Mutualism mediates legume response to microbial climate legacies

Climate change is altering both soil microbial communities and the ecological context of plant-microbe interactions. Predicting how soil microbes modulate plant resilience to climate change is critical to mitigating the negative effects of climate change on ecosystems and agriculture. Previously, it was demonstrated that heat, drought, and their legacies altered soil microbiomes and potential plant symbionts. In this study, we conducted growth chamber experiments to isolate the microbially-mediated indirect effects of heat and drought on plant performance and symbiosis. In the first experiment, we found that drought and drought-treated microbes, along with their interaction, significantly decreased the biomass of Medicago lupulina plants compared to well-watered microbiomes and conditions. In a second experiment, we then tested how the addition of a well-known microbial mutualist, the rhizobium Sinorhizobium meliloti, affected climate-treated microbiomes’ impact on the M. lupulina. We found that drought-adapted microbiomes negatively impacted legume performance by increasing mortality and reducing leaf number early in life, but that adding rhizobia erased climate treatment effects. Drought can negatively affect legume performance through microbial legacy effects alone, but the addition of rhizobia buffers legumes against climate-mediated variation in the microbiome. In contrast, heat-adapted microbiomes did not differ significantly from control microbiomes in their effects on a legume.

Methods
Study system

The temperature-free-air-controlled enhancement (T-FACE) experiment is located in an old field at Koffler Scientific Reserve (KSR, www.ksr.utoronto.ca) in Ontario, Canada (44°01'48”N, 79°32'01”W). The experimental design, treatment effectiveness, and soil microbiome analysis are fully described in Boyle et al. (2024; https://doi.org/10.1101/2023.08.04.551981), but we give a brief overview here. Plots in the T-FACE experiment grew only white spruce, Picea glauca, and were either heated, droughted, heated and droughted, or ambient (3 plots/treatment). Boyle et al. (2024) applied treatments during the growing seasons of 2020 and 2021, with rainout structures present for 8 months and heaters activated for 9 months. During active treatment, the mean soil temperature of heated plots was 3.7 ℃ or 3.6 ℃ hotter than un-heated plots in 2020 and 2021, respectively. In 2020, the mean soil volumetric water content (VWC) during droughts was 0.28 (m3/m3) in non-drought plots and 0.25 (m3/m3) in drought plots, and in 2021, mean VWC was 0.26 (m3/m3) in non-drought plots and 0.21 (m3/m3) in drought plots. We collected sifted soil (passed through a 4.75 mm metal sieve) from each plot on June 5th, 2022, sterilizing tools between each collection. We kept soil bags stored slightly open in the dark at 4 ℃ until application to the plants.

We used seeds of a single genotype of Medicago lupulina, an annual legume that forms indeterminate root nodules with rhizobia in the genus Sinorhizobium. Seeds were collected from KSR and selfed for two generations in the University of Toronto greenhouses. Medicago lupulina is naturalized at KSR near the experimental warming array, and its growing season peaks in June, making the timing of our soil collection and plant system ecologically relevant.

Experiment 1: Effect of drought-treated microbes on legume performance in drought

We tested whether drought-treated microbes affected legumes in both dry and well-watered soil conditions. We implemented a 2×2 factorial design with plants receiving live drought soil or live control soil, and experiencing a terminal drought or well-watered conditions, with 30 replicates per treatment, for a total of 120 plants. We pooled soil within treatment types to generate inocula. We began the gradual terminal drought treatment with 2 weeks of well-watered conditions, followed by 2 weeks of ⅔ volume water, then 3 weeks of ⅓ volume water, and finally no further water. Ten weeks post-planting, just over 80% of terminal drought plants were dead and we ended the experiment. We counted branch number through time and measured the nodule number and dry weight of above- and below- ground biomass of each plant at the end of the experiment.

Experiment 2: Soil climate legacy effects with and without an additional mutualist

We next tested whether heat- and drought-treated microbes affected legumes and rhizobia, under well-watered conditions only. We factorially tested the four climate treatments, each with 3 replicate plots from the warming array, and with sterilized or fresh soil. We used sterilized soil to test for abiotic differences in soil from different field plots and climate treatments. The field soil resulted in poor nodulation in Experiment 1. In Experiment 2, we also inoculated half of the plants with Sinorhizobium meliloti 1021-71 tagged with green fluorescent protein (GFP) (courtesy of Daniel Gage, (Gage et al. 1996) to determine whether sufficient rhizobia buffer legumes against the negative effects of drought-treated microbiomes. In total, we had 48 treatments replicated 15 times for a total of 720 plants. We also included an additional 15 sterile control plants with no additional soil.

Over almost 12 weeks, we surveyed branch number and mortality. For all plants, we counted the number of nodules then dried and weighed above and below ground biomass. At the end of the experiment, we randomly sampled five live plants per treatment receiving biotic soil to sequence the microbial communities in their rhizosphere and nodules (n = 120), prior to drying. 

We extracted DNA from the rhizosphere and nodule microbes, then sequenced the fungal ITS region and bacterial 16S V4 region. We used Quantitative Insights Into Microbial Ecology 2 (QIIME2) v.2022.2 (Bolyen et al. 2019). We trimmed the sequences for quality. For 16S reads we trimmed the left 20 bp from forward and reverse reads and truncated reads at 240 bp, while for ITS reads we trimmed the left 25 bp for forward sequences, trimmed 20 bp for reverse sequences, and truncated at 240 bp for both forward and reverse reads. We denoised the sequences with DADA2 (Callahan et al. 2016) into amplicon sequence variants (ASVs). We removed ASVs that had fewer than 10 reads across all samples, and assigned taxonomy using the ‘sklearn’ feature classifier (Pedregosa et al. 2011) with the 2021 Greengenes 16S V4 region reference for bacteria (McDonald et al. 2012), and UNITE version 9.0 with dynamic clustering of global and 97% singletons for fungi (Abarenkov et al. 2022). After assigning taxonomy, the bacterial rhizosphere had 8,058 ASVs with 3,055,989 reads and a median 26,923 reads/sample. The fungal rhizosphere had 1,016 ASVs and 2,902,562 total reads, and a median of 25,278 reads/sample. Nodule samples had 269 ASVs with 1,842,801 total reads, and median 28,518 reads/sample. Then we filtered out reads assigned as chloroplasts and mitochondria to remove plant DNA. After filtering, the bacterial rhizosphere retained 8,019 ASVs and 3,043,867 total reads, the fungal rhizosphere retained all ASVs and reads, and the nodule data retained 237 ASVs and 1,114,982 reads. We rarefied bacterial and fungal rhizosphere samples to 15,000 reads and 12,400 reads respectively, and rarefied nodule bacteria to 1000 reads. After rarefaction, we had 113 samples for the bacterial rhizosphere, 115 samples for the fungal rhizosphere, and 58 samples for nodules. Finally, we constructed phylogenies for bacterial reads using QIIME2’s MAFFT (Katoh & Standley 2013) and FastTree (Price et al. 2010) functions to obtain rooted trees.