Seasonal Drought Induces Hydraulic Dysfunction, Not Carbohydrate Depletion, for Robinia Pseudoacacia in a Semi-humid Forest

Shifts in rainfall patterns that are associated with climate change are likely to cause widespread forest decline in regions where droughts are predicted to increase in duration and severity. However, causes of forest decline and their physiological mechanisms remain unclear, particularly the roles of carbon metabolism and xylem function. To explore the response of hydraulic architecture and non-structural carbohydrates (NSC) traits under seasonal drought, we conducted a manipulation experiment in a Robinia pseudoacacia plantation in 2015 and 2016 in Loess Plateau of China. Sap-ow, leaf area index, water potential, non-structural carbohydrate concentrations, and hydraulics in different organs were measured.


Background
Reduced precipitation may be responsible for tree mortality events across different forest and woodland ecosystems (Fensham et al. 2009; Allen et al. 2010Allen et al. , 2015Lopez et al. 2021). However, in many cases, mortality rates vary among species or functional types (Breshears et al. 2005;Fensham et al. 2009;Mitchell et al. 2014) suggesting differences in drought resistance among co-occurring species. Despite improved understanding on how water de cits affect plant functioning and physiological mechanisms (Mitchell et al. 2013; Anderegg et al. 2016), the controls on species survival under extreme drought remain poorly resolved and limit our ability to adequately predict future changes in ecosystem structure and function.
Recent researches have studied physiological mechanisms that have been linked to a plants capacity to regulate carbon and water balance under drought conditions (Mitchell et al. 2014;Anderegg et al. 2012Anderegg et al. , 2016. Two main physiological mechanisms to explain the eventual drought-induced mortality of trees have been formalized by McDowell et al. (2008): hydraulic failure and carbon starvation. The framework suggests that, for any given drought event, both plant hydraulic strategy and meteorological conditions will de ne the 'physiological drought' experienced by the plant. Physiological drought is a function of plant regulation of water-use in response to declining soil water potential and thresholds associated with hydraulic-or carbohydrate-mediated mortality (Mitchell et al. 2013). For example, hydraulic failure occurs when plant water potential falls below a critical value, which causes xylem cavitation and conductivity loss (Adams et al. 2009). Consequently, failure to supply su cient carbon substrate to metabolism following drought-related reductions in photosynthesis and increased use of NSC occur, theoretically leading to carbon starvation (McDowell 2008;2011). This conceptual framework has been extensively studied in recent years (Mitchell et  An initial experimental test of these mechanisms was conducted inside an environmentally-controlled greenhouse with transplanted Pinus edulis under two temperature regimes (Adams et al. 2009). Subsequent attempts postulated that tree mortality due to drought is a complex process that can occur by multiple mechanisms (Sala et al. 2010). For example, hydraulic failure is expected to proceed more rapidly leading to fast mortality in plants that keep their stomata open during drought (i.e., relatively anisohydric plants) (McDowell et al. 2008). Under severe drought, incapacity to regulate plant water status above critical thresholds will promote xylem cavitation and death through dehydration (Mitchell et al. 2013). Radial stem growth, transpiration, and leaf water potential are signi cantly affected during chronic drought and as a consequence of these drought responses, temporal patterns of NSC can occur (Mitchell et al. 2014). During mild or short droughts, however, NSC reserves have been found to increase in multiple tree species (Sala et al. 2010;Palacio et al. 2014). Therefore, carbon starvation could occur from a reduction in tree carbohydrate reserves below a survival threshold during severe drought if these resources became inaccessible through inhibition of conversion of stored starch to sugar (Knoblauch and Robinia pseudoacacia is a fast-growing tree with pro igate water use. It has been introduced widely for plantations in the Loess Plateau of China due to its potential to adapt to drought and low soil fertility, and because of its ability to x nitrogen. However, soil desiccation has caused degradation of R. pseudoacacia forest plantations whereby mature trees have developed abnormally short trunks (Chen et al. 2010;Jia et al. 2017). Previous studies examined carbon starvation and hydraulic failure in potted experiments with R. pseudoacacia (Zhang et al. 2015; Yan et al. 2020). They found that only severe drought stress signi cantly decreased NSC reserves, but it increased the loss of conductivity, which re ected the intrinsic drought resistance of xylem tissues. However, no previous studies have measured both carbon stress and hydraulic stress in mature and drought-stressed trees directly, which makes it di cult to distinguish between the interpretations of carbon stress and hydraulic stress.
To address these shortcomings, we conducted a two growing seasons of experiment, and tested the relationships between drought stress, NSC dynamics, and plant-water relations in a mature forest setting. With this experimental design, we examined the response of hydraulic architecture and NSC components under two growing seasons of drought in R. pseudoacacia in a seasonally temperate forest in the Loess Plateau of China. We hypothesized that tight hydraulic regulation of water and carbon uptake reduces the risk of rapid hydraulic failure, but may also result in increased duration of the physiological drought, promoting depletion of nonstructural carbohydrates.

Site description
Our study was conducted at Yehe National Forestry Center in the Qishui watershed, which had a 12-yearold R. pseudoacacia plantation located in Fufeng County, Shaanxi Province, which is situated on the southern Loess Plateau in China (34.55°N, 107.90°E; 1080 m a.s.l.). The forest covers an area of 2980 ha. The soil in the area is classi ed as Gleyic Phaeozems, and according to the World Reference Base for Soil Resources, these soils are 11% sand, 20% clay, and 69% silt (Zhang et al. 2018). Stand age is ten, stand density is ∼1300 trees ha − 1 , mean tree height is 9.23 ± 1.22 m, and mean diameter at breast height (DBH) is 8.65 ± 0.86 cm, with an average leaf area index (LAI) of ~ 2.05 (based on digital hemispherical photographs). Throughout the region, arti cial afforestation communities, which included R. pseudoacacia and Pinus tabuliformis, were established during the end of the 20th century. R. pseudoacacia accounted for more than 90% of the forest coverage. The dominant understory vegetation was Stipa bungeana and Artemisia argyi, which had a coverage of 80% − 90%.
The area has a semi-humid, temperate, continental climate. Mean annual precipitation was 592 mm, 80 % of which falls between June and October based on 45 years of meteorological data . Mean annual temperature was 11.5°C, and mean monthly temperature ranged from − 2°C in January to 26°C in July. From drought and ambient control. Treatments were applied to 20 × 20 m plots (400 m 2 ) that contained 50 target trees each. The two treatments were replicated in three randomly selected blocks for a total of six experimental treatment plots. The blocks were at least 500 m apart, and each had the same soil type and similar topography. In the drought treatment, rainfall was reduced by approximately 50% relative to ambient using a system of plastic panels and plastic-lined gutters that were xed to rails approximately 1 m high. Drainage tunnels (30 cm deep) were trenched to remove the effect of through-ow of soil water. To control for any temporary damage to roots from the trenching, the control plot was also trenched to the same depth.
In 2014, an instrumented ux tower was installed near the tower base which continuously measured major environmental variables . Soil volumetric water content (VWC) was measured continuously using eight electrically conductive sensors (EC-5, Decagon, USA). The sensors were installed around the trees at depths of 10, 20, 40, 60, 80, and 100 cm in one drought plot and one control plot, and measured from June 2015 to December 2016 (Zhang et al. 2018). The VWC were powered and controlled a data logger (CR1000 Campbell Scienti c, Logan UT, USA) which scanned every 30 s, and this was recorded the value of averaging with 10 min. Sap-ow and leaf area index Six R. pseudoacacia trees were selected randomly for sap ow measurements from the drought and control plots. The thermal dissipation probe (TDP) method (Granier 1987) was used to measure the sapow density, and it was calibrated with the empirical equation established by Ma et al. (2017). We used 10-mm-long sensors, and the detailed information about the sensors is described in Peng et al. (2015). All of the TDP sensors were connected to a CR1000 data logger with an AM16/32 multiplexer (Campbell Scienti c, Logan UT, USA). The temperature difference between the upper heated probe and the lower reference probe was measured every 30 s, and this was recorded the value of averaging with 10 min. The stand transpiration (Q, mm) was calculated as the product of sap-ow density and sapwood area of the plot. The sapwood area was determined by cutting neighboring trees and discriminating between the sapwood and the heartwood based on color differences.
The LAI for the two treatments was estimated indirectly twice a month at the center of each plot. The LAI was calculated in June-November 2015 and April-November 2016 by processing digital hemispherical photographs, which were collected under uniform sky conditions near dawn and sunset on clear days. In all cases, current year extension growth from the uppermost branches exposed to full sun was sampled. We limited the measured material to current year growth to avoid embolism that originated from conditions prior to the growing season (i.e. no freeze-thaw embolism) and to facilitate the cleanest cuts possible, thus reducing the possibility that damage to the xylem during harvest would affect our results. Zhang and Wang (2016) showed that the mean vessel length for R. pseudoacacia is 19.6 cm. In our study, therefore, branches longer than 50 cm were cut from the trees in the early morning and placed immediately in black plastic bags to prevent dehydration. We also collected coarse root (diameter about 3 cm) segments longer than 40 cm which were immediately placed in black plastic bags. Branch and root samples were collected using the same protocol as above 'Leaf water potential measurements' occurred during October 2015 and October 2016, and for these measurements, we used the same sample trees as we used for the analysis of leaf water potential. Excised samples were immersed in water immediately and transported to the laboratory within 1 hour.
For branches and roots, we measured native and vacuum-in ltrated hydraulic conductivity. The exposed segments were cut underwater into 30-cm-long segments. The sections were attached to a tube that was lled with degassed and 0.2 µm-ltered 20 mmol/L KCl solution, and a gravitational pressure of 0.01 MPa was used to force the solution through the segments. Hydraulic conductivity was measured with a 0.1 mg precision balance. Then, we vacuum-in ltrated the samples for 1 h to re ll any embolized conduits, and we re-measured conductivity (Kolb et al. 1996). We calculated percent loss of conductivity (PLC, %) as the percent difference between native and vacuum-in ltrated conductivity. Note that in all cases, measurement segments excised under water may prevent artifacts (Sperry 2013).
Branch and root vulnerability curves (VCs) can be used to quantify the percentage loss of conductivity by air-injection method (Cai et al. 2010;Venturas et al. 2015). These partially debarked segments slightly longer than 30 cm were mounted in the double ended pressure collar. Each segment was ushed at 200 kPa water pressure for 30 min to remove embolism. The basal cut end was connected to a 1 m-long tube of approximately 25 mm in internal diameter, which is large enough to allow the escape of air bubbles coming out of the stem. The pressure was increased from 0 to 4 MPa, with each step followed by 10 min of relaxation, and ow was then measured at a constant background of 100 kPa air pressure until 88% of conductivity was lost. From the sigmoid shape of a VCs, the three variables were acquired:(i) the xylem pressure (Ψ x ) that corresponded to 50% loss of hydraulic conductivity (Ψ 50 ), the most commonly used index of embolism resistance; (ii) the air entry threshold (Ψ e ) that corresponded to about 12% loss of conductivity, where PLC begins to rise sharply with declining Ψ X ; and (iii) Ψ e -Ψ 50 , which was de ned as the hydraulic safety margin, where a larger value of Ψ e -Ψ 50 indicated a more gradual rise of PLC after Ψ x had fallen below Ψ e (Meinzer et al. 2009).

NSC analysis
Using samples cut by a tree climber from fully sunlit twigs, ve leaves were taken from the same trees as for PLC analysis. Samples of tree stem tissue were taken using a 4.3 mm diameter increment borer (Häglof Company Group, Långsele, Sweden) at breast height; this sampling was collected from ve different trees to avoid excessive damage that would occur by repeated boring the same tree. The coarse roots (diameter > 5 mm) were also collected from different trees. All these samples were collected late in the morning and were immediately labeled and frozen by immersion in liquid nitrogen. Samples were collected in June and November 2015, and April to October 2016.
Here, NSCs are de ned as starch plus soluble sugars, which included sucrose, glucose, and fructose. All samples were oven-dried at 70°C to a constant mass and then ne ground to a powder with a ball mill (FOSS CT410, Sweden). The analyses for determining NSC followed Andrew et al. (2013). About 0.5 g of each plant organ powder was vacuum-in ltrated with 80% ethanol for 15 min, and then it was centrifuged at 7000 rpm for 10 min. A further two extractions were carried out with 80% ethanol. Supernatants were combined, ltered through a 0.45-µm syringe lter, and analyzed for soluble sugars. The ethanol-insoluble pellets were used to determine starch content.
The ethanol-soluble fractions were analyzed using a Waters Alliance high-pressure liquid chromatographic (HPLC) system (Milford, MA, USA) (Andrew et al. 2013). The separated soluble sugars were identi ed and quanti ed with known standards and expressed as per cent of dry matter.
Starch was extracted from the remaining dry matter in a boiling solution of 0.02 NaOH for 1 h, followed by hydrolysis to glucose with α-amyloglucosidase (EC 3.2.1.3, Boehringer Mannheim Biochemicals, Mannheim, Germany). Glucose that was formed by hydrolysis was measured colorimetrically at 340 nm (spectrophotometer model 2550, Shimadzu, Japan). Starch concentrations were calculated from standard curves and expressed as per cent of dry matter.

Statistical analyses
A one-way analysis of variance (ANOVA) was used to test the effect of treatment on DBH, LAI, leaf water potential, PLC, and plant water use at each sampling date (P < 0.05). A one-way ANOVA was used to test the effect of treatment on the sugar and starch concentrations for each plant organ (leaf, stem, and root) at each sampling date (P < 0.05). Relationships between Ψ x (negative value) and PLC in stems and roots

Plant water status
Daily stand transpiration (Q d ) was obtained for each measurement day. The Q d ranged from 0 to 2.84 mm, and it exhibited pronounced diurnal variations (Fig. 2a). The daily average Q d (1.06 mm day − 1 ) in the drought treatment was signi cantly lower than in the control (1.75 mm day − 1 ) (P = 0.045) due to the signi cant decrease in LAI (Fig. 2b). During the measurement period in 2015, total water use was similar between the drought treatment and the control. In 2016, however, total water use in the drought treatment (272 mm) was signi cantly lower than that of the control (378 mm) (P = 0.038) (Fig. 3).
As the drought treatment progressed, Ψ pd uctuated between − 0.6 and − 2.6 MPa, with an average of -1.7 MPa. The Ψ md exhibited a steady decline throughout the drought and reached a mean minimum of − 3.0 MPa after two growing seasons of drought (Fig. 4). Both Ψ pd and Ψ md in the drought treatment were signi cantly lower than in the control since June 2016 (P = 0.036). Control trees displayed comparatively tighter stomatal regulation, where monthly stand transpiration (Q m ) declined more rapidly with changes in Ψ md than in the drought (Fig. 5). These patterns were also re ected in the rate of water loss in response to declining water availability, as indicated by the slope of Ψ md compared with Q m , which was larger for the control (50 mm) and signi cantly smaller for the drought (22 mm; Fig. 5).
Branches in the drought trees had signi cantly higher percentage loss of conductivity than their paired control baseline in 2016 (P < 0.001). Also, the percentage loss of conductivity of root segments in the drought trees (71%) was signi cantly higher than that of the control trees (37%) in 2016 (P < 0.001, Fig. 6). Thus, drought led to notable increases in loss of hydraulic function in this experiment, especially in roots. We also determined xylem Ψ 50 and Ψ e of branches and roots for the control and drought plots.
After two growing seasons, Ψ 50 in stems and roots exhibited approximately − 1.8 and − 1.4 MPa in control trees, respectively, compared with − 1.2 and − 0.8 MPa in drought trees, respectively. The Ψ e in stems and roots showed similar patterns. The hydraulic safety margin expressed as Ψ e -Ψ 50 in the control was 40% − 70% higher compared with the drought plots (Fig. 7).

Patterns of NSC
The seasonal changes in sugar and starch concentrations, which varied by 50% − 90%, were much larger than any differences associated with the drought treatment (Fig. 8). As the 2016 growing season progressed, leaf sugar concentration increased on average from 3.2-6.6%, but starch decreased on average from 5.9-3.1% (Figs. 8a and 8b). For stems and roots, sugar and starch concentrations increased consistently throughout the growing season.
Different tree tissues (leaves, stems, and roots) showed different responses in NSC concentration to the progression of the drought (Fig. 8). At the start of the drought experiment, leaf sugar concentrations were similar for droughted and control trees (means 2.5 ± 0.2 and 2.6 ± 0.2% dry mass, respectively, P = 0.169). For droughted trees, we found a signi cant decline in leaf sugar concentration over the course of the experiment (Fig. 8a, P = 0.048). This pattern was also found in leaf starch (Fig. 8b). Neither the concentrations of soluble sugar nor starch were depleted signi cantly in stems and roots from the drought treatment relative to the control (Fig. 8c, d, e, f).

Hydraulic responses to drought
To examine the role of hydraulics responses to drought, we measured native and vacuum-in ltrated hydraulic conductivities of stems and roots. Our data showed that two growing seasons of drought could lead to notable increases in loss of hydraulic function of root and branch segments (Fig. 6). It seems that hydraulic failure may have been induced by xylem tension exceeding cavitation thresholds due to progressive water loss, for example, through leaf and cuticular tissues, or when fast decline in water potential occurred (Fig. 4, Sevanto et al. 2014). In addition, in our study, we have identi ed two alternative hydraulic safety parameters (Ψ 50 and Ψ e ) referenced to species-speci c values of daily minimum leaf water potential (Fig. 7). Physiologically, attainment of Ψ 50 indicates that nearly catastrophic hydraulic failure has already occurred and that the risk of further runaway embolism is high because xylem pressure is operating along the steepest portion of the vulnerability curve (Tyree and Sperry 1988). Thus, the relevance of Ψ 50 as an indicator of resistance to hydraulic failure may be restricted to extreme episodic drought conditions under which it becomes physically impossible for the stomata to constrain xylem pressure above or around Ψ e (Meinzer et al. 2009). In the present study, for branches and roots, Ψ 50 and Ψ e in droughted trees were higher than in the control (Fig. 7), which was consistent with the acclimatisation of these traits to drought (Anderegg et al. 2016). These results indicated that branches and roots of R. pseudoacacia may regularly experience a substantial loss of their conductivity even when soil water availability is not severely restricted.
In our study, the hydraulic safety margin Ψ 50 -Ψ e indicated a progressively more gradual slope of the VC after the air entry threshold had been crossed. This can be regarded as another hydraulic safety parameter of defence against catastrophic xylem failure once stomatal regulation can no longer prevent xylem pressure from entering the steeper portion of the vulnerability curve (Meinzer et al. 2009). The speci c xylem structural features responsible for the observed behavior of Ψ 50 -Ψ e are not currently known. Most likely they relate to the incidence of safety features in the population of cells contributing to embolism avoidance, which could include not only vessels, but also parenchyma and bre cells (Meinzer et al. 2009;Adams et al. 2018). In the present study, the branches and roots operated with narrow (< 1MPa) hydraulic safety margins, which suggested that R. pseudoacacia were exposed to the risk of tree decline during two growing seasons of drought (Nardini et al. 2013).

NSC during the drought
The responses of NSC during drought represent the balance among the assimilation, respiration, and export of carbon across different source and sink tissues (da Costa et al. 2010). We demonstrated that after two growing seasons of water de cit, R. pseudoacacia showed no signi cant change in NSC concentrations in the roots and stems (Fig. 8) 2015) reported that R. pseudoacacia had signi cantly lower NSCs under severe drought (i.e., 6% soil gravimetric water content) in potted experiments. In our study, despite two growing seasons of drought, the use of NSCs was unaltered in roots and stems, which implied that R. pseudoacacia did not draw signi cantly upon their NSC reserves to buffer against the short-term effects of soil moisture de cit. Many past investigations have documented no change, and often even an increase, of C reserves in trees exposed to drought (Anderegg et al. 2012;Sevanto et al. 2014). An explanation for such phenomenon is that growth and respiration demand for C increases faster than photosynthetic activity; a response known as source limitation (Korner 2003;Weber et al. 2017). By contrast, Wiley and Helliker (2012) proposed that higher C storage can help species to better survive environmental stress like drought, and increased NSC concentrations might thus even exacerbate C limitation to growth.
However, leaf carbohydrate resources were diminished during drought as leaf NSCs declined by 26% compared with the control, but not in roots and branches (Fig. 8), which indicated a proportionally larger drawdown of leaf NSC after two growing seasons of drought. These differences may be related to characteristics of carbon allocation and transport. Under optimum conditions, tissues near to the leaves, the major carbon source, appear to be favored in carbon allocation at the expense of other distant tissues (Jordan and Habib 1996). Under drought stress, however, trees tend to allocate more carbon to roots and branches to facilitate water absorption and survival (Chapin et al. 1990). For a continuous controlled drought pot experiment, the synthesis of starch was inhibited under drought stress, and the activity of hydrolases was accelerated, which resulted in a decrease in starch (Zhang et al. 2015). The decline in starch may have indicated that there was a relatively small carbohydrate pool for leaves that maintained only a small buffer when the demand for carbon became larger than the supply.

Interplay between hydraulics and carbohydrates
We examined carbon and water status in R. pseudoacacia to develop an understanding of the interdependence of carbon-and water-use strategies. In contrast to existing conceptual models (Amthor and McCree 1990), we found that there was not necessarily a generalized plant response to drought that was characterized by accumulation of carbohydrates during drought (Mitchell et al. 2014). Our results showed that droughted R. pseudoacacia trees succumbed to pronounced hydraulic failure (i.e., relatively rapid decline in hydraulic conductivity) (Fig. 6, Fig. 7), but not to carbon starvation (i.e., lower carbohydrate concentrations in stems and roots) (Fig. 8), which suggests a decoupling of hydraulic failure and carbon starvation (McDowell 2011). The decoupling strategy of R. pseudoacacia may have increased the capacity of hydraulics to recover from the two growing seasons of drought (Galiano et al.

2011
). Generally, carbohydrates supposed to play an important role in re lling embolized conduits (Secchi 2011), and if the carbohydrates can be transported to where they are needed, this may prevent hydraulic failure. However, embolized conduits diminish the hydraulic capacity of the plant, which in turn can limit its photosynthetic capability (Sperry 2000). This may exhaust carbohydrate reserves, which implies that prolonged drought may cause hydraulic failure and limit carbohydrate utilization (McDowell 2011).

Conclusions
The interdependence of carbon-and water-use strategies can help us to understand the physiological mechanisms that underlie plant responses during water de cit. Following two growing seasons of drought, our results suggested that the R. pseudoacacia experienced hydraulic dysfunction, but not carbohydrate depletion. The hydraulic failure was re ected mainly in the accelerating loss of hydraulic conductivity, lower water potential and xylem transpiration, and the narrow hydraulic safety margins. Furthermore, trees maintained a fairly good coordination between carbon supply and carbon demand when confronted with two growing seasons of drought. Overall, drought stress affected hydraulics more than it affected carbon storage. Our results provide a means of identifying the response of physiological mechanisms during two growing seasons of drought. Our ndings also emphasized that hydraulic failure plays the predominant role in causing tree death during highly intense drought, while whether "carbon starvation" occurs during tree mortality remains to be tested in longer (multi-year) but less intense drought.    Relationships between mean mid-day, leaf water potential (Ψmd) and monthly stand transpiration (Qm) for the drought and control treatments. Each treatment data set was tted with a linear regression   Typical xylem VCs that shows the relationship between the percent loss of conductance for branches (a, b) and roots (c, d) and xylem pressure for the drought and control treatments. The xylem pressures that correspond to 50% loss of conductance (Ψ50) and the air entry threshold (Ψe) are shown. The difference between Ψ50 and Ψe (shaded area) corresponds to a "hydraulic safety margin".