To Slip or Skip, Visualizing Frameshift Mutation Dynamics for Error-prone DNA Polymerases*

Three models describing frameshift mutations are “classical” Streisinger slippage, proposed for repetitive DNA, and “misincorporatation misalignment” and “dNTP-stabilized misalignment,” proposed for non-re-petitive DNA. We distinguish between models using pre-steady state fluorescence kinetics to visualize transiently misaligned DNA intermediates and nucleotide incorporation products formed by DNA polymerases adept at making small frameshift mutations in vivo . Human polymerase (pol) (cid:1) catalyzes Streisinger slippage exclusively in repetitive DNA, requiring as little as a dinucleotide repeat. Escherichia coli pol IV uses dNTP-stabilized misalignment in identical repetitive DNA sequences, revealing that pol (cid:1) and pol IV use different mechanisms in repetitive DNA to achieve the same mutational end point. In non-repeat sequences, pol (cid:1) switches to dNTP-stabilized misalignment. pol (cid:2) generates (cid:3) 1 frameshifts in “long” repeats and base substitutions in “short” repeats. Thus, two polymerases can use two different frameshift mechanisms on identical sequences, whereas one polymerase can alternate between frameshift mechanisms to process different sequences. Pre-steady studies of DNA polymerase fidelity have been focused on base substitution mutagenesis mechanisms 2). Simple frameshift mechanisms not yet been addressed despite the destructive biological consequences of having one or a few bases or added. an Quenched-flow Methodology— A Kintek rapid quenched-flow used determine the step for dNTP Primer extension was by equal of 32 and with either or Reactions were quenched with at various onto analyzed on a PhosphorImager were using IV concentrations Real fluorescence analysis of nucleotide was using IV, 2 m dNTP. The in 2AP fluorescence a of or double Steady State 2AP Fluorescence Methodology— Experiments using the non-reiterative p/t and DNA pol (cid:1) were performed using a QuantaMaster QM-1 fluorometer (Photon Technology International) by excit- ing 2AP at 310 nm (6-nm band pass) and monitoring the emission at 370 nm (6-nm band pass). A 1-cm path cuvette containing 200 n M p/t DNA and 1 m M pol (cid:1) was placed in the fluorometer, and the change in fluorescence as a function of time was monitored following the addition of dGTP (400 (cid:1) M ). The data were fit to a single exponential equation. stopped-flow pre-steady state kinetics. Direct incorporation of dTTP opposite 2AP results in a decrease (quench) in the fluorescence signal. Streisinger slippage also causes a decrease in fluorescence when 2AP slips under the primer, whereas dNTP-stabilized misalignment forces 2AP out of the helical plane, thereby reducing 2AP base-stacking interactions and resulting in an increase in fluorescence intensity. Misincorporation misalignment is a two-step process in which an initial dGTP (cid:1) 2AP mispair forms, causing a decrease in fluorescence followed by a realignment of the mismatched G to pair with its correct downstream partner, C, prior to extension. During the mismatch realignment, 2AP is forced out of the template plane, causing a fluorescence signal increase. For purposes of visual clarity, 2AP is denoted 2A p in the diagram.

Three models describing frameshift mutations are "classical" Streisinger slippage, proposed for repetitive DNA, and "misincorporatation misalignment" and "dNTP-stabilized misalignment," proposed for non-repetitive DNA. We distinguish between models using pre-steady state fluorescence kinetics to visualize transiently misaligned DNA intermediates and nucleotide incorporation products formed by DNA polymerases adept at making small frameshift mutations in vivo. Human polymerase (pol) catalyzes Streisinger slippage exclusively in repetitive DNA, requiring as little as a dinucleotide repeat. Escherichia coli pol IV uses dNTP-stabilized misalignment in identical repetitive DNA sequences, revealing that pol and pol IV use different mechanisms in repetitive DNA to achieve the same mutational end point. In non-repeat sequences, pol switches to dNTP-stabilized misalignment. pol ␤ generates ؊1 frameshifts in "long" repeats and base substitutions in "short" repeats. Thus, two polymerases can use two different frameshift mechanisms on identical sequences, whereas one polymerase can alternate between frameshift mechanisms to process different sequences.
Streisinger slippage results in simple deletions by displacement, i.e. the "looping out" of one or more bases as a primer strand slides along a run of reiterated template bases during replication (Fig. 1). Misincorporation misalignment occurs when DNA polymerase initially forms a mismatched base pair at the 3Ј-primer end that subsequently realigns by pairing with a complementary downstream template base prior to undergoing further extension (Fig. 1). Alternatively, DNA misalignment could occur as the first step followed by the "correct" incorporation of an incoming dNTP opposite a complementary downstream template base, a process referred to as dNTPstabilized misalignment (Fig. 1), which has been observed in the crystal structure of the pol IV homolog Sulfolobus solfataricus Dpo4 in ternary complex with DNA and an incoming nucleotide (15). The bottom line is that all three processes can follow different paths to arrive at the same mutational end point, a Ϫ1 deletion. Determining precise frameshifting mechanisms for individual DNA polymerases during replication and repair is an essential step toward understanding the basic principles of mutagenesis.
In this study we perform pre-steady state fluorescence kinetics analysis with three polymerases known to generate small deletions, using the fluorescent base analog 2-aminopurine (2AP) to visualize frameshift intermediates and nucleotide incorporation products as they are occurring during real time catalysis. We report the first pre-steady state observation of two different slippage mechanisms within the same polymerase, pol . The data further reveal that pol and pol IV use different deletion mechanisms on identical repetitive sequences to achieve the same mutational end point, whereas pol ␤ performs a mix of deletions and base substitutions in similar repetitive sequences of different lengths. We derive a minimal kinetic model to account for pol -catalyzed slippage in both reiterated and non-reiterated DNA to explain how end processing in a light chain VJ recombination (16,17) would be aided by the multipotent activities of DNA pol .

EXPERIMENTAL PROCEDURES
Expression and Purification of Native Human pol and E. coli pol IV-pol cDNA was subcloned into vector pET41b (Novagen) and expressed in strain BL21(DE3) R1L Codon Plus (Stratagene). Cells were grown at 30°C in Luria-Bertani medium supplemented with kanamycin (5 g/ml) and chloramphenicol (30 g/ml) to an A 600 of 0.7 and induced with isopropyl-1-thio-␤-D-galactopyranoside (1 mM) for an additional 3 h. Cells were resuspended in 50 mM Tris-Cl (pH 7.5), 1 M NaCl, 2 mM DTT, and 10% sucrose and lysed with lysozyme (2 mg/ml) while stirring for 1 h at 4°C. Soluble protein was recovered by centrifugation (12,000 rpm in a GSA rotor at 4°C) and kept at 4°C for the remaining purification steps. Ammonium sulfate was added to 40% saturation, and pellets were resuspended in PC buffer (50 mM Tris-Cl (pH 7.5), 10% glycerol, 1 mM EDTA, and 2 mM DTT) supplemented with 500 mM NaCl and dialyzed against the same buffer overnight. The sample was diluted in PC buffer to 150 mM NaCl and immediately loaded onto a Whatman P-11 phosphocellulose column, washed with 20 column volumes of PC buffer supplemented with 150 mM NaCl and eluted with a 150 -500 mM NaCl gradient (Ͼ10 column volumes). pol -containing fractions were loaded onto a Superdex 200 column (Amersham Biosciences) in PC buffer plus 250 mM NaCl. The cleanest fractions were applied to a Heparin Hi-Trap column (Amersham Biosciences), washed with 20 column volumes of PC buffer supplemented with 250 mM NaCl, and eluted with a gradient of 250 mM to1 M NaCl (Ͼ15 column volumes). Finally, the cleanest fraction of pol was loaded onto a Superdex 75 column (Amersham Biosciences) equilibrated in PC buffer plus 250 mM NaCl. The eluted fractions were aliquoted and stored at Ϫ80°C until used.
DNA Substrates and Reaction Conditions-Oligonucleotides were synthesized on an Applied Biosystems 392 DNA/RNA synthesizer and purified on denaturing 16% polyacrylamide gels. 2-Aminopurine phosphoramidite and a 5Ј-C6 amino linker phosphoramidite (used in rhodamine X labeling; described previously) (18) were purchased from Glen Research. Primers for quenched-flow and gel assay experiments were labeled with [␥-32 P]ATP (ICN) using T4 polynucleotide kinase (United States Biochemical Corp.). Primers and templates were annealed at a 1:3 primer/template ratio by heating to 95°C for 3 min and slow cooling to room temperature. Ultra-pure deoxynucleoside triphosphates were purchased from Amersham Biosciences and Sigma. Reaction buffer contained 50 mM Tris-Cl (pH 7.5), 50 mM NaCl, 8 mM MgCl 2 , 2 mM DTT, 5% glycerol, and 50 g/ml bovine serum albumin, and reactions were performed at 37°C.
Steady State and Pre-steady State Rotational Anisotropy Measurements-Fluorescence steady state rotational anisotropy measurements were made using a QuantaMaster QM-1 fluorometer (Photon Technology International) with a single emission channel. Samples were excited with vertically polarized light at 580 nm (8-nm band pass), and both vertical and horizontal emissions were monitored at 610 nm (8-nm band pass). The G factor was determined and used to calculate anisotropy. To measure the K d for DNA, the change in the anisotropy of rhodamine-labeled p/t DNA (50 nM) was determined at varying concentrations of pol . Anisotropy data was converted to the concentration of bound DNA as shown in Equation 1, where r is the measured anisotropy, r F is the anisotropy without protein, r B is the anisotropy of fully bound DNA, and D 0 is the initial concentration of p/t DNA. To determine the K d for DNA, the concentration of bound DNA was fit to the quadratic equation shown below as Equation 2, where E 0 and D 0 are initial pol and p/t DNA concentrations, respectively. The off rate (k off ) for pol from rhodamine-labeled p/t DNA was determined using a pre-steady state ⌸*180 stopped-flow instrument from Applied Photophysics equipped for anisotropy by exciting with vertically polarized light at 580 nm (2-nm slit width) and monitoring emissions using a 620 cutoff filter. pol (2 M) was pre-incubated in reaction buffer with p/t DNA (100 nM) in one syringe. The reaction was initiated by mixing equal volumes with a second syringe filled with trapping agent (either heparin (2 mg/ml) or unlabeled p/t DNA (10 M)), yielding final concentrations of 50 nM p/t, 1 M pol , and either 1 mg/ml heparin or 5 M unlabeled p/t trap. The change in anisotropy was monitored as a function of time and fit to an exponential decay.
Pre-steady State Rapid Quenched-flow Methodology-A Kintek rapid quenched-flow apparatus was used to determine the chemical step for dNTP incorporation. Primer extension was measured by mixing equal amounts of syringe 1 containing 32 P-labeled p/t DNA (100 nM) and pol (1.6 M), with syringe 2 containing either dTTP or dGTP (800 M). Reactions were quenched with EDTA (0.5 M) at various time points. Reaction products were loaded onto 16% PAGE, analyzed on a Storm PhosphorImager (Amersham Biosciences), and the data were plotted using Sigmaplot 8.0. Reactions using pol IV contained final concentrations of 200 nM p/t, 2 M pol IV, and 2 mM dNTP. The K d of dGTP for pol was measured by varying the concentration of dGTP. The rate (k obs ) at each dGTP concentration was plotted and fit to a rectangular hyperbola (k obs ϭ k pol [dNTP]/(K d ϩ [dNTP])).
Pre-steady State 2AP Real Time Fluorescence Analysis-Real time fluorescence analysis of nucleotide incorporation was performed using the ⌸*180 stopped-flow instrument from Applied Photophysics. The p/t DNA containing 2AP was excited at 312 nm with a xenon-mercury lamp, and emission was monitored using a 360-nm cutoff filter. Reactions were initiated by combining a syringe filled with 400 nM p/t DNA and 1.6 M pol with a syringe containing either 800 M dTTP or dGTP. Final concentrations were 200 nM p/t DNA, 800 nM pol , and 400 M dNTP. Reactions using pol IV contained final concentrations of 200 nM p/t DNA, 2 M pol IV, and 2 mM dNTP. The change in 2AP fluorescence as a function of time was fit to either a single or a double exponential equation to obtain reaction rates.
Steady State 2AP Fluorescence Methodology-Experiments using the non-reiterative p/t and DNA pol were performed using a Quanta-Master QM-1 fluorometer (Photon Technology International) by exciting 2AP at 310 nm (6-nm band pass) and monitoring the emission at 370 nm (6-nm band pass). A 1-cm path cuvette containing 200 nM p/t DNA and 1 mM pol was placed in the fluorometer, and the change in fluorescence as a function of time was monitored following the addition of dGTP (400 M). The data were fit to a single exponential equation.

RESULTS
pol Frameshifts Occur by a Streisinger Slippage Mechanism in Reiterative DNA-2AP, the base analog of A, forms Watson-Crick base pairs with T (19). When excited at 310 nm, 2AP emits a fluorescent signal at 370 nm, enabling the detection of distortions in local DNA structure by a change in fluorescence intensity (20 -23). The changes in 2AP fluorescence intensity accompanying slippage events during replication are a decrease in fluorescence when T is incorporated opposite 2AP directly or when 2AP slips opposite a 3Ј-primer T and an increase in fluorescence when 2AP "flips" out of the helical plane ( Fig. 1). By placing 2AP adjacent to a homopolymer run of A residues in the p/t DNA such that the analog can easily re-align along the primer end, concurrent changes in the amplitude and direction of 2AP fluorescence intensity that occur during slippage can thereby reveal the mechanisms at work in the different polymerases ( Fig. 1).
FIG. 1. DNA polymerase slippage mechanisms and predicted fluorescent changes. Diagram of a correct incorporation, the three frameshift mechanisms, and the associated changes in 2AP fluorescent intensities as measured by stopped-flow pre-steady state kinetics. Direct incorporation of dTTP opposite 2AP results in a decrease (quench) in the fluorescence signal. Streisinger slippage also causes a decrease in fluorescence when 2AP slips under the primer, whereas dNTP-stabilized misalignment forces 2AP out of the helical plane, thereby reducing 2AP base-stacking interactions and resulting in an increase in fluorescence intensity. Misincorporation misalignment is a two-step process in which an initial dGTP⅐2AP mispair forms, causing a decrease in fluorescence followed by a realignment of the mismatched G to pair with its correct downstream partner, C, prior to extension. During the mismatch realignment, 2AP is forced out of the template plane, causing a fluorescence signal increase. For purposes of visual clarity, 2AP is denoted 2A p in the diagram.
To establish a reference fluorescence profile for incorporation by pol on a five-base repeat sequence without a frameshift, we monitored the addition of dTTP forming a T⅐2AP base pair. This resulted in a biphasic quench in fluorescence fitting to a double exponential with rates of 14 Ϯ 2.5 s Ϫ1 and 0.7 Ϯ 0.15 s Ϫ1 (Fig. 2a, black curve). The second phase of the reaction represents the rate-limiting step for chemistry, which disappears ( Fig. 2a, gray curve) when dideoxy-T is placed at the 3Ј-primer end. Further verification of the chemistry rate was obtained by pre-steady state quenched-flow analysis using single turnover reactions that yielded a rate of 0.68 Ϯ 0.09 s Ϫ1 (data not shown), closely matching the rate observed via fluorescence (0.7 Ϯ 0.15 s Ϫ1 ).
The kinetics for incorporating the "next correct" nucleotide G opposite C is measured by omitting dTTP from the reaction (Fig. 2b, black curve). A biphasic fluorescence quench is observed with rates of 15 Ϯ 3 s Ϫ1 and 1.5 Ϯ 0.35 s Ϫ1 . The slower rate disappears using the dideoxy-T-terminated primer (Fig.  2b, gray curve) and corresponds to the rate-limiting step for chemistry measured by single turnover experiments using quenched-flow (1.2 Ϯ 0.27 s Ϫ1 ) (Fig. 2c). The direct incorporation of T opposite 2AP or the slipped incorporation of G opposite C occurs at similar rates of ϳ1 s Ϫ1 under these conditions.
The biphasic fluorescence quench during the incorporation of G holds true even when the p/t is modified to contain just a dinucleotide repeat (Fig. 3a, black curve). To confirm that the incorporation of G on these repeat sequences is occurring opposite C on a transiently misaligned template and not directly opposite 2AP, we replaced the template C just downstream of 2AP with T (Fig. 3a, gray trace). If the slower fluorescence quench results from direct misincorporation of G opposite 2AP and not slippage, then the chemistry quench should not be governed by the presence of either C or T downstream but should yield essentially similar profiles when copying 3Ј APC 5Ј or 3Ј APT 5Ј templates. Fig. 3a shows that this is not the case. When C is replaced with T in the template, incorporation of dGTP fails to occur, as indicated by a lack of a change in the fluorescence signal (Fig. 3a, gray trace) and confirmed by the absence of product formation using quenched flow on the 3Ј APT 5Ј template (data not shown). The data suggest the incoming dGTP directly forms a correct base pair with the downstream template C immediately before chemistry and that the 2AP is concurrently "slipping" under the primer end T, possibly forcing a more upstream base out of the helical plane in the duplex region of the primer-template. Frameshift events where the template base has "slipped" under the primer were described by Streisinger (11), and this mechanism appears to be at work in pol , even in the case of just a dinucleotide repeat. The likelihood of misinserting G opposite 2AP is extremely small because of the gross instability of G⅐AP mispairs (24), thus strongly favoring Streisinger slippage on the 3Ј APC 5Ј template. In contrast, nucleotide misincorporation, detected as an increase in fluorescence, is favored over Streisinger slippage when pol incorporates A on the 3Ј APT 5Ј template, because A⅐AP forms a relatively stable wobble structure (25) (data not shown).
The pre-chemistry quench could be attributable to either of two causes. One possible explanation is that rapid dNTP binding quenches 2AP through increased template stacking (26,27) as the substrates line up with Mg 2ϩ in preparation for catalysis in the active site (21,28). Supporting this possibility is our observation that the pre-chemistry quench becomes significantly greater as the concentration of dGTP increases (data not shown). Alternatively, the concomitant closure of the pol active site upon dNTP/Mg 2ϩ binding might also account for the initial quench. A similar pre-chemistry 2AP quench was ob-served previously using pol ␤ (29,30) and was attributed specifically to concomitant closure of the active site upon dNTP/ Mg 2ϩ binding based on the crystal structure of the binary versus ternary complex (31). A crystal structure for pol will likely be necessary to help distinguish between active site closure and increased 2AP stacking as the cause for the initial pre-chemistry quench, although the two explanations are not mutually exclusive.
FIG. 2. pol pre-steady state dNTP incorporation on a fivebase reiterative p/t DNA. a, changes in 2AP fluorescence upon the addition of dTTP monitored using stopped-flow kinetics. The primers (30-nucleotide) and template (80-nucleotide) used for the reactions in panels a-c contain the same repeat region sketched in panel a. The data fit to a double exponential decay with rate constants of 14 Ϯ 2.5 s Ϫ1 and 0.7 Ϯ 0.15 s Ϫ1 (black curve). The same experiment was repeated using a primer containing a dideoxy-TMP on the 3Ј-end annealed to the same template (gray curve), resulting in the disappearance of the second slower rate, whereas the initial quench remained relatively unchanged at 18 Ϯ 1.3 s Ϫ1 . 2AP is denoted Ap in the sequence. b, changes in 2AP fluorescence upon the addition of dGTP. The black curve shows a biphasic quench in fluorescence with rates of 15 Ϯ 3 s Ϫ1 and 1.5 Ϯ 0.35 s Ϫ1 . The experiment was also performed on the 3Ј-dideoxy containing p/t (gray curve), resulting in the disappearance of the second phase and leaving the initial quench relatively unchanged at 13 Ϯ 2.4 s Ϫ1 . c, the rate-limiting step for the chemistry of dGTP incorporation was measured by quenched-flow. The data fit to a single exponential rise yielding a rate of 1. pol Uses dNTP-stabilized Misalignment in Non-reiterative DNA-To verify that 2AP movement under the T at the primer end is the cause of the quench observed during dGTP incorporation (Fig. 2b, five-base repeat; Fig. 3a, dinucleotide repeat), we used a p/t DNA sequence containing an A at the 3Ј-primer end rather than a T, eliminating the potential for base pairing between the 3Ј-terminal base of the primer with 2AP in the template (Fig. 3b). If 2AP cannot slip under the primer, then the only means to achieve chemistry with dGTP would be to force 2AP out of the template plane to accommodate the G⅐C pair in a dNTP-stabilized misalignment mode (Fig. 1), causing the disruption of base stacking and an increase in fluorescence. Using the modified p/t sequence, dGTP no longer triggers a quench but rather an increase in fluorescence (Fig. 3b). 2AP is "skipped" by slow displacement in the active site (0.0025 Ϯ 0.001 s Ϫ1 ; Fig. 3b), yielding the predicted fluorescence increase that matches a reduction in the rate of chemistry to 0.0031 Ϯ 0.001 s Ϫ1 as measured by quenched-flow (Fig. 3c). Bond formation thus may not occur until 2AP moves out of the way for proper G⅐C base pair formation. In the non-reiterative sequence, pol is limited to making frameshift mutations via dNTP-stabilized misalignment. We believe that these experiments provide the first "dynamic" (i.e. real time) evidence that a single polymerase (pol ) can use two kinetically distinct mechanisms to achieve the same mutational (Ϫ1 deletion) end point.
Direct Observation of dNTP-stabilized Misalignment on Reiterative p/t DNA Using E. coli pol IV-The crystal structure of S. solfataricus Dpo4 suggests that a dinB (pol IV) homolog favors dNTP-stabilized misalignment as a primary means to produce frameshifts (15). A steady state fluorescence kinetics analysis (32) provides independent support that dNTP-stabilized misalignment is the predominant frameshift mechanism for E. coli pol IV. We have measured the slipped-mispaired intermediates in the pol IV frameshift pathway for direct comparison with pol on an identical p/t DNA sequence (Fig. 4). A rapid quench in fluorescence fitting to a double exponential at rates of 6.4 Ϯ 2.2 s Ϫ1 and 0.31 Ϯ 0.089 s Ϫ1 is observed for a correct T⅐2AP base pair in a three-base repeat p/t DNA sequence (Fig. 4a).
In the non-slippage T⅐2AP reaction, the faster rate of quench for pol IV (6.4 Ϯ 2.2 s Ϫ1 ) closely matches the rate-limiting step for the chemistry obtained by quenched-flow (6.9 Ϯ 0.55 s Ϫ1 , Fig. 4c). In contrast, incorporation of dGTP by pol IV causes an increase in fluorescence on p/t DNA of varying degrees of reiteration ( Fig. 4b) but in the opposite direction of the fluorescence change seen for Streisinger slippage on the same p/t with pol (Fig. 2b). A slow increase in fluorescence is observed for pol IV (Fig. 4b), indicating that 2AP unstacks and is skipped, allowing dGTP to align with the next template base C via dNTP-stabilized misalignment as the Dpo4 crystal structure predicts (15). The fluorescence increase with pol IV and dGTP is abolished when the downstream template C is replaced with T (data not shown) and also with a dideoxy-T on the primer end (data not shown).
The rate of dNTP-stabilized misalignment with pol IV changes as a function of the number of reiterative A residues in the template (Fig. 4b), with a run of five A residues upstream of 2AP causing the fastest rise in signal at 0.065 Ϯ 0.019 s Ϫ1 , a run of three A residues at 0.036 Ϯ 0.018 s Ϫ1 , and a single A residue upstream of 2AP causing little change at all. The ratelimiting chemical step measured by quenched-flow is accordingly also reduced in the run of five-A-residue (0.069 Ϯ 0.011 s Ϫ1 ) and the run of three-A-residue (0.033 Ϯ 0.0053 s Ϫ1 ) sequences (Fig. 4d). When G has been incorporated in identical sequences by pol IV (skip equals an increase in fluorescence) and is compared directly to pol (slip equals a decrease in fluorescence), it is clear that these two enzymes are using different mechanisms to generate the same frameshifts on repetitive DNA.
Terminal Nucleotidyl Transferase Activity Does Not Contribute to the Streisinger Slippage Mechanism-pol contains a FIG. 3. Incorporation of dGTP on p/t DNA containing just a dinucleotide repeat results from a slippage event and not from direct misincorporation opposite 2AP. a, pre-steady state incorporation of dGTP on 2AP containing p/t DNA as a function of the next template base containing C or T immediately downstream of the 2AP target base. The upper plot (black) shows fluorescence quenching upon incorporation of (50 M) dGTP as a double exponential quench when 2AP (denoted Ap in the graph) is followed by C in the template. The lower plot (gray) shows the same reaction performed on an identical DNA p/t, except that the next template base downstream from 2AP is changed to a T. The alteration of C to T in the template results in elimination of dGTP incorporation on the DNA p/t. b, steady state fluorescence change of 2AP in the non-reiterative p/t DNA was monitored during the incorporation of dGTP as a function of time, yielding an increase in fluorescence at a rate of 0.0025 Ϯ 0.001 s Ϫ1 . Fluorescence intensity is monitored in a steady state fluorometer as counts per second using a photon detector, in contrast to the pre-steady state stopped flow fluorometer in which an analog measurement of fluorescence is made in units of volts. c, quenched-flow reactions on the non-reiterative p/t sequence reveal incorporation of dGTP at a rate of 0.0031 Ϯ 0.001 s Ϫ1 . potent nucleotidyl transferase activity reported to be more active on duplex p/t DNA than on single-stranded DNA (7). In quenched flow studies, we have found that the apparent rate of transferase activity on a single-stranded primer sequence is ϳ10-fold faster at 6.8 Ϯ 0.5 s Ϫ1 (data not shown) than the observed rate for single T addition in the polymerase mode on duplex DNA of 0.7 Ϯ 0.15 s Ϫ1 (Fig. 2a).
Could the robust transferase activity observed on the singlestranded primers be contributing to the Streisinger slippage process? Transient melting of the reiterative primer end could generate a short single-stranded region on which a G could be added in the transferase mode. If this extended primer rapidly anneals to a realigned 5Ј APC 3Ј template, it could cause a quench in fluorescence. To test this hypothesis, we performed quenched-flow assays using radiolabeled primer in the absence of template to determine whether the rate of chemistry for the dGTP addition in transferase mode could account for the observed Streisinger slippage rates of 1.5 Ϯ 0.35 s Ϫ1 and 1.2 Ϯ 0.27 s Ϫ1 seen in Fig. 2, b and c, respectively. Our results show that the transferase-based addition of dGTP onto the primer end in the single-stranded form is considerably slower (0.004 Ϯ 0.002 s Ϫ1 ; data not shown), and does not appear to contribute to Streisinger slippage. In the transferase mode, pol strongly favors the addition of T over G on single-stranded DNA with T at the 3Ј-end. The contribution of pol transferase activity to the observed dNTP-stabilized misalignment on non-reiterative p/t DNA cannot, however, be unequivocally excluded, but is not likely considering the presence of a template in the active site of the enzyme.
pol Makes Frameshifts More Readily in Reiterative Sequences Than Does pol ␤-Although we were able to elicit Streisinger frameshifts by pol using only dGTP in the reactions, it was important to determine whether such slippage events are possible in a less stringent system where discrimination between all four nucleotides is necessary. To accomplish this, we turned to an assay in which we allowed replication by pol to proceed from a 32 P-labeled version of the p/t DNA for 30 min with all four dNTPs present and followed with restriction digestion using BsrDI (Fig. 5, a-c). The six-base recognition site of BsrDI is located immediately downstream of the first template C, and the enzyme cuts opposite the 5Ј-side of the template C in the non-realigned template. If the template 2AP slipped under the primer during extension, the BsrDI recognition site would be shifted one base closer to the primer end and thus cut opposite the 5Ј-side of 2AP. Fig. 5, a-c show the results of this experiment when performed on three different lengths of reiterative p/t DNA.
Remarkably, with only a simple dinucleotide (Fig. 5a), pol converts at least 25% of cleavable extension products into frameshift mutations as indicated by an increase in the band FIG. 4. Pre-steady state kinetics of E. coli DNA pol IV-catalyzed, dNTP-stabilized misalignment. a, stopped-flow fluorescence assay for incorporation of dTTP opposite 2AP (denoted Ap in the graph) in a three-base repeat sequence. The curve was fit to a double exponential decay with a rapid rate of 6.4 Ϯ 2.2 s Ϫ1 followed by a slower rate of 0.31 Ϯ 0.089 s Ϫ1 . b, fluorescence studies for dGTP-mediated slippage of 2AP adjacent to five, three, or one reiterated A base(s). 2AP slippage by pol IV was observed in both five-A (black) and three-A (dark gray) DNAs with rate constants of 0.065 Ϯ 0.019 s Ϫ1 and 0.036 Ϯ 0.018 s Ϫ1 , respectively. On the other hand, 2AP slippage was not observed in one-A DNA (light gray). c, rapid-quench assay of dTTP insertion opposite 2AP. The resulting data fit to a single exponential equation with a rate of 6.9 Ϯ 0.55 s Ϫ1 , comparable with that observed in Fig. 4a. d, rapid quench rates for 2AP slippage in p/t DNA containing runs of five A and three A residues were found to be 0.069 Ϯ 0.011 s Ϫ1 and 0.033 Ϯ 0.0053 s Ϫ1 , respectively, and correspond to those observed in Fig. 4b using fluorescence.
intensity appearing at the site opposite template 2AP following digestion (Fig. 5a, lane 3). pol ␤, on the other hand, does not produce frameshift mutations on the same p/t DNA, with the only cut appearing opposite template C, i.e. the non-slipped product (Fig. 5a, lane 6). An increase in the length of repeat sequence to include a run of three A residues on the template enables pol to convert Ͼ90% of all cleavable replication products into frameshift mutations (Fig. 5b, lane 3), whereas pol ␤ continues to replicate in a non-slipped mode (Fig. 5b, lane  6). However, if we further increase the number of repeats to 5 A residues, pol still converts Ͼ90% of the products to frameshift mutations (Fig. 5c, lane 3), but now pol ␤ is able to convert 50% of its replication products to Ϫ1 frameshifts (Fig. 5c, lane  6). pol and pol ␤ thus exhibit very different slippage propensities on the same sequences despite being closely related family X polymerases.
Minimum Kinetic Models and Intermediate Steps Depicting pol -Catalyzed Ϫ1 Deletion Pathways-To complete the characterization of pol -catalyzed Ϫ1 frameshifts, we performed additional measurements to determine the K d for DNA, including individual off-and on-rates (k off and k on ) of the polymerase, the K d for dNTPs, and the rate of pyrophosphorolysis (k pyro ). Fig. 6a shows the binding of pol to p/t DNA (five-base repeat; see sketch in Fig. 2a) as a function of polymerase concentration with the data fit to a quadratic equation yielding a K d for DNA of 92 Ϯ 14 nM. Fig. 6b shows the results of measuring k off for pol from 5Ј-rhodamine labeled p/t DNA. The same off rate was observed using either an unlabeled p/t trap with a k off of 9.0 Ϯ 1.2 s Ϫ1 (Fig. 6b, gray line) or a heparin trap with a k off of 9.5 Ϯ 1.6 s Ϫ1 (Fig. 6b, black line). Using the kinetically measured k off and K d for DNA, an estimated k on was calculated to be 1 ϫ 10 8 M Ϫ1 s Ϫ1 (k off /K d ).
To determine the K d for dGTP during a Streisinger slippage event, the rate-limiting chemical step for nucleotide incorporation was monitored using pre-steady state quenched-flow measurements at various concentrations of dGTP. Fig. 6c shows the hyperbolic plot of the rates of incorporation using dGTP as a function of nucleotide concentration, revealing a value of 58 Ϯ 7.9 M for the K d of dGTP and a maximum polymerization rate (k pol ) of 2 Ϯ 0.7 s Ϫ1 . Because of the propensity of pol to use dTTP in both transferase and polymerase modes on duplex DNA, it was not possible to isolate a K d value for dTTP polymerization alone on the duplex reiterative substrate (data not shown).
Pyrophosphorolysis, the reverse reaction of polymerization (shown in Fig. 6d), was measurable using only the highest soluble concentration of pyrophosphate (2 mM) and occurred so slowly (2.3 ϫ 10 Ϫ3 s Ϫ1 ) that only steady state measurements could be made. A summary of the entire slippage reaction mechanism for pol is summarized for both the favored Streisinger and the less efficient dNTP-stabilized misalignment modes (Fig. 7). FIG. 5. Effect of p/t DNA repeat length on slippage by pol and pol ␤. a, DNA primers and templates (10 nM), were 32 P-5Ј-end labeled and incubated with either pol (40 nM) or pol ␤ (40 nM) in the presence of each dNTP (100 M) for 30 min at 37°C. After 30 min, the reactions were split, and the restriction enzyme BsrDI (0.5 units) was added to one-half of the initial reaction and placed at 65°C for 1 h. For the p/t containing only a single T in the primer and a dinucleotide repeat in the template, digestion with BsrDI revealed that Ͼ30% of the cleavable extension products using pol resulted from a slipped intermediate (left panel), whereas Ͼ99% of the cleaved extension products using pol ␤ yielded a band corresponding to the non-slipped structure (right panel). b, expansion of the p/t to include two T residues on the primer shifted Ͼ95% of all cleaved extension products as the result of a slippage process for pol (left panel), whereas Ͼ99% of the pol ␤ extended and cleaved products remain the result of non-slipped replication events (right panel). c, expansion of p/t to include a run of five T⅐A base pairs in the p/t resulted in the same high rate (Ͼ95%) of cleavage products, revealing a slippage event for pol , whereas pol ␤ only now begins to exhibit a Ͼ50% slippage ratio in the cleavage bands (right panel). 2AP is denoted Ap in panels a-c.

DISCUSSION
In this study we have addressed the mechanism by which pol and pol IV rearrange p/t DNA to cause Ϫ1 frameshifts. Combining fluorescent studies in a stopped-flow instrument with the power of quenched-flow and gel assay experiments using 32 P-labeled p/t DNA, we have captured and identified the major steps in the slippage processes of pol and pol IV, as well as the minimal mechanism of slippage for pol . This study marks the first pre-steady analysis of DNA polymerase frameshift mechanisms.
pol Nucleotide Incorporation Dynamics Reveal a Putative Conformational Change and Chemical Step-Correct incorporation of both dTTP opposite 2AP and dGTP opposite the downstream C in the reiterative p/t DNA resulted in a biphasic quench. The rapid phase using the normal deoxynucleotideand dideoxynucleotide-terminated p/t DNA are dependent upon nucleotide binding, and the slower phase of fluorescence quench represents the rate-limiting chemistry step (Fig. 2). The crystal structure of the closest homologue to pol , TdT, revealed that it appears to be locked in a conformation equivalent to the closed form of pol ␤ (33). It is not known if pol undergoes large scale conformational shifts between open and closed forms like pol ␤ (31) or instead is locked in the closed form like TdT, but our data suggest that pol might undergo some form of closure in the active site prior to chemistry.
pol and pol IV Use Different Slippage Mechanisms in Repetitive DNA-Previous studies suggested that pol and pol IV could both misalign p/t DNA, causing frameshifts (5,34). A pre-steady state comparison of pol and pol IV using 2AP allowed us to visually distinguish the mechanisms at work that produce frameshifts in the active sites of these two enzymes. We demonstrate that, whereas pol re-aligns repetitive sequences to "slip" the template under the 3Ј-primer end to cause a quench in 2AP fluorescence (Fig. 2a), pol IV on the other hand prefers to "skip" the first template base, moving it out of the way in its active site to accommodate the next correct nucleotide pair through dNTP-stabilized misalignment, resulting in 2AP fluorescence increase (Fig. 4b). pol IV is misaligning the FIG. 6. Kinetic steps along the pathway of pol -catalyzed frameshifts. a, the binding ability of DNA pol to a rhodamine (RhX)-labeled reiterative p/t (five-base repeat; see sketch in Fig. 2a) was determined by monitoring steady state changes in anisotropy (converted to units of DNA bound in nM) as a function of pol concentration (in nM) and fit to a quadratic equation yielding a K d for binding DNA of 92 Ϯ 14 nM. b, the off-rate (k off ) of pol from the RhX-labeled reiterative p/t DNA was determined by preincubating p/t and pol and monitoring changes in anisotropy following the addition of a trap (either unlabeled DNA (in 100-fold excess) or heparin (1 mg/ml)) as a function of time. Both traps caused a drop in signal with the initial rate of decay fitting to 9 Ϯ 1.2 s Ϫ1 (DNA) and 9.5 Ϯ 1.6 s Ϫ1 (heparin). c, the K d for incorporation of dGTP during slippage on the reiterative p/t was determined by quenched-flow performing multiple reactions at varying concentrations of dGTP (50 M-1 mM) and fitting the observed rates to a hyperbola to yield a K d of 58 Ϯ 7.9 M and a maximum polymerization rate of 2 Ϯ 0.7 s Ϫ1 . d, pyrophosphorolysis, the kinetics of the reverse of polymerization, was measured as conversion of 32 P-5Ј-end-labeled p/t into shorter products as a function of time in the presence of a saturating amount of PP i (2 mM). Reaction products were run on 16% PAGE, quantitated on a PhosphorImager, and fit to a single exponential rise at a rate of 2.3 ϫ 10 Ϫ3 s Ϫ1 .
incoming nucleotide with the downstream template base in a manner similar to the observed ternary complex for S. solfataricus Dpo4 (15). We found that pol can also make frameshifts by dNTP-stabilized misalignment on non-reiterative p/t DNA sequences, albeit with dramatically reduced efficiency compared with the Streisinger mode of slippage (Fig. 3). We have also shown that pol ␤, another member of the family X polymerases, does not slip as readily as its counterpart, pol , in the same reiterative sequences (Fig. 5), probably a reflection of their evolution toward distinct cellular purposes.
In vivo, pol appears to be involved in a rearrangement of light chain immunoglobulin genes (17,35) likely mediated by its interactions with non-homologous end-joining proteins Ku and DNA-PK (16). The ability of pol to slip efficiently in a simple dinucleotide run and use microhomologies to insert the next base would be extremely beneficial to such a system, enabling two partially complementary broken ends to come together that might otherwise fail to do so. Perhaps as pol assists in VJ recombination (17) a variety of complicated junctions are encountered, requiring it to call upon all of its functions (transferase, polymerase, and microhomology searching) to help process the DNA intermediates properly.
It is important that cells limit frameshift-prone polymerases from wreaking havoc throughout their chromosomes. E. coli has evolved a solution to this problem by limiting the access of pol IV to DNA through the SOS regulon, which up-regulates its expression only during times of extreme stress (36). Human pol may be constrained from working on DNA by the requirement of additional cofactors such as the non-homologous endjoining proteins. Through the comparative analysis of polymerases from humans and E. coli, we have attempted to provide a global perspective on deletion mutagenesis, visualizing two unique frameshift mechanisms in real time and proving the old saying that "there is more than one way to skin a cat." FIG. 7. Minimum kinetic schemes accounting for pol -catalyzed ؊1 deletion mutations. Models summarizing steps in the mechanism of pol Streisinger-based frameshift mutation (top) and dNTP-stabilized misalignment (bottom) are shown. During Streisinger slippage (top) the initial association of enzyme (E) to the DNA is limited by the second order binding constant 1 ϫ 10 8 M Ϫ1 s Ϫ1 and the K d for DNA of 92 Ϯ 14 nM. The enzyme-DNA complex can also dissociate at 9 s Ϫ1 . Binding of the second substrate, dGTP coupled with a magnesium ion, occurs at a rate of 15 Ϯ 3 s Ϫ1 and is limited by the binding constant (K d ) for dGTP on this substrate of 58 Ϯ 7.9 M. Binding of the dGTP causes a concurrent slippage of the p/t DNA into a misaligned conformation that is rapidly fixed into position by chemistry occurring at a maximum rate of 2 Ϯ 0.7 s Ϫ1 . The reverse reaction for polymerization occurs very slowly at a rate of 2.3 ϫ 10 Ϫ3 s Ϫ1 , driving pol in the forward direction to another round of nucleoside addition. DNA pol dNTP-stabilized misalignment (bottom) is identical to the Streisinger model above, except the chemistry rate is effectively reduced by the insertion of a slow misalignment step (0.003 s Ϫ1 ) in which 2AP is displaced from the helical plane prior to chemistry.