Martha L. Ludwig and Rowena G. Matthews are in the Department of Biological Chemistry and Biophysics Research Division, University of Michigan, 930 North University Avenue, Ann Arbor, Michigan 48109-1055, USA.
Correspondence should be addressed to Martha L. Ludwig mlludwig@umich.edu
The monomeric B12-dependent ribonucleotide reductase from L. leichmannii has the central 10-stranded /-barrel found in all ribonucleotide reductases but incorporates two distinctive structural features, a novel cobalamin-binding fold and an insert forming part of a specificity control site that mimics the allosteric site found in the oligomeric di-iron dependent reductases.
"It is usually the comparison of structures... that is most enlightening about function."
Gregory A. Petsko
Ribonucleotide reductases (RNRs) catalyze the reduction of ribonucleotides to deoxyribonucleotides. Since no pathway exists for the de novo biosynthesis of deoxyribonucleotides, which can only be formed by reduction of ribonucleotides, these enzymes play essential roles in all living organisms. All known reductases employ a radical mechanism for ribonucleotide reduction, in which catalysis is initiated by hydrogen atom transfer from the substrate to a thiyl radical1,
2. Quite surprisingly for essential enzymes, ribonucleotide reductases employ three different chemical strategies to generate the essential thiyl radical. The structure of the B12-dependent ribonucleotide reductase from Lactobacillus leichmannii, reported on page 293 in this issue of Nature Structural Biology3, depicts the second of these strategies for thiyl radical formation.
Mechanisms of radical generation Class I enzymes require O2 for radical generation, and employ a di-iron cluster on a separate subunit (R2) to produce a stable tyrosyl radical. Long-range electron transfer from the catalytic thiol to the tyrosyl radical of R2 generates the thiyl radical. Since oxygen was presumably absent from the atmosphere at the time of the transition from the RNA world to the DNA world, this enzyme class is thought to be an evolutionary latecomer. The other two classes of ribonucleotide reductase can generate the catalytic thiyl radical under strictly anaerobic conditions. Class II enzymes use the adenosylcobalamin form of the B12 cofactor (AdoCbl); cleavage of the weak carbon-cobalt bond of AdoCbl forms a 5'-deoxyadenosyl radical that in turn generates the thiyl radical (Fig. 1). These enzymes have only one type of subunit and are monomers or homodimers. Class III enzymes use S-adenosylmethionine (AdoMet), referred to by the late H.A. Barker as "poor man's cobalamin", to generate a radical on a glycyl residue. The glycyl radical is formed by reaction with a 5'-deoxyadenosyl radical derived from AdoMet in a step that requires a second subunit containing an iron−sulfur cluster and uses a reducing equivalent provided by flavodoxin4,
5.
Figure 1. Formation of the thiyl radical in class II ribonucleotide reductases.
Homolysis of the Co−5' CH2 bond, enhanced by addition of effector11, forms the adenosyl radical that abstracts the hydrogen from cysteine (residue 408 in L. leichmannii ribonucleotide reductase).
RNR structures The X-ray structures of the R1 catalytic component and the R2 di-iron component of the class I RNR from E. coli were determined by Hans Eklund and colleagues6,
7. The catalytic subunit was found to be a novel 10-stranded /-barrel with a loop that hosts the thiyl radical protruding into its center. Despite a lack of sequence similarity with other ribonucleotide reductases, Eklund6 and others8 correctly predicted that all ribonucleotide reductases would have similar catalytic subunits, reflecting their similar catalytic strategies. The structure of the catalytic subunit of the class III enzyme was determined by Logan et al.9, and revealed the characteristic 10-stranded /-barrel with a central loop bearing the thiyl radical precursor.
Many features of the structure reported by Drennan and colleagues3 are remarkably similar to the structures of the catalytic subunits of the class I and class III enzymes, even though there is <10% sequence homology among them. The similarities and contrasts with the enzymes of other classes have much to tell us about all ribonucleotide reductases.
The L. leichmannii enzyme is by far the simplest ribonucleotide reductase, substituting a relatively small radical- generating cofactor, AdoCbl, for the complicated radical generating apparati of the class I and class III enzymes. One of the structures reported by Sintchak et al.3 shows the cobalt of a bound cobalamin analog, adeninylpentylcobalamin, located 10 Å from the site of the thiyl radical. Although the adeninylpentyl group is not fully visible in the structure with the cofactor analog bound, the cobalamin is positioned above the /-barrel in a manner that permits access of the 5'- dexoyadenosyl radical to Cys 408 (Fig. 2). Prior to the analysis of the L. leichmannii enzyme, no structure had been solved in which the relative orientations of the radical generating apparatus and the catalytic site of a ribonucleotide reductase could be seen. Superpositions of the class II and class III structures now align the adeninylpentyl group with the glycine that generates the thiyl radical in class III enzymes. Similarly, alignments with the class I enzyme place the tyrosine residues 730 and 731 implicated in electron transfer to R2 (ref. 10) close to the cobalamin. These structural analogies confirm the postulate that the three different radical generators converge on essentially identical mechanisms at the point of thiyl radical formation.
Figure 2. Arrangement of the players in the reduction of ribonucleotides.
The effector in class II ribonucleotide reductases plays multiple roles in catalysis: it not only influences the substrate specificity, but also accelerates cleavage of AdoCbl and generation of the thiyl radical on Cys 408. This figure shows the relative positions of the effector, cofactor, substrate and critical thiol residues. The cobalamin and critical residues from the class II structure are shown with carbons colored gold; the GDP substrate and allosteric effector dTTP, modeled by superimposing the structure of the class I E. coli RNR15, are shown in pale yellow and labeled in gray. The cobalamin cofactor is located above the conserved 10-stranded barrel and near the finger carrying Cys 408. Modeling of the substrate places it next to the site of the thiyl radical and in close proximity to Cys 119 and Cys 419, which in the dithiol form provide electrons for substrate reduction. Part of the four-helix bundle comprising the effector site can be seen at the right.
One of the novel features of this class II structure is a new fold for B12 binding that is unlike any of the other known B12 binding modules. This region is formed from two insertions in the barrel, and is poised to close around the cobalamin. In most AdoCbl-dependent mutases, binding of substrate triggers weakening of the carbon-cobalt bond of the cofactor to initiate radical generation. Uniquely in RNRs, the binding of effector alone can trigger homolysis of the AdoCbl11,
12. The structure analyses3 reveal a displacement of the B12 binding region on addition of cobalamin, and further closure of the active site is likely to be prompted by binding of effector, which is not present in the structures reported here. The authors anticipate rearrangements that would decrease the distance between cobalt and the sulfur of Cys 408 to 6 Å deduced from EPR measurements13. Closure of the active site would ensure that the reactive radical species are never exposed to solvent during the catalytic cycle, but are only generated in a closed ternary complex.
Ribonucleotide reduction consumes two reducing equivalents. In class III enzymes, these are provided by formate, while in class I and class II enzymes, the immediate reductants are a pair of thiols at the active site, which are oxidized to form a disulfide. The disulfides are reduced by thiol/disulfide exchange with thiols near the C-terminus of each of the proteins, and the resultant C-terminal disulfide is in turn reduced by thioredoxin or glutaredoxin4,
14. Reduction of the active site disulfide must take place in a conformation that admits the C-terminal thiols to the interior of the /-barrel (Fig. 2). Thus reduction of the active site disulfide, which appears to be rate limiting in turnover, may occur only after dissociation of AdoCbl12. The orchestration of conformation changes and ligand binding remains to be determined.
Regulation of substrate specificity Each of the four ribonucleotides is a substrate for ribonucleotide reductase, and each class of enzyme has a complicated mechanism for regulating the relative flux of substrates in response to cellular concentrations of deoxyribonucleotides4. All reductases have a specificity site, where the effector nucleotides bind, that is separate from the catalytic site where the substrate ribonucleotides bind. In the class I and class III enzymes the specificity control sites are at a dimer interface where the two chains meet to form a four-helix bundle, and residues from each catalytic chain bind to each effector nucleotide15,
16. The L. leichmannii enzyme is a monomer17, but its specificity site looks remarkably like the specificity sites of the class I 2 dimer. An intricate 130 residue insert into the 10-stranded barrel forms a structure that 'imitates' the four-helix bundle responsible for effector binding in the dimeric class I RNR. Alignment of the four-helix motifs and conservation of some of the residues known to bind dNTP effectors in class I RNR argue for similar modes of effector binding in the class I and class II enzymes. However the exact symmetry of the dimer interface is lost, leaving only a single site for effector binding in the L. leichmannii enzyme3,
17.
An allosterically regulated monomeric enzyme seems to be a conundrum; the models that dominate our thinking about allosteric regulation achieve cooperativity by deploying interactions between the subunits of an oligomer. In fact, recent models for regulation in class I ribonucleotide reductases invoke two modes of allosteric behavior — communication between subunits and ligand-dependent subunit association/dissociation18. Much less is known about the details of allosteric control in L. leichmannii and other class II enzymes, but initial studies have shown that the patterns of activation and inhibition resemble those found in class I17. How then should we think about the action of effectors in the monomeric L. leichmannii enzyme? We suggest that the fundamental phenomenon is simply that of linkage between the binding sites19; in this view the presence of a single effector bound to the same chain may propagate conformation changes that can alter binding specificity and kinetics at the catalytic site.
In the class II enzymes, the effector nucleotides also play a critical role in radical generation. Blakley11, and more recently Stubbe and coworkers12, have shown that effector binding, even in the absence of substrate, leads to rapid cleavage of the carbon-cobalt bond in adenosylcobalamin and generation of the thiyl radical at Cys 408. In addition, effector binding has been shown to influence the rate of reduction of the active site disulfide14. Thus the effector may be an important mediator in the active site closure and reopening that seem to be required for completion of the catalytic cycle. It will be fascinating to see the structural changes that are induced by effector binding.
/
-barrel found in all ribonucleotide reductases but incorporates two distinctive structural features, a novel cobalamin-binding fold and an insert forming part of a specificity control site that mimics the allosteric site found in the oligomeric di-iron dependent reductases.
/
-barrel with a loop that hosts the thiyl radical protruding into its center. Despite a lack of sequence similarity with other ribonucleotide reductases, Eklund
10 Å from the site of the thiyl radical. Although the adeninylpentyl group is not fully visible in the structure with the cofactor analog bound, the cobalamin is positioned above the 
