Expression of a CO2-permeable aquaporin enhances mesophyll conductance in the C4 species Setaria viridis

A fundamental limitation of photosynthetic carbon fixation is the availability of CO2. In C4 plants, primary carboxylation occurs in mesophyll cytosol, and little is known about the role of CO2 diffusion in facilitating C4 photosynthesis. We have examined the expression, localization, and functional role of selected plasma membrane intrinsic aquaporins (PIPs) from Setaria italica (foxtail millet) and discovered that SiPIP2;7 is CO2-permeable. When ectopically expressed in mesophyll cells of Setaria viridis (green foxtail), SiPIP2;7 was localized to the plasma membrane and caused no marked changes in leaf biochemistry. Gas exchange and C18O16O discrimination measurements revealed that targeted expression of SiPIP2;7 enhanced the conductance to CO2 diffusion from the intercellular airspace to the mesophyll cytosol. Our results demonstrate that mesophyll conductance limits C4 photosynthesis at low pCO2 and that SiPIP2;7 is a functional CO2 permeable aquaporin that can improve CO2 diffusion at the airspace/mesophyll interface and enhance C4 photosynthesis.

The roles of the CO 2 permeable aquaporins have been largely characterized in C 3 photosynthetic plants where aquaporins localized in both the plasma membrane and the chloroplast envelope have been shown to facilitate CO 2 diffusion from the intercellular airspace to the site of Rubisco in chloroplasts (Kaldenhoff, 2012;Uehlein et al., 2008). The capacity for CO 2 diffusion to the initial sites of carboxylation influences the amount of water loss through transpiration (Cousins et al., 2020). Therefore, by providing a more efficient pathway for CO 2 diffusion, these membrane pores may contribute to increasing the water-use-efficiency (Groszmann et al., 2016). However, little is known about the role of CO 2 permeable aquaporins and their influence on CO 2 diffusion from substomatal cavities to the first site of carboxylation in C 4 photosynthesis. The C 4 photosynthetic pathway is a biochemical CO 2 pump where the initial conversion of CO 2 to bicarbonate (HCO 3 − ) by carbonic anhydrase (CA) and subsequent fixation of phosphoenolpyruvate (PEP) by PEP carboxylase (PEPC) takes place in the cytosol of mesophyll cells. The pathway requires a close collaboration between mesophyll and bundle sheath cells and this constrains leaf anatomy limiting mesophyll surface area that forms a diffusive interface for CO 2 (Evans and Von Caemmerer, 1996). Mesophyll conductance is defined as the conductance to CO 2 diffusion from the intercellular airspace to the mesophyll cytosol (Evans and Von Caemmerer, 1996;Osborn et al., 2017;Ubierna et al., 2017). Although the rate of C 4 photosynthesis is almost saturated at ambient pCO 2 , current modelling suggests that higher mesophyll conductance can increase assimilation rate and water-use-efficiency at low intercellular CO 2 partial pressures which occur when stomatal conductance is low (von Caemmerer and Furbank, 2016).
Setaria italica (foxtail millet) and Setaria viridis (green foxtail) are C 4 grasses of the Paniceae tribe and Poaceae family, related to important agronomical crops such as Z. mays (maize) and Sorghum bicolor (sorghum). S. viridis is frequently used as a model species for C 4 photosynthesis research as it is diploid with a relatively small genome that is sequenced and can be easily transformed (Brutnell et al., 2010;Ermakova et al., 2019;Osborn et al., 2017). Here, we used a yeast heterologous expression system to examine the permeability to CO 2 of selected PIPs from S. italica. We identified SiPIP2;7 as encoding a CO 2 -permeable aquaporin that, when expressed in the plasma membrane of S. viridis mesophyll cells, increased mesophyll conductance. Our results demonstrate that CO 2permeable aquaporins can be used to increase CO 2 diffusion from the intercellular airspace to mesophyll cytosol to provide higher carboxylation efficiency in C 4 leaves.
RNA-seq data from the publicly available Phytomine database (Phytozome) was examined for tissue-specific expression patterns of the S. italica PIPs (Figure 1a). SiPIP1;1, 1;2, 1;5, and 2;1 were expressed at moderate to high levels and SiPIP2;6 at low to moderate levels, in all tissues analyzed (root, leaves, shoot, and panicle). SiPIP1;6, 2;4, 2;5, 2;7, and 2;3 were expressed predominantly in roots at low to moderate levels. SiPIP2;8 was expressed only in leaves and SiPIP2;2 transcripts were not detected. (c) CO 2 permeability assay on yeast co-expressing SiPIPs and human CARBONIC ANHYDRASE II (hCAII) analyzed by stopped-flow spectrometry (see Figure 1-figure supplement 2 for details). 'hCAII only' expression was used as negative control. Mean±SE, n=3 biological replicates. Two independent experiments are presented. Asterisks indicate statistically significant differences between yeast expressing SiPIPs and 'hCAII only' control (t-test, p<0.05). (d) Yeast water permeability was assessed in the yeast aquaporin deletion background (aqy1/2) by the cumulative growth between untreated and freeze-thawed cells and determined by the percent area under the curve (% AUC). The yeast expressing the β-glucuronidase reporter gene (515.GUS) was used as negative control. Mean±SE, n=4 biological replicates. Asterisks indicate statistically significant differences between yeast expressing SiPIPs and 515.GUS control (t-test, p<0.01).
The online version of this article includes the following figure supplement(s) for figure 1:

Figure 1 continued on next page
Functional characterization of SiPIPs GFP localization of SiPIP-GFP fusions was used to confirm expression and determine targeting to the yeast plasma membrane (Figure 1b). Overall, SiPIP1s had lower GFP signal that was patchy at the cell periphery with strong internal signal consistent with localization to the endoplasmic reticulum. GFP signal was also present diffusively throughout the cytosol suggestive of protein degradation. Overall, SiPIP1s were poorly produced in yeast and were not efficiently targeting the plasma membrane as needed for the functional assays. For the PIP2s, only SiPIP2;1, SiPIP2;4, SiPIP2;5, and SiPIP2;7 showed clear localization to the plasma membrane in addition to other internal structures, and were therefore selected for further functional analyses.
CO 2 permeability was measured in yeast co-expressing a SiPIP along with human CARBONIC ANHYDRASE II (hCAII). A stopped-flow spectrophotometer was used to monitor CO 2 -triggered intracellular acidification via changes in fluorescence intensity of a pH-sensitive fluorescein dye (Figure 1figure supplement 2; Ding et al., 2013;Heckwolf et al., 2011;Uehlein et al., 2008). Importantly for reliable results, all SiPIP yeast lines tested showed similar cell volumes and were not limited by CA activity (Figure 1-figure supplement 2). A screen of the lines revealed that yeast expressing SiPIP2;7 had the highest CO 2 permeability of 1.5×10 -4 m s -1 , which was significantly larger than the negative control expressing hCAII only (Figure 1c). Other SiPIPs displayed comparable CO 2 permeability to the hCAII only control. The changes in CO 2 permeability detected on the stopped-flow spectrophotometer for yeast expressing SiPIP2;7 were not an artifact brought on by an increased permeability to protons causing the intracellular acidification ( Figure 1-figure supplement 3).
Freeze-thaw survival assays, which quantify water permeability of aquaporins (Tanghe et al., 2002), provided further confirmation that the SiPIPs expressed in yeast were functional. Overexpression of water permeable aquaporins greatly improves freeze-thaw tolerance in yeast, especially in the highly compromised aquaporin knockout mutant aqy1/2 (Tanghe et al., 2002). Yeast expressing the β-glucuronidase reporter gene (515.GUS) was used a control to show that the single freeze-thaw treatment was effective in almost killing off the entire yeast population (Figure 1d). Consistent with the poor plasma membrane localization and abundance of SiPIP2;1-GFP (Figure 1b), yeast expressing SiPIP2;1 did not show any protection to freeze-thaw treatments ( Figure 1c). On the other hand, SiPIP2;4, 2;5, and 2;7 all showed some level of protection, indicating that they permeated water and were functional within the plasma membrane of yeast cells. For detailed characterization of water permeability, SiPIP2;7 was expressed in Xenopus laevis oocytes. Swelling assay confirmed that SiPIP2;7 is a functional water channel (Figure 1-figure supplement 4).

Expression of SiPIP2;7 in mesophyll cells of S. viridis
To confirm and exploit the CO 2 permeability characteristic of SiPIP2;7 in planta, we created transgenic S. viridis plants expressing SiPIP2;7 with a C-terminal FLAG-tag fusion and under the control of the mesophyll-preferential Z. mays PEPC promoter (Gupta et al., 2020;Salesse-Smith et al., 2018). Out of 52 T 0 plants analyzed for SiPIP2;7-FLAG protein abundance and the hygromycin phosphotransferase (hpt) gene copy number (      Abundances of photosynthetic proteins PEPC, CA, the Rieske subunit of the Cytochrome b 6 f complex, and the small subunit of Rubisco (RbcS), did not differ between transgenic and control plants ( Figure 2a). In line with the immunoblotting results, measured activities of PEPC and CA, and the amount of Rubisco active sites were not altered in the transgenic plants (Table 1). Chlorophyll content, leaf dry weight per area, and biomass of roots and shoots did not differ between the genotypes either ( Table 1).
To study the impact of SiPIP2;7-FLAG expression on the photosynthetic properties in transgenic plants, we conducted concurrent gas exchange and chlorophyll fluorescence analyses at different intercellular CO 2 partial pressures (C i ) ( Figure 3). No significant changes were detected between transgenic and control plants in CO 2 assimilation rates (A), effective quantum yield of Photosystem II    (φPSII) or stomatal conductance to water vapor ( Figure 3-figure supplement 1). The SiPIP2;7-FLAG protein abundance was compared to the gas exchange phenotype in individual plants ( Figure 3figure supplement 2). A statistically significant polynomial relationship (R 2 =0.345, p<0.05) was found between the initial slopes and the relative protein content, which was significantly better than that achieved using a linear model (p<0.05). No significant relationship was observed between the SiPIP2;7-FLAG abundance and the saturating rates of assimilation (A max ; Mesophyll conductance to CO 2 in plants expressing SiPIP2;7 Next, we analyzed in detail the initial slopes of the AC i curves and mesophyll conductance. Fitting linear regressions indicated that mean±SE values of the initial slopes of the AC i curves for lines 27, 44, and 52 were 0.46±0.03, 0.52±0.01, and 0.53±0.05, respectively, compared to the value of 0.41±0.02 in control plants (Figure 4a). Measurements of ∆ 18 O were used to estimate conductance of CO 2 from the intercellular airspace to the sites of CO 2 and H 2 O exchange in the mesophyll cytosol (g m ) with the assumption that CO 2 was in full isotopic equilibrium with leaf water in the cytosol (Barbour et al., 2016;Osborn et al., 2017). Transgenic lines showed mesophyll conductance of 0.59±0.05, 0.55±0.08, and 0.46±0.04 mol m -2 s -1 bar -1 compared to the mean±SE value of 0.42±0.03 mol m -2 s -1 bar -1 in control plants ( Figure 4b). Two-way ANOVA analysis with Tukey post hoc test on the initial slopes of the AC i curves and g m showed that differences measured in plants of lines 27 and 44 were statistically significant from the control plants (p=0.04573 and 0.03724, respectively). Interestingly, for plants of line 52, only initial slopes were significantly different compared to control plants when compared by one-way ANOVA (Tukey post hoc test, α=0.05, p=0.02993).
We also used the g m calculations proposed by Ogée et al., 2018 which try to account for the rates of bicarbonate consumption by CA. The CA hydration constant (k CA ) of 6.5 mol m -2 s -1 bar -1 was used for these calculations ( Table 1). We found that the g m measured with this method gave on average 1.25 times greater values but did not change the ranking of mesophyll conductance shown in Figure 4a ( Figure 4-figure supplement 1). The C 4 photosynthetic model by von Furbank, 1999 andvon Caemmerer, 2000 relates the initial slope of the CO 2 response curve (dA/C i ) to g m (see Figure 4 caption and Materials and methods). Figure 4c shows that the measured relationship between the initial slope and g m fits closely with model prediction.

Discussion
The diffusion of CO 2 from the Earth's atmosphere to the site of primary carboxylation within leaves of C 3 and C 4 plants often limits photosynthesis and impacts the efficient use of water. Overexpression of CO 2 permeable aquaporins with the aim to enhance leaf CO 2 diffusion has been extensively probed in C 3 plants. Many studies have demonstrated that it was an effective strategy to improve g m , leading to increased assimilation rate or grain yield (Flexas et al., 2006;Hanba et al., 2004;Uehlein et al., 2003; Xu et al., 2019). However, some published research could not identify a functional link between the CO 2 permeable aquaporins and mesophyll conductance, and possible explanations include functional redundancy between aquaporin isoforms (Kromdijk et al., 2020) and complications of expressing membrane proteins in the chloroplast envelope (Fernández-San Millán et al., 2018). In contrast to C 3 plants, where CO 2 needs to cross both plasma membrane and the chloroplast envelope to reach the site of carboxylation, the primary carboxylation step in C 4 plants occurs in the cytosol of mesophyll cells. Due to high photosynthetic rates of C 4 plants, this generates a large CO 2 drawdown between the intercellular airspaces and the cytosol and thus a large mesophyll conductance (Evans and Von Caemmerer, 1996). Because only one membrane needs to be traversed by CO 2 , C 4 plants could serve as a simpler model to demonstrate the effect of CO 2 permeable aquaporins on photosynthesis.    Screening S. italica PIPs for CO 2 permeability in a yeast heterologous system resulted in identification of SiPIP2;7 as a CO 2 pore (Figure 1c). Expression analysis revealed that SiPIP2;7 was almost exclusively expressed in roots under ideal conditions ( Figure 1a, Figure 2-figure supplement 2) which, combined with the water permeability identified in yeast and oocyte assays (Figure 1d, Figure 1-figure supplement 4), suggest that SiPIP2;7 may function in regulating root hydraulic conductivity, a role extensively documented for PIP aquaporins (Gambetta et al., 2017;McGaughey et al., 2018). The physiological relevance of SiPIP2;7's CO 2 permeating capacity is not immediately clear. Gas uptake by roots is well documented (Stemmet et al., 1962) and in C 3 plants CO 2 uptake by roots may contribute to the C 4 photosynthesis-like metabolism detected in stems and petioles (Hibberd and Quick, 2002). It is possible that SiPIP2;7 is conditionally expressed in leaves, or even that its capacity to transport CO 2 is inadvertent and related to the transportation of another yet undetermined substrate; analogous to the uptake of toxic metalloids by some NIP aquaporins due to their capacity to transport boron (Mukhopadhyay et al., 2014). Further work is needed to determine whether PIPs in general function natively as relevant CO 2 pores in C 4 leaves.
We employed the CO 2 transport capacity of SiPIP2;7 to enhance transmembrane CO 2 diffusion from the intercellular airspace into the mesophyll cytosol, where CA and PEPC reside, by overexpressing SiPIP2;7 in S. viridis. We confirmed the localization of SiPIP2;7 within the mesophyll plasma membranes (Figure 2c) and detected the increase in CO 2 diffusion across the mesophyll membranes in transgenic plants by two independent methods. First, we calculated g m from the C 18 O 16 O discrimination measurements ( Figure 4b) and the theory for these calculations Figure 4. Effect of the mesophyll conductance, g m, on the initial slope of the CO 2 assimilation response curve to the intercellular CO 2 partial pressure (AC i curve) in leaves of Setaria viridis expressing SiPIP2;7-FLAG in mesophyll cells. (a) Initial slope of the AC i curves estimated by linear fitting of curves (a subset of curves is presented in Figure 3a inset). (b) Mesophyll conductance, g m , estimated by oxygen isotope discrimination assuming full isotopic equilibrium (Osborn et al., 2017). Measurements were made at ambient CO 2 and low O 2 . (c) Data from (a) and (b) compared to the C 4 biochemical model predictions (von Caemmerer, 2000;von Caemmerer and Furbank, 1999). All graphs show Mean±SE; azygous plants of line 44 were used as control. The model relates the initial slope of the AC i curve (dA/C i ) to g m by: dA , where V pmax and K p denote the maximum PEPC activity and the Michaelis Menten constant for CO 2 taken here as 250 µmol m -2 s -1 and 82 µbar (DiMario and Cousins, 2019;

Figure 4 continued on next page
von Caemmerer, 2021). Asterisks indicate statistically significant differences between the plants of lines 27 (p=0.04573) and 44 (p=0.03724) and control plants (twoway ANOVA with Tukey post hoc test, α=0.05). Plants of line 52 were not significantly different from the control plants (p=0.27518).
The online version of this article includes the following figure supplement(s) for figure 4: Figure supplement 1. Comparison of the g m estimated by oxygen isotope discrimination assuming full isotopic equilibrium (Osborn et al., 2017) and calculations suggested by Ogée et al., 2018. Figure 4 continued has been outlined (Barbour et al., 2016;Ogée et al., 2018;Osborn et al., 2017). Second, we fitted linear regressions to the initial slopes of the AC i curves (Figure 4a), which depend on g m , V pmax , and K p where the two latter parameters denote the maximum PEPC activity and the Michaelis Menten constant of PEPC for HCO 3 − (von Caemmerer, 2000;von Caemmerer and Furbank, 1999). Since PEPC and CA activities were not altered in plants expressing SiPIP2;7 (Table 1), higher initial slopes of the AC i curves in transgenic lines were attributed to the increased g m . When plotted against each other, the initial slopes and g m in transgenic and control plants, fitted the model predictions confirming the hypothesized functional role of g m in C 4 photosynthesis (Pfeffer and Peisker, 1995;Ubierna et al., 2017;von Caemmerer, 2000). Importantly, and in line with the model predictions, expression of SiPIP2;7 had an effect only on the initial slopes of the AC i curves but not on the saturating rates of assimilation (Figure 3-figure supplement 2).
Our findings demonstrate that CO 2 permeable aquaporins can enhance CO 2 diffusion at the airspace/mesophyll interface in C 4 plants. However, overexpression of aquaporins is likely accompanied by confounding factors which could explain the lack of phenotype observed from the plants with the highest SiPIP2;7 expression level (Figure 3-figure supplement 2) and plants of line 52 showing significant increase of only initial slopes (Figure 4). These factors possibly include the transport of multiple substrate(s) by SiPIP2;7 and different efficiency of SiPIP2;7 targeting to the plasma membrane between the individual lines, plants, leaves, and even between different mesophyll cells. Moreover, it is still not clear whether there are other CO 2 -permeable aquaporins present in S. viridis leaves that account for the basal g m level of 0.42±0.03 mol m -2 s -1 bar -1 in control plants. Further research is required to unravel the aforementioned confounding factors. Nevertheless, building on our findings that increasing mesophyll conductance is possible with overexpression of SiPIP2;7, this trait will be a strong candidate to combine with complementary traits such as the overexpressing of Cytochrome b 6 f (Ermakova et al., 2019) and Rubisco (Salesse-Smith et al., 2018), which offer improvements to photosynthesis in C 4 plants.

Materials and methods
Heterologous expression in yeast cDNAs encoding the 12 S. italica aquaporins (Supplementary file 1) and human CARBONIC ANHY-DRASE II (hCAII, AK312978) were codon-optimized for expression in yeast with IDT DNA tool (https:// sg. idtdna. com/ pages/ tools) and a yeast-related Kozak sequence was added at the 5′ end to help increase translation (Nakagawa et al., 2008). For CO 2 permeability measurements, pSF-TPI1-URA3 with an aquaporin and pSF-TEF1-LEU2 with hCAII were co-transformed into the S. cerevisiae strain INVSc1 (Thermo Fisher Scientific, Waltham, MA). For water permeability measurements, pSF-TPI1-URA3 with an aquaporin was transformed into the aqy1/2 double mutant yeast strain deficient in aquaporins (Suga and Maeshima, 2004). The yeast vectors pSF-TPI1-URA3 and pSF-TEF1-LEU2 were obtained from Oxford Genetics (Oxford, UK). Yeast transformation was performed using the Frozen-EZ yeast transformation II kit (Zymo Research, Irvine, CA) and selection of positive transformants was based on amino acid complementation. To ensure CA was not limiting, CA activity was determined using a membrane inlet mass spectrometry as described by Endeward et al., 2006 (Figure 1-figure supplement 2). For CO 2 permeability measurements, an average cell diameter of 4.63 µm was determined by measuring ~100 yeast cells expressing each aquaporin (Figure 1-figure supplement 2). To study the subcellular localizations of aquaporins in yeast, a C-terminus GFP tag was added to the sequences into the pSF-TPI1-URA3 vector (pSF-TPI1-URA3-GFP). The fluorescence signal was observed using a Zeiss 780 confocal laser scanning microscope (Zeiss, Oberkochen, Germany): excitation 488 nm and emission 530 nm. Cytosolic GFP expression was used as control.
CO 2 induced intracellular acidification assay CO 2 intracellular acidification was measured in yeast cells loaded with fluorescein diacetate (Sigma-Aldrich, St. Louis, MO) as described previously (Bertl and Kaldenhoff, 2007;Otto et al., 2010). Briefly, an overnight culture of yeast cells was collected and resuspended in an equal volume of 50 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES)-NaOH, pH 7.0, 50 µM fluorescein diacetate and incubated for 30 min in the dark at 37°C. The suspension was centrifuged and the pellet resuspended in ice-cold incubation buffer (25 mM HEPES-NaOH, pH 6.0, 75 mM NaCl). Cells loaded with fluorescein diacetate were then injected into the stopped-flow spectrophotometer (DX.17MV; Applied Photophysics, Leatherhead, UK) alongside a buffer solution (25 mM HEPES, pH 6.0, 75 mM NaHCO 3 , bubbled with CO 2 for 2 hr). The kinetics of acidification was measured at 490 nm excitation and >515 nm emission (OG515 long pass filter, Schott, supplied by Applied Photophysics). Data were collected over a time interval of 0.2 s and analyzed using ProData SX viewer software (Applied Photophysics). CO 2 permeability was determined using the method of Yang et al., 2000. An average of 75 injections over at least three separate cultures was used for each aquaporin.

Determination of water permeability
A freeze-thaw yeast assay was used to determine water permeability of aquaporins expressed in aqy1/2 based on previous reports (Tanghe et al., 2002). Briefly, an overnight culture was diluted to ~6×10 6 cells (final volume 1 ml) in appropriate selection liquid growth medium and incubated at 30°C for 1 hr. 250 µl of each culture were then aliquoted into two standard 1.5 ml microtubes: the first (control) tube was placed on ice and the second tube was subject to a single freeze-thaw treatment, consisting of 30 s freezing in liquid nitrogen and thawing for 20 min in a 30°C water bath. Following the treatment, the cells were placed on ice. The tubes were then vortexed briefly to ensure even suspension of cells and 200 µl of the culture was transferred to wells of a Nunc-96 400 µl flat bottom untreated plate (Thermo Fisher Scientific, Cat#243656). Yeast growth in control and treated cultures were monitored over a 24-30 hr period in an M1000 Pro plate reader (TECAN, Männedorf, Switzerland) at 30°C with double orbital shaking at 400 rpm and measuring absorbance at 650 nm every 10 min. Growth data were log-transformed and freeze-thaw survival calculated as the growth (area under the curve) of treated culture relative to its untreated control from time zero up until the untreated control culture reached stationary phase.

Construct assembly and S. viridis transformation
The coding sequence of S. viridis PIP2;7 (Sevir.2G128300.1, Phytozome, https:// phytozome. jgi. doe. gov/) has been codon optimized for the Golden Gate cloning (Engler et al., 2014) and translationally fused with the glycine linker and the FLAG-tag coding sequence (Hopp et al., 1988). The resulting coding sequence was assembled with the Z. mays PEPC promoter and the bacterial tNos terminator into the second expression module of the pAGM4723 binary vector. The first expression module has been occupied by the hygromycin phosphotransferase (hpt) gene assembled with the Oryza sativa Actin-1 promoter and the tNos terminator. The construct was transformed into S. viridis cv. MEO V34-1 using Agrobacterium tumefaciens strain AGL1 following the procedure described in Osborn et al., 2017. T 0 plants resistant to hygromycin were transferred to soil and analyzed for SiPIP2;7-FLAG protein abundance and hpt insertion number by droplet digital PCR (iDNA Genetics, Norwich, UK). Lines 27, 44, and 52 were selected for further analysis because they had the strongest FLAG signal per transgene insertion number (Figure 2-figure supplement 1). The T 1 and T 2 progenies of T 0 plants 27, 44, and 52 were analyzed. Azygous T 1 plants of line 44 and their progeny were used as control.

Plant growth conditions
Seeds were surface-sterilized and germinated on medium (pH 5.7) containing 2.15 g L −1 Murashige and Skoog salts, 10 ml L −1 100× Murashige and Skoog vitamins stock, 30 g L −1 sucrose, 7 g L −1 Phytoblend, 20 mg L −1 hygromycin (no hygromycin for azygous plants). Seedlings that developed secondary roots were transferred to 0.6 L pots with garden soil mix layered on top with 2 cm seed raising mix (Debco, Tyabb, Australia) both containing 1 g L −1 Osmocote (Scotts, Bella Vista, Australia). Plants were grown in controlled environmental chambers with 16 hr light/8 hr dark, 28°C day, 22°C night, 60% humidity, and ambient CO 2 concentrations. Light intensity of 300 µmol m −2 s −1 was supplied by 1000 W red sunrise 3200K lamps (Sunmaster Growlamps, Solon, OH). Youngest fully expanded leaves of the 3-4 weeks plants before flowering were used for all analyses.

Chlorophyll and enzyme activity
Chlorophyll content was measured on frozen leaf discs homogenized with a TissueLyser II (Qiagen, Venlo, The Netherlands) (Porra et al., 1989). PEPC activity was determined after Pengelly et al., 2010 from fresh leaf extracts from the plants adapted for 1 hr to 800 µmol photons m -2 s -1 . CA activity was measured on a membrane inlet mass spectrometer as a rate of 18 O exchange from labeled 13 C 18 O 2 to H 2 16 O at 25°C according to Von Caemmerer et al., 2004 by calculating the hydration rate after Jenkins et al., 1989. The amount of Rubisco active sites was determined by [ 14 C] carboxyarabinitol bisphosphate binding as described earlier (Ruuska et al., 2000).

RNA isolation and qPCR
Leaf and root tissue were frozen in liquid N 2 . Leaf samples were homogenized using a TissueLyser II and RNA was extracted using the RNeasy Plant Mini Kit (Qiagen). Roots were ground with mortar and pestle in liquid N 2 and RNA was isolated according to Massey, 2012. Briefly, 150 µl of pre-heated (60°C) extraction buffer (0.1 M trisaminomethane (Tris)-HCl, pH 8.5 mM ethylenediaminetetraacetic acid [EDTA], 0.1 M NaCl, 0.5% sodium dodecyl sulfate [SDS], 1% 2-mercaptoethanol) was added to ~100 mg of fine root powder and incubated at 60°C for 5 min. 150 µl of phenol:chloroform:isoamyl alcohol (25:24:1) saturated with 10 mM Tris (pH 8.0) and 1 mM EDTA was added to the samples, vortexed vigorously for 10 min and centrifuged at 4500×g for 15 min. Aqueous phase was mixed with 120 µl of isopropanol and 15 µl of 3 M sodium acetate and incubated at -80°C for 15 min, then centrifuged at 4500×g (30 min, 4°C). The pellet was washed two times in 300 µl of ice-cold 70% ethanol, air-dried, and dissolved in 60 µl of RNase-free water. After addition of 40 µl of 8 M LiCl, samples were incubated overnight at 4°C. Nucleic acids were pelleted by centrifugation at 16,000×g (60 min, 4°C), washed two times with 200 µl of ice-cold 70% ethanol, air-dried, and dissolved in RNase-free water. DNA from the samples was removed using an Ambion TURBO DNA-free Kit (Thermo Fisher Scientific), and RNA quality was determined using a NanoDrop (Thermo Fisher Scientific). 100 ng of total RNA were reverse transcribed into cDNA using a SuperScript III Reverse Transcriptase (Thermo Fisher Scientific). qPCR and melt curve analysis were performed on a Viia7 Real-Time PCR System (Thermo Fisher Scientific) using the Power SYBR Green PCR Master Mix (Thermo Fisher Scientific) according to the manufacturer's protocol. Primer pairs designed to distinguish between S. viridis PIP2;6 and PIP2;7 using Primer3 in Geneious Prime (https://www. geneious. com) and reference primers are listed in Supplementary file 3.

Gas exchange measurements
Gas exchange and fluorescence analyses were performed at an irradiance of 1500 µmol m −2 s −1 (90% red/10% blue actinic light) and different intercellular CO 2 partial pressures using a LI-6800 (LI-COR Biosciences, Lincoln, NE) equipped with a fluorometer head 6800-01A (LI-COR Biosciences). Leaves were first equilibrated at 400 ppm CO 2 in the reference side, leaf temperature 25°C, 60% humidity, and flow rate of 500 µmol s −1 and then a stepwise increase of CO 2 concentrations from 0 to 1600 ppm was imposed at 3-min intervals. Initial slopes of the CO 2 response curves were determined by linear fitting in OriginPro 2018b (OriginLab, Northampton, MA). Quantum yield of PSII upon the application of multiphase saturating pulses (8000 µmol m −2 s −1 ) was calculated according to Genty et al., 1989 O were made by coupling two LI-6400XT gas exchange systems (LI-COR Biosciences) to a tunable diode laser (TDL; model TGA200A, Campbell Scientific Inc, Logan, UT) to measure C 18 O 16 O discrimination and a Cavity Ring-Down Spectrometer (L2130-i, Picarro Inc, Sunnyvale, CA) to measure the oxygen isotope composition of water vapor (Osborn et al., 2017). Measurements were made at 2% O 2 , 380 µmol mol -1 CO 2 , leaf temperature of 25°C, irradiance of 1500 µmol m -2 s -1 , and relative humidity of 55%. Each leaf was measured at 4 min intervals and 10 readings were taken. Mesophyll conductance was calculated as described by Osborn et al., 2017 with the assumptions that there was sufficient CA in the mesophyll cytosol for isotopic equilibration between CO 2 and HCO 3 − . We also used calculations proposed by Ogée et al., 2018 to estimate g m. These calculations try to account for the rates of bicarbonate consumption by CA. We used the rate constant of CA hydration (k CA ) of 6.5 mol m -2 s -1 bar -1 for these calculations.

Statistical analysis
One-way and two-way ANOVAs with Tukey post hoc test were performed in OriginPro 2018b. A two-tailed, heteroscedastic Student's t-tests were performed in Microsoft Excel. Linear modeling was performed in R (R Development Core Team, 2021); mixed-effects models to test the need to incorporate transgenic event and copy status was performed in lme4 (Bates et al., 2015).

Additional files
Supplementary files • Supplementary file 1. Amino acid similarity of S. italica and S. viridis PIPs.
• Supplementary file 2. Amino acid composition of S. italica PIPs at known substrate selectivity positions. Aquaporins have six transmembrane α-helices (H1-H6) joined by five loops (LA-LE). The monomeric channel is characterized by two NPA (Asn, Pro, Ala) motifs on loops A and E and along with the aromatic/arginine selectivity filter, these largely dictate the transport selectivity of substrates through the monomeric channels (Azad et al., 2016). Froger's positions indicate additional residues that are predicted for substrate specificity (Froger et al., 1998).

• Transparent reporting form
Data availability All data generated or analysed during this study are included in the manuscript and supporting files.