New insights into the mechanisms of plant isotope fractionation from combined analysis of intramolecular 13C and deuterium abundances in Pinus nigra tree-ring glucose

Understanding isotope fractionation mechanisms is fundamental for analyses of plant ecophysiology and paleoclimate based on tree-ring isotope data. To gain new insights into isotope fractionation, we analysed intramolecular 13C discrimination in tree-ring glucose (Δi’, i = C-1 to C-6) and metabolic deuterium fractionation at H1 and H2 (εmet) combinedly. This dual-isotope approach was used for isotope-signal deconvolution. We found evidence for metabolic processes affecting Δ1’ and Δ3’ which respond to air vapour pressure deficit (VPD), and processes affecting Δ1’, Δ2’, and εmet which respond to precipitation but not VPD. These relationships exhibit change points dividing a period of homeostasis (1961-1980) from a period of metabolic adjustment (1983-1995). Homeostasis may result from sufficient groundwater availability. Additionally, we found Δ5’ and Δ6’ relationships with radiation and temperature which are temporally stable and consistent with previously proposed isotope fractionation mechanisms. Based on the multitude of climate covariables, intramolecular carbon isotope analysis has a remarkable potential for climate reconstruction. While isotope fractionation beyond leaves is currently considered to be constant, we propose significant parts of the carbon and hydrogen isotope variation in tree-ring glucose originate in stems (precipitation-dependent signals). As basis for follow-up studies, we propose mechanisms introducing Δ1’, Δ2’, Δ3’, and εmet variability.


Introduction
Analysis of the systematic 13 C/ 12 C variation (commonly termed " 13 C signal"; abbreviations in Table 1) across tree-ring series is widely used to study past climate conditions, plant-environment interactions, and physiological traits such as leaf water-use efficiency (CO2 uptake relative to H2O loss) (Leavitt & Roden, 2022).Signals found at the whole-tissue or whole-molecule level (Fig. 1A, top and middle) are commonly interpreted based on a simplified mechanistic model of 13 C discrimination, Δ (denoting 13 C/ 12 C variation caused by physiological processes) (Farquhar et al., 1982).This model considers isotope effects of CO2 diffusion from ambient air into intercellular air spaces (Craig, 1953) and CO2 assimilation by rubisco (Roeske & O'Leary, 1984) and phosphoenolpyruvate carboxylase (PEPC; Fig. 2) (Farquhar, 1983;Farquhar & Richards, 1984).Manifestation of these effects as 13 C discrimination depends on the ratio of intercellular-to-ambient CO2 partial pressure (pi/pa) (Farquhar et al., 1982), and a highly significant positive relationship between pi/pa and leaf Δ was confirmed experimentally (Evans et al., 1986).Environmental parameters influence pi/pa and thus leaf Δ (Evans et al., 1986) by affecting the stomatal aperture and CO2 assimilation.For instance, in response to drought, isohydric plant species such as Pinus nigra (studied here) close their stomata (McDowell et al., 2008).This can be expected to decrease pi/pa and leaf Δ (Farquhar et al., 1982;Evans et al., 1986).
Isotope fractionation by metabolic processes downstream of CO2 assimilation is complex (Hobbie & Werner, 2004), incompletely understood (Badeck et al., 2005;Cernusak et al., 2009) and has yet to be adequately integrated into 13 C-discrimination models (Ubierna et al., 2022).Specifically, the simple 13 C discrimination model described above requires multiple adaptations to enable correct interpretation of the 13 C composition of tree-ring glucose (studied here).For instance, we recently argued that incorporation of carbon assimilated by PEPC into tree-ring glucose is negligible because leaves lack a high-flux pathway shuttling this carbon into glucose metabolism (Fig. 2) (Wieloch et al., 2022c).Therefore, all carbon in treering glucose proposedly derives from rubisco-assimilated CO2.Rubisco catalyses the addition of CO2 to ribulose 1,5-bisphosphate (RuBP).Since this reaction is essentially the sole carbon source of glucose, 13 C discrimination accompanying CO2 diffusion and subsequent rubisco CO2 assimilation (denoted diffusionrubisco discrimination) is expected to affect all glucose carbon positions equally (Wieloch et al., 2018(Wieloch et al., , 2022c)).
Moreover, we recently measured Δ intramolecularly at all six carbon positions, i, of glucose (Fig. 1A, bottom) extracted across an annually resolved tree-ring series of Pinus nigra (Wieloch et al., 2018).The resultant Δi ' dataset comprises 6*31 values (study period: 1961 to 1995; four years missing: 1977, 1978, 1981, 1982) which were corrected for 13 C signal redistribution by heterotrophic triose phosphate cycling (indicated by prime, Supporting Information Notes S1).We found that, at least, four 13 C signals contribute to the interannual 13 C/ 12 C variability in tree-ring glucose (Fig. 1B) and proposed the following theories on underlying mechanisms.
Hydrogen isotope evidence consistent with these proposed metabolic shifts was reported recently (Wieloch et al., 2022a).Second, the PGI reaction in chloroplasts is usually displaced from equilibrium on the side of F6P whereas the PGI reaction in the cytosol is closer to or in equilibrium (Dietz, 1985;Gerhardt et al., 1987;Leidreiter et al., 1995;Schleucher et al., 1999;Szecowka et al., 2013).This is expected to result in 13 C/ 12 C differences between starch and sucrose at both hexose C-1 and C-2 (Table 2) (Wieloch et al., 2022b).By extension, changes in the relative contribution of starch to the biosynthesis of tree-ring glucose is expected to contribute to the 13 C signals at C-1 and C-2.
In addition to 13 C signals at C-1 and C-2, tree-ring glucose samples discussed here carry deuterium signals caused by metabolic processes at H 1 and H 2 .These signals are strongly correlated and were approximated as where Di denotes relative deuterium abundances at individual H-C positions (Wieloch et al., 2022a,b).
Proposedly, G6PD and PGI-dependent metabolic processes in both leaves and tree rings may contribute to εmet signal introduction (Wieloch et al., 2022b).Interestingly, Wacker (2022) recently reported that the commonly observed whole-molecule deuterium depletion of leaf starch which derives from deuterium depletion at starch glucose H 2 (Schleucher et al., 1999;Wieloch et al., 2022a) is not detectable in nocturnal sucrose.Proposedly, this depletion is either washed out at the level of cytosolic PGI or masked either by the vacuolar sucrose pool or deuterium enrichments at other sucrose hydrogen positions.Washout would imply that any εmet signal present at leaf-level G6P H 2 is lost to the medium.In this case, the εmet signal at tree-ring glucose H 2 may originate outside of leaves.
The 13 C signal at C-5 and C-6 (Fig. 1B) is thought to derive from the postulated (but not yet measured) isotope effects of leaf-level enzymes that modify the carbon double bond in phosphoenolpyruvate (PEP, Fig. 2) (Wieloch et al., 2022c).This includes enolase, pyruvate kinase (PK), PEPC, and 3-deoxy-D-arabinoheptulosonate-7-phosphate synthase (DAHPS), the first enzyme of the shikimate pathway.Breaking the double bond in PEP is thought to proceed faster when 12 C instead of 13 C forms this bond (Wieloch et al., 2022c).Consequently, increasing relative flux into metabolism downstream of PEP is thought to 13 C enrich remaining PEP at the double-bond carbons and their derivatives including glucose C-5 and C-6 (Wieloch et al., 2022c).For example, O3 causes downregulation of rubisco, upregulation of PEPC, and DAHPS expression (Dizengremel, 2001;Janzik et al., 2005;Betz et al., 2009).This is expected to cause increasing relative flux into metabolism downstream of PEP (Wieloch et al., 2022c).Accordingly, we previously found a negative relationship between reconstructed tropospheric O3 concentration and tree-ring glucose Δ5-6' (arithmetic average of Δ5' and Δ6', Table 1) (Wieloch et al., 2022c).
Photorespiration increases with drought which results in increasing supply of mitochondrial NADH via the glycine decarboxylase complex.Since this NADH can feed oxidative phosphorylation, NADH and FADH2 supply by the TCAC which requires injection of PEP into the TCAC via PK and PEPC may be reduced.This should result in Δ5-6' increases counteracting drought-induced decreases in diffusion-rubisco discrimination (see above).
The theories of isotope signal introduction outlined above require further testing.They derive from separate analyses of either the Δi' or deuterium dataset.However, some reactions exhibit both carbon and hydrogen isotope effects (e.g., G6PD at G6P C-1 and H 1 ; PGI at G6P C-1, C-2, and H 2 but not H 1 ) and should therefore introduce intercorrelated 13 C and deuterium signals (suggested terminology: hydro-carbon isotope signals and hydro-carbon isotope fractionation).Combined analysis of intramolecular 13 C and deuterium data can, in principle, help to separate those signals from signals introduced by reactions which merely exhibit either carbon or hydrogen isotope effects.Therefore, we here studied the relationships between Δi' and εmet and their dependence on environmental parameters.Based on our results, we critically examine and revise existing isotope theory and provide new insights into a central open question-whether carbon and hydrogen isotope variability across tree rings derives from leaf-level processes only (as supported by current evidence) or whether processes in the stem contribute as well.

Material and Methods
The Δi' and εmet datasets reanalysed here are described in Wieloch et al. (2018Wieloch et al. ( , 2022b) ) Tank et al., 2002).Air vapour pressure deficit (VPD) was calculated following published procedures (Abtew & Melesse, 2013).Data of the standardised precipitation-evapotranspiration index (SPEIi) calculated for integrated periods of i = 1, 3,6,8,12,16,24,36,48 months were obtained for 48.25° N, 16.25° E (Fan & van den Dool, 2004;Beguería et al., 2010).The SPEI is a multi-scalar drought index approximating soil moisture variability when calculated for short timescales and groundwater variability when calculated for long timescales (Vicente-Serrano et al., 2010).The RAD series starts in 1964 while all other climate series start in 1961.Horizontal distances between the tree site and the climate station and grid point are < 15 km.Vertical offsets are small.Hence, climate data and site conditions are expected to be in good agreement.Analytical procedures are described in Notes S1.

Results
Hydro-carbon isotope signals at tree-ring glucose HC-1 and HC-2 Tree-ring glucose of our Pinus nigra samples exhibits strongly correlated hydrogen isotope signals at H 1 and H 2 (Wieloch et al., 2022b).These signals occur only after crossing a change point in 1980.Isotopeenvironment-relationship analyses indicated that the trees had likely access to groundwater before 1980 which prevented changes of the processes introducing these isotope signals.We proposed the signals derive from the hydrogen isotope effects of G6PD ( H  D ⁄ = 2.97) (Hermes et al., 1982) and PGI (Table 2) (Rose & O'Connell, 1961;Wieloch et al., 2022a) in autotrophic and/or heterotrophic tissue (Fig. 2; 'Introduction') (Wieloch et al., 2022a,b).If this proposal is correct then there should be related signals in Δ1' and Δ2' due to the carbon isotope effects of G6PD affecting C-1 ( C ⬚ 12  C ⬚ 13 ⁄ = 1.0165) (Hermes et al., 1982) and PGI affecting C-1 and C-2 (Table 2) (Gilbert et al., 2012).Several findings support this hypothesis.
Consistent with the findings above (Fig. 3, Table 3), only εmet but not growing season VPD contributes significantly to the Δ2' model whereas only growing season VPD but not εmet contributes significantly to the Δ3' model (Table 4, M2-3).Removing insignificant terms, we find that εmet explains 57% of the systematic variance in Δ2', while growing season VPD explains the entire systematic variance in Δ3' (Table 4, M4-5; Table S4).The effect of VPD on Δ1' is about twice as large as on Δ3' (Table 4, M1 versus M5) while the effect of εmet on Δ1' is about half as large as on Δ2' (M1 versus M4).Intriguingly, Δ1' and Δ3' are affected by processes that respond to growing season VPD.VPD-dependent processes can account for both the clustering and correlation between Δ1' and Δ3' data of 1983 to 1995 (Fig. 4A).By contrast, εmet is significantly correlated only with PRE (especially March to July PRE) but no other climate parameter (Table S5; Table 4, M11).Furthermore, in our Δ1' and Δ2' models, εmet can be substituted by March to July PRE (Table 4, M1 versus M6, M4 versus M7).
Compared to the late period, we found fewer and weaker isotope-climate correlations (Table 5).

Discussion
Intramolecular carbon isotope analysis of tree-ring glucose yields information about metabolic variability and water status of both leaves and stems We found evidence for processes affecting Δ1' and Δ3' which respond to VPD (Table 4, M1, M3, M5).
Intriguingly, we also found evidence for processes simultaneously affecting εmet, Δ1', and Δ2' which respond to PRE but not VPD (Table S5; Table 4, M1, M2, M4, M6, M7, M11).This sensitivity to different hydrological properties may be explained by the fact that stem capacitance can buffer stem water status against changes in VPD (McCulloh et al., 2019), whereas leaf water status is tightly coupled to VPD (Grossiord et al., 2020).
Changes in PRE will affect soil water potential and hence both stem and leaf water status.Variability in leaf water status may be impacted more by VPD than by soil water status which would explain why VPD is the best predictor of the intercorrelated processes affecting Δ1' and Δ3'.By contrast, VPD-insensitive processes affecting εmet, Δ1', and Δ2' may reside in stems.Hence, we propose intramolecular carbon and hydrogen isotope analysis of tree-ring glucose yields information about metabolic variability and water status not only of leaves but also of stems.PRE-dependent systemic changes in enzyme expression can be considered as an alternative explanation.

Isotope fractionation mechanisms in leaves affecting tree-ring glucose C-1 to C-3
The Δ1-2' and Δ1-3' series exhibit change points in 1980, i.e., their frequency distributions do not align with the properties of a single theoretical probability distribution ('Results'; Tables S1-2).Consequently, we investigated the early (1961 to 1980) and late period (1983 to 1995) separately.During the late period, Δ1' and Δ3' are significantly intercorrelated (Fig. 4A) and correlate negatively with VPD and positively with shortterm SPEI whereas Δ2' lacks most of these correlations (Table 3).Furthermore, during the late period, growing season VPD accounts for a significant fraction of the systematic variance in Δ1' and the entire systematic variance in Δ3' but does not contribute significantly to explaining Δ2' (Table 4, M1, M2, M5; Table S4).Hence, increasing VPD during 1983 to 1995 causes 13 C enrichments at tree-ring glucose C-1 and C-3 but not C-2.At C-1, the effect is about twice as large as at C-3 (Table 4, M1 and M5).
As discussed above, the VPD-dependent processes affecting Δ1' and Δ3' are likely located in leaves.
Qualitatively, VPD-induced 13 C enrichments at C-1 and C-3 are consistent with the mechanisms of diffusionrubisco fractionation (see 'Introduction').However, diffusion-rubisco fractionation affects all glucose carbon positions equally (Wieloch et al., 2018).Hence, the unequal VPD response of Δ1', Δ2', and Δ3' points to post-rubisco fractionations.In the following, we assume Δ3' variation derives entirely from diffusion-rubisco fractionation and argue VPD-dependent isotope fractionation at PGI and G6PD in leaf chloroplasts and the cytosol may exert additional control over Δ1' and Δ2' variability.Generally, variability in PGI fractionation depends on three biochemical properties: (i) the equilibrium status of the PGI reaction, and relative flux of the PGI reactants (ii) F6P and (iii) G6P into competing metabolic pathways (Figs. 2 and 5): (i) PGI reversibly converts F6P into G6P (Fig. 5A).Under nonstress conditions, the PGI reaction in chloroplasts is strongly displaced from equilibrium on the side of F6P (Dietz, 1985;Gerhardt et al., 1987;Kruckeberg et al., 1989;Schleucher et al., 1999;Wieloch et al., 2022a;Wieloch, 2022).With decreasing pi, however, the reaction moves towards equilibrium (Dietz, 1985;Wieloch et al., 2022a).This shift is accompanied by 13 C increases at C-1 and C-2 of G6P (Table 2) which will be transmitted to downstream derivatives such as starch and tree-ring glucose (Wieloch et al., 2018).In isohydric species such as Pinus nigra, pi decreases with drought due to stomatal closure (McDowell et al., 2008).Here, we found stronger VPD-induced 13 C increases at tree-ring glucose C-1 than at C-3.This is consistent with the PGI-related isotope shift expected at C-1.However, the apparent absence of the diffusion-rubisco signal from C-2 contrasts with the expected isotope shift.That said, in Phaseolus vulgaris, the ratio of leaf sucrose-to-starch carbon partitioning was shown to increase steeply with decreasing pi (Sharkey et al., 1985).Hence, the relative contribution of chloroplastic G6P and its isotope composition to downstream metabolism may decrease with increasing VPD reducing the influence of the mechanism described on Δ1' and Δ2' variation.
Furthermore, in the dark, the cytosolic PGI reaction was found to be near equilibrium (Gerhardt et al., 1987).
Consequently, F6P would be 13 C depleted at C-1 but 13 C enriched at C-2 relative to the corresponding G6P positions (Table 2).Increasing relative F6P flux into mitochondrial respiration would then result in 13 C increases at C-1 and 13 C decreases at C-2 of G6P and downstream derivatives.Thus, this mechanism is consistent with both observations, stronger VPD-induced 13 C increases at tree-ring glucose C-1 compared to C-3 and the apparent absence of the diffusion-rubisco signal from C-2.
(iii) While carbon assimilation commonly decreases with drought (McDowell et al., 2008), the activity of leafcytosolic G6PD increases (Fig. 5C; Landi et al., 2016).This can be expected to result in increasing relative G6P flux into the OPPP.While some authors reported that the cytosolic PGI reaction in illuminated leaves is in equilibrium (Gerhardt et al., 1987) others found displacements from equilibrium (Leidreiter et al., 1995;Schleucher et al., 1999;Szecowka et al., 2013).Hence, PGI-related isotope shifts in tree-ring glucose resulting from G6P flux into the leaf-cytosolic OPPP are hard to predict (Table 2).By contrast, the unidirectional conversion of G6P to 6-phosphogluconolactone catalysed by G6PD proceeds faster with 12 C- (Hermes et al., 1982).Hence, increasing relative flux through the leaf-cytosolic OPPP may contribute to the stronger VPD-induced 13 C increases at tree-ring glucose C-1 compared to C-3.
Aside from these mechanisms, there are others that might introduce Δ1' and Δ2' variation.For instance, we recently reported evidence consistent with increasing relative flux through the chloroplastic OPPP in response to decreasing pi under illumination (Fig. 5A; Wieloch et al., 2022aWieloch et al., , 2023)).Furthermore, under illumination, chloroplastic F6P is used for both RuBP regeneration and starch biosynthesis (Fig. 5A).
Increasing VPD promotes photorespiration resulting in increasing RuBP regeneration relative to carbon export from the Calvin-Benson cycle into sinks such as starch.
The mechanisms described above should also introduce hydrogen isotope signals because of the hydrogen isotope effects of G6PD affecting G6P H 1 (Hermes et al., 1982) and PGI affecting G6P H 2 (Table 2, Fig. 5).
However, growing season VPD neither correlates with εmet pertaining to tree-ring glucose H 1 nor H 2 (Tables S5-6).Hence, either G6PD and PGI are not the sources of VPD-dependent carbon isotope fractionation in Δ1' (and Δ2') or the corresponding hydrogen isotope signals were washed out after introduction.Washout at H 1 may occur during equilibration of F6P with mannose 6-phosphate by phosphomannose isomerase (cf.Topper, 1957).Similarly, complete washout at H 2 may occur when the leaf-cytosolic PGI reaction is in equilibrium (Notes S3).Previously, this latter process was invoked (among others) to explain why a wholemolecule deuterium depletion observed in leaf starch was not transmitted to nocturnal sucrose ('Introduction'; Wacker, 2022).As each conversion by PGI was found to be associated with a 0 to 50% probability for hydrogen exchange with the medium (Noltmann, 1972), partial washout of existing hydrogen isotope signals may also occur under non-equilibrium conditions (Notes S3).
In the mechanisms described above, we assumed diffusion-rubisco fractionation contributes to VPDdependent Δi' variation.However, diffusion-rubisco fractionation affects all glucose carbon positions equally (Wieloch et al., 2018).Since merely two out of six glucose carbon positions carry VPD-dependent isotope variation, the question arises of whether the diffusion-rubisco signal was already below the detection level on introduction.If this were the case, then VPD-dependent Δ1' and Δ3' variation would originate entirely from post-rubisco processes.Furthermore, post-rubisco processes that were previously invoked to explain the absence of the diffusion-rubisco signal from C-4, C-5, and C-6 (see 'Introduction') would not occur.

Isotope fractionation mechanisms in stems affecting tree-ring glucose HC-1 and HC-2
Previously, we found a change point in εmet in 1980 (Wieloch et al., 2022b).Here, we found the same change point in Δ1-2' ('Results').Consistent with this, Δ1' and Δ2' data of 1983 to 1995 exhibit a significantly lower average value and a significantly larger variance than those of 1961 to 1980 (Tables S1-2).
Furthermore, Δ1' and Δ2' are significantly correlated during the late (Fig. 4B) but not the early period (Fig. 4A), and εmet accounts for a significant fraction of the variance of both Δ1' and Δ2' during the late period ( Previously, we reported evidence suggesting the groundwater table before 1980 was high enough to prevent metabolic changes causing εmet variation (Wieloch et al., 2022b).By extension, this should also explain the properties of Δ1' and Δ2' listed above.That is, since the trees had access to groundwater during the early period, metabolic shifts that can cause intercorrelated variation in εmet, Δ1', and Δ2' were not induced.
Processes causing intercorrelated variation in εmet, Δ1', and Δ2' are probably located in the stem (see the first section of the 'Discussion').The εmet signal is present at glucose H 1 and, considerably more strongly, at H 2 (range: 64‰ and 240‰, respectively; 1983 to 1995).In the biochemical pathway leading to tree-ring cellulose, PGI is the last enzyme acting on precursors of glucose H 2 (Figs. 2 and 5D).With each conversion by PGI, there is a probability for hydrogen exchange with the medium of 0 to 50% (Noltmann, 1972).Thus, if we assume Pinus nigra stem PGI exchanges hydrogen with the medium as does spinach leaf PGI (Fedtke, 1969) and the reaction is in equilibrium, then any deuterium signal at G6P H 2 will be washed out.
Among all H-C positions of tree-ring glucose, the deuterium abundance at H 2 is neither exceptionally high nor low during 1961 to 1980 whereas it is exceptionally high (and exceptionally variable) during 1983 to 1995 (Fig. 6).This indicates that the PGI reaction was close to or in equilibrium during 1961 to 1980 but displaced from equilibrium on the side of G6P during 1983 to 1995 (Table 2).Additionally, shifts of the PGI reaction away from equilibrium towards the side of G6P should cause 13 C enrichment at G6P C-1 and C-2 (Δ1' and Δ2' decreases), and Δ2' should decrease 3 times more than Δ1' (Table 2).Consistent with this, we found negative relationships between εmet and Δ1', as well as Δ2' (Table 4, M1 and M4).However, Δ2' decreases only 1.88 times more than Δ1', but this best estimate is associated with a relatively large error (SE interval: 1.04 to 3.45).That said, the offset from 3 is likely explained by increasing relative flux through the OPPP accompanying the putative PGI reaction shift (Figs. 2 and 5D).This is because G6P to 6phosphogluconolactone conversion by G6PD exhibit 13 C and D isotope effects ( C ⬚ 12  C ⬚ 13 ⁄ = 1.0165,  H  D ⁄ = 2.97) (Hermes et al., 1982).Hence, increasing relative OPPP flux causes 13 C enrichment at G6P C-1 (Δ1' decreases) and deuterium enrichment at G6P H 1 .This is consistent with both the apparently decreased PGI effects ratio (1.88 instead of 3) and, more importantly, εmet increases at glucose H 1 of up to 64‰ during 1983 to 1995 (Fig. 6).
Sucrose translocated from leaves can be split into UDP-glucose and fructose via sucrose synthase or glucose and fructose via invertase (Fig. 2).UDP-glucose entering tree-ring cellulose biosynthesis directly via sucrose synthase is protected from isotope fractionation by PGI and G6PD.However, in stems of juvenile Quercus petraea and Picea abies, at least 79% and 43% of the precursors of tree-ring glucose went through PGI catalysis, respectively (Augusti et al., 2006).Theoretically, shifts of the PGI reaction away from equilibrium towards the side of G6P can cause εmet increases at glucose H 2 of up to 611‰ (Notes S3, hydrogen exchange with the medium not considered).With 43% and 79% of all precursors of tree-ring glucose undergoing PGI catalysis, εmet increases at glucose H 2 of up to 263‰ and 483‰ are possible, respectively.Thus, the PGI-related fractionation mechanism proposed here can potentially cause previously reported εmet increases at glucose H 2 of up to 240‰ (Wieloch et al., 2022b).Shifts in sucrose cleavage by sucrose synthase versus invertase may exert additional control on the εmet signal at glucose H 2 .
Based on results and interpretations presented above, decreasing stem water content is associated with both increasing OPPP flux and a shift of the PGI reaction away from equilibrium towards the side of G6P corresponding to low relative F6P concentration (Fig. 5D).We propose these concerted shifts may ensure redox homeostasis and balanced substrate supply to glycolysis as follows.In heterotrophic tissue, NADPH from the OPPP is believed to be central for maintaining redox homeostasis (Fig. 2) (Stincone et al., 2015).
Flux through the OPPP is regulated at G6PD.Heterotrophic G6PD activity reportedly increases with drought (Liu et al., 2013;Wang et al., 2016Wang et al., , 2020)), oxidative load (Wang et al., 2016(Wang et al., , 2020;;Li et al., 2020), NADPH demand (Wendt et al., 2000;Esposito et al., 2001;Castiglia et al., 2015), and abscisic acid concentration (Cardi et al., 2011;Wang et al., 2016).Decreasing stem water content may cause increasing OPPP flux via increasing abscisic acid concentration (Brunetti et al., 2020), and possibly increasing oxidative load increasing the demand for NADPH.In turn, increasing OPPP flux results in increasing supply of pentose phosphates which may feed into glycolysis via the reductive part of the pentose phosphate pathway (Figs. 2 and 5D).This would reduce the demand for glycolytic substrates supplied via PGI.The shift of the PGI reaction away from equilibrium towards the side of G6P may reflect this decreased demand and result from PGI downregulation by intermediates of the pentose phosphate pathway such as erythrose 4-phosphate, ribulose 5-phosphate, and 6-phosphogluconate (Parr, 1956;Grazi et al., 1960;Salas et al., 1965).
Furthermore, relative changes in G6P-to-F6P supply versus consumption may contribute to the shift of the PGI reaction.For instance, while starch storage consumes G6P, remobilisation supplies G6P (Noronha et al., 2018).Under drought, the storage-to-remobilisation balance may tilt towards remobilisation (Mitchell et al., 2013;Thalmann & Santelia, 2017;Tsamir-Rimon et al., 2021).Consequently, the PGI reaction may move towards the side of G6P.Similarly, we previously reported below-average tree-ring widths for years in which the PGI reaction is on the side of G6P (Wieloch et al., 2022b).Hence, in these years, G6P consumption by growth may have been reduced while F6P consumption by downstream metabolism may have been maintained.

to 1980: A period of homeostasis with respect to processes affecting Δ1', Δ2', Δ3', and εmet
During 1961 to 1980, Δ1', Δ2', and Δ3' are not significantly correlated which contrasts with the period 1983 to 1995 (Figs. 4a-b).Similarly, relationships of Δ1' and Δ3' with VPD and Δ1' and Δ2' with PRE observed during the late period are largely absent during the early period (Tables 3-5) even though there is no difference in the magnitude of VPD and PRE variability between these periods (Fig. S1).Like Δ1-2' and Δ1-3', εmet exhibits a change point in 1980 and responds to PRE after but not before 1980 (Wieloch et al., 2022b).This shift in εmet sensitivity was attributed to long-term drought which intensified over the study 13 period and proposedly lead to a groundwater depletion below a critical level in 1980 (Wieloch et al., 2022b).
By extension, this groundwater depletion might also explain the insensitivity of Δ1 ' andΔ3' to VPD andΔ1' andΔ2' to PRE during 1961 to 1980 and their sensitivity from 1983 onwards.Thus, while the trees had access to groundwater, leaf-and stem-level processes affecting Δ1', Δ2', Δ3' and εmet could apparently maintain homeostasis despite changing atmospheric conditions.

Isotope fractionation mechanisms in leaves affecting tree-ring glucose C-4
As for Δ5' and Δ6', no change point was detected in Δ4' ('Results'; Tables S1-2).Considering the entire study period, Δ4' is weakly associated with Δ5-6' (Fig. 1B).Consistent with this, the Δ5-6'-climate model works reasonably well for Δ4' considering the relatively low systematic variance in Δ4' of 38% (Table 7, M1 and M6; Table S4).Introduction of the Δ4' and Δ5-6' signals proposedly involves leaf-level consumption of PGA and PEP by downstream metabolism, respectively (Wieloch et al., 2021(Wieloch et al., , 2022c)).Since PGA is a precursor of PEP (Fig. 2), our previously proposed theories of signal introduction are in line with the observation that Δ4', Δ5' and Δ6' are associated and respond to the same environmental parameters.

Conclusions and future directions
Dual-isotope-environment analysis was used to deconvolute isotope signals and provide several new insights into plant isotope fractionation.First, the diffusion-rubisco signal was previously shown to be absent from tree-ring glucose C-4 to C-6 (Wieloch et al., 2021(Wieloch et al., , 2022c) but believed to be present at C-1 to C-3 (Wieloch et al., 2018).Here, this signal was found to also be absent from C-2.Second, isotope fractionation beyond leaves is commonly considered to be constant for any given species (Roden et al., 2000;Gagen et al., 2022).However, our results suggest a significant part of the carbon and hydrogen isotope variation in tree-ring glucose originates in stems from processes affecting Δ1', Δ2', and εmet simultaneously.Third, VPD affects Δ1' and Δ3' and PRE affects Δ1', Δ2', and εmet (Table 4).These relationships proposedly reflect water content variability in leaves and stems, respectively.They apply to the late but not the early study period consistent with the finding of a change point in both the εmet (Wieloch et al., 2022b) and Δ1-3' series (see above).This change point proposedly marks the crossing of a physiologically relevant groundwater threshold (Wieloch et al., 2022b).Additionally, we reported Δ5-6' relationships with RAD and TMP which apply to the entire study period (Table 4).These latter relationships are consistent with previously proposed isotope fractionation mechanisms (Wieloch et al., 2022c).By contrast, we here revised and expanded our previous theory on the mechanisms introducing Δ1', Δ2', Δ3', and εmet variability.Given the multitude of isotope-environment relationships (including change-point responses), intramolecular carbon isotope analysis has a remarkable potential for reconstructions of environmental conditions (VPD, PRE, RAD, TMP, soil moisture, groundwater thresholds, tropospheric O3 concentration), tissue water content (leaf, stem), metabolic flux variability (various processes), and ecophysiological properties such as intrinsic water-use efficiency across space and time.Complementing hydrogen isotope analysis is expected to significantly enhance these capabilities.
Understanding isotope fractionation mechanisms is central for retrospective studies of plant physiology and climate based on tree-ring isotope data, and there is considerable room for improvement as shown above.

MAMJ
Climate data were averaged for all ≥ 4-month periods of the growing season (March to November).Months were abbreviated by their initial letters.Δ4', Δ5', and Δ6' denote intramolecular 13 C discrimination at glucose C-4, C-5, and C-6, respectively.Glucose was extracted across an annually resolved tree-ring series of Pinus nigra.Glucose was extracted across an annually resolved tree-ring series of Pinus nigra from the Vienna Basin.(black squares, 1961 to 1980; green circles, 1983 to 1995).Dashed line, relationship between the hydrogen and carbon isotope data of the period 1983 to 1995.

Figure 5 .
Figure 5. Processes invoked to explain isotope fractionation at tree-ring glucose HC-1 and HC-2: (A) in leaf chloroplasts under illumination, (B) in the leaf cytosol in the dark, (C) in the leaf cytosol under illumination, and (D) in the stem cytosol.F6P and G6P carbon atoms 1 to 6 occur in sequentially order from top to bottom.Atom positions affected by G6PD and PGI fractionation are given in blue and green, respectively.In some cases, carbon position 1 is given both as blue letter and green dot to indicate fractionation at both enzymes.Dashed arrows indicate that intermediate reactions are not shown.Wavy lines indicate fractional introduction of hydrogen from water by the PGI reaction.Note, G6PD in stem leucoplasts may additionally contribute to isotope fractionation at tree-ring glucose C-1 and H 1 .Abbreviations: F6P, fructose 6-phosphate; G6P, glucose 6-phosphate; G6PD, G6P dehydrogenase; OPPP, oxidative pentose phosphate pathway; pi, intercellular CO2 partial pressure; PGI, phosphoglucose isomerase; PRE; precipitation; RuBP, ribulose 1,5-bisphosphate.

Figure 6 .
Figure 6.Average intramolecular δDi patterns of the periods 1961 to 1980 and 1983 to 1995 (black and blue, respectively).The data were acquired for tree-ring glucose of Pinus nigra laid down at a site in the Vienna basin.The figure shows discrete data.Dashed and dotted lines were added to guide the eye.Data reference: Average deuterium abundance of the methyl-group hydrogens of the glucose derivative used for NMRS measurements.Modified figure from Wieloch et al. (2022b).
and in Notes S1.

Relationship between the average hydrogen isotope fractionation caused by 776 metabolic processes at glucose H 1 and H 2 (εmet) and 13
C discrimination at C-1, C-2, and C-3 777 (Δ1', Δ2', and Δ3').Glucose was extracted across an annually resolved tree-ring series of Pinus