Individual transgenic lines were generated overexpressing four of the core C₄ cycle enzymes required for a functional NADP-ME C₄ cycle (Kajala et al., 2011; Supplementary Figure 1), ZmPEPC (GRMZM2G083841), ZmPPDK (GRMZM2G306345), ZmNADP-MDH (GRMZM2G129513), and ZmNADP-ME (GRMZM2G085019. Generations of pSC0/ZmPEPC, pSC0/ZmPPDK, pSC0/ZmNADP-MDH, and pSC0/NADP-ME vectors were previously described (Giuliani et al., 2019b; Karki et al., 2020). In almost all cases, three independent single insertion homozygous transgenic lines with high transgene expression were selected for molecular and biochemical evaluation. However, for ZmNADP-ME, protein expression was only detected in a single transgenic line containing >6 copies of the overexpression construct and so this was the only line that could be taken forward (Karki et al., 2020). The overexpression constructs were stacked into single lines through conventional crossing to create two triple cross line (PEPC-28/PPDK-11/MDH-40 and PEPC-62/PPDK-2/MDH-22) each with an independent transgenic event for each gene. The crossing strategy was presented in Supplementary Figure 2. The ZmPEPC (PEPC-28 and PEPC-62) and ZmNADP-MDH (MDH-40 and MDH-22) single transgenic lines were initially crossed to create two double transgene lines (PEPC-28/MDH-40 and PEPC-62/MDH-22) that were then crossed with the ZmPPDK (PPDK-11 and PPDK-2) single transgenic lines. These two triple lines (PEPC-28/PPDK-11/MDH-40 and PEPC-62/PPDK-2/MDH-22) were then crossed with the single ZmNADP-ME (ME-116) transgenic line to produce two quadruple lines (PEPC-28/PPDK-11/MDH-40/ME-116 and PEPC-62/PPDK-2/MDH-22/ME-116). Progeny of the PEPC-28/PPDK-11/MDH-40 showed that detectable ZmPEPC localizes to MCs, but progeny of the PEPC-62/PPDK-2/MDH-22 showed that ZmPEPC localizes to both MCs and BSCs (Supplementary Figure 3). The PEPC-28/PPDK-11/MDH-40/ME-116 line was selected for analysis in the present study since its parent PEPC-28/PPDK-11/MDH-40 showed that correct ZmPEPC MCs expression and its progenitor PEPC-28 had been detail characterized in Giuliani et al. (2019b). Two quintuple crosses (PEPC-28/PPDK-11/MDH-40/ME-116/gdch-31) and (PEPC-28/PPDK-11/MDH-40/ME-116/gdch-38) were then generated by crossing the quadruple F₂ line (PEPC-28/PPDK-11/MDH-40/ME-116) with single Osgdch knockdown lines (gdch-31 and gdch-38) described by Lin et al. (2016). The PEPC-28/PPDK-11/MDH-40/ME-116/gdch-38 line was chosen for analysis in present study since its progenitor gdch-38 had shown a more consistent photorespiratory-deficient phenotype under different O₂:CO₂ growing and measuring conditions compared with gdch-31 line, and had been detail characterized in Giuliani et al. (2019a). The presence of transgenes was determined by genomic PCR and protein accumulation by immunoblotting in each crossed line (Supplementary Figures 4–6).

Plants were grown under natural light conditions in a screenhouse with a day/night temperature of 35/28 ± 3°C at the International Rice Research Institute (Los Baños, Philippines: 14° 10019.900N, 121° 15022.300E). Maximum irradiance was 2000 μmol photons m^–2 s^–1 on a sunny day. Plants were grown in 7-liter pots filled with soil from the IRRI upland farm.

Leaf samples for soluble protein extraction were harvested from the youngest fully expanded leaf at the mid-tillering stage between 09:00 h and 11:00 h, and stored on ice immediately. Leaves were homogenized to a fine powder using a nitrogen-cooled mortar and pestle. Proteins were extracted and fractionated by SDS-PAGE as described previously (Lin et al., 2016). Samples were loaded based on equal leaf area (0.2364 mm² for ZmPEPC and ZmPPDK, and 2.364 mm² for ZmNADP-MDH, ZmNADP-ME and OsGDCH). After electrophoresis, proteins were electroblotted onto a polyvinylidene difluoride membrane and probed with rabbit antisera against ZmPEPC, ZmNADP-MDH, ZmNADP-ME (all provided by Richard Leegood, University of Sheffield, United Kingdom), ZmPPDK (provided by Chris Chastain, Minnesota State University, United States), and OsGDCH protein (provided by Asaph Cousins, Washington State University, United States). The dilutions of ZmPEPC, ZmPPDK, ZmNADP-MDH, ZmNADP-ME, and OsGDCH antisera were 1:20,000, 1:20,000, 1:5,000, 1:2,000, and 1:100, respectively. A peroxidase-conjugated goat anti-rabbit IgG secondary antibody (Sigma-Aldrich, United States)² was used at a dilution of 1:5,000 and immunoreactive bands were visualized with ECL Western Blotting Detection Reagents (GE Healthcare, United Kingdom)³.

The middle portion of the youngest fully expanded leaf at the mid-tillering stage was sampled between 09:00 h and 11:00 h and processed as described previously by Lin et al. (2016). After fixation and cutting, the thin leaf sections were probed with the antisera against ZmNADP-MDH, ZmNADP-ME, ZmPPDK, and ZmPEPC at dilutions of 1:500, 1:25, 1:10, and 1:200, respectively. The secondary Alexa Fluor 488 goat anti-rabbit IgG (Invitrogen, United States)⁴ antibody was used at a dilution of 1:200. The sections were visualized on a BX61 microscope fitted with a Disk Scanning Unit attachment microscope (Olympus, United States)⁵ with fluorescence function under DAPI, RFP, and GFP filters.

Leaf samples were harvested between 09:00 h and 11:00 h from the youngest fully expanded leaf of plants at the mid-tillering stage, and frozen immediately. Leaves were homogenized to a fine powder using a nitrogen-cooled mortar and pestle and extracted in 250 μL of buffer containing: 50 mM HEPES-KOH, pH 7.4, 5 mM MgCl₂, 1 mM EDTA, 1 mM Dithiothreitol, 1% (v/v) glycerol. After centrifugation at 10,000 × g for 2 min at 4°C, the supernatant was collected for enzyme activity measurements. PEPC enzyme activity was assayed using a method modified from Meyer et al. (1988) and Ueno et al. (1997). The PEPC reaction mixture contained: 100 mM HEPES-NaOH, pH 7.5, 10 mM MgCl₂, 1 mM NaHCO₃, 5 mM G6P, 0.2 mM NADH, 12 unit/mL MDH (from pig heart; Roche Diagnostics, Basel, Switzerland)⁶, and the reaction was started by adding PEP to a final concentration of 4 mM. PPDK enzyme activity was assayed as described by Fukayama et al. (2001). NADP-MDH activity was determined by a method modified from Tsuchida et al. (2001). NADP-MDH reaction mixture contained: 50 mM HEPES-KOH, pH 8, 70 mM KCl, 1 mM EDTA, 1 mM DTT, and 0.2 mM NADPH, and the reaction was started by adding OAA to a final concentration of 1 mM. NADP-ME activity was measured by a method modified from Tsuchida et al. (2001; protocol 1). The activities of PEPC, PPDK, NADP-MDH, and NADP-ME were measured spectrophotometrically at 340 nm at 25°C, 30°C, 25°C, and 25°C, respectively.

Leaf gas-exchange measurements were made using a Li-6400XT infrared gas analyzer (LI-COR Biosciences, United States)⁷ fitted with a standard 2 × 3 cm leaf chamber and a 6400-02B light source. Measurements were made at a constant airflow rate of 400 μmol s^–1, leaf temperature of 25°C, leaf-to-air vapor pressure deficit between 1.0 and 1.5 kPa and relative humidity of 60–65%. Data were acquired between 08:00 h and 13:00 h. Measurements were made from the two youngest fully expanded leaves for each plant during the tillering stage. The mid-portions of leaves were acclimated in the cuvette for approximately 30 min before measurements were made. The response curves of the net CO₂ assimilation rate (A, μmol m^–2 s^–1) to changing intercellular pCO₂ concentration (C_i, μmol CO₂ mol air^–1) were acquired by decreasing C_a (pCO₂ concentration in the cuvette) from 2000 down to 20 μmol CO₂ mol air^–1 at a photosynthetic photon flux density (PPFD) of 1500 or 2000 μmol photon m^–2 s^–1. The CO₂ compensation point (Γ) and maximum carboxylation efficiency (CE) were calculated from the intercept (Vogan et al., 2007) and slope (Wang et al., 2006) of the CO₂ response curves. Light response curves were acquired by increasing the PPFD from 0 to 2000 μmol photon m^–2 s^–1 at C_a 400 μbar. The quantum efficiency for CO₂ assimilation (φ) and respiration rates (R_d) were calculated from the slope and intercept of the light-response curves (PPFD < 100 μmol photons m^–2 s^–1).

Carbon flux analysis was performed with a custom-built gas exchange freeze clamp apparatus (Supplementary Figure 7). Measurements were made from two youngest fully expanded leaves for each plant during the tillering stage. Two leaves of up to 22 cm in length were placed inside a gas exchange chamber (23.5 cm × 4.5 cm × 0.4 cm), the top was constructed from a piece of clear flexible plastic to allow light penetration and the bottom from a sheet of aluminum foil to accelerate cooling when freeze clamping. The foil and plastic were attached with foil tape to a three-sided aluminum frame to provide rigidity. Two holes were drilled through the side of the frame to accommodate the air inlet and outlet tubes. A third hole on the end enabled thermocouples to be threaded through the frame to measure leaf and air temperature inside the labeling chamber. The chamber was then placed in a mounting frame allowing the leaf to be inserted prior to sealing the chamber with a foam gasket secured with bulldog clips. The mounting frame was positioned horizontally between two LED banks capable of providing illumination to the upper leaf surface of up to 1,000 μmol photons m² s^–1.

Air was drawn from outside the laboratory through a compressor. The air stream passed through an oil water separator and flow control valve into a copper coil placed in an ice bath for cooling. The air stream could be directed into the leaf chamber or by-passed into a CO₂ conditioning unit. In the latter, CO₂ could be removed from the air with soda lime and then optionally enriched with ¹³CO₂ gas (300 ppm) prepared by mixing NaH¹³CO₃ (Sigma-Aldrich, United States; see text footnote 2) with 2-hydroxypropanoic acid (lactic acid). The flow of air passing over the leaf (3 ml/min) was adjusted with a flow controller, and the CO₂ concentration of the incoming and outgoing air streams measured with two CO₂ analyzers (WMA-5, PP-Systems, United States)⁸. The humidity of the air inside the chamber was maintained at ∼60% with the addition of water to the soda lime chamber or reduced by passing air through a chamber of silica gel.

The leaf chamber was mounted on the stand in such a way that the plane of the leaf was halfway between two pneumatically operated aluminum bars. These were cooled with liquid nitrogen, and when released they clamped together fitting inside the aluminum chamber frame, very rapidly freezing the leaf. A fan was mounted horizontally to the bars to blow the fog from the liquid nitrogen away from the chamber to ensure there was minimal disruption to the environment before freezing occurred. The lower bar was positioned in such a way as to push the chamber up on closure. The cold temperatures and force of the bars closing meant the chamber disintegrated enabling the leaf to be removed with tweezers and placed in a liquid nitrogen bath for 10 s before subsequent storage at −80°C. To perform ¹³CO₂ pulse-labeling, leaves were acclimated at steady-state conditions prior to scrubbing CO₂ from the incoming air stream, then subjected to ¹³CO₂ enriched air for a duration of 0 and 60 s and metabolic activity was quenched at these time points by freeze clamping the leaves as above. Freeze-quenched tissue was ground into a fine powder by mortar and pestle in liquid nitrogen. Finely ground leaf tissues were freeze-dried for 3 days and placed into sealed tubes. The sealed tubes containing finely ground lyophilized leaf tissue samples were shipped to Max Planck Institute of Molecular Plant Physiology, Germany for the metabolite analyses.

Aliquots (3 or 5 mg) of finely ground lyophilized rice leaf tissue were extracted with chloroform-methanol as described in Arrivault et al. (2017), and the lyophilized extracts were resuspended in 300 μL or 600 μL purified (Millipore, United States) water, respectively. Isotopomers were measured by reverse-phase LC-MS/MS (malate, aspartate, 3PGA, PEP, citrate + isocitrate; Arrivault et al., 2017; n.b. citrate and isocitrate were not resolved using this method) and anion-exchange LC-MS/MS (malate, PEP, citrate; Lunn et al., 2006 with modifications as described in Figueroa et al., 2016). Total amounts of malate, aspartate, citrate + isocitrate, and citrate were calculated by summing isotopomers. The total amounts of 3PGA and PEP were determined enzymatically in trichloroacetic acid extracts using a Sigma-22 dual-wavelength photometer (Merlo et al., 1993), with PEP being measured in freshly prepared extracts. ¹³C enrichment and relative isotopomer distribution were calculated as in Szecowka et al. (2013).

Statistical analysis for all experiments was performed in R version 3.0.0 (The R Foundation for Statistical Computing, Vienna, Austria) using a one-way analysis of variance (ANOVA) and a Tukey post hoc test or a Student’s t-test with a p-value of <0.05.

Immunoblotting of F₂ generation transgenic plants showed that the protein of the correct size for ZmPEPC, ZmNADP-MDH, ZmNADP-ME, and ZmPPDK was stably expressed in both the quadruple and quintuple cross rice lines, with OsGDCH protein almost undetectable in the quintuple cross (Figure 1). In all lines, protein abundance was lower than that of wild-type maize plants. We next sought to determine whether these proteins were localized to the correct cell type and subcellular compartment. Immunolocalization analysis of the quadruple cross line revealed that ZmPEPC was localized to the cytosol of MCs similar to the single ZmPEPC transgenic line (Giuliani et al., 2019b) and the triple cross line (Supplementary Figure 3). ZmPPDK, was localized to the chloroplast in both BSCs and MCs (Supplementary Figure 8). The ZmNADP-MDH and ZmNADP-ME antisera cross-reacted with native protein in wild-type rice and so it was not possible to distinguish protein encoded by the endogenous rice gene from that encoded by the maize transgene. Overexpression of these enzymes conferred enhanced enzyme activity (Supplementary Table 1). In the quadruple cross line, PEPC activity was 18.4-fold higher compared to wild-type rice plants, PPDK 5.8-fold higher, NADP-MDH 9.9-fold, and NADP-ME 4.1-fold (Table 1).

Given that protein levels and activities of all four introduced C₄ enzymes were enhanced compared to wild-type plants, we investigated whether this affected growth and photosynthesis. None of the crossed lines consistently showed altered chlorophyll content (Figure 2A). Tiller number in the quintuple cross lines (Figure 2B) and plant height in both quadruple and quintuple crosses (Figure 2C) were significantly reduced. Phenotypic perturbations were most marked in the quintuple cross lines (Supplementary Figure 9), although the plants still developed and flowered at the same time with wild-type plants. These growth perturbations were not observed in the single C₄ gene transgenic lines of ZmPEPC, ZmPPDK, ZmNADP-MDH, and ZmNADP-ME (Giuliani et al., 2019b) but were observed in the single Osgdch knockdown line (Lin et al., 2016).

To investigate whether overexpression of C₄ genes impacted photosynthesis, the response of net CO₂ assimilation rate (A) to CO₂ concentration under photorespiratory conditions (21% O₂) was measured. In the quadruple cross line there were no differences in CO₂ assimilation (Figures 3A,C), Γ, CE, R_d or Φ compared to wild-type plants (Table 2). These results suggested that accumulation of active maize C₄ cycle enzymes in rice does not significantly affect leaf level photosynthetic gas-exchange and that the phenotypic perturbations are not associated with reduced CO₂ assimilation.

For the quintuple cross, the CO₂ assimilation rate in response to increasing intercellular CO₂ concentrations (Ci) was reduced, most notably at lower CO₂ concentrations (<700 μmol CO₂ mol^–1, Figure 3B). Under non-photorespiratory conditions (2% O₂, Figure 3B) CO₂ assimilation rates were similar to wild-type plants. Consistent with this, the quintuple cross had a significantly higher Γ under high photorespiratory conditions but not under low photorespiratory conditions (Table 2). In response to changes in photon flux density, photosynthesis was saturated at lower light levels than in wild-type plants (400 μmol photons m^–2 s^–1 versus 1750 μmol photons m^–2 s^–1, respectively, Figure 3D) with significantly lower Φ and higher R_d than wild-type plants (Table 2).

To investigate these photosynthetic responses for the quintuple cross line in more detail, CO₂ responses were measured under conditions conducive to low and high rates of photorespiration. Under low light (400 μmol photons m^–2 s^–1) and high CO₂ (2000 μmol CO₂ mol air^–1), conditions conducive to low photorespiration, CO₂ responses of the quintuple cross line were similar to wild-type plants (Supplementary Figures 10A,D). Under high light (1500 μmol photons m^–2 s^–1) and low CO₂ (400 μmol CO₂ mol air^–1), conditions conducive to high photorespiration, CO₂ assimilation was lower in the quintuple cross (Supplementary Figures 10B,C). Correspondingly, Γ was higher, and carboxylation efficiency (CE) lower under conditions conductive to high rates of photorespiration but not under non-photorespiratory conditions (Supplementary Table 1). This is consistent with the photorespiratory-deficient phenotype observed in the single Osgdch knockdown lines (Lin et al., 2016).

We performed experiments to measure the flux of ¹³CO₂ through photosynthetic metabolism to investigate whether there was partial functionality of a C₄ pathway in these plants. There was significantly more incorporation of ¹³C into malate in the quadruple and quintuple cross lines than in wild-type plants (Figure 4 and Supplementary Figure 11), with the m₁ isotopomer being more abundant than the other ¹³C-labeled isotopomers (Table 3), consistent with increased fixation of ¹³CO₂ via PEPC in the transgenic lines expressing the maize PEPC enzyme. Labeling of aspartate was also significantly higher than wild-type in the quintuple line (Figure 4 and Supplementary Figure 11), with the m₁ isotopomer being more abundant that m₂-m₄ isotopomers (Table 3), indicating ¹³CO₂ fixation into oxaloacetate by maize PEPC and conversion to aspartate by endogenous rice aspartate aminotransferase activity. There was no evidence of significant difference in labeling of aspartate in the quadruple line. There was almost complete labeling of the 3PGA pool in wild-type plants after a 60 s pulse, and similarly high levels of labeling of 3PGA were observed in the quadruple and quintuple lines (Figure 4 and Supplementary Figure 11). There was also substantial labeling of PEP after a 60 s pulse, with no significant differences between the transgenic lines and wild-type plants (Figure 4 and Supplementary Figure 11). There was almost no labeling of citrate and isocitrate in wild-type plants and the quadruple line (Figure 4). In contrast, the enrichment of ¹³C in citrate and isocitrate in the quintuple line was 10-fold higher than in wild-type plants.