CO2 concentration and water availability alter the organic acid composition of root exudates in native Australian species

Root exudation of organic acids (OAs) facilitates plant P uptake from soil, playing a key role in rhizosphere nutrient availability. However, OA exudation responses to CO2 concentrations and water availability remain largely untested. We examined the effects of CO2 and water on OA exudates in three Australian woodland species: Eucalyptus tereticornis, Hakea sericea and Microlaena stipoides. Seedlings were grown in a glasshouse in low P soil, exposed to CO2 (400 ppm [aCO2] or 540 ppm [eCO2]) and water treatments (100% water holding capacity [high-watered] or 25–50% water holding capacity [low-watered]). After six weeks, we collected OAs from rhizosphere soil (OArhizo) and trap solutions in which washed roots were immersed (OAexuded). For E. tereticornis, the treatments changed OArhizo composition, driven by increased malic acid in plants exposed to eCO2 and increased oxalic acid in low-watered plants. For H. sericea, low-watered plants had higher OAexuded per plant (+ 116%) and lower OArhizo per unit root mass (–77%) associated with larger root mass but fewer cluster roots. For M. stipoides, eCO2 increased OAexuded per plant (+ 107%) and per unit root mass (+ 160%), while low-watered plants had higher citric and lower malic acids for OArhizo and OAexuded: changes in OA amounts and composition driven by malic acid were positively associated with soil P availability under eCO2. We conclude that eCO2 and altered water availability shifted OAs in root exudates, modifying plant–soil interactions and the associated carbon and nutrient economy.


Introduction
Understanding how terrestrial ecosystems respond to global changes requires an accurate prediction of plant responses to increases in atmospheric CO 2 concentrations and changes in water availability. While elevated CO 2 concentrations (eCO 2 ) generally enhance photosynthetic rates and increase plant growth, multiple lines of evidence suggest that the plant response to eCO 2 depends on soil nutrient status: the positive effect of eCO 2 on plant growth weakens under nutrient-limited conditions but strengthens with increased nutrient availability (Reich et al. 2006). Thus, soil nutrient economy and plant nutrient uptake strategy likely drive the magnitude of plant responses to eCO 2 . However, little is known about how future atmospheric CO 2 concentrations will affect these factors (Walker et al. 2021).
Plants exhibit various strategies to acquire nutrients when soil availability is low, for example, by increasing their relative belowground carbon (C) investment (higher root:shoot ratios) or adjusting their root morphology (Hill et al. 2010;Ostonen et al. 2011). Soil nutrient acquisition can also be enhanced via plant investment of photosynthates to support mycorrhizae (Eissenstat et al. 2015;Helmisaari et al. 2009). Other strategies to increase nutrient availability include secretion of readily available organic C (i.e. sugars) from roots that acts as substrates for rhizosphere microbes, enhancing their decomposition of organic matter-so-called 'priming' (Kuzyakov 2010;van Hees et al. 2005). Finally, plants can exude low-molecular-weight organic anions (organic acids; OAs) from roots to mobilise P. Phosphate bound to cations (e.g. Mn 2+ , Al 3+ , Ca 2+ ) or adsorbed onto soil mineral surfaces is inaccessible to plants. However, OAs compete with phosphate, preventing it from forming complexes with cations and thus liberating P (Bais et al. 2006). Interestingly, P mobilisation efficiency varies among different OAs (Mendes et al. 2020). Thus, changes in the amount and composition of root-exuded OAs likely impact soil P cycling and associated plant growth.
Soils in Australia are typically impoverished, particularly in P availability (Walker and Syers 1976). Therefore, native plant species in Australia are welladapted to low P conditions, with a range of P acquisition strategies such as high rates of OA exudation and mycorrhizal symbiosis Ryan et al. 2012). Indeed, some species in the Cyperaceae and Proteaceae families are particularly welladapted to P-limited conditions and form cluster roots that express especially high rates of OA exudation .
Elevated CO 2 likely increases plant nutrient demands due to enhanced growth and/or altered tissue stoichiometry (Bassirirad et al. 1996;Dijkstra et al. 2012;Johnson et al. 2004). Plants typically invest excess C belowground to tackle increased nutrient demands, enhancing microbial decomposition of organic matter and ectomycorrhizal nutrient transport (Terrer et al. 2016). Hasegawa et al. (2016) and Ochoa-Hueso et al. (2017) reported eCO 2 increased soil P availability in a free-air CO 2 enrichment (FACE) experiment established in a P-limited Eucalyptus woodland in Australia (EucFACE). These studies put forward the hypothesis that increased photosynthetic rates under eCO 2 conditions increase root secretion of OAs into the soil, liberating soil P in line with the expectations from Jin et al. (2015), Lambers et al. (2008) and Ryan et al. (2012).
In combination with eCO 2 and global warming, substantial changes in precipitation patterns linked to climate change cause extreme rainfall events in many parts of the world (Sun et al. 2021). In particular, the southeastern parts of Australia have experienced prolonged periods of drought and intensified rainfall accompanied by flooding on an unprecedented scale in recent years (Australian Bureau of Meteorology: http:// www. bom. gov. au). These extreme events are projected to be exacerbated in the future (Grose et al. 2020), and thus there is an urgent need to investigate the role of climate in modulating plant nutrient uptake strategies. However, few studies have scrutinised the responses of OAs in root exudates to different levels of water availability; those that have studied such effects on OA exudation report drought stress increased OA amounts and shifted rhizosphere OA composition for leguminous species, despite lower photosynthetic rates (Sharma et al. 2021;Suriyagoda et al. 2010). Waterlogging stress, on the other, suppressed root exudation for rapeseed (Zhou et al. 1997) and soybean (Shimono et al. 2012). This contrasts with the hypothesised effects of eCO 2 enhancing root exudation via increased photosynthetic rates . Thus, CO 2 -and water-induced changes in OA exudates are likely regulated by different mechanisms but remain largely unexplored. Furthermore, little is known about the effects of CO 2 and water availability on OA exudation.
Here, we implement two extreme scenarios for water availability: high-watered conditions (soil water held close to field capacity) and low-watered conditions (soil water availability maintained at a similar level to that of dry, warm seasons in local Eucalyptus woodlands in southeast Australia). We study how root exudation of OAs responds to eCO 2 under these two water regimes using three plant species native to these woodlands: Eucalyptus tereticornis, Hakea sericea and Microlaena stipoides. Eucalyptus tereticornis thrives across eastern Australia (Drake et al. 2015) and is the dominant canopy species at the EucFACE field site. The genus Eucalyptus forms both arbuscularmycorrhizae and ectomycorrhizae (Boudarga et al. 1990;Chilvers et al. 1987). Hakea sericea is an understorey shrub species in the Proteaceae family. It is typically nonmycorrhizal but produces cluster roots and hence is considered well-adapted to low P conditions (Dinkelaker et al. 1995). Microlaena stipoides is a native, perennial C 3 grass species, typically associated with arbuscular mycorrhizae (Hill et al. 2010). At the EucFACE site, M. stipoides is responsible for ~ 99% of the herbaceous understorey vegetation biomass (Collins et al. 2018;Hasegawa et al. 2018), contributing to 24% of the ecosystem gross primary productivity and thus playing a significant role in C cycling (Jiang et al. 2020).
We grew plants in a glasshouse in soil collected from the EucFACE field site, allowing us to relate results to earlier findings from the field study. We tested the hypotheses that: 1) eCO 2 has positive effects on root exudation; 2) plants growing under low-watered conditions exude larger amounts of OAs than those under high water availability; 3) the magnitude of positive eCO 2 effects on root exudation is smaller for low-watered than high-watered plants as eCO 2 improves plant water use efficiency, leading to slower depletion of soil water under low-watered conditions (Robredo et al. 2007); and 4) CO 2 and water treatments alter OA composition. Release of carboxylic acids from roots serves to maintain the charge balance in plant cells when cation and anion uptakes are unbalanced, and hence available N forms (nitrate or ammonium) can directly influence root exudation (Imas et al. 1997). Therefore, we further hypothesise 5) that CO 2 -/water-induced changes in OA amounts are concomitant with changes in soil nutrient availability, with larger OA amounts associated with higher nitrate and phosphate concentrations and lower ammonium concentrations.

Materials and methods
Experimental design, setup and plant growth Three plant species were examined, each representing one stratum within the native Eucalyptus woodlands in southeastern Australia: E. tereticornis (canopy tree), H. sericea (understorey shrub) and M. stipoides (woodland floor graminoid). Seeds of E. tereticornis were purchased from a commercial supplier (Australian Tree Seed Centre, CSIRO National Research Collections Australia), while those of H. sericea and M. stipoides were collected locally in a native Eucalyptus woodland near Richmond, NSW (adjacent to the EucFACE field site). The seeds were sown and germinated on seed cell trays in growth chambers (30 °C for E. tereticornis, 25 °C for H. sericea and M. stipoides). Following germination, trays were transferred to a glasshouse bay. Once seedlings were well-established (ca. 2-3 weeks after germination), they were transplanted into cylindrical PVC pots (10 cm diameter × 33 cm height): 20 pots for E. tereticornis and 32 each for H. sericea and M. stipoides. The number of pots for E. tereticornis was fewer than for the other two species due to lower germination rates. One seedling was transplanted into each pot for E. tereticornis and H. sericea. Three seedlings for M. stipoides were planted into each pot to achieve a similar level of plant biomass per pot across all three species. Each cylindrical pot was filled with 3.6 kg soil collected adjacent to the EucFACE site (ca. 1.6 g cm -3 bulk density). The soil is a loamy sand Chromosol of the Clarendon Formation (Bannerman and Hazelton 1990; Barton et al. 2010) with a pH of 5.45, 677 mg kg -1 total N, 59 mg kg -1 total P and 1.8% organic matter content. A more detailed description of the soil is available elsewhere Hasegawa et al. 2016). Soil was collected from the top 0-20 cm, air-dried for two days, sieved (4 mm) to remove plant material and rocks, and thoroughly homogenised before use. When transplanted into pots, seedlings were well-watered and allowed to settle for 2-3 days.
Pots were placed in one of four naturally lit glasshouse compartments located at Western Sydney University in Richmond, NSW, Australia, between March and April 2016. Detailed information on the management of glasshouse conditions is available elsewhere (Ghannoum et al. 2010). Briefly, the glasshouse temperature simulated the 30-year average daily temperature of the local region during the growing season (November-May), with average temperatures of 26 and 18 °C for day and night, respectively. The average relative humidity was 70%. Of the four compartments, two received ambient CO 2 (400 ppm), and two received elevated CO 2 (ambient + 140 ppm). CO 2 concentrations in the eCO 2 compartments were raised by injecting CO 2 gas (Food grade, Air Liquid, Australia) from pressurised cylinders. The glasshouse temperature and CO 2 concentrations were constantly monitored and regulated by a control system (Lambda T, ADC BioScientific Ltd., Hoddesdon, Herts, UK). The plants were exposed to two levels of water availability to examine the interactive effects between CO 2 and water treatments. Half of the pots were highly watered (hereafter referred to as 'high-watered'), with the soil water content maintained at 100% water holding capacity (WHC) (ca. 15% gravimetric water content [GWC] or 23% volumetric water content [VWC]). The other half of the pots were maintained at 50% WHC (ca. 7% GWC) for the first three weeks and then 25% WHC for the next three weeks until harvest (hereafter referred to as 'low-watered'). The pots were weighed and watered three times a week with deionised water to meet the target soil water contents. Pots were grouped into blocks, each comprising high-watered and low-watered replicates, resulting in ten blocks for E. tereticornis and 16 blocks for H. sericea and M. stipoides. Half of the blocks received aCO 2 and the other half eCO 2 . Blocks were randomly allocated to one of four glasshouse compartments, rotated within the compartments three times a week and between replicate compartments within the same CO 2 treatment once a week to minimise potential heterogeneity of environmental conditions within and between replicate compartments.
Harvest and root exudate collection After six weeks, the plants were harvested. All plants were watered one day before harvest so that the soil water contents were consistent at 50% WHC. Each pot was upended to remove the full plant and soil contents, and the soil was then gently shaken off the plant. The soil that remained on roots was defined as rhizosphere soil. Root exudates of OAs were collected following the approach in Kidd et al. (2018) using two sequential methods. First, roots were immersed in a known volume (ca. 50 ml) of 0.2 mM CaCl 2 solution and gently shaken for 2-3 min to wash off rhizosphere soil. The collected OAs are referred to as rhizosphere OAs (OA rhizo ). Second, following removal of rhizosphere soil, roots were thoroughly washed with Milli-Q water, placed in a known volume (ca. 60 ml) of 0.2 mM CaCl 2 solution, and incubated for 1 h at room temperature in a laboratory to collect exudates directly from roots, referred to as OA exuded . OA exuded represents the amount of OA exudates from plant roots over time, whereas OA rhizo is the amount of OAs in rhizosphere soil as the net result of exudation and losses through microbial degradation and adsorption onto soil particles. Although carefully executed, the process of harvesting inevitably damaged some root cells, likely resulting in a minute metabolite leak into the collected OA solutions. The OA rhizo and OA exuded solutions were filtered (0.2 µm) before adding a drop of concentrated phosphoric acid to each to minimise microbial decomposition. Solutions were immediately frozen in liquid nitrogen and stored at -20 °C until analysis. The harvested plants were placed in a drying oven at 80 °C for 2-3 days, with dry weight (DW) of above-and below-ground parts measured to obtain total biomass and root:shoot ratios (RSR). For H. sericea, the number of cluster roots was counted, and cluster root counts per unit root biomass were calculated.

Analysis of OAs in root exudates
Organic acids were analysed using Reverse Phase Liquid Chromatography as described in Cawthray (2003) for the following compounds: malic, acetic, citric, cisaconitic, succinic, fumaric and trans-aconitic acids. Malonic and maleic acids were also analysed but not detected in the current study. Oxalic acid was measured following Uloth et al. (2015). Briefly, oxalic acid was separated at 23 °C on a Hypersil Hypercarb column using a mobile phase of 0.1% trifluoroacetic acid at 15 ml min -1 . Detection was at 210 nm, with photodiode array detector acquisition from 195 to 400 nm.
Total amounts of OA rhizo and OA exuded equated to the sum of all detected OAs on a mass basis for each growing pot. Total OA rhizo amount was expressed as µg plant -1 , and total OA exuded amount was expressed as µg plant -1 h -1 . Root-mass based amounts of OA rhizo and OA exuded were expressed as µg g -1 root DW and µg g -1 root DW h -1 , respectively. Total OA amounts were converted to C mass using molecular weights and the number of C atoms for each OA to estimate the amount of C secreted from roots as OAs.

Soil nutrients
Plant-accessible nutrients were evaluated using ion exchange resin membranes (IEMs) buried in the top 0-6 cm of each pot, after modifying the method outlined in Hovenden et al. (2008). Anion and cation IEMs were cut into 1 cm × 6 cm pieces. Anion membranes were placed on one side of a plastic tag and cation membranes on the other side, collectively representing an IEM-tag. Two IEM-tags were inserted into the soil in each pot at the start of the CO 2 and water treatments. These tags were collected at harvest and washed with Milli-Q water to remove adhesive soil. Duplicate IEM-tags for each pot were placed in a falcon tube, mechanically shaken with 40 ml of 0.1 M hydrochloric acid for 1 h to extract nitrate (NO 3 --N), ammonium (NH 4 + -N) and phosphate (PO 4 3--P). This solution was then diluted ten times to determine extracted N and P concentrations using a continuous flow auto-analyser (AQ2 Discrete Analyzer, SEAL Analytical, Mequon, WI, USA).

Statistical analysis
Statistical analyses were performed using R 4.1.2 (R Core Team 2021). We report results at the significance level of ⍺ = 0.05, and marginally significant results (P ≤ 0.1) as an important indicator of treatment effects. Each of the studied species was analysed separately. There were 20 pots for E. tereticornis and 32 for each of M. stipoides and H. sericea, resulting in ten and 16 blocks of paired pots (low-and high-watered pots), respectively. Hence, the water treatments had treatment replicates of ten for E. tereticornis and 16 for M. stipoides and H. sericea. There were two glasshouse compartments for each of the CO 2 treatments, where blocks were regularly moved within and between the compartments to minimise variation in the measurements attributed to different compartments. Hence, each block (instead of the compartment) was treated as a replicate for CO 2 treatment-five replicates for E. tereticornis and eight replicates for M. stipoides and H. sericea.
Linear mixed-effects models (LMMs) were performed using the 'lme4' package (Bates et al. 2015) to examine CO 2 , water and their interactive effects on OA rhizo and OA exuded (total and root-mass based amount), with block as a random factor. LMMs were also used to evaluate treatment effects on plant/ soil variables (total biomass, RSR and soil nitrate, ammonium and phosphate availability, and the number of cluster roots per unit root biomass for H. sericea). Response variables were transformed (log e , square root or Box-Cox transformation) as required to ensure homogeneity of variances and normality of errors prior to analysis (Crawley 2012;Fox and Weisberg 2011). P values of the fixed factors in the LMMs (i.e. CO 2 , water and CO 2 × water) were approximated by F-test using Type II ANOVA tests with Kenward-Roger Degrees of Freedom, using the 'car' and 'lmerTest' packages (Fox and Weisberg 2011;Kenward and Roger 1997;Kuznetsova et al. 2017). The amounts of each OA in OA rhizo and OA exuded were evaluated individually to examine the response of each OA to CO 2 and water treatments. Since there were frequent zeros for some OAs (or not detectable) and the values were bound by zero, the distribution and error structure of the data were not normal. Therefore, we employed generalised linear mixed-effects models (GLMMs) using the Gamma distribution with log link, with block as a random factor. As the Gamma distribution can handle only non-zero positive values, a small value (i.e. the smallest value detected) was added to all OA values that had zeros prior to analysis. P values of the fixed factors in the GLMMs were obtained by Chi-squared tests.
Multiple regression analysis was performed to evaluate the associations between total OA amounts and plant/soil measurements for OA rhizo and OA exuded . RSR and soil ammonium availability were log e -transformed prior to the analysis. Explanatory variables were Z-standardised so that values had mean = 0 and standard deviation = 1. Ninety-five percent confidence intervals (CIs) on the coefficients of the multiple regression model were estimated by parametric bootstrap with 999 simulations using the 'lme4' package. Multicollinearity between the explanatory variables was checked using variance inflation factors (≤ 5). Where block did not explain any of the variations in response variables and did not improve the models, this random factor was removed from the LMMs, GLMMs and multiple regression analysis, and (generalised) linear models were performed instead to avoid a singular fit (Bates et al. 2015).
Redundancy analysis (RDA) was employed to assess CO 2 and water effects on the composition of OA rhizo and OA exuded for each species using the 'vegan' package (Oksanen et al. 2020). Compositional OA data were Hellinger-transformed prior to analysis. The interactive term of the combined CO 2 and water treatments did not improve adjusted R 2 values for the tested models and hence was not included in the analysis (Legendre and Legendre 2012). RDA models were tested with 4,999 permutations. Additional permutation tests were performed when significant associations between the OA composition and CO 2 and water treatments were indicated at ⍺ < 0.05 to analyse the marginal effects of each term. Correlations between the resulting RDA axes (RDA1 and RDA2) and plant/soil variables (total biomass, RSR, soil ammonium, nitrate and phosphate, and the number of cluster roots per unit root biomass for H. sericea) were also assessed. RSR was log e -transformed. Non-metric multidimensional scaling (NMDS) was used for OA rhizo and OA exuded to visualise the pattern of OA composition across the studied species, with the relative abundance of each OA computed and Hellinger-transformed for each plant prior to analysis.

Plant and soil measurements
The effects of CO 2 and water treatments on plant/soil measurements were species-specific. Total biomass generally increased in the eCO 2 treatment, by 20% (P > 0.1), 29% (P < 0.05) and 33% (P < 0.1) relative to aCO 2 , for E. tereticornis, H. sericea and M. stipoides, respectively (Fig. 1a, Table S1). No eCO 2 effects occurred for C investment in belowground biomass (i.e. RSR and the number of cluster roots per unit root biomass, Fig. 1b, c) or for soil nitrate concentrations in any studied species (Fig. 1d). The eCO 2 treatment increased soil ammonium concentrations for E. tereticornis, relative to aCO 2 (+ 139%, P < 0.001), while they did not change for H. sericea and decreased for M. stipoides (-49%, P < 0.001). The eCO 2 effects on soil ammonium concentrations were particularly large for low-watered E. tereticornis and M. stipoides compared to high-watered pots, as indicated by a significant interaction for both species (Fig. 1e). Soil phosphate concentrations were unaffected by the eCO 2 treatment for E. tereticornis or H. sericea, but significantly increased for M. stipoides (+ 79%, P < 0.01) relative to aCO 2 (Fig. 1f, Table S1).

Quantitative responses of organic acids in root exudates
The OA rhizo and OA exuded amounts varied among the three species but were within the same order of magnitude when expressed per unit root biomass (Fig. 2). On average, OA rhizo amounts were 376 ± 73, 836 ± 368 and 479 ± 84 µg g -1 root DW for E. tereticornis, H.

Effect of CO 2 and water treatments on exudation
For E. tereticornis and H. sericea, there was no evidence of eCO 2 effects on the total amounts of OA rhizo or OA exuded regardless of the units of expression (per plant or unit root biomass) (Figs. 2, S1, Table S2). For M. stipoides, eCO 2 did not alter OA rhizo amount but more than doubled OA exuded amount per plant (P < 0.05, Fig. 2) and per unit root biomass (P < 0.05, Fig. S1b, Table S2), increasing the amount of C secreted from roots from 2.6 to 5.6 ug C h -1 per plant (Fig. S2b).
Water treatments did not affect the total amount of OA rhizo or OA exuded for E. tereticornis and M. stipoides, regardless of expressed units (per plant or unit root biomass) (Figs. 2, S1, Table S2). While this was also the case for total OA rhizo amount per plant in H. sericea, (Fig. 2a), the low-watered plants had 77% less total OA rhizo amount per unit root biomass than high-watered plants of this species (P < 0.05, Fig. S1a, Table S2), decreasing the amount of C in OA rhizo from 470 to 104 µg C g -1 root DW (Fig. S2). Lowwatered H. sericea plants also had greater OA exuded amounts per plant (+ 116%, P < 0.05, Fig. 2b) and per unit root biomass (+ 32%, P > 0.1, Fig. S1b) than highwatered plants, although the latter was not statistically significant (Table S2), increasing the amount of C in OA exuded from 4.7 to 9.4 µg C plant -1 h -1 (Fig. S2).

Association between exudates and plant/soil measurements
For E. tereticornis, total OA rhizo amount had a significant positive association with total biomass, but total OA exuded did not correlate with any tested variable (Table S3). No associations were found between total OA rhizo amount and any tested variable for H. sericia, but total OA exuded was positively correlated with total biomass and RSR (Table S3). OA rhizo amount in M. stipoides positively correlated with total biomass and soil phosphate concentrations, and OA exuded amount was positively correlated with total biomass (Table S3).

Effect of CO 2 and water treatments on OA composition
For E. tereticornis, the CO 2 treatment significantly altered OA rhizo composition (P < 0.01, Fig. 4a, Table S5), likely driven by increased malic acid at eCO 2 relative to aCO 2 (+ 322%, P < 0.05, Table S4), but did not affect OA exuded composition (P > 0.1, Fig. 4d). CO 2 treatment did not affect OA rhizo or OA exuded composition for H. sericea (Table S5, Fig. 4b, e). For M. stipoides, CO 2 treatment significantly altered OA rhizo composition (P < 0.05, Fig. 4c), despite no main CO 2 effects on any detected OAs when analysed individually (Table S4). However, an interaction occurred between the CO 2 and water treatments for citric acid in OA rhizo for M. stipoides: eCO 2 significantly increased the amount of citric acid relative to aCO 2 under high-watered but not low-watered  (Table S4). A similar pattern occurred for malic acid in OA rhizo , but this was not statistically significant. For M. stipoides, the CO 2 treatment weakly influenced OA exuded composition (P < 0.1, Fig. 4f, Table S5), possibly due to changes in shikimic acid, as its exudation rate was sevenfold higher under eCO 2 than aCO 2 (Table S4).
The water treatment significantly altered OA rhizo composition (P < 0.05, Fig. 4a, Table S5) in E. tereticornis, likely driven by increased oxalic acid under low-watered compared to high-watered conditions (+ 80%, P < 0.05, Table S4), but not OA exuded composition (P > 0.1, Fig. 4d, Table S5). Hakea sericea OA rhizo and OA exuded composition were unaffected by the water treatment (Fig. 4b, e, Table S5). For M. stipoides, the water treatment significantly altered OA rhizo composition (P < 0.05, Fig. 4c, Table S5), likely driven by increased citric acid (+ 100%, P < 0.05) and decreased malic (-45%, P < 0.01) and shikimic acids (-40%, P < 0.1) under low-watered compared to high-watered conditions (Table S4). Changes in oxalic acid may have also contributed to the shift in OA rhizo composition as it was only observed in low-watered M. stipoides. For this species the water treatment also significantly altered OA exuded composition (P < 0.05, Fig. 4f, Table S5), likely due to increased citric acid under low-watered compared to high-watered plants (+ 215%, P < 0.05, Table S4). Furthermore, malic acid may have contributed to the compositional shift in OA exuded as indicated by its  stipoides (c, f). The results of permutation tests for the significance of constraints are shown for CO 2 (C), water (W) and their interaction (CW) at the top of each panel, with significance codes. Proportion of variability in OA composition attributed to RDA axes is shown within parentheses. The correlation test results for the RDA axes and plant and environmental measurements are shown with arrows: root:shoot ratios (RS), total biomass (Biom), soil nitrate, ammonium and phosphate, and number of cluster roots per g root dry weight (CR; only for H. sericea), with significance codes. The significance codes are P < 0.1 ( †), < 0.05 (*), < 0.01 (**) and < 0.001 (***). OAs are indicated in purple: acetic (Ace.), cis-aconitic (c-Aco.), citric (Cit.), fumaric (Fum.), malic (Mal.), oxalic (Oxa.), shikimic (Shi.) and transaconitic (t-Aco.) acids relatively larger negative value for the species score along the first RDA axis accounting for 11.0% of the total variability (Fig. 4f). The amount of malic acid in OA exuded for M. stipoides was so low that it was undetectable in some pots. While there was no evidence of CO 2 or water treatment effects on malic acid, this OA was detected more often in high-watered plants at eCO 2 (7 of 8 pots), with higher mean levels, than other conditions (2-3 of 8 pots) (Table S4).

Association between OA composition and plant/soil measurements
As mentioned above, the CO 2 and water treatments significantly altered OA rhizo composition in E. tereticornis, with the RDA axes correlated with soil phosphate concentrations (P < 0.05) and plant biomass (P < 0.1, Fig. 4a). Along the first RDA axis, accounting for 24% of the total variation in data, soil phosphate concentrations negatively correlated with oxalic acid concentrations in low-watered pots under aCO 2 and positively correlated with malic acid in high-watered pots under eCO 2 (also see Fig. 3a). For M. stipoides, the CO 2 and water treatments strongly influenced OA rhizo and OA exuded composition, with RDA axes strongly correlated with changes in total biomass and soil phosphate concentrations (Fig. 4c, f); under eCO 2 :High-watered conditions (with a relatively high proportion of malic acid) OA composition correlated with high soil phosphate concentrations, whereas under eCO 2 :Low-watered conditions (with a relatively larger proportion of citric acid) OA composition correlated with high total biomass (also see Fig. 3a, b).

Discussion
We performed a unique study demonstrating CO 2 concentration and water availability caused quantitative and qualitative changes in OA root exudates. Plants grown under low-watered conditions outperformed those under high-watered conditions for all three species. The H. sericea and M. stipoides seeds used in our study were collected from the local native Eucalyptus woodland where soil rarely reaches its full WHC (Hasegawa et al. 2018). Thus, the provenances of H. sericea and M. stipoides used in this study may have been well-adapted to the local hot, dry climate with well-drained sandy soil (Pathare et al. 2017;Piñeiro et al. 2021). Meanwhile, insufficient water availability is generally known to inhibit the growth of a wide range of Eucalyptus species (Atwell et al. 2007;Ngugi et al. 2003). It is possible that highwatered plants in this study experienced anaerobic soil conditions. Associated negative effects of plant pathogens may also have played a role, although there were no visual signs of fungal infection on any of the studied plants.
Organic acid exudation of the studied species Overall, the total amount of OAs in root exudates (on a per plant basis) was mainly driven by plant size: larger plants within the same species had higher exudation rates. Also, significant shifts in OA composition in response to CO 2 and/or water treatments were accompanied by changes in plant biomass. Thus, CO 2 and water availability likely indirectly influence OA amount and composition through altered plant size (Aulakh et al. 2001;Calvo et al. 2019). Kidd et al. (2018) collected OAs in root exudates from 10 pasture legume species using the same technique as the present study, reporting OA rhizo amounts from 12-191 µmol g -1 root DW which are generally larger amounts than herein (3.5-6.7 µmol g -1 root DW; data not shown). However, when expressed on a root length basis, instead of root mass basis, OA amounts in Kidd et al. (2018) ranged from < 1-14 nmol cm -1 root (except Cicer arietinum L. with 68 nmol cm -1 root), which are similar to our amounts (7-13 nmol cm -1 root, data not shown). Our study species had lower specific root lengths (or thicker roots) and hence smaller surface area per unit of root mass than those in Kidd et al. (2018). These differences in root architecture may explain differences between our native woodland species and the pasture legumes reported in Kidd et al. (2018). It should also be noted that the growth media differed between the two studies: Kidd et al. (2018) grew plants in washed river sand, which minimises the adsorption and microbial decomposition of OAs, likely resulting in greater amounts of rhizosphere OA than those in native soil (Ryan et al. 2012). Roelofs et al. (2001) evaluated OA exudation from seven Australian Proteaceae (Banksia, Hakea and Dryandra) species grown in nutrient solutions, reporting OA exudation rates from cluster roots ranging from 1.0-2.5 nmol g −1 root fresh weight s -1 . If root DW is assumed to be 10% of fresh weight, as assumed by Roelofs et al. (2001), these figures equate to 36-90 µmol g -1 root DW h -1 , which is higher than the value for H. sericea in this study (6.7 ± 2.9 µmol g -1 root DW h -1 , mean ± 1 standard error, data not shown). This was likely due to our collection of root exudates from the whole root system, including cluster and non-cluster roots, whereas Roelofs et al. (2001) only sampled from cluster roots. The OA exuded from H. sericea in our study was predominantly represented by trans-aconitic acid, while citric and malic acids were undetected. This compositional pattern is more similar to non-cluster than cluster roots (Roelofs et al. 2001), suggesting that cluster roots may have played a relatively small role in OAs of root exudates in the present study.
Organic acid composition for the two OA collection methods (OA rhizo or OA exuded ) were generally in agreement for the studied species (Fig. 3a, b), suggesting that the loss of each OA by microbial degradation or mineral adsorption in the rhizosphere was in proportion to what were produced by plants and hence the treatment effects were similar for those two measures. However, for M. stipoides acetic acid was observed in OA rhizo but not in OA exuded : this likely reflected acetic acid being derived from rhizosphere microbes (Suriyagoda et al. 2016). On the other hand, fumaric acid from M. stipoides and E. tereticornis was less abundant in OA rhizo than OA exuded , suggesting it was rapidly degraded by microbes or adsorbed onto soil particles.
Effect of CO 2 treatment on exudation For E. tereticornis, there was no evidence of a CO 2 effect on OA amounts in root exudates. Given the lack of CO 2 effect on total biomass or RSR, these results suggest that E. tereticornis did not alter its belowground C allocation to facilitate increased nutrient uptake in response to eCO 2 .
For H. sericea, there was no evidence of CO 2 effects on OA exuded or OA rhizo amount. There was also no evidence of a CO 2 effect on cluster root production, which contrasts with Campbell and Sage (2002) who reported an increased number of cluster roots in white lupin (Lupinus albus) in response to eCO 2 using a hydroponic technique. A plethora of hydroponic studies report increased cluster root production with reduced P supply (e.g. Neumann et al. 1999;Shane et al. 2003). As argued above, the OA composition of H. sericea in this study was similar to non-cluster roots rather than cluster roots. Combined with a significant increase in biomass of under eCO 2 , it appears that H. sericea had adequate access to P without the need to invest C into further cluster root production or root exudation.
Increased OA exuded amounts in M. stipoides under eCO 2 were likely due to increased photosynthetic rates as observed in the field study where there was a 28% increase in net CO 2 assimilation rates (Pathare et al. 2017). Given the lack of CO 2 effects on RSR, larger amounts of OA exuded per unit root biomass suggest a greater proportional C investment into exudation under eCO 2 for this species.

Effect of water treatment on exudation
For E. tereticornis, the water treatment did not change the OA amounts in root exudates, despite higher total biomass and lower RSR for the low-watered plants. Together with slightly smaller amounts of root exudates per unit root biomass, this suggests that E. tereticornis seedlings invested less C in OAs or belowground biomass and more C in aboveground growth under low-watered conditions than high-watered conditions. This may be related to its survival strategy at an early phase of juvenile seedling development when it is typically exposed to intense competition for light with surrounding herbaceous species on the forest floor (Collins et al. 2018).
Hakea sericea, on the other hand, had greater total amounts of OA exuded per plant and lower total OA rhizo amount per unit root mass under low-watered conditions relative to high-watered conditions. While this species allocated more biomass belowground under low-water conditions (increased total biomass and RSR), the number of cluster roots per unit root biomass decreased. These results suggest that H. sericea invested C into developing non-cluster roots under low-watered conditions, and its larger root system could maintain similar whole plant OA exuded levels under both watering regimes.
Higher plant growth for M. stipoides under low-watered than high-watered conditions did not increase the total amount of OAs exudated, as they were counterbalanced by smaller amounts per unit root biomass. Thus, M. stipoides did not invest more C into exudation under low-watered conditions, despite greater plant growth.
Effect of CO 2 and water treatments on OA composition in association with plant/soil measurements CO 2 concentrations and water availability altered OA composition for E. tereticornis and M. stipoides but not for H. sericea, mainly via changes in malic acid, accompanied by changes in soil phosphate concentrations. Given the positive relationship between the total OA rhizo amount for M. stipoides (but not E. tereticornis) and soil phosphate, with malic acid the primary OA for this species, these results suggest that M. stipoides exposed to eCO 2 (or low-watered conditions) secretes a relatively larger (or smaller) quantity of malic acid in the rhizosphere, facilitating (or suppressing) soil P mobilisation (Bais et al. 2006;Dakora and Phillips 2002;Ryan et al. 2001). Additional (or alternative) explanations for decreased soil P availability at lowwatered conditions are that greater plant growth, and therefore P uptake, reduced available P and/or that low water availability restricted P mobility in the soil. Despite E. tereticornis increasing the amount of malic acid exuded in the rhizosphere in response to eCO 2 , it did not change the total OA rhizo amount or soil phosphate concentration. Furthermore, notwithstanding the higher P mobilisation efficiency of oxalic than malic acids (on a molar concentration basis) (Mendes et al. 2020), the increased amount of oxalic acid in OA rhizo in low-watered E. tereticornis plants co-occurred with decreased soil phosphate concentrations. While many hydroponic studies have demonstrated increased rates of root exudation at low P supply to facilitate P uptake (e.g. Khorassani et al. 2011), responses differ among organic compound types. For instance, O'Sullivan et al. (2021) reported increased rates of citric and malic acids in response to eCO 2 and low P supply in white lupin while fumaric acid was unaltered. In our study, CO 2 and water treatment did not influence the C allocation of E. tereticornis to root exudation but shifted OA composition; however, this did not seem to facilitate P mobilisation. The relationship between particular OAs and soil P availability is highly species-specific, especially when using soil from a natural ecosystem.
Estimating C fluxes from root exudates of OAs in a Eucalyptus woodland under eCO 2 Although it varies substantially among plant species, roots secrete approximately 11-40% of assimilated C (Badri and Vivanco 2009;Jones et al. 2009), with OAs accounting for a significant fraction (Jones 1998). For instance, OAs are responsible for 16-31% C in root exudates of maize (Gransee and Wittenmayer 2000). Hence, changes in the exudation rates of OAs could significantly change the amount of C secreted from plant roots to soil. Our study provides empirical estimates of the amounts of C secreted from the studied plant species in the form of OAs. Our study was motivated by the EucFACE experiment where M. stipoides is responsible for ~ 99% of the understorey vegetation biomass (Collins et al. 2018;Hasegawa et al. 2018). Given that the soil and seeds of M. stipoides used in this study were collected from this site, the figures obtained from M. stipoides may be relatively realistic estimates of C secreted from roots under field conditions, while those for E. tereticornis and H. sericea may not be comparable to those in the field as they were juvenile seedlings when harvested. If we focus on the exudation of M. stipoides, we can derive estimates of belowground C inputs from the grassy understorey: OA rhizo amounts were 4.9 ± 1.6 (aCO 2 ) and 9.0 ± 1.7 (eCO 2 ) µg C plant -1 and OA exuded amounts were 2.6 ± 1.2 (aCO 2 ) and 5.6 ± 1.4 (eCO 2 ) µg C plant -1 h -1 , with average aboveground biomass of 0.3 ± 0.03 and 0.4 ± 0.04 g plant -1 , respectively (mean ± 1 standard error). Collins et al. (2018) estimated the aboveground biomass of the understorey at EucFACE to be ~ 329 g m -2 . Using this value, we can approximate that M. stipoides secretes 4.5 ± 1.3 (aCO 2 ) and 9.8 ± 2.5 (eCO 2 ) mg C m -2 in the rhizosphere and 2.1 ± 0.9 (aCO 2 ) and 4.8 ± 1.0 (eCO 2 ) mg C m -2 h -1 from roots at the EucFACE site. The annual gross primary productivity in the understorey estimated from M. stipoides is 497 and 552 g C m -2 year -1 in aCO 2 and eCO 2 plots, respectively (Jiang et al. 2020). Thus, M. stipoides roots may have transferred a non-negligible fraction of assimilated C into this system's soil in the form of OAs. At EucFACE, low soil P availability constrained the positive CO 2 effects on the overstorey growth of E. tereticornis , whereas M. stipoides increased its dominance after three years of CO 2 fumigation (Hasegawa et al. 2018). Given that M. stipoides dominated the understorey coverage of this site, it may have contributed to eCO 2 -induced increases in soil P (as reported in Hasegawa et al. (2016)) via qualitative and/or quantitative changes in root exudates and further supported its growth stimulation by eCO 2 .
It is often argued that additional C investment into mycorrhizae through root exudation at eCO 2 facilitates plant nutrient uptake via enhanced SOM decomposition (priming) ). However, priming or enhanced mycorrhizal activity may not be the driver of the increased P availability in this system because neither mycorrhizal biomass (both arbuscular-and ecto-mycorrhizae), nor extracellular phosphatase activity, were altered by the CO 2 treatment (Castañeda-Gómez et al. 2021). Castañeda-Gómez et al. (2022) further explored the role of arbuscular mycorrhizae in SOM decomposition and P uptake by M. stipoides under eCO 2 conditions with a pot study, demonstrating that eCO 2 did not alter SOM decomposition regardless of the presence of arbuscular mycorrhizal fungi. Thus, the increased P availability at eCO 2 observed at the EucFACE may be derived from inorganic P liberated by OAs from chemical bindings with cations or mineral surfaces and not from the decomposition of organic P by mycorrhizal fungi.

Conclusion
We found that CO 2 and water treatments influenced OA amount and composition via increased plant biomass. In general, bigger plants within a species had greater amounts of exudates. When expressed per unit root biomass, OA amounts increased in response to eCO 2 for M. stipoides, and low-watered H. sericea plants exuded lower amounts of OAs in the rhizosphere than high-watered plants. Compositional changes in OAs driven by altered exudation of malic acid were associated with changes in soil P availability, possibly explaining the decreased P availability under low-watered conditions and increased P at eCO 2 for M. stipoides (but not E. tereticornis). Together with previous studies reporting eCO 2 -induced increases in soil phosphate availability and soil C fluxes at the Euc-FACE experiment, our study suggests that changes in root exudation of OAs in M. stipoides could significantly contribute to soil nutrient and C cycling in this Eucalyptus woodland. Thus, increased CO 2 concentrations and drastic changes in precipitation patterns (and resultant soil water availability) will likely influence OA amount and composition of the studied Australian native species in low P soil. Quantitative and qualitative changes in root exudates can alter abiotic and biotic interactions in the rhizosphere (de Vries et al. 2019;Drake et al. 2013) and modify edaphic structure by influencing soil aggregates and organo-mineral compounds (Bronick and Lal 2005;Keiluweit et al. 2015), such that the observed changes in OAs in root exudates in this study can have cascading impacts on ecosystem functioning.
Author contributions All authors contributed to the study conception and design. Shun Hasegawa led material preparation, data collection and statistical analysis. Megan H. Ryan led root exudate analysis. The first draft of the manuscript was written by Shun Hasegawa and Sally A. Power, and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.
Funding Open access funding provided by Umea University. Shun Hasegawa received Endeavour Scholarships and Fellowships from the Australian Government.

Data availability
The raw data and the R scripts that reproduce the results presented in this paper are available to download from Zenodo: https:// doi. org/ 10. 5281/ zenodo. 75222 77.

Declarations
Competing interests The authors have no relevant financial or non-financial interests to disclose.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.