1 INTRODUCTION
Desertified reclamation for irrigated agricultural use is increasingly necessary to meet with the needs of a growing global population, and large swathes of reclaimed farm and forest land have thus been established throughout the world (Salama, Abd El-Ghani, El-Tayeh, Amro, & Abdrabbu, 2017; Yang et al., 2019). The reclamation of desertified land to generate economic forest lands has been practiced in regions of Northwest China such as the Hongsibu District in Ningxia. This region experiences a very dry climate and significant temperature variability depending on the time of day such that few crops can reliably grow in this area. However, winegrapes can grow well under these conditions, leading to their increasingly widespread cultivation and rising incomes for grape-growers. Through concerted desertified land reclamation efforts over the past two decades in Hongsibu, the winegrape cultivation area had risen to 7067 hm2 as of 2018 (Statistics of Ningxia Wine Bureau). These land reclamation efforts result in well-documented changes in soil quality attributable to changes from desert soil to irrigated cropland (Ferreira, Leite, de Araújo, & Eisenhauer, 2016; Tosi et al., 2016). Crop cultivation, however, can also reduce certain aspects of soil quality including soil organic carbon (SOC) and nitrogen levels (Liu et al. 2005; Bhattacharyya et al., 2014; Sharma et al., 2017), in addition to altering the composition of soil microbial communities (Crecchio et al., 2007). Exploring how soil properties change following crop planting is essential to more fully assess the relative success of different land reclamation strategies in order to support sustainable land management efforts.
Soil quality is an integrative concept regulated by the physical, chemical, and biological properties of the soil (Oliver, Bramley, Riches, Porter, & Edwards, 2013). Bulk soil density, porosity, soil organic matter/SOC content, pH, and available nutrient levels are among the most commonly used metrics to gauge such soil quality (Bünemann et al., 2018; Morvan et al., 2008; Mouazen, Steffens, & Borisover, 2016). Changes in soil properties following reclamation are vital to the monitoring of sustainable land use. Microbial diversity in the context of desertified soils, however, is not well documented and must be further studied to more fully clarify how winegrape planting impacts the properties of the underlying soil.
Reclamation-related shifts in soil microbial communities can be positive or negative depending upon the land-use type. For example, converting land use from a natural ecosystem to a managed agricultural ecosystem can adversely impact soil microbial diversity (Thomson et al., 2015), whereas the conversion of lands with extreme conditions such as high salinity or aridity into cropland can positively impact associated microbial communities (Wang et al., 2012). Many reports have demonstrated the effects of land-use changes on soil microbial community composition and diversity in very arid environments, with several soil properties being closely linked to these microbial shifts. For example, alterations in soil pH and moisture content were shown to be related to bacterial communities therein following 30 years of agricultural practice in desert soil in Egypt (Köberl, Müller, Ramadan, & Berg, 2011). In the Atacama Desert, soil relative humidity was also the primary determinant of microbial community richness and diversity (Neilson et al., 2017). These findings emphasized the overriding impact of water availability and related soil properties on microbial community composition in the context of improving soil quality and productivity in arid regions throughout the world. The Yellow River flows through Ningxia, and owing to the Yellow River Irrigation Project, water is not a major limiting factor associated with soil quality, whereas high salinity, low organic matter, and nutrient levels are likely to be important regulators of microbial community composition and associated shifts in productivity (Cheng, Chen, & Zhang, 2019). Further studies regarding these concomitant changes in soil properties and microbial communities in these desertification areas are thus necessary to ensure sustainable land use.
Soil microbial communities are highly sensitive to variations in soil physicochemical properties (Franco-Otero, Soler-Rovira, Hernández, López-de-Sá, & Plaza, 2012) and land use alterations (Szoboszlay, Dohrmann, Poeplau, Don, & Tebbe, 2017; Tosi et al., 2016; Wu et al., 2017). Recent research suggests that sustained monocropping can reduce the diversity of soil microbial communities and can interfere with soil microbial community structure (Ellouze et al., 2014; Li, Ding, Zhang, & Wang, 2014). In one study, bacterial richness and diversity were shown to rise following the conversion of desert land to 5-year-old cropland, and thereafter remained stable following 5 years of cotton cropping (Li et al., 2020). Land-use type can profoundly shape soil bacterial community composition and diversity owing to the resultant shifts in soil chemical properties, particularly in the context of desertified land reclamation (Li, Pokharel, Liu, & Chen, 2020). There have been few studies to date, however, assessing changes in soil microbial communities and associated soil parameters after desert land reclamation for viticulture.
Herein, we evaluated soil microbial community composition and diversity in vineyards following the reclamation of desertified land in Hongsibu in Northwest China, with the goal of providing a theoretical foundation for sustainable land use efforts. We hypothesized that: (a) soil microbial communities would be significantly altered following desertified land reclamation in a manner associated with the types of grapes grown in different vineyards, and (b) that soil microbial community composition in these reclaimed lands is attributable to the soil chemical properties (such as pH, SOC, nitrogen, and phosphorus levels) associated with these different land use types.
2 MATERIALS AND METHODS2.1 Study area
This study was conducted in the Huida Chateau and Xiaojiayao winegrapes planting area in the Hongsibu District of Northwest China (Figure 1). Hongsibu is the largest wine-growing sub-area in the eastern region of the Helan Mountains, which is a wine region with a nationally protected geographical indication. While previously dominated by natural desert land (DL) populated by a small number of salt-tolerant plants able to grow under arid conditions including Populus L. ,Zygophyllaceae , Ulmusglaucescens , Leguminosae , andElaeagnaceae (Xu, Wang, Jia, & Guo, 2020), the reclamation of this land in the 1990s has led to its use for crop growth. Winegrapes are the crop best-suited to growing in this region owing to its arid conditions and substantial diurnal temperature variability.
The study site was situated within a mountain basin with an area of 2,767 km2 at an altitude of 1,240 - 1,450 m with a typical temperate continental climate. The region remains dry throughout the year with an average annual precipitation of 251 mm, an average annual evaporation of 2,387 mm, an average annual temperature of 8.7°C, a daily temperature difference of 13.7 °C, a sum of accumulated effective temperatures (≥10°C) above 3,200°C, 2,900 - 3,550 h of annual sunshine, and an averageannual wind speed of 2.9 - 3.7 m/s. The soil exhibits a sandy loam texture (50% clay, 30% silt, and 20% sand). The same type of soil was used in the desertified land reclamation associated with all studied vineyards, which are regularly managed, with plants arranged in north-south rows at a row spacing of 3 m, a vine spacing of 0.5 m, and cultivation mode of single cane ‘Dulonggan’, only Italian Riesling cultivation mode of Crawled Cordon Training. Annual fertilization amounts include an estimated 40.18 kg/hm2 of nitrogen fertilizer, 30.21 kg/hm2 of phosphate fertilizer, 8.68 kg/hm2 of potassium fertilizer, and 22.18 kg/hm2 of organic manure, while the annual irrigation water volume is 4000 m3/hm2.
2.2 Experimental design and soil sampling
Winegrape croplands reclaimed from desert land in 2012, 2014, and 2017 were selected for the present study in 2020, with crops of Cabernet Sauvignon (CS), Italian Riesling (IR), Merlot (M), and Chardonnay (C) grapes having been planted 8, 6, 3, and 3 years following reclamation, respectively. Undisturbed, non-cultivated desertified soil was collected at two research sites adjacent to these vineyards, and the same amount of soil from each site was evenly mixed following collection. The distance between these two sampling sites varied from 0.5 – 3 km.
In September 2020, three replicate 10 × 10 m plots separated by at least 100 m were established. Within each plot, a 0-20 cm rhizosphere soil sample was collected using a steel corer (5 cm diameter) following surface litter removal. In total, 10 random cores were collected per plot and combined into a single composite sample. Three samples of undisturbed soil were taken from the middle of these cores (~10 cm depth) to assess soil bunk density using a corer (5 cm diameter; 100 cm3). Soil samples were sealed into plastic bags and kept on ice while in transit. In the laboratory, samples were passed through a 2 mm sieve to remove coarse matter and plant debris, after which samples were split into two parts, one of which was air-dried for physicochemical analyses and the other of which was stored at -80°C and used for DNA isolation within 1 week.
2.3 Soil physicochemical analyses
After soil had been dried to a constant weight at 105°C, its bulk density was established. Soil pH was measured at a water-soil ratio of 2.5:1 without any CO2 water extraction via a potentiometric method. Total nitrogen (TN) was assessed via the micro-Kjeldahl method (Lu 1999). Soil available nitrogen was quantified via an alkaline hydrolysis diffusion approach (Bao 2000). Soil ammonium nitrogen (NH4+-N) and nitrate-nitrogen (NO3--N) were leached with a 2 mol/L potassium chloride solution and then measured via flow injection analysis. Soil total phosphorus and available phosphorus were quantified using a molybdenum-antimony anti-colorimetric approach (Bao 2000). Soil total organic carbon was assessed via an external heating method using potassium dichromate (Carter &Gregorich, 2007). A flame photometer approach was employed to measure total potassium content, while available potassium was quantified via an ammonium acetate extraction-flame photometric method (Bao 2000; Smith & Bain, 1982).
2.4 DNA extractionand sequencing
An OMEGA Soil DNA Kit (Catalog number D5625-01) was used to extract soil DNA, and DNA concentrations were measured with a UV-visible spectrophotometer (Thermo Scientific, DE, USA). DNA quality was assessed via 0.8% agarose gel electrophoresis. The V3 andV4 bacterial 16s rRNA hypervariable regions were amplified with the 5’-ACTCCTACGGGAGGCAGCA-3’ forward primer and the 5’-CGGACTACHVGGGTWTCTAAT-3’ reverse primer. Additionally, the V1 hypervariable region of the fungal ITS gene was amplified with the 5’-GGAAGTAAAAGTCGTAAGG-3’ forward primer and the 5’-GCTGCGTTCTTCATCGATGC-3’ reverse primer. A sequencing library was constructed with the TruSeq Nano DNA LT Library Prep Kit (Illumina). After library quality had been assessed, community DNA fragments were analyzed via paired-end sequencing using an Illumina MiSeq instrument by Shanghai Peisenol Biotechnology Co., Ltd.
2.5 Data processing
Raw high-throughput sequencing data were initially assessed to measure sequence quality. Samples were separated into libraries according to barcode information, and barcode sequences were removed. Those sequences > 200 bp in length with a mean quality score ≥ 20 were retained for further analysis. Primer removal, mass filtration, denoising, splicing, and the removal of chimeric sequences were conducted via the DADA2 method (Callahan et al., 2016). Rather than performing similarity-based clustering, this approach only consisted of de-replication. Amplicon sequence variants (ASVs) or operational taxonomic units (OTUs) were generated according to the QIIMA2 DADA2 quality control results. After rarefaction-based normalization to the same sequencing depth (95% of the lowest sample sequence depth), ASVs/OTUs were predicted for each sample along with relative abundance at that sequencing depth (Kemp, & Aller, 2004).
2.6 Statistical analyses
Data normality and homogeneity of the variance were respectively analyzed via the Shapiro-Wilk and Levene’s tests, after which soil properties were compared by conducting one-way ANOVAs. Venn diagrams were used to identify shared and unique ASVs/OTUs associated with different land-use types. Observed species, Chao1, Shannon, and Simpson indices were used to assess the diversity and richness of soil microbial communities at the ASV/OTU level (Qiao, Zhang, Shi, Song, & Bian, 2018). A nonmetric multidimensional scaling (NMDS) approach was employed to assess differences in soil microbial communities across all soil sample types based upon Bray-Curtis distances, with similarities between samples being displayed via hierarchical clustering analyses. A permutational multivariate analysis of variance (PERMANOVA) approach was used to test for differences in soil microbial composition as a function of land-use type based upon the adonis test (Anderson, 2001; McArdle, & Anderson, 2001). LDA Effect Size (LEfSe) was employed to identify land-use type-related biomarkers (Segata et al., 2011). A redundancy analysis (RDA) was performed to examine correlations between microbial communities and soil chemical properties using Canoco 5.0. Pearson correlation analyses were conducted to examine relationships between soil properties and microbial communities. R, QIIME 2.0, and SPSSS 24.0 were used for all analyses.
3 RESULTS3.1 Soil properties
Significant differences in soil chemical properties were observed as a function of land-use type (Table 1). DL soils exhibited a significantly higher bulk density relative to other land-use types, whereas DL soil had the lowest levels of TN and available potassium (Table 1). TN levels were highest in C soils (0.40 g kg-1), while available nitrogen was the highest in CS soils (37.63 mg kg-1), and TN levels were significantly higher in CS, M, C, and IR soil samples relative to DL samples (p < 0.05). Available phosphorus levels were highest in DL soil (7.86 mg kg-1), with significant differences in available phosphorus in the CS, M, C, IR, and DL samples (p < 0.05) without any concomitant differences in TP. TK exhibited a downward trend from CS, M, C to IR, and then rose to DL. Available potassium levels in CS, M, C, and IR samples were 1.44, 1.42, 1.37, and 1.13 times those in DL samples. The TN, SOC, and pH of CS, M, C, and IR soil samples were all significantly higher than those in DL samples (p < 0.05).
3.2 Bacterial and fungal taxa distributions in soils associated with different land-use types
In total, 1,942,867 original sequences and 1,693,153 effective sequences were derived from the 16S_V3V4 region in sequences soil samples, while 1,233,294 original sequences and 1,085,213 effective sequences were obtained from the ITS_V1 region. In total, 165 bacterial ASVs/OTUs were found to be shared across all five land-use types, with the numbers of unique bacterial ASVs/OTUs in these soil samples being rank-ordered as follows: C > IR > M > CS > DL. Following reclamation, soil bacterial ASVs/OTUs in the CS, M, C, and IR samples rose to 2846, 3191, 7630, and 6373, respectively. A total of 55 shared fungal ASVs/OTUs were detected across land-use types, while numbers of unique fungal ASVs/OTUs associated with these different land-use types were, in rank-order: IR > M > CS > C > DL. Following desertified land reclamation, the number of soil fungi present in CS, M, and IR vineyards rose to 788, 984, and 1114, respectively, but the number of fungi in C was lower than in DL (Figure 2).
Actinobacteria (33.58%), Proteobacteria (30.71%),Acidobacteria (12.67%), Chloroflexi (8.20%),Gemmatimonadetes (6.17%), Firmicutes (2.65%),Bacteroidetes (1.57%), and Rokubacteria (1.14%) were the most abundant bacterial phyla in analyzed soil samples (Figure 3a).Actinobacteria and Proteobacteria were the dominant phyla in CS, M, C, and IR soils, whereas Actinobacteria (24.08%),Proteobacteria (22.47%), and Acidobacteria (25.74%) were dominant in DL soils. Relative Actinobacteria andProteobacteria abundance in CS, M, C, and IR soils was higher than that observed in DL soils. Relative Chloroflexi abundance was significantly higher in CS, C, and IR soils relative to DL soils (p <0.05), whereas Acidobacteria andFirmicutes abundance was significantly lower in CS, M, C, and IR soils relative to DL soils (p <0.05).
Ascomycota (71.50%), Basidiomycota (8.87%),Mortierellomycota (5.87%),Mucoromycota(0.86%), and Rozellomycota (0.17%) were the most dominant fungi in analyzed soil samples (Figure 3b), with Ascomycota ,Basidiomycota, and Mortierellomycota being the most dominant phyla associated with different land-use types. Relative to DL samples, the relative abundance of Ascomycota in CS, M, C, and IR samples rose by 5.45%, 6.06%, 34.47%, and 21.59%, respectively.Ascomycota abundance was significantly higher in C and IR soils relative to DL soil (p < 0.05). Basidiomycotalevels were significantly higher in CS and M soils relative to DL soil (p < 0.05), whereas Mortierellomycota abundance was significantly lower in M, C, and IR samples relative to DL soils (p < 0.05).
3.3 Bacterial and fungal community diversity
Next, bacterial richness and diversity in soils associated with different land-use types were measured using the Chao1, Observed species, Shannon, and Simpson indices (Table 2). The rhizosphere bacterial abundance (Observed species) for the four vineyard soil types was significantly higher than that in DL samples, while bacterial abundance in C and IR samples (Chao1) was significantly higher than that in DL soil (p < 0.05). Relative to DL soils, the bacterial diversity (Shannon) in C and IR samples was significantly increased (p < 0.05), while the Simpson index was significantly increased in CS and C samples. Relative to DL soil samples, viticultural land use was associated with increased soil fungal diversity in all soil types other than the C soil (Table 2).
The NMDS ordination of bacteria and fungi are 0.0837 and 0.0999, respectively, with both of these values being less than 0.2, consistent with good analysis results (Figure 4). NMDS revealed clear separation in bacterial composition across land-use types with the exception of CS and M samples, with close clustering of the community composition profiles associated with these two land-use types.
Hierarchical clustering analyses were then performed using the Bray Curtis distance algorithm, revealing that bacteria in vineyard soil samples were clustered together and separated from those in DL soil (Figure 5a). Similarly, for fungal communities, all vineyard soil samples were separated from all DL soil samples other than sample DL2 (Figure 5b). These data suggested that microbial community structure differed significantly when comparing vineyard soils to DL soils.
A PERMANOVA analysis revealed significant differences in soil bacterial composition rankings for different land-use types (R2= 0.65, p = 0.001), with the same also being observed for soil fungal community composition rankings (R2 = 0.58,p = 0.001) (Table 3). LEfSe analysis results indicated that bacteria that differed significantly between land-use types at the class level included Chloroflexi , Fibrobacteres ,Firmicutes , Gemmatimonadetes , Patescibacteria ,Proteobacteria, and Rokubacteria (Figure 6a), with further details regarding differences at other taxonomic levels being further detailed in the corresponding Figure. Fungi that differed significantly between land-use types at the class level included Rozellomycota ,Trichothecium , Eremomycetaceae , and Selenophoma(Figure 6b), with further details being shown in the corresponding Figure.
3.4 Associations between soil bacterial and fungi communities, diversity, and soil chemical properties
Pearson correlation analyses revealed that several soil chemical properties were correlated with bacterial and fungal diversity (Table S). TN was positively correlated with bacteria Chao1 (r=0.558,p <0.05), Observed species (r=0.562,p <0.05), and Shannon index values (r=0.627,p <0.05). Soil nitrate-nitrogen was significantly positively correlated with bacteria Chao1 (r=0.970,p <0.01), Observed species (r=0.970,p <0.01), and Shannon index values (r=0.823,p <0.01). Soil ammonium nitrogen was significantly positively correlated with bacteria Chao1(r=0.812,p <0.01), Observed species (r=0.823,p <0.01), and Shannon index values (r=0.701,p <0.01). SOC and pH were positively correlated with bacterial Chao1 and Observed species index values.
An RDA analysis was then performed to assess the relationships between soil chemical properties and the composition of soil fungal and bacterial communities at the phyla level (Figure 7). The proportion of the total variance in these bacterial communities explained by all eight measured soil chemical parameters analyzed in this study was 82.91% (Figure 7a). The first constraint axis (RDA1) explained 54.25% of this variance, while the second (RDA2) explained 28.66% of the total variance. Overall, pH explained 42.20% (p = 0.004) of total variance in bacterial community composition, followed by available nitrogen (14.10%, p = 0.032). The measured soil chemical parameters in the present study similarly explained 73.97% of the total variance in soil fungal communities (Figure 7b). The first constraint axis (RDA1) explained 71.35% of this variance, while the second (RDA2) explained 2.64% of this variance. Additionally, pH explained 31.50% of the total variance (p = 0.014), followed by available phosphorus, available nitrogen, and SOC.