INTRODUCTION

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, 8, 9, 10), spinach (11, 12, 13, 14), and a cyanobacterium (15, 16) are consistent with a hexadecameric L₈S₈ 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

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 MgCl₂, collected by centrifugation at 15,000 × 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 × 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 MgCl₂ as above.

Crystals were grown by vapor diffusion with 6-ml drops containing a protein solution (15 mg/ml) in 7% PEG 4000, 20 mM MgCl₂, 20 mM NaHCO₃, 1 mM dithiothreitol, 2 mM 2-carboxyarabinitol 1,5-bisphosphate, 50 mM Bicine (pH 7.9), and reservoir solutions containing 9% PEG 4000, 20 mM MgCl₂, 20 mM NaHCO₃, 1 mM dithiothreitol, and 50 mM Bicine (pH 7.9) at 20 °C. Crystals were of space group C222₁ with unit cell dimensions of a = 157.8, b = 157.8, and c = 200.9 Å, isomorphous to crystals prepared with ammonium sulfate (18). If we assume that there are four large subunits and four small subunits in an asymmetric unit, the solvent content of the crystal is 46% (V_m = 2.27 Å³/D) (19).

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 = 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  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.

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, electrophoresis, and staining were done as recommended by the manufacturer of the system.

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 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 L₈S^I₄S^II₄ structure, not L₈S₈ as reported before (11, 12, 13, 14).

Table I. Refinement statistics Statistic Value Resolution (Å) 10-1.8 Number of reflections used (F > 2(F)) 175,568 Rcryst (%) 15.9 Rfree (%) 18.8 Number of nonhydrogen protein atoms  L1-S1 4,679  L2-S2 4,681  L3-S3 4,679  L4-S4 4,681 Number of water molecules 1,875

Ball-and-stick stereo models around residue 56 of four small subunits with 2F_obs  F_calc electron density maps.

Fig. 3 shows the L₈S^I₄S^II₄ structure schematically. Earlier spinach Rubisco was described as having D₄ 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 ^SIIHis-56N-^LGlu-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)₂S^I₄}{(L^II/L^II)₂S^II₄}, where L^II is the L that has the ^SIIHis-56-^LGlu-259 hydrogen bond and L^I does not have such a bond, and L^I/L^I and L^II/L^II are mean L₂ dimers (12) formed by the same kind of large subunits. ^LGlu-259 participates in a dimer-dimer interaction with ^LArg-258 in the neighboring L₂ dimer and may be involved in the transfer of signals of an L₂ 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.