The Structure of Pyrococcus Furiosus Glutamate Dehydrogenase Reveals

Home , Pyrococcus, Pyrococcus furiosus, Pyrococcus woesei, Thermus

View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Elsevier - Publisher Connector The structure of Pyrococcus furiosus glutamate dehydrogenase reveals a key role for ion-pair networks in maintaining enzyme stability at extreme temperatures KSP Yip', TJ Stillman', KL Britton1 , PJ Artymiuk', PJ Baker'1, SE Sedelnikova1, PC Engel' t, A Pasquo2 , R Chiaraluce2 , V Consalvi2, R Scandurra2 and DW Rice1*

1 The Krebs Institute for Biomolecular Research, Department of Molecular Biology and Biotechnology, University of Sheffield, PO Box 594, Sheffield S10 2UH, UK and 2 Dipartimento di Scienze Biochimiche, Universita La Sapienza, Roma 00185, Italy

Background: The hyperthermophile Pyrococcusfuriosus is glutamate dehydrogenase from the mesophile Clostridium one of the most thermostable organisms known, with symbiosum. an optimum growth temperature of 1000C. The proteins Conclusions: Comparison of the structures of these two from this organism display extreme thermostability. We enzymes has revealed one major difference: the structure have undertaken the structure determination of glutamate of the hyperthermophilic enzyme contains a striking series dehydrogenase from P. furiosus in order to gain further of ion-pair networks on the surface of the protein insights into the relationship between molecular structure subunits and buried at both interdomain and inter- and thermal stability. subunit interfaces. We propose that the formation of such Results: The structure of P. furiosus glutamate dehydro- extended networks may represent a major stabilizing genase, a homohexameric enzyme, has been determined feature associated with the adaptation of enzymes to at 2.2 A resolution and compared with the structure of extreme temperatures. Structure 15 November 1995, 3:1147-1158 Key words: glutamate dehydrogenase, hyperthermophile, Pyrococcus furiosus, thermal stability, X-ray crystallography

Introduction metabolism and which catalyze the oxidative deamina- The Archaea represent a phylogenetically distinct evo- tion of L-glutamate to 2-oxoglutarate and ammonia lutionary domain whose members possess various extra- with the concomitant reduction of one molecule of ordinary properties, which include the ability of some NAD(P)+ . Biochemical studies on the GluDHs from a species to grow at temperatures in excess of 100 0 C. diverse range of sources have shown that the vast major- Common habitats of these remarkable microorganisms ity of these enzymes are based on a hexamer of identical are submarine hydrothermal areas and continental solfa- subunits [8]. Analysis of the amino acid sequence for Pf taric fields [1-3]. Not surprisingly, the proteins from GluDH has shown that it displays considerable similarity these organisms display extreme thermostability and the to other members of this enzyme family [12,13]. In analysis of their structures is expected to provide insights particular, Pf GluDH shows 34% sequence identity with into the molecular mechanisms which Nature employs the GluDH from the mesophile Clostridium symbiosum in order to achieve this. Understanding the basis of (Cs) [14], a molecule of known three-dimensional this thermostability may allow us to engineer proteins to structure [15,16], which has a half-life at 520 C of withstand not only high temperatures but also environ- approximately 20 min [17]. The striking difference in ments even more hostile to life but relevant to the thermostability combined with the sequence similarity conditions used in many industrial processes [4]. between these two enzymes encouraged us to solve the structure of Pf GluDH. In this paper we describe the The archaeon Pyrococcusfuriosus (Pf), is one of the most structure determination of this enzyme and the results thermostable organisms known, with an optimum growth of the comparative analysis with the structure of the temperature of 100°C [5]. The glutamate dehydrogenase GluDH from Cs. This has allowed us to highlight struc- (GluDH) from this organism accounts for up to 20% of tural features of the Pf enzyme which may contribute to the total cell protein and is extremely thermostable, with its superior thermal properties. a half-life at 1000C of 12 h [6,7]. The abundance of GluDH, coupled with the wealth of biochemical data on this enzyme from a diverse range of mesophilic organisms Results and discussion [8], has created considerable interest in it as a model Description of the Pyrococcus furiosus GluDH structure system for the analysis of thermostability [9-11]. Overall, the fold of the GluDH from Pf is very similar to that of the hexameric GluDH from Cs GluDHs are ubiquitous enzymes which occupy an [15,16], enabling us to undertake a detailed comparison important branch point between carbon and nitrogen of their molecular structures. The secondary-structure

*Corresponding author. tPresent address: Department of Biochemistry, University of Dublin, Belfield, Dublin 4, Eire.

assignments and structure-based sequence alignment for In previous studies, a decrease in the lengths of the loops Pf and Cs GluDHs are given in Figure 1. A schematic connecting elements of regular secondary structure has representation of the structure of a single subunit of Pf been suggested to be important for thermostability [18]. GluDH is shown in Figure 2. The structure of a single Comparisons between the two GluDHs show that there subunit of this enzyme consists of two domains sepa- are only two regions of major structural difference. rated by a deep cleft. Domain I is largely composed of Firstly, at the N terminus, the Pf GluDH is some 21 residues from the N-terminal portion of the polypep- residues shorter, corresponding to the loss of helix al in tide chain and is responsible for directing the self- Cs GluDH (15 residues) and the replacement of two assembly of the subunits into a hexamer with 32 short helices, a3 and o4, by a loop. The structural symmetry, whereas domain II contains the nucleotide- rearrangement caused by these changes does indeed lead binding site. The domain interface is dominated by to a more compact character for Pf GluDH in this region interactions involving four long x helices (a6, a12, a13 (Fig. 3a). However, the sequence at the N terminus of and ot14 in Pf GluDH). The active-site pocket is located GluDH is the most variable region across the enzyme in the cleft region and contains residues provided by family. Helix axl is also deleted from the structure of both domains. A description of the structure of the GluDH from the mesophilic organism Neurospora crassa hyperthermophilic Pf GluDH is presented below and is (TJS, DWR, et al., unpublished data) and its absence in compared with Cs GluDH. many other mesophilic species, such as Saccharomyces cere- visiae, is implied from sequence alignment of GluDHs Analysis of secondary structure, hydrogen bonding and loops [14,19,20]. It would therefore appear that this difference The overall similarity between the two enzymes is illus- is not necessarily related to thermostability. Furthermore, trated in Figure 1 and Table 1. The percentage of resi- the shortening of sequences at the N terminus has been dues in helical -or -strand conformations is virtually noted elsewhere and has led to the suggestion that this identical. Furthermore, it is clear from Table 1 that the feature may merely reflect phylogenetic differences [18]. number of main-chain hydrogen bonds, and the average number of protein-protein hydrogen bonds per residue, A second region of significant structural and sequence are very similar and thus the differences between these difference occurs between strands 13i and 3k (residues enzymes in these respects are also very subtle. 267-316 in the Cs enzyme), as shown in Figure 3b. In

GDH(CS) 1 kydrvia.vekkyadEPEFVQtvvls..lgpvvdahpy..vALLMZPVIPRvEFRVP ZDDNGKVHVNTGYRVQFN82 GDH(Pf) 1 veQDPYEIvikqleraaqy ------ieEALEFLPQRIVEVTIPV NDiDGSVKVFTGFRVQHN61

GDH(Cs) 83 GAIGPYKGGLRFAPSVNLSIMKFLGFEQAIKDSLTLPMGGAKGGSOFDPNGKSDREVMRFCQAPIFTELYRHIGPDIDVPAG164 GDH(Pf) 62 WARGPTKGGIRWHPEETLSTVALAA AVMDLPYGKGG IVDPLSDp RLAR YIRAIYDVISYEDIPAP 143

GDH(Cs ) 165 DLGGAREI GYMYGQYRKIVGGf-- yNGVLTGKARSFGGSLVRPEATGYGSVYYVEAVMID -tLVGKTVALAGGNVAW 243 GDH(Pf) 144 DVYTNPQIKAWEDEYETISRRktpaFGIITGK LSIGGSLGRIEATARGASYTIREAAXV ItLKGKTIAIQGYGNAGY 225

GDH(C) 244 GAAKLAE-LGAKAVTLSGPYGYIYDPEGittekiylUasLqrkvqdydkfgqVQFFpg POWIQVDXInPCATQN 324 GDH(Pf) 226 YLATIMSEdFGM VAVSDSKGGIYNP1 ----.dvYkWkneh---- vkWdfgATNIt.. .LLeLVDVILAPAAIZZ 299 Fig. 1. Alignment of the sequences for Clostridium symbiosum (Cs) and Pyro- coccus furiosus (Pf) GluDH based on the superposition of the respective Ca coordinates of each domain. Equivalent residues whose relative positions differ GDH(Cs) 325 DVDLZQAIV VKYIVAnTYYI EVA QNWAPSKA GVLVSGFUMSQNSI IrL8WTAEESVDMNHq 406 GDH(Pf) 300 VITI0XADNI ---kaKIWVANGPVTPEADEILFEK-qILIPDFLNAGGVTVSyFELQNIX CTGYY6IEEVRULD 377 by greater than 2 A are indicated by the use of lower-case letters. The secondary dr1 structural elements of the three-dimensional structure of the two GluDHs are shown with helices represented as cylinders and strands shown as arrows.

GDH(Cs) 407 TDIHDGSAA..ryggynlVAANIVGFQKIADAO Gi-W 449 The residue numbers for the individual CDH(Pf) 378 TXAFYDVYNiakkni--h.RDAAYVVAVQRVYQAILaDRvkh 419 sequences are given at the beginning and end of each line. Positions of del- otirnc ,ro inrlirtrrl hv, - _LV.IU~I~_ 1 ...... ( ._1C .. i P. furiosus glutamate dehydrogenase Yip et a. 1149

Fig. 2. Stereo diagrams of a single subunit of Pf GluDH. The organization of the subunit into two domains, separated by a deep cleft, can be seen. In this view, the threefold axis of the GuDH hexamer runs vertically and domain I, responsible for the major intersubunit interactions lies uppermost with domain II, which contains the nucleotide-binding site, in the lower portion of the figure. (a) Schematic representation with strands (a-m) and helices (1-14) marked. (b) Ca trace with every tenth residue indicated by a black dot. (Figure prepared using MOLSCRIPT [56].)

contrast, PfGluDH can Table 1. Comparison of the secondary structures of the two use either cofactor (though with GluDHs. a preference for NADP) and in that respect has similar coenzyme specificity to the vertebrate GluDHs. The Cs Pf latter show an identical pattern of insertions/deletions with respect to Cs GluDH as is shown by the Pf enzyme. No. of residues per subunit 449 419 No. (%) of residues Therefore the changes in this region are thought to in helical conformation 224 (50%) 208 (50%) reflect differences in nucleotide specificity rather than No. (%) of residues differences in thermostability. in P conformation 67 (15%) 66 (16%) No. (%of total possible) Deletions and insertions at a total of seven other main-chain H-bonds 310 (69%) 287 (68%) sites lead No. of intrasubunit to a net loss of only one residue in the Pfenzyme relative protein-protein H-bonds 582 578 to Cs GluDH. An essential feature of the catalytic cycle No. of intersubunit of GluDH is a large-scale conformational change which protein-protein H-bonds* 171 163 closes the cleft between the two domains and is important No. of protein-protein H-bonds per residue 1.36 1.44 for the appropriate positioning of the partners involved in hydride transfer [16]. Two of these sites of insertion/deletion (after residues 187 and 422 in Cs GluDH) lie close *This is the total for the hexamer. to points in the structure which move during this conformational rearrangement [16]. Clearly, these sequence Cs GluDH, this region folds to form oll, a12, P3j and a differences may modulate the conformational flexibility long loop to 13k. In Pf GluDH, the helix equivalent to of the enzyme. However, within the wider family of al 1 is shortened by the removal of four residues from its GluDHs, small insertions or deletions are often observed N terminus and the deletion of a further four residues by at these points [14,19,20]. Overall, therefore, the lack of the replacement of helix a12 by a loop in the PfYenzyme. any significant differences in the size of loops between the However, sequence similarities [14] suggest that this Pf and Cs enzymes implies that such changes are not a deletion and the consequent structural change also occur major determinant of thermostability of GluDH. in the GluDHs from vertebrate sources which are not known to exhibit enhanced thermostability. In the Analysis of temperature factors NAD-linked Cs GluDH, residues from this region, and We have compared the refined temperature factors of the in particular al 1 to a12, provide important determinants two enzymes in an attempt to see if they are correlated for the binding of the adenine ribose moiety [21]. In with enzyme stability. The results of such an analysis have 1150 Structure 1995, Vol 3 No 11

Fig. 3. Superpositions of Cs (blue) and Pf (green) GluDHs. (a) The N-terminal regions (l-a5 in Cs, a1--a2 in Pf) are shown. Helices a2 (Cs) and al (Pf) superimpose well as do helices 5 (Cs) and a2 (Pf). However, a (Cs) is completely absent in Pf and the loop l-a2 in Pf is much shorter, replacing 3 and a4 in Cs. (b) Superposition of the region 13h-i3j which, in both GluDHs, forms the recognition site for the adenine ribose moiety. (Figure prepared using MIDASPLUS [57,58].)

to be interpreted carefully, because crystal packing con- factors shows that there are 47 and 69 residues in the rep- tacts can have a significant effect on the observed tem- resentative subunits of the Cs and Pfenzymes which have perature factors. The Vm of the Pf enzyme is unusually averaged side-chain B factors exceeding 50 A2. There- high at 4.2 A3 Da-l [22] and the crystal lattice is domi- fore, in both enzymes, equivalent numbers of residues are nated by the presence of cylindrical cavities between the partially disordered. Thus, an anticipated reduction in molecules, each 70 A long and 80 A in diameter, with the mean temperature for the Pf enzyme was not only a small number of contacts between the molecules observed. However, the substantially different packing involving residues from domain II. In contrast, the Vm arrangements in the crystals may well influence these for Cs GluDH is 2.6 A3 Da- l [23], which is close to the results. Analysis of further crystal forms could be an average observed for protein crystals [24] and there are a important next step in furthering our understanding of large number of lattice contacts involving atoms from the relationship between stability and flexibility. both domains. Such contacts might be expected to lower the thermal parameters of the Cs enzyme. Comparisons of accessible surface area and packing density Recent studies on the aldehyde ferredoxin oxidoreduc- The overall B factors for all atoms for representative sub- tase from Pf [25] have led to the suggestion that the units of Cs and Pf GluDHs are 29 A2 and 35 A2. The reduction of the solvent-accessible area and an increase in values for the main-chain atoms of residues from the fraction of buried atoms are possible contributing domains I and II of Cs GluDH are 21 A2 and 27 A2 factors to thermostability. However, no mesophilic coun- respectively, and for Pf GluDH are 26 A2 and 34 A2 terpart of the oxidoreductase was available for compari- respectively. A schematic representation of the variation son. The results of the analysis of the solvent-accessible in B factors of the Coa atoms is given in Figure 4. Despite area and packing density in Cs and PfGluDHs are shown the different crystal environments, this plot shows that a in Table 2. This shows that the nature of the accessible remarkable agreement exists between the pattern of tem- surface area of Cs GluDH is very similar to that found in perature factors. In both enzymes, the N terminus and the Pf enzyme, although a higher fraction of the surface the whole of domain II are considerably more mobile is charged and a lower fraction is polar for the Pf enzyme. than the rest of the structure. This suggests that the flexi- However, we note that the fraction of charged surface in bility of the main chain is not dramatically affected by Pf GluDH is significantly higher (27%) than the value of the stabilization of the enzyme to extreme temperatures. 19% reported in a survey of 37 monomeric proteins Similarly, inspection of the side-chain temperature refined at 2.5 A or better [26].

Fig. 4. A 'glowing coals' representation to illustrate the B factors of the main- chain Ca atoms of (a) Pf GluDH and (b) Cs GluDH. The colours from black to white through red and yellow indicate increasing temperature factors. The colour key is black (0-5 A2), indigo (5-15 A2), dark red (15-25 A2), orange (25-35 A2), yellow (35-45 A2), pale yellow (45-55 A2) and white (above 55 A2). P. furiosus glutamate dehydrogenase Yip et al. 1151

Table 2. Comparison of the molecular surface and packing The solvent accessibility calculations also show that very efficiency of the two GluDHs. similar percentages of atoms are totally buried within the subunits, and at subunit interfaces, with 61% and 62% of Cs Pf the total number of atoms buried in the hexamers of Cs Solvent-accessible area and Pf GluDH respectively (Table 2). Thus, the general of a monomer (A2)* 18500 17000 trends observed in this study for the solvent-accessible Solvent-accessible area area and fraction of buried atoms are in the same direc- of a hexamer (A2)* 88 000 82 500 tion as reported by Chan et al. [25]. However, the differ- Accessible surface comprising: ences between the two GluDHs are very small. In non-polar residues (%) 48 48 polar residues (%) 28 25 agreement with our findings, comparative analysis of charged residues (%) 24 27 glyceraldehyde 3-phosphate dehydrogenase from ther- No./(% of total) of buried mophilic and mesophilic sources, revealed little difference atoms in the monomer 1963/57 1902/58 in surface area [27]. We therefore believe that the differ- No./(% of total) of buried ences we have observed in surface area, and in the num- atoms in the hexamer 12 720/61 12 204/62 No./(% of total) of atoms buried ber of atoms buried, may not be major factors governing between subun.its 157/6 132/7 protein stability of the hyperthermophilic Pf GluDH. Mean packing densityt 0.733 0.739 In a study which compared the structures of citrate syn- *Rounded to the nearest 500 A2. tDefined as the ratio of the thase from the moderate thermophile Thermoplasma aci- volume enclosed by the van der Waals envelope of the given dophilum with the equivalent pig enzyme, Russell et al. atom to the actual volume of space that it occupies. Reported values are calculated for all residues in a monomer that are [18] noted a reduced number of buried cavities in the completely buried in the structure. thermophilic enzyme, implying that superior van der Waals interactions might be a feature of more stable enzymes. Previous studies [28,29] had shown that the The solvent-accessible area for a single subunit of Pf average packing density for the interior of proteins is GluDH is approximately 1500 A2 lower (8% of the total) generally close to 0.75, which is in the middle of the than that of its mesophilic counterpart (Table 2). How- range found for crystals of most organic molecules. ever, comparison of the surface areas for the hexamers, However, within a protein structure, the packing densi- shows that a larger surface area of each subunit is buried ties (averaged over a relatively small number of atoms) on assembly of the Cs enzyme (3900A 2 per subunit, have been shown to vary substantially in different parts of compared with 3350 A2 for Pf GluDH) and thus, the the same protein. This led to the suggestion [28] that low overall difference in surface area for the hexamer, is only densities represent packing defects which may permit rel- 6% lower (5500 A2) in Pf GluDH. Despite this apparent atively easy motion, whereas the surrounding areas of difference, which is in the direction predicted from the high density may serve as relatively incompressible studies on the aldehyde ferredoxin oxidoreductase from regions which transmit or correlate motions over consid- Pf [25], it must be borne in mind that the Pf GluDH erable distances. A natural extension of this argument is subunit is 30 residues shorter than this particular that, at any given temperature, improved packing within mesophilic GluDH and a smaller surface area is therefore the structures of the enzymes from hyperthermophiles to be expected. leads to lower levels of thermal motion than would be observed in their mesophilic counterparts. As observed above, the differences in length between the two proteins are concentrated in two regions, corre- Calculation of the mean packing density of the residues sponding to changes at the N terminus and close to the that are completely buried in the Cs and Pf GluDH adenine ribose binding site. If we calculate the surface monomer structures yields values of 0.733 and 0.739 area of the hexamer of Cs GluDH, having removed the respectively. These values are not only close to each additional 17 residues which lie at the N terminus and other but they are very close to 0.74, the density for which are not found in the Pf enzyme, a reduction in close-packed spheres. Is the slightly higher packing den- surface area of 2900 A2 is observed. In addition, if we sity of the Pf enzyme significant? In a recent study on replace residues 271-300 in the structure of the Cs volume changes on protein folding, Harpaz et al. [30] enzyme, which lie close to the adenine ribose recogni- calculated the average volumes of all the totally buried tion site, by their equivalents in Pf GluDH (residues residues in 108 proteins whose structures had been 254-275), a further reduction of 1400 A2 is seen. Thus, determined at high resolution (between 1.0 A and 1.9 A) the differences in these two regions of the structure alone and had been refined to R factors of 20% or less with account for almost 80% of the difference in surface area good stereochemistry. If we calculate the expected pack- between these two enzymes. It is therefore possible that ing density for Cs and PfGluDH, using the particular set the reduction in surface area observed between Cs and Pf of residues that are completely buried in these two struc- GluDHs merely reflects the insertions and deletions tures, and values for the mean volumes of those residues between the structures. And, as noted above, the inser- taken from the study of Harpaz et al. [30], the expected tions and deletions may be unrelated to thermal stability densities are 0.741 and 0.744 respectively. Thus, whilst because they have counterparts in mesophilic GluDHs. the observed packing density for Pf GluDH is slightly 1152 Structure 1995, Vol 3 No 11

higher than that of the Cs enzyme, it is lower than that calculated from the volumes occupied by residues in a reference set of proteins, very few of which are from moderate or extreme thermophiles. Once again, the marginal differences between the mesophilic and hyperthermophilic enzymes make it unlikely that packing density differences play a major role in determining the thermostability of PfGluDH.

In a previous homology-based modelling study on the hyperthermophilic GluDHs from Thermococcus litoralis and Pf, which are closely related in amino acid sequence but where the Pf enzyme is slightly more thermostable, potential changes in packing density had been highlighted as a contributing factor to this increased stability [14]. In particular, seven valine (less-stable enzyme) Fig. 5. The cluster of isoleucine residues in the core of domain I to isoleucine (more-stable enzyme) substitutions were in the PfGIuDH structure. (Drawn with MIDASPLUS [57,58].) observed among the 53 sequence differences between these two GluDHs. This study also highlighted the formation of a cluster of isoleucines in the buried core of Inspection of Table 3 clearly indicates that a higher pro- one of the two domains of the more stable protein. On portion of charged residues participate in the formation the basis of the modelling, this study concluded that of ion-pairs in the hyperthermostable enzyme, with 40% given the apparent conservation of residues in the regions and 54% of the charged residues being involved in such neighbouring the valine to isoleucine substitutions, the interactions in Cs and Pf GluDH, respectively. In addi- additional methyl groups on the isoleucine might fill tion, despite the similar number of arginines in the two small voids in the structure and lead to stabilization enzymes (20 and 18 in Pf and Cs GluDH respectively), through superior van der Waals interactions. However, there is a strong preference for the utilization of arginine the similarity of the packing density of the Cs and Pf in such ion-pairs, with 64% of the ion-pairs in Pf GluDH GluDHs implies that the contribution played by changes involving arginine compared with only 46% in the Cs in packing density in stabilizing enzyme structures is rela- enzyme. This can be seen most dramatically in the tively small. Nevertheless, as predicted by the modelling involvement of 90% of all the arginine residues in f study, the hyperthermostable enzyme contains a cluster GluDH in ion-pair formation. As a possible explanation of five isoleucines within the core of the N-terminal for this, we note that 14 of the 18 arginines per subunit domain (Fig. 5) and the frequency of valine to isoleucine used in ion-pair interactions in Pf GluDH are associated substitutions could be significant. Further work will be with multiple hydrogen bonds between the guanidinium required in this area to establish the precise role of these group of the arginine and the carboxylate groups of the sequence changes on thermostability.

Analysis of ion-pairs Table 3. Comparison of the ion-pair interactions within the All the analyses described above reveal no significant dif- two GluDHs. and hyperthermophilic t ferences between the mesophilic Cs Pf Survey GluDH structures. However, there is one respect in which the two structures are strikingly different. As No. of ion-pairs per subunit 26 45 shown in Table 3, there is a significant increase in the No. of ion-pairs per residue 0.06 0.11 0.04 % charged residues number, and a difference in the distribution of ion-pair forming ion-pairs 40 54 29 interactions in the hyperthermophilic enzyme relative to % ion-pairs formed by Arg/Lys/His 46/31/23 64/27/9 the mesophilic one. There are 45 ion-pairs (involving 63 % ion-pairs formed by Asp/Glu 46/54 47/53 residues) in total per subunit in the structure of Pf % of all Arg forming ion-pairs 55 90 38 No. of residues forming 2 ion-pairs 6 17 GluDH compared with only 26 (involving 44 residues) No. of residues forming 3 ion-pairs 1 5 for the Cs enzyme, an average of 0.11 and 0.06 ion-pairs No. of 2/3/4 residue networks* 72/24/12 54/24/12 per residue in the Pf and Cs structures respectively. In No. of 5/6/18 residue networks* 0/0/0 12/6/3 % ion-pairs in networks addition, the proportion of ion-pairs in Pf GluDH is sig- of >3 residues* 23 62 17 nificantly higher than the average value of 0.04 per No. of intersubunit ion-pairs* 30 54 residue observed in a survey of 38 high-resolution pro- No. of interdomain ion-pairs 1 7 tein structures conducted by Barlow and Thornton [31] and consistent with the number of ion-pairs found in Ion-pair reactions were identified using the criteria of Barlow P furiosus aldehyde ferredoxin oxidoreductase [25]. The and Thornton [31]. *Comparisons of the ion-pair networks distribution of the residues involved in the formation of and the number of intersubunit ion-pairs refer to the number of such interactions in the hexamer. 1Survey by Barlow and these ion-pairs on a single subunit of the Pf and Cs Thornton [31]. GluDHs can be seen in Figures 6a and 6b respectively. P. furiosusglutamate dehydrogenase Yip et al. 1153

Fig. 6. Ion-pair distribution. (a) Stereo- view of a monomer of Pf GIuDH. (b) Stereoview of a monomer of Cs GluDH. The polypeptide backbone of each enzyme is represented in white. The side chains of amino acid residues which are involved in ion-pairs in each of the respective structures are highlighted in blue (lysine, arginine and histidine) and red (aspartate and glutamate). The viewing direction is approximately the same as in Figure 2. (c) Space-filling representation of the trimer of PfGluDH. (d) Space-filling representation of the trimer of Cs GIuDH. Both trimers are viewed down the threefold axis and show the face of the trimer buried on hexamer formation. The basic and acidic side chains involved in the formation of ion-pairs are coloured as in (a). The higher proportion of ion-pairs in the PfGluDH structure is clear. (Drawn with MIDASPLUS [57,58].) acidic partner in the ion-pair (e.g. see Fig. 7a). In the interface (Fig. 7). Furthermore, the number of ion-pairs ion-pairs involving lysine, no equivalent bidentate inter- participating in intersubunit interactions is dramatically actions are observed. The stabilization afforded by increased in the hyperthermophilic enzyme. Thus, at the these additional interactions may account for the more twofold axis relating dimers, networks of ion-pairs domi- frequent participation of arginine in the ion-pairs. nate the interface with two networks of three residues and one network of eighteen residues being found A second remarkable feature of the difference in ion-pairs (Figs 8,9). Nearby, a further two symmetry-related clus- between the two structures is in their distribution. In the ters of six residues (each formed by Arg164, Arg165 and Pf enzyme, a number of extensive ion-pair networks can Lys166 from subunit A, Arg35 and Glu138 from subunit be found which involve either five, six or eighteen B and Asp132 from subunit A# [nomenclature as in residues. These networks criss-cross the surface of the Pf Fig. 9]) form further intersubunit ion-pair networks. GluDH subunit and straddle the subunit interfaces, to an extent not found in the mesophilic enzyme. For example, The existence of extended ion-pair networks in -Pf in Cs GluDH, there is only one ion-pair between the GluDH necessarily implies that individual residues are two domains (between Lys195 and Glu383). This inter- involved in multiple interactions. Among the triple ion- action is conserved in the Pf enzyme, but a further two pairs seen in PfGluDH, examples are found of both pos- networks involving five residues each contribute a total of itively and negatively charged amino acids at the centre six ion-pairs which link the two domains and the domain of the cluster. Residues at the centre of triple ion-pairs 1154 Structure 1995, Vol 3 No 11

important, as recent site-directed mutagenesis studies have -suggested that small arrays of ion-pairs are not crucial in generating stability [34,35].

Without the advantage of direct comparison with a mesophilic counterpart, other recent analyses of the structures of enzymes from hyperthermophiles have suggested that, rather than being the consequence of one dominant type of interaction, the stabilization of enzymes to extreme temperatures arises from the subtle addition of many different contributing factors [25,36]. This was also the conclusion drawn from the comparison of citrate synthase from mesophilic and moderately thermophilic species [18]. On the basis of those studies, there would appear to be no simple 'prescription' for introdu- cing hyperthermostability into a protein. However, the present analysis provides compelling evidence for a significant difference in the number and nature of ion-pair networks between the structures of the mesophilic and hyperthermophilic GluDHs, and suggests that this difference may represent the dominant interaction in the adaptation ofP-fGluDH to extreme temperatures.

Following an early modelling study of the structures of ferredoxins with differential thermal stability, Perutz [37] proposed ion-pairs as a major contributing factor to thermostability. Moreover, comparison of the structures of mesophilic and thermophilic variants of rubredoxin [38], malate dehydrogenase [39] and ribonuclease H [40] have suggested a role for electrostatic interactions in generating Fig. 7. Ion-pair formation at the interface between the two stability. However, in the absence of dramatic differences domains of the Pf GuDH subunit. The part of the polypeptide between the structures of equivalent enzymes of varying backbone belonging to domain I is shown in yellow whilst that thermostabilities, the validity of these studies have not of domain II is in green. The side chains of amino acid residues been widely accepted. Other researchers have argued that which are involved in ion-pairs in the structure are highlighted in blue (lysine, arginine and histidine) and red (aspartate and gluta- ion-pairs are not crucial in maintaining stability at high mate). Potential hydrogen bonds are indicated by dashed lines. temperatures [34,35,41], and have implicated a diverse (a) The ion-pair interactions centred on a cluster of five residues range of other factors [18,25,42]. Furthermore, some the- at the domain interface and involving a triple ion-pair of Arg396 oretical calculations have been used to show that the pres- with three acidic residues. (b) A second cluster of five residues ence of polar groups at interfaces has a destabilizing effect, which form ion-pairs linking the two domains. (Drawn with MIDASPLUS [57,58].) reducing unfolding free energies (reviewed in [43]). This raises the question as to whether it is even possible for ion-pairs to contribute to the thermodynamic stability of include Glu120, Arg124 and Glu158 which are part of an enzyme. If, as we suggest here, the ion-pairs are the extended network of 18 ion-pairs stabilizing the responsible for maintenance of enzyme activity for long intersubunit interactions (Figs 8b,9); Glu138 which is periods at high temperatures, then it might be that this involved in the intersubunit network involving six stabilization occurs via a kinetic rather than a thermo- residues described above; and Arg396 which is involved dynamic mechanism. However, microcalorimetric data on in a network of five residues at the domain interface PfGluDH have established that the transition temperature (Fig. 7a). These multiple interactions may well provide for unfolding of this enzyme is 113°C, from which the significant stability to critical parts of the structure and undoubted thermodynamic stability of the enzyme can contribute towards increased thermostability. For exam- be seen [44]. Thus, if the ion-pair networks are responsi- ple, studies on glutathione reductase [32] have shown ble for this thermodynamic stabilization, then cooperativ- that the multiple interactions made by the 2'-phosphate ity between the interactions in the ion-pair networks may group of the NADPH cofactor with three positively play an important role. charged residues provides a major part of the binding energy for this moiety; and mutagenic studies on barnase In recent studies on glyceraldehyde 3-phosphate dehydro- [33] have indicated that creating ion-pair networks is genase from Thermotoga maritima (an organism with an energetically more favourable than forming an equivalent optimum growth temperature of 80°C) it was postulated number of single interactions. Furthermore, it may be that ion-pairs may play a role in stability [27], but it was that the extended nature of the networks of ion-pairs is also concluded that increased hydrophobic interactions P. furiosusglutamate dehydrogenase Yip et al. 1155

Fig. 8. The extensive ion-pair network at one of the twofold axes relating dimers of the Pf hexamer. (a) Stereoview of the region. The Ca backbone for each independent subunit involved is shown in a different colour for clarity (subunits A and B are green and white respectively, with their symmetry-related subunits in yellow and magenta). Residues highlighted in the picture correspond to those which participate in ion-pair interactions and are coloured red (aspartate and glutamate) and blue (arginine, lysine and histidine). Cen- tral to the diagram is the extensive network of 18 ion-pairs which stabilizes the intersubunit interface. (b) Close-up view depicting the detailed interactions at the centre of the 18-residue ion-pair cluster, centred on Arg124, Arg128 and Glu158. Residues labelled # are related by the twofold axis. (Drawn with MIDASPLUS 157,58].)

between the subunits, rather than additional intersubunit ion-pairs, might lead to stabilization of these interfaces. In contrast to this, our studies on the GluDH from the more extreme hyperthermophile P? furiosus (an organism capable of growth at 1050 C) demonstrate that a series of highly organized and optimally aligned ion-pairs form critical components of both the subunit surface and the subunit interfaces in this hyperthermophilic enzyme.

For the future, what remains to be verified by mutagenesis of members of the GluDH family and, by the structural analysis of other enzyme families, is whether the introduction of ion-pair networks is indeed the origin of the hyperthermostability in GluDH and whether such a mechanism is generic or one of many possible routes by which proteins may be stabilized. Furthermore, even if the introduction of ion-pair networks is capable of giving rise to extreme thermostability in proteins, it may not be simple to recreate such hyperthermostability in other mesophilic enzymes by the addition of further charged residues by site-directed mutagenesis. The kingdom of the Archaea is thought to include the most ancient organisms on Earth. On an evolutionary timescale, it is probable that environmental pressures have resulted in subtle structural changes elsewhere in the protein which allow the interactions of these ion-pairs to be optimized. It may also be necessary, therefore, in order to stabilize mesophilic enzymes for biotechnological and medical applications, to introduce ancillary mutations that Fig. 9. A schematic representation of the ion-pair network shown improve the geometry of ion-pair networks, and thereby in Figure 8. The twofold axis relating dimers is indicated by the strengthen their interactions. If this is the case, such dyad symbol. Two of the subunits of each trimer are labelled A changes will clearly stretch the limits of our current and B with their respective twofold symmetry-related partners understanding of protein structure and of our ability to labelled A# and B#. The individual amino acid residues which are involved in ion-pair interactions, are indicated using the predict the structural consequences of mutations. How- three-letter amino acid code followed by their sequence number. ever, in the longer term, it is clear that the application of Dotted lines connect ion-pair partners. ideas derived from structural comparisons such as those 1156 Structure 1995, Vol 3 No 11

presented here, will have important consequences for the Materials and methods design of a new generation of 'extremozymes' with Crystallization and X-ray data collection enormous potential in biotechnology and medicine. GluDH from P. furiosus was purified using the procedure described previously [22] and crystallized by the hanging drop method of vapour diffusion using lithium sulphate as the pre- Biological implications cipitant [22]. The crystals belong to space group P4221 2; with cell dimensions a=b=167.2 A, c=173.0 A. There is a trimer in Currently, the prediction of the nature of amino the asymmetric unit and a solvent content of 67%. Native acid changes required to produce a more (115 181 independent reflections to a dmin of 2.2 A, a redun- thermostable protein remains a major unsolved dancy of 2.58, 93.2% complete and Rmerge=4.3%) and deriva- problem in structural biology. Previous studies tive (crystal soaked in 5 mM potassium tetracyanoplatinate for have indicated that protein thermostability arises 24 h; 70477 independent reflections to a dmin of 2.5 A, a from the subtle combination of several forces redundancy of 2.79, 83.5% complete and Rmerge =6.5%) data including electrostatic effects, hydrophobic inter- sets were recorded on station PX9.5 at the Synchrotron Radia- actions, hydrogen bonds, and changes in packing tion Source (SRS), Daresbury using a wavelength of 0.92 A density, ultimately leading to decreased flexibility and a 30 cm diameter Mar-research image plate system. Rmerge of the polypeptide chain. With the aim of pro- is defined as ]hkI-I [/2Ul,, where I and Im are the observed intensity and mean intensity of related reflections, viding further insights into this problem, increas- respectively. The SRS data sets were processed using the ing interest has been shown in the study of MOSFLM package [45] and all subsequent processing involved proteins from extremophilic microorganisms programs from the CCP4 suite [46] unless otherwise stated. which are able to grow above the temperature of boiling water. These hyperthermophiles all Structure determination belong to the Archaea, a group of organisms The structure was initially solved by molecular replacement with unique physiological, biochemical and using the C. symbiosum GluDH structure [15,16] as a search genetic properties. The analysis of the structures model. The cross-rotation function gave a clear solution which of the proteins from these organisms is expected produced a peak in the translation function with a height of 18c. This model was refined using TNT [47], but the subsequent to provide crucial insights into the molecular map calculated using the refined phases was not of good quality. mechanisms which underlie their adaptation to However, the phases were sufficient to reveal the heavy-atom such extreme conditions. positions of the single derivative by difference Fourier methods. Heavy-atom refinement of the derivative was performed using The glutamate dehydrogenases from the hyper- MLPHARE [48], and this 12-site derivative had a phasing thermophile Pyrococcus furiosus and the mesophile power of 1.84 and 1.58 for the acentric and centric reflections, Clostridium symbiosum differ markedly in their ther- respectively. The overall figure of merit for the subsequent mostability, with the former having a half-life at isomorphous map for all reflections to 2.5 A resolution was 0.44. 100°C of 12 hours, and the latter a half-life at Construction of the model and refinement 52°C of approximately 20 minutes. Comparing The isomorphous map was of sufficient quality to unambigu- the structures of these two enzymes has shown ously position a model for a trimer of P. furiosus GluDH, and that many of the physical parameters thought to molecular masks for the molecule were generated using the be involved in the maintenance of thermostability program MAMA [49]. Forty cycles of solvent flattening, are, in fact, very similar between them. Thus, threefold molecular averaging and histogram matching using there are only subtle differences in the extent the program DM (K Cowtan, unpublished program) reduced of secondary structure formation, the number the free R factor to 0.34. In the subsequent electron-density of protein-protein hydrogen bonds, the solvent- map, calculated using the averaged phases, clear density could accessible area, the fraction of buried atoms, and be found for all but the first two residues of the structure and the packing density. One difference in structure therefore a model comprising 417 of the 419 amino acids of between these enzymes, the origins of which are the P. furiosus GluDH was built using the graphics program FRODO [50], and coordinates for the second and third sub- currently unclear, involves the formation of a units were generated by non-crystallographic symmetry. Sev- cluster of isoleucine side chains in the core of one eral cycles of rebuilding and refinement using TNT and of the domains of the P furiosus enzyme. However, FRODO, gave a final R factor of 0.17 (115153 reflections a much more significant and striking difference 10-2.2 A, 10224 atoms including 318 water molecules and six concerns the formation of a series of extensive sulphate ions), an rms bond deviation of 0.011 A, an rms angle networks of ion-pairs in the hyperthermophilic deviation of 2.85°, an rms planar group deviation of 0.018 A, enzyme. Arginine side chains, on the surface of an rms trigonal atom non-planarity of 0.013 A and an average 2 the protein subunits and buried at both inter- B factor for the trimer of 38 A . A representative portion of domain and intersubunit interfaces, are preferen- the final electron density map is shown in Figure 10. tially involved in these ion-pair networks. We Structure analysis and comparison suggest that the formation of such extended net- Main-chain hydrogen bonds and the extent of the elements of works of ion-pairs may represent a major stabiliz- secondary structure in the two GluDHs were assigned using ing feature associated with the adaptation of the program PROCHECK [51] coupled to a visual inspection enzymes to extreme temperatures. of the coordinates. The overall number of protein-protein hydrogen bonds was determined using the criteria of Baker and P. furiosus glutamate dehydrogenase Yip et al. 1157

6. Consalvi, V., Chiaraluce, R., Politi, L., Vaccaro, R., De Rosa, M. & Scandurra, R. (1991). Extremely thermostable glutamate dehydrogenase from the hyperthermophilic archaebacterium Pyrococcus furiosus. Eur. J. Biochem. 202, 1189-1196. 7. Robb, F.T., Park, J.B. & Adams, M.W.W. (1992). Characterization of an extremely thermostable glutamate dehydrogenase; a key enzyme in the primary metabolism of the hyperthermophilic archaebacterium Pyrococcus furiosus. Biochim. Biophys. Acta 1120, 267-272. 8. Smith, E.L., Austen, B.M., Blumenthal, K.M. & Nyc, J.F. (1975). Glutamate dehydrogenases. In The Enzymes (3rd edition). (Boyer, P.D., ed), vol. 11, pp. 293-367, Academic Press, New York. 9. Maras, B., et al., & Barra, D. (1992). The protein sequence of glutamate dehydrogenase from Sulfolobus solfataricus, a thermo- acidophilic archaebacterium. Eur. J. Biochem. 203, 81-87. 10. DiRuggiero, J., et al., & Adams, M.W.W. (1993). Characterization, cloning and in vitro expression of the extremely thermostable glutamate dehydrogenase from the archaeon, ES4. J. Biol. Chem. 268, 17767-17774. 11. Ohshima, T. & Nishida, N. (1994). Purification and characterization of extremely thermo-stable glutamate dehydrogenase from a hyperthermophilic archaeon Thermococcus litoralis. Biocatalysis 11, 117-129. 12. Eggen, R.I.L., Geerling, A.C.M., Waldkotter, K., Antranikan, G. & de Fig. 10. A representative portion of the final 2Fo-Fclexp(iac) Vos, W.M. (1993). The glutamate dehydrogenase-encoding gene of map calculated using the refined model at 2.2 A. the hyperthermophilic archaeon Pyrococcus furiosus: sequence, transcription and analysis of the deduced amino acid sequence. Hubbard [52] and the program CONTACT [46]. The solvent- Gene 132, 143-148. accessible surface of the GluDHs from Cs and Pf was calculated 13. Maras, B., et al., & Barra, D. (1994). The amino acid sequence of furiosus, a hyperther- using the algorithm of Lee and.Richards [53]. The fraction of glutamate dehydrogenase from Pyrococcus mophilic archaebacterium. J. Protein Chem. 13, 253-259. the surface area defined as non-polar, polar and charged was 14. Britton, K.L., et al., & Yip, K.S.P. (1995). Insights into thermal stabil- calculated following the method described by Miller et al. [26]. ity from a comparison of the glutamate dehydrogenase from Pyro- The volumes of atoms buried in the interior of the two coccus furiosus and Thermococcus litoralis. Eur. J. Biochem. 229, GluDH structures were calculated by the Voronoi procedure 688-695. 15. Baker, P.J., et al., & Stillman, T.J. (1992). Subunit assembly and [28,30,54,55]. The mean packing densities (defined as the ratio active site location in the structure of glutamate dehydrogenase. of the volume enclosed by the van der Waals envelope of the Proteins 12, 75-86. given atom to the actual volume of space that it occupies [28]) 16. Stillman, T.J., Baker, P.J., Britton, K.L. & Rice, D.W. (1993). Confor- were calculated for all residues in a monomer that are cornm- mational flexibility in glutamate dehydrogenase. Role of water in substrate recognition and catalysis. J. Mol. Biol. 234, 1131-1139. pletely buried in the structure. Ion-pair interactions in the two 17. Syed, S.E.H. (1987). Glutamate dehydrogenase from Clostridium structures were identified using the criterion suggested by symbiosum [PhD Thesis]. University of Sheffield, UK. Barlow and Thornton [31] of a distance less than or equal to 18. Russell, R.J.M., Hough, D.W., Danson, M.J. & Taylor, G.L. (1994). 4 A between charged groups. The crystal structure of citrate synthase from the thermophilic archaeon, Thermoplasma acidophilum. Structure 2, 1157-1167. 19. Britton, K.L., Baker, P.J., Rice, D.W. & Stillman, T.J. (1992). Struc- The coordinates for Pyrococcusfuriosus and Clostridium symbiosum tural relationship between the hexameric and tetrameric family of glutamate dehydrogenase have been submitted, with a one year glutamate dehydrogenases. Eur. J. Biochem. 209, 851-859. delay on release, to the Brookhaven Protein Data Bank. 20. Teller, J.K., Smith, R.M., McPherson, M.J., Engel, P.C. & Guest, J.R. (1992). The glutamate dehydrogenase gene of Clostridium symbiosum - cloning by polymerase chain reaction, sequence analysis Acknowledgements: We would like to thank Drs N Raven and and over-expression in Escherichia coli. Eur. J. Biochem. 206, R Sharp at the Public Health Laboratory Service Centre for 151-159. Applied Microbiology and Research for providing cell paste of 21. Baker, P.J., Britton, K.L., Rice, D.W., Rob, A. & Stillman, T.J. (1992). P.furiosus, the support staff at the Synchrotron Radiation Source at Structural consequences of sequence patterns in the fingerprint region of the nucleotide binding fold. J. Mol. Biol. 228, 662-671. DRAL Daresbury Laboratory for assistance with station alignment 22. Yip, K.S.P., et al., & Scandurra, R. (1995). Crystallization of the and Dr M Gerstein for use of his program to calculate packing NAD(P)-dependent glutamate dehydrogenase from the hyperther- densities. PJA is a Royal Society University Research Fellow. mophilic Pyrococcus furiosus. Acta Cryst. D 51, 240-242. This work was supported by grants from the EC Biotechnology 23. Rice, D.W., Hornby, D.P. & Engel, P.C. (1985). Crystallization of an Program, the Biotechnology and Biological Sciences Research NAD+-dependent glutamate dehydrogenase from Clostridium symbiosum. J. Mol. Biol. 181, 147-149. Council (BBSRC) and the Italian MURST. The Krebs Institute is 24. Matthews, B.W. (1977). X-ray structure of proteins. In The Proteins a designated BBSRC Biomolecular Sciences Centre. (3rd edition). (Neurath, H. & Hill, R.L., eds), vol. 3, pp. 468-477, Academic Press, New York. 25. Chan, M.K., Mukland, S., Kletzin, A., Adams, M.W.W. & Rees, D.C. References (1995). Structure of a hyperthermophilic tungstopterin enzyme, alde- 1. Stetter, K.O., Fiala, G., Huber, G., Huber, R. & Segerer, A. (1990). hyde ferredoxin oxidoreductase. Science 267, 1463-1469. Hyperthermophilic micro-organisms. FEMS Microbiol. Rev. 75, 26. Miller, S., Janin, J., Lesk, A.M. & Chothia, C. (1987). Interior and 117-124. surface of monomeric proteins. J. Mol. Biol. 196, 641-656. 2. Neuner, A., Jannasch, H.W., Belkin, S. & Stetter, K.O. (1990). Ther- 27. Korndorfer, I., Steipe, B., Huber, R., Tomschy, A & Jaenicke, R. mococcus litoralis sp. nov. A new species of extremely thermophilic (1995). The crystal structure of holo-glyceraldehyde-3-phosphate marine Archaebacteria. Arch. Microbiol. 153, 205-207. dehydrogenase rom the hyperthermophilic bacterium Thermotoga 3. Segerer, A.H., et al., & Stetter, K.O. (1993). Life in hot springs and maritima at 2.5 A resolution. J. Mol. Biol. 246, 511-521. hydrothermal vents. Orig. Life Evol. of the Biosph. 23, 77-90. 28. Richards, F.M. (1974). The interpretation of protein structures: total 4. Flam, F. (1994). The chemistry of life at the margins. Science 265, volume, group volume distributions and packing density. J. Mol. 471-472. Biol. 82, 1-14. 5. Ohshima, T. & Nishida, N. (1993). Purification and properties of 29. Richards, F.M. (1977). Areas, volumes, packing and protein struc- extremely thermostable glutamate dehydrogenases from two hyper- ture. Annu. Rev. Biophys. Bioeng. 6, 151-175. thermophilic Archaebacteria Pyrococcus woesei and Pyrococcus 30. Harpaz, Y., Gerstein, M. & Chothia, C. (1994). Volume changes on furiosus. Biosci. Biotech. Biochem. 57, 945-951. protein folding. Structure 2, 641-649. 1158 Structure 1995, Vol 3 No 11

31. Barlow, D.J. & Thornton, J.M. (1983). Ion-pairs in proteins. J. Mol. 45. Leslie, A.G.W. (1992). Recent changes to the MOSFLM package for Biol. 168, 857-885. processing film and image plate data. In Joint CCP4 and ESF-EACBM 32. Pai, E.F., Horn, E. & Schulz, G.E. (1984). The coenzyme site of glu- Newsletter on Protein Crystallography, no.26. SERC Daresbury tathione reductase. In Flavins and Flavoproteins. (Bray, R.C., Engel, Laboratory, UK. P.C. & Mayhew, S.C., eds), pp. 139-146, Walter de Gruyter & Co, 46. Collaborative Computing Project Number 4 (1994). The CCP4 suite: Berlin. programs for protein crystallography. Acta Cryst. D 50, 760-763. 33. Horovitz, A., Serrano, L., Avron, B., Bycroft, M & Fersht, A.R. 47. Tronrud, D.E., Ten Eyck, L.F. & Matthews, B.W. (1987). An efficient (1990). Strength and co-operativity of contributions of surface salt general-purpose least-squares refinement program for macromolecu- bridges to protein stability J. Mol. Biol. 216, 1031-1044. lar structures. Acta Cryst. A 43, 489-501. 34. Lumb, K.J. & Kim, P.S. (1995). Measurement of interhelical electro- 48. Otwinowski, Z. (1991). Maximum likelihood refinement of heavy static interactions in the GCN4 leucine zipper. Science 268, atom parameters. In Proceedings of the CCP4 study weekend 436-439. DL/SCI/R32. SERC Daresbury Laboratory, Warrington, UK. 35. Waldburger, C.D., Schildbach, J.F. & Sauer, R.T. (1995). Are buried 49. Kleywegt, G.J. & Jones, T.A. (1994). Halloween... masks and bones. salt bridges important for protein stability and conformational speci- In From First Map to Final Model. (Bailey, S., Hubbard, R. & Waller, ficity? Nat. Struct. Biol. 2, 122-128. D., eds), pp. 59-66, SERC Daresbury Laboratory, Warrington, UK. 36. Rees, D.C. & Adams, M.W.W. (1995). Hyperthermophiles: taking 50. Jones, T.A. (1985). Interactive computer graphics: FRODO. Methods the heat and loving it. Structure 3, 251-254. Enzymol. 115, 157-171. 37. Perutz, M.F. (1978). Electrostatic effects in proteins. Science 201, 51. Laskowski, R.A., MacArthur, M.W., Moss, D.S. & Thornton, J.M. 1187-1191. (1993). PROCHECK: a program to check the stereochemical quality 38. Day, M.W., et al., & Rees, D.C. (1992). X-ray crystal structures of the of protein structures. J. Appl. Cryst. 26, 283-291. oxidised and reduced forms of the rubredoxin from the marine 52. Baker, E.N. & Hubbard, R.E. (1984). Hydrogen bonding in globular hyperthermophilic archaebacterium Pyrococcus furiosus. Protein proteins. Prog. Biophys. Molec. Biol. 44, 97-179. Sci. 1, 1494-1507. 53, Lee, B. & Richards, F.M. (1971). The interpretation of protein struc- 39. Kelly, C.A., Nishiyama, M., Ohnishi, Y., Beppu, T. & Birktoft, J.J. tures: estimation of static accessibility. J. Mol. Biol. 55, 379-400. (1993). Determinants of protein thermostability observed in the 54. Chothia, C. (1975). Structural invariants in protein folding. Nature 1.9 A crystal structure of malate dehydrogenase from the ther- 254, 304-308. mophilic bacterium Thermus flavus. Biochemistry 32, 3913-3922. 55. Bernal, J.D. & Finney, J.L. (1967). Random-close-packed, hard- 40. Ishikawa, K., et al., & Morikawa, K. (1993) Crystal structure of sphere model II. Geometry of random packing of hard spheres. ribonuclease H from Thermus thermophilus HB8 refined at 2.8 A Discussions Faraday Soc. 43, 62-69. resolution. J. Mol. Biol. 230, 529-542. 56. Kraulis, P.J. (1991). MOLSCRIPT: a program to produce both 41. Stark, W., Pauptit, R.A., Wilson, K.S. & Jansonius, J.N. (1992). The detailed and schematic plots of protein structures. J. Appl. Cryst. 24, structure of neutral protease from Bacillus cereus at 0.2-nm resolu- 946-950. tion. Eur. J. Biochem. 207, 781-791. 57. Ferrin, T.E., Huang, C.C., Jarvis, L.E. & Langridge, R. (1988). The 42. Adams, M.W.W., Perler, F.B. & Kelly, R.M. (1995). Extremozymes: MIDAS display system. J. Mol. Graphics 6, 13-27. expanding the limits of biocatalysis. Biotechnology 13, 662-668. 58. Huang, C.C., Pettersen, E.F., Klein, T.E., Ferrin, T.E. & Langridge, R. 43. Honig, B. & Nicholls, A. (1995). Classical electrostatics in biology (1991). Conic: a fast renderer for space filling molecules with and chemistry. Science 268, 1144-1149. shadows. J. Mol. Graphics 9, 230-236. 44. Klump, H., et al., & Robb, F.T. (1992). Glutamate dehydrogenase from the hyperthermophile Pyrococcus furiosus. Thermal denatura- Received: 4 Aug 1995; revisions requested: 17 Aug 1995; tion and activation. J. Biol. Chem. 267, 22681-22685. revisions received: 30 Aug 1995. Accepted: 5 Sep 1995.