Xinwei Guo, Shirong Liu,＊, Hui Wng, Zhicheng Chen, Jinglei Zhng, Lin Chen,Xiuqing Nie, Lu Zheng, Doxiong Ci, Hongyn Ji, Boling Niu

a Key Laboratory of Forest Ecology and Environment of National Forestry and Grassland Administration, Ecology and Nature Conservation Institute, Chinese Academy of Forestry, Beijing, 100091, China

b Experimental Center of Tropical Forestry, Chinese Academy of Forestry, Pingxiang, 532600, China

c Youyiguan Forest Ecosystem Research Station, Pingxiang, 532600, China

Keywords:Photosynthesis Carbon balance Tree growth Nonstructural carbohydrates Carbon limitation

ABSTRACT Nonstructural carbohydrates(NSC)are indicators of tree carbon balance and play an important role in regulating plant growth and survival. However, our understanding of the mechanism underlying drought-induced response of NSC reserves remains limited. Here, we conducted a long-term throughfall exclusion (TFE) experiment to investigate the seasonal responses of NSC reserves to manipulative drought in two contrasting tree species (a broadleaved tree Castanopsis hystrix Miq.and a coniferous tree Pinus massoniana Lamb.)of the subtropical China.We found that in the dry season, the two tree species differed in their responses of NSC reserves to TFE at either the whole-tree level or by organs,with significantly depleted total NSC reserves in roots in both species.Under the TFE treatment,there were significant increases in the NSC pools of leaves and branches in C.hystrix,which were accompanied by significant decreases in fine root biomass and radial growth without significant changes in canopy photosynthesis;while P.massoniana exhibited significant increase in fine root biomass without significant changes in radial growth.Our results suggested that under prolonged water limitation,NSC usage for growth in C.hystrix is somewhat impaired, such that the TFE treatment resulted in NSC accumulation in aboveground organs(leaf and branch); whereas P. massoniana is capable of efficiently utilizing NSC reserves to maintain its growth under drought conditions. Our findings revealed divergent NSC allocations under experimental drought between the two contrasting tree species,which are important for better understanding the differential impacts of climate change on varying forest trees and plantation types in subtropical China.

1. Introduction

The spatiotemporal patterns of precipitation are shifting under ongoing climate change,such that some of the terrestrial ecosystems may experience increased frequency,intensity and duration of drought in the coming decades(Dai,2021;McDowell et al.,2018).Drought episodes,as a result of climate change,have shown to alter the carbon balance of trees(Granda and Camarero, 2017; Hartmann et al., 2013; Jin et al., 2018),potentially leading to a reduction in the carbon sink capacity in forests.For example,drought-induced tree mortality may cause forests to switch from a carbon sink into a carbon source(Allen et al.,2010;Wang et al.,2018). Forests, as largest carbon sink and carbon pool of terrestrial ecosystems, store large amounts of assimilated carbon in leaves, stems,and roots.Drought may inhibit photosynthetic C assimilation leading to insufficient carbon supply, while C storage in organs may play a major role in maintaining tree function during such period (McDowell et al.,2011;Wiley and Helliker,2012).

Non-structural carbohydrates (NSC) are important forms of tree C storage, which are mainly composed of soluble sugar and starch. The concentration of NSC represents a trade-off between carbon source gain via photosynthesis and carbon sink costs through metabolism and growth,and reflects the relationship between supply and demand across plant tissues(Hartmann and Trumbore,2016).NSC pools in plant tissues can serve as buffers to compensate for the reduction in photosynthetic carbon supply caused by a decline in photosynthetic capacity during drought (Chapin et al., 1990); NSC fluctuation under shifting drought regimes is often regarded as a fundamental mechanism for plant survival(Yang et al.,2019).However,regulation of NSC storage in natural forests and the extent to which it is a passive process vs.an active process have been vigorously debated (Resco de Dios and Gessler, 2021; Sala et al.,2012).

The first and most sensitive response to moderate drought stress in plants is a reduction in turgor-driven expansion of cells (Woodruff and Meinzer, 2011), such that the cessation of growth is earlier than the decline in photosynthetic capacity. The asynchrony in the temporal pattern of growth (carbon sink) and photosynthesis (carbon source)provides the possibility for the passive accumulation of NSC in the early onset of drought or at mild drought stress (Piper et al., 2017). NSC depletion may occur if drought persists over longer periods or drought is more intense when carbon demands exceed the carbon supply, eventually leading to carbon limitation (McDowell et al., 2008, 2018). These hypotheses may partially explain the divergent findings in the previous publications(see schematic illustrations in Fig. 1.), with results ranging from NSC increase (McDowell et al., 2011; Piper et al., 2017) to no changes at all(Jin et al., 2018;Sala and Hoch, 2009) or even decreases(Galiano et al., 2011) during drought. Moreover, there are studies suggesting that the different responses of NSC to drought may be the result of species-specific variation such as leaf phenology (Newell et al., 2002),soil water uptake depth(Pivovaroff et al.,2021),and drought resistance(Piper, 2011; Regier et al., 2010). In short, the NSC dynamics can be regarded as a result of a series of coordinated traits that trees use to deal with drought,which is species-specific.

Research on the NSC dynamics have mostly been focused on seedlings or coniferous evergreen and deciduous trees with distinct growing seasons,with limited information on seasonal patterns of NSC dynamics at a whole tree and organ levels under long-term controlled field studies.The significant differences in eco-physiological characteristics between seedlings and mature trees may lead to the differentiated NSC responses to drought(Hartmann et al.,2018;Smith et al.,2018;Zhang et al.,2020).Furthermore, compared with deciduous and evergreen coniferous trees of the northern Hemisphere (temperate biomes), evergreen broadleaf tree species in Mediterranean, subtropical, and tropical biomes have various growth phenology and no apparent seasonality (Griebel et al.,2017;Peng et al.,2021;Smith et al.,2018).In such forests,the growth of trees may be more susceptible to seasonal drought disturbances rather than growth phenology (Smith et al., 2018; Würth et al., 2005). Moreover, previous studies have shown that seasonal drought appear to be a stronger driver causing seasonal fluctuations in NSC than phenological rhythms(Liu et al.,2018; Newell et al.,2002;Würth et al.,2005).

It is difficult to accurately measure NSC storage at a whole tree-level because that NSC fluctuation is easily affected by water regimes or by growth phenology (Furze et al., 2019; Smith et al., 2018; Würth et al.,2005). Previous studies have shown that it is a more common and feasible method to infer the seasonal accumulation and depletion of NSC in mature trees in-situ by monitoring changes in the concentration of NSC in all organs(Fermaniuk et al.,2021;Furze et al.,2019;Jin et al.,2018;Mei et al.,2015).The dynamics of whole-tree NSC pools over time can be regarded as a representative of the complex integration of source-sink relationships, storage strategies and all potential functions of sugar and starch (Furze et al., 2019; Martínez-Vilalta et al., 2016). Hence, a more comprehensive evaluation of the response of NSC to drought is very fundamental for better understanding the functional significance of NSC storage and the adaptive mechanism of trees.Up to now,changes in NSC under drought conditions have been investigated only in mature trees of few species in subtropical regions (Lin et al., 2018; Zhang et al., 2020).Therefore,there is a need to explore the effects of inter-seasonal climatic variability on NSC, and the relationships between changes in NSC and tree growth in response to seasonal drought.

Fig. 1. A conceptual framework of NSC responses to drought and how source activity and sink activity affect NSC reserves. The thickness of the arrows and the size of oval (blue) and rectangular (gray) boxes are proportional to the amount of activity. Under drought stress, we expect (a) source-sink co-limitation, (b) source limitation or (c) sink limitation in different tree species. When source activity (photosynthesis) and sink activity (growth) are co-limited,the NSC reserve may remain unchanged (a). When sink activity is greater than source activity, the NSC reserves is expected to decrease(b).Conversely,when source activity is greater than sink activity, the NSC reserves is expected to increase(c).(For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

In this study,we selected two contrasting tree species,a broadleaved tree Castanopsis hystrix Miq.and a coniferous tree Pinus massoniana Lamb.that are typically used in subtropical plantations, to elucidate drought effects on the seasonal NSC dynamics at a whole tree- and organ-level with an 8-year throughfall exclusion (TFE) experiment. The two tree species have different life-history traits and environmental requirements due to the distinguished differences in leaf habits, wood anatomy and hydraulic characteristics (Carnicer et al., 2013; Zhao et al., 2021).Species-specific test provides an opportunity to investigate the drought response and adaptability of tree species characterized with different life-history traits. The objectives of study are to answer the following questions: (1) How do seasonal allocation patterns of NSC at the whole-tree and the organ-level vary with tree species? (2) How is the discrepancy in NSC correlate with the growth responses of the two species under the TFE treatment? We hypothesized that: the limitation of NSC utilization caused by season or drought would lead to growth suppression, hence an observation of NSC accumulation in various storage organs.

2. Materials and methods

2.1. Study site

The field experiment was carried out at the Experimental Center of Tropical Forestry, Chinese Academy of Forestry, located in the south of Guangxi Zhuang Autonomous Region of southern China (latitude 22°10′N, and longitude 106°50′E).It has a typical subtropical monsoon climate with hot, humid summers and cool, dry winters. The annual average precipitation is approximately 1,360 mm.There is a pronounced wet season from May through September,with 78%of the total rainfall.The mean annual air temperature is about 21.7°C (Chen et al., 2019).According to the USDA Soil Taxonomy, the loamy-textured soil is classified as Oxisol(Wang et al.,2017).It contains 34%sand,33%silt,and 33% clay. The soil pH ranges from 3.8 to 4.8 (Chen et al., 2019; Yang et al.,2019).

2.2. Experimental design

The throughfall exclusion (TFE) experiment was established in September 2012, with treatments nesting within a separate block for each species.Each block has six plots(each of 20 m×20 m),comprising three TFE plots and three controls in random arrangement with at least 30 m spacing between adjacent plots.TFE plots were installed with subcanopy roofs made of supporting frames and 150 μm thick transparent film for excluding about 50%of throughfall.Each film shelter was placed 0.5–1.5 m above the ground at an interval of 0.3 m,covering 50%of the ground area.The intercepted throughfall was excluded with gutters.The roofs with white nylon nets were placed over the control plots at the same height as the sub-canopy roof of the TFE plots to control for the shading effects (Fig. 2). The litterfall on the roofs was manually picked and returned to forest floor monthly. Detailed information on the design of TFE refers to Chen et al.(2019)and Yang et al.(2019).

The same treatments were implemented in both P. massoniana and C. hystrix plantations on the same ridge with the same management history; both were established in 1983 after the clear-cutting of a Cunninghamia lanceolata plantation. The C. hystrix plantation had a stand density of 334±92 ttrreeeess· hhaa-–11,average diameter at breast height(DBH)of 27.15 ± 1.96 cm, and average tree height of 19.84 ± 2.96 m; the P. massoniana plantation had a stand density of 275 ± 45 trees?ha?1,average DBH of 31.26 ± 2.15 cm, and average tree height of 18.81 ±1.30 m.

2.3. Measurements of photosynthetic gas exchange

Photosynthetic gas exchange was measured on detached branches from 3 to 5 trees per plot in wet season (August 2019) and dry season(January 2020), respectively. The sunlit southern branches were cut using long-handled shears or by climbing and immediately placed in a bucket of water to avoid water loss. The net photosynthetic rate (Asat)was measured using a portable gas exchange system equipped with a redblue LED source (Li-6400, LI-COR Inc., Lincoln, NE, USA) from 9:00 to 11:30 a.m. under a pre-determined saturated irradiance of 1,500 μmol?m?2?s?1PPFD and a CO2concentration matching the ambient air,with air temperature inside the cuvette set for 25°C.A previous study has established that the values of photosynthetic measurements on detached branches can be similar to that of in-situ photosynthesis in mature trees(Zhang et al.,2020).

2.4. NSC measurements

Fig. 2. Geographical location of the study site (left) and the setup of throughfall exclusion treatment and control (right).

To evaluate the seasonal changes in NSC at the organ-level, field samplings were implemented in the wet(June and August of 2019)and dry season(November 2019 and January 2020),respectively.Three trees were randomly selected from each plot for collection of plant materials used in NSC measurements. Leaf and branch samples were collected between 9:00 and 11:00 a.m., from fully sunlit branches on the upper crown using a long-handle shears or by climbing.Fully expended,healthy leaves were collected from branches and mixed to form a composite sample. Current-year and two-year branches were sampled, and then mixed to form a single branch sample.A stem-wood increment core were collected at the breast height from the south side using an increment borer (5 mm in diameter)(Furze et al.,2019).After removing the bark,the increment cores of 3 cm in length from the most recent growth were separated and used for NSC analysis. Root samples were obtained by manual excavation.A tap-root was located,and all lateral roots attached to it were traced carefully with screw drivers,with careful removal of the soil around the roots. The root samples included coarse (>5 mm diameter), medium (2–5 mm diameter), and fine (<2 mm) fractions in NSC analysis(Huang et al.,2021).All tissue samples were microwaved at 600 W for 90 s to terminate enzymatic activity,and then oven-dried at 65°C for 48 h to constant mass. The oven-dried samples were ground and sieved through a 100-mesh screen, homogenous powder for chemical analysis.

Soluble sugar and starch content were measured using the anthrone method (Yemm and Willis, 1954). Approximately 0.1 g of the homogenous powder was extracted three times with 80%ethanol for 30 min in a 100°C water bath and centrifuged for 5 min at 5,000 g.The supernatant was collected for soluble sugar analysis,and the residue was washed with distilled water and dried at 60°C to remove any residual ethanol for measurement of starch content.The starch in the residue was hydrolyzed by boiling in 2 mL distilled water for 15 min. After cooling to room temperature, 2 mL 9.2 mol?L-1HClO4was added, and then 4 mL of distilled water was added after 15 min.The mixture was centrifuged for 5 min at 5,000 g. A further extraction was carried out with 2 mL 4.6 mol?L-1HClO4. The supernatants were analyzed for starch. The sugar content (%) was calculated from the regression equations based on glucose standard solutions,and the starch content(%)was multiplied by 0.9(Osaki et al.,1991)on a dry-matter basis.

2.5. Estimating NSC pools of leaves,branches,stems,roots,and whole-tree

Biomassofeachorgan(leaves,branches,stemsandroots)wasestimated using allometric equations (He et al., 2013). We paired the NSC and its componentsfor eachsample and thencalculated the NSC poolineachtissue(biomass×concentration).The whole-treetotal NSC pool was calculated as the sum of the four organs.To eliminate the effect of individual tree size on comparison of the NSC and its components among different tree species,we used standardized DBH (30 cm) and standardized tree height (18 m) for both tree species to calculate the individual tree biomass(Furze et al.,2019;Yu et al.,2011).Therefore,differences in organ NSC storages truly reflected changes in NSC concentration.Stem samples were collected only in August 2019(wet season)and January 2020(dry season),to avoid multiple samplingofstemthatwouldaffectthe transportofNSC.Therefore,the NSCpool at whole-tree and organ-level in the wet season and dry season are based on above two periods.

2.6. Tree and soil measurements

Soil cores were taken at the 0–10 cm depth using the soil auger with a diameter of 5 cm,within 0.5–1 m from the base of five randomly selected tree trunks in each plot. They were carefully washed by wet sieving to remove the soil attached to root.Live fine roots were carefully picked out and sorted according to the color, elasticity, and morphology (Persson,1983), and oven-dried at 65°C for 48 h, and weighed to calculate the biomass.

Litter were collected from five litterfall traps(1 m×1 m)installed in each plot. Field sampling was implemented in the wet seasons (June to August) and dry seasons (November to January) of 2019–2020, respectively.Litter from litterfall traps were aggregated into a bag,oven-dried at 65°C for 48 h, and weighed to calculate the litterfall dry matter production.

All mature trees within plots were marked and measured for DBH(>15 cm)at the onset of the TFE treatment(2012).The measurements on DBH were repeated in December of 2015, 2018, and 2020. Relative increment in stem diameter was calculated based on the onset of the experiment by [(DBH-DBH2012)/DBH2012×100%]in all plots.

Soil moisture and temperature were recorded continuously in each plot at 10 cm depth by HOBO data loggers (Onset Computer Corp.,Bourne,MA,USA)throughout the study period.Data of daily precipitation, air temperature, and relative air humidity were obtained from a nearby meteorological station.

2.7. Statistical analyses

Treatment effects on soil temperature,soil moisture and mean annual DBH increment were tested using a student's t-test. The linear mixedeffects model (LMM) was fitted with maximum likelihoods in the‘lme4’ package of R 3.6.2 software (R Core Team,2019). The treatment and measurement season were treated as fixed factor and plots as random factors to test the effects of sampling date,treatment and their interaction on NSC,leaf litter,Asat,and find root biomass(FRB).Tukey's post hoc test following linear mixed-effects model analyses was performed to evaluate significant differences among treatments and seasons for all response variables. For both LMM, residuals were calculated for normality using the Shapiro-Wilks test at P < 0.05, and data were logarithmically transformed when necessary.Origin 9.8.0.200 software was used to draw diagrams.

3. Results

3.1. Soil moisture and temperature

The precipitation and soil properties during the study period are shown in Fig. 3. Soil moisture content and soil temperature in the wet season were significantly greater than that in the dry season(P<0.05).Soil water content at 10 cm depth was significantly lower in the TFE than that in the control, with 10.63% reduction in the wet season (May to September 2019) and 15.69% in the dry season (October 2019 to April 2020) in the C. hystrix(P <0.05; Fig.3a),and 14.14%reduction in the wet season and 14.82%in the dry season in the P.massoniana(P<0.05;Fig. 3b), respectively. There was no significant difference in soil temperature between the TFE treatment and controls(P>0.05,Fig.3).

3.2. Photosynthesis and growth

The canopy photosynthesis in C.hystrix and P.massoniana decreased by 21.8%and 50.4%respectively,during the wet season compared to the dry season. TFE treatment had no effect on light-saturated net photosynthetic rate(Asat)in both species(P> 0.05;Fig.4a and b).

Seasonal differences in fine root biomass(FRB)were significant(P<0.05;Fig.4c and d)in both tree species.TFE treatment decreased FRB in C. hystrix by 33.4% in the wet season and 20.4% in the dry season,respectively, but increased FRB in P. massoniana by 11.2% in the wet season and 37.7% in the dry season, respectively. TFE treatment markedly increased the leaf litter production in C. hystrix during the study period (P < 0.05; Fig. 4e). Compared with the controls, a significant decrease in mean DBH increment in C.hystrix was observed with the TFE treatment(P<0.05;Fig.4g).

3.3. Variations of NSC with season and TFE treatment

The NSC concentrations in all of the organs in both species varied significantly among sampling dates(P<0.05;Fig.5 and Table 1).Foliar NSC concentration was obviously lower in the wet season than in the dry season (Fig. 5 and Table 1). The NSC concentrations in branches and roots displayed consistent seasonal patterns, with a significant decrease in the wet season and a gradual increase in the dry season (P < 0.05;Fig.5).Whole-tree NSC pools were significantly higher in the dry season than in the wet season(NSC 29.6 kg vs 15.2 kg in C.hystrix,80.9 vs 42.9 in P.massoniana; P < 0.05;Fig. 6e and f). The seasonal patterns of NSC pools mostly differed among organs(Fig.7).In the wet season,the rank of organ contributions to whole-tree NSC pool in both species were:stem>branch>root>leaf,with largest fraction of NSC allocated in branches or roots, followed by stems; in the dry season, leaves had the smallest NSC allocation(Fig.8e and f).

Fig. 3. Daily mean soil temperature and moisture at 10 cm depth in controls and throughfall exclusion(TFE)plots in(a)Castanopsis hystrix plantation and(b)Pinus massoniana plantation from January 2019 to September 2020. Values are the average of control plots or TFE plots(n ＝3).The precipitation shown in the figure is the daily accumulative rainfall during the period 2019–2020.The red arrows indicate the timing of measurements.(For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Fig.4. Light saturated net photosynthetic rate(a,b),fine root biomass(c,d),leaf litter(e,f)and relative increment of stem diameter(g,h)in Castanopsis hystrix and Pinus massoniana in the controls and throughfall exclusion treatment (TFE) plots. Values are means ± SE (N ＝3 plots). The effects of treatment, season and their interaction on each species are displayed in the figure. (n.s., P > 0.05; ＊＊＊, P < 0.001; ＊＊, P < 0.01; ＊, P < 0.05).

TFE treatment did not significantly affect sugar pool at both the whole-tree and organ level in P.massoniana in most instances(Fig.6a and b;Fig.7a and b;P>0.05).In contrast,either the starch pool or the total NSC pool experienced fluctuations with TFE treatment in P.massoniana,with respectively 19% and 20% reductions in the whole-NSC pool and the starch pool, and the corresponding reduction in the contribution of roots to whole-tree NSC pool from 39.7%to 29.3%in the dry season(P<0.05; Fig. 7d, f; 8f). In C. hystrix, however,both the starch pool and the total NSC pool remained stable with TFE treatment in the dry season.This stability of whole-tree NSC pool attributed to the reversed changes in NSC pool among organs in C.hystrix(Fig.7c,e).In the wet season,TFE treatment increased root NSC pool in C.hystrix by 1.2 times(P<0.05),while decreased branch NSC pool by 14.2%(P ＝0.662).On the contrary,in the dry season, TFE treatment reduced root NSC pool by 27.8%, but increased leaf and branch NSC pools by 11.1% and 35.9%, respectively(P<0.05;Fig.8c,e).

Fig. 5. Variations in nonstructural carbohydrate (%) concentration at organ-level in Castanopsis hystrix and P. massoniana in controls and TFE treatment plots from June 2019 through January 2020.Values are the means±SE for both species on each sampling date.The asterisk indicates significant difference between TFE and the control (＊＊＊, P < 0.001;, ＊＊, P < 0.01; ＊P < 0.05).

4. Discussion

We investigated a prolonged drought effect on NSC storage, depletion, and allocations in two contrasting subtropical tree plantations through a field manipulative experiment with throughfall reduction.We found that the seasonal fluctuations of NSC in both tree species were similar, and that the NSC allocations among organs and tree growth,inferred by changes in fine root biomass and DBH, had species-specific response to drought. These results suggested that the shift of NSC allocations would dominate the response of tree growth to drought (Würth et al.,2005;Doughty et al.,2014).

4.1. Species-specific responses of NSC to drought

In our study, P. massoniana (a typical gymnosperm tree species in subtropical China)experienced a significant decrease in NSC reserves atwhole-tree level in the dry season under TFE condition,while C.hystrix(a common angiosperm tree species in subtropical China) did not show significant treatment response.These results are in line with the findings of Adams et al.(2017)and Piper and Paula(2020)that reduction in NSC storage was more common for gymnosperms than angiosperms in terms of NSC responses to drought. This discrepancy can be attributed to the wider hydraulic safety boundary of gymnosperms relative to angiosperms(Johnson et al.,2012),as gymnosperms could maintain growth at the expense of stored NSC during drought (Doughty et al., 2014; Piper and Paula,2020).However,a recent meta-analysis showed that total NSC in gymnosperms increased under drought stress, while total NSC in angiosperm decreased under both slight-to-moderate and severe drought(He et al., 2020), suggesting differential impacts of drought on carbohydrate assimilation and growth demand between gymnosperms and angiosperms under varying intensity and duration of drought.

Table 1 F values in the linear mixed-effects model on the effects of treatment(T),month(M),and their interaction on sugar(%),starch(%),and total NSC concentration(%)in Castanopsis hystrix and Pinus massoniana plantations.

Fig.6. Allocations of whole-tree nonstructural carbohydrate(NSC)pool in response to TFE in Castanopsis hystrix and Pinus massoniana.(a,b)Sugar pools,(c,d)starch pools, (e, f) total NSC pools. Values are means ± SE (N ＝3 plots). Lowercase letters above the bars indicate significant differences between the control and TFE treatment plots. Uppercase letters above the bars indicate significant differences among seasons (P < 0.05).

Fig.7. Seasonal dynamics of sugar pools(a,b),starch pools(c,d)and the total NSC(e,f)pools at organ-level in Castanopsis hystrix and P.massoniana in the controls and the TFE treatment plots.Values are means±SE for both species by season.The asterisk indicates significant difference between TFE and control(＊＊＊,P<0.001;,＊＊, P < 0.01; ＊P < 0.05).

We found that at the organ-level, there was no apparent effect of a prolonged experimental drought on NSC concentrations in leaf, branch and stem of P. massoniana, as reported similarly by Lin et al. (2018). A likely explanation for this phenomenon is that an imbalance between carbon supply and carbon demand depend on the duration of drought,i.e., a NSC accumulation at early stage of drought due to growth inhibition and a depletion of NSC under a prolonged drought(K¨orner,2015;McDowell et al.,2011;Sch¨onbeck et al.,2018).Although we cannot rule out the above possibility,it seems an unlikely explanation for findings in this study given the results of decreased NSC concentration in roots and a significant increased fine root production in P. massoniana under TFE treatment. We postulated that P. massoniana subjected to TFE invested newly assimilated NSC preferentially into root growth, rather than into higher NSC reserves. Thus, NSC storage in P. massoniana under TFE appears to be an active regulation process between C supply via photosynthesis and growth(Sala et al.,2012)and decreases in root NSC could be related to increased root metabolism associated with water-uptake(Fan et al., 2017). Surprisingly, although the response of NSC to seasonal drought was similar between P. massoniana and C. hystrix, their responses of NSC to TFE significantly differed, especially in the dry season with observation of severe reduction in root NSC and significant increases in NSC in branches and leaves in C. hystrix subjected to TFE.One possible explanation for this finding is that TFE treatment could intensify soil water shortage in dry season with less rainfall, and then very likely promoted the decoupling between photosynthesis and growth in C.hystrix.Results on increased production of leaf litter supported the view of increased intensity of drought stress to C.hystrix during the dry season under TFE condition (Souza et al., 2019). Drought-induced growth inhibition may lead to NSC accumulation as carbon supply exceeded carbon demand. In addition, an intensified drought stress would constrain water movement from xylem into phloem, impairing NSC movement through the vascular network and consequently causing the root NSC to decline (Hartmann et al., 2013; H¨oltt¨a et al., 2009).Alternatively,an increased carbon allocated to above-ground organ(leaf and branch) under drought may be used to maintain the hydraulic integrity of xylem.This view seems to be supported by our observation of no significant difference in the percentage loss of conductivity (PLC) in C. hystrix between the TFE treatment and the controls (unpublished data).These results suggested little use of NSC in above-ground organs in

Fig. 8. Comparison of allocations of sugar (a, b), starch (c, d), and total NSC (e, f) among organs in Castanopsis hystrix and Pinus massoniana between TFE treatment and controls in wet season and dry season, respectively. Values are means ± SE.C. hystrix to meet the growth demand of roots. The patterns of NSC allocation under TFE illustrated that NSC storage in C. hystrix were not always a purely passive process.

4.2. Trade-offs of NSC allocation with growth

We found a remarkable feature that NSC reserves in both species was higher in the dry season than in the wet season, while canopy photosynthesis showed an opposite patterns of seasonal effects,indicating that NSC reserves decreased when source activity(photosynthesis)was high and increased when source activity was low.Difference in the timing of depletion and accumulation of NSC stocks may be explained by water condition-related growth phenology(seasonal growth pattern;Liu et al.,2018).Tree growth seasonality in subtropical forests matched well with the rainfall seasonality,as tree growth was larger in the wet season than in the dry season(Gheyret et al.,2021).This could explain an increased usage of assimilated C for growth rather than NSC storage in the wet season; but in the dry season, water deficit leads to decreased turgor in the cambial cells, consequently suppressing tree growth (Peters et al.,2021). When tree growth is inhibited, carbon demand can lag behind carbon supply, hence a surplus carbon storage in plant tissues as NSC.Our results appeared to support the view proposed by Würth et al.(2005)that active sinks (growth) stimulate source activity (photosynthesis),whereas sink limitation facilitates NSC reservation.Contrary to what has been reported for several temperate tree species(Furze et al.,2019),this study shows that NSC pools in P.massoniana and C.hystrix of subtropical forests were depleted during the growing season and increased during the dry season.This discrepancy in NSC storage dynamics may be related to tree functional types (deciduous trees verses evergreen trees) and could be largely explainable by leaf phenology(Newell et al.,2002).As evergreen trees retained their leaves throughout a year, they have the opportunity to continuously provide carbon through photosynthesis to C sink processes during the dry seasons (Smith et al., 2018). However,deciduous trees necessarily rely on stored carbon to maintain growth during leafless seasons,causing depletion of NSC reserves(Newell et al.,2002).Additionally,the seasonal patterns of NSC at the whole-tree level in both tree species subjected to 8-year TFE remained unchange compared with the controls, indicating the phenological rhythm of growth was not altered following long-term TFE treatments.

Another notable feature in our study is that the seasonally relative changes of NSC pool significantly differed among organs. This seasonal difference in NSC allocations among organs may reflect the unique functions of organs in tree growth (Furze et al., 2019; Landh¨ausser and Lieffers, 2003; Martínez-Vilalta et al., 2016). Due to the seasonal climate characteristics in subtropical regions, the growth of evergreen tree species typically goes through three distinct phases: an early dormancy period in spring, followed by a rapid growth period from spring to autumn, and then the subsequent plateau period (Gheyret et al., 2021). During rapid growth period (from June through August),decreases in NSC concentration in branches and roots are attributable to preferential use of photosynthetically fixed carbon for growth and a dilution effect caused by increased biomass accumulation (Furze et al.,2019; Landh¨ausser and Lieffers, 2003). During the plateau and dormancy periods, growth slows down or even ceases, and the excess carbon supply is then stored in branches and roots as NSC, causing an increase in NSC pool in the two organs (Furze et al., 2019). Therefore,the NSC pools in branches and roots can be buffers when the supply of assimilated carbon does not meet the growth and maintenance demand. In addition, when carbon supply exceeds its demand, branches and roots can be used as NSC storage organs. The branch NSC pool mainly supports the aboveground process, including budding, leafing,flowering, fruiting and branch growth and maintenance respiration(Huang et al., 2021; Palacio et al., 2018), while root NSC pool is responsible for maintaining growth and respiration of belowground processes (Landh¨ausser and Lieffers, 2003). At no point did we detect a reduction in leaf NSC concentration during rapid growth period, suggesting that the major function for leaf is to fuel growth through current photo-assimilates at expenses of NSC allocation to belowground. The accumulation of leaf NSC in dry season might be a result of reduced sink demand of the growth (Palacio et al., 2018).

A third noteworthy and most important feature found in this study is that TFE treatment caused species-specific shift in NSC allocation during the dry season.P.massoniana allocated more NSC to belowground organ for root growth, while C. hystrix allocated more NSC to aboveground organs.A significant increase of fine root biomass in P.massoniana under TFE treatment was also reported in Yang et al.(2021).Not surprisingly,trees can increase NSC allocation to root growth in order to reduce the effect of water shortage (Doughty et al., 2014; Hertel et al., 2013). The greater network of fine roots in P.massoniana subjected to TFE could help facilitate water and nutrient uptake under severe drought, so there was no significant decline in the stem growth.At the same time,the case with P.massoniana in this study also provides additional experimental support for the optimal partitioning theory that plant would increase belowground C allocation when water is limiting(Bloom and Mooney,1985).In contrast, TFE reduced the fine root biomass in C. hystrix, suggesting that the declines in soil water would inhibit the production of fine roots in angiosperms(Fan et al.,2021).This conflicting result with the optimal partitioning theory could be attributable to the duration of drought.Zhou et al. (2020) found that fine root biomass in a subtropical forest substantially increased after three years of rainfall reduction,but no further change was discovered in the fourth year. However, our results did not seem to support this view, because the change direction of fine root biomass in both tree species after eight years of TFE is consistent with the result in the second year of TFE treatment.Obviously,the response of fine root biomass to TFE is species-specific, reflecting the differential strategies of growth response to drought between the two tree species.Increased leaf shedding in C. hystix subjected to TFE could be another mechanism that protects water and carbon balance in other organs by reducing water loss through transpiration and carbon consumption through respiration.Zhang et al.(2009)found that tree crown dieback is conducive to maintaining sufficient water supply to other parts of tree in a neotropical savanna. In addition, our results showed that C. hystrix subjected to TFE allocated more NSC to aboveground organs during the dry season. Similar findings were reported for several evergreen shrub species at the end of an extreme drought in a savanna ecosystem in Southwest China (Shen et al., 2021). The increased NSC under drought may contribute to maintenance of xylem hydraulic integrity(Wang et al.,2018), ensuring normal leaf functioning for carbon assimilation. However,the change of NSC allocation in C.hystix subjected to TFE during the dry season occurred at the expenses of growth. Therefore, a continuous and long-term drought would inevitably lead to a significant decline in forest growth.

5. Conclusion

In this study, we provided direct evidence that a prolonged drought inhibited the utilization of NSC (growth), resulting in the accumulation of surplus carbon storage in roots and stems in C. hystrix. The seasonal trend of NSC reserves at the whole-tree level in both species did not respond to experimental drought imposed via partial throughfall exclusion, suggesting that carbon limitation did not occur during the dry season under reduced water supply. Under an eight-year TFE treatment, C. hystrix allocated more NSC to aboveground organs during the dry season, while P. massoniana allocated more NSC to roots. These findings indicate that drought can alter the transport, usage and distribution of NSC among organs, and ultimately affects tree growth.P. massoniana used more C reserves to meet sink demand (growth),resulting in a decrease in NSC reserves under long-term drought, while the carbon sink activity was reduced in C. hystrix that led to NSC accumulation. In summary, NSC accumulation could occur due to lack of sink demand in response to growth constraint under prolonged drought. Our study provided experimental insight into the mechanisms underlying forest responses to intensified drought in subtropical regions.

Availability of data and materials

The datasets used during the current study are available from the corresponding author on reasonable request.

Declaration of competing interest

The authors declare that they have no competing interests.

Acknowledgements

We would like to thank the Youyiguan Forest Ecosystem Research Station for experimental and logistic support,and Chaoyin Li,Zhanchao Song and Weijian Jiang for assistance in fieldwork and laboratory analyses. This study was jointly supported by the National Natural Science Foundation of China (Grant No. 31930078)and the Ministry of Science and Technology of China for Key R&D Program (Grant No.2021YFD2200405). Authors are also grateful to two anonymous reviewers for constructive comments and editorial suggestions over a previous version of the manuscript.