Agricultural land-use change in a Mexican oligotrophic desert depletes ecosystem stability

Soil nutrient and enzymatic analyses

Soil pH was measured in deionized water (1:2 w:v) using a digital pH meter (Corning) and soil electrical conductivity was measured by conductivity meter (Hannan Instruments Inc., Houston, USA). A subsample (100 g) was oven-dried at 75 °C to constant weight for soil moisture determination using the gravimetric method in order to adjust for water content when expressing nutrient concentration on the basis of dry soil mass. All C forms analyzed in all samples were determined in a total carbon analyzer (UIC model CM5012, Chicago, USA), while the N and P forms analyzed were determined colorimetrically in a Bran-Luebbe Auto analyzer 3 (Norderstedt, Germany). Prior to the total soil nutrient analyses, soil samples were dried and ground with a pestle and mortar. Total carbon (TC) and inorganic carbon (IC) were determined by combustion and coulometric detection (Huffman, 1977). Total organic carbon (OC) was calculated as the difference between TC and IC. For total N (TN) and total P (TP) determination, samples were acid digested with H₂SO₄, H₂O₂, K₂SO₄ and CuSO₄ at 360 °C. Soil N was determined by the macro-Kjeldahl method (Bremmer, 1996), while P was determined by the molybdate colorimetric method following ascorbic acid reduction (Murphy & Riley, 1962).

Available, dissolved and microbial nutrient forms were extracted from field moist soil samples. Available inorganic N (NH${}_4^{+}$ and NO${}_3^{-}$) was extracted from 10 g of fresh soil subsamples with 2M KCl, followed by filtration through a Whatman No. 1 paper filter (Robertson et al., 1999) and determined colorimetrically by the phenol-hypochlorite method. Available (inorganic) and labile (organic) P was determined by extraction with 0.5M NaHCO₃ at pH 8.5 according to Hedley sequential P fractionation (Tiessen & Moir, 1993) and quantified as described above for orthophosphate.

Dissolved nutrients were extracted with deionized water after shaking for 45 min and filtering through a Millipore 0. 42 00B5m filter (Jones & Willett, 2006). Prior to acid digestion, one aliquot of the filtrate was used to determine dissolved ammonium (DNH${}_4^{+}$) and inorganic P (IP) in deionized water extract. Total dissolved nitrogen (TDN) was digested using the macro-Kjeldahl method. Total dissolved P (TDP) was also acid digested and determined by colorimetry. Total dissolved carbon (TDC) was measured with an Auto Analyzer of carbon (TOC CM 5012) module for liquids (UIC-COULOMETRICS). Inorganic dissolved carbon (IDC) was determined in an acidification module CM5130. Dissolved organic carbon (DOC), dissolved organic nitrogen (DON) and dissolved organic phosphorous (DOP) were calculated as the difference between total dissolved forms and inorganic dissolved forms.

Microbial C (C_mic), N (N_mic) and P (P_mic) concentrations were determined by the chloroform fumigation extraction method (Vance, Brookes & Jenkinson, 1987). Fumigated and non-fumigated samples were incubated for 24 h at 25 °C and constant moisture. Microbial C was extracted from fumigated and non-fumigated samples with 0.5 M K₂SO₄ and filtered through Whatman No. 42 filters (Brookes et al., 1985). The concentration of C was measured in each extract as total and inorganic C concentration by the method described before. Microbial C was calculated by subtracting the extracted carbon in non-fumigated samples from that of fumigated samples and dividing the result by a K_EC value (the extractable part of microbial biomass C) of 0.45 (Joergensen, 1996). Microbial N was extracted with the same procedure used for C_mic, but the extract was filtered through Whatman No. 1 paper. The filtrate was acid digested and determined as TN by Macro-Kjeldahl method (Brookes et al., 1985). Microbial N was calculated as for C_mic, but divided by a K_EN value (the extractable part of microbial biomass N after fumigation) of 0.54 (Joergensen & Mueller, 1996). Microbial P was extracted using NaCO₃ 0.5M at pH 8.5, after which the fumigation-extraction technique involving chloroform was performed (Cole et al., 1978). Microbial P was calculated as for C_mic and N_mic and converted using a K_P value (the extractable part of microbial biomass P after fumigation) of 0.4 (Lathja et al., 1999). Microbial P was determined colorimetrically by the molybdate-ascorbic acid method using an Evolution 201 Thermo Scientific Inc. spectrophotometer (Murphy & Riley, 1962). Finally, C_mic, N_mic and P_mic values were normalized on a dry soil basis.

Because P is considered the most limited soil nutrient in the east-side of the CCB (Tapia-Torres et al., 2015), alkaline phosphatase activity was analyzed colorimetrically using ρ-nitrophenol (ρNP) substrates, according to Tabatabai & Bremner (1969) and Eivazi & Tabatabai (1977). For this analysis, 2 g of fresh soil and 30 ml of modified universal buffer (MUB) at pH 9 were used for the exoenzyme extraction. Three replicates and one control (sample without substrate) were prepared per sample. Three substrate controls (substrate without sample) were also included per assay. We centrifuged the tubes after the incubation period and then 750 µl of supernatant was diluted in 2 ml of deionized water and absorbance of ρ-nitrophenol (ρNP) measured at 410 nm on an Evolution 201 Thermo Scientific Inc, spectrophotometer. Exoenzyme activities were expressed as micromoles of ρNP formed per gram dry weight of soil per hour (µmol ρNP [g SDW]⁻¹ h⁻¹). This value was standardized by C_mic concentration for expression as a specific enzyme activity (µmol ρNP [mg C_mic]⁻¹ h⁻¹).

Molecular analyses

Total DNA was extracted using the hydroxyapatite spin-column method (Purdy et al., 1996). DNA molecular weight and quality were confirmed using agarose gel electrophoresis. The 16S rRNA gene was amplified from each sample using a polymerase chain reaction (PCR) with the universal primers F27 (5′AGAGTTTGATCMTGGCTCAG3′) and R1492 (5′GGTTACCTTGTTACGACTT3′). Three independent PCRs were performed for each sample. The PCR reactions were 50 µl in volume and contained 2µl of DNA, 1 µl PCR buffer 1× 0.5 mM MgCl₂, 0.2 mM dNTP mixture, 0.2 mM of each primer, 1 unit of platinum Taq DNA Polymerase High Fidelity (Invitrogen), 5% DMSO and 0.05 mg of BSA. The PCR was performed in a thermal cycler (MJ Research, Watertown, MA) under the following cycling program: initial denaturation step at 94 °C for 5 min, then 30 cycles at 94 °C for 1 min, 52 °C for 1 min,and 72 °C for 1 min 20 sec, with a final extension step at 72 °C for 30 min and storage at 4 °C. The three reactions were pooled and purified in a 1% agarose gel using the QIAquick gel extraction kit (Qiagen). The purified fragment was cloned into the vector PCR 2.1 and transformed into Escherichia coli following the manufacturer’s instructions (Invitrogen). Only plasmids containing inserts were isolated for sequencing with the Montage Plasmid Miniprepkit (Millipore). The insertion within the plasmids was sequenced with the Sanger method using the vector-based primer 27F.

Stoichiometric homeostasis

The degree of community-level microbial C:N and C:P homeostasis (H′) by soil microorganisms was calculated with the formula proposed by Sterner & Elser (2002): (1)${\displaystyle H^{\prime }=1∕m.}$ In Eq. (1), m is the slope of log_e C:N_R (Carbon and Nitrogen in the resources) versus log_e C:N_B (Carbon and Nitrogen in the microbial biomass) or slope of log_e C:P_R (Carbon and Phosphorus in the resources) versus log_e C:P_B (Carbon and Phosphorus in the microbial biomass) scatterplot. H′ ≫ 1 represents strong stoichiometric homeostasis, while H′ ≈ 1 represents weak or no homeostasis (Sterner & Elser, 2002).

Resistance and resilience index

Nutrient concentration and enzymatic activity data were both analyzed for resistance and resilience using the indices proposed by Orwin and Warlde (2004). The grassland site was considered as the control (C₀), the cultivated site as the disturbance (P₀) and the abandoned plot was used for measuring resilience 30 years after the cessation of agriculture management (P_x). Resistance (RS) was calculated as follows: (2)${\displaystyle \mathrm{RS}=1-\left(\left(2\left|{\mathrm{D}}_0\right|\right)∕\left({\mathrm{C}}_0+\left|{\mathrm{D}}_0\right|\right)\right).}$

In Eq. (2), C₀ represents the control soil and D₀ is the difference between C₀ and the disturbed plot (P₀). In addition, resilience (RL) was calculated as follows: (3)${\displaystyle \mathrm{RL}=\left(\left(2\left|{\mathrm{D}}_0\right|\right)∕\left(\left|{\mathrm{D}}_0\right|+\left|{\mathrm{D}}_X\right|\right)\right)-1\left(3\right).}$ In Eq. (3), D_X is the difference between C₀ and P_x. Both indexes are bounded by −1 and +1, if the value is −1 means less resistance or resilience, while the +1 value means maximal resistance or resilience.

Bioinformatics analysis

Sequencing quality evaluation as well as cloning vector removal were performed using the sorftware PHRED (Ewing & Green, 1998). For processing and classification of the sequence data, the open source software package Mothur (v 1.15.0; Schloss et al., 2009) was used. Sequences were screened for potential chimeric reads using Chimera.slayer (Haas et al., 2011) and the linked SILVA template database. High-quality sequences were compared against the SILVA database in order to obtain their taxonomic rank. A pairwise distance matrix was calculated across the non-redundant sequences, and reads were clustered into operational taxonomic units (OTUs) at 3% distance, using the furthest neighbor method (Schloss & Handelsman, 2005). In addition, the Simpson and Shannon (H) indices, Chao species richness estimator and rarefaction curves were estimated.

Statistical analysis

One-way ANOVA was used to identify differences in nutrient concentrations and enzymatic activity between the sites of the agricultural gradient (grassland, cultivated field and abandoned field). Log-transformations were applied where the data deviated from normality. When ANOVA indicated a significant site effect, mean comparisons were performed with Tukey’s multiple comparisons test (Von Ende, 1993).

Pearson correlations were used to explore relationships among soil parameters. Principal Components Analysis (PCA) was conducted in order to group soil samples with active nutrients forms (dissolved, available and microbial) and enzymatic activity. Similarly, Canonical Analysis was conducted with soil nutrients (available, dissolved organic and pH) as the independent variables and nutrients within microbial biomass and phosphatase activity as dependent variables. All analyses were performed using R software 2.10.1 (R Development Core Team, 2009).

Soil nutrients

The abandoned and cultivated plots had the highest and the lowest soil pH and soil electrical conductivity, respectively (P < 0.0001 and P = 0.0002 for pH and electrical conductivity, respectively; Table 1). Total organic C, N and P concentrations differed among management gradient plots. Total organic C was almost two times greater in the cultivated plot than in the other two plots (P < 0.0001; Table 1), whereas the cultivated and grassland plots presented the highest and the lowest N and P concentrations, respectively (P < 0.001 and P < 0.0001 for N and P, respectively; Table 1). As a consequence, the highest C:P and N:P ratios were in the grassland plot (P < 0.0001 for both C:P and N:P), while the C:N ratio did not differ among plots (Table 1).The cultivated plot presented higher DOC and DOP than the other two plots (P < 0.0001 and P < 0.001 for DOC and DOP, respectively), but DON presented no differences among plots (Table 1). Similarly, the cultivated plot presented a greater concentration of ammonium than the other two plots (P < 0.0001), but the highest values of nitrate and available P were in the abandoned and the grassland plots, respectively (P < 0.0001 for both NO₃ and available P; Table 1).

Nutrients within microbial biomass

The cultivated plot had higher C and N concentrations within the microbial biomass (P < 0.0001 for both C_mic and N_mic), but did not differ from the abandoned plot in terms of microbial P (Table 1). However, the grassland plot had higher N_mic concentration than the abandoned plot and, consequently, the C:N and C:P ratios of the microbial biomass did not differ among plots, but the N:P ratio was highest in the cultivated plot (P = 0.05).

Using the equation for C:N and C:P homeostasis (H′), the soil microbial community did present a strong elemental homeostasis for phosphorus acquisition in the three sites (H′ = 6.25, 9.35 and 12.9 respectively for cultivated, grassland and abandoned plots). For nitrogen acquisition, however, the microbial community of the cultivated soil presented a weak homeostasis (H′ = 0.63), while the grassland (3.23) and abandoned plot (5.29) presented higher homeostasis.

Enzymatic activity

The grassland soil had higher specific phosphatase activity than the other two managed plots (P < 0.0001; Table 1). The DOC correlated positively with DOP, ammonium, nutrients within microbial biomass and phosphanatase activity, while nitrate correlated negatively with available P and phosphanatase activity (Table 2). The first two principal components explained 74% of the total variance, in which 54% was explained by the first component. In the first component, the cultivated plot differed statistically to the other two non-cultivated plots, while all three plots were significantly different in the second component (Fig. 1). These results suggest that the difference between the cultivated plot and the other two plots explained 54% of the total variance in the soil nutrient dynamic. The dynamic forms of soil nutrients strongly correlated with nutrients within microbial biomass and phosphatase activity as determined by canonical analysis (Canonical R = 0.98, P < 0.0001). The eigenvalue of root 1 was 0.95 and pH and POD had the highest canonical weight in root 1.

Composition of bacterial communities

A total of 111 sequences were obtained for the grassland, 107 sequences for the cultivated plot and 93 sequences for the abandoned site. In the grassland, we obtained a clone library with 111 sequences, while the cultivated plot had 107 sequences and the abandoned plot had 93. In the grassland, the sequences were distributed among 12 phyla and 19 classes, while the cultivated plot sequences comprised 9 phyla and 14 classes, and those of the abandoned plot comprised 9 phyla and 12 classes. These results suggest that the bacterial community of the grassland soil was distributed in higher phyla than was the case in the other two managed plots. For example, Protobacteria was the more abundant bacteria phylum in the three plots, accounting for 50% of the results in the grassland and the abandoned plot, but representing only 35% in the cultivated plot (Fig. 2). Similarly, Actinobacteria was the second most dominant phylumin both the grassland and abandoned plot (20% and 21%, respectively), but only represented 15% in the cultivated plot. The two most important phototrophic phyla (Chloroflexi and Cyanobacteria) were not found in the cultivated plot, but Cyanobacteria was found in both the grassland soil and abandoned plot (Fig. 2).

Diversity of bacterial communities

Rarefaction curve analysis showed that the cultivated plot had the richest bacterial community, followed by the abandoned plot and finally the grassland soil (Fig. 3). In addition, the cultivated plot had the highest expected OTUs by the Chao analyses (659), while the abandoned plot had the lowest expected value of OTUs (179). The latter plot also had the lowest values of Simpson and Shannon indices (D = 0.025 and H = 3.8, respectively), suggesting that the bacterial community of the abandoned plot was dominated by fewer OTUs in comparison with the bacterial communities in the cultivated plot and the grassland soil (D = 0.04, H = 4.4 and D = 0.013, H = 4.2; respectively).

From the total of 307 sequences obtained for all sites, 223 OTUs were recognized at 97% of similitude. The cultivated plot again had the highest number of OTUs (92), followed by grassland (84 OTUs) and finally the abandoned plot with the lowest number of OTUs (59). The three sites shared four OTUs corresponding to the Proteobacteria (Rhizobiales, Pseudomonadales, Burkholderiales and Xanthomonadales). The abandoned plot shared two OTUs with the other sites, but there were no OTUs shared between the grassland and the cultivated plot. Finally, the grassland soil and abandoned plot presented higher similitude between them relative to the cultivated plot, using the 16S rRNA community composition at 97% similarity based on the Bray-Curtis algorithm.