Nucleotide-dependent conformational change governs specificity and analog discrimination by HIV reverse transcriptase

Abstract

Single turnover studies on HIV reverse transcriptase suggest that nucleoside analogs bind more tightly to the enzyme than normal substrates, contrary to rational structural predictions. Here we resolve these controversies by monitoring the kinetics of nucleotide-induced changes in enzyme structure. We show that the specificity constant for incorporation of a normal nucleotide (dCTP) is determined solely by the rate of binding (including isomerization) because isomerization to the closed complex commits the substrate to react. In contrast, a nucleoside analog (3TC-TP, triphosphate form of lamivudine) is incorporated slowly, allowing the conformational change to come to equilibrium and revealing tight nucleotide binding. Our data reconcile previously conflicting reports suggesting that nucleotide analogs bind tighter than normal nucleotides. Rather, dCTP and 3TC-TP bind with nearly equal affinities, but the binding of dCTP never reaches equilibrium. Discrimination against 3TC-TP is based on the slower rate of incorporation due to misalignment of the substrate and/or catalytic residues.

The contribution of substrate-induced structure changes toward enzyme specificity and efficiency has long been debated (1). Although considerable attention has been given to the small conformational changes that are thought to effect catalysis from the closed enzyme-substrate complex (2, 3), little is known about the role of the larger changes in structure from an open to a closed complex following substrate binding. Like most enzymes, structural analysis of HIV reverse transcriptase (RT) reveals a large conformational change after nucleotide binding (4, 5). However, the role of substrate-induced conformational changes in specificity is controversial.

Nucleotide binding and incorporation by HIV RT was initially characterized by Kati et al. (6) using rapid chemical-quench-flow methods. Examination of the nucleotide concentration dependence of the rate of polymerization under single turnover conditions provided an apparent nucleotide dissociation constant (K_d,app) and a maximum rate of nucleotide incorporation (k_pol) according to Scheme 1.

Scheme 1.

Interpretation of these data depended upon the simplifying assumption that polymerization was governed by a single rate-limiting step (k_pol), that nucleotide binding occurred as a rapid equilibrium (K_d,app), and that reactions following polymerization (pyrophosphate release and translocation) were fast. These parameters provide the best measurements to define k_cat and K_m values that apply to processive DNA synthesis where the polymerase incorporates nucleotides sequentially to extend a growing polymer. This model and method of analysis have since been adopted throughout the polymerase field to assign values for nucleotide binding (K_d,app) and incorporation (k_pol) governing specificity, where k_cat/K_m = k_pol/K_d,app. Although the kinetics of inhibition of HIV RT by chain terminators in the steady state are a complex function of the length of the template and the ratio of enzyme to primer template (7), single turnover kinetic studies define the parameters governing nucleotide incorporation at a single site. However, problems arise when attempting to interpret k_pol and K_d,app mechanistically and to quantitatively evaluate discrimination in terms of the individual kinetic parameters and kinetic checkpoints (8). In particular, it is paradoxical that most nucleoside reverse transcriptase inhibitors (NRTIs) appear to bind more tightly to the enzyme (lower K_d,app) than normal nucleotides (9–14) even though steric effects are expected to cause weaker binding.

Analysis of the mechanism of inhibition of HIV RT by nonnucleoside inhibitors implied the existence of a thermodynamically favorable conformational change following nucleotide binding (15, 16), but methods have not allowed the direct measurement of the dynamics of nucleotide-dependent changes in enzyme structure. In order to resolve these controversies and address the larger issues regarding the role of induced fit in enzyme specificity, in general, we have developed a method to measure the time dependence of changes in enzyme structure induced by nucleotide binding. In this report, we show that we can fluorescently label HIV RT at a position on the fingers domain to provide a signal to monitor the enzyme closure that occurs during nucleotide binding and subsequent reopening after incorporation.

We show that nucleotide binding occurs in two steps: A weak rapid equilibrium binding (K₁) to the open state (ED_n) is followed by a rapid change in enzyme structure to form a closed state (FD_ndNTP) at rate k₂, and then a slower chemical reaction at rate k₃ (Scheme 2). Because nucleotide release is slow relative to chemistry (k₃≫k_-2), nucleotide specificity is defined simply by the product K₁k₂ = k_cat/K_m. In contrast, incorporation of a nucleoside analog [3TC-triphosphate (3TC-TP)] is slow so that binding comes to equilibrium, explaining the apparently tighter binding of nucleoside analogs.

Scheme 2.

Results

Nucleotide Binding and Incorporation Kinetics.

We fluorescently labeled HIV RT on the fingers domain at a position that provides a change in fluorescence that is correlated with the conformational change induced by nucleotide binding. To measure the kinetics of the reaction, the labeled enzyme-DNA complex was rapidly mixed in the stopped flow with various concentrations of correct nucleotide (dCTP) to get the results shown in Fig. 1A. There was a rapid decrease in fluorescence followed by a slower increase returning to the starting intensity, defining the kinetics of a single turnover. We show that the decrease in fluorescence defines the rate of the nucleotide-induced change in enzyme structure from the open to the closed state, whereas the chemical reaction defines the rate at which the enzyme returns to the open state as described in Scheme 2.

Fig. 1.

Time dependence of fluorescence change during dCTP binding and incorporation. (A) An E·DNA complex was formed using MDCC-labeled HIV RT (0.4 μM) and 25/45(C) DNA oligonucleotide (0.6 μM) and then mixed 1∶1 in the stopped flow with various concentrations of dCTP (to achieve final concentrations of 2, 4, 8, 10, 20, 40, 60, 80, 200 μM dCTP, 200 nM enzyme, and 300 nM DNA). Data from each concentration were fit to a double exponential to derive rates of reaction. (B) Rates of the fast and slow reaction phases (λ₁ and λ₂) from fitting data in A are graphed as a function of concentration. The slope of the fast phase defines the apparent second-order rate constant for dCTP binding (K₁k₂ = 7 ± 0.5 μM^-1 s^-1), whereas the maximum rate of the slow phase defines k₃ = 19 ± 0.5 s^-1 according to Scheme 2. (C) The experiment shown in A was repeated as lower temperatures (5, 10, 18 °C, lower to upper curves) to estimate the maximum rate of the conformational change, which is too fast to observe at 37 °C. The dashed line corresponds to data collected at 37 °C at sufficiently low concentrations to accurately define the binding rate.

The concentration dependence of the rates of the two reaction phases are shown in Fig. 1B. The slope of the concentration dependence of the fast phase defines the second-order rate constant for dCTP binding K₁k₂ = 7 ± 0.5 μM^-1 s^-1 (Scheme 2). The slow phase reached a maximum rate of 16 ± 2 s^-1, which corresponds to the rate of incorporation measured by rapid quench methods (Fig. 2A and Fig. S1). At high dCTP concentrations, the fast fluorescence signal was lost due to the dead time of the instrument, preventing accurate measurement of the maximum rate of isomerization. Therefore, we examined the reaction at lower temperatures. Fig. 1C shows the results obtained at temperatures of 5, 10, and 18 °C where we could resolve the maximum rates of the fast phase to obtain values of 126 ± 3 s^-1, 233 ± 6 s^-1, and 444 ± 13 s^-1, respectively. The temperature dependence of the rate was analyzed on an Arrhenius plot (Fig. S2) and extrapolation afforded an estimate of a maximum rate of 2,000 ± 200 s^-1 at 37 °C. This is consistent with global fitting of the fluorescence data (shown below) which provided a minimum estimate of k₂ > 1,600 s^-1. Using the value of k₂ = 2,000 ± 200 s^-1 in global fitting of the data allows estimation of the dissociation constant for the initial binding (1/K₁ = 200 ± 20 μM).

Fig. 2.

Global fit to kinetics of dCTP incorporation. In each figure, the smooth lines show the best fit obtained by nonlinear regression in fitting all data simultaneously to the model shown in Scheme 3 using KinTek Explorer software. (A) Data from a chemical-quench-flow experiment in which an E·DNA complex (150 nM enzyme, 100 nM DNA) was mixed with various concentrations of nucleotide (0.25, 0.5, 1, 10, 25 μM), then quenched with 0.5 M EDTA. (B) Fluorescence stopped-flow signal after mixing E·DNA with various nucleotide concentrations (final concentrations: 2, 4, 6, 12, 28, 40, 60, 80, 200 μM dCTP, 200 nM enzyme, 300 nM DNA). (C) Time course of nucleotide dissociation as described in the text; here we only fit the slow phase of the reaction (black line) which occurred at a rate of 0.06 s^-1.

Nucleotide Binding and Release.

Stopped-flow experiments were performed using a 3′-dideoxy-terminated primer (DNA_dd) in order to resolve the kinetics of nucleotide binding and to determine whether the observed fluorescence change was correlated with binding or chemistry. These studies produced a biphasic nucleotide-concentration-dependent decrease in fluorescence as shown in Fig. S3A. The rate of the fast phase increased linearly with nucleotide concentration (Fig. S3B) and defines the second-order rate constant for nucleotide binding of 10.5 ± 0.5 μM^-1 s^-1, which is comparable to that observed with a native primer (Fig. 1B). The slow phase (40 s^-1) was slightly larger than the rate of chemistry observed with the native DNA primer and may suggest an additional isomerization in the pathway preceding chemistry or an additional step to accommodate the dideoxy-nucleotide primer (see SI Text). Nonetheless, these data demonstrate that the decrease in fluorescence is due to nucleotide binding to the enzyme and the subsequent isomerization, not the chemistry step.

Next, we measured the rate of nucleotide release from an E·DNA_dd·dNTP complex by mixing the complex with an excess of unlabeled enzyme to trap free nucleotide. The time dependence of the fluorescence change was fit to a double exponential equation yielding rates of 6.7 ± 0.3 s^-1 and 0.065 ± 0.01 s^-1 (Fig. 2C). The fast phase had a smaller amplitude and could be attributed to binding of excess DNA with the labeled enzyme following mixing. Thus, we assign an estimate for the rate of nucleotide dissociation of 0.065 s^-1.

Global Data Fitting.

The kinetic data defining nucleotide binding and incorporation were fit globally to a two-step nucleotide binding mechanism as shown in Fig. 2. This analysis demonstrates that simultaneous fitting of rapid chemical-quench-flow data to monitor the rate of polymerization (Fig. 2A), the fluorescence data to monitor conformational changes (Fig. 2B), and the rate of nucleotide release (Fig. 2C). The results show that the data can be accurately fit to the minimal model shown in Scheme 3.

Scheme 3.

In this scheme, we list 1/K₁ = 200 μM. Pyrophosphate release and translocation are presumed to be rapid following chemistry (17, 18) and, therefore, this model would account for sequential incorporation events during processive synthesis, although the rates of pyrophosphate release and translocation have not yet been measured for HIV RT.

This model could be expanded to include an additional isomerization step preceding chemistry to account for the slow phase of fluorescence data seen with the dideoxy-terminated primer (Fig. S3), as described in SI Text. Alternatively, the slower phase could be due to translocation of the DNA. However, given the possible artifacts introduced by the artificial primer, we have chosen to omit from the global fitting the binding data obtained with the dideoxy-terminated primer. Rather, we present the minimal model that accounts for the time dependence of reactions with a native primer. The approximate agreement for the rate constant for nucleotide binding in comparing the native primer with the dideoxy-terminated primer is sufficient to justify use of the dideoxy-terminated primer to estimate the nucleotide dissociation rate.

The rate constants given in Scheme 3 demonstrate that the specificity constant governing dCTP incorporation is determined solely by the second-order rate constant for nucleotide binding: k_cat/K_m = K₁k₂ = 10 μM^-1 s^-1, and that the apparent dissociation constant measured by quench-flow methods is defined by K_d,app = k₃/K₁k₂ = 1.6 μM. The net dissociation constant for nucleotide binding would be 1/K₁K₂ = 6 nM if the two-step binding reaction came to equilibrium, but it does not. Therefore, it is clear that previous rapid quench kinetic estimates of “ground-state” nucleotide binding are misleading and should be considered as a Michaelis constant rather than an estimate of a dissociation constant. This then raises question regarding the kinetics of incorporation of nucleoside analogs, which appear to bind more tightly than normal nucleotides in rapid quench kinetic studies. We address these questions by analysis of the incorporation of 3TC-TP (triphosphate form of lamivudine), an analog of dCTP.

3TC-TP Incorporation Kinetics.

Quench-flow analysis of 3TC-TP incorporation under rapid quench conditions yielded values of k_pol and K_d,app of 0.03 s^-1 and 0.04 μM, respectively (Fig. 3A). Kinetics of the fluorescence change during 3TC-TP binding are shown over a timescale of the first 100 ms to illustrate nucleotide binding and the conformational change to form the closed state (Fig. 3B). The subsequent incorporation and recovery of fluorescence occur on a 120 s timescale and fit a rate of 0.029 ± 0.001 s^-1 corresponding to the rate of the chemistry step. The binding data were fit to a double exponential equation to obtain the concentration dependence of the rates governing the two conformational change steps. The concentration dependence of the rates fit a straight line up to 20 μM 3TC-TP, defining an apparent second-order rate constant (K₁k₂) of 37 ± 2 μM^-1 s^-1. The slow phase reached a maximum rate of 65 ± 15 s^-1. At high concentrations of nucleotide, the rate of the fast phase of the reaction was too fast to measure by stopped-flow methods at 37 °C. Therefore, stopped-flow fluorescence data monitoring the fast phase of 3TC-TP binding were analyzed at several lower temperatures and the results extrapolated to estimate the maximum rate of k₂ = 4500 ± 300 s^-1 at 37 °C (see Fig. S2).

Fig. 3.

Global fit to kinetics of 3TC-TP incorporation. (A) Data from a chemical-quench-flow experiment in which an E·DNA complex (150 nM enzyme, 100 nM DNA) was mixed with various concentrations of nucleotide (0.01, 0.02, 0.04, 0.06, 0.1, 1, 2.5, 5, 7.5 μM), then quenched with 0.5 m EDTA. (B) Fluorescence stopped-flow signal after mixing E·DNA (200 nM enzyme, 300 nM DNA) with various nucleotide concentrations (2, 4, 8, 10, 20, 40, 60 μM). For figures A and B, the smooth lines show the best fit obtained by nonlinear regression fitting all data simultaneously to the model shown in Scheme 4 using KinTek Explorer software. (C) Time course of nucleotide dissociation from the E.DNA_dd complex was performed as described in the text; fitting the slow phase (black line) required a value of k_-4 = 0.6 s^-1 (Scheme 4).

Kinetics of 3TC-TP Release.

3TC-TP release from the closed tertiary complex was measured taking advantage of the differing fluorescence states of bound dCTP versus bound 3TC-TP. Although both nucleotides produce a decrease in fluorescence upon binding, the magnitude of the fluorescence change is greater for dCTP. This difference in fluorescence provides a measurable signal when 3TC-TP dissociates and dCTP binds following the addition of 1 mM dCTP to an E·DNA_dd·3TC-TP complex (Fig. 3C). The fast phase can be attributed to the rapid binding of dCTP to free E·DNA_dd due to incomplete saturation of the E·DNA_dd complex with 3TC-TP prior to mixing. The slow phase of fluorescence decrease corresponding to 3TC-TP release occurred at a rate of 0.25 ± 0.01 s^-1.

Model for 3TC-TP Incorporation.

The minimal model for 3TC-TP incorporation is shown in Scheme 4. Analysis of stopped-flow incorporation of 3TC-TP yielded a second-order rate constant K₁k₂ = 37 ± 2 μM^-1 s^-1. Temperature dependence of k₂ yielded a rate of 4,500 ± 300 s^-1. Data could be fit approximately to the two-step binding model shown in Scheme 4 (summarized in Table 1). However, this model does not account for the biphasic fluorescence change and does not provide an optimal fit (Fig. S4).

Scheme 4.

Table 1.

Rate constants derived in fitting to a two-step binding model

Nucleotide	1/K1, μM	k2, s-1	k-2, s-1	k3, s-1
dCTP	200 ± 20	2000 ± 200	0.06 ± 0.02	15 ± 2
3TC-TP	120 ± 20	4500 ± 300	1 ± 0.2	0.028 ± 0.02

Rate constants were derived by a combination of conventional analysis of the concentration dependence of the observed rates and global fitting to the entire dataset as described in SI Text. Rate constants listed for dCTP correspond to the fitted curves shown in Fig. 2. Rate constants listed for fitting 3TC-TP to Scheme 2 produce a less than optimal global fit, shown in Fig. S4.

We fit the data globally to a three-step binding model as shown in Scheme 5 to give the fitted curves shown in Fig. 3 (summarized in Table 2).

Scheme 5.

Table 2.

Rate constants derived in fitting to a four-step model^*

Nucleotide	1/K1, μM	k2, s-1	k-2, s-1	k3, s-1	k-3, s-1	k4, s-1
dCTP†	200 ± 20	2000 ± 200	(1–20)	(100–300)	0.08 ± 0.02	15 ± 2
3TC-TP	120 ± 70	4500 ± 300	70 ± 10	100 ± 20	1 ± 0.2	0.027 ± 0.002

^*Rate constants were derived by a combination of conventional analysis of the concentration dependence of the observed rates and global fitting to the entire dataset as described in SI Text. Rate constants are according to Scheme 5 where the integer subscript refers to the step number and a negative integer refers to the reverse reaction as written.

^†These values are based upon globally fitting the dCTP data to a model with two isomerization steps. The values for k_-2 and k₃ were not well constrained by the data, and a range of values are listed that give a satisfactory fit.

Because the rate of incorporation is so slow, unlike the case with dCTP, the binding reaction comes to equilibrium and the K_d,net = 1/K₁K₂K₃ = 50 nM, which is approximately equal to the apparent dissociation constant estimated by quench-flow methods. Global fitting the rapid quench data (Fig. 3A) and the fluorescence stopped-flow data (Fig. 3B) required a rate of 2.6 s^-1 for the step governing the release of nucleotide (k_-3), whereas fitting the data defining the release of nucleotide from the E·DNA_dd complex (Fig. 3C) to the same model required k_-3 = 0.6 s^-1. We attributed the small difference to the effect of the dideoxy-terminated primer (see SI Text).

Discussion

We show that nucleotide selectivity by HIV RT is due to an initial weak substrate binding followed by a fast isomerization to a tightly bound state leading to incorporation. Although previous kinetic work had predicted the existence of this conformational change step (15, 16) which was then seen structurally (4, 19), we now define the role that the conformational change plays in nucleotide selectivity based upon the observed reaction kinetics. A single fluorophore was attached to the outside of the fingers domain to monitor changes in enzyme structure and had no significant effect on the kinetics of incorporation measured by rapid quench kinetic methods.

The nucleotide-induced change in structure plays a pivotal role in enzyme specificity. Because the kinetic partitioning of the FD_nN state favors chemistry over nucleotide release (k₃≫k_-2), the specificity constant governing incorporation of a normal correct nucleotide (dCTP) is determined solely by the net rate of binding (20). For a two-step binding model, the apparent second-order rate constant for substrate binding is determined by the product K₁k₂. This parameter defines the specificity constant, k_cat/K_m, because k₃≫k_-2. Thus, even though the rate of chemistry (k₃) determines k_cat in this system, k_cat/K_m is independent of the rate of chemistry. Moreover, the apparent dissociation constant determined by rapid quench methods (K_d,app) can now be understood as the ratio of the rate of product formation divided by the substrate binding rate: K_d,app = k₃/K₁k₂. Thus the rapid quench data do not define the K_d for nucleotide ground-state binding, but more accurately reflect the Michaelis constant governing substrate binding and incorporation.

Analysis of the kinetics of the fluorescence change using a dideoxy-terminated primer (DNA_dd) support our conclusion that the fluorescence signal monitors changes in enzyme structure rather than chemistry. The biphasic kinetics observed during nucleotide binding to E·DNA_dd imply that there are two isomerization steps following nucleotide binding. However, the presence of the dideoxy-nucleotide at the 3′ terminus may alter the interaction of the nucleotide at the active site, leading to a second reaction phase. We attempted to address this question in globally fitting the fluorescence data obtained during nucleotide incorporation (Fig. 2B). At the highest nucleotide concentrations, there appeared to be a brief lag after the decrease in fluorescence and prior to the subsequent increase in fluorescence after chemistry. A slightly better fit to the data could be obtained by including a second isomerization step of approximately 200 s^-1 preceding chemistry. However, the improvement in the fit was marginal and was deemed to be insufficient to warrant the inclusion of an additional step. Rather, in our global fitting, we have shown that we can adequately fit the kinetic data governing nucleotide binding and incorporation to the minimal model shown in Scheme 3 without an additional isomerization step. Additional experimentation will be required to test whether an additional isomerization step precedes the chemistry step with a normal primer or whether DNA translocation could account for the slower fluorescence change.

As a result of the isomerization after nucleotide binding, the net dissociation constant, K_d,net = 1/[K₁(1 + K₂)], is 280-fold lower than the K_m (K_d,app), but the two-step binding reaction never comes to equilibrium during enzyme turnover. Rather the concentration dependence of the observed rate is determined by the balance between the rate of binding (K₁k₂) and the rate of product formation (k₃). Using rate constants derived from global fitting for dCTP the K_d for formation of the weak collision complex (step 1) for nucleotide binding is only 200 μM, the K_d,net = 1/[K₁(1 + K₂)] = 6 nM and K_d,app = k₃/K₁k₂ = 1.5 μM (Scheme 3). In contrast, conventional fitting of data derived from quench flow provides a value of K_d,app = 1.2 ± 0.1 μM. The slight discrepancy between these two estimates reflects errors in conventional fitting of data to a simple burst equation compared to global fitting to the model.

The apparent tighter binding by 3TC-TP is due to the 500-fold slower forward rate of chemistry compared to correct nucleotide. This greatly reduced rate of chemistry allows the binding reaction to come to equilibrium resulting in a decreased K_m (K_d,app) value. Accordingly, an NRTI will have an apparently tighter binding (K_d,app) compared to a native nucleotide when changes in structure of the nucleotide have greater effects on the rate of chemistry (k₃) than they do on the rate of binding (K₁k₂). Thus the only significant differences between 3TC-TP and dCTP are in the rate of incorporation, which, in turn, must reflect alterations in the alignment of reactive groups at the active site. The greater bulk of the pyranose ring in 3TC-TP and the altered stereochemistry do not alter the binding rate or affinity. It is noteworthy that, because of the very tight binding of 3TC-TP (6 nM) and slow rate of incorporation, in vivo 3TC-TP could act as a competitive inhibitor, in addition to chain termination. Studies currently underway will address how kinetic parameters governing 3TC-TP incorporation are altered in the evolution of resistance by HIV RT.

For both dCTP and 3TC-TP, there is evidence suggesting that the conformational change may occur in two kinetically significant steps rather than one, although the data were insufficient to define the rate of the second isomerization after dCTP binding. We and others (20–23) have tended toward simple interpretations based upon a single kinetic step from the open to the closed state following nucleotide binding. However, the energy landscape defining the trajectory between open and closed states is likely to be complex. The fluorescence signal afforded by the exogenous label provides only one readout of this complex process, but shows clearly that changes in enzyme structure provide essential elements by which enzymes achieve their extraordinary substrate specificity. The conformational change affords optimal alignment of catalytic residues to grip a correct substrate tightly and align it for catalysis. Our data imply that conformational changes following binding of the nucleoside analog 3TC-TP lead to tight binding, but the less than optimal geometry of the analog precludes fast catalysis.

In light of these findings, we must reassess the mechanisms for discrimination against NRTIs in order to guide the design of new analogs. Questions remain regarding the evolution of resistance to NRTIs because the enzyme could evolve to weaken binding to further slow the rate of chemistry, or both. An advanced understanding of how nucleoside analogs interact with HIV RT could be used in the design of the next generation of NRTIs to overcome the evolution of resistance. It should also be noted that the fluorescence methods we have developed could easily be adapted for use in rapid screens for new inhibitors.

Materials and Methods

HIV RT (HXB2) was bacterially expressed and purified by ion exchange chromatography followed by ssDNA affinity chromatography. The protein was labeled with 7-diethylamino-3-((((2-maleimidyl)ethyl)amino)carbonyl)coumarin (MDCC) at the site of cysteine mutation E36C in p66 as described by Tsai et al. (24). A second mutation was introduced to eliminate on cysteine (C280S), but serine is frequently found at this position.

Pre-steady-state measurements were conducted using an RQF-3 rapid quench flow and SF-2004 stopped flow (KinTek Corp.). The resulting data were initially fit by nonlinear regression using Grafit 5 (Erithacus software), followed by more rigorous global fitting to a single model using KinTek Global Explorer. See SI Materials and Methods.