Real-time carbon allocation into biogenic volatile organic compounds (BVOCs) and respiratory carbon dioxide (CO2) traced by PTR-TOF-MS, 13CO2 laser spectroscopy and 13C-pyruvate labelling

Our understanding of biogenic volatile organic compound (BVOC) emissions improved substantially during the last years. Nevertheless, there are still large uncertainties of processes controlling plant carbon investment into BVOCs, of some biosynthetic pathways and their linkage to CO2 decarboxylation at central metabolic branching points. To shed more light on carbon partitioning during BVOC biosynthesis, we used an innovative approach combining δ13CO2 laser spectroscopy, high-sensitivity proton-transfer-reaction time-of-flight mass spectrometry and a multiple branch enclosure system in combination with position-specific 13C-metabolite labelling. Feeding experiments with position-specific 13C-labelled pyruvate, a central metabolite of BVOC synthesis, enabled online detection of carbon partitioning into 13C-BVOCs and respiratory 13CO2. Measurements of trace gas emissions of the Mediterranean shrub Halimium halimifolium revealed a broad range of emitted BVOCs. In general, [2-13C]-PYR was rapidly incorporated into emitted acetic acid, methyl acetate, toluene, cresol, trimethylbenzene, ethylphenol, monoterpenes and sesquiterpenes, indicating de novo BVOC biosynthesis of these compounds. In contrast, [1-13C]-pyruvate labelling substantially increased 13CO2 emissions in the light indicating C1-decarboxylation. Similar labelling patterns of methyl acetate and acetic acid suggested tightly connected biosynthetic pathways and, furthermore, there were hints of possible biosynthesis of benzenoids via the MEP-pathway. Overall, substantial CO2 emission from metabolic branching points during de novo BVOC biosynthesis indicated that decarboxylation of [1-13C]-pyruvate, as a non-mitochondrial source of CO2, seems to contribute considerably to daytime CO2 release from leaves. Our approach, combining synchronised BVOC and CO2 measurements in combination with position-specific labelling opens the door for real-time analysis tracing metabolic pathways and carbon turnover under different environmental conditions, which may enhance our understanding of regulatory mechanisms in plant carbon metabolism and BVOC biosynthesis.


Introduction
Our understanding on emissions of biogenic volatile organic compounds (BVOCs), including thousands of different compounds from different chemical classes, has improved substantially during the last decades, partially due to new developments in measuring techniques. Therefore, the biosynthetic pathways of a great number of volatile compounds such as isoprenoids, wound induced compounds or several oxygenated compounds (e.g. acetaldehyde, ethanol, methanol) are well-but still not fully-understood. In contrast, very little is known on the underlying biochemical production processes of many other volatiles (e.g. benzenoids, methyl acetate), the regulation of plant carbon investment into such BVOCs, as well as environmental factors controlling biosynthesis and emission of these compounds. Moreover, unknown processes and key determinants controlling BVOC release from terrestrial vegetation account for large uncertainties in our understanding of global carbon cycling [1,2]. Although 13 C-labelling experiments already improved our knowledge on carbon allocation into BVOCs [3][4][5], there are still open questions regarding carbon partitioning into plant primary and secondary (e.g. BVOC emissions) metabolism with special regard to the biochemical link between both.
Noteworthy, the complex biosynthetic pathways of many volatiles produced in plants are tightly linked to CO 2 exchange between the plant and the atmosphere. For example, biosynthesis of monoterpenes and isoprene, the latter being the most important volatile emitted by vegetation, takes place in the chloroplasts and uses Calvin cycle intermediates (pyruvate and glyceraldehyde-3-phosphate) as direct precursors. Hence, production of these compounds is tightly linked to photosynthetic CO 2 fixation [5,6]. On the other hand, use of the key metabolite pyruvate as a precursor often involves a decarboxylation step during biosynthesis, in which the carboxyl group of pyruvate is released as CO 2 [7][8][9]. Thus, many pathways producing BVOCs constitute a source of CO 2 even in the light, besides the main CO 2 releasing process in plants, mitochondrial dark respiration. Hence, there are at least four main processes determining the CO 2 exchange in plants, i.e. CO 2 fixation during photosynthesis, photorespiration, CO 2 generating processes in metabolic pathways and mitochondrial respiration mainly taking place in the tricarboxylic acid (TCA)-cycle. From an analytical point of view, mitochondrial respiration can easily be determined in the absence of light, however, it is at least partially inhibited in the light [7,[10][11][12][13][14]. In contrast, analysing CO 2 evolving processes during the day is tricky due to co-occurring photosynthesis. Moreover, in the light multiple sources of CO 2 production exist, as during production of BVOCs whose link to CO 2 production has not been well established so far [15]. Tcherkez et al. (2017) point out that a comprehensive analysis of metabolic fluxes, including partitioning at branching points, is still lacking and that the specific origin of carbon atoms fuelling day-respiration is still uncertain.
To unravel metabolic networks, the use of central metabolites, which are position-specific labelled with stable isotopes, reflects a very effective tool in order to follow the metabolic fate of the tracer into different anabolic and catabolic pathways [15,16]. As mentioned above, pyruvate is a central metabolite tightly linking plant primary and secondary metabolism. In the present study, we therefore fed plants with position-specific labelled pyruvate (pyruvate 13 Clabelled at the C1 or C2-carbon position). This approach enables tracing the fate of single 13 Catoms of pyruvate through metabolic branching points into volatile plant metabolites and CO 2 and allows to investigate their metabolic pathways.
Preliminary experiments indicated the vast potential of combining proton-transfer-reaction mass spectrometry (PTR-MS) with molecular tracer approaches for better understanding carbon partitioning in plants [17][18][19][20]. Because incorporation of tracers into individual volatiles happens at short temporal scales and dynamics are diverse for different compounds, quantification of these processes needs precise and synchronized analytical systems. Therefore, we used highly sensitive proton transfer reaction time-of-flight mass spectrometry (PTR-TOF-MS) with stable isotope CO 2 infrared laser spectroscopy (IRIS) and combined this technological infrastructure with position-specific molecular labelling. We aimed at analyzing the trace gas emissions of a Mediterranean shrub Halimium halimifolium under standardized environmental conditions. On the one hand, Mediterranean shrubs from the Cistaceae family possess a very active secondary metabolism resulting in the biosynthesis of a plethora of plant secondary metabolites including many BVOC [21,22]. On the other hand, H. halimifolium exhibits large variations in day respired δ 13 CO 2 [8,23,24], suggesting a, so far speculative, link to plant secondary metabolism [15,16,25,26].
In particular, we aim to identify and quantify the range of BVOC emitted from H. halimifolium shoots, link BVOC emission with day released CO 2 and explore the underlying pathway. To this end, we take advantage of our system, which enables to trace the fate of 13 C derived from either [1-13C]-pyruvate (PYR) or [2-13C]-pyruvate into BVOCs and CO 2 in real-time. We anticipated that traceability of 13 C in different volatile compounds may help elucidating biosynthetic pathways particularly of less studied compounds that are still unclear or not understood at all.

Plant material and labelling experiments
Three-year-old, 50-70 cm high Halimium halimifolium L. plants, a Mediterranean drought semi-deciduous shrub, were grown in 3 L plastic pots, filled with two parts of potting soil and one part of sand. Plants were fertilized weekly with one-quarter diluted, modified Hoagland´s fertilizer solution [27]. For the experiments, H. halimifolium plants were transferred from a greenhouse into the walk-in climate chambers, where they were kept at day/night cycles of 12:12 h, 25:25˚C, 60:60% relative humidity and 500 μmol m -2 s -1 PPFD during light period. Acclimation to these conditions lasted for one week; plants were supplied with tap water according to their demand and were fertilized as given above.
For labelling experiments, branches of H. halimifolium were placed into self-constructed enclosures (see below) the evening before the measurements maintaining the natural orientation of the branches by adjusting the enclosure position. Ninety minutes before 13 C-labelling, the branches were cut carefully at the petiole without changing the branch position in the enclosure; they were immediately recut under water and the cut end of the petiole was placed in deionized water. Thereafter (after 90 min) deionized water was replaced by a 10 mM pyruvate solution, i.e. the solution was fed to the shoot via the transpiration stream. Control branch samples were left in deionized water for the entire duration of the experiment. Pyruvate (Cambridge Isotope Laboratories, Andover, MA, USA) was 99% 13 C-labelled at either the C1 or C2 carbon position dissolved in deionized water. Throughout the entire experiment, continuous online measurements of the isotopic composition of BVOCs and CO 2 as well as photosynthetic gas exchange fluxes were conducted. Two branches were alternately investigated to increase the number of replicates in a given period of time. A picture of an attached H. halimifolium branch inside the enclosure system can be found as supplementary information (S1 Fig). All branches were in a good phenological condition and were not flowering.
( Fig 1). The analytical section combined, (i) a proton-transfer-reaction time-of-flight mass spectrometer (PTR-TOF-MS, Ionicon Analytic, Innsbruck, Austria), (ii) an isotope ratio infrared spectrometer (IRIS, Thermo Fisher Scientific, Bremen, Germany), and (iii) an infrared gas analyzer (LI-7000 CO 2 /H 2 O Analyzer; LI-COR, Lincoln, NE, USA). To ensure controlled environmental conditions, the plant enclosures and measuring devices were placed within two walk-in climate chambers (ThermoTec, Weilburg, Germany), entirely light, temperature and humidity controlled. Standardized conditions allowed for monitoring of trace gas emissions in response to changes of single environmental parameters such as temperature or light. The climate chambers were illuminated with LED panels with a maximum photon flux density (PFD) of 1600 μmol m -2 s -1 , which generated high and uniform illumination without overheating the enclosures.
We constructed a custom-made zero air generator for continuous supply of hydrocarbonfree, humidified and CO 2 controlled zero-air at a flow rate of up to 20 l min -1 . It combined a gas pump, a cold trap, in-line vessels filled with soda lime and activated charcoal, a custom made humidifier and a gas cylinder containing synthetic air mixed with 5 Vol.% CO 2 (Messer Austria, Gumpoldskirchen, Austria) for CO 2 enrichment. In addition, the system contained a catalyst module (HPZA-30000, Parker Hannifin Corporation, Lancaster, NY USA) filled with platinum catalyst pellets (Infiltec, Speyer am Rhein, Germany), which efficiently removes nearly all remaining hydrocarbons in the air stream.
The enclosure system consisted seven 600 ml borosilicate glass enclosures (Kummer, Freiburg, Germany), six of them were used as plant enclosures and one was kept empty as a reference. Branches were inserted into the bottomless enclosures and tightly closed with PTFE-foil  at the plant stem. The inlet flow was controlled by mass flow controllers (Omega Engineering, Stamford, CT, USA) and kept constant at 600 ml min -1 resulting in a calculated residence time in the enclosures of 60 s. The outlet flow was split to channel enclosure air to the different analytical devices. The response-time of air leaving the enclosure and reaching the PTR-TOF-MS and IRIS/IRGA was less than 4 and 10 seconds, respectively. In operation mode, the analytical instruments were supplied with an air flow of~400 ml min -1 resulting in a slight overpressure in the enclosures. To assure chemical inertness, all cuvette system parts were made of glass or PFA, and outlet lines were continuously isolated and heated up to 60˚C to prevent BVOCs adsorption at the tube walls or water condensation. Inside the PTR-TOF-MS mainly polyetheretherketon (PEEK) and metal is used.
The enclosure air temperature increased by 2˚C (from 25±0.5˚C to 27±0.5˚C) during light period, which is an inhered property of enclosure systems and was taken into account in data analysis. A moderate light reduction of 15-20% was compensated by adjusting the height of the enclosures.
The zero air generator and enclosure system were tested by supplying the empty enclosure system with 1000 ml min -1 of purified air. All parameters remained stable during one hour of measurements (CO 2 400±1 ppm; H 2 O 15000±500 ppm; e.g. acetone 0.2±0.1 ppb; δ 13 CO 2 −4 ±0.5 ‰), there were no substantial differences between single enclosures, nor did the enclosure system have any significant influence on the zero air itself. Compared to climate chamber or even bottled hydrocarbon free synthetic air (Messer Industriegase, Bad Soden, Germany), BVOCs were markedly lower after passing the zero air generator, indicating that the instrument efficiently removed BVOCs from ambient air (S2 Fig)

BVOC detection by PTR-TOF-MS and TD-GC-MS
Online BVOC measurements were conducted with a 4000ultra PTR-TOF-MS (Ionicon Analytic, Innsbruck, Austria). This PTR-TOF-MS version has a built-in internal mass calibration standard (diiodobenzene, m/z: 330.848 and fragments on 203.943), as well as an ion funnel at the end of the drift tube, which focuses the ion beam to the detector allowing BVOCs with higher molecular masses to be analysed more efficiently. We used a multi-position valve controlled by the PTR-TOF-MS software for switching between the enclosures. All sample lines connected to the multi-position valve were continuously flushed to avoid dead end influences.
The PTR-TOF-MS was operated at 2.7 mbar drift pressure, 600 V drift voltage, at an E/N of 120 Td, and drift tube heated to 80ºC. Water impurities were kept on average below 4%, m/z 29.99 was kept on average below 0.2% and oxygen impurities were kept below 6% relative to the primary ions. The mass resolution was 2000 ± 500 m/Δm depending on the compound.
PTR-TOF-MS data post-processing consisted of (i) correction for non-extending and extending dead times as well as the correction for Poisson statistics [28] and iterative residual analysis and cumulative peak fitting [29] using the PTR-TOF Data Analyzer software version 4.48; (ii) normalization of the data to primary ions and water; (iii) background subtraction; (iv) application of calibration factors and (v) synchronization of PTR-TOF-MS data with the IRIS and IRGA data.
We measured non-13 C-labelled BVOCs and their isotopologues containing one 13 C-atom (denoted as 13 C-BVOCs in the following), which were detected at m/z +1. The PTR-TOF-MS technique is a mass-selective technique, thus isomers could be interfering in the measured masses. Therefore, where possible, thermodesorption-gas chromatography mass-spectrometry (TD-GC-MS) was used to verify the results of PTR-TOF-MS. Polar compounds and those with molecular mass <80 amu were not analysed by TD-GC-MS and we assigned these masses to specific compounds according to literature. These compounds were either previously reported to be emitted by vegetation and/or based on mass accuracy of previously reported PTR-TOF-MS measurements. An overview of possible compounds can be found as supporting information (S1 Table). Moreover, specific compounds such as methyl acetate or kaurene emissions by our H. halimifolium plants were validated in previous studies by TD-GC-MS measurements and calibration standards [22,30].
Calibration of the assigned BVOCs was done either with a multicomponent calibration gas standard (1000 ppb ± 5%, Ionicon Analytic, Innsbruck, Austria), liquid standards or via transmission calculations. For selected compounds (gas and liquid water based calibrations) humidity dependent calibrations were done. Liquid calibrations with water-based solutions were done for acetic acid and methyl acetate (Sigma-Aldrich, Taufkirchen, Germany). For sesquiterpenes, trimethylbenzene, 2-hexenal and cis-3-hexenol (Sigma-Aldrich, Taufkirchen, Germany) hexanal-based solutions were used. All calibrations were performed using the Liquid Calibration Unity (LCU, Ionicon Analytic, Innsbruck, Austria). Information on transmission calculations can be found elsewhere [31].
Furthermore, potential losses of some BVOCs in the system were tested by injecting calibration gas via a PFA-tube into the enclosures under normal operation mode (~600 ml min -1 inlet flow,~400 ml min -1 sample flow). For this purpose, we used compound concentrations of 20 ppb, which was within the magnitude of most of the measured compounds. Comparison of calibration gas measurements (normalized counts per second (ncps)) with and without enclosure system influence revealed that losses for toluene, isoprene, α-pinene, acetone, methanol, acetaldehyde and crotonaldehyde were negligible (S3 Fig). Negligible losses can be assured only for these compounds, but given that also low-volatility compounds such as the volatile diterpene (C 20 ) kaurene could be detected (calibrated and cross-validated with GC-MS measurements) [22], makes us confident that strong influences of the measurement system on the considered BVOCs are unlikely. Furthermore, if there were wall losses, the actual emission rates would be even higher than reported. 13 CO 2 fluxes [nmol min -1 m -2 ] were quantified based on the differences in 13 CO 2 isotopic composition and their concentration between empty and plant-containing enclosures as analyzed by a Delta Ray Isotope Ratio Infrared Spectrometer (IRIS, Thermo Fisher Scientific, Bremen, Germany). IRIS is an optical analyzer for continuous, online measurements of δ 13 C-CO 2 , δ 18 O-CO 2 and CO 2 mixing ratios in sample gas. The analyzer operates at a sample flow rate of 80 ml min -1 with a temporal resolution down to one second and a precision of 0.15 ‰. Net fluxes of 13 CO 2 (e 13CO2 ) were calculated per projected leaf area (s) as:

CO 2 detection by IRIS and IRGA
where u in is the molar flow of incoming air calculated as: with V being the gas volume, t being the time, p being the gas pressure, R being ideal gas constant and T being the temperature, and 13 c out / 13 c in is the molar fraction of 13 CO 2 at the inlet and outlet of the enclosure calculated as: where c total is the mole fraction of total CO 2 in the enclosure and 13 C/ 12 C VPDB is the isotope ratio of the international standard Vienna Pee Dee Belemnite.
The IRIS is equipped with reference gases of different isotopic compositions (δ 13 C VPDB of -9.7±0.3‰ /-27.8±0.3‰ and δ 18 O VPDB-CO2 of -27.2±0.3‰ /-17.5±0.3‰) (Thermo Fisher Scientific, Bremen, Germany) and CO 2 concentrations (CO 2 mixing ratio of 408 and 1008 ppm) (Messer, Bad Soden, Germany). The isotope reference of CO 2 is automatically diluted with synthetic air carrier gas (Messer, Bad Soden, Germany) to the current CO 2 concentration of the sample gas. Samples were calibrated for isotopic composition throughout the measurements whenever any significant change in the CO 2 mixing ratio in the sample gas occurred, e.g. during the change between plant and blank cuvette. Additionally, automatic calibration for concentration dependency of the analyzer was conducted every night. To cross-validate IRIS measurements, especially temporal dynamics in isotope signature of leaf dark-respired CO 2 , we used the in-tube incubation technique as described elsewhere [32].
Photosynthetic gas exchange parameters at branch level were quantified using a differential infrared gas analyzer (LI-7000 CO 2 /H 2 O Analyzer; LI-COR, Lincoln, NE, USA), which analyzes the CO 2 and H 2 O differences between empty and plant containing enclosures. Transpiration rate (E) [mmol m -2 s -1 ], assimilation rate (A) [μmol m -2 s -1 ], and stomatal conductance to water vapor (G H2O ) [mmol m -2 s -1 ] were calculated according to Caemmerer and Farquhar [33]. The IRGA was calibrated at least once per week with synthetic air (0 and 408 ppm; (Messer, Bad Soden, Germany)) for CO 2 and just before the experiment for H 2 O by a manufacture calibrated GFS3000 (Walz GmbH, Effeltrich, Germany).

Incorporation of 13 C label into BVOCs and respiratory CO 2
The PTR-TOF-MS revealed a broad spectrum of over 60 different masses emitted by H. halimifolium branches at a considerable amount under controlled conditions (S2 Table); a selection of the most abundant compounds was identified and listed in Table 1 and a characteristic TOF spectrum is shown as supporting information (S4 Fig). Each detected m/z was assigned to specific BVOCs, which were classified as terpenoids (isoprene, monoterpenes, sesquiterpenes), benzenoids (1.3-cyclopentadiene, benzene, toluene, o-xylene, cresol, trimethylbenzene, ethylphenol), oxygenated BVOCs (methanol, acetone, acetic acid, methyl acetate) and green leaf volatiles (hexenal, hexanal, hexenol). Some of the compounds could be cross-validated by GC-MS analysis (Table 1) others were putatively identified on the basis of literature references (S1 Table) or in case of ethylphenol tentatively identified based on mass accuracy.
To shed more light on the biosynthesis pathways of selected BVOC emitted by H. halimifolium, position-specific labelling of pyruvate was used. As our measurement system is able to distinguish between 12 C and 13 C-BVOC fluxes as well as δ 13 CO 2 emitted, these experiments were set up to test whether BVOC emissions are linked to CO 2 released during daytime.
In response to cutting, a slight decrease of stomatal conductance and assimilation rates occurred (9 and 11%) during the first 2 hours of labelling. Plants, which did not reach stable values, were discharged from the analysis. The response was similar in control plants (deionized water) and labelled branches, suggesting no appreciable influence of the pyruvate solution. While the 12 C-BVOC flux of acetic acid (0.14 to 0.15 nmol m -2 s -1 ), methyl acetate (0.34 to 0.32 nmol m -2 s -1 ) and toluene (0.45 to 0.51 nmol m -2 s -1 ) remained constant, the fluxes of cresol (0.002 to 0.006 nmol m -2 s -1 ), trimethylbenzene (0.01 to 0.04 nmol m -2 s -1 ), ethylphenol (0.001 to 0.003 nmol m -2 s -1 ), monoterpenes (0.21 to 0.56 nmol m -2 s -1 ) and sesquiterpenes (0.19 to 0.50 nmol m -2 s -1 ) increased. In unlabelled control plants the 13 C/C total -BVOC ratio did not change significantly over time (Figs 2C-2J and 3). During [1-13C]-PYR labelling slight but significant increase in the heavier BVOC isotopologue (0.5-1%) was recorded (starting a  (11 a.m.) under controlled conditions. The m/z, assigned species and their molecular mass are given. Calibration was performed via calibration gas (gaseous), liquid solution (liquid) or via transmission. Some compounds were cross-validated via GC and the match factor given (NIST17 mass spectral library), were nd-not detectable in GC spectra, na-not analyzed. The GC Match for monoterpenes and sesquiterpenes refers to α-pinene and β-caryophyllene. n = 15± standard error (SE). Sensitivities are exemplary shown at a humidity level of 8000 ppm and limits of detection (LODs) are set to 3σ.  Real-time 13 C-BVOC and 13 CO 2 measurements few minutes after labelling) in almost all considered compounds, indicating a slight progressive incorporation of 13 C (Figs 2C-2J and 3). However, even after 3 h of labelling, the total incorporation of 13 C remained minor (<0.1%) in some compounds such as 13 C-acetic acid and methyl acetate. Since the labelled C1 position was markedly decarboxylated, a significant increase in 13 Fig 2B). Within 3 h of [2-13C]-labelling, significantly increased (2-6%) emission of the heavier BVOC isotopologues were observed for most compounds. As mentioned above, the standard error for acetic acid was high due to large intraspecific variability in the emission rates of individuals for this compound, which, nevertheless, showed an unequivocal incorporation of [2-13C]-PYR label. Fig 4 shows the temporal dynamics of the label incorporation into BVOCs and CO 2 during [2-13C]-PYR feeding. All considered BVOCs showed an increase in the heavier isotopologue already within the first 10 minutes. All benzenoid/terpenoid compounds (Fig 4 purple and  green lines) showed an almost linear incorporation, not reaching a steady state within 50 minutes, whereas the oxygenated compounds (Fig 4 red lines) revealed the strongest incorporation in the first 25 min, reaching almost an equilibrium thereafter. We observed clear differences in the velocity (slope) of 13 C-label incorporation into oxygenated compounds and benzenoids/ terpenoids, being fastest into methyl acetate and acidic acid, followed by the others. Interestingly, the terpenoid as well as the benzenoid compounds showed patterns comparable to the [2-13C]-label incorporation into emitted CO 2 (Fig 4 black line).

Discussion
In the present study we used position-specific labelling of the key metabolite pyruvate (either [1-13C] or [2-13C]-labelled), which is an efficient tool to trace carbon partitioning at metabolic branching points [15], to elucidate which moiety of the pyruvate molecule is incorporated into volatile compounds or released as 13 CO 2 in the light. The latter can be quantified as measurable 13 CO 2 emission from decarboxylating enzymatic steps mostly of the [1-13C]-atom of pyruvate. If, however, the entire pyruvate molecule is used as a precursor for volatiles, we expect similar 13 C-incorporation from both positions. In contrast, if the 13 C-tracer is only detected after feeding of [2-13C]-PYR, pyruvate must have been decarboxylated and only the C2-C3-moiety of the molecule is used for production of the volatile compound.
In general, labelling via the transpiration stream in cut branches induces mechanical stress to the plant and should therefore not be applied over many hours. Moreover, additional added metabolites to the plant due to labelling might cause changes in metabolic pathways. Therefore, low concentrated labelling solutions were used to reduce the influences and photosynthetic parameters were continuously controlled and remained relatively stable (Fig 2). On the other hand, a great advantage of position-specific metabolite labelling is, that it enables to trace the fate of specific carbon atoms through plant biochemical cycles, a helpful tool to study Real-time 13 C-BVOC and 13 CO 2 measurements metabolic branching points. Such position-specific labelling cannot be achieved via fumigation experiments.
The broad range of volatiles and their isotopoloques emitted from the Mediterranean shrub Halimium halimifolium indicated a rapid incorporation of 13 C to a different extend for each labelled carbon position, which elucidating the production pathways of certain compound groups (Fig 5), as discussed below in detail. Overall, the mass spectra revealed a broad range of organic compounds (> 30) emitted by H. halimifolium (S2 Table and S4 Fig), a selection of them could be identified and quantified in this study (Table 1). We chose the compounds with highest emission rates, clearest peak for 12 C-and 13 C-mass and thorough calibration. The list does not claim completeness but shall give an idea of compounds that are getting labelled via pyruvate during de novo biosynthesis. Apart from the well-studied and usually strongly emitted BVOCs like isoprene, monoterpenes, sesquiterpenes, acetaldehyde, formaldehyde and methanol, H. halimifolium shoots also released less well studied compounds such as methyl acetate and several benzenoids (benzene, toluene, o-xylene, cresol, trimethylbenzene and ethylphenol).

Biosynthesis of short-chained oxygenated volatiles
We observed a fast incorporation of 13 C into short-chain molecules such as acetic acid after feeding the plants with [2-13C]-PYR, which was partially released into the atmosphere due to its high volatility. Such connection of pyruvate with formation and emission of oxygenated compounds has been described in several studies [18,19,34]. The 13 C-labelled pyruvate fed to the plant is transported to the leaves via the transpiration stream where it is taken up by leaf cells [19]. In the cytosol, pyruvate can be converted to acetyl-CoA or acetaldehyde as catalyzed by pyruvate dehydrogenase (PDH) and pyruvate decarboxylase (PDC), respectively [3,18]. Real-time 13 C-BVOC and 13 CO 2 measurements Both reactions involve decarboxylation of pyruvate accompanied by a release of CO 2 . Studies causing enhanced acetaldehyde levels in mesophyll cells demonstrated that this toxic compound is further oxidized by aldehyde dehydrogenase (ALDH) to yield acetate [18]. On the other hand, acetate can also be produced during the exchange of C 2 -units between organelles like mitochondria and chloroplasts by acetyl-CoA hydrolysis [35]. Elucidation of the main source of acetate would need further knowledge on the 13 C-labelling of acetaldehyde, which, unfortunately, was not provided by our data.
While the biosynthesis of acetic acid via acetaldehyde and acetyl-CoA is well described [3,17,36], much less information is available on the formation of the volatile ester methyl acetate in plants [30]. H. halimifolium strongly emitted this compound at rates of 0.6±0.2 nmol m -2 s -1 , which was an order of magnitude higher than mono-(0.07±0.002 nmol m -2 s -1 ) or sesquiterpenes (0.01±0.0004 nmol m -2 s -1 ) emission and a little higher than reported rates for this species (0.39 ±0.22 nmol m -2 s -1 ) [30]. Noteworthy, methyl acetate (and also acidic acid) emission rates in our study were quite variable between plant individuals in contrast to other emitted compounds (Fig  2). H. halimifolium has been characterized as a species with high phenotypic and genotypic plasticity [27,37,38], however, though not being the focus of this study, we were not able to clearly assign high and low emission pattern to individual plants. Since the plants were kept under identical conditions in a controlled climate chamber, we cannot currently disentangle its reasons.
Our labelling patterns suggest that methyl acetate was either derived directly from acetic acid or it was formed from acetyl-CoA, which was produced by conversion of pyruvate and/or activation of acetic acid [30,36,39]. Early work suggested that volatile esters can be synthesised from acetyl-CoA by a transfer of the acetyl moiety to an alcoholic substrate as catalysed by an alcohol acetyltransferase [40]. It has also been proposed that such esters are produced from the lipoxygenase (LOX) pathway, which includes oxidation of cell membrane constituents [41,42]. However, this pathway seems unlikely considering the fast incorporation of the [2-13C]-PYR label in the present study (Fig 4); as we assume that the turnover rates of cell membrane components are much slower. Therefore, the results of our study suggest a link of methyl acetate biosynthesis with the production and emission of acetic acid and confirm earlier assumptions that either acetate or acetyl-CoA are enzymatically methylated [40]. Furthermore, tight linear correlation between 12 C-and 13 C-fluxes of acetic acid and methyl acetate (R 2 = 0.92 and R 2 = 0.97) support a strong metabolic connection between both compounds (S8 and S9 Figs).

Metabolic link between the biosynthesis of terpenes and volatile aromatics
Besides these short-chained oxygenated compounds, we observed emission and 13 C-incorporation into several larger volatile compounds including aromatics and terpenes (Figs 2 and 3). Interestingly, the experiments revealed incorporation of 13 C into sesquiterpenes, which has been barely reported before. It is generally assumed that sesquiterpenes are mainly synthesised in the cytosol via the mevalonate pathway (MVA) [43], although there are also some reports about a production in the chloroplasts via the methylerythrol-phosphate (MEP) pathway [44,45]. As expected from the intermediates used in these pathways (MVA: acetyl-CoA, MEP: pyruvate), sesquiterpenes showed a strong incorporation of the C2-C3 moiety of the pyruvate applied. The 13 C from [2-13C]-PYR is used in both pathways to form DMAPP and subsequently farnesyl pyrophosphate (FPP), the universal precursor of all sesquiterpenes [46,47], which ultimately results in biosynthesis of 13 C-labelled sesquiterpenes [43,48]. In contrast, the weak but distinct incorporation of 13 C from [1-13C]-PYR may be explained by photosynthetic re-fixation of emitted 13 CO 2 [17]. The fixed 13 CO 2 contributes to production of glyceraldehyde-3-phosphate, which can be transported to the cytosol, is incorporated into pyruvate derived from glycolysis and subsequently leads to formation of 13 C-labelled sesquiterpenes.
Furthermore, crosstalk between MEP-and MVA-pathway via IPP transport into the cytosol can provide an alternative explanation as well [49,50]. Nevertheless, the quick incorporation of the 13 C-tracer into sesquiterpenes suggests that these compounds were at least partially produced from de novo biosynthesis, which is also supported by our recent study on the same species [22] and is consistent with other work [44,[51][52][53]. However, we detected sesquiterpenes (mainly β-caryophyllene and farnesene) also in storage pools of H. halimifolium leaves [22], which indicates that emission is driven by both, a temperature dependent release from stored compounds plus de novo production of sesquiterpenes.
Similar to sesquiterpenes, we found strong 13 C-labelling of emitted monoterpenes if [2-13C]-PYR was fed to the plants, and a weaker 13 C-signal if [1-13C]-PYR was applied. Again, such labelling patterns suggest that the pyruvate taken up by the mesophyll cells was channelled into the plastids and used in the MEP-pathway to form geranyl diphosphate (GPP), the precursor of monoterpenes [43,54]. The MEP pathway starts with the formation of 1-deoxy-dxylulose 5-phosphate (DXP) by an acyloin condensation of hydroxyl-ethyl-thiamine, derived from the decarboxylated pyruvate, with the C1-aldehyde group of glyceraldehyde 3-phosphate (GA-3P) [49,55]. Thus, due to decarboxylation of pyruvate, the [1-13C]-atom of pyruvate was released as 13 CO 2 and could only be used for terpene biosynthesis after photosynthetic re-fixation. Because of the strong 13 C-labelling of monoterpenes, we further conclude that emission is at least partially driven by de novo biosynthesis. In good agreement with this assumption, we recently demonstrated that many volatile terpenes (in the order of abundance ρ-cymen-8-ol > α-pinene > eugenol > limonene > terpinene-4-ol > camphene > p-cymene > rcymene > myrcene) were present in leaf tissue of H. halimifolium and many of them seemed to be produced de novo in the leaves of this species [22]. Noteworthy, a major portion of the terpenes found in H. halimifolium leaves were aromatics such as p-cymen-8-ol, eugenol, pcymene and m-cymene. Radio-labelling experiments and enzymatic evidence suggest that the aromatic monoterpene p-cymene is produced by an aromatization of γ-terpinene [56]. This compound together with the structurally very similar and biochemically related p-cymene-8-ol and terpinen-4-ol was highly abundant in H. halimifolium leaves [22], suggesting a closely linked production.
The labelling patterns observed for these compounds (Figs 2 and 3) indicate incorporation of 13 C from [2-13C]-PYR, whereas, similar to mono-and sesquiterpenes, the [1-13C]-PYR label was not intensively incorporated into volatile benzenoids. The fast incorporation of [2-13C]-PYR label into benzenoids suggest that these compounds are produced de novo and emission is not mainly driven from storage pool release. Earlier 13 CO 2 fumigation experiments confirm this assumption at least for toluene [57,58].
Even though benzenoids are assumed to be synthesised in the shikimic acid pathway [47,54], the exact biosynthesis pathway of the selected compounds is still unclear [57,58,60]. Importantly, the shikimic acid pathway starts with the addition of pyruvate to erythrose-4-phosphate to eventually form shikimate. In subsequent reactions another PEP is added to shikimate-3-phosphate to generate chorismate. This compound, therefore, should contain two 13 C-atoms independent if [1-13C]-PYR or [2-13C]-PYR is fed to the plants. Before chorismate is converted to phenylalanine, the assumed precursor for volatile benzenoids, one CO 2 is decarboxylated either by arogenate dehydrogenase or by prephenate dehydratase (PDT), so that phenylalanine would either be double-labelled with 13 C if [2-13C]-PYR is fed, or single 13 C-labelled if [1-13C]-PYR is fed. The aromatic amino acid phenylalanine is converted to cinnamic acid by the action of phenylalanine ammonia lyase (PAL); volatile benzenoids are then produced by further changes and substitutions on the aromatic ring. Importantly, these modifications involve the release of one 13 C from cinnamate most likely leading to benzenoids, which are labelled by one 13 C if [2-13C]-PYR was fed and without 13 C-label if [1-13C]-PYR was fed. Our results support the idea that volatile benzenoids are formed via the shikimate pathway. On the other hand, they show a very close correlation of benzenoids 13 C-labelling with the labelling of terpenes (Fig 4), a high abundance of aromatic terpenes in leaves of H. halimifolium, and a very similar chemical structure of these aromatic monoterpenes with the released benzenoids. This highlights the possibility that these compounds might also be derived from the MEP pathway as suggested by Misztal et al. [58]. This hypothesis is supported by the correlation matrix of benzenoids and monoterpenes fluxes ( 12 C-monoterpenes to 12 Ctoluene (R 2 = 0.98), 13 C-monoterpenes to 13 C-toluene (R 2 = 0.99), to 13 C-cresol (R 2 = 0.99) or to 13 C-trimethylbenzene (R 2 = 0.99)) (S8 and S9 Figs), indicating a strong connection between these biosynthesis pathways. However, further labelling studies of important intermediates, such as phenylalanine or shikimate, is required to elucidate the pathways responsible.

Day respiratory CO 2 efflux from metabolic branching points
In general, feeding of [1-13C]-labelled pyruvate resulted in substantially increased 13 CO 2release from the shoots. This increase was threefold stronger as compared to [2-13C]-PYR labelling and detectable immediately after applying the label (Fig 2B), indicating the quantitative relevance of decarboxylation processes during metabolic pathways in plants. As pointed out above, the incorporation of the [1-13C]-atom of pyruvate into the emitted BVOCs was much weaker than the [2-13C]-atom, suggesting that the entire pyruvate molecule is generally not directly used as a precursor of volatile compounds. The remaining weak 13 C-label may originate from photosynthetic re-fixation of the decarboxylated gaseous 13 CO 2 . Similar CO 2 re-fixation and incorporation into volatiles has been reported in 13 C-labelling experiments with other plants [17,19,63]. However, it must be denoted that incorporation of 13 C into acetic acid and methyl acetate during [1-13C]-PYR feeding was minor, which is not in line with substantial re-fixation, which would lead to measurable incorporation into these oxygenated compounds. A possible reason for the stronger labelling of e.g. monoterpenes could be the metabolic and spatial proximity to the location of CO 2 fixation in the Calvin cycle. Acetaldehyde, however, a direct precursor of acetic acid, was found to be strongly labelled during 13 CO 2 fumigation of Prosopis veluntina [64]. Thus, freshly fixed 13 CO 2 is incorporated into oxygenated volatiles, suggesting [1-13C]-PYR labelling of terpenoids and benzenoids might not just be explained by re-fixation.
Moreover, the strong release of 13 CO 2 after [1-13C]-PYR feeding indicates that decarboxylation of pyruvate is not restricted to dark period. Thus, this non-mitochondrial source of CO 2 seems to contribute considerably to the CO 2 release of leaves during the light period. This observation underlines the importance of CO 2 production during the day, which is still a highly debated topic [12,65]. It is widely accepted that mitochondrial respiration is partly inhibited in the light compared to the dark due to a reorchestration of major pathways, resulting in a lower CO 2 efflux [10][11][12][13][14]. But it is also assumed that day respiration can account e.g. for at least 5% of the net assimilation in leaves [13]. Though, at the first glance, it might be of minor importance for a total plant carbon balance, nevertheless under unfavourable environmental conditions with low assimilation rates, day respired CO 2 may become an important player in the carbon budget [12]. Nevertheless, since the measurement of day respiratory CO 2 efflux is a nontrivial problem due to strong interaction with photosynthesis and photorespiration, quantifying the amount of day released carbon remains challenging. Hence, precise estimations of day respiratory CO 2 efflux from plants is still scarce [13] and even less is known on its regulation under changing environmental conditions. Here, we can show that one source of non-mitochondrial CO 2 release in the light is the decarboxylation of the C1-atom of pyruvate during biosynthesis of a plethora of plant secondary metabolites such as BVOCs, a carbon source which our approach can help to account for.

Conclusion and perspective
Real-time, synchronised analysis of BVOCs and δ 13 CO 2 in combination with position-specific labelling of central metabolites provide an enormous potential for biochemical research. Here we demonstrated that position-specific pyruvate labelling combined with on-line BVOC and 13 CO 2 measurement helps to disentangle possible biosynthetic pathways of less studied BVOCs such as methyl acetate or toluene. Methyl acetate might have a biochemical origin strongly connected to acetic acid and methanol and there are hints for a synthesis of some benzenoids via the MEP-pathway. Furthermore, during many of these BVOC biosynthesis pathways the C1-carbon atom of the central metabolite pyruvate is decarboxylated, while the remaining C2-C3-moiety is incorporated. Given the fact, that day respiratory CO 2 efflux is still a big uncertainty in the plant carbon budget, our approach can provide new insights in key determinants controlling the emission of CO 2 in the light period. Despite of increasing awareness of fast reactions in plant emitted trace gases and a tight linkage between primary and secondary carbon metabolism, there is still a lack of understanding the regulation of carbon allocation into these pathways and driving factors, environmental constrains or differences among plant functional groups. Further investigations opens the door for real-time analysis tracing metabolic pathways and carbon turnover under different environmental conditions, which may enhance understanding of regulatory mechanisms in plant carbon metabolism.
Supporting information S1