Brassicaceae display variation in efficiency of photorespiratory carbon-recapturing mechanisms

Abstract Carbon-concentrating mechanisms enhance the carboxylase efficiency of Rubisco by providing supra-atmospheric concentrations of CO2 in its surroundings. Beside the C4 photosynthesis pathway, carbon concentration can also be achieved by the photorespiratory glycine shuttle which requires fewer and less complex modifications. Plants displaying CO2 compensation points between 10 ppm and 40 ppm are often considered to utilize such a photorespiratory shuttle and are termed ‘C3–C4 intermediates’. In the present study, we perform a physiological, biochemical, and anatomical survey of a large number of Brassicaceae species to better understand the C3–C4 intermediate phenotype, including its basic components and its plasticity. Our phylogenetic analysis suggested that C3–C4 metabolism evolved up to five times independently in the Brassicaceae. The efficiency of the pathway showed considerable variation. Centripetal accumulation of organelles in the bundle sheath was consistently observed in all C3–C4-classified taxa, indicating a crucial role for anatomical features in CO2-concentrating pathways. Leaf metabolite patterns were strongly influenced by the individual species, but accumulation of photorespiratory shuttle metabolites glycine and serine was generally observed. Analysis of phosphoenolpyruvate carboxylase activities suggested that C4-like shuttles have not evolved in the investigated Brassicaceae. Convergent evolution of the photorespiratory shuttle indicates that it represents a distinct photosynthesis type that is beneficial in some environments.


Introduction
The majority of plant species on Earth, including many crops, employ C 3 photosynthesis.In these plants, under the pre sent environmental conditions, the central photosynthetic enzyme Rubisco fixes approximately one molecule of oxy gen for every three molecules of CO 2 (Sharkey, 1988).Here, whilst the carboxylase reaction of Rubisco produces two mol ecules of 3phosphglycerate (3PGA) which feeds into the Calvin-Benson-Bassham cycle (CBB), the oxygenase reac tion produces 2phosphoglycolate (2PG).2PG is a competi tive inhibitor of some CBB enzymes (Flügel et al., 2017) and hence must be rapidly removed.Further, carbon contained in 2PG must be recycled into 3PGA to prevent depletion of CBB intermediates.These functions are fulfilled by the pho torespiratory pathway.Photorespiration consists of coordinated enzyme activities that are located in different cellular compart ments.In the plastids, 2PG is converted into glycolate followed by oxidation and transamination in the peroxisome, produc ing glycine.Glycine is transported into the mitochondria and metabolized into serine by glycine decarboxylation.Serine is finally converted into glycerate in the peroxisome and into 3PGA in the plastid.During glycine decarboxylation, previ ously fixed carbon and nitrogen are converted into CO 2 and NH 3 , respectively.The refixation of carbon and nitrogen into organic forms, however, requires energy.Therefore, photorespi ration is often considered a wasteful process in terms of energy, carbon, and nitrogen balance.Further, the affinity of Rubisco for O 2 increases with rising temperatures; in addition, stom atal closure during water shortages can lead to a drop in CO 2 concentrations inside the leaf.Thus, climate change could con tribute to increased activity of the Rubisco oxygenase reaction.
Given the negative impact of photorespiration on plant productivity, there has been considerable interest in reducing flux through this pathway.For instance, various avenues to re duce photorespiratory losses which are being explored include increasing the capacity of the plant to recapture photorespira tory CO 2 , modifying Rubisco kinetic properties, and introduc ing carbonconcentrating mechanisms to limit the oxygenase reaction of Rubisco by creating an CO 2 rich environment around the enzyme (Busch et al., 2013).A better understanding of naturally occurring carbonconcentrating mechanisms will help in the design of these biotechnological approaches.
In C 3 species, photosynthesis and photorespiration mainly take place in the mesophyll (M) of the leaves.Recapture of photorespiratory CO 2 can be facilitated by arrangement of a continuous layer of plastids at the cell periphery next to the intercellular space.This arrangement acts to block diffusion of CO 2 out of the cell because the CO 2 would need to pass through the plastids where it can be reassimilated by Rubisco (Sage and Sage, 2009;Busch et al., 2013).In rice, Rubisco containing extensions (stromules) can increase the area of the plastidial barriers, preventing the efflux of CO 2 produced in the mitochondria (Sage and Sage, 2009).C 3 species such as wheat and rice achieve photorespiratory reassimilation rates of 24-38% (Busch et al., 2013).
The number of chloroplasts in the C 3 bundle sheath (BS) varies between species (Leegood, 2008), but they tend to be smaller and fewer in number compared with the M. The con tribution of BS chloroplasts to leaf photosynthesis is consid ered to be small (Kinsman and Pyke, 1998;Janacek et al., 2009;Aubry et al., 2014).Nevertheless, BS cells possess important roles in leaf hydraulics, phloem loading, intraleaf signalling, and transport processes (Leegood, 2008;Aubry et al., 2014;Lundgren et al., 2014).Increases in BS organelle numbers in dicate enhanced photosynthetic and photorespiratory activity.The centripetal arrangement of BS organelles helps to reduce loss of photorespiratory CO 2 (Muhaidat et al., 2011;Sage et al., 2014).Such BS cells have also been labelled as 'activated' or 'protoKranz' anatomy (Gowik and Westhoff, 2011;Sage et al., 2014).
Further increases in BS CO 2 concentration are possible by shifting the glycine decarboxylation step from the M to the BS (Monson et al., 1984;Rawsthorne et al., 1988a;Sage et al., 2014).This rearrangement forces photorespiratory gly cine produced in the M to diffuse to the BS, where the tissue specific increase in glycine decarboxylation activity promotes an elevated concentration of CO 2 around the BS Rubisco and thus supresses its oxygenase reaction.BSlocalized activity of glycine decarboxylation is mainly associated with cell speci ficity of the glycine decarboxylase (GDC) Pprotein (GLDP) (Rawsthorne et al., 1988a;Schulze et al., 2013).The installation of a glycine shuttle is accompanied by a further increase in organelle numbers in the BS.The majority of mitochondria are therefore located between the centripetally arranged plas tids and the veinorientated cell wall (Sage et al., 2014).Such a combination of carbon supply by the glycine shuttle and efficient CO 2 scavenging by an adequate organelle arrange ment improves the leaf carbon conservation and can be meas ured as reduction in the carbon compensation point (CCP or Г).Plants employing the photorespiratory glycine shuttle are often classified as C 2 species or C 3 -C 4 intermediates of type I (Edwards and Ku, 1987;Sage et al., 2014).Their CO 2 reassimi lation capacity was estimated to be ~73% in Moricandia arvensis (Hunt et al., 1987).
GDC activity is thought to be absent or considerably reduced in the M cells of welldeveloped C 2 species, perhaps as a result of a lossoffunction mutation or insertion of a transposable element early in C 2 evolution (Rawsthorne, 1992;Sage et al., 2012;Adwy et al., 2015).Consistently, preferential BS localiza tion of GLDP has been observed in many welldeveloped C 2 species (Lundgren, 2020;Schlüter and Weber, 2020).The as similatory power of the BS can generally be further enhanced by decarboxylation of additional metabolites.Organelle accu mulation and enhancement of organellar metabolism in the BS could increase the availability of such compounds.The glycine shuttle not only transports carbon between the M and BS, but also creates a nitrogen imbalance between these cells where adjustment of leaf nitrogen metabolism in C 3 -C 4 intermedi ates was proposed to occur by additional metabolite shuttling between M and BS cells (Mallmann et al., 2014).Shuttling of malate, aspartate, pyruvate, αalanine, αketoglutarate, and glu tamate could contribute to rebalancing of nitrogen and energy balances between the two cell types (Mallmann et al., 2014;Johnson et al., 2021).
In the M cells, a rise in phosphoenolpyruvate carboxylase (PEPC) activity can contribute to the provision of these shuttle metabolites.PEPC fixes carbon by catalysing the addition of bicarbonate to phosphoenolpyruvate (PEP), forming the C 4 acid oxaloacetate that is usually quickly converted into malate or aspartate.In contrast to Rubisco, PEPC possesses higher substrate specificity and affinity.Combined with decarboxyl ation reactions in the BS, high PEPC activity in the M cells can implement a carbon shuttle mechanism transporting C 4 metabolites into the BS where CO 2 is released.Plants using the glycine shuttle in combination with such a C 4 shuttle have been identified mainly in the Asteraceae genus Flaveria.They are also termed C 2 +C 4 or C 3 -C 4 type II species (Edwards and Ku, 1987;Sage et al., 2014;Bellasio, 2017).
Additional anatomic rearrangements and consequent sep aration of the PEPC and Rubisco reactions into the M and BS cells finally support an efficient C 4 cycle (Taniguchi et al., 2021).In the M cells, CO 2 is converted into bicarbonate by carbonic anhydrase and is then fixed by PEPC.The bound carbon diffuses into the BS mainly in the form of malate or as partate.Decarboxylation is then mediated by the NADPmalic enzyme in plastids, NADmalic enzyme in the mitochondria, or phosphoenolpyruvate carboxykinase in the cytosol.The cycle is completed by diffusion of a C 3 metabolite back to the M cells where ATP is needed for PEP regeneration by pyru vate phosphate dikinase.Plants with a strong C 4 shuttle, but which still exhibit Rubisco in the M, are classified as C 4 like (Moore et al., 1989).In bona fide C 4 species, all CO 2 is first assimilated by PEPC, with subsequent shuttling of the resulting C 4 acid to the BS where CO 2 is then delivered to Rubisco by a decarboxylase reaction.High CO 2 partial pressure in the BS strongly represses the oxygenase reaction, the following pho torespiratory pathway, and the concomitant loss of CO 2 and NH 3 .
The efficiency of the C 4 shuttle also depends on anatom ical features, especially the close connection between M and BS cells.In the majority of C 4 species, the BS forms a tight cell layer around the veins without direct exposure to the intercellular space (Sage et al., 2014).The proportion of M tissue is reduced to a second cell layer around the BS cells, and C 4 species usually have high vein densities.Since CO 2 fixation in C 4 species can continue at lower internal CO 2 concentration (Ci) and stomatal conductance, water use efficiency (WUE) is improved compared with C 3 spe cies.Operation of Rubisco under high CO 2 partial pressure allows high efficiency for the carboxylase reaction with lower amounts of protein, thus also improving the nitrogen use efficiency of C 4 photosynthesis.
The continuous fitness gain in the intermediate forms seems to have been an important prerequisite for the evolution of the complex C 4 biochemistry and anatomy (Heckmann et al., 2013;Williams et al., 2013;Mallmann et al., 2014;Blätke and Bräutigam, 2019;Dunning et al., 2019a).Species using the photorespiratory glycine shuttle have been identified in var ious monocot and dicot plant lineages, often but not always in phylogenetic proximity to C 4 species (Sage et al., 2011).Convergent evolution of the C 3 -C 4 pathway indicates sub stantial improvement of leaf carbon economy at least under certain environmental conditions (Bellasio and Farquhar, 2019;Lundgren, 2020).Reduction in the CCP and centripetal ac cumulation of BS organelles seem to be general features of C 3 -C 4 plants, but knowledge about the anatomical and bio chemical plasticity of the pathway and their influence on leaf physiological is still limited (Schlüter and Weber, 2016).
In our study, we concentrated on the Brassiceae tribe that evolved ~23 million years ago in the circumMediterranean region (Arias and Pires, 2012).It includes multiple lineages of C 3 -C 4 intermediates, but no known C 4 species (Apel et al., 1997).Aiming at large sample sizes from the group of C 3 and well as C 3-C 4 species, we selected taxa from all currently known C 3 -C 4 intermediates and related C 3 species from the genera Moricandia, Diplotaxis, and Brassica.Among the Brassiceae are also numerous crops species such as canola or rapeseed (Brassica napus), cabbage (Brassica oleracea), radish (Raphanus sativus), mustard (Sinapis alba), and the salad vegetable rocket or arugula (Eruca sativa, Diplotaxis tenuifolia).With the exception of D. tenuifolia, these are all C 3 crops.
In the present study, we analysed 34 taxa representing 28 Brassicaceae species.Our investigation of taxa from diverse pho tosynthesis types allowed us to assess the variation in 75 photo syntheticrelated parameters across C 3 and C 3 -C 4 species.The Cleomaceae Gynandropsis gynadra was included for comparison with the anatomy, biochemistry, and physiology of C 4 species.We use the CCP to rank the species and accessions according to their carbonconcentrating capacity.By parallel analysis of leaf structural features, we investigate the correlation between carbonconcentrating capacity and vein density, leaf thick ness, and organelle arrangement in the BS.As the installation of the glycine shuttle requires further metabolic adjustments in the leaf such as the nitrogen balancing between M and BS cells, the primary metabolite pattern of the leaf sections was also analysed.Consequences of the anatomical and biochem ical changes for the leaf physiology including assimilation and WUE were assessed.The large number of plant taxa will allow us also to learn about lineagespecific developments of the C 3 -C 4 pathways and potential variation within the trait.Detailed knowledge about the carbonconcentrating mechanisms ex isting in Brassiceae can help to identify interesting traits for en gineering or breeding approaches.

Plant cultivation
All seeds used in this study were obtained from either Botanical gardens, seed stock centres, or seed companies.Since physiology can vary between populations from the same species (Lundgren et al., 2016;Yorimitsu et al., 2019;Gomez et al., 2020), multiple accessions were used from some spe cies.The complete list of plant taxa comprised: Arabidopsis thaliana (L.) Heynh.(At), Brassica gravinae Ten.four accessions, Bg1, Bg2, Bg3, and Bg4 From the Cleomaceae, the C 4 species Gynandropsis gynandra (L.) briq.(Gg) was also included in the present study.A complete list of origins for these seeds can be found in Supplementary Table S1.
All seeds were vapour sterilized by incubation in an exicator with a fresh mixture of 100 ml of 13% Nahypochloride with 3 ml of 37% HCl for 2 h.The sterilized seeds were then germinated on plates con taining 0.22% (w/v) Murashige and Skoog medium, 50 mM MES pH 5.7, and 0.8% (w/v) agar.After 7-10 d, the seedlings were transferred to pots containing a mixture of sand and soil (Floraton 1 soil mixture, Floraguard, Germany) at a ratio of 1:2.All plants were firstly cultivated in climate chambers (CLF Mobilux Growbanks, Germany) under 12 h day conditions with 23 °C/20 °C day/night temperatures and ~200 µmol s -1 m -2 light.After establishment in soil, 2weekold plants were trans ferred to the greenhouse of the Heinrich Heine University with a 16 h day/8 h night cycle.Natural light conditions in the greenhouse were supplemented with metal halide lamps (400 W, DH Licht, Germany) so that the plants received between 250 µmol m -2 s -1 and 400 µmol m -2 s -1 .Minimum temperatures of the greenhouse were controlled to 21 °C during the night and 24 °C during the day.
The initial main experiment was conducted between October 2018 and March 2019.A small number of additional accessions were also studied between July and October 2020 following the same protocol.As controls, G. gynandra, D. tenuifolia, and H. incana (HIR3) were included in both experiments.Gas exchange parameters obtained for these three species, especially CCPs, were stable across the experiments.Thus, results from both experiments were considered comparable.Gas exchange was measured on the youngest fully expanded rosette leaves before onset of flowering.After gas exchange measurements were performed, plants were taken back to the greenhouse for 2 d.Following that, leaf material was harvested for metabolite analysis (only experiment in 2018/2019), el emental analyser isotope ratio mass spectrometry (EAIRMS) analysis, leaf vein determination, and embedding for light microscopy.A third experiment was conducted on plant accessions selected from the initial experiments in September to November 2021 in the new greenhouses of the HeinrichHeine University equipped with natural light LED lamps using the same experimental design.Leaves were snapfrozen in liquid nitrogen directly in the greenhouse in the late morning Samples for pro tein and PEPC assay were snapfrozen in liquid nitrogen at midday.An additional leaf was used for determination of specific leaf area (SLA).For the majority of taxa, 4-6 plants were analysed as replicates in the different experiments; for a few taxa, only two plants were available for analysis.In the graphs, each data point represents a measurement from a single plant.

Phylogenetic inference
Plants for genome sequencing were grown in a climatecontrolled chamber from the same seed stock used for physiological analysis.Linked read sequencing was performed by 10× Genomics and BGI, comple mented by PacBio and Nanopore long read sequencing for some species (Guerreiro et al., 2023).Most assemblies are linked read assemblies, some being scaffolded and gapfilled with the long read data, while two assem blies are long read assemblies polished and scaffolded by linked reads.
Finally, the filtered proteomes were fed into Orthofinder v2.5.1 (Emms andKelly, 2015, 2019) for orthology identification based on all versus all sequence BLASTp searches and MCL clustering (Emms and Kelly, 2015).Multiple sequence alignments for identified orthogroups (HOGs) were produced with MAFFT and used for creating gene trees with RaxML with the PROTGAMMALG substitution model.The gene trees of HOGs with singlecopy genes for at least 80% of species (102 HOGs) were fed to ASTRALpro (Zhang et al., 2020) for the creation of a species phylogeny with quartetbased local posterior probability values (Sayyari and Mirarab, 2016) for each node.

Photosynthetic gas exchange
Gas exchange was measured on the youngest fully expanded rosette leaf ~6-10 weeks after sowing and before the onset of flowering.The settings of the LI6800 (LICOR, Lincoln, NE, USA) were as follows: flow of 300 µmol s -1 , fan speed of 10 0000 rpm, light intensity of 1500 µmol m -2 s -1 , leaf temperature of 25 °C, and vapour pressure deficit of 1.5 kPa.After adjustment of leaves to the conditions in the leaf chamber, A-Ci curves were measured at reference atmospheric CO 2 concentrations of 400, 200, 100, 75, 50, 40, 30, 20, 10, 0, 400, 400, 600, 800, 1200, and 1600 ppm.For the experiment in 2018/2019, the LI6800 was equipped with a fluo rescence head measuring F v ʹ/F m ʹ and electron transport rate (ETR) at each CO 2 level.
For calculation of the CCP and the carboxylation efficiency (CE=initial slope of A-Ci), a minimum of four data points in the linear range close to the interception with the Ci axis were used.Maximal assimilation was determined at CO 2 concentrations between 1200 ppm and 1600 ppm.Assimilation (A), stomatal conductance (g sw ), Ci, WUE=A/g sw , and the ratio between internal and external CO 2 concentrations (Ci/Ca) from the measurements at 400 (ambient CO 2 ), 200, 100, and 50 ppm CO 2 were used for more detailed physiological analysis of the investigated plant accessions.

Metabolite and element analysis
After the gas exchange measurements, plants were allowed to read just to greenhouse conditions before sampling for metabolite patterns.Leaves were snapfrozen in liquid nitrogen directly in the greenhouse in the late morning and stored at -80 °C.The leaf samples then were homogenized into a fine powder by grinding in liquid nitrogen.Soluble metabolites were extracted in 1.5 ml of extraction solution consisting of water:methanol:chloroform in a 1:2.5:1 mixture following the method of Fiehn et al. (2000).A 30 µl aliquot of the supernatant was dried com pletely in a vacuum concentrator and derivatized for GCMS measure ments (Gu, 2012).GCMS measurements were performed as described by Shim et al. (2020) using a 5977B GCMSD (both Agilent Technologies).Metabolites were identified via MassHunter Qualitative (v b08.00,Agilent Technologies) by comparison of spectra with the NIST14 Mass Spectral Library (https://www.nist.gov/srd/niststandardreferencedatabase1av14).A standard mixture containing all target compounds at a concentration of 5 µM was processed in parallel to the samples as a response check and retention time reference.Peaks were integrated using MassHunter Quantitative (v b08.00,Agilent Technologies).For rel ative quantification, all metabolite peak areas were normalized to the corresponding fresh weight used for extraction and to the peak area of the internal standard ribitol or dimethylphenylalanine (SigmaAldrich).The same homogenized leaf material was used for determination of δ 13 C and CN ratios.After lyophilization, the material was analysed using an Isoprime 100 isotope ratio mass spectrometer coupled to an ISOTOPE cube elemental analyser (both from Elementar, Hanau, Germany) ac cording to Gowik et al. (2011).

Vein density measurements
The top third of mature rosette leaves were used for vein density mea surements.The leaf material was cleared in an acetic acid:ethanol mix (1:3) overnight followed by staining of cell walls in 5% safranin O in eth anol, and destaining in 70% ethanol.Pictures were taken using a Nikon eclipse TiU microscope equipped with a ProgRes MF camera from Jenoptik, Germany, at ×4 magnification.The vein density was analysed with ImageJ software determining the total vein length per total micro graph area.In most cases, six leaves were analysed for vein density per line with a minimum of three pictures measured and averaged per leaf.

Specific leaf area
Whole mature rosette leaves were cut and their outlines were copied on checked paper.The FW was measured immediately after and DW was subsequently determined after 48 h at 60 °C.The leaf area was deter mined using ImageJ software.For calculation of SLA, the area was divided by the dry weight.

Analysis of leaf cross-section
For light microscopy, sections of ~1 × 2 mm were cut from the top third of mature rosette leaves and immediately fixed in 2% paraformalde hyde, 2% glutaraldehyde, 0.1% Triton X100 in phosphatebuffered saline (PBS; 137 mM NaCl, 2.7 mM KCl, 12 mM H 2 PO 4 -/HPO 4 2-, pH 7.4).A vacuum was applied to the reaction tubes until all leaf sections sank to the bottom.The sections were incubated in the primary fixation solu tion overnight followed by washing once with PBS solution, pH 7.4, and twice with distilled H 2 O.For postfixation, the sections were incubated in 1% OsO 4 for 45 min followed by washing again three times with distilled H 2 O.A dehydration series was performed ranging from 30% to 100% acetone, followed by incubation in increasing proportions of Araldite resin until 100% Araldite was reached.The sections were finally positioned into flat embedding moulds and polymerized at 65 °C for at 48 h.
After cutting, the leaf sections were stained with toluidine blue so lution (0.5% toluidine blue, 0.5% methylene blue, 6% Na 2 B 4 O 7 , 1% H 3 BO 3 ) and studied under a light microscope (Zeiss Axio Observer, Carl Zeiss, Germany).For quantitative analysis, pictures of at least three BSs per biological replicate were taken and analysed with ImageJ software.The following parameters were determined per BS for quan titative analysis: crosssection area of the BS cell (BS_cell_area), area of organelles orientated towards the vein (V_organelle_area), area of organelles orientated towards the intercellular space (ICS), and M (M_ organelle_area).The following parameter were calculated: percentage of veinorientated organelle area per BS cell area (percent_V_organ elle), percentage of organelle area orientated towards the ICS and M (percent_M_organelles), the total organelle area per cell (V_organ elle_area + M_organelle_area = total_organelle_area), and the ratio between percentage values for vein and ICS/Morientated organelles.Furthermore, the leaf thickness was measured at the site of the selected BS.Three representative cells were analysed per BS, and three BSs were analysed per plant.

PEPC activity
Total soluble proteins were extracted from the homogenized leaf mate rial in 50 mM HEPESKOH pH 7.5, 5 mM MgCl 2 , 2 mM DTT, 1 mM EDTA, 0.5% Triton X100.For the PEPC assay, 10 µl of the extract were mixed with assay buffer consisting of 100 mM TricineKOH pH 8.0, 5 mM MgCl 2 , 2 mM DTT, 1 mM KHCO 3 , 0.2 mM NADH, 5 mM glucose6phosphate, and 2 U ml -1 malate dehydrogenase in a microtitre plate.The reaction was started after addition of PEP to a final concen tration of 5 mM in the assay.Protein content of the solutions was deter mined with the BCA assay (Thermo Fischer Scientific).

Statistical analysis
Data analysis was performed using R (www.Rproject.org).Statistical dif ferences between the measured parameters in the accessions were calcu lated by oneway ANOVA followed by Tukey's posthoc test.Differences between parameters in C 3 and C 3 -C 4 photosynthesis types were deter mined by a twotailed ttest.

Assessment of CO 2 -concentrating efficiency by measuring CO 2 compensation points
The CCP is a measure of the internal leaf CO 2 level at which photosynthetic CO 2 fixation is equal to the CO 2 re lease by photorespiration, day respiration, and other catalytic processes (i.e. the concentration at which net CO 2 assimila tion is zero).In the present study, the CCP was determined from A-Ci curves across a diverse range of 33 Brassicaceae taxa representing 28 species to assess their carbon usage ef ficiency (Fig. 1).
When sorting all sampled taxa in the Brassicaceae according to their CCP, a range of CCP values from 60 ppm to 12 ppm was detected.The Cleomaceae C 4 species G. gynandra with highly efficient CO 2 concentration showed CCP values <10 ppm.Plants with CCP values between 10 ppm and 40 ppm are pre dicted to utilize less efficient CO 2 concentrating mechanisms, such as the photorespiratory shuttle, and were hereby classi fied as C 3 -C 4 intermediates.In contrast, all accessions and spe cies with a CCP >40 ppm were classified as C 3 species.This grouping was supported by ANOVA with posthoc Tukey's HSD test (α=0.05;Fig. 1).The same threshold values were also used in the survey by Krenzer et al. (1975).Among the C 3 -C 4 intermediates, the lowest CCP value of 12 ppm was measured in D. tenuifolia.In contrast, the highest CCP value recorded among the C 3 -C 4 intermediates at ~40 ppm was measured in H. incana HIR3.Importantly, the identification of this taxon as a C 3 -C 4 intermediate species which presumably operates a CO 2 concentrating mechanism is described here for the first time.Interestingly, another taxon assigned to the same species (H.incana HIR1) exhibited a CCP value within the range of C 3 species (Fig. 1).In all other spe cies, different taxa were assigned to the same photosynthetic type.Altogether, a wide range of CCPs was observed among Brassicaceae and especially the C 3 -C 4 intermediates.

Phylogeny suggests multiple origins of C 3 -C 4 photosynthesis in the Brassiceae
The phylogenetic relationship among the plant species selected for this study was investigated using sequence data from 102 orthogroups (Fig. 2).When investigating the distribution of species classified as C 3 -C 4 intermediates based on CCP data, the tree reveals multiple origins of CO 2 concentrating mechanisms in the Brassiceae.We are aware that the presented tree includes only a small subset of species from the Brassiceae group.However, a more densely sampled phylogenetic tree by Koch and Lemmel (2019) suggests the same number of origins because our predicted C 3 -C 4 lineages all have common ances tors with C 3 species.
The phylogenetic positions of B. gravinae, H. incana HIR3, and D. erucoides on the tree suggest independent origins of C 3 -C 4 features.In the Moricandia group, C 3 -C 4 features are observed among the close relatives M. arvensis, M. suffruticosa, M. nitens, M. sinaica, and M. spinosa, but not the sister species M. moricandioides which is C 3 .Thus, C 3 -C 4 most probably evolved once in this group, in a common ancestor preceding the speci ation of M. arvensis, M. suffruticosa, M. nitens, M. sinaica, and M. spinosa but after diversification from M. moricandioides.Finally, C 3 -C 4 like CCPs were observed in D. tenuifolia and D. muralis.Both of these respective species are closely related to the C 3 species D. viminea.Diplotaxis muralis is a natural hybrid between the C 3 parent D. viminea and the C 3 -C 4 parent D. tenuifolia.Therefore, the C 3 -C 4 features in D. muralis are assumed to be inherited from D. tenuifolia (Ueno et al., 2006).In summary, our phylogenetic data indicate that C 3 -C 4 features have independ ently evolved up to five times in the Brassicaceae, in B. gravinae, in D. erucoides, in H. incana HIR3, in the Moricandia group, and in D. tenuifolia.

Physiology of C 3 , C 3 -C 4 , and C 4 leaves under different CO 2 concentrations
Efficiency of photosynthetic gas exchange in the different C 3 and C 3 -C 4 Brassicaceae was assessed under ambient CO 2 (400 ppm) and saturating light (1500 µmol m -2 s -1 ).Here, under these conditions, no association between photosynthesis type and net assimilation could be observed (Fig. 3).For in stance, in C 3 plants, assimilation rates under ambient CO 2 varied between 12.3 µmol m -2 s -1 (A.thaliana) and 28.1 µmol m -2 s -1 (D. tenuisiliqua) (Fig. 3; Supplementary Table S2), whilst among C 3 -C 4 intermediates, assimilation rates varied between 17.3 µmol m -2 s -1 (B.gravinae accession 2) and 26.1 µmol m -2 s -1 (D. erucoides).Moreover, assimilation rates achieved in the C 4 species G. gynandra of 23.7 µmol m -2 s -1 were similar to rates in nonC 4 plants.Thus, assimilatory capacity under ambient CO 2 conditions appears to be species specific, rather than de termined by the activity of a metabolite shuttle mechanism.
In contrast to the above, enhanced rates of assimilation were discovered in plants operating a CO 2 concentrating mech anism under lower atmospheric CO 2 concentrations (Fig. 3; Supplementary Table S2).For instance, at preindustrial levels of 200 ppm CO 2 , the C 4 G. gynandra showed higher assimila tory capacity compared with any other C 3 or C 3 -C 4 species (Fig. 3; Supplementary Table S2).Further, this elevated assim ilation rate observed in G. gynandra became even more pro nounced under 100 ppm CO 2 (Fig. 3; Supplementary Table S2).Interestingly, C 3 -C 4 intermediate species tended to per form better than C 3 plants under subambient CO 2 conditions (Fig. 3; Supplementary Table S2).On average, across all plants of each photosynthesis type, assimilation was significantly higher in the C 3 -C 4 group compared with the C 3 group under CO 2 conditions of ≤200 ppm (ttest, P<0.05; Supplementary Fig. S1).Thus, although CO 2 concentrating mechanisms do not improve net assimilation under present atmospheric CO 2 , they appear to be advantageous under former preindustrial levels of CO 2 .
In addition to the above, operating a CO 2 concentrating mechanism also yielded benefits in terms of improved WUE (ratio between CO 2 assimilation and water stomatal conduct ance).For example, under ambient 400 ppm CO 2 , the WUE was significantly higher in the C 4 species G. gynandra as com pared with all other C 3 and C 3 -C 4 species (Fig. 4).On av erage, WUE did not differ between C 3 and C 3 -C 4 accessions at 400 ppm CO 2 .However, C 3 -C 4 plants were found to exhibit a significantly improved WUE at both 200 ppm and 100 ppm CO 2 compared with C 3 species (ttest, P>0.05; Supplementary Fig. S1), recapitulating the trend observed for the assimila tion rate.In addition, a strong negative correlation between WUE and Ci was found across species at all atmospheric CO 2 concentrations (Supplementary Fig. S2).Given that changes in stomatal conductance exhibited no photosynthesis type specific pattern (Supplementary Fig. S1), this result suggests that higher WUE is achieved across species in the Brassicaceae by CO 2 assimilation at lower internal Ci, Thus, C 3 -C 4 species are able to operate at lower internal CO 2 concentrations than species from the C 3 group.
Next, to determine the effect of C 3 , C 3 -C 4 , and C 4 me tabolism on the light reactions of photosynthesis, chlorophyll fluorescence parameters were also measured.In general, a pos itive correlation was found between net assimilation and ETR, as well as between assimilation rate and effective quantum efficiency F v ʹ/F m ʹ (Supplementary Fig. S2).However, fluo rescence parameters were less affected under reduced atmos pheric CO 2 concentrations as compared with the assimilation rate (Supplementary Fig. S2).
To further assess how different photosynthesis types are characterized by differences in leaf physiological parameters, a principal component analysis (PCA) was performed.Since the importance of carbonconcentrating mechanisms becomes more obvious when CO 2 is limited, gas exchange measure ments at CO 2 concentrations of 200, 100, and 50 ppm were included in this analysis in addition to those measured under ambient 400 ppm CO 2 (Fig. 5A, B).In this PCA, the first prin cipal component was found to explain 65.5% of the variation, and separates the C 4 species G. gynandra from all other C 3 and C 3 -C 4 plants.To a lesser extent, the same component also sepa rates C 3 and C 3 -C 4 plants, though these groups do overlap on this axis (Fig. 5A, B).As expected from the above results, the first principal component is driven by WUE, Ci, Ci/Ca ratio, CCP, and CE.In contrast, the second principal component, driven by stomatal conductance and assimilation at higher CO 2 concentrations, has no effect on separating plants across different photosynthesis types and is driven by speciesspecific variation.
As CCP was used above to classify different photosynthesis types, correlations were investigated between CCP values across species and all other measured leaf physiological param eters.A strong positive correlation was observed between CCP and both Ci and the Ci/Ca ratio at low CO 2 concentrations, respectively (Fig. 5C, D).In contrast, a negative correlation was found between CCP and both WUE and assimilation at low CO 2 (Fig. 5C, D).These relationships were independently  observed irrespective of whether the C 4 species G. gynandra was included in the analysis or not.Conversely, CE (=initial slope of the A-Ci curve) was negatively correlated with CCP only when the C 4 G. gynandra was included in the analysis (Fig. 5C, D).This indicates that CE was not influenced considerably during the transition from C 3 to C 3 -C 4 , but only during tran sition from C 3 -C 4 to C 4 .

Metabolite profiles of leaves with different photosynthesis pathways
To assess the identity of potential transport metabolites used in C 3 -C 4 intermediates, leaf primary metabolites of sampled spe cies were also quantified and analysed by PCA (Fig. 6A, B).In this analysis, the first principal component explained 28.25% of variation and distinguishes C 3 /C 3 -C 4 and C 4 leaf biochem istry.Mainly responsible for this separation are high levels of αalanine, αketoglutarate, aspartate, glycine, glutamate, pyru vate, phenylalanine, and γaminobutyric acid (GABA) in the C 4 G. gynandra compared with the C 3 and C 3 -C 4 background (Fig. 6B; Supplementary Table S3).In contrast, the second principal component sorts the majority of C 3 species (clustered to the top of this axis) from C 3 -C 4 species (clustered to the bottom of this axis) (Fig. 6A, B).Here, on this axis, the C 3 -C 4 plants tend to have higher levels of serine, branched amino acids, and proline, whilst the C 3 species are characterized by higher levels of glucose, sucrose, and myoinositol.Correlation analyses of CCP values with primary metab olite levels were also performed across species (Fig. 6C, D; Supplementary Fig. S3).A negative correlation was observed between CCP and the C 4 related metabolites αalanine, αketoglutarate, aspartate, glutamate, and pyruvate (Figs 6, 7; Supplementary Table S3).However, this relationship seems to be driven by the strong accumulation of these metabolites in the C 4 species alone (Fig. 6C).To identify metabolites that are potentially specific to C 3 -C 4 photosynthesis, the correlation analysis was repeated without the C 4 outgroup species.This resulted in the reduction of the strength of all statistical associa tions.Specifically, only serine and glycine showed significant negative correlations with CCP (Fig. 6D; Supplementary Table S3).Thus, this suggests that serine and glycine have a ubiqui tous role in the glycine shuttle across all C 3 -C 4 intermediates in the Brassiceae.Interestingly, however, glycine was the only metabolite that increased between C 3 and C 3 -C 4 species which  ) Fig. 4. Water use efficiency (WUE) under different CO 2 concentrations.WUE was calculated as the ratio between assimilation (A) and stomatal conductance (g sw ).The gas exchange parameters were measured under conditions of ambient CO 2 (400 ppm) or reduced CO 2 concentrations of 200 ppm and 100 ppm.The tested taxa were sorted according to their CO 2 compensation points and coloured according to the photosynthesis type as C 3 (grey), C 3 -C 4 (blue), and C 4 (red).Taxa names have been abbreviated for legibility and are provided in Fig. 2 and the Materials and methods.
was also high in the C 4 species (Fig. 7A; Supplementary Figs S3 , S4).In contrast, serine was enhanced among C 3 -C 4 species compared with C 3 species, but was detected at a C 3 level in the C 4 G. gynandra (Fig. 7B; Supplementary Figs S3, S4).In the present study, glutamate, αalanine, aspartate, pyruvate, malate, and αketoglutarate formerly predicted to be involved in ni trogen shuttling of C 3 -C 4 leaves (Mallmann et al., 2014) were not associated with CCP (Figs 6D, 7).Instead, levels of these metabolites were high in only some, but not all, C 3 -C 4 taxa.For instance, glutamate and aspartate levels were relatively high in M. arvensis, D. muralis, and D. tenuifolia, but not in the other C 3 -C 4 Moricandia species M. nitens and M. suffruticosa (Fig. 7).
In contrast, D. erucoides separated from the majority of other C 3 -C 4 species in the PCA (Fig. 6A, B), showing relatively high levels of glycerate, glycolate, and malate (Fig. 7; Supplementary Table S3).In summary, the present results describe a general role for only glycine and serine as predicted shuttle metabolites in C 3 -C 4 biochemistry across all species.

Association of CCP with structural features of the bundle sheath
Given that leaf and BS cell architecture play an important role in underpinning CO 2 concentrating mechanisms by enabling adequate metabolite transport between M and BS tissue, we also sought to characterize the leaf anatomy of our Brassicaceae species.In the present study, it was observed that vein density was highest in the C 4 G. gynandra compared with all other spe cies.However, no difference in vein density was observed be tween C 3 and C 3 -C 4 plant accessions (Fig. 8A; Supplementary Fig. S5).To determine whether differences in BS structure were present between photosynthetic types, a representative subset of plant accessions were studied in more detail by light mi croscopy (Supplementary Fig. S6).In this analysis, although BS crosssection area was high in the C 4 species as well as in sev eral C 3 -C 4 species, it was not found to be significantly different between C 3 and C 3 -C 4 plants (Fig. 8B; Supplementary Fig. S7).Moreover, within the BS cells, the areas occupied by plastids and other organelles with either vein (inner half) or ICS/M orientation (outer half) were determined.Areas with ICS/M oriented organelles did not differ between C 3 and C 3 -C 4 leaf crosssections (Fig. 8E).In the C 4 leaf, none of the BS organ elles faced the outer ICS/M side.On the other hand, all plant accessions with a CCP <40 ppm featured enhanced organ elle accumulation around the vein (Fig. 8H).This resulted in higher total organelle area in the C 3 -C 4 BS cells.Thus, organ elle abundance and orientation probably played a decisive role for the functioning of weak CO 2 concentrating mechanisms.C 4 anatomy consists of just one layer of BS and M cells around the veins, which limits the total number of cell layers.In our study, the C 4 leaves of G. gynandra were comparably thin.However, leaf thickness within the C 3 and C 3 -C 4 groups showed speciesspecific variation.For instance, independent of photosynthesis type, all Moricandia species possessed thick suc culent leaves.Values of SLA are usually greater for C 4 than for C 3 leaves (Atkinson et al., 2016), but no pronounced photo synthesis typerelated differences in SLA could be observed in our study (Fig. 8I).
The C 4 pathway allows plants to fix CO 2 with lower nitrogen input.This means that typically, C 4 plants have lower leaf nitrogen concentrations compared with C 3 spe cies (Long, 1999;Craine et al., 2005;Gowik et al., 2011).Interestingly, however, C 4 G. gynandra in our analysis had surprisingly high leaf nitrogen concentrations and low leaf CN ratios (Fig. 8D).This result could be influenced by the slow growth rate of this species in comparison with the ma jority of Brassicaceae species in this study.However, leaf pro tein concentrations in this C 4 G. gynandra were low relative to the background of other species, indicating C 4 specific differences in nitrogen allocation (Fig. 8F).Interestingly, no difference in CN ratios or leaf protein concentrations could be observed between the C 3 and C 3 -C 4 species (Fig. 8D; Supplementary Fig. S7).
Operation of the C 4 pathway required increased activity of PEPC, but allows reduction in concentrations of Rubisco and CBB cycle enzymes (Brautigam et al., 2011;Gowik et al., 2011).In our study, PEPC activity was 8 to 20fold higher in the C 4 G. gynandra leaves as compared with the leaves of the C 3 and C 3 -C 4 species (Fig. 8C; Supplementary Table S4).PEPC activities varied in the individual plant taxa, but were not significantly different between the C 3 and C 3 -C 4 groups (Supplementary Fig. S7).Eruca sativa and M. moricandioides, in particular, showed PEPC activities similar to the C 3 -C 4 taxa M. arvensis, M. suffruticosa, and D. tenuifolia (Fig. 8; Supplementary Fig. S7).These results emphasize the power of our multispecies analysis that allows distinction between species and photosyn thesis typerelated variation.
Summarizing the abovementioned structural and leaf com positionrelated parameters in a PCA, the C 4 G. gynandra can be separated from the rest of the Brassicaceae plants (Fig. 9A,  B).This was mainly driven by high values for δ 13 C, vein den sity, and veinorientated organelles in the BS as well as low values for CCP and ICS/Morientated organelles in the BS (Fig. 9A, B).C 3 and C 3 -C 4 accessions separated along the same line in a combination of PC1 and PC2, but an overlap between the two groups was nevertheless observed.Correlation of the CCP to the selected components supported the importance of organelle accumulation and orientation in the BS for the activity of the C 4 as well as the C 3 -C 4 pathway (Fig. 9C, D; Supplementary Fig. S8).

Physiological and phylogenetic analysis indicate evolution of multiple independent C 3 -C 4 lineages in the Brassiceae
Our survey revealed multiple origins of C 3 -C 4 photosynthesis in the Brassiceae tribe (Figs 1, 2), ranging from very efficient photo respiratory shuttles in D. tenuifolia and the Moricandia genus (M.suffruticosa, M. arvensis, M. sinaica, M. nitens, and M. spinosa), to rela tively weaker carbonconcentrating mechanisms in B. gravinae, D. erucoides, and H. incana HIR3.The carbonconcentrating mech anism in D. muralis is assumed to be inherited from the C 3 -C 4 parent D. tenuifolia during natural hybridization with the C 3 spe cies D. viminea (EschmannGrupe et al., 2003;Ueno et al., 2006).
Interestingly, from the two taxa assigned as H. incana, only one (HIR3) showed C 3 -C 4 like features such as a CCP below 40 ppm and the typical organelle arrangement in the BS cells.Comparison of chloroplast sequences from both accessions re vealed that only HIR1 clustered together with other accessions of this species while HIR3 sequences clustered closer to Sinapis pubescencs and Brassica procumbens (Guerreiro et al., 2023).This suggests that HIR3 and H. incana belong to different species, and the former represents a new C 3 -C 4 lineage in the Brassiceae.

Brassicaceae display large variation in efficiency of the carbon conservation mechanism but no C 4 -like shuttles
Our survey of CO 2 concentrating mechanisms in the Brassicaceae confirmed that measurements of the CCP rep resent a valuable tool for the identification of C 3 -C 4 inter mediate plant accessions.In agreement with the large CCP screening study by Krenzer et al. (1975) and our own statistical analysis, taxa with CCPs between 10 ppm and 40 ppm were classified as C 3 -C 4 intermediates.In this group, we observed gradual changes in the CO 2 concentrating capacity.Our study therefore supports models claiming that after establishment of the basic photorespiratory shuttle, multiple metabolic and anatomic adjustments can contribute to the efficiency of the pathway, resulting in additive small fitness gains (Heckmann et al., 2013).
The lowest CCPs in the present investigation were measured in D. tenuifolia and the C 3 -C 4 Moricandia species.Although var ious accessions of these species were used in different studies (Hylton et al., 1988;Rawsthorne et al., 1988a;Apel et al., 1997;Ueno et al., 2003Ueno et al., , 2006;;Schlüter et al., 2017), low CCPs seem to be a ubiquitous trait of these respective species.Moreover, low CCPs in these species were supported by BSspecific local ization of the GLDP protein (Hylton et al., 1988;Rawsthorne et al., 1988a;Ueno et al., 2003).Further, especially in D. tenuifolia, CCP values were observed as very low and close to those  7. Metabolites in mature leaves of selected Brassicaceae.Relative amounts of glycine (A), serine (B), glycerate (C), malate (D), aspartate (E), glutamate (F), α-ketoglutarate (G),and pyruvate and α-alanine in selected plant accessions.The tested taxa were sorted according to their CO 2 compensation points and coloured according to photosynthesis type as C 3 (grey), C 3 -C 4 (blue), and C 4 (red).Taxa names have been abbreviated for legibility and are provided in Fig. 2 and the Materials and methods.typical of C 4 species.However, the strict separation of the C 4 G. gynandra in all PCAs and especially the PEPC and 13 C mea surements support previous observations claiming absence of C 4 like shuttles in the Brassicaceae (Holaday and Chollet, 1984;Hunt et al., 1987;Sage et al., 2011).

Reduction in CCP correlates negatively with organelle accumulation and arrangement in the BS
Despite the differences in efficiency of the photorespiratory shuttle, changes in organelle arrangement were observed in all taxa classified as C 3 -C 4 .For instance, all C 3 -C 4 taxa possessed an enhanced BS area occupied by organelles in the centripetal position and a higher total organelle area per BS cell compared with C 3 species (Figs 8,9).This underlines the importance of anatomical features for carbonrecapturing mechanisms.A strong correlation between reduction in CCP and increased organelle accumulation facing the vein in the BS was also pre viously observed in interspecific hybrids between D. tenuifolia (C 3 -C 4 ) and R. sativus (C 3 ) (Ueno et al., 2003).The BS structural features appeared to be genetically encoded and are inherited independently from the GLDP localization (Ueno et al., 2003).Residual expression of GLDP was also observed in M cells of C 3 -C 4 intermediate Flaveria species (Schulze et al., 2013).This shows that structural modifications can underpin an effective CCP without complete suppression of GLDP in the M cells.
C 3 -C 4 intermediates in our study contained several layers of M cells such that many do not directly border BS cells.So complete absence of GDC activity in the M cells would require transport of photorespiratory glycine through other M cell layers prior to entering the BS for metabolization.However, accumulation of mitochondria in the BS might create a gly cine sink supporting glycine diffusion to the BS, and partial reduction of M GLDP expression would enforce the shuttle.TEM studies of centripetally localized organelles from C 3 -C 4 Brassiceae (Ueno et al., 2006;Schlüter et al., 2017), Asteraceae (McKown and Dengler, 2007), Boraginaceae (Muhaidat et al., 2011), Scrophulariaceae (Khoshravesh et al., 2012), Arthropogoninae (Khoshravesh et al., 2016), and Chenopodiaceae (Yorimitsu et al., 2019) have shown a close arrangement of mitochondria and chloroplasts.Thus, BS ultrastructure seems to play a major role in prevention of photorespiratory CO 2 and NH 3 loss and in im provement of leaf carbon and possibly also nitrogen economy.
In contrast to the C 4 species in our study, BS cells in C 3 -C 4 Brassicaceae exhibited organelles facing the ICS and M cells (Fig. 8; Supplementary Fig. S6).This amount of ICS/M cellfacing organ elles decreased in C 3 -C 4 species with higher carbonconcentrating Fig. 8. Leaf structure-and composition-related parameters and PEPC activity.Mature leaves were used for determination of vein density (A), bundle sheath cell area in micrographs (B), PEPC activity (C), carbon to nitrogen ratio (D), area of bundle sheath organelles with vein orientation in micrographs (E), protein content (F), 13 C signature (G), area of bundle sheath organelles with orientation to intercellular space or mesophyll in micrographs (H), and specific leaf area (I).The tested taxa were sorted according to their CO 2 compensation point and coloured according to photosynthesis type as C 3 (grey), C 3 -C 4 (blue), and C 4 (red).Taxa names have been abbreviated for legibility and are provided in Fig. 2 and the Materials and methods.efficiency.Our results suggest that accumulation of centripetal organelles and reduction of peripheral organelles are not neces sarily regulated by the same process.Additional structural features of C 4 species such as enlarged BS cell area and higher vein density did not differ between the tested C 3 and C 3 -C 4 Brassicaceae taxa.Further, leaf thickness, SLA, and FW/DW ratios were also not different between the leaves of the C 3 , C 3 -C 4 , and C 4 taxa (Fig. 8; Supplementary Fig. S8).Thus, despite leaf anatomy and BS ar chitecture being important requirements for evolution of carbon concentrating mechanisms (Christin et al., 2013), modifications in leaf succulence parameters do not appear to be essential for an ef ficient photorespiratory carbonconcentrating pathway.Plasticity in some morphological parameters could also play a role in fur ther evolution towards the C 4 leaf, and it could be speculated that limited genetic potential for the adjustment of vein density and mesophyll structure could be connected to the absence of C 4 like shuttles in the Brassicaceae.

Brassicaceae C 3 -C 4 metabolism had only a minor influence on leaf steady-state metabolite patterns
Beside organelle arrangement in the BS, the shift of GDC ac tivity to this tissue influences leaf biochemistry (Rawsthorne, 1992;Schlüter et al., 2017).Relocation of the GLDP protein to the BS has been observed in all investigated C 3 -C 4 classi fied species to date (Schlüter and Weber, 2016) and is therefore seen as the decisive step for the evolution of a photorespiratory carbon concentration shuttle.In the Brassicaceae, BS specificity of GLDP was shown for different Moricandia species (Rawsthorne et al., 1988a), Diplotaxis tenuifolia (Ueno et al., 2003), and Brassica gravinae (Ueno, 2011).Lack or reduction of GDC activity in the M causes accumulation of photorespiratory glycine and its transport along the concentration gradient to the BS.The GDC reaction converts two molecules of glycine into one molecule of serine, but also utilizes NADH and releases NH 3 alongside CO 2 .This imbalance requires further metabolic readjustment between the two cell types.Nevertheless, beyond GLDP locali zation, not much is known about the cellspecific metabolism or the nature of additional metabolite shuttles in C 3 -C 4 Brassicaceae.
If metabolite exchange between M and BS cells is realized by a concentration gradient, high concentrations of these trans ported metabolites would be expected in the leaves (Leegood and von Caemmerer, 1989).However, it should be noted that high metabolic flux and cellspecific metabolite accumulation might mask these gradients in total leaf extracts.Our metab olite analysis did not identify preferential metabolite shuttles operating across all C 3 -C 4 Brassicaceae species.Steadystate gly cine concentrations were generally enhanced in the C 3 -C 4 species compared with C 3 species, supporting the hypothesis that glycine is transported from the M cells to the BS for decar boxylation.High glycine was, however, also found in leaves of the C 3 D. tenuisiliqua and the C 4 species G. gynandra, indicating that glycine accumulation is not a distinct C 3 -C 4 feature (Fig. 7A).Further uncertainty exists around the metabolites trans ported back from the BS to the M for rebalancing of carbon, nitrogen, and energy metabolism (Borghi et al., 2022).Beside glycine, serine accumulation also exhibited a negative correla tion with CCP values (Fig. 6D).This strongly supports the in volvement of serine as a metabolite transported back from the BS to the M cells (Rawsthorne, 1992;Mallmann et al., 2014), although variation in serine levels suggests that the contribu tion of serine transport could vary between the different taxa.
High variation between the individual taxa also existed for other shuttle metabolite candidates.Modelling approaches have previously predicted the involvement of glutamate, αketoglutarate, alanine, pyruvate, aspartate, and malate in shut tling processes for rebalancing of nitrogen metabolism between the M and BS (Mallmann et al., 2014).Malate and aspartate could also be involved in rebalancing of reducing power be tween the two cell types (Johnson et al., 2021).Contributions of glutamine/glutamate and asparagine/aspartate to intercel lular shuttles were suggested for the C 3 -C 4 species Flaveria anomala (Borghi et al., 2022).Here, enhanced levels of these various metabolites could be observed in some, but not all, C 3 -C 4 taxa (Fig. 7).For example, high concentrations of malate, aspartate, and glutamate were found in species displaying very low CCPs such as M. arvensis and D. tenuifolia.Interestingly, the C 3 -C 4 Moricandia species which supposedly share a single C 3 -C 4 evolutionary origin also showed strong variation in the metabolite pattern.A similar absence of main shuttle metabo lites has also been described for C 3 -C 4 Flaveria species (Borghi et al., 2022).Our data generally support the hypothesis that multiple metabolites are transported between the M and BS (Schlüter et al., 2017;Borghi et al., 2022).The contribution of the different metabolites could differ in the individual taxa depending on genetic as well as environmental influences.Such a multitude of solutions indicates that metabolite and energy balancing does not represent a limiting step during ev olution of carbonconcentrating pathways.
To date, enzyme localization studies have mostly focused on the GLDP protein, and much less is known about whether other reactions are shifted to the BS in C 3 -C 4 species.In M. arvensis, other tested photorespiratory enzymes such as glyco late oxidase, serine hydroxymethyl transferase, and other sub units of the GDC complex were present in both cell types (Morgan et al., 1993).Enzyme activities in M. arvensis in M and BSenriched fractions were also equally distributed for glyoxylate aminotransferases, glycolate oxidase, and hydroxypy ruvate reductase (Rawsthorne et al., 1988b), supporting the crucial role of GLDP for uneven distribution for glycine shuttle operation in M. arvensis.On the other hand, all GDC subunits were preferentially expressed in the BS in C 3 -C 4 Flaveria and Panicum species (Morgan et al., 1993).Shifting of additional photorespiratory steps could considerably influence the metab olite shuttles.In our study, some species, especially D. erucoides and D. tenuifolia, showed high levels of glycolate and glycerate.Interestingly, intercellular transport of glycerate and glycolate was predicted in a constraintbased modelling approach for weak carbonconcentrating mechanisms on the evolutionary path to C 4 photosynthesis (Blätke and Bräutigam, 2019).Exchange of these metabolites between M and BS would re duce the need for intercellular nitrogen recycling (Borghi et al., 2022).Part of the photorespiratory metabolites could also feed into additional pathways in the BS.It has been estimated that 1-5% of the photorespiratory glycine and ~30% of serine can be metabolized outside the photorespiratory cycle in processes such as protein biosynthesis (Busch et al., 2018).The high or ganelle accumulation would increase the demand for protein synthesis in the C 3 -C 4 BS.Furthermore, the BS is also respon sible for loading of assimilation products into the phloem, and part of the carbon and nitrogen transported into the BS by the glycine shuttle could support metabolite export to the sink tissue of the plants.

C 3 -C 4 photosynthesis is associated with reduced Ci and enhanced WUE especially under limiting CO 2
In the Brassicaceae, the presence of C 3 -C 4 metabolism did not translate into improved photosynthetic assimilation under am bient environmental conditions (Figs 3,4).For instance, across the Brassicaceae species analysed in the current study, assimila tion rates appeared to be genotype specific rather than related to photosynthesis type under ambient CO 2 .This lack of cor relation between assimilation and photosynthesis type has also been previously described in the Chenopodiaceae (Yorimitsu et al., 2019).
Interestingly, however, C 3 -C 4 taxa in the Brassicaceae adjusted leaf Ci to lower levels compared with C 3 taxa in this clade.The difference between these photosynthesis types was mar ginal under ambient conditions, but became more pronounced under CO 2 conditions of ≤200 ppm (Fig. 3).The ability to assimilate CO 2 at lower Ci translated into higher WUE in the C 3 -C 4 taxa compared with the C 3 species.This increase in WUE observed was underpinned by enhanced assimilation, as stomatal conductance was similar among the C 3 and C 3 -C 4 taxa under all tested conditions (Fig. 4).It should be noted, however, that the differences observed for Ci and WUE be tween C 3 and C 3 -C 4 taxa were small in comparison with the difference between all C 3 and C 3 -C 4 taxa and C 4 G. gynandra, thus underlining the superiority of the C 4 pathway as a CO 2 concentrating mechanism compared with C 3 -C 4 me tabolism.Similar observations have been previously made in Heliotropium and Flaveria, in which C 3 -C 4 species achieved WUE values between those of the C 3 and C 4 species.This was also due to higher assimilation rather than modified conduct ance (Huxman and Monson, 2003;Vogan et al., 2007).These results support an advantage of the C 3 -C 4 pathway in high photorespiratory conditions which cause CO 2 restriction due to stomatal closure.
Evolution of the glycine shuttle often appears to be connected to an enlargement of the growth habitat (Lundgren and Christin, 2017).The C 3 species M. moricandioides for instance seems to be geographically restricted to the Iberian Peninsula, while the closely related C 3 -C 4 species M. arvensis has spread into north west Africa, Southern Europe, and other parts of the planet where it is mostly associated with cultivated areas and disturbed sites (Perfectti et al., 2017).Diplotaxis tenuifolia also often grows as an invasive species occupying sunny, harsh, and arid areas (Nicoletti et al., 2007) in which water, nutrient, and temperature conditions can change rapidly.It is therefore possible that C 3 -C 4 species profit from high environmental plasticity of the trait.
Ecological studies which have investigated the adaptation of C 3 -C 4 species to specific environmental conditions are unfor tunately still rare (Oono et al., 2022).In contrast to C 3 and C 4 species, the C 3 -C 4 compensation points are strongly influenced by environmental conditions, especially light, temperature, and nitrogen (Brown and Morgan, 1980;Holaday and Chollet, 1983;Hunt et al., 1987;Schuster and Monson, 1990;Gomez et al., 2020;Oono et al., 2022).In Chenopodium album, the CCP was lowest under high temperature and low nitrogen condi tions, which was connected to accumulation of the GLDP protein preferentially in the BS (Oono et al., 2022).Moricandia arvensis leaves also had lower CCPs and higher WUE under hotter and more arid summer conditions than in milder spring climates (Gomez et al., 2020).Plasticity of photosynthetic traits under stress conditions was recently also reported for the C 2 species Sasola divaricate (Tefarikis et al., 2022).Our results in dicate that gradual and even facultative implementation of carbon shuttles between the M and BS are possible and should be considered in future experiments.
Knowledge about the distribution of species with glycine shuttle metabolism is generally still limited to studies among relatives of C 4 species.This is mainly due to the dependence on gas exchange equipment and timeconsuming measurements.It is therefore assumed that the frequency of species with weaker carbonconcentrating mechanisms is greatly underestimated (Sage et al., 2011;Lundgren, 2020).Identification of C 3 -C 4 features in a H. incana HIR3 and recently also in some C. album accessions (Yorimitsu et al., 2019) supports this hypothesis.As such, faster methods for identification of C 3-C 4 intermediates could help to close this knowledge gap.Here, our correlation analysis showed that measurements of assimilation at low CO 2 are sufficient for detection of C 3 -C 4 phenotype and would save considerable time as opposed to having to calculate CCP by measuring assimilation across a range of CO 2 concentra tions (Fig. 5D).For example, a very strong positive correlation in the present results was found to exist between CCP and assimilation rate at 50 ppm CO 2 , which is close to the CCP of C 3 species.High and significant negative correlation to CCP also existed for WUE under CO 2 conditions of ≤200 ppm.As in our experiments assimilation generally correlated positively with photosynthesis efficiency F v ʹ/F m ʹ, fluorescence combined with stomatal conductance measurements could possibly also be used in a fast initial screening experiments for identification of C 3 -C 4 intermediates in the future.

Conclusions
Our survey revealed that photorespiratory shuttles evolved up to five times in the Brassiceae tribe in different genetic back grounds.Measurements of the CCP indicated considerable variation in the pathway in the different tested taxa.Reduction in CCP was generally associated with organelle arrangement in the BS.Thus, elucidation of regulatory mechanisms under lying organelle multiplication and arrangement in the BS ap pear to be crucial for engineering an efficient glycine shuttle pathway into the leaf.
Although CCPs as low as 12 ppm were observed in D. tenuifolia, there was no evidence for the operation of C 4 like shuttles in the tested taxa, supporting its classification as a dis tinct pathway.All C 3 -C 4 classified taxa belong to the Brassiceae tribe which appears to have lost one GLDP gene copy, suggest ing that this event facilitated evolution of the glycine shuttle (Schlüter et al., 2017).Additional lossoffunction mutations or insertion of a transposable element are thought to be in volved in loss or reduction of GDC activity in the M cells (Rawsthorne, 1992;Sage et al., 2012;Adwy et al., 2015;Triesch et al., 2023, Preprint).In D. muralis, transfer of weak carbon concentrating mechanisms seem to have been inherited during hybridization from a C 3 -C 4 parent (Ueno et al., 2003).The contribution of hybridization to distribution of carboncon centrating pathways has been discussed for several plant groups including Sasola and Flaveria (Kadereit et al., 2017;Tefarikis et al., 2022;MoralesBriones and Kadereit, 2023, Preprint).In some grasses, lateral gene transfer has been shown to sup port the rapid and successful establishment of the C 4 pathway (Dunning et al., 2019b).Such scenarios would nevertheless re quire donor species that are able to successfully transfer essen tial features into the receiving genetic background.
Our results reveal that photorespiratory carbonconcen trating mechanisms in the Brassiceae show large variation in their biochemical and physiological features.C 3 -C 4 Brassiceae species are often associated with fast changing temperature, water, and nutrient conditions.Metabolic plasticity could also be advantageous in crop species challenged by dynamic cli matic variability.Brassica napus or B. oleaceae are closely re lated to the described C 3 -C 4 species and would be prime targets for transfer of this trait.Recent progress in sequenc ing the genomes of these species and related species in the Brassicaceae (Guerreiro et al., 2023) can help to identify the molecular mechanisms behind BSspecific C 3 -C 4 architecture and biochemistry.

Fig. 1 .
Fig.1.CO 2 compensation points in selected Brassicaceae.CO 2 compensation points were measured in young, fully expanded leaves of greenhousegrown plants.The letters above each box indicate the statistical grouping determined by ANOVA followed by HSD post-hoc test with α=0.05.The tested taxa are coloured according to photosynthesis type as C 3 (grey), C 3 -C 4 (blue), and C 4 (red).Plant names have been abbreviated for legibility and are provided in full in the Materials and methods.

Fig. 2 .
Fig. 2. Phylogeny and photosynthesis types.The numbers on the nodes represents quartet-based local posterior probability values.Species names are coloured according to the photosynthesis type as C 3 (grey), C 3 -C 4 (blue), and C 4 (red).

Fig. 3 .
Fig.3.Net assimilations under different CO 2 concentrations.Assimilation was measured under conditions of ambient CO 2 (400 ppm) or reduced CO 2 concentrations of 200 ppm and 100 ppm.The tested taxa were sorted according to their CO 2 compensation points and coloured according to the photosynthesis type as C 3 (grey), C 3 -C 4 (blue), and C 4 (red).Taxa names have been abbreviated for legibility and are provided in Fig.2and the Materials and methods.

Fig. 5 .
Fig. 5. Principal component analysis (PCA) and correlations of the CO 2 compensation point (CCP) with selected gas exchange parameters.Average values for the selected photosynthetic parameters determined under 50-400 ppm of CO 2 were used for the analysis.(A) Localization of the plant lines in the PCA, (B) PCA including parameter loadings, (C) Pearson correlation coefficients demonstrated as heatmaps using all taxa, and (D) Pearson correlation coefficients demonstrated as heatmaps using only C 3 and C 3 -C 4 lines.The tested taxa were coloured according to photosynthesis types as C 3 (grey), C 3 -C 4 (blue), and C 4 (red).Taxa names have been abbreviated for legibility and are provided in Fig. 2 and the Materials and methods.The dataset included CO 2 compensation point (CCP), carboxylation efficiency (CE), assimilation (A), internal CO 2 concentration (Ci), stomatal conductance (gsw), water use efficiency (WUE), ratio of internal to external CO 2 concentrations (CiCa), electron transport rate (ETR), and quantum efficiency F v ʹ/F m ʹ (FvFm).The numbers after the parameter abbreviation indicate the CO 2 concentration in the outside the leaf in the measuring cuvette.

Fig. 6 .
Fig. 6.Principal component analysis (PCA) and correlations of the CO 2 compensation point (CCP) with specific metabolites.Average values for the selected metabolites per taxon were used for the analysis.(A) Localization of the plant lines in the PCA, (B) PCA including metabolite loadings, (C) Pearson correlation coefficients demonstrated as heatmaps using all plant lines, and (D) Pearson correlation coefficients demonstrated as heatmaps using only C 3 and C 3 -C 4 lines.The tested taxa were coloured according to photosynthesis type as C 3 (grey), C 3 -C 4 (blue), and C 4 (red).Taxa names have been abbreviated for legibility and are provided in Fig. 2 and the Materials and methods.

Fig. 9 .
Fig. 9. Principal component analysis and of the CO 2 compensation point (CCP) with leaf structural and compositional components.Average values for the selected parameters measured by EA-IRMS analysis of leaf cross-sections by light microscopy.(A) Localization of the taxa in the PCA, (B) PCA including parameter loadings, (C) Pearson correlation coefficients demonstrated as heatmaps using all plant lines, (D) Pearson correlation coefficients demonstrated as heatmaps using only C 3 and C 3 -C 4 lines.The tested taxa were coloured according to photosynthesis type as C 3 (grey), C 3 -C 4 (blue), and C 4 (red).Taxa names have been abbreviated for legibility and are provided in Fig. 2 and the Materials and methods.