The spruce-fir forest, naturally distributing between 1,100 and 1,800 m a.s.l. (above sea level), is one of the main forest types in CMNR in the north temperate climate zone of eastern Eurasia (41°31′-42°28′ N and 127°09′-128°55′ E). The climate here is classified as the temperate continental monsoon climate with the characteristics of warm and rainy summer, and cold and dry winter. The monthly mean temperature ranges between −17.5°C in January and 20.1°C in July. The mean annual precipitation is 680 mm, with 80% of precipitation in the growing season (Yu et al., 2011, 2013).

In April 2019, we established the experiment at the Changbai Mountain Forest Ecosystem Research Station, Chinese Academy of Sciences. The ambient climate data in the station were shown in Supplementary Figure S1.

Three-year-old Jezo spruce seedlings were collected from a local nursery garden and planted into 31 wooden rectangle containers (length × width × height: 120 × 25 × 25 cm) with 7 individuals per container on 30th April 2019. Wooden rectangle containers were placed under a transparent rain roof in a forest gap, avoiding rapid water loss due to strong sunlight exposure. The light condition was similar in different positions under the roof according to condition premeasurement. The soil used in the experiment was obtained from local forests' topsoil without the large stone and roots. Since 5th July 2019, 132 healthy seedlings in 19 containers were not watered throughout the experiment as the drought treatment, while water was continually supplied to the amount of 84 homogeneous seedlings in 12 containers in the control treatment (see the initial seedling status in Table 1). The significant difference in the size (stem base diameter and height) between control and drought was not detected (p > 0.05).

The experiment period was from 5th July 2019 to 30th July 2020 (392 days in total), including the first growing season (GS2019: D1–D88, 5th July to 30th September), winter dormant season (DP2019: D89–D180, 1st October to 31st December; DP2020: D181–D301, 1st January to 30th April), and the second growing season (GS2020: D302–D392, 1st May to 31st July) referring to Yu et al. (2021). In the dormant season, irrigation was stopped as the stomatal closure and transpiration stopped, referring to previous studies (Fajstavr et al., 2019).

During the drought process, the soil water content was measured with the gravimetric method. The soil water content was determined as the ratio of water weight to the soil sample weight (Piper and Fajardo, 2016). The soil water content was not measured in the winter dormant season due to low evapotranspiration and soil frost in the study area (Yang et al., 2006).

According to the phenology, we harvested seedlings on the D49 (August 2019, the middle growing season), D127 (November 2019, the early dormant period), D270 (March 2020, the late dormant period), and D392 (July 2020, the middle growing season). For each harvest, one container of each treatment was randomly selected, and all living seedlings in that container were sampled and then transported to the laboratory in a cooling box. Collected seedlings were divided into leaves, branches (including stems and twigs), and roots. Samples were dried at 105°C for 30 min to stop the enzymatic activity, then oven-dried at 65°C for 48 h, and finally stored at −20°C for further NSC analyses (Huang et al., 2019).

To evaluate the effects of drought on seedlings' survival, we examined seedling mortality weekly during the growing season and biweekly during the dormant season. It is difficult to determine the exact death time of the evergreen species. In this study, dead seedlings were identified when all leaves turned yellow and fallen (referring to previous studies, e.g., Hartmann et al., 2013; Ivanov et al., 2019). The mortality rate of seedlings was counted during the experimental period. In this work, we calculated the mortality rate on the D49, D127, D350, D372, and D392 as:

where i is the days after the treatment; n_i is the number of dead seedlings from D₁ to D_i; and N is the total number of seedlings of each treatment at the beginning of the experiment.

Leaf predawn water potential (ψ_pd) between 3:00 and 4:00 and middy water potential (ψ_md) between 12:00 and 13:00 from 4 to 7 living seedlings were measured on sunny days (D17, D47, D349, D370, and D390) using a pressure chamber (PMS1000; Albany, OR, USA; maximum measurement: 8 MPa). One twig from each seedling was selected for the measurement. No repeated measurement was conducted on the same seedling throughout the experiment.

The hydraulic conductivity was measured on the same seedlings that were used for leaf water potential measurement at D350, D371, and D391, respectively. The seedlings were cut at the stem base in the next early morning (before sunrise) after leaf water potential measurement. Stems were cut immediately under water to avoid the formation of embolism during sampling, and were transported to the laboratory (<100 m). In the laboratory, the stem segment was recut repeatedly under water at two ends to release the tension. The segment with 7 cm length was used to measure native hydraulic conductivity (K_h, kg m s⁻¹ MPa⁻¹). The K_h was measured using a pressure induced by the gravity of a hydraulic head of 50 cm with the 20 mM KCl solution. The K_h was calculated by: K_h=J_v / (F / L), where J_v is the flow rate (kg s⁻¹), F is the gravity-induced driving pressure (MPa), and L is the length of the segment (m). The K_h was divided by the xylem wood area to calculate the stem wood-specific hydraulic conductivity (K_w, kg m⁻¹ s⁻¹ MPa⁻¹) (Fang et al., 2018).

Net photosynthesis rate (A_net) and stomatal conductance (g_s) were measured between 9:00 and 12:00 on sunny days (D17, D47, D349, D370, and D390). The measurements were done under the ambient light (c. 1,000 μmol m⁻² s⁻¹) and CO₂ concentration (c. 400 μmol mol⁻¹). The relative difference in A_net and g_s between drought and control was calculated: Relative to control (%) =100 × (Value_drought / Mean_control), where the Value_drought is the value of drought seedling and Mean_control is the mean value of control at the same time.

Nonstructural carbohydrate (NSC) concentrations (soluble sugars + starch) were quantified according to the standardized protocols by Landhausser et al. (2018). For soluble sugars extraction, approximately 30 mg sample was bathed at 90°C for 10 min after mixing with 1.5 ml 80% (v/v) ethanol. The mixture was centrifuged at 3,500 rpm for 10 min, and the supernatant was used for soluble sugars quantification. For soluble sugars quantification, the solution was colored with phenol-sulfuric and the absorbance was determined at 490 nm with a microplate reader [Multiskan Sky, Thermofisher Scientific (China) Co., Ltd.]. For starch digestion, the pellet after soluble sugars extraction was digested at 85°C for 1 h with the α-amylase from Aspergillus oryzae (600 U/ml, Macklin A861434). The supernatant obtained after centrifuging was further digested with amyloglucosidase from Aspergillus niger (100 U/ml, Macklin A800618). For starch quantification, the solution color was regulated by peroxidase-glucose oxidase, and the absorbance was determined at 525 nm. Concentrations of soluble sugars, starch, and NSC were expressed as a percentage of dry matter (% d.w.).

Two-way ANOVA was used to analyze the effects of season, drought, and their interaction on concentrations of NSC, soluble sugars, and starch. The difference in variables among control, drought, and dead seedlings was assessed using least significant difference (LSD) with package “agricolae” (Mendiburu, 2021). The difference in soil water content, A_net, and g_s between control and drought seedlings was evaluated using Wilcoxon rank-sum test with the package “stats” (R Core Team, 2020). In addition, the difference between ψ_pd and ψ_md at the same date was evaluated by Wilcoxon rank-sum test. The starch concentration in the root and the K_w were logarithmic (ln) transformed to meet normality and homogeneity before the analysis. Statistical analysis for all data was conducted in R 4.0.0 (R Core Team, 2020).

Soil water content in control ranged from 47.1 to 35.8% in the GS2019 and stabled at 38% in the GS2020, but in drought condition rapidly decreased from 47.1 to 19.0% in GS2019 and from 10.8 to 4.0% in the GS2020 (Figure 1A). Accordingly, from the DP2019, seedlings exposed to drought showed a considerable increase in cumulative mortality rate toward the experimental end (D392) from 9.1 to 35.6% (Figure 1B).

Effects of continuous drought on seedling biomass varied significantly depending on season and organ. The whole-plant biomass was similar between control and drought treatments in the GS2019 and DP2019 (Figures 2A,B), but significantly different in the GS2020 (Figure 2C). Seedlings exposed to drought were categorized into living and dead individuals to clarify the detailed response in biomass. Branch biomass of dead but not living individuals was lower than that of seedlings in control in the DP2019 (Figure 2B). However, leaf and branch biomass of both living and dead seedlings was significantly lower than those in control in the GS2020 (Figure 2C and Supplementary Table S1). A similar trend was observed in leaf biomass in the GS2020. Root biomass had no significant variation between treatments and among organs across the experimental period.

Leaf water potential was significantly affected by the interaction between treatment and season (Figure 3). Leaf ψ_pd and ψ_md showed no significant difference between control and drought treatments in the GS2019 (Figures 3A,B), while both variables in seedlings exposed to drought were much lower than those to control especially in the middle and late GS2020 (Figures 3A,B). Specifically, ψ_pd and ψ_md significantly decreased from −0.9 and −1.4 MPa on D349 to −4.3 and −4.0 MPa on D390, respectively. ψ_pd was significantly higher than ψ_md in both control and drought treatments in the GS2019, while ψ_pd and ψ_md existed no significant difference under drought condition since middle GS2020 (Figures 3A,B). In terms of stem hydraulic conductivity, K_w in seedlings exposed to drought was significantly lower than those of control in the middle and late GS2020 (Figure 3C). Specifically, K_w was 27.4 and 20.2% of the control on D371 and D391, respectively (Figure 3C).

Seedlings exposed to drought showed an evident decrease in A_net and g_s from the GS2019 to GS2020 (Figure 4). A_net was significantly lower under drought than control since D17 and accounted for 55.1% of that in control, and reached below zero in the GS2020. g_s under drought had a similar trend with A_net since D47. The g_s under drought decreased to 10.6% of that in control at the end of GS2020.

Drought treatment, season, and their interaction significantly affected concentrations of NSC and its components, but the effect magnitude depended on organs (Table 2 and Supplementary Tables S2–S4). In two growing seasons, NSC concentrations were similar irrespective of drought treatment or organs (Figure 5). The exception was that dead individuals exposed to drought had higher NSC concentrations in leaves than control (Figure 5J). However, such response in concentrations of NSC and its components was opposite in woody organs across the winter dormant season (except for leaf in the DP2019). Especially in roots, concentrations of soluble sugars, starch, and NSC for dead individuals from drought averagely reached 0.6, 0.4, and 1.0%, respectively. Meanwhile, the concentrations of soluble sugars, starch, and NSC were 9.8, 15.4, and 11.5% of the control (Figures 5F,I).

The stomatal closure, one of the earliest reactions to decrease in the soil water (Figure 4), reduced the canopy water loss and maintained high water potential in the GS2019 and the early GS2020 (Figure 3). This behavior suggests a protective mechanism against embolism (McDowell et al., 2008; Xiong and Nadal, 2020), which is consistent with the finding from another spruce species (P. abies) (Hajickova et al., 2021). However, the water loss did not stop due to passive water loss and incomplete stomatal closure (Duursma et al., 2019). Moreover, water loss could be exacerbated by the relatively high temperature in the growing season (Hartmann, 2015; Yan et al., 2020). The ongoing water loss could lead to massive embolism, which constrains xylem water transport (Gebauer et al., 2020). In this study, two results could support the xylem hydraulic dysfunction in the GS2020.

First, drought significantly decreased K_w in the middle and late GS2020 (Figure 3). In this study, the K_w in living seedlings exposed to drought reduced c. 80% compared to the control, and complete loss of hydraulic conductivity was detected in dead seedlings. This is consistent with the previous results that severe long-term drought led to great damage in stem hydraulic integrity and hydraulic conductivity (Chen et al., 2020; Li et al., 2020).

Second, ψ_pd was similar to ψ_md in droughted seedlings since D370 (Figure 3). ψ_pd and ψ_md correspond to the daily maximum and minimum leaf water potentials, respectively (Donovan et al., 2001). ψ_pd recovers overnight through xylem water transport under normal water conditions (Gleason et al., 2017). Similar values of ψ_pd and ψ_md in this study suggest that massive embolism impaired water transport. Therefore, the hydraulic dysfunction might be a primary reason for mortality in the GS2020, which has been proven in field trees and potted seedlings in many extreme drought events (Adams et al., 2017; Arend et al., 2021).

Unexpectedly, living seedlings exposed to drought had similar hydraulic conductivity with control at D351 (Jun, 2020). This might indicate that the hydraulic conductance is above the critical threshold (McDowell et al., 2019) since the xylem embolism formed in the winter dormant season might be repaired (Maruta et al., 2020). It is necessary to frequently measure hydraulic dynamics after winter drought in future work.

Under continuous drought, trees may adjust phenology to maintain the C dynamic (Jin et al., 2018). In the GS2020, seedlings exposed to drought showed higher NSC but lower growth than control (Figures 2, 5). It is possible that seedlings sacrificed growth to maintain higher NSC for subsequent C use (e.g., osmoregulation) (Huang et al., 2021; Luo et al., 2021). In addition, as a drought-defoliation species, Jezo spruce fell off leaves to protect the hydraulic system, which may also stimulate C accumulation in branches and roots (Santos et al., 2022). However, C reserve may be unavailable under drought stress, and less change at the time of death (Sala et al., 2010; Jin et al., 2018; Wiley et al., 2019), which may lead to high-level NSC in drought and dead seedlings (Figure 5). Alternatively, seedlings that died in the GS2020 were collected earlier than the drought and control seedlings. The higher NSC in dead seedlings than living seedlings exposed to drought might result from an earlier occurrence of use constraint and (or) the shorter time duration of negative A_net.

The mortality rate increased in the winter dormant season (Figure 1), suggesting that droughted seedlings might be more vulnerable in the winter dormant season. The temperature below −32°C is expected to induce frost damage to Jezo spruce (data from https://www.worldplants.ca). The minimum temperature reached −33.7°C during the experiment. Furthermore, the winter drought might aggravate the freezing damage of plants (Charrier et al., 2015; Fernández-Pérez et al., 2018).

Nonstructural carbohydrate (NSC) plays a critical role in drought and cold tolerance of temperate trees (Charrier et al., 2015). In contrast to the growing season, we detected NSC depletion in the roots of dead seedlings in the winter dormant season (Figure 5). This is consistent with the results of Galvez et al. (2013), which showed that insufficient NSC reserve in roots might be the primary reason for seedling mortality of two poplar species in the winter dormant season. However, the concentrations of soluble sugars and starch did not decrease to zero in dead seedlings, indicating that there may be a minimum threshold of NSC reserve for tree survival (McDowell, 2011). In this work, NSC concentration in dead seedlings is consistent with the life-threatening NSC level in previous drought studies (Schönbeck et al., 2020), defoliation (Barker Plotkin et al., 2021), and shade (Weber et al., 2018, 2019).

In this study, drought (living) seedlings had higher or similar soluble sugars than control (Figure 5). Similar results were also observed in previous studies, e.g., Chuste et al. (2020) and Schönbeck et al. (2020). These support the theory that NSC concentration initially increases and shifts to decrease with drought persisting (McDowell, 2011). It is possible that NSC reserve might increase or be stable during the early drought stage by limiting growth respiration and consuming with drought persisting (McDowell, 2011). In this work, different death times among seedlings under drought treatment might indicate the different exact physiological stress stages among seedlings (Zhang et al., 2021). However, the reason why seedlings exposed to the same drought showed different death times should be further investigated.

The insufficient NSC reserve and severe embolism may threaten tree survival in the winter (Galvez et al., 2013). Since the lack of hydraulic conductivity and PLC data in the winter dormant season, it is unable to assess the direct contribution of hydraulic dysfunction to seedling mortality during the dormant season. However, almost complete loss of hydraulic conductance may not be damaging due to stomatal closure and reduction in water uptake (Beikircher et al., 2016). As shown in previous studies (Ogasa et al., 2019; Mayr et al., 2020), trees that experienced near to 100% PLC in winter recovered growth in the following seasons, suggesting the whole tree and tissues were still alive in the winter dormant season. Thus, future studies are required to explore the appropriate methods for assessing the linkage between tree mortality and hydraulic dysfunction in the winter dormant season under drought stress.