RNRs (ribonucleotide reductases) are key players in nucleic acid metabolism, converting ribonucleotides into deoxyribonucleotides. As such, they maintain the intracellular balance of deoxyribonucleotides to ensure the fidelity of DNA replication and repair. The best-studied RNR is the class Ia enzyme from Escherichia coli, which employs two subunits to catalyse its radical-based reaction: β2 houses the diferric-tyrosyl radical cofactor, and α2 contains the active site. Recent applications of biophysical methods to the study of this RNR have revealed the importance of oligomeric state to overall enzyme activity and suggest that unprecedented subunit configurations are in play. Although it has been five decades since the isolation of nucleotide reductase activity in extracts of E. coli, this prototypical RNR continues to surprise us after all these years.
In all organisms, propagation of genetic material from one generation to the next requires DNA replication. The accuracy of the genetic copy and the repair of mistakes depend on balanced levels of ribonucleotides and deoxyribonucleotides [1–3]. One remarkable enzyme, RNR (ribonucleotide reductase), converts all four ribonucleotides into deoxyribonucleotide precursors for DNA synthesis and repair. Because of its crucial metabolic role in dividing cells, RNR has an important place in medicine as the target of several clinically used cancer drugs, such as gemcitabine and clofarabine [4,5].
All RNRs employ a radical-based mechanism for catalysis and are classified by the metallo-cofactor used to generate an essential thiyl radical within a structurally conserved α/β-barrel substrate-binding domain [6–8]. Once the thiyl radical is generated, nucleotide reduction proceeds through a conserved mechanism [9,10] (Figure 1). Eukaryotes and some aerobic prokaryotes employ class Ia RNRs, which utilize a diferric-tyrosyl radical cofactor for the reduction of NDPs (ribonucleoside diphosphates) . In addition to having a complex catalytic mechanism, RNRs employ sophisticated allostery to maintain the appropriate nucleotide balance in the cell (recently summarized in ). In the present paper, we focus on the Escherichia coli enzyme, which has long served as the prototype for class Ia RNRs, and review how structural biology techniques have illuminated its catalytic mechanism and allosteric regulation.
Catalytic mechanism of
E. coli class Ia RNR
Surprising structures of domains
In class Ia RNRs, thiyl radical generation in the active site is achieved by intermolecular interactions between two separate proteins: the nucleotide-binding protein encoded by the gene nrdA and the radical-generating protein encoded by the gene nrdB . For the canonical class Ia enzymes, including the prototypical E. coli enzyme, the NrdA and NrdB proteins form homodimers called α2 and β2 respectively (Figure 2). A diferric-tyrosyl radical cofactor is assembled in β2 at Tyr122 in the presence of molecular oxygen, Fe(II) and an electron source [14–16]. For each turnover, the formation of the active holocomplex triggers long-range PCET (proton-coupled electron transfer), reducing the Tyr122 radical in β2 and oxidizing Cys439 in the active site of α2 to form the essential thiyl radical .
Structures of the class Ia
E. coli RNR subunits
A number of crystal structures of the individual α2 and β2 subunits have been determined for the E. coli enzyme. The β2 structure is often described as a heart-shaped dimer (Figure 2B) . One surprise from the β2 structure was that the tyrosine residue (Tyr122) that harbours the tyrosyl radical is buried in the centre of each β protomer, adjacent to its di-iron centre, 10 Å (1 Å=0.1 nm) from the surface. This observation suggested the need for the radical to transfer through other residues to reach the surface of β2en route to the substrate-binding site on α2. Such a long-range electron transfer was unprecedented at the time that it was proposed , but is now well supported by experimental evidence [19–21]. Another finding from the β2 structure was that the 32 C-terminal residues were not observed owing to their thermal lability (Figure 2B, broken curves) . Earlier studies demonstrated that C-terminal residues of β2 were critical for its interaction with α2 [22,23]. Just upstream of this C-terminal sequence within the unstructured tail is the absolutely conserved Tyr356 that subsequent studies have shown to be part of the radical relay between the RNR subunits [19,24,25].
Shortly after the first β2 structure was determined, an equally surprising α2 structure was reported  (Figure 2A). The α2 structure was the first example of the ten-stranded α/β barrel that is now known to be characteristic of all three classes of RNR [6,7]. Cys439, the residue that harbours the catalytic thiyl radical, was located at the tip of a ‘finger loop’ poking into the middle of the barrel. This residue was located adjacent to the absolutely conserved Cys225 and Cys462 that provide the reducing equivalents for nucleotide reduction  (Figures 1 and 2A). In this structure, Cys225 and Cys462 were observed in their oxidized state as a disulfide-bonded cystine. Furthermore, the positions of Tyr730 and Tyr731 relative to Cys439 implicated their involvement in PCET from β2 to α2 (Figure 2A). Like β2, α2 has an unstructured C-terminal tail that also contains residues involved in catalysis. Two cysteine residues, Cys754 and Cys759, in the flexible α2 tail are responsible for accessing the active site to reduce the Cys225–Cys462 disulfide bond at the end of each catalytic cycle [10,26] (Figure 1). Additionally, the structure of α2 was obtained by co-crystallization with a 20-residue C-terminal peptide of β2, revealing an important determinant of the α2–β2 interaction , but again density for Tyr356 was not observed (Figure 2A, orange ribbons).
Early experiments with the E. coli class Ia enzyme revealed that deoxynucleotide effectors and ATP are responsible for regulating substrate specificity as well as overall enzyme activity [12,27]. Nucleotide-binding sites in E. coli α2 were located by further crystallographic work . Crystals of α2 co-crystallized or soaked with the effector dTTP yielded information about the allosteric site known as the ‘specificity site’. The resultant structures showed dTTP bound at either end of a four-helix bundle formed at the α2 dimer interface, approximately 15 Å from the active site  (Figure 2A). The specificity effector bound at this site appears to communicate with the active site by ordering the conformation of flexible loops, including one termed ‘loop 2’, which is adjacent to the active site . Consistent with the ancient origin of RNR, this specificity-regulating feature is conserved in the other RNR classes [7,12,29,30]. However, although GDP reduction is promoted by dTTP, structures determined from crystals soaked with this substrate–effector pair showed low substrate occupancy, even for a C292A mutant with enhanced GDP affinity and using dithiothreitol to reduce the active-site disulfide . Thus we do not yet have a molecular-level picture of how effectors at this site determine substrate specificity for the E. coli enzyme. The α2 subunit also has a third nucleotide-binding site that accepts ATP to activate or dATP to inhibit overall catalytic activity  (Figure 2A). This ‘activity site’ was located in the same study using crystals soaked with the substrate, CDP, and the ATP analogue, AMP-PNP (adenosine 5′-[β,γ-imido]triphosphate) . In the resulting structure, AMP-PNP was observed at the N-terminus of the protein in the so-called ‘cone domain’, a four-α-helix bundle that is found across class Ia enzymes . However, it was not evident from this structure how binding of ATP or dATP to the cone domain would communicate a signal to the distant active site over approximately 40 Å. In addition, although ATP is also the specificity effector for CDP, AMP-PNP was only weakly present at the specificity site, and the substrate molecule was not observed in the active site .
Insight into the structural components of an active RNR complex
To afford enzymatic turnover, α2 and β2 must interact allowing for radical transfer to the active site. However, structural data describing the active complex of E. coli class Ia RNR has been challenging to obtain. Seminal studies on this RNR led to the proposal that active enzyme exists as an α2β2 complex [32,33]. Using sucrose gradient centrifugation, mixtures of α2 and β2 in the absence of nucleotides or in the presence of ATP or dTTP were shown to sediment at around 8.8–9.7 S with an apparent 1:1 subunit stoichiometry . In a later study, AUC (analytical ultracentrifugation) was repeated in the absence of sucrose, which can perturb protein–protein interactions, to show again that an equimolar mixture of α2 and β2 sedimented at around 10 S with an apparent molecular mass that is consistent with an α2β2 oligomerization state . Together, these studies suggested that the active quaternary structure is an α2β2 complex.
The determination of the individual α2 and β2 structures that followed two decades later showed that the two dimeric proteins were complementary in shape and therefore could plausibly form an active α2β2 complex [8,18]. When the two structures are docked along their two-fold symmetry axes, the lobes of the heart-shaped β2 neatly fit in the concave pockets containing the active sites of α2  (Figure 3A). Importantly, the subunit arrangement in this model places residues that are important for radical transfer in close proximity across the α2–β2 interface. In the proposed PCET pathway, radical propagation occurs along conserved aromatic residues: from Tyr122→Trp48→Tyr356 in β2 to Tyr731→Tyr730→Cys439 in α2 . These residues form a conduit in the α2β2 docking model, with the surface-exposed Trp48 of β2 within ~23 Å of Tyr731 of α2, a reasonable distance for PCET if bridged by Tyr356 on the flexible C-terminal tail of β2 (Figure 3A). Site-directed mutation of each of these conserved residues eliminated activity to background levels, consistent with their proposed involvement in the PCET pathway [25,34–36]. More recently, incorporation of unnatural amino acids that act as radical traps at Tyr356, Tyr731 and Tyr730 has permitted the direct demonstration of redox coupling between these residues [20,21,24]. Furthermore, these mutants enabled measurements of Tyr122–Tyr731 and Tyr122–Tyr730 distances by PELDOR (pulsed electron–electron double resonance) spectroscopy [19,37]. These distances measured were consistent with those predicted by the α2β2 docking model and thus provided the first experimental support for the subunit arrangement in this model.
Structural insight into the active complex
Recently, AUC studies of class Ia RNR were reinitiated, and showed that the sedimentation value of equimolar mixtures of α2 and β2 in the presence of activating allosteric effectors is highly dependent on protein concentration and able to exceed 11.5 S, the theoretical value for the globular α2β2 complex predicted by the docking model . This concentration-dependent behaviour indicated that the active complex is embedded in an equilibrium mixture and able to interconvert not only with dissociated species, but also with a higher-order complex, offering an explanation for the elusiveness of the active complex to crystallographic techniques that benefit from homogeneous samples.
SAXS (small-angle X-ray scattering) was used to extract structural information for the active complex from the equilibrium mixture and provide the first structural evidence that the active complex is well described by the α2β2 docking model . Under assay conditions where α2 and β2 should be 97% in an active conformation, the resultant shape reconstruction shows a globular molecule, consistent with the close subunit arrangement in the docking model (Figure 3B, left-hand panel). The excess density seen in this shape reconstruction was shown to be due to the presence of a small population of the higher-order complex interconverting with the predominant α2β2 complex . When data are subjected to a global analysis, the contribution of the higher-order complex can be quantified and therefore subtracted from the SAXS data, yielding a shape reconstruction that represents the active complex on its own (Figure 3A, right-hand panel).
First structures of the dATP-inhibited state of
E. coli RNR complex
In addition to the quest for a structure of active E. coli class Ia RNR, a structure depicting both α2 and β2 in a dATP-inhibited state has been long sought. Early ultracentrifugation studies provided the first evidence that allosteric inhibition of class Ia RNR by dATP is coupled with higher-order oligomerization [32,33]. More recently, the role of oligomerization in the activity regulation of the E. coli enzyme was revisited with multiple biophysical techniques [38,39]. These studies have converged on a consistent picture involving the formation of an α4β4 complex under dATP inhibition [38,39] and have led to the elucidation of the first structures of this complex .
A number of low-resolution techniques were used to determine the oligomerization state of the dATP-inhibited E. coli enzyme. In a GEMMA (gas-phase electrophoretic mobility macromolecule analysis) study, a heavy complex was observed in the presence of inhibiting dATP concentrations with an electrophoretic mobility diameter that is consistent with a molecular mass of 510 kDa . This heavy complex was shown to have an α4β4 subunit stoichiometry with the use of a β construct tagged with the 57 kDa NusA protein in order to distinguish the α monomer from the similarly sized β2 . More recently, AUC and SAXS were used to confirm the molecular mass and subunit composition of the dATP-inhibited α4β4 complex and show further the stability of this species in solution over a wide range of RNR concentrations . Furthermore, both the power-law-dependence observed in the X-ray scattering and the sedimentation rate indicated that this α4β4 complex is highly non-globular in shape .
The first structures of the dATP-inhibited α4β4 were determined by single-particle EM (electron microscopy) and X-ray crystallography, revealing a ring-shaped complex with an alternating arrangement of α2 and β2 subunits  (Figure 4). In this study, the α4β4 rings were the dominant species observed in EM images of α2 and β2 mixtures under dATP inhibition. A three-dimensional EM reconstruction of this species generated at 23 Å resolution was fitted with crystal structures of the individual subunits, yielding a pseudo-atomic model. This pseudo-atomic model was nearly superimposable with a 5.7 Å resolution crystal structure of the α4β4 ring obtained by co-crystallizing α2 and β2 with dATP, demonstrating the structural stability of this complex even at the dilute protein concentrations used for EM. Furthermore, the theoretical sedimentation rate and X-ray scattering profile predicted from these structures closely matched experimental AUC and SAXS data, indicating that this α4β4 ring exists in solution at physiologically relevant RNR and dATP concentrations . Moreover, a low-resolution molecular envelope obtained from SAXS resembled both the EM model and crystal structure, showing consistency among the three structural techniques .
Crystal structure of the dATP-inhibited α4β4 complex
The structures obtained from this study provide insight into how dATP inhibits the prototypic class Ia enzyme from E. coli . The ring is formed through contacts made by β2 to the α2 N-terminal cone domains housing the activity sites, such that Trp48 in β2 and Tyr731 in α2 face a large solvent-exposed hole (Figure 4). In the crystal structure, dATP is observed in both the activity and specificity sites. Consistent with this result, the α4β4 complex was shown by SAXS to maximally form at above 4 molar equivalents of dATP per α2. Additionally, in the crystal structure, C-terminal residues of β2 were observed at the peptide-binding site of each α-protomer (Figure 4). However, there is broken electron density between Leu339 and Ser363 that again prevented visualization of the elusive Tyr356. Although nucleotide reduction by the class Ia RNR depends on the juxtaposition of α2 and β2 to permit the round-trip radical relay from Tyr122 to Cys439, the distance between the closest visible residues in the PCET pathway (Trp48 in β2 and Tyr731 in α2) in the α4β4 complex is ~55 Å (Figure 4). The ~55 Å distance is too great to accommodate the radical transfer even with a bridging Tyr356 and would require the implausible scenario of radical transfer through bulk solvent . Together, these structures suggest that the dATP binding at the activity sites inhibits RNR by stabilizing an open subunit arrangement of α2 and β2 that physically separates the PCET residues, preventing radical transfer.
Model for allosteric regulation
These recent developments in the structural characterization of E. coli class Ia RNR have provided a new understanding of allosteric activity regulation for this protein. We now know that the prototypical class Ia RNR from E. coli exists in at least three states that are supported by direct structural observation: the dissociated α2 and β2 subunits [8,18], the active α2β2 complex , and the inhibited α4β4 complex . Furthermore, it was shown using AUC and SAXS that, in the presence of activating effectors or in the absence of any effectors, the E. coli enzyme exists as an equilibrium mixture of α2, β2, α2β2 and α4β4, whose distributions are sensitive to protein concentration. In contrast, dATP is uniquely able to push the equilibrium towards a single α4β4 state . From these data, a three-state model emerges for the allosteric regulation of overall activity in E. coli class Ia RNR. Here, active α2β2 complex is an intermediate between the dissociated state and the inhibited α4β4 complex (Figure 5). Although increases in protein or dATP levels shift the equilibrium to the right, according to this model, ATP may up-regulate the E. coli enzyme by competing with dATP for binding at the activity site and shifting the equilibrium away from the inhibited α4β4 complex. Thus, in the cell, RNR may act as a molecular thermostat that senses the balance between ribonucleotide and deoxyribonucleotide pools by the competitive binding of ATP and dATP for the activity site, which in turn modulates the equilibrium between the active α2β2 and inhibited α4β4 complexes.
Model for allosteric regulation of activity in
E. coli class Ia RNR
In E. coli, the nrdA and nrdB genes are on the same operon and are thus co-expressed . The ability to form α4β4 provides the E. coli enzyme with a clever mechanism for regulating activity by keeping its subunits together, but in a non-productive arrangement. The structure of the dATP-inhibited α4β4 suggests that interconversion between these states can readily occur. Each contact at the cone domains containing the activity site buries ~1000 Å2 involving only a handful of residues from each subunit. In addition to the contacts involving the activity site, interconversion between α2β2 and α4β4 may be facilitated by the flexible C-terminal tails of β2. Although both C-terminal tails from a single β2 may be interacting with a single α2 in the compact active complex, the crystal structure of the α4β4 complex suggests that the C-terminal tails of a single β2 may be used to recruit two α2 subunits. The C-terminal tails may allow α2 and β2 to remain flexibly linked in order to rapidly switch from a closed to open conformation and vice versa (Figure 5).
Recent advances bring clarity to the early observations that suggested a link between activity regulation and oligomerization in the class Ia RNR of E. coli, yet many important problems remain to be solved . High-resolution structures of α2 both alone and in complex with β2 in the presence and absence of the activity site effectors, ATP and dATP, are needed to understand at an atomic level the conformational changes in the activity site. In existing E. coli α2 crystal structures, the cone domains containing the activity sites make lattice-packing interactions [8,28], and therefore new crystallization conditions might be required to unambiguously resolve details of effector-induced rearrangements at the cone domain. Although structures of α2 with specificity effector–substrate pairs have been determined for the yeast class Ia RNR , the effect of β2 in an active α2β2 complex on specificity regulation is not yet known for any class Ia enzyme. Furthermore, species-specific differences between the sequences at the specificity site [41,42] suggest that a more complete model of specificity regulation in the E. coli enzyme will require structures with full occupancy substrate–effector pairs.
Eukaryotic class Ia RNRs subscribe to a qualitatively similar regime of activity and specificity regulation as their prokaryotic homologues [43–45]. However, the relationship between oligomerization and allostery in the eukaryotic class Ia RNRs is not as well defined as in prokaryotes. In contrast with the E. coli homologue that requires both subunits to form the inhibited oligomer [33,38,39], the eukaryotic α2 alone can oligomerize under both ATP activation and dATP inhibition [42,43,46–48]. Recently, the dATP-stabilized complex from yeast was shown to be a hexameric ring of α subunits with the cone domains again forming intersubunit contacts . Without structural information on the active form of eukaryotic class Ia RNRs, however, it is not yet clear why this dATP-induced α6 arrangement would be inactive. A comparison between the structures of the dATP- and ATP-induced complexes is needed to understand its unique twist to this developing saga of allosteric regulation. Excitingly, differences in the allosteric regulation of eukaryotic and prokaryotic class Ia RNRs may permit the future design of allosteric inhibitors that target specific RNRs, such as the RNRs of pathogens.
Finally, little is known about the interaction between the α2 and β2 subunits in the active complex. The long-sought atomic-resolution structure of the active α2β2 would reveal the positions of residues involved in PCET and might provide critical insight into the conformational changes that trigger radical transmission. In particular, the position of Tyr356 on the C-terminal tail of β2 has not yet been visualized in any of the available structures. Moreover, mounting evidence establishes that only one of the two active sites in RNR is engaged in radical transfer and substrate reduction at a time [37,49–52]. Such half-sites reactivity suggests asymmetry in the interaction between α2 and β2 that might finally be explained by a high-resolution structure of the active complex. Presumably, the difficulty in obtaining a high-resolution structure of an active complex derives from the dynamic properties of the subunit interactions, the full nature of which was emphasized recently in AUC and SAXS studies .
Although five decades have passed since the discovery of the prototypical class Ia enzyme from E. coli, RNRs continue to amaze us with their elaborate modes of allosteric regulation, dramatic conformational changes and complex radical transfer pathways. In the latest chapter, two decades since the first structures of E. coli α2 and β2 were determined [8,18], we now have the first low-resolution reconstruction of an active α2β2 complex and the first X-ray and EM structures of the dATP-inhibited α4β4 complex for the E. coli enzyme . These studies reveal the transient nature of the active complex and an unanticipated ring-shaped structure for the dATP-inhibited complex, adding a new twist to the RNR story and promising more surprises to come in the future.
Frontiers in Biological Catalysis: Biochemical Society Annual Symposium No. 79 held at Robinson College, Cambridge, U.K., 10–12 January 2012. Organized and Edited by David Leys (Manchester, U.K.), Andrew Munro (Manchester, U.K.), Emma Raven (Leicester, U.K.) and Martin Warren (University of Kent, U.K.).
We are indebted to our collaborators for their contributions to much of the science we describe: Michael A. Funk, Kenichi Yokoyama, JoAnne Stubbe and Francisco J. Asturias. We also thank JoAnne Stubbe, Joseph A. Cotruvo and Patrick G. Holder for helpful discussion in preparing this mini-review.
This work was supported by the National Institutes of Health [grant numbers T32GM08334 (C.M.Z), F32GM904862 (N.A.), K99GM100008 (N.A.), F32DK080622 (E.J.B.) and P30-ES002109 (C.L.D.)]. C.L.D. is a Howard Hughes Medical Institute Investigator.
These authors contributed equally to this work.