ZHANG Zhi-jian , WANG Xian-zhe LIANG Lu-yi HUANG En, TAO Xing-hua

1 College of Environmental and Resource Science, Zhejiang University, Hangzhou 310058, P.R.China

2 China Academy of West Region Development, Zhejiang University, Hangzhou 310058, P.R.China

3 Hangzhou Gusheng Agricultural Technology Company Limited, Hangzhou 311108, P.R.China

Abstract The structural complexity of soil dissolved organic matter (DOM) may reflect soil biogeochemical processes due to its spectral characteristics. However, the features of DOM structural complexity in paddy soil amended with long-term chemical P fertilization are still unclear, which may limit understanding of nutrient-related soil C cycle. We collected soil samples from field experiments receiving application of 0, 30, 60, and 90 kg P ha-1 yr-1 to assess the effect of exogenous P on the complexity of soil DOM structure. Three-dimensional excitation-emission matrix fluorescence analysis and enzymatic activity assay were used to determine the features of soil DOM molecular structure and the associated microbial reactions.The results showed that P input increased the biodegradability of DOM, indicating by the increased lower molecular weight components and decreased humic degree in the DOM. P input also reduced the structural complexity of DOM with blue shifts of fluorescent signals. The fluorescence index and β/α index of DOM increased with increasing P application by 4-5% and 3-11%, respectively, while humification index decreased by 8-13%. The P input increased the abundance of bacteria and fungi by 34-167% and 159-964%, respectively, while 29-54% increments were found for the β-1,4-glucosidase activities. These results implicated that P fertilization accelerated the soil DOM cycle, although the structural complexity of DOM declined, which potentially benefits soil C sequestration in paddy fields and may be a C sequestration mechanism in the P-dependent paddy.

Keywords: soil carbon, fertilization, fluorescence, humification, agroecosystem

1. Introduction

Soil carbon (C) is critical for regulating global climate changes (Bailey et al. 2018). Dissolved organic matter(DOM) is usually the key component of the exchange of soil C pool and atmosphere CO2in the biosphere (Gao et al. 2018). DOM defines a compound of atypical size with molecular weights ranging from simple carbohydrates to highly complex molecules with different aromatics and humus (Li et al. 2018). The dynamics of soil DOM can be changed by land usage, vegetation cover, and agricultural activities (Avery et al. 2003; Wilson and Xenopoulos 2009;Singh et al. 2014a; Gao et al. 2018). Meanwhile, phosphorus(P) fertilizer input is an indispensable agricultural practice to increase crop yield, but excessive P export from agricultural field contributes to freshwater eutrophication (Sharpley 2016). Thus, it is vital to illuminate the correlation between soil DOM dynamics and P fertilization, in the context of addressing climate change and improving sustainability of agro-ecosystem.

The storage of DOM in soil and/or water is defined not only by the source of C input (Fellman et al. 2008; Gao et al. 2018), but also by the structural complexity of DOM itself (Li et al. 2018). Carbohydrates generated from primary productivity have a higher proportion of hexose and deoxysugars than pentose sugars and were preferentially respired by soil microbes (Wilson and Xenopoulos 2009;Li et al. 2018). The portion of aromatic compounds mostly driven from lignin are relatively stable and difficult to be utilized by microbes (Song et al. 2012). Products of microbial metabolism constitute a significant proportion of DOM and act as an important energy source (Fellman et al.2008). These compounds could strongly influence the global C movement by controlling intensities of C uptake, retention,and release (Battin et al. 2008). The structural complexity of soil DOM, which was mirrored by spectral and fluorescence indexes (Wilson and Xenopoulos 2009; Wang et al. 2015;Gao et al. 2017), could indicate changes in soil DOM stability in the environment (Fellman et al. 2008). The indices of fluorescence index (FI), β/α index or freshness index (β/α),and humification index (HI) can reliably refer to degree of structural conjugation, autochthonous C inputs from microbial-originated residues, and aromaticity ratio as to the C humicity (Olefeldt et al. 2013; Wang et al. 2015). These parameters may indicate differences in the composition of soil DOM in response to environmental changes (Singh et al.2014b; Wang et al. 2015; Gao et al. 2017).

Soil bacteria are the main degraders of lignin and its derived compounds (Lipson and Schmidt 2004), while fungi generally contribute to the incomplete degradation of organic matter (Datta et al. 2017). Therefore, in addition to understanding DOM dynamics itself, it is necessary to further reveal the impact of microbial community changes on the complexity of soil DOM structure. Similarly, organic matter depolymerization by enzymatic catalytic reaction may contribute to low molecular weight compounds in soil, which are easily used as sources of C and nutrients for microbial assimilation (Tian et al. 2010). Thus, enzymatic reactions may represent both processes of soil DOM consumption and production, and alter structural complexity of soil DOM(Liu et al. 2014). The widely assayed eco-enzymes, β-1,4-glucosidase (BG) and acid phosphatase (AP), degrade plant and microbial biomass into soluble compounds for microbe assimilation and investigate the rate limitations of enzymatic catalysis in term of soil C storage (Tian et al. 2010; Sinsabaugh and Follstad 2012). Thus, these enzymes may be sensitive to the effects of P input on soil microbe community and act as a mechanistic indicator for resource acquisition. Till now, the relationship between extracellular enzymes activities and soil DOM features is poorly understood (Wilson and Xenopoulos 2009; Zhang et al. 2015, 2019; Li et al. 2018). Furthermore, using soil DOM as the source of microbial energy and biomass growth is controlled by the ratio of carbon to nitrogen (C/N) (Zhang et al. 2015), crop systems (Gao et al. 2017), limiting nutrients(Sinsabaugh et al. 2008), and DOM composition (Liu et al.2014) in an agricultural environment. P is the key element that controls these properties (Liu et al. 2012). Moreover,P can regulate and control soil C dynamics by microbial acquisition and storage (Bradford et al. 2008) according to ecological stoichiometry (Sinsabaugh and Follstad 2012). The above reported investigations are helpful for researchers and farmers to understand the basic regularity on soil P input to crop yield, soil nutrition and soil carbon balance. However, few studies to date have integrated the biochemical and microbial data to evaluate the effect of P application on the complexity of soil DOM structure and associated soil C cycle.

The total area of Chinese paddy fields is approximately 28.4×107ha, contributing to approximately 30% of the world's total rice yield (Cheng et al. 2010). To date, few studies have probed the chemical P application on the structural complexity of soil DOM in paddy ecosystems. Since 2005, we have carried out the long-term field experiment of superphosphate application in a paddy field located in southeastern China to study the soil P biogeochemical cycle. We hypothesized that P application could reshape the structural complexity of DOM, preferentially through the enhancement of enzymatic catalysis and microbial reactions. Filling the gap between the structural complexity of soil DOM and P-C biogeochemical features can assist in understanding the role of long-term P input in large-scale soil C cycle in paddy ecosystems.

2. Materials and methods

2.1. Study location and field experimental design

A P fertilization experiment in the paddy fields was established in April 2005 at the Demonstration Park of Agricultural Research Station in Yuhang County (30°18′51.84′′N,119°54′13.37′′E) in Zhejiang, in the Yangtze River delta in southeastern China. The experimental area has a subtropical monsoon climate with an average annual rainfall of 1 450 mm and an average temperature of 17.8°C. Blue-purple paddy soil (Mollic endoaquoll) is the dominant soil type in this region. The 150-mm-top soil (before experiment operation) was composed of 3% sand, 47% silt, and 50%clay, while levels of soil total P, total C and total N were 13.7, 2 087, and 148 mmol kg-1, respectively. In this region,local farmers regularly apply 20-50 kg P ha-1yr-1, either as chemical P fertilizer or mixture of chemical P and animal manure, to support one season of rice and an over-wintering crop, such as rape (Brassica napus) or wheat (Triticum aestivum).

The design and construction of the field experimental plots, including layouts of plot ridges, trenches, berms,and inlets/outlets, was previously reported by Zhang et al.(2007). In brief, twelve 4 m×5 m field plots were set in two parallel rows. In order to make these plots hydrologically isolated, a polyethylene-made impermeable membrane of 0.75 mm thickness×105 cm depth was first inserted among every neighboring plot. Then, 12 cm (width)×105 cm (depth)concrete-brick walls were closely attached to sides of every membrane. The experiment was operated according to a completely randomized block design, with three replicates for each treatment. P was applied annually for 10 years at the early summer by input of superphosphate at rates of 0,30, 60, and 90 (P-0 to P-90, respectively) kg P ha-1(6.2%of available P content) to the experimental plots, where rice was cultivated as single rice. P fertilizer was broadcast and incorporated into soil to a depth of 100 mm, and the intensities of P input covered the range from the routine rate of P application of local farmers to the excessive P rates for field experiments. Each plot also received the same input rates of 170 kg N ha-1(urea) and 50 kg K ha-1(KCl) during the experimental period. On the day after P fertilization, rice seedlings (Oryza sativa L., 25-day-old) were transplanted at spacing of 150 mm×150 mm in June, and mature rice was harvested in November, 2015. Our previous investigation(Zhang et al. 2015) shown that rice yield was significantly increased from 6 100 kg ha-1(P-0) to 9 500 kg ha-1(P-60),afterwards down to 9 145 kg ha-1(P-90).

2.2. Soil sampling

To investigate the structural complexity of DOM, topsoil samples (0-15 cm) were collected on 21 May, 2015 (the day before rice transplantation), and soil samples were collected again on November 12 before rice harvest. Six topsoil cores, 10 cm deep cores×2 cm diameter, were collected from each plot firstly. After full mixing, a uniform sample was taken for each plot. Samples were stored in a 25-L portable freeze container during 2 h shipping. Air-dried soils were screened passing 2-mm sieves for lab assay of soil chemical properties. The fresh soil samples were stored at -80°C for the real-time quantitative PCR (qPCR) assay and at 4°C before microbial C, N, P and enzyme activities analyses.

2.3. Laboratory measurements for soil samples

Soil chemical propertiesConcentrations of total P (TP),total organic C (TOC), and available P (Olsen-P) were tested according to standard methods (Bolt 1956; Jones 2001).According to the decomposition of organic C to KMnO4oxidation, the levels of three fractions of labile organic components in soil samples, labile organic carbon (LOC),mid-labile organic carbon (MLOC), and namely highly labile organic carbon (HLOC) were determined using 333, 167 and 33 mmol L-1KMnO4, respectively (Loginow et al. 1987).

Soil DOM was extracted by making slurry with 5-fold volumes of Milli-Q water (Chapin et al. 2000). After acidifying(10% HCl) and purging with inert gas to remove any inorganic C, the DOM concentration was then determined using a TOC analyzer (Shimadzu Scientific Instruments,Columbia, USA). Based on the reported analytic procedure(Nishijima and Speitel 2004; Wilson and Xenopoulos 2009), DOM spectral characteristics, i.e., FI, β/α, HI, were measured using a three-dimensional excitation-emission matrix (3D-EEM) fluorescence spectrophotometer (Hitachi Inc., F-4500, Japan). The specific ultraviolet absorbance at 280 nm (SUVA280) and at 254 nm (SUVA254), ratio of light absorbance at 250 and 365 nm (E250/365), and at 253 and 203 nm (E253/203) were measured using an ultraviolet visible scanning spectrophotometer (Shimadzu Inc., UV-2550, Japan). The indices of SUVA254and SUVA280were positively related to the molecular weight and aromaticity of DOM, and the E253/203value was reported to reflect the degree of substitutability of aromatic rings in DOM and the types of substituents, while lower E250/365ratio reflects higher average molecular weight and aromaticity (Fellman et al.2008; Wilson and Xenopoulos 2009; Valencia et al. 2013;Wang et al. 2015).

Microbial biomassFresh soil samples (stored at 4°C less than 7 days) were used for microbial biomass (C and P) measurements, according to the chloroform fumigationextraction method (Wu et al. 1990). In brief, soil samples were split into two subsamples: 1) one was immediately extracted with 0.5 mol L-1K2SO4for microbial C and 0.5 mol L-1NaHCO3for microbial P; 2) the other was fumigated with chloroform and then extracted. Following centrifugation,concentrations of biomass and microbial biomass elemental content (C and P) were calculated from the difference between the non-fumigated and fumigated soil subsamples.

Enzyme activity assayActivities of extracellular enzymatic reaction were measured under saturated substrate concentrations to represent in situ enzymatic processes with no substrate constraint (Wallenstein and Weintraub 2008).Fluorescence-based methods were used for measurements of BG and AP activities by using methylumbelliferyl-linked artificial substrates(Saiya-Cork et al. 2002; Sinsabaugh et al. 2008), as indicators or the microbial nutrient demand in the C and P cycles, respectively. Enzyme activities were calculated and expressed as nmol h-1g-1. After thawing, soil DNA was extracted from 0.2 g of fresh soil, according to the guideline of Ezup Soil Genomic DNA Extraction Kit (Sangon Biotech, Shanghai). To determine the relative abundances of bacteria, fungi, and archaea communities in each soil sample,the previously described qPCR-based protocol was used (Cytryn et al. 2000;Fierer et al. 2005). Briefly, individual quantification of bacteria, archaea, and fungi was generated, according to a plasmid standard containing a specific target region of the ribosomal gene from E. coli, Fusarium, and Halobacteriales,respectively. Primer sets for the bacteria (Eub338, Eub518), fungi (ITS1f, 5.8s)and archaea (ARCH21f, ARCH958r) were described previously (Cytryn et al.2000; Fierer et al. 2005). Each 20 μL qPCR reaction was operated by the Bio-Rad CFX96 System (Bio-Rad, Munich, Germany). All of qPCR reactions for each DNA sample were simultaneously run in quadruplicate. Because of the different amplification efficiencies of some taxa and potentially measurement deviations of nucleic acid concentrations for each qPCR reaction, we determined our results to the relative ratios of gene copies. The raw sequences have been uploaded into the DDBJ Sequence Read Archive (DRA) database with the accession number of DRA001286.

2.4. Statistical analysis

The experiment was operated in a completely randomized block design in order to test the significant differences in soil chemical composition, microbiological properties, and ecoenzymes among the P fertilization treatments (P-0, P-30,P-60 and P-90). Two-way ANOVA at P＜0.05 level was used to test these repeated measure analysis of variances. Soil organic C fractions, microbial biomass, and enzyme activities were examined with rates of P applications and sampling time as factors by two-way ANOVA. A paired Student's t-test was further performed to examine significant changes among the P fertilization treatments at a P-value of 0.05. Pearson correlation statistics and univariate regression analyses were performed using SPSS Statistical Software (version 17.0) for correlation matrix analysis. Before analysis, both soil Olsen-P and soil DOM variables were checked for normality and were log10transformed. Origin 8.0 Software (Origin Lab Corporation, USA) was also adopted for data mapping.The correlation matrix circles were computed using R Software (version 2.14.0).

Ta tio n1)licappP armg-tefter lon5 a01er 2mbveNonday ad in M P-902 a 0.0 57±1.83 2 a0.028±0.76 0 c0.027±0.36 3 b0.006±0.03 5 b0.009±0.04 c8 b0.003±0.08 0 a0.439±8.33 ermbve P-600 b 0.0 14±1.73 1 a0.023±0.76 9 c0.038±0.36 1 c0.002±0.02 4 c0.003±0.03 6 c0.005±0.07 8 a0.469±8.21 No P-30 c2 b0.015±1.70 8 b0.007±0.69 1 b0.010±0.45 b1 a0.008±0.03 1 b0.004±0.05 b5 a0.002±0.09 7 a0.762±8.15 P-0 2 c0.024±1.66 3 c0.015±0.63 5 a0.024±0.52 1 a0.000±0.04 7 a0.005±0.06 2 a0.005±0.09 2 b0.162±7.11 td aannm03d 2an53t 2nce absorbaht af ligtio ora5, thectively.0/3 6E25, respend 3 aP-90nd ld so ils co lle ctedy fie 3/20 P-90 bcb 3 a ctively; E25 0.01.0 a V ANOy Ae-waonnd's test ancanDusings uceereniff) d 7 a 0.05 3±0.005 c 1.538±0.011 a 0.667±0.008 b 0.517±0.012 a 0.061±0.012 a 0.101±0.09 0.151±6.85-60, P , respe a dd pM) in tester (DOattenic morgalved yMa P-608 b 0.0 01±1.51 8 a0.012±0.65 7 c0.002±0.51 3 b0.006±0.06 5 c0.004±0.09 9 c0.006±0.12 6 a0.007±6.84 P-308 b 0.0 08±1.50 6 b0.013±0.62 3 b0.013±0.54 b4 a0.008±0.06 1 b0.013±0.12 c6 b0.002±0.13 1 a0.132±6.63-3nm, P54-0as P d 2aned 8 0am t 2 1 yr-1, nnce aa-P h b so rb at akg le90 iond0 a u ltra v0, 6 e cific, 3 sps 0 with isso 0, the t (P＜f d nifican eters o n t sigse aram rers rept lette P-0 ut a e re niffn. Dviatioderdda 4 c0.01 3 b stan 1 a 4±3 a 0.01 0.01 4±1.45 0.602±0.006 a 2±0.013 a 0.07 0.599±0.018 b s inp VA28 0.166±0.33waSU 2±tionnd 0 a alue 5.50 0.16lica A28ea tra l pecSpble 1 r2)temeraPa xdeinceenscreFluo β/α e xn indmificatioHu 0VA28SU 4VA25SU 3/20 3 E250/36 5 E25 p pate a U Vx; S ctively.n v phde os in erphss s mm, respe d a ne5 n upsh 1) Field s e xp re ss e36 re 2) β/α, frend 25e d 0 aata a Th

3. Results

3.1. The effect of phosphorus fertilization on structural complexity of soil DOM

The FI and β/α value of soil DOM in May increased with increasing P application by 4-5% and 3-11%, respectively (Table 1). However, the HI in May following P treatment significantly (P＜0.01) decreased by a factor of 8-13% (Table 1).Meanwhile, SUVA280and SUVA254decreased by 6-9%, and 27-45%,respectively. The E253/203following P treatment significantly (P＜0.05) decreased by 9-24% compared to the P-0 treatment. In addition, a significant (P＜0.05)increase of 21-25% were observed for E250/365(Table 1). These DOM structural characteristics also showed similar trends in the November samples (Table 1).

Rates of P treatment and date of sampling had significant(r2=0.529;P＜0.001) impacts on soil DOM spectral characteristics, according to two-way ANOVA analysis.The interactions of the two variables were also significant(r2=0.335;P＜0.05), except forE250/365(Table 2). The FI,β/α,andE250/365values showed a significant (r2=0.468;P＜0.05)positive relationship with the Olsen-P, while negative relationships (r2=0.261;P＜0.05) were found for SUVA254and HI (Fig. 1). Soil Olsen-P accounted for 18% (FI), 54% (β/α),41% (HI), 16% (SUVA254), and 22% (E250/365) of the variance in the data, respectively. The Pearson correlation showed positive correlations between TP and FI,β/α,E250/365, and BG, and negative correlations with HI and SUVA254(Fig. 2).Significant (r2=0.316;P＜0.05) correlations between DOM spectral characteristics and microbial biomass (i.e., MBC and MBP) were observed, with the exception ofβ/αandE253/203. DOM spectra characteristics, except for SUVA280,were also significantly (r2=0.281;P＜0.05) correlated with BG (Fig. 2). In addition, significant (r2=0.197;P＜0.05)positive relationships were found between FI andβ/αas well as HI and SUVA254. Meanwhile, FI andβ/αwere negatively (r2=0.273;P＜0.05) correlated with HI and SUVA254(Fig. 2). The 3D-EEM images showed that P application caused humic-like peaks and a related microbial peak,represented byαandβ, blue-shifted (Mobedet al.1996) to the wavelengths nearby the lower end of emission (Fig. 3).

Table 2 Results of two-way analysis of variance on soil dissolved organic matter changes responding to P application treatments

3.2. Features of soil carbon, phosphorus, and enzyme activity

Phosphorus input significantly (r2=0.268;P＜0.05) increased soil DOM by 10-35% in May relative to P-0 treatment(Table 3). A similar phenomenon was found for TOC, MBC,and BG, which was 10-19%, 20-62% and 12-29% higher than that in the treatment of P-0, respectively (Table 3).Three parts of labile organic C showed that LOC, MLOC,and HLOC were significantly (r2=0.186;P＜0.05) increased by 15-45%, 23-34%, and 1-19% in May, respectively(Fig. 4). Soil Olsen-P and TP was significantly (r2=0.311;P＜0.05) increased with the increasing P application (in May), except for the Olsen-P in P-30 treatment (Table 3).The P application also significantly (r2=0.235;P＜0.05)increased the MBP by 35-89%, while BG activities were correspondingly increased by 29-54%. However, the opposite trend occurred for AP, which had reductions of 7-37% in May. In addition, a similar trend was observed for these C- and P-related properties in November, which were slightly higher than that in May (Table 3).

3.3. The effect of phosphorus fertilization on microbial abundance

Phosphorus treatment significantly increased the relative abundance of soil bacteria by 40-167% (May) and 34-118%(November) relative to P-0 treatment (Table 4). Similarly,significant increases of 2.58-9.64 times and 1.59-4.35 times were observed in the fungal abundance for these two seasonal samplings, respectively. Moreover, P treatment increased the ratio of fungi to bacteria by 97-206% (May)and 91-147% (November) compared to P-0 treatment(Fig. 5). However, P treatment decreased the abundance of soil archaea by 23-61% (May) and 16-69% (November)(Table 4), thus reducing the corresponding ratios of archaea to bacteria by 47-84% and 38-84%, respectively (Fig. 5).

4. Discussion

4.1. Phosphorus fertilization decreased in structural complexity of soil DOM

Fig. 1 Relationship between spectral parameters of soil dissolved organic matter (DOM) and Olsen-P using selected univariate linear regressions with the highest r2 shown in each panel. β/α, freshness index; SUVA280 and SUVA280, the specific ultraviolet absorbance at 280 and 254 nm, respectively; E250/365, the ratio of light absorbance at 250 and 365 nm, respectively.

The spectroscopic characteristics of DOM were highly responsive to P application and sampling time in the paddy ecosystem (Tables 1 and 2; Figs. 1 and 3). FI serves as a simple index to characterize the features of the DOM skeleton, particularly the ratio of aromaticity and degree of structural conjugation (Wilson and Xenopoulos 2009; Liuet al.2014), whileβ/αmay indicate the relative contribution of recently microbially-produced DOM (Wilson and Xenopoulos 2009; Wanget al.2015). We found that the FI andβ/αvalues increased significantly (P＜0.05) with P input (Table 1),thus representing a more microbially-derived DOM and a nutrient-related increase in the portion of autochthonous DOM. This indicates that P inputs induce more labile C sources for microbial acquisition (Fig. 1), which is basically consistent with a similar watershed-scale and lab-scale investigation (Wilson and Xenopoulos 2009; Liuet al.2014).This was most likely caused by changes in the composition of microbial community as a result of the greater P availability(Battinet al.2008; Gaoet al.2018). Moreover, the increased fractions of labile organic C contents (i.e., LOC, MLOC,and HLOC) (Fig. 4) also verified that P fertilization might accelerate the generation of labile C sources and contribute to the higher input of microbially-derived DOM. Humic degree indicators (i.e., HI, SUVA280, SUVA254, andE250/365)correlate with the degree of soil humic degree and could reflect aromaticity as well as the molecular complexity of DOM components (Fellmanet al.2008; Valenciaet al.2013; Wanget al.2015). In our study, the decrease in HI,SUVA280, SUVA254, andE250/365humic degree indicators after P application (Table 1) indicate a breakdown of aromatic nuclei and decomposition of humic constituents (Fellmanet al.2008; Wilson and Xenopoulos 2009; Valenciaet al.2013). Moreover, the lower UV indices, especiallyE253/203and SUVA254, indicate a decline in unsubstituted aromatic ring structures as well as carbonyl, carboxylic, ester, and hydroxyl functional groups (Hyun-Chul and Yu 2007). Therefore, the organic matter is predominantly composed of hydrophobic components with low molecular weights after the addition of P to paddy fields (Table 1; Fig. 1), correspondingly increasing the DOM biodegradability. Since the molecular structure of DOM determines the decomposition rates of soil humic substances, the reduced levels of DOM humic degree and aromaticity have lower chemical stability (Schmidtet al.2011; Liet al.2018). The significant negative (r2=0.41;P＜0.001) linear relationships between these humic degree indicators and Olsen-P (Fig. 1) indicate that P application caused the simple structural complexity of DOM, which contributed to the enhanced biodegradability of DOM(Fellmanet al.2008; Wanget al.2015).

Phosphorus application caused a blue shift in DOM structure (Fig. 3). This type of shift toward the left side of the excitation wavelength typically reflects lower molecular weight fractions and a lower degree of aromaticity as well as humic degree of DOM structural complexity (Valenciaet al.2013). Field application of animal manure caused soil DOM with relatively high portion of low aromaticity and strengthened a microbial/tryptophan-like character (Singhet al.2014a; Gaoet al.2018), increasing bioavailability for microbial growth (Germanet al.2011) and biodegradation of DOM (Kothawalaet al.2012). As revealed by the 3D-EEM,the blue shift feature (Fig. 3) verified that P application promoted a breakdown in humic carbon into lower molecular weight compounds, which illustrated the decreased structural complexity of DOM and indicated the enhanced biodegradability and bioavailability of DOM (Valenciaet al.2013; Liuet al.2014; Liet al.2018). Combined with the DOM spectroscopy characteristics (Table 1; Fig. 1), we confirmed our hypothesis that P fertilization caused significant shifts in soil DOM dynamics and induced a decrease in the structural complexity of soil DOM.

Fig. 2 Pearson correlation coefficients between soil biochemical properties and dissolved organic matter (DOM) spectral parameters responding to four varied phosphorus applications in the tested paddy field. TOC, soil total organic carbon;DOM, dissolved organic matter; TP, total phosphorus; MBC,microbial biomass carbon; MBP, microbial biomass phosphorus;BG, β-1,4-glucosidase; AP, acid phosphatase; SUVA280 and SUVA254, the specific ultraviolet absorbance at 280 and 254 nm,respectively; E253/203 and E250/365, the ratio of light absorbance at 253 and 203 nm and at 250 and 365 nm, respectively;β/α, freshness index; HI, humification index; FI, fluorescence index. The Pearson correlation coefficients are represented by correlation matrix circles. Red and blue colours represent positive and negative correlations, respectively, while size saturation is proportional to the magnitude of the correlation.Asterisks in the circles represent significant correlations(＊, P＜0.05).

4.2. Phosphorus fertilization accelerated soil carbon cycling

As discussed above, P fertilization induced a trend toward a simpler structural DOM complexity (Fig. 3). These low molecular compounds among DOM composition, such as carbohydrates, are the major sources of substrates for microbial growth (Lipson and Schmidt 2004) and may also comprise the predominant sites of microbial origin (Cooperet al.2016). Therefore, the increase in soil DOM content(Table 3) may partially be due to high microbial growth,since P fertilization increased the production of microbialderived DOM sources (Table 1). However, in theory, DOM concentrations should be low under nutrient-saturated conditions due to stimulated microbial activity (Cooperet al.2016; Gaoet al.2017, 2018), and nutrient-induced suppression of ligninase could result in an accumulation of moderately degraded compounds (Sinsabaughet al.2008;Cooperet al.2016; Dattaet al.2017). However, we had not found consistent evidence showing soil DOM decreased with nutrient availability. Fungi were the main degraders of lignin and lignin-derived compounds, and increasing abundance of bacteria may accelerate the degradation of recalcitrant C fraction (Lipson and Schmidt 2004; Bradfordet al.2005;Fiereret al.2005). In our study, the higher relative abundance of fungi subjected to P fertilization (Table 4) suggest a mechanism for the decline in DOM structural complexity(Fig. 3) by fungi-derived DOM (Table 1). The increased abundance of fungi and the ratio of fungi to bacteria following P input (Table 4; Fig. 5) also suggests that P fertilization contributed to DOM production, due to the incomplete degradation of organic matter by fungi (Dattaet al.2017).Other studies have shown that P input can accelerate the growth of bacteria and fungi and alter their community structure (Hanet al.2012), thus leading to preferential degradation of cellulose and phenolic compounds (De Graaffet al.2010) and a greater C pool (Bradfordet al.2008).The increased MBC and MBP confirmed the increase in bioavailable DOM for microbial utilization of C (Table 3),which was a result of the P fertilization-induced increase in biodegradability of DOM (Fig. 3). Our previous study shown that P application was apt to enhance relative abundances of C-decomposing bacteria (such asPropionibacteriales,Acidimicrobiales,Cytophagales, andGemmatimonadales),promoting breakdown in the plant-derived production of carbon sources for both microbial growth and soil C storage(Zhanget al.2015). As indicated by higher soil MBC and TOC in Table 3, changes of microbial structure and the corresponding C-related metabolisms may contribute to the dynamic change of DOM in soil (Liet al.2018). Thus, DOM consumption in soil was accelerated, and DOM molecules with relatively simple structure were preferentially selected and left behind.

Fig. 3 Examples of three-dimensional excitation-emission matrix (3D-EEM) for soil dissolved organic matter (DOM) under phosphorus application treatments. The solid lines denote the emission wavelengths used in the calculation of β/α and the fluorescence index and are placed at emission wavelengths of 380, 450, and 500 nm; the dashed line represents the maximum emission wavelength for the α component. Field superphosphate application was input as 0, 30, 60 and 90 kg P ha-1 yr-1, named as P-0 (A), P-30 (B),P-60 (C) and P-90 (D), respectively.

Table 3 Properties of soil chemical, microbial biomass, and enzyme activity in the experimental paddy field after long-term P application in 20151)

Fig. 4 Effects of varied phosphorus application on different soil labile organic carbons in May and November, 2015. HLOC, highly labile organic carbon; MLOC, mid-labile organic carbon; LOC, labile organic carbon. Field superphosphate application was input as 0, 30, 60 and 90 kg P ha-1 yr-1, named as P-0, P-30, P-60 and P-90, respectively. Error bars are the standard deviations of the mean. Different letters listed above bars represent significant differences at P＜0.05 (Duncan's test).

Table 4 qPCR analysis for microbial abundance (copies g-1 dry soil), including bacteria, fungi, and archaea under P application treatments (P-0, P-30, P-60, and P-90) observed in May and November, 20151)

Soil biochemical properties were also a function of DOM spectral characteristics, and the strong and positive relationships between BG and FI,β/α, orE250/365(Fig. 2)suggest that the BG was also a function of microbe-derived DOM. Apparently, our findings are consistent to the fact that DOM is the end product of enzymatic reactions of compounds with varied molecular weights (Cooperet al.2016; Liet al.2018). The significantly positive relationship between BG and DOM verified their positive relationship(Fig. 2), as DOM acts as the substrate for synthesis of BG (Songet al.2012). Moreover, because BG activity contributed to the degradation of cellulose and hydrolysis of cellobiase to glucose for microbial assimilation (Sinsabaughet al.2008; Peoples and Koide 2012), the application of P in paddy fields may stimulate a greater microbial C demand.Such enzymatic catalysis might have led to an increased degradation of DOM as a result of the increased BG activity(Table 3), and reduced the structural complexity of DOM.The humic degree indicators HI, SUVA280, SUVA254, andE253/203showed significantly negative relationships with BG (Fig. 2), suggesting that BG contributed to the DOM loss. This may be because DOM was metabolized at a faster rate due to increased activities of C-related enzymes(Table 3), which stimulated the degradation processes of macromolecule substances (Peoples and Koide 2012; Zhao et al. 2019). Shifts in microbial community composition(Table 4) due to differences in carbon availability selectively influenced the organic matter decomposition (Leff et al.2012), partially by the regulation of enzyme-catalyzed depolymerization (Bradford et al. 2008). These results indicate that consumption of soil DOM was enhanced through enzymatic catalysis when the soil was subjected to P fertilization.

Fig. 5 Relative abundances of soil bacterial, fungal, and archaeal genes estimated using qPCR assays under four P application treatments in the tested paddy field. Field superphosphate application were input as 0, 30, 60 and 90 kg P ha-1 yr-1, named as P-0, P-30, P-60 and P-90, respectively.Error bars are the standard deviations of the mean. Different letters listed above bars represent significant differences at P＜0.05 (Duncan's test).

By combining the increased trends of both soil DOM production and decomposition, we conclude that P fertilization accelerated DOM cycling by microbial reactions and enzymatic catalysis. Although the DOM cycling was accelerated, the DOM content significantly (P＜0.05)increased (Table 3), indicative of fertilization contributes to a net DOM sink. Because of the increased levels of TOC and labial C with DOM content (Table 3; Fig. 4), P fertilization may increase soil C sequestration in paddy field ecosystems, beneficial for offsetting the impacts of global warming. However, soil CO2emission could increase as soil organic C increases (Zhang et al. 2019); and C sequestration by soil microorganisms is generally controlled by balances of biogenetic elemental acquisition (C, N and P) according to ecological stoichiometry theory (Sinsabaugh and Follstad 2012; Zhang et al. 2015). Thus, further studies are needed to investigate P biogeochemical effects on soil C sequestration based on the ecological stiochiometry for soil microorganisms. In addition, others have reported that growth of specific bacterial microorganisms, such as Bacteroidetes and Gemmatimonadetes, may be accelerated in the decomposition of recalcitrant C compounds (Lipson and Schmidt 2004; Li et al. 2018). Therefore, a further investigation on the functional attributes of microbial communities in both functional and taxonomic levels is required to quantify soil C cycling subject to P fertilization in paddy field ecosystems.

5. Conclusion

The structural complexity of soil DOM significantly decreased in the paddy field after continuous P fertilization.The blue shift, coincident with decreased humic degree indicators, provided strong evidences that P enhanced the biodegradability of DOM by decreasing the structural complexity of DOM. Because this is beneficial for microbial growth, P fertilization induced microbially-derived DOM and autochthonous DOM production, resulting in enhanced generation of soil DOM and TOC contents. Therefore,although the structural complexity of DOM declined, P fertilization contributed to the acceleration of DOM cycling and subsequently enhanced soil C sequestration through a synthesis coordinating mechanism of microorganism and enzyme activities in paddy fields. This work provides a broad picture of soil C cycle response to P input into paddy fields, which has implications for climate change prediction.

Acknowledgements

The authors wish to thank the National Natural Science Foundation of China (41373073 and 41673081) and the Key Research and Development Projects in Zhejiang Province,China (2015C03SA420001). We also express our thanks to Dr. Jeff Muehlbauer (University of California, Davis, USA)for his professional editing on English writing.