1. Human land uses, such as agriculture, can leave long-lasting legacies as ecosystems recover. As a consequence, active restoration may be necessary to overcome land-use legacies; however, few studies have evaluated the joint effects of agricultural history and restoration on ecological communities. Those that have studied this joint effect have largely focused on plants and ignored other communities, such as soil microbes.
2. We conducted a large-scale experiment to understand how agricultural history and restoration tree thinning affect soil bacterial and fungal communities within longleaf pine savannas of the southern United States. This experiment contained 64 pairs of remnant (no history of tillage agriculture) and post-agricultural (reforested following abandonment from tillage agriculture >60 years prior) longleaf pine savanna plots. Plots were each 1-ha and arranged into 27 blocks to minimize land-use decision making biases. We experimentally restored half of the remnant and post-agricultural plots by thinning trees to reinstate open-canopy savanna conditions and collected soils from all plots five growing seasons after tree thinning. We then evaluated soil bacterial and fungal communities using metabarcoding.
3. Agricultural history increased bacterial diversity but decreased fungal diversity, while restoration increased both bacterial and fungal diversity. Both bacterial and fungal richness were correlated with a range of environmental variables including aboveground variables like leaf litter and plant diversity, and belowground variables such as soil nutrients, pH, and organic matter, many of which were also impacted by agricultural history and restoration.
4. Fungal and bacterial community compositions were shaped by restoration and agricultural history resulting in four distinct communities across the four treatment combinations.
5. Past agricultural land use left persistent legacies on soil microbial biodiversity, even over half a century after agricultural abandonment and after intensive restoration activities. The impacts of these changes on soil microbe biodiversity could play important roles in the functioning of ecosystems following agricultural abandonment and during restoration.

In Fall 2015 we collected ~1.2 L of soil from each of the 126 1-ha plots. Each soil sample was an aggregate of 30 1.6 cm wide by 20 cm deep soil probes collected along two 50 m transects through the middle of each plot (Fig. 1). The soil sampling transects ran on both sides of our already-present vegetation sampling transects (Turley and Brudvig 2016; Fig. 1). Before each probe the leaf litter, duff, and sticks were brushed aside. To minimize contamination we used one soil probe for all remnant sites and another for all post-agricultural sites and between each plot we rinsed the inside and outside of the probe with a 10% bleach solution and then water. Aggregate soil samples were mixed thoroughly and split up for different purposes. 50 mL was stored in a -20 C freezer for microbial analysis, and two other subsamples were used for environmental sampling.

For microbial analysis, we extracted soil DNA using MoBio PowerSoil Extraction Kit following the manufacture’s instructions. We submitted DNA to the Michigan State University Core Genomics Facility for Illumina sequence library construction. Following their standard protocols, bacterial 16S V4 (515f/806r) and ITS (ITS-F/ITS2) Illumina compatible libraries were prepared using primers containing both the target sequences and the dual indexed Illumina compatible adapters. The 16S and ITS1 amplicon pools were sequenced independently in a 2x250bp paired end format using independent v2 500 cycle MiSeq reagent cartridges.

The first of the soil subsamples was analyzed by Brookside Laboratories Inc. (New Bremen, Ohio, USA) for soil texture (percent sand, clay, and silt), pH, organic matter, and nutrients and minerals. On the second subsample we measured soil water-holding capacity (proportionate difference between saturated wet and oven dry weight) and gravimetric soil moisture using the same methods as Brudvig and Damschen (2011). Soil pH, water holding capacity, organic matter, and several soil nutrients all decreased with agricultural history while soil phosphorus was strongly increased (Table S1).

Environmental data collection

We measured a set of environmental variables within each experimental plot at 10 m intervals along the 100 m vegetation transects (Fig. 1) during the 2015 growing season. In 1 × 1 m plots we visually estimated the percent cover of leaf litter, down woody debris, bare ground, and understory vegetation. At each of these plots we also measured the depth of leaf litter and canopy cover of overstory trees using a spherical densiometer. In 1 × 1 m and also 10 × 10 m plots we recorded all plant species and calculated plant species richness. For all these environmental variables we averaged the 10 measurements across each transect to get one value per 1-ha plot. Restoration thinning resulted in strong declines in leaf litter and canopy cover and large increases in vegetation cover and understory plant richness (Table S1). Units and methods for measuring all of our environmental variables are available in supplementary material (Table S6).

Bioinformatics

We processed and clustered bacterial and fungal reads into Operational taxonomic units (OTUs). Reads from the bacterial community were chimera checked, quality filtered, and merged using Trimmomatic and Pandaseq (Masella et al. 2012, Bolger et al. 2014). Processed reads were clustered into OTUs at 97% identity level using UCLUST6.1 with the default settings (Edgar 2010). Singletons were removed and contigs were screened using QIIME 1.9.1  (Caporaso et al. 2010) with the default parameters. OTUs classified to Chloroplast, Mitochondria, or with less than four reads across all samples were filtered out to avoid over splitting (Thiéry et al. 2012) and sequencing errors (Dickie 2010). The resulting community was composed of 90,103 OTUs and 1,650,420 reads. Fungal reads were quality filtered and merged using the USEARCHv10 pipeline (http://drive5.com/usearch/; Edgar 2010, 2013). Merged sequences were quality filtered to an expected error threshold of 1.0 fastq_filter (Edgar and Flyvbjerg, 2015) and primer sequences bases were removed. The combined reads were clustered into OTUs at 97% identity level then reference based chimera checked (Edgar 2016) and classified against the UNITE 7.1 ITS1 chimera and reference databases, respectively (Kõljalg et al. 2013). All non-fungal OTUs and those with less than four reads were filtered from the community matrix. The resulting fungal community had 10,285 OTUs and 584,113 reads.

See read me file named: turley_etal_jae_2020_read_me_explinations_of_data_files.txt