The function of the small subunits of ribulose bisphosphate carboxylase-oxygenase.

When ribulose bisphosphate carboxylase-oxygenase from Synechococcus (strain RRIMP N1) was precipitated under mildly acidic conditions, most of its small subunits remained in solution. The precipitated enzyme readily redissolved at neutral pH and remained as an octamer of large subunits with some small subunits still attached. Optimum pH for this separation was 5.3 and disulfide-reducing reagents were not necessary. The fraction of small subunits removed by a single precipitation increased with increasing NaCl concentration. Catalytic activity of small subunit-depleted enzyme was linearly proportional to the fraction of small subunits remaining, while the carboxylase:oxygenase activity ratio and the affinity for CO2 remained constant. When isolated small subunits were added back to depleted enzyme preparations at neutral pH, reassociation occurred with return of catalytic activity. Under the usual assay conditions at pH 7.7, the binding constant of the small subunits was estimated to be about 10(-9) M. The small subunits also bound avidly to surfaces. However, loss of small subunits from the enzyme during the course of purification was minimal. The results are consistent with a simple model in which only those large subunits which have a small subunit bound to them are catalytically competent. Thus, an essential, even if indirect, role for the small subunits in catalysis is indicated.

Rbu-P2' carboxylase-oxygenase (EC 4.1.1.39) is the bifunctional enzyme which catalyzes the initial COs-fixing reaction of photosynthesis as well as the initial 02-fixing reaction of photorespiration. In many prokaryotes and all eukaryotes the enzyme is a hexadecamer, consisting of eight large (about 55 kDa) and eight small (about 15 kDa) subunits, subsequently referred to as L and S. The catalytic site and the associated site of carbamate formation involved in activation both reside on L, which is present in the enzyme from all sources. Some bacterial Rbu-Pz carboxylases lack S and its function in those enzymes in which it, is present was hitherto unknown (for reviews see Lorimer and Andrews, 1981;Lorimer, 1981).
Previous work in this laboratory established that Rbu-P, carboxylase from the unicellular marine cyanobacterium, Synechococcus strain RRIMP N1, has the hexadecameric L plus _ _ " I _ * Contribution No. 129 from the Australian Institute of Marine Science. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore he hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
S subunit structure and that the association between L and S could be broken under mildly acidic conditions where Ls precipitated leaving S predominantly in solution (Andrews and Abel, 1981;Andrews et al., 1981). Loose binding of S has also been reported by Akazawa e t al. (1982) for the enzyme from another cyanobacterium, Aphunothece halophytica. In the present studies, dissociation and reassociation of the Synechococcus enzyme were studied quantitatively using improved methods for measuring the L and S content of enzyme preparations. Results support the view that S is just as essential for catalysis as L.

EXPERIMENTAL PROCEDURES
Enzyme Purification-Synechococcus Rbu-Pz carboxylase was purified either as described previously (Andrews and Abel, 1981) or by the following modification of that procedure. The 25-50% saturated (NH,),SO, pellet was prepared as before and redissolved in 250 mM K phosphate buffer, pH 7.6, containing 1 mM EDTA and 10 mM dithiothreitol, diafiltered against the same buffer with 1 mM dithiothreitol using an Amicon XM-300 membrane, and applied directly to the DEAE-Sephacel column. The phycobiliproteins were not retained by the column under these conditions. Rbu-Pp carboxylase was eluted with a gradient of 250-750 mM K phosphate in the same buffer. Peak fractions were concentrated and diafiltered with a YM-30 membrane against 50 mM Na phosphate buffer, pH 7.6, containing 1 mM EDTA, centrifuged to remove slight turbidity, and chromatographed on a column (2.6 X 86 cm) of Ultrogel AcA22 equilibrated with the same buffer at a flow rate of 12 ml. h-*. Active fractions were pooled and concentrated with a YM-30 membrane prior to storage at -80 "c in the presence of 20% (v/v) glycerol in the same buffer. Specific carboxylase activity of this preparation was 3.1 r.tmol.min"~mg" at 25 "C.
Subunit Quuntitation-A reliable method for measuring the L and S content of enzyme preparations was essential for these studies. Conventional SDS-gel electrophoresis followed by densitometry of Coomassie blue-stained bands frequently underestimated S, particularly when present in the gels at low levels, despite fixation of the bands with 10% (w/v) trichloroacetic acid before staining. Scanning of fixed but unstained gels at 280 nm improved reproducibility but diminished sensitivity. A rapid, sensitive, and quantitative procedure was developed using H P gel filtration. An LKB Ultropac TSK-G 3000SW (7.5 X 600 mrn) column was used, equilibrated with 20 mM Na phosphate buffer, pH 6.8, containing 100 mM Na2S04 and 0.1% (w/v) SDS. Samples were dissociated at 100 "C in buffer solution, pH 7.6, containing 1% (w/v) SDS and 1% (v/v) 2-mercaptoethanol and diluted 10-fold immediately prior to injection. Column flow rate was 1 ml. min" with an inlet pressure of 2.8 megapascals. The eluate was monitored a t either 210 or 280 nm. The peaks of L and S were well separated and their areas were linearly related to the amount of sample applied. The peak area ratio, S:L, of the purified enzyme was similar a t the two wavelengths (0.25 and 0.28 at 210 and 280 nm, respectively) and close to the ratio between the subunit molecular weights. Therefore, calibration was effected assuming that the mass extinction coefficients of L and S are equal. While this may introduce a slight systematic error, the data will be internally consistent. The same procedure was used to estimate subunit molecular weights. The plot of log M, uersus retention time for several standard proteins was linear and yielded M, estimates (60,300 for L and 14,600 for s) which were similar to those obtained by SDS-gel electrophoresis (Andrews and Abel, 1981).
Other Methods-Protein content of purified enzyme preparations was measured spectrophotometrically using an &of 12.6, determined by amino acid analysis (Andrews et al., 1981). Carboxylase and oxygenase activities were measured at 0.25 mM 0 2 as PreviouslY described by Lorimer et al. (1977) and modified by Andrews and Abel (1981). The pH of the carboxylase assay buffer was altered to 7.7 and a glass Rank On-electrode vessel was used for the oxygenase assay. Previously described methods were used for SDS-gel electrophoresis and nondenaturing gradient gel electrophoresis (Andrews and Abel, 1981).

Conditions
Favoring Separation of L and S-As shown in previous studies (Andrews and Abel, 1981), a fraction of the small subunits of Synechococcus Rbu-Pp carboxylase remained in solution when the enzyme was precipitated under mildly acidic conditions. A more detailed study of conditions favoring this separation of L and S was conducted with the aim of effecting as large a depletion of S as possible, consistent with good recoveries and minimal irreversible denaturation. The optimum pH for precipitation was 5.25 where a satisfactory compromise between these requirements was attained ( Table   I). Depletion of S was a little greater at pH 5.6 but precipitation was slow and incomplete, leading to poor recovery of L in the pellet. Furthermore, irreversible denaturation of the recovered protein was extensive, perhaps due to the long period at acidic pH required for precipitation to occur. Precipitation did not occur at pH 6 or above. Precipitation was more rapid at pH values below the optimum, being virtually instantaneous at pH 4.5. However, irreversible denaturation became progressively more serious as the pH was lowered and

TABLE I Effect of p H and dithwthreitol on subunit separation and recovery
Native enzyme (0.8-0.9 mg.ml-') was incubated with 10 mM ditbiothreitol and diafiltered against 10 mM Na phosphate, 1 mM EDTA, pH 7.6. NaCl and dithiothreitol were added to the stated concentrations and the pH was adjusted to the stated values with 0.25 M acetic acid. After 45-55 min at 0 "C, the precipitated protein was recovered by centrifugation, dissolved in the above buffer, clarified by centrifugation if necessary, and assayed for carboxylase activity and subunit content as described under "Experimental Procedures." Three parameters were calculated: the percentage recovery of L after precipitation and redissolution (column 4); the peak area ratio of the subunits, S:L, obtained from SDS-HP gel filtration analysis of the precipitated protein, monitored at 210 nm (column 5); and the specific carboxylase activity of the recovered protein expressed per unit of S content (column 6). The latter parameter is a useful index of the extent of irreversible denaturation of the precipitated protein.
Since the activity of S-depleted, but otherwise undenatured, preparations was proportional to their S content (Fig. lb), denaturation is reflected by a reduction in specific activity expressed in this form. The numbers in parentheses beside columns 5 and 6 are the ratios, expressed as percentages, of the values shown to the comparable value for the original unprecipitated enzyme. The S:L peak area ratio of the unprecipitated enzyme was 0.25 and its specific activity per unit s was 13.3 pmol. min". mg S-'. tion was carried out after 7 days at 4 "C.
Precipitation occurred extremely slowly. In this case, centrifuga-Precipitate only partially redissolved when resuspended in the pH recovery of L declined because an increasing fraction of the precipitated protein did not redissolve when resuspended at pH 7.6.
At optimum pH, the fraction of S released during precipitation increased with increasing NaCl concentration (Fig. l a ) . At high NaCl concentrations, two precipitations were sufficient to achieve a virtually inactive preparation, containing <8% of the native S content. Further precipitations were deleterious because of the cumulative effects of concomitant irreversible denaturation. Isolated S was retained in the supernatant after precipitation. Even under optimum conditions, it was usually accompanied by traces of undissociated enzyme, resulting in L protomer levels in supernatants which ranged from the barely detectable up to one-tenth (on a molar ratio basis) of the amount of S present. This contamination was least when the native enzyme was reduced with dithiothreitol, which was then removed by diafiltration, immediately prior to precipitation. The presence of dithiothreitol during the precipitation itself did not increase the fraction of S released (Table I). Isolated S was difficult to manipulate because it has a considerable affinity for surfaces, including those of glass and ultrafiltration membranes.
Correlation of Catalytic Activities with S Content-Carboxylase activities of enzyme variably depleted in S (by precipitation over a range of NaCl concentrations) were found to be exactly proportional to S content (Fig. lb). This indicates that S, as well as L, probably is required absolutely for activity. Effect of NaCl concentration during precipitation on S content (a) and relationship between S content and carboxylase activity (b). Precipitation was carried out at pH 5.2 and a range of NaCl concentrations as described for Table I and the redissolved enzyme was assayed for subunit content and carboxylase activity. In one experiment (A) precipitation was allowed to proceed for 55 min, the pellet was dissolved in 10 mM Na phosphate, 1 mM EDTA, pH 7.6, and the SDS-HP gel filtration to measure subunit content was monitored at 210 nm. In another (O), precipitation proceeded for 90 min, the pellet was dissolved in 50 mM Bicine-NaOH, 1 mM EDTA, pH 7.6, and 280 nm detection was used. The S:L peak area ratios and carboxylase activities (pmol .ruin". mg L-') of the untreated enzyme were, respectively, 0.25 and 3.3 (A) and 0.27 and 2.7 (0) for the two experiments. Surface adsorption was minimized by assaying for carboxylase activity at high enzyme concentrations (180 nM L protomers).   Effect of addition of isolated S on carboxylase activities of S-depleted and native enzyme preparations in the absence (a) and presence (b) of bovine serum albumin. Sdepleted or native enzyme preparations were mixed with increasing amounts of isolated S (supernatants from first precipitation step, adjusted to pH 7.6 with 1 M Tris free base) in a I-ml assay solution in a sealed glass scintillation vial. The solution contained 50 mM Bicine-NaOH, pH 7.7, 20 mM MgC12, 30 or 60 m M NaCI, 70 mM NaH"C03 (440 dpm.nmol"), with (b) and without (a) 0.1 mg.ml-' bovine serum albumin. After activation for at least 2 min, carboxylation was initiated by addition of Rbu-Pz to 0.4 mM, terminated with acid 1 or 2 min later, and otherwise treated similarly to carboxylase assays. Similar results were obtained regardless of whether the enzyme preparations and the isolated S preparations were mixed together for up to 30 min before addition to the assay solution or added separately immediately prior to the preactivation period. The total concentration of S, st, derived from the S-depleted or native enzyme preparation as well as the isolated S preparation, is plotted as the abscissa. Each activity was corrected by subtracting activity due to traces of undissociated enzyme present in the isolated S preparations (measured in controls) and u, the specific activity in terms of L, calculated. Each data set, i.e. set of u,st pairs for a particular enzyme preparation and concentration, was then fitted to Equation 1 as described under "Discussion." The resultant estimates of V and KO are listed in Table 111. Each value of u was normalized bv dividine it  Extrapolation of the regression line in Fig. l b to the S content of unprecipitated enzyme results in an activity about 12% below that of unprecipitated enzyme. This is consistent with the data of Table I which show that a small amount of irreversible denaturation accompanies precipitation even at optimum pH.
Oxygenase and carboxylase activities were affected similarly by depletion of S so that the carboxy1ase:oxygenase activity ratio remained constant, within experimental error, for preparations with different S:L subunit ratios ( Table 11).
The affinity of the enzyme for COz in the carboxylase reaction was not changed significantly by depletion of S. The K,,, (COz) of a preparation with about 50% of the native enzyme's S content was 96 p~ compared to 115 pM for the untreated enzyme.
Reconstitution-Earlier studies (Andrews and Abel, 1981) showed that S-depleted enzyme existed as La with varying degrees of saturation with S. Depleted species were distinguishable from the native enzyme by their electrophoretic mobility and sedimentation coefficient. When remixed with isolated S, reconstitution occurred to a form electrophoretically indistinguishable from the native enzyme. This was accompanied by a return of catalytic activity. These observations have been extended by the present studies. The reconstitution process was quantitated by varying the concentrations of L and S, and their ratio, on remixing. When a fixed quantity of highly S-depleted enzyme was mixed with increasing quantities of isolated S, activity increased with increasing S concentration until saturation was reached (Fig.   2a). The reassociation process was too rapid to be detected under normal assay conditions. No preincubation of L-S mixtures prior to assay was necessary beyond the usual period required for activation by CO, and M e (minimum 2 min).
However, S-depleted enzyme deteriorated with storage. After 24 days at 4 "C, activity after full saturation with S was reduced by 40% although the shape of the reconstitution curve was unaltered, indicating that the binding constant was similar (data not shown). Nondenaturing gradient-gel electrophoresis of aged preparations revealed multiple bands both higher and lower in molecular weight than the predominant native enzyme band (not shown). Therefore, both disruption and aggregation of the La core must be involved in the deteriora- Surface Adsorption Phenomena-Surprisingly, the activity of native enzyme preparations was stimulated by addition of S, especially when the former was present at low concentration (Fig. 2a). Direct assays of the native enzyme also gave progressively lower activities with decreasing enzyme concentration, with highly variable results being observed at low concentration (Fig. 3). When bovine serum albumin (0.1 mg. ml-') was included both the variability and the decline in activity at low concentrations were eliminated (Fig. 3). These observations are probably attributable to surface adsorption phenomena. Presumably albumin competes successfully with S for the binding sites on the surface of the glass assay vessel. In view of the already noted affinity of S for surfaces, it is likely that S was sequestered by surface-binding sites, thus depleting the enzyme of S and reducing its activity. However, the possibility that the holoenzyme was also bound to surface sites and denatured cannot be excluded. When binding experiments similar to those of Fig. 2a were repeated with serum albumin included in the solution, the data were consistent with much tighter binding between L and S, and additional S caused little or no stimulation of the activity of native enzyme, even when the latter was present at very low concentration (Fig. 26).
Retention of S during Purification-In view of the apparent ease with which S dissociates from the enzyme under some conditions and the observed affinity of S for surfaces, the possibility that the enzyme might have been partially depleted of S during purification must be considered. However, present evidence is against this possibility. Firstly, the purified enzyme appeared to have a molar ratio of S:L close to unity. This is indicated by the similarity between the molecular weight ratio of the subunits (0.24) and their peak area ratios measured by SDS-HP gel filtration, monitored at both 210 and 280 nm (0.25 and 0.28, respectively; see under "Experimental Procedures"). Secondly, the purified enzyme appeared  FIG. 3. Effect of enzyme concentration during assay on carboxylase activity. The native enzyme was assayed at various concentrations using the procedure described in the legend of Fig. 2 without addition of isolated S. Bovine serum albumin (0.1 mg.ml-') was absent (0) or present (0). The symbols represent the means of the observations and the bars indicate the range (where larger than the symbols).
to be fully saturated with S since, when bindlng to surfaces was suppressed, additional S did not stimulate its activity (Fig. 2b). Thirdly, the final gel filtration step did not cause any detectable loss of S. The preparation was pure enough before this step to estimate its S:L ratio by SDS-HP gel filtration and the ratio, 0.27 at 280 nm, was insignificantly different from that of the final preparation.

DISCUSSION
Some limited information about the nature of the forces involved in maintenance of the quaternary structure of synechococcus Rbu-P2 carboxylase may be gleaned from the conditions which promote dissociation of S from the L octamer. As expected, disulfide bridges are clearly not involved. The need for high NaCl concentrations suggests that ionic interactions may be important. This is also borne out by the efficacy of mildly acidic conditions where the subunits are probably approaching their isoelectric points, thus reducing ionic interactions.
The simplest hypothesis consistent with the linear correlation of catalytic activity with S:L ratio (Fig. lb) is that only those large subunits which have a small subunit bound to them are catalytically competent. This conflicts with previous conclusions that catalytic activity was not directly related to S content (Andrews and Abel, 1981). These conclusions were based on two lines of evidence. The first was data showing a nonstoichiometric relation between activity and degree of saturation with S (Andrews and Abel, 1981). These data must now be disregarded because they were based on SDS-gel electrophoresis measurement of S content, which are now known to be unreliable (see under "Experimental Procedures"). The second line of evidence was based on the observation that, after centrifugation of an S-depleted preparation through a sucrose gradient, the peaks of catalytic activity and protein were coincident (Andrews and Abel, 1981). In the light of present data, this constancy of specific activity across the peak is difficult to reconcile with the presumption that there was a gradient in S:L ratio across the peak with the leading edge being richest in S. However, this presumption may not be correct if S partially dissociated from the fastersedimenting S-rich complexes and reassociated with the slower-sedimenting S-depleted complexes during the course of the experiment. This would tend to even out the S:L ratio across the peak. Dissociation of S during sedimentation would have been promoted by the low enzyme concentrations used in the experiment in question.
Binding of S to S-depleted enzyme may be represented as follows.
LsSn-1 + S + L S n (n = 1-8) If all binding sites act independently with similar binding constants, this model simplifies to L + s e LS.
In this case, the binding constant at each site, KO, is given by where I, s, and k are the concentrations of unbound L and S and the LS complex, respectively, expressed in terms of protomers, and It and st are the total concentrations of L and S protomers. This gives rise to a quadratic expression for k where the applicable root is

1t,st]")/2.
If the LS complex is the only active species, the specific activity of a preparation per unit L, u, is given by where V is the specific activity per unit L after saturation with S. Therefore,

.Zt.st]"}/(2.II) (1)
In the experiments reported in Fig. 2 and Table 111, lt was held constant and u measured as a function of st. Each data set was fitted to Equation 1 by means of a nonlinear regression (PAR program of the BMDP package developed by the Health Sciences Computing Facility, University of California, Los Angeles) and the parameters V and KD estimated (Table 111). A convenient normalization of the data sets for different preparations and different lt values was obtained by plotting the observed u divided by the estimated V of each data set (Fig. 2). While the estimates of V were reasonable, with small standard deviations, the estimates of K D varied widely, ranging from 2 to 130 nM, and the standard deviations were large (Table IIIa and other data not shown). This variation was probably caused by surface binding phenomena with the apparent KD values reflecting the affinity of S for surface sites as well as its affinity for L. When surface binding was suppressed with albumin, the KO values were much lower (about lo-' M) and the standard deviations much smaller ( Table   IIIb). The standard deviations in the presence of albumin were still large relative to the estimated K O but this is to be expected since the K D values are much lower than even the lowest protomer concentrations which could be used. Therefore, the K D estimates should be considered significant at the order of magnitude level only. However, it is clear that the simple model embodied in Equation 1 provides a reasonable description of L-S interactions when surface binding is suppressed. The observed KD value, measured at pH 7.7, is at least four orders of magnitude lower that the enzyme protomer concentrations commonly prevailing during the purification procedures, which were conducted at a similar pH. Therefore, very little dissociation of S is to be expected under these conditions and retention of S during purification is understandable.
A clear picture thus emerges from this study. The small subunits of Rbu-Pp carboxylase are essential for catalytic activity. The data are consistent with a very simple model where only the complex between L and S is catalytically active. Both carboxylase and oxygenase activities are expressed by LS pairs present in the LsS, complex regardless of the value of n. The Cop affinity is also independent of n. All active site-directed probes so far tested, as well as the COz molecule which becomes bound during activation, react with amino acid residues of L (Lorimer, 1981). Therefore, at least part of the active site must reside on L. Since it appears that only the LS complex is active, two possible roles for S in catalysis may be inferred. The first possibility is that the active site occurs at a point of interaction between L and S with residues of both participating in catalysis; the second is that binding of S, at a site remote from the potential active site, induces a conformational change in L which promotes activity. Until an active site probe which labels S is found, the latter possibility must be favored. Regardless of which possibility is true, it is clear that the structure of S and the manner of its binding to L could have a profound effect on the kinetic properties of the active complex. The close similarity between the amino acid sequences of L in cyanobacterial and higher plant Rbu-Pp carboxylases, inferred from the nucleotide sequences, has been emphasized by the recent studies of Reichelt and Delaney (1982). It is possible that the large differences in kinetic properties between cyanobacterial and higher plant enzymes (e.g. the poorer affinity for COz and higher turnover number of the former (Badger, 1980;Andrews and Abel, 1981)) could be due to differences between the small subunits rather than the relatively few sequence differences between the large subunits.