In summer 2018, we conducted two measurement campaigns in a Q. suber ecosystem in Eastern Portugal (Vila Viçosa, Alentejo, 38° 47′ N, 7° 22′ W, 430 m a.s.l.). The mean annual air temperature is 16.5 °C with a mean annual precipitation of 585 mm (1981–2010, www.ipma.pt). The ecosystem is dominated by Q. suber trees, which have been progressively invaded by C. ladanifer, an invasive shrub species naturally occurring in Portugal. The experimental site consists of three blocks, which all include Q. suber and C. ladanifer individuals. All blocks were fenced in winter 2017/2018 to avoid disturbance by fauna. We retrieved meteorological data such as precipitation, air temperature, relative humidity and photosynthetically active photon flux density (PPFD) from small meteorological stations installed on each block. Vapor pressure deficit (VPD) was calculated from air temperature and relative humidity values. Volumetric soil water content was measured continuously with 5TM probes (METER Group, Inc. USA, Pullmann, WA, United States) in 0.20 m depth. The soil on site is poorly developed and classified as a haplic Leptosol (FAO, 2006), derived from schist with a high proportion of soil skeleton (see Caldeira et al., 2015; Haberstroh et al., 2018 for further details).

BVOC sampling took place from 19 to 26 June and 16–18 July 2018. On each of the three blocks, we chose four trees (n = 12) and two shrub individuals (n = 6) for measurements. The sampling started between 8 and 9 am and proceeded to 5 or 6 pm to cover a daily course of emissions. On each day, BVOCs were sampled with a dynamic enclosure system (Haberstroh et al., 2018) along the diurnal course with five measurements (n = 5) for 1–2 h on the same leaf or branch. This study set-up allowed us to collect 60 BVOC samples for Q. suber (n = 60) and 30 samples for C. ladanifer (n = 30) during each of the sampling campaigns. For Q. suber, custom-made enclosures with a volume of ~30 ml were used to fit a single, south exposed, current-year leaf at ~3 m height. In rare cases, two leaves were enclosed, which did not shade each other. Enclosures (~460 ml) for C. ladanifer were made to fit a south exposed branch with 8 to 19 current year leaves. Leaves and branches were enclosed at least 30 min before the first sampling period and remained inside the enclosure until the last measurement in the late afternoon. Reinstalling the enclosures for each measurement point raised concerns about the mechanical disturbance of the leaves and the production of artifacts from induced emissions. Air sampling pumps (210-1003MTX, SKC, Germany) were connected to the enclosures, which were permanently flushed with an ambient airflow rate of 200 ml min⁻¹. Although the enclosures were flushed permanently, an enclosure effect on the measurement temperature cannot be excluded. All materials in contact with the gas stream were made of chemically inert PET foil (Bratschlauch, Toppits®, Minden, Germany) or perfluoroalkoxy tubing (Swagelok, Karlsruhe, Germany). During sampling, an air sampling glass tube (Gerstel, Mülheim a.d.R., Germany) filled with Tenax TA (Sigma Aldrich, Munich, Germany) was placed between enclosure and air sampling pump to accumulate the emitted BVOCs. Enclosed leaf samples were occasionally shaded with neutral density meshes in case of condensation effects on the enclosure walls. Natural shading effects occurred throughout the day depending on neighboring trees and solar altitude. PPFD was measured with a Ceptometer (ACCUPAR LP-80, METER Group, Inc., Washington, USA) parallel to the leaf or branch once during the sampling period. Ambient BVOC concentrations were determined 2 m above ground and sampled simultaneously with plant BVOCs. All adsorbent tubes were shaded during sampling with aluminum foil. After sampling, adsorbent tubes were stored at +4°C in Labco Exetainers (Labco Limited, Lampeter Ceredigion, United Kingdom) until further analysis. We harvested leaves of Q. suber and C. ladanifer after the last sampling period for determination of dry weight after oven drying (65°C, 48 h). Further description of the sampling procedure and sampling material is given in Haberstroh et al. (2018).

During each sampling campaign, net carbon assimilation, stomatal conductance and transpiration were measured with a portable LI-6400XT photosynthesis system (LICOR Inc., Nebraska, USA) equipped with a CO₂ mixer and light source. Gas exchange measurements were always conducted on leaves close to the leaf enclosures at a PPFD of 300 μmol m⁻² s⁻¹ to match the conditions under natural or artificial shading and to allow for a general comparison of net carbon assimilation values. Due to the inaccessibility of Q. suber leaves with the heavy and sensitive gas exchange equipment in the tall canopies, gas exchange was measured on freshly detached branches (Haberstroh et al., 2018). For C. ladanifer, attached leaves were chosen, as shrubs were easily accessible. Photosynthetic discrimination (δ¹³C_plant) of both species was calculated in‰ according to Farquhar et al. (1982) for C₃-plants as in equation 1 for values obtained from measurements at a PPFD of 300 μmol m⁻² s⁻¹:

where δ¹³C_atm is the carbon signature of the ambient air, a is the fractionation factor due to diffusion of CO₂ in air, b is the discrimination factor of RuBisCO, c_i the internal CO₂ partial pressure in the intercellular space and c_a the atmospheric CO₂ partial pressure. δ¹³C_atm was taken as −8‰, a as 4.4‰ (Farquhar et al., 1982) and b as 27‰ (Cernusak et al., 2013). c_i and c_a were obtained from gas exchange measurements. Values for δ¹³C_atm were taken from the literature, as no field measurements were available. We are aware, that δ¹³C_atm is not always stable; however, the difference should have a minor effect in an open evergreen Mediterranean ecosystem. In addition, both measurement campaigns were conducted in close temporal proximity. The calculated daily mean values of photosynthetic discrimination were considered to be equal to the δ¹³C signature of the recently fixed carbon in photosynthesis. In addition, pre-dawn (Ψ_PD) and midday leaf water potential (Ψ_MD) were determined with a Scholander-type pressure chamber (PMS 1,000, PMS Instruments, Corvalis, Orgeon, OR, USA). The meteorological conditions during all sampling days were stable with clear skies, seldom intercepted by minor clouds. No precipitation occurred during the measurements.

Emitted BVOCs were analyzed on a gas chromatograph (GC, 6890A, Agilent Technologies Böblingen, Germany) with a mass-selective detector (MSD, 5975C, Agilent Technologies Böblingen, Germany) containing a thermodesorption/cold injection system (TDU-CIS, Gerstel, Germany). Briefly, adsorbent tubes were heated to 220°C to thermodesorb the volatiles, which were then cryotrapped at −50°C in the CIS. Heating to 240°C released the BVOCs onto the separation column (DB-5ms UI, 30 m × 0.25 mm ID, 0.25 μm film thickness, Agilent Technologies Böblingen, Germany). Helium was used as carrier gas at a flow rate of 1 ml min⁻¹. The MSD operated at 70 eV with an ion source temperature of 230°C and a quadrupole temperature of 150°C. Further GC oven and MSD settings are given by Kleiber et al. (2017) and Haberstroh et al. (2018). Obtained mass spectra were processed with the MassHunter Software (Agilent Technologies, Böblingen, Germany). For compound identification mass spectra were compared to the NIST library together with available authentic standards, which were further used for quantification. Emission rates (E_m) in μg g⁻¹ DW h⁻¹ of chosen BVOCs were calculated as in formula 2:

where c_o and c_c (μg) are the amounts of BVOCs trapped on adsorbent tubes from plant enclosures and ambient air, respectively, d_w (g) the dry weight of plant leaves in enclosures and t (h) the sampling time. We selected the acyclic monoterpenoids myrcene (MC), linalool (LN), and trans-β-ocimene (TBO), the cyclic monoterpenes α-pinene (AP), camphene (CP), limonene (LM), sabinene (SB), the monoterpene related alkyl benzene p-cymene (PC), the monoterpenoids verbenone (VB) and bornyl acetate (BAC), the sesquiterpenes δ-cadinene (DC) and ledene (LD) and two benzenoid aromatic compounds benzaldehyde (BZA) and durene (1,2,4,5-tetramethyl-benzene) (DU) for further investigation. All compounds were identified with a match factor of at least 90%. These compounds were chosen, as they were mostly present in considerable amounts in both species and represent chemically different BVOC classes. In addition, a certain amount of BVOCs is necessary to obtain a reliable and reproducible carbon isotopic ratio measurement (section carbon isotope ratio measurements). This was not the case for many BVOCs emitted at low rates.

δ¹³C values of plant volatiles were determined by gas chromatography—combustion interfaced—isotope ratio mass spectrometry (GC-C-IRMS). This system allows a high precision and accuracy of carbon isotopic ratios at natural abundance (Meier-Augenstein, 1999). For compound-specific δ¹³C analysis, the separation column ended at a split channeling 10% of the eluting compounds to the MSD for compound identification and quantification (see above), and 90% to a combustion furnace (GC5 interface, Elementar, Hanau, Germany) and ultimately into the IRMS (Isoprime precisION, Elementar, Hanau, Germany) where the carbon isotopic ratios of individual BVOCs were analyzed. The eluting compounds were combusted continuously at 850°C by CuO producing CO₂ and H₂O. After passing a Nafion H₂O trap, CO₂ entered the continuous flow IRMS to determine δ¹³C signatures for all compounds separated by GC. As reference compounds, we used octane (δ¹³C: −31.75 ± 0.01‰) and octadecane (δ¹³C: −32.70 ± 0.01‰) (available from Schimmelmann lab, Indiana University, https://arndt.schimmelmann.us/compounds.html). Data were analyzed with the Software IonIOS (Elementar Analysesysteme GmbH, Langenselbold, Germany). Due to small emission rates, it was not possible to obtain reliable δ¹³C values in some cases. Peaks with a height <0.5 nA and > 50 nA were excluded from the analysis.

We applied a two-way repeated measure analysis of variance (RMANOVA) to identify significant differences in assimilation, stomatal conductance, photosynthetic discrimination, Ψ and total BVOC emission rates between species and sampling dates. In case of statistical significance, we applied the post hoc Tukey's test. Beforehand the data was tested for normal distribution and equality of variances. If assumptions for an ANOVA were not met, data was log or square root transformed. Due to normality violations, differences in the emission of single BVOCs between months were tested for each species separately with the Wilcoxon signed-rank test to account for the repeated measures design. The δ¹³C values of single BVOCs were tested for significant differences between months with a one-way RMANOVA for each species separately. In rare cases, the assumptions for a RMANOVA were violated and results crosschecked and confirmed with the Wilcoxon signed-rank test. The δ¹³C values of BVOC were tested with Grubbs' test for outliers in a data sample (package “outliers” in R). In 37 out of 54 datasets, no outlier could be identified. The number of omitted data points (1–5) in the remaining datasets is given in Table 2. Afterwards, missing data was imputed with the package “mice” (multivariate imputation by chained equations) in R. The complete datasets for June and July were used to perform Ward hierarchical clustering (method: “Euclidean”) with the package “stats” in the statistical software R. All statistical analysis was conducted in R (version 3.4.3).

The two sampling dates were characterized by very similar meteorological conditions (Table 1). Air temperatures reached maximum values of 33.5 and 33.0°C in June and July, respectively. VPD and air temperatures were moderate in the morning and highest in the afternoon between 2 and 5 p.m. The light conditions were highly variable during the measurements ranging from 30 to 1,860 μmol m⁻² s⁻¹ in June and from 40 to 1,940 μmol m⁻² s⁻¹ in July, as sampled leaves were partially shaded in their natural position at some point during the day.

Precipitation was twice as high 30 days prior to the first campaign compared to the second campaign in July (Table 1). Hence, the soil water content in 0.20 m depth decreased on average by 0.03 m³ m⁻³ from June to July. This trend was reflected onto the net carbon assimilation rates of both species, which significantly (p < 0.001) declined (Figure 1A). While assimilation of C. ladanifer was clearly higher compared to Q. suber in June (p < 0.01), differences vanished in July. The same pattern was observed for stomatal conductance, though values were not significantly different between species neither in June nor in July (Figure 1B). Likewise, photosynthetic discrimination increased significantly (p < 0.05) from −27.7 ± 1.7 to −26.2 ± 1.9‰ in Q. suber and non-significantly from −27.0 ± 1.8 to −26.1 ± 1.5‰ in C. ladanifer (Figure 1C). A similar trend is indicated by the leaf water potential measurements (Figure 1D), where Ψ_PD declined significantly in both species (p < 0.001), although more profoundly for C. ladanifer from −1.00 ± 0.05 MPa to −1.45 ± 0.03 MPa. Ψ_MD was significantly lower for C. ladanifer (p < 0.01) compared to Q. suber in both months. Between months, significant changes of Ψ_MD did not occur for either species.

Total BVOC emissions of Q. suber and C. ladanifer were highly variable and clearly differed (p < 0.05) between the two species (Figure 2). Total emissions of Q. suber slightly increased (p > 0.05) on average from 5.6 ± 1.0 μg g⁻¹ h⁻¹ in June to 7.0 ± 1.3 μg g⁻¹ h⁻¹ in July. Emissions were dominated by monoterpenoids, which made up over 99% of emissions in both months (Figures 2A,C). Other BVOCs such as bornyl acetate, ledene, benzaldehyde or durene were only detected in few plant samples and in low quantities (Figure 2E). Emissions of plant volatiles only increased significantly between June and July for sabinene (p < 0.001) and durene (p < 0.05) and decreased significantly for ledene (p < 0.001), benzaldehyde and bornyl acetate (both p < 0.05). The sesquiterpene δ-cadinene could not be detected in the emissions of Q. suber, and hence was only emitted by C. ladanifer. In June, total emissions of C. ladanifer were low with 1.2 ± 0.1 μg g⁻¹ h⁻¹ and almost doubled to 2.3 ± 0.3 μg g⁻¹ h⁻¹ in July (p < 0.01). This increase was mostly attributed to a significant increase in emissions of α-pinene, sabinene, camphene, p-cymene, limonene, myrcene, and bornyl acetate (p < 0.05). The dominant BVOCs were p-cymene and limonene with a share of at least 50% of emissions (Figures 2B,D). In contrast to Q. suber, emissions of myrcene, trans-β-ocimene and linalool were low and not always detected in C. ladanifer. On the other hand, the monoterpenoids, verbenone, and bornyl acetate had a considerable share of total emissions with 13% in June and 19% in July (Figure 2F). Sesquiterpene emissions were low, but constant from June to July. The two benzenoid aromatic compounds benzaldehyde and durene were emitted at similar rates by both species. It must be noted that we focused on BVOCs mostly present in both species for comparative reasons. However, the total emissions of C. ladanifer were more diverse in terms of total emitted compounds.

Between sampling dates and species, the obtained δ¹³C values of BVOCs varied along a very broad range from −23.7 to −42.6‰ (Figure 3, Table 2). Since no clear diurnal pattern of δ¹³C values could be detected for BVOCs or for the potential carbon source (Figure S1), we focused on differences between sampling campaigns, species and BVOC compound classes. In both species, δ¹³C values of most BVOCs became more enriched from June to July (Figure 3). This pattern reflected the calculated decrease in photosynthetic discrimination. Yet, the change in δ¹³C values was stronger for BVOCs, compared to the change in the potential carbon source. On average, δ¹³C values of BVOCs emitted by Q. suber increased by 3.4 ± 0.5‰ and by C. ladanifer by 1.8 ± 0.5‰ from June to July. For all BVOCs with sufficient number of replicates emitted by Q. suber, the carbon isotope composition changed significantly from June to July (p < 0.05). For verbenone, bornyl acetate, ledene, benzaldehyde, and durene (Figure 3E) a statistical test was not applied due to the small sample size (Table 2), since not all plant individuals emitted these compounds at rates above detection limits (Figure 2E, Table 2). In C. ladanifer, only 8 out of 14 compounds became significantly (p < 0.05) more enriched from June to July, namely myrcene, α-pinene, limonene, ledene, δ-cadinene, verbenone, bornyl acetate, and durene.

In Q. suber, the most enriched compounds were linalool, myrcene and trans-β-ocimene with values of −28.3 to −28.8‰ in June and −23.7 to −25.9‰ in July. These BVOCs were either similar or even slightly ¹³C-enriched compared to the potential carbon source in Q. suber (Figure 3A). In C. ladanifer, δ¹³C values of myrcene were comparable (June: −30.5‰; July: −25.8‰), while trans-β-ocimene and linalool were more ¹³C-depleted (Figure 3B, June: −28.5 to −31.3‰; July: −29.3 to −30.5‰), yet emissions were low and not detected in all samples (Figure 2B). While δ¹³C values differed between species for p-cymene in June (Q. suber: −32.5‰; C. ladanifer: −28.0‰), they were in the same range of −27.0 to −27.3‰ in July (Figures 3A,B). The cyclic monoterpenes limonene, α-pinene, sabinene, and camphene were depleted in ¹³C compared to the potential carbon source with δ¹³C values between −29.4 and −32.9‰ in June and −27.6 and −29.8‰ in July, independent of the species (Figures 3C,D, Table 2). Of these four cyclic monoterpenes, α-pinene was the most enriched compound (−28.4 to −29.5‰), while camphene was the most depleted (−28.9 to −32.9‰). The oxygenated monoterpenoids verbenone and bornyl acetate were clearly depleted compared to all other terpenoids with values between −31.5 and −37.0‰ (Figures 3E,F). The two sesquiterpenes ledene and δ-cadinene, however, expressed δ¹³C values more similar to the cyclic monoterpenes (Figures 3C,D). The aromatics benzaldehyde and durene were also ¹³C-depleted, especially compared to the monoterpene related alkyl benzene p-cymene (Figures 3A,B). Although emissions of Q. suber were low and not present in all samples, δ¹³C values for ledene, verbenone, bornyl acetate, benzaldehyde, and durene were similar to the values measured for C. ladanifer (Figures 3E,F).

Given the overall similarities of δ¹³C values between the two species, we decided to conduct hierarchical clustering analyses independent of the species for June and July (Figures 4A,B). In June, two main clusters could be identified, of which the first (I) included the monoterpenoids verbenone, bornyl acetate and the benzenoid compound durene (Figure 4A). The remaining compounds were allocated to the second cluster (II), which split into two sub-clusters. All acyclic monoterpenoids clustered into the second of these two sub-clusters (ii). Moreover, the clustering approach for July yielded an even more consistent picture (Figure 4B), as the variability of δ¹³C values of most BVOCs was reduced from June to July (Figure 3). Both oxygenated monoterpenoids and benzenoid aromatic compounds were allocated to the first cluster (I). In the second cluster (II), all cyclic monoterpenes except camphene were found in the same sub-cluster (ii) with the acyclic monoterpenoids myrcene, linalool and trans-β-ocimene. The second sub-cluster (ii) included both sesquiterpenes and the cyclic monoterpene camphene. Hence, BVOCs with a similar chemical structure were allocated to the same group, showing similar δ¹³C values. It must be denoted that the dendrograms were created using imputed datasets and require careful interpretation. Hence, they only give a first indication of similarities and dissimilarities of the investigated BVOCs.

BVOCs are either emitted de novo or from storage pools (i.e., Fineschi et al., 2013). Independently of the emission source, the very first isotopic imprint on the carbon source for the production of plant volatiles derives from isotopic discrimination during photosynthetic carbon fixation by RuBisCO in the primary metabolism of C₃-plants. This process itself strongly reflects changes in environmental conditions, as both the stomatal conductance to CO₂ as well as assimilation rate determine the effective ¹³C-discrimination by RuBisCO (Farquhar et al., 1982). Hence, under favorable environmental conditions, relatively ¹³C-depleted carbon molecules are produced, as seen for both species in this study in June. From June to July, environmental conditions became slightly drier, which resulted in decreased net assimilation, stomatal conductance and photosynthetic discrimination of Q. suber and C. ladanifer. This ¹³C-enrichment of freshly produced sugars was clearly reflected in the carbon isotope composition of BVOCs in July, pointing mainly toward de novo emission from both species. Q. suber does not possess specialized storage structures for BVOCs (Loreto et al., 1996) and can, per se, only emit de novo produced BVOCs. Several studies also indicate that BVOC storing plants, such as C. ladanifer, are also able to emit substantial amounts of BVOCs from de novo production (Lluisà et al., 2010; Yáñez-Serrano et al., 2018). Hence, the changes of the δ¹³C-values of BVOCs over time support our first hypothesis that changing environmental conditions evoke changes in the carbon isotope composition of recently fixed carbon and de novo emitted BVOCs. Yet, the carbon isotope composition was not only different between months, but also between the investigated BVOC compound classes. These differences point toward various metabolic steps and processes in the formation of these BVOCs. Thus, we suggest a scheme for different metabolic pathways independently of the species for BVOCs investigated in this study based on the obtained δ¹³C values and available literature (Figure 5). As the variability of δ¹³C values for BVOCs was lower in July and yielded a clearer picture in the clustering approach (Figure 4B), this scheme is based on the δ¹³C-values obtained from samples with sufficient replicates in July.

Given the large amount of microbes on leaf surfaces (Lindow and Brandl, 2003) and their ability to produce, absorb and alter volatiles (e.g., Korpi et al., 2009; Junker and Tholl, 2013), a contribution to the observed emission rates and δ¹³C values cannot be excluded and even seems likely in some cases. All VOCs presented in this study, except durene, have been detected to be also emitted from either bacteria or fungi as microbial VOC (mVOC) and can be found in the mVOC 2.0 (http://bioinformatics.charite.de/mvoc/) database (Lemfack et al., 2017). Especially, for substances with relatively low emission rates, the contribution of mVOC could affect the measured δ¹³C values and could add to the high variability of δ¹³C values of some VOCs, even though very little is known to date. Hence, the δ¹³C values of myrcene, trans-β-ocimene and linalool emitted by C. ladanifer and verbenone, bornyl acetate and benzaldehyde emitted by Q. suber could point toward a mixed or microbiome dominated origin of emissions. It is intriguing that the variability of δ¹³C values for those substances decreased strongly from June to July. The microbiome is known to change in richness and diversity in response to changes in UV light (Jacobs and Sundin, 2001; Lindow and Brandl, 2003) and seasonality, such as progressing drought conditions (Peñuelas et al., 2012). Thus, potential changes in the microbial composition and their emission spectrum from June to July could explain the reduced variability in δ¹³C values, but also the re-allocation of benzaldehyde from the second to the first cluster (Figure 4), however further research is needed. Benzaldehyde is a compound which is produced by many microbes present in the phyllo- and rhizosphere, e.g., Pedobacter sp. (Garbeva et al., 2014) or Pseudomonas sp. (Hunziker et al., 2015). Yet, to fully evaluate the contribution of the microbiome, the calculation of flux weighted effects would be necessary.

A contribution of mVOC emissions to total measured VOC emissions and δ¹³C values for other VOCs can also not be fully excluded, but might be negligible in some cases due to high plant emission rates. This seems in general to be the case for most monoterpenes found in the emissions of Q. suber, as α-pinene, limonene, myrcene, and sabinene are commonly the main BVOC components detected in this species (Staudt et al., 2004; Loreto et al., 2009; Haberstroh et al., 2018). To elucidate the influence and contribution of the microbiome to VOCs emitted at high rates by plants, more studies that are specific targeted at the microbiome, its emissions, and metabolism are certainly necessary. However, as the metabolism for mVOC for terpenoids and aromatic benzenoid compounds is believed to be similar to the secondary plant metabolism (Korpi et al., 2009), the scheme in Figure 5 might also be valid for the microbial secondary metabolism.

The results of our study demonstrate that the investigation of the carbon isotope composition at natural abundance has a great potential to elucidate different metabolic pathways and fractionation processes during the biosynthesis of BVOCs under field conditions. Up to date, we are only aware of a few studies taking advantage of this carbon isotope composition natural abundance approach in BVOC studies, namely Affek and Yakir (2003), Jux et al. (2001) and Tan et al. (2018). The advantages of this approach lie in the simplicity of the application, where no costly carbon labeling is necessary (Tan et al., 2018). In vivo sampling in field studies is especially important as conditions reflect the real world more closely than laboratory approaches, where most environmental factors are generally kept constant. Although we observed some variability in our carbon isotope data (Figure 3), the hierarchical clustering confirmed the validity of our approach, allocating BVOCs with a similar chemical structure into the same cluster (Figure 4). This is especially intriguing, as we sampled BVOCs under highly varying environmental conditions, such as PPFD (Figure S2) and air temperatures (Table 1) as well as emission rates (Figure 2) over diurnal courses. In addition, we investigated two species, which are known to possess dissimilar emission patterns (Haberstroh et al., 2018) and different reactions to changing environmental conditions (Caldeira et al., 2015). Moreover, single plant individuals of the same species can have a very distinct fingerprint regarding the composition and amplitude of their emissions, even at controlled environmental conditions (Loreto et al., 2009; Yáñez-Serrano et al., 2018). Nevertheless, both species and all individuals reacted similar in their photosynthetic discrimination to decreasing water availability, which reflected synonymously the carbon isotope ratio of emitted BVOCs. In conclusion, we show that the investigation of the carbon isotope composition at natural abundance can be a robust measure to further advance our knowledge on similarities and dissimilarities in the biosynthetic pathway of BVOCs. Moreover, the δ¹³C values detected in this study and the proposed metabolic pathways demonstrate how much there is still to learn about the secondary metabolism of plants.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.