Divergent effects of short-term warming on microbial resource limitation between topsoil and subsoil in a young subtropical Chinese fir forest

Microbes depolymerize soil organic matter to supply themselves with carbon (C), and nutrients such as nitrogen (N) and phosphorus (P). In this way, microbial resource limitations play an important role in characterizing the biogeochemical cycle dynamics of the ecosystem. However, the effect of warming on microbial resource limitations, especially in resource-poor subsoils, remains unclear. Therefore, this study aimed to examine the effects of warming (+ 5 °C above ambient) on microbial resource limitation and explored their relationships with soil physicochemical properties and microbial community structure in topsoil (0–10 cm) and subsoil (40–60 cm) in a Chinese fir (Cunninghamia lanceolata) plantation in southern China. Microbial resource limitation with warming treatment was assessed via vector analysis of soil extracellular enzymatic stoichiometry after two years. We found warming aggravated microbial C limitation in the topsoil but alleviated microbial C limitation in the subsoil, while it shifted the relative nutrient limitation from P limitation to N limitation in the subsoil. Soil microbial C limitation was explained by soil properties (specifically, ammonium nitrogen) in the topsoil while by microbial community composition in the subsoil based on variance partition analysis. The soil microbial nutrient limitation was explained by soil properties in the topsoil and subsoil. The decrease in actinomycetes abundance in the warming treatment may have led to a decreased microbial C limitation in the subsoil. Our study highlighted the differences in warming effects between the topsoil and subsoil. We argue that the microbial resource demand of the subsoil should be further implemented in the study of soil biogeochemistry to improve the prediction of the impact of climate warming on soil C dynamics.

limitation and explored their relationships with soil physicochemical properties and microbial community structure in topsoil (0-10 cm) and subsoil (40-60 cm) in a Chinese fir (Cunninghamia lanceolata) plantation in southern China. Microbial resource limitation with warming treatment was assessed via vector analysis of soil extracellular enzymatic stoichiometry after two years. We found warming aggravated microbial C limitation in the topsoil but alleviated microbial C limitation in the subsoil, while it shifted the relative nutrient limitation from P limitation to N limitation in the subsoil. Soil microbial C limitation was explained by soil properties (specifically, ammonium nitrogen) in the topsoil while by microbial community composition in the subsoil based on variance partition analysis. The soil microbial nutrient limitation was explained by soil properties in the topsoil and subsoil. The decrease in actinomycetes abundance in the warming treatment may have led to a decreased microbial C limitation in the subsoil. Our study highlighted the differences in warming effects between the topsoil and subsoil. We argue that the microbial resource demand of the subsoil should be further implemented in the study of soil biogeochemistry to improve the prediction of the impact of climate warming on soil C dynamics.

Introduction
Climate warming may affect soil carbon (C) sequestration and nutrient cycling via affecting soil organic matter decomposition (Fontaine et al. 2007;Liang et al. 2019;Meyer et al. 2018). Soil microorganisms, as the soil organic matter decomposers, produce various extracellular enzymes to depolymerize macromolecular soil organic matter into absorbable substrates to fulfill their demands for C, nitrogen (N), and phosphorus (P) for growth and metabolism (Sinsabaugh et al., 2008;Allison et al. 2011). Therefore, abundance and composition of extracellular enzymes can reflect microbial resource constraints (Cui et al. 2021;Sinsabaugh et al., 2008) and are widely used to assess resource limitations (Chen et al. 2018a, b;Cui et al. 2019;Guan et al. 2020;Jing et al. 2020;Moorhead et al. 2016;Peng et al., 2016;Waring et al., 2014). Although microbial enzyme activity is known to be sensitive to temperature (Stone et al. 2012), the effects of warming on microbial resource limitations remain unclear.
The soil extracellular enzyme stoichiometric theory combines ecological metabolism theory with the ecological stoichiometric theory to evaluate resource limitation of microbial metabolism through the ratios between various enzyme activities (Chen et al., 2018a, b;Cui et al., 2018;Cui et al., 2021;Sinsabaugh and Shah 2012;Sinsabaugh et al., 2008;Yuan et al., 2019;Zhou et al., 2020). Enzyme stoichiometry is regarded as an effective tool for assessing the environmental drivers and resource constraints of microbial metabolism and has been widely used to explore biogeochemical cycles (DeForest and Moorhead 2020; Fanin et al. 2022;Nottingham et al. 2015). The enzymes commonly used in enzymatic stoichiometry are mainly involved in the microbial acquisition of C (β-1,4-glucosidase, BG), N (β-1,4-N-acetylglucosaminidase, NAG), and P (acid phosphatase, AP) (Sinsabaugh et al., 2008). Moorhead et al. (2013) proposed a vector model to identify the relative resource limitation of microorganisms by using the vector length and angle calculated from the enzymatic C:P versus C:N acquisition ratio. Using this approach, Moorhead et al. (2016) showed that soil microbial communities are more limited by P in montane tropical forests than in lowland tropical forests. Zheng et al. (2020) reported that warming reduced microbial C limitation in the mineral soil layer and increased P limitation in the organic layer in an alpine shrubland ecosystem. Lie et al. (2019) found that warming mitigated P limitation and increased N consumption in tropical forests by increasing the soil P availability. However, these studies have mainly focused on the 0-20 cm topsoil. Our understanding on the effects of warming on microbial resource limitation below the 20 cm soil depth subsoil remains limited.
Subsoil (soil at > 20 cm) accounts for over 60% of the soil C stock in the top 1 m of soil (Jobbágy and Jackson 2000). Subsoil and topsoil differ in many crucial physical and chemical properties. Compared to topsoil, subsoil tends to have a lower temperature (Zogg et al., 1997) and oxygen concentration (DeAngelis et al. 2010) and higher soil moisture (Serna-Chavez et al., 2013) and pH (Fierer and Jackson 2006). These physical and chemical properties play important roles in the regulation of extracellular enzyme activity by modifying the diffusion and adsorption of available substrates (Deng et al. 2019;Feng et al. 2019). Thus, the resource limitation of soil microorganisms can be substantially different between topsoil and subsoil. For example, fresh C input is more limited in subsoil than topsoil such that the subsoil may experience greater C limitation than the topsoil does. On the other hand, several recent studies have shown that warming decreases topsoil moisture, and deepens root growth and as a result C input via root exudation increases and C limitation decreases in subsoil (Querejeta et al. 2021;Zheng et al. 2020). The relief of C limitation in subsoil has been suggested to increase soil organic matter (SOM) decay (Shahzad et al. 2018;Sullivan et al. 2020). However, the understanding of how warming affects C and nutrient limitations in subsoil relative to topsoil is incomplete because few studies have focused on subsoil. Considering the subsoil C storage, slight changes in subsoil C may have significant effects on C cycle. Therefore, it is critical to clarify the impact of warming on microbial resource limitations in the subsoil.
Microbial community composition, which is recognized to be sensitive to environmental conditions (e.g., climate and soil properties; Fierer and Jackson 2006), has been shown to be a major factor regulating enzyme production (McGuire and Treseder 2010;Schnecker et al., 2015;Stone et al. 2014;Strickland et al. 2009, Li et al. 2019. Gram-positive (GP) and gram-negative (GN) bacteria are oligotrophic and copiotrophic bacteria, respectively Fanin et al. 2018). GP uses recalcitrant C compounds, exhibiting a higher C limitation than GN do. Comparatively, GN uses more labile C compounds (Fanin et al. 2018;Naylor and Coleman-Derr 2018), showing a lower C limitation than GP. In addition, warming has been shown to increase, decrease, or have no effect on the abundances of GP and GN (Feng and Simpson 2009;Frey et al. 2008;Karhu et al. 2010;Rinnan et al. 2008Rinnan et al. , 2009Vanhala et al. 2011). Moreover, the impact of warming on the soil microbial community composition is likely depth-dependent (Dove et al. 2021). Although many studies have focused on the effects of warming on microbial community structure, the relationship between microbial communities and microbial resource limitation has rarely been examined. Examining the relationship between microbial community structure and resource limitations will help to advance our understanding of the effects of warming on C, N, and P cycles.
In the current study we investigated how warming affects microbial resource limitation and explored its relationships with microbial communities and soil properties in the topsoil and subsoil in a subtropical Chinese fir (Cunninghamia lanceolata) plantation. Based on the commonly reported differences in abiotic factors and microbial communities between topsoil and subsoil, we hypothesized that warming would have different effects on microbial resource limitation between topsoil and subsoil (H 1 ). Because of the critical role of microbial enzymes in nutrient acquisition, we also hypothesized that the microbial community structure plays a primary role in regulating microbial resource limitation (H 2 ).

Study site and experimental design
The study site was located at the Chenda Research Station (300 m above sea level) of Fujian Normal University, Fujian Province, southeast China (26°19N, 117°36E). The site has a subtropical monsoon climate, with a mean annual air temperature of 19.1 °C, annual precipitation of 1750 mm, and annual evapotranspiration of 1585 mm. The soil at the study site has a sandy texture and is classified as red soil on the basis of China's soil classification systems, equivalent to Oxisols in the United States Department of Agriculture Soil Taxonomy (State Soil Survey Service of China 1998; Soil Survey Staff of USDA, 2014).
We established ten 2 m × 2 m plots and then randomly and evenly assigned them to the warming and control treatments. Within each warmed plot (W), a long resistance-heating cable (Nexans-type TXLP, Oslo, Norway) was installed 10 cm into the soil forming nine parallel lines with 20 cm between two adjacent lines. Heating cables were also installed in the unwarmed plots (CT) but were not heated, to account for the effect associated with the installation of the heating cables. The temperature difference between the CT and W was maintained at 5 °C at a soil depth of 10 cm. The cable was installed in August 2013 and the heating began at March 4, 2014. The details of the experimental setup were described in Liu et al. (2017).
Each 2 m × 2 m plot was divided into four 1 m × 1 m subplots, and four Chinese-fir seedlings were planted in each subplot in November 2013. The soil moisture of each plot was measured year-round at 10 cm and 60 cm depths each with one ECH 2 O-5 soil moisture probe (Decagon, Pullman, Washington, 138 USA). The soil temperature was measured using temperature sensors (T109; Campbell Scientific Inc., Logan, UT, USA) placed between the cables. Three temperature sensors were installed at 10 cm soil depth and three at 60 cm soil depth in the W, and two were installed at each of the two soil depths in the CT. Soil temperature and moisture were recorded every 30 min using a computer-based control system. Detailed information on the soil temperature and moisture at 10 cm and 60 cm was described in Liu et al. (2017) and Lin et al. (2018).

Soil sampling and biogeochemical analyses
Bulk soil samples of topsoil (0-10 cm) and subsoil (40-60 cm) were collected with a 3.5 cm soil corer on April 20, 2016 (about two years after warming). Because enzymes produced by plants concentrated in the rhizosphere (Ma et al. 2018), we believe that extracellular enzymes that we measured mostly originate from microorganisms. Upon sampling, six soil cores in each plot were randomly collected between heating cable lines to ensure that all samples received similar heat input. Soil samples from each depth in the same plot were mixed to form a composite sample. The soil samples were stored in a cooling box and immediately transported to the laboratory at the site. The soil was cleared of visible living plant materials and stones, then sieved through a 2 mm mesh. Before further analysis, soil water content was gravimetrically determined (105 °C for 24 h). A soil subsample of approximately 5 g from each soil sample was stored at 4 °C for enzyme activity assessment within one week. The remaining sample was divided into two subsamples, one freeze-dried for phospholipid fatty acid (PLFA) analysis and the other for chemical analyses. All data are expressed on a dry soil weight basis.
Soil pH was determined using a pH meter with a soil:water ratio of 1:2.5. Before measurements of total C and N, a subsample of air-dried soil was ground (< 0.15 mm) using a mortar and pestle. Soil organic C (SOC) and total N (STN) were determined using a CN Autoanalyzer (Elementar Vario MAX, Germany). For soil total P (STP) analysis, 0.25 g air-dried soil was digested with 2 mL HClO 4 and 3 mL H 2 SO 4 for 180 min at 120-130 °C and then diluted with deionized water to 100 mL (Kovar and Pierzynski 2009). After overnight stratification of the digestion liquid, 5 mL supernatant liquid was added to 5 mL of molybdenum antimony reagent, and water was added to 50 mL. The total P concentration of the solution was measured using an ultraviolet spectrophotometer (Hitachi UV2300) at 700 nm. For soil active P (SAP) analysis, 3 g of dry soil was added to 30 mL of M3 extraction solution (soil: solution ratio 1:10), shaken for 5 min (120 oscillations min −1 ), and centrifuged for 10 min. The supernatant was filtered through a 2.5 µm Whatman no. 42 filter paper. The SAP concentration in the filtrate was determined using a Skalar SAN plus Segmented Flow Analyzer. Soil ammonium N (NH 4 + -N) and nitrate N (NO 3 − -N) were determined in the KCl extract (Zhou et al. 2012). Briefly, each fresh soil sample (5 g) was added to a 50 mL tube with 20 mL 2 M KCl, shaken for 0.5 h, and then centrifuged for 10 min. The supernatant was filtered through a 2.5 µm Whatman no. 42 filter paper. The concentrations of inorganic N in the supernatants were measured using a continuous flow analyzer (Skalar san + + , Netherlands). Dissolved organic C (DOC) and organic N (DON) were measured using K 2 SO 4 extract. Briefly, 5 g of a soil sample was extracted with 0.5 M K 2 SO 4 solution (1:4, w/v, soil/extractant ratio), shaken for 0.5 h and centrifuged for 10 min, and then filtered using a 0.45 mm nitrocellulose filter. The organic C concentration in the supernatant was measured using a TOC analyzer (TOC-VCPH/CPN, Shimadzu, Japan) as the DOC concentration. Mineral nitrogen (NH 4 + -N, NO 3 − -N) and total N concentration in the supernatant were measured using a continuous flow analyzer (Skalar san + + , Netherlands). The organic N concentration in the supernatant was derived from the difference between the total N concentration and mineral N concentration as the DON concentration.

Potential enzyme activities analysis
The activities of four enzymes, BG (EC 3.2.1.21), NAG (EC 3.1.6.1), AP (EC 3.1.3.2), and peroxidase (PER, EC 1.11.1.7), were measured following German et al. (2011), which is a modified method of Saiya-Cork et al. (2002). BG, NAG, and AP are related to microbial C, N, and P acquisition, respectively (Sinsabaugh et al. 2009), and PER is related to the decomposition of recalcitrant organic C (Sinsabaugh 2010). Briefly, 1 g fresh soil was homogenized in 125 mL sodium acetate buffer (pH 5.0) using a magnetic blender. Fifty μL of fluorescent substrate proxies specific to each enzyme (BG: 4-MUB-β-D-glucoside, NAG: 4-MUB-N-acetylβ-D-glucosaminide, AP: 4-MUB-phosphate, PER: L-DOPA) were added to eight replicate assay wells with 200 µl soil slurry at optimal concentrations to measure the total potential activity (optimal concentrations were determined before the experiment). Assays were performed using two standard columns containing soil homogenates and methylumbelliferone (MUB). Each assay microplate also contained blank substrate columns containing 50 μL substrate and 200 μL sodium acetate buffer. Soil homogenate blanks were measured simultaneously. Assays for BG, NAG, and AP were incubated at 20 °C for 4 h (optimal duration of assay determined before the experiment). Next, 10 μL 1 M NaOH was added to each well to terminate the reaction. Within 1 min of NaOH addition (DeForest 2009), and activity was measured fluorimetrically (excitation 365 nm and emission 450 nm). Plates for PER activity measurements were incubated for 18 h at 20 °C; fluorescence was measured (absorbance 450 nm).

Vector analysis for measurement of resource limitation
To measure the extent of soil microbial C and nutrient limitation, we followed the vector analysis of soil enzymatic stoichiometry proposed by Moorhead et al. (2013). Vector lengths and angles were measured by plotting the proportional enzyme C:P to C:N ratios, which were calculated to illustrate the potential and relative resource use limitations for soil microorganisms. Vector length represents the relative C and nutrient investment (Deng et al. 2019), a longer vector length means a higher C limitation. The vector angle represents the relative P and N investment (Moorhead et al. 2013). Vector angles greater than 45 degrees indicates P limitation, and the larger the angle, the stronger P limitation. Otherwise, it means N limitation. The smaller the angle, the stronger N limitation We calculated the vector length and vector degree as follows: where atan2 is a trigonometric function that provides the angles in radians between the x-axis and the vector from the plot origin to point (x, y), and 180/π is used to convert the angles in radians into degrees.

Statistical analysis
All results are reported as the mean ± standard error (SE). We used two-way analysis of variance followed by Tukey HSD tests to examine differences in measured items between the topsoil and subsoil and between warmed and control treatments. Differences between groups were considered statistically significant at P < 0.05. Before analysis, the data were log transformed or rank normalized to fulfill the normal distribution assumption and homogeneity of variance. Statistical analyses were performed using the vegan package (Oksanen et al. 2019).
To examine the relative importance of soil physiochemical properties and microbial communities on regulating soil microbial C and nutrient limitation, using vector length and vector angle as indicators, respectively, we performed multiple linear regression analysis followed by variation partitioning analysis between the two explanatory variables (Borcard et al. 1992). Firstly, we conducted principal component analyses for (a) physicochemical properties and (b) microbial community composition separately for the topsoil and subsoils. The first two principal components (PCs) of the soil variables explained 65% of the variation in the soil physicochemical properties in the topsoil [ Fig. S1] and 64% in the subsoil [ Fig.  S2]). The first two PCs of the microbial communities explained 95% of the variation in microbial community composition in the topsoil (Fig. S3) and 94% in the subsoil (Fig. S4). Then, the first two PCs were used for multiple linear regression analysis. The R 2 of each multiple linear regression model was then used for variation partitioning analysis to estimate the relative importance of the two types of variables in regulating soil microbial C and nutrient limitations, and Vector Length = √ (BG/AP) 2 + (BG/NAG) 2 Vector Angle = a tan 2 (BG/AP, BG/NAG) × 180∕ the variable with higher R 2 were selected as the independent variable in next multiple regression analysis to examine the relationships among soil microbial C, nutrient limitation, and soil physiochemical properties (all measured soil variables except SOC, STN, and STP, which were conjugated to other soil variables) and the five soil microbial groups.
In multiple regression analysis, to evaluate potential collinearity of relationships among model covariates (Schmidt-Nielsen 1984), we calculated variance-inflation factors (VIFs) for each covariate in each model and excluded the covariates with a VIF greater than 10 (Dormann et al. 2013;Schmidt-Nielsen 1984). A VIF was calculated by using the function "vif" from the package "car" in R. Next, a model selection process was used to select the best explanatory variable based on corrected Akaike's information criterion (AICc; ΔAICc < 4; Burnham and Anderson 2002). The procedures were conducted for the topsoil and subsoil separately (Table S1-S4), using the function "dredge" in the R package "MuMIn" (Bartoń, 2020). Model averaging was performed also based on the AICc weights when multiple models were selected. All predictors and response variables were standardized before analysis by using Z scores to interpret parameter estimates on a comparable scale. Predictors were log transformed when necessary before the analysis to fulfill the assumptions of the tests used. All statistical analyses were conducted using R software (version 3.4.1, The R Foundation for Statistical Computing, Vienna, Austria).

Soil physiochemical properties
Warming increased topsoil DOC concentrations, decreased topsoil NH 4 + -N concentration and subsoil DOC and DON concentrations, but had no significant    (Table 1).

Microbial C and nutrient limitation
Warming increased microbial C limitation in the topsoil (i.e., increased vector length), but decreased it in the subsoil (i.e., decreased vector length, Fig. 1a). Warming alleviated microbial P limitation (i.e., decreased vector angle) in both the topsoil and subsoil (Fig. 1b). The vector angle of enzyme activities was smaller than 45° in the warmed subsoil, indicating a greater N than P limitation for subsoil microbes (Fig. 1b). Across the CT and W, vectors were longer in the subsoil than in the topsoil, while the vector angle was smaller in the subsoil than in the topsoil in the control and warming treatments (Fig. 1). Warming significantly increased topsoil BG (26%, P = 0.002) and NAG (12%, P = 0.041) activity while it decreased topsoil AP activity (− 50%, P = 0.005) (Table 1). However, warming did not affect subsoil NAG activity and decreased subsoil BG activity by 13% (P = 0.037) and AP activity by 45% (P < 0.001) ( Table 1). Warming increased the ratios of BG:AP and NAG:AP in the topsoil and subsoil, decreased the ratio of BG:NAG in the subsoil, and had no effect on the BG:NAG ratio in the topsoil (Table 1).

Microbial community composition
Warming did not affect the microbial community composition in the topsoil but decreased the abundance of ACT in the subsoil (Fig. 2). Microbial PLFA contents are shown in Table S5.

Variation partitioning on vector analysis
Soil physicochemical properties and microbial community composition accounted for 87% of the variation in soil microbial C limitation (i.e., variance in vector length) and N versus P limitation (i.e., variance in vector angle) in the topsoil. Variance decomposition showed that topsoil microbial C and nutrient limitations were primarily explained by soil physicochemical properties. The soil physicochemical properties accounted for 54% of the variation in microbial C limitation (Fig. 3a) and 45.4% of the variation in microbial nutrient (P vs. N) limitation (Fig. 3b) in the topsoil. By contrast, microbial communities only accounted for ~ 1% of the variation in microbial C Fig. 1 Vector lengths and angle in topsoil and subsoil in control and warming treatments. a) vector lengths; b) vector angle. Values (means ± SE, n = 5). Different uppercase letters and lowercase letters indicate significant differences between control and warming at the same depth and between two depths in the same treatment, respectively relative to the nutrient limitation and P relative to the N limitation in the topsoil.
Soil physicochemical properties and microbial community composition accounted for 72% of the soil microbial C limitation (Fig. 3c) in the subsoil and 85% of the nutrient (P vs. N) limitation (Fig. 3d) in the subsoil. The majority (64%) of the variation in microbial C limitation in the subsoil was explained Fig. 2 Proportional mol % of microbial communities' PLFA in topsoil and subsoil in control and warming treatments. Values (means ± SE, n = 5). Different asterisks indicate significant differences between control and warming at the P < 0.05 significance level at the same depth. GP: Gram-positive, GN: gramnegative, ACT: actinomycetes Fig. 3 Venn diagram showing variation partitioning in soil microbial C and nutrient limitations in two sets of explanatory variables (soil physicochemical properties and microbial communities). a Soil microbial C limitation in topsoil, b soil microbial P vs. N limitation in topsoil. c Soil microbial C limitation in subsoil, d soil microbial P vs. N limitation in subsoil. Values denote the proportion of variance accounted for by each of the explanatory variables by microbial communities, and a much smaller proportion (8%) was explained by soil physicochemical properties in the subsoil (Fig. 3c). Similarly, the majority (59%) of the variation in microbial nutrient limitation was explained by soil physicochemical properties, and a much smaller proportion (12%) was explained by microbial communities in the subsoil (Fig. 3d).
In the topsoil, microbial C limitation was negatively related to NH 4 + -N concentration (Fig. 4a), while microbial N versus P limitation was positively related to NH 4 + -N concentration (Fig. 4b). The subsoil microbial C limitation was positively related to ACT (Fig. 4c), and microbial nutrient limitation was negatively related to SAP concentration but positively related to DON concentration and pH (Fig. 4d).

Effects of warming on microbial resource limitation
Warming increased the vector length in topsoil (Fig. 1a), which suggests that warming increases the microbial C limitation. This result is contrary to the findings that warming decreases soil labile C in subtropical ecosystems Melillo et al., 2017). The increased PER activity (Table 1) in topsoil under the warming treatment relative to the control supports that labile SOC was largely depleted by warming. Thus, warming is likely to aggravate than mitigate microbial C limitation by reducing soil carbon availability in the topsoil in subtropical ecosystems. In contrast, microbial C limitation was decreased by warming in subsoil. Previous studies in our study site found that warming increased fine root biomass in deep soils with the potential of increasing C input to the subsoil (Xiong et al. 2018), which may explain the warming-induced decreases of microbial C limitation in subsoil (Giardina et al. 2014;Melillo et al. 2002). However, warming decreased DOC in the subsoil, possibly because DOC was quickly consumed by organisms in the C-limited subsoil despite the increased C input associated with the warminginduced fine root biomass increase. The greater P than N limitation for microbial communities in the studied subtropical ecosystem (Fig. 1b) is consistent with the findings of many studies that point to P limitation of soil microbial communities in tropical and subtropical ecosystems Fig. 4 Model-averaged effect size of the predictors on microbial C and P vs. N limitation (based on Z scores with linear mixedeffects models). a Soil microbial C limitation in topsoil, b soil microbial P vs. N limitation in topsoil. c Soil microbial C limitation in subsoil; d soil microbial P vs. N limitation in subsoil. M: soil moisture (Camenzind et al. 2018;Sinsabaugh et al. 2008;Xu et al. 2017). The observed warming alleviated microbial P limitation in the topsoil is in agreement with a translocation warming experiment in tropical China (Li et al. 2019). Soil microbial turnover was accelerated by warming (Hagerty et al. 2014), through which P was provided for other soil microorganisms (Hagerty et al. 2014), thus P limitation of soil microorganisms was alleviated. In addition, our results suggested that warming shifted the microbial nutrient limitation from P to N in subsoil. The relative limitation of microbial N vs P evaluated by the vector analysis in 45 degrees is based on the study of Sinsabaugh et al. (2008), which reported that soils from a variety of terrestrial ecosystems have an approximately 1:1:1 intersection of C:N:P acquiring enzyme activity. In the study of Sinsabaugh et al. (2008), NAG and leucine aminopeptidase (LAP) were both considered as microbial N acquisition enzymes. However, Rothstein et al. (1996) found that LAP only contributed 10-20% of the sum of NAG + LAP at pH < 7. Moreover, LAP is positively correlated with soil pH (Sinsabaugh et al. 2008). Thus, LAP likely only provided a minor portion of N-acquiring enzymes in the current study as the soil pH in our study site ranged from 4.11 to 5.04 (Table 1). Thus, the lack of including LAP as N-acquisition enzyme in our study should have limited impact on our inference of microbial resource limitation but it likely resulted in underestimation of the microbial N limitation.
The greater soil microbial C limitation in the subsoil than in the topsoil in our study may have resulted from strong organo-mineral associations in the subsoil (Dungait et al. 2012;Fontaine et al. 2007;SalomÃ et al. 2010). In addition, we found that the subsoil PER activity is higher than that of the topsoil (Table 1), which is consistent with the results of other studies (Kramer et al. 2013;Uksa et al. 2015). This probably derives from the fact that the pH of subsoil (5.02) is closer to the optimum pH of PER, which is considered to be 8 ± 1 (Sinsabaugh et al. 2008), than that of topsoil (4.31) ( Table 1). The smaller soil microbial P limitation in the subsoil than in the topsoil (based on the vector angle) (Fig. 1) can be explained by the similar total and available P between the topsoil and subsoil but lower total N and DON in the subsoil than in the topsoil (Table. 1).
Despite our findings suggesting that microbial resource limitation differed between the topsoil and subsoil, and in response to warming, our study has some potential limitations. Firstly, the evaluation of microbial C and N limitations is complicated by the fact that the N-acquisition enzyme (NAG) also supplies C for microbes (Moorhead et al. 2016;Mori et al., 2020Mori et al., , 2021. As a result, increases in the activity of NAG does not always reflect N limitation. Secondly, microbial relative N limitations may be underestimated because of disregarding LAP, another N-acquisition enzyme, as described above. Thirdly, a single enzyme, BG, was utilized to characterize microbial C acquisition in vector analysis, which may result in a misinterpretation of the microbial C limitation. Because BG is the primary enzyme for catalyzing cellulose decomposition (Sinsabaugh and Shah 2012), it mainly captures the breakdown of cellulose but not all types of soil organic carbon. Although our study also incorporated enzymes (e.g., PER) involved in the decomposition of recalcitrant C, such as lignin and humus (Sinsabaugh et al. 2008), it may not fully reflect microbial C limitation. Therefore, it is important to proceed cautiously when concluding how warming affects microorganisms' C and N vs P limitations based on vector analysis. Experimental C, N, and P additions are suggested to validate our findings.
Soil properties and microbial communities in relation to microbial C and nutrient limitation Warming had major effects on the physical and chemical properties of soil, such as increasing soil temperature and decreasing soil moisture (Table 1). We found that soil physicochemical properties accounted for ~ 90% of the variation in microbial C limitation in the topsoil (Fig. 3a), suggesting that changes in soil physicochemical properties have major impacts on the microbial C limitation of the topsoil. In contrast, soil microbial community composition accounted for ~ 64% of the variation in subsoil microbial C limitation (Fig. 3c), implying that warming-induced changes in the microbial community (Biasi et al. 2005;Creamer et al. 2015;Schindlbacher et al. 2011) probably affect subsoil C limitation. Notably, soil physicochemical properties accounted for the majority of microbial nutrient limitation in the topsoil and subsoil, highlighting the potential effects of warming on microbial nutrient limitation by altering soil physicochemical properties. These results only partially rejected H 2 , which predicts that microbial community structure plays a primary role in regulating microbial resource restriction.
Based on the regression models, NH 4 + -N explained the majority of warming-induced increases in microbial C and nutrient limitation in the topsoil under warming (Fig. 4a). Microbial C limitation has been suggested to be affected by soil N availability due to stoichiometrically coupled C and N cycling (Moorhead et al. 2016;Sinsabaugh et al. 2008). When N is limited, microbes may use C acquisition enzymes to break down organic matter for N mining (Craine et al. 2007), which leads to the aggravation of microbial C limitation. Thus, the increase in microbial C limitation under warming may have resulted from the reduced NH 4 + -N under warming. In the subsoil, microbial C limitation was positively correlated with ACT abundance (Fig. 4c). Actinomycetes are filamentous GP bacteria that favor consuming recalcitrant organic C in subsoil (Potthoff et al. 2006). The reduction in ACT abundance means that the microbial community has less demand for recalcitrant organic carbon and microbial carbon limitation is alleviated. Notably, the shift in microbial nutrient limitation from P to N may have resulted from increased SAP and decreased DON caused by warming (Fig. 4d).

Conclusions
Our study provides empirical evidence that the microbial C and nutrient limitation of topsoil and subsoil responded differently to warming in subtropical Chinese fir plantations. Warming aggravated the microbial C limitation in the topsoil. However, warming mitigated C limitation in the subsoil. In addition, warming shifted the subsoil nutrient limitation from P-to N-limited soil. Furthermore, we found that microbial C limitation was more related to soil physicochemical properties than microbial community in the topsoil but closely related to microbial community composition in the subsoil. Soil properties explained most of the variation in microbial nutrient limitation in the topsoil and subsoil.
In summary, our study provides insights into soil C and nutrient limitations of microbial metabolism in subtropical forests and highlights that the main regulatory factors of microbial C and nutrient limitation differ between topsoil and subsoil. These findings advance the understanding of soil microbial resource limitations and the potential effects of warming on soil C sequestration and nutrient cycles in subtropical forest ecosystems.