Effects of nitrogen and phosphorus additions on soil microbial biomass and community structure in two reforested tropical forests

Elevated nitrogen (N) deposition may aggravate phosphorus (P) deficiency in forests in the warm humid regions of China. To our knowledge, the interactive effects of long-term N deposition and P availability on soil microorganisms in tropical replanted forests remain unclear. We conducted an N and P manipulation experiment with four treatments: control, N addition (15 g N m−2·yr−1), P addition (15 g P m−2·yr−1), and N and P addition (15 + 15 g N and P m−2·yr−1, respectively) in disturbed (planted pine forest with recent harvests of understory vegetation and litter) and rehabilitated (planted with pine, but mixed with broadleaf returning by natural succession) forests in southern China. Nitrogen addition did not significantly affect soil microbial biomass, but significantly decreased the abundance of gram-negative bacteria PLFAs in both forest types. Microbial biomass increased significantly after P addition in the disturbed forest but not in the rehabilitated forest. No interactions between N and P additions on soil microorganisms were observed in either forest type. Our results suggest that microbial growth in replanted forests of southern China may be limited by P rather than by N, and this P limitation may be greater in disturbed forests.


Results
Soil properties. There was no significant difference in the chemical parameters for soil between the disturbed and rehabilitated forests, except for SMC and soil NO 3 --N concentrations (Table 1). Nitrogen addition did not change soil properties, except for NO 3 --N and NH 4 + -N concentrations. Soil NO 3 --N concentrations were also altered in the P-addition plots. Soil available P was significantly elevated in the P-amended plots in both the disturbed and rehabilitated forests (Table 1). Soil net mineralization rate and nitrification rate were different between forest types, and were lower in the disturbed than the rehabilitated forest. Statistical analysis from a three-way analysis variance (ANOVA) showed that there were no significant interactions between N and P additions on the chemical parameters of soil (Table 1).

Soil microbial biomass.
There was no significant effect on soil microbial biomass after N addition in either of the forest types. Response of soil microbial biomass to P addition, however, varied depending on forest type (Figs 1 and 2). P addition significantly increased microbial biomass, including total PLFAs, bacterial PLFAs, and fungal PLFAs in the disturbed forest, but not in the rehabilitated forest. Soil microbial biomass did not respond to NP addition in either forest, except for fungal PLFAs, which was significantly increased after NP addition in the disturbed forest (Fig. 1). Although there was no significant treatment responses in the rehabilitated forest (Fig. 2), the trend in the responses were identical to that of the disturbed forest (Fig. 1).
Scientific RepoRts | 5:14378 | DOi: 10.1038/srep14378 Soil microbial community. The mean abundance of gram-negative bacteria PLFAs were significantly decreased in the N-amended plots in both forests (Fig. 3). The lower abundance of gram-negative bacteria PLFAs in the N-amended plots was mainly reflected in the relative abundance of the individual PLFAs 16:1w7c and 18:1w7c (Fig. 4). Moreover, the ratio of cyclopropyl fatty acid to their precursors (cy17:0/16:1w7c and cy19:0/18:1w7c) was also significantly affected by N addition in both forests (Table  S1, S2). Conversely, the mean abundance of actinomycetes PLFAs in the P-amended plots was significantly increased in the disturbed forest, but not in the rehabilitated forest (Fig. 3). There were no interactions between N and P addition on soil microbial community structure, except for a few individual PLFAs (Appendix S1, S2).
In addition, the mean abundance of gram-positive PLFAs was significantly higher in the P-amended plots than in the N-amended plots in the rehabilitated forest, whereas the mean abundances of arbuscular mycorrhizal (AM) PLFAs was significantly higher in the P-addition plots compared with the N-addition plots in the disturbed forest (Fig. 3).
The first two axes produced by principal components analysis (PCA) accounted for 47.1% of the total variation in the PLFA profile; the forest types were separated along the PC1 and PC2 axes (  Table 1. Effects of forest, N addition, P addition, and two-way interactions of N and P addition on soil properties (0-10 cm depth), n = 5. N × P: interactions between N addition and P addition; SMC: soil moisture content; SOC: soil organic carbon; Avail P: available P; NMR: net mineralization rate; NNR: net nitrification rate. Significant P-values (P < 0.05) shown in bold face type.  Table 2). When comparing the effect of soil chemical properties on soil microbial community composition, soil pH and inorganic N had no effect on soil microbial community composition. Soil organic carbon (SOC) and available P significantly influenced soil community composition (Table 3).     Table 2. Effects of forest, N addition, P addition, and two-way interactions of N and P addition on soil microbial PLFAs (n = 5). N × P: interactions between N addition and P addition; F:B indicates the percent of ratio of fungal to bacterial PLFAs; PC1 and PC2 indicate the first two axes produced by principal components analysis based on the total variation in the PLFA profile.

Discussion
Interactive effects of N and P addition. As interactive effects of combined N and P enrichment have been frequently observed in many terrestrial ecosystems 1 , we expected that simultaneous N and P addition would produce notable interaction effects on soil microbial biomass or community composition in these tropical forests. Conversely, the expected interaction effects were not found in the present study (Appendix S1, S2). Similarly, interaction effects were not encountered in an old-growth forest in the same area 21 . Fanin et al. (2015) also reported that no evidence for N and P interactions on soil microbial structure and functioning in an Amazonian rain forest 28 . Compared to other resources, however, N is relatively abundant at our study forests due to high inorganic N deposition 2 . As a result, N may not be a limiting nutrient for soil microorganisms, and consequently, additional N fertilization may not increase soil microbial biomass or activities.
Effect of N addition. Nitrogen addition generally had a negative effect on microbial biomass in both field and lab-based studies 29 . A number of mechanisms have been proposed for the decline in microbial biomass under N fertilization, such as direct inhibition, indirect effect due to magnesium or calcium deficiency caused by soil acidification, or alteration of C availability 29 . On the other hand, aboveground litter production usually increases and litter quality may improve under N fertilization 30,31 . In this case, soil microbial biomass may increase because of alleviated C and N limitation as shown by results from temperate forests, which are typically N-limited under natural conditions 32,33 . Contrary to our original hypotheses, no significant effect of N addition on soil microbial biomass was found over a long-term (52-month) period in either forest (Figs 1 and 2). Li et al. (2015) also reported that N addition did not significantly affect microbial biomass in a secondary tropical forest of China 34 . The lack of a clear effect of N addition on soil microbial biomass, as mentioned above, is unlikely because of the relatively high N status in the studied forests. Although long-term N addition had no effect on microbial biomass, it had an impact on microbial community composition. Similar results were reported in the study of three Hawaiian forests 35 . We found that N addition altered microbial community structure by significantly decreasing the relative abundance of gram-negative bacteria (e.g. 16:1w7c and 18:1w7c) in both forests. Gilliam et al. (2011) found that the gram-negative PLFA 18:1n7c (gram-negative bacteria) was predominant in soils with the highest rates of net nitrification at Fernow Experimental Forest, a central Appalachian hardwood forest in West Virginia, USA 16 . In the present study, N addition actually decreased the relative abundance of  gram-negative bacteria with lower net nitrification, which may include NH 4 -oxidizing and NO 2 -oxidizing (ammonia-oxidizing and nitrite-oxidizing) gram-negative bacteria. In a parallel investigation in an old growth forest, a similar negative response of gram-negative bacteria to N addition was observed 21 . These results indicated that N is unlikely a limiting factor, but rather a negative factor for soil microorganisms in these forests. We further observed a shift in bacterial communities in response to N fertilization as indicated by an increase of the cy:pre ratio (cy 17:0/16:1w7c and cy19:0/18:1w7c), which is an indicator of physiological stress or C limitation 36,37 . Higher values of these ratios have been associated with decreased bacterial growth rate and increased C limitation 37,38 . In this study, the cy:pre ratios were higher in N-amended plots than in control plots for both forests (Appendix S1, S2). These results indicated that the microbial community in N-amended plots may be in physiological stress, and be C limited in these reforested tropical forests. Fanin et al. (2015) also reported that the responses of microbial community structure to N fertilization were controlled by C availability in an Amazonian rain forest 28 .
Effect of P addition. The observed strong microbial responses to P addition are in line with the results of previous experiments in other tropical lowland forest in France Guiana, Costa Rica, and Panama 28,39,40 . There are several mechanisms that cause P limitation in terrestrial ecosystems. In the present study, the N deposition reaches 30-70 kg N ha −1 ·yr −1 , such high N deposition in this region may increase N availability, but also aggravates P deficiency in these old weathered soils 2 . Therefore, the relieved P constraints could increase soil microbial biomass, and a previous study found that microbial utilization of soil C in these tropical forests was P limited 39 . However, the P limitation of microbial biomass was only found in the disturbed forest and differed from the short-term results, which did not indicate a significant effect of P addition on microbial biomass in either forest 30-month after P addition 11 .
The lag effect of P addition on soil microbial biomass in this study is possibly that soil microorganisms are C limited in the initial years following P addition. The annual total litterfalls increased, and thus alleviated C limitation for soil microorganisms under P fertilization 11 . The lag effect of P addition on soil microbial biomass was also found in a nearby old growth forest (no effect after 18-month, but significant effect after 30-month) 11 The different response of soil microbial biomass to P addition between these two forests may be related to the differences in tree species composition, and soil nutrient status caused by different land-use practices. In the present study, Pinus massoniana was the dominant tree in the disturbed forest (> 80% of basal area) in comparison to the rehabilitated forest (< 40% of basal area) 11 . The concentration of N and P in the pine needles was higher in the rehabilitated forest than in the disturbed forest 26 . Moreover, N addition increased mass loss rate and C release rate, but suppressed the release rate of N and P from decomposing needles in both disturbed and rehabilitated forests 26 . Therefore, there was less N and P released from litter decomposition in the disturbed forest than in the rehabilitated forest. These results indicated that soil microbial growth maybe more P-limited in disturbed forests because of the low P release rate from litter decomposition. In addition, the understory and litter harvesting practice removed 44 to 73% of the nutrients (N, P, K, Ca, and Mg) from the annual production of litter and understory biomass until the late 1990 s 22 . Therefore, P was one of the limiting factors in the disturbed forest rather than in the rehabilitated forest.

Conclusions
In conclusion, we found that there were no interaction effects of N and P addition on soil microbial biomass or community composition. N availability was not a limiting factor for microbial growth in the reforested sites of the studied region and the disturbed forest may be more sensitive to nitrogen addition. However, P availability was one of the limiting factors for microbial growth in the disturbed forest, but not in the rehabilitated forest. These results imply that nutrient limitation on soil microorganisms depend on the changes in tree species composition and soil nutrient status caused by the degree of human disturbance and land-use practices under high N deposition. In addition, our results suggested that there is a lag time in the effect of P addition on soil microorganisms in these tropical forests.

Materials and Methods
Site description. This study was conducted in Dinghushan Biosphere Reserve, which covers an area of 1,200 ha and is located in the central part of Guangdong Province, south China (112°10′ E, 23°10′ N). This Reserve is about 90 km west of metropolitan Guangzhou (10 million inhabitants). In the Reserve, we have identified two types of reforest forests: a mixed pine and broadleaf forest (rehabilitated) and a pine forest (disturbed). The rehabilitated forest, at about 200 m above sea level (asl) occupied approximately 50% of the Reserve, and the disturbed forest, at about 50-200 m asl occupied approximately 20% of the reverse 25 . The remaining 30% of the Reserve is covered by the undisturbed old growth forest where a parallel study was completed by Liu et al. (2013) 21 . Both rehabilitated and disturbed forests originate from the clear-cuts of the 1930 s and subsequent pine plantation establishment. They were under continuous human disturbance (generally the harvesting of understory and litter) from 1930 to 1956 (rehabilitated forest) and 1998 (disturbed forest). In the rehabilitated forest, after cessation of the Scientific RepoRts | 5:14378 | DOi: 10.1038/srep14378 disturbance, colonization from the natural dispersal of regional broadleaf species altered its plant community. Thus, these forests vary both in the level of human impacts, as well as stages of succession, site conditions, and species assemblages 25 . Dominant tree species in the rehabilitated forest are Pinus massoniana, Schima superba Chardn. & Champ., Castanopsis chinensis Hance, Craibiodendron kwangtungense S. Y. Hu, Lindera metcalfiana Allen, and Cryptocarya concinna Hance, whereas, P. massoniana remained the dominant tree in the disturbed forest 11 Further details on the tree structure at the sites can be found in Liu et al. (2012). The selected sites of the two forest types were approximately 4 km from each other.
The Reserve has a monsoon season and humid climate. The average annual precipitation of 1,927 mm has a distinct seasonal pattern with 75% from March through August and only 6% from December through February 25 . The mean annual temperature is 21 °C, with the coldest and warmest months being January (12.6 °C) and July (28.0 °C), respectively. Annual mean relative moisture is 80%. The N deposition measured as inorganic N in throughfall was 24 and 26 kg N ha −1 ·yr −1 in 2004 and 2005 for the rehabilitated and disturbed forests, respectively, with an additional input of 15-20 kg N ha −1 ·yr −1 as dissolved organic N 42 . The soil in the Reserve is lateritic red earth formed from sandstone 25 . The soil depth ranges from 30 to 60 cm (to the top of the C horizon) in the rehabilitated forest and is generally less than 30 cm in the disturbed forest 25 .
Experimental treatment. In 2007, four treatments (each in five replicates) were established in both forests: Control, N-addition (15 g N m −2 ·yr −1 ), P-addition (15 g P m −2 ·yr −1 ), and NP-addition (15 g N m −2 ·yr −1 plus 15 g P m −2 ·yr −1 ). Each of the 20 plots was 5 m × 5 m and surrounded by a 5-m wide buffer strip separating the next plot. Field plots and treatments were laid out randomly. Plots size and fertilizer level were similar to those in the experiment in Costa Rica by Cleveland and Townsend (2006) 43 . Applications of N and P were made as NH 4 NO 3 and NaH 2 PO 4 solutions sprayed in two monthly portions below the canopy with a backpack sprayer starting from January 2007 and continuing through the end of this project (June 2011). Fertilizer r was weighed and mixed with 5 L of water for each plot. Each control plot received 5 L of water without fertilizer.
Field sampling and measurements. Soil sampling of the upper 10 cm was conducted in the warm and wet season, June 2011. From each plot, 5 soil core samples (2.5 cm inner diameter) were collected randomly and combined into one composite sample. The litter layer was carefully removed before sampling. After removing stones and coarse roots, soils were sieved through a 2-mm mesh and divided into two parts, one retained for measuring soil chemical parameters and the other for analysis of microbial biomass and community structure. At the same time, an in situ soil-core technique 44 was used to estimate soil net nitrification rates.
Soil moisture content (SMC) was measured gravimetrically using 10 g of moist soil that was oven dried at 105 °C for 24 h. Soil pH was measured in a 1:2.5 soil/water suspension. Soil organic C (SOC) was determined by dichromate oxidation and titration with ferrous ammonium sulfate. Dissolved organic carbon (DOC) in filtered 0.5 M K 2 SO 4 -extracts of fresh soil was measured with a TOC analyser (TOC-VCPH Shimadzu Corp., Japan). NH 4 + -N and NO 3 --N in filtered 2 M KCl-extracts of fresh soil were measured with a flow injection autoanalyser (FIA, Lachat Instruments, USA). Available P concentration was analysed colorimetrically after acidified ammonium persulfate digestion 45 .