Nitrogen Addition Affects Soil Respiration Primarily through Changes in Microbial Community Structure and Biomass in a Subtropical Natural Forest

Forest soil respiration plays an important role in global carbon (C) cycling. Owing to the high degree of C and nitrogen (N) cycle coupling, N deposition rates may greatly influence forest soil respiration, and possibly even global C cycling. Soil microbes play a crucial role in regulating the biosphere–atmosphere C exchange; however, how microbes respond to N addition remains uncertain. To better understand this process, the experiment was performed in the Castanopsis kawakamii Hayata Nature Reserve, in the subtropical zone of China. Treatments involved applying different levels of N (0, 40, and 80 kg ha−2 year−1) over a three-year period (January 2013–December 2015) to explore how soil physicochemical properties, respiration rate, phospholipid fatty acid (PLFA) concentration, and solid state 13C nuclear magnetic resonance responded to various N addition rate. Results showed that high levels of N addition significantly decreased soil respiration; however, low levels of N addition significantly increased soil respiration. High levels of N reduced soil pH and enhanced P and C co-limitation of microorganisms, leading to significant reductions in total PLFA and changes in the structure of microbial communities. Significant linear relationships were observed between annual cumulative respiration and the concentration of microbial biomass (total PLFA, gram-positive bacteria (G+), gram-negative bacteria (G−), total bacteria, and fungi) and the microbial community structure (G+: G− ratio). Taken together, increasing N deposition changed microbial community structure and suppressed microbial biomass, ultimately leading to recalcitrant C accumulation and soil C emissions decrease in subtropical forest.


Introduction
Anthropogenic reactive nitrogen (N) production originated primarily from agricultural activities, fossil fuel combustion, and the growing popularity of biofuels, and has increased three-to five-fold over the past century [1]. By 2050, N deposition is projected to reach 200 Tg N year −1 , especially in forest ecosystems [1,2]. Approximately twice as much C is stored in soils compared to that in the atmosphere. Soil respiration (Rs) is the primary pathway through which C is released from the soil system into the atmosphere [3]. Thus, even minor changes in Rs would have significant effects on C The experiment was carried out in the Castanopsis kawakamii Hayata Nature Reserve, which is in central Fujian Province, China (117 • 28 E, 26 •  The climate is classified as a subtropical monsoon, with mean annual precipitation of 1552 mm, 2141 mm, and 2025 mm in 2013, 2014, and 2015, respectively. It has distinct seasons, with most rain falling between March and August. Mean annual temperature, potential evapotranspiration, and relative humidity in this region were 18.7 • C, 1585 mm, and 79%, respectively [31]. Regional soils are Oxisols, formed from sandstone (based on the United States Department of Agriculture Soil Taxonomy), and are about 30-70 cm deep [32].

Experimental Design
Three N addition treatments (with four replicates each) were established in this forest, consisting of 0 kg ha −2 year −1 (control, CT), 40 kg ha −2 year −1 (low N, LN), and 80 kg ha −2 year −1 (high N, HN). Treatment levels were based on known background atmospheric N deposition rates in subtropical regions of China (18-53 kg ha −2 year −1 ), with an average deposition rate of~40 kg ha −2 year −1 [33]. A total of 12 plots (20 m × 20 m) were established, each surrounded by a 10-m wide buffer zone and unshielded from natural atmospheric N deposition. The plots and treatments were set randomly. Beginning in November 2012, a solution of ammonium nitrate (NH 4 NO 3 ), and 20 L of deionized water was distributed monthly below the canopy with a backpack sprayer, totaling 12 applications of equal volume annually, and an equivalent volume of deionized water was sprayed on the control plots.

Soil Sample Collection
Five soil cores were collected from each subplot with a 3.5-cm-diameter corer in January 2016. We removed the surface litterfall and collected soil samples from the A horizon (0-10 cm). Soil cores were then kept in portable refrigerated box until being processed in the laboratory. After removal of plant roots and stones, soil samples were sieved through a 2-mm mesh and stored at 4 • C prior to the analysis of inorganic N, dissolved organic C, and N and microbial phospholipid fatty acid (PLFA) content. Part of the soil was air-dried for measuring its pH. The remaining soil was air-dried and ground (<150 µm) for determination of total C and total N.

Soil Respiration Rate Measurement
Soil respiration rate (R s ) was measured using an automated CO 2 efflux system (LI-8100, LI-COR Inc., Lincoln, NE, USA). Eight polyvinyl chloride (PVC) collars (diameter: 20 cm; height: 10 cm) were fixed in each plot in August 2011. Living plants inside the collars were removed and kept for almost 1.5 years to minimize disruption. Soil respiration was assessed once every two weeks over the course of the experimental period. Measurements were taken between 09:00 and 12:00, as soil flux over these hours has been shown to represent the mean of the whole day [34]. Soil temperatures and moisture were simultaneously monitored using a hand-held long-stem thermometer (Model SK-250WP, Sato Keiryoki Mfg. Co. Ltd, Tokyo, Japan) and a time-domain reflectometer (TDR) (Model TDR300, Spectrum Technologies Inc., Plainfield, IL, USA), respectively. The data of monthly soil temperature and annual moisture during study period are shown in Figures S2 and S3.
To examine the effects of N addition rate on R s in sub/tropical forests, data were obtained from 15 peer-reviewed articles (Table S4) by searching Web of Science. The searched key words were combinations of, "nitrogen (N) addition," "nitrogen (N) deposition," "soil respiration," "subtropical forest" and "tropical forest". Data were selected based on the following criteria: (1) from a field study (data from incubation studies were excluded); (2) from control and simulated N addition treatments in multifactorial studies. Response ratios of R s to N addition rate (RRs) were calculated using the following equation: where R T is the treatment mean and R C is the control mean. Here, RRs > 0 means N addition increased soil respiration; RRs = 0, means N addition has no effect on soil respiration; RRs < 0 means N addition reduced soil respiration.

Phospholipid Fatty Acid Analysis
The soil microbial community was characterized using a phospholipid fatty acid (PLFA) analysis, as previously described by Wan et al. [35]. In brief, a solvent consisting of a 2:1:0.8 mixture of methanol (CH 3 OH), chloroform (CHCl 3 ), and phosphate buffer (pH 7.4) was used to extract 10 g of freeze-dried soil by shaking for 2 h. The samples were centrifuged at 3500 g for 10 min, and then the supernatant was transferred to a new tube. The remaining soil was re-extracted as described above. The extracted solvents from both steps were combined and then evaporated to 1 mL under N 2 gas. Then, neutral glycolipids, glycolipids-, and polar lipids were separated over a silicon hydroxide column eluted with chloroform, acetone, and methanol, respectively. Polar lipids were methylated to form fatty acid methyl esters (FAMEs) by subjecting them to 0.2 M methanolic KOH. Individual FAMEs were identified by Hewlett Packard 5890 gas chromatography, equipped with a 6890 series injector, a flame ionization detector, and an Ultra 2 capillary column (25 m × 0.2 mm inner diameter, film thickness, 0.33 µm) based on their retention times and in combination with the MIDI Sherlock Microbial Identification System (MIDI Inc., Newark, DE, USA).

Solid-State 13 C Nuclear Magnetic Resonance Spectroscopy Analysis
Soil samples for solid state 13 C cross polarization magic angle spinning (CP-MAS) nuclear magnetic resonance (NMR) analysis were repeatedly treated with 2% hydrofluoric acid, then rinsed with deionized water, freeze-dried, and ground into powder [42]. The powdered samples were packed into 4-mm zirconium rotors. Solid state 13 C NMR spectra were acquired on a 500 MHz Bruker BioSpin Avance III spectrometer (Bruker BioSpin, Rheinstetten, Germany) equipped with a 4-mm probe. The parameters used to obtain the spectra consisted of a 13 kHz spinning rate, 1 ms ramp-CP contact time, 1 s recycle delay, and 4096 scans. Glycine was used as the external reference for chemical shift. NMR spectra were processed using a zero filling factor of 2 and 75 Hz line broadening. The NMR spectra were divided into seven regions representing the different chemical environments of the 13 C nucleus (Table S1). The ratios were calculated using percentage intensity values as follows [43,44]: aromaticity = aromatic C/(alkyl C + methoxyl and N-alkyl C + O-alkyl C + (3) Di-O-alkyl + phenolic C + aromatic C)

Additional Soil Analysis
Soil inorganic N was treated with 2 M KCl and analyzed using a Continuous Flow Analytic System (Skalar san++, Skalar, Breda, Netherlands); soil organic C (SOC) and total N (TN) were determined with an elemental analyzer (Elementar Vario EL III, Elementar, Langenselbold, Germany). Dissolved organic C (DOC) and N (DON) were extracted from 10 g of field-moist soil by mixing the soil with 40 mL deionized water at 20 • C and shaking for 30 min, then filtering the supernatant through a 0.45-µm filter membrane [45]. Soil pH was determined with a pH meter (STARTER 300, OHAUS, Pine Brook, NJ, USA) in a 1:2.5 soil: water solution. Soil moisture content was measured gravimetrically by drying for 48 h at 105 • C.

Statistical Analyses
Daily cumulative respiration (R c ) was calculated from R s as follows: Soil respiration was measured every two weeks with a total of 24 samplings/year. Annual cumulative respiration was calculated by daily R c multiplied by the number of sampling interval days. Equation used was as follows: The relationship between the R s and soil temperatures was performed using following widely exponential regression model [46]: where R s is the soil respiration rate, t is the soil temperature at 5 cm depth, a and b are the model coefficients.
The apparent temperature sensitively (Q 10 ) was calculated as follows: All statistical analyses were performed using SPSS v.21.0 (SPSS Inc., Chicago, IL, USA). All response variables were tested for normality and homoscedasticity prior to statistical analyses, and data were log-transformed when the assumptions were not met (DON, ammonium nitrogen [NH 4 + -N]). One-way analysis of variance (ANOVA) with the Tukey's HSD test was used to evaluate the differences in soil physicochemical properties, annual cumulative respiration, Q 10 , and PLFA in response to different N-addition treatments. Linear regression model analyses were conducted to explore the relationships among PLFA and annual cumulative respiration.

Response of Soil Physicochemical Properties to N Deposition
No significant differences in SOC, TN, ammonium N (NH 4 + -N), or nitrate N (NO 3 − -N) were detected after three years of N deposition (Table 1). Soil pH decreased with increasing N, with pH significantly reduced (by 0.13 units) in the HN treatment. Soil DOC was significantly higher in the LN treatment, and soil DON increased with higher N addition, with levels in the HN treatment being significantly higher than those in the CT and LN treatments.

Response of Soil Respiration Rate and Temperature Sensitivity to N Addition
Monthly dynamics of Rs rate showed a strong seasonal pattern, with the highest rate observed in July-August and the lowest in January-February for all treatments (Figure 1). Annual cumulative Rs in the LN treatment was 15.62%, 19.16%, and 23.29% higher than in the CT treatment in 2013, 2014, and 2015, respectively, while annual cumulative Rs in the HN treatment was 11.86% and 16.68% lower relative to the CT treatment in 2014 and 2015, respectively ( Figure 2). However, during the period of 2013 to 2015, the sensitivities of Rs to soil temperature were not significantly different among N-addition treatments ( Table 2 and Table S2).

Response of Microbial Community to N Addition
The concentrations of G + , G − , fungi, and total PLFA were significantly decreased in the HN treatment, but no significant differences were observed for any of the PLFA between the LN and CT treatments ( Figure 3; Table S3). The ratio of G + : G − was significantly higher in the HN treatment than in the CT treatment (1.5 and 1.1, respectively), but no significant difference in F:B ratio was observed among the treatments.

Response of Soil Chemical Characteristics to N Addition
The most dominant component was the alkyl C region (Figure 4; Table S1). Alkyl C, aromatic C, phenolic C, and carboxyl/carbonyl C exhibited major differences in the HN treatment. Thus, differences in these chemical shift regions caused a 9% reduction in A/O-A and a 5% increase in aromaticity.

Correlation between Annual Cumulative Soil Respiration and Microbial Biomass, Root Biomass and Microbial Community Structure Ratio
Significant linear relationships were found between annual cumulative Rs and the concentration of microbial biomass (total PLFA, G + , G − , total bacteria, and fungi) and the G + : G − ratio according to linear regression. However, neither root biomass nor the F:B ratio was significantly correlated with annual cumulative Rs in 2015 ( Figure 5).

Effects of N Addition on Soil Respiration
Nitrogen deposition did not affect the seasonal patterns of Rs rates, with the highest rate observed in July-August and the lowest in January-February (Figure 1). This seasonal pattern has been reported in previous studies and can be ascribed to differences in mean temperature [47,48]. N addition induced significant changes in Rs rates (p < 0.05); however, the different rates of N addition exhibited distinct trends in terms of their effect on Rs. This discrepancy in annual cumulative respiration among the treatments continued to diverge over time, which was consistent with results reported by Allison et al. [28], Hasselquist et al. [30], Bowden et al. [48], and Maaroufi et al. [49]. This is because a certain threshold for N addition exists. For instance, in a subtropical Moso bamboo forest ecosystem, Li et al. [6] found that a N addition rate of 60 kg N ha −1 year −1 may reflect a N saturation threshold. When N addition rate exceed 60 kg N ha −1 year −1 , N addition still increased Rs, but the positive effects diminished. However, in our study, a significant decrease in Rs was observed in 2014 and 2015 when N addition rate exceed 60 kg N ha −1 year −1 , possibly because the bamboo plantation was N-limited and had a high demand for N [50], unlike the natural forests examined in previous research.
Previously, there have been six meta-analyses with regard to the effects of N addition on Rs [9][10][11][12]51,52]. However, these studies paid little attention to subtropical forests owing to the smaller sample sizes reported. With the increasing attention paid to subtropical forests in recent years, several studies have been carried out. We recollected a series of data from 16 study sites (including this study) and defined <60 kg N ha −1 year −1 as a low N addition rate according to a N saturation threshold of 50-60 kg N ha −1 year −1 for global aboveground net primary production [53]. We found strong evidence to indicate that N addition significantly reduces Rs in subtropical forests as observed in all examined studies (n = 58, p = 0.011; Figure 6, Table S4). However, the magnitude of N addition rate also affected the response of Rs, which mainly showed as a high N addition rate reducing Rs (n = 37, p = 0.046), while the response of Rs to low N addition rate varied (n = 21, p = 0.052). It is noteworthy that N addition rate is a dominant factor affecting soil acidification and total microbial biomass; however, Zhou et al. [54] indicated that the effect of N addition rate was ignored in several of the previous meta-analyses, such as Janssens et al. [11], Treseder [51], and Lu et al. [52]. In particular, in subtropical forest ecosystems, rapid N addition exacerbates the loss of NO 3 − combined base cations (K + , Na + , Ca 2+ , and Mg 2+ ) through leaching, which in turn causes nutrient cations to be lost at a faster rate than minerals can be replenished [55]. A lack of base cations can be harmful to vegetative and microbial growth in subtropical forest ecosystems [56,57] leading to lower Rs. In our control treatment, the Q 10 value (2.07) was similar to that reported from a subtropical rehabilitated forest (2.1) [58], but lower than those from a subtropical disturbed forest (2.3) [58], a Moso bamboo forest (2.29) [6], a sweetgum forest (2.73) [59], and a larch forest (3.24) [14] across the temperate zone. This is in line with the observations of Wang et al. [60], in which Q 10 showed a positive relationship with latitude in forest ecosystems. In addition, the C:N ratio is a dominant factor for regulating Q 10 , owing to the shift from C limitation to nutrient limitation with increasing latitude [60,61]. This result may support the microbial N mining theory, which suggests that microbes decompose more SOM to obtain sufficient N at high temperature in high latitude forests with low N availability [62].

Microbial Community and Carbon Structure under N Addition
In accordance with our first hypothesis, we found that the HN treatment reduced the concentration of bacteria, fungi, actinomycetes, and unclassified biomarkers (Figure 3; Table S3), leading to a significant reduction in total PLFA. Both incubation and field studies have definitively shown that Rs and microbial biomass are consistently suppressed following N addition [11,51,63]. However, a recent meta-analysis by Zhou et al. [54] revealed that a decrease in microbial biomass is not always associated with N addition suppressing microbial activity. Interestingly, the companion study reported that the C:N:P stoichiometry in microbial biomass was significantly altered in the HN treatments and enhanced microbial P limitation [32]. Additionally, high N addition significantly reduced soil pH. Therefore, these results collectively demonstrate that high N addition inhibits microbial growth.
Ramirez et al. [63] suggested that understanding how N addition induces change in soil microbial communities is imperative for better understanding soil C storage dynamics. High availability of N could alter the microbial process of SOC that is controlled by the microbial community [64,65]. In our study, the HN treatments not only inhibited microbial biomass but also shifted microbial composition (increased G + :G − ratio). The two different groups of Gram-stained bacteria, classified by their cell wall compositions, have been shown to differ in their preferences regarding substrate conditions and living strategies in a changing environment [66]. G + are well-adapted to low SOM substrates, while G − prefer conditions with high organic matter availability [67,68]. Increase in the G + :G − ratio under the high N addition treatment not only indicated a low quality substrate but also an acclimation of microbes to changes in substrate and nutrient availability.
Undoubtedly, the inhibition in microbial biomass and shift in soil microbial community structure affected the soil C structure, to an extent. The results of the solid state 13 C NMR spectroscopy analysis showed that the relative proportions of aromatic C and phenolic C, which originate from lignin and amino acids of peptides [43,69], increased under the HN treatment ( Figure 4; Table S1). Wang et al. [65] also observed that lignin-derived phenols accumulated in soil with long-term N addition (22 years). There are two plausible reasons for this observation. First, that N addition significantly increased the lignin content of the plant and litter [21], leading to lower substrate decomposition rates with higher lignin content [70]. Second, phenolic compounds are highly resistant to degradation, being susceptible only to a handful of fungal species that are more efficient at lignin decomposition [71]. Thus, non-preferred microbial substrates would accumulate under high N addition.

Correlation between Soil Respiration and Microbial Biomass, Root Biomass, and Microbial Community Structure Ratio under N Addition
A significant correlation was observed between microbial traits (biomass and community structure) and cumulative Rs rather than root biomass ( Figure 5). These results supported our second hypothesis, that microbial traits are primary factors affecting Rs under N addition, which is to say that the decrease in Rs is mainly due to microbes. A reasonable explanation is that microbes may be subject to co-limiting factors in subtropical forests, such as C and P, rather than N [65,72,73]. This pattern of Rs under N addition is different in N limited ecosystems, which have often been reported to increase Rs via root products and biomass [74]. Importantly, heterotrophic respiration dominates Rs (almost 72%) [75], and the positive relationships were observed between root/fine root biomass and autotrophic respiration [15,76] and microbial biomass and heterotrophic respiration [77]. Additionally, N addition significantly increased root biomass ( Figure S1) and decreased microbial biomass, which collectively suggested that heterotrophic respiration may have decreased even further under N addition.
Decreasing microbial biomass is always accompanied by decreasing microbial diversity [78]. Microbial diversity is a vital determinant of ecological function that cannot be obtained using the PLFA method [79]. Thus, it is necessary to employ nucleic-acid-based methods to link microbial diversity to function. In this study, N addition inhibited microbial biomass and shifted microbial community structure, which disrupted the microbial process of SOM, leading to a decrease in Rs and an increase in recalcitrant C accumulation, but was also beneficial to an increase in forest soil C sequestration. However, nutrient release from microbial decomposition will slow down as SOC storage increases. Especially in natural forests without intensive management practices (such as fertilization), shifts in biogeochemical cycling will alter the productivity of ecosystems.

Conclusions
The effects of N addition on Rs are subject to an N saturation threshold, which triggered low N addition to increase Rs and high N addition to decrease Rs in the subtropical C. carlesii forest investigated. High N addition increased P limitation and decreased pH. Additionally, high N addition led to recalcitrant C accumulation, restricting microbial utilization. These processes collectively reduced the total PLFA concentration and shifted microbial community structure (G + : G − ratio). Our results suggest that microbial traits are dominant factors affecting Rs. As such, additional research on the transformations in soil microbial traits in response to increasing rates of N deposition may provide further insights into soil C emission dynamics in subtropical forests.  Table S1. Relative carbon (C) distribution (%) in different chemical shift regions in 13 C cross-polarization magic-angle spinning of soil under different nitrogen addition treatments, Table S2. The exponential relationship between soil CO 2 emission rate and soil temperature in each subplots under different nitrogen addition treatments, Table S3 Effects of N addition on the phospholipid fatty acid biomarker concentration (nmol g −1 soil). Error bars represent standard deviation (n = 4). The different letters indicate significant differences between treatments at p < 0.05. G + : gram-positive bacteria; G − : gram-negative bacteria; Bacteria: sum of G + and G − ; Unclassified: unclassified biomarkers; ACT: actinomycetes; Total: total PLFA, Table S4 Characteristics