Soil Chemistry and Soil History Significantly Structure Oomycete Communities in Brassicaceae Crop Rotations

ABSTRACT

Oomycetes are critically important in soil microbial communities, especially for agriculture, where they are responsible for major declines in yields. Unfortunately, oomycetes are vastly understudied compared to bacteria and fungi. As such, our understanding of how oomycete biodiversity and community structure vary through time in the soil remains poor. Soil history established by previous crops is one factor known to structure other soil microbes, but this has not been investigated for its influence on oomycetes. In this study, we established three different soil histories in field trials; the following year, these plots were planted with five different Brassicaceae crops. We hypothesized that the previously established soil histories would structure different oomycete communities, regardless of their current Brassicaceae crop host, in both the roots and rhizosphere. We used a nested internal transcribed spacer amplicon strategy incorporated with MiSeq metabarcoding, where the sequencing data was used to infer amplicon sequence variants of the oomycetes present in each sample. This allowed us to determine the impact of different soil histories on the structure and biodiversity of the oomycete root and rhizosphere communities from the five different Brassicaceae crops. We found that each soil history structured distinct oomycete rhizosphere communities, regardless of different Brassicaceae crop hosts, while soil chemistry structured the oomycete communities more during a dry year. Interestingly, soil history appeared specific to oomycetes but was less influential for bacterial communities previously identified from the same samples. These results advance our understanding of how different agricultural practices and inputs can alter edaphic factors to impact future oomycete communities. Examining how different soil histories endure and impact oomycete biodiversity will help clarify how these important communities may be assembled in agricultural soils.

IMPORTANCE Oomycetes cause global plant diseases that result in substantial losses, yet they are highly understudied compared to other microbes, like fungi and bacteria. We wanted to investigate how past soil events, like changing crops in rotation, would impact subsequent oomycete communities. We planted different oilseed crops in three different soil histories and found that each soil history structured a distinct oomycete community regardless of which new oilseed crop was planted, e.g., oomycete communities from last year’s lentil plots were still detected the following year regardless of which new oilseed crops we planted. This study demonstrated how different agricultural practices can impact future microbial communities differently. Our results also highlight the need for continued monitoring of oomycete biodiversity and quantification.

INTRODUCTION

Soil history established by previous plant-soil microbial communities conditions future generations (1–4). These communities are subject to various biotic and abiotic factors—including the initial soil chemistry and microbes present, changes in the plant communities, and environmental extremes, such as droughts—and subsequently reflect them through a plant-soil microbial feedback process (5–7). Plant hosts alter soil chemistry, first through their capacity to take up nutrients from the soil (8). Second, through rhizodeposition, the host plant can vary the quantity and array of compounds released into the rhizosphere as required, thereby changing the soil chemistry (9–12). Modifying a rhizodeposition profile also permits plants to actively tailor the structure of their microbial rhizosphere community in response to variable conditions and the plant’s needs (9–12). For example, soil microbes increase the host plant’s access to nutrients (12–14), temper environmental change (15) or stress (16, 17), and protect against pathogens (18, 19). Soil bacterial communities in particular help integrate these diverse signals and modulate the plant’s responses (17, 20).

As such, the plant-soil microbial community generates a reciprocal feedback process that incorporates various biotic and abiotic factors (5–7) and impacts future plant-soil microbial generations and their composition (1, 21, 22). Thus, information from one plant-soil microbial community is transmitted through time to impact subsequent plant-microbial generations, i.e., the soil history, also referred to as soil legacy, of previous plant-soil microbial communities conditions future ones (1–3). However, how different biotic and abiotic factors interact to establish soil histories are not well understood.

Drought, or water stress, for example, is an increasingly important abiotic factor for establishing plant-soil microbial communities, as well as its growing impact on global agricultural production (23). On the Canadian prairies, drought is a common experience. During the last event of the 2021 season, major crop production plunged nationally to a 15-year low due to severe drought conditions: wheat decreased by 38.5% to 21.7 M tons, while canola decreased by 35.4% to 12.6 M tons (24). Water availability is a key determinant of plant nutrition and performance, such that plant growth becomes restricted due to drought, as plants are constrained in nutrient uptake from the soil (22). Soil moisture is also critical for microbial communities (reviewed in reference 25); bacteria depend on water for nutrient diffusion and mobility (23). Soil moisture is also a key promoter of phytopathogenic oomycete growth and dispersion (5, 26, 27). As such, there is interest in investigating how soil history and plant-microbial communities interact under dry, water-stressed conditions in an agricultural setting.

Crop rotations, complete with their agricultural inputs, model how a previous crop plant establishes a soil history by altering the biotic and abiotic soil conditions for future plant-microbial communities (5–7). For example, when lentils, or other legumes, are introduced into a rotation, they shift the soil microbial community, which establishes more bioavailable nitrogen and moisture in the soil (6, 28–30). These benefits translate to the subsequent crops, which tend to have higher yields (28, 30).

Beneficial soil histories have also been established by canola (cultivars of Brassica napus, Brassica rapa, or Brassica juncea) rotations, as they are thought to reduce the growth of cereal-specific pathogens. As such, cereals tend to have higher yields when they are planted after canola (6, 31). As demand for vegetable oil and biofuels increases, Brassicaceae oilseed-based rotations are increasingly common throughout the world, such as the frequent canola-wheat rotation in Canada (6). Increasing Brassicaceae oilseed crop diversity has been ongoing in order to improve production by identifying and breeding crop plants better adapted to the drought stress of the Canadian prairies, as well as cultivars that are more resistant to pathogens (32–34). Though extensive work has investigated how Brassicaceae oilseed rotations benefit the crop plants involved (recently reviewed in reference 6), less is known how these crops and their agricultural inputs impact beneficial, or pathogenic, soil microbial communities.

Brassicaceae oilseed crops are known hosts to many microbial pathogens, including fungi such as Fusarium sp., Leptosphaeria sp. (blackleg), Rhizoctonia solani, and Sclerotinia sclerotiorum (stem rot) and oomycetes, like Pythium sp., Phytophthora sp., Albugo sp. (staghead), and Peronospora and Hyaloperonospora spp. (downy mildews) (35, 36). Oomycete phytopathogens alone have had an historically outsized impact on global agriculture (37). Although particular oomycete strains have been fairly well studied as model patho-systems with specific Brassicaceae species (37–41), as a whole soil oomycete biodiversity remains relatively understudied, notably among Brassicaceae crops (36).

A few high-throughput sequencing studies have yielded some clues concerning soil oomycete communities (36, 42); those studies illustrated oomycete communities present in agricultural and natural Brassicaceae soil samples, respectively. They identified root and rhizosphere communities dominated by Pythiales (Pythium sp. or Globisporangium sp.), with minorities of Peronosporales (Phytophthora sp., Peronospora sp., and Hyaloperonospora sp.) and Saprolegniales (Aphanomyces sp.), though those authors indicated substantial difficulties in assigning taxa (36, 42, 43). Taheri et al (2017) also yielded further insights by also identifying oomycete root and rhizosphere communities dominated by Pythiales from various agricultural pea fields across the Canadian prairies (43). Beyond these examples, there is a lack of baseline knowledge concerning how oomycete communities are structured around diverse Brassicaceae oilseed crops.

Crop rotations, and the soil histories they establish, highlight important factors to investigate in order to better understand oomycete community dynamics. First, different crops, along with their agricultural treatments, alter edaphic factors, which may have important consequences for oomycete communities. For example, legumes tend to retain soil moisture, which is a key factor for oomycete growth, as discussed previously (5, 26, 27). Second, crop rotations are known to shift bacterial communities, which may interact to suppress or promote oomycete communities (44). Both of these examples stress the impacts crop rotations might have on the subsequent crop plant-oomycete community.

We took advantage of an existing agriculture experiment to investigate the impact of soil history established by the previous year’s crop, and the current Brassicaceae oilseed host crops, on the soil oomycete communities. Three crop histories were established by growing wheat or lentils or being left fallow (see Fig. S1A in the supplemental material). The following year, each crop history plot was divided into five subplots and planted with a different Brassicaceae oilseed crop. At full flower, the root systems of these Brassicaceae host crops were harvested and divided into root and rhizosphere compartments, from which environmental DNA was extracted. We hypothesized that the three soil histories established by the previous crops would structure different oomycete communities, regardless of their current Brassicaceae host, in both the roots and rhizosphere. We used a MiSeq metabarcoding approach to specifically target the oomycete communities for each root and rhizosphere sample, where the sequencing data were used to infer amplicon sequence variants (ASVs) to identify the oomycete biodiversity and how this understudied group varies in agricultural soils.

RESULTS

Brassicaceae rhizospheres had significantly different oomycete communities compared to the roots.

To identify the composition of the oomycete rhizosphere and root communities from the test phase Brassicaceae crop species, we inferred ASVs from the retained oomycete-specific internal transcribed spacer (ITS) amplicons using the DADA2 pipeline (45, 46). The four replicates of the oomycete mock community were sequenced to an average of ~23,000 reads and closely resembled each other in ASV composition (Fig. S5B and C). From the retained MiSeq reads of the mock community, DADA2 inferred 316 individual ASVs which were assigned taxa to at least the order level as 2 Pythiales, 2 Peronosporales, 1 Saprolegniales, and 1 Albuginales (Fig. S5B). The mock community was composed of 21 individuals from these four orders (Table S2); this provided some reassurance that our pipeline was effective in identifying the oomycetes present in each experimental sample.

We retained 8,222,283 high-quality ITS MiSeq amplicons through the pipeline, with more reads retained in the rhizosphere samples of the test Brassicaceae crop species in both trials (Table 1; Fig. S5). The 1,037 ASVs inferred from the retained reads were subsequently filtered to 412 oomycete ASVs identified among the test phase samples. Differences between the rhizosphere and root test phase oomycete communities were highly significant in both field trials (trial 1, permutational multivariate analysis of variance [PERMANOVA] R² = 0.1226, P < 0.0001; trial 2, PERMANOVA R² = 0.0751, P < 0.0001). The majority of oomycete ASVs were found in the rhizosphere compared to their cognate root samples (Table 1). The oomycete rhizosphere communities were also consistently more phylogenetically diverse than the root communities (Fig. 1), which reflected the greater species richness observed in the rhizosphere (Fig. S7).

TABLE 1.

The oomycete rhizosphere communities had more unique ASVs than the root communities of five Brassicaceae host plants^a

Trial	Source	No. of retained ITS reads (mean ± SD per sample)	ASV occurrence (mean ± SD per sample)	ASV abundance (mean ± SD per sample)
Trial 1	Rhizosphere	2,178,736 (36,312 ± 12,084)	227 (69 ± 34)	1,353,787 (22,563 ± 15,585)
	Roots	1,785,905 (29,765 ± 11,416)	89 (63 ± 36)	81,396 (1,357 ± 4,627)
Trial 2	Rhizosphere	2,355,829 (39,263 ± 14,105)	218 (74 ± 40)	1,237,771 (20,629 ± 16,799)
	Roots	1,747,543 (29,126 ± 11,252)	43 (56 ± 39)	1,820 (33 ± 147)

^aCommunities were obtained from root communities of five Brassicaceae host plants in the test phase of a 2-year crop rotation, harvested from two field trials (trial 1, 2016; trial 2, 2017) from Swift Current, SK, Canada. A total of 17,656,076 raw reads were produced via Illumina’s MiSeq at Génome Québec and processed through DADA2, where 8,068,013 reads were retained (the ITS reads reported here) for ASV inference. A total of 1,037 ASVs were identified across the entire data set, which were filtered to 412 oomycete ASVs.

FIG 1.

Phylogenetic diversity tended to be higher in the oomycete rhizosphere communities (A, B, E, and F) than in the root communities (C, D, G, and H) from five Brassicaceae host plants in the test phase of a 2-year rotation, harvested in 2016 (trial 1, A to D) and 2017 (trial 2, E to H) from Swift Current, SK, Canada. Phylogenetic diversity also tended to be higher in the Brassicaceae host plants in trial 2 (E and G) than trial 1 (A and C). Diversity tended to be higher in the root communities in Camelina sativa compared to the corresponding rhizosphere communities but only in trial 2 (G versus E). As the transformed data did not adhere to assumptions of normality, the nonparametric Kruskal test was used to test for significance among the test phase oomycete communities grouped by Brassicaceae host crops, or by their conditioning phase soil histories. No significant differences were detected.

When the oomycete ASVs were plotted as taxa clusters, we observed similar taxonomic composition between the oomycete test phase rhizosphere and root communities from trials 1 and 2 (Fig. 2). Pythium species dominated the roots and rhizosphere in both trials in terms of relative abundance, while the order Peronosporales consistently had the most taxa across each community (Fig. 2). In trial 1, Pythium and Peronospora genera were significantly enriched (P < 0.01) in the test phase rhizosphere communities compared to the roots, while in trial 2 only Pythium species were significantly more abundant in the rhizosphere (Fig. S8).

FIG 2.

Taxa clusters of the oomycete ASVs inferred from among the rhizosphere and roots of five Brassicaceae host crops in the test phase of a 2-year rotation from Swift Current, SK, Canada. The abundance and composition of the oomycete communities are represented to the genus level, where the size of each taxonomic group (bubble) represents the abundance of inferred ASVs and the color scale represents the number of unique taxa. (A) In trial 1 (harvested 2016), Pythium dominated in both the rhizosphere (left) and root (right) communities, while the genera Aphanomyces, Lagena, and Peronospora were dramatically reduced between the rhizosphere and the root communities. (B) Significantly enriched taxa, labeled in bold, were identified between each pair of Brassicaceae host crops in trial 1 rhizosphere (top panel) and root (bottom panel). Taxa that were significantly more abundant are highlighted brown or green, following the labels for each compared factor. Differential taxa clusters were identified to be significantly enriched (i.e., abundant) taxa, using the nonparametric Kruskal test, followed by the post hoc pairwise Wilcox test, with an FDR correction. No enrichment was detected for trial 2, nor did soil histories enrich any taxa in either experiment. (C) In trial 2 (harvested 2017), similar to the results shown in panel A, Pythium dominated in both the rhizosphere and root communities, although there were very few oomycete ASVs detected in the Brassicaceae root communities.

Indicator species analysis identified oomycete ASVs specific to the test phase rhizosphere communities in both trials, but none were identified in the root communities. Forty-one ASVs were specific to the oomycete rhizosphere communities in trial 1 (P < 0.005) (Table 2), while no ASVs were specific to the root communities. Thirty-four ASVs belonged to the Pythiaceae, of which half were further identified as Pythium sp.; two ASVs were Lagena sp., and one was Aphanomyces sp. (Table 2). The final four indicator ASVs were unknown oomycetes (Table 2). In trial 2, indicator ASVs in the rhizosphere communities were similar to those from the rhizosphere of trial 1; 45 ASVs were specific to the test phase rhizosphere communities in trial 2 (P < 0.05) (Table 2), whereas none was identified in the cognate root communities. The Pythiaceae accounted for 37 of these ASVs, of which 19 were further identified as Pythium sp. (Table 2). Two ASVs were recognized as Lagena sp. and one was Aphanomyces sp., while the remaining five ASVs were unknown oomycetes (Table 2). All of the oomycete ASVs, including those identified by indicator species analysis as significant, were matched to plant pathogen functional lifestyles.

TABLE 2.

Indicator species were identified exclusively in the oomycete rhizosphere communities^a

Closest taxon	Trial 1	Trial 2
By compartmentb
Aphanomyces sp.	Rhizosphere	Rhizosphere
Lagena sp.	Rhizosphere, 2 ASVs	Rhizosphere, 2 ASVs
Pythium sp.	Rhizosphere, 17 ASVs	Rhizosphere, 19 ASVs
Pythiaceae	Rhizosphere, 17 ASVs	Rhizosphere, 18 ASVs
Unknown oomycetes	Rhizosphere, 4 ASVs	Rhizosphere, 5 ASVs
By soil historyc
Pythium sp.	Rhizosphere/lentil
Pythiaceae	Rhizosphere/lentil, 4 ASVs	Rhizosphere/lentil, 11 ASVs
Aphanomyces sp.		Rhizosphere/wheat

^aCommunities were obtained from the test phase of a 2-year crop rotation, harvested in 2016 (trial 1) and 2017 (trial 2) from Swift Current, SK, Canada. In trial 1, 41 ASVs were identified as indicator species in the test phase rhizosphere communities, while 45 ASVs were identified in trial 2, notably from the same taxonomic groups. ASVs from the rhizosphere communities were also significant in specific soil histories established in the previous conditioning phase: in trial 1, five ASVs were associated with the lentil soil history, while in trial 2, 11 ASVs were again associated with the lentil soil history and 1 ASV was associated with wheat soil history. No indicator species were identified for any of the five Brassicaceae host crops. Indicator species analysis relied on abundance and site specificity to statistically test each ASV, which we report here as P < 0.05 with an FDR correction.

^bRhizosphere communities, but never root communities.

^cFallow, lentil, or wheat, grown the previous year.

Soil history significantly structured the soil oomycete communities of Brassicaceae crops.

Next, we tested if the three soil histories established by the previous crops structured significantly different oomycete communities. Soil history (trial 1, PERMANOVA R² = 0.0292, P < 0.0015; trial 2, PERMANOVA R² = 0.0300, P < 0.0018) was significant in structuring the test phase oomycete communities in both field trials (Table 3), though the soil history and crop host interaction was not significant in either trial. We complemented the PERMANOVA results with a variance partition to model the explanatory power of each factor (soil histories, soil chemistry, and the Brassicaceae host crops). Variance partitioning found that soil history explained similar amounts of the oomycete rhizosphere community data in both field trials (trial 1 R² = 0.0453, P < 0.001 [Fig. 3A]; trial 2 R² = 0.0476, P < 0.001 [Fig. 3C]). We also quantified how the experimental factors impacted the oomycete community structure with a redundancy analysis (RDA). Soil history was highly significant for the test phase oomycete rhizosphere communities in both field trails (trial 1 adjusted R² = 0.0539, P < 0.001 [Fig. 4A]; trial 2 adjusted R² = 0.0727, P < 0.001 [Fig. 4B]), where the communities were grouped by soil history (Fig. 4).

TABLE 3.

PERMANOVA results for oomycete community structures^a

Analysis category	Results for indicated trialb
	Trial 1			Trial 2
	F model	R 2	Pr(>F)	F model	R 2	Pr(>F)
Compartmentc	17.8018	0.12264	0.0001	8.3717	0.07508	0.0001
Soil historyd	2.1235	0.02926	0.0015	1.6600	0.02977	0.0018
Brassicaceae hoste	1.8876	0.05202	0.0003	0.9364	0.03359	0.6464
Interactions
Compartment-soil history	1.3196	0.01818	0.0941	1.6202	0.02906	0.0020
Compartment-Brassicaceae host	1.7741	0.04889	0.0008	1.0426	0.03740	0.3031
Soil history-Brassicaceae host	1.0734	0.05916	0.2367	0.8299	0.05954	0.9822
Compartment-soil history-Brassicaceae host	0.9040	0.04982	0.7576	0.8778	0.06298	0.9235

^aOur PERMANOVA analysis identified that the compartment and soil history were significant experimental factors in structuring the oomycete communities from the test phase of a 2-year crop rotation, harvested in 2016 (trial 1) and 2017 (trial 2) from Swift Current, SK, Canada. Brassicaceae host crops were only significant in the test phase oomycete communities of trial 1, while the Brassicaceae host-soil history interaction was never significant. The PERMANOVA was calculated using a Bray-Curtis distance matrix, with 9,999 permutations.

^bValues in bold indicate significant factors or interactions.

^cRhizosphere or roots.

^dFallow, lentil, or wheat.

^eBrassica carinata, B. napus, B. juncea, Sinapis alba, or Camelina sativa.

FIG 3.

Soil chemistry, soil history, and the Brassicaceae host crops were each influential in structuring the oomycete communities in the rhizosphere and roots from the test phase of a 2-year crop rotation, harvested in 2016 (trial 1, A and B) and 2017 (trial 2, C and D) from Swift Current, SK, Canada. Soil history explained a consistent amount of variance in the oomycete communities (A, B, and C). The current soil chemistry consistently explained the most variance in the oomycete communities (A, B, and D), except in the roots of Trial 1, where soil chemistry explained the least variance. In Trial 1 (A and B), the influence of the host plants increased between the rhizosphere communities (A) and the roots (B). Bray-Curtis distances were used in the variance partition.

FIG 4.

Soil history was significant in structuring the oomycete rhizosphere communities in both field trials of a 2-year crop rotation, harvested in 2016 (trial 1, A) and 2017 (trial 2, B) from Swift Current, SK, Canada. Distance-based redundancy analyses quantified how soil history impacted the oomycete community structure, where communities with similar compositions appear closer together.

Soil history was less consistent in the oomycete root communities of both field trials. First, it explained a similar amount of the variance in the test phase root community data (R² = 0.0418, P < 0.001) (Fig. 3B) as in the rhizosphere communities in trial 1, but was not significant in the root community data of trial 2 (Fig. 3D). Second, RDAs demonstrated the importance of soil history in the oomycete root communities of both trials (trial 1 adjusted R² = 0.0429, P = 0.007 [Fig. S9A]; trial 2 adjusted R² = 0.0403, P = 0.005 [Fig. S9B]). Though they were less significant compared to the rhizosphere (Fig. 4), they explained a similar amount of the data (R² = ~0.04) (Fig. 4; Fig. S9). A gradient separating the oomycete root communities based on their previous soil history was observed in trial 2 (Fig. S9B), similar to the corresponding rhizosphere communities (Fig. 4), but was less obvious in trial 1 (Fig. S9A).

Soil history also determined indicator species identified in the test phase rhizosphere communities in both trials. In trial 1, five oomycete ASVs were specific to rhizosphere communities with lentil crop histories; four were recognized as Pythiaceae and the fifth as a Pythium sp. (Table 2). In trial 2, 11 Pythiaceae ASVs were specific to rhizosphere communities with the lentil soil history (P < 0.05) (Table 2), while another Pythiaceae ASV was specific to rhizosphere communities with the wheat soil history (Table 2).

Brassicaceae crop hosts had limited influence on their oomycete communities.

Brassicaceae hosts had a significant effect (R² = 0.0520, P < 0.0003) on oomycete community structure in trial 1 (Table 2), but not in the dry year of trial 2. The variance partition illustrated that the Brassicaceae crop hosts accounted for 4.67% of variance of the test phase rhizosphere community data (Fig. 3A) in trial 1. However, Brassicaceae crop hosts were not significant in the variance partition of trial 2 (Fig. 3C and D). RDA also supported the significance of the Brassicaceae crop host (R² = 0.0454, P = 0.006) (Fig. S10A) in the oomycete rhizosphere communities in trial 1 but was not significant in trial 2. The first RDA axis showed a gradient among oomycete communities between Camelina sativa and Sinapis alba, with a notable amount of overlap. Interestingly, the second axis showed a gradient between soil histories, with the majority of communities from lentil sites clustered in the bottom left (Fig. S10A).

In the oomycete root communities, Brassicaceae crop hosts explained the most variation in trial 1 (R² = 0.0719, P < 0.001) (Fig. 3B) but were not significant in the variance partition of the root communities in trial 2, similar to the findings for the rhizosphere in trial 2. RDA also illustrated the importance of the Brassicaceae crop hosts in structuring the test phase oomycete root communities (R² = 0.0961, P = 0.002) (Fig. S10B) in trial 1, but not in trial 2. Oomycete root communities from C. sativa and Brassica carinata had more distinct clusters than the other three crop hosts in trial 1 (Fig. S10A).

Differential taxa clusters from trial 1 identified variations between Brassicaceae hosts in trial 1, but not in the dry year of trial 2. The test phase oomycete communities in S. alba rhizospheres were depleted in Lagena sp. ASVs relative to the rhizosphere communities of both B. napus and C. sativa (P < 0.06) (Fig. 2B). S. alba root (P < 0.01) and rhizosphere (P < 0.06) communities were also depleted in Pythiaceae ASVs relative to the oomycete communities of B. carinata (Fig. 2B). Test phase oomycete root communities from B. carinata were enriched in Pythium sp. ASVs compared to those for B. napus (P < 0.01) (Fig. 2B). Indicator species analysis did not identify any oomycete ASVs as specific to any of the five Brassicaceae host crops in either field trial.

Soil chemistry significantly influenced the oomycete rhizosphere and root communities.

Variance partitioning revealed that the soil chemistry was the most significant factor in the oomycete rhizosphere communities in both field trials (trial 1 R² = 0.0945%, P < 0.001 [Fig. 3A]; trial 2 R² = 0.2024, P < 0.001) (Fig. 3C). RDA also supported that soil chemistry was the most explicative experimental factor of the test phase oomycete rhizosphere communities in both trials (trial 1 adjusted R² = 0.1004, P = 0.012 [Fig. 5A]; trial 2 adjusted R² = 0.271, P < 0.001 [Fig. 5B]). These data indicated that the test phase oomycete rhizosphere communities were strongly shaped by soil chemistry in both field trials. This effect was stronger in the rhizosphere during the dry year of trial 2.

FIG 5.

Soil chemistry was the most significant factor for oomycete community structures in the rhizosphere (R² = 0.1004, p - 0.012 [A] and R² = 0.271, P < 0.001 [B]) from both field trials, as well as in the roots from five Brassicaceae crop hosts from trial 2 (R² = 0.1148, P < 0.001 [C]) in the test phase of a 2-year crop rotation, harvested in 2016 (trial 1) and 2017 (trial 2) from Swift Current, SK, Canada. Distance-based redundancy analyses quantified how soil chemistry impacted oomycete community structure, where communities with similar compositions appear closer together. The largest factors in the rhizosphere were iron, calcium, nitrate, and manganese, which contrasted with pH (A) and calcium contrasting with manganese, and to a smaller extent zinc (B); magnesium opposed organic carbon, conductivity and pH contrasted total nitrogen; iron was also a strong factor. (C) The largest factor in the root communities was magnesium, which weakly contrasted with calcium, while conductivity contrasted with iron. Note that in panels B and C total carbon was explained by organic carbon, while Zn was explained by Cu in panel C.

In the oomycete root communities, soil chemistry explained the least amount of variation in trial 1 (R² = 0.0330, P < 0.001) (Fig. 3B) but was the only significant factor during the dry year in trial 2 (R² = 0.0576, P < 0.035) (Fig. 3D). RDA also supported the significance of soil chemistry on the test phase oomycete root communities in trial 2 (adjusted R² = 0.1148, P < 0.001) (Fig. 5C) but not in trial 1. The data suggested that soil chemistry was only influential in the oomycete root communities during the dry year of trial 2, when the effects of soil history and Brassicaceae crop host were reduced on structuring the communities.

Coinertia analysis between Brassicaceae oomycete and bacterial communities.

We used a coinertia test to investigate how influential soil history was on the relationship between the test phase oomycete communities investigated here and the previously identified bacterial communities (4). In trial 1, the root and rhizosphere samples had similar RV coefficients, 0.6291 and 0.7049, respectively, suggesting that the oomycete and bacterial communities did have significant relationships in both compartments. The rhizosphere samples were plotted on the first and second axes, which represented 8.185% and 6.551% of the coinertia. The low inertia suggested little influence of the two sets of ASVs on the samples. This was further illustrated in that the majority of the 60 samples in trial 1 appeared quite similar, as they remained clustered together toward the center of the plot. Rhizosphere samples from plots 17, 41, 46, and 51 were relatively more influenced by the presence of particular microbial ASVs, as they were further from the center (Fig. S11A). Only samples 17 and 51 illustrated any noticeable divergence between which microbial ASVs influenced their structure, given their appreciable arrow lengths.

The root samples from trial 1 captured 9.404% and 7.682% of the coinertia in the first and second axes, respectively, which suggested minimal influence of the ASVs on each sample, similar to their corresponding rhizosphere communities. Most of the root samples were also clustered together in the coinertia plot; only samples 47 and 51 were noticeably influenced by the presence of particular microbial ASVs, given their positions. The arrow length of the root sample from plot 47 suggested it diverged between which microbial ASVs influenced their structure (Fig. S11B).

In trial 2, the root and rhizosphere samples had less similar correlation and vectorial (RV) coefficients; the oomycete and bacterial communities had a significant relationship (RV = 0.8307) in the rhizosphere samples, but not in the root samples (RV = 0.5767). The first and second axes for the rhizosphere samples represented 12.776% and 4.578% of the coinertia, respectively. Given the low inertia, the rhizosphere samples appeared to be weakly influenced by the two data sets of microbial ASVs. Nonetheless, trial 2 rhizosphere samples were more heterogenous, as they were more dispersed than the rhizosphere samples from trial 1, with samples 22 and 30 being distinctly different (Fig. S11C). Finally, the influence of the microbial ASVs only appeared to shift in samples 19 and 30, as shown by their arrow lengths.

The trial 2 root samples captured 8.880% and 8.131% of the coinertia in the first and second axes, respectively. Again, the low inertia suggested a minimal influence of microbial ASVs on each sample, which was similar to their corresponding rhizosphere communities. Unlike the rhizosphere samples, however, the root samples were tightly clustered; only sample 51 was noticeably influenced by particular microbial ASVs. The arrow length of sample 51 also suggested that the influence of the microbial ASVs shifted between the bacterial and oomycete communities (Fig. S10C).

DISCUSSION

How the soil history established by previous plant-soil microbial communities conditions future generations of oomycete communities remains relatively unknown. Oomycetes are vastly understudied compared to bacteria and fungi, yet they are important microbial communities, especially for agriculture where many oomycetes are responsible for severe declines in yields. In this study, we investigated the impact of three different soil histories established by the previous year’s crops on the soil oomycete communities associated with five Brassicaceae oilseed host crops. The seminested ITS amplicon strategy we incorporated with MiSeq metabarcoding specifically targeted oomycetes and has been previously shown to limit off-target amplification from the ITS region of other eukaryotes (47, 48). The oomycete metabarcoding data illustrated that soil history had a greater influence on the communities than the Brassicaceae host crops, while soil chemistry structured the oomycete communities more during the dry field trial. Our results highlighted the impact of edaphic factors over different growing seasons and the importance of monitoring and quantifying oomycete biodiversity.

Previous crops significantly impacted the oomycete rhizosphere communities.

The previous crops, and their agricultural treatments, impacted the subsequent oomycete communities through plant-soil microbial community feedback. We hypothesized that the three soil histories established by the previous crops would structure significantly different oomycete communities, regardless of their current Brassicaceae host, in both the roots and rhizosphere. Our data illustrated that this was largely sustained; we found consistent support for soil history influencing the structure of the oomycete rhizosphere communities of both field trials, as well as the root communities in trial 2 (Table 2, Fig. 3). Moreover, gradient analysis (Fig. 4 and Fig. S9) highlighted how different oomycete communities tended to cluster according to the soil histories established by the previous crops, especially among the rhizosphere communities. These are exciting results, as they raise more questions about oomycete community dynamics and their interactions with different soil histories established through crop rotations.

In this study, we found that the oomycete communities were significantly structured by each of the three previously established soil histories. We observed little effect from the Brassicaceae crop hosts to restructure the oomycete communities according to their current hosts. Crop rotations have previously been shown to quickly adjust subsequent bacterial communities via plant-microbial community feedback mechanisms (4, 30). Important fractions of bacterial rhizosphere communities tend to be fast-growing and have rapid turnover, which may allow bacterial communities to be more responsive to the dynamic needs of their host plants (17, 20). Thus, plant-microbial community feedback mechanisms from new host plants appear sufficient to quickly erase the soil history established by a previous plant and modify the bacterial communities (1, 3, 4).

Conversely, experimental evidence has suggested that such feedback mechanisms are not sufficient to alter fungal communities (1, 3). One suggestion for why the influence of established soil history varies between bacterial and fungal communities has been due to their different growth rates (3, 49). Compared to the rapidly growing parts of plant bacterial communities, fungal communities remain more stable through time. Fungal growth seems more steady and less influenced by host plant feedback mechanisms, which limits how responsive fungal communities might be to the influence of new crop hosts (1, 3). Although oomycetes are not fungi and have vastly different evolutionary origins (37, 50, 51), our data illustrated a similar trend, where oomycetes, like fungi, remained relatively unaffected by changes in crop hosts, possibly due to their growth rate. Complementary to this idea is that oomycete oospores can persist from year to year and are constitutively dormant, such that not all oospores germinate at the same time, even under optimal conditions (52–55). This could also help account for why oomycete communities may appear less affected by the influence of the new Brassicaceae hosts feedback mechanisms (1, 3).

In fact, our results illustrated that the soil history established by the previous lentil and wheat crops helped to structure distinct oomycete communities that were still detectable the following year (Fig. 4, Fig. S9, and Table 2). The lentil-specific oomycete rhizosphere community we detected in both field trials may be unsurprising, since legumes, including lentils, tend to retain more soil moisture than other crops, and soil moisture is a key factor for oomycete growth. Recent studies have also suggested that lentils have an increased vulnerability to oomycete outbreaks (5, 26, 27). Finally, Pythiaceae have previously been reported in Canadian pea fields (43); thus, detecting a variety of Pythiaceae ASVs specific to the lentil soil history seems reasonable.

Conversely, the Aphanomyces ASV specific to the rhizosphere communities in trial 2 with wheat soil history (Table 2) was somewhat more unexpected. Most of the interest concerning Aphanomyces has focused on Aphanomyces cochlioides and Aphanomyces euteiches, which are well-described pathogens specific to sugar beets and legumes, respectively (56). However, there is a divergent lineage that consists of saprotrophs and opportunistic plant pathogens that are not known to maintain specific hosts (56). A saprotrophic oomycete capable of degrading wheat residues might explain the Aphanomyces ASV identified among trial 2 rhizosphere communities with wheat soil history.

However, oomycete functional lifestyles can actually be rather diverse (57). Even though our putative lifestyles identified the ASVs we reported as plant pathogens—which is concordant with the number of Pythium and Phytophthora ASVs we identified, and with other metabarcoding surveys (57)—these lifestyle assignments can still contain a range of functions (57). For example, oomycete plant pathogens exist in an array of biotrophic and hemibiotrophic capacities (57) in terms of timing, specificity, and duration (57). Furthermore, hemibiotrophs can also live saprotrophically in the soil in the absence of a plant host and have even been shown to play important roles in decomposition (58, 59).

Oomycete communities were not influenced by Brassicaceae crops during the drier field trial.

Although our initial hypothesis concerning soil history was largely supported, we did nonetheless observe an influence of the Brassicaceae crop hosts on the oomycete communities, but only during trial 1 (Fig. 3 and Table 2), and particularly in their roots (Fig. S10). This was an interesting finding, given that the root communities we identified were noticeably reduced and less diverse than their cognate rhizosphere communities. Although we scrapped off attached soil from the roots and repeatedly washed the roots to remove the rhizosphere, additional surface washes, or surface sterilization, could have further reduced any residual rhizosphere-rhizoplane influence on the root communities.

Plant hosts ought to have the most influence to select for microbial communities in their roots, compared to the rhizosphere or leaf surface (36, 60). While our data illustrated that the Brassicaceae host crops were quite significant in the roots during trial 1 (Fig. 3 and Fig. S9), we did not detect any oomycete ASVs as indicator species from any of the five hosts (Table 2). Furthermore, we observed a considerable amount of overlap among the oomycete root communities, notwithstanding the more distinct clusters of communities from C. sativa and B. carinata (Fig. S10). Our data indicated that the influence of the Brassicaceae hosts on the oomycete root communities was insufficient to structure more distinct groups of oomycetes. Similar results were found for the communities from different cultivars of Rhododendron (61). This weaker effect of plant hosts could be due to other competing factors, such as the previous soil history or current soil chemistry. Alternatively, the close genetic relationship of the host plants may preclude us from identifying more specific oomycete assemblages (4, 61).

Moreover, there was no influence of any of the Brassicaceae crops on the oomycete communities during trial 2 that we observed, despite following identical experimental protocols and using the same agricultural management practices and inputs. The disparate observations between the two field trials could have been due to the environmental conditions being 6 times drier during trial 2: 55.0 mm of precipitation versus 328.4 mm in trial 1 (4). The Brassicaceae host plants appeared to be restricted in growth due to the dry conditions (Fig. S3), which would also constrain their nutrient uptake from the soil and rhizodeposition (22). Therefore, if the reciprocal plant-soil microbial community feedback was impaired due to the availability of water, it could account for the absent influence of the Brassicaceae hosts on the oomycete communities in Trial 2.

The drier conditions of field trial 2 may also have had an impact on the oomycete community itself. Oomycetes prefer high soil moisture for motility, subsequent infection, and growth and the completion of their life cycle (51, 52, 62, 63). Our data demonstrated a similar community composition between both field trails, though with reductions in sequencing reads and diversity in trial 2. This could be evidence for how the drier conditions impacted the community. Quantifying community sizes could help determine this in future experiments.

Nonetheless, the impact of the Brassicaceae crop hosts remained limited, as the structure of the oomycete communities remained significantly influenced by the previous crops. This could indicate that these specific crops may not be effective as a strategy to limit the accumulation of potentially pathogenic oomycetes in the soil over the short term. Various crop rotations, including those involving Brassicaceae, have been shown to help control phytopathogens by restructuring the microbial communities from one season to the next (6, 31). Such shifts generally occur through plant-soil feedback processes, such as rhizodeposition or by producing antimicrobial compounds (9–12, 64, 65). For example, Brassicaceae crops produce antimicrobial glucosinolates, which have been used to control phytopathogens, including oomycetes (64). However, our data illustrated that the five Brassicaceae crops were unable to sufficiently alter the soil history established by the previous crops, given that the oomycete communities remained significantly structured by soil history. This suggested that these crop rotations may be an insufficient strategy to control oomycete phytopathogens in the short term.

In addition, we observed that B. carinata crop hosts were significantly enriched in Pythiaceae ASVs in their root and rhizosphere communities compared to the other Brassicaceae hosts (Fig. 2B). This suggested that B. carinata is more susceptible to oomycete accumulation. Conversely, we also noted that S. alba hosts were depleted in Lagena and Pythiaceae ASVs in their rhizosphere communities (Fig. 2B), compared to the other Brassicaceae crop hosts. This might demonstrate increased resistance to accumulating these oomycetes in their rhizosphere. These two examples warrant further study to help evaluate how effective these particular crop rotations may be at limiting oomycete infections.

Further to this, our study points out three recommendations needed to better understand the phytopathogenicity of oomycetes: first, biodiversity monitoring should inventory the oomycete communities established at the end of each growing season and observe which ASVs persisted in a given plot (39, 66). Second, quantifying the size of each community would help determine if crop rotations actually limit, or reduce, the growth of the oomycete communities. Third, since the impact of a crop may not be observed during the active growing season (30), longer field trials with multiple time points could help confirm our findings. These additional steps may provide a more nuanced understanding of the dynamics within oomycete communities and help determine the utility of crop rotations as a strategy to limit the accumulation of oomycete phytopathogens in agricultural soil.

Soil chemistry constrained oomycete community structure.

Although we initially sought to test the influence of soil history on structuring oomycete communities, our data revealed that the soil chemistry had the strongest influence in the rhizosphere during both field trials and among the root communities during the dry season in trial 2 (Fig. 3 and 5). In our agricultural setting, soil chemistry was largely a synthesis of the previous soil history, current agricultural management practices, and the plant-microbial community feedback mechanisms (67). These processes interacted to yield a number of edaphic conditions previously identified to promote oomycete growth.

For example, soils with excessive or insufficient nutrients for their local microbial communities are prone to outbreaks of oomycete infections, as a soil nutrient imbalance provides a niche space for them (26, 44). Indicators of soil nutrient balance may include conductivity, cation exchange capacity (EC), total nitrogen, and total carbon (26, 27, 44). Although none of the measured edaphic factors was particularly related to the communities observed in trial 1, oomycete rhizosphere and roots communities with lentil soil histories were strongly associated with EC in trial 2 (Fig. 3). Soil moisture is another key factor in promoting oomycete growth and is compounded by seeding into cool (<16°C) soils (5, 26, 27). These conditions favor the release and chemotaxis of oomycete zoospores (51, 52). Therefore, we might have expected to observe a more dramatic change in the oomycete rhizosphere community between the wetter season of trial 1 and the dry season of trial 2.

The importance of soil chemistry may also be reflected in the results of our coinertia analysis. This analysis illustrated that although the oomycete and bacterial data tended to have a significant relationship, neither community was particularly impactful in the roots nor the rhizosphere, nor did any of the three soil histories influence their relationship (Fig. S11). Given that both of these microbial communities were derived from the same soil samples, they were more likely to experience the same edaphic factors. Therefore, the lack of obvious influences in the coinertia analysis could have been due to the oomycete and bacterial communities being similarly constrained by their common soil chemistry.

Microbes largely share basic biological reactions to abiotic factors, such as changes to pH, temperature, or water availability (62). For example, bacteria, fungi, and oomycetes, among others, require water for chemotaxis and locomotion, as well for maintaining turgor pressure (62, 63). Cellular function requires the correct regulation of pressure, without which cells are unable to grow, divide, or move (63). Although microbes have evolved a number of specialized tactics to regulate osmolarity, water stress, among other limitations imposed by soil chemistry, remains a common constraint (62). Therefore, the homogeneity we observed from the coinertia analysis could reasonably have been due to the oomycete and bacterial communities being similarly constrained by their common soil chemistry.

Conclusion.

Oomycetes are major global phytopathogens, yet they are understudied compared to other microbes. Here, we have shown for the first time the important role of soil history in structuring oomycete rhizosphere and root communities. We tested three different soil histories and found that none of the five planted Brassicaceae oilseed crops was able to restructure the oomycete community the following year. We also took a novel approach in investigating how oomycete and bacterial communities may have structured one another. To our knowledge, this is the first demonstration of the weak impact between the two microbial communities. Rather, the similarities between the two microbial communities may have been due to being constrained by common edaphic factors. This study advances our understanding of how different agricultural practices can impact future microbial communities differently. Our results also highlight the need for continued monitoring of oomycete biodiversity and quantification.

MATERIALS AND METHODS

Site and experimental design.

A field experiment was conducted at the experimental farm of Agriculture and Agri-Food Canada’s Research and Development Centre in Swift Current, Saskatchewan (50°15′N, 107°43′W). The site is located in the semiarid region of the Canadian prairies; according to the weather station of the research farm, the 2016 and 2017 growing seasons (May, June, and July) had 328.4 mm and 55.0 mm of precipitation, respectively; compared to the 30-year average (1981 to 2010) of 169.2 mm. The daily temperature averages for the 2016 and 2017 seasons were 15.6°C and 15.9°C, respectively, while the 30-year average was 14.93°C. The farm is on a brown Chernozem with a silty loam texture (46% sand, 32% silt, and 22% clay) and has been well described previously (34, 68).

The experiment was established in a field previously used for growing spring wheat (Triticum aestivum cultivar AC Lillian). A two-phase cropping sequence, consisting of a conditioning phase the first year and a test phase in the second year (see Fig. S1 in the supplemental material), was repeated in two field trials: trial 1, 2015 to 2016, and trial 2, 2016 to 2017, on adjacent sites (Fig. S1B and C). On each site, the experimental design was a split-plot replicated in four complete blocks. In the conditioning phase, three soil history treatments were randomly assigned to the main plots, consisting of spring wheat (Triticum aestivum cv. AC Lillian), red lentil (Lens culinaris cv. CDC Maxim CL), or left fallow (Fig. S1). Thus, the conditioning phase established a soil history composed of either wheat, lentil, or fallow, plus the respective management plans as described below (4, 33, 34).

In the test phase, the 12 conditioning phase plots were each subdivided, and five Brassicaceae oilseed crop species were randomly assigned to one of these five subplots (Fig. S1). The Brassicaceae crops seeded were Ethiopian mustard (Brassica carinata L. cv. ACC110), canola (B. napus L. cv. L252LL), oriental mustard (B. juncea L. cv. Cutlass), yellow mustard (Sinapis alba L. cv. Andante), and camelia (Camelina sativa L. cv. Midas). Therefore, the test phase established the Brassicaceae host plant-soil microbial community feedback, composed of the individual Brassicaceae genotypes, their soil microbial community, and their respective management plans, as described below (4, 33, 34). In total, each field trial had 60 subplots to sample (Fig. S1 and S2). For further details of this well-described experiment, including its design and treatments, see references 4, 33, and 34.

Crop rotation management and sampling.

Crops in both field trials were grown and maintained according to standard management practices, as previously described (4, 33, 34). A preseed “burn-off” herbicide treatment using glyphosate (Roundup, 900 g acid equivalent [a.e.] ha⁻¹) was applied to all plots each year to ensure a clean starting field prior to seeding. Lentil seeds were treated with a commercial rhizobium-based inoculant (TagTeam, 3.7 kg ha⁻¹). Lentil and wheat were direct-seeded into wheat stubble from late April to mid-May, depending on the crop and year. The herbicides glyphosate (Roundup; 900 g a.e. ha⁻¹), Assure II (36 g active ingredient [a.i.] ha⁻¹), and Buctril M (560 g a.i. ha⁻¹) were applied to the fallow, lentil, and wheat plots, respectively, for in-season weed control, while fungicides were only applied as needed. Soil tests were used to determine the rates of in-season nitrogen, phosphorus, and potassium application; no synthetic nitrogen fertilizer was applied to the lentil plots during the conditioning phase. Both lentil and wheat were harvested between late August and early October, depending on the crop and year.

The subsequent test phase Brassicaceae plant hosts were subjected to the same standard management practices as in the conditioning phase, including preseed burn-off, in-season herbicide and fungicide treatments as needed, and fertilized as recommended by soil tests (Table S1) (33, 34, 69). Additionally, all Brassicaceae crops, except B. napus, were treated with Assure II mixed with Sure-Mix or Merge surfactant (0.5% [vol/vol]) for postemergence grass control; Liberty (glufosinate, 593 g a.i. ha⁻¹) was used for B. napus.

Test phase Brassicaceae plants were sampled in mid-late July at full flowering, i.e., when 50% of the flowers on the main raceme were in bloom, as described by the Canola Council of Canada (70), with flowering corresponding to higher activity in rhizosphere microbial communities (71). Four plants from two different locations within each subplot were excavated and pooled together as a composite sample (4, 33, 34, 69). In the field, each plant had its rhizosphere soil divided from the root material by gently scraping it off using bleach-sterilized utensils into fresh collection trays. The roots were then gently washed three times with sterilized distilled water to remove any soil. Both the rhizosphere and root portions were immediately flash-frozen and stored in liquid nitrogen vapor shipping containers until stored in the lab at −80°C (72). Based on the sampling strategy, in this study we defined the rhizosphere microbiome as the microbial community in the soil in close contact with the roots (3) and the root microbiome as the microbial community attached to, and within, the roots (21). Two additional soil cores were sampled from each plot, pooled, and kept on ice in coolers. These samples were homogenized in the lab and sieved to remove rocks and roots. They were then used for soil chemistry analyses, including total carbon, nitrogen, pH, and micronutrients (see reference 69 for details). Aerial portions of each harvested plant sample were retained to determine dry weight (Fig. S3).

DNA extraction from test phase Brassicaceae root and rhizosphere samples.

Nucleic acids were extracted from trial 1 test phase Brassicaceae samples for both rhizosphere and root portions. First, all the root samples were ground in liquid nitrogen via sterile mortar and pestle (Fig. S2). Total DNA and RNA were extracted from ~1.5 g of rhizosphere soil using the RNA PowerSoil kit with the DNA elution kit (Qiagen, Germany). DNA and RNA were extracted using ~0.03 g of roots using the DNeasy plant DNA extraction kit and RNeasy plant minikit (Qiagen, Canada), respectively, following the manufacturer’s instructions (see reference 69 for use of the RNA samples). All remaining harvested materials from trials 1 and 2, as well as the extracted DNA from trial 1, were kept at −80°C before being shipped to Université de Montréal’s Biodiversity Centre, Montréal (QC, Canada) on dry ice for further processing (72, 73).

Total DNA was extracted from the trial 2 test phase samples; ~500 mg of rhizosphere soil was used with the NucleoSpin Soil gDNA extraction kit (Macherey-Nagel, Germany), and ~130 mg of roots was used with the DNeasy plant DNA extraction kit (Qiagen, Germany) (73). A no-template extraction negative control was used with both the root and rhizosphere extractions and included with the test phase samples (Fig. S2) to assess the influence of the extraction kits on our sequencing results and the efficacy of our lab preparation. All 242 extracted DNA samples (60 plots × 2 parallel field trials × 2 compartments [rhizosphere and root] + 2 no-template extraction control samples) were quantified using the Qubit dsDNA high sensitivity kit (Invitrogen, USA) and qualitatively evaluated by mixing ~2 μL of each sample with 1 μL of GelRed (Biotium) and running it on a 0.7% agarose gel for 30 min at 150 V. The no-template extraction negative controls were confirmed to not contain DNA after extraction, where the detection limit was >0.1 ng (Qubit, Invitrogen, USA). Samples were kept at −80°C (72, 73).

Assembly of oomycete mock community.

To assess potential bias caused by lab manipulations, sequencing, and downstream bioinformatic processing, we assembled an oomycete mock community of known composition from 20 species with staggered copy numbers of the ITS1 region. To do so, we followed a method described previously (74), beginning with generating a standard curve of the copy numbers of the ITS region from DNA extracted from Pythium ultimum strain 6358.Ba.B, generously provided by S. Chatterton (Agriculture and Agri-Food Canada, Lethbridge Research and Development Centre). We then used this standard curve to estimate the ITS copy numbers from a diversity of oomycete DNA samples provided by the Ministry of Agriculture and Fisheries of Quebec and by Agriculture and Agri-Food Canada (Table S2).

A ~1-kb fragment containing the entirety of the ITS 1 region, the 5.8S gene, and the ITS2 region was PCR amplified from P. ultimum 6358.Ba.B using the ITS4 and ITS6 primers (Alpha DNA, Montréal, Canada) (Table 4) (47). The ITS4/6 PCR consisted of 11.5 μL distilled H2O (dH₂O), 5.0 μL of 10× buffer (Qiagen, Canada), 2.5 μL of 10 μM ITS4 and ITS6 primers (Alpha DNA, Montréal, Canada), 1.0 μL of deoxynucleoside triphosphates (Qiagen, Canada), 0.5 μL of Taq polymerase (Qiagen, Canada), and 2 μL of a 1:10 dilution of the template, for a total volume of 25 μL. The PCR was run in an Eppendorf Mastercycle ProS thermocycler (Mississauga, ON, Canada) and consisted of an initial denaturation of 2 min at 95°C, followed by 30 cycles of 30 s of denaturation at 95°C, 30 s of annealing at 55°C, and 1 min of elongation at 72°C, before a final elongation of 5 min at 72°C (47). Four microliters of the ITS PCR product was mixed with 1 μL of loading dye containing Gel Red (Biotium) and visualized on a 1% agarose gel after 60 min at 100 V. The amplified ITS fragment was quantified using the QuBit dsDNA high sensitivity kit (Invitrogen, USA) and then serially diluted to 10⁻⁹, where 1 μL of each dilution was then used as template in a 10-μL qPCR mixture.

TABLE 4.

ITS primers used in this study

Name	Sequence (5′–3′)	Reference
ITS6 F	GAAGGTGAAGTCGTAACAAGG	78
ITS4 R	TCCTCCGCTTATTGATATGC	102
ITS7-ae	WGYGKTCTTCATCGATGTGC	48

The ITS1 region qPCRs amplified an ~350-bp region using the oomycete-specific ITS6 forward and ITS7-a.e. reverse primers (Table 4) (48). All qPCRs were setup in triplicate in 96-well plates using the Freedom EVO100 robot (Tecan, Switzerland), with a triplicate no-template negative control included on each plate. The reaction mixtures consisted of 5.0 μL of Maxima SYBR green/ROX qPCR mix (ThermoFisher Scientific, Canada), 3.4 μL dH₂O, and 0.3 μL of 10 μM ITS6 and ITS7-a.e. primers (Alpha DNA, Montréal, Canada) and were run in a ViiA 7 real-time PCR system (ThermoFisher Scientific, Canada). The cycling conditions consisted of an initial denaturation of 2 min at 94°C, followed by 30 cycles of 30 s of denaturation at 94°C, 30 s of annealing at 59°C, and 1 min of elongation at 72°C, before a final elongation of 10 min at 72°C (48). The number of ITS1 region copies present in the serially diluted standards were calculated using the following formula (75): Number of ITS1 copies μL⁻¹ = [Avogadro’s constant × DNA (g μL⁻¹)]/[Number of base pairs × 600 Da]. A standard curve was plotted, with an R² value of 0.9938 and an amplification efficiency of −3.389 (Fig. S4) falling within acceptable values (76).

Copy numbers of the ITS1 region were then estimated for each oomycete sample (Table S2). One microliter of a 1:10 dilution of extracted oomycete DNA was used as template for the ITS1 qPCR, following the cycling conditions described for generating the standard curve. Melt curves generated by 0.5°C increments at the end of the qPCR program confirmed amplicon specificity. The mean cycle threshold was then calculated from the qPCRs for each oomycete sample, and the corresponding ITS1 copy number was estimated from the standard curve (Fig. S4) (74). Finally, the oomycete mock community was assembled with staggered ITS1 copy numbers, where different oomycete community members had different ranges of ITS1 copy numbers (74).

ITS amplicon generation and sequencing to estimate composition of the oomycete communities.

To estimate the composition of the oomycete communities in the rhizosphere and roots from the test phase Brassicaceae species, extracted DNA from all samples was used to prepare ITS amplicon libraries following Illumina’s MiSeq protocols. First, all DNA samples were diluted 1:10 into 96-well plates using the Freedom EVO100 robot (Tecan, Switzerland). To assess potential bias caused by lab manipulations, sequencing, and downstream bioinformatic processing, we included a no-template DNA extraction control and a mock community on each plate.

The prepared plates of the test phase DNA samples were submitted to Génome Québec (Montréal, Québec) for ITS amplicon generation and sequencing (73, 77). In order to preferentially target the oomycete community and exclude other eukaryotes, as well as ensure sufficient quantity for detection, a seminested PCR strategy was used to generate the ITS1 amplicons from each test phase DNA sample (47). Each sample was used as template in a PCR consisting of 15 cycles using the ITS6 and ITS4 primers (Table 4) (47, 78) to amplify an ~1-kbp fragment. A second-round PCR was then done using the oomycete-specific ITS6 and ITS7a.e. primers (Table 4) (47, 48) to amplify an ~350-bp fragment of the ITS1 region. This seminested PCR strategy has previously been shown to enrich for oomycete sequences and limit off-target amplification from other eukaryotes (47, 48). The degenerate primer design has been suggested to allow for the wider capture of oomycete-specific ITS1 sequences, but not at the cost of being biased for a particular oomycete group (47, 48).

These amplicons were then prepared for paired-end 250-bp sequencing using Illumina’s MiSeq platform and the MiSeq reagent kits v3 (600 cycles) (Génome Québec, Montréal) (73, 77). We estimated this would provide a mean of 40,000 reads per sample, which is in line with previous studies that described microbial eukaryote communities (4, 73, 77).

Estimating ASVs from MiSeq ITS amplicons.

The ITS amplicons generated by Illumina MiSeq sequencing were used to estimate the diversity and structure of the oomycete communities present in both the rhizosphere and roots of each test phase Brassicaceae sample. The integrity and totality of the ITS MiSeq data downloaded from Génome Québec, all 17,656,076 reads, were confirmed using their MD5 checksum protocol (79). Subsequently, all data were managed and analyzed in R (80) and plotted using ggplot2 (81).

Instead of generating OTUs from the ITS amplicon data, we opted to use DADA2 for ASV inference, as it generates fewer false positives than OTUs, reveals more low-abundant, or cryptic, microbes, and as ASVs are unique sequence identifiers, they are directly comparable between studies, unlike OTUs (22, 45, 46). ASV inference has also been successfully used with oomycete data sets to discriminate to the species level (36, 61). Here, due to the variable length of the ITS region, we first used cutadapt (82) to carefully remove primer sequences from all the ITS reads generated from the control samples, the mock communities, and the experimental test phase Brassicaceae samples, including any primer sequences generated due to readthrough. The filterAndTrim function from the DADA2 package (83) was then used for all reads, following the default settings, including removal of reads shorter than 50 nucleotides or of low quality (Q ≤ 20). Filtered and trimmed reads were then processed through DADA2 for ASV inference (Fig. S2 and S5). Default settings were used throughout the DADA2 pipeline, except for the DADA inference functions dadaF and dadaR, for which we used the “pool = pseudo” argument, to increase the likelihood of identifying rare taxa. Consequently, the chimera removal function removeBimeraDenovo included the method “=pooled” argument (83).

The unique ASVs inferred from the ITS amplicon data were assigned taxonomy classifications using the UNITE database for all eukaryotes (84). ASVs were assigned species when possible, though UNITE may not account for all the latest taxonomic revisions to individual groups between versions, as reported elsewhere (85). Moreover, it is important to keep in mind that species names reported here for each ASV are the closest designations after comparisons with reference sequences. Thus, even at 100% similarity, species names remain approximations. Data quality was assessed using the included controls (Fig. S5B), any off-target ASVs assigned to the taxa Viridiplantae, Alveolata, Fungi, Heterolobosa, or Metazoa were subsequently removed, and the remaining oomycete ASVs were assigned functional lifestyles based on the FungalTraits database (86). Rarefaction curves confirmed that we captured the majority of the oomycete communities in both root and rhizosphere samples from both field trials (Fig. S6). The oomycete-specific ITS sequencing data were subsequently reanalyzed independently following the described protocol to avoid any biases from the four no-template negative controls and the four mock communities. These are the test phase oomycete ASVs which are reported hereafter and in the supplemental material.

Inferring phylogenetic trees.

We assembled phylogenies for each compartment, from both trials, in order to infer phylogenetic diversity of the test phase Brassicaceae oomycete communities. Following the method described by Callahan et al., the ITS region sequences for each ASV inferred from the test-phase Brassicaceae data were aligned using a profile-to-profile algorithm (87), with a dendrogram guide tree using the decipher package (88). With the phangorn package (89), the maximum likelihood of each site was calculated using the dist.mL function with a JC69 equal base frequency model, before assembling phylogenies using the neighbor-joining method. An optimized general time-reversible nucleotide substitution model was fitted to the phylogeny using the optim.pml function. Phylogenies were subsequently added to each phyloseq object (90).

α-Diversity of the test phase Brassicaceae rhizosphere and root communities.

First, to visualize taxonomic diversity, ASVs were plotted as taxa cluster maps using heat_tree from the metacoder package (91) for the rhizosphere and roots of both experiments, where nodes represent class to genus: node colors represent the number of unique taxa, while node size indicates the relative abundance of each ASV. Taxa cluster maps facilitated visualizing abundance, as well as diversity across taxonomic hierarchies (91).

Second, in order to estimate the coverage of the oomycete class, we incorporated the oomycete phylogenies into the phyloseq object following a method described previously (83). Faith’s phylogenetic diversity was calculated as an α-diversity index from the test phase Brassicaceae samples using the pd function from the picante package (sum of all branch lengths separating taxa in a community) (92). For comparison, Simpson and Shannon’s α-diversity indices were also calculated (Fig. S7).

We assessed differences between the mean phylogenetic diversities between soil histories and Brassicaceae hosts using the nonparametric Kruskal-Wallis rank sum test, kruskal.test, as the transformed data did not respect the assumptions for normality. Specific groups of statistical significance were identified with the post hoc pairwise Wilcoxon rank sum tests, pairwise.wilcox.test, with the false-discovery rate (FDR) correction on the P values to account for multiple comparisons.

Identification of differentially abundant ASVs and specific indicator species.

To refine our understanding of the abundance and composition of the test phase Brassicaceae oomycete communities, we used two complementary methods to identify taxa specific to soil histories or Brassicaceae hosts. First, taxa cluster maps were used to calculate the differential abundance of ASVs between experimental groups, including rhizosphere and root compartments, Brassicaceae host plants, and soil histories. Taxa cluster maps were generated using compare_groups in the metacoder package (91), where the nonparametric Wilcoxon rank sum tests determined if a randomly selected abundance from one group was greater on average than a randomly selected abundance from another group. As the statistical test was performed for each taxon, we used an FDR correction on the P values to account for the multiple comparisons. When the comparison was between more than two groups, the differential abundances were plotted onto the taxa cluster map using heat_tree_matrix (91).

Second, indicator species analysis was used to detect ASVs that were preferentially abundant in predefined environmental groups (roots or rhizosphere, soil histories, or Brassicaceae host). A significant indicator value was obtained if an ASV had a large mean abundance within a group, compared to another group (specificity), and had a presence in most samples of that group (fidelity) (93). The fidelity component complemented the differential abundance approach between taxa clusters, which only considered abundance. We performed an indicator species analysis for the ASVs identified in the test phase of trial 1, and then trial 2. From the indicspecies package (94), we used the multipatt function with 9,999 permutations. As the statistical test was performed for each ASV, we used the FDR correction on the P values to account for multiple comparisons.

β-Diversity of the test phase Brassicaceae rhizosphere and root communities.

To test for significant community differences between trials, compartments, soil histories, and Brassicaceae hosts, we used nonparametric permutational multivariate ANOVA (PERMANOVA), where any variation in the ordinated data distance matrix was divided among all the pairs of specified experimental factors. The PERMANOVA was calculated using the adonis function in the vegan package (95), with 9,999 permutations, and the experimental blocks were included as strata. Our PERMANOVA used a distance matrix calculated with the Bray-Curtis formula and tested the significance of the effects of soil history, Brassicaceae host, and compartment.

We used a variance partition, as a complement to the PERMANOVA, to model the explanatory power of soil history, Brassicaceae host, and soil chemistry in the structure of the test phase Brassicaceae oomycete communities. We then quantified how each significant factor (i.e., the explanatory variables) impacted oomycete community structure with a distance-based redundancy analysis (db-RDA) (93). First, singleton ASVs were removed before the phyloseq data were transformed using Hellinger’s method, such that ASVs with high abundances and few zeros were treated equivalently to those with low abundances and many zeros (96). With the vegan package (95), soil chemistry was standardized (93) using the decostand function. We modeled the explanatory power of each experimental factor in each compartment from both experiments with a variance partition of a partial RDA, using the varpart function, and a Bray-Curtis distance matrix (97). Variation in the oomycete community data not described by the explanatory variables were quantified by the residuals. Finally, to quantify the amount of variation described by each explanatory factor, db-RDA results were calculated using the capscale function. Colinear variables were only identified in the trial 2 soil chemistry, such that specific variables were removed without a loss of information. We subsequently removed total carbon from both the trial 2 root and rhizosphere RDAs, as well as the zinc concentration from the root RDA. The final plots were generated using phyloseq (90).

Coinertia analysis of relationship between oomycetes and bacterial communities.

We used a coinertia analysis (93, 98) to compare how each sample was influenced by different ASVs, as a means to evaluate the strength of the soil history effect. Briefly, this analysis identifies the relationship between two data sets from a common sample by projecting that sample into a common multivariate space. For this analysis, the two data sets were the oomycete ASVs identified in this study—and structured by soil history—and the bacterial ASVs identified from the same experiment and extracted DNA samples, but these were not structured by soil history (4). This type of analysis is appropriate for exploring relationships in species-rich data sets, for example, where there are more ASVs than sites, and it imposes no assumptions on the data sets, such as co-occurrence or interactions (93).

The analysis identified the axes of the common coinertia space that represented the greatest inertia, or spread, of the common data. The analysis then compared how the positions of each sample in the new coinertia space were influenced by particular bacterial or oomycete ASVs. The directions of the arrows indicate how a sample was influenced by bacterial ASVs (tail) compared to oomycete ASVs (head); samples with shorter arrows are more similar (93, 99). Coinertia analysis also evaluated a RV coefficient (R = correlation, V = vectorial), a multidimensional correlation coefficient equivalent to the Pearson correlation coefficient. A higher RV indicates a stronger relationship between the oomycete and bacterial community matrices (93, 100).

Therefore, if soil history is particularly significant to the relationship between the bacterial and oomycete ASVs, the samples might be clustered by soil history on the common coinertia space. Alternatively, a weaker soil history may only be reflected in a shift from bacterial ASVs to oomycete ASVs. This might be plotted by longer arrows oriented toward common oomycete ASVs, which represent the differences by soil histories. Finally, if soil history has little to do with the relationship between the bacterial and oomycete ASVs, there may be no discernible pattern in how the samples are plotted, and the arrows between communities would be short.

First, to facilitate the analysis, we reduced the bacterial phyloseq objects by removing any ASVs that occurred only once. The phyloseq objects for the oomycete and bacterial communities from the roots and rhizosphere from both field trials were transformed using Hellinger’s method. Finally, each oomycete-bacteria sample pair were subjected to coinertia analysis using the coinertia function from the ade4 package (101). The large number of ASVs identified here and from the previous study (4) precluded us from plotting the ASVs onto the coinertia plane. However, as noted previously (90), these are not essential to the coinertia analysis.

Data availability.

Sequencing data and metadata are available at NCBI Bioproject under accession number PRJNA849532.