Subunit Interaction Dynamics of Class Ia Ribonucleotide Reductases: In Search of a Robust Assay

Abstract

Ribonucleotide reductases (RNRs) catalyze the conversion of nucleotides (NDP) to deoxynucleotides (dNDP), in part controlling the ratios and quantities of dNTPs available for DNA replication and repair. The active form of E. coli class Ia RNR is an asymmetric α₂β₂ complex in which α₂ contains the active site and β₂ contains the stable diferric-tyrosyl radical cofactor responsible for initiating the reduction chemistry. Each dNDP is accompanied by disulfide bond formation. We now report that β₂ can act catalytically when assaying RNR in either the absence or presence of an external reductant. In the absence of reductant, rapid chemical quench analysis of a reaction of α₂ (10 μM), substrate, and effector with variable amounts of β₂ (0.1, 1 or 10 μM) yields 3 dCDP/α₂ at all ratios of α₂:β₂ with a rate constant of 8–9 s⁻¹, associated with a rate limiting conformational change(s). Stopped-flow fluorescence spectroscopy with a fluorophore-labeled β reveals that the rate constants for subunit association (163 ± 7 μM⁻¹ s⁻¹) and dissociation (75 ± 10 s⁻¹) are fast relative to turnover, consistent with catalytic β₂. When assaying in the presence of an external reducing system, the turnover number is dictated by the ratio of α₂:β₂, their concentrations, and the concentration and nature of the reducing system; the rate-limiting step can change from the conformational gating, to a step or steps involving disulfide re-reduction, dissociation of the inhibited α₄β₄ state or both. The issues encountered in the E. coli RNR assays are likely of importance in all class I RNR assays and are central to understand in developing screening assays for inhibitors of these enzymes.

Graphical Abstract

INTRODUCTION

E. coli class Ia ribonucleotide reductase (RNR) catalyzes the conversion of nucleoside 5′-diphosphates (CDP, UDP, GDP, ADP) to their corresponding deoxynucleotides (dNDPs), providing the essential building blocks for DNA replication and repair.^1,2 The E. coli Ia enzyme, a paradigm for class I RNRs, consists of two homodimeric subunits, α₂ and β₂, that are active in an asymmetric, tetrameric α₂β₂ complex.^3–6 The diferric-tyrosyl radical (Y₁₂₂•) located in β₂ generates a transient thiyl radical in α₂ (C₄₃₉•) subsequent to its binding NDP and effector, which then initiates NDP reduction. This reversible oxidation process occurs over a distance of ~35 Å and involves a pathway of conserved aromatic amino acid residues and a cysteine.^6,7

The central role of RNRs in nucleic acid metabolism has made the human RNR the target of five clinically used therapeutics that inhibit distinct steps in the reduction process.^8–10 The limiting factor in identifying new types of RNR inhibitors has been the development of rapid in vitro and in vivo screens for enzyme activity. In a recent study with an RNR containing a site-specifically incorporated unnatural amino acid, we identified the ability of a mutant β₂ to act catalytically in α₂ turnover.¹¹ We now report that wt β₂ can also act catalytically in pre-steady state and steady state assays. In support of this result, the kinetics of α/β subunit interactions are fast relative to enzymatic turnover. Consideration of the dynamics of these inter-subunit interactions are important in designing assays for RNR activity and screening for RNR inhibitors.

In early studies to purify E. coli RNR,¹² the two subunits separated and required recombination for activity measurement. The K_d for subunit interaction was subsequently established to be 0.4 μM in the apo form and to increase to 0.2 μM on binding of NDP and its matched allosteric effector.¹³ Since that time, RNR subunits have typically been assayed independently, using a physiological concentration of one subunit (0.1–0.5 μM) and a 5 to 10-fold excess of the second subunit to ensure “active” α₂β₂ complex formation. However, the activity of α₂ is always less than that of β₂,¹⁴ suggesting that the amount of α₂β₂ complex alone does not govern activity.

RNR assays are typically carried out under two sets of conditions in the absence and presence of endogenous reductant. Our recent model for dNDP formation in the absence of endogenous reducing system (steps A-G) and in its presence (step H) is shown in Scheme 1.¹⁵ Many distinct perturbative biochemical studies and recent structural studies of an active α₂β₂ double mutant (F₃Y₁₂₂/E₅₂Q-β₂/α₂/GDP/TTP) establish asymmetry within the “active” complex.¹⁵ We propose a model in which α_2, loaded with substrate (CDP) and allosteric effector (ATP), binds to β₂ and that one dCDP is formed in an α/β pair before triggering formation of a second dCDP in the adjacent α/β pair. In the wt enzyme, 2 dCDPs are generated at ~5–10 s⁻¹.¹⁴ In all the perturbative systems examined, the asymmetry becomes apparent as the switching between the two α/β pairs is altered. For the wt system, each dCDP is accompanied by the oxidation of a C₂₂₅,C₄₆₂ pair to a disulfide. The C-terminal tail of each α monomer contains two additional conserved cysteines (C₇₅₄, C₇₅₉) that are involved in re-reduction of the active site disulfide in their own monomer.¹⁶ While the details of the re-reduction are not established, the structure suggests that the subunits likely dissociate or may undergo a slow conformational reorganization so that two more dCDPs can be generated (step G, Scheme 1) at a rate constant of ~ 0.1 s⁻¹. While a maximum of 4 dCDPs/α₂ can be generated, only 2.5–3 dCDPs are routinely measured. We attribute this observation to α₂’s sensitivity to oxidation during its purification and/or handling. Under typical assay conditions for E. coli RNR, in the presence of endogenous reductants thioredoxin (TR), thioredoxin reductase (TRR) and NADPH, α₂ generates many dCDPs (step H) with a rate constant of 1–2 s⁻¹, that is likely associated with the complexity of C-terminal tail mediated re-reduction and the role of the thioredoxin.^17,18

Scheme 1.

Updated model for kinetics of dNDP formation in the absence of endogenous reductant at 5–10 s⁻¹ (steps A-G), and in the presence of endogenous reductant and multiple turnovers (step H) at 1–2 s⁻¹.

Unlike α₂, the Y• in β₂ is regenerated immediately following dCDP formation.¹¹ This scenario makes it feasible that each β₂ can service multiple α₂s and that fast subunit dissociation and re-association to α₂β₂ followed by rapid chemistry, could preclude observation of these dynamics. We hypothesize that catalytic β₂ might account for the differences in the turnover number of α₂ and β₂.

In the current study, we describe experiments to examine the ability of β₂ to act catalytically in the absence and in the presence of an external reducing system. In the absence of reductant and at ratios of α:β from 1:1, 10:1 and 100:1, ~3 dCDP/α₂ are generated with a rate constant of 8 s⁻¹ relative to Y^•. When the α:β ratio is 100:1, 200 turnovers occur. Rate constants for association (163 ± 7 μM⁻¹ s⁻¹) and dissociation (75 ± 10 s⁻¹) of α₂ and β₂, measured by stopped-flow fluorescence spectroscopy using β₂ modified site-specifically with an environmentally sensitive fluorophore, support a catalytic role for β₂. In the presence of endogenous reductant, we also establish that β₂ can function catalytically. However, under the conditions of limiting α₂ or at high concentrations of the subunits (1 – 10 μM), the rate-limiting step switches from the conformational gating of radical transfer (RT) chemistry to re-reduction of α₂, likely involving conformational change(s)¹⁵ or interconversion of active and inactive quaternary structures of RNR (α₂β₂ and α₄β₄ in E. coli)¹⁹ or both.

The observations that (i) β₂ can function catalytically, (ii) the discoveries of the complexity of α’s quaternary structures and their dependence on concentration of protein and (deoxy)nucleotides, (iii) incompletely loaded metallocofactor in β, and (iv) the instability of α to oxidation and its propensity to aggregate require consideration when developing assays for both old and new RNRs. These issues lead us to suggest a standard way to assay RNRs that is distinct from the one that has been used for the past few decades.

MATERIALS AND METHODS

Materials

His₆-α₂, specific activity (SA) of 2500 nmol(min-mg)⁻¹ and wt-β₂ with 1.2 Y•/β₂, SA of 6000 nmol(min-mg)⁻¹ were expressed from the pET28a-nrdA and pTB2-nrdB plasmids and purified as previously described.^20,21 The diferric-tyrosyl radical in β₂ is always generated by self-assembly,²² and by our method²³ gives 1.2 Y•s that we believe are unequally distributed between each β₂. Sixty percent of β₂ is active and 40% contains a diferric-center with no radical, is inactive and termed met-β₂. α₂ was pre-reduced prior to use with dithiothreitol (DTT) and treated with hydroxyurea (HU) to reduce the tyrosyl radical in the small amount of β₂ that always co-purifies with α₂.²⁴ E. coli thioredoxin, TR (40 U/mg) and thioredoxin reductase, TRR (1400 U/mg) were purified using established protocols.^25,26 6-Bromoacetyl-2-dimethylaminonaphthalene (BADAN) was purchased from Molecular Probes (Eugene, OR). C₂₆₈S/C₃₀₅S/V₃₆₅C-β₂ was isolated and labeled with BADAN to make dansyl-β₂ as previously described.²⁷ For stopped flow (SF) fluorescence spectroscopy experiments, Y₁₂₂• of wt-β₂ or (dansyl-β₂) was reduced by treatment of β₂ with HU. [5-³H]-CDP was obtained from Vitrax (Placentia, CA). Calf alkaline phosphatase was purchased from Roche. Assay buffer is 50 mM HEPES, 15 mM MgSO₄ and 1 mM EDTA (pH 7.6).

dCDP formation kinetics as a function of subunit ratio with no endogenous reductant

Rapid Chemical Quench (RCQ) experiments were performed on a Kintek RQF-3 instrument attached to an external circulating water bath set at 25 °C. To assess the effect of [β₂] on kinetics, wt-α₂ (20 μM) and ATP (6 mM) in assay buffer in syringe A was rapidly mixed with wt-β₂ (0.2, 2 or 20 μM and [5-³H]-CDP (1 mM, 20,000 cpm/nmol) in an equal volume from syringe B. The reaction was quenched in the instrument (5 ms – 100 s) or by hand (> 100 s) in 2% HClO₄ and samples were dephosphorylated and worked up as previously described.^14,28 To assess the effect of [α₄β₄] on dCDP formation kinetics, a control reaction of wt-α₂ (1 μM), wt-β₂ (1 μM), [5-³H]-CDP (0.5 mM) and ATP (3 mM) was also monitored by RCQ. Data were fit to Eq. 1 (for 1:1 and 10:1 α₂:β₂), and Eq. 2 (for 100:1 α₂:β₂) to obtain k_cat from the sum of A₁k₁ + A₂k₂.

$$y=A_1\left(1-e^{-k_{1}t}\right)+A_2\left(1-e^{-k_{2}t}\right)$$ (1)

$$y=A\left(1-e^{-kt}\right)$$ (2)

Association and dissociation rate constants by SF fluorescence spectroscopy

k_assoc and k_dissoc for the α₂β₂ complex in the presence and absence of CDP, ATP, and dATP were determined by SF fluorescence spectroscopy using an Applied Photophysics DX.17MV instrument equipped with the Pro-Data upgrade. The temperature was maintained at 22.0 ± 0.1 °C with a Lauda RE106 circulating water bath and all samples were incubated for 3 min at this temperature prior to measurement. Total fluorescence at λ > 420 nm was detected by a PMT with a long pass filter, using an excitation bandwidth of 40 nm centered at 390 nm. All solutions were prepared in assay buffer. Association rate constants were determined by rapid mixing of 1–6 μM of α₂, ± CDP (0.4 mM), and ± ATP (1.2 mM) in one syringe with an equal volume of 0.16 μM dansyl-β₂ in a second syringe. The reaction was iterated to obtain ~50 traces that were then averaged and fit to Eq. 2 to obtain k_obs. Plotting k_obs as a function of [α₂] provided k_assoc from the slope of the linear fit to Eq. 3 derived from a one-step reversible binding model for pseudo-first order conditions. This process was repeated three times and reported error reported accounts for error between replicates as well as error associated with the goodness of fit.

$$k_{obs}=k_{assoc}\left[{\alpha }_2\right]+k_{dissoc}$$ (3)

As determination of k_dissoc by extrapolation of Eq. 3 to [α₂] = 0 gives significant error, k_dissoc was measured directly by mixing equal volumes of dansyl-β₂ (0.2 μM), α₂ (0.5 μM), ± CDP (0.5 mM) and ± ATP (1.2 mM) or ± dATP (100 μM) in one syringe with a large excess of met-β₂ (100 μM) in a second syringe with the same CDP/ATP. A minimum of 15 traces was averaged per trial and 3–6 independent trials were conducted for each condition (in the presence and absence of CDP and/or ATP). With the exception of dissociation rate constants measured in the presence of dATP, all SF fluorescence traces were well fit to a monoexponential function (Eq. 2) over 1–75 ms. In the presence of dATP, dissociation kinetics were biphasic and fit to Eq. 1 over 0.001–8 s. All fitting was performed using OriginPro 8.0 software (OriginLab). Acceptability of fitting was determined on the basis of qualitative symmetry of residuals about zero amplitude and the R² factor.

dCDP formation kinetics as a function of subunit ratio in the presence of TR/TRR/NADPH

RCQ experiments were performed identically to those described above in the absence of reductant with the following minor modifications. Syringe A included TR (80 μM), TRR (1.6 μM) in addition to wt-α₂ (20 μM) and ATP (6 mM), while syringe B contained wt-β₂ (0.2 or 20 μM), [5-³H]-CDP (2 mM, 20,000 cpm/nmol) and NADPH (2 mM). Data were fit to Eq. 4 (for 1:1 and 10:1 α₂:β₂), and Eq. 5 (for 100:1 α₂:β₂) where A and k_burst are the amplitude and observed rate constant for the burst phase, respectively, and k_linear is the steady-state turnover number (k_cat).

$$y=A\left(1-e^{k_{burst}t}\right)+k_{linear}t$$ (4)

$$y=k_{linear}t+c$$ (5)

Steady-state turnover kinetics

Assays contained a reaction mixture of 170 μL: wt-α₂ (1–10 μM), wt-β₂ (0.1–10 μM), [5-³H]-CDP (8,000 cpm/nmol, 0.5 mM), ATP (3 mM), TR (40 μM), TRR (0.8 μM) and NADPH (1 mM). Aliquots (40 μL) were quenched at 30 s, 1, 2 and 3 min in 25 μL of 2% HClO₄. Product separation was performed as described by Steeper and Steuart.²⁸

Kinetics of dimanganese tyrosyl radical RNR (NrdF2-NrdE) with NrdH exogenous reductant

NrdI was purified to homogeneity and used to assemble the dimanganese-tyrosyl radical cofactor in NrdF2. Typically, 0.7 Y•/β were obtained for cluster assembled in NrdF2. In our hands, NrdF1 was always in inclusion bodies. NrdE was purified under a variety of conditions and in all cases its solubility was limited (~10 μM). Full length NrdE, with no tag, was used in these assays. Purification of thioredoxin equivalents (TrxA, TrxB1, Trxc and NrdH) as well as thioredoxin reductase were carried out and examined for activity. NrdH was established to be the physiological reductant.

RESULTS

Effect of catalytic β₂ on inactive site-directed mutants of RNR

Since RNR is essential for E. coli viability, over-expression of either mutant α₂ or β₂ always results in small amounts of co-purifying endogenous subunits.²⁹ For example, Y₃₅₆F-β₂, a variant incapable of dNDP formation,³⁰ typically contains between 0.1–5% contaminating wt-β₂. When product formation is examined for a reaction mixture containing 0.34 nmol each of Y₃₅₆F-β₂ and wt-α₂, CDP, and ATP in the absence of a reducing system, increasing amounts of dCDP are observed with time such that 1.5 dCDP/α₂ are produced within 15 min (Fig. S1). If β₂ is not catalytic with respect to α₂, one would predict that turnover in the absence of a reducing system would result in 0.003–0.15 dCDP/α₂. The observation of a 10–500-fold enhancement over the expected number of dCDPs suggests that one β₂ can service multiple α₂s. In contrast to Y₃₅₆F-β₂, when similar experiments are performed with another catalytically inactive mutant, Y₇₃₁F-α₂,³¹ only 0.05–0.09 dCDP/α₂ are generated with no visible time dependence. This observation is consistent with 2–3% contaminating wt-α₂ and a non-catalytic role for α₂ with respect to β₂ turnover. These observations together with the observation that α₂ in the presence of excess β₂ vs β₂ in the presence of excess α₂ (Fig. S2) give different turnover numbers provided the impetus for examining the reaction of wt RNR at different α:β ratios.

dCDP formation kinetics as a function of subunit ratio in the absence of reductant

To test the hypothesis that β₂ is catalytic, dCDP was measured at α:β = 1:1, 10:1 or 100:1 (Fig. 1) with 10 μM α₂, and β₂ varied (0.1, 1 or 10 μM). The total amount of dCDP was normalized per α₂ or Y• and the rate constants calculated from the sum of A₁k₁ + A₂k₂ in Eqs. 1 and 2, where A_n and k_n represent the amplitude and rate constant of each kinetic phase (Table 1). As seen in Fig. 1, 2.5–3 dCDP/α₂ are produced at all α:β ratios. At α:β = 1:1 and 10:1, total dCDP production occurs over two kinetic phases: 1.3–1.4 dCDP/α₂ are generated in the first phase and 1.1–1.6 dCDP/α₂ in the second phase. As we previously proposed,¹⁴ the first phase reports on the generation of the first two dCDPs by each α monomer, and the second phase reports on re-reduction of the active site disulfide by the C-terminal dithiols of α₂ (Scheme 1). Support for this interpretation is provided by the ability to generate 1.4–1.7 dCDP/α₂ in a single kinetic phase in a mutant α₂ containing serines in place of the C-terminal cysteines.¹⁴ At α:β = 100:1, a single kinetic phase producing 2.5 dCDP/α₂ is observed.

Figure 1.

Product formation kinetics measured by RCQ in the absence of reductant. Final reaction mixtures contained 10 μM α₂, 0.5 mM CDP, 3 mM ATP and (a) 10 μM, (b) 1 μM, or (c) 0.1 μMβ₂ as indicated. Circles and error bars represent the mean and average error of 2 trials, respectively. Black traces represent fits to Eq. 1 (a and b) or Eq. 2 (c) and produced the rate constants and amplitudes listed in Table 1.

Table 1.

dCDP formation kinetics at varying ratios of α:β in the absence of reductant.

α2 (μM)	β2 (μM)	First phase			Second phase			kcatb (s−1)
		k1 (s−1)	dCDP/ α2	dCDP/Y•a	k2 (s−1)	dCDP/α2	dCDP/Y•a
10	10	6 (1)	1.3 (2)	1.1 (1)	0.51 (8)	1.6 (2)	1.3 (1)	8 (1)
10	1	0.8 (2)	1.4 (1)	10 (1)	0.030 (7)	1.1 (1)	11 (1)	8 (2)
10	0.1	0.046 (2)	2.5 (1)	207 (3)	–	–	–	9.5 (4)
1	1	6 (1)	1.5 (2)	1.2 (1)	0.9 (1)	1.5 (2)	1.2 (1)	8 (1)

^aY• amount used in the dCDP/ Y• calculation was obtained from the moles of β₂ (87,000 g/mol) in the reaction mixture and the amount of radical per dimer (1.2).

^bk_cat = A₁k₁ + A₂k₂, where A_n refers to the amplitude with respect to Y•.

In contrast, normalizing product formation relative to equivalents of Y• reveals that a total of 1.1, 10, and 207 dCDP/Y• are formed at α:β = 1:1, 10:1, and 100:1, respectively (Table 1). Recalling that ~60% of β₂ is active and contains 2Y•, the remaining 40% is inactive with only a diferric cluster.¹¹ Thus, a single active β₂ performs many turnovers during the same time α₂ makes 2.5–3 dCDP/α₂ (Fig. 1, Table 1). Calculating the turnover number relative to Y• reveals that product formation occurs at 8–9 s⁻¹ in all cases (Table 1). This rate constant is within the 5–10 s⁻¹ range reported in previous studies¹⁴ and is associated with the protein conformational gating that occurs prior to initiation of the reaction. As the ratio of α:β increases, the second phase (re-reduction of the α₂ active site) becomes unencumbered by the issues associated by re-reduction of the active site disulfide (Table 1, Scheme 1).

Our previous studies on wt E. coli RNR showed that the activity of each subunit in the presence of a 5-fold excess of the second subunit increased with concentration up to 0.5 μM (Fig. S2), but that by 2 μM it decreased by 50%.¹⁴ We recently reported, using size exclusion chromatography analysis, that α₂β₂ can be converted in part into an inactive α₄β₄ state at concentrations >10 μM or in the presence of the negative effector dATP.²⁷ To investigate the potential effects of the α₄β₄ state on our pre-steady state kinetics measurements, we performed the 1:1 α₂:β₂ experiment at low (1 μM), and high (10 μM) subunit concentrations (Fig. S3, Table 1). No differences were observed in dCDP formation. This result suggests that if α₄β₄ forms under these experimental conditions that it rapidly reverts back to α₂β₂.

Measurement of α₂β₂ subunit association (k_assoc) and dissociation (k_dissoc) rate constants by stopped-flow fluorescence spectroscopy

For β₂ to function catalytically, dissociation and re-association of the active α₂β₂ complex must be fast relative to the rate limiting processes. To study this possibility, we prepared β₂ that contains an environmentally sensitive fluorophore site-specifically attached to a cysteine at position 365 within its C-terminal tail that is largely responsible for the binding affinity to α₂.^13,30 This dansyl-β₂ exhibits increased fluorescence intensity when the probe is situated in a hydrophobic environment compared to its fluorescence intensity free in buffered solution, providing a method to explore subunit oligomerization.^19,27

To measure the rate constant for k_assoc for α₂β₂, stopped-flow (SF) fluorescence experiments were performed at varying concentrations of α₂ in > 10-fold excess relative to dansyl-β₂ (80 nM) for a range of [α₂] (Fig. 2a). The subunits were rapidly mixed and the increases in fluorescence, shown in Fig. 2b, were fit to Eq. 2 for each [α₂]. Plotting the resulting k_obs versus [α₂] (Fig. 2c) provided k_assoc from a linear fit to Eq. 3. These experiments were performed in the presence and absence of CDP/ATP and yielded k_assoc that were similar and range from 150–200 μM⁻¹ s⁻¹ (Table 2).

Figure 2.

Diagram describing SF fluorescence experiment to measure (a) k_assoc, where green indicates the highly fluorescent state. (b) To extract k_assoc, the fluorescence intensity was measured under pseudo-first order conditions for a range of [α₂], and fit to Eq. 2 (black traces, 1 set of 3 data sets shown). (c) k_obs (points) obtained in this way were then plotted versus [α₂], and fit (lines) to Eq. 3 to obtain k_assoc from the slope of fitted lines.

Table 2.

Association and dissociation rate constants as a function of CDP and effector measured by SF fluorescence spectroscopy.

Substrate/Effector	kassoc (μM−1 s−1)	kdissoc (s−1)
CDP/ATP	163 (7)	75 (10)
CDP	193 (7)	93 (3)
ATP	156 (14)	86 (10)
None	179 (11)	109 (6)
dATPa		0.46 (4), 7 (2)

^a50 μM, sufficient to saturate both the specificity- and activity-allosteric binding sites.

Dissociation rate constants (k_dissoc) were determined by rapid mixing of α₂(dansyl-β₂) complex (or α₄(dansyl-β₂)₂ in the case with 50 μM dATP) with a large excess of wt-β₂ (Fig. 3a). In the absence and presence of CDP and ATP, the loss in fluorescence intensity was fit to a monoexponential decay, Eq. 2, to provide k_dissoc values of 75–100 s⁻¹ (Fig. 3b, Table 2). k_dissoc of the active α₂β₂ complex is ~8–12-fold faster than the conformational change(s) (8–9 s⁻¹), supporting the proposal that a single β₂ can service multiple α₂s without attenuating turnover kinetics. In contrast, when the experiment is performed in the presence of the inhibitory effector dATP, dissociation kinetics are slow and biphasic (Fig. 3c).

Figure 3.

Scheme for SF fluorescence experiment to determine k_dissoc (a), where green indicates the highly fluorescent state and n = 1 in (b) and 2 in (c). Fluorescence decay kinetics were measured in the presence of 0.5 mM CDP and 1.5 mM ATP (b), and 50 μM dATP (c). Black circles represent averages of 15 SF traces, and lines are fits to Eq. 2 (b) or 1 (c). The data shown represent 1 of 3 independent trials, the averages and sd of which are presented in Table 2.

The oligomeric state (α₂β₂ vs α₄β₄) of E. coli RNR can be controlled by protein concentration and dATP concentrations.^3,19,32,33 We therefore also determined the subunit dissociation rate constant for α₄β₄ generated in the presence of 50 μM dATP. At these concentrations of dATP, both the specificity and activity sites of α₂ are saturated,^34,35 resulting in a shift toward the inactive α₄β₄ state.^3,36 Fitting these data to a biexponential decay (Eq. 1), each phase represents ~50% of the total amplitude change and the rate constants obtained are 10- and 100-fold slower than the corresponding dissociation rate constants from the α₂β₂ complex (Table 2). This is consistent with previous reports of enhanced subunit affinity in the α₄β₄ state.^19,32 As stated previously, α₄β₄ formation that may result from high protein concentrations (10 μM of each subunit) does not affect turnover in the absence of TR/TRR/NADPH. However, the slower dissociation kinetics of α₄β₄ could explain the range of values observed for steady state turnover in the regime where α₂ becomes limiting (Table 3).

Table 3.

Steady-state turnover numbers for different α:β ratios.

β2 (μM)	α2 (μM)	α2β2a (% β2 in complex)	kcat b (s−1)
0.1	10	98	9.0 (3)
1	10	98	7.8 (1)
10	10	96	1.45 (4)
1	1	64	4.3 (2)
0.2c	1	81	6.8 (3)

^aAssuming a K_d for α₂β₂ of 0.2 μM (ref 13).

^bTurnover number in the presence of TR/TRR/NADPH. k_cat has not been scaled for the percentage of active complex.

^cThese are our standard assay conditions to assess β₂ activity.

Steady state turnover kinetics

Studies by Ge et al. revealed that turnover number of RNR in the presence of TR/TRR/NADPH declines with increasing protein concentration (Fig. S2).¹⁴ With these data, we proposed that the conformational change(s) prior to turnover are rate-limiting under physiological subunit concentrations (0.1–0.5 μM) while some aspect of the re-reduction process governs turnover under higher protein concentrations (1–10 μM). To determine if β acts catalytically in the presence of reductant, standard RNR assays were performed at different concentrations and ratios of α₂ and β₂ (Table 3). In contrast to turnover in the absence of reductant (8–9 s⁻¹), the steady-state turnover varied from 1.4 to 9 s⁻¹. Maximal α₂β₂ activity with reductant is realized when [α₂] > [β₂]. These data combined with our previous studies¹⁴ provide further support for the catalytic role of β₂ during turnover. Under conditions of α:β = 1:1 and subsequent to complete oxidation of α₂, additional turnovers can only occur upon re-reduction by the TR/TRR/NADPH system. However, in the presence of excess α₂, a single β₂ can “rapidly” dissociate and re-associate with multiple α₂s prior to requirement for re-reduction. Catalytic β₂ could explain the discrepancy between α₂ and β₂ activities as measured by standard assays at physiological concentrations. These data further suggest that re-reduction is not only rate-limiting during steady-state turnover at high protein concentrations, but also under physiological concentrations when α₂ is limiting.

dCDP kinetics as a function of subunit ratio in the presence of TR/TRR/NADPH

Our previous studies monitoring dCDP formation in the presence of the TR/TRR system with 15 μM α₂ and 32 μM β₂, reported 1.4–1.7 dCDP/α₂ are generated in a burst phase at ~6 s⁻¹, followed by a linear phase reporting on steady-state turnover at 0.9 s⁻¹ (Table 4 (top row) and Fig. S2).¹⁴ These experiments provided evidence that re-reduction of α₂ by TR/TRR/NADPH is rate-determining under steady-state turnover at high protein concentrations. Given these results, and the studies above, we predicted that altering the concentration of β₂ in the reaction mixture would not alter the amplitude of the burst phase with respect to α₂, but that it would change when calculated relative to Y•.

Table 4.

dCDP formation kinetics at varying ratios of α:β in the presence of TR/TRR/NADPH.

[α2] (μM)	[β2] (μM)	Burst phase			Linear phase
		kburst (s−1)	dCDP/α2	dCDP/Y•	kcat/α2 (s−1)	kcat/Y• (s−1)
15	32a	6.1	1.4	0.7	0.9	–
10	10b	9 ± 2	1.9 ± 0.1	1.5 ± 0.1	1.85 ± 0.05	1.45 ± 0.04
10	0.1b	–	–	–	–	6.8 ± 0.08

^aAdapted from reference 14, 0.9 Y•/β₂.

^bCurrent work, 1.2 Y•/β₂.

RCQ experiments were therefore carried out in the presence of α:β = 1:1 ([α₂] = [β₂] = 10 μM) with the prediction that the burst phase should remain constant with respect to α₂ but minimally double with respect to Y•. We found that 1.9 ± 0.1 dCDP/α₂ and 1.5 ± 0.1 dCDP/ Y• (relative to 0.7, see Table 4 top two rows in middle column) are generated in the burst phase at 9 ± 2 s⁻¹ (Fig. 4a,). This result is similar to product formation kinetics and stoichiometry in the absence of TR/TRR where β₂ performs several turnovers, servicing all the α₂s prior to re-reduction of the active site disulfide.

Figure 4.

dCDP formation kinetics measured by RCQ in the presence of TR/TRR/NADPH. Reaction mixtures contained α₂ (10 μM), [5-³H]-CDP (1 mM), ATP (3 mM), TR (40 μM), TRR (0.8 μM), NADPH (1 mM), and β₂ at either 10 μM, (a), or 0.1 μM, (b). Data points represent the averages of two trials. Black traces represent fits to Eq. 4 (a) or 5 (b).

Finally, based on our current model, the 8–9 s⁻¹ observed rate constant (Table 1) reports on the protein conformational change(s) that gates radical propagation. To support this hypothesis, we measured turnover kinetics in the presence of TR/TRR/NADPH at α:β = 100:1. As shown in Fig. 4b, no burst phase is observed under these conditions. Instead, a linear phase for formation of product with a rate constant of 6.8 ± 0.1 s⁻¹ is observed. These results are consistent with the rate-determining step occurring prior to the first turnover.

DISCUSSION

Our studies in 1999 on the adenosylcoblamain (AdoCbl)-dependent Lactobacillus leichamannii ribonucleoside triphosphate reductase (RTPR) showing that AdoCbl can act catalytically,³⁷ were in part, the inspiration for thinking about whether β₂ could function in a similar capacity. The RTPR system lent itself to careful kinetic analysis under two sets of conditions: [AdoCbl] >> [RTPR] and [RTPR] >> [AdoCbl] in the presence of the TR/TRR reducing system. Our studies in the former case revealed that carbon-cobalt bond reformation and AdoCbl dissociation from AdoCbl•RTPR into solution can accompany every turnover, that is the radical chain length of the RTPR catalyzed dNTP reaction is ~1. In that case, the rate-limiting step is disulfide re-reduction by TR/TRR at 2 s⁻¹. Under conditions where [RTPR] >> [AdoCbl], k_cat is 15–20 s⁻¹ and is governed by the rate constant for AdoCbl dissociation. In this situation, AdoCbl services more than one RTPR, that is, it functions catalytically. The ability of RTPR to release AdoCbl after each turnover can have important implications as previous physiological studies in L leichamannii suggested that cells produce RTPR in molar excess of AdoCbl molecules taken into cells. This strategy would thus increase the effective concentration of holo enzyme and enable a higher rate of DNA biosynthesis; elevated levels of RTPR can optimize use of scarce nutrients such as AdoCbl. The turnover number of 2 s⁻¹ is very similar to the turnover number of many class I RNRs (1–2 s⁻¹) when re-reduction occurs by endogenous proteins. Thus, catalytic AdoCbl in the L leichmannii class II RNR set a precedent for catalytic β₂ in E. coli Ia RNR.

In most cases, the intracellular concentrations of the α and β subunits of class I RNRs have not been measured, although with bacterial systems, where their genes reside in the same operon, it is likely that the α/β ratio is ~1:1. However, as in the case of RTPR, in the class Ia E. coli RNR, the availability of the active form of the metallocofactor is carefully controlled based on our studies in which the amount of β₂ was varied genetically and the amount of diferric cluster and tyrosyl radical was measured quantitatively.³⁸ Future studies will be required to determine not only the amounts of RNR subunits, but the amounts of active, diferric-tyrosyl radical loaded β₂. If future studies reveal that substoichiometric amounts of Y• relative to α₂ is possible, catalytic β₂ could play a physiological role. We note, however, that regardless of the physiological role of catalytic β₂, the ability of E. coli RNR to use β₂ catalytically provides an explanation for the differences in activity of α₂ and β₂ under standard assay conditions in which each subunit is assayed in the presence of an excess of the second subunit to form the “same” active complex.

Based on studies of class Ia E. coli and human RNRs and class Ib S. sanguinis, B. subtilis, M. tuberculosis, and E. coli RNRs, there are important lessons to be learned for performing reliable assays.

(1) Both α and β subunits must be carefully characterized as well as their complexes:

α subunit characterization

α in general is very air sensitive and its oxidation is often irreversible within a few hours if a reductant such as dithiothreitol (DTT) is not always present during purification and protein storage. α is now known to reside in many distinct quaternary structures (monomer, dimer, non-canonical dimer, tetramer, hexamer, fibrils) and the basis for these changes reside, in part in its N-terminal, ATP cone domain or partial cone domains.¹⁵ Purification of α is often facilitated by appendage of an N-terminal (His)₆-tag. An altered N-terminally tagged α must thus be used with caution. The recent availability of (His)₆-SUMO-tags, which allows purification by Ni affinity column, followed by complete removal of the tag with the SUMO protease is recommended.³⁹ The C-terminal tail of all αs is always disordered and houses the two essential cysteines required for the re-reduction of the active site disulfide (Scheme 1). Thus, a tag should never be placed at the C-terminal end. Anion exchange purification has worked well for all class I RNRs that we have studied. Purification of α in the presence of small molecule reductant and under anaerobic conditions without a tag, might well avoid many problems.

β subunit characterization

β₂ is a homodimer in all class I RNRs and thus quaternary structure is not an issue. The problems routinely encountered with β₂s are the stability of the Y• in the metallocofactor and the quantity of metallocofactor inserted during recombinant expression. First, the tyrosyl radicals of the dimetallo-Y• cofactors have widely varying stability. While the E. coli Ia Y• has a half-life of 4 days at 4 °C, the human enzyme has a half-life of 30 min at 37 °C. Since β is almost always isolated from recombinant sources at an elevated level of expression, it almost never has a full complement of cofactor.⁴⁰ Thus, the cofactor must be assembled and characterized with respect to the metal content and Y• per β. The loading of β with 0.5 to 1 Y•/β, from the self-assembly process in vitro²² is distinct from in vivo studies^38,41 thus far reported; more basic biology and biochemical studies are required to understand how the cofactors are assembled in cells. In the optimized assembly of the E. coli β₂ cofactor, while there are 3.6 irons/β₂, there are typically 1.2 Y•/β₂.⁴² Sixty percent of the cluster is the active Fe³⁺₂-Y• cofactor and 40% is inactive, tyrosyl radical reduced, Fe³⁺₂-YOH. Thus, in most assays that are 1:1 α/β, β₂ always acts catalytically! The cofactor assembly has been optimized for the E. coli RNR, but each β₂ is distinct. Since the Y• is a key indicator of activity, optimizing cluster assembly for assays is essential.

α/β complex characterization

Although the α and β subunits in all class I RNRs are structurally homologous, the quaternary structures of α and the α/β complex are unique and appear to be protein concentration- and nucleotide-dependent. While α alone has been found as noted above in many states, it also forms inactive complexes with β (α₄β₄) in E. coli¹⁹ and in a helical α/β fibril in B. subtilis.⁴³ Some have proposed that α₆/β₂ forms exist in S. cerevisiae and human RNRs,⁴⁴ but in our opinion, the activity of this state is in question. Not much is known about the equilibria of these species and how other factors in cells affect them. Thus, the conditions used to study RNRs must be carefully described.

K_d for subunit interactions

The K_d for subunit interaction is also essential to determine. If the K_d is weak, (0.2 μM) and the protein is diluted during purification or in crude lysates, activity can be lost because the active complex does not form or the subunits dissociate and the active complex cannot reform. An important issue to consider is that K_dS are challenging to determine, if one does not understand the quaternary structure of each subunit.

(2) Redox reductants to cycle RNRs for multiple turnovers must be identified:

Given the requirement for re-reduction of the disulfide in α accompanying formation of each dNDP, the endogenous reductant in each organism needs to be determined. DTT is often used to reduce previously unidentified αs because the protein reductant(s) are initially unknown. For this case, the efficiency varies widely and is dependent on the concentration of reductant.⁴⁵ As almost all organisms have multiple candidates for α reduction including TR/TRR, glutaredoxin/glutaredoxin reductase/glutathione, NrdH, etc., identification of the most efficient system is very important to obtain the most robust assay for inhibitor(s) discovery.

(3) Assay conditions musts have optimized α:β and reductant ratios:

Knowledge about the RNR and the organism from which it is obtained is a pre-requisite for the best experimental design. Our experiences with class Ia RNRs (E. coli, human, and S. cerevisiae) and class Ib RNRs (S. sanquinis, B. subtilis, and M. tuberculosis) have suggested the following general protocol. First, assume that the ratio of α:β is 1:1 and that α₂β₂ is the active form. We believe this is a reasonable assumption based on the recent structures of an active E. coli Ia RNR¹⁵ and structures of B. subtilis⁴³ and S. typhimurium⁴⁶ Ib RNRs. Second, determine the concentration of RNR in the organism of interest and identify the endogenous reductant. Third, measure the activity of each RNR over the protein concentration range from 0.5 to 5× estimated physiological levels. The results from some recent studies on S. sanquinis,⁴⁷ B. subtilis^39,48 and M. tuberculosis dimanganese-tyrosyl radical RNRs using this approach are shown in Fig. S4. The apparent K_m values obtained from these analyses are lower (K_m of 6.4 nM, 0.025 μM, and 0.16 μM, respectively) than a similar experiment with E. coli and human Ia RNRs (apparent K_m of 0.2 μM) suggesting “tighter” α/β affinity. However, efforts to isolate active α/β complexes of these Ib RNRs using several biophysical methods (size exclusion chromatography and analytical ultracentrifugation) all revealed that the subunit behavior, as with Ia E. coli RNR, is dynamic. Our studies on M. tb RNR have shown that it uses a dimanganese-tyrosyl radical cofactor, possesses high activity and behaves similarly to the B. subtilis and S. sanguinis enzymes, in contrast with the recent report of Fe-loaded NrdF2.⁴⁹ In the case of the B. subtilis RNR, the assay approach described above appears to avoid the known complexities associated with complex quaternary structure⁴³ in this system.

(4) The rate limiting steps for single vs multiple turnovers of RNR must be considered:

Studies on E. coli Ia RNR have shown that the rate limiting steps for single turnover is described by a rate constant of 8–9 s⁻¹ whereas multiple turnovers of RNR yields a rate constant of 1–2 s⁻¹. Thus, experiments on RNR are typically performed under two general conditions: one in the steady state with endogenous reductant and the other under “single” turnover in the absence of a reducing system. In the former case, low protein amounts are sufficient and the protocol described in (3) is recommended in search of small molecule RNR inhibitors. Once an inhibitor is identified, the target of the inhibition (tyrosyl radical reduction in β, active site modification in α, altered quaternary structure of α or α/β, etc.) can be identified using the “single” turnover approach. In this case, elevated levels of protein are required so that the method of analysis can detect a single equivalent or less of intermediate or product.

CONCLUDING REMARKS

We highlight here that studies on newly discovered RNRs must identify the system and describe the assay conditions carefully. Heroic studies are now turning to examining RNR in vivo owing to the importance of inhibiting the enzyme for development of antibacterial and anticancer therapies. Considering the complexity of in vitro RNR assays, as shown herein for E. coli class I RNR, the design of in vivo assays⁵⁰ for such studies will be extremely challenging. Understanding RNR’s properties is important for uncovering its role in nucleotide metabolism in general, and key for development of assays for new RNR inhibitors in vitro and in vivo.

ABBREVIATIONS

RNR

class Ia ribonucleotide reductase from E. coli

SF

stopped flow

RCQ

rapid chemical quench

TR

thioredoxin

TRR

thioredoxin reductase