Detecting long-term metabolic shifts using isotopomers: CO₂-driven suppression of photorespiration in C₃ plants over the 20th century

Significance

Decadal-scale metabolic responses of plants to environmental changes, including the magnitude of the “CO₂ fertilization” effect, are a major knowledge gap in Earth system models, in agricultural models, and for societal adaptation. We introduce intramolecular isotope distributions (isotopomers) as a methodology for detecting shifts in plant carbon metabolism over long times. Trends in a deuterium isotopomer ratio allow quantification of a biogeochemically relevant shift in the metabolism of C₃ plants toward photosynthesis, driven by increasing atmospheric CO₂ since industrialization. Isotopomers strongly increase the information content of isotope archives, and may therefore reveal long-term acclimation or adaptations to environmental changes in general. The metabolic information encoded in isotopomers of plant archives bridges a fundamental gap between experimental plant science and paleoenvironmental studies.

Abstract

Terrestrial vegetation currently absorbs approximately a third of anthropogenic CO₂ emissions, mitigating the rise of atmospheric CO₂. However, terrestrial net primary production is highly sensitive to atmospheric CO₂ levels and associated climatic changes. In C₃ plants, which dominate terrestrial vegetation, net photosynthesis depends on the ratio between photorespiration and gross photosynthesis. This metabolic flux ratio depends strongly on CO₂ levels, but changes in this ratio over the past CO₂ rise have not been analyzed experimentally. Combining CO₂ manipulation experiments and deuterium NMR, we first establish that the intramolecular deuterium distribution (deuterium isotopomers) of photosynthetic C₃ glucose contains a signal of the photorespiration/photosynthesis ratio. By tracing this isotopomer signal in herbarium samples of natural C₃ vascular plant species, crops, and a Sphagnum moss species, we detect a consistent reduction in the photorespiration/photosynthesis ratio in response to the ∼100-ppm CO₂ increase between ∼1900 and 2013. No difference was detected in the isotopomer trends between beet sugar samples covering the 20th century and CO₂ manipulation experiments, suggesting that photosynthetic metabolism in sugar beet has not acclimated to increasing CO₂ over >100 y. This provides observational evidence that the reduction of the photorespiration/photosynthesis ratio was ca. 25%. The Sphagnum results are consistent with the observed positive correlations between peat accumulation rates and photosynthetic rates over the Northern Hemisphere. Our results establish that isotopomers of plant archives contain metabolic information covering centuries. Our data provide direct quantitative information on the “CO₂ fertilization” effect over decades, thus addressing a major uncertainty in Earth system models.

Atmospheric CO₂ levels have increased from ∼200 ppm during the last ice age to currently 400 ppm, and they may, according to pessimistic scenarios, exceed 1,000 ppm in the year 2100 (1). Understanding plant responses to increasing CO₂ is currently hampered by two fundamental limitations: First, it is unknown how well manipulation experiments represent responses to the gradual CO₂ increase over decades and centuries. In Free-Air CO₂ Enrichment (FACE) experiments, which most closely mimic natural conditions, increases in [CO₂] generally increase plant growth, but this “CO₂ fertilization” effect often declines after a few years of enrichment (2). Such transient responses may be related to the step increases in [CO₂] used in the experiments, their limited duration (2), or factors other than CO₂ becoming limiting (3). Second, in response to the [CO₂] increase since industrialization, genetic (4) and phenotypic plant responses (5–7) have been observed. Although century-scale changes have been detected in carbon isotopes (δ¹³C) and attributed to [CO₂], these responses are tied to differences in intercellular substrate concentrations that reflect several metabolic fluxes and diffusion processes (8). However, the responses do not measure the metabolic fluxes directly. The consequently poor constraint of the magnitude of the CO₂ fertilization effect over decadal to century time scales (1, 9) is a major source of uncertainty in parameterizations of Earth system models (1, 10) and crop productivity models (11, 12). Therefore, observational data are essential to enable global carbon models to provide reliable long-term predictions.

Plants that fix carbon by the C₃ photosynthetic pathway (C₃ plants) account for most (ca. 75%) global primary production and human food production (13). In C₃ plants, CO₂ initially reacts with D-ribulose-1,5-bisphosphate (RubP) catalyzed by the enzyme D-ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco), but this enzyme also has a competing O₂ fixation activity (14). The resulting carboxylation and oxygenation processes create photosynthetic and photorespiratory metabolic fluxes (15), which, respectively, cause C gains and losses for plants (Fig. 1), and which, globally, are among the largest biogeochemical C fluxes (16). Thus, the balance between these metabolic fluxes is a major determinant of the net primary productivity of C₃ plants in response to increasing [CO₂], and therefore is a major influence on the CO₂ sink strength of terrestrial vegetation, including the performance of C₃ crop plants.

Fig. 1.

Overview of metabolic pathways emanating from RubP by Rubisco activity. Upon carboxylation, 3-phosphoglycerate (3-PGA) molecules are formed from C1−C2 and C3−C5 of RuBP. Upon oxygenation, one 3-PGA molecule is formed from C3−C5 of RubP, and up to one-half 3-PGA is formed via the photorespiration pathway through the peroxisome and mitochondrion. The color coding of hydrogen atoms tracks the biochemical origins of hydrogens at C3 of 3-PGA, which give rise to the isotopomer signal encoded in the colored C6H₂ group of glucose.

Stable isotopes such as deuterium (D) are key tools for probing plants’ ecological and biogeochemical interactions with their environment. Conventional compound-specific stable isotope methods provide the D/H ratio, usually expressed as δD (17), of an entire molecule, that is, averaged over all intramolecular positions in the molecule. This means that conventional stable isotope methods are based on isotopologues, which, by definition, differ in the number of isotopic substitutions. The same situation applies to the carbon isotope ratio ¹³C/¹²C expressed as δ¹³C. The δD or δ¹³C signals of plant metabolites largely depend on three mechanisms: the isotope signatures of the plant’s H₂O and CO₂ sources, isotope fractionation during substrate uptake, and isotope fractionation during biosynthesis, primarily enzyme isotope effects.

It is well established that enzyme isotope effects influence stable isotope abundance in specific intramolecular positions (18, 19). A molecule carrying an isotope substitution in a specific intramolecular position is termed an isotopomer. Deuterium isotopomer abundances can be measured by NMR. Glucose, measured as a derivative that gives highly resolved deuterium NMR spectra (see Methods), gives rise to seven signals with variable integrals (Fig. 2A), which reflect the abundance of each of the seven D isotopomers of glucose. Isotopomer variation encodes metabolic information (18, 19), but isotopomers have not yet been used for unraveling long-term metabolic changes induced by environmental drivers.

Fig. 2.

Effect of [CO₂] on the D6^S/D6^R isotopomer ratio of photosynthetically generated glucose moieties in sunflower (H. annuus). (A) Deuterium NMR spectrum of a glucose derivative displaying one signal for each of the seven isotopomers of glucose. The signals’ integrals are proportional to the isotopomer abundances. (B) Excerpts of deuterium NMR spectra of glucose prepared from sunflower leaf starch, showing signals arising from the D6^S and D6^R isotopomers of the C6H₂ group of glucose. The solid and dashed spectra were acquired from glucose formed at 280 ppm and 1,500 ppm CO₂, respectively. The dashed spectrum has been shifted sideways to ease comparison. (C) Dependence of the D6^S/D6^R ratio of glucose from sunflower leaf starch on 1/[CO₂] (in units of 10⁻³ ppm⁻¹) during growth [r² = 0.88, slope 0.057 ±0.006 (SEM), P < 10⁻⁷, n = 17, individual plants except for 180 and 280 ppm, where material from two to four plants had to be pooled for each sample].

Here, we establish isotopomers as a ground-breaking technique for evaluating long-term effects of gradual [CO₂] increase on plant metabolism at the biochemical level. We first use CO₂ manipulation experiments to identify an isotopomer signal that reflects the Rubisco oxygenation/carboxylation ratio of C₃ plants. We then trace the signal in herbarium samples of wild plants, crops, and Sphagnum mosses, to estimate changes in metabolic flux ratios in photosynthetic carbon metabolism of C₃ plants over the past century.

Results

An Isotopomer Signal of the Photorespiration/Photosynthesis Ratio.

We first exposed sunflower (Helianthus annuus), a C₃ plant, to CO₂ concentrations ranging from 180 ppm to 1,500 ppm. Expansions of deuterium NMR spectra of a glucose derivative (Fig. 2B) show signals that reflect the abundances of the D isotopomers of the C6H₂ group of glucose; these hydrogens are stereochemically and biochemically distinct, which is indicated by the labeling as D6^S and D6^R. The integral ratio of the signals, the D6^S/D6^R ratio, differs between the two spectra, which were obtained from glucose samples formed at 280 ppm and 1,500 ppm [CO₂], respectively. When CO₂ is varied, the D6^S/D6^R ratio shows a linear dependence on 1/CO₂ (Fig. 2C and Table S1). Because sunflower is a C₃ plant, the oxygenation/carboxylation metabolic flux ratio at Rubisco is proportional to the concentration ratio of the competing substrates O₂/CO₂ (20, 21), which, at constant [O₂], yields a linear dependence on 1/[CO₂]. That the D6^S/D6^R ratio reflects this metabolic flux ratio is supported by leaf gas exchange measurements (Fig. S1).

Table S1.

Gas exchange and isotopomer data obtained from experiments with sunflower leaves (group A)

Variable	CO2 treatment, ppm
	180	280	450	700	1,500
Ci, µM	5.2 ± 0.1	7.8 ± 0.2	11.65 ± 0.3	19.7 ± 0.6	49.1 ± 0.7
ϕ	0.51	0.34	0.23	0.14	0.05
D6S/D6R	1.187	1.107	0.944	0.962	0.883
	1.108	1.097	1.014	0.877	0.893
			1.007	0.925	0.904
				0.931	0.863
				0.923	0.921

The leaf temperature was 22 °C, giving an intercellular (dissolved) O₂ concentration of 277 μM and K_sp of 104 (59). C_i was derived from gas exchange measurements on sunflower leaves, and D6^S/D6^R ratios were measured on leaf starch. Values of C_i are averages ± SE. Samples represent individual plants, except for 180 ppm and 280 ppm, where material from two to four plants had to be pooled to obtain sufficient sample amount.

Fig. S1.

The D6^S/D6^R isotopomer ratio as a function of ϕ (the oxygenation to carboxylation ratio). Glucose derivative was prepared from structural carbohydrates of sunflower leaves; input data are given in Table S1. The red curve shows a root-least-squares fit of the data to Eq. S1, and the black line shows a linear fit of the data.

Photosynthetic oxygen production first appeared ca. 3 billion years ago, before which atmospheric [O₂] was low (22). Since then, it has risen greatly, with a concomitant increase in the wasteful Rubisco oxygenation reaction. To cope with oxygenation, C₃ plants have evolved the photorespiration pathway to recycle oxygenation products (23). It appears that evolutionary changes could not suppress Rubisco’s oxygenation activity, because of a strong trade-off between Rubisco’s reaction rate and its CO₂/O₂ selectivity (8, 23–25). In contrast, C₄ plants largely suppress oxygenation by locally increasing [CO₂] (23). Thus, [CO₂] dependence of the D6^S/D6^R ratio should be observable in all C₃ plants but should be absent in C₄ species. Accordingly, analysis of the D6^S/D6^R ratio of chloroplast starch of two additional C₃ species grown at experimentally increased [CO₂], spinach (Spinacia oleracea) and bean (Phaseolus vulgaris) (Fig. 3A and Table S2), confirmed the findings from sunflowers. In both species, the D6^S/D6^R ratio decreased by ∼0.1 upon doubling [CO₂], in accordance with the results from the experiment with sunflower (Fig. 2C). In marked contrast, but in agreement with the suppression of photorespiration in C₄ plants, the D6^S/D6^R ratios of glucose formed under 450 ppm and 1,200 ppm in the C₄ species maize (Zea mays) were not significantly different (P = 0.9). Furthermore, reducing the O₂ concentration influenced the D6^S/D6^R ratio in the same way as increasing [CO₂] in the tested C₃ plants (Fig. 3B), as expected for suppression of photorespiration. These mechanistic tests lead us to conclude that the D6^S/D6^R ratio is indeed a robust measure of the oxygenation/carboxylation metabolic flux ratio in C₃ plants.

Fig. 3.

Response of the D6^S/D6^R ratio of different species and metabolites to manipulation of the [CO₂]/[O₂] ratio. (A) [CO₂] manipulation; white bars represent low and black bars represent high [CO₂] of 360 ppm and 700 ppm, respectively, except for sunflower (200 ppm and 1,000 ppm). (B) O₂ manipulation; white bar represents ambient atmosphere, and gray bar represents ambient [CO₂] with reduced [O₂] (12%); glucose from soluble sugars was analyzed. Values are averages ± SEM (n = 2–5), except for one single sample pooled from several plants. *P < 0.05; **P < 0.01; ***P < 0.001 (ANOVA).

Table S2.

Summary of plant material, growth conditions, CO₂ treatments, and measured D6^S/D6^R values for greenhouse and growth chamber experiments

Group	Species	Growth conditions, duration of CO2 treatment	Sugar analyzed	Samples	CO2 treatment, ppm	D6S/D6R
						Low CO2	High CO2	Difference	ANOVA P
A	sunflower	greenhouse, 2 d	leaf starch	a	180/280/450/700/1,500	see Table S1	see Table S1	see Table S1	—
B	sunflower	growth chamber, from germination	leaf structural carbohydrates	b	200 and 1,000	1.043 ± 0.018	0.862	0.0181	—
C	maize*	greenhouse, 2 d	leaf starch	c	450 and 1,200	1.269 ± 0.013	1.270 ± 0.001	−0.001	0.966
D	spinach	growth chamber, from germination	leaf starch	d	360 and 700	0.950 ± 0.007	0.865 ± 0.005	0.085	<0.001†
			leaf soluble sugars	e	360 and 700	0.968 ± 0.014	0.877 ± 0.015	0.091	0.007
			leaf soluble sugars	f	360‡		0.849	0.101‡	—
E	bean	growth chamber, from germination	leaf starch	g	360 and 700	1.032 ± 0.006	0.923 ± 0.007	0.109	<0.001†
			leaf soluble sugars	h	360 and 700	1.004 ± 0.022	0.906 ± 0.017	0.098	0.024
			seed starch	i	360 and 700	0.920 ± 0.013	0.868 ± 0.011	0.052	0.037

^*Plants with C₃, C₄, and crassulacean acid metabolism (CAM) show distinct overall D isotopomer patterns, and high D6^S/D6^R values are characteristic of C4 metabolism (28). Importantly, the D6^S/D6^R ratio of the maize metabolite is independent of [CO₂].

^†Two-way ANOVA with species (spinach or bean) and CO₂ treatment as variables shows that D6^S/D6^R ratios differ between species (P = 10⁻⁹) and depend on [CO₂] (P = 10⁻⁷), with no interaction between the variables (P = 0.14), indicating that although D6^S/D6^R values are species-dependent, their shifts as function of [CO₂] are conserved among species. Such variation may reflect species-dependent D fractionation connected to the regulation of carbon metabolism.

^‡Twelve percent O₂ and difference is taken relative to 360-ppm treatment, sample e.

Next, we tested if the 1/[CO₂] dependence of the D6^S/D6^R ratio is preserved in glucose units of metabolites formed after export of carbohydrates from the chloroplast (leaf soluble sugars, leaf structural carbohydrates, and bean seed starch). The D6^S/D6^R ratio in all metabolites depended on [CO₂] as well (Fig. 3A), and statistical comparison of metabolites in bean and spinach revealed the same CO₂ response in leaf starch and soluble sugars, whereas, in bean seed starch, the response was smaller (two-way ANOVA, Table S2). This suggests that the [CO₂] dependence of the D6^S/D6^R ratio in initial photosynthates of C₃ plants may be weakened by subsequent metabolic transformations. A likely explanation is hydrogen isotope exchange with cellular water, which is known to occur during many enzyme reactions (26). If so, observed variations in the D6^S/D6^R ratio would put lower bounds on the underlying change in the oxygenation/carboxylation ratio. Despite the exchange reactions, the data on exported metabolites show that physiological information encoded in the isotopomers is transmitted to long-lived metabolites.

CO₂ Increase Since Industrialization Has Shifted the Photorespiration/Photosynthesis Ratio.

The D6^S/D6^R ratio is sensitive to [CO₂] across a concentration range including preindustrial to current levels (Fig. 2C). Therefore, analysis of the isotopomer ratio may retrospectively reveal metabolic shifts induced by increasing atmospheric [CO₂]. For a series of 18 archived sucrose samples from sugar beet, spanning the period 1890–2012 and a [CO₂] increase from ∼295 ppm to ∼395 ppm, the D6^S/D6^R ratio shows a highly significant linear dependence on 1/[CO₂] (Fig. 4 and Table S3). The regression slope (0.048 ± 0.012, SEM) is not significantly different from the slope determined for glucose moieties in greenhouse-grown sunflowers (Fig. 2C), showing that increasing atmospheric [CO₂] has reduced the oxygenation/carboxylation ratio in sugar beet to the same degree as in our greenhouse manipulation experiments. The [CO₂] dependence in archived beet sugar samples demonstrates that the photorespiration signal is highly robust, despite probable differences among the samples in agricultural practices, cultivars, and locations. This is remarkable, given that the modern samples and their progenitors experienced increasing [CO₂] over more than 100 growing seasons. Furthermore, all modern plant breeding has occurred during this time period—after rediscovery of Mendel’s laws in 1900—but has not detectably influenced the CO₂-driven change of the photosynthesis/photorespiration ratio. Finally, the agreement implies that the [CO₂]-driven metabolic shift during the past 100 y has not been counteracted by homeostatic adjustments of physiological properties, such as stomatal function or regulation of photosynthesis, in sugar beet.

Fig. 4.

Isotopomer response to changes in [CO₂] in archived samples of beet sugar. D6^S/D6^R ratios of the glucose moiety of beet sucrose samples as a function of 1/[CO₂] (in units of 10⁻³ ppm⁻¹) during the year of growth, shown on the upper x axis (r² = 0.49, P = 0.001, n = 18).

Table S3.

Origin, growth year, and D6^S/D6^R values of samples from spinach, beet sugar, E. angustifolium, and S. fuscum

Species/sample ID	Origin or growth location	Growth year*	D6S/D6R ratio± SEM†
Sugar beet	Danmarks Sukkermuseum, Nakskov, Denmark (sugar from “Odense Sukkerfabrik”)	1890	1.031 ± 0.008
Sugar beet/ NM.0209.401a	Nordiska Museet, Stockholm, Sweden (sugar from Tanto sugar factory)	1900–1906	1.025 ± 0.004
Sugar beet/ NM.0261.873	Nordiska Museet, Stockholm, Sweden	1920–1925	1.016 ± 0.003
Sugar beet	Zuckerforschung Tulln, Austria	1940	1.023 ± 0.008
Sugar beet	Karin Rädel, Zuckersammler-Klub, Germany	1950–1960	1.033 ± 0.008
Sugar beet	Karin Rädel, Zuckersammler-Klub, Germany	1950–1960	1.048 ± 0.004
Sugar beet	Karin Rädel, Zuckersammler-Klub, Germany	1950–1960	1.012 ± 0.011
Sugar beet	Haus der Geschichte der Bundesrepublik Deutschland, Bonn, Germany	1955–1962	1.012 ± 0.004
Sugar beet	“Kölner Zucker” brand, Pfeifer & Langen Germany	1970–1980	0.997 ± 0.005
Sugar beet	Zuckerforschung Tulln, Austria	1992	1.035 ± 0.003
Sugar beet	Zuckerforschung Tulln, Austria	1994	1.025 ± 0.006
Sugar beet	Zuckerforschung Tulln, Austria	1996	0.986 ± 0.006
Sugar beet	Zuckerforschung Tulln, Austria	2002	1.010 ± 0.004
Sugar beet	“Dansukker” brand, sugar cube	2011	0.981 ± 0.004
Sugar beet	“Dansukker,” granulated sugar, Sweden	2011	0.960 ± 0.008
Sugar beet	Diamant Zucker, Pfeifer & Langen, Germany	2011	0.995 ± 0.007
Sugar beet	Südzucker, Germany	2011	0.997 ± 0.008
Sugar beet	Zuckerforschung Tulln, Austria	2012	0.993 ± 0.004
Spinach	Stadsliden, Umeå, Sweden; herbarium UME, Umeå	1905	0.961 ± 0.008
Spinach	Orsa, Dalarna, Sweden; herbarium UME, Umeå	1908	0.912 ± 0.004
Spinach	Sparreholm, Södermanland Sweden; herbarium UME, Umeå	1903	0.931 ± 0.010
Spinach	Lidingö, Stockholm, Sweden; Naturhistoriska riksmuseet, Stockholm	1892	0.956 ± 0.010
Spinach	commercial: Findus, grown in Sweden	2011	0.888 ± 0.005
Spinach	commercial: Findus, grown in Sweden	2014	0.875 ± 0.003
Spinach	commercial: ICA, grown in EU	2014	0.918 ± 0.004
Spinach	commercial: Findus, grown in Sweden	2014	0.885 ± 0.003
Spinach	commercial: COOP, grown in EU	2014	0.914 ± 0.004
Spinach	commercial: COOP, grown in EU	2014	0.915 ± 0.003
E. angustifolium	Jukkasjärvi, Sweden; herbarium UME, Umeå	1949	0.968 ± 0.006
E. angustifolium	Åvike, Medelpad, Sweden; herbarium UME, Umeå	1944	0.942 ± 0.009
E. angustifolium	Grangärde, Sweden; herbarium UME, Umeå	1943	0.986 ± 0.004
E. angustifolium	Åby parish, Kalmar län, Sweden; herbarium UME, Umeå	1954	0.991 ± 0.004
E. angustifolium	Umeå, Sweden	2014	0.942 ± 0.003
E. angustifolium	Umeå, Sweden	2014	0.936 ± 0.002
E. angustifolium	Umeå, Sweden	2014	0.916 ± 0.002
E. angustifolium	Umeå, Sweden	2014	0.932 ± 0.002
S. fuscum UME 176929	Dorotea, Västerbotten, Sweden; herbarium UME, Umeå	1921	0.877 ± 0.004
S. fuscum	Östergotland, Sweden, Naturhistoriska riksmuseet, Stockholm	1888	0.875 ± 0.003
S. fuscum	Västergotland, Sweden, Naturhistoriska riksmuseet, Stockholm	1923	0.874 ± 0.002
S. fuscum	nutrient-poor minerogenic mire, Småland, Sweden (56°56’00N; 15°25’25E)	2013	0.836 ± 0.001
S. fuscum	nutrient-poor minerogenic mire, Småland, Sweden (56°56’00N; 15°25’25E)	2013	0.865 ± 0.002
S. fuscum	nutrient-poor minerogenic mire Västerbotten, Sweden (63°52’01N; 20°30’02E)	2014	0.824 ± 0.004
S. fuscum	nutrient-poor minerogenic mire Västerbotten, Sweden (63°52’01N; 20°30’02E)	2014	0.831 ± 0.002
S. fuscum	nutrient-poor minerogenic mire Västerbotten, Sweden (63°52’01N; 20°30’02E)	2014	0.852 ± 0.002

^*When only an interval could be determined, the middle of the interval was assumed to be the growth year. Atmospheric [CO₂] for growth years is from Carbon Dioxide Information Analysis Center (cdiac.ornl.gov/ftp/trends/co2/lawdome.smoothed.yr20).

^†SEM represents measurement precision from five replicate spectra.

The magnitude of CO₂ fertilization is a major uncertainty in models of the global carbon cycle and of crop productivity (1, 11). Assuming constant leaf and environmental properties, oxygenation/carboxylation ratios and net photosynthesis of C₃ plants can be modeled as a function of [CO₂] using Rubisco kinetic parameters (21, 27). For the [CO₂] increase from 295 ppm (∼A.D. 1900) to 395 ppm (∼A.D. 2012), such modeling predicts that the oxygenation/carboxylation ratio declined by ∼25%, which contributed to an increase in net photosynthesis of about 35%. The agreement of the regression slopes obtained for our greenhouse-grown plants and the beet sugar samples indicates that this putative shift in metabolism has actually occurred in field-grown sugar beet between 1890 and 2012.

To further explore the generality of our observations, we compared herbarium and modern samples of an additional crop and two undomesticated C₃ species: the herbaceous crop spinach (S. oleracea L.), the vascular plant rosebay willowherb (or fireweed, Epilobium angustifolium), and a common peat moss species, Sphagnum fuscum. The modern samples of each species show significantly lower D6^S/D6^R ratios, by 0.034–0.041, compared with the older herbarium samples (Fig. 5 and Table S3) that were produced at lower [CO₂]. This reduction is again consistent with a reduction in oxygenation due to the [CO₂] increase during their respective growth periods. These results provide experimental evidence that the CO₂ increase since industrialization has shifted the photorespiration/photosynthesis ratio toward photosynthesis not only in crops grown under fertilized conditions but also in natural vegetation. E. angustifolium and S. fuscum both occur widely across the Northern Hemisphere. Because the underlying mechanism based on Rubisco properties is general for C₃ plants, these observations suggest that a substantial shift in the oxygenation/carboxylation ratio has occurred in terrestrial C₃ plants and it can be quantified using isotopomers.

Fig. 5.

Isotopomer response comparing herbarium and modern samples. D6^S/D6^R ratios of structural carbohydrates of herbarium samples (open bars) and modern samples (black bars). Modern samples grew between 2011 and 2014, herbarium spinach samples grew between 1892 and 1908 ([CO₂] ∼295 ppm), herbarium Epilobium samples grew between 1943 and 1954 ([CO₂] ∼310 ppm), and herbarium Sphagnum samples grew between 1888 and 1923 ([CO₂] ∼295 ppm). Values are averages ± SEM (n = 3–6); *P < 0.02.

Discussion

Origin of the Isotopomer Signal.

We observed a dependence of the D6^S/D6^R ratio on 1/[CO₂] in controlled experiments and in archival plant material. In the following, we explain this dependence by differences in isotope fractionation in the metabolic pathways originating from carboxylation and oxygenation, respectively, at Rubisco. The hydrogen atoms of the C6H₂ group of glucose can be traced back to their origins after carboxylation or oxygenation of ribulose bisphosphate (RuBP) in C₃ metabolism, as shown in Fig. 1. Although photorespiration leads to C loss, 3-phosphoglycerate (3-PGA) is formed after both carboxylation and oxygenation, but with differing stoichiometries and along differing paths. Numerous observations have shown that synthesis of a compound via different pathways induces different isotopomer distributions (18, 19, 28–32). No exception to this rule has been observed, and it agrees with the theory of isotope effects. Therefore, it is rational to assert that the PGA molecules formed from C3−C5 of RuBP, from C1−C2 of RuBP, and via the photorespiration pathway have differing D isotopomer ratios in their C3H₂ groups, as indicated by color coding in Fig. 1. In the photorespiration cycle, the C3H₂ group of 3-PGA originates from serine, which is formed by serine hydroxymethyl transferase. Commercial serine is produced by the same enzyme (33) and shows a strongly uneven isotopomer distribution in its C3H₂ group (Fig. S2). This supports the proposition that 3-PGA originating from photorespiration has a distinct isotopomer composition, and that serine hydroxymethyl transferase contributes to creation of the oxygenation/carboxylation signal. The ratio of metabolic fluxes originating from carboxylation and from oxygenation determines the contributions of the alternative PGA sources to the PGA pool, and the D isotopomer distribution of the PGA pool is a flux-weighted average of the PGA sources. During the conversion of PGA into glucose, no reaction occurs at the C3H₂ group; thus, there is no scrambling of hydrogen isotopes (34), and the D isotopomer ratio of the C6H₂ group of the resulting glucose will be identical to that of the PGA pool. Hence the weighted isotopomer distribution is transmitted from the 3-PGA pool to glucose, so that the D6^S/D6^R ratio of glucose becomes dependent on the oxygenation/carboxylation ratio.

Fig. S2.

Deuterium NMR spectrum of O-Phospho-l-serine (Santa Cruz Biotechnolgy), indicating D isotopomer abundances of serine formed by serine hydroxymethyl transferase during fermentation (33). The integrals of the signals, obtained by deconvolution, are proportional to the abundances of the Cα-D isotopomer and the two D isotopomers at Cβ, labeled Cβ-D¹ and Cβ-D². Although the stereochemical assignments of Cβ-D¹ and Cβ-D² are unknown, their strongly differing integrals show that serine hydroxymethyl transferase creates a strongly uneven D isotopomer ratio at C3 of serine, which, in the photorespiration pathway, is converted into an uneven D isotopomer ratio at C3 of 3-phosphoglycerate.

This dependence is the first example, to our knowledge, of an isotopomer shift induced by an environmental driver that can be mechanistically interpreted in terms of a metabolic response to this driver. We conclude that isotopomers carry metabolic signals that can be identified using laboratory experiments, and may be retrieved from archives of plant material.

Methodological Advantages of Isotopomer Signals.

An important advantage of isotopomer ratios is that they can be interpreted without any reference to the δD of source water. This is possible because the δD of a plant’s source and leaf water affect the abundance of all isotopomers to the same degree, so isotopomer ratios are independent of δD (35). Therefore, isotopomer ratios are exclusively determined by biochemical isotope effects. In contrast, interpretations of δD of plant archives are often complicated (36) because the required data on δD of source water and on evaporative enrichment of leaf water are not available. Because hydrogen isotope signatures are stable for >10⁴ y (37), isotopomers may yield information on physiological changes on long time scales, including interglacial cycles. The complete D NMR spectrum of glucose (Fig. 2A) is fully described by seven isotopomer abundances, or by six isotopomer ratios and the overall amplitude of the spectrum, which is given by δD of the entire molecule. Thus, six out of seven degrees of freedom that describe the spectrum are isotopomer ratios, and hence the complete isotopomer pattern contains much more information than δD of the entire molecule.

Isotopomer information complements manipulation studies such as FACE experiments, in that information on metabolism can be retrieved on long time scales. Here we show that the primary metabolic response to elevated CO₂—suppression of photorespiration—has, in the species studied here, persisted during the 20th century. In contrast to the persistent suppression of photorespiration observed here, CO₂-driven increases in plant growth observed in FACE experiments often decline after a few years of CO₂ enrichment (2). Thus, combining FACE and isotopomer data may allow assessment of the role of factors besides photosynthesis that limit growth on the ecosystem level (38). Further studies will have to address whether long-term suppression of photorespiration has occurred for C3 plants in general, how photorespiration will develop under scenarios for future CO₂ levels and climate change, and how the global photorespiration flux will influence efforts to stabilize CO₂ levels.

Properties of Rubisco and the photorespiration cycle are current targets for crop improvement (20, 39). The D6^S/D6^R ratio may be used to detect differences in CO₂ responses between crop species, or cultivars, and isotopomers in general to monitor metabolic effects of crop engineering. Isotopomer variation exists for many other metabolites and isotopes, notably ¹³C (40, 41). Therefore, time series of δ¹³C of entire molecules may also be affected by isotopomer trends, and ¹³C isotopomer analysis of plant archives can yield much more information than is available from δ¹³C.

Implications for Biogeochemistry and Plant Ecophysiology.

Among the little that is known about responses of plants to increasing CO₂ over recent centuries, δ¹³C time series of annual plant material or tree rings are arguably the strongest evidence. From δ¹³C, increases in assimilation have been deduced, and the reduction in the oxygenation/carboxylation ratio is a large part of this. Our isotopomer results provide the first empirical data on the magnitude of the shift in this central biogeochemical flux ratio.

Forests make a big contribution to the global carbon cycle, and tree rings are one of the most prominent paleoarchives. Therefore, it is important to note that isotopomer information may be retrieved from tree rings, because the glucose derivative can be produced from whole wood or cellulose (35). Furthermore, we have previously shown that metabolic signatures produced on the leaf level are preserved in the glucose moieties obtained from tree rings (42). This indicates that it will be possible to derive long-term physiological information from tree rings. Because glucose can sometimes be produced from fossil leaves, paleoatmospheric reconstruction of atmospheric CO₂ may also be possible.

The peat mosses belonging to the genus Sphagnum are the main constituents of peat of high-latitude mires (43, 44). This peat has accumulated during the Holocene and currently represents ∼25% of the global soil carbon pool. The amount of carbon stored in peat has caused a lowering of the contemporary atmospheric CO₂ concentration by 35 ppm (45). Due to the extraordinarily large carbon store in these peatlands, their feedback to changing climate and rising CO₂ constitutes a major scientific and societal concern with respect to predicting future biosphere−atmosphere CO₂ exchange and storage (44, 46). Recent studies have revealed net primary production as the main driver for peatland carbon accumulation (46, 47), which is also supported by the observation that incoming photosynthetic active radiation during the growing season is the main driver for geographic variation in Sphagnum moss productivity (48). The studied species S. fuscum grows in hummocks and is thus directly exposed to atmospheric CO₂. Such hummock-forming sphagnum species dominate peat formation at high-latitude mires (48). The trend in the isotopomer ratio observed in S. fuscum constitutes the first evidence of increased photosynthesis among peat-forming Sphagnum mosses in response to the century-scale CO₂ increase under natural conditions. Thus, by applying isotopomers to a geographic and species range of peat-forming species, it may be possible to assess general trends in the physiology of peat-forming species. Trends in net photosynthesis may then be scaled up to the global scale, to obtain a mechanistic understanding of the carbon balance of peatlands.

Reducing the photorespiration/photosynthesis ratio increases the light use efficiency of photosynthesis. Some measure of light use efficiency is frequently used to parameterize photosynthetic algorithms in models of ecosystem response to climate change (49–51). These parameterizations have been based on models of Rubisco substrate specificity (27) scaled up to the ecosystem level or, more often, empirical fitting from remote sensing (52), eddy covariance (53), or biomass (54) data that was scaled downward and thus lacked a mechanistic description of this key metabolic flux ratio. The isotopomer approach thus provides a mechanistically based means of parameterizing light use efficiency models, allowing us to improve our predictions. These improvements will appear in both paleoarchive data, looking backward, and the use of experimental CO₂ enrichment experiments, to look forward.

Information on effects of long-term increases in [CO₂] on the metabolism of crops and natural vegetation is essential for understanding underlying mechanisms and robustly modeling C exchange fluxes between terrestrial vegetation and the atmosphere. The presented methodology may be used to detect variation in the CO₂ responses between species or crop lines, and changes in isotopomer patterns may be used to detect long-term acclimation or adaptation. Retrieving such information will enable the time frames of plant physiological studies to be extended to centuries, thereby bridging the gap between manipulation experiments and paleoenvironmental studies.

Methods

CO₂ Manipulation Experiments.

In greenhouse and growth chamber experiments, Helianthus annuus, Z. mays, S. oleracea, and P. vulgaris plants were exposed to different atmospheric [CO₂] levels, and S. oleracea was exposed to reduced [O₂]. Gas exchange parameters of H. annuus leaves were determined using a two-channel fast response measurement system (Fast-Est), described in detail in ref. 55. Details of growth and measurement conditions are provided in Supporting Information.

Historic Plant Material.

Eighteen archived sugar beet sucrose samples that grew between 1890 and 2012 were obtained from sources listed in Table S3 (specified growth years of these samples are as stated by the contributors). Current samples (2011−2012) are dated to the previous growing season. Herbarium samples of S. oleracea, E. angustifolium, and S. fuscum were obtained from the herbarium at Umeå University and from Naturhistoriska Riksmuseet. Modern samples for comparison were either commercial frozen products (spinach) or were collected in Sweden. Exact origins, growth locations, and growth years are provided in Table S3. Structural carbohydrates were analyzed.

Sample Preparation for NMR Isotopomer Measurements.

For deuterium NMR measurements, pure samples of a glucose derivative were prepared from all samples, as follows. Soluble sugars were extracted from leaf material according to previously used methods (35, 56), starting with 10 g of fresh material. Starch from chloroplasts was extracted from the pellet resulting from extraction of soluble sugars, according to ref. 57, with the following modifications: 200 mL of 50 mM KOH was added, and the solution was kept at 90 °C for 2 h. After cooling on ice, the pH was adjusted to 4.7 with 1 M acetic acid, and 42 units of amyloglucosidase (Roche) were added. After incubation at 37 °C overnight, samples were centrifuged for 20 min at 4,000 × g, and the supernatant was collected. Starch from bean fruits was hydrolyzed to glucose by amyloglucosidase treatment. Leaf structural carbohydrates were obtained by removing soluble sugars and starch from leaves by extraction and amyloglucosidase hydrolysis, respectively. Structural carbohydrates were hydrolyzed to glucose according to Betson et al. (35). Sugar samples prepared from plants were dried in vacuum and were converted into a pure glucose derivative suitable for deuterium NMR analysis (35).

Isotopomer Quantification.

Deuterium NMR measurements and subsequent data analysis followed Schleucher et al. (19), modified as previously described (35). For NMR measurements, we used a DRX600 spectrometer (Bruker) equipped with a 5-mm broadband observe probe and a ¹⁹F lock device, or an AVANCE III 850 spectrometer (Bruker) equipped with a cryogenic probe optimized for deuterium detection and equipped with a ¹⁹F lock. Deuterium NMR spectra were integrated by deconvolution with a Lorentzian line shape fit, using TopSpin software (version 3.1; Bruker). The D6^S/D6^R isotopomer ratio was determined as the ratio of the integrals of the D6^S and D6^R signals.

Statistical Analysis.

For each sample, five or six replicate spectra were recorded, the average isotopomer ratio among the spectra was used, and its SE was calculated to estimate measurement precision. For biological replicate samples, SEs were calculated. The significance of between-treatment differences in isotopomer ratios was tested using ANOVA, and linear regression analysis was applied to test relationships between the ratios and [CO₂] (in both cases using Excel).

Description of CO₂ Manipulation Experiments

In the following text, “groups A−E” refer to independent laboratory experiments and “samples a−i” refer to samples of materials subjected to various treatments (as described in the following text) applied in the experiments. Isotopomer results are summarized in Tables S2 and S3.

Group A.

Sunflower (H. annuus L. cv Zebulon) plants were grown in 1.4-L pots in a greenhouse in Umeå, Sweden, during fall 2000. The photoperiod was set at 16 h with a light intensity of 300–400 µmol photons m⁻²⋅s⁻¹. The [CO₂], as measured by an LCA2 infrared gas analyzer (ADC), was slightly higher than atmospheric (around 450 ppm). The plants were watered daily and fertilized, twice a week, with a commercial nutrient solution (RIKA-S, NPK 7–1-B; Weibulls). After 7–8 wk in the greenhouse, plants were transferred to a growth chamber in groups of eight, where they were exposed to a range of CO₂ concentrations (180 ppm, 280 ppm, 450 ppm, 700 ppm, or 1,500 ppm). CO₂ concentrations of 450 ppm and higher were created by introducing pure CO₂ via a valve system into the growth chamber, while concentrations of 180 ppm and 280 ppm were attained by mixing flows of air and CO₂-free air. Temperature and relative humidity in the growth chamber were 22/18 °C and 60/70% during day/night, respectively. During the first day in the growth chamber, the plants were kept in the dark at the same [CO₂] as in the greenhouse (450 ppm). During the following 2 d, the plants were exposed to the aforementioned photoperiod and the selected [CO₂]. During the second day of CO₂ treatment, gas exchange measurements were performed and leaf samples were collected. Leaf material was stored at −20 °C until leaf starch was extracted (Table S2, samples a).

Group B.

Sunflower plants were grown, from seeds, during summer 2003 in Freising, Germany, in two growth chambers with [CO₂] set at 200 ppm and 1,000 ppm. Further details about the experiment and growth conditions can be found in ref. 58. Glucose moieties in structural carbohydrates of freeze-dried leaves were isolated and prepared as described in Methods (Table S2, samples b).

Group C.

Maize (Z. mays) plants were grown in 5-L pots in a greenhouse in Umeå, Sweden, during summer 2000. The growth conditions and [CO₂] treatments were identical to those for sunflowers (group A) except that the maize plants were exposed only to 450 ppm and 1,200 ppm [CO₂], and leaf samples were collected during the second day of CO₂ treatment. Leaf material was stored at −20 °C until its soluble sugars and starch contents were extracted (Table S2, samples c).

Group D.

Spinach (S. oleracea cv. Giant noble) plants were grown from seeds in a growth chamber providing the following conditions: 11-h photoperiods (600 µmol photons m⁻²⋅s⁻¹), 22/18 °C light/dark temperatures and 60% relative humidity. Two [CO₂] concentration treatments (ambient ≈ 360 ppm and 700 ppm) were applied from germination. In a third treatment, five spinach plants (∼6 wk old, grown at 360 ppm [CO₂]) were exposed for 2 d to air with the oxygen content reduced to 12%. Plants subjected to each of the three treatments were harvested 6 wk after germination and stored at −20 °C until leaf soluble sugars and leaf starch were extracted (Table S2, samples d−f).

Group E.

Bean (P. vulgaris L. cv. Linden) plants were grown from seeds under the same conditions as for spinach (group D). Leaves and fruits of mature plants were harvested and stored at −20 °C until leaf soluble sugars and leaf starch were extracted (Table S2, samples g−i).

Leaf Gas Exchange Measurements

Gas exchange parameters of sunflowers’ leaves (group A) were determined using a two-channel fast-response measurement system (Fast-Est), described in detail in ref. 55. For each measurement, part of the leaf (8.04 cm²) was enclosed in a leaf chamber, its upper part was attached with starch gel to a thermostated glass window maintained at 22 °C [to calculate the dissolved O₂ concentration and CO₂/O₂ specificity constant of Rubisco (59)] and it was exposed to an airflow of 0.5 mmol⋅s⁻¹. Light was provided by a Schott KL 1500 source (Heinz Walz) and measured by a Li-Cor Quantum Sensor. Gas exchange was measured only through the lower side of the leaf, although sunflower leaves have stomata on both sides. Gas exchange measurements were performed at 300 µmol photons m⁻²⋅s⁻¹ and the [CO₂] to which plants had been exposed for 2 d in the growth chamber. From the gas exchange data, the intercellular CO₂ concentration (C_i) was obtained. In the afternoon of the day of gas exchange measurements, leaves were harvested and stored at −20 °C, until extraction of their starch contents (Table S2, sample a).

Functional Dependence of the D6^S/D6^R Ratio on the Oxygenation to Carboxylation Ratio

The 3-PGA molecules derived from oxygenation and carboxylation differ in their isotopomer ratios (Fig. 1). This leads to a functional dependence of the D6^S/D6^R isotopomer ratio on the oxygenation/carboxylation flux ratio. The ratio of oxygenation flux (V_o) to carboxylation flux (V_c) is ϕ = V_o/V_c = 1/K_sp [O₂]/C_i, where K_sp is the specificity constant of Rubisco (21, 60), and [O₂] is the dissolved oxygen concentration. Two 3-PGA molecules are formed upon RuBP carboxylation. Upon RuBP oxygenation, one 3-PGA is formed directly, and up to 0.5 additional 3-PGA are formed via the photorespiration cycle (Fig. 1). We assign the value m, 1 ≤ m ≤ 1.5, as the total 3-PGA production due to oxygenation. If the oxygenation/carboxylation ratio is ϕ, then, in total, (2 + mϕ) 3-PGA are formed. Of these, the fractional contribution due to oxygenation is mϕ/(2 + mϕ), and the fractional contribution due to carboxylation is 2/(2 + mϕ). For 3-PGA molecules formed via oxygenation and carboxylation, we define R_o and R_c, respectively, as variables describing their D isotopomer ratios at the C3H₂ group. The isotopomer ratio of the PGA pool used to synthesize glucose will be an average of R_o and R_c, weighted by the respective fractional contributions,

$$R=\left(m\varphi R_o+2R_c\right)/\left(2+m\varphi \right).$$ [S1]

Experimental values for C_i and ϕ from gas exchange measurements and associated D6^S/D6^R ratios are given in Table S1.

From the intercellular O₂ and CO₂ concentrations and the specificity constant of Rubisco, the ratio of oxygenation to carboxylation, ϕ, was calculated as ϕ = V_o/V_c = 1/K_sp × [O₂]/C_i. In Fig. S1, the D6^S/D6^R ratios are plotted as a function of ϕ. Photorespiratory metabolites such as Gly and Ser may be removed from the PCO pathway, but only in small amounts (61); therefore, the number of PGA molecules formed following oxygenation (m) will be close to 1.5. Assuming m = 1.5, the data can be fitted to Eq. S1 (best-fit parameters R_c = 0.83, R_o = 2.06). The curvature of the fitted line is small; therefore, the fit to Eq. S1 is well approximated by a linear function. Thus, the D6^S/D6^R ratio can be described by a linear function of ϕ. Because ϕ is proportional to 1/C_i, and typical variation of the C_i/C_a (C_a, atmospheric CO₂ concentration) ratio is small, the D6^S/D6^R ratio is also a linear function of 1/C_a.