig 2B). So far, it is the only polymerase in human cells that can efficiently perform both insertion and extension reactions. POLQ is involved in the maintenance of genome integrity, as reticulocytes from mice deficient in this polymerase show abnormal nuclear division or increased chromosome breakage (Shima et al, 2003; Shima et al, 2004).
As POLQ is highly expressed in lymphoid tissues, including the spleen and germinal centres, the enzyme might be involved in SHM (Kawamura et al, 2004; Shima et al, 2004). Extension of a primer from a misincorporated adenine opposite an AP site would stably generate a C:G to T:A transition mutation, which is common in SHM. POLQ might also be a candidate for involvement in A:T mutagenesis, as a low‐fidelity DNA polymerase is proposed to be involved in a gap‐filling reaction in this process. Recent experiments with gene‐deficient mice have directly tested whether POLQ is involved in SHM. One series of experiments from Casali, Schimenti and colleagues used mice carrying a gene disruption that completely eliminates the POLQ enzyme (Zan et al, 2005). The overall frequency of SHM was reduced by 60–80%. The spectrum of mutations was moderately shifted towards more transitions at both A:T and C:G base pairs, but there was no overall change in the proportion of events at A:T and C:G positions. An independent study from the laboratory of O‐Wang (Masuda et al, 2005) used mice carrying an in‐frame deletion of part of the DNA polymerase domain, so that a catalytically inactive POLQ protein was still produced. In this mouse, the frequency of SHM was only reduced slightly, with normal levels of A:T mutations but fewer mutations of C:G base pairs. In follow‐up studies, O‐Wang and collaborators constructed a second Polq‐null mouse, which did not synthesize Polq mRNA or protein. In this mouse, a mild overall reduction in mutation frequency was seen with little shift in the mutation spectrum (J. O‐Wang, personal communication). The effects in all of these studies were relatively moderate and it is not certain whether POLQ has a major role in SHM.
As the mechanism of SHM emerges, it is proving to be a fascinating intersection of biochemistry associated with DNA repair, transcription and replication enzymes. However, there are many unanswered questions about the pathway. For example, once an AP site is generated in DNA through the sequential activities of AID and UNG, why does the active base‐excision‐repair pathway fail to repair the AP site in an error‐free manner? Moreover, what is the mutagenic pathway that involves the activities of MSH2 and MSH6? The latter does not seem to be normal mismatch repair, as it does not involve other mismatch‐repair proteins, such as MLH1 and PMS2. The MRE11–RAD50–NBS1 (MRN) nuclease complex also accelerates SHM, but the biochemical mechanism is unknown (Yabuki et al, 2005).
In summary, it seems that several DNA polymerases are involved in different stages of the hypermutation process. Further experiments using double‐gene disruptions (for example, Polq–Polh double‐mutant mice) should be informative in exploring which DNA polymerases can participate in the process and at which steps. It is also possible that specialized DNA polymerases might substitute for one another in SHM when one enzyme is completely deleted. One way to explore this subject might be to examine mice or cell lines in which DNA polymerase catalytic sites are specifically inactivated, without total deletion of the enzyme.
We thank G. Gan for assistance with editing, and members of our laboratories for comments on the manuscript. This work was supported by US National Institutes of Health grants CA098675 and CA101980 to R.D.W., by the NIH Intramural Research Program and by the University of Pittsburgh Cancer Institute.
- Copyright © 2005 European Molecular Biology Organization