Mutants with Temperature-Sensitive Defects in the Escherichia coli Mismatch Repair System: Sensitivity to Mispairs Generated In Vivo

Abstract

We have used direct selections to generate large numbers of mutants of Escherichia coli defective in the mismatch repair system and have screened these to identify mutants with temperature-sensitive defects. We detected and sequenced mutations that give rise to temperature-sensitive MutS, MutL, and MutH proteins. One mutation, mutS60, results in almost normal levels of spontaneous mutations at 37°C but above this temperature gives rise to higher and higher levels of mutations, reaching the level of null mutations in mutS at 43°C. However, at 37°C the MutS60 protein can be much more easily titrated by mispairs than the wild-type MutS, as evidenced by the impaired ability to block homeologous recombination in interspecies crosses and the increased levels of mutations from weak mutator alleles of mutD (dnaQ), mutC, and ndk. Strains with mutS60 can detect mispairs generated during replication that lead to mutation with much greater sensitivity than wild-type strains. The findings with ndk, lacking nucleotide diphosphate kinase, are striking. An ndk mutS60 strain yields four to five times the level of mutations seen in a full knockout of mutS. These results pose the question of whether similar altered Msh2 proteins result from presumed polymorphisms detected in tumor lines. The role of allele interactions in human disease susceptibility is discussed.

The repair of DNA replication errors is crucial for the avoidance of heritable mutations. The postreplication mismatch repair (MMR) system, characterized in bacteria, yeasts, and humans, plays a central role in the repair of replication errors (see reviews in references 25, 37, and 38). Cells lacking this system are mutators, with high rates of certain base substitutions and of frameshifts at repeat-tract sequences (28, 37, 38, 46, 51). In humans, the loss of the MMR system can lead to certain types of cancer. Lynch syndrome results in an increased susceptibility to colon (human nonpolyposis colon cancer [HNPCC]) and ovarian cancer, due to the inheritance of one defective copy of one of the genes involved in MMR (16, 27, 31, 42, 53). When a somatic cell loses or suffers inactivation of the other copy, a mutator cell results, accelerating the accumulation of the mutations needed to result in a tumor cell line. While 2 to 7% of all colorectal cancers result from HNPCC (31), as many as 15% of sporadic colon cancer lines are mutators with defects in expression or activity of the MMR system (1, 23, 52). The MMR system also plays a role in limiting recombination between related but divergent DNAs. For instance, homeologous recombination, such as occurs in interspecies crosses, is greatly enhanced in strains lacking MutS, since MutS binds to the frequent mismatches and limits the size of the heteroduplex DNA that is formed (44).

The biochemistry of MMR has been the subject of extensive study (25, 37, 38). In Escherichia coli, the A residues at GATC sequences are methylated at the 6 position by DNA adenine methylase. Immediately after replication, the new strand is unmethylated. Mispairs are recognized by the MutS protein, which then recruits the MutH and MutL functions. MutH cuts the hemimethylated DNA, on the unmethylated strand, and ultimately the mismatched base is excised, exonuclease action removes additional bases, and the gap is filled in and ligated. In humans, homologs of MutS and MutL (Msh2 and Mlh, respectively) function together with other proteins (G-T-binding protein [GTB]) to affect a similar repair (25). Recently, Yang and coworkers have elucidated the three-dimensional structure of the MutH and MutL proteins from E. coli and the MutS protein from Thermus aquaticus bound to a heteroduplex with an unpaired base (4-6, 41). Also the structure of the E. coli MutS protein bound to a mispair has been determined (26).

It would be a great advantage for mutator studies to have a set of conditional mutants defective in MMR, and particularly in MutS. Although several temperature-sensitive mutants with defects in MutH have been described in E. coli (19), there are no temperature-sensitive mutants with defects in MutS or MutL in E. coli. Alani and coworkers (3) have recently described mutations in the MLH1 gene in Saccharomyces cerevisiae that result in a temperature-sensitive Mlh protein, a homolog of the E. coli MutL protein. We developed a direct plate selection for MMR-deficient mutators (36), based on the principle that the mutator subpopulation increases in a population with each successive selection (32). We have used this selection to screen large numbers of mutators for those with temperature-sensitive defects. Here we describe mutations in mutS, mutL, and mutH that result in temperature-sensitive mutators. We report the sequence change resulting from each mutation and examine some of the characteristics of cells carrying these mutations. We find that strains with a temperature-sensitive mutS allele allow more frequent homeologous recombination at temperatures where mutations occur at relatively low frequencies and show that the altered MutS protein is titrated by the mispairs encountered in an interspecies mating. The altered MutS protein studied here becomes a sensitive biosensor for mispairs generated by any of a number of pathways, including weak mutator alleles of mutD (dnaQ), mutC, and also ndk. The effects of ndk mutations, which result in loss of nucleotide diphosphate kinase and altered nucleotide triphosphate pools, are dramatic on strains with mutS60. We consider these findings with respect to the concept of allele interactions and disease susceptibility.

MATERIALS AND METHODS

Bacterial strains and strain construction.

Table 1 lists the strains used in this work. AY102 was constructed by transduction of PA102 with P1vir lysates grown on a strain carrying a mini-Tn10 inserted into mutS (J. H. Miller, P. Funchain, and A. Yeung, unpublished data). Strains EH102, EH1, EH2, and EH5 were constructed in several steps. First, AS18-29 derivatives carrying either mutS60 or mutS64 were transduced to Tet^r (tetracycline resistance) with lysates grown on CAG12173, which carries a Tn10 inserted into cysC. Those strains that still carried a temperature-sensitive mutator were used to prepare new P1vir lysates that were used to transduce either PA102 or CC107 to Tet^r. Tet^r Cys⁻ transductants that retained the temperature-sensitive mutator character were then transduced to Cys⁺ with P1vir lysates grown on P90C, and strains that had become Tet^s but remained temperature-sensitive mutators were kept. Strains EH3 and EH5 were transduced to mutL35 and the linked zje-2241::Tn10. Tet^r transductants were screened for the temperature-sensitive mutator character. In the case of EH5, P1vir lysates were prepared on candidate strains to verify the transfer to CC107 of the temperature-sensitive mutator character linked to Tet^r. EH4 was constructed by transduction of CC107 with P1vir lysates grown on a derivative of AS18-29 that carried both mutH126 and the linked recD1901::Tn10. AY307 was constructed by transducing CC107 with a P1vir lysate grown on a strain carrying mutC (33) and the linked uvrC::Tn10. Ninety-eight percent of the Tet^r transductants carry mutC. All transductions and methods used in strain construction were performed as described by Miller (34).

TABLE 1.

Bacterial strains used in this study

Strain	Genotype	Reference
CC107	ara Δ(gpt-lac)5/F′ 128 lacIZ proA+B+ (the lac region on F′ carries a lacI mutation and also a frameshift in the lacZ gene that reverts by the addition of a GC base pair to a monotonous run of 6 GC base pairs)	11
P90C	ara Δ(gpt-lac)5	34
AS18-29	CC107 metE? bglA	36
MS107	CC107 mutS::mini-Tn10tet	35
PA102	ara Δ(gpt-lac)5 supE gyrA metB	18
AY102	PA102 mutS::mini-Tn10tet	This work
EH102	PA102 mutS60	This work
CAG12173	MG1655 cysC95::Tn10	48
CAG18427	MG1655 zje-2241::Tn10	48
DPB267	MG1655 recD1901::Tn10	48
SA975	HfrK13 thrA49 leuBCD39 ara-7 (Salmonella serovar Typhimurium)	45
GS1	CC107/pBR329dnaQ	50
GSI129	CC107/pBR329dnaQ129	50
GS70	CC107/pBR329dnaQ70	50
AY207	CC107ndk::mini-Tn10tet	35
AY307	CC107mutC uvrC279::Tn10	This work
EH1	CC107 mutS60	This work
EH2	CC107 mutS64	This work
EH3	CC107 mutL35 zje-2241::Tn10	This work
EH4	CC107 mutH126 recD1901::Tn10	This work
EH5	CC107 mutS60 mutL35 zje-2241::Tn10	This work

Mutagenesis.

Ethyl methanesulfonate (EMS) and 2-aminopurine (2AP) mutagenesis was carried out as described by Miller (34). EMS was used at a dose of 30 and 60 min of exposure to 0.03 ml of EMS added to 2 ml of resuspended washed cells in minimal phosphate buffer, pH 7.0. Cells were diluted 1:10 and grown overnight in Luria-Bertani broth. 2AP was used at 700 μg/ml in Luria-Bertani broth.

Detection of mutators.

Strain AS18-29 was mutagenized as described above, and lactose minimal plates containing limiting amounts of glucose and methionine were used to select mutators as described previously (36). Mutators were identified by an increased frequency of Rif^r (rifampin-resistant) mutants. Mutation tests for determining Lac⁺ and Rif^r mutants were carried out as described in the work of Miller et al. (35). Mutation rates were determined by the method of Drake (14) and evaluated according to the work of Dixon and Massey (13). Mutators were mapped by using Tn10 inserts linked to either mutL, mutS, or mutH (Table 1). Transductants to Tet^r from lysates made on either CAG12173 (for linkage to mutS), CAG18427 (for linkage to mutL), or DPB267 (for linkage to mutH) were screened for loss of the mutator effect.

Conjugational matings.

Conjugational matings were carried out and scored as described by Funchain et al. (18).

RESULTS

Selection and screening for MMR-deficient mutants.

We used the direct selection that we previously described (36) to isolate mutators that lack the MMR system. Briefly, this selection relies on the principle that successive selections on a plate containing limiting nutrients enhance the mutators in the population. When the phenotypes being selected for are generated by mutations that occur frequently in MMR-deficient backgrounds (such as frameshifts at repeat-tract sequences), then most of the colonies surviving the selection are MMR⁻ mutators. Using either EMS or 2AP, we mutagenized strain AS18-29, which carries frameshifts in the metE and lacZ genes (36), and plated it on lactose minimal medium containing trace amounts of glucose and methionine. Colonies appearing after 3 to 4 days were picked and purified and tested by replica plating (see Materials and Methods) for increased frequency of Rif^r (rifampin-resistant) mutants within patches of cells at different temperatures. We screened approximately 500 mutants from several different mutagenesis experiments. At least 90% of the mutants detected on the selection plates after mutagenesis and outgrowth were mutators, based on their increased Rif^r relative to the starting control. Five to ten percent showed possible temperature effects, but on retesting in more detail, only 1 to 2% of the total mutants screened showed significant temperature-sensitive mutator activity. These were picked for further study.

Mapping the temperature-sensitive mutations.

We used P1 cotransduction with Tn10 inserts linked to the MMR loci (mutS, mutH, and mutL) to map the temperature-sensitive mutations (see Materials and Methods). To confirm the mapping assignment, we transferred each mutation to an unmutagenized strain by P1 cotransduction with the respective linked Tn10. Subsequent DNA sequencing (see below) verified the assignments. We confined our studies to one or two examples of temperature-sensitive mutations located in each of the three loci mutS, mutL, and mutH.

Amino acid changes resulting in temperature-sensitive mutants.

Table 2 lists the changes that we detected in mutH, mutL, or mutS in the alleles mutH126, mutL35, mutS60, and mutS64. Note that mutS60 and mutS64 result from different changes in the same codon that specifies residue 134 in the MutS protein. The change of alanine to valine results in mutS60, and the change of alanine to threonine results in mutS64. The change in the MutH protein (Table 2) is different from the two previously reported (19).

TABLE 2.

Amino acid changes resulting in temperature-sensitive MMR

Mutation	Gene	Amino acid change	Base pair change
mutS60	mutS	Ala 134→Val	GC→AT
mutS64	mutS	Ala 134→Thr	GC→AT
mutL35	mutL	Gly 62→Ser	GC→AT
mutH126	mutH	Glu 149→Lys	GC→AT

Temperature-sensitive mutagenesis.

Table 3 shows the mutation rates for base substitutions leading to Rif^r in strains carrying each of the temperature-sensitive alleles. We also examined frameshifts in some cases, using the strain CC107 (11) to detect Lac⁺ revertants resulting from the addition of a G to a run of six G's. The mutS alleles mutS60 (Table 3) and mutS64 (data not shown) both exhibit their main effects between 37 and 42°C. On the other hand, the mutL allele that we studied causes its main temperature effect between 34.5 and 37°C. A double mutant with defects in both mutS and mutL is also temperature sensitive between 34.5 and 37°C. The mutation rates for base substitution mutations and frameshifts follow virtually identical curves, as shown in Fig. 1, which plots the data for CC107 carrying the temperature-sensitive mutS60, with additional temperature points, for base substitutions.

TABLE 3.

rpoB and lacZ rates of mutations at different temperatures^a

Strain	Mutation rate (μ) per replication (10−8) at temp (°C)
	Rifr				Lac+
	30	34.5	37	43	34.5	37	43
WTb (CC107)		1.3	1.5	2.6	3.8	8.9	6.7
mutS::Tn10			120	93		3,600
mutS60		1.3	4.3	144	1.5	39	4,570
mutH126		3.9	8.7	46
mutL35	1.3	1.0	27	37
mutS60 mutL35	0.8	2.3	67	52

^aThe rpoB and lacZ mutation rates (μ) per replication were determined by first dividing the median number of Rif^r or Lac⁺ mutants by the average number of cells in a series of cultures to obtain the mutation frequencies and then calculating the mutation rate (μ) shown in this table from these values by the method of Drake (14).

^bWT, wild type.

FIG. 1.

Rates of base substitution and frameshift mutations at different temperatures in strain EH1, carrying mutS60. (A) Rates of rpoB mutations leading to Rif^r (base substitutions). (B) Rates of lacZ mutations leading to Lac⁺ (frameshifts).

Blocking recombination after interspecies mating.

Recombination between related but divergent DNAs (termed homeologous recombination) is greatly reduced compared to recombination between identical DNA molecules (homologous recombination). Strains lacking MutS or MutL have greatly increased recombination after interspecies matings between Salmonella species and E. coli, presumably because MMR proteins recognize the mismatches and limit the heteroduplex region (44). Table 4 (rows 1 and 3) shows this effect in crosses with a Salmonella Hfr that donates metB early and E. coli recipients that are either wild type or mutS. In this case there is a 670-fold increase in a mutS strain in the level of Met⁺ recombinants detected. When we look at crosses with a recipient carrying mutS60 (row 2), we note that the level of recombinants, though not as high as that for a complete mutS knockout, is still close to 80-fold higher than the wild type. Yet, as shown in Table 3, at 37°C there is only a very low level of mutations in strains with mutS60. The right-hand portion of Table 4 shows the ratios to wild type for mutations after overnight growth (base substitutions, followed by frameshifts) and for recombinants in the interspecies cross in the mutS knockout and in strains with mutS60. For the mutS knockout, the ratios are high for both mutations (80 to 405, depending on the assay) and recombinants (670). However, for strains with mutS60, the ratios are low for mutations (2.9 and 4.4) but high for recombinants (79).

TABLE 4.

Efficiency of recombination (Met⁺/Hfr) in crosses between Salmonella Hfr and E. coli F⁻ srains at 37°C compared with mutation rates^a

Cross and/or strain (recipient genotype)	No. of Met+ recombi- nants (106)	Ratio, Met+/Met+ WT	Ratio, mutation rate (μ)/WTb
PA102 (WT) × SA975	2.9	1.0	1.0, 1.0
EH102 (mutS60) × SA975	230	79	2.9, 4.4
AY102 (mutS::Tn10) × SA975	1,950	670	80, 405

^aMeasured after crosses with SA975 as Hfr. WT, wild type.

^bThe first ratio in each row is the rpoB (Rif^r) rate, and the second is the lacZ (lac⁺) rate.

Reduced capacity to deal with mispairs.

How can we explain the fact that strains with mutS60 have little or no mutator activity at intermediate temperatures such as 37°C but have lost the ability to prevent homeologous recombination after interspecies crosses? In both cases the MMR system recognizes base mispairs. One possibility lies in the fact that the number of mispairs encountered during an interspecies mating is much greater than those generated as replication errors during a normal cell cycle. With 18% sequence divergence, there would be close to 9,000 mispairs per min of transferred DNA. Although not every mispair within a region is a binding target for MutS, still, with 7 to 15 min of DNA transferred in each cell in the cross depicted in Table 3, the number of mispairs may be just near the point of saturating the MMR capacity in a wild-type strain. In strains with a lower population of functional MutS molecules, saturation may occur. At intermediate temperatures, the active fraction of the temperature-sensitive MutS molecules resulting from the mutS60 allele should be reduced compared to wild type and susceptible to titration by the large number of mispairs encountered during an interspecies mating.

Mispairs resulting from dnaQ alleles.

To test whether the mutS60 allele renders the MMR system more susceptible to saturation by mispairs at 37°C, we looked at several pathways that generate increased mispairs. First, we employed different engineered mutations in the dnaQ gene. Mutations in the dnaQ gene can affect the ɛ subunit of DNA polymerase III, which provides a proofreading function during DNA synthesis (15, 47). Cells with an impaired function of ɛ are mutators (mutD). Strong mutD mutators have been described that are dominant in the presence of a wild-type dnaQ gene (10, 12). We constructed a set of dnaQ mutations by changing, in series, each histidine or aspartic acid codon in the gene to a glycine codon (50). The set of 23 altered genes, each on a plasmid, when transferred to a wild-type strain produced four strong mutD mutators and several weak mutators (50). Here we have used two of these weak mutators to ask whether a plasmid carrying a mutD-dnaQ allele could provoke more mutations in a strain carrying the mutS60 allele at 37°C than in a wild-type strain. Table 5 (rows 1 to 3) shows the mutation rates for base substitutions leading to Rif^r found in either wild-type or mutS60 strains carrying plasmids with two different engineered mutD alleles. It can be seen that in each case the mutation frequency is higher in the mutS60 strain than in the corresponding wild-type strain. The effects range from 4- to 150-fold.

TABLE 5.

Effect of different mutations on rpoB rates of mutations in wild-type (WT) and mutS60 strains at 37°C^a

Mutation	μ (10−8)
	WT	mutS60
None (WT)	1.5	4.3
mutD (D129G)	2.3	360
mutD (D70G)	2.5	10
mutC	9.1	44
ndk	26	420

^aSee footnote a, Table 3.

Mispairs resulting from mutator tRNAs.

We examined the effect of mutS60 on mutC, a mutator resulting from an altered tRNA that causes mistranslation, in this case resulting in the insertion of glycine in place of aspartic acid approximately 1% of the time. This may cause mutator effects either by generating small levels of an altered polymerase proofreading subunit that allows more mispairs (33, 49) or by a different mechanism affecting polymerase III (22). Table 5 (row 4) shows that the presence of mutS60 increases the mutation rate in strains with mutC.

Mispairs resulting from the loss of nucleotide diphosphate kinase.

We also tested the effects of an insertion in the ndk locus on mutation levels in wild-type and mutS60 strains. The ndk gene encodes nucleotide diphosphate kinase, which is involved in maintaining proper pool sizes of each of the nucleotide triphosphates. Mutations in ndk have been shown to increase base substitutions (30, 35). Table 5 (row 5) shows that the level of base substitution mutations at 37°C leading to Rif^r mutants is significantly increased (16-fold) in ndk strains that carry mutS60 relative to strains with a normal mutS. Clearly, many of the mispairs generated in ndk strains are corrected by the MMR system, and this system is partly saturated in ndk mutS60 strains.

DISCUSSION

Genetic susceptibilities to human disease such as cancer have been well studied in cases where an inherited mutation results in complete loss of function, the high penetrance leading to identifiable family histories of susceptibilities. Examples are xeroderma pigmentosum- and UV-induced skin cancer (9), HNPCC and colon cancer (16, 27, 42), BRCA1 and BRCA2 and breast cancer (reference 21 and references therein), and more recently, increased colon cancer susceptibilities in families with inherited defects in MYH (2, 24), the human mutY gene. However, these inherited susceptibilities represent only a small percentage of cancer cases, perhaps nearly 10% in the case of breast and colon cancer, but not more than 5% overall (reference 40 and references therein). Although some of the remaining so-called sporadic cancers are caused by totally defective genes (1, 23, 52), a number of studies suggest that many of the remaining sporadic cancers may be the result of increased susceptibilities caused by mutations that only slightly impair repair or other functions. These alleles create a smaller increase in susceptibility than those mentioned above, but they are much more widespread in the population. Also, in combination with other mutations that slightly impair different functions, the resulting susceptibility could be significantly increased. Thus, Mohrenweiser et al. examined 37 different repair genes in 36 to 164 individuals and detected 127 amino acid substitution variants, many of which appear likely to affect function (40). In most cases more than one repair gene showed variation. Other studies have revealed that 10% of the population has reduced capacity (60 to 75% of normal) to repair DNA damage, as judged by examining lymphocytes in vitro (17, 20), and also suggest a relationship between reduced repair capacity and increased cancer susceptibility (7, 8, 39, 54, 55). More complex are the interactions between multiple alleles (43). In the work reported here we have developed a bacterial model system for looking at allele interactions in inactivating DNA repair. This system is based on a mutation in the E. coli mutS gene that reduces MMR capacity at high temperatures. At 37°C, cells have normal mutation rates, since the reduced activity is still sufficient to repair replication errors under normal conditions. However, in the presence of other mutations that cause additional mispairs, MMR is easily titrated and large increases in mutagenesis occur. The apparently silent allele can trigger large effects in the presence of certain conditions or genetic backgrounds.

In order to find temperature-sensitive mutations affecting MMR, we took advantage of specific bacterial genetic selections that we designed (36) to generate a large number of mutants deficient in the E. coli MMR system. Subsequent screening of these mutations allowed the identification of several temperature-sensitive mutants with defects in either mutH, mutS, or mutL. DNA sequence analysis pinpointed the exact base and amino acid change resulting from each mutation examined. Different temperature-sensitive mutations in the mutH gene of E. coli have been described elsewhere (19) as have those in the MLH1 (mutL homolog) gene in yeast (3). Such mutations in mutS have not been described in any organism, nor have temperature-sensitive mutations in mutL in E. coli been reported. Strains carrying these mutations have higher levels of both frameshift and base substitution mutations (Table 3). We focused on a mutation in mutS, mutS60, that has almost wild-type levels of spontaneous mutagenesis at 37°C but at temperatures above 42°C gives levels of mutagenesis similar to that found in complete knockouts of mutS.

In addition to repairing replication errors, the MMR system also acts as a barrier to recombination with divergent chromosomes (homeologous recombination [44]). However, at 37°C, E. coli recipient strains with the mutS60 allele have significantly higher levels of homeologous recombination in interspecies crosses with donor Salmonella enterica serovar Typhimurium Hfr strains than wild-type E. coli recipients (Table 4). A plausible explanation for these results is that, in strains with mutS60, the active MutS protein is nearly titrated out by the replication mispairs, and the increased load of mispairs generated during an interspecies cross is enough to overload the system and partially break down the barrier to homeologous recombination. We therefore looked for synergistic effects between mutS60 and mutations that presumably result in low levels of mispairs, such as weak mutD (dnaQ) alleles, mutC (encoding mutator tRNAs), and ndk (encoding nucleotide diphosphate kinase). In all cases the level of spontaneous mutagenesis was significantly higher in mutS60 than in wild-type backgrounds, and the effect seen with ndk mutS60 double mutants is dramatic. E. coli strains defective in nucleotide diphosphate kinase have altered deoxynucleoside triphosphate pools and have been shown to have modest increases in base substitutions leading to the Nal^r or Rif^r phenotype (30, 35) and in frameshifts (35). However, ndk strains have increased levels of frameshift mutations and ndk mutS::mini-Tn10 double mutants have extraordinary levels of both base substitutions and frameshifts (35). These levels approach 10 times those found in strains with complete knockouts of mutS (mutS::mini-Tn10). The levels seen in the double mutant ndk mutS60 are also extremely high (Table 5), exceeding those resulting from a complete knockout of mutS in a strain with a normal NDK (Table 3). Here, the mutS60 allele yields an apparently normal MutS protein at 37°C, yet in the presence of increased mispaired bases the MutS60 protein is titrated out and mutagenesis levels rise dramatically. The MutS60 protein acts as an enhanced biosensor at 37°C. Thus, in a wild-type background, ndk strains have only modest mutation rates, but in the presence of mutS60, ndk stimulates very high mutation rates.

How many alleles similar to mutS60 are there in the human population that appear normal in one background or set of conditions but that can generate high mutation levels either in combination with certain mutations in other pathways, such as ndk, or in the presence of certain heterozygous loci, or perhaps transiently in the presence of temporary perturbations in the cellular milieu, such as deviations in pool size? There are known susceptibilities to disease that depend on a synergistic interaction of two mutations. A recent example is the work by Levy-Lahad et al. (29) that reports a rad51 mutation that interacts with a BRCA2 defect to increase the risk of breast cancer in BRCA2-defective individuals but not in BRCA1-defective individuals. The challenge for the next phase of work in this field is to identify the full range of synergies between mutations in different genes and the role that this plays in human disease susceptibility.