Site description

This experiment was conducted at the Russell E. Larson Agricultural Research Center in Pennsylvania Furnace, PA (40.723165, -77.929840). This experiment is approximately 4 hectares in area, has been managed according to USDA organic regulations since 2011, and received organic certification in 2014 [42]. In 2023, this site was in Zone 6b of the USDA Plant Hardiness Zones [43]. Annually, this site experiences an average of 1,000 mm of precipitation. Average annual temperatures range from 5˚C to 28˚C. Soils at the site are representative of the Hagerstown Soil Series according to the USDA Natural Resources Conservation Service soil classification system and includes mostly silt loam [44].

Experimental design

The field experiment was comprised of four organic cropping systems that differed in cover crop species, establishment and terminations method; tillage tool, timing, and frequency of tillage; and in-season crop management. Winter cover crops were grown between each annual cash crop. Systems 1 through 3 consisted of a three-year, annual crop rotation of corn (Zea mays L., Viking O.45-88-P)—soybean (Glycine max L. Merr., Viking O.2155 N)—wheat (Triticum aestivum L., Malabar) in which all crops in the rotation were present in each year (i.e., full entry design), and a fourth system comprised of a mixture of alfalfa (Medicago sativa L., King’s 544 PLH) and orchardgrass (Dactylis glomerata L., King’s Echelon). To maintain the crop rotation in the full entry experiment, each crop entry was initiated with a different crop in the sequence in 2021. For ease, where crop entry is discussed, the abbreviations C-S-W, S-W-C, or W-C-S will be used to indicate the crop sequence, where C denotes corn, S denotes soybean, and W denotes wheat, listed in order of year in which the crop was grown in 2021, 2022, and 2023. The experiment was implemented in a randomized complete block design. Each treatment was replicated four times totaling 48 experimental plots, each measuring 6.1m by 48.7m (S Tables 1 and 2 in S1 File).

The four systems, or treatments, were designed as follows. System 1 used full inversion tillage before corn and soybean similarly to commercial organic grain farms in the Northeast US, while also integrating cover crops using reduced tillage methods. Winter wheat was sown in October following the use of a chisel plow to a depth of 15 cm at 202 kg ha^-1 (S Tables 2 and 3 in S1 File). In early March, medium red clover (Trifolium pratense L, Albert Lea Seed, VNS) was frost-seeded at a rate of 17 kg ha^-1 into winter wheat using a no-till grain drill. After wheat was harvested in July, the medium red clover was allowed to grow and then mowed to a height of 5 cm in early October. Cereal rye (Secale cereale L., Aroostook) was then no-till drilled at a seeding rate of 33 kg ha^-1 in mid-October into established red clover. In late May to early June, corn was planted at 86,000 seed ha^-1 into a seedbed created using a moldboard plow set to a depth of 20 cm. A cover crop mixture of annual ryegrass (Lolium perenne spp. multiflorum, Kodiak), forage radish (Raphanus sativus, Organic Tapmaster), and crimson clover (Trifolium incarnatum, Dixie) was interseeded into standing corn (V4 vegetative growth stage) after last cultivation at 28 kg ha^-1 using a high-clearance, no-till grain drill. The interseeded cover crop mixture was allowed to growth through the following fall and spring, and in late May soybean was planted 590,000 seed ha^-1 into a seedbed created using a moldboard plow set to a depth of 20 cm.

System 2 used shallow tillage with a compact high-speed disk (HSD), intended to reduce the intensity and depth of soil disturbance before each cash crop and to understand the effects of the HSD on multiple soil health and agronomic performance indicators. The HSD uses shallow non-inversion tillage that mixes the soil to a depth of approximately 10 cm. The number of HSD passes necessary to achieve an adequate seedbed differed by crop and year, depending on soil and crop residue conditions. The HSD was used before all cash crop plantings in System 2. Wheat was planted in late October at a seeding rate of 202 kg ha^-1. After wheat harvest in July, a cover crop mixture of oat (Avena sativa L., Jerry), forage radish, and Austrian winter pea (Pisum sativum L., Albert Lea Seed, VNS) was drill-seeded at a seeding rate of 22 kg ha^-1, 1 kg ha^-1, 32 kg ha^-1, respectively. Corn was planted in late May or early June, at a rate of 86,000 seed ha^-1. Soybeans were planted in late May or early June following high-speed disk tillage at a rate of 590,000 seed ha^-1. After corn harvest, cereal rye was seeded at a rate of 67 kg ha^-1.

System 3 was a reduced tillage system intended to reduce disturbance to the extent possible in an organic grain rotation and used no-till planting in soybean and relay cover cropping practices in wheat. System 3 had the lowest tillage intensity among the annual cropping systems, and the longest period between primary tillage events. A moldboard plow was used before planting corn at a seeding rate of 86,000 seed ha^-1. This was the only inversion tillage event in the 3-yr crop rotation. Cereal rye was drill seeded at a seeding rate of 135 kg ha^-1 following harvest of corn. The cereal rye was terminated using a roller-crimper in the following spring before no-till planting soybean at a rate of 590,000 seed ha^-1 into the mat of roll-crimped cereal rye. In October, wheat was drilled at a seeding rate of 202 kg ha^-1 following a single pass with the HSD set at 5–10 cm depth. In March, medium red clover was frost drill-seeded into wheat at a rate of 16 kg ha¹.

System 4 used perennial alfalfa-orchardgrass as a minimal soil disturbance baseline to compare to the three annual cropping systems. In System 4, the alfalfa (29.1 kg ha^-1])–orchardgrass (8.9 kg ha^-1) mixture followed wheat harvest and chisel plowing and disking. Because the experiment was full entry, and alfalfa followed wheat in the rotation, the number of alfalfa plots differed by year. In the crop rotation entry that started with corn in 2021, there were four alfalfa plots. In 2022, an additional four plots were planted to alfalfa, and in 2023, four more plots were planted to alfalfa to total 12 alfalfa plots (S Tables 2 and 3 in S1 File).

Soil disturbance

We used the number of field operations and a soil disturbance rating (SDR), from the USDA Natural Resources Conservation Service, to represent the frequency and intensity of soil disturbance from machinery for each system [45]. The six components of the SDR include soil inversion, soil mixing, soil compaction, soil shattering, soil lifting, and soil aeration. Each component of the SDR is assigned an intensity rating that ranges from 0 (least intense) to 5 (most intense) for each field operation. The total rating for an operation is the sum of the six components. Therefore, SDR ranges from 0, which is the least intense level of disturbance from a field operation, to 30, which is the most intense level of disturbance. The tillage implement with the highest SDR was the moldboard plow, with a rating of 29. The high-speed disk was assigned a rating of 24. No-till planting had an associated SDR value of 5.

The number of operations and SDR values for each system were summed starting on January 1 of each year and accumulated through December 31 to provide an annual rating (Table 1). Annual number of operations and SDR were accumulated across each year from 2021 through 2023 to provide estimates of disturbance across the crop sequence [rotation] in a system. To determine the effects of in-season and rotation number and intensity on M. robertsii during a year, we accumulated the numbers of disturbances and SDR that occurred before each soil sample from January 1 of the sample year and from January 1, 2021, respectively. We included number of days from the most recent field operation as a factor in analyses to determine the effects of the time elapsed since disturbance because we considered that recency of disturbance could affect observed relative prevalence.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Annual and accumulated number of disturbances and soil disturbance ratings (SDR) by crop entry, year, crop and experimental system.

https://doi.org/10.1371/journal.pone.0317149.t001

Multiple consecutive passes using the same equipment was common during the experiment. For example, in System 2 multiple passes with the high-speed disk were sometimes needed to sufficiently incorporate cover crop residue. These multiple passes can account for the greater annual and rotation numbers of disturbances and SDR in System 2 than in System 1 (inversion), even though the HSD has a lower SDR than the moldboard plow used in System 1. System 2 was designed to represent an intermediate level of disturbance. Even though management in System 2 sometimes resulted in the greatest overall SDR, we still considered it to represent an intermediate level of disturbance due to the shallow non-inversion action of the HSD. The multiple consecutive uses of the same implement within the same field operation may or may not have similar effects on biological communities as a single use [46]. Therefore, while the number of disturbances and their associated SDR are useful measures for quantifying and comparing the relative levels of disturbance across systems, we acknowledge that there are challenges to its use and interpretation that remain to be addressed through research.

Relative prevalence of M. robertsii

To assess the effects of system on the relative prevalence of M. robertsii, we conducted a standard soil bioassay using larval insects as a sentinel bait [47]. We collected pre-plant (April) and pre-harvest (September) composite soil samples comprised of 11 cores with a soil probe (2.5 cm x 20 cm) from all treatment plots. We collected post-plant (June) samples from plots in the corn and soybean phases of the rotation in Systems 1–3 in 2021, 2022, and 2023 (S Table 3 in S1 File). In the laboratory, we homogenized the composite sample from each treatment plot by crumbling soil clumps and removing plant residue and rocks. From each sample, we removed two, 250 ml subsamples for soil assays to detect entomopathogenic fungi (EPF) and one 250 ml subsample for soil fertility analysis.

We transferred the two, 250 ml subsamples for the entomopathogen assays to two 500 ml plastic deli containers (12 cm x 8.5 cm). Previous studies have found that results of sentinel bait assays to detect entomopathogenic fungi could vary by insect used [16]. Therefore, we conducted the assays concurrently and separately using last instar T. molitor L. (Coleoptera:Tenebrionidae) and last instar G. mellonella L. (Lepidoptera:Pyralidae). Fifteen last instar G. mellonella or 15 last instar T. molitor were added to each assay arena. We stored the prepared assay arenas at room temperature in the dark for ten days, after which, we retrieved the larvae from the soil for assessment. We classified the larvae as alive, dead from causes other than EPF, or potentially infected with M. robertsii and other EPF based upon morphological characteristics [48]. We recorded and then discarded larvae that were alive or dead from other causes. We rinsed potentially infected larvae with tap water to remove soil particles, then rinsed with 80% ethanol and then tap water. We then placed the larvae in humid chambers consisting of plastic Solo® condiment cups (7 cm x 3 cm) with a small piece of Whatman No. 1 filter paper to maintain humidity and allow for sporulation. After 7–10 days at room temperature, we initially determined infection by M. robertsii by morphology and the presence of characteristic green spores on the surface of the insect cadavers. We confirmed the identity of fungi initially identified as M. robertsii by sequencing the translation elongation factor-1 alpha (TEF-1α) of a random subset of isolates from infected G. mellonella and T. molitor from each System by the methods of Kepler et al. [49] (S Table 4 in S1 File). We observed infection by Beauveria Vuill. (Hypocreales: Cordycipitaceae) in only four assay insects among the 15,120 insects used in assays during the experiment. Therefore, we excluded Beauveria from further consideration. We did not detect any other EPF with the assay method used.

Soil properties

We submitted a 250 ml subsample of soil from each plot to the Agricultural Analytical Services Laboratory of The Pennsylvania State University (University Park, PA) for analysis of the following characteristics: proportions of sand, silt, and clay; pH, electrical conductivity, salts, phosphorus (P), potassium (K), magnesium (Mg), calcium (Ca), cation exchange capacity (CEC), soil organic matter by loss-on-ignition (SOM-LOI), and the trace elements zinc (Zn), copper (Cu), and sulfur (S). We used 5 g of each soil sample to determine the concentration of permanganate oxidizable carbon (POX-C) concentrations as an indicator of labile, biologically active soil carbon [50]. We used a 500 ml subsample to determine gravimetric moisture [51] and soil matric potential [52].

On two dates (4 October 2022, 11 October 2023), we measured soil compaction in each plot using a digital recording penetrometer (CYNST, Anhui East Electronic Technology Co., Ltd.) to a maximum depth of 45.7 cm. We measured compaction in three random locations in each plot and averaged the values to produce mean compaction value for the plot. Penetrometer data included: compaction in mPa in 2.54 cm increments to a maximum depth of 45.7 cm, maximum depth of penetrometer readings, and depth of maximum compaction.

Statistical analysis

In univariate analyses, we used generalized linear mixed-effect models to analyze relative prevalence of M. robertsii, expressed as the proportion of T. molitor and G. mellonella on which the fungus was detected [53]. We compared the null random-effects only model to fully fitted models containing main effects and interactions by adding one effect at a time. The function “anova” [54] was used to determine significance of fixed main effects and interactions using log-likelihood ratio tests and the Wald χ² test statistic. To meet assumptions of normality, equality of variances, and to reduce heterogeneity of variances, we utilized distribution options in R 4.2.2 [54]. We transformed percentages and proportions using square root arcsine transformation. The generalized linear mixed effects models were created with the function “glmmTMB” in the package glmmTMB [55] or used binomial distributions with the function “weights” set to the total number of G. mellonella or T. molitor used in each plot to account for the proportion of sentinel insects infected.

When models were significant, we used the package emmeans [56], to compare system effects with Tukey HSD post hoc tests. To determine mean prevalence for each individual grouping [by cash crop, system], we used the function “group_by” in the package dplyr [57]. Where the model was significant, we conducted pairwise comparisons with a Tukey HSD post hoc test in the package emmeans [56]. We considered the results of analyses significant at p < 0.05. Untransformed data are presented in the tables and figures.

We conducted data analyses to answer the following questions:

1. Did the relative prevalence of M. robertsii differ among systems?
   The relative prevalence based on the proportions of T. molitor and G. mellonella infected by M. robertsii in bioassays of soil were modeled separately using a generalized linear mixed effects model. System, entry point (C-S-W, S-W-C, W-C-S), year, and interactions between system and entry point, system and year, and entry point and year, and the three-way interaction between, system, entry, and year were used as fixed effects. To account for non-independence, random effects included assay date nested within plot, within entry, within block. This accounted for repeated measures of the same plot three times in each year over three years and block effects. We report statistical values only for significant effects in the results.
2. Did the relative prevalence of M. robertsii based on sentinel T. molitor and G. mellonella differ?
   We compared the proportions of M. robertsii-infected T. molitor and G. mellonella using a generalized linear mixed effects model. The proportion of infected T. molitor was set as the response variable and proportion of infected G. mellonella was set as a predictor variable. Year and system were set as fixed effects. Random effects included assay date nested within year, within crop, within block to account for repeatedly measuring the same plot three times per year and to account for non-independence.
3. Did the legacy of differences in tillage used in soybeans in 2021 affect the relative prevalence of M. robertsii in corn in 2023?
   The main tillage treatments were implemented in the soybean phase of the experiment while management of the wheat and corn phases were similar among the systems (S Table 3 in S1 File). To determine the presence of a legacy effect due to management differences in soybean on the relative prevalence of M. robertsii in corn, we used the function “glmmTMB” in the package glmmTMB [55] with data from sentinel assays of soil from soybean plots in 2021 and corn plots in 2023. We ran two generalized linear mixed effects models. Both models included assay date nested within block as the random effect to account for non-independence and repeated sampling, and system as the fixed effect. Only data from 2023 was analyzed to determine potential legacy effects from previous management in the system for the complete 3-year crop sequence. When the effects of system were significant for the relative prevalence of M. robertsii, we used a Tukey pairwise post hoc test to compare the legacy effect of system with the function “emmeans” from the package emmeans [56].
4. Did the environmental conditions associated with each system affect the prevalence of M. robertsii?
   To understand the relationship between disturbance and environmental variables associated with relative prevalence of M. robertsii in each system, we used forward selection stepwise multiple regression in JMP®, Version 17 (SAS Institute Inc., Cary, NC, USA). The initial pool of disturbance and soil variables included: the day of year that samples were collected and number of days since a field operation (disturbance); the number of disturbances and SDR accumulated from January 1 of the sample year until the date of soil sample (annual disturbance); the number of field operations and SDR accumulated from the beginning of the experiment until the date of soil sample (rotation disturbance); soil moisture (gravimetric soil moisture and matric potential); soil pH; electrical conductivity and salt content; cation exchange capacity; percent organic matter by loss-on-ignition and POX-C; P, K, Mg, Ca, zinc, copper and sulfur content; and percent sand, silt and clay. A similar multivariate analysis was conducted for the dates on which soil compaction data was available, limiting the measures of the relative prevalence of M. robertsii, disturbance, and soil properties to the dates that corresponded most closely to those on which compaction was measured. The final models were those with the lowest Akaike information criterion. We tested for multicollinearity of variables using pairwise correlations among the disturbance and soil variables. When a significant correlation between variables was detected, the variable with the greatest contribution (greatest F value, lowest p value) to variation in percentage of insects infected by M. robertsii was retained for use in further analyses to understand the combined effects of disturbance and environmental factors on the relative prevalence of M. robertsii.

To visualize the relationships between the systems, crop and environmental factors we used principal components analysis (PCA) in CANOCO for Windows version 5.0 [59]. The initial pool of disturbance and soil variables included: the number of disturbances and SDR accumulated from January 1 to December 31 of the sample year [annual disturbance]; the number of field operations and SDR accumulated from the beginning of the experiment until December 31 of the sample year (rotation disturbance); mean soil moisture (matric potential); soil pH; electrical conductivity and salt content; cation exchange capacity; percent organic matter by loss-on-ignition and POX-C; P, K, Mg, Ca, zinc, copper and sulfur content; and percent sand, silt and clay. Only environmental variables with at least 20% fit to the ordination space are visualized on biplots [59].

Relative prevalence of M. robertsii

Based on morphological and molecular methods, all fungal isolates from the sentinel insect assays were confirmed as M. robertsii [49]. These results are consistent with extensive sampling of soil from this site over the last 10 years [24,25,34].

1. 1) Did relative prevalence of M. robertsii differ among systems?

Based on sentinel assays with G. mellonella, system had a significant effect on the relative prevalence of M. robertsii (χ² = 12.61, df = 3, p = 0.0056). Relative prevalence was significantly lower in System 2 than in Systems 1 (z = 3.04, p = 0.0127) and 3 (z = -2.80, p = 0.0261). Mean relative prevalence was 22.30 ± 2.19%, 14.30 ± 1.72%, 21.60 ± 2.17%, and 15.80 ± 2.06% in Systems 1, 2, 3, and 4, respectively (Fig 1 and Table 2). Based on sentinel assays with T. molitor, system had a significant effect on the relative prevalence of M. robertsii (χ² = 11.94, df = 3, p = 0.0076). Similar to assays using G. mellonella, relative prevalence was significantly lower in System 2 (z = 3.489, p = 0.0027) than in System 1. Relative prevalence was 34.00 ± 2.43%, 23.50 ± 2.28%, 30.70 ± 2.53%, and 28.20 ± 2.28% in Systems 1, 2, 3, and 4, respectively (Fig 1 and Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Proportion of mealworms (T. molitor, left panel) and waxworms (G. mellonella, right panel) infected by M. robertsii by system, entry, and year.

Red circles represent System 1 (inversion tillage), green triangles represent System 2 (high-speed disk), and blue squares represent System 3 (no-till planting). Entries include crop sequences of corn-soybean-wheat (C-S-W), soybean-wheat-corn (S-W-C), and wheat-corn-soybean (W-C-S) in 2021, 2022, and 2023, respectively. Error bars represent the 95% confidence level.

https://doi.org/10.1371/journal.pone.0317149.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Mean infection rates of G. mellonella and T. molitor by M. robertsii in each cash crop by system.

System had a significant effect for infection rates of G. mellonella (P = 0.0056) and T. molitor (P = 0.0076). Mean infection rates for the interaction of system and crop were not significantly different.

https://doi.org/10.1371/journal.pone.0317149.t002

Year had a significant effect on relative prevalence of M. robertsii in sentinel assays with T. molitor (χ² = 14.94, df = 2, p = 0.0005), but not G. mellonella. Relative prevalence in 2021 was greater than in 2022 (z = 2.45, p = 0.0377). Relative prevalence was significantly lower in 2022 than 2023 (z = -3.82, p = 0.0040). No other differences were detected between years. Mean relative prevalence of M. robertsii based on T. molitor infection rate was 28.9 ± 1.90%, 24.60 ± 2.32%, and 33.70 ± 2.18% in 2021, 2022, 2023, respectively (Fig 1 and Table 2). All other pairwise comparisons, including crop and crop within system, were not significant for either insect species.

1. 2) Did relative prevalence of M. robertsii based on sentinel T. molitor and G. mellonella differ?

The estimation of relative prevalence based on mean infection rates of T. molitor was greater than estimations of prevalence based on G. mellonella (χ² = 8.43, df = 1, p = 0.0037). The mean relative prevalence based on T. molitor was 29.16 ± 1.26%, whereas relative prevalence based on G. mellonella was 18.73 ± 1.04% (Table 2).

1. 3) Did the legacy of differences in tillage used in soybeans in 2021 affect the relative prevalence of M. robertsii in corn in 2023?

In corn in 2023, soybean management in Systems 1, 2, or 3 did not have a significant effect on the relative prevalence of M. robertsii based on infection rates of T. molitor (χ² = 3.31, df = 2, p = 0.1909). However, system had a significant effect on the relative prevalence of M. robertsii based on infection rates of G. mellonella (χ² = 14.32, df = 2, p = 0.0008). Relative prevalence in System 1, in which inversion tillage was used in soybean in 2021 and in corn in 2023, was significantly greater than in System 2, which used the HSD in soybean and inversion tillage in corn (z = 3.326, p = 0.0025). Similarly, relative prevalence in System 3, which used no-till planting of soybean and inversion tillage in corn, was greater than in System 2 (z = -3.11, p = 0.0053). Relative prevalence in System 1 was not significantly different than in System 3 (z = 0.211, p = 0.9757). Based on sentinel assays with G. mellonella, relative prevalence of M. robertsii in corn in 2023 was 24.05 ± 5.79%, 11.46 ± 5.79%, and 23.34 ± 5.79% in Systems 1, 2, and 3, respectively.

1. 4) Did environmental conditions associated with each system affect the relative prevalence of M. robertsii?

Principal components analysis explained 12.9% of the adjusted explained variation in disturbance indicators and soil properties among the system-crop combinations (Fig 2). Axis 1 explained 16.61% of the adjusted explained variation and represents level of disturbance and was related to soil organic matter content, cation exchange capacity, and Ca and Mg content. Axis 2 explained 15.84% of the explained variation salts and Zn. The centroids for the corn phase in Systems 1, 2, and 3, which were all planted following inversion tillage, clustered in one quadrant of the ordination space and were associated with high levels of disturbance and soil P, Zn, and salts. The centroids for soybean in System 3, managed with no-till planting, and wheat in System 3, which followed no-till planting in the soybean phase, clustered with System 4, perennial alfalfa, which experienced soil disturbance only when planted, and were associated with soil pH and low disturbance. The centroids for the soybean phase in Systems 1 (inversion tillage), 2 (shallow high-speed disk), and 3 (no-till planting) were dispersed across the ordination space according to level of disturbance. The centroid for System 1 soybean, managed with inversion tillage was associated with the soil disturbance rating accumulated across the three-year rotation.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Biplot from PCA visualizing the relationship between system, crop, and environmental variables.

The interaction of system*crop accounted for 12. 9% of the adjusted variation in disturbance indicators and soil properties. Axes1 and 2 represents 16.61% and 15.84% of the explained variation, respectively. Only environmental variables with at least 20% fit to the ordination space are shown. CEC = Cation exchange capacity, Rotation SDR = Soil disturbance rating accumulated through the three-year rotation, SOM = Soil organic matter (%).

https://doi.org/10.1371/journal.pone.0317149.g002

The associations between environmental variables and estimated relative prevalence of M. robertsii varied between the two sentinel insect species (Table 3). Across all systems, G. mellonella produced lower estimates of relative prevalence of M. robertsii than T. molitor, but the association with environmental variables was greater for G. mellonella. Across all systems, environmental variables explained 28.4% of the variation in estimated prevalence of M. robertsii in assays using G. mellonella, but only 4.8% of the variation in assays using T. molitor. Five environmental variables were significant predictors of the relative prevalence of M. robertsii by G. mellonella and T. molitor. Three predictors—soil moisture, day of year of assay, and soil S content, were common to both sentinel species, although all three were in opposite direction from the other species.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Environmental factors contributing to the variation in relative prevalence of M. robertsii in assays with G. mellonella and T. molitor, determined by of stepwise forward selection multiple regression across all systems and in each of the four cropping systems.

https://doi.org/10.1371/journal.pone.0317149.t003

The associations of M. robertsii with disturbance indicators and soil properties varied among systems and insect species. In assays using G. mellonella, environmental variables explained 35.8%, 28.7%, 10.6% and 32.4% of variation in estimated relative prevalence of M. robertsii in Systems 1, 2, 3, and 4, respectively (Table 3). The day of year that the soil was collected and assayed was a significant positive predictor in all four systems, indicating that prevalence of M. robertsii increased over the growing season. Soil moisture was a significant predictor in Systems 2 (HSD) and 4 (alfalfa), indicating greater prevalence with increases in soil moisture. Other soil properties and disturbance measures associated with relative prevalence of M. robertsii but unique to single systems included: a negative predictor, CEC, and two positive predictors, POX-C and percent clay, in System 1; soil Ca and Cu content as negative predictors in System 2; SDR as a positive predictor in System 3; and soil S and pH as positive predictors, and days since disturbance as a negative predictor, in System 4 (Table 3).

In assays using T. molitor, environmental variables explained 22.1%, 18.5%, 13.5%, and 2.8% of variation in estimated relative prevalence of M. robertsii in Systems 1, 2, 3, and 4, respectively (Table 3). Soil S content was a negative predictor in Systems 2 and 3. The direction of associations, positive or negative, differed among systems for some shared predictors. For example, POX-C was a negative predictor in Systems 1 and 2 and a positive predictor in System 3. Indicators of disturbance were significant predictors in Systems 1 and 2 where annual SDR was a positive predictor, while the number of disturbances was a negative predictor in System 1. No disturbance indicators or soil properties were associated with relative prevalence of M. robertsii in System 4.

Soil compaction at 2.5 to 10.2 cm depth differed significantly among systems at 2.5 cm (F_3,35.7 = 3.66, p = 0.0213), 5.1cm (F_3,40 = 3.96, p = 0.0145); 7.6cm (F_3,36.3 = 12.95, p < 0.0001), and 10.2 cm (F_3,36.3 = 8.48, p = 0.0002) (Table 4). In soybean, the crop phase in in which the main difference in soil management was imposed, compaction in Systems 2 was greater on the two dates measured than in Systems 1 and 3 at 5.1, 7.6, and 10.2 cm depths, but was not different from compaction in System 4 (alfalfa) (Table 4). No measures of compaction were significantly associated with prevalence of M. robertsii in assays with either G. mellonella or T. molitor on the two dates on which compaction was measured.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Mean soil compaction in soybean in systems 1–3 and alfalfa (System 4).

https://doi.org/10.1371/journal.pone.0317149.t004

System effect

We quantified the effect of three annual and one perennial cropping systems that varied in intensity and depth of soil disturbance during a three-year crop rotation on the prevalence of entomopathogenic fungus, M. robertsii. We hypothesized that tillage frequency and intensity, as estimated by the SDR, and depth of soil disturbance would negatively affect the relative prevalence of M. robertsii. We expected that System 1, which used inversion tillage in the soybean and corn phases of the rotation, would result in the lowest relative prevalence of M. robertsii. Use of the moldboard plow to a depth of 20 cm in all crop phases was associated with the greatest SDR of the tillage tools utilized. However, we detected greater relative prevalence of M. robertsii by both sentinel insect species in System 1, which used a moldboard plow in corn, chisel plow in wheat, and no-till planting in soybean, compared to System 2, which utilized a high-speed disk set at a depth of 5 to 10 cm in soybeans and wheat, and a moldboard plow in corn. We expected that System 3, which utilized no-till planting in the soybean phase, a high-speed disk in wheat, and a moldboard plow in the corn phase, would have a greater relative prevalence of M. robertsii than Systems 1 or 2, as System 3 experienced less soil disturbance than Systems 1 or 2. However, because multiple passes were needed to manage plant residues with the high-speed disk, System 2 experienced greater frequency and intensity (SDR) of disturbance than the other three system, although to a shallower depth compared to System 1 (Table 1). Therefore, unexpectedly, the relative prevalence of M. robertsii was, on average, lower in System 2 than in Systems 1 and 3, where we observed similar relative prevalences of M. robertsii.

In assays with G. mellonella, relative prevalence of M. robertsii in System 4, which experienced the lowest frequency and intensity of disturbance, was low. Other studies suggest that some level of soil disturbance benefits the detection of Metarhizium spp. in agricultural soil. In a survey of Metarhizium spp. in soils sampled from a long-term experiment in the Mid-Atlantic region, organically managed plots managed with a chisel plow harbored greater numbers of Metarhizium colony-forming units than no-till plots [49]. Another study reported greater prevalence of Metarhizium spp. in an organic feed grain rotation in inversion tillage treatments compared to reduced tillage treatments [33]. In that study, the authors suggested that the detection of Metarhizium was facilitated by the mixing of spores throughout the soil profile in disturbed systems compared to the reduced tillage systems where fungal spores would remain relatively aggregated around infected insects. Our findings support the hypothesis proposed by Jabbour and Barbercheck [33] in that the reduced tillage system did not provide the soil mixing and distribution of fungal spores throughout the soil profile that the inversion tillage system provided, therefore resulting in lower detection of M. robertsii in System 4 than Systems 1 and 3. We hypothesized that System 2 (HSD) would have an intermediate prevalence of M. robertsii compared to System 1 (inversion) or 3 (reduced tillage). However, among the three annual cropping systems, we observed the lowest relative prevalence of M. robertsii using both G. mellonella and T. molitor in System 2 (HSD) compared to System 1. We suggest that that some level of soil mixing benefits the detection of M. robertsii, while a greater level of disturbance reduces prevalence, especially in systems with relatively low disturbance, such as Systems 3 (no-till planting in soybean) and 4 (perennial alfalfa). We suggest that the greater frequency and intensity of disturbance associated with System 2 resulted in the observed lower prevalence of M. robertsii in that system. Additionally, the shallower disturbance of the soil by the HSD compared with the moldboard plow may have also contributed to the lower detection of M. robertsii in System 2, as a greater volume of soil would have been mixed in System 1 compared to System 2 [33].

We expected that the greatest prevalence of M. robertsii would occur in System 4 (alfalfa) as it was managed as a perennial forage crop that experienced the least frequent and intense level of disturbance compared to Systems 1, 2, and 3, in which annual crops were grown. However, there were no significant difference in mean relative prevalence associated with crop alone. Another study in the mid-Atlantic US similarly found greater numbers of colony forming units (CFU) of Metarhizium spp. under tilled soybean compared with alfalfa [49]. We suggest that M. robertsii is well-adapted to the range of levels of disturbance in agricultural systems and can tolerate a wide range of environmental conditions due to its diverse roles as a generalist entomopathogen, saprophyte and endophyte [63–65].

Legacy effect

We hypothesized that the legacy of the tillage management in soybean in 2021 would affect prevalence of M. robertsii in corn in 2023. In System 1, soybeans were managed with inversion tillage in 2021, whereas in System 3, soybeans were managed with no-till planting. This difference in soybean management provided an opportunity to compare the legacy of no-tillage and inversion tillage in soybeans on the corn phase of the rotation. Wheat, the crop following soybeans and preceding corn in 2022, was managed with non-inversion tillage in all systems, including the use of the chisel plow in System 1 and the HSD in Systems 2 and 3. Corn in both Systems 1 and 3 were managed with inversion tillage. We hypothesized we would measure greater relative prevalence of M. robertsii in corn after no-till soybean than in corn after soybeans managed with inversion tillage due to the potential negative legacy effects of soil disturbance. However, we did not detect legacy effects of soil disturbance on prevalence of M. robertsii in these cropping systems, suggesting that in-season management has a greater effect on the relative prevalence of M. robertsii than does management legacy, at least within a single three-year rotation.

Sentinel insects

We found that relative prevalence based on infection rates of G. mellonella by M. robertsii, was lower than that for sentinel T. molitor. Both sentinel species showed the same general trend in prevalence among systems with lowest relative prevalence in System 2, but no significant differences among crops. In a survey of vineyard and semi-natural habitat soils, Sharma et al. [16] reported that T. molitor was more likely than G. mellonella to become infected with M. robertsii [16]. The authors found that T. molitor and G. mellonella worked well together to estimate relative prevalence of M. robertsii and suggested that using both species as sentinel larvae would provide a greater understanding of the distribution of entomopathogenic fungi than either one alone. Similarly, Vänninen et al. [66] observed high sensitivity of T. molitor to M. anisopliae in assays of agricultural soils.

Environmental effects

The four management systems in our study resulted in different soil characteristics and levels of disturbance (Table 1 and Fig 2). Even though our estimates of relative prevalence of M. robertsii using G. mellonella were lower than in assays using T. molitor, the relative prevalence determined from assays using G. mellonella was more closely associated with particular soil properties (Table 3). We identified all Metarhizium in this study and in previous studies conducted at this site since 2010 as M. robertsii by morphological characteristics and translation elongation factor 1-alpha (TEF 1-α) sequencing [24,25,34,49,67].

We used a commonly employed assay method based on infection of sentinel insects to detect relative prevalence of M. robertsii. Using this method, and consistent with our previous results [33,34], we found that M. robertsii was the overwhelmingly dominant EPF at our site. Others have used culture-based methods alone or in combination with sentinel insect assays [e.g., 32,49,64,65,68–70] to investigate the prevalence and diversity of EPF in soil and detected greater diversity of EPF than we detected in our study site. It is possible that if we had used additional types of assays, we may have detected a greater diversity of EPF species.

In an assessment of the Metarhizium community in soil from a single agricultural field and surrounding hedgerows in Denmark using T. molitor as a sentinel insect, Steinwender et al. [64,65] identified multiple genotypes of M. robertsii using multilocus simple sequence repeat typing. The authors suggested that multilocus genotypes (MLGs) of M. robertsii may have originated from multiple immigration events or that the multiple genotypes represented a long evolutionary history of local lineages at the site. The authors concluded that agricultural practices appear to allow for co-occurrence of MLGs of multiple Metarhizium species and that knowledge of the factors responsible for the dominance of particular species or genotypes could be used to either improve conditions for frequently occurring MLGs or to enhance conditions for the addition of rarer MLGs [64,65]. Similarly, Kepler et al [49], using multilocus microsatellite genotyping, found significant genotypic diversity among M. robertsii isolates, which fell into multiple MLGs in two clades. Those authors [49] suggested that the multiple levels of diversity could be used to inform strategies by which Metarhizium populations in soil could be manipulated to suppress diverse pest organisms and promote plant health. In our study, we did not identify MLGs of M. robertsii, so we do not know its genotypic diversity at this site. We speculate that due to its apparently greater susceptibility to M. robertsii, T. molitor may have “captured” a greater diversity of genotypes of M. robertsii, each with different tolerances to specific environmental conditions, than did G. mellonella, giving the appearance of non-responsiveness to environmental conditions specific to each crop management system. Our results suggest that the two sentinel insect species provided different but complementary information. T. molitor may have provided a better estimate of overall relative prevalence, but G. mellonella may have provided more information on the environmental tolerances of the particular genotypes infecting it.

Although we did not detect a significant effect of compaction on relative prevalence of M. robertsii on the two dates on which we measured compaction, soil in System 2 was more compacted in the surface 10.2 cm than in Systems 1 and 3 (Table 4). Compaction in the surface 10.2 cm in System 4 was not different from System 2 and mean relative prevalence of M. robertsii in assays with G. mellonella was low and similar in Systems 2 and 4. The mean compaction in System 2 at 7.6 and 10.2 cm depth exceeded 2.07 mPa, a level of compaction that can inhibit root growth [71]. In a laboratory experiment, Viera Tiago et al. [72] observed a decrease in colony forming units of Metarhizium sp. after accidental soil compaction occurred. We suggest that although we did not observe a significant relationship between compaction and relative prevalence of M. robertsii for the two corresponding dates on which sentinel assays and measurements of compaction were conducted, more observations and experiments focusing on the effects of soil compaction on relative prevalence are needed to better understand the effects of soil disturbance and compaction on those relationships.