All Arabidopsis (A. thaliana) Wild Type (WT) and the plants containing T-DNA insertions used in this work are of the Columbia (COL) ecotype. These plants were grown under ambient CO₂ (400 μL CO₂ per L air or 400 μL L⁻¹), low CO₂ (200 μL L⁻¹), and high CO₂ (1,000 μL L⁻¹) with an 8-h light, 16-h dark cycle with a light intensity of 120 μmol photons m⁻² sec⁻¹. All plants used in growth studies were watered every other day, alternating between distilled H₂O (dH₂O) and a 1:3 dilution of Hoagland’s nutrient solution in distilled H₂O (Epstein and Bloom, 2005).

Amplicons for pENTR™ Gateway construction were generated using Phusion polymerase (New England Biolabs). Primers for amplifying the coding regions and promoter regions of αCA2 (At2g28210) and βCA4 (At1g70410) were designed using Integrated DNA Technologies primer design tools and were generated by Integrated DNA Technologies (see Supplementary Table S1 for primer sequences). PCR fragments were gel purified using the Qiaquick Gel Extraction kit (Qiagen). 1–2 μL of purified PCR product was added to a pENTR™ master mix (1 μL of a 1.2 M NaCl and 0.06 M MgCl₂ mix [Invitrogen], 1 μL pENTR™/dTOPO^® vector mix [Invitrogen], and dH₂O [Invitrogen] to a final volume of 6 μL) for pENTR™ vector construction. Vectors were transformed into E. coli TOP10 chemically competent cells and plated onto YEP plates (for 1 L: 10 g peptone, 5 g NaCl, 10 g yeast extract, 15 g agar) supplemented with 50 μg mL⁻¹ kanamycin. pENTR™ vectors were subjected to restriction digestion and sequencing to confirm the correct orientation and sequence of the construct. eGFP amplicons were recombined into the pDEST vector pB7FWG2 (Karimi et al., 2002) and GUS amplicons were recombined into the pDEST vector pKGWFS7 (Karimi et al., 2002). The correct orientation of the pDEST™ vector was confirmed via restriction digestion.

Constructs used for the complementation studies were assembled using the GoldenGate modular cloning system (Weber et al., 2011; Patron et al., 2015). Linear Level 0 gene fragments (promoters, coding regions, terminators) were synthesized by Twist Biosciences (San Francisco, CA, United States) with defined Golden Gate compatible overhangs and cloned via traditional digestion andligation into Golden Gate compatible acceptor plasmids. Next, Level 1 constructs were generated to express the αCA2 or βCA4.1 genes under the control of the desired promoters. Finally, Level 1 constructs were assembled into Level 2 constructs to include a constitutively expressed BASTA resistance cassette. Correct assembly of Level 2 backbones was confirmed via restriction digestion.

Stable eGFP, GUS, and complementation lines were created following a modified procedure (Weigel and Glazebrook, 2002). A total of 200 μL of transformed A. tumefaciens was used to inoculate 200 mL of LB medium supplemented with antibiotics (30 μg mL⁻¹ gentamycin and 10 μg mL⁻¹ rifampicin for A. tumefaciens helper plasmids and either 100 μg mL⁻¹ spectinomycin for the eGFP and GUS vectors or 50 μg mL⁻¹ kanamycin for the complementation vector). The cultures were grown overnight at 28°C with vigorous shaking, and cells were pelleted in the morning by centrifugation at 7,250 RCF for 10mins at 20°C using a Beckman J2-HS centrifuge and JA-10 rotor. Pelleted cells were resuspended in 400 mL of A. tumefaciens infiltration medium (one-half-strength Murashige and Skoog medium with Gamborg’s vitamins from Caisson Laboratories, 5% [w/v] Sucrose, 0.044 μM benzylaminopurine suspended in dimethyl sulfoxide, and 50 μL L⁻¹ Silwet L-77 from Lehle Seeds). Inflorescences of Arabidopsis plants were dipped in the A. tumefaciens infiltration medium for approximately 40 s and then laid sideways in a flat tray with a covered dome to recover overnight, before incubating in constant light at 21°C (Weigel and Glazebrook, 2002). Positive transformants were selected on soil by spraying seedlings with a 1:1,000 dilution of BASTA (AgrEvo).

Four-to five-week-old Nicotiana tabacum (tobacco) plant leaves were used for transient eGFP expression. Two A. tumefaciens strains were used to generate transient eGFP expression in tobacco leaves. When infiltrating tobacco leaves, one strain, GV3101 containing the αCA2-mTurquoise construct, was combined with a second A tumefaciens strain, AGL-1 [p19], containing a suppressor to gene silencing construct (Voinnet et al., 2003). Two days before infiltrating tobacco leaves with an A. tumefaciens solution, 2 mL of YEP liquid culture was inoculated with either a single A. tumefaciens colony from an agar plate or from a glycerol stock and placed in a 28°C shaker overnight. The medium which was used to grow GV3101 contained 100 μg mL⁻¹ Spectinomycin (MP Biomedicals, LLC), 30 μg mL⁻¹ Gentamycin (GOLDBIO), and 10 μg mL⁻¹ Rifampicin (GOLDBIO). The medium which was used to grow AGL-1 [p19] also contained 50 μg mL⁻¹ Kanamycin (Sigma). The following afternoon, 100 μL of the GV3101 and AGL-1 [p19] cultures were transferred to 5 mL of YEP solution containing the appropriate antibiotics and were placed in a 28°C shaker overnight. Once an OD600 of 2.0 was reached,0.5 mL of the GV3101 culture and 0.5 mL of the AGL-1 [p19] culture were combined in a 2 mL microcentrifuge tube. The sample was spun at 3,625 RCF for 10 min and the supernatant was discarded. The A. tumefaciens pellet was resuspended in 1 mL of 10 mM MgCl₂ to remove the antibiotics. The A. tumefaciens solution was spun again at 3625 RCF for 10 min and the supernatant was discarded. The A. tumefaciens pellet was resuspended in 2 mL of an A. tumefaciens resuspension solution (1 mL of 100 mM MES; 1 mL of 100 mM MgCl₂; 100 μL of 1.5 Acetosyringone (Sigma); 7.9 mL dH₂O). Using a needle-less syringe, the 2 mL of A. tumefaciens solution was infiltrated into the abaxial side of multiple tobacco leaves. Infiltrated tobacco plants were returned to normal growth conditions and were imaged by confocal microscopy 3 days later.

GUS staining was visualized following a modified protocol of Jefferson et al. (1987). Plants and inflorescences were submerged in a GUS staining solution (0.1M NaPO₄ pH 7, 10 mM EDTA, 0.1% [v/v] Triton X-100, 1 mM K₃Fe(CN)₆, 2 mM 5-bromo, 4-chloro, 3-indolβ-D-glucuronic acid [X-Gluc, from GoldBio] suspended in N, N-dimethylformamide) and were placed in a 37°C incubator in the dark overnight. The following morning, plants were taken out of the incubator and the GUS staining solution was aspirated. Plant tissues were incubated in 100% methanol at 60°C for 15 min repeatedly until all chlorophyll was removed.

Following the protocol of Wu et al. (2009), 2 g of leaf tissue was incubated in 10 mL of enzyme solution (1% [w/v] cellulase from Trichoderma viride [Sigma], 0.25% [w/v] pectinase from Rhizopus spp. [Sigma], 0.4 M mannitol, 10 mM CaCl₂, 20 mM KCl, 0.1% [w/v] bovine serum albumin, and 20 mM MES at pH 5.7) for 1 h in light after placing Time Tape on the upper epidermis of the leaves and removing the lower epidermis of the leaves via Magic Tape. Protoplasts were then pelleted by centrifugation at 73 RCF for 3 min at 4°C using a Beckman J2-HS centrifuge and JS-13.1 rotor. Protoplasts were resuspended in a solution containing 0.4 M mannitol, 15 mM MgCl₂, and 4 mM MES at pH 5.7. eGFP fluorescence was visualized using protoplasts and leaves from stable eGFP plants with a Leica SP2 confocal microscope. The white light laser was used with 5% laser power and smart gain was adjusted to 100%. A 40X oil-emersion lens was used to visualize protoplasts and a ×20 objective lens was used to visualize intact cells from leaf samples. eGFP and chlorophyll were excited using a krypton/argon laser tuned to 488 nm, and eGFP and chlorophyll fluorescence were observed between the wavelengths of 500–520 nm and 660–700 nm, respectively.

DNA for genomic PCR was isolated from Arabidopsis leaves ground with a mortar and pestle and incubated in Edward’s extraction buffer (200 mM Tris-Cl, pH 7.5, 250 mM NaCl, 25 mM EDTA, and 0.5% [w/v] SDS). DNA was precipitated using 100% isopropanol followed by 70% [v/v] ethanol washes. RNA for reverse transcription was isolated from 80 mg of leaf tissue from 6-week-old Arabidopsis plants grown under ambient CO₂ and short days using the Qiagen RNeasy Plant minikit. Three micrograms of RNA were used for the reverse transcription reaction, and cDNA was generated using the SuperScript First-Strand RT-PCR kit and protocol (Invitrogen). cDNA at 0.5 μL was used for a 25-μL PCR using the standard protocol for One Taq (New England Biolabs).

Individual plants from each line were photographed weekly, and rosette areas were measured by tracing the outlines of the plants and obtaining the projected rosette area within each outline in ImageJ (National Institutes of Health). Rosette areas were measured on three plants per line. Fresh weights of above-ground plant mass were measured every week for each plant line for 5 weeks. Three plants per line were used for fresh weight analysis.

Leaf gas-exchange rates were measured with a LI-COR 6800 gas analyzer system and the 6800-fluorometer chamber. Photosynthetic response was characterized by construction of net assimilation rate versus leaf intercellular CO₂ concentration (A/Ci) curves for the 16th youngest leaf from four, separate, 7-week-old plants from each genotype studied. These measurements were made on plants grown at 200, 400 and 1,000 µL L^–1 CO₂. Leaves were first allowed to acclimate in the leaf cuvette at 400 μL L^–1 CO₂, 1,000 μmol photons m^–2s^–1 (saturating irradiance for these leaves), and 23°C–25°C until a steady-state for A was reached. Steady-state A was then measured at 400, 300, 250, 200, 100, 50, 400, 500, 600, 700, 800, 1,000, 1,300, 1,600 and 1,800 μL L^–1 CO₂ in air. We statistically analyzed the effects of genotype and growth CO₂ concentration on three characteristics of the A/Ci curves. The first characteristic was the CO₂ saturated A (A_csat) which occurred at Ci values of 1,300–1,600 μL L^–1 CO₂. The slope of the initial, linear portion of the A/Ci curve was the second characteristic. The CO₂ compensation point was the third characteristic. The initial linear portion for each A/Ci curve was determined from a linear regression fitted to the first 3-4 points on the curve (where r ² for the regression was 0.95–0.99). The compensation point was the Ci calculated from each regression by setting A equal zero.

Average abaxial and adaxial stomatal density values were taken from 20 WT, αca2, βca4, and αca2βca4 Arabidopsis leaf peels. Student’s t-test was performed to analyze the significant difference in the values among the four genotypes.

Statistical analysis for the growth data (Figure 8) was performed using One-way ANOVA in GraphPad Prism 8. Statistical significance is defined as p < 0.05. Two-factor ANOVA was performed for the CO₂ saturated A (Table 1), The slope of the initial, linear portion of the A/Ci curve (Table 2), and the CO₂ compensation point (Table 3). Statistical significance is defined as p < 0.01. Stomatal density (Supplementary Table S2) and stomatal conductance (Supplementary Table S3) values were analyzed using Student’s t-test. Statistical significance is defined as p < 0.05. At least three replicates of WT, αca2, βca4, and αca2βca4 plants were used for all the analysis.

The αCA2 gene (At2G28210) is composed of five exons with a relatively large intron between exons two and three (Figure 1A). The gene encoding βCA4 (At1g70410) is more complex as it has two active transcription start sites (Figure 1B). The longer transcript, βCA4.1, has ten exons while the shorter form, βCA4.2, is composed of nine exons. Both βCA4.1 and βCA4.2 are present in leaf tissue but only βCA4.2 is found in root tissue (DiMario et al., 2016). Previous work has shown that the shorter version of βCA4.2 is cytoplasmic while βCA4.1 is bound to the plasma membrane (DiMario et al., 2016).

To determine the cellular location of αCA2, αCA2 was initially fused with GFP, but this resulted in faint fluorescent signals around the periphery of the leaf cells. To better determine the localization of αCA2 in plant cells, the coding region of αCA2 was cloned upstream of a C-terminal m-Turquoise tag to generate the 35S::αCA2-m-Turquoise construct. We switched to m-Turquoise as it gives a better fluorescent signature under acidic conditions. Transiently expressed tobacco leaves with 35S::αCA2m-Turquoise showed the fluorescence signal on the outline of the leaf cells as seen by the clear “jigsaw” pattern in Figure 2. The clear fluorescent signals from m-Turquoise were similar to those seen when a C-terminal eGFP tag was added to βCA4.1 (Figure 2). When protoplasts were made from leaf cells expressing βCA4.1-GFP, the GFP signal was clearly seen around the plasma membrane (Figure 3) confirming the earlier reports by Fabre et al. (2007), Hu et al. (2010), and Hu et al. (2015) that βCA4.1 is located on the plasma membrane. The results shown in Figure 3 are consistent with the RNAseq data (Figure 1) that βCA4.1 is expressed in leaves and it is located on the plasma membrane.

αCA2 is predicted to move through the secretory pathway of Arabidopsis (TargetP 2.0) and the fluorescence pattern seen in plants expressing 35S::αCA2-m-Turquoise is consistent with the protein going through the secretory pathway to the plasma membrane or cell wall. However, while protoplasts generated from stably transformed Arabidopsis plants expressing βCA4.1-eGFP gave a plasma membrane signal (Figure 3), the m-Turquoise signal attached to αCA2 was missing in the transfected protoplasts confirming the cell wall localization (Figure 3). We hypothesize that the m-Turquoise linked to αCA2 is released from the cell as the cell wall is digested during protoplast generation. These results are consistent with αCA2 being a cell wall protein and agree with the software location predictions.

To determine which plant tissues expressed αCA2 and βCA4.1, promoter regions upstream of the αCA2 and βCA4.1 genes were used to generate promoter GUS constructs. Expression of the pαCA2::GUS construct showed GUS staining in roots and at the base of the young, developing leaves of the rosette (Figure 4A). In addition, there was higher expression of GUS in trichomes in plants with the pαCA2::GUS construct (Supplementary Figure S1). Plants containing the construct pβCA4.1::GUS showed strong GUS expression in both roots and leaves (Figure 4A) as has been reported earlier (DiMario et al., 2016). This result is consistent with the RNAseq data showing strong expression of βCA4 in both shoots and roots (Figure 1B). There was no significant expression of pβCA4.1::GUS in flowers whereas pαCA2::GUS showed expression in filaments and stigma (Figure 4B) in addition to the higher GUS expression in trichomes (Supplementary Figure S1).

Alleles containing T-DNA disruptions in each gene, SALK_080341 for the αca2 line and CS859392 for the βca4.1 line, were obtained from TAIR to determine the effect of αCA2 and βCA4.1 on plant growth. The SALK_080341 insert is located in the second intron of the αCA2 gene and the CS85939 insert is located within the fourth intron of the βCA4.1 gene (Figure 5A). Genomic PCR using either αCA2 or βCA4.1 gene-specific primers was used to confirm the specific T-DNA gene disruptions. In addition, a primer specific to the T-DNA insert was paired with a gene-specific primer to confirm the location of each T-DNA in its respective gene (Figures 5A, B). To confirm that these mutants are T-DNA KO lines, reverse transcription PCR (RT-PCR) was performed. The results indicated that αCA2 and βCA4.1 transcripts were present in the WT but absent in their respective mutant lines (Figure 6).

Plants missing αCA2 (αca2), βCA4 (βca4), or both CAs (αca2βca4) were compared with WT plants when grown under different CO₂ levels in an 8-h photoperiod with a light intensity of 120 μmol of photons m^-2 sec^-1. When these plants were grown under low CO₂, (200 μL L^–1 CO₂) the double mutant exhibited reduced growth when compared to WT plants or the single mutants (Figure 7A and Figure 8A). When grown at low CO₂, (200 μL L^–1 CO₂) the αca2βca4 plants had a smaller rosette area and low fresh weight when compared to WT plants and the αca2 and βca4 lines (Figure 7A and Figure 8A). The growth of αca2βca4 plants improved in 400 μL L^–1 CO₂ (ambient) or 1,000 μL L^–1 CO₂ (high) as it grew about the same as the WT or single mutants (Figures 7B, C; Figures 8B, C). The rosette areas and above ground fresh-weights of the WT, αca2, and βca4 plants were similar at 400 μL L^–1 CO₂ and 1,000 μL L^–1 CO₂. The reduced growth was only observed in the double mutant when grown on low CO₂.

To confirm whether the slow growth exhibited by the αca2βca4 double mutant at 200 μL L^–1 CO₂ was due to the missing CA genes, each gene was seperately put back into the double mutant (Supplementary Figure S2). When either the WT αCA2 or WT βCA4.1 was transformed into the double mutant, normal growth at 200 μL L^–1 CO₂ was restored (Figure 9). These results agree with our earlier observations that either single mutant was indistinguishable from the WT plants under the growth conditions we employed.

The photosynthetic response of the αca2βca4 plants was clearly different from responses of the other three genotypes when plants were grown at a CO₂ concentration of 200 μL L^–1 (Figure 10A). For plants growing at 400 and 1,000 μL L^–1 the A/Ci curves were similar for all genotypes (Figures 10B, C). We further examined three characteristics of the A/Ci curves to better define the differences in photosynthetic response. At 200 μL L^–1 the mean Acsat of the αca2βca4 plants was 39%–42% lower than Acsat in plants from the other three genotypes (Table 1). At 400 μL L^–1 the mean Ascat was similar among all genotypes, although the Acsat of αca2βca4 plants increased over 100% and the Acsat of the other three genotypes increased 20%–25% from values from plants grown at 200 μL L^–1 (Table 1). At 1,000 μL L^–1 the Acsat of the four genotypes were again similar, however, the Acsat values again increased 6%–17.6% over the values at 400 μL L^–1 (Table 1). For plants grown at 200 μL L^–1, the mean initial slope of the A/Ci curves from αca2βca4 plants was approximately 37% lower than the initial slopes from the other three genotypes (Table 2). The initial slopes from the A/Ci curves from all four genotypes were similar for plants grown at 400 and 1,000 μL L^–1. Like Acsat, there was an increase in the initial slopes of all genotypes from 200 to 400 μL L^–1 (57% in the αca2βca4 and 17%–25% in the other three genotypes, Tables 1, 2). Unlike the case in Acsat, there was not an increase in the initial slopes among genotypes from 400 to 1,000 μL L^–1 (Table 2). CO₂ compensation points were higher in all genotypes grown at 200 μL L^–1 CO₂ as compared to CO₂ compensation points in plants grown at 400 and 1,000 μL L^–1 CO₂. ANOVA showed the differences among growth CO₂ concentrations was highly significant (p < 0.01). The mean CO₂ compensation point for αca2βca4 at 200 μL L⁻¹ was higher than the other three genotypes but there was not a statistical difference due to variability among replicates. The CO₂ assimilation curves generated from the αca2βca4 complemented lines Ubi::αCA2 and Ubi::βCA4.1 showed that putting back either gene results in the recovery of normal photosynthesis (Figure 11).