Orderly disposition of heterogeneous small subunits in D-ribulose-1,5-bisphosphate carboxylase/oxygenase from spinach.

We determined the crystal structure of spinach ribulose-1, 5-bisphosphate carboxylase/oxygenase (Rubisco) by x-ray diffraction at 1.8-A resolution and found that the enzyme contained two kinds of S, SI and SII, present in equal number and disposed in an orderly way within the Rubisco holoenzyme. The electron density maps suggested that leucine was at residue 56 in SI, although histidine was at that position in SII. There were other residue differences. Thus, spinach Rubisco has a L8SI4SII4 subunit structure. The orderly disposition of the heterogeneous small subunits in the Rubisco holoenzyme provides accounts of a multigene family of S in plants.

Ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco, 1 EC 4.1.1.39) is the key enzyme catalyzing the primary reactions in photosynthesis as well as photorespiration (1). This enzyme is an important enzyme involved in regulation and synthesis in complex ways (1). Cells of green plants contain 50 -100 chloroplasts, each of which has 20 -900 copies of the genome (2). Because of these large numbers, cells can synthesize much enzyme rapidly. Rubisco from some bacteria and eukaryotes is composed of eight large (L) and eight small (S) subunits (3). The gene for L is encoded in the plastid genome, and 4 -13 S genes compose the multigene family in higher plants (4). A multigene family of S subunits in plants probably facilitates fine tuning of the rate of synthesis of these subunits relative to L subunits (5). All S genes are expressed in green leaves of plants. The multiplicity in the genome construction may make possible transitory, organ-specific, or signal-specific expression of different genes that have individual promoters (6). Where do the translations of this multigene family in green leaves reside in the structure of Rubisco holoenzyme? Crystal structures of Rubiscos from tobacco (7-10), spinach (11)(12)(13)(14), and a cyanobacterium (15,16) are consistent with a hexadecameric L 8 S 8 structure in which all S subunits are identical. We analyzed the crystal structure of spinach Rubisco at 1.8-Å resolution to answer the question.

EXPERIMENTAL PROCEDURES
Purification and Crystallization-Rubisco was purified from spinach leaves with polyethylene glycol 4000 (PEG 4000) (17) instead of ammonium sulfate as reported previously. The enzyme was stored as a precipitate in a mixture of 20% PEG 4000 and 20 mM MgCl 2 , collected by centrifugation at 15,000 ϫ g for 20 min, and dissolved in 10 mM potassium phosphate buffer (pH 7.0). The enzyme was dialyzed against the same buffer overnight and put on a column (1.5 ϫ 7 cm) of hydroxylapatite that had been washed with the phosphate buffer. The effluent from the column contained Rubisco, and the enzyme, free from contaminants, was collected by precipitation in a mixture of PEG 4000 and MgCl 2 as above.
Data Collection and Refinement of Structure-Diffraction data from the Rubisco crystals were collected at room temperature with a Weissenberg camera for macromolecules (20) at the Photon Factory. A total of 586,168 observations was recorded from two crystals and was reduced to 216,085 unique reflections. The data were 67% complete to 1.6-Å resolution with an R merge ϭ 7.5%.
The crystal structure to 2.4-Å resolution of spinach Rubisco was used as an initial model (Brookhaven Protein Data Bank code, 8RUB) (12). After rigid body and positional refinement, a simulated annealing method (21) was used on SGI indigo2 and NEC EWS4800 workstations. Data between 6.0 and 2.5 Å were used for these calculations. After positional refinement and individual temperature factor refinement, an atomic model was fitted to a 2F obs Ϫ F calc electron density map with the program FRODO (22). Noncrystallographic restraints were used throughout the refinement process. Gradual expansion of the resolution range gave the final model.
Two-dimensional Gel Electrophoresis-Thirty micrograms of purified spinach Rubisco was analyzed by two-dimensional electrophoresis and Coomassie Blue staining with a horizontal electrophoresis system (Multiphore II, Pharmacia Biotech Inc.). A broad pH gradient gel ranging from pH 3 to 10.5 (Immobiline DryStrip, Pharmacia), was used for isoelectrofocusing in the first dimension, and a 15% SDS-polyacrylamide gel was used in the second dimension. Sample preparation, * A part of this study was supported by the Kansai Research Foundation for Technology Promotion, the Foundation for Earth Environment, and the Petroleum Energy Center subsidized by the Ministry of International Trading and Industry of Japan. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

RESULTS AND DISCUSSION
The results of the structure analysis of the four pairs of an L subunit with an S subunit in an asymmetric unit of the crystal lattice are shown in Table I. While refining the structures, we found that the side chain skeletons of some residues in the S subunits, reported by Martin (23), deviate significantly from our electron density maps, especially at residue 56, and that the shapes of the maps of the four crystallographically independent small subunits, named as S1, S2, S3, and S4, were different. Examination of the electron density maps suggested that residue 56 in S1 and S3 was leucine instead of the aspartate reported elsewhere (23) and that residue 56 in S2 and S4 was histidine (Fig. 1). The electron density maps at residue 93 suggested further structural differences; the reported alanine side chains of S2 and S4 fit the maps well, but S1 and S3 had a much longer side chain at this position (the residue could not be identified). Still other differences were observed in residue 8. These findings are evidence that spinach Rubisco had two S FIG. 1. Ball-and-stick stereo models around residue 56 of four small subunits with 2F obs ؊ F calc electron density maps. A, S1; B   Heterogeneous Small Subunits in Rubisco 26450 chains. Two-dimensional electrophoresis showed two peptides of S with one L peptide (Fig. 2) as reported before (24). With the pI of L taken to be 6.13, as calculated from the reported amino acid sequence (25) with GENETYX-MAX, Version 8 (Software Development Co., Ltd.), the pI points of the two S peptides were 6.10 and 6.42. The observed difference in the pIs was partly due to the residue at position 56 being leucine in one peptide and histidine in the other. This difference should give rise to a difference in the pI of 0.13. The larger difference (0.32) actually found might be explained by other differences in the amino acid residues at positions 8 and 93 or elsewhere; these could not be identified in this study. Thus, the results of x-ray diffraction analysis showed that spinach Rubisco had two kinds of small subunits: S I , with Leu-56, and S II , with His-56. Spinach Rubisco, therefore, has a L 8 S I 4 S II 4 structure, not L 8 S 8 as reported before (11)(12)(13)(14). Fig. 3 shows the L 8 S I 4 S II 4 structure schematically. Earlier spinach Rubisco was described as having D 4 point symmetry (11)(12)(13)(14), which would be possible only if Rubisco were composed of eight identical large subunits and eight identical small subunits. Our x-ray results showed that what might be the 4-fold symmetry of spinach Rubisco was broken by the heterogeneity of the S subunits. The same kind of small subunits occupies the positions furthest from each other, maintaining 2-fold symmetry along the central core.
The N⑀ atom of His-56 in the S II subunits forms a hydrogen bond with the carbonyl oxygen atom of Glu-259 in the neighboring large subunit (Fig. 1). On the other hand, the side chain of Leu-56 of the S I subunits does not interact electrostatically. The additional SII His-56N⑀-L Glu-259-O hydrogen bond must cause the difference in the dissociation constants; the L-S II interaction must be more stable because of the hydrogen bond between His-56 and Glu-259, and the S II subunit will construct the L-S pair more readily than S I . In this context, it is interesting to recall the finding that a highly conserved sequence of 16 amino acids, including that at position 56, is essential for the assembly of L and S in the plant enzyme (26).
In terms of its different interactions of the two S subunits, L also may be of two kinds; the structure of spinach Rubisco may be {(L I /L I ) 2 S I 4 }{(L II /L II ) 2 S II 4 }, where L II is the L that has the SII His-56 -L Glu-259 hydrogen bond and L I does not have such a bond, and L I /L I and L II /L II are mean L 2 dimers (12) formed by the same kind of large subunits. L Glu-259 participates in a dimer-dimer interaction with L Arg-258 in the neighboring L 2 dimer and may be involved in the transfer of signals of an L 2 dimer to the next one. Plant Rubisco gradually decreases in activity to a constant level during reaction (17). The decrease is smaller if there is binding of the substrate ribulose 1,5-bisphosphate to the noncatalytic substrate-binding sites (27). Binding of ribulose 1,5-bisphosphate to these sites proceeds cooperatively; binding to the first four sites suppresses binding to the remaining four sites (28,29). The grouping of eight large and eight small subunits into two different structures as described above may give an account for the cooperativity in plant Rubisco. Thus, a multigene family for S may be related to a genetic mechanism that has given the enzyme the ability to fine tune its own catalysis.