Copyright (cid:211)
`
`1998 by the Genetics Society of America
`
`DNA Polymerase Fidelity: From Genetics Toward
`a Biochemical Understanding
`
`Myron F. Goodman and D. Kuchnir Fygenson
`Department of Biological Sciences, Hedco Molecular Biology Laboratories, University of
`Southern California, Los Angeles, California 90089-1340
`
`ABSTRACT
`This review summarizes mutagenesis studies, emphasizing the use of bacteriophage T4 mutator and
`antimutator strains. Early genetic studies on T4 identified mutator and antimutator variants of DNA
`polymerase that, in turn, stimulated the development of model systems for the study of DNA polymerase
`fidelity in vitro. Later enzymatic studies using purified T4 mutator and antimutator polymerases were
`essential in elucidating mechanisms of base selection and exonuclease proofreading. In both cases, the
`base analogue 2-aminopurine (2AP) proved tremendously useful—first as a mutagen in vivo and then as
`a probe of DNA polymerase fidelity in vitro. Investigations into mechanisms of DNA polymerase fidelity
`inspired theoretical models that, in turn, called for kinetic and thermodynamic analyses. Thus, the field
`of DNA synthesis fidelity has grown from many directions: genetics, enzymology, kinetics, physical biochem-
`istry, and thermodynamics, and today the interplay continues. The relative contributions of hydrogen
`bonding and base stacking to the accuracy of DNA synthesis are beginning to be deciphered. For the
`future, the main challenges lie in understanding the origins of mutational hot and cold spots.
`
`THE development of molecular biology has been
`
`profoundly influenced by genetic and biochemical
`studies using the bacteriophage T4. In particular, T4
`has served as an invaluable tool for testing new ideas and
`refining concepts of mutagenesis and DNA polymerase
`fidelity. Through his studies on T4 mutagenesis, Jan
`Drake, to whom this issue of Genetics is dedicated,
`played a central role in initiating the remarkably fertile
`area of research into the biochemistry of fidelity.
`In 1968, Drake reported the surprising discovery of
`antimutagenic T4 polymerase mutants (Drake and
`Allen 1968). Until then, mutations in the structural
`gene coding for the T4 polymerase, gene 43 (de Waard
`et al. 1965), had only been reported to generate mutator
`phenotypes (Speyer 1965; Speyer et al. 1966; Freese
`and Freese 1967). The notion that a “defective” (i.e.,
`mutant) polymerase might replicate DNA with higher
`fidelity than the wild-type was revolutionary.
`Reversion frequencies in the nonessential rII region
`of T4, used by Seymour Benzer in his classic studies
`on genetic fine structure (Benzer 1961), were the phe-
`notype of choice for determining the effects of muta-
`tions in the T4 pol gene. The various T4 mutant poly-
`merases exhibited very different mutation rates. While
`the effect depended somewhat on which rII reversion
`was investigated, for several of the rII alleles reversion
`frequencies in the tsL56 mutator and tsCB120 antimuta-
`tor backgrounds differed by as much as 103–104-fold
`
`Corresponding author: Myron F. Goodman, University of Southern
`California, Department of Biological Sciences, SHS Rm. 172, Univer-
`sity Park, Los Angeles, CA 90089-1340.
`E-mail: mgoodman@mizar.usc.edu
`
`Genetics 148: 1475–1482 (April, 1998)
`
`(Drake and Allen 1968; Drake et al. 1969; Speyer
`1965). The mutations in tsL56 are A89T⫹D363N, and
`the mutation in tsCB120, also known as tsL141, is A737V
`(Reha-Krantz 1988, 1989).
`Such large variation in error frequencies suggested
`that the polymerase may play an active role in base
`selection during DNA synthesis. To quote Speyer’s pa-
`per “Mutagenic DNA Polymerase” (Speyer 1965)
`. . . the replicating enzyme is involved more directly in
`the selection of the base . . . [such that] the information
`of the parental DNA strand is transmitted sequentially by
`the enzyme to an allosteric site where selection of the
`nucleotide . . . occurs. Such an enzymic mechanism may
`permit selection by criteria other than the relatively weak
`hydrogen bonds postulated in the template hypothesis
`and account for the high accuracy of DNA replication.
`
`However, as the mechanisms of exonuclease editing
`and mismatch repair emerged, the contribution of the
`polymerase active site to fidelity was deemphasized. But,
`recently, Speyer’s conclusion is regaining prominence.
`For example, Eric Kool has constructed a base ana-
`logue of T that is geometrically similar to T but cannot
`form H-bonds with A (Figure 1), and has shown that it
`is nevertheless incorporated opposite A almost as well
`as T by DNA polymerase I Klenow exo⫺ (Moran et al.
`1997).
`Pioneers in genetic fidelity, such as Speyer, Drake,
`Freese and, of course, Watson and Crick, set the stage
`for three decades of ongoing research into the question
`of how DNA polymerases synthesize DNA with such
`exquisitely high accuracy. What follows is a review of key
`results from those decades and a personal assessment of
`how the fidelity field evolved from the early genetic
`experiments.
`
`Columbia Ex. 2084
`Illumina, Inc. v. The Trustees
`of Columbia University in the
`City of New York
`IPR2020-00988, -01065,
`-01177, -01125, -01323
`
`

`

`Copyright (cid:211)
`
`1998 by the Genetics Society of America
`
`DNA Polymerase Fidelity: From Genetics Toward
`a Biochemical Understanding
`
`Myron F. Goodman and D. Kuchnir Fygenson
`Department of Biological Sciences, Hedco Molecular Biology Laboratories, University of
`Southern California, Los Angeles, California 90089-1340
`
`ABSTRACT
`This review summarizes mutagenesis studies, emphasizing the use of bacteriophage T4 mutator and
`antimutator strains. Early genetic studies on T4 identified mutator and antimutator variants of DNA
`polymerase that, in turn, stimulated the development of model systems for the study of DNA polymerase
`fidelity in vitro. Later enzymatic studies using purified T4 mutator and antimutator polymerases were
`essential in elucidating mechanisms of base selection and exonuclease proofreading. In both cases, the
`base analogue 2-aminopurine (2AP) proved tremendously useful—first as a mutagen in vivo and then as
`a probe of DNA polymerase fidelity in vitro. Investigations into mechanisms of DNA polymerase fidelity
`inspired theoretical models that, in turn, called for kinetic and thermodynamic analyses. Thus, the field
`of DNA synthesis fidelity has grown from many directions: genetics, enzymology, kinetics, physical biochem-
`istry, and thermodynamics, and today the interplay continues. The relative contributions of hydrogen
`bonding and base stacking to the accuracy of DNA synthesis are beginning to be deciphered. For the
`future, the main challenges lie in understanding the origins of mutational hot and cold spots.
`
`THE development of molecular biology has been
`
`profoundly influenced by genetic and biochemical
`studies using the bacteriophage T4. In particular, T4
`has served as an invaluable tool for testing new ideas and
`refining concepts of mutagenesis and DNA polymerase
`fidelity. Through his studies on T4 mutagenesis, Jan
`Drake, to whom this issue of Genetics is dedicated,
`played a central role in initiating the remarkably fertile
`area of research into the biochemistry of fidelity.
`In 1968, Drake reported the surprising discovery of
`antimutagenic T4 polymerase mutants (Drake and
`Allen 1968). Until then, mutations in the structural
`gene coding for the T4 polymerase, gene 43 (de Waard
`et al. 1965), had only been reported to generate mutator
`phenotypes (Speyer 1965; Speyer et al. 1966; Freese
`and Freese 1967). The notion that a “defective” (i.e.,
`mutant) polymerase might replicate DNA with higher
`fidelity than the wild-type was revolutionary.
`Reversion frequencies in the nonessential rII region
`of T4, used by Seymour Benzer in his classic studies
`on genetic fine structure (Benzer 1961), were the phe-
`notype of choice for determining the effects of muta-
`tions in the T4 pol gene. The various T4 mutant poly-
`merases exhibited very different mutation rates. While
`the effect depended somewhat on which rII reversion
`was investigated, for several of the rII alleles reversion
`frequencies in the tsL56 mutator and tsCB120 antimuta-
`tor backgrounds differed by as much as 103–104-fold
`
`Corresponding author: Myron F. Goodman, University of Southern
`California, Department of Biological Sciences, SHS Rm. 172, Univer-
`sity Park, Los Angeles, CA 90089-1340.
`E-mail: mgoodman@mizar.usc.edu
`
`Genetics 148: 1475–1482 (April, 1998)
`
`(Drake and Allen 1968; Drake et al. 1969; Speyer
`1965). The mutations in tsL56 are A89T⫹D363N, and
`the mutation in tsCB120, also known as tsL141, is A737V
`(Reha-Krantz 1988, 1989).
`Such large variation in error frequencies suggested
`that the polymerase may play an active role in base
`selection during DNA synthesis. To quote Speyer’s pa-
`per “Mutagenic DNA Polymerase” (Speyer 1965)
`. . . the replicating enzyme is involved more directly in
`the selection of the base . . . [such that] the information
`of the parental DNA strand is transmitted sequentially by
`the enzyme to an allosteric site where selection of the
`nucleotide . . . occurs. Such an enzymic mechanism may
`permit selection by criteria other than the relatively weak
`hydrogen bonds postulated in the template hypothesis
`and account for the high accuracy of DNA replication.
`
`However, as the mechanisms of exonuclease editing
`and mismatch repair emerged, the contribution of the
`polymerase active site to fidelity was deemphasized. But,
`recently, Speyer’s conclusion is regaining prominence.
`For example, Eric Kool has constructed a base ana-
`logue of T that is geometrically similar to T but cannot
`form H-bonds with A (Figure 1), and has shown that it
`is nevertheless incorporated opposite A almost as well
`as T by DNA polymerase I Klenow exo⫺ (Moran et al.
`1997).
`Pioneers in genetic fidelity, such as Speyer, Drake,
`Freese and, of course, Watson and Crick, set the stage
`for three decades of ongoing research into the question
`of how DNA polymerases synthesize DNA with such
`exquisitely high accuracy. What follows is a review of key
`results from those decades and a personal assessment of
`how the fidelity field evolved from the early genetic
`experiments.
`
`

`

`1476
`
`M. F. Goodman and D. Kuchnir Fygenson
`
`balance between the polymerase and 3⬘-exonuclease re-
`actions was fundamentally linked to the overall accuracy
`of DNA synthesis.
`But was the N/P ratio actually determining the accu-
`racy of DNA synthesis or was it merely correlated with
`increased accuracy in the individual polymerization and
`excision reactions? Bessman and co-workers addressed
`this question by measuring the specificity of the individ-
`ual nuclease and polymerase reactions (Bessman et al.
`1974). They showed that mutator, antimutator, and
`wild-type T4 pols (L56, L141, and 43 ⫹) inserted the
`mutagenic base analogue 2-aminopurine (2AP) oppo-
`site T with similar frequencies. What’s more, the three
`polymerases were also similarly specific in removing
`2AP: excising one correctly inserted A for every two to
`three “misinserted” 2AP molecules. The difference was
`in the overall activity of two reactions. The L141 antimu-
`tator pol excised about 91% of the misinserted 2AP,
`resulting in a “low” net misincorporation frequency of
`about 3%, whereas the L56 mutator excised only 20%
`of the 2APs, resulting a “high” error frequency of 10%.
`The relevance of data using 2AP in vitro to the bacte-
`riophage T4 system in vivo was documented in experi-
`ments showing that 2AP incorporation into T4 DNA in
`vivo was highest for tsL56 mutator and very low for
`tsL141 antimutator relative to 43 ⫹ (Goodman et al.
`1977), and that the mutant and wild-type strains con-
`verted 2AP-free-base to 2AP-triphosphate with roughly
`similar efficiencies, giving rise to similar d(2AP)TP/
`dATP pool ratios for the three strains infecting E. coli
`in vivo (Hopkins and Goodman 1985).
`from Bessman’s
`Concurrently with experiments
`group, Nancy Nossal and her students at NIH were
`also using the T4 system to study polymerase fidelity
`(Hershfield 1973; Hershfield and Nossal 1972).
`Gillen and Nossal (1976) found that L141 (CB120)
`polymerase had difficulty carrying out strand displace-
`ment, suggesting that an impediment to forward trans-
`location may enable the enzyme to proofread more
`effectively. Indeed, it has been shown that the A737V
`mutation in L141 causes an increase in exonuclease
`processivity at the expense of polymerase processivity
`(Spacciapoli and Nossal 1994). These results provide
`a mechanistic explanation for the increase in nuclease/
`polymerase ratio for the L141 antimutator relative to
`wild-type polymerase.
`To test and refine this mechanistic link between N/P
`ratio and polymerase fidelity, we carried out a kinetic
`analysis of the fidelity of L141, wild-type, and L56 poly-
`merases, comparing the incorporation of 2AP in direct
`competition with A opposite a template T (Clayton et
`al. 1979). We found that although 2AP misinsertion
`frequencies were the same for each enzyme at all dNTP
`concentrations, 2AP misincorporation frequencies were
`highly dependent on substrate concentration. At satu-
`rating dNTP concentrations, 2AP misincorporation fre-
`quencies were higher for mutator (L56) and lower for
`
`Figure 1.—Difluorotoluene, a non-hydrogen bonding base
`analogue of T. Chemical structures of thymine and difluoro-
`toluene, an isosteric analog for thymine, used to demonstrate
`the relatively small influence of hydrogen bonding in DNA
`polymerase base selection (Goodman 1997; Moran et al.
`1997).
`
`Studies on the biochemical basis of mutation
`The role of 3ⴕ-exonuclease proofreading in reducing
`polymerase errors: Two important papers published in
`1972 suggested the existence of a polymerase-associated
`3⬘→5⬘ exonuclease, which could increase fidelity by ex-
`cising misincorporated nucleotides at their point of ori-
`gin. Brutlag and Kornberg showed that Escherichia
`coli Pol I excised mispaired nucleotides in preference
`to correctly paired nucleotides from primer-3⬘-termini
`(Brutlag and Kornberg 1972). Bessman and co-work-
`ers, building on the work of Speyer and Drake, purified
`mutant and wild-type T4 polymerases and showed that
`the nuclease-to-polymerase (N/P) activity ratio was high
`intermediate for wild type
`for antimutator (L141),
`(43 ⫹), and extremely low for mutator (L56) strains
`(Muzyczka et al. 1972).
`In the latter experiments, polymerase and 3⬘-exo-
`nuclease activities were measured on an oligo dT-polydA
`primer-template, using saturating dTTP substrate con-
`centrations. Individual phosphocellulose column frac-
`tions of the three T4 pols showed N/P ratios that were
`constant across each chromatographic peak but varied
`between peaks. Wild-type T4 pol excised 1 molecule
`dTMP per 25 molecules inserted. In contrast, the L141
`antimutator T4 pol excised 10 out of 11 dTMPs inserted,
`while the L56 mutator polymerase excised only one out
`of 200. The apparent correlation between N/P ratio and
`polymerase fidelity was very suggestive and demanded
`further substantiation.
`In 1972, Linda J. Reha-Krantz joined Bessman’s
`laboratory as a graduate student and embarked on a
`thesis project of heroic proportions. She grew T4 gene
`43 amber mutants in E. coli suppressor strains and mea-
`sured their mutation frequencies. She then purified the
`mutant polymerases and determined their N/P ratios.
`She observed a near-perfect correlation between anti-
`mutator and mutator behavior in vivo and correspond-
`ingly high and low N/P ratios (Reha-Krantz and Bess-
`man 1977). These results were solid evidence that the
`
`

`

`Accuracy of DNA Replication
`
`1477
`
`antimutator (L141) compared to wild type but all three
`converged to the same value at low-dNTP concentra-
`tions. The effect of dNTP concentration was most pro-
`nounced for the relatively inactive L56 exonuclease. The
`relatively active L141 exonuclease was only marginally
`affected. Thus, we concluded that when low dNTP con-
`centrations limit polymerase activity, even inactive exo-
`nucleases are able to edit out the majority of polymerase
`errors.
`The logic can be seen by analogy to quality control
`along an assembly line. A polymerase is like a machine
`that makes widgets and sends them down the line at a
`certain rate. An exonuclease is like a worker responsible
`for removing defective widgets that come down the line.
`The worker sometimes removes perfect widgets by mis-
`take. (The fewer such mistakes, the more “specific” the
`worker.) However, the number of defective widgets that
`get past the worker depends primarily on how many
`widgets the worker checks as the assembly line rolls by.
`If the assembly line slows down (i.e., there arises an
`impediment to forward translocation), the worker will
`be able to check more widgets and therefore let fewer
`defective ones go by.
`It should be noted, however, that N/P ratio is not a
`fail-safe indicator of a mutator phenotype. As Jan Drake
`has pointed out, it was fortunate that A·T→G·C muta-
`tions were investigated early on for the tsL141 allele,
`otherwise it may not have been identified as an antimu-
`tator (Drake 1992).
`From the beginning (Drake and Allen 1968; Drake et
`al. 1969), it was clear that [antimutators] consistently
`reduce A·T→G·C transition rates (sometimes by more
`than 100-fold), reduce some but not all base-addition and
`base-deletion rates, but tend either not to affect or else
`to increase G·C→A·T transition rates.
`Of course, antimutators will always exhibit some muta-
`tional specificity in the sense that they will only be found
`for alleles that are not well corrected in the wild type
`(Reha-Krantz 1995). And N/P ratio may not reflect
`on the ability to correct mutations templated by unusual
`(e.g., slipped out) primer/template structures. L141, for
`example, exhibits an increased mutagenicity for simple
`frameshifts (Ripley and Shoemaker 1983), perhaps
`due to the altered processivity of its polymerase. Despite
`this lack of universality (Drake 1993), the N/P ratio
`continues to serve as an important enzymatic “marker”
`of polymerase fidelity.
`
`Studies on the biophysical basis of mutation
`
`Models of DNA polymerase fidelity: The discovery
`of proofreading spurred the development of theoretical
`models to account for polymerase fidelity. John Hop-
`field proposed that polymerases might rely on “kinetic
`proofreading” to edit out miscreant base pairs (Hop-
`field 1974). The key idea was that, after binding a
`dNTP in the polymerase active site, the enzyme might
`
`irreversibly enter an activated state, perhaps driven by
`hydrolysis of ATP. Discrimination between correct and
`incorrect nucleotides could then occur twice: first, upon
`entering the active site, where difference in the free
`energy of binding of right vs. wrong dNTPs would favor
`the correct nucleotide, and again upon leaving the acti-
`vated state, where the reaction rates of hydrolysis or
`unbinding might also distinguish between correct and
`incorrectly bound nucleotides. Jacques Ninio pro-
`posed a similar model, invoking a “time delay” that
`facilitated nonproductive hydrolysis of a wrongly bound
`nucleotide (Ninio 1975). These models offered a means
`for reducing the number of nucleotides misinserted by
`a DNA polymerase without resorting to “brute force”
`excision by a dedicated proofreading exonuclease
`(Hopfield 1974).
`We now know that Nature has found “brute force”
`acceptable, however, and a model which explicitly in-
`vokes a 3⬘→5⬘ exonuclease to excise polymerase inser-
`tion errors has proven most useful. The model was pro-
`posed by Galas and Branscomb (1978) in the context
`of analyzing the data of Bessman and co-workers (Bess-
`man et al. 1974) for the incorporation and proofreading
`of 2AP using T4 L56 mutator, 43 ⫹, and L141 antimutator
`polymerases. A simplified sketch of this polymerase-
`proofreading model is presented in Figure 2.
`The model treats polymerization and proofreading
`as two possible outcomes of a series of random events,
`which take place after a dNTP (right or wrong) binds
`to the enzyme. In the sketch, polymerization occurs in
`the lower reaction pathway and proofreading takes
`place in the upper pathway. Connecting the two path-
`ways are the states (A) and (M), referring to annealed
`and melted primer-3⬘-termini, respectively. No distinc-
`tion is made between right and wrong base pairs, except
`to recognize that Watson-Crick (WC) pairs favor the
`annealed state, kA ⬎ kM, while non-WC pairs tend to be
`melted out, kM ⬎ kA. However, a non-WC pair may,
`with low probability, be in the annealed state and get
`incorporated into a growing DNA chain, while a proper
`WC pair may be melted out and get excised. The model
`therefore suggests that it is the equilibrium between
`melted and annealed primer-3⬘-termini, rather than any
`intrinsic/geometric difference between WC and non-
`WC base pairs, that determines whether proofreading
`is likely to occur.
`It was originally assumed that following either an in-
`corporation or excision the system was constrained to
`begin a new polymerization-proofreading cycle starting
`from the annealed state (A). This assumption led to
`the prediction that saturating concentrations of a next-
`correct dNTP (complementary to the template base im-
`mediately downstream from the initial dNMP incorpora-
`tion site) would completely suppress proofreading. How-
`ever,
`the experimental data clearly showed that,
`although the excision of dNMP by the proofreading
`exonuclease diminished at saturating next-nucleotide
`
`

`

`1478
`
`M. F. Goodman and D. Kuchnir Fygenson
`
`Figure 2.—Polymerase-proofreading model. Sketch of a simple model illustrating insertion and 3⬘-exonuclease proofreading
`of right and wrong nucleotides. State (M) refers to a melted primer terminus from which exonucleolytic excision takes place;
`state (A) refers to an annealed primer terminus along the polymerization pathway. Selective hydrolysis of misincorporated
`nucleotides results from the ratio, kM/kA, being much larger for mismatches than for correct matches. Polymerization from state
`(P) is favored over proofreading from state (M) as the concentration of rescue dNTP is increased. Following either excision or
`insertion, a shift occurs one base backward or forward to allow the cycle to repeat. When cycling occurs, the terminal base is
`assumed to reach an equilibrium distribution between states (A) and (M), explaining why proofreading is not entirely suppressed
`even at saturating concentrations of rescue dNTP (Clayton et al. 1979).
`
`concentrations, it was nevertheless present to a signifi-
`cant extent (Clayton et al. 1979). A refinement of the
`model, allowing the system to reach an equilibrium dis-
`tribution of melted and annealed primer termini follow-
`ing nucleotide incorporation and excision, resolved the
`problem. Partial suppression of proofreading in the
`presence of high dNTP concentrations is referred to
`as the “next-nucleotide effect” (Clayton et al. 1979;
`Fersht 1979), and has come to be recognized as a basic
`hallmark of proofreading (Echols and Goodman 1991;
`Goodman et al. 1993).
`The Galas-Branscomb model highlights the impor-
`tance of the interactions between polymerases, proof-
`reading exonucleases, and primer-template DNA. It has
`served as a starting point for investigations into why
`mutational spectra and error rates differ substantially
`among polymerases in different sequence contexts.
`Sequence context effects on DNA polymerase fidelity:
`One of the most general and important sequence con-
`text effects can be understood by examining the influ-
`ence of local DNA stability on N/P ratios. Simply stated,
`stable regions are less frequently melted out, and so less
`available to exonuclease. Consequently, base substitu-
`tion mutations tend to occur more frequently in more
`stable (e.g., G·C-rich) sequences and less frequently in
`less stable (A·T-rich) regions. For example, it has been
`shown that T4 mutation frequencies in vivo and misin-
`corporation of 2AP by T4 pol in vitro decrease with
`increasing temperature (Bessman and Reha-Krantz
`1977). Were it not that higher temperatures made stable
`regions more accessible to exonuclease proofreading,
`one might expect mutations to increase because of
`higher rates of deamination and depurination reac-
`tions. The same study also showed that sites on DNA
`which are relatively insensitive to temperature also did
`
`not show an appreciable difference in mutation compar-
`ing 43 ⫹ and antimutator L141 alleles.
`The ambiguous base pairing properties of 2AP make
`it a useful compound for studying fidelity in vitro and
`mutagenesis in vivo (Echols and Goodman 1991;
`Ronen 1979). However, 2AP has another extremely use-
`ful property; it is moderately fluorescent and can there-
`fore be used to study polymerase mechanisms by observ-
`ing its
`insertion by polymerase and excision by
`exonuclease on a pre-steady-state time scale (Bloom et
`al. 1993; Bloom et al. 1994; Frey et al. 1995).
`Further evidence for the effect of local DNA stability
`on mutagenesis came from such pre-steady-state mea-
`surements. Excision of 2AP was measured on a millisec-
`ond time scale by its increase in fluorescence upon
`excision from a primer-3⬘-terminus and a concomitant
`increase in rotation, as measured by fluorescence depo-
`larization (Bloom et al. 1994). 2AP was placed at a
`primer-3⬘-terminus opposite template T, C, A or G, while
`maintaining a constant surrounding sequence context.
`The observed excision rate correlated inversely with the
`stability of the base pair. Thus, removal of 2AP was
`slowest when paired opposite T, with the order of exci-
`sion being 2AP·T ⬍ 2AP·A ⬍ 2AP·C ⬍ 2AP·G. Measure-
`ments were then made of the hydrolysis of 2AP·N base
`pairs placed proximal to either A·T- or G·C-rich neigh-
`boring sequences. It was found that a proper Watson-
`Crick 2AP·T base pair in an A·T-rich environment was
`actually excised faster than a wobble 2AP·C mispair in
`a G·C-rich environment (Bloom et al. 1994).
`Another important sequence context effect on fidelity
`comes from the influence of base-stacking interactions.
`Ronen and Rahat (1976) first showed that neighboring
`base pairs influenced 2AP-induced base substitution
`mutation rates. Later, Pless and Bessman (1983) cre-
`
`

`

`Accuracy of DNA Replication
`
`1479
`
`ated an extensive data set for misincorporation of 2AP
`at 57 different template T sites on f X DNA, using T4
`wild-type and L141 DNA polymerases. These data
`showed that 2AP misincorporation frequencies varied
`from 0 to 20% when T was located at the primer-3⬘-
`terminus, 0 to 14% for C nearest-neighbors, and from
`0 to ⵑ7% for both G and A primer-termini. At first,
`the 2AP misincorporation frequency did not appear
`to correlate strongly with the identity of the nearest-
`neighbor base-stacking partners on the primer-3⬘-termi-
`nus. However, John Petruska and M.F.G. showed that
`base-stacking interactions between an incoming 2dAPTP
`or dATP and the base at the primer-3⬘-terminus were
`a significant factor in explaining the data, once the
`influence of DNA duplex stability on exonuclease activ-
`ity in the vicinity of the primer terminus was taken into
`account (Petruska and Goodman 1985). The data for
`the L141 antimutator fit best to a model in which five
`base pairs, both upstream and downstream from the site
`of 2AP misincorporation, contributed to neighboring
`sequence stability. The requirement to include contri-
`butions from downstream DNA implied that synthesis
`can continue for as many as five base pairs beyond the
`misincorporation site before the enzyme enters a pro-
`cessive peelback mode to excise the errant base pair, as
`previously observed in vitro (Goodman et al. 1974).
`While duplex stability and base-stacking are certainly
`fundamental, sequence context effects on DNA synthe-
`sis fidelity can be considerably more complex, leading
`to differences that persist irrespective of proofreading
`(Echols and Goodman 1991). For example, Miller
`and co-workers (Coulondre et al. 1978) showed that
`deamination of 5-methyl cytosine gives rise to strong
`mutational hot spots in E. coli. Koch (1971), Strei-
`singer et al. (1966), and Kunkel and colleagues (Kun-
`kel 1985, 1986; Kunkel and Soni 1988) showed that
`frameshift and base substitutions could result from local
`primer-template slippage. We have shown that polymer-
`ase-dNTP interactions can stabilize slipped primer-tem-
`plate sequences and lead to misincorporations (Bloom
`et al. 1997, 1998; Efrati et al. 1997). Polymerase-cata-
`lyzed primer-template slippage is also a plausible expla-
`nation for the occurrence of triplet repeat expansions
`that are implicated in causing neurodegenerative dis-
`eases (Mitas 1997).
`
`Studies on the physical chemical basis of mutation
`
`Kinetics of fidelity: Insight into nucleotide misinser-
`tion and proofreading mechanisms has come from en-
`zyme kinetic analysis on steady state (Bloom et al. 1997,
`1998; Goodman et al. 1993) and pre-steady-state (Bloom
`et al. 1994; Johnson 1993; Kuchta et al. 1987) time
`scales. Pre-steady-state analysis provides detailed mecha-
`nistic information on individual steps in polymerization
`and proofreading pathways, while steady-state measure-
`ments are used to determine transition and transversion
`
`misincorporation rates for polymerases in different se-
`quence contexts. In one complementary series of stud-
`ies, mutational spectra obtained in vitro (Kunkel 1985,
`1986; Kunkel and Soni 1988) and in vivo (Schaaper
`1988) offer a large-scale view of base substitution muta-
`tions and small and large additions and deletions, which
`occur spontaneously in different target genes.
`A “gold standard” measurement of fidelity is made
`by having 3H-labeled right and 32P-labeled wrong dNTPs
`compete directly for incorporation into DNA. While
`this method has proven to be excellent for measuring
`fidelity using base analogues such as 5BU (Trautner
`et al. 1962) and 2AP (Bessman et al. 1974), which are
`incorporated with reasonably high efficiencies, it is dif-
`ficult, if not impossible, to measure misincorporations
`occurring at frequencies less than 10⫺4, which are char-
`acteristic of natural base mispairs (Echols and Good-
`man 1991).
`We have replaced the nucleotide competition method
`to measure fidelity by a kinetic approach originally sug-
`gested by Fersht (1985). In the kinetic method, incor-
`poration of right and wrong nucleotides are measured
`in separate reactions as a function of dNTP concentra-
`tion to obtain Vmax/K m values for incorporating each
`nucleotide. The ratio of Vmax/K m’s for right and wrong
`incorporations measures polymerase fidelity in either
`the presence or absence of proofreading (Creighton
`and Goodman 1995; Fersht 1985). We have built on
`the ideas of Fersht to develop a gel-kinetic fidelity assay
`in which polymerases, including those with proofread-
`ing, processivity clamps, and clamp loading proteins
`can be included in the assay (Bloom et al. 1997, 1998;
`Creighton and Goodman 1995). Recently, we have
`used the assay to determine the base substitution error
`frequency of the E. coli pol III holoenzyme complex and
`found it to fall within in a range from 5 ⫻ 10⫺6 to 4 ⫻
`10⫺7 (Bloom et al. 1997, 1998).
`Relating kinetics to thermodynamics: Perhaps the
`the Galas-Branscomb
`most
`important
`feature of
`model is that it yields estimates of the differences be-
`tween free energy of matched and mismatched base
`pairs in the polymerase active site and in the exo-
`nuclease active site. These ⌬⌬Gpol and ⌬⌬Gexo parame-
`ters can be compared with free energy differences,
`⌬⌬G0, obtained using van’t Hoff (Aboul-ela et al. 1985;
`Petruska et al. 1988) or calorimetric analyses (Bres-
`lauer 1995). The comparisons reveal a sizeable discrep-
`ancy.
`Measurements of 2AP insertion opposite template T
`in competition with A (Clayton et al. 1979) or insertion
`of C opposite 2AP in competition with T (Watanabe
`and Goodman 1981, 1982) yielded free energy differ-
`ences too large by about a factor of two compared with
`⌬⌬G0’s obtained from melting heteroduplex DNA con-
`taining 2AP·T and 2AP·C base pairs (Law et al. 1996).
`Similar conclusions were reached using the gel kinetic
`assay to measure the fidelity of DNA polymerases for a
`
`

`

`1480
`
`M. F. Goodman and D. Kuchnir Fygenson
`
`wide variety of naturally occurring transition (Pu·Py)
`and transversion (Pu·Pu, Py·Py) mispairs (Mendelman
`et al. 1989; Petruska et al. 1988).
`The notion that thermodynamically derived ⌬⌬G0 val-
`ues might govern polymerase insertion specificity was
`based on the fact that on-off rates for binding of dNTP
`to the polymerase-DNA complex are extremely rapid
`compared to phosphodiester bond formation, thereby
`allowing right and wrong nucleotides to reach an effec-
`tive equilibrium at the pol active site (Clayton et al.
`1979; Galas and Branscomb 1978). Similarly, the melt-
`ing, kM, and annealing, kA, of primer-3-termini (Figure
`2) are rapid relative to excision of a previously inserted
`nucleotide or insertion of a next-correct nucleotide,
`so that both pol and exo sites sample an equilibrium
`population of melted and annealed primer termini
`(Clayton