nature structural & molecular biology volume 18 number 3 mArCH 2011 251

The Karolinska Institute in Stockholm cele­brated its bicentennial in 2010—a year that also marked the golden jubilee of the discov­ery of a complex and fascinating enzyme, ribo­nucleotide reductase (RNR), by the Institute’s eminent emeritus, Peter Reichard1. RNR is an essential enzyme for every living organ­ism (with the exception of a very few obligate intracellular symbionts), catalyzing the first committed step in production of the build­ing blocks of DNA, the dNTPs, from their NTP precursors. RNR continues to fascinate biologists and chemists five decades after its discovery because of its complex radical chem­istry and intricate allosteric regulation2. It thus seems fitting that early in 2011, the article on page 316 by Fairman et al.3 opens a window onto one of the least understood aspects of RNR function: the structural basis for its over­all activity regulation.

RNRs are divided into three classes, mainly based on their radical generation chemistry. Their core functionality lies in their reductase subunits (α2 or α), which have co­opted a diverse array of radical generation strategies (Fig. 1). In the class I RNRs, a tyrosine radi­cal is generated in the β2 subunit, a di­iron­oxo enzyme; in class II the radical is generated directly on α or α2 by cleavage of adenosyl­cobalamin; and in class III a glycyl radical is gen­erated on α2 when a radical SAM protein cleaves S­adenosylmethionine. In all three classes, the radical is channeled to a cysteine in the active site of the α subunit to initiate catalysis4.

The structures of the active holoenzymes of RNR have unfortunately remained obscure, with the exception of the simpler class II type5. In 1994 Uhlin and Eklund presented

a model for an α2β2 complex of class I RNR6 from Escherichia coli, based on crystal struc­tures of the individual subunits6,7 and bio­chemical information (Fig. 1), that remains the most plausible view of the active enzyme to date. Nevertheless, it has proven very dif­ficult to confirm this model using crystallo­graphy. The previously solved low­resolution crystal structure of an α2β2 complex from Salmonella typhimurium rather muddied the waters, as it appeared to show the β2 sub­unit trapped in an intermediate conformation in which it did not make the expected degree of contact with the α2 subunit8.

RNRs are allosterically regulated on two levels: overall activity and substrate specific­ity (Fig. 2). The overall activity is regulated by the binding of dATP (inhibition) or ATP (stimulation) to the so­called activity site in the ATP cone domain of the α2 subunit of RNRs from classes Ia and III (Fig. 1). The substrate specificity is regulated by the binding of dNTPs to the specificity site: ATP and dATP upregu­late the reduction of CDP and UDP, whereas dTTP upregulates GDP reduction and dGTP increases the rate of ADP reduction. This regulation is essential to maintain balanced dNTP pools for DNA synthesis and repair.

Derek T. Logan is in the Department of Biochemistry and Structural Biology at Lund University, Lund, Sweden. e-mail: derek.logan@biochemistry.lu.se

Closing the circle on ribonucleotide reductasesDerek T Logan

In this issue, a wide array of structural and biochemical techniques are applied to reveal the molecular details of activity regulation in one of life’s most essential enzymes, the ribonucleotide reductase.

α2

??

Probablelocation ofATP cone

β2

α2

ATP cone

Class Ia, aerobic Class III, anaerobic

4Fe-4S 4Fe-4S

Class II, dimeric Class II, monomeric

α

β β

β

Figure 1 The three different classes of RNR6,7,11,20,21. Models of the active class I α2β2 complex (based on the crystal structures of the R1 and R2 proteins, PDB 1RLR and 1RIB); class III (1HK8), dimeric class II (PDB 30O0) and monomeric class II (PDB 1LIL; C. Drennan, Massachusetts Institute of Technology, personal communication). The monomeric class II enzymes have maintained within one polypeptide chain the minimal structural elements required for substrate specificity regulation21. Substrates and allosteric effectors are shown as space-filling models. The adenosylcobalamin cofactors in the class II enzymes are shown as sticks. The box at the upper right shows a homology model of a β-monomer for the class III enzymes.

n e w s a n d v i e w s©

201

1 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

252 volume 18 number 3 mArCH 2011 nature structural & molecular biology

Remarkably, the overall pattern of allosteric regulation is the same across all three classes, despite their different radical chemistries and quaternary associations.

Although our understanding of the sub­strate specificity mechanisms is approach­ing maturity through structural studies on all three classes9–12, we do not yet under­stand how the minimal chemical differences between dATP and ATP can have opposing effects on activity or how the pattern has been conserved over large evolutionary dis­tances. For a long time, it was believed that dATP and ATP exerted subtle effects on the structure or affinity of the α2β2 complex; however, recently the importance of higher­order oligomers was highlighted, particularly for eukaryotic class I RNRs. Cooperman and Stubbe independently suggested α6β6 com­plexes13–15, but Hofer and co­workers pro­posed, using the highly sensitive gas­phase electrophoretic molecular mobility analysis (GEMMA) method, that both active and inactive complexes had α6β2 stoichio metry16 at physiological nucleotide concentrations.

In this issue, Fairman et al. demonstrate the value of a hybrid approach to structural analysis. They combine crystal structures with electron microscopy and back up the con­clusions with biophysical, biochemical and

mutational analysis. They present for the first time the structure of the human RNR and fur­ther fill in some details of specificity regula­tion. However, the major achievement of this research is in illuminating activity regulation: human RNR is described in complex with both dATP and ATP at the activity site in the ATP cone domain. dATP is seen to bind deeper into the pocket owing to the lack of a 2′­OH group, providing some clues as to its higher affinity.

Most importantly, Fairman et al. crystal­lized the dATP­induced α6 hexamer of the yeast enzyme. Although the structural model obtained was only at 6.6 Å resolution, this was sufficient to analyze the quaternary asso­ciation. In a fine piece of detective work, the authors used mutational analysis to distinguish between two possible hexamerization modes observed in the crystal. Only mutations based on one set of crystal contacts disrupted the oligomerization, as gauged by size­ exclusion chromatography. This established the geo­metry of the α6 hexamer, and based on the previously proposed interaction surfaces for α2 and β2, it was obvious that β2 would bind on the inside of the ring, excluding stoi­chiometries other than α6β2, as there is only space for one β2 dimer inside the ring (Fig. 3). This conclusion was supported by the final piece in the puzzle, a negatively stained electron microscopy reconstruction of the human α6β2 complex induced by dATP.

As often happens, this study raises as many questions as it answers. The low­resolution structure of an inactive class Ia RNR has been

revealed, but what about the active ATP com­plex? Interestingly, the mutations made by Fairman et al. to disrupt dATP­dependent hexamerization do not prevent the formation of α6 complexes by ATP, implying that these may have a different architecture. GEMMA analysis suggests that the active form also has an α6β2 composition16. How do the subtle con­formational differences observed between ATP and dATP binding to the ATP cone translate to macroscopic changes in the architecture of the α6β2 complex? Although Fairman et al. used the S. typhimurium intermediate com­plex8 to model the position of β2 inside the α6 ring, the α2β2 model for E. coli RNR6 (Fig. 1) also fits very well, suggesting that this associa­tion mode may be a common denominator for active class I RNR states across organisms and that small reorganizations of the α6 ring could induce different β2 binding modes. Higher­resolution studies will be required to analyze such changes. Finally, the possibility of an α6β6 stoichiometry for the ATP­activated active eukaryotic enzyme13–15 cannot be completely excluded and needs to be investigated further.

What about prokaryotic RNRs? Recent GEMMA studies on E. coli RNR suggest that the active and inactive complexes have α2β2 and α4β4 stoichiometry, respectively17. A simple modeling exercise (Fig. 3) seems to exclude a closed, symmetric form for the α4β4 complex. Because of the S­shaped α2 dimer, with ATP cones at each tip (Fig. 2), if both ATP cones contact each other, binding of β2 is excluded (Fig. 3). A more open structure

C Catalytic site

S Speci�city site

A Activity site

S

C

Loop 2

Loop 2

NDP

dATP, ATP

A

dATP, dTTP, dGTP

Figure 2 Schematic overview of the two modes of allosteric regulation in ribonucleotide reductases. The overall activity is governed by the binding of dATP (inhibition) or ATP (stimulation) to the activity site, located in a small N-terminal ATP cone domain of the α2 subunit of RNRs from class Ia and III. With very few exceptions, class II RNRs lack ATP cones and are not activity regulated. The substrate specificity is regulated by the binding of dNTPs to the specificity site at the dimer interface: ATP and dATP upregulate the reduction of CDP and UDP, dTTP upregulates GDP reduction and dGTP increases the rate of ADP reduction. Loop 2 is a flexible loop involved in transmission of the specificity signal.

Yeastmammalian

Dictyostelium discoideum

E. coliP. aeruginosa

β2

α6

β2β2

dATP

?

dATP

β2 bindingsurface blocked

α6β2

Figure 3 Different architectures of inactive RNR complexes. In eukaryotes, the α6β2 complex seems to predominate3,16,22. In E. coli17 and Pseudomonas aeruginosa23 RNR (M. Crona and B.-M. Sjöberg, Stockholm University, and A. Hofer, Umeå University, personal communication), dATP probably does not induce an α4β4 complex in which all ATP cones are in contact, as this would exclude β2 binding. Instead, a more open complex in which only two ATP cones (represented by the blue shapes in the lower part of the figure) are in contact seems more likely. Such organization would also explain why α4 tetramers are not observed alone but only in complex with β2.

n e w s a n d v i e w s©

201

1 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

nature structural & molecular biology volume 18 number 3 mArCH 2011 253

11. Larsson, K.M. et al. Nat. Struct. Mol. Biol. 11, 1142–1149 (2004).

12. Xu, H. et al. Proc. Natl. Acad. Sci. USA 103, 4022–4027 (2006).

13. Kashlan, O.B., Scott, C.P., Lear, J.D. & Cooperman, B.S. Biochemistry 41, 462–474 (2002).

14. Cooperman, B.S. & Kashlan, O.B. Adv. Enzyme Regul. 43, 167–182 (2003).

15. Wang, J., Lohman, G.J. & Stubbe, J. Proc. Natl. Acad. Sci. USA 104, 14324–14329 (2007).

16. Rofougaran, R., Vodnala, M. & Hofer, A. J. Biol. Chem. 281, 27705–27711 (2006).

17. Rofougaran, R., Crona, M., Vodnala, M., Sjöberg, B.-M. & Hofer, A. J. Biol. Chem. 283, 35310–35318 (2008).

18. Torrents, E., Eliasson, R., Wolpher, H., Gräslund, A. & Reichard, P. J. Biol. Chem. 276, 33488–33494 (2001).

19. Brown, N.C. & Reichard, P. J. Mol. Biol. 46, 39–55 (1969).

20. Logan, D.T., Andersson, J., Sjöberg, B.-M. & Nordlund, P. Science 283, 1499–1504 (1999).

21. Sintchak, M.D., Arjara, G., Kellogg, B.A., Stubbe, J. & Drennan, C.L. Nat. Struct. Biol. 9, 293–300 (2002).

22. Crona, M. Ph.D. thesis, Stockholm University (2010).

23. Torrents, E., Westman, M., Sahlin, M. & Sjöberg, B.M. J. Biol. Chem. 281, 25287–25296 (2006).

ACKNOWLEDGMENTSThe author thanks Anders Liljas for help with Figure 2.

COMPETING FINANCIAL INTERESTSThe author declares no competing financial interests.

1. Reichard, P. Biochem. Biophys. Res. Commun. 396, 19–23 (2010).

2. Nordlund, P. & Reichard, P. Annu. Rev. Biochem. 75, 681–706 (2006).

3. Fairman, J.W. et al. Nat. Struct. Mol. Biol. 18, 316–322 (2011).

4. Eklund, H., Uhlin, U., Färnegårdh, M., Logan, D.T. & Nordlund, P. Prog. Biophys. Mol. Biol. 77, 177–268 (2001).

5. Larsson, K.M., Logan, D.T. & Nordlund, P. ACS Chem. Biol. 5, 933–942 (2010).

6. Uhlin, U. & Eklund, H. Nature 370, 533–539 (1994).

7. Nordlund, P., Sjöberg, B.-M. & Eklund, H. Nature 345, 593–598 (1990).

8. Uppsten, M., Färnegårdh, M., Domkin, V. & Uhlin, U. J. Mol. Biol. 359, 365–377 (2006).

9. Eriksson, M. et al. Structure 5, 1077–1092 (1997).10. Larsson, K.-M., Andersson, J., Sjöberg, B.-M.,

Nordlund, P. & Logan, D.T. Structure 9, 739–750 (2001).

based on an incomplete version of the α6β2 ring seems more plausible (Fig. 3).

Finally, how has the pattern of activity regu­lation been conserved throughout evolution? The class III enzymes dimerize very differently and interact with a different type of β2 subunit (Fig. 1). Moreover, the radical is generated on α2 and can be used for many catalytic cycles before having to be regenerated18, unlike class Ia enzymes, whose subunits have to reassociate on each cycle. Does the ATP cone affect subunit interactions or act on the α2 subunit itself?

Given that α4β4 oligomers of E. coli RNR, the first sign of higher­order association, were first observed by Brown and Reichard more than 40 years ago19, it is exciting that the circle is now beginning to close on the structural basis for RNR activity regulation. The window that has just been opened is also a window of opportunity for future studies.

Glutamate receptor ion channels (iGluRs) are complex allosteric proteins that have gener­ated enormous interest in the neuroscience community because of their role in synaptic transmission, learning, memory formation and brain development1. The emergence of a comprehensive iGluR pharmacology in the 1980s, combined with the cloning of 18 iGluR genes 10 years later, led to the identification of numerous subtype­selective ligands and allosteric modulators. Current medicinal chemistry research programs in the pharma­ceutical industry continue to develop new iGluR ligands, several of which have thera­peutic potential for a range of neurological and psychiatric diseases. Questions not answered by these studies include what the precise mechanisms underlying the affinity of ligand

binding may be and why ‘partial agonists’ pro­duce less activation than ‘full agonists’ such as the neurotransmitter glutamate2,3.

On page 283 of this issue, Lau and Roux4

use a comprehensive set of all­atom molecular dynamics simulations for a panel of nine iGluR agonists, partial agonists and antagonists to address these and other important problems in the binding­gating conundrum for ligand­activated ion channels. Using this approach, they dissect the energetics of binding into components resulting from ligand interac­tions with the protein, on the one hand, from the free­energy changes resulting from ligand­induced conformational modifications on the other. Surprisingly, they find that the binding of glutamate is energetically unfavorable when it is docked into the ligand­binding domain (LBD), but this is compensated for by a large free­energy change produced by closure of the LBD (Fig. 1). They find that the energy result­ing from LBD closure is less for partial agonists than for full agonists, but there are unique landscapes for each of the nine ligands studied, highlighting the complexity of mechanisms that underlie the binding of small molecules to

proteins. Such a comprehensive analysis of the energetics of ligand binding to a neurotrans­mitter receptor has not been attempted before. It will undoubtedly be applied to other systems when crystal structures have been solved to provide the necessary molecular templates for molecular dynamics simulations.

Lau and Roux study the α­amino­3­hydroxy­5­methyl­4­isoxazolepropionic acid (AMPA) receptor GluA2, a model system for the receptor family that mediates fast synaptic transmission throughout the brain, and the only iGluR for which a full­length structure has been solved5. By contrast, during the past decade, numerous biochemical, crystallographic and NMR studies have focused on the LBDs of iGluRs, which can be genetically isolated and expressed as soluble recombinant proteins. GluA2, in particular, has been studied extensively, and more than 80 ligand complexes have been solved by X­ray crystallography6. These studies have generated models for activation7 and desensitization8 that have been amply substantiated by experimental tests9–12 and have also identified binding sites for allosteric modulators8,13,14. Despite these impressive advances, our understanding of the

Mark L. Mayer is at the Laboratory of Cellular and Molecular Neurophysiology, Porter Neuroscience Research Center, National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, Maryland, USA. e-mail: mayerm@mail.nih.gov

Glutamate receptor ion channels: where do all the calories go?Mark L Mayer

Glutamate receptor ion channels use the free energy of ligand binding to trigger ion channel activation and desensitization. In this issue, an analysis of all-atom molecular dynamics simulations dissects the binding process, reveals a substantial gain in free energy produced by domain closure for agonists and reports unique energy landscapes for individual ligands.

n e w s a n d v i e w s©

201

1 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.