Ribonucleotide Reductases (RNRs): Structure, chemistry, and metabolism suggest new therapeutic targets

Abstract

Ribonucleotide reductases (RNRs) catalyze the de novo conversion of nucleotides to deoxynucleotides in all organisms, controlling their relative ratios and abundance and in doing so play an important role in fidelity of DNA replication and repair. RNR’s central role in nucleic acid metabolism has resulted in five therapeutics that inhibit human RNR. In this review we discuss the structural, dynamic, and mechanistic aspects of RNR activity and regulation, primarily for the human and E. coli class Ia enzymes. The unusual radical-based organic chemistry of nucleotide reduction, the inorganic chemistry of the essential metallo-cofactor biosynthesis/maintenance, the transport of a radical over a long distance and the dynamics of subunit interactions all present distinct entry points toward RNR inhibition relevant for drug discovery. Our current mechanistic understanding of small molecules that target different elements of RNR function are described, including downstream pathways that lead to cell cytotoxicity. We conclude by summarizing novel and emergent RNR targeting motifs for cancer and antibiotic therapeutics.

Introduction.

The availability of adequate and balanced deoxynucleotide pools is essential for accurate DNA replication and repair and consequently genome stability. Deoxynucleotides are supplied universally in all organisms by a de novo pathway catalyzed by ribonucleotide reductases (RNRs) that convert RNA building blocks to DNA building blocks (1–3). Deoxynucleotides can also be generated in an organism-, environment-, and disease-specific fashion by nucleoside (or nucleotide) salvage pathways (4). Our current understanding of the unique organic (5) and inorganic chemistry (6) of RNRs, have been revealed, in part, by our understanding of clinically used therapeutics that target the universal radical-mediated nucleotide reduction mechanism, and the specific metallo-cofactor biosynthetic and repair pathways. The ensemble of studies led to the first structures of class I RNRs at low resolution (7–10), and more recently, to high resolution structures in trapped active and inhibited states (8, 11, 12). These new studies suggest, in combination with inhibitors of specific signaling pathways downstream of RNR, that the time is right to revisit RNR as a target for antibacterial, antiviral, as well as anticancer agents.

All RNRs catalyze the conversion of nucleoside diphosphates (NDPs) or triphosphates (NTPs) to deoxynucleotides (dNDP or dNTP, Figure 1A). The RNR’s share a common active site architecture located in subunit α that houses 3 essential cysteines (Figure 1B) (13–15). Two cysteines (bottom face) provide the reducing equivalents to make dNDPs and the third cysteine (top face) is transiently oxidized to a thiyl radical (−S•) that initiates NDP reduction (16). Distinct metallo-cofactors catalyze this oxidation (Figure 1C) and they are the main basis for RNR classification (Ia-e, II, III), though a recently discovered non-metallo-cofactor, 2,3-dihydroxyphenylalaninine radical (DOPA•) breaks this paradigm (17–19). This review focuses on the class I RNRs that share a distinct mechanism by which a transient thiyl radical is generated, and whose formation requires a second subunit β that houses the cofactor oxidant (Figure 1C).

Figure 1.

A RNRs catalyze the conversion of nucleoside di- or triphosphates, ND(T)Ps, to deoxynucleoside di- or triphosphate, dND(T)Ps. B The reduction occurs in the active site in subunit α composed of a 10 stranded β barrel with three cysteines and the conserved placement of the oxidant (gray circle, panel B) involved in thiyl radical formation (−S•, top face in A) that initiates NDP reduction. The bottom face thiols in A deliver the reducing equivalents and themselves become oxidized. C The oxidants are distinct among the RNR classes (I, II and III) represented here by a gray circle that is juxtaposed with the thiyl radical loop. Substrate and four essential residues, including the three essential cysteines and E₄₄₁, are shown as sticks. C The class Ia RNRs use a diferric-tyrosyl radical (Y•) cofactor (M₁, M₂ = Fe³⁺) that is located in subunit β (left, bottom) to regenerate a radical species in the active site in subunit α. The oxidation occurs over a distance of ~33 Å by long range radical transfer to first generate a Y• in subunit α (under the gray circle), and second generate -S• on an adjacent cysteine (top face in A). In other class I RNRs (Ib-Ie) the oxidation also occurs by long range radical transfer across α and β, but involves distinct metallo-oxidants (X, M₁, M₂). In the case of the class II and III RNRs the oxidants, the 5′-deoxyadenosyl radical generated from adenosylcobalamin (class II) and the glycyl radical (class III) generated from S- adenosylmethionine and an FeS cluster, are located adjacent to the cysteine to be oxidized (gray circle). A = adenine base.

Docking model and Radical Transfer (RT) pathway.

Reichard and coworkers in 1969 discovered the class Ia E. coli RNR and proposed that active enzyme is an α2β2 complex (20, 21). However, it wasn’t until 1994 that Eklund et al. (13) reported the X-ray structure of α2 (Figure 2B), which together with their earlier structure of β2 (Figure 2A) led to a symmetrical “docking model” based on subunit shape complementarity (Figure 2C). This model has guided experimentation until recently. A fascinating feature of the docking model is that the diferric-tyrosyl radical cofactor (Fe³⁺₂-Y₁₂₂•, Figure 2C) in β is ~ 35 Å away from C₄₃₉ (E. coli numbering), which is oxidized in the α subunit. A turnover frequency for dNDP production of 2 to 10 s⁻¹, together with the long distance between Y₁₂₂• and C₄₃₉, engender a RT pathway (7, 14): Y₁₂₂•[β] ⇌ [W₄₈[β]] ⇌ Y₃₅₆[β] to Y₇₃₁[α] ⇌ Y₇₃₀[α] ⇌ C₄₃₉[α] (Figure 3, noting that [W₄₈] involvement has not yet been demonstrated).

Figure 2.

Structural models of class Ia RNR from E. coli. X-ray structures of A β2 (15), B α2 (13), C the Eklund docking model of α2β2 (13), and D a cryo-EM structure of an active α2β2 with two mutations in β2: F₃Y₁₂₂/E52Q (11). A β2, a homodimer in red/orange with disordered C-terminal tail residues (dashed lines 341–375 E. coli). B α2, a homodimer in light and dark blue with disordered C-terminal tail residues (dashed lines 737–761) that houses the two cysteines (red balls) which re-reduce the active site disulfide formed on NDP reduction (Figure 1A). α2 also houses the A-site (activity site or cone domain) that binds ATP (that activates RNR) or dATP (that inactivates RNR) in green; the C-site (catalytic site that binds CDP, UDP, GDP, and ADP) in magenta; the S-site (specificity site that binds the effectors dATP, ATP, TTP dGTP) in yellow. C Docking model of α2β2 with the long-range radical transfer pathway (left) (14). Shown also is a peptide in gray (residues 360–375 of β2) proposed to represent the tail of β2 responsible for α2 binding. D Asymmetric complex formed when F₃Y₁₂₂/E52Q-β2, interacts with α2, GDP, and TTP. 3.6-Å resolution cryo-EM density shown in transparent gray. This structure of the active α2β2 can be compared with the symmetric docking model in C.

Figure 3.

Proposed radical transfer pathway in Ia RNRs (E. coli numbering) within the α2β2 complex. Binding of substrate and allosteric effector (not shown) to α2, triggers radical transfer from the Y₁₂₂• of the Fe³⁺₂-Y₁₂₂• cofactor in β, through three transient Y•s (356-β across the subunit interface to 731-α and 730-α). The Y₇₃₀•- α (under the gray circle in Figure 1B) then oxidizes the active site cysteine to a −S• that initiates NDP reduction (Figure 1A). Subsequent to dNDP formation, the Y₁₂₂• is regenerated by reverse radical transfer. W₄₈-β is bracketed as its role in the pathway has not been established. Each step in the pathway is proposed to involve distinct proton (H⁺, blue arrows) coupled electron (e⁻, red arrows) transfer (PCET) steps indicated by blue and red arrows (7).

Rate limiting physical step(s) mask both the NDP reduction and the RT chemistry. These processes are conformationally gated by proper substrate and effector binding to α2 and its association with β2 (22). The “stable” Y₁₂₂• in β2 is transiently reduced and re-oxidized on each turnover, and RT through the pathway involves distinct proton-coupled electron transfer (PCET) steps at each pathway residue (Figure 3). The first step in RT is proposed to occur at the metal cofactor in β2, triggered by substrate and effector binding in α2 more than 35–40 Å away. Studies using site-specifically incorporated tyrosine analogs with altered reduction potentials, high field, multifrequency, electron paramagnetic resonance (EPR) methods, structural analysis, and RT photoinitiation in photosensitized RNRs (photo-RNRs) have provided insight into each of the proposed steps (7). Our current understanding of this pathway suggests that the thermodynamic landscape of the radical transfer process (Y₁₂₂• to C₄₃₉) is uphill by greater than 200 mV, and that the NDP reduction reaction, which also involves an uphill 3′-hydrogen atom abstraction, is driven to the right by rapid and irreversible loss of water during NDP reduction (Figure 1A). This pathway design avoids buildup of the highly reactive protein radical intermediates such as Y•, which has a reduction potential of 0.96 V vs the normal hydrogen electrode). Reduction of any of the Y• intermediates in the pathway (Figure 3) would inactivate RNR ending in catastrophic consequences for the organism (23, 24). Accordingly, the RT pathway provides a target of opportunity for future drug design.

Evidence for the docking model (Figure 2C) has been provided by trapping pathway radicals using tyrosine analogs with perturbed reduction potentials (Figures 3 and 4A), and active site radicals in α2 using MBIs (Figure 4B) (25). As briefly summarized, pulsed electron-electron double resonance (PELDOR) spectroscopy and negative stain electron microscopy (EM) have allowed spectroscopic analysis of these trapped α2β2 complexes.

Figure 4.

Structures of unnatural amino acids and nucleotide analogs used to study class Ia RNRs. A Unnatural amino acids that have been site-specifically incorporated in place of the tyrosines or cysteine within the radical transfer pathway (Figure 3) and their reduction potentials vs. tyrosine for NO₂Y, DOPA, NH₂Y, F_nYs and vs. cysteine for seleocysteine (Sec) at pH 7 (24). 3-Aminotyrosine (NH₂Y) is 590 mv easier to oxidize than Y. Fluorinated Ys (F_nY where n = 2 or 3) have allowed for tuning of the reduction potential over 170 mV depending on the number of Fs and their substitution pattern. B Nucleoside 5′-diphosphate are irreversible and reversible inhibitors of RNR. The irreversible inhibitors are mechanism based as the 3′ C-H bond (red) of the inhibitor must be cleaved as with the normal substrate (Figure 1A), before distinct radical chemistry in each case occurs that causes enzyme inactivation.

3-Aminotyrosine (NH₂Y)-RNR (Figure 4) and PELDOR analysis: low resolution evidence for the docking model.

NH₂Y is easier to oxidize than Y by 590 mV (Figure 4A). When NH₂Y replaces a pathway Y in α or β and is incubated with the second subunit, substrate, and specificity effector, it functions as an efficient radical trap forming a 3-aminotyrosyl radical (NH₂Y•) (10, 25) that is unable to oxidize the next residue in the pathway. Under these conditions, 0.5 equivalents of total Y₁₂₂• in β2 is reduced and a stoichiometric amount of NH₂Y• is formed in (α/β pair, one side). Because the chemistry in one α/β pair is incomplete, 0.5 equivalents of Y₁₂₂• resides in the adjacent α/β pair (on the other side), which is unable to carry out chemistry. These results require asymmetry within α2β2 and are described as half-sites reactivity.

In these trapped complexes, PELDOR spectroscopy has been used to measure the distance between the NH₂Y• (located at 356-β or 731-α or 730-α) in one α/β pair and the Y₁₂₂• on the adjacent pair, as shown in Figure 5A. Studies with the MBI N₃CDP (Figure 4B), which forms a nitrogen centered radical (N•) covalently bound to a cysteine in the active site of α2, also allows a distance measurement. Additionally, RNR mutants in which Y₁₂₂• is replaced with 2,3,5-F₃-Y₁₂₂• or NO₂Y₁₂₂• (Figure 4A), which are “hotter” oxidants than Y₁₂₂•, also generate pathway radicals and demonstrate half-site reactivity (26). Distances measured to date are summarized in Table 1 and they are consistent with the docking model (Figure 2C) (25, 26). An unexpected outcome of these experiments was that when radicals are trapped within the pathway, the α2β2 complex exhibits increased subunit affinity, thus enabling its isolation. A number of these complexes have been examined by negative stain EM and have revealed structures resembling the docking complex (15–30 Å resolution (Figure 5B) (7). Our current model from these studies with perturbants is that for wild type RNR, radical transfer and catalysis occur initially within one α/β pair of an asymmetric α2β2, which triggers rapid chemistry within the second α/β pair. The mechanism of switching remains to be determined.

Figure 5.

Support for the Eklund docking model (Figure 2C). A PELDOR spectroscopy used to measure distances between Y₁₂₂• in the unreacted α/β pair (right) and the trapped radicals (NH₂Y• or N•) in the reactive α/β pair where Y₁₂₂ is reduced (YOH, left). B Representative negative stain EM 2D class averages of the structures of the NH₂Y₇₃₀• trapped in α2β2 complex (25). The view with the yellow star resembles the Eklund docking model shown in Figure 2C.

Table 1.

Pathway radicals trapped in E. coli class Ia RNR by site-specifically incorporating unnatural amino acids or reaction with N₃UDP; PELDOR distances (< ± 1 Å) are given in last column. ● corresponds to forward RT and ● to reverse RT, as indicated by the direction of the arrows on the pathway shown below the table.

α	β	Y122	Y356	Y731	Y730	C439	NDP	Y122•-X• distance (Å)
WT	WT	●/●	-	-	-	-	-	33
WT	DOPA356	-	●	-	-	-	-	30
WT	NH2Y356	-	●	-	-	-	-	30
Y731F	F3Y122	-	●	-	-	-	-	30
NH2Y731	WT	-	-	●	-	-	-	38
NH2Y731/R411A	WT	-	-	●	-	-	-	35
NH2Y730	WT	-	-	-	●	-	-	39
WT	WT	-	-	-	-	-	N3UDP●	48
WT	F3Y122	-	●	-	-	-	-	30
WT	NO2Y122	-	●	-	-	-	-	30

Higher resolution structures of RNRs

Inhibited structures in vitro.

dATP is a universal inhibitor of all class Ia RNRs. It binds to the N-terminal domain of α (cone domain, green, Figure 2B). Two independent studies by the Dealwis and Walt groups (2011) (8) and Drennan and Asturias groups (2018) (27) revealed structures of eukaryotic dATP-inhibited states. Fairman et al studying S. cerevisiae RNR observed crystallographically, an α6 hexameric ring structure (6.6 Å). Brignole et al studying human α with CDP, dATP and a small amount of ATP, observed by cryo-EM, a similar, but higher resolution hexameric ring structure (Figure 6B, 3.3Å). Both groups also reported negative stain EM studies (28 and 30 Å) (8, 27) of a dATP inhibited state in the presence of both α and β (1:1). Despite the stoichiometry of the subunits, in both cases, less than 1β2/α6 was observed. Both eukaryotic dATP inhibited states are a trimer of dimers with the cone domains responsible for the dimer interfaces. In addition, with the human RNR, small-angle X-ray scattering data on this state suggested that β2 could not enter the hole in the α6 ring structure, and hence that an active α2β2 is not accessible (28).

Figure 6.

Structures of dATP inhibited states of class Ia RNRs. A X-ray structure of dATP inhibited E. coli class Ia RNR (29) is an α4β4 ring structure with a hole in the middle, composed of alternating α2 (light and dark blue with the cone domains in green) and β2 (orange and red) subunits. Note the importance of the cone domain in the α/β interaction. B Cryo-EM structure of dATP inhibited human class Ia RNR (27) is a hexameric α6 ring with a hole in the middle. A subunits are in light blue and dark blue with cone domain in light and dark green and a three-helix insertion in purple (residues 638–681). Note the importance of the cone domain in the α/α interactions. C) Cryo-EM structure of α/β (1:1)clofarabine triphosphate (ClFTP) inhibited human class Ia RNR (27). Top panel is a representative cryo -EM 2D class average image generated from α, β and clofarabine triphosphate (ClFTP) that shows β (arrow) interacting with α. The middle and bottom panels are two views of the 3D reconstruction of the same data set. The bottom is rotated 90° from the middle image. Only a fraction of the α6 rings in these images have a single and variably positioned β.

The structure of the dATP inhibited E. coli Ia RNR generated from α:β (1:1) in the presence of dATP is distinct from its eukaryotic counterparts. Drennan and collaborators using a variety of biophysical methods, reported an α4β4 ring complex with alternating α2 and β2s and a hole in the center (Figure 6A) (29, 30). In this structure, in contrast with the eukaryotic inhibited state, the cone domain (green) interacts with β2. The most intriguing result is that the distance between Y₁₂₂•-β2 and C₄₃₉ in α2 (Figure 2C) has increased from 35 to 60 Å, shutting down radical transfer and consequently nucleotide reduction. Thus, despite the distinct quaternary structures of the dATP-inhibited states, a common mechanism of inhibition emerges that involves inability of β2 to form an active α2β2 state.

Inhibited structures in vivo.

The presence of these inhibited states in cells (Figure 6) are important to establish. Drennan et al. used their structural insight from the E. coli α4β4 complex and site-directed mutagenesis to disrupt the α2-cone domain-β2 interface. Activity assays and negative stain EM analysis of several mutants showed that dATP no longer inhibited RNR and, that no α4β4 was detected (31). This study, in concert with genetic experiments on E. coli using a random mutagenesis protocol, a screen for altered dNTP pools, and genome sequencing, identified RNR with mutations at the same interface (32). The biochemical and genetic studies together suggest that the dATP-inhibited state of E. coli α4β4 occurs in vivo. Studies with clofarabine (ClF) (33) and other nucleoside therapeutic (cladribine (ClA) and fludarabine (FlU)) inhibitors of RNR (Figure 4B) (34) have demonstrated α6 formation in several human cell lines treated with sublethal doses of the nucleosides (35). The distinct inhibitory structures of the Ia RNRs (Figure 6) will be discussed further in the section on mechanism-based and reversible inhibitors as a new strategy to target RNRs.

Toward Active α2β2 structures.

Our studies using fluorinated tyrosine analogs (Figure 4A) combined with bioinformatics and the docking model of α2β2 (Figure 2C) to identify residues within the α/β subunit interface including E52 in β, led us to investigate the E. coli double mutant of β2 (E52Q and F₃Y₁₂₂•) (11). The F₃Y₁₂₂ substitution allowed trapping of the Y₃₅₆• on the RT pathway (Figure 3) resulted in a tighter subunit affinity, K_d < 0.4 nM as compared with 0.2 μM for the control with E52Q-β2 (36). Although E52Q-β2 when incubated with any substrate and effector resulted in completely inactive RNR, incubation of the double mutant β2 (E52Q and F₃Y₁₂₂•) with α2 (or His₆-α2), GDP, and TTP produced 0.5 equivalents of Y₃₅₆• and 0.5 equivalents of dGDP consistent with half-sites reactivity (36). The resulting α2(E52Q and F₃Y₁₂₂• β2) complex gave rise to a near atomic resolution (3.3 to 5 Å) cryo-EM structure shown in Figure 2D which is asymmetric, consistent with the half-sites reactivity. In line with the biochemistry, the structure of α (left, Figure 2D) has generated a disulfide in the active site that we presume gives rise to the dGDP. With the α/β pair on the right side of the complex in Figure 2D, residues 341–375 in β are visualized for the first time (Figure 2A). In addition, GDP/TTP are apparent and the location of Y₃₅₆ is finally revealed for the first time as part of the entire radical transfer pathway (Figures 2C and 3) (11). The details of this structure have recently been submitted for publication. Our ability to trap radicals at different residues within the pathway summarized in Table 1 and the increased subunit affinity observed under these conditions, suggests that this approach may lead to additional cryo-EM structures that will provide insight about the dynamics of this amazing machine.

New mechanistic insight about the chemistry of NDP reduction

Model for disulfide re-reduction and conformational gating.

In the E. coli RNR, the rate-limiting step(s) are physical (Figure 7 step B, E and G), involving conformational changes that mask the chemistry of long-range reversible RT and dNDP formation (22, 37). The observed rate constants (k_cat) for dNDP formation in the absence of an external reductant (steps A through E), range from 2 to 5–10 s⁻¹ with substrate and substrate/effector, respectively. In the presence of the physiological reductants (in cells or in steady state assays), additional conformational changes become rate limiting (1–2 s⁻¹, step H) and likely involve either α/β subunit dissociation, conformational changes associated with the re-reduction of the active site disulfide by the C-terminal tail of α (steps F, G, H), or both, and are protein concentration dependent. Different αs have distinct cysteine configurations within their C-terminal tails (red balls in the C-terminal tails in Figure 2B) and require organism-specific reductants (e.g. thioredoxin (TR), NrdH, glutaredoxin (Grx), and thioredoxin reductase (TRR) and glutaredoxin reductase) (38–41).

Figure 7.

Resetting RNR for single and multiple turnovers. The model assumes a 1:1 α to β ratio, that Y₁₂₂• is distributed equally between each β, and that the wt-α2β2 complex is asymmetric. In the absence of external reductants, 2 dCDPs are generated at 2 (substrate only) to 5–10 s⁻¹ (substrate and effector) that arise from chemistry at each α of α2 (step E). Steps B and E are rate-limiting and conformationally gated. In an assay in the absence of an external reductant, two additional dCDPs are formed at 0.1 s⁻¹ (step G). In the presence of an external reductant such as thioredoxin (TR) and TR reductase (TRR), under steady state conditions, step H becomes rate-limiting.

To unmask the mechanistic details of dNDP formation, multiple methods have been employed including site-directed mutagenesis, insertion of unnatural amino acids (Figure 4A), photosensitized RNRs (photo-β2), and MBIs (Figure 4B). In subsequent sections, we describe how this information combined with structural studies have provided insight into new therapeutic strategies.

NDP reduction mechanism.

Our current mechanism for nucleotide reduction is shown in Figure 8 (5, 43). The important features are that a −S• (C₄₃₉, E. coli) initiates the reduction by removal of the NDP 3′-H (step A) (44). E₄₄₁ facilitates this step by functioning as a base catalyst for 3′-OH deprotonation (1 to 2) (45). This reaction is driven to the right by the rapid, irreversible loss of water catalyzed by C₂₂₅ (step B). The proposal for the reductive half-reaction (steps C and D) is that the 3′-keto-2′-radical [3] is reduced by PCET to generate the 3′-ketodeoxynucleotide and the three electron, disulfide radical anion [4]. This species then reduces the 3′-ketone by another PCET step (step D) where the proton is supplied by the protonated E₄₄₁. In the last step, the H-atom abstracted from the 3′ position of NDP is returned to the same position to form dNDP and the C₄₃₉•, that reoxidizes Y₁₂₂ in β2 on each turnover.

Figure 8.

Mechanism of −S• mediated NDP reduction by most RNRs (5). Steps A–F are the proposed steps in the reaction and 1–6 represent the active site participants in each step. Note that steps A, C, D, and E involve either hydrogen atom transfer (HAT) or proton coupled electron transfer (PCET) processes.

Role of multiple thiyl radicals.

The role of the −S• initiator (C₄₃₉, Figure 8) was previously established based on studies with the class II, adenosylcobalamin-dependent ribonucleotide triphosphate reductase. An exchange coupled −S•-cob(II)alamin species, detected by EPR and UV-vis absorption stopped flow spectroscopies, was shown to be chemically and kinetically competent in deoxynucleotide formation (46, 47). The structural homology and conserved residues in the active site of all RNRs (Figure 1B) have thus been used to infer the universal involvement of −S• in initiating 3′-H atom abstraction (16). Support for −S• involvement in the reductive half-reaction (Figure 8 step C and D) comes from studies with E441Q-α/β/CDP/TTP where a disulfide radical anion was spectroscopically identified due to the absence of a required proton from E₄₄₁ for the PCET step D (48). Although −S• chemistry has been proposed for many enzymatic reactions, RNR is the only enzymatic system where this intermediate has been detected (46, 47).

PhotoRNRs unmask rate constants for NDP reduction chemistry.

The development of methods to uncouple conformational gating and unmask chemistry has allowed unprecedented insight into the active site chemistry, including thiyl radical mediated hydrogen atom abstraction (Figure 8, step A) (49, 50) and the subsequent rate-limiting 3′-ketodeoxynucleotide reduction (step D, 4 to 5) (23,26). In the former case, we designed a method for photosensitization of RNR (Figure 9A) in which a photooxidant, bromomethylpyridyl rhenium(I) tricarbonyl phenanthroline ([Re]), is covalently attached to a single surface-exposed cysteine in S355C-β2 mutant; the Y₁₂₂• in β2 is reduced and Y₃₅₆ in β2 is replaced with a fluorinated tyrosine (F_nY_356, Figure 4A). This photoβ2 in complex with α2, substrate, and effector, can be rapidly (ns) oxidized to a F_nY₃₅₆•-β2 state upon illumination (Figure 9B). The photochemically generated radical rapidly equilibrates with the radical transfer pathway in α2, ultimately oxidizing C₄₃₉, initiating cleavage of the 3′ C-H bond of NDP. Comparison of the F_nY₃₅₆• decay, observed by transient absorption spectroscopy, in the presence of 3′-[¹⁽²⁾H]-CDP established a lower limit for the −S• mediated H-atom abstraction (step A) of 1.3 × 10⁴ s⁻¹ and an isotope effect of ≥7 (51)! Note that the k_cat for RNR is 2–10 s⁻¹. The RT chemistry is thus very fast (!) and unmasked for the first time using this method.

Figure 9.

Use of photoRNRs to unmask rates of chemical steps. A Schematic of photoβ2 with rhenium photooxidant [Re] attached covalently to C₃₅₅ and Y₁₂₂OH-β2 (that is, Y₁₂₂• is reduced) complexed with CDP, ATP, and α2 (49). In the case shown, Y₃₅₆ in β is replaced with the unnatural amino acid 2,3,5-F₃Y₃₅₆. (Figure 4). B Light initiates the reaction and the presence of the flash quencher, Ru(NH₃)₆Cl₃, prevents charge recombination (left) and generates [Re^II]-F₃Y-O⁻ (right), that rapidly drives 2,3,5-F₃Y₃₅₆ oxidation to the 3,5-F₃Y₃₅₆• that initiates chemistry within α2. Different mechanisms of oxidation shown are electron transfer (ET), PCET, proton coupled electron transfer, and HAT, hydrogen atom transfer.

The subsequent rate-limiting 3′-ketodeoxynucleotide reduction (step D, Figure 8) has been examined by incorporating tyrosine analogs with altered reduction potentials in place of Y₁₂₂ in β. Use of these “hotter” oxidants drives radical transfer and also uncouples conformational gating. Specifically, F₃Y₁₂₂• and NO₂Y₁₂₂• (vida supra) have higher reduction potentials than the native Y₁₂₂•(β2) by 80 and >200 mV, respectively (23, 26), as determined from the independent measurements of formal reduction potentials in the small 3-helix bundle protein (24). These β2 mutants have been studied in an effort to observe the slow step(s) within the proposed chemistry, specifically the PCET reduction of the 3′-ketone by the disulfide radical anion (Figure 8, 4 to 5, step D). When NO₂Y₁₂₂•-β2 (or F₃Y₁₂₂•-β2) is mixed with α2/CDP/ATP, dCDP formation occurred at ~150 s⁻¹ (30 s⁻¹). This rate constant is similar to the 50 s⁻¹ measured for dCTP formation; the latter is catalyzed by class II RNR (52) where the active site chemistry is not masked by physical steps and is much faster than the wt turnover of 1–2 s⁻¹.

MBIs and reversible inhibitors to understand mechanism and design of new therapeutics.

The MBIs 2′-halo (X)-2′-deoxyNDPs (XNDP, X = Cl, F; Figures 4B irreversible and Figure 10) have played a pivotal role in our current understanding of the mechanism of nucleotide reduction (53). In 1976 Thelander and Eckstein et al. reported that ClCDP incubated with E. coli RNR resulted in time-dependent release of Cl⁻ and cytosine, and that the α subunit was inactivated (54). These observations provided the impetus for studies using labeled nucleotide analogs, which led to the general model for inhibition shown in Figure 10. As with the NDP substrate, the −S• abstracts the 3′-H (step A) to generate 2. The outcome of the reaction depends on whether and how the loss of X at 2′-C is catalyzed by the enzyme. From 3, the 2′-delocalized radical can be reduced from the top face by H-atom transfer mediated by C₄₃₉ (red H) or the bottom face (blue H) facilitated by C₂₂₅. With Cl(F)NDP, a 3′-ketodeoxynucleotide is generated (7) that dissociates from the active site (when X is not protonated). Intermediate 7 can decompose on a minute time scale to nucleic acid base, pyrophosphate (PP_i), and a furanone (8) that non-specifically alkylates the α subunit. If the reduction of 3 is by C₄₃₉, then reverse RT can effectively regenerate the Y₁₂₂• in β2. However, if reduction occurs from the bottom face, Y₁₂₂ remains reduced and β2 is inactivated. Thus, α and/or β can be inactivated via distinct mechanisms. With both XNDPs (X = Cl, F), if X is protonated (XH, step B above arrow), then dNDP is formed. The details of RNR inactivation in vitro and in vivo depend on the identities of substrate and effector, leaving group (X), and the reductant. In all cases, α inactivation requires Y₁₂₂• reduction. These inhibitors inactivate class I, II, and III RNRs by a common mechanism suggesting similar active sites (Figure 1B). The involvement of the α C-terminal cysteines in enzyme inhibition (Figure 2B, red balls) is not well understood as their covalent linkage to 8 is reversible, precluding isolation and characterization of alkylated α.

Figure 10.

Generic mechanism for 2′ X-dNDP (X = F, Cl, N₃, F₂ (Figure 4)) mechanism-based inhibition of RNRs with loss of X⁻ in step B (43, 53). 3 is formed with bottom face protonation states of the thiols unknown, which can be reduced from the top face by SH to produce 7 (right) or the bottom face to produce 7 (left) which dissociates from the active site and then decomposes to generate the products (PP_i. Base and 8). 8 can alkylate the α subunit. Alternatively, if in conversion of 2 to 3 (step B), XH is eliminated, and the normal product dNDP is formed.

In contrast to XNDPs (X = Cl, F), there are a number of MBIs (X = N₃, F₂, VF, Figure 4B) (55–57) that share similar chemistry in steps A and B (Figure 10), but then undergo distinct chemistry controlled by X and the active site cavity. Unraveling the mechanism by which N₃NDP inactivates all RNRs defined the strategy to study the mechanism of action of the clinically used nucleoside therapeutics, gemcitabine (F₂C) and clofarabine (ClF). F₂CDP is an irreversible inhibitor (58) and, in contrast to expectations, ClFDP and ClFTP are reversible, non-covalent inhibitors (Figure 4B) (59–61).

2′-Azido-2′-deoxynucleotide (N₃NDP).

N₃NDPs (N = C (shown in Figure 4B), U or A) are substoichiometric MBIs that were also first reported by Thelander and Eckstein (54). Extensive studies with N₃UDP revealed that its incubation with α2β2 resulted in rapid loss of ~90% RNR activity concomitant with loss of only 0.5 equivalents of Y₁₂₂•. The Y• loss was biphasic with the fast phase accompanied by formation of a nucleotide-based nitrogen centered radical (N•) derived from the N₃ moiety that was structurally identified using isotopically labeled N₃UDPs and EPR methods (Figure 11) (62, 63). The N• species then slowly decomposes to form nucleoside base (blue N in Figure 11), PP_i, and 8. The α/β subunits dissociate, and subsequent to α2β2 complex reformation, more Y• is lost and N• formed. While these observations were perplexing at the time, the many recent examples of “half-site” reactivity (Table 1) and RNR asymmetry (Figure 2D), now place these observations on firm footing.

Figure 11.

N₃NDP (Figure 4B) is a potent inhibitor of all Ia RNRs (62,63). Studies of this inhibitor provided a glimpse of unprecedented chemistry associated with reactive radical species in an active site cavity and the challenges associated with radical structure elucidation. It also provided the first evidence for half-sites reactivity.

In vitro N₃CDP inhibits β2 by reduction of the essential Y₁₂₂• whereas in vivo, the nucleoside analog N₃C is not cytotoxic. In cells, N₃C is not readily phosphorylated to N₃CMP by deoxycytidine kinase, presaging the importance of the specificity of kinases in generation nucleotide (di and triphosphate) therapeutics.

Gemcitabine (F₂C) and clofarabine (ClF), clinically used nucleoside therapeutics that inhibit human RNR.

F₂C and ClF are used clinically as cancer therapeutics (Figure 4B). F₂C targets a broad spectrum of solid tumors (pancreatic, metastatic breast, lung) and hematological cancers. In the clinic, this compound is used in combination with DNA damaging agents such as cis-platin or small molecule inhibitors of signaling pathways that affect the cells response to DNA replication stress (3, 64–66). ClF is limited to hematological cancers (AML, ALL). Both agents inhibit DNA synthesis. RNR is the upstream target of the diphosphate forms of these compounds (F₂CDP, ClFDP), whose inhibition alters dNTP pools. Additionally, the triphosphate forms of these compounds (F₂CTP, ClFTP) inhibit DNA polymerases by incorporation into DNA. The mechanisms, however, by which these compounds inhibit RNR and DNA synthesis, are distinct.

F₂CDP.

F₂C was synthesized independently by research groups at Merrill Dow and Lilly (58, 67). Studies from Plunkett and the Lilly group demonstrated that F₂C inhibited growth of a variety of tumor cell lines and that the cytotoxicity resulted from inhibition of multiple targets including DNA polymerases and RNR (67–69). Biochemical studies on E. coli and human RNR established that F₂CDP is a time-dependent irreversible inhibitor and that inactivation occurs with one equivalent per α2 subunit. Studies using isotopically labeled F₂CDPs established that the products of the inactivation were distinct depending on whether they were carried out in the presence or in the absence of reductant (70–72).

In the presence of reductant (TR/TRR or DTT), 2F⁻, cytosine, PP_i and one alkylated α-cysteine (C₂₂₅) per α2 were identified and no Y₁₂₂• in β2 was lost. Under these conditions, while only 0.5 equivalents of α is inactivated and β remains active, all enzymatic activity is lost. Analysis of the inhibited reaction mixture by SDS-PAGE with no heating revealed that α migrated as a 60:40 ratio of 80 kDa (endogenous α molecular weight) to 110 kDa (modified α) (Figure 12A and B). In cancer cell lines incubated with F₂C, SDS PAGE analysis of cell lysate revealed that the α subunit also migrated in a 60:40 ratio (Figure 12B), similar to the in vitro studies (73). Similar experiments in the absence of reductant, resulted in 50% loss of the β2-Y₁₂₂• and formation of an equivalent amount of a new, nucleotide-based radical (structure shown in Figure 12C). This radical slowly breaks down to cytosine and PP_i. Inhibition was accompanied by loss of 2F⁻, but the α subunit was not covalently labeled. Thus, both in the presence and absence of reductant, 1 F₂CDP/α2 is sufficient for inhibition, although the underlying mechanisms of inactivation are distinct.

Figure 12.

Products of RNR inhibition experiments depend on whether reductant is used. A F₂CDP is a MBI of E. coli and human RNRs in the presence of reductant (protein TR or dithiothreitol) in vitro (70–72), B in cell lines (73), and C in the absence of reductant in vitro. A and C show the products produced during RNR inactivation. B shows SDS-PAGE of RNR from studies in vitro and cell lysate without boiling after incubation with F₂CDP (in vitro) or F₂C (in cells). Wt-α (80 kDa, C), migrates as 110 kDa.

To account for the complete inactivation of RNR with only 0.5 equivalents F₂CDP/α, we proposed that the α/β subunit affinity increased and switching to the second α/β pair for additional chemistry is prevented. To test this possibility, inactivated E. coli and human RNR were subjected to size exclusion chromatography (SEC) analysis. The former showed a species consistent with an apparent molecular weight for α2β2 and the latter with α6β6 (74). In the control, in the absence of F₂CDP, the subunits separate with β eluting as a dimer and α as a mixture of monomers and dimers consistent with weak subunit interactions. Based on our recent EM analyses of a mixture of α and β (1:1) with ClFTP (Figure 6C and the next section) (27), the size with human-RNR attributed to α6β6 is not possible. Altered molecular weights using SEC analysis can be attributed to unusual, non-globular, shapes (α6β2) or altered quaternary structure(s). Fibril structures for example have been reported with human α and ATP (27) and with Bacillus subtilis class Ib (α/β) RNR (75).

ClFDP and ClFTP (reversible).

Early studies by Plunkett et al. established the toxicity of ClF towards many cell lines (CEM, K562, Hep2). In cell-free systems, ClFTP inhibits RNR and DNA polymerase α and ε (59, 60, 66, 76). The observation that the ClFTP:ClFDP ratio in some cells was 7:1, led to the proposal that ClFTP was a reversible inhibitor of ATP binding to the A-site of α (Figure 2B, cone domain (green)). To better understand how RNR is targeted, kinetic and biochemical studies were undertaken with both ClFDP and ClFTP (61). ClFDP was shown to be a reversible, time-dependent, slow binding, inhibitor to the C-site. The kinetic analysis revealed a two-step binding mechanism with a K_I* of 17 nM. ClFTP exhibits reversible, time-independent A-site binding. With ClFTP in 5-fold excess relative to RNR under physiological conditions, RNR activity was rapidly and completely lost with a K_I of 40 nM. With sample dilution and follow-up assays, enzyme activity was recovered over 30 min but only to 50% of the initial value. The t_1/2 of the human Y• in β2 is 30 min at 37 °C (61) and the α subunit is prone to oxidation, making the kinetic measurements challenging; further studies are required. To determine if the observed inhibition was associated with changes in RNR’s quaternary structure, studies of ClFTP (ClFDP) with α, with and without allosteric effector (dGTP), were each examined by SEC. In the absence of the corresponding nucleotide in the elution buffer, α migrated as α6 in the presence of either ClFTP and ClFDP. This result is distinct from dATP-α6 When dATP is absent from the elution buffer during SEC, the hexamer rapidly reverts to α monomer. Thus, the presence of ClFDP or ClFTP alters α’s quaternary structure such that, even subsequent to ClFD(T)P dissociation, α6 remains trapped in the inactive state! β2 had no effect on the inhibition or migration on SEC analysis. Structures of ClFTP mixed with human α/β (1:1) were examined by cryo-EM and solved to 30-Å resolution (Figure 6C). Less than 1 β₂ per 3 α2 was observed and it appeared randomly positioned on the exterior or on top of a hexameric ring structure (Figure 6C) (27). In support of this model, experiments with D57N-α, where the mutation in the cone domain prevents hexamerization of α, revealed that neither ClFD(T)P treated mutant-RNR, nor ClF treated cells with mutant RNR were inhibited. Finally, E. coli RNR, which does not form α6 structures, is not inhibited by ClFDP (61).

The dynamics of quaternary structure interconversions offer an opportunity to inhibit RNRs through unconventional mechanisms. The flexible cone domains (Figure 2, Figure 6A and B) (27, 30) play critical, but distinct roles in these states. Strengthening or weakening the interactions responsible for these quaternary structures with small molecules could alter RNR activity.

To assess the importance of the hexameric state of human-RNR, studies were carried out with His₆-α expressed at 3 or 30× endogenous levels in COS cells that were then treated with noncytotoxic levels of ClF for 3 h. Analysis of the 30× material purified by Ni-affinity chromatography revealed the α6 state was present; with 3× endogenous levels crosslinking was required to detect α6. The α6 state from these and other studies is likely the inhibited state inside the cell in the presence of ClFTP and dATP (33).

A recent extension of this strategy to other adenosine analog therapeutics, cladribine (ClA) and fludarabine (FlU) (Figure 4B), was reported (34). In vitro studies of ClADP and ClATP interactions with human-α revealed α6 formation. Further assessment of the hexameric structures and their relationship to cell cytotoxicity is an ongoing challenge. Collectively, the results in cells and in vitro with these adenosine inhibitors suggest a potential new way to target RNR: trapping α in an inhibited state with a small molecule.

Pleiotropic modes of cytotoxicity of F₂C and ClF.

With both F₂C and ClF, the mechanisms of cytotoxicity require nucleoside uptake and metabolism (64, 65, 77). As noted above, the diphosphates and triphosphates of F₂C and ClF inhibit RNR and DNA polymerases, respectively (60, 78), the latter by chain termination. The consequences of DNA inhibition involving both targets are DNA replication stress that manifests as stalled or collapsed DNA replication forks, DNA single or strand breaks, which can lead to cell cycle arrest, DNA repair or programmed cell death (64) (see Figure 13).

Figure 13.

A general scheme for metabolism of nucleosides using F₂C as an example (64). F₂C and ClF therapeutics require cellular uptake and phosphorylation to the appropriate state recognized by target enzymes. The former is mediated by nucleoside transporters ENT1, ENT2, and CNT [1]. Once inside the cell, both F₂C and ClF are phosphorylated to the monophosphate by deoxycytidine kinase [3] and subsequently to the di and triphosphates by cellular kinases [4 and 6]. Deoxycytidine kinase has unusual specificity in that it phosphorylates both pyrimidines and purines. The concentrations of the monophosphates are in general > than the triphosphates ≫ than the diphosphates, are cell type distinct, and influence therapeutic outcomes. [5] is RNR, [7] is DNA polymerase, and [2] is cytidine deaminase.

F₂CDP, a potent MBI of RNR, results in lower dNDP and consequently dNTP pools. Reduced dCTP, a feedback inhibitor of deoxycytidine kinase (3, Figure 13), results in enhanced production of F₂CMP, leading to elevated levels of F₂CTP. The F₂CTP is then able to more effectively compete with lowered dNTPs pools to inhibit DNA synthesis. F₂C’s broad spectrum of solid tumor inhibition, distinct from other nucleoside therapeutics such as araC, may be associated with the pleiotropic metabolic effects (Figure 13) resulting in its self-potentiation (69).

ClF is also phosphorylated by (3, Figure 13) and subsequently by distinct kinases to afford ClFDP and ClFTP. Its stability (due to F/Cl substitution) is increased relative to other adenosine analogs (ClA, FlU, Figure 4B) by its resistance to metabolism by purine nucleoside phosphorylase and adenosine deaminase. Down-stream consequences of DNA synthesis inhibition by F₂C and ClF are actively being pursued. F₂C is being investigated in combination with DNA damage response inhibitors of Chk1 (64, 68, 79, 80), with inhibitors of ATR in the same pathway (81), and with DNA repair enzyme inhibitors (65, 82). In addition, F₂C is often used in combination with cis-platin that enhances DNA damage and alters the downstream consequences. The ability to monitor the consequences of treatment with combinations of therapeutics using genomics, phospho-transcriptomics and metabolomics, have and will continue to aid in new approaches (65, 83).

Reversible C-site binders lacking phosphoryl groups.

Two compounds (I and II) have recently been reported to inhibit human RNR by binding reversibly to the C-site of α. In contrast to ClFDP and ClADP (Figure 4B), these small molecules lack the diphosphate moiety thought to be essential for substrate recognition. A 5′-substituted amine of F₂C (I), for example, is reported to inhibit RNR in vitro and in vivo (85). The unusual diphosphate binding site for NDP in α (no lysine, arginine or Mg²⁺) suggests that amine substitution might avoid issues associated with cellular uptake and phosphorylation (Figure 13). In a second case, Dealwis et al., reported molecules with a naphthyl salicyl hydrazone scaffold (II) that target the C-site of α and bind reversibly (86, 87). Decorating the scaffold, with an appropriately placed electrophilic moiety such as ClCH₂CO-(R″), could result in alkylation of one of the C-site cysteines, analogous to the mechanisms of covalent protein kinase inhibitors (88).

Reversible inhibitors that disrupt α/β subunit affinity.

The C-terminal tails (30 to 35 amino acids) of all β2 subunits are disordered (Figure 2A) (14, 89–93), distinct, and predominantly responsible for subunit affinity (Figure 14). Early studies of Herpes simplex viruses (HSV-1, HSV-2), which encode for their own RNRs, provide an example of this approach (94, 95). Peptidomimetics of their tail were successfully developed that disrupted the α/β subunit interaction in vitro and in a murine ocular model of HSV-1 induced keratitis (96). The new structure of the E. coli α2β2 (Figure 2D) , which reveals for the first time the tail interaction (residues 341 to 375) with the α subunit (11), may suggest new approaches to disruption of this interface.

Figure 14.

Targeting the α/β interface of active RNR to prevent active complex formation with peptidomimetics and COH29. Targeting formation and repair of Me₂³⁺-Y• (Me is Fe)-cofactor of β2 with HU and triapine, 3-AP.

A second example of subunit disruption was reported by Yen’s group (82, 97). They used the structure of human-p53β2 and computer modeling to identify a pocket in each β subunit close to the C-terminal tail but removed from the buried Fe³⁺₂-Y• cluster (Figure 1C) essential for β2 stability. Virtual screening and additional experiments led to the identification of COH29 (Figure 14) that exhibited cytotoxicity to many of the NIH 60 cancer cell lines and caused S-phase cell cycle arrest. COH29 enhanced cytotoxicity of BRAC1 deficient HCC1937 cells. This report provides an example of DNA repair inhibition (98) (in this case genetically) that potentiated the effects of the RNR inhibitor COH29.

Biosynthesis and repair of the essential diferric-Y• cofactor of Ia RNRs: targeting the β2 cofactor.

Whereas ClF and F₂C target α and the α/β subunits of RNR, respectively, hydroxyurea (HU) and triapine (3-AP, a thiosemicarbazone) (Figure 14) target reduction of the essential Y• (83, 99, 100) cofactor in β2 and/or interfere with cofactor assembly and/or its repair if the essential Y• gets reduced (Figure 1C). HU is used clinically, predominantly in combination with other therapeutics (65), although recent studies suggest that RNR is not a key target of its cytotoxicity (99, 101, 102). Triapine (3-AP) continues to be examined in clinical trials, but has not yet been approved for clinical use (83, 100). Although the upstream target of both these compounds is RNR, the downstream pathways that lead to cytotoxicity are pleiotropic and distinct in different organisms. Herein we focus on HU and 3-AP inhibition of RNR in vitro and in the early stage of cell culture where cell viability remains high. Even under these conditions, their detailed mechanism(s) of RNR inhibition require further exploration.

Background for metallo-cluster metabolism.

The class Ia RNRs require a Fe³⁺₂-Y• cofactor in β2 to initiate NDP reduction in α2 with activity being directly proportional to the concentration of Y• (Figure 1C and Figure 3). The t_1/2 of the Y• in the cluster of different Ia β2s is variable, ranging from 4 days in E. coli at 4° C to 30 min in humans at 37 °C. In addition, recombinant expression of β from different organisms results in variable amounts of active cofactor (0 to 1 Fe³⁺₂-Y•/ β2) (6, 103). In general, therefore, the β2 cofactor must be loaded by self-assembly using Fe²⁺and O₂ with variable outcomes (6, 104). In the past two decades, the importance of biosynthetic pathways has been established for FeS cluster cofactor assembly that, in turn, has been linked to mono- and dinuclear non-heme iron cofactor formation including the RNR cofactor (105). Although much remains to be learned, genetic studies in E. coli and S. cerevisiae, and biochemical studies in vitro on these Ia β2s, have suggested that there are pathways not only for cofactor biosynthesis, but also for its maintenance, and activity regulation (Figure 15). Our general model for Fe³⁺₂-Y• cofactor biosynthesis indicates the requirement for a chaperone protein(s) (106) to alter the apo-β2 conformation for optimized Fe²⁺ loading, an Fe²⁺ carrier protein or small molecule that delivers Fe²⁺ to apo-β2, and a reducing equivalent delivery mechanism required for cluster assembly with O₂ as the oxidant (107). Studies in vivo in E. coli (108) and S. cerevisiae (109) reveal that cluster assembly can yield β2 with each β subunit having 2 Fe²⁺ and 1 Y•, that is, quantitative loading. In vitro however, E. coli β2, loading gives rise to 66% of active cofactor and 33% of inactive diferric cluster with no Y•. In both in vivo and in vitro loading, the activity of RNR per Y• is the same, suggesting identical cofactor structures.

Figure 15.

Model for Me₂³⁺-Y• (Me is Fe or Mn) cofactor biosynthesis (black), maintenance (green) and regulation (red) with factors identified from E. coli in orange and from S cerevisiae in blue (6). S. cerevisiae counterparts are found in humans. Regulation can occur by endogenous reductants or by therapeutics such as HU and triapine (Figure 14).

Hydroxyurea (HU).

HU (Figure 14) has been studied since the 1960s. Based on EPR analysis of prokaryotic and eukaryotic cells or of purified β2 with a self-assembled Fe³⁺₂-Y•, HU treatment results in reduction of the Y• to YOH. In vitro, the iron cluster of human β2 is also reduced (Fe²⁺₂-YOH), whereas in E. coli it remains in the Fe³⁺ state (Fe³⁺₂-YOH). HU reduction of β2 alone is slow (0.45 M⁻¹ s⁻¹) and there is no evidence for its binding to either E. coli or human-β2 (107, 110, 111). The chemical mechanism of Y• reduction and the structure of the resulting cluster remain unknown (112). Studies by the Sjöberg lab in vitro have shown that HU mediated loss of RNR activity is potentiated ten-fold by complexation of β2 with α2, substrate and effector (113). This result led to their suggestion that its reduction of Y₁₂₂• is not direct, but instead might be trapping a “transient” pathway radical at the α/β subunit interface (Figure 3). Studies on the reduction by HU of the Mn⁴⁺-Fe³⁺-β2 cofactor in the C. trachomatis Ic RNR (Figure 1C), an Fe³⁺₂-Y• surrogate, were also interpreted to suggest that HU intercepts a pathway radical at the α/β interface (112). Furthermore, Mn⁴⁺-Fe³⁺-β2 in the presence of α2, CDP, and ATP is reduced by HU to a Mn³⁺-Fe³⁺ cluster with half-sites reactivity involving a fast phase and a slow phase, with apparent saturation by HU for the fast phase. These studies support HU binding and targeting of the radical transfer pathway (112, 114). The consequences of the HU reduced cofactor state in E. coli and mammalian cells are still unclear; however, since the proteins identified in S. cerevisiae for β2 cofactor biosynthesis and maintenance are also found in mammalian cells (107), Y• regeneration is one possible fate (maintenance pathway, Figure 15) and requires further investigation.

RNR inhibition by HU blocks DNA replication. Two papers have suggested that cytotoxicity from extended HU exposure of E. coli (100) or S. cerevisiae (102) cells is linked to reactive oxygen species (such as HO•) mediated damage. Vernis et al. showed that HU resistance in S. cerevisiae led to enhanced production of the cytosolic FeS cluster biosynthetic machinery including Dre2/Tah18 (102). We have shown the importance of these two proteins in the assembly of the β2 cofactor in S. cerevisiae (107).

Triapine.

Triapine (3-AP, Figure 14) has been extensively investigated since its introduction in the 1990’s and its cytotoxic effects have inspired the synthesis of many additional thiosemicarbazones. However, studies on these analogs reveal that the mechanism of cytotoxicity changes with structure. The complexity arises from their distinct abilities to bind Fe²⁺ and Fe³⁺ (also Cu²⁺ and Zn²⁺, not discussed) and the resultant ligand field imposed iron redox chemistry (100). The results reported by different groups (115, 116) in different mammalian cell lines, primarily at late stages of 3-AP treatment have thus made it challenging to compare and evaluate the outcomes between the different studies.

Our recent studies in cell culture in the “early” stages of 3-AP treatment are summarized and provide a framework for thinking about the issues and evaluation of their potential as therapeutics that target β2 (115). Three mammalian cell lines (K565, COS-1 and HU resistant TA3) treated with 3-AP and analyzed by whole cell EPR, revealed loss of the RNR Y• and assays of the corresponding cell lysates revealed loss of RNR activity. Immunoprecipitation of β2 from ⁵⁵Fe treated and non-treated cells revealed similar iron content. These and additional studies suggested that the Y• loss is the major mode of RNR inhibition with iron loading remaining unchanged. Although the oxidation state of the bound iron is unknown, we know from in vitro studies that the Fe²⁺ loaded β2 can assemble rapidly into the native Fe³⁺₂-Y• cofactor, consistent with a maintenance pathway (Figure 15, green). Our model is that Fe²⁺-(3-AP) is the active species involved in β2 inhibition and that in the continued presence of Fe²⁺/ Fe³⁺, RNR is susceptible to Fe²⁺-(3-AP) inhibition by direct Y• reduction. Recent studies by Gräslund et al. using [³H]-3-AP and a docking model of 3-AP to mouse β2, resulted in the proposal of a specific 3-AP binding site (116). However, neither 3-AP or Fe²⁺-(3-AP) binding to β2 has been observed. In our opinion, the mechanism of action of these compounds requires further study. Finally, our studies at early times subsequent to 3-AP treatment, in contrast with latter stage studies of others, indicate that reactive oxygen species are not responsible for RNR activity loss.

From the above discussion, how 3-AP and HU inhibit RNR and the relationship between their RNR inhibition and cell cytotoxicity still remain a mystery. Although interference with cluster assembly/maintenance might yield effective therapeutics, better understanding of the biology of Fe³⁺₂-Y• pathways is required. The recent discovery, however, of Mn³⁺₂-Y• cofactors in β2 of Ib RNRs (Figure 1C) and the identification of a NrdI-β2 interaction essential for oxidant delivery (O₂•⁻) for active cofactor formation (Figure 15) (6, 117), suggests that disruption of this protein/protein interface could provide proof of principle for targeting cofactor pathways in pathogenic bacteria. The link between the Ia Fe³⁺₂-Y• pathway, iron homeostasis and oxidative stress will make selective targeting difficult. However, for pathogenic organisms with Mn and or Mn/Fe clusters (Figure 1C Ib–Id), interference with cluster assembly, may well provide a new therapeutic target.

Summary points.

1. The quaternary structures of the class Ia RNR α subunit are nucleotide-dependent and distinct. α structures detected to data are: α2, non-canonical α2 (118), α4, α6, and fibrils (27, 75).

2. dATP-inhibited RNR structures include α6 (human), α4β4 (E. coli), α4 (P. aeruginosa) (119), and a double helical fibril of canonical and non-canonical α2s (Ib, B. subtilis) (75).

3. dATP inhibited states appear to interfere with the radical transfer pathway and thiyl radical formation by preventing β2 from achieving a productive α2β2 complex.

4. ClFDP (ClFTP) bind to human-RNR and form conformationally “stable” α6 state(s), even subsequent to their dissociation.

5. N₃NDP and F₂CDP are mechanism-based inhibitors of class Ia RNRs with one inhibitor/α2 in the α2β2 complex, half-sites reactivity; half-sites reactivity.

6. Incorporation of unnatural amino acids (F₃Y₁₂₂•or NO₂Y₁₂₂•-β or NH₂Y₃₅₆-β, NH₂Y ₇₃₁ or NH₂Y ₇₃₀-α) and incubation with the second subunit, substrate, and effector, trap radicals within the pathway and increase α/β subunit affinity.

7. The reaction of F₃Y₁₂₂•/ E52Q-β2 with α2, substrate, and effector, results in an asymmetric, active, and kinetically trapped α2β2 complex, whose structure has been determined by cryo-EM.

Future studies.

Continued studies of the structure, dynamics, and biology of RNRs

1. Trapping of additional α2β2 complexes of RNR with mechanism-based inhibitors, and unnatural amino acid-mutants may provide distinct and higher resolution structures.

2. The active form of a mutant E. coli Ia RNR has been shown by cryo-EM analysis to be an asymmetric and dynamic α2β2. The relationship of this structure to the wt enzyme and the structure of the human active complex remain to be established.

New ways to target RNRs

1. Identification of small molecules that can trap human and bacterial RNRs in distinct inhibited quaternary structures.

2. The discovery of biosynthetic pathways for dimetallo-Y• cluster assembly in Ia and Ib RNRs suggest that targeting the metal center formation, such as disruption of NrdI-NrdF interaction in the assembly of the class Ib, Mn³⁺₂-Y• cofactor, might be possible.

3. The omics revolution (proteomics, phosphomics, transcriptomics) and a refined understanding of nucleotide metabolism are providing new insight into RNR regulation. This knowledge will lead to combination chemotherapies using RNR inhibitors in conjunction with inhibitors of downstream signaling pathways.

BOX 1 The model (Figure 7) encompasses weak, dynamic, subunit interactions that change with subunit and dNTP concentrations by altering RNR quaternary structure and activity. The assays/issues for the E. coli RNR are described in BioRxiv (42) and are essential for development of in vitro and in vivo high throughput screens for RNR inhibitors. The same issues are likely to be encountered, to different extents, with other Ia and Ib RNRs.

BOX2 Moonlighting function of α. A recent report (35) and review (84) provide support for a “moonlighting” function for α, independent of its ability with β to make dNDPs in the cytosol of the cell. Using a C→S mutant that inactivates formation of −S• in α (Figure 1A), a small amount of α was detected in the nucleus of the cell in an α6 state. Yeast two hybrid experiments with cDNA from HELA cells revealed that α interacts with ZRANB3, a protein that forms a complex with PCNA, the sliding clamp that together with DNA polymerase promotes DNA synthesis in non-stressed cells. Nuclear localized α inhibits the interaction of ZRANB3 with PCNA resulting in inhibition of DNA synthesis. This study potentially provides an explanation for the tumor suppressor activity reported for α (3).