Photosynthesis Research 17:145-157 (1988) © Kluwer Academic Publishers, Dordrecht - Printed in the Netherlands Minireview
Synthesis and assembly of bacterial and higher plant Rubisco subunits in Escherichia coli A N T H O N Y A. GATENBY Central Research and Development Department, Experimental Station, E.L du Pont de Nemours & Co., Wilmington, DE 19898 U.S.A. Received 2 September 1987; accepted 17 December 1987
Key words: Carbon fixation, enzyme assembly, gene expression, recombinant D N A Abstract. The synthesis in Escherichia coli of both the large and small subunits of cereal ribulose bisphosphate carboxylase/oxygenase has been obtained using expression plasmids and bacteriophages. The level and order of synthesis of the large and small subunits were regulated using different promoters, resulting in different subunit pool sizes and ratios that could be controlled in attempts to optimize the conditions for assembly. Neither assembly nor enzyme activity were observed for the higher plant enzyme. In contrast, cyanobacterial large and small subunits can assemble to give an active holoenzyme in Escherichia coli. By the use of deletion plasmids, followed by infection with appropriate phages, it can be demonstrated that the small subunit is essential for catalysis. However, the small subunit is not required for the assembly of a large subunit octomer core in the case of the Synechococcus enzyme; self-assembly of the octomer will occur in an rbcS deletion strain. The cyanobacterial small subunits can be replaced by wheat small subunits to give an active enzyme in Escherichia coli. The hybrid cyanobacterial large/wheat small subunit enzyme has only about 10% of the level of activity of the wild-type enzyme, reflecting the incomplete saturation of the small subunit binding sites on the large subunit octomer, and possibly a mismatch in the subunit interactions of those small subunits that do bind, giving rise to a lower rate of turnover at the active sites.
Abbreviations: IPTG-isopropyl-fl-D-thiogalactopyranoside, L-large ribulose bisphosphate carboxylase/oxygenase, S-small subunit
subunit,
Rubisco-
Introduction
One objective of the synthesis and assembly of Rubisco in Escherichia coli is to obtain an understanding of the assembly process itself. Once assembly can be demonstrated, then experiments can be designed to understand how the subunits interact with each other, and the importance of particular structures in catalysis, and its possible modification. Following the initial report on the successful expression of a chloroplast rbcL gene in E. coli (Gatenby et al. 1981) there have been rapid developments which have
146 expanded the range of rbcL and rbcS genes from both plant and prokaryotic sources that can be expressed in E coli, resulting in the synthesis of Rubisco large (L) and small (S) subunits. This has enabled experimental systems to be developed that result in the assembly of catalytically active dimeric (Somerville and Somerville 1984), hexameric (Quivey and Tabita 1984) and hexadecameric (reviewed by Bradley et al. 1986) enzymes. These systems can now be used to introduce site specific mutations to investigate the active site, to examine the assembly pathway, and the formation of hybrid enzymes. The isolation from its normal cell of a pathway leading from gene expression to subunit assembly, and its reconstitution in E. coli enables a precise approach to finding what the requirements are for successful assembly by mutating, using different subunit genes and otherwise modifying the system. Another feature, particularly important where subunits are encoded by a multigene family, is that assembly from a cloned gene will provide an enzyme with identical, well characterized subunits of a particular type, rather than the plant derived enzyme which may be a heterogeneous mixture of subunits giving rise to an average value when used in enzyme assays. This review examines some of the strategies that have been adopted to asesemble active Rubisco in E. coli.
Expression of higher-plant Rubisco genes in E. coli The maize and wheat chloroplast rbcL genes were the first plant genes to be expresed in E. coli (Gatenby et al. 1981). The original experiments were designed to detect the synthesis of the L subunit when the appropriate chloroplast DNA fragment was cloned into the plasmid pBR322. The L subunit synthesized had a similar Mr to that observed in chloroplasts, suggesting that translation was initiating correctly at the chloroplast ribosome binding site. Furthermore, the levels of expression observed were similar for both orientations of the rbcL gene in the plasmid, indicating that transcription was probably initiating within chloroplast DNA, and was not the result of readthrough transcription from plasmid promoters. This was later substantiated by the observation that plasmid encoded lambda PL promoters could not direct transcription through the maize rbcL gene unless a phage encoded transcription antitermination protein was present (Gatenby and Cuellar 1985), and that chloroplast promoters could be recognized by the bacterial RNA polymerase (Bradley and Gatenby 1985). It was clear, therefore, that the close similarities between the chloroplast and bacterial transcription and translation signals enabled the plant genes to be readily expressed in E. coli. The expression of chloroplast rbcL genes in E. coli has
147 also been observed for tobacco (Fluhr et al. 1983, Zhu et al. 1984), petunia (Bovenberg et al. 1984) and Chlamydomonas (Zhu et al. 1984). The levels of synthesis of the L subunit, using chloroplast promoters, is low. It was estimated that only about 100-200 monomers were synthesized with the plasmids pZmB 1A and pZmB 1B using chloroplast promoters (Ellis and Gatenby 1984). The level of synthesis could be substantially improved by increasing the rate of transcription of the gene. This was initially achieved by cloning the maize rbcL gene into bacteriophage lambda, and then relying on the early PL promoter, or the late P~t promoter to transcribe the gene during lytic infection (Gatenby et al. 1981). Although rbcL expression levels achieved 0.5-1.0% of total protein synthesis following phage infection, the protein did not accumulate to high levels due to the transitory nature of expression in the lytic cycle. By using a plasmid vector containing the lambda PL promoter, a transcription antiterminator, and the maize rbcL gene, it was possible to induce synthesis of L subunits by thermal denaturation of the cI repressor. Under these conditions synthesis of the L subunit continued for more than 120min and the percentage of L in the cell increased after induction to reach a maximum of 2% of total E. coli protein (Gatenby and Castleton 1982). This represents approximately 60,000 subunit monomers per bacterial cell. High level expression of the maize rbcL gene has also been achieved by Somerville et al. (1983, 1986), using a different strategy. In their experiment a lac promoter was used and a hybrid ribosome binding site was formed using the sites of lacZ and rbcL fused together. The levels of expression obtained with plasmid pPBI3 (Gatenby and Castleton 1982) enabled studies to be carried out on the properties of the L subunit synthesized in the absence of the S subunit. An advantage with this approach is that the use of denaturing conditions which are used to separate the L and S subunits in the holoenzyme, can be avoided. Pulse-chase experiments showed that the L subunit was stable in E. coli for at least 4 h (Gatenby 1984). The protein was also insoluble, whether expressed at high or low levels. Attempts to detect enzyme activity, or to form a stable enzyme.metal.CO2.[~4C] carboxyarabinitol bisphosphate quaternary complex were unsuccessful. The presence of the substrate ribulose bisphosphate, or the positive effector fructose 1,6-bisphosphate, did not improve solubility, although it was found that pH 11-12 solubilized L slowly suggesting that aggregate structures may have formed. It would appear that maize L adopts an inactive, insoluble conformation after, or during, synthesis. Similar results have been reported by others for maize L (Somerville et al. 1983, 1986).
148
Fig. 1. Fluorograph showing the synthesis of Rubisco subunits in E. coli following induction of plasmid pLSS308. Cells were labelled with [35S] methionine in the presence of IPTG to induce wheat small subunit (SS) synthesis, and either at 30°C or 41 °C to obtain repression or induction respectively of maize large subunit (LS) synthesis. The 35S-labelled polypeptides were incubated with anti-wheat RuBPCase serum and protein A Sepharose before electrophoresis on polyacrylamide gels containing SDS. Reprinted from Gatenby et al. (1987).
In an attempt to overcome the problem of L insolubility, and to assemble a plant holoenzyme, experiments were designed to co-express the L and S subunit genes in E. coli (Bradley et al. 1986, Gatenby et al. 1987). These experiments required that the nuclear encoded S subunit gene be cloned into an E. coli vector such that the S polypeptide is synthesized without the amino-terminal transit peptide, which is presumably only necessary for transport of the protein into chloroplasts (Broglie et al. 1983). These constructions have been made by subcloning a region of the wheat rbcS gene, which encodes the mature S subunit, from a c D N A clone into plasmid and M13 vectors (Bradley et al. 1986, van der Vies et al. 1986). The resulting
149 expression clones (p565, pLSS308, and M13-72) produce the S subunit as a lacZ::rbcS fusion protein in E. coll. The amino-terminals of the S fusion protein contain 11 (p565 and pLSS308) or 10 (M13-72) amino acids from fl-galactosidase. The first four amino acids of the mature S polypeptide have also been deleted. Transcription and translation of the rbcS gene fusions are initiated from the lacZ promoter and ribosome-binding sites respectively, resulting in synthesis of an S polypeptide of 15.6kDa, which migrates slightly slower than authentic wheat isolated from chloroplasts (14.7 kDa) as expected from the fusion at its N-terminus. The rbcS expression clones were used to study plant holoenzyme formation in two ways, either using heterologous subunit mixtures, or homologous subunits. In plasmid pLSS308 the wheat rbcS gene was placed under the control of Plac. and the maize rbcL genes were placed under the control of PL" The use of different promoters allows the independent expression of the rbcS and rbcL genes. The expression of rbcL from pLSS308 relied on translational intiation at the chloroplast ribosome binding site (Gatenby et al. 1981), and temperature induced transcription obtained by shifting cultures to 41-42 °C in the presence of the anti-terminating protein p N (Gatenby and Cuellar 1985). Fig. 1 shows the result of a typical induction experiment in which S synthesis is constitutive, but the synthesis of L has been increased at 41 °C. The advantage of this approach is that by altering the pool sizes of the two subunits it might be possible to arrive at an optimal ratio of subunit types for assembly to occur. When the S subunit is synthesized in the absence of the L subunit it is found in the soluble fraction, but the polypeptide is unstable and has a half-life of less than 15 min (Gatenby et al. 1987). Its size on sucrose gradients indicates a monomeric or dimeric form. When L subunit synthesis is induced in cells containing the S subunit both subunits are found predominantly in the insoluble fraction and are fully stable for more than 120 min, suggesting that aggregation of the subunits may occur. The two subunits do not assemble to form an active holoenzyme in vivo even when nascent L subunits are synthesized in a pool of mature S subunits. To test if the synthesis of homologous subunits in E. coli would provide a more favorable opportunity for assembly, the wheat rbcS and rbcL genes were expressed together (Bradley et al. 1986). A bacteriophage M I3 clone (M68) expressing the wheat L subunit under the control of Plac was constructed. Cells were allowed to first synthesize the S subunit by means of a plasmid-borne rbcS gene (p565), and were then infected with the bacteriophage M68 (rbcL). Phage M68 was also used to infect plasmid-free E coil cells to examine the properties of the wheat L subunit in the absence of S. In contrast to the maize L subunit insolubility in E. coli, about 60% of
150
Fig. 2. Velocity sedimentation in sucrose gradients of cyanobacterial Rubisco synthesized a n d assembled in E. coli. Soluble protein samples were prepared from E. coli cells that had synthesized the Synechococcus L and S subunits following induction of plasmid pSV55. In panel A the bacterial extract, or purified spinach Rubisco, was incubated with the transition state analogue [~4C] carboxyarabinitol bisphosphate before centrifugation. In panel B the distribution of Rubisco activity was measured in a similar sucrose gradient using the induced bacterial extract. In panel C an immunoblot analysis of the fractions from panel B was carried out using anti-Rubisco serum. The M track is a maize leaf track used as a marker. The Mr markers used to calibrate the sucrose gradients are shown by arrows. Reprinted from Gatenby et al. (1985).
151 the wheat L subunit was found in the soluble fraction (Bradley et al. 1986). The S subunit did not appear to influence the amount of L that was found in the soluble fraction. It was also observed that about 80% of the wheat S subunit was in the soluble fraction from E. coli cells. Rubisco activity was undetectable. To examine subunit assembly, the soluble proteins from cells expressing both L and S were subjected to velocity sedimentation on sucrose gradients. Most of the S subunit remained at the top of the gradient. The wheat L subunit, although distributed throughout the gradient, sedimented with a peak abundance indicating a molecular mass of 750kDa. This probably represents a large protein aggregate. Therefore, even with a homologous combination of subunits, it was not possible to detect assembly or activity of the higher plant enzyme.
Synthesis of bacterial Rubisco genes in E. coli
An attractive feature of synthesizing bacterial Rubisco in E. coli is that the prokaryotic structure of the genes can facilitate their expression. A range of different forms of active Rubisco has been synthesized in E. coil using the isolated genes from several different species. These include the homodimeric L subunit enzyme from Rhodospirillum rubrum (Somerville and Somerville 1984, Latimer et al. 1986), the hexadecameric L subunit enzyme from Rhodopseudomonas sphaeroides (Quivey and Tabita 1984) and the hexadecameric enzyme containing both L and S subunits from cyanobacteria (Christeller et al. 1985, Gatenby et al. 1985, Gurevitz et al. 1985, Tabita and Small 1985), Chromatium (Viale et al. 1985) and Rhodopseudomonas sphaeroides (Gibson and Tabita 1986). The prokaryotic nature of cyanobacteria also simplifies the expression of their genes in E. coli, because the absence of an organdie, and therefore an S subunit transit peptide, enables the direct synthesis of mature S. In addition, the rbcL and rbcS genes are physically linked on a contiguous fragment of DNA. For expression of the Synechococcus Rubisco genes they were initially transcribed from a bacteriophage lambda PL promoter following temperature induction (Gatenby et al. 1985). Synthesis of the L and S subunits could be detected, together with enzyme activity and [14C] carboxyarabinitol bisphosphate binding. In addition, sucrose gradient centrifugation resolved a peak of enzyme activity that was similar in size to the spinach holoenzyme (Fig. 2). The large size indicated that the L 8 structure was formed, but an L 8Sn complex where n is less than 8 would be difficult to resolve from L8S8 because of the small size of S. Although S must be present to obtain an active enzyme (Andrews and Abel 1981), it was suspec-
152ted that the S binding sites on the L octomer were not fully saturated because of two observations. First, the specific activity in cell extracts is lower than expected based on the known amount of L8 present (as determined by [14C] carboxyarabinitol bisphosphate binding). Second, the amount of 35S present in the purified L and S polypeptides indicated that L 8Sn structures were formed where n = 2 or 3. More recently it has been shown that transfer of the appropriate Synechococcus D N A fragment from the PL expression plasmid to a Plac expression plasmid increases the amount of Rubisco synthesized and enables either the assembly of L8 $8 structures, or structures almost completely saturated with S subunits (Bradley et al. 1986, van der Vies et al. 1986). Christeller et al. (1985) demonstrated that cloning the Anacystis nidulans rbcL and rbcS genes separately resulted in the absence of enzyme activity, indicating that both L and S subunits were required for activity. We have obtained similar results using plasmid or phage encoded Synechococcus rbcL and rbcS genes (van der Vies et al. 1985). Plasmid pDB50 contains both rbcL and rbcS and directs the synthesis of active Rubisco. If rbcS is deleted, the resulting plasmid (pDB53) synthesized only the L subunit which is inactive. Since this deletion does not result in any addition, or noticeable reduction, in the amount of soluble L subunit synthesized, it is concluded that the product of rbcS is essential for activity. If this is correct, infection E. coli cells containing the S subunit plasmid delection (pDB53) with M 13 phages encoding the rbcS gene should restore Rubisco activity by complementation. This is exactly what happens. Following phage infection, Rubisco activity can be detected. These deletion and complementation experiments have also enabled us to obtain information about the assembly process of Rubisco in E. coil. It appears that the major soluble protein structures formed in cells synthesizing the L subunit alone (pDB53) are oligomeric forms, principally an L octomer and an L dimer (van der Vies et al. 1986). Therefore, the formation of a Synechococcus L8 structure does not require S subunits to be present. The L2 structure may represent an intermediate which then selfassembles to the L8 form, or may be the result of disassembly of the L octomer. When both cyanobacterial L and S subunits are synthesized in the same cell the L dimer is not present, and the main L peak shifts further down the gradient indicating the binding of S subunits. The presence of S subunits appears to reduce the pool of unassembled L2 and promote formation of LB.
Assembly of a hybrid Rubisco in E. coli
Since it was possible to demonstrate complementation of Rubisco activity
153
Fig. 3. Analysis of heterologous Rubisco assembly in E. coli. A soluble protein extract of E. coli cells containing plasmid pDB53 (encoding the cyanobacterial rbcL gene), and infected with the phage ML3-72 (encoding the wheat rbcS gene) was fractionated on a sucrose gradient. Panel a shows an immunoblot of the fractions using anti-Rubisco serum, and panel b shows the determination of Rubisco activity. The gradient was calibrated with Mr markers shown by arrows. Wheat leaf protein was used as a marker in the right lane of panel a. Reprinted from van der Vies et al. (1986).
in a n S s u b u n i t p l a s m i d deletion strain b y infection w i t h M 13 r b c S + p h a g e s e n c o d i n g a h o m o l o g o u s S s u b u n i t ( v a n der Vies et al. 1986), it w a s d e c i d e d to a t t e m p t c o m p l e m e n t a t i o n using the h e t e r o l o g o u s w h e a t r b c S gene. T h e p l a s m i d p D B 5 3 e n c o d e s the S y n e c h o c o c c u s r b c L gene t r a n s c r i b e d b y Plac
154 (Gatenby et al. 1985). Cells containing pDB53 were infected with the phage M13-72 that encodes the wheat rbcS gene, and after an appropriate time cells were assayed for Rubisco activity. Activity was detected at a level of 3 nmol of CO2 fixed per min per mg total protein for the hybrid enzyme. Under identical conditions the control c~nobacterial rbcS phage infection resulted in a level of Rubisco activity of 32 nmol of COs fixed per min per mg total protein (van der Vies et al. 1986). Therefore, the higher plant S subunit can substitute for the cyanobacterial S subunit in assembly of an active Rubisco molecule in vivo. However, the degree of heterologous activity observed is not as high as that obtained using the homologous gene and the specific activity of the hybrid is about 10% of that of the wild-type. Extracts of cells synthesizing the hybrid enzyme were examined by sucrose gradient centrifugation (Fig. 3). A peak of activity was observed and its position in the gradient indicated that the enzyme is a Synechococcus L subunit octomer, and that wheat subunits have assembled with it to produce activity. The binding of the wheat S subunit is non-stoichiometric since only trace amounts of S were present in the active fractions. Most of the wheat S subunits were located near the top of the gradient. It is not clear whether the active hybrid enzyme has an LsSs or L8S
Synthesis and assembly of bacterial and higher plant Rubisco subunits in Escherichia coli A N T H O N Y A. GATENBY Central Research and Development Department, Experimental Station, E.L du Pont de Nemours & Co., Wilmington, DE 19898 U.S.A. Received 2 September 1987; accepted 17 December 1987
Key words: Carbon fixation, enzyme assembly, gene expression, recombinant D N A Abstract. The synthesis in Escherichia coli of both the large and small subunits of cereal ribulose bisphosphate carboxylase/oxygenase has been obtained using expression plasmids and bacteriophages. The level and order of synthesis of the large and small subunits were regulated using different promoters, resulting in different subunit pool sizes and ratios that could be controlled in attempts to optimize the conditions for assembly. Neither assembly nor enzyme activity were observed for the higher plant enzyme. In contrast, cyanobacterial large and small subunits can assemble to give an active holoenzyme in Escherichia coli. By the use of deletion plasmids, followed by infection with appropriate phages, it can be demonstrated that the small subunit is essential for catalysis. However, the small subunit is not required for the assembly of a large subunit octomer core in the case of the Synechococcus enzyme; self-assembly of the octomer will occur in an rbcS deletion strain. The cyanobacterial small subunits can be replaced by wheat small subunits to give an active enzyme in Escherichia coli. The hybrid cyanobacterial large/wheat small subunit enzyme has only about 10% of the level of activity of the wild-type enzyme, reflecting the incomplete saturation of the small subunit binding sites on the large subunit octomer, and possibly a mismatch in the subunit interactions of those small subunits that do bind, giving rise to a lower rate of turnover at the active sites.
Abbreviations: IPTG-isopropyl-fl-D-thiogalactopyranoside, L-large ribulose bisphosphate carboxylase/oxygenase, S-small subunit
subunit,
Rubisco-
Introduction
One objective of the synthesis and assembly of Rubisco in Escherichia coli is to obtain an understanding of the assembly process itself. Once assembly can be demonstrated, then experiments can be designed to understand how the subunits interact with each other, and the importance of particular structures in catalysis, and its possible modification. Following the initial report on the successful expression of a chloroplast rbcL gene in E. coli (Gatenby et al. 1981) there have been rapid developments which have
146 expanded the range of rbcL and rbcS genes from both plant and prokaryotic sources that can be expressed in E coli, resulting in the synthesis of Rubisco large (L) and small (S) subunits. This has enabled experimental systems to be developed that result in the assembly of catalytically active dimeric (Somerville and Somerville 1984), hexameric (Quivey and Tabita 1984) and hexadecameric (reviewed by Bradley et al. 1986) enzymes. These systems can now be used to introduce site specific mutations to investigate the active site, to examine the assembly pathway, and the formation of hybrid enzymes. The isolation from its normal cell of a pathway leading from gene expression to subunit assembly, and its reconstitution in E. coli enables a precise approach to finding what the requirements are for successful assembly by mutating, using different subunit genes and otherwise modifying the system. Another feature, particularly important where subunits are encoded by a multigene family, is that assembly from a cloned gene will provide an enzyme with identical, well characterized subunits of a particular type, rather than the plant derived enzyme which may be a heterogeneous mixture of subunits giving rise to an average value when used in enzyme assays. This review examines some of the strategies that have been adopted to asesemble active Rubisco in E. coli.
Expression of higher-plant Rubisco genes in E. coli The maize and wheat chloroplast rbcL genes were the first plant genes to be expresed in E. coli (Gatenby et al. 1981). The original experiments were designed to detect the synthesis of the L subunit when the appropriate chloroplast DNA fragment was cloned into the plasmid pBR322. The L subunit synthesized had a similar Mr to that observed in chloroplasts, suggesting that translation was initiating correctly at the chloroplast ribosome binding site. Furthermore, the levels of expression observed were similar for both orientations of the rbcL gene in the plasmid, indicating that transcription was probably initiating within chloroplast DNA, and was not the result of readthrough transcription from plasmid promoters. This was later substantiated by the observation that plasmid encoded lambda PL promoters could not direct transcription through the maize rbcL gene unless a phage encoded transcription antitermination protein was present (Gatenby and Cuellar 1985), and that chloroplast promoters could be recognized by the bacterial RNA polymerase (Bradley and Gatenby 1985). It was clear, therefore, that the close similarities between the chloroplast and bacterial transcription and translation signals enabled the plant genes to be readily expressed in E. coli. The expression of chloroplast rbcL genes in E. coli has
147 also been observed for tobacco (Fluhr et al. 1983, Zhu et al. 1984), petunia (Bovenberg et al. 1984) and Chlamydomonas (Zhu et al. 1984). The levels of synthesis of the L subunit, using chloroplast promoters, is low. It was estimated that only about 100-200 monomers were synthesized with the plasmids pZmB 1A and pZmB 1B using chloroplast promoters (Ellis and Gatenby 1984). The level of synthesis could be substantially improved by increasing the rate of transcription of the gene. This was initially achieved by cloning the maize rbcL gene into bacteriophage lambda, and then relying on the early PL promoter, or the late P~t promoter to transcribe the gene during lytic infection (Gatenby et al. 1981). Although rbcL expression levels achieved 0.5-1.0% of total protein synthesis following phage infection, the protein did not accumulate to high levels due to the transitory nature of expression in the lytic cycle. By using a plasmid vector containing the lambda PL promoter, a transcription antiterminator, and the maize rbcL gene, it was possible to induce synthesis of L subunits by thermal denaturation of the cI repressor. Under these conditions synthesis of the L subunit continued for more than 120min and the percentage of L in the cell increased after induction to reach a maximum of 2% of total E. coli protein (Gatenby and Castleton 1982). This represents approximately 60,000 subunit monomers per bacterial cell. High level expression of the maize rbcL gene has also been achieved by Somerville et al. (1983, 1986), using a different strategy. In their experiment a lac promoter was used and a hybrid ribosome binding site was formed using the sites of lacZ and rbcL fused together. The levels of expression obtained with plasmid pPBI3 (Gatenby and Castleton 1982) enabled studies to be carried out on the properties of the L subunit synthesized in the absence of the S subunit. An advantage with this approach is that the use of denaturing conditions which are used to separate the L and S subunits in the holoenzyme, can be avoided. Pulse-chase experiments showed that the L subunit was stable in E. coli for at least 4 h (Gatenby 1984). The protein was also insoluble, whether expressed at high or low levels. Attempts to detect enzyme activity, or to form a stable enzyme.metal.CO2.[~4C] carboxyarabinitol bisphosphate quaternary complex were unsuccessful. The presence of the substrate ribulose bisphosphate, or the positive effector fructose 1,6-bisphosphate, did not improve solubility, although it was found that pH 11-12 solubilized L slowly suggesting that aggregate structures may have formed. It would appear that maize L adopts an inactive, insoluble conformation after, or during, synthesis. Similar results have been reported by others for maize L (Somerville et al. 1983, 1986).
148
Fig. 1. Fluorograph showing the synthesis of Rubisco subunits in E. coli following induction of plasmid pLSS308. Cells were labelled with [35S] methionine in the presence of IPTG to induce wheat small subunit (SS) synthesis, and either at 30°C or 41 °C to obtain repression or induction respectively of maize large subunit (LS) synthesis. The 35S-labelled polypeptides were incubated with anti-wheat RuBPCase serum and protein A Sepharose before electrophoresis on polyacrylamide gels containing SDS. Reprinted from Gatenby et al. (1987).
In an attempt to overcome the problem of L insolubility, and to assemble a plant holoenzyme, experiments were designed to co-express the L and S subunit genes in E. coli (Bradley et al. 1986, Gatenby et al. 1987). These experiments required that the nuclear encoded S subunit gene be cloned into an E. coli vector such that the S polypeptide is synthesized without the amino-terminal transit peptide, which is presumably only necessary for transport of the protein into chloroplasts (Broglie et al. 1983). These constructions have been made by subcloning a region of the wheat rbcS gene, which encodes the mature S subunit, from a c D N A clone into plasmid and M13 vectors (Bradley et al. 1986, van der Vies et al. 1986). The resulting
149 expression clones (p565, pLSS308, and M13-72) produce the S subunit as a lacZ::rbcS fusion protein in E. coll. The amino-terminals of the S fusion protein contain 11 (p565 and pLSS308) or 10 (M13-72) amino acids from fl-galactosidase. The first four amino acids of the mature S polypeptide have also been deleted. Transcription and translation of the rbcS gene fusions are initiated from the lacZ promoter and ribosome-binding sites respectively, resulting in synthesis of an S polypeptide of 15.6kDa, which migrates slightly slower than authentic wheat isolated from chloroplasts (14.7 kDa) as expected from the fusion at its N-terminus. The rbcS expression clones were used to study plant holoenzyme formation in two ways, either using heterologous subunit mixtures, or homologous subunits. In plasmid pLSS308 the wheat rbcS gene was placed under the control of Plac. and the maize rbcL genes were placed under the control of PL" The use of different promoters allows the independent expression of the rbcS and rbcL genes. The expression of rbcL from pLSS308 relied on translational intiation at the chloroplast ribosome binding site (Gatenby et al. 1981), and temperature induced transcription obtained by shifting cultures to 41-42 °C in the presence of the anti-terminating protein p N (Gatenby and Cuellar 1985). Fig. 1 shows the result of a typical induction experiment in which S synthesis is constitutive, but the synthesis of L has been increased at 41 °C. The advantage of this approach is that by altering the pool sizes of the two subunits it might be possible to arrive at an optimal ratio of subunit types for assembly to occur. When the S subunit is synthesized in the absence of the L subunit it is found in the soluble fraction, but the polypeptide is unstable and has a half-life of less than 15 min (Gatenby et al. 1987). Its size on sucrose gradients indicates a monomeric or dimeric form. When L subunit synthesis is induced in cells containing the S subunit both subunits are found predominantly in the insoluble fraction and are fully stable for more than 120 min, suggesting that aggregation of the subunits may occur. The two subunits do not assemble to form an active holoenzyme in vivo even when nascent L subunits are synthesized in a pool of mature S subunits. To test if the synthesis of homologous subunits in E. coli would provide a more favorable opportunity for assembly, the wheat rbcS and rbcL genes were expressed together (Bradley et al. 1986). A bacteriophage M I3 clone (M68) expressing the wheat L subunit under the control of Plac was constructed. Cells were allowed to first synthesize the S subunit by means of a plasmid-borne rbcS gene (p565), and were then infected with the bacteriophage M68 (rbcL). Phage M68 was also used to infect plasmid-free E coil cells to examine the properties of the wheat L subunit in the absence of S. In contrast to the maize L subunit insolubility in E. coli, about 60% of
150
Fig. 2. Velocity sedimentation in sucrose gradients of cyanobacterial Rubisco synthesized a n d assembled in E. coli. Soluble protein samples were prepared from E. coli cells that had synthesized the Synechococcus L and S subunits following induction of plasmid pSV55. In panel A the bacterial extract, or purified spinach Rubisco, was incubated with the transition state analogue [~4C] carboxyarabinitol bisphosphate before centrifugation. In panel B the distribution of Rubisco activity was measured in a similar sucrose gradient using the induced bacterial extract. In panel C an immunoblot analysis of the fractions from panel B was carried out using anti-Rubisco serum. The M track is a maize leaf track used as a marker. The Mr markers used to calibrate the sucrose gradients are shown by arrows. Reprinted from Gatenby et al. (1985).
151 the wheat L subunit was found in the soluble fraction (Bradley et al. 1986). The S subunit did not appear to influence the amount of L that was found in the soluble fraction. It was also observed that about 80% of the wheat S subunit was in the soluble fraction from E. coli cells. Rubisco activity was undetectable. To examine subunit assembly, the soluble proteins from cells expressing both L and S were subjected to velocity sedimentation on sucrose gradients. Most of the S subunit remained at the top of the gradient. The wheat L subunit, although distributed throughout the gradient, sedimented with a peak abundance indicating a molecular mass of 750kDa. This probably represents a large protein aggregate. Therefore, even with a homologous combination of subunits, it was not possible to detect assembly or activity of the higher plant enzyme.
Synthesis of bacterial Rubisco genes in E. coli
An attractive feature of synthesizing bacterial Rubisco in E. coli is that the prokaryotic structure of the genes can facilitate their expression. A range of different forms of active Rubisco has been synthesized in E. coil using the isolated genes from several different species. These include the homodimeric L subunit enzyme from Rhodospirillum rubrum (Somerville and Somerville 1984, Latimer et al. 1986), the hexadecameric L subunit enzyme from Rhodopseudomonas sphaeroides (Quivey and Tabita 1984) and the hexadecameric enzyme containing both L and S subunits from cyanobacteria (Christeller et al. 1985, Gatenby et al. 1985, Gurevitz et al. 1985, Tabita and Small 1985), Chromatium (Viale et al. 1985) and Rhodopseudomonas sphaeroides (Gibson and Tabita 1986). The prokaryotic nature of cyanobacteria also simplifies the expression of their genes in E. coli, because the absence of an organdie, and therefore an S subunit transit peptide, enables the direct synthesis of mature S. In addition, the rbcL and rbcS genes are physically linked on a contiguous fragment of DNA. For expression of the Synechococcus Rubisco genes they were initially transcribed from a bacteriophage lambda PL promoter following temperature induction (Gatenby et al. 1985). Synthesis of the L and S subunits could be detected, together with enzyme activity and [14C] carboxyarabinitol bisphosphate binding. In addition, sucrose gradient centrifugation resolved a peak of enzyme activity that was similar in size to the spinach holoenzyme (Fig. 2). The large size indicated that the L 8 structure was formed, but an L 8Sn complex where n is less than 8 would be difficult to resolve from L8S8 because of the small size of S. Although S must be present to obtain an active enzyme (Andrews and Abel 1981), it was suspec-
152ted that the S binding sites on the L octomer were not fully saturated because of two observations. First, the specific activity in cell extracts is lower than expected based on the known amount of L8 present (as determined by [14C] carboxyarabinitol bisphosphate binding). Second, the amount of 35S present in the purified L and S polypeptides indicated that L 8Sn structures were formed where n = 2 or 3. More recently it has been shown that transfer of the appropriate Synechococcus D N A fragment from the PL expression plasmid to a Plac expression plasmid increases the amount of Rubisco synthesized and enables either the assembly of L8 $8 structures, or structures almost completely saturated with S subunits (Bradley et al. 1986, van der Vies et al. 1986). Christeller et al. (1985) demonstrated that cloning the Anacystis nidulans rbcL and rbcS genes separately resulted in the absence of enzyme activity, indicating that both L and S subunits were required for activity. We have obtained similar results using plasmid or phage encoded Synechococcus rbcL and rbcS genes (van der Vies et al. 1985). Plasmid pDB50 contains both rbcL and rbcS and directs the synthesis of active Rubisco. If rbcS is deleted, the resulting plasmid (pDB53) synthesized only the L subunit which is inactive. Since this deletion does not result in any addition, or noticeable reduction, in the amount of soluble L subunit synthesized, it is concluded that the product of rbcS is essential for activity. If this is correct, infection E. coli cells containing the S subunit plasmid delection (pDB53) with M 13 phages encoding the rbcS gene should restore Rubisco activity by complementation. This is exactly what happens. Following phage infection, Rubisco activity can be detected. These deletion and complementation experiments have also enabled us to obtain information about the assembly process of Rubisco in E. coil. It appears that the major soluble protein structures formed in cells synthesizing the L subunit alone (pDB53) are oligomeric forms, principally an L octomer and an L dimer (van der Vies et al. 1986). Therefore, the formation of a Synechococcus L8 structure does not require S subunits to be present. The L2 structure may represent an intermediate which then selfassembles to the L8 form, or may be the result of disassembly of the L octomer. When both cyanobacterial L and S subunits are synthesized in the same cell the L dimer is not present, and the main L peak shifts further down the gradient indicating the binding of S subunits. The presence of S subunits appears to reduce the pool of unassembled L2 and promote formation of LB.
Assembly of a hybrid Rubisco in E. coli
Since it was possible to demonstrate complementation of Rubisco activity
153
Fig. 3. Analysis of heterologous Rubisco assembly in E. coli. A soluble protein extract of E. coli cells containing plasmid pDB53 (encoding the cyanobacterial rbcL gene), and infected with the phage ML3-72 (encoding the wheat rbcS gene) was fractionated on a sucrose gradient. Panel a shows an immunoblot of the fractions using anti-Rubisco serum, and panel b shows the determination of Rubisco activity. The gradient was calibrated with Mr markers shown by arrows. Wheat leaf protein was used as a marker in the right lane of panel a. Reprinted from van der Vies et al. (1986).
in a n S s u b u n i t p l a s m i d deletion strain b y infection w i t h M 13 r b c S + p h a g e s e n c o d i n g a h o m o l o g o u s S s u b u n i t ( v a n der Vies et al. 1986), it w a s d e c i d e d to a t t e m p t c o m p l e m e n t a t i o n using the h e t e r o l o g o u s w h e a t r b c S gene. T h e p l a s m i d p D B 5 3 e n c o d e s the S y n e c h o c o c c u s r b c L gene t r a n s c r i b e d b y Plac
154 (Gatenby et al. 1985). Cells containing pDB53 were infected with the phage M13-72 that encodes the wheat rbcS gene, and after an appropriate time cells were assayed for Rubisco activity. Activity was detected at a level of 3 nmol of CO2 fixed per min per mg total protein for the hybrid enzyme. Under identical conditions the control c~nobacterial rbcS phage infection resulted in a level of Rubisco activity of 32 nmol of COs fixed per min per mg total protein (van der Vies et al. 1986). Therefore, the higher plant S subunit can substitute for the cyanobacterial S subunit in assembly of an active Rubisco molecule in vivo. However, the degree of heterologous activity observed is not as high as that obtained using the homologous gene and the specific activity of the hybrid is about 10% of that of the wild-type. Extracts of cells synthesizing the hybrid enzyme were examined by sucrose gradient centrifugation (Fig. 3). A peak of activity was observed and its position in the gradient indicated that the enzyme is a Synechococcus L subunit octomer, and that wheat subunits have assembled with it to produce activity. The binding of the wheat S subunit is non-stoichiometric since only trace amounts of S were present in the active fractions. Most of the wheat S subunits were located near the top of the gradient. It is not clear whether the active hybrid enzyme has an LsSs or L8S
