The size and the age of the metabolically active carbon in tree roots

Little is known about the sources and age of C respired by tree roots. Previous research in stems identified two functional pools of non-structural carbohydrates (NSC): an “ active ” pool supplied directly from canopy photo-assimilates supporting metabolism and a “ stored ” pool used when fresh C supplies are limited. We compared the C isotope composition of water-soluble NSC and respired CO 2 for aspen roots ( Populus tremula hybrids) cut off from fresh C supply after stem-girdling or prolonged incubation of excised roots. We used bomb radiocarbon to estimate the time elapsed since C fixation for respired CO 2 , water-soluble NSC and structural α -cellulose. While freshly excised roots (mostly <2.9 mm in diameter) respired CO 2 fixed <1 year previously, the age increased to 1.6 – 2.9 year within a week after root excision. Freshly excised roots from trees girdled ~3 months ago had respiration rates and NSC stocks similar to un-girdled trees but respired older C (~1.2 year). We estimate that over 3 months NSC in girdled roots must be replaced 5 – 7 times by reserves remobilized from root-external sources. Using a mixing model and observed correlations between Δ 14 C of water-soluble C and α -cellulose, we estimate ~30% of C is “ active ” (~5 mg C g (cid:1) 1 )

The "bomb-radiocarbon" approach permits quantification of respiration sources and NSC dynamics in mature trees by tracing excess 14 C created during atmospheric nuclear weapons tests that nearly doubled background values of radiocarbon signature (Δ 14 C) of atmospheric CO 2 (Δ 14 C atm ) in the early 1960s (Trumbore, Sierra, & Pries, 2016).Since then, decline in Δ 14 C atm reflect uptake of bombradiocarbon into the oceans and terrestrial biosphere, and dilution by radiocarbon-free CO 2 originating from fossil fuel emissions (Levin & Hesshaimer, 2016).Since 1964, an annual unique Δ 14 C atm signature is transferred to biomass C via photosynthesis.A comparison of the Δ 14 C signature of any C pool in plants to the current year's Δ 14 C atm thus provides a means to estimate the mean time elapsed since the C was fixed.
Plants use soluble sugars both to transport C (predominantly sucrose) and as primary substrates in cellular metabolism (e.g.glucose and fructose).Starch is insoluble and thus cannot be translocated but can serve as local storage.These different properties are often mistakenly implied to reflect differences in turnover time, with sugars are assumed to comprise a fast-cycling pool and starch a slow-cycling pool (Dietze et al., 2014).However, in tree stems and large roots no systematic differences in radiocarbon signatures have been observed between the water-soluble (representing sugars) and insoluble (starch) C fractions, suggesting fast C exchange between the two pools (Richardson et al., 2013;Richardson et al., 2015;Trumbore, Czimczik, Sierra, Muhr, & Xu, 2015).
Nevertheless, NSC pools are not uniform in age.The Δ 14 C of stem respired CO 2 (Δ 14 C resp ) was lower than the water-soluble Δ 14 C (Δ 14 C ws ) in the outermost 2 cm of the stem, reflecting age differences of ~10 year (Carbone et al., 2013).Combined with evidence from 13 Clabelling of fresh photo-assimilates from the canopy, it is clear that some young sugars support respiration, while other, older, sugars are stored intact (Epron et al., 2011).Accordingly, C allocation models usually distinguish two functional NSC-sub pools: a "fast" pool of recently fixed C (assumed age < 1 year) supporting metabolism and respiration, and a "slow" NSC pool for storage (Herrera-Ramirez et al., 2020;Richardson et al., 2015).However, we are still missing ways to quantify and differentiate between these two pools.
Starch is usually slightly enriched (~1‰) while lipids are more depleted (~À5‰) in 13 C compared to sugars (Bowling, Pataki, & Randerson, 2008), and are expected to accordingly affect the δ 13 C signature of the respired CO 2 (δ 13 C resp ) when decarboxylated.The "apparent isotopic fractionation" of respiration (Δ R ) is defined as the difference in δ 13 C between the putative respiratory substrate (usually sugars) and δ 13 C resp (Ghashghaie et al., 2003).For roots of woody C 3 plants mainly negative Δ R values (δ 13 C substrate < δ 13 C resp ) have been reported (Ghashghaie & Badeck, 2014).The 13 C enrichment in respired CO 2 is commonly explained by the dominance of CO 2 emitted from decarboxylation of pyruvate via pyruvate dehydrogenase (PDH) over CO 2 emitted from the tricarboxylic acid (TCA) cycle reactions (Tcherkez et al., 2003).The activity of phosphoenolpyruvate carboxylase (PEPC), which re-fixes CO 2 to replenish TCA cycle intermediates, was suggested to decrease respired CO 2 enrichment and increase Δ R (Gessler et al., 2007;Ghashghaie & Badeck, 2014).To our knowledge, only one study to date has evaluated Δ R in roots of mature trees (Gessler et al., 2007).
Here, we investigated the sources of respiration for fine and coarse roots of mature temperate aspen trees (Populus tremula hybrids).We performed a stem-girdling experiment (complete circumferential removal of bark, cambium, and phloem), terminating belowground transport of fresh photo-assimilates and forcing mobilization of storage NSCs for survival beneath the girdling.Girdled Amazonian tree stems quickly shifted from using current year photosynthetic products as respiratory substrates in un-girdled trees to C fixed ~5 year previously a month after girdling (Muhr, Trumbore, Higuchi, & Kunert, 2018).We expected increases in Δ 14 C resp for roots of girdled trees, but with potentially smaller magnitude, since trees from the genus Populus often have root connections that enable C transfer from healthy trees (Gaspard & DesRochers, 2020;Pregitzer & Friend, 1996).To characterize the usage of NSC reserves in the roots themselves we conducted additional experiments where the age of C respired from detached roots was followed for a week.This experiment is also the base for our first approach to estimate the size and isotopic signature of the fast-cycling "active" C pool that feeds respiration and metabolism.The relationship between Δ 14 C ws and the Δ 14 C signature of the α-cellulose (Δ 14 C cell , representing structural C) allowed us to develop the second approach to estimate the size and isotopic signatures of the "active" C, as well as the slow-cycling "stored" C pool.

| Study site and experimental design
We sampled 12 Eurasian aspen trees (Populus tremula hybrids) with estimated age of 60-70 year growing in a forest stand located on a slope of the Großer Hermannsberg Mountain (867 m a.s.l), Germany (50 42 0 50 00 N, 10 36 0 13 00 E, site elevation 616 m a.s.l).Soils at the site are developed on volcanic parent rock, mean annual temperature is ~7 C and annual rainfall is 800-1,200 mm (Bouriaud, Marin, Bouriaud, Hessenmoller, & Schulze, 2016).During our field campaign in the summer of 2018, an extreme "hot drought" occurred in central Europe (Bastos et al., 2020), also observed in a nearby weather station situated 812 m a.s.l (https://www.bgc-jena.mpg.de/freiland/index.php/Sites/Hermannsberg).During the 2018 growing season (May and October), mean monthly temperatures were warmer by 1.5-3.5 C than those for the years 2010-2017, while total monthly rainfall was 260 mm, compared to the average of 323 mm in the years 2010-2017 (Figure 1).
The size of the forest stand where our study took place is ~40 Â 40 m 2 with ca.20 mature and 50 younger trees.Control and girdled trees were all mature with diameter at breast height of 87-129 cm, and interspersed.Distances between girdled trees and the closest un-girdled trees, including unstudied trees, were 1.2-5.8m.
After sampling all 12 trees on 26 June and 4 July, six of the trees were girdled on July 4th by removing a ~4 cm wide band of bark, cambium and phloem from the stem 1.5 m above the ground.During subsequent samplings (25 September and 2 October), we differentiated "Girdling" and "Control" trees.
To ensure we sampled roots specific to the chosen trees, we tracked their connections back to the main root or stem.Roots were collected from the top 10 cm of the mineral soil.We aimed at finding two root clusters for each tree with total mass of several grams.In most trees, we harvested individual roots growing directly from large lateral roots to gain the desired mass.Hence, a sample from a single tree usually combined roots from several spatial locations.The samples were put on ice immediately after harvest until analysis.On the subsequent day, the roots were thoroughly washed to remove any remaining soil particles and then separated into two size classes: coarse roots (> 2 mm, mean 2.9 mm, max 6 mm) and fine roots (≤ 2 mm, almost all were suberized); each size fraction consisted of roots from different orders, and not necessarily from the same root cluster.
Each size class was split for (a) NSC and α-cellulose extractions, and (b) respiration incubations (details below).
Incubations of excised roots lasted for 2 days and integrated CO 2 respired between the start of incubation (time 0) to day 2. To test roots' ability to utilize their own C reserves for metabolism during C starvation, we made additional incubations that sampled integrated respiration between days 6 and 8.We report these using the mean incubation times (e.g., days 1 and 7).In addition, we performed short-term incubations (up to 1.5 hr) on days 0, 2 and 6 to measure the respiratory quotient (RQ), that is, the ratio CO 2 efflux/O 2 influx.The RQ is mainly defined by the respiratory substrate, which in plants is assumed to be carbohydrates with RQ = 1.Compounds more oxidized than carbohydrates like organic acids are expected to yield RQ > 1, whereas amino acids and lipids yield RQ values of 0.9 and 0.7, respectively (Masiello, Gallagher, Randerson, Deco, & Chadwick, 2008).Thus, RQ measurements can indicate the substrate used for respiration.For the correction of CO 2 efflux rates measured at laboratory temperatures (22 C) to field temperatures, we conducted short-term incubations at different temperatures to calculate Q 10 , the factor by which CO 2 efflux increases with 10 C warming.
Because regional fossil fuel emissions have the potential to affect Δ 14 C atm , we collected additional samples at the site to reconstruct the recent history of local Δ 14 C atm from tree rings.Two stem cores from nearby tree were extracted in 2019 using a 5.15 mm increment borer.
We visually identified the annual rings for the last 9 years and sampled their outermost halves, which presumably are dominated by latewood produced during midsummer-autumn mainly from C fixed in the current growing season, thus approximating the current year's Δ 14 CO atm (Kudsk et al., 2018;Pilcher, 1995).Samples from identical rings from both stem cores were pooled for α-cellulose extraction and Δ 14 C cell analysis (Hoper, McCormac, Hogg, Higham, & Head, 2016).In addition, assuming leaves respire C fixed recently, we analysed Δ 14 C resp from aspen leaves as a proxy for Δ 14 C atm during the 2019 growing season.
We combined our data with the mean Δ 14 C atm of the northern hemisphere zone 1 during the growing season (May-October) as published by Hua, Barbetti, and Rakowski (2016).We further compared our estimation for 2018 Δ 14 C atm with direct measurements of atmospheric air in different sites in Europe that are part of ICOS (the Integrated Carbon Observation System, see Table S1 for PID numbers).

| NSC analysis
Aliquots of 50 mg oven-dried root tissue (≥2 d at 60 C), were analysed for (a) sugars and starch concentrations, (b) Δ 14 C ws , (c) δ 13 C ws, and (d) Δ 14 C cell .Since Δ 14 C ws is usually correlated with the age of the containing tissue (Furze et al., 2020;Richardson et al., 2015;Trumbore et al., 2015), for correct interpretation of Δ 14 C ws variance it is important to account to the root's mean C age estimated using Δ 14 C cell .Therefore, we coarsely cut rather than milled the roots since in the α-cellulose extraction the sample is rinsed with reagents through 16-40 μm mesh that might allow fine material to pass and clog the system's tubing (Steinhof, Altenburg, & Machts, 2017).We decided only after the pre-girdling campaign to measure starch concentration, hence data are available only for roots from the girdled and control trees.
The methods for NSC, sugar and starch extractions are based on protocols S1 and S2 from Landhausser et al. (2018) with some modifications necessary for minimizing extraneous C additions that would affect the 14 C measurement.We avoided plastic vials that contain C and used only glass vials that were pre-baked at 550 C to eliminate any C residuals.For sugar extraction we used water as a solvent instead of ethanol, which is slightly more efficient (Landhausser et al., 2018), to avoid possible C addition from the ethanol.To reduce starch dissolution in water, the extraction temperature was lowered from 90 C to 65 C. Extraction of water-soluble C from the 50-mg samples was carried out by shaking the samples in 5 ml deionized water for 10 min (65 C), in three repetitions.After each repetition the vials were cooled, centrifuged and the supernatant was transferred to water-soluble glucans and then to glucose by amyloglucosidase (Sigma cat.no.ROAMYGLL) (Landhausser et al., 2018).The remaining pellet was used for Δ 14 C cell analysis (Steinhof et al., 2017).
To measure the concentrations of the soluble sugars and the glucose hydrolysate from the starch digestion we used high-performance anion-exchange chromatography with pulsed amperometric detection (HPLC-PAD) device equipped with autosampler (Dionex ® ICS 3000, Thermo Fisher GmbH, Idstein, Germany) (Raessler, Wissuwa, Breul, Unger, & Grimm, 2010).The starch and sugar concentrations are calculated as glucose-equivalent weight per sample dry weight (mg glucose g À1 ) (Landhausser et al., 2018), and further multiplied by 0.4 to express as C mass per dry weight (mg C g À1 ).
Then the set-up was closed quickly by connecting two flasks (pre-filled with the same synthetic air) at both ends of the chamber.Incubations were conducted at room temperature (22 C) in the dark and ended by closing the flask valves.Leaf incubations aimed to estimate the local Δ 14 CO atm , used the same setup and lasted 1 day.
Incubations were conducted at room temperature that was higher than average field soil temperatures.While direct effects on substrate identity and thus age (Δ 14 C resp ) are not expected initially, higher temperatures could lead to faster depletion of "fast" sources and thus ageing of the substrate pools, expected to affect both Δ 14 C resp and δ 13 C resp (Tcherkez et al., 2003).In addition, the temperature-induced metabolic change can affect δ 13 C resp directly reflecting temperature sensitivity of the processes emitting CO 2 (Kodama et al., 2008).Thus, the δ 13 C resp values should be regarded as room-temperature acclimated values and not extrapolatable to field values.

| Short-term incubations
Short-term incubations were done in a solid aluminium chamber equipped with an NDIR CO 2 sensor (COZIR 0%-1% CO 2 Sensor, CO 2 Meter, Inc., Ormond Beach, FL) and a quenching-based O 2 sensor (LuminOx, Coatbridge, UK).CO 2 and O 2 fluxes were calculated by linear fit of the concentration change with time, where the slope is equivalent to the term ΔCO 2 /I t in Equation ( 2).Incubations with R 2 < 0.9 in any of the gases were discarded.
Flask-measured CO 2 efflux rates (measured at room temperature) were corrected to field (in situ) temperatures using Q 10 values estimated using incubations of roots collected in the September-October at two or three different temperatures (ranging between 5 C and 22 C).We used the R package respirometry that fits the measured CO 2 efflux (R) at a given temperature (T) with the Equation R = a Â e b Â T and calculates Q 10 = e 10 Â b .We assume the computed Q 10 value is also valid for the June-July campaign following Burton and Pregitzer (2003) who observed little to no seasonal temperature acclimation in fine roots.Soil temperature in the top 10 cm where the roots were collected is coupled to changes in air temperature with some lag time (Brown, Pregitzer, Reed, & Burton, 2000).To estimate air temperature at the site we added 1.6 C to the measured temperature at the weather station due to altitude difference (196 m) and assumed mean lapse rate of 0.8 C every 100 m.Soil temperature (0-10 cm) was estimated as the average air temperature over the previous 7 days.

| Radiocarbon analysis
For radiocarbon analysis, respired CO 2 from the flasks or CO 2 from combusted solid samples was cryogenically purified and graphitized on iron in the presence of H 2 at 550 C (Muhr et al., 2018;Steinhof et al., 2017).
The graphitized samples were analysed by accelerator mass spectrometry (AMS; Micadas, Ionplus, Switzerland) in the radiocarbon laboratory in Jena, Germany (Steinhof et al., 2017).Radiocarbon data are expressed as Δ 14 C (‰) and calculated according to Trumbore et al. (2016): where R À25 is the sample's 14 C/ 12 C ratio corrected for mass dependent fractionation by normalizing the sample's δ 13 C to a δ 13 C of À25‰.R oxalic,-19 is the 14 C/ 12 C ratio in the standard, oxalic acid, normalized to δ 13 C of À19‰, and the 0.95 term converts to the absolute radiocarbon standard (1890 wood) activity in 1950.The exponent corrects for decay of 14 C in the standard between 1950 and the year of measurement (x), to provide the absolute amount of 14 C in our samples.For individual tree rings x matched the estimated year.
The measurement precision for each AMS measurement of 14 C was 2-3‰.The estimated mean age of measured C is calculated from the difference between its Δ 14 C value and the Δ 14 C atm during the 2018 growing season, divided by the mean annual decline in Δ 14 C atm (see Equation [5] in Results).
Our samples were assumed to have Δ 14 C ≥ Δ 14 C atm , therefore, samples with Δ 14 C < Δ 14 C atm were considered contaminated with extraneous fossil C. One batch of nine samples was discarded due to many results with highly negative values.
2.5 | CO 2 efflux rate and δ 13 C analysis The CO 2 concentration and δ 13 C in the two-day incubations were determined from a sub-sample of air from one flask analysed by Isotope Ratio Mass Spectrometer (IRMS; Delta+ XL; Thermo Fisher Scientific) coupled to a modified gas bench with Conflow III and GC (Thermo Fisher Scientific, Bremen, Germany).Air from the flask was expanded into a pre-evacuated 12-ml Labco extainer (Labco Ltd, Lampeter, UK) equipped with a septum cap.Argon gas (Ar) was added to the exetainer to create a small over pressure to enable passive sampling by the auto-sampler (CTC Combi-PAL autosampler, CTC-Analytics, Zwingen, Switzerland).The air pressure values in the exetainer before and after the Ar addition were recorded for pressure correction.Each exetainer was sampled twice (30 or 50 μl per aliquot) by the auto-sampler and the aliquots were injected into the IRMS.Samples were analysed against a laboratory air standard on the Vienna Pee Dee Belemnite (VPDB) scale [Jena Reference Air Set-06 (JRAS-06), (Wendeberg, Richter, Rothe, & Brand, 2013)].For estimating the CO 2 concentration the m/z peaks 44, 45 and 46 were integrated and calibrated against samples of standard gas with a known concentration of 2,895 ppm.δ 13 C resp and CO 2 concentration were determined from the mean of two sequential injections.For δ 13 C resp the mean standard deviation (SD) of the duplicate samples was 0.04‰.For CO 2 concentration the mean SD for the duplicate samples was 14 ppm, regardless of the samples' concentration.
We tested the CO 2 concentration evaluation using the calibrated IRMS peak area by analysing flasks filled with known CO 2 concentrations using the same IRMS measurement protocol and found the IRMS overestimated CO 2 concentration by 4% at 10,000 ppm and 18% at 50,000 ppm.Empirical corrections based on a polynomial fit were used to correct reported sample concentrations.After correction, differences between known to corrected CO 2 concentrations were 0% on average and for 90% of the measurements the difference was < 2% of the measured concentration.In the same test, we found that the δ 13 C was stable when varying the injection volume up to 200 μl and in flask air pressures between 800 and 1,000 hPa, suggesting there are no isotopic fractionations in the sub-sampling procedure.
The mean CO 2 efflux during the incubation period normalized to the sample dry weight per day (mg C g À1 day À1 ) was calculated using Equation (2): where ΔCO 2 is the net change in CO 2 concentration (ppm/10 6 ) during the incubation (equal to the measured CO 2 concentration), I t is the incubation time (days), V HS is the volume of the headspace (269 ml), BP is the local barometric pressure (hPa), M C is the molar mass of C (12 mg mmole À1 ), T is the temperature of incubation (295 K), M dm is the dry mass of roots (g) and R is the ideal gas constant (83.14 ml hPa k À1 mmol À1 ).The error in the CO 2 efflux due to propagated uncertainties in ΔCO 2 , V hs , BP and M dm is estimated to be 4% of the reported value.
To measure δ 13 C ws, the capsule containing water-soluble C was combusted using an elemental analyser (NA 1110, CE Instruments, Milan, Italy) coupled to a Delta + XL IRMS (Thermo Finnigan, Bremen, Germany) via a ConFlow III.Samples were analysed against laboratory standards on the VPDB scale.
The apparent isotopic fractionation (Δ R , ‰) in respiration was estimated with the equation (neglecting the denominator): 2.6 | Estimations of C pool sizes and age The first approach (approach 1, Table 1) is based on the view that storage C age follows "last in, first out" dynamics (Lacointe, Kajji, Daudet, Archer, & Frossard, 1993), in which the most recent C transported to the root is the most accessible for respiration (Carbone et al., 2013).Accordingly, two pools were defined: (a) an "active" pool that supports metabolism and respiration, derived from recently transported C to the roots (in intact roots under no C limitation, recently fixed C < 1 year), and (b) a "stored" pool that only becomes active when the supply of transported C is reduced (Figure 2) (Herrera-Ramirez et al., 2020;Richardson et al., 2013).We measured respired CO 2 as a proxy to estimate the age and size of the metabolically "active" pool, with Δ 14 C resp representing its age.To estimate the amount of storage C < 1 year old (the active pool size in intact roots) we used the repeated respiration measurements.First, C with Δ 14 C signature of 1 year was defined as the 2018 Δ 14 C atm + the mean yearly Δ 14 C atm change (see Equation [5] in Results).Assuming excised roots will access increasingly older (higher Δ 14 C) C, we estimated when the CO 2 respired by the roots increased beyond this 1 year threshold (t depletion ).Integration of the fitted CO 2 efflux versus time curve between time = 0 until t depletion yielded an estimate of the size Note: Approach 1 is based on the Δ 14 C signature of root-respired CO 2 (Δ 14 C resp ), and approach 2 is based on the intercept and slope estimates of the fitted linear line between the Δ 14 C signatures of the water-soluble C (Δ 14 C ws ) and α-cellulose C (Δ 14 C cell ) extracted from roots.Those C extractions represent soluble sugars and structural C, respectively.
of the "active" pool of C younger than 1 year in intact roots.This approach is indifferent to the actual storage compound that supports respiration.
Our second approach (approach 2, Table 1) is based on the strong correlation between Δ 14 C ws and tissue age and the observation that Δ 14 C ws is often less than (younger) than Δ 14 C cell (Furze et al., 2020;Richardson et al., 2015;Trumbore et al., 2015).
For a root that has lived for multiple years, we expect that Δ 14 C cell will reflect the 14 C of substrates used for growth that can vary from year to year, while the Δ 14 C ws will reflect mixing of C that may date from the year of root formation with soluble C translocated from other parts of the plant.The mean Δ 14 C ws signature of a root thus comprises a spectrum of Δ 14 C values, ranging from recently assimilated C in young tissues to older storage C in the innermost ring.
Hence, in simplified view, the measured Δ 14 C ws is a weighted mean of the two functional C fractions (Figure 2): a proportion of active C, F active , with radiocarbon signature Δ 14 C active and the stored fraction C, F stored (= 1 -F active ), with radiocarbon signature Δ 14 C stored .The mass balance is described by (Equation [4]): We assume that the stored pool C consists of C deposited as storage reserves concurrent with root growth, thus Δ 14 C stored can be approximated by the measured Δ 14 C cell .Even if old C with high Δ 14 C signature is allocated to a root, the formed structural C and storage C are expected to share the high isotopic signature and the assumed approximation will still hold.Using the linear regression estimates for the relation between the measured Δ 14 C ws with Δ 14 C cell enables us to solve the other variables of Equation ( 4); The slope of the linear equation is equal to F stored , F active equals 1 -F stored and the intercept equals to F active Â Δ 14 C active , thus Δ 14 C active = intercept/F active .Using this approach, we do not assume an age for the active C (e.g., < 1 year); instead, we estimate the Δ 14 C of F active from the It is important to note that the water-soluble fraction contains sugars, but also many other soluble compounds like tannins and amino and organic acids.Trumbore et al. (2015) estimated that ~50% of the C in the soluble fraction extracted from wood (they used methanol: water mixture as a solvent) originates from the sugars sucrose, glucose and fructose.Here, we assume that on timescales of > 1 year exchange of C between cellular metabolites means that the Δ 14 C ws signature approximates that of sugars.

| Statistical analysis
All data were analysed using R (R Core Team, 2019).We tested effects of treatment (Pre-girdling, girdling and control) and root class (coarse, fine) on the different measures.Normality was tested with the Shapiro-Wilk test and equality of variance with Leven's test.When both assumptions were not violated we proceed with one or two-way ANOVA followed by Tukey's HSD post-hoc test.When the assumptions of normality and/or homogeneity were violated, we used the non-parametric Kruskal-Wallis rank-sum test, followed by pairwise Wilcoxon Rank Sum Tests.For the linear models, we used the lm function.

| CO 2 efflux rate
The CO 2 efflux rate in the two-day incubations was significantly higher in the fine roots than in the coarse roots (p < .01,Kruskal-Wallis) with mean ± SE values of 3.0 ± 0.3 and 2.0 ± 0.2 mg C g À1 d À1 , respectively (Figure 3b).The mean Q 10 measured in September-  3b).We used the mean efflux rates corrected to the daily mean of soil temperature to estimate the total amount of C respired by fine and coarse roots during the 82 days between the girdling and the first post-girdling campaign; these were 150 ± 19 and 100 ± 12 mg C g À1 , respectively (uncertainty was calculated by varying the mean efflux rates and Q 10 with their uncertainties).Accordingly, the mean daily C respired in the fine roots is 1.8 ± 0.2 mg C g À1 and 1.2 ± 0.1 mg C g À1 in the coarse roots.

| Radiocarbon
Combining measures of Δ 14 C cell in tree-ring late wood, leaf Δ 14 C resp and the atmospheric record in the region, the annual decline of Δ 14 C atm averaged 4.7‰ per year during the last two decades (Figure 4).Our estimate for the atmospheric Δ 14 C resp during the 2018 growing season was +2.3‰, the mean Δ 14 C resp of the roots of the control trees (Table 2).This value is within the 0.6-4.4‰range of Δ 14 C atm mean values measured during the 2018 growing season at several ICOS stations (Table S1).Thus, we calculated the mean age (year) of a C pool with radiocarbon signature Δ 14 C using Equation ( 5): Root respired Δ 14 C resp values were mostly similar to the Δ 14 C atm in the year of collection, with overall mean ± SE of 4.8 ± 0.9 (0.5 ± 0.2 year; n = 36) and range of À5.2‰ to +23.7‰ (0-4.6 years).
The mean Δ 14 C resp when both root classes are pooled together was higher in the girdled trees (7.9 ± 1.9; n = 12) than the pre-girdling (4.2 ± 0.8; n = 12) and the control (2.3 ± 1.8, n = 12) (Figure 5).Differences between the treatments were not significant (p = .14-.20, Wilcoxon).However, the mean Δ 14 C resp of the girdled trees was significantly higher than the control trees (represent current atmosphere) when compared exclusively (p = .04,Wilcoxon).The mean age of the respired Δ 14 C resp in the girdled trees was 1.2 ± 0.4 year, compared to 0 ± 0.4 year in the control roots.
The Δ 14 C ws values were higher than Δ 14 C resp for the same root samples, ranging between 0.8 and 90.7‰ (0-18.8year).The linear regression model equations for the Δ 14 C ws versus Δ 14 C cell relationship, its statistical information and approaches 1 and 2 predictions are presented in Table 2 and Figures 5 and 6.The variability in Δ 14 C cell explained the majority of the variability in Δ 14 C ws (r 2 > 0.8 in most groups; Table 2; Figure 6).The slope estimates based on the linear fit equations were highly significant in most subgroups (p < .01),except in the fine roots of the girdled trees due to one outlier with much higher Δ 14 C ws than Δ 14 C cell (Figure 6).According to approach 2, F stored  2).The uncertainties in Δ 14 C active estimations (= Intercept/F active ) are also large as a result of the propagation of the intercept errors.The overall Δ 14 C active calculated by approach 2 is 5.7 ± 5.9‰ with an estimated age of 0.7 ± 1.3 year (Table 2).

| Repeated incubations
Temporal changes in all measured parameters were observed between day 1 and day 7 of the repeated incubations (Figure 7).
Mean Δ 14 C resp (± SE) increased more rapidly in the fine roots (from 3.2 ± 0.8 to 16.1 ± 2.1‰) than in the coarse roots (from 4.4 ± 0.3 to 10.1 ± 0.9‰), corresponding to estimated ages of 2.9 ± 0.5 and 1.6 ± 0.2 year (Figure 7a) at the end of incubation, respectively.The CO 2 efflux rates declined over time, but remained higher in fine roots compared to coarse (Figure 7b).
Following approach 1, any carbon with Δ 14 C < 7‰ [the sum of Δ 14 C atm (2.3‰) and annual decline (4.7‰)] was defined as younger than 1 year.Using linear regression, we estimated that the 7‰ value was exceeded after 2.3 and 4.3 days (= t depletion ) in fine and coarse roots, respectively, indicating time to deplete the active C pool (Figure 7a).Integrating the total CO 2 efflux over time (fitted curves in Figure 7b) between time = 0 and time = t depletion gives an estimate of the active pool size: 5.3 and 5.1 mg C g À1 for fine and coarse roots, respectively.The total amount of C respired over the 7 days of incubation was calculated as 13.4 and 7.1 mg C g À1 for fine and coarse roots, respectively, which is equivalent to 66 and 35% of the total mean NSC-C in fine and coarse roots, respectively.

| δ 13 C results
The δ 13 C of the water soluble fraction extracted from roots varied significantly with treatment (p < .001;two-way ANOVA) and root T A B L E 2 A summary of approaches 1 and 2 results size (p < .01),with no interaction effect (p = .833)(Figure 8a, Table S2).The post-hoc tests (Tukey's HSD) showed δ 13 C ws in the pre-girdling roots were significantly lower than the girdling and control roots that did not differ significantly.Water soluble C extracted from fine roots was 0.74‰ more enriched than the coarse root extracts.The treatment had significant effect also for δ 13 C resp (p < .001,Kruskal-Wallis), and the post-hoc test (Wilcoxon) indicates significant difference between all three campaigns: the pre-girdling roots had the lowest mean value, with girdled trees on average 1.15‰ higher and the control trees 2.64‰ higher than the pregirdling (Figure 8b, Table S2).The coarse root δ 13 C resp was 0.80‰ more enriched than the fine roots (p = .052,Kruskal-Wallis).The mean Δ R value (δ 13 C ws À δ 13 C resp ) for fine roots was 1.53‰ higher than for coarse roots (p < .01;one-way ANOVA; Figure 8c).The treatment had marginal effect (p = .053;Kruskal-Wallis), where the Δ R in the control roots was significantly lower (wilcoxon test) than in the pre-girdling and girdling (Figure 8c, Table S2).

| DISCUSSION
4.1 | Size and age estimates for active and stored C pools using two approaches Our results provide unique estimates for the size and age of the functional sub-pools of NSC soluble in water using two distinct approaches.
The correlation between Δ 14 C ws and Δ 14 C cell (approach 2) suggests that the C stored over multiple years on average makes up 70 ± 5% (range of 55-80%) of the extracted, water-soluble C (Figure 6).Sugars make up only a fraction of the water-soluble C used for the Δ 14 C ws measurement.
Assuming sugars contribute in the same proportion to both active and stored pools found in water-soluble C, of the total average sugar concentration (17.1 ± 1.2 mg C g À1 ; Table S2), 30% (5.1 ± 0.5 mg C g À1 ) is found in the active pool and 70% (12.0 ± 0.5 mg C g À1 ) in the stored pool (Table 1; approach 2).Based on the amount of C respired in repeated incubations (approach 1), we estimated the pool of C younger than 1 year as 5.1 mg C g À1 for coarse roots and 5.3 mg C g À1 for fine roots.Both approaches thus agree well in estimating the active pool size, suggesting that the < 1 year criterion provides a way to estimate the age and size of the active pool in intact roots.Both approaches (1 and 2) also agree with regard to the active pool age being younger than 1 year, that is, 0.5 ± 0.2 and 0.7 ± 1.3 year, respectively (Table 2).Our results in roots F I G U R E 5 The Δ 14 C signatures and mean age estimations of the active C pool according to two approaches; approach 1 assumes Δ 14 C active equals to the Δ 14 C signature of respired CO 2 (Δ 14 C resp ); approach 2 is based on the intercept and slope estimates of the fitted linear line between the Δ 14 C signatures of the extracted watersoluble C (Δ 14 C ws ) and α-cellulose C (Δ 14 C cell ).According to this approach the fraction of the active C pool F active = 1 À slope.Roots collected before girdling (Pre-girdling), ~3 months after girdling (Girdling) and ~3 months after girdling but in un-girdled trees (Control).One set of roots was used for respiration incubations (n = 12) and second set was used for the C extractions (n = 23, 12, 11, respectively).Error bars of approach 1 is the standard errors of the Δ 14 C resp results, and of approach 2 are the cumulative standard errors of the slope and intercept estimates  46, 11, 12, 6, 6, 5, 6, respectively).For example, the slope in the equation in panel (a) is 0.7 therefore F stored for the subgroup "All" is 0.7.Labels within the bars are the r 2 of the linear regression.Error bars are the standard error of the slope estimate [Colour figure can be viewed at wileyonlinelibrary.com] corroborate those found by Richardson et al. (2013) for tree stems.
When they constrained active ("fast") NSC pool age to < 1 year, they found this pool contributed 23-56% of the total NSC, compared with our estimates of 20-45% for roots (Table 2).
We also found disagreements between the two approaches, namely for individual subgroups (Figure 5, Table 2).There are apparently two reasons.The first reflects large uncertainties in the linear model's prediction of the intercept value that approach 2 is based on, which leads to extreme and unrealistic Δ 14 C active values, for example, equivalent to À5.4 year for coarse roots during pre-girdling (Table 2).Still, considering the uncertainties, the two approaches to estimate Δ 14 C active mostly agree (Figure 5).In addition, the tendency to use older C to fuel respiration (i.e., transferring older C to the active pool) in the girdled-trees roots that were predicted and observed in approach 1 (Δ 14 C resp ), was also observed in approach 2 (Figure 5).
The second source for the disagreement between the methods is potentially true difference, as apparent for the control roots, where the estimate from approach 2 for Δ 14 C active (based on the Δ 14 C ws that is not explained by variability in Δ 14 C cell ) is significantly higher than approach 1 considering the uncertainties (Figure 5, Table 2).A change in the age of the incoming C to the roots during the recent past might have impact on that discrepancy.Further understanding would require better estimation of the composition of water soluble components and the potential for their makeup and age to change with season.

|
The stored fraction should be also considered in driving Δ R values Water deficit in the soil increased from midsummer to the end of the growing season because of the 2018 hot drought conditions (Figure 1).In control (un-girdled) trees, we observed enrichment in both δ 13 C ws and δ 13 C resp measured in Sept/Oct compared to the pregirdling roots measured in June-July.These δ 13 C resp increases are in line with expected water shortage effects on leaf-level fractionation (Figure 8) (Farquhar & Sharkey, 1982;Madhavan, Treichel, & Oleary, 1991;Pate & Arthur, 1998;Scartazza, Moscatello, Matteucci, Battistelli, & Brugnoli, 2015).Smaller enrichments in δ 13 C ws could be explained by dilution with storage pools.A simple mass balance calculation using the relative enrichments in δ 13 C ws (+1.12‰) and δ 13 C resp (+2.64‰) between control and pre-girdling provides an estimate for F active : 1.12/2.64= 0.42, nearly equalling the 0.43 predicted by approach 2 for the control trees (Table 2).This suggests that 13 Cenriched sugars produced in control-tree leaves are mixed into the active pool in the roots where they support respiration, while the stored water-soluble C integrates more depleted δ 13 C values inherited from previous years.Thus the apparent decrease in Δ R in the control roots most likely does not reflect a metabolic shift (e.g., reduced PEPC activity) in the roots.In contrast, the enrichment between pre-girdled and girdled roots is larger for δ 13 C ws (+1.52‰) than for δ 13 C resp (+1.15‰), perhaps reflecting differences in transported substrates.S2).The estimated ages of the respired C increased from < 1 year to 1.6 and 2.9 year, for coarse and fine roots, respectively.This is consistent with roots respiring an initial active pool fed by freshly fixed C and switching to older substrates as this pool is exhausted (Herrera-Ramirez et al., 2020).In the case of excised roots, the source of the older C substrate was clearly in the roots themselves.
During the months that elapsed after girdling, root respiration rates did not decline as they did for excised roots (Figure 3b).During this period, assuming constant respiration rates, an estimated 100 and 150 mg C g À1 were respired in the coarse and fine roots, respectively.
This was roughly 5-7 times higher than the C stored in the NSC pools, which also did not decline following girdling (Figure 3a).Therefore, it is clear that the C being respired in the girdled roots must be replaced by C translocated from root-external sources.The C may originate from storage in below-girdling parts of the tree, and from neighbouring healthy trees.Populus trees are known to have root connections with shared root systems.Populus root systems can remain functional for at least 20 year after shoot removal, and can persist for millennia while shoots are repeatedly resprouting (Pregitzer & Friend, 1996).Thus, the root systems of our girdled trees may be older than the ~60-70 year stems and have disproportionally large NSC stocks with possible connection from girdled to healthy trees.Natural root grafting, the phenomena where two roots are pressed together to form vascular continuity that enables C transfer between trees, is also common in Populus trees and could be a source for the replenished C (Fraser, Lieffers, & Landhausser, 2006;Gaspard & DesRochers, 2020;Mudge, Janick, Scofield, & Goldschmidt, 2009).Carbon transfer from neighbouring trees can occur also via mycelial networks (Rog, Rosenstock, Korner, & Klein, 2020).Indeed, the girdled trees in our study have sur- Whatever the source of translocated C to girdled tree roots, our isotopic measurements indicate that it is not derived from fresh photo assimilates.Based on Δ 14 C resp , the mean age of C respired from girdled roots was ~1.2 year older than C respired from control roots (Figure 5).This is younger than the respired C at the end of the repeated incubations (1.6-2.9 year).Difference in the δ 13 C resp between the girdling and control roots further indicates they do not share the same respiratory C source (Figure 8b).The increase in δ 13 C resp between pre-girdling and girdling can be explained by hydrolysis of starch, which tends to be more enriched than sugars (Brugnoli, Hubick, von Caemmerer, Wong, & Farquhar, 1988;Damesin & Lelarge, 2003;Gleixner, Danier, Werner, & Schmidt, 1993;Maunoury-Danger et al., 2010).

| Hints for high PEPC activity in the fine roots
The Δ R mean values by root class and treatment ranged between (À0.35‰) and (À4.38‰) (Figure 8, Table S2), a slightly wider range than the (À0.7‰) to (À3.1‰) values measured in mature Eucalypt roots (Gessler et al., 2007).While some of this could be due to differences in temperature effects, these likely do not explain the range in Δ R values.
As discussed earlier, the driver for the most negative Δ R values measured in the control trees probably reflects differences in δ 13 C between the active and the stored C pools ("mixing"), and not a change in respiration metabolism (Figure 8).In fact, it is difficult to resolve whether the main driver for a given Δ R value is mixing or metabolism, especially in perennial plants with old reserves.However, differences in metabolism might be inferred in some circumstances, for example when comparing Δ R in fine and coarse roots from the same trees where uniform mixing could be assumed.Δ R was higher for fine (À1.1‰) than coarse (À2.67‰) roots, a difference reflecting lower δ 13 C resp (À0.80‰) and higher δ 13 C ws (+0.74‰) in the fine roots (Figure 8).This pattern is in agreement with the expected net isotopic effect of refixation of internal CO 2 by PEPC: a 13 C depletion of the respired CO 2 and a 13 C enrichment of the products (Werner & Gessler, 2011).
Additional support for CO 2 fixation with high PEPC activity in roots comes from the decline in the respiration quotient (RQ) from ~1 to ~0.6 during our repeated incubations (Figure 7d), since PEPC CO 2 re-fixation would reduce CO 2 efflux and thereby RQ (Hilman et al., 2019).The effect of PEPC on respiration grows as its activity as a fraction of total respiration increases (Badeck et al., 2005), as might occur with repeated incubations.Our results mirror those of Bathellier et al. (2009) who also observed a decrease in RQ from 1.1 to 0.8-0.9 while δ 13 C resp remained stable in roots of French bean during 6 days of carbohydrate starvation by darkening.Alternative explanations for the change in RQ exist, including a shift from carbohydrates to lipids as the main respiration substrate (Figure 7d).However, we observed no simultaneous decline in δ 13 C resp that would be expected to accompany such a substrate shift, given the depleted δ 13 C value of lipids (Fischer et al., 2015;Tcherkez et al., 2003).Other factors that could cause a decline in RQ include a relative increase of O 2 uptake [e.g., through production of reactive oxygen species associated with cell death during the experiment (Chae & Lee, 2001)].
While not conclusive, our results suggest that the potential role of PEPC in tree roots deserves further exploration.PEPC has an important role in plant metabolism, synthesizing C 4 organic acids from HCO 3 À and phosphoenolpyruvate (PEP).The carbon skeletons of the organic acids can be also used to generate amino acids.High activity of PEPC in roots of model plants is well documented and linked, among other factors, with exudation of organic and amino acids to the rhizosphere (Neumann & Römheld, 2007;O'Leary, Park, & Plaxton, 2011).The addition of Δ R and RQ analysis to the PEPC assessments toolbox that contains enzymatic and genetic assays can provide realistic estimations for the rate of PEPC activity, although this first requires resolving mixing effects on Δ R .Compound-specific analysis (using GC-C-IRMS) for δ 13 C of the putative respiratory substrates and metabolites might provide more accurate Δ R values and information about PEPC metabolism.
Mean monthly rainfall (mm, bars) and air temperature ( C, circles) during 2018 growing season, and in the previous 8 years (2010-2017) in the study site (Großer Hermannsberg, Germany).Error bars for the 2010-2017 data represent one standard deviation from the mean a glass vial kept on ice (to slow-down microbial degradation).A subsample of 2 ml from the total 15 ml was used for quantification of the sugars sucrose, glucose and fructose.For the δ 13 C and Δ 14 C analysis, the rest of the solution was concentrated by freeze-drying and pipetted into tin capsules for δ 13 C analysis, and into pre-baked silver capsules for Δ 14 C analysis.The starch in the pellet from the water extraction was converted by α-amylase (Sigma cat.no.A4551) into

Δ
14 C ws value that is not explained by Δ 14 C cell .The model estimates are based on roots from different trees hence they represent averaged values for several trees (n = 5-12).
U R E 3 Means ± SE of coarse (> 2 mm) and fine (2 ≤ mm) roots collected before girdling (Pre-girdling), ~3 months after girdling (Girdling) and ~3 months after girdling but in un-girdled trees (Control).(a) NSC concentrations separated by molecules.Starch was not measured in pre-girdling roots.n = 11, 12, 6, 6, 6, 6, respectively; (b) CO 2 efflux rates measured in two-day incubations in room temperature (empty bars) and corrected to field temperature (full bars).n = 10, 11, 6, 6, 6, 6, respectively.Below the legends the respective significant statistical tests [Colour figure can be viewed at wileyonlinelibrary.com]The estimated atmospheric Δ 14 C signature (Δ 14 C atm ) in the study site during the last two decades.The atmospheric record in grey is the mean Δ 14 C atm of the northern hemisphere zone 1 after Hua et al. (2016).Tree rings in orange is the α-cellulose Δ 14 C signature measured for the late wood in annual rings from years 2010 to 2018.The blue point is the Δ 14 C signature of respired CO 2 from leaves harvested in July 2019, which assumed to represent recent photo-assimilates thus current Δ 14 C atm .The linear equation indicates the mean annual decline in Δ 14 C atm is 4.7‰ [Colour figure can be viewed at wileyonlinelibrary.com] equals the fitted line slope, hence the overall F stored (± SE of the model) is 0.7 ± 0.05 and values in the subgroups range between 0.55 ± 0.03 and 0.8 ± 0.1 (without the exceptional girdled fine roots).The overall F active , which equals 1 -F stored , is 0.3 with range of 0.2-0.45 in the subgroups.Significant intercept estimates were computed only for the coarse roots and coarse + fine roots of the control trees (p < .05),while in the other subgroups the errors were rather large (Table The linear model equations are for the relationship between the Δ 14 C signatures of the water-soluble C (Δ 14 C ws ) and α-cellulose C (Δ 14 C cell ) extracted from roots.Those C extractions represent soluble sugars and structural C, respectively.Approach 2 is based on the linear model and predicts the fractions of the stored (Fstored) and active (Factive) C pools from the total water-soluble C, and the Δ 14 C signature of the active pool (Δ 14 Cactive).a Values are means ± SE. b Bold numbers indicate significant (p < .05)estimate of the model.The standard errors of the coefficients are in parentheses.c The error is the model's slope SE, F stored equals the equation's slope, equation's intercept/Factive.Errors are the propagated standard errors of the intercept and slope estimates.
U R E 6 (a) A scatter plot of the Δ 14 C signatures of the watersoluble C (Δ 14 C ws ) and α-cellulose C (Δ 14 C cell ) extracted from roots, with the equation of the linear model for all results pooled together.The shapes indicate the root class (coarse roots, > 2 mm, fine roots, ≤ 2 mm), and colours indicate the treatment: before girdling (Pregirdling), ~3 months after girdling (Girdling) and ~3 months after girdling in un-girdled trees (Control).Error bars are the analitycal uncertainty; (b) the stored fraction (F stored ) from the total watersoluble C as estimated from the slope when the linear model is applied to the different subgroups (n = U R E 7 Results (mean ± SE) for coarse roots (> 2 mm) and fine roots (≤ 2 mm) from repeated respiration incubations during 7 days.(a) the Δ 14 C signature of the respired CO 2 (Δ 14 C resp ).Presented are equations of linear regression models; (b) CO 2 efflux rates.Presented are equations that best fitted the results to the equation y = a Â b Àx ; (c) the δ 13 C signature of respired CO 2 (δ 13 C resp ); (d) RQ (ratio CO 2 efflux/O 2 uptake).Results in panels (a) and (c) were measured in two-day flask incubations (n = 4, 3 for coarse and fine roots, respectively), results in panels (b) and (d) were measured in short-term incubations (~20 min, according to the consecutive incubations order, n = 3, 4, 3 for coarse roots, and n = 3, 4, 4 for fine roots) [Colour figure can be viewed at wileyonlinelibrary.com]4.3| The carbon balance in girdled-tree roots was maintained by translocation of stored NSCThe ability of roots to access older C for respiration when new C supply is cut off was demonstrated in the repeated incubations and in the girdling experiments.However, the supply of older C was not the same in both cases.Over the 7 days of the repeated incubations, the coarse and fine roots respired 7.1 and 13.4 mg C g À1 , respectively, roughly 35-70% of the 20.3 mg C g À1 of NSCs contained in the incubated roots (overall sugars and starch mean, Table

vived for 2
years so far without visible signs of crown damage.Tree responses to girdling vary with species, stand age and experimental design that affect the size of the storage reserves and the connections between non-girdled and girdled trees (Levy-Varon, Schuster, & Griffin, 2012).The fact that our girdled trees were located among the un-girdled trees without physical or substantial spatial separation increase the chances for C sharing via roots.A way to reduce such effects in future is to girdle trees as a block of forest stand and isolate from root connections by trenching at the border.

1
Methods for C pool size and age estimation