The bacterial strains, vectors, and constructs used in this study are summarized in Table 1. Escherichia coli cultures were routinely grown on lysogeny broth (LB) medium (Bertani, 1951). For cloning procedures cultures were incubated at 37°C. If cultivated for measuring recombinant RubisCO activities, the growth temperature was lowered to 28°C, while cultures grown as part of over expression experiments were incubated at 17°C or 22°C. If required, the following supplements were added: ampicillin, 100 μg ml^-1; 5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside (X-gal), 50 μg ml^-1, chloramphenicol, 12.5 μg ml^-1; isopropyl β-D-1-thiogalactopyranoside (IPTG), 100 μg ml^-1 (cloning) or 0.1–1 mM (expression); kanamycin, 50 μg ml^-1, and tetracycline, 10 μg ml^-1.

Two double transposon mutant libraries were constructed from two versions of the 13 kb metagenomic fragment consisting of the RubisCO gene cluster (cbbO-mQ-mM lysR2 lysR1 cbbLSQ-1O-1; accession: KJ639815.1) using the EZ-Tn5^TM <TET-1> Tnp Transposome^TM Kit (epicentre^®, Madison, WI, United States) according to manufacturer’s instructions, with chemically competent Epi300^TM – T1^R (epicentre^®) as the host. One library was constructed with transposon clone 22II, where the cbbM structural gene was deleted (ΔcbbM). The second library was constructed using transposon clone 24II, where the cbbL structural gene was impaired (ΔcbbL). Clones containing fosmids with <TET-1> insertions were selected on LB agar plates using the following antibiotic additions: (i) chloramphenicol (12.5 μg ml^-1) for selecting the fosmid vector, (ii) kanamycin (50 μg ml^-1) to verify the presence of the first insertion, i.e., ΔcbbM or ΔcbbL, and (iii) tetracycline (100 μg ml^-1) to verify the insertion of the second transposon element. Fosmids of double transposon clones were isolated from autoinduced cultures (for detailed information on autoinduction procedure see the manual for the CopyControl^TM Fosmid Library Production Kit, epicentre^®) using the High-Speed Plasmid Mini Kit (Geneaid, New Taipei City, Taiwan) according to manufacturer’s instruction. Isolated fosmids were sequenced starting from the <TET-1> insertion using the TET-1 FP-1 forward and TET-1 RP-1 reverse primers (see manual of the EZ-Tn5^TM <TET-1> Insertion Kit, epicentre^®) to identify the exact insertion position. Selected clones were tested for their RubisCO activities.

For RubisCO activity measurements double transposon clones were cultivated at 28°C on 200 ml pre-heated LB medium supplemented with chloramphenicol (12.5 μg ml^-1), kanamycin (50 μg ml^-1), tetracycline (10 μg ml^-1), and autoinduction solution [1x final concentration (epicentre^®)] in 1 l flasks with shaking (130 rpm) and harvested after 18 h by centrifugation (9,800 × g, 10 min, and 4°C). Subsequently crude extracts were prepared. For this purpose, cell pellets were washed twice with buffer A [100 mM Tris-HCl (pH 7.8), 10 mM MgCl₂, 1 mM EDTA, 25 mM NaHCO₃ and 1 mM DTT] before resuspension in 2 ml of the same buffer. Cells were disrupted by the French pressure cell press method, followed by centrifugation (19,580 × g, 20 min, and 4°C), as described before (Böhnke and Perner, 2015). The generated crude extracts were finally used as template to perform the RubisCO activity assay, where the concentrations of the reactant (RuBP) and the product (3-PGA) of RubisCO reaction were quantified over time using High-Performance Liquid Chromatography (HPLC) (Böhnke and Perner, 2015). At least two biological replicates and three technical replicates were used for the RubisCO activity assay. Mean values of technical replicates were used to calculate the overall mean. Errors of RubisCO activity measurements were calculated with the Gaussian propagation of error. Standard derivations of technical replicates were propagated forward and are thus entered into the equation. Significant differences were calculated using an unpaired t-test with equal variance and two-tailed distribution. For each performed HPLC run different controls were tested additionally to the measured samples. The crude extract of the metagenome derived fosmid clone 71C2 containing the RubisCO gene cluster (cbbO-mQ-mM lysR2 lysR1 cbbLSQ-1O-1) and 22.2 kb flanking DNA serves as positive control and the crude extract of an E. coli fosmid clone without RubisCO genes encoded on its fosmid insert serves as a negative control. A protein free reference sample with 5 mM RuBP and 5 mM 3-PGA dissolved in buffer A were furthermore applied through the assay and used (i) for sample peak assignment and (ii) to gather the non-enzymatic degradation of educts and products. The latter was used to calculate the pseudo-activity which is subtracted from each sample activity.

Clones were cultivated in 100 ml flasks on 20 ml LB media supplemented with autoinduction solution [1x final concentration (epicentre^®)] and the following antibiotics: chloramphenicol (12.5 μg ml^-1) for the fosmid subclone 71C2II, chloramphenicol (12.5 μg ml^-1) and kanamycin (50 μg ml^-1) for transposon clones 22II (ΔcbbM), 24II (ΔcbbL), 6II (ΔlysR1), and 149II (ΔlysR2), and chloramphenicol (12.5 μg ml^-1), kanamycin (50 μg ml^-1), and tetracycline (10 μg ml^-1) for the double transposon clones 22II2B2 (ΔcbbM ΔlysR1), 22II3A3 (ΔcbbM ΔlysR2), 24II1H1 (ΔcbbL ΔlysR1), and 24II1H7 (ΔcbbL ΔlysR2). Cultures were allowed to grow until an optical density (λ = 600 nm) between 2.0 and 3.0 was reached [for clone 6II (ΔlysR1) 24 h, all other clones 16 h]. Total RNA was isolated with the UltraClean^® Microbial RNA Isolation Kit (MO BIO Laboratories, Inc., Carlsbad, CA, United States) according to manufacturer’s instructions with the exception that only 1 ml cell culture was harvested instead of the recommended 2 ml. Subsequently, genomic DNA was removed by using the RTS DNase^TM Kit (MO BIO Laboratories, Inc.) following the provided protocol, but with the modification that after half an hour an additional microliter RTS DNase was added. The reaction was incubated at 37°C for further 30 min followed by RTS DNA removal using 10 μl instead of 5 μl RTS DNase removal resins. One thousand two hundred microgram isolated RNA was used to synthesize cDNA with Invitrogen’s SuperScript^® VILO^TM cDNA Synthesis Kit (Life Technologies^TM, Darmstadt, Germany), according to manufacturer’s instructions. The generated cDNA was used to examine the fold change of RubisCO form I (cbbL) and form II (cbbM) structural genes during expression in above mentioned transposon and double transposon clones relative to the intact version 71C2II. The expression data were normalized to the transcripts of three different genes, namely (i) the chloramphenicol-acetyltransferase (cat) gene, which is encoded on the fosmid vector and reflects its copy number, (ii) the RNA polymerase sigma factor rpoD, which is a housekeeping gene, and (iii) the 16S rRNA encoding gene. For this purpose, cDNA was diluted 1–10 and the cDNA that was derived from transcripts was used as a template for the amplification of cbbL and cbbM genes as well as the three different housekeeping genes. The SYBR^® Select Master Mix, CFX (Applied Biosystems^® by Life Technologies^TM) and the following primer pairs were used: for (i) cbbL – cbbL_810F and cbbL_1115R, for (ii) cbbM – cbbM_647F and cbbM_976R, for (iii) cat – ChlR_821F and ChlR_1104R, for (iv) rpoD – rpoD_416F and rpoD_720R, and for (v) the 16S rRNA gene – 16S_280F and 16S_564R (for details on primer characteristics see Table 2). The qRT-PCR on the MJ Mini^TM Gradient Thermal Cycler (Bio-Rad, Hercules, CA, United States) was performed under the following conditions: 95°C for 2 min followed by 40 cycles of 98°C for 15 s, 51°C for 20 s, and 72°C for 30 s. Each run contains, next to the samples, various controls like (i) the non-template controls, (ii) the no reverse transcriptase control as well as (iii) an inter run calibrator to ensure comparability between different runs, i.e., one reaction from the previous plate was repeated on the new plate. At least two biological and three technical replicates were measured and used to calculate fold changes (2^-ΔΔCt). Technical replicates were arithmetically averaged and resulting mean values were used to calculate an overall mean. Errors were calculated with the Gaussian propagation of error. Standard derivations of technical replicates were entered into the equation and thus propagated forward. Significant differences were calculated from log transformed values using an unpaired t-test with equal variance and two-tailed distribution.

Polar effects were investigated to determine whether transposon insertions have an impact on transcript abundances of genes located downstream of an insertion site. Therefore, transcript abundances of genes located downstream of cbbM and cbbL were measured for ΔcbbM (22II) and for ΔcbbL (24II), and compared with the transcript abundances in the intact version 71C2II. Investigated genes were (i) cbbO-m and (ii) cbbQ-m in ΔcbbM and (iii) cbbQ-1 and (iv) cbbO-1 in ΔcbbL. The cDNA used as template was the same as that isolated before for qRT-PCR of cbbL and cbbM. The qRT-PCR conditions were the same as mentioned above for the amplification of cbbL and cbbM, but with different primers: (i) cbbO-m – cbbO-m_700F and cbbO-m_922R, for (ii) cbbQ-m – cbbQ-m_2828F and cbbQ-m_2974R, for (iii) cbbQ-1 – cbbQ-1_9714F and cbbQ-1_9912R, and for (iv) cbbO-1 – cbbO-1_11263F and cbbO-1_12525R (for details on primer characteristics see Table 2). Three biological and three technical replicates were measured and used to calculate fold changes (2^-ΔΔCt). Statistics were calculated in the same way as has been described for qRT-PCR data of cbbL and cbbM.

The four genes encoding CbbL, CbbM, LysR1, and LysR2 were cloned in the expression vector pet21a (Novagen/Merck, Darmstadt, Germany). The coding regions of the targeted genes were amplified from the fosmid DNA of the metagenome derived clone 71C2, whereby restriction sites for NheI and BamHI were inserted using following primer pairs: (i) for cbbL – CbbLF_NheI and CbbLR_BamHI, (ii) for cbbM – cbbMF_NheI and cbbMR_BamH1, (iii) for lysR1 – LysR1F_NheI and LysR1R_BamHI, and (iv) for lysR2 – LysR2F_NheI 5′- and LysR2R_BamHI (primer sequences, annealing temperatures and product length are listed in Table 2). Amplification was done with the Pfu DNA Polymerase (Thermo Fisher Scientific, Waltham, MA, United States), following manufacturer’s instructions. The amplified fragments were ligated in the pet21a expression vector (Novagen/Merck) using the previously inserted restriction sites (NheI and BamHI). This vector has a His-tag coding sequence for the C-terminus of the cloned protein. The constructed plasmids (i) pet21a::cbbL, (ii) pet21a::cbbM, (iii) pet21a::lysR1, and (iv) pet21a::lysR2 were transformed into E. coli Rosetta-gami2 host strains. Verified clones were cultured at 17°C (CbbM) or at 22°C (CbbL, LysR1, and LysR2) in 200 ml LB supplemented with ampicillin (100 μg ml^-1), tetracycline (10 μg ml^-1), and chloramphenicol (12.5 μg ml^-1) to an optical density (λ = 600 nm) of 0.7–0.8. IPTG was added to a final concentration of 0.1 mM for CbbL, 1 mM for CbbM, 1 mM for LysR1, and 0.1 mM for LysR2. The cultures were then grown over night at 17°C (CbbM) or at 22°C (CbbL, LysR1, and LysR2). Cells were harvested by centrifugation (7,600 × g, 8 min, and 8°C) and washed twice with 1x PBS buffer (137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, and 2 mM KH₂PO₄, pH = 7.4). Cell pellets were stored at -20°C until proceeding with His-tag purification using Ni-NTA agarose (Qiagen, Hilden, Germany) as described in protocol 14 of TheQiaexpressionist (Qiagen, 2003), with some modifications: Initially cell pellets were resuspended in 2 ml lysis buffer (containing 10 mM imidazole). Each lysate was passed through the French press in order to disrupt cells. Cellular debris were removed by centrifugation (19,580 × g, 20 min, and 4°C) and supernatant volumes were brought to 20 ml with lysis buffer. Afterward 2 ml Ni-NTA agarose was added to the diluted lysates, which were incubated for 2 h on ice (shaking). After pelleting (1,000 × g, 30 s, and 8°C) Ni-NTA resins were washed twice with washing buffer (containing 20 mM imidazole). The protein was eluted from the column with five volumes of elution buffer (containing 250 mM imidazole). The concentration of the total purified protein was measured by performing the Bradford assay as described previously (Bradford and Williams, 1976) using bovine serum albumin as a standard. The proteins were further analyzed by SDS polyacrylamide gel electrophoresis using 12% (w/v) gels and Western-immunoblotting using 6-His-specific antibodies (see Supplementary Figure 1).

In preparation for the mobility shift assay four non-coding regions located within the metagenome derived RubisCO gene cluster were Cy3-labeled, namely the non-coding regions between: (i) cbbQ-m and cbbM, (ii) cbbM and lysR2, (iii) lysR1 and cbbL as well as (iv) cbbS and cbbQ-1. Labeling was done during amplification using Cy3-labeled dCTP’s (1 mM, GE Healthcare, Little Chalfont, United Kingdom), a mixture of dATP, dTTP, and dGTP (2 mM), the Phusion DNA Polymerase (Thermo Fisher Scientific) and the following primer pairs: (i) ncrQm-M_3427F and ncrQm-M_3797R for the non-coding region between cbbQ-m and cbbM, (ii) ncrM-R2_4993F and ncrM-R2_5361R for the non-coding region between cbbM and lysR2, (iii) ncrR1-L_7191F and ncrR1-L_7465R for the non-coding region between lysR1 and cbbL, and (iv) ncrS-Q1_9151F and ncrS-Q1_9531R for the non-coding region between cbbS and cbbQ-1 (see Table 2 for primer sequences). PCR conditions were: Denaturation at 98°C for 10 s, primer annealing for 30 s at appropriated annealing temperatures (see Table 2), and elongation at 72°C for 12 s (32 cycles).

The mobility shift assay was based on a previously published protocol (Charoenpanich et al., 2013) but with modifications: Purified His₆-CbbL, His₆-CbbM, His₆-LysR1, and His₆-LysR2 were investigated for the ability to bind at the four non-coding regions amplified from the metagenome derived RubisCO gene cluster. For this purpose, proteins were tested (i) individually but also (ii) pairwise in combination with each other. The protein concentrations used for approaches with individual proteins ranged from 0 to 1,000 ng per 30 μl reaction mixture. For approaches with two different proteins up to 2,000 ng total protein per 30 μl were used in one reaction, which corresponds to a maximum of 1,000 ng of each protein and thus ensures the comparability with the single protein approaches. Regardless of whether one or two proteins were used for the assay, protein(s) was/were firstly incubated with a total of 200 ng salmon sperm DNA for 5 min at room temperature in binding buffer (50 mM Tris-HCl, 250 mM KCl, pH = 8.5) to prevent unspecific DNA shifts. After this Cy3-labeled DNA fragments were added (200 ng per 30 μl reaction) and reaction mixture was incubated at room temperature in the dark for further 20 min. Subsequently 5 μl loading dye (20% TBE buffer and 80% glycerol) was added and samples were loaded on a 5% TBE-polyacrylamide gel. Following electrophoresis at 50 mV for 3 h in cooled TBE-buffer, gels were visualized on a VersaDoc^TM MP4000 (Bio-Rad) at 550 nm and an exposure time of 300 ms.

Distinct regulatory features were predicted for the DNA of the metagenome derived RubisCO gene cluster using different online tools. Promoter regions were predicted for all non-coding regions with the SoftBerry program BProm (Solovyev and Salamov, 2011). We also searched for inverted repeats, which are putatively able to fold into stem-loop structures using Emboss Palindrome (Rice et al., 2000), with a minimum length for repeats of 8 nt and a maximum gap between repeated elements of 100 nt. With respect to the formation of stem-loop structures, inverted repeats with loops less than three bases were not taken into account, because they are thought to be sterically impossible and thus are believed not to be formed (Bon and Orland, 2011). LysR binding sites were identified manually by searching for the typical LysR binding motif TnA-n7/8-AnT which has been identified in other RubisCO harboring organisms before (van Keulen et al., 2003; Maddocks and Oyston, 2008 and references therein).

cbbL and cbbM transcription abundances were tested for eight mutants and normalized to three different reference genes (cat, rpoD, and 16S rRNA) (Figure 1). Generally, transcript levels of cbbL in ΔcbbM and of cbbM in ΔcbbL remained unchanged relative to the undeleted metagenomic fragment (71C2II). cbbL gene transcription was only significantly downregulated in ΔlysR1 if cbbM was expressed, since no changes in cbbL transcript levels were observed in ΔlysR1 ΔcbbM. In contrast, cbbM gene transcription was significantly upregulated in ΔcbbL ΔlysR1 double transposon clone 24II1H1 (3-fold) and in ΔcbbL ΔlysR2 double transposon clone 24II1H7 (15-fold).

All RubisCO activities of clones from the double mutant libraries where either ΔcbbM (22II) or ΔcbbL (24II) was used for the construction of the double mutants can be viewed in Figures 2A,B, respectively. Total RubisCO activity of the undeleted 13 kb fragment (71C2II) increased considerably in ΔcbbM (22II) (5-fold). Additional deletions in lysR1, lysR2, cbbQ-m, and cbbO-m resulted in a significant decrease of RubisCO activity.

Total RubisCO activity of the undeleted 13 kb fragment (71C2II) was significantly reduced in ΔcbbL (24II). When additionally deleting lysR2 (24II1H7), the RubisCO activity increased (3.5-fold), restoring the original activity of clone 71C2II. In four of the tested double mutant ΔcbbL clones the RubisCO activity was considerably reduced. These were clone 24II1G2, where parts of the intergenic region of cbbM and lysR2 were deleted, clone 24II2G1, where parts of the non-coding region between cbbS and cbbQ-1 were deleted, and the two ΔcbbL ΔcbbQ-1 clones (24II6H6 and 24II5G11). As expected, no RubisCO activity was measured for ΔcbbL ΔcbbM.

We searched the intergenic regions of our metagenomic fragment for structures which encode putative promoter regions that provide potential LysR binding sites or that may form stem-loops (Figure 3). We found 6 putative promoter binding sites, 15 putative LysR binding sites, and 18 putative stem-loop forming structures (for exact positions on the metagenomic fragment see Supplementary Figure 2).

We also performed mobility shift assays to test whether LysR1, LysR2, CbbL, and CbbM or a combination of these proteins bind to the non-coding regions cbbQ-m and cbbM, cbbM and lysR2, lysR1 and cbbL or cbbS and cbbQ-1 (Figure 4). LysR1 binds to all tested non-coding regions (Figure 4A). Its binding ability is enhanced for the intergenic region lysR1 and cbbL if CbbL or LysR2 are additionally present (Figure 4B). LysR2 alone appears to only bind to two non-coding regions: between cbbQ-m and cbbM and between cbbM and lysR2 (Figure 4A). CbbM addition intensifies the binding ability to the cbbM and lysR2 intergenic region (Figure 4B). The presence of CbbM also enables LysR2 to bind to two further non-coding regions, namely lysR1 and cbbL as well as cbbS and cbbQ-1 (Figure 4B). Other protein combinations likely reflect binding of one of the proteins alone.

In the intergenic region of lysR1 and cbbL two promoters were predicted: (i) one could be for cbbL transcription (with the -10 box ‘AGGAATCAT’ at position 7,271 bp and the -35 box ‘TTGATA’ at position 7,250 bp) and (ii) the other for lysR1/lysR2 transcription with the -10 box at position 7,275 bp ‘ATCATATAC’ and with the -35 box at position 7,302 bp ‘TAACAA’ (Supplementary Figure 2). This is in line with previous predicted functions for the non-coding region between lysR and cbbL in other organisms, where promoters for both directions were identified (Kusano and Sugawara, 1993; Wei et al., 2004). Additionally, three and two putative LTTR binding sites upstream of the putative cbbL and lysR1/lysR2 promoters, respectively, were recognized (Figure 3). These sites may be involved in LysR1 and/or LysR2 regulated cbbL transcription as well as autoregulation of their own transcription, as has been commonly demonstrated for enzymes of the LysR family (Schell, 1993; Maddocks and Oyston, 2008). The mobility shift assay verified that binding sites are located in this non-coding region (Figure 4). Here, DNA binding of LysR1 is intensified by the presence of LysR2 or CbbL (Figure 4B). LysR2 is also capable of binding to this region, but only when CbbM proteins are available. Promiscuous heterotypic interactions between different LTTRs in E. coli have been shown before, but the relevance of such cross-interactions remains unknown (Knapp and Hu, 2010). However, since our experiment showed that LysR1’s DNA binding ability is increased by LysR2, one may conclude that LysR1 and LysR2 are also able to cross-interact and form heteromultimers with its non-cognate partner. The heteromultimer (LysR1+LysR2) may cause different regulatory effects relative to the homomultimers (LysR1+LysR1 or LysR2+LysR2). The role that CbbL and CbbM play for intensified LysR binding currently remains unclear but may be related to DNA or RNA stability. Mobility shift assays with RNA and the large RubisCO subunit of Chlamydomonas reinhardtii demonstrated CbbL’s ability to bind to RNA in a sequence-independent manner under certain conditions (Yosef et al., 2004).

Although cbbL transcription levels remained unchanged in ΔcbbM ΔlysR1 and ΔcbbM ΔlysR2 (Figure 1), these clones exhibited reduced CbbL activity (Figure 2A). This discrepancy may be explained if LysR proteins also act on the transcription of genes encoding proteins, which influence CbbL activity. Likely candidates encoded on this metagenomic fragment are CbbO-m, CbbQ-m, CbbQ-1, and CbbO-1, previously shown to be involved in post-translational activation of RubisCO enzymes (Tsai et al., 2015). Indeed LysR1 proteins are demonstrated to bind upstream of cbbQ-m/cbbO-m and cbbQ-1/cbbO-1 regions (Figure 4A). LysR2 can also bind upstream of cbbQ-m/cbbO-m and if CbbM is present can bind upstream of cbbQ-1/cbbO-1, too (Figures 4A,B, respectively). While the deletion of cbbQ-m and cbbO-m in ΔcbbM illustrated a RubisCO activity loss, CbbQ-1 and CbbO-1 did not have an effect on CbbL activity (Figure 2A). Based on the available information, we posit that CbbQ-m/CbbO-m activates CbbL and that LysR1 and LysR2 proteins promote CbbQ-m/CbbO-m expression (compare model in Figure 5).

Intriguingly, CbbM also appears to play a role for cbbL expression, which has not been observed in any other study before. The deletion of cbbM leads to a 5-fold RubisCO activity increase (Figure 2A) indicative of CbbM’s repressive nature for cbbL expression. However, in ΔcbbM cbbL transcript levels are not elevated (Figure 1). Possible scenarios include that CbbM scavenges post-translational activators (CbbQ-m/CbbO-m), which are then not available for CbbL activation or that CbbM is involved in repressing CbbQ-m/CbbO-m, which may be needed for CbbL activation (Figure 5). For information on putative polar effects for cbbQ-m/cbbO-m transcript abundance caused by transposon insertion in the upstream neighboring cbbM gene see Supplementary Results and Discussion and Supplementary Figure 3.

A putative promoter was identified upstream of cbbM (Figure 3), confirming recent results indicative of cbbM being transcribed alone in this fragment (Böhnke and Perner, 2015). In the intergenic region of cbbM and lysR2 three putative LysR binding sites were predicted, two of which partially overlap (‘LysR bs1_cbbM,’ ‘LysR bs2/3_cbbM,’ see Supplementary Figure 2). In Xanthobacter flavus the same arrangement of a single LysR binding site followed by two overlapping binding sites, between a RubisCO structural gene and a LysR transcriptional regulator was also identified and all three sites were evidenced to be functional (van Keulen et al., 2003). In our fragment LysR binding sites upstream of cbbM exist, which allow binding of LysR1 and LysR2 proteins (Figure 4A). LysR2 binding was even intensified if CbbM was present (compare Figure 4B). Upregulation of the cbbM transcript (Figure 1) and increasing RubisCO activity in the double mutant ΔcbbL ΔlysR2 (Figure 2B) strongly suggest that LysR2 acts as a repressor for cbbM gene expression where CbbM itself contributes to intensified LysR binding ability and thus its own transcriptional regulation (Figure 5). The combination of CbbL and LysR1 also appears to result in a repressive cbbM transcriptional regulation but single mutations in cbbL and lysR1 did not cause upregulation of the cbbM gene (see Figure 1). However, despite higher cbbM transcript levels in ΔcbbL ΔlysR1, this clone did not demonstrate an increase in RubisCO activity (Figure 2B), contrasting the transcriptional data at first glance. These results can be explained though if LysR1 also controls the expression of post-translational activators, which here only seems applicable if CbbL is also present. And indeed LysR1 appears to be able to bind upstream of cbbQ-m/cbbQ-m and of cbbQ-1/cbbO-1 (Figure 4), where also putative LysR binding sites were recognized (Figure 3). Of these likely post-translational activators only the deletion of cbbQ-1 caused a CbbM activity loss and thus makes its LysR regulated role in CbbM activation under the provided conditions highly likely. We thus suggest that CbbQ-1 activates CbbM and that LysR1 proteins may prevent CbbQ-1 expression if CbbL is present. Additionally, an insertion at position 9,463 bp, i.e., 15 bp upstream of the cbbQ-1 transcription start in the ΔcbbL clone 24II, resulted in significantly reduced CbbM activity (see Figure 2B, clone 24II2G1). Even though this insertion is located downstream of putative regulatory features, the 1,674 bp comprising <TET> insertion represents a barrier the RNA polymerase most likely cannot simply overcome, which would result in an impaired cbbQ-1 transcription and thus no CbbQ-1 would be present. Unexpectedly, the deletion in cbbO-1, downstream of cbbQ-1, did not alter RubisCO activity (Figure 2B). However, the elevated number of inverted repeats (5/18 identified on the metagenomic fragment) in the cbbO-1 gene, which are theoretically capable of forming stem-loop structures, may indicate some fine-tuned transcriptional regulation (Treangen et al., 2009 and references therein). Under the provided conditions cbbO-1 may be downregulated and under other environmental conditions may well be important for RubisCO activation.

In the intergenic region of cbbM and lysR2 a large inverted repeat flanking 86 nt (IR8, see Supplementary Figure 2) also exists. Such inverted repeats often form stem-loop structures that are important for controlling transcription initiation and termination, mRNA stabilization or genome plasticity (Treangen et al., 2009 and references therein). They also play roles in supporting DNA binding proteins in finding their binding sites (Frost et al., 1994). The IR8 in our metagenomic fragment may well represent such a signaling stem-loop structure that guides LysR to the LysR binding sites. An insertion of a tetracycline cassette in the left arm of the IR8 (clone 24II1G2), directly between the LysR binding sites ‘bs1_cbbM’ and ‘bs2/3_cbbM’ (see position 5,232 nt in Supplementary Figure 2) resulted in a significant RubisCO activity loss (Figure 2B). One explanation could be that this structure is necessary for the expression of a functional CbbM. Since this insertion also separates the putative -10 from the -35 promoter box, the RubisCO activity loss may be due to the impaired promoter region.

Restricting the work to culture-depended approaches, neglects the large majority of RubisCO gene clusters from uncultured organisms. However, working with metagenomes in non-native systems holds both promise and pitfalls. The benefits of using E. coli as a host organism are well-known: E. coli has an unrivaled fast growth on inexpensive media and the genetics are very well-known, making transformations with exogenous DNA simple and straightforward (Rosano and Ceccarelli, 2014). In contrast, expression of a metagenomic fragment in a surrogate host may also entail cross-talks, inhibitions and unspecific reactions. Recombinant gene expression in E. coli and other surrogate hosts might be troublesome due to, e.g., unrecognized intrinsic promotors and associated factors, a diverging codon usage or problems with correct protein folding (Perner et al., 2011).

One major advantage of working with a metagenomic fragment in fosmid clones (besides gaining access to the world of the unculturables) is its relatively small size (in this case: 13 kb metagenomic fragment). The small size infers clear gene arrangements and a limited number of possible gene and/or protein interactions relative to the (hardly tangible) complexity in a native system. The same work in a cultured representative is considerably more difficult and time consuming, because genes with yet unknown functions, which are not necessarily located in the vicinity of the gene cluster under investigation, may well participate in/contribute to the gene regulation and activation of the enzyme (indirectly), as has recently been shown for orf06 (Böhnke and Perner, 2015). Although the metagenomic approach with a defined number of genes can simplify first insights into regulatory mechanisms, it can also hinder the understanding of the mechanisms given that some vital genes/respective products cannot be expressed/synthesized as they are located on parts of the genome not present on the captured fraction of the metagenome.

To overcome such limitations the use of a host with the genomic inventory to operate the CBB cycle may be a viable option. In our case a cultured Thiomicrospira strain could be used and the genes under investigation deleted. However, deleting gene clusters in T. crunogena which are comparable to our 13 kb metagenomic fragment and expressing the latter heterologously in the Thiomicrospira host or alternatively constructing double mutants as we did in our metagenomic fragment (i.e., nine deletions in ΔcbbL and eight deletions in ΔcbbM) to investigate regulatory mechanisms is hardly feasible, given that Thiomicrospira’s genetic accessibility is not understood and thus any transformation with exogenous DNA becomes challenging. An alternative host that operates the CBB cycle and where mutations have been successfully constructed is R. capsulatus (Alphaproteobacteria) (Paoli et al., 1998; Witte et al., 2010; Dangel et al., 2014; Varaljay et al., 2016). However, this potential host encodes different types of RubisCOs and has other RubisCO gene cluster arrangements and likely different gene regulation mechanisms than the organism encoding our metagenomic fragment.

For future work one may consider combining studies in a genetically accessible surrogate host such as E. coli, naturally incapable of operating the CBB cycle, with subsequent investigations in a closely related cultured representative. Thus, the first insights of complex RubisCO regulatory mechanisms obtained through studies dealing with RubisCO gene expression in a non-native system could be used in further studies where, e.g., the role of external factors could be studied in a native system.