Molecular Models That May Account for Nitrous Acid Mutagenesis in Organisms Containing Double-Stranded DNA

,


I NTROD UCTlO N
Spontaneous deaminations of C -t U occur at biologically significant frequencies and are mutagenic unless the U is excised through the action of uracil-DNA glycosylase [Duncan and Miller, 1980;Duncan and Weiss, 1982; reviewed by Lindahl, 19931. Direct oxidative deamination of bases by nitrous acid (NA) exposure of single-stranded RNA or DNA and of organisms containing these nucleic acids as genetic material also can explain the bulk of NA-induced mutants detected [Zimmermann, 1977;Frankel et al., 19801. On the other hand, data on NA treatment of duplex DNA in vitro and on mutations obtained after NA treatment of intact organisms containing duplex DNA have been confusing and often not consonant with the simple deamination theory [Zimmermann, 1977;Thomas et al., 1979b;Frankel et al., 19801. Here we present data on additional tests and formulate models that suggest further experimentation which may help in the synthesis of a unified description of NA mutagenesis of duplex DNA.

Bacterial Strains
Escherichia coli K12 strains matched for concordant background genotypes were of two sets: a) E. coli Genetic Stock Center strain CGSC6078 (strain BW310 of B. Weiss) carrying mutation nng-1 and defective in Detroit, MI). Liquid minimal E medium [Vogel and Bonner, 19561 was supplemented with d-biotin (2 pg/ml), L-histidine (20 pg/ml), and 0.2% glucose (pH 7.1). Minimal plus biotin plates contained E medium plus 2 pglml d-biotin, 0.2% glucose, and 1.5% Difco agar. Soft agar contained E medium supplemented with 9.6 pg/ml L-histidine, 2 pg/ml d-biotin, 0.2% glucose, and 0.5% agar. Nutrient agar was supplemented with 50 pg/ml rifampicin when used for selection of Rif' mutants. The rifampicin stock solution was made up in methanol (5 mg/ml).

Nitrous Acid Mutagenesis
Aerated cultures in nutrient broth were grown overnight at 37°C. For log phase cultures, aliquots were diluted 1:lO into fresh nutrient broth and aerated 90 min at 37°C. Bacteria were sedimented at 3,500 rpm for 10 min in a model NSE "Servall" centrifuge (Ivan Sorvall, Inc., Norwalk, CT) and resuspended in 0.5 volume of 0.05 M KH,PO, buffer at pH 4.7. A 500 mM aqueous solution of sodium nitrite was diluted 1:lO into portions of the bacterial suspension in buffer and the mixture incubated at 37°C. Samples were removed at intervals. For assays of viability, dilutions were spread on the surfaces of nutrient agar plates. In experiments with Salmonella strains hisG46 and TA1950, 0.1 ml undiluted samples were added to each of two tubes containing 2.5 ml soft agar and poured onto minimal plus biotin plates to select for His' revertants. Revertant colonies were hand counted on a New Brunswick colony counter after incubation at 37OC for 2 days.
The same protocol was used for E. coli strains except for mutant selection. To score for rifampicin-resistant (Rif? mutants, 0.1 ml undiluted samples for each time point were spread evenly on each of two nutrient agar plates and incubated at 37°C for 24 hr. This growth allowed phenotypic expression of rifampicin resistance. The resulting bacterial lawn was then replica plated [Lederberg and Lederberg, 19521 onto nutrient agar plus rifampicin plates and incubated for 2 days at 37°C before scoring Rif' mutants.

Nitrous Acid + Spermidine Mutagenesis
Aerated cultures of TA1950 were grown overnight in minimal medium supplemented with biotin, glucose, and L-histidine. Aqueous solutions of 5 M sodium nitrite and 1 M spermidine were diluted 1:lO into citrate-phosphate buffer, pH 4.2, in the following combinations: control (distilled water only added), nitrite alone, nitrite plus spermidine, and spermidine alone. Reaction mixtures were incubated for 20 min at 37°C and then diluted 1: 10 into aliquots of strain TA1950 in minimal medium. After incubation for 10 min at 37°C [Hartman and Hartman, 19871, viability was determined and His+ revertants were scored as described under Nitrous Acid Mutagenesis, above. Reaction mixtures containing 500 mM nitrite plus 150,50, or 16.7 mM spermine or putrescine were similarly tested.

Mutation Analysis
hisG46 revertants from mutagenicity assays were single-colony isolated on Minimal plus biotin plates. Sequence changes in representative populations of spontaneously derived and NA-induced and NA + spermidineinduced hisG46 revertants were determined using allele-specific colony hybridization [Cebula and Koch, 19901, except that psoralen was not used to cross-link the probe to the immobilized target DNA. Two revertants, not identified by our probing analysis, were characterized by directly sequencing a 187 single-stranded KR-generated fragment encompassing the hisG46 locus using previously described primers and amplification condi-

Effects of Defects in DNA Repair
Howard-Flanders and Boyce [ 19661 noted that a strain of E. coli K12 (strain AB1886) carrying mutation uvrA6, and thus defective in nucleotide excision repair, was much more sensitive to killing by NA than its repair-competent parent strain (strain AB 1 157). Clarke [ 19701 demonstrated an analogous effect in another pair of E. coli strains and showed that mutagenesis by NA also was considerably enhanced in the excision repair-defective strain. Our data c o n f i i and extend these earlier results. Figure 1 shows killing curves and Figure 2 shows mutation curves for two parental repair-competent E. coli K12 strains (#2 and #3) and derivatives defective in correction of alkylated bases (#4 = ada-), excision of U's via uracil-DNA glycosylase (#1 = ung-l), and the broad-spectrum nucleotide excision repair system (#5 = uvrB5; #6 = uvrA6). Bacteria were exposed to 50 mM NA, pH 4.7, at 37°C. Only the two uvr-strains (#5 and #6) exhibited pronounced sensitivity to killing by NA (Fig. 1). Using a procedure different from that utilized here, Da Roza et al. [1977] indicated a modest increase in sensitivity of ung-bacteria to NA. The same two uvr-strains also exhibited greatly increased induction of Rif' by NA (Fig. 2). A surprisingly modest increase in mutagenesis by NA occurred in the case of the uracil-DNA glycosylase-deficient mutant (# 1) (Fig. 2). mutants are induced very slowly in repair-proficient strains, slightly but significantly more rapidly in the Ung-(uracil-DNA glycosylase negative) strain, but much more effectively in the two Uvr-(nucleotide excision repair negative) strains. With these two strains a decrease in mutants at long exposure is attributed to enhanced killing of the bacterial population (see Fig. 1).
Spontaneous Rif' mutants have been shown to arise from base-substitutions (both transitions-specially C -T-and transversions) at a variety of sites in the rpoB gene coding for a subunit of RNA polymerase [Jin and Gross, 19881. From the above, we conclude that the nucleotide excision repair pathway is responsible for removal of an otherwise potentially lethal lesion(s) induced at significant frequency in DNA of bacteria exposed to NA. In addition, persistence of this class of lesion appears critical for a high level of mutagenesis by NA. In contrast, a defect in uracil-DNA glycosylase has strikingly little impact on killing or mutagenesis by NA.
A requirement for the presence of a uvr mutation for cell sensitivity to NA and a high-level induced-mutant yield is also found in Salmonella (Table I). Bacteria exposed to 50 mM NA, pH 4.7, at 37°C show appreciable killing and reversion of the hisG46 mutation only in the strain TA1950 that carries a deletion of the uvrB gene. The repair-proficient strain, hisG46, still failed to exhibit significant NA-induced reversion even when exposed to 100,200,400, and 800 mM NA for up to 36 min, although killing was evident at these higher NA concentrations even after only 6 min incubation (data not shown). In contrast, after 20 min exposure to 50 mM NA, the uvrB-defective strain TA1950 exhibited an over 8,000-fold increase in the frequency of revertants per surviving bacterium over the frequency found among spontaneous mutants (Table I). Previously, Murphey-Corb et al.
[ 19801 had noted NA-induced reversion only in hisG46 strains also carrying a uvrB mutation.

Mutational Spectra
Spontaneous revertants of strain TA1950 as well as those present after NA treatment for 5 min (100% survival) and after 20 min (0.26% survival) ( Table I) were purified and spectra of reverse mutations to His+ determined (upper portion of Table 11). The vast majority of NA-induced reversions are due to C -t T transition mutations with an approximate twofold preference for the second position of the hisG46 mutant codon CCC. Such mutations are to be expected if NA deaminates C to U. Transversions (ACC, CAC , GCC, and an AiT 4 C/G in a suppressor gene :sup) are present either at a very low level (e.g., CAC) or are absent in the majority of assays (upper portion ofTable 11). Thomas et al. [ 1979bl found that purified duplex bacterial transforming DNA is relatively insusceptible to NA mutagenesis in vitro but is readily mutated following denaturation or after addition of various alcohols, glycols, phenols, or amines. Of the three common polyamines, spermine was most effective in enhancing the mutation rate, followed by putrescine and spermidine in that order. Murphey-Corb et al. [1980, 19831 noted comutagenic activity with NA by only a limited number of polyamines in reversion studies of hisG46 in Salmonella; spermidine was an active comutagen whereas putrescine was inactive.

Enhancement of NA Mutagenesis by Spermidine
Indeed, spermidine-NA mixtures are highly potent mutagens for Salmonella strain TA1950 (Table 111; also see Hartman and Hartman [ 19871). Similar mixtures containing either putrescine or spermine failed to show a level of mutagenesis above that elicited by NA alone (data not shown). Base changes effecting reversion were analyzed from spontaneous, NA-treated, and NA + spermidinetreated bacteria (lower portion of Table 11). As previously found for NA (upper portion of Table II), both NA and NA + spermidine mixtures elicit C -T transition mutations in abundance, again with a decided preference for the base substitution in the second position of the codon (lower portion of Table 11).

DISCUSSION
Our data and those in the literature cited in Results clearly demonstrate two key points with regard to NA mutagenesis --   and an aliquot added to a bacterial culture and incubated at 37°C for an additional 10 rnin before plating. Experiment I = stationary phase culture; Experiment I1 = log phase culture.
in bacteria: a) elimination of the nucleotide excision repair (UvrABC) system is requisite for demonstration of a high level of mutagenesis whereas there is little influence by the absence of uracil-DNA glycosylase ( Fig. 2; Table I) and b) transition mutations in a highly characteristic spectrum predominate among revertants. Essentially the same spectrum is found when mutagenesis is by NA alone or by NA supplemented with the "comutagen" spermidine ( Table 11).
The prevalent idea that NA is mutagenic through deamination of C, leading to transition mutations, is consonant with the base changes we observe. However, the lack of impact of a defect in uracil-DNA glycosylase (Fig. 2) is enigmatic. Furthermore, two studies indicate that in NA treatment of duplex DNA there is a very pronounced lag in discernible deamination of C's [Litman, 1961;Frankel et al., 19801. In fact, NA mutagenesis in duplex DNA can hardly be detected in the absence of comutagens such as polyamines [Thomas et al . , 1979bl.

Model 1 : Mutation via Alkylation
One initially attractive idea would be that NA does not work directly but acts indirectly to nitrosate compounds which then lead to formation of DNA base adducts repairable by the Uvr system. Rapid repair of DNA alkylation damage is dependent on nucleotide excision repair [Samson et al., 1988; reviewed by Van Houten, 19901. Thomas et al. [ 1979bl observed strong NA mutagenesis of cell-free DNA in vitro when any one of several polyamines were present. In hisG46 bacteria in vivo, only spermidine and not spermine or putrescine acts as a "comutagen" [Murphey-Corb et al., 1980this paper]. This could merely indicate more ready cellular uptake of spermidine reaction products, either as free spermidine [Kashiwagi et al., 19931 or via a glutathionyl conjugant [Tabor and Tabor, 19751. Both spermine and spermidine form a variety of products of nitrosation [Hildrum et al., 1976[Hildrum et al., , 1977Hotchkiss et al., 19771, a number of which are direct-acting mutagens [Hotchkiss et al., 1979;Thomas et al., 1979al. A major mutagenic product of spermine nitrosation has a half-life of about 4 min at 37°C in aqueous solution [Thomas et al., 1979b], the same as one of two major mutagenic products of spermidine nitrosation [Murphey-Corb et al., 19801. Any hypothetical adducts formed by NA reaction products, however, would have to result in the spectrum of mutations described here, a pattern decidedly different from mutational spectra found for DNA methylating agents and for several bulkier adduct-forming mutagens [Koch et al., 19941. The mutational spectra we observe for NA and NA + spermidine closely mimic those found after treatments, for example with nitroglycerin, that are attributed to deamination of C residues [Wink et al., 1991;Maragos et al., 19931. While nitric oxide does lead to cytosine deamination in duplex DNA [Wink et al., 19911, it is not firmly established that deamination is a major determinant in mutagenesis by nitric oxide in vivo, although substantial data point in that direction [Routledge et al., 1993.

Facilitate Deamination
A second model that would account for the results described here, earlier advanced by Murphey-Corb et al. [ 19801, is that NA-induced intrustrand cross-links are formed in DNA. Such cross-links, designated pX*pX*, were observed in NA-treated DNA by Dubelman and Shapiro [ 19771 but not further characterized with respect to their structure, their kinetics of formation, or their possible sitespecificity. One could suppose that this cross-linked product, presumed to arise from two adjacent G residues on one DNA strand, would distort the duplex enough to facilitate deamination of C residues on the opposing strand ( Fig. 3; Model 2). Thus, deamination effects on mutation would be most striking when the cross-links were left unrepaired by the absence of the Uvr nucleotide excision repair system, allowing time for deaminations to occur. It is established that deamination of C residues in denatured DNA occurs much more rapidly than in double-stranded DNA [Schuster, 1960a,b;Schuster and Vielmetter, 1961;Frankel et al., 19801.
While polyamines do not seem to enhance NA mutagenesis by eliminating the lag or stimulating the overall rate of deamination of C residues, in duplex DNA [Frankel et al., 19801, it is possible that they could have impacts at specific sites on modified DNA. Stimulation of cross-link formation is unlikely since spermine fails to enhance NA inactivation of transforming DNA [Thomas et a). , 1979bl.
As a variation of this second model, one may hypothetically picture the formation of Uvr-repairable G adducts that would similarly allow enhanced deamination of cytosines in the immediate vicinity when left unrepaired. A possible sequence-specificity in NA-induced mutations is discussed below.

Model 3: lnterstrand Cross-Links at Preferred Sites
In Figure 3 (Model 3) we propose a variant of the line of thought used to construct Model 2, but one that has some experimental support. Models 2 and 3 are not mutually exclusive.
G-G interstrand cross-links [Shapiro et al., 19771 are known to be formed in vitro in duplex DNA with only a barely perceptible lag following exposure to NA; interstrand cross-links are formed in high yield, namely 1 cross-link per 4 total deamination events [Becker et al., 1964;Burnotte and Verly, 19711. Interestingly, such cross-links (step 1 in Figure 3) are not formed randomly in duplex DNA; rather, interstrand cross-links preferentially occur in specific sequences of which the sequence found at hisG46 [Hartman et al., 19861, namely 5' CCGG, is considered as an optimal "hotspot" [Kirchner and Hopkins, 1991;Kirchner et al., 1992a,b]. A possible role of polyamines in site-selection of cross-links has not been explored.
Susceptibility of special sequences to NA mutagenesis is consistent with the observation that NA-induced mutations are located non-randomly in genes, i.e., at specific "hotspot" sites in bacteriophage [Benzer, 19611 and Srreprococcus pneumoniue DNA [Lacks, 19661. The NA-induced mutations in S. pneurnoniue DNA so far examined are exclusively G/C -A/T transition mutations [Lacks, 1966;Chen and Lacks, 19911. One sequenced NAinduced mutation, 582nr, involves the first base pair of a ' GCC sequence [Lacks et al., 19821. Three mutants in-

3' CGG
duced by other agents [Chen and Lacks,199 11 were found to revert with NA via transition mutations at the middle base sequence (M597), at the second base 5' GGGCC pair of a ~ 3 ' CCCGG pair of a 5' -GGCC sequence (M532), and at the third base pair of a 5' ~ CGG sequence (M.510) (sequences in Lacks et al. [1982]). All four NA-mutable sites have sequences that would theoretically allow formation of intraor of interstrand cross-links in close proximity to the base pair susceptible to NA mutagenesis. Since S. pneumoniue DNA is 40% G/C, the probability that four NA mutation-prone sites would have such non-random basepair configurations is well below 1%. Model building indicates that G-G inter- with respect to its base pairs 1,2, and 3. Nitrous acid is pictured as forming intrastrand cross-links between G residues as a first step in nitrous acid mutagenesis. This lesion can be repaired by the Uvr nucleotide excision repair system. If unrepaired, DNA structure is distorted, facilitating deamination of one or another C residue. Quite possibly, this cross-link hinders base excision of any U residues formed by deamination of C. By-pass replication would ensure mutation fixation through insertion of an A on the newly synthesized strand opposite the U. Recombination would allow formation of a viable double helix free of cross-links throughout the continuity of the chromosome which would carry a G/C -+ A/T transition mutation at a site "preordained" by the initial intrastrand cross-link. strand cross-links perturb DNA structure only slightly, with significant effects extending only a few base pairs away [Kirchner et al., 1992b], perhaps enough to facilitate deamination of C residues (second step in Model 3, Figure 3).
Retention in Uvr-bacteria of interstrand cross-links would certainly block excision of U's by uracil-DNA glycosylase in the vicinity of the cross-link. As final steps in the process, the interstrand cross-link would need to be resolved and the NA-induced mutation fixed through replication. Wild-type bacteria are apparently able to withstand a large number of cross-links and survive, while much more sensitive, uvr-bacteria still show multi-hit inactivation kinetics (Fig. 1) [Clarke, 19701. A Salmonella double mutant defective both in nucleotide excision repair (uvr-) and recombinational repair (recA-) is ultrasensitive to lethal damage by NA + spermidine mixtures [Murphey-Corb et al., 19801,

Model 3
Model 3 is analogous to Model 2 except that the initial event that enhances opportunity for C deamination is a G-G interstrand cross-link preferentially produced at a '' -cG or a 5' site. The codomation at this covalent interstrand cross-link, when unrepaired by the Uvr nucleotide excision repair complex, both exposes nearby C residues to deamination and blocks excision of the resulting U's by uracil DNA glycosylase. Resolution of the otherwise lethal interstrand cross-link in Uvr-bacteria would have to be more complex than that pictured in Model 2. It would have to be initiated by a recombinational process that allows expression of altered bases in close proximity to the cross-link, ultimately resulting in a C -. T transition mutation at a "nitrous acid hotspot."

3' CG
and an analogous E. coli K12 double mutant is rapidly killed by NA in a kinetically single-hit process [Howard-Flanders and Boyce, 19661. We conjecture that a slow or late-acting, possibly inducible, recombinational system could be active in allowing mutational expression of the targeted, NAdeaminated bases. NA and NA + spermidine mutagenesis is absent in uvrB-recA-Salmonella double mutants although spontaneous mutation still occurs [Murphey-Corb et al., 19801. In keeping with this model, recombinational repair of DNA damage is considered to be a very sluggish process [Sinden and Cole, 19781. The majority, but not all, GC-AT transition mutations in the supF gene of a NA-treated shuttle vector replicated in transformed human cells [Routledge et al., 1994bl can be accounted for by Models 2 and/or 3. However, in this system, numerous transversion mutations, including GC-TA and GC-CG, were also detected [Routledge et al., 1994bl.

Model 4: Xanthine as a Mutagenic lesion
Model 4 acknowledges a deficiency of cytosine deamination during mild nitrous acid treatment of duplex DNA and the proficiency of guanine deamination in such DNA [Litman, 19611. Ability of nitrous acid to induce G -A base substitutions, most likely via oxidative deamination of guanosine to xanthosine (X), was clearly demonstrated in NAtreated single-stranded bacteriophage DNA [Vanderbilt and Tessman, 19701. Kamiya et al. [ 19921 detected substitution of adenine for X at a precise site in a synthetic duplex DNA c-Ha-rus gene transfected into NIH3T3 cells. In addition, Drosophila DNA polymerase cy has been shown to incorporate thymine as well as cytosine at high frequency opposite deoxyxanthine in synthetic polymers in vitro [Eritja et al., 19861.
Thus, oxidative deamination of G to X could explain the G/C-A/T mutations reviewed above. This would be particularly applicable if there were hotspots for deamination at particular base sequences noted in the discussion of Model 3. Kinetic experiments on loss of bases in duplex DNA indicate that a minority of G residues, perhaps not all of them involved in cross-links, is very rapidly lost upon initial exposure to NA [Litman, 19611. Absence of the nucleotide excision repair system in facilitating mutation could result if the UvrABC system actively eliminates T from X:T base pairs. Such X:T base pairs are hypothesized to exist in an equilibrium between two alternate conformations [Eritja et al., 1986;Kamiya et al., 19921. "Wobble" between conformers might be expected to interfere with base stacking. Base stacking is believed to be the major source of DNA duplex stability. Perturbations in base stacking have been indicated as the predominant signal for recognition by the UvrABC nucleotide excision repair system [Van Houton and Snowden, 19931. One might additionally conjecture that cells possess a dX glycosylase or analogous activity. In fact, Oeda et al. [ 19781 detected an enzyme activity in E. coli extracts which was active on NA-treated duplex DNA, an activity still present in a uracil DNA glycosylase negative mutant. This activity could well have included a glycosylase for hypoxanthine [Karren and Lindahl, 1978;Dianov and Lindahl, 19911 as well as a xanthine-specific glycosylase active on X:C basepairs. In contrast, Oeda et al. [ 19781 observed that partially purified uracil DNA glycosylase failed to exhibit activity on NA-treated duplex DNA.

Future Prospects
Numerous aspects of the models proposed here are subject to critical experimental tests. In fact, a major purpose of this paper is to question current textbook renditions of the mutagenic action of NA on duplex DNA, leading to its reexamination by both in vivo and in vitro experimentation.