Abstract
Engineering C4 photosynthesis into C3 crops such as rice or wheat could substantially increase their yield by alleviating photorespiratory losses1,2. This objective is challenging because the C4 pathway involves complex modifications to the biochemistry, cell biology and anatomy of leaves3. Forward genetics has provided limited insight into the mechanistic basis of these characteristics and there have been no reports of significant quantitative intra-specific variation of C4 attributes that would allow trait-mapping4,5. Here we show that accessions of C4 Gynandropsis gynandra collected from locations across Africa and Asia exhibit natural variation in key characteristics of C4 photosynthesis. Variable traits include bundle sheath size and vein density, gas exchange parameters and carbon-isotope discrimination associated with the C4 state, but also abundance of transcripts encoding core enzymes of the C4 cycle. Traits relating to water use showed more quantitative variation than those associated with carbon assimilation. We propose variation in these traits likely adapted the hydraulic system for increased water use efficiency rather than improving carbon fixation, indicating that selection pressure may drive C4 diversity in G. gynandra by acting to modify water use rather than photosynthesis. As these accessions can be easily crossed and produce fertile offspring, our findings indicate that natural variation within a C4 species is sufficiently large to allow genetic-mapping of key anatomical C4 traits and regulators.
Plants that use C4 photosynthesis can effectively abolish photorespiratory losses caused when Ribulose 1,5-Bisphosphate Carboxylase/Oxygenase (RuBisCO) fixes oxygen rather than CO2 6,7. In C4 plants, RuBisCO is typically sequestered in bundle sheath (BS) cells that are concentrically arranged around the vasculature. Establishment of a molecular CO2 pump delivers carbon to RuBisCO from Mesophyll (M) cells via C4 acid intermediates8. C4 photosynthesis relies on an increased importance of the BS for photosynthesis, reduced dependence on M cells, more chloroplasts in BS cells, increased proliferation of plasmodesmata between M and BS cells, and a higher vein density to increase the volume of the leaf occupied by the BS. These morphological alterations to the leaf that facilitate the C4 cycle are known as Kranz anatomy9. Moreover, photosynthesis gene expression is modified such that genes encoding components of the C4 and Calvin-Benson-Bassham cycles are strongly and preferentially expressed in either M or BS cells8,10.
Despite the complex modifications associated with C4 photosynthesis, current estimates are that the C4 pathway has evolved independently more than sixty times in angiosperms11, which suggests a relatively straightforward route must allow the transition from the ancestral C3 to the derived C4 state. Genome-wide analysis of transcript abundance in multiple C3 and C4 species has provided unbiased insight into processes that likely change in C4 compared with C3 leaves12–14. Furthermore, cis-elements that control expression of genes encoding the C4 cycle have been documented. To date however, the regulators that recognize these motifs have not been isolated15. Despite progress in our understanding of C4 photosynthesis, it is currently not possible to rationally design a C4 pathway in a C3 leaf.
When natural variation is present, it enables quantitative methods such as Genome-Wide Association Studies (GWAS) and/or the development of a mapping population. Molecular marker-trait associations on the population allow identification of the causal genes underpinning the variation, which has been used extensively to map loci responsible for numerous complex traits in plants16. If such an approach could be applied to study C4 photosynthesis, then it would expedite discovery of key regulators to engineer increased photosynthetic efficiency in C3 plants. Interspecific hybrids have been generated between C3 and C4 species of the dicotyledon Atriplex17. Although progeny possessed variation in C4 phenotypes, specific traits showed limited penetrance and there were high rates of sterility18. In the grasses, Alloteropsis semialata shows natural variation in C4 parameters and has been classified into C3 or C4 subspecies19,20, but there are currently no reports that these populations have been bred. Thus, to our knowledge there are currently no examples that variation in C4 traits within a single species is sufficient to allow breeding and then molecular trait-mapping. We therefore investigated the extent to which key C4 traits varied in the C4 dicotyledonous Gynandropsis gynandra, which is a leafy green vegetable21 in a clade with both C3 and C4 species22–24. Here we show that accessions of G. gynandra show significant variation in both anatomical and physiological aspects associated with C4 photosynthesis. These accessions have short generation spans, are sexually compatible and produce fertile offspring. These findings indicate that in a dicotyledonous species that is phylogenetically close to the model Arabidopsis thaliana there is sufficient natural variation to allow the use of classical genetics to identify loci controlling the multifaceted C4 syndrome.
Accessions of G. gynandra were collected from African and Asian sub-continents were used (Supplementary Table 1). DNA sequencing and phylogenetic reconstruction generated a taxonomy that was generally consistent with geographical origin but also indicated that the accession from Benin in West Africa was more like the Asian accessions than those from East Africa (Fig. 1a). These accessions displayed considerable variation in macroscopic characters associated with leaf appearance (Fig. 1b, Supplementary Fig.1a). For example, fully expanded leaflets varied in size and shape, and there was also variation in petiole length, presence of trichomes and anthocyanin pigmentation. As there was considerable macroscopic variation in leaf characteristics, we then evaluated these accessions for variation in features of Kranz anatomy. Interestingly, there were statistically significant differences in vein density (Fig. 1c&d, Supplementary Fig.1b), cross-sectional area of BS strands (Fig. 1c&e, Supplementary Fig. 1c), size of individual BS cells (Fig. 1f) and stomatal density (Fig. 1c&g). Furthermore, East African accessions showed higher vein density, reduced distance between veins, and a greater stomatal density than Asian lines (Supplementary Fig. 3a-c). Asian accessions typically had larger BS areas and cell sizes than those from the African continent (Supplementary Fig. 3d&e). Vein density was inversely correlated with BS area and BS cell size but positively with stomatal density (Supplementary Table 2). The average number of BS cells around each vein showed no statistically significant differences between lines (Supplementary Fig. 3f), but cross-sectional area of the BS and the size of individual BS cells were positively correlated (ρ=0.8, P<0.0001). We therefore conclude that the area of individual BS cells, rather than the number of these cells per vein bundle, drives the increased BS strand area. This suggests that genetic determinants of cell size rather than cell proliferation are involved in the variation in BS tissue in G. gynandra. Thus, despite the lower phenotypic variation associated with C4 compared with C3 leaves25, our findings demonstrate flexibility is still possible within individual species that are fully C4.
We next investigated whether differences observed in Kranz anatomy affected photosynthetic performance. For all accessions, their CO2 response curves (assimilation (A) response to the concentration of CO2 inside the leaf Ci) were typical of C4 plants with high carboxylation efficiencies and low CO2 compensation points Γ (Fig. 2a, Supplementary Fig. 4a). Although parameters associated with instantaneous gas exchange such as maximum rate of photosynthesis (Amax), rate of photosynthesis under the conditions of growth (A400), CO2 carboxylation efficiencies and Γ showed little variation between accessions (Fig. 2b-e, Supplementary Fig. 4b-e), there were statistically significant differences in transpiration (Fig. 2f), stomatal conductance (Fig. 2g) and water use efficiency WUE (Fig. 2h). Furthermore, there was also significant variation in the carbon isotope discrimination against 13C (δ13C) in leaf dry matter (Fig. 2i), which is a measure of the efficiency of the C4 carbon pump over the life-time of the leaf. Asian accessions showed reduced discrimination against δ13C compared with East African lines (Supplementary Fig. 4i). These data therefore indicate that the accessions of G. gynandra possess significant variation in parameters linked to the balance between water use and photosynthesis that influenced the efficiency of the C4 cycle over the lifetime of a leaf.
We next sought to investigate the extent to which transcript abundance of core genes of the C4 cycle differed between the accessions. Interestingly, there were statistically significant differences in the abundance of transcripts encoding Phosphoenolpyruvate carboxylase (PEPC) which catalyses the first committed step of the C4 cycle, the BS-specific decarboxylase NAD-dependent Malic Enzyme (NAD-ME), the small subunit of RuBisCO (RbcS), and pyruvate,orthophosphate dikinase (PPDK) that regenerates PEP the primary acceptor of (Fig. 3a,c,e,g). In all cases, these differences in C4 transcript abundance were associated with geographical location and phylogenetic position of the accessions, with Asian and West African accessions accumulating greater levels of C4 transcripts than East African accessions (Fig. 3b,d,f,h). Understanding how photosynthesis enzymes become strongly expressed and patterned to either mesophyll or bundle sheath cells of C4 species is a longstanding area of research. However, although progress has been made in understanding cis-elements responsible, there is little known about the transcription factors involved. The intraspecific variation in expression of genes encoding enzymes of the C4 cycle in G. gynandra therefore provides an opportunity to identify trans-factors important for C4 photosynthesis.
Despite accessions functioning with similar photosynthetic efficiencies under ambient CO2 and light conditions, when assessed by phylogenetic grouping those with more pronounced Kranz traits (e.g., larger BS tissues and lower vein densities) exhibited increased Amax, WUE and δ13C (Supplementary Fig. 4c,h,i) and stronger expression of the C4 cycle (Fig. 3b,d,f,h). To summarize, compared with East African accessions, Asian and West African accessions tended to have higher WUE, lower density of stomata and veins, and larger BS areas and cell sizes. Lastly, consistent with the higher δ13C, which is indicative of a stronger C4 cycle, the Asian and West African accessions had increased expression of genes encoding C4 enzymes (Fig. 4).
The considerable variation reported in this study offers a valuable germplasm resource to identify regulators of the C4 pathway and Kranz anatomy through genetic mapping. All accessions in this study hybridize easily. Emasculation and pollination need only take 15-30 seconds per flower. For example, the most divergent accessions regarding anatomy ‘Malaysia-1’ X ‘Malawi’, ‘Malaysia-2’ X ‘Malawi’, and their reciprocal crosses produce an average 52 ± 11 seeds per silique (n=6), whose offspring are fully fertile. These F1 hybrid populations provide an excellent breeding foundation to delineate regulatory mechanisms, and also provide an opportunity to test whether the C4 trait is induced by a master switch26, or the action of multiple independent processes15,27. The discovery of intra-specific variation in a C4 grass would be particularly useful in mapping traits relevant to improving photosynthesis in cereals and thus introduce C4 photosynthesis into C3 crops.
While our understanding of the regulatory mechanisms underlying C4 metabolism is growing, there is still a significant gap in tools to expand understanding of the regulation behind Kranz anatomy and the C4 biochemical cycle. Methods such as Quantitative Trait Loci (QTL) mapping or GWAS in G. gynandra or in an equally diverse C4 species may provide beneficial insights for the regulation of Kranz development. Most trait variation in G. gynandra was associated with characteristics relating to water use that impact on carbon capture. It is noteworthy that modifications to C3 leaves considered to represent early steps on the path towards the C4 phenotype are also associated with water use rather than CO2 fixation27,28. As natural vegetation is not considered to be under strong selection pressure to optimize photosynthesis29,30, it seems likely that C4 trait variation continues to be driven by optimizing water use rather than photosynthesis per se.
AUTHOR CONTRIBUTIONS
GR, PS and JMH designed the study. GR, PS, TAR and EODS carried out experimental work. GR, PS, TAR, MES and JMH wrote the manuscript.
MATERIALS AND METHODS
Plant accessions and growth conditions
A selection of nine diverse accessions of G. gynandra were made from a larger germplasm collection based on initial phenotypic and genetic screening. Five accessions were from Africa and four from Asia (Supplementary Table 1, materials available on request from MES). Plants from all G.gynandra accessions were grown under identical conditions prior to sampling. After germination, all seeds were planted in 5:1 F2 compost (Levington Advance, UK) to fine vermiculite premixed with 0.17 g/L insecticide (Imidasect 5GR, Fargro, UK). Seedlings were kept in a growth cabinet at 350 µmol photons m-2 s-1 light with a 16 h photoperiod, at 25 °C, 60% relative humidity (RH), ambient [CO2]. A single dose of 3 mL/L slow release 17N-9P-11K fertilizer (All Purpose Continuous Release Plant Food, Miracle-Gro, UK) was applied after 1.5 weeks. Plants for physiological measurements were grown under identical conditions to those for Kranz measurements for the first three weeks, after which they were re-planted in 13 cm3 pots with 5:1 M3 soil (Levington Advance, UK) to medium vermiculite soil mixture and moved to a growth room set to 23 °C, 60% RH, ambient [CO2], 350 µmol photons m-2 s-1 PAR with a 16 h photoperiod.
Preparation of leaf tissue sections
Three weeks after germination, tissue was harvested from healthy plants from the centre trifoliate leaves of the second pair of fully expanded true leaves. A 3 mm2 rectangle was cut from the leaf adjacent to the midvein with a razor blade for transverse sections. Two slightly larger rectangles were cut from identical regions for paradermal sectioning and qRT-PCR analysis.
For transverse sections, leaf tissue in plastic cuvettes were submerged in a 4% paraformaldehyde in PBS solution (Sigma-Aldrich, St. Louis, MO, USA), placed in a vacuum chamber for 1 h, and incubated at 4 °C overnight for fixing. Cuvettes then underwent an ethanol dehydration series from 30% to 90% (v/v) ethanol solutions (Thermo Fisher Scientific, Waltham, MA, USA) in 10% (v/v) increments for 45 minutes each at 4 °C with a final overnight treatment at 4 °C in 95% ethanol with 0.1% eosin dye solution (Sigma-Aldrich, St. Louis, MO, USA). The dye solution was washed thrice with 100% (v/v) ethanol at room temperature. The samples were embedded in resin in accordance with the Technovit 7100 (Kulzer GmbH, Wehrheim, Germany) manufacturer’s protocol. Hardened resin blocks were cut with a manual rotary microtome (Thermo Fisher Scientific, Waltham, MA, USA). Sections were placed on microscope slides and stained with 0.1% (w/v) toluidine blue solution (Sigma-Aldrich, St. Louis, MO, USA) prior to imaging on a light microscope.
For paradermal sections, fresh tissue samples were placed in plastic cuvettes and incubated in 3:1 100% (v/v) ethanol to acetic acid solution before treatment with 70% (v/v) ethanol solution (refreshed once) at 37 °C overnight. To clear the samples, cuvettes were submerged in 5% NaOH solution for three hours at 37 °C. After storage in 70% ethanol solution, the samples were stained with 95% (v/v) ethanol and 0.1% (v/v) eosin dye solution (Sigma-Aldrich, St. Louis, MO, USA). Samples were stored overnight at 4 °C and washed with 70% (v/v) ethanol thrice before transfer to slides for imaging. To determine stomatal density impressions of the abaxial epidermis of each central leaflet were generated by applying a thin coat of transparent nail varnish (Boots, Nottingham, UK). After drying, the varnish was peeled off and mounted onto glass slide for imaging.
Measurement of Kranz anatomy traits
Slides of all leaf sections were imaged with an Olympus BX41 light microscope with a mounted Micropublisher 3.3 RTV camera (Q Imaging, Surrey, BC, Canada). Images were captured with Q-Capture Pro 7 software, and measurements were analyzed with the software ImageJ31. To maximize comparability, strict criteria were applied for all image analyses. Microscopy of transverse leaf sections was used to quantify the BS both in terms of average BS tissue area (the total cross-sectional area of all BS cells immediately surrounding a vein) and BS cell size (the average cross-sectional area of individual BS cells around the vein). To quantify BS tissue area, the freehand selection tool was used to subtract the integrated area of each vein from the integrated area of all BS cells in direct contact with the vein on images with 200X total magnification. This value was divided by the number of BS cells in each vein bundle to obtain the average BS cell size. For intervein distance (the distance between the centers of adjacent veins in transverse sections), only vein bundles were measured for which the following criteria did not apply: wide (indicates branching) or extremely large veins, veins with distorted BS cells due to contact with adjacent BS tissues (indicates merging), veins with damaged BS cells. The line selection tool was used to measure the linear distance between the centers of adjacent veins on images with 40X total magnification. Vein density (vein length per unit area of leaf) was quantified on paradermal sections on images with 100X total magnification. Slides were imaged with the same microscopy equipment as transverse sections but set to Ph3 (phase contrast). Three images (from three different leaves per plant) were randomly selected for measurement. The freehand line tool was used to trace all veins (both major and minor) along their center. As it was not possible to trace all veins in an image simultaneously, individual vein sections were progressively measured without overlap and the individual lengths summed. The total vein length was divided by the image area to obtain the density. Stomatal density (the number of stomata per unit area of leaf) was measured on three subsampled leaves from three random plants on images with 200X total magnification. The total number of stomata were divided by the image area to obtain the density.
Photosynthetic performance
A LI-6800 portable photosynthesis infrared gas analyzer (IRGA) system (LI-COR, Lincoln, NE, USA) equipped with a multiphase flash fluorimeter was used to assess physiological differences for photosynthetic parameters between G. gynandra accessions. All physiological measurements were performed on the central leaflet of five-week old plants with three biological replicates. For stomatal conductance, transpiration (E), and assimilation (A400), measurements were taken during ambient conditions of growth (400 ppm atmospheric [CO2], Ca; photosynthetic photon flux density (PPFD) 350 µmol m-2s-1). Water use efficiency (WUE) was defined as A400/E. A combination chlorophyll fluorescence and assimilation / intracellular CO2 concentration (A/Ci) curve was measured for three plants from each accession. Atmospheric CO2 (Ca) reference values were: 400, 400, 300, 200, 100, 50, 25, 400, 400, 400, 600, 800, 1000, 1200, 400 ppm, with a saturating rectangular pulse of 12,000 µmol m-2s-1 at each reference point. Otherwise, measurements were made at a PPFD of 2000 µmol m-2s-1, 23 °C and 60% RH at each reference point. All leaves covered the full area of the fluorimeter/IRGA cuvette. Measurements were carried out on consecutive days between one and eight hours post dawn, measuring one random plant from each accession per day. Maximal assimilation (Amax) was calculated as the asymptote of the A/Ci response curve. The CO2 compensation point (Γ) was calculated from the regression of A and Ci measurements ranging between C a values of 200 and 25 ppm at A= 0. Adjusted R2 values for the regression line ranged between 0.9932 and 0.9967. Carboxylation efficiency was calculated as the partial derivative at A = 0. Stable carbon isotope (δ13C) analysis was performed according to methods previously described32 on three biological replicates per accession with 500 µg of dried leaf tissue.
Statistical Analysis
For all tests, individual plants were considered experimental units in a complete randomized design. Data were analyzed in SAS (University Version, SAS Institute, Cary, NC, USA) and in R (Version 3.4.2, R Studio, Inc., Boston, MA, USA). A One-Way Analysis of Variance (ANOVA) compared all means from anatomical and physiological measurements among G. gynandra accessions and a Student’s t-test was used to compare means of accessions by continent (α=0.05). Null hypotheses were rejected for specific ANOVA or t-tests for any population with P value ≤ 0.05. Levene’s Test was used to evaluate homoscedasticity33. Duncan’s Multiple Range post-hoc test was used for mean separation on accessions (α=0.05) with statistically significant ANOVAs34. Pearson product-moment correlation coefficients were calculated to find associations among features of Kranz traits35.
Analysis of transcript abundance
Leaf tissue samples for RNA extraction were harvested simultaneously with samples for Kranz trait measurements. The fresh samples were immediately frozen with liquid nitrogen and stored at −80 °C. Total RNA was extracted from three tissue samples per accession with a RNeasy Mini Kit (QIAGEN, Hilden, DE) according to the manufacturer’s instructions. An On-Column DNase Digestion protocol was applied to remove genomic DNA contamination (QIAGEN, Hilden, DE) before cDNA was synthesized with Invitrogen Superscript II RT enzyme according to the manufacturer’s instructions (Thermo Fisher Scientific Inc., Waltham, MA, USA). All cDNA samples were stored at −20 °C before qRT-PCR. Primers were designed for Quantitative PCR of C4 cycle genes PEPC, NAD-ME, RbcS and PPDK (Supplementary Table 3), and reactions carried out as reported previously36 on three biological and three technical replicates.
ACKNOWLEDGEMENTS
We thank Frank Becker and Beatrice Landoni (Wageningen University) for providing ITS sequence data and initial information on the selected lines, Blandine Gilbert for help with image processing and James Rolfe for carbon isotope analysis. GR was supported by a Gates Cambridge Trust PhD Fellowship and PS by Advanced ERC grant 694733 Revolution to JMH.