Article No. jmbi.1999.3145 available online at http://www.idealibrary.com on J. Mol. Biol. (1999) 293, 57±66

Ribulose Bisphosphate Carboxylase/Oxygenase fromthe Hyperthermophilic Archaeon Pyrococcuskodakaraensis KOD1 is Composed Solely of LargeSubunits and Forms a Pentagonal Structure

Norihiro Maeda1, Ken Kitano2, Toshiaki Fukui1, Satoshi Ezaki1

Haruyuki Atomi1, Kunio Miki2 and Tadayuki Imanaka1*

1Department of SyntheticChemistry and BiologicalChemistry, Graduate School ofEngineering, Kyoto UniversityYoshida-Honmachi, Sakyo-kuKyoto, 606-8501, Japan2Department of ChemistryGraduate School of ScienceKyoto UniversityKitashirakawa-Oiwake-choSakyo-ku, Kyoto606-8502, Japan

E-mail address of the correspondimanaka@sbchem.kyoto-u.ac.jp

Abbreviations used: bicine, N,N-bhydroxyethyl]glycine; Ches, N-cycloaminoethanesulfonic acid; Rubisco,bisphosphate carboxylase/oxygenasRubisco from Pyrococcus kodakaraens

0022-2836/99/410057±10 $30.00/0

We previously reported the presence of a highly active, carboxylase-speci®c ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco) in ahyperthermophilic archaeon, Pyrococcus kodakaraensis KOD1. In thisstudy, structural analysis of Pk-Rubisco has been performed. Phylogeneticanalysis of Rubiscos indicated that archaeal Rubiscos, including Pk-Rubisco, were distinct from previously reported type I and type IIenzymes in terms of primary structure. In order to investigate the exis-tence of small subunits in native Pk-Rubisco, immunoprecipitation andnative-PAGE experiments were performed. No speci®c protein other thanthe expected large subunit of Pk-Rubisco was detected when the cell-freeextracts of P. kodakaraensis KOD1 were immunoprecipitated with poly-clonal antibodies against the recombinant enzyme. Furthermore, nativeand recombinant Pk-Rubiscos exhibited identical mobilities on native-PAGE. These results indicated that native Pk-Rubisco consisted solely oflarge subunits. Electron micrographs of puri®ed recombinant Pk-Rubiscodisplayed pentagonal ring-like assemblies of the molecules. Crystals ofPk-Rubisco obtained from ammonium sulfate solutions diffracted X-raysbeyond 2.8 AÊ resolution. The self-rotation function of the diffraction datashowed the existence of 5-fold and 2-fold axes, which are located perpen-dicularly to each other. These results, along with the molecular mass ofPk-Rubisco estimated from gel ®ltration, strongly suggest that Pk-Rubiscois a decamer composed only of large subunits, with pentagonal ring-likestructure. This is the ®rst report of a decameric assembly of Rubisco,which is thought to belong to neither type I nor type II Rubiscos.

# 1999 Academic Press

Keywords: Rubisco; crystallization; hyperthermophile; Pyrococcus; archaea

*Corresponding author

Introduction

Ribulose-1,5-bisphosphate carboxylase/oxygen-ase (Rubisco; EC 4.1.1.39) is the key enzyme of theCalvin cycle and is arguably the most abundantprotein on earth (Ellis, 1979). In higher plants andphotosynthetic bacteria, Rubisco has been demon-strated to catalyze two competing reactions

ing author:

is[2-hexyl-2-ribulose 1,5-e; Pk-Rubisco,is KOD1.

(reviewed by Hartman & Harpel, 1994; Schneideret al., 1992; Shively et al., 1998; Watson & Tabita,1997). A carboxylase reaction produces two mol-ecules of 3-phosphoglycerate from one molecule ofribulose 1,5-bisphosphate, carbon dioxide andwater. This reaction is the major link betweenatmospheric inorganic carbon and organic carbonin the food chain. Rubisco also catalyzes anoxygenase reaction, which generates one moleculeof 3-phosphoglycerate and 2-phosphoglycolatefrom ribulose 1,5-bisphosphate and oxygen, lead-ing to photorespiration. Other than photosyntheticorganisms including higher plants, algae and cya-nobacteria, Rubisco has been detected in non-photosynthetic chemoautotrophic bacteria, such asThiobacillus denitri®cans (Hernandez et al., 1996).

# 1999 Academic Press

58 Decameric Assembly of Archaeal Rubisco

Although Rubiscos are supposed to be responsiblefor carbon dioxide ®xation in these bacteria, directevidence that speci®es their physiological roles islacking at present.

All Rubiscos mentioned above from bacteria andeukaryotes can be fundamentally classi®ed into twotypes (type I and type II) (Watson & Tabita, 1997).Type I Rubisco, found in almost all photosyntheticorganisms, is composed of eight large (L) subunitsand eight small (S) subunits, and comprises an L8S8

hetero-hexadecamer. In the holoenzyme, four L2

dimers are assembled into a core of eight large sub-units, (L2)4, displaying a 4-fold axis with local 422symmetry. The small subunits compose two separ-ate clusters of four subunits each (S4)2, which tightlyinteract with the large subunits. In the center of theL8S8 holoenzyme, along the 4-fold axis, a channelextends throughout the molecule (Knight et al.,1990). Based on the results of biochemical studies,along with structural analyses of the enzyme alone(Curmi et al., 1992; Knight et al., 1990; Schneideret al., 1986, 1990a; Shibata et al., 1996), or complexedwith inhibitors or reaction intermediates(Andersson, 1996; Newman & Gutteridge, 1993;Taylor & Andersson, 1997a,b), it is clear that allamino acid residues involved in catalytic activityare located on the large subunit molecule. Onedimer unit of large subunits has been supposed tobe responsible for catalytic activity. The smallsubunit is not directly involved in catalysis, but issupposed to affect the structure of the large subunit.It has been reported that the position of helix 8 withrespect to the rest of the barrel is different in thetype I and type II enzymes, possibly leading to thehigher carboxylase activity of type I enzymes(Schneider et al., 1990b).

In contrast to the type I enzyme, type II Rubisco,discovered in some purple non-sulfur bacteria,symbiotic and free-living chemoautotrophic bac-teria, is composed of large subunits only (reviewedby Hartman & Harpel, 1994; Schneider et al., 1992;Shively et al., 1998; Watson & Tabita, 1997). Thetype II Rubisco from Rhodospirillum rubrum hasbeen extensively studied and has been reported toexist as dimers (L2) (Ranty et al., 1990; Tabita &McFadden, 1974). There have also been indicationssuggesting the presence of an L4 or L8 type enzymein other photosynthetic bacteria (reviewed byTabita, 1988). Compared to the large subunit of thetype I spinach enzyme, that of type II enzyme ofR. rubrum had an extension at the C terminus.Interestingly, when the C-terminal extension of theR. rubrum enzyme was truncated at position 441 or449, the dimers assembled into L8 octamers. Elec-tron micrographs of the truncated Rubisco showedthat the shape of the L8 octamer was ring-like witha large central channel (Ranty et al., 1990).

Recently, some Rubiscos from the third king-dom, archaea, have been characterized. Rubiscofrom halophilic archaeon, Haloferax mediterranei,consists of large subunits and small subunits(Rajagopalan & Altekar, 1994), suggesting theenzyme to be a type I Rubisco. In the case of

Rubiscos from hyperthermophilic archaea, genomeanalysis of Methanococcus jannaschii (Bult et al.,1996) and Archaeoglobus fulgidus (Klenk et al., 1997)showed that putative Rubisco large subunit geneswere present on their chromosomes. They dis-played approximately 50 % similarity with knownsequences of type I and type II large subunits, andwhether their protein products harbored Rubiscoactivity had been an attractive subject in the ®eld(Watson & Tabita, 1997). In our previous study, weprovided initial evidence that a putative Rubiscogene from the hyperthermophilic archaeon, Pyro-coccus kodakaraensis KOD1, actually encodes anactive Rubisco (Pk-Rubisco) (Ezaki et al., 1999). Fol-lowing this, Tabita and co-workers determinedthat the protein product of the putative gene fromM. jannaschii also displayed Rubisco activity(Watson et al., 1999). Both were recombinantenzymes composed only of large subunits and thepresence of small subunits in the correspondingnative enzymes has not been investigated. The factthat homologous small subunit genes have notbeen found in the entire genomes of M. jannaschiiand A. fulgidus makes it tempting to speculate thatthe corresponding native enzymes are composedsolely of large subunits. However, studies on thenative enzymes are essential to elucidate the pre-sence of small subunits in these archaeal Rubiscos.

Pk-Rubisco displayed characteristic biochemicalproperties compared to previously reported Rubis-cos (Ezaki et al., 1999). The optimal temperature ofcarboxylase activity was an extremely high 90 �C,and the speci®c activity was 19.8 � 103 nmol ofCO2 ®xed per minute/mg, approximately tenfoldhigher than the speci®c activity of Rubisco fromspinach at 25 �C. Furthermore, the t value of Pk-Rubisco, which is the ratio of the speci®cities of thecarboxylase and oxygenase reactions of Rubisco,was an unprecedented 310 at 90 �C, displayingextreme speci®city towards the carboxylaseactivity. Due to these extraordinary properties ofPk-Rubisco, we conducted structural analysis ofthe recombinant Pk-Rubisco. Furthermore, as wehave previously been able to detect the presence ofthe native enzyme in the host cells, we pursuedanalysis of the native enzyme from P. kodakaraensisKOD1, focusing on the presence or absence of thesmall subunit.

Results

Phylogenetic analysis

We have previously compared the amino acidsequence of Pk-Rubisco with the large subunit ofRubisco from spinach, a typical type I enzyme, andthat from R. rubrum, a type II enzyme. Althoughall residues that are supposed to be directlyinvolved in catalysis were conserved in all threeenzymes, the enzymes showed only 50 % similaritywith each other (Ezaki et al., 1999). Type I and typeII enzymes are distinguished by their primaryamino acid sequences, as well as the presence or

Decameric Assembly of Archaeal Rubisco 59

absence of small subunits. As Pk-Rubisco and theenzyme from M. jannaschii both proved to be bona®de Rubiscos, we constructed a phylogenetic treeincluding type I and type II proteins.

An unrooted composite phylogenetic tree(Figure 1) was calculated on the basis of the neigh-bor-joining method by the ClustalW programavailable at DNA Data Bank of Japan (DDBJ).Archaeal Rubiscos showed an early divergencefrom the type I and type II enzymes and com-prised a distinct major branch of the tree. Whetherthe archaeal Rubiscos should be classi®ed as typeIII Rubiscos is discussed below.

Assembly of Pk-Rubisco

We next focused on the presence of the smallsubunit in the native Pk-Rubisco. Polyclonal anti-bodies against the recombinant enzyme has madeit possible to detect (Ezaki et al., 1999) and furtheranalyze the native enzyme in cell-free extracts ofP. kodakaraensis KOD1. Immunoprecipitation withthese antibodies was performed in order to detectproteins that have physical interaction or contactwith Pk-Rubisco. Precipitates were denatured andSDS-PAGE was performed (Figure 2(a)). As shownin lane 2, a protein with identical mobility withrecombinant Pk-Rubisco (lane 1) was present in theimmunoprecipitate. The antibody chains were alsodetected. Other major proteins were found toderive from the antisera (lane 3). No band could bedetected with a molecular mass of approximately15 kDa, corresponding to the molecular mass ofsmall subunits from previously characterized typeI Rubiscos (spinach, 14,281 Da; R. capsulatus, 13,547Da). Very faint bands could be detected with

Figure 1. Phylogenetic tree of selected type I, type II anR. sphaeroides, R. capsulatus and T. denitri®cans are indicateMultiple sequence alignments were conducted using Clustalwere done by the neighbor-joining method. This tree is unroare indicated at the nodes of the tree and are expressed as pe

molecular masses of 43, 21, and 18 kDa, but theamounts of these proteins were too trivial for con-sideration as candidates for a small subunit. Thesame results were obtained after silver staining(data not shown). This indicates the absence of atightly binding small subunit in native Rubiscofrom P. kodakaraensis, contrary to the case of type Ienzymes. The same conclusion was obtained byWestern blot analysis after native-PAGE(Figure 2(b)). Cell-free extracts of P. kodakaraensisand recombinant Pk-Rubisco were electrophoresedon a non-denaturing polyacrylamide gel. Detectionwith the antibodies displayed identical mobilitiesof native and recombinant Rubiscos. The results ofthe two experiments strongly indicate that theRubisco of P. kodakaraensis in the native host cellwas composed solely of large subunits. The latterresult suggests also that the assembly of subunitsof native and recombinant Pk-Rubisco are similar.

Electron microscopy

As the native Rubisco from P. kodakaraensis wasindicated to be composed only of large subunits,and that the assembly of subunits was similar tothat of the recombinant enzyme, further structuralstudies were carried out with the recombinantenzyme. This allowed large quantities of theenzyme to be available for analysis. We attemptedto visualize the enzyme structure by use of electronmicroscopy. Recombinant Pk-Rubisco was nega-tively stained with uranyl acetate and recorded bya transmission electron microscope. To our sur-prise, pentagonal structures were observed, asshown in the representative grid (Figure 3(a)). InFigure 3(b), one such structure is shown at high

d archaeal Rubiscos. Type I and type II enzymes fromd by 1 and 2 following the organism nomenclature.W. Tree topology and evolutionary distance estimationsoted. Bootstrap values calculated from 1000 replicationsrcentages rounded to the nearest whole number.

Table 1. Crystal data and data collection statistics of Pk-Rubisco

Crystal dataSpace group P3121 (or P3221)Cell parameters (AÊ ) a � b � 233.7, c � 93.1

Data collection statisticsResolution limit (AÊ ) 2.8Number of reflections 637,536 (I/s > 1.0)Number of unique reflections 63,731 (I/s > 1.0)Redundancy 10.0Completeness (%) 88.5 (100-2.8 AÊ )

63.1 (2.9-2.8 AÊ )hI/si 12.2Rmerge (%) 8.4 (100-2.8 AÊ )

29.9 (2.9-2.8 AÊ )

Rmerge(%) � �(�ijIi ÿ hIij)/�hIi, where hIi is the mean of theintensity measurements, Ii, and the summation extends over allre¯ections.

Figure 2. (a) SDS-PAGE after immunoprecipitation.Lane 1, recombinant Pk-Rubisco; lane 2, immunoprecipi-tated fraction from cell-free extracts of P. kodakaraensisKOD1 using anti-Pk-Rubisco antibodies; lane 3, antiserafraction bound to Protein A Sepharose. Arrow indicatesPk-Rubisco and arrowheads indicate antibody chains.(b) Native PAGE-Western blot analysis. Lane 1, 80 ng ofrecombinant Pk-Rubisco; lane 2, 10 mg of the cell-freeextracts of P. kodakaraensis KOD1; lane 3, 60 mg of thecell-free extracts of P. kodakaraensis KOD1. Arrow indi-cates Pk-Rubisco.

60 Decameric Assembly of Archaeal Rubisco

magni®cation. One can clearly recognize a penta-gonal ring-like structure with a 5-fold axis. Withinthe ring-like structure, we observed a large centraldark area, which can be presumed to be a solventchannel parallel with the 5-fold axis. In the pre-

vious report (Ezaki et al., 1999), we estimated themolecular mass of native Pk-Rubisco to be450 kDa, which is approximately nine times themolecular mass of a single subunit (49.7 kDa). Asall active-center residues were conserved amongtype I, type II Rubiscos and Pk-Rubisco, it can beassumed that the holoenzyme is formed by theassembly of dimer units. We initially proposedthat the native enzyme was a homo-octamer, con-sidering that type I enzymes are composed of eightsubunits (and eight small subunits) without excep-tion. Moreover, it has been reported that a mutantof the type II Rubisco from R. rubrum with aC-terminal truncation assembles to form an octa-mer. Taking into account the results of electronmicrography, we must now presume that thenative enzyme is a homo-decamer, consisting of®ve dimer units, forming a ring-like pentagonalstructure.

Crystallization and X-ray diffractiondata collection

Pk-Rubisco was crystallized with ammonium sul-fate as precipitant. Well-shaped hexagonal crystalswith dimensions of 0.5 mm � 0.5 mm � 0.5 mmwere obtained within three days at 20 �C from aprotein solution of approximately 10 mg/ml, 1.8 Mammonium sulfate and 10 mM MgCl2 in 100 mMChes-NaOH buffer at pH 9.0 (Figure 4). The crystalsdiffracted X-rays beyond 2.8 AÊ (Figure 5), indicat-ing that they were suitable for X-ray crystal struc-ture determination at an atomic resolution. Thespace group was determined to be P3121 (or P3221)by the systematic absences. Crystal data and datacollection statistics are summarized in Table 1.

Self-rotation function and molecular symmetry

The assembly of recombinant Pk-Rubisco wascon®rmed by preliminary X-ray crystallographicstudy, where the self-rotation function of thecrystal provides information on the molecular

Figure 3. Electron micrographs of recombinant Pk-Rubisco stained with uranyl acetate. (a) Electron micrographwas recorded at a magni®cation of 50,000 �. (b) Expanded electron micrographs of clear pentagonal structures of Pk-Rubisco.

Decameric Assembly of Archaeal Rubisco 61

symmetry. Fifteen 2-fold axes were identi®ed atk � 180 � (in Polar angles) by strong peaks on asingle plane in the self-rotation map (Figure 6(a)).Because of the crystallographic 3-fold screw axisthat is perpendicular to these 2-fold axes, ®ve (�15/3) axes are independent of each other. A5-fold axis was also identi®ed at k � 72 �(Figure 6(b)) and 144 �, perpendicularly to the 152-fold axes. These non-crystallographic symmetriessuggest that Pk-Rubisco exists as a decamer wherethe ®ve dimers of large subunits form a pentagonalring-like assembly. This is consistent with the resultobtained by electron micrography. Five monomersare expected in the asymmetric unit of the crystal,where one of the ®ve 2-fold axes of the decamercoincides with a crystallographic 2-fold axis. Inthis case, the VM value is calculated to

be 3.0 AÊ 3/Da, which lies in the range of1.8-3.5 AÊ 3/Da for usual protein crystals (Matthews,1968).

Discussion

This study has elucidated two major points con-cerning the Rubisco from a hyperthermophilicarchaeon P. kodakaraensis KOD1. One was that thenative Rubisco from the host cell was composedsolely of large subunits, and did not harbor smallsubunits. The other point was that Pk-Rubisco wasindicated to assemble as an (L2)5 decamer, distinctfrom all previously characterized Rubiscos.

Amino acid alignment (Ezaki et al., 1999) andphylogenetic studies clearly indicated that Rubis-cos from hyperthermophilic archaea, including Pk-

Figure 4. Crystals of recombinant Pk-Rubisco by thehanging drop vapor diffusion. The bar indicates0.5 mm.

62 Decameric Assembly of Archaeal Rubisco

Rubisco, were distinct from the type I and type IIRubiscos in terms of primary structure. ArchaealRubiscos showed wide diversity. Pk-Rubiscoand the putative Rbcl2 protein (ORF AF1638) ofA. fulgidus were the most closely related enzymes.

Further accumulation of archaeal Rubiscosequences and the characterization of their proteinsare necessary to accurately categorize theseenzymes. The only other archaeal Rubisco knownto harbor enzymatic activity is the M. jannaschiienzyme (Watson et al., 1999). Surprisingly, the bio-chemical properties of the enzyme drastically dif-fered from Pk-Rubisco in many points. TheM. jannaschii enzyme was highly sensitive to oxy-gen, with a large loss of activity when exposed toair. This inhibitory effect of oxygen was found tobe reversible. In contrast, Pk-Rubisco showed noinhibition by oxygen, as in the case of all pre-viously known Rubiscos. The second major differ-ence was between the t values of each enzyme.The M. jannaschii enzyme displayed t value of 0.5,the lowest for any Rubisco reported hitherto. Thet value of Pk-Rubisco was 310 at 90 �C, the highestof all known enzymes. A third difference was thesubunit composition; the M. jannaschii enzyme isan L2 dimer, while the present study has shownthat Pk-Rubisco is an (L2)5 decamer. Thesebiochemical differences might re¯ect the fact thatPk-Rubisco and the M. jannaschii enzyme showedonly 61.3 % similarity. Moreover, studies on theM. jannaschii enzyme were performed at roomtemperature, which may contribute to these differ-

Figure 5. X-ray diffraction photo-graph of recombinant Pk-Rubiscocrystals. The diffraction was takenby the Weissenberg method withsynchrotron radiation at PF-BL6B,KEK, Japan. Rotation range is 5.3 �and the crystal-to-®lm distanceis 573 mm. The arrow indicatesdiffraction at 3.0 AÊ resolution.

Figure 6. Self-rotation function of the crystal at (a) k � 180 � (�360 �/2) and (b) 72 � (�360 �/5). The orientation ofthe rotation axis k is de®ned by o (angle from z axis) and f (angle in the x, y plane) with respect to the orthogona-lized axes. The stereographic projections of each k section are shown with o and f as the radial and angular coordi-nates, respectively. Contour lines are drawn starting at 10 % with 10 % increments.

Decameric Assembly of Archaeal Rubisco 63

ences. In the case of Pk-Rubisco, carboxylaseactivity and its speci®city were indeed signi®cantlylower at 50 �C compared to those at 90 �C (Ezakiet al., 1999).

Concerning subunit composition, the lack of asmall subunit in Pk-Rubisco resembles the proper-ties of type II enzymes. The type II R. rubrumenzyme has been found to exist as an L2 dimer. Bycomparison of three-dimensional structures, theC-terminal domain of the R. rubrum enzymeincluded two additional a-helices that were notfound in the type I spinach Rubisco. Deletion ofone or both of these helices caused the L2 dimersto assemble into (L2)4 octamers. The mutantholoenzyme formed a ring-like structure with a 4-fold axis, and contained a large central cavity. Itwas suggested that the dimers showed end-to-endinteraction, in contrast to the side-by-side inter-actions observed in type I enzymes (Ranty et al.,1990). It can be noted that the C-terminal region ofPk-Rubisco corresponds well to the truncatedR. rubrum enzyme, lacking the ®nal two a-helices.Determination of the three-dimensional structurewill elucidate the subunit interactions in the penta-gonal Pk-Rubisco holoenzyme. Further biochemicalstudies at higher temperatures will also help to elu-cidate the subunit assembly of the enzyme at tem-peratures where P. kodakaraensis KOD1 can grow.

We have previously characterized the enzymaticproperties of Pk-Rubisco. The enzyme showed ahigh speci®c activity for the carboxylase reactionwith a t value of 310 at 90 �C. The type I spinachenzyme displays a t value of 80-90, while that ofthe type II R. rubrum enzyme is 10-20 (Tabita,1995). These values re¯ect those of most type I and

type II enzymes, and has led to the assumptionthat the small subunit increases the speci®city forthe carboxylase reaction (Schneider et al., 1990b,1992). Structural studies have shown that the smallsubunit interacts with the large subunit helix a8 ofthe spinach enzyme, resulting in a local confor-mation that signi®cantly differs from that of theR. rubrum enzyme. This conformational change hasbeen proposed to be responsible for the differencein carboxylase/oxygenase speci®city between thetwo enzymes. The high t value of Pk-Rubisco,without small subunits, makes the enzyme anattractive target for future research towards theunderstanding of enzymatic structures responsiblefor the t values of Rubiscos.

We have proven that the Pk-Rubisco protein waspresent in P. kodakaraensis KOD1 (Ezaki et al.,1999). This study has indicated that the enzyme inthe native host cell was composed solely of largesubunits, and should therefore presumably displaythe enzymatic properties of the recombinant pro-tein. Identi®cation of the physiological role of Pk-Rubisco, especially its contribution to carbon diox-ide ®xation, is still unclear and should beaddressed in future studies. At present we have noevidence that the strain can grow chemolithotro-phically. Genome analysis of M. jannaschii, whoseputative Rubisco gene has been shown to encodean active Rubisco, and A. fulgidus has revealed thepresence of putative ribose 5-phosphate isomerase,phosphoglycerate kinase, glyceraldehyde-3-phos-phate dehydrogenase and triose phosphate isomer-ase genes of the Calvin cycle. However, a putativeribulose-5-phosphate kinase, whose productshould be the substrate of Rubisco, has not been

64 Decameric Assembly of Archaeal Rubisco

identi®ed in either strain (Bult et al., 1996; Klenket al., 1997). Further metabolic studies of P. kodaka-raensis KOD1 are necessary to identify the pathwayto which Pk-Rubisco mainly contributes.

Materials and Methods

Strains, plasmids, and media

P. kodakaraensis KOD1 was cultivated in a previouslydescribed medium (Morikawa et al., 1994). Escherichia coliBL21(DE3) (Stratagene, La Jolla, CA) and the vectorpET21a(�) (Novagen Inc., Madison, WI) were used foroverexpression of the Pk-Rubisco gene (Pk-rbc1) (Ezaki,et al., 1999).

Gene expression and protein purification

The plasmid harboring the Pk-rbc1 was transformedinto E. coli BL21(DE3). Gene expression was inducedwith 0.1 mM IPTG for four hours. After harvesting,recombinant Pk-Rubisco was puri®ed as described(Ezaki et al., 1999) with some modi®cation. After harvest-ing, the cells were suspended in 100 mM bicine/KOH(pH 8.3), 10 mM MgCl2 and sonicated. The soluble cell-free extract was heat-precipitated for 30 minutes at 85 �C.The supernatant was applied to a Resource Q column(Amersham Pharmacia Biotech, Uppsala, Sweden), andeluted with a linear gradient of 0-1.0 M NaCl. Furtherpuri®cation was carried out by gel ®ltration on Superdex200 HR 10/30 (Amersham Pharmacia Biotech). The pro-tein was dialyzed against 100 mM bicine/KOH (pH 8.3),10 mM MgCl2 solution. Puri®ed recombinant Pk-Rubiscowas concentrated with Centricon 30 (Millipore, Bedford,MA), when necessary.

Phylogenetic analysis

The multiple alignment of protein sequences and thenumber of similarity between sequences were obtainedwith the program ALIGN contained within the ClustalWprogram provided by DNA Databank Japan (DDBJ). Thephylogenetic tree was constructed by the neighbor-join-ing method after alignment. Bootstrap resampling wasperformed with the BSTRAP program 1000 times.

Immunoprecipitation

Protein A Sepharose Fast Flow (Amersham PharmaciaBiotech) was used as the af®nity resin for immunopreci-pitation experiments. Rabbit antisera including polyclo-nal antibodies against recombinant Pk-Rubisco (Ezakiet al., 1999) were mixed with the resin. Antibodies boundto the resin were centrifuged and the supernatant wasremoved. The pellet was washed ®ve times with 20 bedvolumes of 100 mM bicine/KOH (pH 8.3), 10 mMMgCl2, followed by addition of the cell-free extracts ofP. kodakaraensis KOD1 in 100 mM bicine/KOH (pH 8.3),10 mM MgCl2. After removal of the supernatant by cen-trifugation, the pellet was washed ®ve times. Antibody-bound proteins were extracted (90 �C, two minutes) withSDS sample buffer (50 mM Tris-HCl (pH 6.8), 10 %(w/v) glycerol, 2 % (w/v) SDS, 2 % (v/v) mercaptoetha-nol), and applied to SDS-PAGE. Gels were stained withCoomassie brilliant blue.

Electrophoresis and Western blot analysis

SDS and non-denaturing polyacrylamide gel electro-phoresis (native PAGE), respectively, were conducted asdescribed (Maniatis et al., 1989). In this work, 12.5 % geland 10 % gel were utilized in SDS-PAGE and nativePAGE, respectively. Proteins were blotted from the poly-acrylamide gel to polyvinylidene di¯uoride membranes(Millipore). Antisera, containing polyclonal rabbit anti-bodies against recombinant Pk-Rubisco, were diluted1000-fold and used to detect native and recombinant Pk-Rubisco.

Electron microscopy

Drops of puri®ed recombinant Pk-Rubisco in 100 mMbicine/KOH (pH 8.3), 10 mM MgCl2 was deposited oncarbon ®lm-coated copper grids and then allowed todry. These grids were negatively stained with 2 % (w/v)uranyl acetate. Electron micrographs were recorded witha transmission electron microscope H7500 (Hitachi,Tokyo, Japan) operated at an acceleration voltage of 80kV. All images were taken at room temperature.

Crystallization

Initial crystallization conditions for Pk-Rubisco werescreened using Crystal Screen and Grid-Screen kits(Hampton Research, Riverside, CA) by the sitting orhanging drop vapor diffusion at 20 �C. The mixed sol-utions of 2-6 ml of protein solution and the same volumeof a reservoir solution were equilibrated against0.5-1.0 ml of the reservoir.

X-ray diffraction

The crystals of Pk-Rubisco were mounted in glasscapillaries with a trace amount of the mother liquor.Intensity data of the native crystals were collected atroom temperature with synchrotron radiation at theBL-6A and 6B beam lines of the Photon Factory (PF), theHigh Energy Accelerator Research Organization (KEK),Tsukuba, Japan (Sakabe et al., 1995). The diffractionintensities were recorded on Imaging Plates (Fuji PhotoFilm, Japan), which were digitized at 100 mm intervalson a Fujix BAS2000 IP reader. Data processing was per-formed using the program DENZO and SCALEPACK(Otwinowski & Minor, 1997). The intensity data werecollected using ®ve crystals with different rotation axes,which were merged into one data set.

Self-rotation function of Pk-Rubisco

The self-rotation function (Rossman & Blow, 1962)was calculated to determine the molecular symmetry ofPk-Rubisco, using the program PORARRFN in the CCP4Program Suite (Collaborative Computational Project,Number 4, 1994). Orthogonalized axes were chosenaccording to the Brookhaven format (x � a, y � c* � aand z � c*). Diffraction data from 15-8.5 AÊ resolutionand a 50 AÊ integration radius were used.

Acknowledgements

We thank Drs N. Sakabe, N. Watanabe, M. Suzuki,and N. Igarashi of the Photon Factory for their kind help

Decameric Assembly of Archaeal Rubisco 65

in the X-ray diffraction work, which was performedunder the approval of the Photon Factory AdvisoryCommittee (proposal no. 99G106). K.M. is a member ofthe Sakabe Project of TARA (Tsukuba AdvancedResearch Alliance), University of Tsukuba. This workwas partly supported by the ``Research for the Future''Program (97L00501) to K.M. from the Japan Society forthe Promotion of Science (JSPS), and by a Grant-in-Aidfor JSPS fellows (no. 9605) to K.K. from the Ministry ofEducation, Science, Sports and Culture. This work wasalso partly supported (to T.I.) by Japan Science and Tech-nology Corporation (JST) for Core Research for Evolu-tional Science and Technology (CREST).

References

Andersson, I. (1996). Large structure at high resolution:the 1.6 AÊ crystal structure of spinach ribulose-1,5-bisphosphate carboxylase/oxygenase complexedwith 2-carboxyarabinitol bisphosphate. J. Mol. Biol.259, 160-174.

Bult, C. J., White, O., Olsen, G. J., Zhou, L.,Fleischmann, R. D., Sutton, G. G., Blake, J. A.,FitzGerald, L. M., Clayton, R. A., Gocayne, J. D.,Kerlavage, A. R., Dougherty, B. A., Tomb, J.-F.,Adams, M. D., Reich, C. I. & Overbeek, R., et al.(1996). Complete genome sequence of the methano-genic archaeon, Methanococcus jannaschii. Science,273, 1058-1073.

Collaborative Computational Project, Number 4. (1994)The CCPA suite: programs for protein crystallogra-phy Acta Crystallog. sect. D, 50, 760-763.

Curmi, P. M. G., Cascio, D., Sweet, R. M., Eisenberg, D.& Schreuder, H. (1992). Crystal structure of theunactivated form of ribulose-1,5-bisphosphate car-boxylase/oxygenase from tobacco re®ned at 2.0-AÊ

resolution. J. Biol. Chem. 267, 16980-16989.Ellis, R. J. (1979). The most abundant protein in the

world. Trends Biochem. Sci., 241-244.Ezaki, S., Maeda, N., Kishimoto, T., Atomi, H. &

Imanaka, T. (1999). Presence of a structurally noveltype ribulose-bisphosphate carboxylase/oxygenasein the hyperthermophilic archaeon, Pyrococcus koda-karaensis KOD1. J. Biol. Chem. 274, 5078-5082.

Hartman, F. C. & Harpel, M. R. (1994). Structure, func-tion, regulation, and assembly of D-ribulose-1,5-bisphosphate carboxylase/oxygenase. Annu. Rev.Biochem. 63, 197-234.

Hernandez, J. M., Baker, S. H., Lorbach, S. C., Shively,J. M. & Tabita, F. R. (1996). Deduced amino acidsequence, functional expression, and unique enzy-matic properties of the form I and form II ribulosebisphosphate carboxylase/oxygenase from the che-moautotrophic bacterium Thiobacillus denitri®cans.J. Bacteriol. 178, 347-356.

Klenk, H.-P., Clayton, R. A., Tomb, J.-F., White, O.,Nelson, K. E., Ketchum, K. A., Dodson, R. J.,Gwinn, M., Hickey, E. K., Peterson, J. D.,Richardson, D. L., Kerlavage, A. R., Graham, D. E.,Kyrpides, N. C. & Fleischmann, R. D., et al. (1997).The complete genome sequence of the hyperthermo-philic, sulphate-reducing archaeon Archaeoglobus ful-gidus. Nature, 390, 364-370.

Knight, S., Andersson, I. & BraÈndeÂn, C.-I. (1990). Crys-tallographic analysis of ribulose 1,5-bisphosphatecarboxylase from spinach at 2.4 AÊ resolution: sub-unit interactions and active site. J. Mol. Biol. 215,113-160.

Maniatis, T., Fritsch, E. F. & Sambrook, J. (1989). Mol-ecular Cloning: A Laboratory Manual, 2nd edit., ColdSpring Harbor Laboratory Press, Cold Spring Har-bor, NY.

Matthews, B. W. (1968). Solvent content of protein crys-tals. J. Mol. Biol. 33, 491-497.

Newman, J. & Gutteridge, S. (1993). The X-ray structureof Synechococcus ribulose-bisphosphate carboxylase/oxygenase-activated quaternary complex at 2.2-AÊ

resolution. J. Biol. Chem. 268, 25876-25886.Morikawa, M., Izawa, Y., Rashid, N., Hoaki, T. &

Imanaka, T. (1994). Puri®cation and characterizationof a thermostable thiol protease from a newly iso-lated hyperthermophilic Pyrococcus sp. Appl.Environ. Microbiol. 60, 4559-4566.

Otwinowski, Z. & Minor, W. (1997). Processing of X-raydiffraction data collected in oscillation mode.Methods Enzymol. 276, 307-326.

Rajagopalan, R. & Altekar, W. (1994). Characterisationand puri®cation of ribulose-bisphosphate carboxy-lase from heterotrophically grown halophilic archae-bacterium, Haloferax mediterranei. Eur. J. Biochem.221, 863-869.

Ranty, B., Lundqvist, T., Schneider, G., Madden, M.,Howard, R. & Lorimer, G. (1990). Truncation ofribulose-1,5-bisphosphate carboxylase/oxygenase(Rubisco) from Rhodospirillum rubrum affects theholoenzyme assembly and activity. EMBO J. 9,1365-1373.

Rossman, M. G. & Blow, D. M. (1962). The detection ofsub-units within the crystallographic asymmetricunit. Acta Crystallog. 15, 24-31.

Sakabe, N., Ikemizu, S., Sakabe, K., Higashi, T.,Nakagawa, A., Watanabe, N., Adachi, S. & Sakaki,K. (1995). Weissenberg camera for macromoleculeswith imaging plate data collection system at PhotonFactory: present status and future plan. Rev. Sci.Instrum. 66, 1276-1281.

Schneider, G., Lindqvist, Y., BraÈndeÂn, C.-I. & Lorimer,G. (1986). Three-dimensional structure of ribulose-1,5-bisphosphate carboxylase/oxygenase from Rho-dospirillum rubrum at 2.9 AÊ resolution. EMBO J. 5,3409-3415.

Schneider, G., Lindqvist, Y. & Lundqvist, T. (1990a).Crystallographic re®nement and structure of ribu-lose-1,5-bisphosphate carboxylase from Rhodospiril-lum rubrum at 1.7 AÊ resolution. J. Mol. Biol. 211,989-1008.

Schneider, G., Knight, S., Andersson, I., BraÈndeÂn, C.-I.,Lindqvist, Y. & Lundqvist, T. (1990b). Comparisonof the crystal structures of L2 and L8S8 Rubiscosuggests a functional role for the small subunit.EMBO J. 9, 2045-2050.

Schneider, G., Lindqvist, Y. & BraÈndeÂn, C.-I. (1992).RUBISCO: structure and mechanism. Annu. Rev.Biophys. Biomol. Struct. 21, 119-143.

Shibata, N., Inoue, T., Fukuhara, K., Nagara, Y.,Kitagawa, R., Harada, S., Kasai, N., Uemura, K.,Kato, K., Yokota, A. & Kai, Y. (1996). Orderly dis-position of heterogeneous small subunits in D-ribu-lose-1,5-bisphosphate carboxylase/oxygenase fromspinach. J. Biol. Chem. 271, 26449-26452.

Shively, J. M., van Keulen, G. & Meijer, W. G. (1998).Something from almost nothing: Carbon dioxide®xation in chemoautotrophs. Annu. Rev. Microbiol.52, 191-230.

Tabita, F. R. (1988). Molecular and cellular regulation ofautotrophic carbon dioxide ®xation in microorga-nisms. Microbiol. Rev. 52, 155-189.

66 Decameric Assembly of Archaeal Rubisco

Tabita, F. R. (1995). The biochemistry and metabolicregulation of carbon metabolism and CO2 ®xationin purple bacteria. In Anoxygenic PhotosyntheticBacteria (Blankernship, M. T., Madigan, M. T. &Bauer, C. E., eds), pp. 885-914, Kluwer AcademicPublishers, Dordrecht, The Netherlands.

Tabita, F. R. & McFadden, B. A. (1974). D-Ribulose 1,5-diphosphate carboxylase from Rhodospirillumrubrum II. Quaternary structure, composition, cata-lytic, and immunological properties. J. Biol. Chem.249, 3459-3464.

Taylor, T. C. & Andersson, I. (1997a). Structure of aproduct complex of spinach ribulose-1,5-bispho-

sphate carboxylase/oxygenase. Biochemistry, 36,4041-4046.

Taylor, T. C. & Andersson, I. (1997b). The structure ofthe complex between Rubisco and its naturalsubstrate ribulose 1,5-bisphosphate. J. Mol. Biol. 265,432-444.

Watson, G. M. F. & Tabita, F. R. (1997). Microbial ribu-lose 1,5-bisphosphate carboxylase/oxygenase: amolecule for phylogenetic and enzymological inves-tigation. FEMS Microbiol. Letters, 146, 13-22.

Watson, G. M. F., Yu, J.-P. & Tabita, F. R. (1999).Unusual ribulose 1,5-bisphosphate carboxylase/oxy-genase of anoxic archaea. J. Bacteriol. 181, 1569-1575.

Edited by R. Huber

(Received 15 June 1999; received in revised form 18 August 1999; accepted 19 August 1999)