Hyperlocal Variation in Soil Iron and Rhizosphere Microbiome Determines Disease Development in Amenity Turfgrass

Dollar spot, caused by the fungal pathogen Clarireedia spp., is an economically important disease of amenity turfgrass in temperate climates worldwide. This disease often occurs in a highly variable manner, even on a local scale with relatively uniform environmental conditions. The objective of this study was to investigate mechanisms behind this local variation, focusing on contributions of the soil and rhizosphere microbiome. Turfgrass, rhizosphere, and bulk soil samples were taken from within a 256 m2 area of healthy turfgrass, transported to a controlled environment chamber, and inoculated with C. jacksonii. Bacterial communities were profiled targeting the 16s rRNA gene, and 16 different soil chemical properties were assessed. Despite their initial uniform appearance, the samples differentiated into highly susceptible and moderately susceptible groups following inoculation in the controlled environment chamber. The highly susceptible samples harbored a unique rhizosphere microbiome with lower relative abundance of antibiotic-producing bacterial taxa and higher predicted abundance of genes associated with xenobiotic biodegradation pathways. In addition, stepwise regression revealed that bulk soil iron content was the only significant soil characteristic that positively regressed with decreased dollar spot susceptibility during the peak disease development stage. These findings suggest that localized variation in soil iron induces the plant to select for a particular rhizosphere microbiome that alters the disease outcome. More broadly, further research in this area may indicate how plot-scale variability in soil properties can drive variable plant disease development through alterations in the rhizosphere microbiome. IMPORTANCE Dollar spot is the most economically important disease of amenity turfgrass, and more fungicides are applied targeting dollar spot than any other turfgrass disease. Dollar spot symptoms are small (3-5 cm), circular patches that develop in a highly variable manner within plot-scale even under seemingly uniform conditions. The mechanism behind this variable development is unknown. This study observed that differences in dollar spot development over a 256 m2 area were associated with differences in bulk soil iron concentration and correlated with a particular rhizosphere microbiome. These findings provide important clues for understanding the mechanisms behind the highly variable development of dollar spot, which may offer important clues for innovative control strategies. Additionally, these results also suggest that small changes in soil properties can alter plant activity and hence the plant-associated microbial community which has important implications for a broad array of important agricultural and horticultural plant pathosystems.

Soil spatial variation in microbial properties is often studied at multiple levels, including micro, 92 plot, field, landscape and regional scales (15,16). Over a small plot-scale, spatial variation of 93 smut disease (Ustilago syntherismae) on crabgrass (Digitaria sanguinalis) was influenced by 94 both pathogen spore density and spatial location (17). However, soil property influences were 95 not investigated in this study and spores or other long-distance dispersal mechanisms have never 96 been associated with dollar spot in a field environment (2). High spatial variations in soil 97 physicochemical and microbial properties were observed in a managed grassland, including a 98 wide range of soil pH, nitrogen content, microbial biomass, and microbial catabolism profiles 99 within the scales of several centimeters to meters (18), but the impact of these variations on the rhizosphere microbiome on plant health over a field-scale or smaller. However, it remains 108 unknown whether the rhizosphere and/or bulk soil microbiome impacts the disease severity of a 109 foliar fungal pathogen when interacting with specific soil chemical properties.

110
In this study, various factors contributing to the localized variation in dollar spot development on 111 monocultured turfgrass was studied. Rhizosphere and bulk soil microbiomes as well as soil 112 chemical properties were examined to determine possible causes for the highly variable spatial 113 nature of dollar spot development. We hypothesized that soil chemical properties and the 114 rhizosphere microbiome are both significant variables in determining dollar spot disease 115 susceptibility in a uniformly managed and monocultured turfgrass system. Turfgrass is an 116 excellent system to study this phenomenon because the high plant density allows for robust 117 sampling over a small scale. The initial 132 cm 2 surface area turfgrass soil plug harbored an 118 estimated 1,200 individual creeping bentgrass plants, and each sub-sample derived from the soil 119 core contained 10 to 15 individual plants. By understanding the factors that drive variation in 120 dollar spot disease development within a plot-scale in a high-density monoculture system, we 121 may discover mechanisms that can be targeted for improved biological management of a number 122 of important plant pathogens. be difficult to determine with simple visual assessments. The resulting greenness decay curve followed a sigmoidal decay pattern (r=0.9286 and p-value<0.0001) (Fig.1). Disease symptoms 129 initially developed within two days after inoculation (DAI), then increased rapidly over the next 130 four to twelve DAI, before slowing during the saturation phase on 14 to 16 DAI. Substantial 131 differences in symptom severity between samples started showing up on four DAI and 132 differences remained apparent throughout the incubation.

133
Attributing soil bacterial community difference as a function of disease variability. Turf 134 samples were grouped into high, medium, and low disease according to the disease severity of 135 each DAI. The bacterial microbiome from rhizosphere and bulk soil associated with each sample, 136 which had been separated prior to inoculation, was then assessed to see if the microbiome 137 structure explained turfgrass responses to C. jacksonii inoculation. The rhizosphere bacterial 138 community differed between high and low disease severity groups when categorized based on 139 severity between 4 and 10 DAI according to permutational analysis of variance (PERMANOVA) 140 (Table 1). There were no differences in bacterial community structure found between high and 141 low disease severity groups when categorized according to initial disease development (DAI 0-2) 142 or the disease saturation phase . In addition, no differences in the bulk soil bacterial 143 community were found among the disease severity groups throughout the entire incubation 144 (Table 1). The period that the rhizosphere soil microbiome showed differences in structure 145 between the high and low disease groups (4-10 DAI) matched the backslope of the disease 146 development curve (Fig. 1), which suggested that the initial soil rhizosphere microbiome can 147 affect the peak dollar spot development. The samples were then re-categorized according to their 148 disease status during the peak disease development stage (4-10 DAI) to make the peak disease 149 development period as the target of prediction instead of any single day within this period. The 150 samples initially categorized as high disease during the 4 to 10 DAI period never shifted into the low severity group and vice versa, so the 18 samples naturally broke into two groups except for 152 one sample that stayed in the medium disease group throughout the study and was excluded from 153 further analysis. Further analyses were performed based on breaking the samples into nine highly with more ASVs being unique to HS (1181) than MS (347) (Fig. 4). Highly susceptible turfgrass 169 samples also had a higher species richness and β-diversity as shown using the Shannon index 170 (Fig. 5).

171
In the rhizosphere, there were 28 families and 32 genera different in relative abundance between 172 discerning the high and low disease rhizosphere bacterial community. The signatures were determined by searching the association between the factor for overall microbiome difference 176 with the bacterial taxa balances defined as normalized log ratio of the geometric mean of the 177 numerator and denominator bacterial taxa. The results showed that relative abundance log ratio

183
A co-occurrence network analysis was performed to visualize the microbial interaction of HS 184 and MS turf rhizosphere soil bacteria and showed different network patterns (Fig. 7a). The co-185 occurrence networks were then further analyzed using "NetShift" to quantify the differences and 186 identify the keystone microbial taxa that triggered the shift of the microbial networking between 187 HS and MS rhizosphere bacterial communities when clustered at the Family and Genus level 188 (Fig. 7b). There were 55 families and 28 genera identified as driver taxa when comparing HS and 189 MS co-occurrence networks aggregated at each taxonomic level.

190
Rhizosphere soil bacterial function was predicted using Tax4Fun2 (21) to explore the potential 191 microbial functional differences between HS and MS samples during the peak disease 192 development period. Predicted functional pathways at level-two according to KEGG reference 193 for molecular functions of genes (22) including nucleotide metabolism, folding, sorting and 194 degradation, cell motility, translation, transcription, replication and repair, and metabolism of 195 cofactors and vitamins associated genes were found to be more abundant in rhizosphere of MS samples (Fig 8). In the HS samples rhizosphere genes associated with xenobiotic biodegradation 197 and metabolism pathways associated genes were more abundant (Fig. 8).  (Table 3). Pseudomonas. In our study, differential analysis revealed that certain families and genera were 229 higher in relative abundance in the rhizosphere of MS samples compared to HS samples. These   Data analysis. The raw sequences were processed using package "DADA2" in R 3.6.0. Forward 392 and reverse reads were quality filtered according to average quality score and merged. The 393 taxonomy levels associated with each amplicon sequence variant (ASV) was assigned according 394 to SILVA database (v.132) after removing the chimeras. The ASV and taxonomic tables were 395 then exported as .txt files and analyzed using R packages "phyloseq" and "vegan." The reads for 396 each sample were normalized using variance stabilizing transformation with the "DeSeq2" 397 package due to a relatively even reads variation among the samples in the library (54). Microbial 398 compositional differences and correlations were analyzed using Bray-Curtis dissimilarity.  Since the overall ASVs were comprised of approximatly 90% of the ASVs having less than 406 0.02% of overall reads, ASVs that represent less than 0.02% of the total reads after normalization 407 for each sample were filtered out to make the result more readable. The core community of the 408 HS and MS microbial networks were compared to quantify the rewiring of the taxa in the