Intramolecular carbon isotope signals reflect metabolite allocation in plants.

Stable isotopes at natural abundance are key tools to study physiological processes occurring outside the temporal scope of manipulation and monitoring experiments. Whole-molecule carbon isotope ratios (13C/12C) enable assessments of plant carbon uptake yet conceal information about carbon allocation. Here, we identify an intramolecular 13C/12C signal at tree-ring glucose C-5 and C-6 and develop experimentally testable theories on its origin. More specifically, we assess the potential of processes within C3 metabolism for signal introduction based (inter alia) on constraints on signal propagation posed by metabolic networks. We propose that the intramolecular signal reports carbon allocation into major metabolic pathways in actively photosynthesizing leaf cells including the anaplerotic, shikimate, and non-mevalonate pathway. We support our theoretical framework by linking it to previously reported whole-molecule 13C/12C increases in cellulose of ozone-treated Betula pendula and a highly significant relationship between the intramolecular signal and tropospheric ozone concentration. Our theory postulates a pronounced preference for leaf cytosolic triose-phosphate isomerase to catalyse the forward reaction in vivo (dihydroxyacetone phosphate to glyceraldehyde 3-phosphate). In conclusion, intramolecular 13C/12C analysis resolves information about carbon uptake and allocation enabling more comprehensive assessments of carbon metabolism than whole-molecule 13C/12C analysis.

Introduction

Plant carbon metabolism is a central component of the global carbon cycle. It both depends on and affects environmental properties. Improved understanding of long-term plant–environment interactions relies on information from plant archives (such as tree rings) because manipulation and monitoring experiments can only cover short to medium time scales. Stable carbon isotope (13C/12C) analysis is among the most advanced tools to extract physiological and environmental information from plant archives. Conventionally, average 13C/12C ratios of whole-plant metabolites are analysed. However, this approach neglects 13C/12C differences known to occur among individual carbon positions of plant metabolites (Abelson and Hoering, 1961). In contrast, we recently analysed intramolecular 13C/12C ratios in glucose extracted across an annually resolved Pinus nigra tree-ring time series (1961–1995) and reported intramolecular 13C signals (i.e. systematic 13C/12C variation confined to individual glucose carbon positions; Wieloch ). Only after their ecophysiological origins have been elucidated can these archived signals become useful for applications within the plant and Earth sciences.

Based on our previous dataset (Wieloch ), we have already pinpointed a 13C signal at tree-ring glucose C-4 and proposed that it informs about carbon flux around leaf cytosolic glyceraldehyde-3-phosphate dehydrogenases and associated energy metabolism (Wieloch, 2021; Wieloch ). Here, we utilize the same dataset to isolate a 13C signal at tree-ring glucose C-5 and C-6. Since intramolecular 13C variation is governed (inter alia) by enzyme isotope effects and metabolite partitioning (Hayes, 2001), we hypothesize that the signal can be linked to shifts in carbon allocation and underlying environmental controls. Thus, we develop experimentally testable theories on ecophysiological mechanisms that can introduce the signal at glucose C-5 and C-6. To this end, we consider all enzyme reactions within central carbon metabolism of C3 plants. This includes the Calvin–Benson cycle (CBC), the photosynthetic carbon oxidation (PCO) cycle, starch and sucrose synthesis and degradation, cellulose synthesis, the pentose phosphate pathway, glycolysis, and carbon metabolism downstream of phosphoenolpyruvate (PEP). Carbon exchange between other biochemical pathways and the pathway leading to the formation of tree-ring glucose are presumably small, particularly when integrated over the course of growing seasons, the time frame of tree-ring formation. Thus, these processes cannot introduce 13C signals of substantial size into tree-ring archives. Furthermore, we only consider primary isotope effects (which occur at atoms with altered binding after chemical reactions). Sizes of secondary isotope effects (which occur at atoms with unaltered binding after chemical reactions due to indirect involvement in reaction mechanisms) are usually small and therefore unlikely to introduce detectable 13C signals into tree-ring archives. Finally, we present evidence supporting our theory. For this part, we reanalyse our own tree-ring dataset (Wieloch ) in combination with publicly accessible climate data and 13C/12C data from an ozone treatment experiment published by Saurer .

We distinguish two major types of 13C fractionation: diffusion–Rubisco fractionation, and post-Rubisco fractionation (Wieloch ). Diffusion–Rubisco fractionation accompanies CO2 diffusion from ambient air into plant chloroplasts and subsequent carbon fixation by Rubisco (Figs 1, 2; Farquhar ). It affects all carbon positions of plant glucose equally (Wieloch ). In contrast, post-Rubisco fractionation results from metabolic processes downstream of Rubisco and is position specific (Figs 1–3). Deconvolution of the two fractionation types requires the intramolecular approach.

Fig. 1.

Central carbon metabolism in trees. Solid and dashed arrows represent substantial and negligible metabolite fluxes, respectively. Red arrows: reactions introducing the Δ5–6ʹ signal. CBC, Calvin–Benson cycle; PCO cycle, photosynthetic carbon oxidation cycle (photorespiration); MEP, non-mevalonate pathway. Metabolites: 2PGA, 2-phosphoglycerate; 3PGA, 3-phosphoglycerate; DAHP, 3-deoxy-d-arabino-heptulosonic acid 7-phosphate; DHAP, dihydroxyacetone phosphate; E4P, erythrose 4-phosphate; FBP, fructose 1,6-bisphosphate; GAP, glyceraldehyde 3-phosphate; OAA, oxaloacetate; PEP, phosphoenolpyruvate; PYR, pyruvate. Enzymes: (1) Rubisco; (2) TPI, triose-phosphate isomerase; (3) aldolase, fructose-bisphosphate aldolase; (4) np-GAPDH, irreversible non-phosphorylating glyceraldehyde-3-phosphate dehydrogenase; (5) p-GAPDH, reversible phosphorylating glyceraldehyde-3-phosphate dehydrogenase; (6) PGK, phosphoglycerate kinase; (7) PGM, phosphoglycerate mutase; (8) enolase, (9) PK, pyruvate kinase; (10) PPDK, pyruvate orthophosphate dikinase; (11) PEPC, phosphoenolpyruvate carboxylase; (12) DAHPS, 3-deoxy-d-arabino-heptulosonate 7-phosphate synthase. Figures 2 and 3 show the reactions in more detail. In contrast to its representation here, parts of the PCO cycle reside outside chloroplasts, in peroxisomes, and mitochondria. Localization of parts of the shikimate pathway in the cytosol is being debated (Maeda and Dudareva, 2012). To avoid clutter, not all metabolic intermediates are shown. For instance, conversion of 3PGA to GAP proceeds via 1,3-bisphosphoglycerate.

Fig. 2.

Formation and cleavage of bonds involving carbon atoms in central carbon metabolism. Carbon numbering is according to carbon positions in tree-ring glucose. Metabolites: 1,3BPGA, 1,3-bisphosphoglycerate; 3PGA, 3-phosphoglycerate; DHAP, dihydroxyacetone phosphate; FBP, fructose 1,6-bisphosphate; GAP, glyceraldehyde 3-phosphate; Pi, inorganic phosphate; RuBP, ribulose 1,5-bisphosphate. Numbers in parentheses denote enzymes: (1) Rubisco; (2) TPI, triose-phosphate isomerase; (3) aldolase, fructose-bisphosphate aldolase; (4) np-GAPDH, irreversible non-phosphorylating glyceraldehyde-3-phosphate dehydrogenase; (5) p-GAPDH, reversible phosphorylating glyceraldehyde-3-phosphate dehydrogenase; (6) PGK, phosphoglycerate kinase.

Fig. 3.

Formation and cleavage of bonds involving carbon atoms in central carbon metabolism. Solid and dashed arrows represent substantial and negligible metabolite fluxes, respectively. Red: carbon bond modifications possibly accompanied by primary isotope effects. Carbon numbering is according to carbon positions in tree-ring glucose. Metabolites: 2PGA, 2-phosphoglycerate; 3PGA, 3-phosphoglycerate; DAHP, 3-deoxy-d-arabino-heptulosonic acid 7-phosphate; E4P, erythrose 4-phosphate; HCO3–, bicarbonate; OAA, oxaloacetate; PEP, phosphoenolpyruvate; Pi, inorganic phosphate; PPi, pyrophosphate; PYR, pyruvate. Numbers in parentheses denote enzymes: (7) PGM, phosphoglycerate mutase; (8) enolase, (9) PK, pyruvate kinase; (10) PPDK, pyruvate orthophosphate dikinase; (11) PEPC, phosphoenolpyruvate carboxylase; (12) DAHPS, 3-deoxy-d-arabino-heptulosonate 7-phosphate synthase.

Our work makes several conceptual advances. (i) We show how constraints on signal propagation posed by metabolic networks can be used to narrow down signal origins. (ii) A conceptual model describes how the signal propagates from its origin to other glucose carbon positions and metabolite pools. Due to space restrictions, this model is presented in Supplementary Protocol S1. (iii) We revise current theory on plant isotope fractionation by ozone exposure. (iv) The present paper and a companion paper on the C-4 signal (Wieloch ) develop theories that consider all relevant parts of metabolism and link intramolecular 13C signals with specific shifts in carbon allocation and their environmental causes. For isotope signals generated within complex metabolic networks, such comprehensive theories are required as a starting point for subsequent tailored experimental tests.

Materials and methods

Intramolecular 13C/12C ratios in tree-ring glucose of P. nigra from Vienna (Austria) were reported in Wieloch . They are expressed in terms of intramolecular 13C discrimination, Δʹ, where i denotes individual carbon positions in tree-ring glucose (Wieloch ; abbreviations and symbols are given in Table 1). In this notation, positive values denote discrimination against 13C. The prime denotes measurements subjected to a procedure that removes the 13C redistribution effect by triose phosphate cycling (Supplementary Protocol S2; Wieloch ). This correction restores leaf-level 13C signals. The dataset comprises six annually resolved time series (one per glucose carbon) each covering the period 1961–1995 and containing 31 time points (n=6 × 31=186).

Table 1.

Abbreviations, terminology, and identifiers.

Abbreviation	Metabolite	Identifier
1,3BPGA	1,3-Bisphosphoglycerate	CAS 1981-49-3
2PGA	2-Phosphoglycerate	CAS 2553-59-5
3PGA	3-Phosphoglycerate	CAS 820-11-1
ADP	Adenosine diphosphate	CAS 58-64-0
ATP	Adenosine triphosphate	CAS 56-65-5
DAHP	3-Deoxy-d-arabino-heptulosonic acid 7-phosphate	CAS 2627-73-8
DHAP	Dihydroxyacetone phosphate	CAS 57-04-5
E4P	Erythrose 4-phosphate	CAS 585-18-2
FBP	Fructose 1,6-bisphosphate	CAS 488-69-7
GAP	Glyceraldehyde 3-phosphate	CAS 591-59-3
NAD+, NADH	Nicotinamide adenine dinucleotide	CAS 53-84-9
OAA	Oxaloacetate	CAS 328-42-7
PEP	Phosphoenolpyruvate	CAS 138-08-9
Pi	Inorganic phosphate	CAS 14265-44-2
PYR	Pyruvate	CAS 127-17-3
Abbreviation	Enzyme	Identifier
	Enolase	EC 4.2.1.11
Aldolase	Fructose-bisphosphate aldolase	EC 4.1.2.13
DAHPS	3-Deoxy-d-arabino-heptulosonate 7-phosphate synthase	EC 2.5.1.54
np-GAPDH	Non-phosphorylating glyceraldehyde-3-phosphate dehydrogenase	EC 1.2.1.9
p-GAPDH	Phosphorylating glyceraldehyde-3-phosphate dehydrogenase	EC 1.2.1.12/13
PEPC	Phosphoenolpyruvate carboxylase	EC 4.1.1.31
PEPCK	Phosphoenolpyruvate carboxykinase	EC 4.1.1.49
PGK	Phosphoglycerate kinase	EC 2.7.2.3
PGM	Phosphoglycerate mutase	EC 5.4.2.11
PK	Pyruvate kinase	EC 2.7.1.40
PPDK	Pyruvate orthophosphate dikinase	EC 2.7.9.1
Rubisco	Ribulose-1,5-bisphosphate carboxylase/oxygenase	EC 4.1.1.39
TPI	Triose-phosphate isomerase	EC 5.3.1.1
Abbreviation	Other
CBC	Calvin–Benson cycle
MEP pathway	Non-mevalonate pathway
[O 3 ]	Tropospheric ozone concentration
PCO cycle	Photosynthetic carbon oxidation cycle
rH	Relative humidity
SD	Sunshine duration
TPC	Triose phosphate cycling
VPD	Air vapour pressure deficit
ΔCi	Difference of intercellular CO2 concentrations
δ 13Ca	13C abundance of atmospheric CO2
Δδ13Cp	Difference of whole-molecule 13C abundances of plant matter
Δ	Whole-molecule 13C discrimination
Δ i	Intramolecular 13C discrimination
Δ i ʹ	Intramolecular 13C discrimination corrected for TPC
ΔΔ	Difference of 13C discrimination
ΔΔDR	Difference of 13C discrimination due to diffusion–Rubisco fractionation

Additionally, we reanalysed published differences in intercellular CO2 concentration, ΔCi, and whole-molecule 13C abundance, Δδ13Cp, between ozone-treated and control plants of Betula pendula grown in 1992 (Table 2; 90/40 nl O3 l–1 day/night versus <3 nl O3 l–1; Saurer ). These authors used two different methods to determine ΔCi. Here, we calculated ΔCi averages. Corresponding differences in 13C discrimination by the diffusion–Rubisco interface were estimated as

Table 2.

Reanalysed data originally published by Saurer

	C/LF	O3/LF	C/HF	O3/HF
β±SD	0.006 ± 0.002	0.013 ± 0.007∗∗∗	0.014 ± 0.005	0.017 ± 0.005∗
ΔC i ±SD (ppm)	O 3 /LF–C/LF		O 3 /HF–C/HF
Steady state	10 ± 16∗		34 ± 41∗∗
Diurnal course	5 ± 28		9 ± 35∗
Δ δ 13 C p ±SD (‰)	O 3 /LF-C/LF		O 3 /HF–C/HF
Leaf cellulose	1.1 ± 0.7∗		0.4 ± 0.9
Stem cellulose	1.3 ± 0.6∗∗		1.1 ± 0.6∗∗

Treatments: control group, C (<3 nl O3 l–1); ozone-treated group, O3 (90/40 nl O3 l–1 day/night); low fertilization, LF; high fertilization, HF. β denotes the carboxylation rate of phosphoenolpyruvate carboxylase relative to the total carboxylation rate of phosphoenolpyruvate carboxylase and Rubisco measured in vitro. ΔCi denotes differences in leaf intercellular CO2 concentrations between ozone-treated and control plants measured by two different methods (‘steady state’ and ‘diurnal course’). Δδ13Cp denotes differences in carbon isotope ratios between ozone-treated and control plants in leaf and stem cellulose. Significance levels of differences between ozone-treated and control plants: ∗P≤0.05; ∗∗P≤0.01; ∗∗∗P≤0.001.

where a and b denote discrimination factors of CO2 diffusion (4.4‰) and Rubisco carboxylation (29‰), respectively, and Ca denotes atmospheric CO2 concentration (356 ppm in 1992; Farquhar ). Reported Δδ13Cp values (Table 2) were transformed to ΔΔ values as

where Ra denotes the 13C/12C ratio of atmospheric CO2, and ΔRp denotes the difference in 13C/12C ratios between ozone-treated and control plants. To utilize Δδ13Cp values reported in per mill, we transformed Equation 2 as

where δ13Ca denotes the 13C abundance of atmospheric CO2 (–8‰ in 1992; Leuenberger, 2007).

For the regression analysis and linear modelling, we used 13C/12C data from Wieloch and publicly accessible climate data. Data of sunshine duration (SD) and relative humidity (rH) were acquired from the climate station Hohe Warte in Vienna (Klein Tank ). Tropospheric ozone concentrations, [O], were acquired from Stephansplatz, Laaer Berg, Hermannskogel, Hohe Warte, and Lobau (City of Vienna, Municipal Department 22).

Results and discussion

Tree-ring glucose exhibits a post-Rubisco signal at C-5 and C-6

Figure 4 shows results of a hierarchical cluster analysis which groups Δʹ time series according to co-variability (Wieloch ); that is, Δʹ time series carrying common 13C signals form clusters. Primary separation occurs between the Δ1ʹ to Δ3ʹ cluster and the Δ4ʹ to Δ6ʹ cluster. Average time series pertaining to these clusters are entirely uncorrelated (r=0.08, P>0.65, n=31). Thus, these clusters convey entirely different ecophysiological information. Wieloch justified the use of air vapour pressure deficit (VPD) as a proxy of diffusion–Rubisco fractionation. While the average time series of the Δ1ʹ to Δ3ʹ cluster correlates highly significantly with VPD (r= –0.70, P=0.00001, n=31), the average time series of the Δ4ʹ to Δ6ʹ cluster is not significantly correlated (r= –0.30, P>0.05, n=31). This indicates that the diffusion–Rubisco signal is preserved at glucose C-1 to C-3 but not at C-4 to C-6. Among all Δʹ, Δ5ʹ and Δ6ʹ exhibit the most significant correlation (r=0.61, P≤0.001, n=31). Since the diffusion–Rubisco signal is confined to glucose C-1 to C-3, we argue that C-5 and C-6 exhibit a strong post-Rubisco signal denoted the Δ5–6ʹ signal.

Fig. 4.

Clustering of Δʹ time series due to co-variability. Δʹ denotes time series of intramolecular 13C discrimination corrected for triose phosphate cycling (Wieloch ). Red: time series discussed here. Data were measured on tree-ring glucose of Pinus nigra laid down from 1961 to 1995 at a dry site in the Vienna basin (n=6 × 31). Members of clusters marked by asterisks are correlated at the following significance levels: ∗P≤0.05; ∗∗ P≤0.01; and ∗∗∗P≤0.001. Precursors of tree-ring glucose, dihydroxyacetone phosphate (DHAP), and glyceraldehyde 3-phosphate (GAP) are shown as molecular structures. Modified figure from Wieloch .

Exclusion of metabolic locations as origin of the Δ5–6ʹ signal

Much is known about plant carbon metabolism. Based on this knowledge, we can exclude several metabolic locations as the origin of the Δ5–6ʹ signal as a first step in development of the theory. Note that the Δ5–6ʹ signal is introduced at the level of three-carbon compounds because reactions at other levels do not modify carbon bonds that become glucose C-5 and C-6.

Exclusion of the tree-ring cell as origin of the Δ5–6ʹ signal

13C labelling experiments provide compelling evidence for a complete or near-complete equilibration of triose phosphates in tree-ring cells of Quercus robur (Figs 1, 2; Hill ). Numerous similar reports for other species and tissues suggest that triose phosphates in the cytosol of heterotrophic tissues are generally substantially equilibrated (Brown and Neish, 1954; Edelman ; Neish, 1955, 1958; Shafizadeh and Wolfrom, 1955; Altermatt and Neish, 1956; Seegmiller ; Shibko and Edelman, 1957; McConnell ; Wolfrom ; Keeling ; Hatzfeld and Stitt, 1990; Viola ). Only two 13C labelling studies report no evidence of substantial heterotrophic triose phosphate equilibration (Greathouse, 1953; Kikuta and Erickson, 1969). However, these authors analysed tissues during stages of exceptionally rapid fruit development with either high hexose phosphate flux into cotton-boll cellulose or high triose phosphate flux into avocado lipids (Greathouse, 1953; Brown and Neish, 1954; Kikuta and Erickson, 1969). Taken together, these reports suggest that heterotrophic triose phosphates are generally substantially equilibrated during normal growth.

The raw dataset of intramolecular 13C discrimination, Δ, in tree-ring glucose (Wieloch ) exhibits significant correlations among all pairs of symmetry-related time series (Table 3); that is, significant correlations occur between Δ1 and Δ6, Δ2 and Δ5, and Δ3 and Δ4. These correlations probably result from carbon redistribution by triose phosphate cycling (TPC) which involves triose phosphate equilibration. Wieloch describe this process mathematically and used the model to remove the TPC effect from Δ, yielding a TPC-free dataset, Δʹ. In this latter dataset, significant correlations among pairs of symmetry-related time series are absent (Table 4). This provides strong evidence for the occurrence of substantial triose phosphate equilibration in tree-ring cells of the samples discussed here.

Table 3.

Correlation coefficients and significance levels obtained from cross-correlation analysis on Δi

	Δ 1	Δ 2	Δ 3	Δ 4	Δ 5	Δ 6
Δ 1	1
Δ 2	0.60∗∗∗	1
Δ 3	0.31	0.52∗∗	1
Δ 4	0.00	0.31	0.38∗	1
Δ 5	0.37∗	0.42∗	0.24	0.39∗	1
Δ 6	0.55∗∗	0.48∗∗	0.31	0.11	0.69∗∗∗∗	1

∗P≤0.05; ∗∗P≤0.01; ∗∗∗P≤0.001; ∗∗∗∗P≤0.0001; n=6 × 31. Δi denotes time series of intramolecular 13C discrimination (Wieloch ). Bold numbers refer to pairs of time series at symmetry-related glucose carbon positions. Data were measured on tree-ring glucose of Pinus nigra laid down from 1961 to 1995 at a dry site in the Vienna basin. This table was originally published as table 1 in Wieloch and is provided here for convenience.

Table 4.

Correlation coefficients and significance levels obtained from cross-correlation analysis on Δʹ

	Δ 1ʹ	Δ 2ʹ	Δ 3ʹ	Δ 4ʹ	Δ 5ʹ	Δ 6ʹ
Δ 1 ʹ	1
Δ 2 ʹ	0.54∗∗	1
Δ 3 ʹ	0.31	0.48∗∗	1
Δ 4 ʹ	–0.12	0.10	–0.12	1
Δ 5 ʹ	0.11	–0.07	0.03	0.32	1
Δ 6 ʹ	0.08	0.19	0.21	0.06	0.61∗∗∗	1

∗∗P≤0.01; ∗∗∗P≤0.001; n=6 × 31. Δʹ denotes time series of intramolecular 13C discrimination corrected for triose phosphate cycling (Wieloch ). Bold numbers refer to pairs of time series at symmetry-related glucose carbon positions. Data were measured on tree-ring glucose of Pinus nigra laid down from 1961 to 1995 at a dry site in the Vienna basin. This table was originally published as table 2 in Wieloch and is provided here for convenience.

If a process in tree-ring cells had introduced a signal at carbon positions corresponding to glucose C-5 and C-6, triose phosphate equilibration would have transmitted it to carbon positions corresponding to glucose C-2 and C-1. The signal at C-5 would have had the same size as the signal at C-2, and the signal at C-6 would have had the same size as the signal at C-1. Please note that equally sized signals at symmetry-related glucose carbon positions are not removed by the method removing TPC effects (Wieloch ). Since the Δ5–6ʹ signal is absent in Δ1ʹ and Δ2ʹ (Fig. 4), it must have been introduced at the leaf level.

Exclusion of the CBC and PCO cycle as origin of the Δ5–6ʹ signal

Introduction of the Δ5–6ʹ signal within the CBC or PCO cycle can be excluded because hexose phosphate synthesis includes conversion of photosynthetic/photorespiratory glyceraldehyde 3-phosphate (GAP; Δ4ʹ to Δ6ʹ) to dihydroxyacetone phosphate (DHAP; Δ3ʹ to Δ1ʹ) by triose-phosphate isomerase (TPI; Figs 1, 2). This would transmit any 13C signal present at GAP carbon positions corresponding to glucose C-5 and C-6 to DHAP carbon positions corresponding to glucose C-2 and C-1. More generally, metabolites feeding into the stromal GAP pool can be excluded as the origin of the Δ5–6ʹ signal based on the same reasoning.

Exclusion of reactions downstream of OAA, pyruvate, and DAHP as Δ5–6ʹ signal origin

Pyruvate kinase (PK) and pyruvate orthophosphate dikinase (PPDK) interconvert PEP and pyruvate (Figs 1, 3). The PK reaction is strongly on the side of pyruvate and considered nearly irreversible (Nageswara Rao ; Tcherkez ). In illuminated leaves of C3 plants, PPDK activity is either very low or undetectable, except for orchids and grasses (Hocking and Anderson, 1986). In illuminated leaves of Xanthium strumarium, flux from pyruvate to PEP is very small at ~0.05% of net CO2 assimilation (Tcherkez ). In Arabidopsis thaliana, PPDK is up-regulated during leaf senescence which is believed to facilitate nitrogen remobilization (Taylor ). This is of minor importance here because leaf senescence occurs during a short period relative to the multiyear life span of conifer needles. In Nicotiana tabacum, PPDK activity is increased up to 2.7-fold under strong drought (Doubnerová Hýsková ). However, this should not result in significant flux in relation to fluxes in carbohydrate metabolism since basal PPDK activities in C3 plants are generally low (Hocking and Anderson, 1986; Tcherkez ). Thus, flux from pyruvate to PEP should be small, and transmission of 13C signals in pyruvate to cytosolic carbohydrates by the PK/PPDK interface should be negligible.

Phosphoenolpyruvate carboxylase (PEPC) and phosphoenolpyruvate carboxykinase (PEPCK) interconvert PEP and oxaloacetate (OAA). Conversion of PEP to OAA by PEPC is irreversible (Chollet ). To our knowledge, there are no reports of PEPCK activity in mesophyll cells where bulk carbohydrate synthesis takes place (Pyke, 2001). PEPCK RNA and protein were not detected in leaves of Solanum lycopersicum irrespective of their developmental stage (Bahrami ; Famiani ). PEPCK protein or activity were not detected in leaves of Hordeum vulgare (Chen ). In mature leaves of A. thaliana, PEPCK protein amount and activity are low and probably confined to specific cell types (Malone ). In leaves of N. tabacum, PEPCK occurs in trichomes and stomata (Leegood ; Malone ). In leaves of Cucumis sativus, PEPCK occurs in trichomes and phloem cells (Leegood ; Chen ). In leaves of Oryza sativa, PEPCK occurs in hydathodes, stomata, and the vascular parenchyma (Bailey and Leegood, 2016). Thus, transmission of 13C signals in OAA to cytosolic carbohydrates by the PEPC/PEPCK interface should not occur.

3-Deoxy-d-arabino-heptulosonate 7-phosphate synthase (DAHPS) catalyses the irreversible reaction from PEP and erythrose 4-phosphate (E4P) to 3-deoxy-d-arabino-heptulosonic acid 7-phosphate (DAHP; Herrmann, 1995). To our knowledge, there are no reports of enzymes catalysing the back reaction. Thus, transmission of 13C signals in DAHP to cytosolic carbohydrates should not occur.

Taken together, leaf-level carbon fluxes from OAA, pyruvate, and DAHP to PEP are negligible or absent (Figs 1, 3). Therefore, reactions downstream of OAA, pyruvate, and DAHP cannot feed significant amounts of carbon and associated 13C signals into carbohydrate metabolism.

Exclusion of starch, sucrose, and cellulose metabolism, and the pentose phosphate pathway as origin of the Δ5–6ʹ signal

Reactions leading directly from stromal GAP to the formation of starch, sucrose, and cellulose, reactions remobilizing starch, and reactions of the pentose phosphate pathway do not simultaneously modify carbon bonds that become glucose C-5 and C-6. This excludes these pathways for Δ5–6ʹ signal introduction.

Origin of the Δ5–6ʹ signal

After excluding several metabolic locations as the origin of the Δ5–6ʹ signal (see ‘Exclusion of metabolic locations as origin of the Δ5–6ʹ signal’), the glycolytic pathway and PEP metabolism in leaves are left for consideration. Within this system, cytosolic GAP is used for PEP metabolism and sucrose synthesis (Fig. 1). Thus, GAP constitutes a central branch point in leaf carbon metabolism enabling isotope fractionation.

Leaf-level enolase, PEPC, PK, and/or DAHPS introduce the Δ5–6ʹ signal

Enolase, PEPC, PK, and DAHPS are the only enzymes which simultaneously modify carbon bonds that become glucose C-5 and C-6 (Figs 1–3). Reactions catalysed by these enzymes may be accompanied by 13C effects of substantial size and may thus introduce the Δ5–6ʹ signal. Enolase interconverts 2-phosphoglycerate (2PGA) and PEP (Figs 1, 3). In vivo, the reaction operates close to equilibrium (Kubota and Ashihara, 1990) and might thus be accompanied by an equilibrium isotope effect. Formation of the C=C double bond in PEP probably favours turnover of 13C isotopologues of 2PGA, leading to 13C enrichment in PEP. A 13C signal might then arise from varying allocation of PEP to downstream processes (Fig. 1). Increased downstream consumption would remove more 13C-enriched PEP and leave behind more 12C-enriched 2PGA for glucose synthesis.

Kinetic isotope effects may accompany the unidirectional conversions of PEP to OAA by PEPC, PEP to pyruvate by PK, and PEP and E4P to DAHP by DAHPS (Figs 1, 3). These reactions break the C=C double bond in PEP and can therefore be expected to favour turnover of 12C-isotopologues of PEP, leaving behind 13C-enriched PEP for glucose synthesis. Due to the usually larger size of kinetic isotope effects compared with equilibrium isotope effects, effects by PEPC, PK and DAHPS can be expected to outweigh any reciprocal effect by enolase. Thus, considering all four enzymes together, increasing turnover of PEP by PEPC, PK, and DAHPS can be expected to result in 13C-enriched tree-ring glucose (i.e. Δ5–6ʹ decreases).

Signal transmission to tree-ring glucose

Isotope signals generated at the level of leaf-cytosolic PEP or 2PGA need to be transmitted to GAP to then enter hexose phosphates and tree-ring glucose (Fig. 1). Transmission of a signal introduced by cytosolic enzymes is straightforward since the cytosolic glycolytic reactions between PEP and GAP are at equilibrium (Kubota and Ashihara, 1990). Transmission of a signal introduced by stromal enzymes is more intricate. First, it requires an incomplete or low-activity glycolytic pathway in leaf chloroplasts because signal equilibration with stromal triose phosphates would result in even signal distribution over all glucose carbon positions (Supplementary Protocol S1.6). An incomplete glycolytic pathway in leaf chloroplasts is supported by a reported lack of enolase in A. thaliana and O. sativa (van der Straeten ; Prabhakar ; Fukayama ). Second, signal transmission from stromal PEP to C-5 and C-6 of cytosolic hexose phosphate requires chloroplast export of PEP. Transport of PEP across the chloroplasts’ inner membrane is mediated by the PEP/Pi translocator as counter-exchange with Pi, PEP, or 2PGA, and the putative in vivo preference for the transport of Pi, and PEP (Fischer ; Flügge ). Numerous stromal processes, such as the shikimate pathway and fatty acid biosynthesis, rely on PEP import from the cytosol (Streatfield ; Flügge ). Therefore, a net flux of PEP from the cytosol to chloroplasts can be expected. However, members of the phosphate translocator family are believed to be highly inefficient. For instance, merely 10% of the activity of the triose phosphate translocator is used for net export of triose phosphate from chloroplasts; 90% is wasted on futile counter-exchanges (Flügge, 1987, 1999). In addition, low stromal and high cytosolic Pi levels (Sharkey and Vanderveer, 1989) can be expected to promote chloroplast export of PEP. Consequently, efficient equilibration of cytosolic and stromal PEP pools, and 13C signals therein, can be expected. Thus, both cytosolic and stromal enzymes may contribute to the Δ5–6ʹ signal.

Signal introduction requires substantial carbon fluxes and flux variability

For the introduction of a 13C signal, a substantial share of the photosynthetically fixed carbon must be directed towards enolase, PEPC, PK, and/or DAHPS and their downstream derivatives. This share must vary substantially; in the present case, on the interannual time scale. Therefore, we will now discuss carbon fluxes through enolase, PEPC, PK, and DAHPS.

Commonly, PEPC is localized in the cytosol both in dispersion and bound to the outer mitochondrial membrane (Figs 1, 3; O’Leary ). In leaf mesophyll cells of O. sativa, a putatively rare additional isoform occurs in chloroplasts (Masumoto ; O’Leary ). In C3 plants, PEPC provides OAA to replenish tricarboxylic acid cycle intermediates, and to support nitrogen assimilation and biosynthetic processes (O’Leary ; O’Leary and Plaxton, 2017). On average, leaf carbon fixation by PEPC is believed to account for up to 5% of net CO2 assimilation (Melzer and O’Leary, 1987). Up-regulation of PEPC occurs (inter alia) with drought, salinity, ozone, nitrogen assimilation, and virus infections (see ‘Ecophysiological effects’; O’Leary ). For instance, ozone triggers both an up-regulation of PEPC and a down-regulation of Rubisco (Dizengremel, 2001). In forest trees, the Rubisco/PEPC activity ratio can change from up to 25 in ozone-free air to ~2 under realistic levels of ambient ozone, redirecting carbon flux to maintenance and repair processes (Dizengremel, 2001).

Isoforms of PK are localized in both the cytosol and chloroplasts (Figs 1, 3; Ambasht and Kayastha, 2002). They provide pyruvate (inter alia) for mitochondrial respiration, fatty acid biosynthesis, and the non-mevalonate pathway (MEP). To our knowledge, estimates of the respiratory flux via PK in actively photosynthesizing leaves are unavailable. However, this flux may be substantial when photorespiration is low and thus co-vary with photorespiration and its environmental controls (Supplementary Protocol S3).

In illuminated photosynthetic tissue of A. thaliana, fatty acid biosynthesis can occur at a rate of 2.3 µmol C mg chlorophyll–1 h–1 (Bao ). Based on this, we estimate an ~2% carbon flux relative to net CO2 assimilation into fatty acid biosynthesis (Supplementary Protocol S4). In leaves, this flux is predominantly controlled at the level of acetyl-CoA carboxylase (Page ; Harwood, 2005; Ohlrogge ). It responds to the stromal redox state (energy status) and associated environmental controls (Rawsthorne, 2002; Harwood, 2005; Geigenberger and Fernie, 2014).

The plastid-localized MEP pathway is yet another metabolic route carrying substantial flux. With isoprene as a major pathway product in some trees, it commonly consumes ~2% of net assimilated CO2 (Sharkey and Yeh, 2001). In forest trees, high temperature can enhance this fraction to up to 15% (Sharkey ); a plant response believed to mitigate short-term high-temperature stress (Sharkey and Yeh, 2001).

DAHPS, located in both chloroplasts and the cytosol, is the first enzyme of the shikimate pathway (Figs 1, 3; Maeda and Dudareva, 2012). In vascular plants, 20–50% of the photosynthetically fixed carbon enters the pathway (Tohge ). In trees, most of the flux can be expected to occur in heterotrophic tissues supporting lignin biosynthesis. To our knowledge, flux estimates for actively photosynthesizing leaves are unavailable. However, the shikimate pathway provides precursors for (inter alia) the aromatic amino acids phenylalanine, tryptophan, tyrosine, and their numerous derivatives. Thus, it should carry substantial flux in most tissues. In leaves of Prunus persica fed 13CO2, <6% of the label accumulated in a metabolite fraction comprising lipids, proteins, and residual compounds (Escobar-Gutiérrez and Gaudillère, 1997). Since the shikimate pathway contributes to the biosynthesis of this metabolite fraction among other pathways, its flux must be markedly below 6% of net assimilated CO2. In leaves of Helianthus annuus, Abadie reported a flux of ~1% relative to net CO2 assimilation into the shikimate pathway product chlorogenate under normal growing conditions. Regulation of the shikimate pathway is primarily exerted by gene expression and post-translational modification in response to developmental and environmental cues (Entus ; Mir ). Relative carbon flux through the shikimate pathway can be expected to (inter alia) vary with light (Henstrand ; Logemann ; Entus ), ozone (Janzik ; Betz ), physical wounding (Dyer ; Keith ), bacterial infection (Keith ; Truman ), fungal infestation (McCue and Conn, 1989; Henstrand ; Görlach ; Bischoff et al., 1996, 2001; Ferrari ), and nitrogen availability (Scheible ). For instance, in leaves of N. tabacum, induction of DAHPS increased up to 5-fold under ozone fumigation (160 nl l–1), and an increase in flux through the shikimate pathway was corroborated by increased levels of pathway products (Janzik ). Performing an 83 d ozone fumigation experiment (160–190 nl l–1, 8 h d–1), Betz reported evidence for increased carbon flux into the shikimate pathway in leaves of Fagus sylvatica.

Since PEPC, PK, and DAHPS are located downstream of enolase (Figs 1, 3), all four enzymes may contribute to the Δ5–6ʹ signal. Based on arguments given above, associated carbon fluxes and their variability can be expected to be substantial. Other leaf-level pathways consuming PEP, such as the cytosolic mevalonate pathway, may exert additional control over the Δ5–6ʹ signal.

Ecophysiological effects

The Δ5–6ʹ signal is independent of the diffusion–Rubisco signal at C-1 and C-2 (Fig. 4). Since diffusion–Rubisco fractionation initially affects all carbon entering glucose synthesis equally (see the Introduction), we propose that the Δ5–6ʹ signal exhibits two components of variance. The first component is inversely correlated with diffusion–Rubisco fractionation and removes the diffusion–Rubisco signal from glucose C-5 and C-6. The second component constitutes systematic variation independent of diffusion–Rubisco fractionation. In the following, we propose ecophysiological mechanisms for the introduction of each component starting with the independent component. Please note that, in the present case, the Δ5–6ʹ signal can be expected to be under environmental rather than developmental control (Supplementary Protocol S5).

Wieloch studied effects of VPD, precipitation, soil moisture, temperature, and global radiation on the diffusion–Rubisco signal in their P. nigra samples. These authors found that VPD, a measure of environmental drought, exerts predominant control and pointed out that this agrees with expectations for the generally dry study site. Thus, the independent component of the Δ5–6ʹ signal is governed by environmental factors other than VPD.

The study site is ~10 km away from the city centre of Vienna and frequently exposed to substantial levels of tropospheric ozone (Oltmans ; Ainsworth ). Lefohn (1992) classified P. nigra as an ozone-sensitive tree species. Radiation stimulates the photochemical reactions of ozone formation (Ainsworth ). Ozone triggers relative flux increases through the anaplerotic and shikimate pathways via PEPC and DAHPS, respectively (see ‘Signal introduction requires substantial carbon fluxes and flux variability’) and may thus cause 13C increases at PEP carbon positions that become glucose C-5 and C-6 (see ‘Leaf-level enolase, PEPC, PK, and/or DAHPS introduce the Δ5–6ʹ signal’). This may introduce an isotope signal independent of the diffusion–Rubisco signal due to independence at the level of environmental controls.

In contrast, a process mitigating ozone entry into plant leaves may explain the component of the Δ5–6ʹ signal which is inversely correlated with the diffusion–Rubisco signal. Dizengremel (2001) proposed that drought and ozone combined is a main recurring stress factor in forest ecosystems. Isohydric plant species, such as P. nigra, respond to drought by closing their stomata (Sade ). Reduced stomatal conductance impedes ozone uptake (Tingey and Hogsett, 1985; Dobson ; Dizengremel, 2001). In needles of Pinus halepensis, PEPC activities in control plants and plants exposed to mild drought stress were similar, strongly increased under ozone stress, but significantly less so under combined ozone and drought stress (Fontaine ). Thus, anaplerotic flux rates can be expected to be highest under ozone stress but lower when ozone stress is accompanied by drought. While drought causes 13C enrichments at all glucose carbon positions due to diffusion–Rubisco fractionation (Wieloch ), it can be expected to reduce ozone-induced 13C enrichments at glucose C-5 and C-6. This drought component of the ozone response may remove the diffusion–Rubisco signal from glucose C-5 and C-6. In Supplementary Protocol S3, we discuss how changes in substrate supply to mitochondrial oxidative phosphorylation (glycolytic pyruvate versus photorespiratory glycine) may additionally contribute to the component of the Δ5–6ʹ signal that is inversely correlated with diffusion–Rubisco fractionation.

Experimental evidence

Effects of tropospheric ozone on whole-molecule 13C/12C composition of plant cellulose

Growing B. pendula at increased ozone levels, several authors reported decreased 13C discrimination, Δ, in leaf and stem cellulose (Matyssek ; Saurer ). Intriguingly, these Δ decreases coincided with increased ratios of intercellular to ambient CO2 concentrations, Ci/Ca. As pointed out by Matyssek and Saurer , this cannot be explained by the standard model of diffusion–Rubisco fractionation which predicts a positive correlation between Ci/Ca and Δ (Farquhar ). Thus, post-Rubisco fractionation can be expected to cause these ozone-related isotope effects.

Matyssek and Saurer proposed that increased relative carbon fixation by PEPC due to ozone explains the Δ decreases because carbon fixed by PEPC is strongly 13C enriched compared with carbon fixed by Rubisco (Melzer and O’Leary, 1987). While this proposal is in line with significantly increased relative PEPC activities observed under ozone (Table 2), it conflicts with the set-up of carbon metabolism. PEPC-fixed carbon supplies downstream metabolism, yet no pathway carrying substantial flux exists that could transfer it into carbohydrate metabolism (see ‘Exclusion of reactions downstream of OAA, pyruvate, and DAHP as Δ5–6ʹ signal origin’). Above, we propose an ozone-dependent mechanism for the introduction of the Δ5–6ʹ signal which reconciles observations of Matyssek and Saurer with the set-up of carbon metabolism (see ‘Ecophysiological effects’).

Saurer reported differences in intercellular CO2 concentration, ΔCi, and whole-molecule 13C discrimination, ΔΔ, between ozone-treated and control plants. Plants grown with lower amounts of fertilizer (LF) exhibited ΔCi=7.5 ± 2.6 SE ppm, while plants grown with higher amounts of fertilizer (HF) exhibited ΔCi=21.5 ± 4.5 SE ppm. This corresponds to estimated increases in 13C discrimination by the diffusion–Rubisco interface of ΔΔDR=0.52 ± 0.18 SE ‰ and 1.49 ± 0.31 SE ‰, respectively (Fig. 5, dashed bars; Equation 1). However, Saurer reported ΔΔ decreases in leaf and stem cellulose under both fertilization regimes (Fig. 5, solid bars; Equations 2, 3). With respect to ΔΔDR, these decreases are statistically significant (one-tailed t-test: P<0.05) except for leaf cellulose synthesized under HF conditions which comes, however, close to being statistically significant (P<0.08).

Fig. 5.

Differences in 13C discrimination between ozone-treated and control plants, ΔΔ (90/40 day/night versus <3 nl O3 l–1). ‘Leaf’ and ‘Stem’ refer to leaf and stem cellulose of Betula pendula, respectively. LF and HF refer to plants grown with low and high amounts of fertilizer, respectively. Solid bars: measured ΔΔ values. Dashed bars: expected ΔΔ values. Expected ΔΔ values were estimated using a model by Farquhar . This model describes 13C discrimination associated with plant carbon uptake including CO2 diffusion into plant leaves and assimilation by Rubisco. Numbers inside bars denote differences between measured and expected ΔΔ values. Statistically significant differences are marked by asterisks (one-tailed t-test: ∗P<0.05; ∗∗P<0.01). The difference of the Leaf/HF treatment is close to being statistically significant (P<0.08). This analysis is based on data published by Saurer .

In B. pendula, post-Rubisco fractionation causes average whole-molecule ΔΔ decreases of approximately –1.98 ± 0.58 SE ‰ (Fig. 5). Below, we propose that a fraction of the Δ5–6ʹ signal enters glucose C-1 to C-4 through indirect signal propagation via chloroplast metabolism (see ‘Signal propagation to all glucose carbons via chloroplast metabolism’). We estimate that the signal at C-5 and C-6 is 6.625-fold larger than at C-1 to C-4 (Supplementary Protocol S1.8). Thus, an approximately –1.98 ± 0.58 SE ‰ effect at the whole-molecule level scales to approximately –4.56 ± 1.34 SE ‰ effects at cellulose glucose C-5 and C-6 and to approximately –0.69 ± 0.20 SE ‰ effects at C-1 to C-4 (Supplementary Protocol S6). In P. nigra, measured Δ5–6ʹ values fall within a 5.80 ± 1.55 SE ‰ range (maximum=22.71 ± 0.99 SE ‰, minimum=16.91 ± 0.56 SE ‰). Wieloch estimated that the Δ5–6ʹ time series contains 79% systematic and 21% error variance. Assuming the error is fully expressed in both the maximum and minimum value, we estimate a systematic time series range of ~4.58 ± 1.22 SE ‰ (5.80 ± 1.55 SE ‰×0.79). This largely equals the estimated effect at glucose C-5 and C-6 in ozone-treated B. pendula, corroborating the theory proposed above. Notably, occurrence of the post-Rubisco fractionation effect in leaf cellulose of B. pendula corroborates the proposed leaf-level origin of the Δ5–6ʹ signal.

Effect of tropospheric ozone on the Δ5–6ʹ signal in tree-ring glucose

In Vienna, [O] is measured at five sites. Complete time series for all sites are available since 1992. Intra-annually, the highest [O] occurs during the period April to August (Fig. 6A) which can be expected to affect tree metabolism. Therefore, we calculated an April to August average time series for the Vienna region covering the period 1992–2020 (Fig. 6B, solid black line). We found that April to August SD and rH explain 59% of the time series variability (Fig. 6B, dashed black line, P<0.00001, n=29):

Fig. 6.

(A) Average monthly ozone concentrations, [O], measured in Vienna over the period 1992–2020. (B) April to September [O] in Vienna. Black solid line, measured [O]; black and red dashed lines, modelled and reconstructed [O], respectively, based on the relationship of [O] with sunshine duration, SD, and relative humidity, rH: [O]=58.28 + 0.02951SD–0.3697rH, R2=0.59, P<0.00001, n=29. (C) 13C discrimination at glucose C-5 and C-6, Δ5–6ʹ, as a function of [O]: Δ5–6ʹ=35.24–0.2258[O], R2=0.33, P<0.001, n=31. Isotope data were measured on tree-ring glucose of Pinus nigra laid down from 1961 to 1995 at a dry site in the Vienna basin. SD and rH were measured at the climate station Hohe Warte (Vienna). [O] was measured at five stations in Vienna (Stephansplatz, Laaer Berg, Hermannskogel, Hohe Warte, and Lobau).

Other studies report similar relationships (Felipe-Sotelo ; Kovač-Andrić ). Based on Equation 4, we reconstructed [O] for the Vienna region over the period 1961–1991 (Fig. 6B, dashed red line). We found that April to August [O] (reconstructed for 1961–1991 and measured for 1992–1995) explains 33% of the Δ5–6ʹ time series variability (Fig. 6C, P<0.001, n=31):

Accounting for measurement error, 75% of the variance in Δ5–6ʹ is explainable by modelling (cf. Nilsson ). Thus, [O] explains 44% of the explainable variability in Δ5–6ʹ (33%/75%×100). This is substantial considering that measured and modelled [O] time series merely exhibit 59% co-variability (Fig. 6B, solid and dashed black line). In summary, these results corroborate the proposed negative relationship between ozone stress and the Δ5–6ʹ signal (see ‘Ecophysiological effects’), and, by extension, the theoretical framework developed above.

Implications of the theory

Signal propagation at the level of TPI in the cytosol of leaves

The post-Rubisco signal at glucose C-5 and C-6 is independent of a signal at glucose C-1 and C-2 (Fig. 4; Wieloch ); that is, substantial signal propagation from C-5 and C-6 to C-2 and C-1 is not supported by the data. This is surprising for the following reason: transmission of the Δ5–6ʹ signal from its origin, the lower end of the glycolytic pathway, into carbohydrate metabolism occurs via GAP (Fig. 1). Hence, signal independence requires negligible conversion of GAP (a precursor of glucose C-4 to C-6) to DHAP (a precursor of glucose C-1 to C-3) via leaf-cytosolic TPI (Figs 1, 2). Since TPI is often referred to as the prime example for the efficiency of enzyme catalysis, one would expect full equilibration of GAP and DHAP and inherent isotope signals. This view, however, is based on in vitro measurements of TPI kinetics. The following mechanisms may explain the apparent lack of equilibration and signal propagation in vivo.

In the light, chloroplast export of DHAP is favoured by the equilibrium position of stromal TPI, which is strongly on the side of DHAP (Walker, 1976; Knowles and Albery, 1977). Sharkey and Weise (2012) calculated that there should be >20 times more DHAP than GAP at equilibrium (Meyerhof and Junowicz-Kocholaty, 1943; Bassham and Krause, 1969). The substrate affinities of the triose phosphate translocator for DHAP and GAP are similar at Km=0.13 mM and Km=0.08 mM, respectively (Fliege ). Thus, DHAP and GAP will be transported according to their concentrations; that is, 20 times more DHAP will be exported from chloroplasts to the cytosol. However, synthesis of fructose 1,6-bisphosphate uses DHAP and GAP at a 1:1 ratio. This may keep the concentration of leaf cytosolic GAP low. Flux of GAP into glycolysis and processes consuming glycolytic intermediates will additionally contribute to low cytosolic GAP concentrations. This may restrict the GAP to DHAP back-conversion. Furthermore, numerous common metabolites inhibit TPI competitively (Anderson, 1971; Grüning ; Flügel ; Li ). In addition, the activity of cytosolic TPI decreases significantly upon treatment with reactive oxygen species, especially H2O2 (Lopez-Castillo ). Thus, during active photosynthesis, a lack of isomeric and isotopic equilibrium between leaf cytosolic GAP and DHAP is conceivable. This would block the propagation of 13C signals in GAP to DHAP and enable independent 13C signals in Δ5–6ʹ and Δ1–2ʹ as observed.

Signal propagation to all glucose carbons via chloroplast metabolism

Cytosolic PEP, 2PGA, 3-phosphoglycerate, 1,3-bisphosphoglycerate, and GAP carry the Δ5–6ʹ signal from its putative origin (enolase, PEPC, PK, and DAHPS) directly into C-5 and C-6 of hexose phosphates (Figs 1–3). In addition, indirect signal propagation to C-1 to C-6 can be expected via import of 3-phosphoglycerate into chloroplasts (Supplementary Protocol S1.2). A model of signal propagation described in Supplementary Protocol S1 has several implications. First, observed Δ5–6ʹ signals at C-5 and C-6 are 6.625-fold larger than at C-1 to C-4 (Supplementary Protocol S1.8). Second, the original signal as introduced at the level of enolase, PEPC, PK, and DAHPS is 1.4-fold larger than the signal observed at glucose C-5 and C-6 (Supplementary Protocol S1.9). Third, clustering of Δ4ʹ with the Δ5ʹ to Δ6ʹ cluster (Fig. 4) can be explained by signal propagation via chloroplast metabolism (Supplementary Protocol S1.10).

Signal propagation to other plant metabolites

We propose that carbon flux changes around leaf cytosolic enolase, PEPC, PK, and DAHPS introduce the Δ5–6ʹ signal. Hence, upstream derivatives of 2PGA carbons corresponding to glucose C-5 and C-6 will inherit the signal (Fig. 1). Compared with plant cellulose, the signal will be distinctly smaller in chloroplast derivatives, such as starch, and distinctly larger in leaf sucrose synthesized during the photoperiod (Supplementary Protocol S1). This latter metabolite may be used to follow the Δ5–6ʹ signal on an hourly basis. Downstream derivatives of PEP carbons corresponding to glucose C-5 and C-6 will obtain an inverse Δ5–6ʹ signal. These differences may help to test the theory.

Turnover by PEPC, PK, and DAHPS can be expected to result in 13C-enriched PEP (see ‘Leaf-level enolase, PEPC, PK, and/or DAHPS introduce the Δ5–6ʹ signal’). Thus, our theory can help to explain the 13C/12C separation observed among plant compounds, specifically between 13C-enriched carbohydrates and 13C-depleted metabolites downstream of PK, PEPC, and DAHPS (Sharkey ; Bathellier ).

Implications for whole-molecule 13C/12C analysis

The Δ5–6ʹ signal has two components of variance (see ‘Ecophysiological effects’). One is inversely correlated with diffusion–Rubisco fractionation, and the other is independent. Both components have implications for studies of plant carbon uptake and associated properties by whole-molecule 13C/12C analysis. The inversely correlated component removes the diffusion–Rubisco signal from glucose C-5 and C-6. In addition, this signal is absent at glucose C-4 (Wieloch ); that is, three out of six glucose carbon positions lack the diffusion–Rubisco signal. Thus, whole-molecule 13C/12C analysis captures an attenuated diffusion–Rubisco signal and underestimates the variability of the original signal and associated physiological properties, such as Ci/Ca and photosynthetic water use efficiency.

The independent component of the Δ5–6ʹ signal weakens signal extractions from whole-molecule 13C/12C measurements because it constitutes pseudorandom noise with respect to diffusion–Rubisco fractionation. This may explain why models of whole-molecule diffusion–Rubisco fractionation as functions of environmental properties often suffer from low explanatory powers, R2≤0.5 (Barbour and Song, 2014). In contrast, intramolecular 13C/12C analysis resolves information about distinct ecophysiological processes; a fundamental conceptual advancement enabling more adequate modelling of the variability of plant carbon uptake and associated environmental/developmental controls.

Tracking carbon allocation in other biological organisms

Whole-molecule 13C/12C analysis enables assessments of plant carbon uptake (Farquhar ). According to theory reported here, intramolecular 13C/12C analysis enables additional assessment of downstream carbon allocation in actively photosynthesizing leaves. This includes carbon flux into the anaplerotic, shikimate, MEP, and fatty acid biosynthesis pathways, and mitochondrial respiration (Fig. 1). Intramolecular 13C signals are governed by a small set of physicochemical principles that apply generally (Schmidt ). Thus, intramolecular 13C/12C analysis can be expected to enable retrospective assessment of carbon allocation in any biological organism including, for instance, disease-related shifts.

Utility of the Δ5–6ʹ signal

Laboratory experiments are limited to short time scales and in their capabilities to reproduce complex natural systems. Manipulation experiments on natural systems are limited to time scales of years and may suffer from spurious effects due to unnatural step changes in ambient conditions. In contrast, tree-ring analysis offers extensive temporal, spatial, species, and genotype coverage of natural systems that have not been subjected to unnatural step changes.

We propose that the Δ5–6ʹ signal reports flux into the anaplerotic pathway including CO2 uptake by PEPC (Figs 1, 3, 5, 6). In addition, it may report flux into mitochondrial respiration (Supplementary Protocol S3). Thus, signal analysis may enable a better understanding of plant and ecosystem carbon balances including the so-called CO2 fertilization effect.

Furthermore, intramolecular 13C/12C analysis enables analysis not only of carbon uptake–environment relationships but also of carbon allocation–environment relationships (Figs 5, 6) and thus more comprehensive assessments of flux-level plant performance. For instance, atmospheric CO2 and ozone concentrations have increased over recent years (Fig. 6B). Under business-as-usual scenarios, this will continue over the next decades (Turnock ). While CO2 promotes leaf photosynthesis and net primary productivity, ozone has the very opposite effect (Ainsworth ; IPCC, 2014). When this highly reactive chemical enters plant leaves through stomata, it causes harm to structure and function, and leads to major rearrangements in carbon metabolism. While ozone decreases carbon fixation, it increases carbon allocation to costly maintenance and repair processes (Dizengremel, 2001; Ainsworth ). This includes increased carbon flux into the anaplerotic and shikimate pathway, and this resource investment is likely to be recorded in the Δ5–6ʹ signal. Ozone tolerance varies among species, with metabolic changes depending on the duration of ozone exposure (Fontaine ; Ainsworth ). Thus, the Δ5–6ʹ signal may support flux-level screenings for species/genotypes with the capacity to optimally adjust to prolonged ozone exposure (requires further investigation).

While glucose positions C-1 to C-3 preserve the VPD-dependent carbon uptake signal (Wieloch ), this signal was removed from glucose C-5 and C-6 and replaced by an independent ozone-sensitive carbon allocation signal (Figs 5, 6). Thus, intramolecular 13C/12C analysis yields information about several environmental variables and may enable more powerful paleoenvironment reconstructions than whole-molecule analysis.

Lastly, sampling glucose at different developmental stages may enable the detection of shifts in carbon uptake and allocation over the life span of plants to better understand basic physiological processes such as plant senescence. In conclusion, intramolecular 13C/12C analysis opens up numerous new avenues of research within the plant and Earth sciences.

Supplementary data

The following supplementary data are available at JXB online.

Protocol S1. Signal propagation to all glucose carbons via chloroplast metabolism and implications.

Protocol S1.1. Signal propagation upon reimport of cytosolic metabolites into chloroplasts.

Protocol S1.2. Predominant import molecules.

Protocol S1.3. Predominant export molecules.

Protocol S1.4. Signal import into chloroplasts.

Protocol S1.5. Signal dilution and partial signal loss.

Protocol S1.6. Stromal signal redistribution

Protocol S1.7. Signal size in stromal triose phosphates.

Protocol S1.8. Signal size in cytosolic hexose phosphates.

Protocol S1.9. Original size of the Δ5–6ʹ signal.

Protocol S1.10. Signal propagation can explain the clustering of the Δ4ʹ to Δ6ʹ time series.

Protocol S1.11. The role of chloroplast starch in signal propagation.

Protocol S1.12. Signal import into chloroplasts via cytosolic 1,3BPGA, 2PGA, and PEP.

Protocol S1.13. Assumptions.

Protocol S2. Correction for 13C signal redistribution by triose phosphate cycling in tree-ring cells.

Protocol S3. Alternating substrate supply to oxidative phosphorylation may contribute to the Δ5–6ʹ signal.

Protocol S4. Carbon flux into fatty acid biosynthesis in illuminated photosynthetic tissue.

Protocol S5. Potential influence of environmental and developmental variables on the Δ5–6ʹ signal.

Protocol S6. Intramolecular isotope effects in response to ozone.

Click here for additional data file.