Engineering of chimeric eukaryotic / bacterial Rubisco large subunits in Escherichia coli

Ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco) is a rate-limiting photosynthetic enzyme that catalyzes carbon fixation in the Calvin cycle. Much interest has been devoted to engineering this ubiquitous enzyme with the goal of increasing plant growth. However, experiments that have successfully produced improved Rubisco variants, via directed evolution in Escherichia coli, are limited to bacterial Rubisco because the eukaryotic holoenzyme cannot be produced in E. coli. The present study attempts to determine the specific differences between bacterial and eukaryotic Rubisco large subunit primary structure that are responsible for preventing heterologous eukaryotic holoenzyme formation in E. coli. A series of chimeric Synechococcus Rubiscos were created in which different sections of the large subunit were swapped with those of the homologous Chlamydomonas Rubisco. Chimeric holoenzymes that can form in vivo would indicate that differences within the swapped sections do not disrupt holoenzyme formation. Large subunit residues 1–97, 198–247 and 448–472 were successfully swapped without inhibiting holoenzyme formation. In all ten chimeras, protein expression was observed for the separate subunits at a detectable level. As a first approximation, the regions that can tolerate swapping may be targets for future engineering.


INTRODUCTION
Ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco,E.C. 4.1.1.39)is the world's most abundant enzyme, and can account for up to 65% of total soluble protein in leaf extracts (Ellis, 1979).It is also arguably the most important enzyme in the biosphere, acting as the gateway for inorganic carbon assimilation through carbon fixation in the first step of the Calvin cycle (reviewed in Cleland et al., 1998).Rubisco catalyzes the carboxylation of ribulose-1,5-bisphosphate with CO 2 , producing two molecules of 3-phosphoglycerate, which are eventually used for carbohydrate synthesis (reviewed in Gutteridge and Gatenby, 1995).
Rubisco is a complex multi-subunit enzyme assembly of large (RbcL) and small subunits (RbcS), with its active site located between large subunits (Curmi et al., 1992;Newman and Gutteridge, 1993).Hexadecameric Form I Rubisco has been intensely studied because it is the most abundant form, found mostly in higher plants, various algae and many photosynthetic bacteria, and it is also a prime target for genetic engineering (Andersson and Backlund, 2008).Form I Rubisco is composed of eight RbcL clustered together at the core, arranged as a tetramer of dimers, with four RbcS located at opposite poles of the large subunit octamer, stabilizing the quaternary complex (Andersson et al., 1989;Knight et al., 1990).In land plants and green algae, RbcL is located in the multicopy chloroplast genome while RbcS are present as a multigene family within the nucleus (reviewed in Spreitzer and Salvucci, 2002).In cyanobacteria, red algae and most Chromista, both RbcL and RbcS are located within the same operon in the bacterial or organellar chromosome, with an RbcX chaperone gene positioned between RbcL and RbcS in several cyanobacterial species (Onizuka et al., 2004;Saschenbrecker et al., 2007).
The world's most abundant enzyme is also notorious for being catalytically sluggish and inefficient as compared to all other photosynthetic enzymes in plants, with the carboxylation turnover rate (K cat ) ranging only from 1-13 per second (reviewed in Tcherkez et al., 2006).Moreover, Rubisco is bifunctional in nature: it fixes O 2 instead of CO 2 onto ribulose-1,5-bisphosphate in up to one of every three catalytic cycles in land plants grown under current atmospheric conditions (500-fold more O 2 than CO 2 ).The resultant side product, 2-phosphoglycolate, is cytotoxic and has to be metabolized at the expense of additional cellular energy in a process known as photorespiration (reviewed in Ogren, 1984).Therefore, despite the pivotal role of Rubisco in sustaining nearly all life on earth, this enzyme is heavily flawed and is rate-limiting to the entire photosynthesis process, which is a direct restriction to the synthesis of most world biomass (Parry et al., 2003).
Rubisco has been intensively studied in recent decades and is becoming a favorite target for genetic manipulation to improve its catalytic capacity and efficiency (Spreitzer et al., 2005;Parry et al., 2007;Whitney et al., 2011).Rubiscos with the highest CO 2 /O 2 specificities are not found in crop species but rather in thermophilic red algae, with Galdieria partita having the highest recorded CO 2 / O 2 specificity factor of 238 (Uemura et al., 1997).It is estimated that if the kinetic properties of Rubisco from the marine red alga Griffithsia monilis (CO 2 /O 2 specificity factor of 167) were engineered into C 3 crops, yields could increase by up to 30% (Long et al., 2006).Various strategies have been deployed to enhance the kinetic properties of Rubisco, targeting both RbcL (Sharwood et al., 2008;Satagopan and Spreitzer, 2008) and RbcS (Karkehabadi et al., 2005;Genkov et al., 2010;Ishikawa et al., 2011).Although genetic engineering of hybrid Rubiscos partially imparted the desired kinetic properties of the foreign Rubiscos, the effects were not always beneficial because of trade-offs in kinetic efficiency (Spreitzer et al., 2005;Satagopan and Spreitzer, 2008;Ishikawa et al., 2011), impaired photosynthetic growth (Sharwood et al., 2008;Genkov et al., 2010), loss of chloroplast pyrenoid and CO 2 -concentrating mechanisms (Genkov et al., 2010;Meyer et al., 2012), and, more commonly, problems with Rubisco folding and assembly (Sharwood et al., 2008;Whitney et al., 2009;Genkov et al., 2010).
Non-assembly of Rubisco subunits is a common problem encountered when a phylogenetically distant Rubisco is expressed in a target organism.When RbcL and RbcS from wheat were inserted into E. coli, the genes were expressed but the polypeptides failed to productively fold and assemble into active enzymes (Cloney et al., 1993).A similar problem was encountered when Form I Rubiscos from the non-green algae Galdieria sulphuraria and Phaeodactylum tricornutum were expressed in tobacco, perhaps because the foreign RbcL could not enter the host chaperone-mediated folding pathway (Whitney et al., 2001).On the other hand, heterologous expression and assembly of cyanobacterial (Synechococcus PCC6301) Rubisco in E. coli is invariably succesful (Goloubinoff et al., 1989;Parikh et al., 2006;Mueller-Cajar and Whitney, 2008).Different but possibly homologous assemblyassisting factors have evolved, which assist with higher plant Rubisco folding and assembly in chloroplasts even though these plastids originated from a cyanobacterial endosymbiont (reviewed in Nishimura et al., 2008).
Because higher plant and Synechococcus RbcL sequences are approximately 80% identical (Shinozaki et al., 1983), subtle residue differences between eukaryotic and bacterial RbcL may account for the success or failure of holoenzyme formation for Rubiscos expressed in E. coli.Holoenzyme formation itself is a series of steps involving post-translational protein processing, recognition-folding by host chaperones and assembly events, most of which involve a growing list of assisting-protein partners that is yet to be fully defined (Onizuka et al., 2004;Saschenbrecker et al., 2007;Kolesinski et al., 2014).
In this study, we aimed to identify the RbcL domains of cyanobacterial Rubisco, as opposed to eukaryotic Rubisco, that may be critical for successful holoenzyme formation in E. coli.Sections spanning the entire coding length of cyanobacterial RbcL from Synechococcus PCC6301 were systematically replaced with the homologous sections of eukaryotic RbcL from the unicellular green alga Chlamydomonas reinhardtii to create ten chimeric mutants.Because successful holoenzyme formation and production of eukaryotic Rubisco in E. coli has yet to be reported, we hypothesized that chimeras which still assembled must contain eukaryotic sections that are not responsible for the contrast in holoenzyme formation.Through this systematic swapping, Synechococcus RbcL residues 1-97, 198-247 and 448-472 (more specifically, the variant sequences within these residue ranges) can be ruled out as key determinants for successful holoenzyme formation within the E. coli host.

MATERIALS AND METHODS
Chimeric RbcL plasmid construction and E. coli transformation PCR amplicons of Synechococcus RbcL fragments and the RbcS gene from pTrcSynLS (Mueller-Cajar and Whitney, 2008), and Chlamydomonas RbcL fragments from pLS-WT (Du and Spreitzer, 2000), were digested with combinations of NcoI, BsmBI and PstI (Tables 1 and 2).The digested fragments were then ligated into pTrcHisB (Invitrogen), and then electroporated into E. coli XL-1 Blue, transformants of which were selected overnight on LB plates containing 200 μg/ml ampicillin at 37 °C.Recombinant plasmids were screened by plasmid size, restriction enzyme digestion and PCR, and the bicistronic RbcL-RbcS operon was fully sequenced to verify the chimeric constructs.

Heterologous expression and assembly analysis of chimeric Rubiscos
XL-1 Blue E. coli cells were transformed with the sequenced RbcL-RbcS constructs and selected on 200 μg/ml ampicillin LB plates at 37 °C prior to protein induction and extraction.Single colonies were picked and grown overnight in LB broth containing 100 μg/ml ampicillin at 37 °C, after which an aliquot was transferred to fresh 100 μg/ml ampicillin LB broth and grown to an OD 600 of 0.5 at 37 °C.Rubisco gene expression was induced for 16 h with 0.5 mM IPTG.After protein induction, E. coli cells were harvested by centrifugation (5 min, 5000 g, 4 °C), resuspended to 10% (w/v) in ice-cold extraction buffer (50 mM Bicine-NaOH, pH 8.0, 10 mM MgCl 2 , 10 mM NaHCO 3 , 2 mM DTT) and sonicated.Total cellular proteins (crude lysate) were resolved on 7.5% native-and 12% SDS-polyacrylamide gels.For SDS-PAGE, the crude lysate was denatured beforehand by mixing with sample loading buffer (30% (w/v) sucrose, 5% (w/v) SDS, 0.05% (w/v) bromophenol blue, 100 mM DTT) at a 3:2 ratio and boiling for 30 min.For Western blotting (Towbin et al., 1979), resolved proteins were probed using rabbit anti-Synechococcus PCC6301 Rubisco IgG (Parikh et al., 2006).

Construction of chimeric Rubiscos
Sequence alignment between Synechococcus PCC6301 RbcL and Chlamydomonas reinhardtii RbcL indicates that 85 residues differ between the two species (Fig. 1).Also, Chlamydomonas RbcL has three additional residues at the N terminus (Fig. 1).One or more of these differences must be responsible for the species specificity of holoenzyme formation in vivo.The whole RbcL was divided into nine sections of 50 amino acids and a tenth section of 25 amino acids (Fig. 2).Each section in Synechococcus RbcL was separately replaced with the collinear section of Chlamydomonas RbcL to create 10 chimeric mutants (Fig. 2).Section demarcations were based on the Chlamydomonas sequence, as most previous phylogenybased mutational analyses of RbcL have used the eukaryotic Form I Rubisco numbering (Zhu and Spreitzer, 1996;Du et al., 2003;Spreitzer et al., 2005;Satagopan and Spreitzer, 2008;Whitney et al., 2011).

Expression of chimeric Rubiscos
To determine whether the chimeric Rubiscos were still expressed despite alterations in the gene and protein sequence, SDS-PAGE was carried out on total cellular protein from each chimeric construct (Fig. 3).Both large and small subunits were detected in every mutant (Fig. 3).Therefore, despite extensive alterations to the original Synechococcus RbcL sequence, the mutant genes can still be transcribed and translated.
Hence, the aforementioned residues on Synechococcus RbcL are not essential determinants for cyanobacteriaspecific holoenzyme formation in E. coli since the four mutant RbcLs were still capable of post-translational folding and assembly into hexadecameric holoenzymes.
Among the assembled chimeric holoenzymes, striking differences in amounts were observed (Fig. 4A).Holoenzyme level for the Section 5 chimera was comparable to wild-type, but levels for Sections 1, 2 and 10 were lower (Fig. 4A).For Sections 1 and 2 chimeras, the decrease may be due to reduced RbcS translation for reasons that are unclear, especially for Section 1, which has barely detectable levels of RbcS (Fig. 3).As for Section 10, the phylogenetic changes may have compromised holoenzyme stability (Guo et al., 2004) due to the foreign Chlamydomonas residues within the predominantly Synechococcus large subunit.Moreover, among the assembled chimeric holoenzymes, comparing the relative amounts of RbcL that were expressed (Fig. 3) to those of assembled soluble holoenzymes (Fig. 4A), only small amounts of the expressed RbcL were soluble with the remainder most likely insoluble and aggregated, consistent with results from a previous study using this expression system (Mueller-Cajar and Whitney, 2008).
Mutations in Sections 3, 4, 6, 7, 8 and 9 eliminate holoenzyme formation (Fig. 4A), even though RbcS and RbcL monomers accumulated to levels sufficient to form detectable Rubisco holoenzymes, if they were able to do so (Fig. 3).It is likely that some of these changes had no role in post-translational processing, folding or assembly events, but instead effected a global disturbance to protein tertiary structure.For the Section 4 mutant, which has three substitutions, because substituted residues Leu-161 and Met-169 (numbering based on Chlamydomonas RbcL) can be mutated without affecting function (Durão et al., 2015), the remaining P168G (Chlamydomonas RbcL numbering) mutation may be responsible for destabilizing the large subunit fold as a result of the loss of the bulkier and more rigid proline pyrrole group.As for the Section 7 mutant, which has six substitutions, four of the substituted residues on the catalytic Loop 6 were previously changed to spinach-specific residues, but an assembled holoenzyme with decreased activity was still produced (Gutteridge et al., 1993).Therefore, the remaining two C317A and L320M (Chlamydomonas RbcL numbering) mutations must be responsible for the lack of mutant assembly.The other non-assembling chimeras have seven to twelve substitutions each, which remain to be explored.
The observed non-assembly in some of these mutants may also be caused by disruption of inter-subunit interactions.Eleven of the 85 residues that are different (Fig. 1) are located at interfaces between large and small subunits, but only four are substituted among the nonassembling Rubisco mutants.Comparing the Rubisco crystal structures of Synechococcus (PDB ID 1RBL) and Chlamydomonas (PDB ID 1GK8), Synechococcus to Chlamydomonas substitution E351D (Chlamydomonas RbcL numbering), which replaces the carboxyethyl sidechain of Glu-351 with the shorter carboxymethyl sidechain of aspartate, potentially disrupts the salt bridge with Synechococcus small subunit Lys-96.Another sub- stitution, T418A (Chlamydomonas RbcL numbering), which replaces the indole side-chain of Trp-418 with the smaller methyl side-chain of alanine, could eliminate its van der Waals contact with small subunit Leu-6, although, in the Chlamydomonas structure, Ala-418 is in contact with small subunit Trp-4.However, large-small subunit interactions are quite robust and malleable, evidenced from the myriad of engineered hybrid enzymes with evolutionarily distant large-small subunits that have been expressed in higher plants, algae and bacteria (van der Vies et al., 1986;Kanevski et al., 1999;Wang et al., 2001;Sharwood et al., 2008;Genkov et al., 2010;Ishikawa et al., 2011;Joshi et al., 2015).Similarly, we discovered that Chlamydomonas RbcS can assemble with Synechococcus RbcL in E. coli to form hexadecameric holoenzyme (data not shown).The possibility that changes at large subunit interfaces compromised assembly was also considered, but no obvious loss in interaction between large subunits as a result of the phylogenetic changes can be deduced by comparing the Synechococcus and Chlamydomonas structures.
The lack of holoenzyme may also be attributed to posttranslational modification systems that are missing in E. coli (reviewed in Kamionka, 2011) but are present in Chlamydomonas (reviewed in Houtz et al., 2008), which might be essential for subtle structural changes to complement the differences in primary structure (Fig. 1) and encourage proper folding or assembly.Endogenous wildtype Chlamydomonas RbcL is acetylated at Pro-3 (Houtz et al., 1992), hydroxylated at Pro-104 and Pro-151, and methylated at Cys-256 and Cys-369 (Taylor et al., 2001), whereas none of these modifications have been reported in Synechococcus RbcL.However, among post-translational modifications, glycosylation is believed to be the most important modification for folding and stability of polypeptides (Hebert et al., 2014).Since glycosylation does not naturally occur in wild-type endogenous Chlamydomonas RbcL, the limited post-translational modification systems in E. coli may ultimately have less of an impact on the ability of the chimeric Rubiscos to fold heterologously.
Another protein processing system crucial to folding and assembly of Synechococcus Rubisco is the molecular chaperone network within the host E. coli.In general, chaperone-mediated folding first involves the recognition of substrate protein in order for the substrate to gain access to the molecular chaperone (Chaudhuri and Gupta, 2005).Of the E. coli chaperones, the GroEL-GroES chaperonin system is one of the most intensively studied molecular chaperone systems (Lin and Rye, 2006;Ruiz-González and Fares, 2013).The role of GroEL-GroES chaperonins in enabling Synechococcus Rubisco holoenzyme formation is well-documented (Goloubinoff et al., 1989;Greene et al., 2007;Saschenbrecker et al., 2007;Mueller-Cajar and Whitney, 2008).Overexpression of GroEL-GroES in E. coli increased the amount of catalyt-ically competent cyanobacterial Rubisco (Goloubinoff et al., 1989).Conversely, when deleterious mutations were introduced into the GroE genes, neither assembled holoenzyme nor Rubisco activity could be detected (Goloubinoff et al., 1989).
Hence, understanding the structural properties common to proteins targeted and folded by GroEL-GroES might explain the discrimination between eukaryotic and bacterial Rubisco in E. coli.GroEL discerns and binds substrate proteins only in their non-native form because the GroEL binding sites are only exposed in the unstructured non-native states of these substrate proteins (Goloubinoff et al., 1989;Stan et al., 2006).Sequence similarity to the mobile-loop peptide of GroES and hydrophobicity are important characteristics for the recognition of binding sites by GroEL (Chaudhuri and Gupta, 2005;Stan et al., 2006).
However, even upon successful entry of the substrate protein into the GroEL cavity, formation of the complete holoenzyme is not guaranteed.The next challenge is for the substrate protein to fold into its native tertiary structure and exit the chaperone (Weissman et al., 1995(Weissman et al., , 1996)).RbcL chimeras (Fig. 2) that gain entry into the chaperone might still misfold, condemning the mutant protein to a perpetual cycle of misfolding and unfolding within the chaperone until its degradation or aggregation in vivo (reviewed in Lin and Rye, 2006).Moreover, RbcL chimeras that are successfully folded by GroEL-GroES might still fail to assemble as complete holoenzymes in the absence of interaction with other Rubisco assemblyassisting proteins (Cloney et al., 1993;reviewed in Gutteridge and Gatenby, 1995;Saschenbrecker et al., 2007;Bracher et al., 2011;Kolesinski et al., 2014).
The large size (Spreitzer and Salvucci, 2002), multiplicity of large and small subunits (Curmi et al., 1992;Newman and Gutteridge, 1993;Taylor et al., 2001), and complex folding pattern of Rubisco (Goloubinoff et al., 1989;Durão et al., 2015) remain a great challenge in understanding and manipulating this enzyme.The true reasons for the absence of holoenzyme formation due to swapping of Synechococcus RbcL Sections 3, 4, 6, 7, 8 and 9 remain uncertain until further testing.Nevertheless, our experimental results suggest that Synechococcus RbcL Sections 1, 2, 5 and 10 (residues 1-47, 48-97, 198-247 and 448-472) are amenable to phylogenetic swapping without gravely affecting folding and assembly in E. coli; hence, omitting these sections should assist future exploration of species-specific assembly or chaperone recognition sections.

Fig. 1 .
Fig. 1.Sequence alignment of Synechococcus and Chlamydomonas RbcL.Numbering is based on the Chlamydomonas sequence.The Synechococcus sequence has a shorter N terminus, with the missing three amino acids represented by (~).There are 85 residue differences between the two sequences.Identical residues are highlighted in black.

Fig. 2 .
Fig. 2. Schematic illustration of chimeric Synechococcus RbcL genes with sections replaced from Chlamydomonas RbcL.Sections of Chlamydomonas RbcL (gray bars) were amplified from pLS-WT (Du and Spreitzer, 2000) and recombined with the remaining sections of Synechococcus RbcL (white bars) from pTrcSynLS (Mueller-Cajar and Whitney, 2008), creating a total of ten chimeric section constructs.In each RbcL fragment, the first and last codon numbers (in lieu of base pair numbers) are indicated (white: Chlamydomonas; black: Synechococcus).

Fig. 3 .
Fig. 3. SDS-PAGE and Western blot analysis of total protein from E. coli expressing Synechococcus wild-type or chimeric mutant Rubiscos.Equal volumes of denatured total cell lysate from E. coli harboring RbcL-RbcS were separated on denaturing 12% polyacrylamide gels and stained with Coomassie Brilliant Blue R-250 (top), or transferred to a nitrocellulose membrane and probed with antibody against Synechococcus Rubisco (middle and bottom).The protein marker (lane M) with relevant sizes (in kDa, left), and the Rubisco large subunit (LS) and small subunit (SS), are indicated.

Fig. 4 .
Fig.4.Non-denaturing-PAGE and Western blot analysis of total protein from E. coli expressing Synechococcus wild-type or chimeric mutant Rubiscos.(A) Equal volumes of total cell lysate from E. coli harboring RbcL-RbcS were separated on non-denaturing 7.5% polyacrylamide gels and stained with Coomassie Brilliant Blue R-250 (upper), or transferred to a nitrocellulose membrane and probed with an antibody against Synechococcus Rubisco (lower).Assembled hexadecameric Rubisco is indicated (L 8 S 8 ).(B) The L 8 S 8 wild-type Rubisco protein was excised from the non-denaturing gel and further resolved by SDS-PAGE.Large (LS) and small (SS) subunits were visualized by Coomassie staining.

Table 1 .
Construction of chimeric Rubisco plasmids