MainA promising strategy to realize higher microbial yields on renewable, electrochemically derived formate is the implementation of more ATP-efficient, synthetic pathways1. One of the most promising synthetic pathways with potential to show this yield improvement is the reductive glycine pathway (rGlyP), which is the most ATP-efficient aerobic pathway for formate assimilation (Fig. 1a,b). We recently demonstrated full synthetic pathway operation, partly expressed from plasmids, via modular and evolutionary engineering in Cupriavidus necator, but showed that biomass yields were so far still not exceeding the upper bounds of natural formatotrophy via the Calvin–Benson–Bassham (CBB) cycle2. Here we show the full integration of the synthetic rGlyP into the genome of C. necator. In chemostat experiments, we demonstrate a yield of 4.52 g cell dry weight (CDW) mol−1 formate for the engineered strain. This yield is 17% higher than that of the wild type and higher than any reported yield for growth on formate via engineered pathways as well as the CBB cycle. This study shows that superior microbial growth yields via synthetic one-carbon (C1)-assimilation pathways are feasible, paving the way for more efficient, sustainable bioproduction.Fig. 1: The rGlyP compared with the CBB cycle.a, In the CO2 module, formate is generated from electrochemical reduction of CO2. Formate can then be used to form pyruvate as indicated by the rGlyP or the CBB cycle, resulting in different formate requirements. In the rGlyP, the C1 module activates and reduces formate to 5,10-methylene-THF via three heterologous enzymes from M. extorquens: FtfL (Ft), FchA (Fc) and MtdA (M). Next, methylene-THF is converted to glycine by the glycine cleavage system (GCS, composed of the enzymes GcvT, H and P) operating in the reductive direction (C2 module). Finally, glycine is condensed with another 5,10-methylene-THF via serine hydroxymethyltransferase (GlyA, G) to yield serine, which is then dehydrated by serine deaminase (SdaA, S) to pyruvate. NADH is regenerated from formate via formate dehydrogenase. NADPH is regenerated by proton-translocating, membrane-bound transhydrogenase (PntAB). Formate and pyruvate are referred to as such in the text, but are here depicted in their protonated forms, in which they are metabolized. In the CBB cycle, three ribulose-5-phosphate molecules and CO2 are converted into 3-phosphoglycerate via the CBB cycle signature enzymes phosphoribulokinase (Prk) and RuBisCO. Subsequently, one of six generated 3-phosphoglycerate molecules can be used in metabolism, for example, via conversion to pyruvate. The other five 3-phosphoglycerate molecules are recycled into ribulose-5-phosphate via gluconeogenesis and the pentose phosphate pathway. For formatotrophic growth via the CBB cycle, all formate is oxidized into NADH and CO2 to supply energy and carbon to the CBB cycle. b, Comparison of the rGlyP with the CBB cycle for the number of enzymatic reactions required, cost of ATP and reducing equivalents, minimal thermodynamic driving force MDF (Supplementary Fig. 1) and biomass yield predicted by the model at a doubling time of 14 h and assuming GAM = 135 mmol gCDW−1 and NGAM = 3 mmol gCDW−1 h−1). Values were calculated for both routes from formate to pyruvate at 10% CO2 (3.4 mM). The GAPDH reaction was set as NAD+-dependent for the CBB cycle, and the standard NADH/NAD+ ratio of 0.1 was used, as the ratio has not been determined for C. necator grown on formate.Source dataFull size imageTo generate a genome-integrated rGlyP strain, we first cured the previously engineered Cupriavidus reductive glycine 4 (CRG4) strain2 of the plasmid expressing the first (C1) module of the rGlyP (Extended Data Fig. 1 and Methods). As this strain still overexpressed the remaining C2 and C3 modules of the rGlyP, it was an ideal platform to select for optimal, genomically expressed C1 module activity on formate. We then integrated a library of operons expressing the C1 module from different strength promoters into the genome using the Tn5-transposon machinery and subsequent selection for fast growth of clones on formate minimal media (Extended Data Fig. 1 and Supplementary Note 1). A fast-growing strain was selected (CRG5).Next, we also integrated the C3 module that was still expressed from a plasmid into the genome of CRG5, using the same approach as above (Supplementary Note 2). This led to the selection of a fast-growing CRG6 strain with a completely genome-integrated rGlyP. The strains were characterized for their growth and genome sequences (Extended Data Figs. 2–5 and Supplementary Tables 1–3). Strain CRG6 clearly grew faster than CRG4 at a doubling time in batch cultures of ~11 h, which was (still) slower than that of the wild-type CBB cycle strain (~6 h doubling time; Fig. 2a). However, both CRG5 and CRG6 consistently grew to a higher maximum biomass optical density than both CRG4 and the wild type (Fig. 2a and Extended Data Fig. 2).Fig. 2: Characterization of the rGlyP and CBB cycle C. necator strains.a, Growth of C. necator H16 ΔphaC1 strains harbouring the CBB cycle compared with that of CRG4, CRG5 and CRG6 in M9 minimal media supplemented with 80 mM formate and 100 mM bicarbonate with 10% CO2 in the headspace. The doubling time in hours of the strains is presented in the designated strain colour. Curves depict the mean of at least two technical replicates and are representative of three experiments conducted in the same conditions to ensure reproducibility. b, Relative protein intensity changes of the rGlyP modules in strain CRG6 relative to CRG4 both grown on formate minimal media. c, Fractions of the quantified proteome associated with various metabolic tasks in the strains CBB, CRG4 and CRG6 all grown on formate minimal media. Clustering criteria for grouping of proteins by metabolic tasks are provided in Methods. d, Measured biomass yields in grams CDW per mole formate consumed are shown, both for the CBB cycle (C. necator ΔphaC1) and rGlyP (CRG6 ΔdadA6) strains grown in formate minimal media in bioreactors in chemostat mode at a dilution rate of 0.05 h−1 (14 h doubling time). Predicted biomass yields are shown in a shaded pattern and are derived from FBA simulations run with a fixed growth rate corresponding to a doubling time of 14 h and maintenance costs of GAM = 135 mmol gCDW−1 and NGAM = 3 mmol gCDW−1 h−1. Measured data represent samples (n = 9) obtained during steady state, each having a volume of 50 ml or 100 ml taken over the course of 3 days. All data points and their mean value are given. Error bars indicate standard deviation. Significance was tested via two-tailed unpaired t-test. ****P = 1.33 × 10−6; 95% confidence interval = 0.4600–0.8244; difference between means ± s.e.m. = 0.6422 ± 0.08594; d.f. = 8.Source dataFull size imageTo investigate and compare the cellular proteome allocation in the rGlyP strains CRG4 (plasmid expressed) and CRG6 (genomically expressed), as well as the CBB cycle strain, we performed quantitative proteomics (Fig. 2b,c, Supplementary Note 3 and Extended Data Fig. 6a–c). This analysis showed that the proteome allocation of the C1 module after genome integration decreased only slightly, whereas the allocation of the C3 module was reduced more than tenfold, suggesting that the latter module was redundantly high expressed in CRG4.High C1 module expression was achieved via transposon insertion downstream of the native phaP1 promoter, which decreased phaP1 expression at the same time (Source Data). We confirmed that deletion of phaP1 alone did not allow for improved growth rate and yields (Extended Data Fig. 7). Overall ~13% of the proteome allocation associated with the rGlyP in CRG4 was freed up in CRG6, which probably allowed allocation of proteome resources to other processes such as ribosome biosynthesis (Fig. 2c), allowing for the increased growth rate of the CRG6 strain.We next investigated at two critical rGlyP bifurcation points (formate and glycine) whether wasteful oxidative sinks were limiting biomass formation using in vivo formate dehydrogenase inhibition assays and deletion of glycine oxidase (ΔdadA6) (Fig. 2a, Supplementary Note 4 and Extended Data Figs. 2 and 8a–e), but did not observe further improvements in growth rate or yield. This probably reflects that the cells already properly balanced formate assimilation flux with energy generation.With the seemingly energetically optimal CRG6 ΔdadA6 strain, we now wanted to accurately determine whether the rGlyP could also support a higher biomass yield on formate than the CBB cycle. Other than energy consumption differences in the pathway, also growth rate differences between the strains may impact yields. Typically, faster growth rates decrease the relative impact of maintenance energy requirements, which can further increase biomass yield. Furthermore, for formatotrophic growth of C. necator, a negative effect on yield has been reported due to high residual formate levels in the medium3, as formate is generally a toxic substrate that can inhibit growth rates and yields, especially at higher concentrations.Hence, we decided to accurately determine the formatotrophic yield of the CRG6 ΔdadA6 and CBB cycle strains by using bioreactors in chemostat mode at the same dilution or growth rate (0.05 h−1, that is, a doubling time of ~14 h). The same dilution rate was used in a previous study that characterized the yield via the CBB cycle on formate for C. necator, which reported a yield of ~4 g CDW mol−1 (ref. 3). The chemostat cultivation also ensures that the formate levels stay low (below the high-performance liquid chromatography (HPLC) detection limit) and hence prevents reductions in yield or growth rate due to toxicity.In chemostat bioreactor experiments, we obtained biomass yields of 3.88 ± 0.19 g CDW mol−1 for the CBB cycle strain and 4.52 ± 0.17 g CDW mol−1 for the rGlyP strain, showing a 16.6% higher yield compared with the wild type in the same condition (Fig. 2d). The measured yields are in accordance with the theoretically predicted yield by a genome-scale metabolic model (GEM) of C. necator for both the CBB cycle (3.92 g CDW mol−1) and the rGlyP (4.54 g CDW mol−1) at 14 h of doubling time. This confirms that the energetically superior formatotrophic pathway indeed can support the predicted yield increase.At the end of the chemostat experiment, the dilution rate was increased to 0.087 h−1, which corresponds to 8 h of doubling time. Both strains were maintained at this dilution rate (Extended Data Fig. 9), which for the rGlyP strain marks the fastest measured growth of C. necator strains growing via the rGlyP so far. At a dilution rate of 0.1 h−1, however, the CRG6 ΔdadA6 strain was washed out, indicating that it cannot yet sustain the same growth rate as the CBB-employing wild type (fastest doubling time ~4 h)3. The lower growth rate of the rGlyP strain compared with the CBB strain could still limit bioproduction rates, but probably can be further improved by long-term evolution as shown recently for the wild-type strain4.Our study shows that an engineered C1-assimilation pathway can outcompete the biomass yield of a natural pathway in the same organism. In fact, the measured formatotrophic yield of 4.52 g CDW mol−1 is, to our best knowledge, higher than any so far reported yield on formate for natural organisms using the CBB cycle, as well as for engineered formatotrophs (Supplementary Table 4). Notably, the relative yield increase of ~17% can in certain conditions even be higher, as in our comparison experiments the CBB cycle is not performing wasteful oxygenation and photorespiration (due to the high CO2).The measured yield, being very similar to the theoretically predicted yield, indicates that the metabolic network of the engineered strain including the synthetic assimilation pathway is well performing. This suggests that there is no major remaining losses or energy-wasting metabolic processes. This is in contrast to the so far demonstrated low yields for several recently engineered synthetic Escherichia coli and Saccharomyces cerevisiae formatotrophs and methylotrophs (Supplementary Table 4). This is probably because the metabolic networks of the naturally strict heterotrophs are less suited for autotrophic metabolism. An example is the relatively high remaining activity of the tricarboxylic acid (TCA) cycle in the E. coli strain growing via the rGlyP5. During formatotrophic growth, the TCA cycle is a redundant process, wasting acetyl-CoA and hence decreasing yields on formate. C. necator seemingly already evolved to suppress TCA cycle activity efficiently during formatotrophic and autotrophic growth with the CBB cycle6, and this regulation seems sustained in the rGlyP strains as indicated via 13C-labelling in our previous study2.Our study shows that designed pathways with theoretically higher efficiencies can indeed improve yields in vivo. In the future, other promising energy-efficient, synthetic C1-fixation pathways, which were only shown in vitro such as the synthetic crotonyl-CoA-ethylmalonyl-CoA-hydroxybutyryl-CoA (CETCH) cycle for CO2 fixation7, may outperform the CBB cycle in vivo, once successfully established. Implementation of such CO2 fixation pathways in autotrophic hosts, such as C. necator or photosynthetic microorganisms, may similarly benefit from the natively already well-wired autotrophic metabolic network to efficiently realize higher yields than the CBB cycle, as observed here. The (randomized) transposon integration technique and growth-coupled selection approach in this study could furthermore serve as a blueprint to realize efficient pathway implementation, also in more difficult-to-engineer bacteria, as well as eukaryotes, such as yeast and microalgae. Realizing higher yields and faster growth on formate, CO2 and other C1 substrates will aid to make the electro-microbial production route an attractive, economically feasible reality for sustainable production of carbon-based chemicals, fuels, feed and food.MethodsBacterial strains and conjugationA complete list of strains and plasmids used in this study can be found in Supplementary Tables 5 and 6. C. necator H16 deleted in polyhydroxybutyrate biosynthesis (ΔphaC1) served as the base strain (‘wild type’) in this work. All CRG strains are also deleted for carbon fixation via the CBB cycle by deletion of both ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCO) subunits on chromosome 2 and megaplasmid (ΔcbbSL2, ΔcbbSLp). Routine cloning was performed in E. coli DH5α, while E. coli ST18 cells were used for conjugation of mobilizable plasmids to C. necator via biparental spot mating.Cultivation conditionsC. necator and E. coli were grown on lysogeny broth (LB; 10 g l−1 NaCl, 5 g l−1 yeast extract and 10 g l−1 tryptone) for routine cultivation and genetic modifications. When appropriate, the antibiotics kanamycin (100 μg ml−1 for C. necator and 50 μg ml−1 for E. coli), chloramphenicol (30 μg ml−1), tetracycline (10 μg ml−1), ampicillin (100 μg ml−1 for E. coli) and gentamycin (20 μg ml−1 for C. necator) were added. Growth experiments were conducted in M9 minimal medium (47.8 mM Na2HPO4, 22 mM KH2PO4, 8.6 mM NaCl, 18.7 mM NH4Cl, 2 mM MgSO4 and 100 μM CaCl2), supplemented with trace elements (134 μM EDTA, 31 μM FeCl3·6H2O, 6.2 μM ZnCl2, 0.76 μM CuCl2·2H2O, 0.42 μM CoCl2·2H2O, 1.62 μM H3BO3, 0.081 μM MnCl2·4H2O). Routine cultivation was performed in 4 ml medium in 12-ml glass tubes in an orbital shaker incubator at 240 rpm at 30 °C and 37 °C for C. necator and E. coli, respectively. M9 minimal medium was supplemented with 80 mM sodium formate and 100 mM sodium bicarbonate, with the pH adjusted to 7.2, under a headspace of 10% CO2 (v/v) for formatotrophic growth. Strictly seen enough CO2 is generated intracellularly to drive the GCV carboxylation of the rGlyP by formate oxidation. However, in relatively low-biomass-density cultures, with high aeration, as performed in this study, this may be insufficient. Hence, we supplement sodium bicarbonate in the medium (only during batch cultivations) and CO2 in the headspace or gas supply during the bioreactor cultivation.No antibiotics were added during growth characterization experiments in the plate reader. Growth measurements were obtained from 96-well-plate experiments (Nunc transparent flat bottom, Thermo Scientific). Strains were typically pre-cultured in M9 minimal medium supplemented with 20 mM pyruvate. Cells were collected, washed twice and inoculated at an optical density at 600 nm (OD600) of 0.01. To avoid evaporation while maintaining diffusion of O2 and CO2, 150 μl of cell medium mix was topped with 50 μl transparent mineral oil (Sigma-Aldrich). The 96-well plates were incubated at 30 °C with continuous shaking (alternating between 30 s orbital and 30 s linear) in a Tecan infinite M200Pro plate reader (Tecan). OD600 values were measured every 8 min. Growth data were blanked and converted from plate reader OD600 to cuvette OD600 by multiplication with a factor of 4.35 via a Matlab script. All growth experiments were repeated at least three times, and the growth curves shown are representative curves of these experiments.Plasmid curingCells containing plasmids to be cured were propagated in LB media without antibiotics. From each grown passage, 50 colonies were streaked out on LB agar plates. These were then replica plated on LB agar plates containing the antibiotic for which the plasmid would provide resistance and on non-antibiotic-containing plates. Once colonies were obtained that did not grow on the antibiotic-containing plates, the respective colonies from the non-selective plate were investigated via PCR targeting the plasmid to confirm the curing of the clone.Tn5 vector constructionThe vector pBAMD1-4 was provided by P. I. Nikel. The vector backbone was amplified with primers pBAMD-for (GCGCGGCCGCATAAAATCTCTGAAGATGTG) for the rGlyP-C1 module or pBAMD-for2 (GCGATGCATATAAAATCTCTGAAGATGTG) for the rGlyP-C3 module and pBAMD-rev (GCGGCTAGCGCCTGAGACACAAAGATGTG) without the antibiotic resistance gene in the cargo module (module between the transposon recognition sites ME1 and ME2), and NotI and NheI restriction sites were attached. The operons mtdA-fch-ftfL and sdaA-glyA were previously cloned under control of the promoters, from weakest to strongest, P14/PPhaC1/P3/P4/P2 and Pcat/PPhaC1/P3/P4, respectively. These were then amplified from existing pSEVA221- and pSEVA331-based expression plasmids respectively with primers C1-for (GCGGCTAGCTCTAGGGCGGCGGATTTGTC) and C1-rev (GCGCGGCCGCTTGGGGACCCCTGGATTCTC) attaching NotI and NheI restriction sites or C1-for and C3-rev (GACATGCATTTGGGGACCCCTGGATTCTC) to attach NheI and NsiI sites. Promoter-FCM/SG operons were cloned into pBAMD vectors using restriction ligation.All genes of the C1 module were previously placed behind a synthetic RBS designed with an RBS Calculator with a translation initiation rate of 30,000 arbitrary (arb.) units (ref. 8).Gene deletionsThe plasmid pLO3-dadA6 was used to delete the dadA6 gene in the CRG6 strain via allelic replacement based on sucrose counter selection with SacB as described previously2,9. For the phaP1 deletion, we used the recently established ‘Self-Splicing Intron-Based Riboswitch’ (SIBR) system for Cas9-based counter selection10. With the use of HR1_phaP1_F (GAGCAAGCCCGTAGGGGGGGAACTGGGCATCAGGAC), HR1_phaP1_R (CTGACATCTAGGCGGCTTTGATAACTGCCTGCG), HR2_phaP1_F (TTCAACGCAGGCAGTTATCAAAGCCGCCTAGATGTCAG) and HR2_phaP1_R (CGCCGCCCTAGACAGCTGGGAGGTCGCTGGCCTCTTTG), 1-kb flanking arms were amplified and cloned into pSIBR004 together with spacer PhaP1-spacer1 (TGCCAACAACGCCTACGAGT) or PhaP1-spacer2 (TTTCCGAAGCGATTTCATAC). The deletions were confirmed by colony PCR using the primers dadA6-for (TGGAAGGCTACCCCTACTTC) and dadA6-rev (ATAGAAACTCAGCGGCTGGC) or phaP1-for (ACGTAGCCGATGCCTG) and phaP1-rev (TAGGTATCGTCGTCGC) and whole-genome sequencing.Tn5-mediated knock-in coupled with liquid selective conditionsE. coli ST18 strains harbouring the P14/PPhaC1/P3/P4/P2-C1 and Pcat/PPhaC1/P3/P4-C3 constructs in pBAMD-Tn5 vectors served as donor strains. C. necator strains CRG4.5 and CRG5.5 served as recipient strains for the conjugation. In a 1.5-ml plastic tube, 100 µl of LB-grown dense overnight cultures of C. necator and E. coli ST18 (supplemented with 50 µg ml−1 5-amniolevulinic acid (ALA)) were mixed. Of this cell mixture, 100 µl was plated on LB + ALA agar plates and dried for 30 min before incubating overnight at 30 °C. The next day, the grown cell lawn was resuspended in LB and a 100-µl inoculum of an OD600 = 1 mixture of C. necator recipient and E. coli ST18 donor cells was used to inoculate 4 ml of liquid M9 minimal medium supplemented with 80 mM formate and 100 mM bicarbonate in 10% CO2 (v/v). When cellular growth was observed and the population reached late log to stationary phase, 1 µl of OD600 = 1 culture was used to reinoculate into 4 ml of selective medium. This population was passaged 10 times to allow the Tn5 transformants with better pathway expression to overtake the population (Extended Data Fig. 3). After passage 10, the population was dilution streaked two consecutive times to single colony on LB agar plates with 20 µg ml−1 gentamicin. The isolated single clones were compared with the populations for growth behaviour in the selective formate media and saved in 25% glycerol at −80 °C (Extended Data Fig. 4).Bioreactor experiments and biomass yield determinationBacterial strains were streaked out on LB agar plates containing antibiotics when appropriate. Single clones were inoculated from the plate into 12-cm glass tubes containing 4 ml of M9 minimal medium supplemented with 80 mM formate and 100 mM HCO3 and cultivated at 10% CO2, 30 °C and 250 rpm. Growing cultures were then used to inoculate 50 ml of the same media in 250-ml flasks and cultivated in the same conditions. Pre-cultures were then used to inoculate 1:10 in M9 80 mM formate media and grown in batch at 30 °C and 400 rpm in 400 ml volume in 1 l DASGIP bioreactors (Eppendorf). The reactors were sparged at a constant flow of 60 sl h−1 (2.5 volume gas per volume culture per min) with a gas mixture of 10% CO2 and 90% air. Dissolved oxygen was measured online and was maintained above 30% via stirrer speed control (400–800 rpm).Following batch growth, the OD was increased using fed batch by adding 4 ml or 8 ml of 4 M formic acid once per day. After an OD600 of >1.2 was reached, the dilution of the reactor was started with M9 80 mM formate media (pH = 3.7) by setting the pumps to a flow rate of 20 ml h−1. The waste pump flow rate was set to 100 ml h−1. The waste pipe was positioned at the surface level of the 400 ml culture volume. The lag phase of the CRG6 ΔdadA6 strain caused the pH to quickly drop, as formate could not be oxidized fast enough, which resulted in a positive feedback loop of lower pH and more lag. Hence, pH-neutral M9 80 mM formate was supplied at the same dilution rate of 0.05 h−1 with 1 M formic acid being used to maintain the pH at 7.1. After the strain reached steady state, the medium was switched to the low-pH M9 80 mM formate medium, which during exponential growth did not cause a wash-out of the strain. Steady state for this medium was reached after more than 7 reactor turnovers (140 h). Every day for 3 days, 50- and 100-ml samples were taken. Formate-grown cells were collected by centrifugation at 3,220 g for 20 min and washed 3 times in 50 ml double distilled water to remove residual medium components. Washed cells were then pipetted into custom-made pre-dried and pre-weighed aluminium trays. The samples were dried for a period of 24 h at 90 °C and were then weighed to obtain the additional weight from the dry cells.Formate concentrations were quantified through HPLC using the Shimadzu LC-2030C Plus system equipped with a Shodex SH1821 column (8 mm × 300 mm, at 45 °C), using refractive index detection. Elution was performed with 0.01 N sulfuric acid at a flow rate of 1 ml min−1, and 10 µl of internal standard (0.1 M sulfuric acid) was co-injected with all sample injections. For construction of a calibration line, two standards were used containing, respectively, 10 mM and 100 mM of formic acid as well as several other organic acids (lactate, acetic acid, propionic acid, isobutyric acid and butyric acid). First, 1 µl, 2 µl and 5 µl of the 10 mM standard solution were injected to get calibration points for 1 mM, 2 mM and 5 mM formate, respectively. In addition, 1 µl, 2 µl, 5 µl and 10 µl of the 100 mM standard solution were injected to get calibration points for 10 mM, 20 mM, 50 mM and 100 mM formate, respectively. Calibration was validated using a commercial formate standard containing 1,000 mg l−1 ± 5 mg l−1 (or 22.21 mM ± 0.11 mM) formate in water (44293 from Merck), of which 10 µl was injected. For sample measurement, 10 µl of sample was injected. No residual formate could be detected in the bioreactor samples. The biomass yield in grams CDW per mole formate was then calculated by dividing the CDW concentration (g l−1) by the consumed formate concentration (mol l−1).Whole-genome sequencingC. necator genomes and plasmids were extracted from LB-grown cells for whole-genome sequencing using the NucleoSpin Microbial DNA Kit (Macherey-Nagel). Samples were sent for library preparation (Nextera, Ilumina) and sequencing at an Ilumina NovaSeq 6000 platform to obtain 150-bp paired-end reads (Novogene). Samples were paired, trimmed and assembled to the C. necator reference genome using the Geneious 8.1 software (Biomatters) or the Breseq pipeline11. Mutations (frequency above >60%) were identified based on comparative analysis with the parental strains.Max–min driving force analysisMax–min driving force (MDF) analysis12 was used to compare the thermodynamic feasibility of the CBB cycle and rGlyP for pyruvate formation from formate in an atmosphere of 10% CO2 (v/v) (Supplementary Fig. 1). The Python packages equilibrator_api (v 0.4.7) and equilibrator_pathway (v 0.4.7) were used. Changes in Gibbs free energy of the reactions were estimated using the component contribution method13. Default values were used, metabolite and cofactor concentrations were constrained to the range 1 μM to 10 mM, pH was set to 7.5, ionic strength was assumed to be 0.25 M and magnesium concentration was 1 mM.Constraint-based metabolic modellingThe genome-scale metabolic model (GEM) of C. necator, RehMBEL1391_sbml_L3V1, was retrieved in SBML format level 3 version 1 from the public repository provided in ref. 6. Model simulations were performed using COBRApy 0.24.0 (ref. 14) and Python 3.9.Flux balance analysis (FBA) was implemented to simulate formate assimilation for both the CBB cycle and the rGlyP. To simulate formate assimilation through the CBB cycle, the glycine cleavage system reaction (‘GLYAMT’) was set as non-reversible to prevent any flux through the rGlyP. To support formate assimilation through rGlyP, two additional reactions were added to the GEM: formate tetrahydrofolate ligase (‘Ftl’) and methenyltetrahydrofolate cyclohydrolase (‘Fch’). In addition, the GLYAMT reaction was set as reversible, and the flux through the ribulose-bisphosphate carboxylase reaction (‘RBPC’) was set to 0 to prevent flux through the CBB cycle.The ATP hydrolysis part of the biomass synthesis reaction was amended with an H2O molecule and H+ to balance the reaction. The growth-associated maintenance (GAM; 135 mmol gCDW−1) and non-growth-associated maintenance (NGAM; 3.0 mmol gCDW−1 h−1) values used to run the simulations were retrieved from ref. 6.More information on their fine-tuned GAM parameter can be found between lines 101 and 104 of the script called ‘run_simulations.py’ within the reported GitLab repository (https://github.com/m-jahn/genome-scale-models/blob/master/Ralstonia_eutropha/).To calculate the theoretical biomass yield, we constrained the biomass reaction to the growth rate corresponding to 14 h doubling time, the growth rate used in the bioreactors (Fig. 2d). The formate uptake rate reaction (‘EX_formate_e’) was used as the objective function, and fluxes were computed for each scenario under the specified conditions. The ratio between growth rate (h−1) and the maximum predicted formate uptake rate (mmol g−1 CDW h−1) was used to calculate the biomass yields in g CDW mol−1 formate.Proteomic analysisC. necator CBB, CRG4 and CRG6 strains were pre-cultured in 4 ml M9 minimal medium supplemented with 80 mM formate and 100 mM HCO3 in 15-ml glass tubes. Then, 1 ml of cells in the late exponential phase was collected and washed three times in M9 medium without a carbon source. From here, 50 ml of M9 minimal medium supplemented with 80 mM formate and 100 mM HCO3 was inoculated to a starting OD600 of 0.01 in 250-ml non-baffled shake flasks. Cultures were incubated in an Infors Minitron at 30 °C in an atmosphere of 10% CO2 (v/v) at 200 rpm. This flask-adapted culture was then used to inoculate a second flask culture (the third consecutive formate cultivation) in the same way. Cells from 3 biological replicates were collected in the mid-log phase (OD600 = 0.3–0.5) and washed twice with phosphate buffer (12 mM phosphate buffer, 2.7 mM KCl, 137 mM NaCl, pH = 7.4). Cell pellets corresponding to 1 ml of OD600 = 3 were flash-frozen in liquid nitrogen and stored at −70 °C until further use. Cell lysis and protein solubilization were conducted as previously reported. In brief, cells were incubated 15 min at 90 °C in 2% sodium lauroyl sarcosinate (SLS) and 100 mM ammonium bicarbonate and then sonicated for 15 s (Vial Tweeter, Hielscher). Soluble proteins were reduced via incubation with 5 mM Tris (2-carboxy-ethyl) phosphine (TCEP) at 90 °C for 15 min, followed by alkylation with 10 mM iodoacetamide for 15 min at 25 °C. Protein concentrations were quantified via a bicinchoninic acid assay (BCA) protein assay kit (Thermo Fisher Scientific). Then, 50 µg of protein was digested with 1 µg trypsine (Promega) in 0.25% SLS (diluted with 100 mM ammonium bicarbonate) overnight at 30 °C. Following SLS removal via centrifugation, trifluoroacetic acid (TFA) was added to a final concentration of 1.5% and samples were incubated at room temperature for 10 min. The supernatant was purified using C18 Micro Spin Columns (Harvard Apparatus) according to the manufacturer’s instructions, dried and resuspended in 0.1% TFA. Peptide mixtures were then analysed using liquid chromatography–mass spectrometry carried out on an Exploris 480 instrument connected to an Ultimate 3000 RSLC nano with a Proflow upgrade and a nanospray flex ion source (all Thermo Scientific). Peptide separation was performed on a reverse-phase HPLC column (75 μm × 42 cm) packed in-house with C18 resin (2.4 μm, Dr. Maisch). The following separating gradient was used: 94% solvent A (0.15% formic acid) and 6% solvent B (99.85% acetonitrile, 0.15% formic acid) to 25% solvent B over 95 min and to 35% B for an additional 25 min at a flow rate of 300 nl min−1. The data-independent acquisition-mass spectrometry (DIA-MS) acquisition method was adapted from ref. 15. In short, the spray voltage was set to 2.0 kV, the funnel radio frequency (RF) level at 45 and the heated capillary temperature at 275 °C. For DIA experiments, full MS resolutions were set to 120,000 at m/z 200, and the full MS automatic gain control (AGC) target was 300% with an injection time (IT) of 50 ms. The mass range was set to 350–1,400. The AGC target value for fragment spectra was set at 3,000%. A total of 49 windows of 15 Da were used with an overlap of 1 Da. The resolution was set to 15,000 and the IT to 22 ms. Stepped higher-energy collisional dissociation (HCD) collision energy of 25%, 27.5% and 30% was used. MS1 data were acquired in profile, MS2 DIA data in centroid mode.Analysis of DIA data was performed using DIA-NN version 1.8 (ref. 16), using the UniProt protein database from C. necator H16 and added sequences for formate–tetrahydrofolate (THF) ligase (ftfL, UniProt: Q83WS0), 5,10-methenyl-THF cyclohydrolase (fchA, UniProt: Q49135) and 5,10-methylene-THF dehydrogenase (mtdA, UniProt: P55818) from Methylorubrum extorquens AM1, RK2 plasmid replication protein (trfA, UniProt: P07676), pBBR1 replication protein (pSEVA331 derived AA sequence), aminoglycoside 3′-phosphotransferase (aphA1, Uniprot: P00551) and chloramphenicol acetyltransferase (cat, Uniprot: P62580). Full tryptic digest was allowed with three missed cleavage sites, and oxidized methionines and carbamidomethylated cysteines. Match between runs and remove likely interferences were enabled. The neural network classifier was set to the single-pass mode, and protein inference was based on genes. Quantification strategy was set to any LC (high accuracy). Cross-run normalization was set to retention time (RT) dependent. Library generation was set to smart profiling. DIA-NN outputs were further evaluated using a SafeQuant version modified to process DIA-NN outputs17.The different metabolic groups were defined as follows. ‘rGlyP-C1’ is composed of FtfL, Fch and MtdA, ‘rGlyP-C2’ of GcvT1HP and ‘rGlyP-C3’ of SdaA and GlyA. The ‘plasmid backbone’ proteome is composed of TrfA, AphA1, pBBR Rep and Cat. ‘CBB cycle’ proteins contain when applicable CbbL1, CbbL2, CbbS, CbbS2, CfxP, CbxXC, CbbYC, CbbAC, CbbAP, Fbp2, Fbp3, Rpe1, Rpe2, CbxXP, CbbTC, CbbTP, CbbZC, CbbZP, CbbKC, CbbKP, CbbGC and CbbGP. The ‘hydrogen utilization’ proteome contains the proteome fractions of HoxA, HoxB, HoxC, HoxF, HoxG, HoxH, HoxI, HoxK, HoxL, HoxM, HoxN, HoxO, HoxQ, HoxR, HoxU, HoxV, HoxW, HoxY, HoxZ, HypA, HypB, HypB2, HypC, HypD, HypE, HypF1, HypF2 and HypX. ‘Formate utilization’ is composed of the proteins FdsD, FdsA, FdsB, FdsG, FdsR, FdoI, FdoH, FdoG, FdhA1, FdhA2, FdhB1, FdhC, FdhD, FdhD1, FdhD2, FdhE, FdwA, FdwB and CbbB. rpsA, rpsB, rpsC, rpsD, rpsE, rpsF, rpsG, rpsH, rpsI, rpsJ, rpsK, rpsL, rpsM, rpsN, rpsO, rpsP, rpsQ, rpsR, rpsS, rpsT, rpsU, rpsU, rplA, rplB, rplC, rplD, rplE, rplF, rplI, rplJ, rplK, rplL, rplM, rplN, rplO, rplP, rplQ, rplR, rplS, rplT, rplU, rplV, rplW, rplX, rplY, rpmA, rpmB, rpmC, rpmD, rpmE2, rpmF, rpmG, rpmH, rpmI and rpmJ make up the ribosome group. All other detected proteins (~3,500) make up the ‘other’ category.Figure preparationPerseus 1.5.1.6 was used to analyse and plot proteomics data. Biomass yield data were plotted and checked for significance in GraphPad Prism 10.1.0 using a two-tailed, unpaired t-test. Adobe Illustrator 2020 was used to make the illustrations for the workflow, the transposon insertion sites and the formate dehydrogenase inhibition trials, as well as to prepare and export the final figures.Reporting summaryFurther information on research design is available in the Nature Portfolio Reporting Summary linked to this article.Data availabilityData supporting the findings of this study are found in the Article and Supplementary Information. Sequence data of plasmids constructed and/or used in this study were made available via EDMOND (https://doi.org/10.17617/3.FSBOQE). Whole-genome sequencing data have been deposited in NCBI Sequence Read Archive (SRA) under BioProject number PRJNA1209586. The mass spectrometry proteomics data have been deposited in ProteomeXchange Consortium via the PRIDE partner repository with dataset identifier PXD059545. Source data are provided with this paper. Further information, bacterial strains and materials related to this study are available from the corresponding authors.Code availabilityThe GEM of C. necator -RehMBEL1391_sbml_L3V1, the code and the associated results obtained in this study are available via GitHub at https://github.com/benit005/Formate_assimilation_rGlyPvsCBB_Cupriavidus_necator.ReferencesBar-Even, A., Noor, E., Flamholz, A. & Milo, R. Design and analysis of metabolic pathways supporting formatotrophic growth for electricity-dependent cultivation of microbes. Biochim. Biophys. Acta 1827, 1039–1047 (2013).Article CAS PubMed Google Scholar Claassens, N. J. et al. Replacing the Calvin cycle with the reductive glycine pathway in Cupriavidus necator. Metab. Eng. 62, 30–41 (2020).Article CAS PubMed Google Scholar Grunwald, S. et al. Kinetic and stoichiometric characterization of organoautotrophic growth of Ralstonia eutropha on formic acid in fed‐batch and continuous cultures. Microb. Biotechnol. 8, 155–163 (2015).Article CAS PubMed Google Scholar Calvey, C. H. et al. Improving growth of Cupriavidus necator H16 on formate using adaptive laboratory evolution-informed engineering. Metab. Eng. 75, 78–90 (2023).Article CAS PubMed Google Scholar Kim, S. et al. Optimizing E. coli as a formatotrophic platform for bioproduction via the reductive glycine pathway. Front. Bioeng. Biotechnol. 11, 1091899 (2023).Article PubMed PubMed Central Google Scholar Jahn, M. et al. Protein allocation and utilization in the versatile chemolithoautotroph Cupriavidus necator. eLife 10, e69019 (2021).Article CAS PubMed PubMed Central Google Scholar Giaveri, S. et al. Integrated translation and metabolism in a partially self-synthesizing biochemical network. Science 385, 174–178 (2024).Article CAS PubMed Google Scholar Farasat, I. et al. Efficient search, mapping, and optimization of multi‐protein genetic systems in diverse bacteria. Mol. Syst. Biol. 10, 731 (2014).Article PubMed PubMed Central Google Scholar Lenz, O. & Friedrich, B. A novel multicomponent regulatory system mediates H2 sensing in Alcaligenes eutrophus. Proc. Natl Acad. Sci. USA 95, 12474–12479 (1998).Article CAS PubMed PubMed Central Google Scholar Della Valle, S. et al. Streamlined and efficient genome editing in Cupriavidus necator H16 using an optimised SIBR-Cas system. Preprint at bioRxiv https://doi.org/10.1101/2024.11.24.625072 (2024).Deatherage, D. E. & Barrick, J. E. in Engineering and Analyzing Multicellular Systems Vol. 1151 (eds. Sun, L. & Shou, W.) 165–188 (Springer, 2014).Noor, E. et al. Pathway thermodynamics highlights kinetic obstacles in central metabolism. PLoS Comput. Biol. 10, e1003483 (2014).Article PubMed PubMed Central Google Scholar Noor, E., Haraldsdóttir, H. S., Milo, R. & Fleming, R. M. T. Consistent estimation of Gibbs energy using component contributions. PLoS Comput. Biol. 9, e1003098 (2013).Article CAS PubMed PubMed Central Google Scholar Ebrahim, A., Lerman, J. A., Palsson, B. O. & Hyduke, D. R. COBRApy: COnstraints-Based Reconstruction and Analysis for Python. BMC Syst. Biol. 7, 74 (2013).Article PubMed PubMed Central Google Scholar Bekker-Jensen, D. B. et al. A compact quadrupole-orbitrap mass spectrometer with FAIMS interface improves proteome coverage in short LC gradients. Mol. Cell. Proteomics 19, 716–729 (2020).Article CAS PubMed PubMed Central Google Scholar Demichev, V., Messner, C. B., Vernardis, S. I., Lilley, K. S. & Ralser, M. DIA-NN: neural networks and interference correction enable deep proteome coverage in high throughput. Nat. Methods 17, 41–44 (2020).Article CAS PubMed Google Scholar Ahrné, E., Molzahn, L., Glatter, T. & Schmidt, A. Critical assessment of proteome-wide label-free absolute abundance estimation strategies. Proteomics 13, 2567–2578 (2013).Article PubMed Google Scholar Download referencesAcknowledgementsWe dedicate this work to the memory of A. Bar-Even (1980–2020) who was involved in the design and the initial phase of this study. Above all, he had been our exceptional and dear mentor and colleague. We thank O. Lenz for providing plasmid pLO3 and advice on working with C. necator. In addition, we acknowledge P. Nikel for providing the pBAMD plasmids and W. Newell and J. Kahnt for experimental assistance. We also thank N. Paczia and P. Pfister for technical assistance with the bioreactor cultivations. We thank M. Jahn for input on the quantitative proteomic data analysis and G. Angelidou for help with bioinformatic analysis. Lastly, we thank A. Satanowski and S. Wenk for critical reading of the paper. This study was funded by the Max Planck Society (to B.D., E.O., H.S.-M., T.G., A.B.-E. and T.J.E.), the German Ministry of Education and Research via the Transformate grant (grant number 033RC023G to B.D., E.O. and A.B.-E.), an Add-On Fellowship for Interdisciplinary Life Science of the Joachim Herz Foundation and the Bosch Research Foundation (to H.S.-M.) and the Dutch Science Organization (NWO) with a Veni grant (grant number VI.Veni.192.156 to N.J.C.) and a Gravitation Project BaSyC (grant number 024.003.019 to S.Y. and N.J.C.).FundingOpen access funding provided by Max Planck Society.Author informationAuthors and AffiliationsSystems and Synthetic Metabolism, Max Planck Institute of Molecular Plant Physiology, Potsdam, GermanyBeau Dronsella, Enrico Orsi & Arren Bar-EvenBiochemistry and Synthetic Metabolism, Max Planck Institute for Terrestrial Microbiology, Marburg, GermanyBeau Dronsella, Helena Schulz-Mirbach & Tobias J. ErbLaboratory of Systems and Synthetic Biology, Wageningen University, Wageningen, The NetherlandsSara Benito-VaquerizoGenome Biology Unit, European Molecular Biology Laboratory, Heidelberg, GermanySara Benito-VaquerizoLaboratory of Microbiology, Wageningen University, Wageningen, The NetherlandsSuzan Yilmaz & Nico J. ClaassensCore Facility for Mass Spectrometry and Proteomics, Max Planck Institute for Terrestrial Microbiology, Marburg, GermanyTimo GlatterCenter for Synthetic Microbiology, Marburg, GermanyTobias J. ErbAuthorsBeau DronsellaView author publicationsYou can also search for this author in PubMed Google ScholarEnrico OrsiView author publicationsYou can also search for this author in PubMed Google ScholarHelena Schulz-MirbachView author publicationsYou can also search for this author in PubMed Google ScholarSara Benito-VaquerizoView author publicationsYou can also search for this author in PubMed Google ScholarSuzan YilmazView author publicationsYou can also search for this author in PubMed Google ScholarTimo GlatterView author publicationsYou can also search for this author in PubMed Google ScholarArren Bar-EvenView author publicationsYou can also search for this author in PubMed Google ScholarTobias J. ErbView author publicationsYou can also search for this author in PubMed Google ScholarNico J. ClaassensView author publicationsYou can also search for this author in PubMed Google ScholarContributionsA.B.-E., B.D., E.O. and N.J.C. designed and conceived the study. A.B.-E., N.J.C. and T.J.E. supervised the project. B.D. performed cloning, strain engineering and NGS analysis. B.D., S.Y. and H.S.-M. performed experimental characterizations. T.G. and B.D. performed the proteomic analysis. S.B.-V. and H.S.-M. conducted the modelling. B.D., E.O., T.J.E. and N.J.C. wrote the paper. The other authors provided feedback and approved the paper.Corresponding authorsCorrespondence to Beau Dronsella or Nico J. Claassens.Ethics declarationsCompeting interestsThe authors declare no competing interests.Peer reviewPeer review informationNature Microbiology thanks Kyle Lauersen and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.Additional informationPublisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.Extended dataExtended Data Fig. 1 The genetic integration workflow employed in this study.The two synthetic operons previously expressed from plasmids were genomically integrated via random Tn5-transposon integration with a library of different strength promoters in two consecutive rounds. The generated genetic libraries were grown on formate medium in 10 rounds to select for the fastest growing strains.Extended Data Fig. 2 Overview of strains and their phenotypes in batch cultivation.Growth of various C. necator strains used in this study on M9 minimal media supplemented with 80 mM formate and 100 mM bicarbonate with 10 % CO2 in the headspace. The doubling time in hours (h) of the strains is presented in the designated strain color. Curves depict the mean of 2–3 technical replicates and are representative of 3 experiments conducted in the same conditions to ensure reproducibility. Less than one doubling of the bacterial strains was considered no growth and indicated as such via an overlayed shaded box.Source dataExtended Data Fig. 3 Creation of CRG5 and CRG6 strains.Liquid selection test tube experiment on M9, 80 mM formate, 100 mM HCO3 minimal media in 10% CO2 for creation of the CRG5 gP14-C1 (a) and CRG6 gP3-C3 strains (b). Following transposon integration of module 1 or module 3 of the rGlyP and upon reaching early stationary phase cells were re-inoculated into fresh formate media. Dots represent the measured glass tube OD600 values of two biological replicates per promoter operon transposon insertion. The lines correspond to their averages. 4 passages are depicted and are indicative of the improved growth over the 10 passages.Source dataExtended Data Fig. 4 Characterization of CRG5 and CRG6 lineages with different promoters.Growth profiles of isolated single clones after the liquid selection experiment of (a) CRG5 and (b) CRG6 strains with different promoter C1 and C3-module transposon insertions on M9, 80 mM formate, 100 mM HCO3 minimal media in 10% CO2. Promoters P14, PPhaC1, P3, P4 and P2 were used to control the C1 operon, while the promoters Pcat, PPhaC1, P3, P4 were used to control the C3 operon. Two individually transposed lineages are shown per promotor, solid lines indicate the first clone and the dotted line the second. Doubling time in hours is given in the respective strain color beside the growth curve. Curves are means of 3 or 4 technical replicates for CRG5 and CRG6 respectively and representative of 3 experiments conducted in the same conditions. 1st clone strains CRG5 gP14-C1 and CRG6 gP3-C3 were the focus of this study and have the corresponding color shade associated with them in the main text figures.Source dataExtended Data Fig. 5 Transposon insertion loci of the genomic rGlyP modules in this study.(a) the CRG5 lineage carries two insertions of the Tn5-P14-C1 construct. One in the acnM gene and one upstream of the phaP1 gene. (b) the transposon expressing the P3-C3-module inserted in the gene H16_B1976 in strain CRG6.Extended Data Fig. 6 Proteomic comparison of rGlyP and CBB cycle strains.Global changes in the proteomes of (a) CRG4 and (b) CRG6 compared to the CBB cycle strain and of CRG6 to CRG4 (c). Log2-fold changes in protein intensity are plotted against significance (-log 10 q-Values). Relevant proteins are labeled, shown as a filled-out circles and colored according to their metabolic task. rGlyP proteins associated with module 1 are shown in blue, module 2 in green and module 3 in purple. CBB cycle and hydrogen utilization proteome are depicted in black and brown respectively. Different formate utilization proteins are highlighted in orange. Loci hit by transposon insertions and the frequently observed mutations as in H16_A0258 are shown in turquois. The most up and downregulated proteins if not already categorized, are colored in grey. All other proteins are not labeled, depicted as empty squares and colored grey.Source dataExtended Data Fig. 7 Effect of phaP1 deletion on growth of the CBB cycle strain.Strains were cultivated in M9 media supplemented with 80 mM formate and 100 mM HCO3 with 10 % CO2 in the headspace in a 96-well plate reader. Two biological clones of the ΔphaC1 ΔphaP1 strain are shown. Curves depict the mean of 4 technical replicates.Source dataExtended Data Fig. 8 Effect of sFDH inhibition on biomass yield (OD600).Depending on the presence of the activating or inhibiting metals molybdenum and tungsten, the sFDH pool can either be fully active (a), inactive (b), or partially inactive (c), when both are present in varying concentrations. (d) FBA-based prediction of the required oxidative and assimilatory fluxes for Calvin cycle (CBB) and reductive glycine pathway (rGlyP) at 14 h doubling time. (e), C. necator H16 CBB (WT), CBB (ΔphaC1), CRG4, CRG5 and CRG6 strains were grown on M9, 80 mM formate, 100 mM HCO3 minimal media in 96-well plates. Cells were either supplemented with varying concentrations of sodium tungstate to inhibit sFDH or sodium molybdate to potentially further increase sFDH activity with the appropriate metal ion.Source dataExtended Data Fig. 9 Chemostat cultivations of C. necator ΔphaC1 and CRG6 ΔdadA6 on formate.Strains were cultivated in a carbon limited chemostat at a dilution rate of 0.05 h−1 with a 80 mM formic acid feed in 10 % CO2 (at 2.5 vvm). Process data from the entire ΔphaC1 and CRG6 ΔdadA6 cultivations are shown. Infrared based cell density data and pH values are depicted. The 3 day harvest window during which samples were taken for cell dry weight analysis is indicated in grey. The dilution rate was set to 0.087 h−1 equal to 8 h doubling time after 1000 h of cultivation and to 0.1 h−1 after ~1300 h of cultivation, after which the cells washed out.Source dataSupplementary informationSupplementary InformationReporting SummaryPeer Review FileSource dataSource Data Fig. 1Source data for Fig. 1b.Source Data Fig. 2Source data for Fig. 2a–d.Source Data Extended Data Fig. 2Source data for Extended Data Fig. 2.Source Data Extended Data Fig. 3Source data for Extended Data Fig. 3.Source Data Extended Data Fig. 4Source data for Extended Data Fig. 4.Source Data Extended Data Fig. 6Source data for Extended Data Fig. 6.Source Data Extended Data Fig. 7Source data for Extended Data Fig. 7.Source Data Extended Data Fig. 8Source data for Extended Data Fig. 8.Source Data Extended Data Fig. 9Source data for Extended Data Fig. 9.Rights and permissionsOpen Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.Reprints and permissionsAbout this article