Engineering the cyanobacterial ATP-driven BCT1 bicarbonate transporter for functional targeting to C3 plant chloroplasts

Abstract The ATP-driven bicarbonate transporter 1 (BCT1) from Synechococcus is a four-component complex in the cyanobacterial CO2-concentrating mechanism. BCT1 could enhance photosynthetic CO2 assimilation in plant chloroplasts. However, directing its subunits (CmpA, CmpB, CmpC, and CmpD) to three chloroplast sub-compartments is highly complex. Investigating BCT1 integration into Nicotiana benthamiana chloroplasts revealed promising targeting strategies using transit peptides from the intermembrane space protein Tic22 for correct CmpA targeting, while the transit peptide of the chloroplastic ABCD2 transporter effectively targeted CmpB to the inner envelope membrane. CmpC and CmpD were targeted to the stroma by RecA and recruited to the inner envelope membrane by CmpB. Despite successful targeting, expression of this complex in CO2-dependent Escherichia coli failed to demonstrate bicarbonate uptake. We then used rational design and directed evolution to generate new BCT1 forms that were constitutively active. Several mutants were recovered, including a CmpCD fusion. Selected mutants were further characterized and stably expressed in Arabidopsis thaliana, but the transformed plants did not have higher carbon assimilation rates or decreased CO2 compensation points in mature leaves. While further analysis is required, this directed evolution and heterologous testing approach presents potential for iterative modification and assessment of CO2-concentrating mechanism components to improve plant photosynthesis.


Introduction
A crop improvement approach of ongoing global interest is the utilization of cyanobacterial and algal CO 2 -concentrating mechanisms (CCMs) to enhance photosynthetic performance through improved carbon fixation (Price et al., 2013;Long et al., 2016;Hennacy and Jonikas, 2020;Nguyen et al., 2024).Carboxylation of ribulose-1,5-bisphosphate by the bifunctional enzyme ribulose-1,5-bisphosphate carboxylase/ oxygenase (Rubisco) is a major limitation to efficient carbon acquisition by crops (Long et al., 2015).Cyanobacterial and algal CCMs, however, have evolved to actively accumulate bicarbonate (HCO 3 − ) within cellular compartments to supply high CO 2 concentrations to fast Rubisco enzymes for highly efficient carbon acquisition (Rae et al., 2017).A number of strategies exist for the engineering of a functional biophysical CCM in C 3 crop plants (Moroney et al., 2023), but crucial to all of these is a requirement to increase HCO 3 − concentration in the chloroplast stroma to supply either the native plant Rubisco, or an introduced Rubisco having a faster K cat , so that the CO 2 fixation reaction is optimized (Price et al., 2011;Rottet et al., 2021).
Cyanobacterial and algal CCMs utilize a suite of dedicated bicarbonate transporters that consume cellular energy to elevate HCO 3 − ion concentrations inside cellular membranes (Rae et al., 2017;Rottet et al., 2021), to levels up to 1000-fold higher than the external environment (Price et al., 2008).Since passive diffusion of HCO 3 − across membranes is very slow compared with CO 2 (Tolleter et al., 2017), a key to the function of the cyanobacterial CCM is that active bicarbonate pumping leads to the successful elevation of HCO 3 − inside cells.Then, specific carbonic anhydrase (CA) enzymes located with Rubisco interconvert the accumulated HCO 3 − to CO 2 , enabling a localized elevation of CO 2 for use by Rubisco (Moroney et al., 2023).Within crop-CCM strategies, the successful elevation of chloroplastic HCO 3 − concentrations via bicarbonate transporters alone is expected to provide increased photosynthetic output through provision of a net increase in CO 2 supply to Rubisco (Price, 2011;McGrath and Long, 2014;Wu et al., 2023).
To date, efforts to successfully express and deliver functional bicarbonate transporters to the correct location in plants have highlighted complexity with respect to protein targeting and function in crop systems (Pengelly et al., 2014;Atkinson et al., 2016;Rolland et al., 2016;Uehara et al., 2016Uehara et al., , 2020;;Nölke et al., 2019;Förster et al., 2023).Those studies predominantly addressed the use of relatively simple, single or dual gene bicarbonate pump systems (e.g.SbtA/B, BicA, LCIA, HLA3), as opposed to the more complex higher-order bicarbonate pumps and CO 2 -to-HCO 3 − conversion complexes found in native CCMs (Rottet et al., 2021).Despite these complexities, some higher-order bicarbonate pumps present desirable characteristics for HCO 3 − accumulation in the chloroplast stroma such as energization and no ion co-transport dependencies (Rottet et al., 2021).
Here we address the potential to make use of a relatively complex bicarbonate pump system, bicarbonate transporter 1 (BCT1), in the engineering of crop chloroplast CCMs.BCT1 is an ideal candidate for HCO 3 − accumulation in the chloroplast stroma, owing to its high affinity for bicarbonate, its ability to transport HCO 3 − against a concentration gradient, and because it is energized by ATP hydrolysis.In cyanobacteria, BCT1 is a low-inorganic carbon (Ci)-inducible ATPbinding cassette (ABC) transporter encoded by the cmpABCD operon under the control of the transcriptional regulator CmpR (Omata et al., 1999b(Omata et al., , 2001;;Nishimura et al., 2008;Pan et al., 2016).The operon gives rise to the expression of four protein components, CmpA, CmpB, CmpC, and CmpD, which occupy different locations associated with the cyanobacterial plasma membrane (Fig. 1A).The cmpABCD operon is found in both αand β-cyanobacterial species (Rae et al., 2011;Sandrini et al., 2014;Cabello-Yeves et al., 2022), and is therefore a ubiquitous element in cyanobacterial CCMs.
BCT1 is a high affinity transporter, exhibiting an apparent K m of 15 µM for HCO 3 − (Omata et al., 1999b), and is a multisubunit ABC transporter, closely related to the nitrate transporter NrtABCD (Omata, 1995;Klanchui et al., 2017).The substrate-binding protein component, CmpA, binds HCO 3 − with high affinity (K d =5 µM) and transfers it to the membrane transport complex (Maeda et al., 2000).The first 28 N-terminal residues of CmpA form a lipoprotein signal peptide, which when removed results in a functional soluble protein in Escherichia coli (Maeda et al., 2000).The signal peptidase II recognizes the cleavage site 26 LKGC 29 , which, following cleavage and removal of the lipoprotein, creates a covalent bond between CmpA and lipids via Cys 29 (Maeda and Omata, 1997;Tjalsma et al., 1999).CmpB is the transmembrane domain component of BCT1 and is likely to form a homodimer that functions as the channel for the transport of HCO 3 − across the plasma membrane (Omata et al., 2002).Finally, the nucleotide-binding domain (NBD) proteins, CmpC and CmpD, are likely to form a heterodimer that hydrolyses ATP to power the transport of HCO 3 − (Omata et al., 1999a;Smith et al., 2002).In cyanobacteria, both sit on the cytoplasmic side of the plasma membrane (Fig. 1A).CmpD is a canonical NBD containing highly conserved ATP binding motifs (i.e.Walker A, Walker B, ABC signature; Schneider and Hunke, 1998).In contrast, CmpC is a non-canonical NBD harbouring an additional C-terminal domain that is 50% similar to NrtA, homologous to CmpA, and thought to act as a solutebinding regulatory domain.
The engineering complexity of constructing a functional form of BCT1 in a crop plant chloroplast is evidenced by the requirement for each protein component of the BCT1 complex to be targeted to a specific sub-compartment of the chloroplast.Given the limited applicability of plastome transformation technologies across diverse crop species (Hanson et al., 2013), we here use a nuclear transformation approach, which has broader applicability (Fig. 1B).Our previous work demonstrated that unmodified BCT1 had no bicarbonate uptake activity when expressed in E. coli (Du et al., 2014).This highlights the potential requirement for regulatory systems that exist in cyanobacteria to modify BCT1 function, such as post-translational phosphorylation (Spät et al., 2021), suggesting the requirement of other factors external to the complex itself in order to control function.Moreover, CmpC regulatory domain function is not yet fully understood, potentially due to the absence of native regulatory mechanisms.Cyanobacterial bicarbonate transporters are subject to poorly understood allosteric controls that limit futile uptake in the dark (Price et al., 2013).In the absence of this information, it is desirable to evolve 'always-active' bicarbonate transporters that can be used in initial plant engineering approaches.
Here, we investigated strategies for targeting Synechococcus sp.PCC7942 BCT1 subunits to plant chloroplast locations and used directed evolution to obtain variants of BCT1 that were 'always-active' in E. coli.A synthetic biology approach that combined chloroplast sub-compartment targeting peptides with fluorescent reporter proteins was used to identify the best targeting systems for each BCT1 component.We also employed a directed evolution approach in a specialized E. coli strain that lacks CA (hereafter CAfree) and requires high levels of CO 2 for growth (Du et al., 2014;Desmarais et al., 2019;Förster et al., 2023).We were therefore able to control the function of a stand-alone BCT1 complex and eliminate regulatory requirements absent in heterologous systems.This resulted in the generation of constitutively active forms of BCT1 in E. coli.However, the expression of BCT1 in Arabidopsis did not result in the expected elevation in CO 2 supply to Rubisco.Although the tested BCT1 constructs did not exhibit functionality in Arabidopsis at this stage, our work has established a framework to assess correct protein targeting in N. benthamiana, activity in E. coli, and eventual functionality in plants.We have developed tools for assessing bicarbonate uptake activity in vivo in both E. coli and Arabidopsis.Moving forward, this process is likely to be iterative, with the next steps involving the evaluation of constructs generated through directed evolution to ascertain their targeting efficiency and expression levels in Arabidopsis.

Construction of BCT1 expression vectors
DNA plasmid constructs were produced using type IIS cloning strategies adapted from Golden Gate cloning and Loop Assembly (Engler et al., 2014;Pollak et al., 2019).BCT1 genes were amplified from Synechococcus sp.PCC7942 and domesticated to remove type IIS restriction sites.Primers were designed around the gene of interest with BpiI recognition sites and an appropriate 4 bp overhang (Supplementary Table S1).PCR was performed using Phusion High-Fidelity DNA Polymerase (Thermo Fisher Scientifc, USA), and the bands of desired sizes were gel-purified using Promega Wizard SV Gel and PCR Clean-Up System (Promega, USA).PCR fragments were assembled into the Universal Level 0 vector (pAGM9121) under cyclical digestion and ligation condition (37 °C for 3 min, 16 °C for 4 min for 25 cycles) followed by heat inactivation (50 °C for 5 min, 80 °C for 5 min).The same cyclical digestion and ligation condition with heat inactivation was performed when assembling Level − across the plasma membrane.First, HCO 3 − is captured by the substrate-binding protein CmpA and delivered to the membrane protein CmpB.HCO 3 − travels across the plasma membrane through the channel formed by a homodimer of the protein CmpB.CmpC and CmpD are nucleotidebinding proteins or ATPases that sit inside the cyanobacterial cell and hydrolyse ATP to provide the energy for the transport of HCO 3 − across the plasma membrane.Once in the cell, HCO 3 − diffuses into the carboxysome where it is converted into CO 2 by a carbonic anhydrase.(B) The strategy for the installation of the cyanobacterial BCT1 complex in the chloroplast envelope is based on nucleus-encoded CmpA, CmpB, CmpC, and CmpD.Each protein is individually targeted to the appropriate chloroplast sub-compartment using three different chloroplast transit peptides (cTP).CmpA is targeted to the intermembrane space (IMS).CmpB is sent to the inner envelope membrane (IEM), while CmpC and CmpD are targeted to the stroma.3-PGA, 3-phosphoglyceric acid; OEM, outer envelope membrane; RuBP, ribulose 1,5-bisphosphate; TIC, translocon on the inner chloroplast membrane; TOC, translocon on the outer chloroplast membrane.

Plant growth conditions
Nicotiana benthamiana plants used for infiltration were grown under 400 μmol photons m −2 s −1 light intensity, 60% relative humidity, a 16 h light/8 h dark photoperiod and 25 °C day/20 °C night temperatures.Only the first, second, and third true leaves from 4-to 5-week-old plants were kept for infiltration, while the rest were discarded.The plants were germinated and grown on pasteurized seed raising mix supplemented with 3 g l −1 Osmocote Exact Mini.

Agroinfiltration of Nicotiana benthamiana leaves
Constructs for BCT1 localization studies (Supplementary Table S2) were transiently expressed in 4-to 5-week-old N. benthamiana leaf tissue via Agrobacterium infiltration, as described previously (Rolland, 2018).Briefly, A. tumefaciens GV3101 (pMP90) was transformed with BCT1 constructs and grown in lysogeny broth (LB) medium supplemented with 25 µg ml -1 rifampicin, 50 µg ml -1 gentamycin, and 50 µg ml -1 kanamycin or 100 µg ml -1 spectinomycin for 24 h at 28 °C and 200 rpm.A vector encoding the tomato bushy stunt virus P19 protein was used to inhibit post-transcriptional gene silencing and to enable the expression of our constructs of interest (Roth et al., 2004).For each infiltration, p19 culture was mixed with each construct of interest at an OD 600 of 0.3 and 0.5, respectively.A p19-only control was prepared to an OD 600 of 0.8 as a negative control.All cells for infiltration were pelleted at 2150 g for 8 min and resuspended in 5 ml of infiltration solution (10 mM MES pH 5.6, 10 mM MgCl 2 , 150 µM acetosyringone).The solutions were incubated at room temperature for 2 h with occasional swirling, then infiltrated into the abaxial surface of 4-week-old N. benthamiana leaves using a 1 ml slip tip syringe.Infiltrated plants were grown for another 3 d before protein expression was assessed via confocal microscopy.
Arabidopsis plants were transformed using the method described by Zhang et al. (2006).A 5 ml starter culture of A. tumefaciens was grown in LB medium with antibiotics overnight at 28 °C.This starter culture was used the following morning to propagate a larger 250 ml A. tumefaciens culture overnight at 28 °C.The next day, the cells were harvested by centrifugation at 4000 g for 10 min.The pelleted cells were resuspended in freshly prepared 5% (w/v) sucrose solution with 0.02% (v/v) of Silwet L-77.The resuspended cultures were generously applied to the Arabidopsis flower buds using transfer pipettes.Afterwards, the plants were placed sideways into the trays and were covered and allowed to recover in darkness overnight.Following recovery, the plants were grown in 21 °C in continuous light.Mature seeds were collected from plants, and positive transformants were selected on soil by spraying seedlings with a 1:2000 dilution of BASTA (AgrEvo, Berlin, Germany).The presence of the transgene was also confirmed via gene-specific PCR for cmpA with the primer pair CmpA-F1 and CmpA-R1 and for the bar gene with the primer pair Basta-F and Basta-R (Supplementary Table S1).DNA was extracted using the protocol described by Edwards et al. (1991).For this about 20 mg of plant tissue was ground using micropestles in 1.5 ml centrifuge tubes.These were further macerated in 400 µl of extraction buffer [200 mM Tris-HCl pH 7.5, 250 mM NaCl, 25 mM EDTA, 0.5% (w/v) SDS].The samples were then centrifuged at 13 000 g for 5 min, and the supernatant was collected into a new tube.An equal volume (~400 µl) of isopropanol was added and mixed with the supernatant.The resulting mixture was again centrifuged at 13 000 g for 5 min.The resulting supernatant was discarded afterwards, and the pellet was allowed to air dry.After drying, the pellet was dissolved in 50 µl of 1× TE buffer (10 mM Tris-HCl, 1 mM Na 2 EDTA, pH 8.0) and was used for subsequent confirmation of transformation.

Bacterial strains and growth conditions
Escherichia coli CAfree strain, kindly provided by Dave Savage (Desmarais et al., 2019), was used for directed evolution and complementation assay.Escherichia coli DH5α strain was used for cloning and protein expression for immobilized metal affinity chromatography (IMAC) purification.Unless otherwise stated, bacteria were grown at 37 °C in LB medium (10 g l −1 tryptone, 10 g l −1 NaCl, and 5 g l −1 yeast extract), supplemented with 15 g l −1 agar for solid medium on plates.For culturing transformants with spectinomycin, ampicillin, and kanamycin resistant genes, the medium was supplemented with 100 µg ml -1 , 100 µg ml -1 , and 50 µg ml -1 of the antibiotics, respectively.

Directed evolution of BCT1 in carbonic anhydrase-free E. coli
By controlling the CO 2 supply, we determined that 0.85% (v/v) CO 2 allowed CAfree to survive for extended periods in liquid culture, providing an opportunity for random mutations in the BCT1 plasmid to confer growth advantages.
A starter culture of CAfree harbouring unmodified BCT1 genes in pEven1 backbone (GM186) was prepared by growing the cells from a glycerol stock in LB medium supplemented with 100 µg ml -1 spectinomycin at 37 °C in the presence of 4% CO 2 for approximately 18 h.This starter culture was then diluted 100 µl into 5 ml of liquid medium consisting of M9 minimal medium supplemented with 1% LB, 100 µg ml -1 spectinomycin, and 20 µM isopropyl β-D-1-thiogalactopyranoside (IPTG).The cultures were incubated at 37 °C with agitation at 120 rpm under a 0.85% CO 2 atmosphere.Regular subculturing in fresh medium was performed while maintaining permissive CO 2 conditions, until the cultures were able to fully grow overnight.
The overnight culture was diluted 100-fold and placed under ambient CO 2 conditions (0.04%).Regular subculturing was again performed until the cultures reached a dense overnight growth.From the cultures that grew under ambient CO 2 , 100 µl was plated on solid LB medium supplemented with 100 µg ml -1 spectinomycin and 100 µM IPTG, and incubated at ambient CO 2 for 18 h.Eight single colonies were selected and cultured, and their plasmid DNA was extracted using the QIAprep Spin Miniprep Kit (Qiagen).The pDNA from these colonies was pooled together and used to retransform new CAfree cells.The transformed cells were then plated onto LB agar supplemented with 100 µg ml -1 spectinomycin and 100 µM IPTG and incubated at ambient CO 2 for 18 h.Growth confirmed that the mutation(s) conferring the advantage at air levels of CO 2 was carried by the plasmid and not in the genome of the CAfree strain.Twelve colonies were selected from this plate, and the pDNA was isolated from each colony for DNA sequencing (Macrogen Inc., Seoul, South Korea).
For the isolation of the strain harbouring the CmpCD chimera, cells were subcultured four times at 0.85% CO 2 and three times at air levels of CO 2 before it grew overnight.For the isolation of the strain harbouring the CmpCD fusion, cells were subcultured six times at 0.85% CO 2 before it grew overnight at air levels of CO 2 .

Complementation assay in carbonic anhydrase-free E. coli
We developed a high-throughput complementation assay to rapidly assess BCT1 function in CAfree.This involved cultivating liquid cultures at 4% CO 2 for 6 h at 37 °C, spotting 5 µl onto four plates of LB agar with or without 0.1 mM IPTG, and incubating them overnight in selective (air, 0.04% CO 2 ) or permissive (4% CO 2 ) conditions.To mitigate the negative growth effects caused by BCT1 overexpression, a modified plasmid backbone with lower copy number was employed in this assay (pFA-Odd and pFA-Even).After the overnight growth, the plates were imaged with a Bio-Rad ChemiDoc XRS+ imaging system (Bio-Rad Laboratories, USA) under white epifluorescence.

Bicarbonate uptake assay in E. coli
Inorganic carbon uptake assays were carried out as described by Förster et al. (2023) with some modifications.The assays were performed in CAfree, and the cultures were induced with 100 µM IPTG.The assay buffer used consisted of 20 mM bis-tris propane-H 2 SO 4 supplemented with 0.5 mM glucose and 1 µM CaCl 2 with a pH of 7.5.
To prepare the cells for the assay, they were first washed twice with the assay buffer.Subsequently, the cells were incubated for 10 min before undergoing a third round of washing with the assay buffer.Following the washing steps, the assay was performed according to the protocol outlined in Förster et al. (2023).This method allowed us to measure the rates of bicarbonate uptake and determine kinetic parameters such as the Michaelis constant (K M ) and maximum velocity (V MAX ) for bicarbonate transport.

Protein induction and immobilized metal affinity chromatography purification
Overnight cultures from glycerol stocks were used to inoculate 40 ml of LB medium supplemented with 100 µg ml -1 spectinomycin to an OD 600 of 0.1-0.2.Cultures were grown at 37 °C until OD 600 reached 0.4-0.6.
To induce protein expression, IPTG was added to a final concentration of 50 µM.Cultures were returned to grow at either 37 °C for 2-3 h or 28 °C for 4-5 h.To prepare the cells for IMAC purification, the OD 600 of each culture was measured and used to normalize the number of cells to pellet.The cell pellets were then harvested by centrifugation at 4800 g for 10 min at 4 °C and subsequently stored at −20 °C until further use.

Physiological measurements on Arabidopsis
Images of different Arabidopsis genotypes were taken weekly, and rosette areas were measured as pixel area using the PhenoImage software (Zhu et al., 2021) and Fiji ImageJ (Schindelin et al., 2012).Rosette areas were measured on six plants per line.Four plants were later harvested for measuring fresh weights.Measurements of photosynthetic parameters were conducted using a LI-COR LI-6800 system (LI-COR, USA).Plants grown for rosette area and biomass measurements were also used for photosynthesis measurements.CO 2 response curves for assimilation (A; µmol m⁻² s⁻¹) in response to intercellular CO 2 (C i ; µmol mol⁻ 1 ) curves were generated from 50 to 1700 µmol mol⁻ 1 ambient CO 2 .A light input of 1000 µmol photons m⁻² s⁻¹ was used for generating CO 2 response curves.This light input was determined to be saturating when tested via the generation of light response curves.The initial linear portion of the CO 2 response curves (points that had R 2 values from 0.95 to 0.99, 50 to 200 µmol mol −1 CO 2 ) were used to determine CO 2 compensation points.CO 2 compensation points were determined from the x-intercept of the linear regressions created from the response curves.

Immunoblots of Arabidopsis extracts
Frozen leaf material was ground to fine paste in microcentrifuge tubes.Lysis buffer (62.5 mM Tris, 300 mM sucrose, 10 mM DTT, 2% lithium dodecyl sulphate, 50 mM EDTA, pH 6.8) was added, and the leaf material was resuspended.The suspension was then centrifuged at 12 000 g for 10 min at 4 °C.The supernatant was then used to measure chlorophyll as previously described (Porra et al., 1989).Three microgram chlorophyllequivalent protein samples were prepared by mixing with loading buffer and incubating at 65 °C for 10 min.The samples were then loaded on a 4-20% mPAGE gel (Merck) with MOPS buffer as the running buffer.Induced E. coli lysate was used to serve as a positive control for immunoblots later.The proteins were then transferred to PVDF membranes (Amersham Hybond, GE Healthcare Life Sciences).The membranes were blocked in 5% milk in TBS-T for 1 h, followed by an overnight incubation at 4 °C with primary antibodies.The primary antibodies used were: α-FLAG (Merck F3165, 1:1200), α-Myc (Merck M4439, 1:6000), α-HA (Merck H9658, 1:6000).Following primary antibody incubation and washing, membranes were incubated with conjugated secondary antibody (Cell Signaling Technology anti-mouse IgG, horseradish peroxidase-linked antibody no.7076, 1:5000) for 1 h.Chemiluminescence detection was performed using ECL reagent (Bio-Rad, Clarity Western ECL Substrate) and images were taken using a ChemiDoc XRS+ imaging system.

Data visualization and statistical analysis
Growth and physiological parameter data were initially visualized using the R software environment and the ggplot2 package (Wickham, 2016;R Core Team, 2019).One-way ANOVA followed by Tukey's multiple comparisons test was performed using GraphPad Prism version 10.0.3 for Windows (GraphPad Software, USA).

Individual targeting of BCT1 components to the chloroplast
To determine the optimal route for installing BCT1 subunits to the correct location in chloroplasts, we employed a transient expression approach in Nicotiana benthamiana combined with fluorescent reporter constructs and confocal microscopy.Targeting foreign proteins to specific chloroplast subcompartments is a significant engineering challenge as there are at least six sub-compartments [i.e.outer envelope membrane (OEM), intermembrane space (IMS), inner envelope membrane (IEM), stroma, thylakoid membrane, and thylakoid lumen; Rolland et al., 2017].Specifically, we targeted nucleusencoded CmpA, CmpB, CmpC, and CmpD individually to the chloroplast IMS, IEM, or stroma using a variety of chloroplast transit peptides (cTP; Fig. 1B).
To date, the targeting of only two IMS proteins, Tic22 and MGD1, has been studied (Kouranov et al., 1999;Vojta et al., 2007;Chuang et al., 2021).While AtMGD1 cTP targeted CmpA to the stroma, Tic22 isoforms from Arabidopsis and Pisum sativum proved more successful in targeting CmpA to the IMS (Supplementary Fig. S1).Notably, the first 64 residues of the protein AtTic22-IV appeared to efficiently target CmpA to the IMS in N. benthamiana (AtTic22-IV 64 -CmpA, Fig. 2).PsTic22 68 and AtTic22-III 69 also exhibited potential in targeting CmpA to the IMS, albeit with lower efficiency, resulting in some accumulation of fluorescence signal in the cytosol (Supplementary Fig. S1).However, the resolution provided by confocal microscopy does not definitively confirm IMS localization for CmpA.Nonetheless, as depicted in Fig. 2, the co-localization of CmpA with the IEM marker AtTGD2 supports an IMS localization, given that CmpA is not inherently a membrane protein.Future investigations should encompass biochemical localization techniques, such as dual protease assays and chloroplast fractionation, to validate the localization of CmpA, ideally in stable transformants.
In cyanobacteria, CmpC and CmpD are cytoplasmic NBD components of the BCT1 complex and are expected to bind transiently to their membrane anchor, CmpB.As a result, in a chloroplastic CCM, targeting of CmpC and CmpD to the IEM is unnecessary.Instead, we attempted to target them to the chloroplast stroma.To achieve this, we employed the wellestablished stromal targeting sequence from AtRecA (Köhler et al., 1997).CmpC and CmpD were both efficiently targeted to the stroma after 3 d post-infiltration (AtRecA 68 -CmpD and AtRecA 54 -CmpC, Fig. 2).

Recruitment of CmpC and CmpD to the inner envelope membrane by CmpB
To determine whether CmpB is properly oriented in the membrane to interact with its stromal NBDs, a strategy involving co-expression of individual NBD with CmpB in N. benthamiana was employed.CmpC and CmpD were tagged with fluorescent reporters, while CmpB carried a small nonfluorescent label (HA-H 6 ) to reduce potential interference (Fig. 3A).Confocal microscopy was used to track NBD localization and detect a shift from the stroma to the IEM.
AtRecA 54 -CmpC (GL418) relocalized from the stroma to the IEM when co-expressed with AtABCD2 97 -CmpB (Fig. 3B, row 2).The removal of the regulatory domain in CmpC (AtRecA 68 -CmpC 263 , GL371) often resulted in aggregates in the stroma when expressed alone.When co-expressed with AtABCD2 97 -CmpB, AtRecA 68 -CmpC 263 was also recruited to the IEM (Fig. 3B, row 4).While AtRecA 68 -CmpD (GL372) alone was targeted to the stroma, it was successfully recruited to the IEM when co-expressed with AtABCD2 97 -CmpB (GL239; Fig. 3B, row 6).Successful recruitment of the two NBDs suggests that AtABCD2 97 -CmpB not only sits in the chloroplast IEM but is also in the correct orientation to allow appropriate protein-protein interactions with stromal CmpC and CmpD.
Considering the limited understanding of membrane protein orientation determinants, we utilized our system to explore the influence of various targeting sequences on the orientation of CmpB in the membrane.Although some targeting sequences were less efficient in delivering CmpB to the IEM, they did not affect its orientation.All seven tested targeting sequences for CmpB triggered the relocalization of AtRecA 54 -CmpC 263 (GL199) to the IEM (Supplementary Fig. S3).In contrast, the control construct lacking a targeting sequence for CmpB (GL234, no signal peptide) did not induce the shift of CmpC 263 from the stroma to the IEM (Supplementary Fig. S3).

Generation of active BCT1 mutants by rational design
Previous results showed that unmodified BCT1 is inactive in E. coli (Du et al., 2014).To address this, we initially removed putative regulatory requirements of BCT1 by rational design.For this purpose, we used a loop assembly (Pollak et al., 2019) approach, enabling high throughput design and construction of flexible linkers, point mutations, and domain deletions (Fig. 4).
Since we hypothesized the lack of BCT1 function in heterologous systems may be due to the absence of regulatory mechanisms present in cyanobacteria, an obvious rational design approach was to remove the regulatory domain of CmpC (Fig. 4C).CmpC is a 663-residue protein, of which only 263 residues fold into a canonical NBD (Supplementary Fig. S4).
The additional C-terminal domain is thought to be involved in BCT1 regulation (Omata et al., 2002;Koropatkin et al., 2006).We therefore generated a construct that only encoded the first 263 residues of CmpC, namely CmpC 263 .
In prokaryotes, the two transmembrane domains and two NBDs of ABC transporters are often encoded by separate genes, while in eukaryotes, these domains are typically connected by linker region(s) to form so-called 'full-' or 'halftransporters' (Theodoulou and Kerr, 2015;Ford et al., 2019).Half-transporter fusions of CmpB with CmpC (hereafter CmpBC) and CmpB with CmpD (hereafter CmpBD) were generated using flexible linkers of approx.40 residues (Fig. 4E).This should ensure domain assembly when expressed in more complex heterologous systems and reduce targeting complexity (Ford et al., 2019).
In ABC transporters, it is accepted that ATP hydrolysis is carried out by the NBDs and that a glutamate-to-glutamine substitution in the conserved Walker B motif causes ATP hydrolysis deficiency (Orelle et al., 2003).A putatively inactive BCT1 mutant was created as a negative control (Fig. 4F) by mutating the catalytic glutamate in both CmpC E164Q and CmpD E179Q .

Generation of active BCT1 mutants by directed evolution
We also employed a directed evolution approach to evolve functional forms of BCT1.A genetic construct was engineered to closely mimic the native BCT1 operon except for its promoter (GM186, Supplementary Table S2).This BCT1 construct was transformed into a specialized E. coli strain lacking CAs (CAfree; Desmarais et al., 2019).This strain only grows under high levels of CO 2 or in the presence of a functional bicarbonate transporter or CA (Du et al., 2014;Förster et al., 2023).By controlling the CO 2 supply, we determined that a 0.85% (v/v) CO 2 allowed CAfree to survive for extended periods in liquid culture with slow growth, providing an opportunity for random mutations in the BCT1 plasmid to confer growth advantages.Upon improved growth, cells were transferred to air levels of CO 2 to increase selection pressure and select functional mutants.The culture was further incubated until it exhibited consistent overnight growth.The duration of the entire process varied from days to weeks.BCT1 plasmids were isolated from single colonies, sequenced and retransformed into CAfree to confirm the mutations were responsible for the observed growth.This directed evolution approach led to the generation of two distinct BCT1 mutants (Fig. 4G, H).In the first, the deletion of the last 450 residues of CmpC (including the regulatory domain) and the first 240 residues of CmpD resulted in a CmpCD chimera of 263 residues (29 kDa).This mutant also harboured a point mutation in the non-coding intergenic sequence between cmpA and cmpB.In the second, the deletion of the intergenic space between cmpC and cmpD produced a CmpCD fusion of 942 residues (105 kDa) that maintained the integrity of both CmpC and CmpD.This mutant also harboured a point mutation in the regulatory domain of CmpC H409Q .
High-throughput screening of BCT1 mutants in carbonic anhydrase-free E. coli A high-throughput complementation plate assay was used to rapidly assess BCT1 function of rational design and directed evolution mutants in CAfree.We screened 72 genetic constructs of BCT1 (Supplementary Table S2), of which 14 are shown in Fig. 5, with each construct identified by a unique identification number.Initially, we confirmed that the unmodified BCT1 construct (GN18) failed to complement CAfree at ambient CO 2 (0.04%, Fig. 5).We also evaluated our rational designs, including without a regulatory domain (GN109), phosphorylation mimic (GN113), and half-transporter (GN135).None of these designs supported growth at air levels of CO 2 (Fig. 5).However, removing CmpC regulatory domain in our halftransporter design enabled growth at ambient CO 2 (GN138; Fig. 5).The addition of small epitope tags to this improved half-transporter design still allowed partial growth at air levels of CO 2 (GN133; Fig. 5).
Among the selected BCT1 constructs, two directed evolution mutants [CmpCD chimera (GN19) and CmpCD fusion (GN128)], exhibited successful complementation of CAfree at air levels of CO 2 .Given its robust complementation ability, we focused our efforts on the CmpCD fusion construct (also containing the H409Q mutation in the regulatory domain of CmpC; Fig. 4H).First, we demonstrated that the complementation depends on BCT1's ability to hydrolyse ATP by mutating the catalytic glutamate in CmpC E164Q and CmpD E179Q .The ATPase deficient CmpCD fusion failed to complement CAfree (GM322; Fig. 5).Second, we teased apart the influence of the fusion event (between cmpC and cmpD) and the H409Q mutation in the regulatory domain of CmpC.When the residue Q409 was mutated back into a histidine, the resulting fusion construct failed to complement CAfree (GM321; Fig. 5).But when a stop codon and an intergenic space were reintroduced between cmpC H409Q and cmpD, the resulting construct weakly complemented CAfree (GN130; Fig. 5).Finally, we looked at the influence of epitope tags on the CmpCD fusion revealing that while the addition of a tag on CmpA and/or CmpCD fusion had little impact on BCT1 function (GM310, GM319, Fig. 5; GM315, GM317, Supplementary Table S2), a C-terminal tag on CmpB always resulted in a loss of function (GN129, Fig. 5; GM316, GM318, GM320, Supplementary Table S2).

Functional analysis of selected BCT1 mutants in E. coli
To gain insights into the functional properties of some BCT1 mutants, we conducted H 14 CO 3 − uptake assays in E. coli as described by Förster et al. (2023).Bicarbonate uptake rates were measured for a subset of seven genetic constructs (Fig. 6A), with the CmpCD fusion exhibiting the highest uptake rate (GN128, 104.2 ± 4.6 nmol OD 600 −1 h −1 ).The addition of a myc tag on CmpA and an mCitrine tag on CmpCD led to a 1.5-fold decrease in uptake rate (GM319, 69.3 ± 10.1 nmol OD 600 −1 h −1 ).Furthermore, replacing mCitrine with HA-H 6 on CmpCD resulted in a total loss of activity (GM310, 5.5 ± 1.6 nmol OD 600 −1 h −1 ).While performance appeared to be slightly better in the improved half-transporter design (GN138, 22.5 ± 9.3 nmol OD 600 −1 h −1 ), the results were not significantly different from the unmodified BCT1 (GN18, 10.8 ± 4 nmol OD 600 −1 h −1 ).The addition of tags reduced the transporter's activity (GN133, 10 ± 2.3 nmol OD 600 −1 h −1 ) to the same negligible level observed with the unmodified BCT1.The CmpCD chimera also exhibited a negligible uptake rate (GN19, 13.4 ± 2.8 nmol OD 600 −1 h −1 ).The negligible uptake rates observed for GN19, GN138, GN133, and GM310 show how sensitive our high throughput screening system is.It suggests that CAfree require very little Ci to survive at reduced CO 2 levels.The kinetic constants were determined for the two best constructs (GN128, GM319) and a low-performing construct (GN19) as a comparison.This revealed a bicarbonate affinity of approximately 150 µM for both the CmpCD fusion with tags (GM319; K M =0.17 ± 0.03 mM) and the CmpCD fusion without tags (GN128; K M =0.12 ± 0.02 mM; Fig. 6B).
We also explored the assembly of the BCT1 complex in E. coli.To facilitate this assessment, each BCT1 protein was tagged with a small epitope (Fig. 6C).CmpC, the bait protein, was purified by virtue of its C-terminal hexa-histidine tag using IMAC, with the expectation that interacting proteins (prey) would co-purify.A negative control involved using a BCT1 construct with identical tags, except for the absence of the hexa-histidine tag on CmpC (GM336).This control confirmed the effectiveness of the column washes, as no signal was detected in the eluate fraction for GM336 (Fig. 6D).The IMAC pull-down was then repeated with three different BCT1 constructs.The eluate of the unmodified BCT1 (GM337) contained all four proteins, indicating that the presence of a tag on CmpB does not obstruct transporter assembly.The NBD-only construct (GM339) revealed that CmpC and CmpD can directly interact without necessitating CmpB to form a heterodimer.Lastly, in GM341, where CmpC is fused with CmpD, the interaction with CmpB persisted, with both CmpB and CmpA detected in the eluate fraction.
To further our analysis, the two half-transporter constructs (GN64, GN65), and CmpCD fusion (GN139) were introduced into WT Arabidopsis.These plants were then grown on air levels of CO 2 (400 ppm) or low CO 2 (200 ppm) to determine whether the BCT1 constructs might enhance growth.None of these constructs enhanced growth of WT Arabidopsis when grown at these CO 2 levels (Fig. 8; Supplementary Fig. S6).In addition, the mature leaves of the transformed plants displayed lower or similar CO 2 assimilation rates (A/C i curves; CO 2 assimilation rate as a function of intercellular CO 2 ) as compared with WT, with CO 2 compensation points unchanged or higher than WT (Supplementary Table S3).On the left-hand side, cultures were plated in 5 µl spots on LB agar containing 0 or 100 µM isopropyl β-d-1-thiogalactopyranoside (IPTG) and incubated overnight at 37 °C in high (4%) or ambient (0.04%) CO 2 .Successful complementation was achieved when the induced cells (100 µM IPTG) were able to grow at ambient CO 2 (as observed in the last column).While unmodified BCT1 (GN18) was inactive, seven out of 13 mutants were able to complement CAfree to different extents at ambient levels of CO 2 (GN138, GN133, GN19, GN128, GN130, GM310, GM319).The corresponding schematic representation (see Fig. 4) to which each plasmid relates to or derives from (indicated by an apostrophe) is presented on the far left as the panel letter from Fig. 4. The black stars represent point mutations that are labelled, unless falling into a non-coding region (e.g.mutation between cmpA and cmpB in GN19), to show the change in residues (e.g.H409Q in GN128).
BCT1 protein expression in transformed Arabidopsis was evaluated using immunoblots, with antibodies targeting the epitope tags appended to each subunit (see Supplementary Fig. S5).As illustrated in Supplementary Fig. S7, all subunits were detected in plants transformed with GN64.However, plants transformed with GN65 or GN139 exhibited minimal protein expression.The underlying reasons for the low protein levels observed in plants carrying GN65 or GN139 remain unclear, but this phenomenon was observed in both βca5 and WT plants.

Discussion
In this study, we present compelling evidence supporting (i) the precise subcellular targeting, and (ii) the functional evolution of a complex cyanobacterial bicarbonate transporter.Our primary objective was to introduce a functional Ci transporter into plants, aiming to enhance CO 2 assimilation in C 3 crops (Price et al., 2013).Previous research in this field predominantly focused on simpler single or dual-gene bicarbonate pump systems, often encountering difficulties related to targeting or additional ion requirements for function (Pengelly et al., 2014;Atkinson et al., 2016;Rolland et al., 2016;Uehara et al., 2016Uehara et al., , 2020;;Nölke et al., 2019;Rottet et al., 2021;Förster et al., 2023).The successful integration of the Chlamydomonas passive channel LCIA into C 3 plant chloroplasts was previously accomplished; however, this transporter's inherent characteristics as a passive channel limit its capacity for high-rate bicarbonate transport (Atkinson et al., 2016;Nölke et al., 2019;Förster et al., 2023).In this work, we addressed two significant challenges that researchers have noted in previous reports.The first challenge is to correctly target bicarbonate transporters in plant chloroplasts.The second challenge is to demonstrate the functionality of the transporter in a heterologous system, in this case in E. coli.Notably, we directed the ABC transporter BCT1 to the chloroplast envelope, a complex task given its four subunits, each needing precise localization (Fig. 1).BCT1 was previously reported to be inactive in E. coli, potentially due to unknown regulatory mechanisms likely present in its native cyanobacterial cellular environment (Du et al., 2014).To remove regulatory requirements, BCT1 was engineered, and its functionality assessed in a specialized E. coli strain.Despite these complexities, BCT1 possesses favourable attributes, including a high affinity for bicarbonate and the reliance on ATP as its sole power source (Omata et al., 1999a, b), eliminating the need for co-transported ions, as is the case for the single gene transporters SbtA and BicA (Price et al., 2004(Price et al., , 2008)).
The ability to import nuclear-encoded proteins into chloroplasts has a broad application to the majority of globally important crops.The assembly of a multi-protein membrane complex in a heterologous system is a significant engineering challenge.It requires the components to be co-localized and for the membrane proteins to be inserted in the correct orientation (Wojcik and Kriechbaumer, 2021).Factors such as stoichiometry and chaperones may also have to be considered (Barrera et al., 2009;Bae et al., 2013;Hallworth et al., 2013;Thornell and Bevensee, 2015).We found that the BCT1 complex assembled in E. coli (Fig. 6).While co-immunoprecipitation was not performed in plants, a critical observation made in N. benthamiana was the recruitment of CmpC and CmpD to the chloroplast IEM when co-expressed with CmpB, which suggests that CmpB is oriented correctly in the membrane irrespective of which leader sequences was used (Fig. 3; Supplementary Fig. S3).This is not only essential for the complex formation but also guarantees the intended direction of transport.Notably, we observed that not all leader sequences were equally effective at targeting BCT1 component proteins to the correct locations within the chloroplast.For example, AtMGD1 failed to target CmpA to the IMS (Supplementary Fig. S1).As more leader sequences become available, our toolkit for subcellular targeting will expand, and the use of modular cloning will enable rapid screening of additional sequences.
Initially, native BCT1 was inactive in E. coli (Figs 5, 6) (Du et al., 2014) and so was its codon-optimized version.We hypothesized the lack of function was due to the absence of regulatory factors in heterologous systems (e.g.specific activation kinases; Spät et al., 2021).To overcome this problem, we used two approaches.Logical changes were made to the proteins by rational design, and directed evolution was employed to evolve active forms of BCT1 (Figs 4,5).Directed evolution led to large changes such as the fusion of the two NBDs in a CmpCD fusion.With rational design, we explored the fusion of the transmembrane domain with each NBDs in the CmpBC and CmpBD half-transporter design (Theodoulou and Kerr, 2015;Ford et al., 2019).In both approaches we obtained some level of activity, suggesting that subunit stoichiometry plays an important role in the functionality of BCT1, as protein fusion likely altered the CmpB:CmpC/D ratio.We also hypothesized that eliminating the CmpC regulatory domain could produce an active transporter.While this rationally designed form, CmpC 263 , did not show the predicted activity, directed evolution produced a CmpCD chimera that had measurable activity in the absence of this regulatory domain (Fig. 5).Additionally, a CmpCD fusion, which was the best-performing mutant, harboured a point mutation in the regulatory domain of CmpC H409Q .This mutation played a more significant role than the fusion event itself.However, both the mutation and the fusion were found to be necessary for achieving maximal activity of the transporter.A multiple sequence alignment (Supplementary Fig. S8) showed that residue H409 in CmpC corresponds to putative ligand-binding residues in NrtA (H196) and CmpA (Q198; Koropatkin et al., 2006Koropatkin et al., , 2007)).Considering this, we speculate that the H409Q mutation might interfere with ligand binding in some manner.Further research is needed to understand the role of H409, but with eight potential binding sites identified in CmpA and NrtA, we predict there are still many unexplored rational designs that could lead to an improved functionality of BCT1.
Based on functional modification of BCT1 through rational design and directed evolution, and the ability to successfully target BCT1 components to their destinations within the chloroplasts, we generated transgenic Arabidopsis lines expressing several modified BCT1 constructs (Supplementary Fig. S5).Notably, none of these, either in βca5 mutant (Fig. 7) or in WT plants (Fig. 8), displayed phenotypes consistent with bicarbonate uptake into the chloroplast.Also, the expected decrease in CO 2 compensation point was not apparent (Supplementary Table S3).Bicarbonate uptake into the chloroplast should enhance chloroplastic CO 2 concentrations, elevating Rubisco carboxylation even at low ambient CO 2 supply (Price et al., 2011).The lack of a CO 2 compensation point reduction in our BCT1 lines and the failure of these constructs to enhance the growth of plants indicate that BCT1 is not significantly changing chloroplast Ci uptake in these plants.
While all three BCT1 subunits were detected in plants transformed with the CmpBC/CmpBD half transporter (GN64), we did not expect it to be active in plants given its poor functionality in E. coli (see GN135, Fig. 5).In E. coli, the presence of the CmpC regulatory domain appeared to hinder the activation of BCT1, as did the addition of epitope tags (Figs 5,6).The absence of growth or photosynthesis enhancement in both βca5 and WT plants by the CmpBC 263 /CmpBD half transporter (GN65) or the CmpCD fusion (GN139) is consistent with their low expression levels in Arabidopsis (Supplementary Fig. S7).The reasons behind the observed low protein levels in plants remain unclear.Perhaps some modifications, such as the removal of the regulatory domain, affect either the expression or stability of BCT1 in Arabidopsis.To enable the screening of a large number of constructs, localization analysis was carried out entirely in N. benthamiana, but subtleties of cTP requirements could result in mis-localization or poor targeting in Arabidopsis.The focus of future work will need to be expanded to include localization and improved expression specifically tailored for Arabidopsis.
We hypothesise that further evolution and refinement of function of BCT1 in the CAfree E. coli system may also be required to deliver improved function in planta.Notably, the large sequence changes observed using directed evolution in this study, and the similarity between evolved outcomes and some of the rational designs, highlights two things.First, that well-considered rational design approaches using known variation in evolution of ABC transporter systems (e.g.halftransporter protein fusion arrangement) is a valid approach to modify this type of transporter.Second, our directed evolution approach enabled the generation of large and unexpected changes in sequence length and gene fusion that would have not been found in the screen of a sequence variant library.We are therefore encouraged that a combination of rational design and directed evolution of both existing chloroplast membrane proteins and bacterial bicarbonate uptake systems will allow significant progress in enabling the elevation of chloroplastic Ci using synthetic biology tools.In combination with high throughput DNA assembly technologies and plant-based platforms that enable functional testing, we expect significant progress toward this goal.

Supplementary data
The following supplementary data are available at JXB online.Table S1.List of primers used in this study.
Table S2.List of constructs used in this study.Table S3.CO 2 compensation points for BCT1 transformants in WT Arabidopsis.

Fig. 1 .
Fig. 1.Structure of bicarbonate transporter 1 (BCT1) and strategy for its installation in the chloroplast envelopes.(A) In cyanobacteria, BCT1 transports HCO 3− across the plasma membrane.First, HCO 3 − is captured by the substrate-binding protein CmpA and delivered to the membrane protein CmpB.HCO 3 − travels across the plasma membrane through the channel formed by a homodimer of the protein CmpB.CmpC and CmpD are nucleotidebinding proteins or ATPases that sit inside the cyanobacterial cell and hydrolyse ATP to provide the energy for the transport of HCO 3 − across the plasma membrane.Once in the cell, HCO 3 − diffuses into the carboxysome where it is converted into CO 2 by a carbonic anhydrase.(B) The strategy for the installation of the cyanobacterial BCT1 complex in the chloroplast envelope is based on nucleus-encoded CmpA, CmpB, CmpC, and CmpD.Each protein is individually targeted to the appropriate chloroplast sub-compartment using three different chloroplast transit peptides (cTP).CmpA is targeted to the intermembrane space (IMS).CmpB is sent to the inner envelope membrane (IEM), while CmpC and CmpD are targeted to the stroma.3-PGA, 3-phosphoglyceric acid; OEM, outer envelope membrane; RuBP, ribulose 1,5-bisphosphate; TIC, translocon on the inner chloroplast membrane; TOC, translocon on the outer chloroplast membrane.

Fig. 2 .
Fig. 2. Individual targeting of CmpA, CmpB, CmpC, and CmpD to Nicotiana benthamiana chloroplasts.(A) Schematic representation of the genetic constructs used in the figure.The chloroplast transit peptides (cTPs) originate from Arabidopsis (At).The proteins used are AtTic22-IV (At4g33350, GL202), AtABCD2 (At1g54350, GL273), and AtRecA (At1g79050, GL418, GL372).The length of the cTPs is shown as the number of residues in subscript.BCT1 genes are coloured as in Fig. 1.CmpC nucleotide-binding domain and regulatory domain are shown in light and dark green, respectively.(B) Confocal microscopy images of N. benthamiana leaf surfaces transiently expressing BCT1 proteins fused with mCitrine and the relevant compartment marker fused with mTurquoise [AtTGD2 for inner envelope membrane (IEM) and AtRbcS for stroma].CmpA localization is consistent with chloroplast intermembrane space when co-expressed with AtTGD2.CmpB co-localized with AtTGD2 at the inner envelope membrane.CmpC and CmpD both co-localized with AtRbcS in the stroma.

Fig. 3 .
Fig. 3. Combinatorial targeting of CmpC or CmpD with CmpB to the chloroplasts of Nicotiana benthamiana.(A) Schematic representation of the genetic constructs used in the figure.The chloroplast transit peptides (cTPs) originated from AtABCD2 (At1g54350, GL239) and AtRecA (At1g79050, GL371-372, GL418).The length of the cTPs are shown as the number of residues in subscript.CmpB (GL239) is tagged with the non-fluorescent HA-H 6 epitope, while CmpC (GL418), CmpC 263 (GL371), and CmpD (GL372) are fused with mCitrine.(B) Confocal microscopy images of N. benthamiana leaf surfaces transiently expressing CmpC and CmpD in isolation or combined with CmpB.When CmpC was co-expressed with CmpB (row 2), CmpC relocalized from the stroma to the inner envelope membrane (IEM; arrowhead).Individual targeting of CmpC without its regulatory domain (CmpC 263 , row 3) resulted in a stromal localization pattern, while co-expression with CmpB (row 4) led to the relocalization of CmpC 263 to the IEM.When CmpD and CmpB were co-expressed (row 6), CmpD relocalized from the stroma to the IEM (arrowhead).

Fig. 5 .
Fig.5.High-throughput spot test screening of BCT1 mutants in carbonic anhydrase (CA)-free E. coli.Plasmids carrying BCT1 variants, depicted on the right-hand side, were introduced into CAfree.The plasmid backbone used is a loop-compatible, modified version of pFA31, featuring a LacIQ-pTrc-pLac repressor/promoter cassette (grey arrow) and rrnB T1 and T2 terminator (grey box).On the left-hand side, cultures were plated in 5 µl spots on LB agar containing 0 or 100 µM isopropyl β-d-1-thiogalactopyranoside (IPTG) and incubated overnight at 37 °C in high (4%) or ambient (0.04%) CO 2 .Successful complementation was achieved when the induced cells (100 µM IPTG) were able to grow at ambient CO 2 (as observed in the last column).While unmodified BCT1 (GN18) was inactive, seven out of 13 mutants were able to complement CAfree to different extents at ambient levels of CO 2 (GN138, GN133, GN19, GN128, GN130, GM310, GM319).The corresponding schematic representation (see Fig.4) to which each plasmid relates to or derives from (indicated by an apostrophe) is presented on the far left as the panel letter from Fig.4.The black stars represent point mutations that are labelled, unless falling into a non-coding region (e.g.mutation between cmpA and cmpB in GN19), to show the change in residues (e.g.H409Q in GN128).

Fig. 6 .
Fig. 6.Functional analysis of BCT1 mutants in E. coli by uptake (A, B) and pull-down (C, D) assays.(A) Representative bicarbonate uptake rates measured in E. coli in the presence of 0.5 mM Ci for a subset of seven BCT1 mutants.The constructs used here are shown in Fig. 5 and corresponding depictions of BCT1 mutants are above each column (see Fig. 4 for legend).The values obtained with an empty vector, representing background CO 2 diffusion, have been subtracted.Statistical differences across mutants were assessed with a one-way ANOVA followed by pairwise multiple comparisons.Asterisks are an indication of the P-value (***P<0.001)relative to the unmodified BCT1 (GN18).Data are means ±SD (n=4).(B) Representative bicarbonate uptake curves for selected BCT1 mutants measured in E. coli.The Michaelis-Menten equation was fitted to the data by non-linear regression to obtain the maximal velocity (V MAX ) and affinity constant (K M ).Individual data points represent the mean of four technical replicates at each bicarbonate concentration (±SD).(C) Depiction of the constructs used for immobilized metal affinity chromatography (IMAC) pull-downs.(D) Western blot of the IMAC eluate showing co-purification of the BCT1 complex in E. coli.Loaded 10 µl of the concentrated eluate.Note that GM341 lacks a flag tag because CmpD is fused to CmpC and is detected with HA-H 6 around 107 kDa.

Fig. 7 .
Fig. 7. Complementation of the Arabidopsis βca5 mutant.Plants were grown at ambient (400 ppm), high (4000 ppm), or very high (40 000 ppm) CO 2 concentrations to assess the complementation ability of various BCT1 mutants.The genetic constructs used to transform the βca5 mutant are depicted in Supplementary Fig. S5.Depictions of BCT1 mutants is on the right-hand side.Colours are used consistently between the three panels.(A) Images of wild-type (WT; Col-0) and transformed βca5 mutant (SALK_121932) Arabidopsis plants 8 weeks after germination.The images are representative of six plants.Scale bar: 2 cm.(B) Overhead images of plants grown at ambient CO 2 were taken weekly, and rosette areas were measured using the PhenoImage and ImageJ software.Data are means ±SE (n=6).(C) Plants were harvested for fresh weight 8 weeks after germination.Statistical differences across genotypes were assessed with a one-way ANOVA followed by pairwise multiple comparisons between plants at each CO 2 concentration.Red asterisks are an indication of the P-value relative to wild type (*P<0.05;**P<0.01;***P<0.001).Data are means ±SE (n=3).

Fig. 8 .
Fig. 8. Functional analysis of BCT1 transformants in wild type (WT) Arabidopsis.(A) Images of 8-week-old Arabidopsis (Col-0) plants transformed with three BCT1 constructs (GN64, GN65, GN139).The plants were grown at ambient (400 ppm) or reduced CO 2 concentrations (200 ppm).The images are representative of six plants.Scale bar: 2 cm.Depictions of BCT1 mutants is on the right-hand side.GN64 and GN65 are translational fusions of CmpBC and CmpBD (reflecting a half-transporter design, Ford et al., 2019) and GN139 is a CmpCD fusion obtained by directed evolution.In the half-transporter design, GN64 harbors full-length CmpC while in GN65 CmpC has no regulatory domain (i.e.CmpC 263 ).BCT1 subunit colours are as described in Fig. 4A.(B) Overhead images of the plants were taken weekly, and rosette areas were measured using the PhenoImage and ImageJ software.Data are means ±SE (n=6).(C) Plants were harvested for fresh weight 8 weeks after germination.Statistical differences across genotypes were assessed with a one-way ANOVA followed by pairwise multiple comparisons between plants at each CO 2 concentration.No statistical difference was recorded.Data are means ±SE (n=4).Colours are used consistently between the three panels and are the same as used in Fig. 7.