Functional metagenomic selection of ribulose 1, 5-bishosphate carboxylase/oxygenase from uncultivated bacteria

Summary

Ribulose 1,5-bisphosphate carboxylase/oxygenase (RubisCO) is a critical yet severely inefficient enzyme that catalyses the fixation of virtually all of the carbon found on Earth. Here, we report a functional metagenomic selection that recovers physiologically active RubisCO molecules directly from uncultivated and largely unknown members of natural microbial communities. Selection is based on CO₂-dependent growth in a host strain capable of expressing environmental deoxyribonucleic acid (DNA), precluding the need for pure cultures or screening of recombinant clones for enzymatic activity. Seventeen functional RubisCO-encoded sequences were selected using DNA extracted from soil and river autotrophic enrichments, a photosynthetic biofilm and a subsurface groundwater aquifer. Notably, three related form II RubisCOs were recovered which share high sequence similarity with metagenomic scaffolds from uncultivated members of the Gallionellaceae family. One of the Gallionellaceae RubisCOs was purified and shown to possess CO₂/O₂ specificity typical of form II enzymes. X-ray crystallography determined that this enzyme is a hexamer, only the second form II multimer ever solved and the first RubisCO structure obtained from an uncultivated bacterium. Functional metagenomic selection leverages natural biological diversity and billions of years of evolution inherent in environmental communities, providing a new window into the discovery of CO₂-fixing enzymes not previously characterized.

Introduction

Ribulose 1,5-bisphosphate carboxylase/oxygenase (RubisCO) is the most abundant protein on earth and catalyses CO₂ fixation onto the enzyme-bound enediolate of ribulose 1, 5-bisphosphate (RuBP) to produce organic carbon in plants, algae and autotrophic bacteria (Tabita et al., 2008). However, RubisCO is a sluggish catalyst and its efficiency is further limited by its poor ability to discriminate between the gaseous substrates CO₂ and O₂ (Spreitzer and Salvucci, 2002; Andersson, 2008), with the oxygen fixation reaction leading to energetically wasteful metabolism. Because of RubisCO’s importance in CO₂ bioconversions, a number of studies have endeavoured to increase carbon capture through various artificial evolution and bioengineering methods to produce RubisCO enzymes with increased activity or higher CO₂ specificity (Spreitzer and Salvucci, 2002; Smith and Tabita, 2003; Parikh et al., 2006; Yoshida et al., 2007; Mueller-Cajar and Whitney, 2008; Satagopan et al., 2009; 2014; Cai et al., 2014; Lin et al., 2014). While these and recent studies to manipulate the properties of RubisCO are especially promising (Lin et al., 2014; Hauser et al., 2015), it is clear that sequence or structural analyses alone are not accurate predictors of enzyme function. The molecular mechanisms underpinning the vast differences in catalytic properties between highly related RubisCO protein sequences and structural homologues from different taxa remain enigmatic (Tabita, 1999; Tabita et al., 2008).

Most of what is known regarding bacterial RubisCO function is derived from pure culture analyses and relatively few model enzymes. However, recent massive deoxyribonucleic acid (DNA) sequencing efforts have shown the abundant presence of diverse and novel RubisCO-encoding genes in environmental communities of largely uncultivated bacteria and new candidate phyla, the so-called ‘microbial dark matter’ (Wrighton et al., 2012; submitted; Campbell et al., 2013; Guo et al., 2013; Castelle et al., 2015; Tebo et al., 2015). Although the physiological significance of these novel RubisCO enzymes is unquantified, these organisms likely have evolved significant structural and functional adaptations to allow CO₂ to be metabolized in diverse environments such as in marine surface waters, hydrothermal vents and the terrestrial subsurface (Witte et al., 2010; Wrighton et al., 2012; submitted; Böhnke and Perner, 2015; Castelle et al., 2015). Sequencing, cloning and activity screening of purified recombinant proteins is possible on a small scale (Böhnke and Perner, 2015). However, it is not always possible to discern physiological function by sequence-gazing alone, nor is it practical to produce and individually screen and characterize recombinant proteins from the large number of novel RubisCO-encoding genes recently observed (Wrighton et al., 2012; submitted; Campbell et al., 2013; Guo et al., 2013; Castelle et al., 2015; Tebo et al., 2015). Whereas a previous study demonstrated the feasibility of studying environmentally derived RubisCO genes in a recombinant photosynthetic host (Witte et al., 2010), a directed functional metagenomic selection approach is necessary for highthroughput retrieval of desirable new RubisCO enzymes from uncultivated organisms. The isolation and analysis of such proteins can ultimately provide additional biochemical and structural insights with respect to key aspects of catalysis (Satagopan et al., 2009; 2014). In this study, we implemented such a streamlined functional metagenomic approach to isolate diverse RubisCOs from three environments and demonstrate that the approach is applicable for gaining novel biochemical and structural properties from uncultivated ‘microbial dark matter’.

Results and discussion

Metagenomic functional selection approach and validation

The feasibility of isolating novel functional enzymes encoded by metagenomic DNA from soil and other environmental samples has been previously demonstrated (Rondon et al., 2000; Handelsman, 2004; Leis et al., 2015; Ufarté et al., 2015). In order to capture functional RubisCOs directly from uncultivated members in microbial communities, we developed a specific functional metagenomics approach that capitalizes on previously described autotrophic CO₂-dependent growth selection using a host bacterium that expresses exogenous RubisCO genes under aerobic and anaerobic conditions (Smith and Tabita, 2003; Satagopan et al., 2009; 2014). DNA was extracted from microbial community biomass obtained from various environments, enzymatically digested to partial completion, and subsequently, high-performance liquid chromatography (HPLC) purified and size fractionated (Fig. 1A and B). The purified DNA was cloned into the dual promoter expression vector p83 (Fig. 1C), allowing the expression of cloned sequences from either orientation, and then transformed into Escherichia coli. The cloned sequences from the E. coli library were conjugated en masse into the host RubisCOdeletion strain SB I/II⁻ of Rhodobacter capsulatus (Smith and Tabita, 2003; see Experimental procedures). This system has been used previously to successfully support autotrophic photosynthetic and chemoautotrophic CO₂-dependent growth when complemented with RubisCO genes from diverse microbes, including Proteobacteria, Cyanobacteria and Archaea under aerobic or anaerobic conditions with varying CO₂/O₂ concentrations (Finn and Tabita, 2003; Smith and Tabita, 2003; Satagopan et al., 2009; 2014; Witte et al., 2010). Because the host strain cannot grow on CO₂ unless complemented with an exogenous and active RubisCO, clones encoding physiologically functional RubisCOs are automatically selected from the environmental DNA pool (Fig. 1D). Additionally, unlike Cyanobacteria and green algae that are known to have carbon concentrating mechanisms (Giordano et al., 2005), the relative ability of the complemented RubisCO to support growth in strain SB I/II⁻ is directly related to the properties of the enzyme and response to supplied external CO₂ and O₂ concentrations.

Fig. 1.

Selection approach to recover functional metagenomic RubisCO sequences. A. Metagenomic DNA was extracted from environmental samples and partially digested with Sau3AI.

B. Digested DNA was size fractionated (~2–8 Kb) and purified via HPLC using a Waters Gen-Pak FAX DNA column. Selected fractions (green boxes) were combined and concentrated.

C. HPLC purified DNA was cloned into a dual cbb operon promoter vector, p83, containing a R. rubrum cbbR and cbbM promoter (Smith and Tabita, 2003) (A) in one orientation and the R. capsulatus SB1003 cbbL promoter (Vichivanives et al., 2000) (B) in the opposite orientation, ensuring that recovered RubisCO genes are expressed regardless of their orientation. Plasmid p83 cloned DNA was transformed into E. coli to create a metagenomic library.

D. The E. coli library was conjugated directly en masse into R. capsulatus strain SB I/II⁻ and plated onto minimal medium under a CO₂/H₂ atmosphere. SB I/II⁻ colonies expressing a functional RubisCO gene on plasmid p83 supported growth using CO₂ as the sole carbon source.

The functional selection protocol was first vetted by interrogating extracted genomic DNA from five laboratorycultured organisms (Rhodobacter capsulatus SB1003 [form I and II], Rhodospirillum rubrumATCC 11170 [form II], Rhodopseudomonas palustris 010 [form I and II], Rhodobacter sphaeroides 2.4.1 [form I and II] and Ralstonia eutropha H16 [two form I]) in the presence and absence of oxygen, demonstrating the successful recovery of divergent RubisCOs (Fig. 2). From two selection experiments, eight of the nine possible form I and form II RubisCO sequences from the five organisms were recovered after cloning into the p83 dual promoter plasmid. The only RubisCO sequence not recovered was the megaplasmid-borne R. eutropha gene, which is basically identical [99% amino acid (aa) identity] in sequence to the chromosomal copy of this gene from this organism. Growth complementation occurred in strain SB I/II⁻ on plates flushed with either 5% CO₂/H₂ for anaerobic photoautotrophic (PA) growth or flushed with 2.5–5% CO₂/H₂ mixed with an equal proportion of air (~10.5% O₂) for aerobic chemolithoautotrophic (CA) growth. As expected, these experiments verified with certainty that our autotrophic selection approach successfully recovered both forms of RubisCO enzymes from a mixed pool of organisms with no false positives. Subsequently, the functional selection was employed with DNA from diverse natural microbial communities that showed promise for obtaining uncharacterized enzymes from RubisCO-harbouring organisms based on prior PCR and shot-gun metagenomic sequencing surveys (Kulp et al., 2008; Hoeft et al., 2010; Dourado-Ribeiro, 2013; Wrighton et al., 2012; submitted). These environments included a freshwater river and soil autotrophic enrichment, a photosynthetic biofilm derived from a hot spring within Mono Lake, CA, and a groundwater subsurface aquifer (see Experimental procedures for details). Selections were performed initially under PA growth conditions, allowing for fast growth and rapid analysis of any RubisCO that complements strain SB I/II⁻ for functionality (Fig. 1). Once PA growth was achieved, selected colonies were further tested for the ability to complement under CA growth conditions.

Fig. 2.

RubisCO amino-acid tree of functionally selected RubisCOs. The following colour scheme was used: RubisCO form I sequences (green), form II (blue), form II Gallionellaceae (pink), hybrid form II/III (black), form III (red), and form IV (gray). Sequences retrieved as part of this study are marked with a star filled with various colours identifying the nature of the strain: Enrichment (orange), Biofilm (gray), and Groundwater aquifer (black). RubisCO sequences employed as part of the five-organism control experiment are indicated by gray-filled circles. Enrichment RubisCO sequences noted with asterisks are partial sequences encompassing aa 130–458 according to R. rubrum numbering for the form II and aa 80–410 according to Synechococcus PCC6301 numbering for the form I sequences. The rest of the sequences are full-length RubisCO sequences. Representative full-length RubisCO sequences were obtained from the NCBI ref seq database (Table S1). The distance-based RubisCO amino acid tree was made in GENEIOUS using the Jukes-Cantor genetic distance model and the neighbour-joining tree building method. Bacillus subtilis form IV RubisCO served as the outgroup. The tree was re-sampled using bootstrapping with 100 replicates.

Environmental RubisCO selection

Using the procedure described above, the selections resulted in 10 functional RubisCO sequences from the river and soil enrichments, four from the hot spring photosynthetic biofilm and three from the groundwater aquifer (Fig. 2). Several of the 17 recovered RubisCO sequences were identical or nearly identical to each other at the level of nucleotide (nt) or the encoded aa sequence but were isolated as unique clones with varying fragment lengths or intergenic regions (Table 1); the results suggested that these RubisCOs were derived from either the same or highly similar strains of bacteria. We selected for colonies able to complement strain SB I/II⁻ under both anaerobic PA and aerobic CA growth conditions without a priori knowledge of the sequences or laborious screening of multiple clones. Functional RubisCOs of the form I type (CbbLCbbS) from the biofilm (Fig. 3A and C) and of the form II type (CbbM) from the groundwater aquifer sample (Fig. 3B and D) also complemented for RubisCO-dependent growth in liquid cultures under both PA and CA growth conditions; moreover, extracts from these cells exhibited measurable levels of RubisCO activity (Fig. 3E and F) under aerobic conditions (21% O₂). The liquid growth complementation experiments showed that these selected RubisCOs, as compared with a positive control using a Rhodosprillum rubrum RubisCO construct, did not have unusually long lag times to achieve exponential growth and all reached similar maximum optical densities (ODs) (1.0–2.0 at OD_660nm).

Table 1.

Functional RubisCO-encoded metagenomic clones recovered from three environmental samples.

Clone name	Fragment clone size (Kb)a	Form	Large subunit RubisCO isolate similarity (NCBI blastx)	No. of identical aa residues/total no.	% aa identity
Soil and river autotrophic enrichment culturesb
E1A	7	I	Rhodobacter capsulatus	315/317	99
E1B	5.5	I	Rhodobacter blasticus	318/321c,d	99
E1F	4.5	I	Hydrogenophaga pseudoflava	452/473c	96
E1I	5.5	I	Hydrogenophaga pseudoflava	315/327	96
E4A	5.5	I	Hydrogenophaga pseudoflava	316/329	96
E4B	5.6	I	Hydrogenophaga pseudoflava	305/316	97
E2A	3.5	II	Rhodobacter capsulatus	331/331	100
E3D	2.2	II	Rhodobacter capsulatus	327/327	100
E4K	3.4	II	Rhodobacter capsulatus	330/333	99
E3A	2.8	II	Rhodopseudomonas palustris	331/334	99
Mono Lake photosynthetic biofilm culture from hot springb
MA1	2.8	I	Ectothiorhodospira PHS-1 sp.	473/473c	100
MA3	2.0	I	Ectothiorhodospira PHS-1 sp.	473/473c	100
MC1	2.4	I	Ectothiorhodospira PHS-1 sp.	473/473c	100
MD1	2.5	I	Ectothiorhodospira PHS-1 sp.	473/473c	100
Groundwater aquiferb
GWS1B	3.1	II	Sulfuricella dentrificans	408/459b	89
GWS6B	4.5	II	Sulfuricella dentrificans	406/459b	88
GWS6D	6	II	Sulfuricella dentrificans	406/459b	88

^a.Enrichment cloned fragment lengths were approximated via restriction digest and agarose gel analysis. The entire cloned fragments were not sequenced.

^b.All clones were able to complement for both PA and CA growth in strain SB I/II⁻.

^c.Full-length large subunit sequence and alignment.

^d.The full-length sequence was recovered via functional selection but the alignment was with the partial sequence of R. blasticus DSM 2131 RubisCO available in the NCBI database. There is no genome sequence available yet for R. blasticus sp. (as of 8/31/2015).

Fig. 3.

Autotrophic growth phenotypes and RubisCO activities of enzymes encoded by metagenomic DNA fragments. Anaerobic PA growth (bubbled with 5% CO₂/95% H₂) of: (A) Mono Lake biofilm form I cbbLS fragments; and (B) Rifle groundwater aquifer form II cbbM fragments, each complemented in strain SB I/II⁻.

C. Aerobic CA growth (2.5% CO₂/47.5% H₂/air) of Mono Lake biofilm form I cbbLS fragments.

D. Rifle groundwater aquifer form II cbbM fragments complemented in strain SB I/II-(bubbled with 5% CO₂/45% H₂/50% air). Rhodospirillum rubrum cbbM was cloned into plasmid p83 and served as a positive control for complementation of strain SB I/II⁻. Empty vector (p83) served as a negative control (data not shown). RubisCO activities of crude protein lysates of cells harvested from liquid autotrophic cultures that reached an OD of > 1.0 at 660 nm E. strain SB I/II⁻ expressing Mono Lake biofilm form I RubisCO (encoded by metagenomic cbbLcbbS genes) and (F) strain SB I/II expressing Rifle groundwater aquifer form II (encoded by metagenomic cbbM gene). The data represent the averages of numbers obtained from two enzyme assay determinations, which both gave similar values.

The river and soil enrichment and biofilm samples identified RubisCOs most closely related to form I and form II sequences from the Alphaproteobacteria, Betaproteobacteri and Gammaproteobacteria (Fig. 2 and Table 1). The enrichment selection and subsequent DNA sequencing provided the first full-length open reading frame (ORF) encoding a form I RubisCO that is highly similar to that of Rhodobacter blasticus; which is ~97% identical to the partial R. blasticus RubisCO sequence available in the National Center for Biotechnology Information (NCBI) database (985 bp) (Uchino and Yokota, 2003). The selection also identified a RubisCO with high similarity of that from a Hydrogenophaga pseudoflava species, which has only been partially characterized based on limited in vitro assays (Kim et al., 1997); the present study represents the first physiological complementation of a RubisCO-encoding sequence from this organism. The biofilm selection provided preliminary evidence that a form I RubisCO 100% identical to that of a moderately halophilic organism, Ectothiorhodospira sp. PHS1 (Kulp et al., 2008; Hoeft et al., 2010), could wholly complement the non-halophilic host strain SB I/II⁻ in a medium with low salt (~50 mM); however, follow-up experiments would be needed to determine the salt requirements for this RubisCO.

The groundwater-aquifer metagenomic sample provided DNA that was not derived from autotrophic enrichments. Selection with this sample resulted in the isolation of three RubisCO-encoding sequences harboured on 3 Kb (GWS1B), 4.5 Kb (GWS6B) and 6 Kb (GWS6D) metagenomic fragments all able to complement for CA growth (Fig. 3); in addition, cell extracts showed activity under full aerobic conditions (21% O₂), indicating that the respective enzymes are likely oxygen tolerant. Each of these groundwater aquifer metagenomic sequences consists of a full-length cbbM (form II RubisCO) gene, cbbR transcriptional regulator gene and potential CbbR binding sites upstream of the cbbM gene (Fig. S1); GWS6B also includes ~800 nt of a partial sequence encoding a putative CbbQ RubisCO activation protein (Fig. 4). GWS6D and GWS6B are 100% identical fragments, but the GWS6D gene fragment has an additional ~2 Kb, which includes a partial sequence for the groEL gene (Fig. 4). GWS1B and GWS6B/D fragments are 94% identical across an overlapping ~3 Kb region; the 6% difference mostly resides in the intergenic regions upstream and downstream of cbbM, suggesting that GWS1B and GWS6D/B might be from two closely related bacterial strains. The translated amino-acid sequences of CbbM (form II RubisCO) (99.5%) and CbbR (98.6%) proteins are highly similar between GWS1B and GWS6D/B. The best matching NCBI BLAST hits show that the recovered form II RubisCO sequences are most similar to Sulfuricella dentrificans (~88–89% identity; Table 1), whereas the CbbR sequences are most similar to Gallionella capsiferriformans ES-2 (~84% identity).

Fig. 4.

Operon structure alignments of the recovered metagenomic fragments from the groundwater aquifer. Sequences are from the Rifle groundwater aquifer, along with closest RubisCO hits and related sequences available in the NCBI database. The acronym pcrf stands for peptide chain release factor.

Uncultivated Gallionellaceae sp. RubisCO analyses

Sampled time points near the selected groundwater aquifer site were previously subjected to deep sequencing, assembly and metagenomic reconstruction (Wrighton et al., 2012; Brown et al., 2015; Castelle et al., 2015); thus we queried the functionally selected RubisCO sequences against the binned genomic sequenced data in order to obtain the genomic context. All three recovered RubisCO fragments had best hits to metagenomic scaffolds with 83–95% nt identities across aligned regions (Table 2) and with similar gene organization (Fig. 4). Using a combination of tetranucleotide frequency (Dick et al., 2009) and data series abundance patterns (Brown et al., 2015), these RubisCO scaffolds were confidently assigned to two genomic bins (GWA2_Gallionellales_59_43, GWB2_Gallionellales_58_32). Phylogenetic analyses of concatenated ribosomal proteins extracted from these genomic bins demonstrate that these constitute distinct species within the family ‘Gallionellaceae’ in the order Betaproteobacteria (Fig. S2). The RubisCO sequences are nearly identical at the aa level (~98% identity; Table 2) and cluster together within the form II RubisCOs (Fig. 2). The family Gallionellaceae has only five sequenced genomes in NCBI, and the genomic scaffolds recovered here are derived from uncultivated lineages that lack representation in existing genomic databases.

Table 2.

Sequence comparisons of groundwater aquifer RubisCO sequences obtained from functional selection with RubisCO scaffolds obtained from deep sequencing.

	Scaffold GWB2_18588		Scaffold GWD2_39307		Scaffold GWA2_17992
	(7.1 Kb)		(2.3 Kb)		(9 Kb)
	No. of identical residues/total no.	% identity	No. of identical residues/total no.	% identity	No. of identical residues/total no.	% identity
Entire fragmenta
GWS1B 3 Kb	3054/3239	94.3	1629/1689	96.4	2229/2654c	84
GWS6B 4.5 Kb	4093/4324	94.7	2301/2563	90	2973/3571c	83
GWS6D 6 Kb	4088/4336	94.3	1881/2138	88	2533/3096c	82
ORFsb
GWS1B CbbM	458/460	99.6	442/460	96.3	455/460	99.1
GWS6B/D CbbM	458/460	99.6	440/460	96	455/460	99.1
GWS1B CbbR	301/306	98.4	n/ad	n/ad	270/306	90.6
GWS6B/D CbbR	303/306	99	n/ad	n/ad	268/306	89.9

^a.nt sequence comparisons.

^b.aa sequence comparisons; note that GWS6B and 6D are identical aa sequences for CbbM and CbbR.

^c.Excluding PilYI signal peptide and peptide chain release factor nt regions in alignment comparisons.

^d.Scaffold ends ~200 bp upstream of the cbbM gene and thus does not have a cbbR gene for comparison.

Because the uncultivated Gallionellaceae RubisCO sequences represent a poorly characterized group of form II RubisCOs, the GWS1B CbbM structural and biochemical properties were further analysed. Recombinant hexahistidine-tagged GWS1B RubisCO was purified from E. coli (Fig. S3A), and its enzymatic properties were evaluated in vitro. The specific activity was 1 (±0.1) μmol/ min/mg at pH 8.0 and 2.8 (±0.3) μmol/min/mg at pH 7.2, and the substrate specificity factor (Ω) was determined to be 12 (±0.3) (Table S2). These values are highly similar to those obtained for the well-studied form II RubisCO enzymes from Rhodopseudomonas palustris 010 (71% aa identity) and Rhodospirillum rubrum ATCC 11170 (66% aa identity) (Satagopan et al., 2014).

Non-denaturing gel electrophoresis indicated that the GWS1B enzyme had a marked propensity to oligomerize into larger molecular weight moieties, which was not observed for R. rubrum or R. palustris enzymes under similar electrophoretic conditions (Fig. S3B). The oligomeric state of GWS1B was further evaluated by size exclusion chromatography at pH 8.0 (Fig. S4) and with size-exclusion chromatography with multi-angle light scattering (SEC-MALS) at pH 7.2 (Fig. S5). Contrary to the non-denaturing gel electrophoresis results, it could be concluded from these experiments that the enzyme assembles as an unusual hexameric structure with a molecular mass of ~320 kDa, similar to the recently studied R. palustris form II enzyme (Satagopan et al., 2014).

The hexameric structure was further substantiated when the crystal structure of the enzyme was solved both in the ‘apo’ form (i.e. with no ligands) and the activated form complexed with 2-carboxyarabinitol 1,5-bisphosphate (CABP), a transition-state substrate analogue (Fig. 5 and Table S3). An alignment of the apo and liganded monomer backbone traces of the GWS1B crystal structures indicated that they are similar (Fig. 5A). Consistent with the relatively high-observed levels of sequence identity between them (> 65% aa identity), the tertiary structure of a monomeric GWS1B subunit aligns well with a subunit from either the R. palustris or the R. rubrum enzyme (Fig. 5B and C). The strong structural similarity between the CABP-bound forms of GWS1B and R. palustris CbbM is reflected in the ‘closed’ conformation for loop 6 (Fig. 5C), an important structural element in catalysis, and a virtually identical conformation for the form II-specific C-terminal domain. As with the R. palustris enzyme, the carboxy terminus points away from the loop 6 region (Fig. 5C), and the active-site geometry is virtually identical in both the CABP-bound R. palustris and GWS1B enzymes (Fig. 5D). Thus, despite apparent differences in amino acid sequence identity (~29%) and non-denaturing gel molecular weight sizes, our further biochemical analyses support that the GWS1B and R. palustris RubisCOs are similar in activity levels, specificity values and structure. Overall, the discovery of a second hexameric form II enzyme suggests that hexameric assemblies of form II enzymes may be more prevalent than originally thought.

Fig. 5.

Structural characteristics of form II RubisCO (GWS1B) from uncultivated Gallionellaceae species. A. Superposition of activated apo (PDB code 5C2C; orange) and CABP-bound (PDB code 5C2G; purple) GWS1B form II RubisCO. The N-, C-termini and loop 6 are labelled.

B. Top views down the threefold symmetry axis of activated, CABP-bound hexameric form II RubisCO from uncultivated Gallionellaceae sp. (PDB code 5C2G, alternating subunits in purple and grey) and R. palustris (PDB code 4LF1, alternating subunits in green and grey).

C. Superimposition of a catalytic large subunit from activated, CABP-bound, x-ray crystal structures of form II CbbM RubisCO from R. rubrum (PDB code, 9RUB; teal), R. palustris (PDB code, 4LF1; dark green) and GWS1B enzyme (PDB code 5C2G, purple) characterized in this study. For clarity, CABP, the transition-state substrate analogue, bound in the active site of GWS1B is shown in stick representation in black. The loop 6, N- and C-termini of GWS1B are labelled for comparing of their positions in the other two structures.

D. Stereo view of the superimposed active sites of activated, CABP-bound GWS1B (purple) and R. palustris (green) form II RubisCOs. Active-site residues and CABP are shown in stick representation in lighter shades. Numbering of active site residues is according to the GWS1B amino acid sequence. Magnesium ion is shown as a sphere and coloured pink (GWS1B) or cyan (R. palustris). Residues from both subunits in a RubisCO dimer contribute to the active site but, for clarity, only residues from one of the subunits are shown in each structure.

Conclusions

Functional metagenomics selection can be used to recover catalytically active RubisCOs from multiple microbial communities adapted to a wide range of environmental conditions. We recovered several new RubisCO variants and characterized one such enzyme in some detail, shedding valuable structure-function insights into currently undersampled microbial functional diversity. Since the bulk of microbial diversity in the environment is uncultivated and unknown, the present results and other metagenomic studies (Wrighton et al., 2012; Campbell et al., 2013; Castelle et al., 2015) suggest it is likely that much of the diversity pertaining to microbial RubisCO structure and function is still yet to be discovered. Functional selection offers a straightforward way to access and obtain enzyme variants from uncultivated bacteria that have been subjected to billions of years of evolution. This study represents an important first step in selecting specific enzymes adapted to function under highly oxic environments since most microbial RubisCO enzymes are either sequestered and protected from oxygen inhibition or possess low tolerance to oxygen due to optimally functioning under anaerobic or semi-anaerobic conditions. Selections could be refined using RubisCO-deletion host strains under specific CO₂ and O₂ regimes. The isolation of such enzymes will complement and enhance current studies for understanding RubisCO catalysis and ultimately provide advantageous catalysts for biosynthetic pathway optimization for CO₂ gas removal and conversion to useful end products.

Experimental procedures

Plasmids, strains and growth conditions

The E. coli One Shot Top10 strain (Invitrogen) was used for all cloning and transformation steps, the E. coli pRK2013 helper plasmid was used for tri-parental matings and the E. coli BL21 (DE3) strain for gene expression. Escherichia coli strains were grown on lysogeny broth (LB) broth or agar plates at 37°C supplemented with the appropriate antibiotics. Laboratory culture strains of wild-type R. capsulatus SB1003, R rubrum ATCC 11170, R. palustris 010, R. sphaeroides 2.4.1 and R. eutropha H16 were used for test DNA extractions and simultaneous recovery of functional RubisCO genes from these five organisms. These organisms were grown with shaking at 30°C under chemoheterotrophic conditions on liquid peptone yeast extract (PYE), LB broth, or on Ormerod’s defined minimal medium (Ormerod et al., 1961) supplemented with malate.

The selection host, strain SB I/II⁻, is a RubisCO deletion mutant strain of R. capsulatus SB1003, a purple non-sulfur photosynthetic organism capable of heterotrophic and autotrophic growth strategies. In this mutant strain, both endogenous form I and form II RubisCO genes have been inactivated with partial deletions and insertion of kanamycin and spectinomycin-resistant cassettes (Smith and Tabita, 2003).

Chemoheterotrophic growth of strain SB I/II⁻ was carried out at 30°C in the dark either on PYE plates or shaking in liquid PYE with 3 mM KCl and 10 mM NaCl supplemented with kanamycin (25 μg/ml) and spectinomycin (25 μg/ml) as described previously (Smith and Tabita, 2003; Satagopan et al., 2009; 2014). For PA growth complementation, Ormerod’s minimal medium plates with no organic carbon source were incubated in illuminated anaerobic sealed jars, which were periodically flushed with a gas mixture of 5% CO₂/H₂. Liquid Ormerod’s minimal medium was used for anaerobic PA cultures, which were bubbled continuously with 5% CO₂/H₂ for growth analysis and harvested for RubisCO activity assays, as described below. For aerobic CA growth complementation, sealed jars were periodically flushed with a gas mixture of 2.5% or 5% CO₂/47.5% or 45% H₂, mixed with an equal proportion of air and grown in the dark.

The cloning vector, p83 (Fig. 1C), was constructed as part of this study from a modified version of the broad-host range plasmid 3716 (courtesy of Oliver Lenz and Bärbel Friedrich). The p83 plasmid has the R. rubrum form II cbbM promoter region and the cbbR gene (Smith and Tabita, 2003) on one side of the multiple cloning site and the form I RubisCO (cbbLS) promoter from R. capsulatus SB1003 (Vichivanives et al., 2000) on the other side of the multiple cloning site (Fig. 1C). Vector p83 also harbours a tetracycline antibiotic resistance cassette.

Environmental and enrichment samples

For autotrophic enrichment samples, water and soil samples were collected from the Olentangy River and embankment on The Ohio State University campus in Columbus, OH in October and December 2011 as described (Dourado-Ribeiro, 2013). Enrichment cultures were grown similarly as described above with the following modifications. Samples were inoculated into 1 L bottles of Ormerod’s minimal medium and subjected to PA growth under an atmosphere of 20% CO₂/H₂ in the light or CA growth in the dark under an atmosphere of 10% CO₂/H₂ mixed with an equal proportion of air. When enrichment cultures reached an OD_660nm of > 1.0, the cultures were saved as frozen stocks. Enrichment cultures were resurrected from frozen stocks, inoculated into fresh medium and incubated under similar PA and CA growth conditions. These cultures were then harvested at an OD_660nm of > 1.0 for DNA extractions and combined in equal concentrations for downstream analyses.

The photosynthetic biofilm sample was collected from direct scrapings of rocks and cobble from Paoha Island hot springs located in Mono Lake, CA, an alkaline, hypersaline lake (Kulp et al., 2008; Hoeft et al., 2010) and the scrapings stored in a liquid mineral salts medium (Kulp et al., 2008) under N₂ in a sealed vial.

Groundwater aquifer samples were collected from the Rifle Integrated Field Research Challenge site in Rifle, CO as described (Wrighton et al., 2012; Brown et al., 2015; Castelle et al., 2015). Approximately 100 L of water were concentrated onto 0.2 μM Supor filters (Pall), using a 1.2 μM pre-filter, at sequential time points over a 4-month acetate-amended experiment from August to November 2011. The time point and filter used for this present study was the 0.2 μM filter from 9 September 2011, 13 days after the start of the acetate amendment when biomass was high.

Construction of metagenomic libraries

Genomic DNA extractions were conducted according to previous extraction methods (Zhou et al., 1996; Crump et al., 2003) with the following modifications. Five to 10 ml of pelleted cells or ~1.5–2.0 g of filter were re-suspended in a ‘modified DNA Extraction Buffer’ [mDEB, 100 mM Tris-Cl (pH 7.5), 50 mM Na₂EDTA (pH 8.0) and 50 mM sodium phosphate buffer (pH 7.5)]. Samples were amended with lysozyme (final concentration 2 mg/ml) and freeze/thawed (80°C/37°C) three times followed by several enzymatic lysis incubations (Zhou et al., 1996; Crump et al., 2003). Following these cell lysis steps, DNA was further extracted twice with equal volumes of phenol : chloroform : isoamyl alcohol and twice with chloroform : isoamyl alcohol. DNA was precipitated overnight at room temperature in 0.6 volumes of isopropanol and linear polyacrylamide carrier. Precipitated DNA was washed twice in 70% ethanol, after which the pellet was dried and re-suspended in 0.2 × TE buffer pH 8. Extracted DNA samples were verified for high MW composition via agarose ethidium bromide gel electrophoresis and quantified on a Nanodrop (ThermoScientific) and a Qubit flourometer (Life Technologies). Approximately 2 μg of DNA were used per Sau3AI (NEB) partial restriction digest with 1–2U of enzyme for a 7 min 37 °C incubation.

In order to remove contaminants and small DNA fragments (< 1 Kb) from the Sau3AI partial digests, between 8 μg and 20 μg of the digested DNA was purified and size fractionated on a Shimadzu HPLC using a Waters Gen-Pak FAX anion-exchange column per manufacturer’s instructions. Eluted fractions were collected and the size distribution in each of the fractions analysed via ethidium bromide or SYBR Gold agarose gel electrophoresis. Selected fractions representing the desired size range (> 1 Kb and < 8 Kb) were then combined and concentrated using an Amicon 30 kDa (Millipore) column.

The HPLC/Amicon-concentrated Sau3AI-restricted DNA was ligated into the BamHI (NEB) digested and phosphatased (NEB Antarctic phosphatase) dual promoter p83 vector. Ligations were carried out in 10–40 μl volumes using T4 DNA ligase and buffer (NEB) and incubated overnight at 16°C. Typically, insert::vector ratios of 2:1 were used (assuming an average insert size of 3 Kb). Ligations were desalted using a DNA clean and concentrator kit (Zymo).

Functional growth selection

Desalted plasmid p83 ligated DNA was transformed into electrocompetent Top10 (Invitrogen) E. coli cells using an Eppendorf Electroporator 2510 (Eppendorf). Transformations were incubated for 1 h at 37°C shaking in super optimal broth with catabolite repression (SOC) medium without antibiotics prior to a 4–6 h tetracycline (tet) (12.5 μg/ml) enrichment; this ensures enrichment of E. coli cells with transformed p83 tet resistant plasmid prior to mating with strain SB I/II⁻. A small aliquot of every transformation and post-enrichment transformation mixtures were plated onto LB-tet plates (12.5 μg/ml) to verify transformation efficiencies and tet enrichment efficiency (Table S4).

Following the tet enrichment and for each experiment, the ~100 000 electrocompetent E. coli member library clone pool (Table S4) was conjugated in a tri-parental mating, as previously described (Smith and Tabita, 2003), with strain SB I/II⁻ and the E. coli strain carrying the pRK2013-helper plasmid kept overnight at 30°C on PYE plates. The following day, matings were re-suspended and washed in Ormerod’s minimal medium, and a small aliquot was plated onto PYE supplemented with tetracycline [2 μg/ml] to verify that the mating was successful; all SB I/II⁻ cells that received the tet plasmid grew under these conditions. The rest of the re-suspended mating was plated onto Ormerod’s minimal medium plates and placed into sealed jars for either photoautotrophic or chemoauototrophic growth selection. Only colonies with a cloned fragment containing a functional RubisCO gene allowed for growth under these selection conditions. For all experiments, the p83::R. rubrum cbbM construct served as a positive control, and the p83 empty vector served as a negative control. Colonies that arose on plates were re-streaked to gain enough material to re-isolate the plasmid for DNA sequencing and to obtain individual colonies for starting liquid autotrophic growth.

Bioinformatic analyses

Plasmid encoding functional RubisCOs were sequenced using primer walking off the ends of the dual promoter vector into the fragment inserts with Sanger sequencing at The Ohio State Plant-Microbe Genomics Facility and analysed using NCBI blast for homology, BioEdit for ClustalW multiple sequence alignments and SnapGene Viewer 2.7.1 or Geneious version 7 for manual ORF annotations. For some recombinant clones recovered from the enrichment samples, only the partial RubisCO gene sequences were obtained using either a conserved form I forward primer (Alfreider et al., 2003) or form II forward primer (Kato et al., 2012).

For the Rifle groundwater aquifer sample selection, recovered RubisCO fragments were compared with assembled genomic scaffolds; information regarding the Illumina sequencing was described previously (Brown et al., 2015). Similar scaffolds were retrieved via in-house blastn or blastp searches using the recovered RubisCO fragment as the query and subsequently analysed in Geneious for alignments and manual ORF annotations. The RubisCO amino acid tree was made in Geneious using the Jukes–Cantor genetic distance model and the neighbour-joining tree building method. Bacillus subtilis form IV RubisCO served as the outgroup. The tree was re-sampled using bootstrapping with 100 replicates.

Phylogenetic analysis of the Gallionellaceae bins (GWA2_Gallionellales_59_43, GWB2_Gallionellales_58_32) was performed using a syntenic block of 16 universal ribosomal proteins (RP) (L2, L3, L4, L5, L6, L14, L15, L16, L18, L22, L24 and S3, S8, S10, S17, S19). Each ribosomal protein (amino acid) was aligned along with 44 reference sequences using MUSCLE (Edgar, 2004) with default parameters. Alignments were manually curated to trim start and end gaps and remove ambiguously aligned regions. Individual ribosomal protein alignments were concatenated in GENEIOUS version 7 (Kearse et al., 2012). In total, the alignment of 46 sequences spanned 13 087 columns. Phylogenetic analysis of RP was inferred by raxml (Stamatakis, 2014) using the protgammagtr algorithm with a total of 100 bootstraps. raxml was called as follows:raxmlHPC-PTHREADS -f a -s input -n result -m PROTGAMMAGTR -x 777 -# 100 -p 333. Brevundimonas subvibrioides ATCC 15264 (Alphaproteobacteria) (Genbank: CP002102) was included as the root for the tree. The entire ribosomal protein alignment used in phylogenetic analysis is provided in fasta format as Supplementary Data File S1.

RubisCO gene expression and purification of recombinant proteins

Hexa-histidine purified proteins were obtained as previously described (Satagopan et al., 2009; 2014). Briefly, genes were expressed via IPTG induction using pET-28a constructs, with the induction initiated at an OD_600nm between 0.5 and 0.7. After overnight growth at 25°C post-induction, the cells were rinsed and re-suspended in 50 mM bicine-NaOH, 10 mM MgCl₂, at pH 8.0, supplemented with 2 mM NaHCO₃, 1 mM DTT and 15 mM imidazole and lysed using a French-pressure cell at 1000 psi. The proteins were clarified using high-speed centrifugation and purified by one-step nickle-affinity gravity-flow chromatography (Ni-NTA Agarose, Qiagen). The GWS1B enzyme was further purified and size fractionated on a Superose 6 (GE Healthcare) gel filtration or a Q sepharose HP ion exchange column. Eluted fractions with peak 280 nm absorbance were pooled, concentrated and tested for RubisCO activity. All assay determinations are the average of two different enzyme preparations done in technical duplicates or triplicates. The R. palustris CbbM, R. rubrum CbbM and R. eutropha CbbLS purified recombinant hexahistidine purified enzymes served as positive controls for downstream activity, specificity and gel filtration assays. Assays were run in technical duplicates or triplicates as necessary.

SEC-MALS

A 1 mg sample (100 μL sample at 10 mg/ml) of the GWS1B RubisCO protein was analysed by size-exclusion chromatography with inline static light-scattering and refractive index detectors. The protein sample was injected onto a Superdex 200 10/300 GL analytical size-exclusion column (GE Healthcare) equilibrated in 20 mM Tris-Cl, pH 7.2, 300 mM NaCl, 10% glycerol, 10 mM MgCl₂, 20 mM NaHCO₃ using an AKTA purifier (GE Healthcare) and a flow rate of 0.3 ml/min. Static light scattering was measured with a miniDAWN TREOS detector, and refractive index was measured using an Optilab T-rEX detector (both from Wyatt Technology, Santa Barbara, CA). The eluted protein peak was analysed using astra software (Wyatt Technology) to determine molecular mass.

Crystallization and structure determination of GWS1B RubisCO

For crystallization screening, hexahistidine-tagged protein was purified as described above with the exception that the size exclusion chromatography step used a HiLoad 16/600 Superdex 200 column (GE Healthcare) equilibrated in crystallization buffer (20 mM Tris, pH 8.0, 300 mM NaCl, 10% glycerol, 10 mM MgCl₂, 20 mM NaHCO₃). The protein was concentrated to ~21 mg/ml and subjected to crystallization screening at 18°C using the hanging-drop vapour diffusion method and a variety of commercially available crystallization screens. Full details of the crystallization and structural determination methodology of the GWS1B RubisCO can be found in the Supporting Information.

RubisCO enzyme assays

RubisCO activity and substrate specificity assays were performed as previously described (Satagopan et al., 2009; 2014). For specific activities, cell pellets or purified proteins were re-suspended in 50 mM bicine-NaOH, 10 mM MgCl₂, at pH 8.0 or 50 mM Tris-Cl, 10 mM MgCl₂, 1 mM EDTA, at pH 7.2, both supplemented with 2 mM NaHCO₃ and 1 mM DTT. Crude protein lysates were obtained via sonication. Assay reaction mixtures contained 50 mM NaHCO₃, 2 μCi NaH¹⁴CO_3, 10 mM MgCl₂ and 0.8 mM RuBP in 50 mM bicine-NaOH buffer, pH 8.0. Reactions were initiated with the addition of RuBP at 30°C in 5 min. time course reactions terminated with acid. Assays with enzyme but without the addition of RuBP served as negative controls. Assays were run in either duplicate or triplicate; exceptions are noted. For substrate specificity assays (Ω), ~50 μg of purified proteins were used under saturating O₂ concentrations (1170 μM) and in the presence of 400 μM [1-³H] RuBP. Product areas were fractionated and integrated via a MonoQ ion exchange column (GE Healthcare) using HPLC (Shimadzu). Peaks corresponding to 3-PGA and 2-PG were identified using β-ram detection (Lab Logic), and the peak areas were obtained after integrating the peaks using a radiochromatography data collection and analysis LAURA software (Lab Logic). These integrated peak areas were used to calculate substrate specificity factor values as described previously (Smith and Tabita, 2003; Satagopan et al., 2009; 2014).

Supplementary Material

SI sequences

Data File S1. Full concatenated ribosomal protein alignment in fasta format used to generate Figure S2.

SI figures

Fig. S1. GWS1B CbbR potential transcriptional binding sites. Potential CbbR binding sites were determined using the standard consensus sequence T-N₁₁-A (Dubbs and Tabita, 1998) as indicated by the gray boxes.

Fig. S2. Phylogenetic tree of 16 concatenated ribosomal proteins inferred by maximum likelihood. Sequences from Rifle are highlighted in bold, and the family Gallionellaceae in the Betaproteobacteria is highlighted in the blue. Bootstrap support values of > 50% are reported. The entire ribosomal protein alignment used in phylogenetic analysis is provided in fasta format as Supplementary Data File S1

Fig. S3. Denaturing and native PAGE of purified hexahistidine tagged recombinant RubisCO proteins. A. denaturing and B. native gel (pH 8.8) images. Lane 1: Ralstonia eutropha form I, Lane 2: Rhodopseudomonas palustris form II, Lane 3: GWS1B form II and Lane 4: Rhodospirillum rubrum form II.

Fig. S4. Gel filtration determination of GWS1B RubisCO molecular mass. A. gel filtration standards (BioRad) with molecular weights in kDa indicated above each peak; B. Ralstonia eutropha form I (green arrow); C. Rhodopseudomonas palustris form II (maroon arrow); D. GWS1B form II (black arrow); and E. Rhodospirillum rubrum form II (teal arrow). The green trace indicates 280 nm absorbance, and the purple trace indicates 260 nm absorbance. One hundred microlitres of approximately 1 mg of hexahistidine tagged purified protein were loaded onto the column for each run. The GWS1B RubisCO peak elution was similar to that of the hexameric form II protein from R. palustris.

Fig. S5. Size-exclusion chromatography with multi-angle light scattering determination of GWS1B RubisCO molecular mass. Refractive index (black trace) and molar mass (red line) are plotted versus time for purified GWS1B RubisCO run on an analytical Superdex S-200 10/300 GL column. An approximate molecular mass of 325 kDa was obtained indicating that form II GWS1B RubisCO is a hexamer. Inset, SDS-PAGE of purified GWS1B RubisCO used for molecular mass determination.

Table S1. Reference organism RubisCO sequences and accession numbers used for constructing the RubisCO large subunit amino acid tree.

Table S2. Specific activities and specificity factors of purified RubisCO enzymes.

Table S3. Data collection and refinement statistics for apo and CABP-bound GWS1B form II RubisCO structures (numbers in parentheses refer to the high-resolution data shell).

Table S4. Cloning and transformation efficiencies for RubisCO selection.

Data accessibility

RubisCO sequences obtained from this study were deposited in NCBI with accession numbers (KT749899 – KT749915). Protein Data Bank Accession numbers: 5C2C (GWS1B apo) and 5C2G (GWS1B CABP bound). Genomes are also available through ggKbase: GWA2 _Gallionellales_59_43 (http://ggkbase.berkeley.edu/organisms/1596), GWB2_Gallionellales_58_32 (http://ggkbase.berkeley.edu/organisms/2084). ggKbase is a ‘live’ site, thus genomes may be improved after publication.