Introduction

Plant parasitic nematodes (PPN) pose a serious threat to crop production in intensive cropping systems (Molendijk and Korthals 2005). Among others, root-lesion nematodes (Pratylenchus spp.) and root-knot nematodes (Meloidogyne spp.) are especially problematic in the Netherlands. After the reduction of the use of chemical control agents, current strategies for nematode control in practice mostly consist of crop-rotations with non-host crops and cover-cropping with marigold and other cover crops (Hamont 2010). However, these measures often perform insufficiently in the long-term and restrict the selection of crops that can be planted, especially for more generalist PPN like Pratylenchus penetrans. Therefore, research is still ongoing to identify sustainable solutions for PPN control.

It has been reported that PPN can be controlled by natural suppressiveness of soil, suppressive soils being characterized as soils where pathogens cannot establish, persist or cause disease (Silva et al. 2018; Topalović et al. 2020). PPN suppressiveness specifically is reported to rely on the presence of specific microorganisms with antagonistic activity against nematodes. Those can include parasitizing bacteria, such as Pasteuria penetrans, nematode trapping fungi, microorganisms producing nematocidal compounds, but also those inducing systemic resistance in plants (Topalović et al. 2020). This suppressiveness can develop as a consequence of continuous monocropping and plants recruiting beneficial microorganisms from the soil. Attempts have been made to isolate microorganisms with direct suppressive effects for application as biological control agents (Akhtar and Malik 2000). But also, several soil management strategies have been investigated as more feasible approaches for inducing suppressiveness.

Organic as opposed to conventional land management has been proposed to increase soil suppressiveness to several soil borne pathogens as the addition of organic matter in the form of compost or farmyard manure increases the soil microbial activity and can increase microbial diversity (Letourneau and van Bruggen 2006). However, not all species of nematodes are sufficiently suppressed by organic management practices (Briar et al. 2016). A number of studies suggest that amendment with organic substances can lead to PPN suppressiveness, both by toxic compounds released from incorporated plant material and by stimulating antagonistic microorganisms. Specifically, chitin amendment was demonstrated to be effective against nematodes as it stimulated microorganisms with the ability for chitin solubilization (Cretoiu et al. 2013). Other soil treatments, such as anaerobic disinfestation or solarization (i.e. solar heating of soil covered by a plastic film) have also occasionally been reported to reduce nematode numbers (Butler et al. 2012). In addition, it has been suggested to use treatment combinations, for example by integrating marigold cover-cropping with organic amendments and solarization as this combination might result in longer-lasting effects, but this possibility has been poorly studied yet (Hooks et al. 2010).

In spite of many studies on the role of organic amendments and cover-crops to control PPN populations, long-term dynamics under different arable land management practices and soil treatments are not well understood. On the one hand, some practices might take several years to show effect, such as the adoption of organic management practices. On the other hand, some treatments might only show transient effects, requiring continuous cost-intensive intervention. The Soil Health Experiment in the Netherlands was established in 2006 with the purpose of studying long-term changes in soil biological parameters and PPN numbers in response to both land management practices and different soil health treatments (Korthals et al. 2014) applied to the same soil type, specifically a Gleyic Podzol with a low clay content (Kurm et al. 2023). Here we analyzed data from two land management systems (organic vs. conventional) and three intermittent soil health treatments (SHTs), anaerobic disinfestation (AD), a combination of marigold cover-cropping, compost and chitin amendment (CB) and a control (CT) from seven years to assess the effect on the soil microbial community and the population of P. penetrans (Pp). These three treatments were selected since both CB and AD have previously been demonstrated to reduce Pp numbers in the conventional field plots compared to CT in same SHE (Korthals et al. 2014). A notable feature of this long-term experiment is that while conventional and organic land management systems are continuously applied in all years (as is done in practice), the SHTs are applied twice in the seven years reported in this study.

We investigated the following hypotheses: (1) Organic land management decreases Pp numbers and increases yield; (2) Anaerobic disinfestation and a combination of marigold, compost and chitin decreases Pp numbers and increases yield; and (3) Decrease in Pp numbers is correlated with changes in the soil microbial community composition.

Materials and methods

Experimental setup

The present study made use of a long-term Soil Health Experiment (SHE) at research station of Wageningen University & Research in Vredepeel, The Netherlands (51° 32’ 27.10’’ N and 5° 51’ 14.86’’ E) between 2006 and 2013. For more information on the SHE we refer to Korthals et al. (2014) and Lupatini et al. (2017). Briefly, the gleyic podzol has been in use as an arable field since 1955 and had a history of problems with Pp. In 2006 the field was divided into 4 blocks, within each block two main plots with an organic or conventional farming system and 10 subplots (6 × 6 m) with different soil health treatments (SHTs) in a split-plot design. Both land management systems received the same amount of nutrients per hectare according to fertilizer recommendations per crop, yearly. The conventional system plots were given liquid cattle manure and mineral fertilizer and the organic system plots, organic cattle farmyard and/or liquid manure (Korthals et al. 2014; Kurm et al. 2023; Lupatini et al. 2017). For the present study, three SHTs were selected, the control (CT), anaerobic soil disinfestation (AD) and a combination of compost, chitin and marigold (CB); for a detailed description see Table 1. The SHTs were applied twice within the period of the study, between cropping seasons after harvesting of barley/wheat (see below), in 2006–2007 and in 2009–2010. The crop rotation between 2006 and 2011 was as follows: 2006: cereal (wheat in conventional/ barley in organic); 2007: potato; 2008: lily; 2009: cereal (wheat in conventional/ barley in organic); 2010: potato; 2011: carrot and followed by 2012 and 2013: maize. Only after harvest of potato and maize, green manure was applied, i.e. forage rye in the conventional and black oat in the organic system plots.

Table 1 Soil Health treatments
Full size table

Soil sampling

Soil was sampled in 2006 in July after harvest and before application of the SHTs. In the following years, sampling was done in spring, before the new cropping season. Due to practical management requirements sampling in 2007, 2008 and 2010 was done shortly after the yearly fertilization and in 2009, 2011 and 2013 shortly before. In all years, soil was sampled from the upper 25 cm with a soil auger (diameter 13 mm), each sample consisted of 30 cores taken in a regular pattern from the center (1.5 m × 2.67 m) of each plot. The sampled soil was mixed per plot and subsamples were used for molecular analyses and reference purposes (2 samples of 10 ml), nematode extraction (100 ml) and, for a limited set of samples, also biochemical analyses (1 kg). For a complete description of the biochemical analyses, nematode extraction and microscopic quantification and identification of Pp we refer to Korthals et al. (2014). The duplicate samples for either molecular analyses or reference purposes were stored at − 80 °C, until further use. Biochemical parameters are reported in table S1. Available indications (based on van Beers et al. 2010) on damage thresholds and multiplication of Pp, sensitivity and potential loss in agricultural soils in the Netherlands for the crops used in this SHE are listed in table S2.

Yield

For a detailed description of harvest and determination of yield see Korthals et al. (2014). In this study, the following parameters were used for net yield, for all crops as ton/ha: potato 2007 and 2010: marketable yield for tubers > 30 mm; lily bulbs 2008: marketable yield; barley/wheat 2009: dry matter; carrot 2011: marketable yield of carrots between 20 and 60 mm. Maize in 2013 was harvested with a forage harvester and fresh weight was determined. A subsample of approx. 1 kg was used for determining the dry weight, which was used to calculate the yield as dry matter. In addition, for carrot in 2011, visible damage (i.e. presence of stubby roots) caused by Pp was assessed.

In 2007, potatoes in the organic plots were infected with Phytophtora infestans and the crop had to be destroyed according to Dutch law. Therefore, no yield data is available for these plots. Moreover, it was observed that the application of chitin (in CB) shortly before planting resulted in a temporary phytotoxic effect on potato. In the second application of SHTs, chitin was applied earlier (November 2009) and no phytotoxic effect was observed.

DNA extraction, amplification and sequencing

Soil DNA was extracted using the MoBio PowerMag soil DNA Isolation KF Kit (MoBio Laboratories, Inc., Carlsbad, CA, USA) as described by Andreo-Jimenez et al. (2021). Briefly, the manufacturers protocol was adapted to facilitate a higher input weight of 1 g of soil per sample. The DNA concentration of the DNA eluates was measured with a Pico Green assay (Quant-IT Pico Green dsDNA assay kit; Invitrogen) on a Tecan Infinite M200Pro (Tecan Group, Ltd.); DNA was diluted to 4 ng/µl with the elution buffer of the DNA extraction kit. The DNA eluates and diluted DNA samples were stored at -20 °C.

For the PCR amplification of the bacterial 16 S V3-V4 rRNA gene region, universal primers E341F (S-D-Bact-0341-b-S-17: 5′-CCTACGGGNGGCWGCAG-3′) and 805R (S-D-Bact-0785-a-A-21: 5′-GACTACHVGGGTATCTAATCC-3′) were used (Herlemann et al. 2011; Klindworth et al. 2013), and primers ITS86F (5′-GTGAATCATCGAATCTTTGAA-3′) and ITS4 (5′-TCCTCCGCTTATTGATATGC-3′) were used for the fungal ITS2 region (Op De Beeck et al. 2014; White et al. 1990). All primers were synthesized with the universal adapters for Illumina MiSeq sequencing (Integrated DNA Technologies, BVBA, Leuven, Belgium). To reduce bias, PCRs were designed with a high DNA target input, in combination with low number of PCR cycles and the use of a proof-reading polymerase (Gohl et al. 2016; Lindahl et al. 2013). PCR mixtures contained 1× Q5 reaction buffer, 200 µM deoxynucleoside triphosphates (dNTPs), 0.5 µM of each primer, 20 ng template DNA, 1 U Q5 Hot Start High-Fidelity DNA polymerase (New England Biolabs, MA, USA), and nuclease-free water to a final volume of 50 µl. The PCR conditions for 16 S V3-V4 samples were 98 °C for 30 s, followed by 10 cycles of 98 °C for 10 s, 55 °C for 30 s and 72 °C for 30 s, and a final elongation at 72 °C for 2 min in a Veriti Thermo Cycler (Thermo Fisher Scientific, USA). For ITS the same protocol was used with 16 cycles and with an annealing temperature of 51 °C. Eight replicate PCRs were performed in separate runs for each sample for both primer sets, and PCR replicates were pooled and stored at − 20 °C. Amplicons were purified and libraries were prepared according to Illumina guidelines (Illumina, San Diego, CA, USA) and paired-end sequenced with 2 × 250 cycles on an Illumina MiSeq platform and all reads were demultiplexed (Next Generation Sequencing Facilities, Wageningen University & Research, Wageningen, The Netherlands). Because the total number of samples to be sequenced exceeded the number of Illumina barcode combinations (96) available at time of sequencing, the samples were sequenced in 2 separate runs. For 16 S, samples were resequenced when the number of reads was below 60,000, to ensure sufficient sequencing depth. An overlap of 16 samples (year 2011, combination and control treatments) was integrated in the set-up of the PCRs described above, for both sequencing runs to check for potential significant differences in bacterial and fungal beta diversity. When there were no significant differences, repeated samples from different sequence runs could be pooled. The raw sequencing data used for this study are available on the NCBI sequence read archive (SRA) under BioProject number PRJNA1047570.

Sequence processing

The demultiplexed sequencing data was preprocessed with Qiime2 version 2019.10 (Bolyen et al. 2019). The reads were denoised with DADA2 (Callahan et al. 2016) using the q2-dada2 plugin, including merging of the forward and reverse reads for 16 S rRNA gene V3-V4 data. For ITS2 data only forward reads were used, due to variable length of the ITS2 region and prior to DADA2 processing potential read-through into opposite primer region was removed with cutadapt (Martin 2011) as a qiime2 plugin. Sequencing replicates for the same sample were pooled. Resulting amplicon sequence variants (ASVs) were filtered to remove low abundant ASVs with total abundance below 10. Taxonomy was assigned to the ASVs using the q2-feature‐classifier (Bokulich et al. 2018). The naïve Bayes classifiers for the bacterial and fungal data were trained (Pedregosa et al. 2011) on extracted V3-V4 sequences of the Silva release 132 16 S/18S reference database (Quast et al. 2012) and the full length ITS Unite database (QIIME release for all eukaryotes v.8.0) with dynamic use of clustering thresholds, respectively (Nilsson et al. 2019). ASVs were additionally filtered on taxonomy. ASVs generated from 16 S rRNA gene sequences that remained unassigned or were assigned to Eukaryotes, Chloroplast or Mitochondria and ASVs without taxonomic information at phylum level were removed. For ITS2, only ASVs that were identified as belonging to the Kingdom Fungi were kept.

Data analysis

All further analyses were carried out with R (version 4.3.1) in RStudio 2023.06.1. The effect of land management system, SHT and year on densities of Pp (number/100 ml soil) and on yield was analyzed with a linear mixed model (package lme4 vs. 1-1.34, Bates et al. (2015) using the log10 of Pp densities as the response variable and land management system, SHT, and year as the explanatory variable and system nested in block as random variable. Pairwise differences were determined with the emmeans function (package emmeans vs. 1.8.7, (Lenth 2018).

Alpha diversity of the bacterial and fungal communities was calculated from rarefied data using the package microeco vs. 1.0.0 (Liu et al. 2021). Both the Shannon H/H’ and inverse Simpson indexes were calculated per system and per soil health treatment per year.

For analyzing the beta-diversity the unrarefied data was used as relative abundances (proportions). Permutational multivariate analysis of variance (permanova) was conducted to determine the significance of the effects of several factors on the bray-curtis distance of the fungal and bacterial community using the function adonis2 (package vegan vs. 2.6-4) (Oksanen 2010). The factors analyzed were land management system, SHT, year, year after treatment and timepoint of sampling. SHT was coded as a separate factor for the year 2006 since in that year no treatment had been applied yet. Year was always coded as a categorical factor. Year after treatment was defined as the number of years after the SHTs were applied. This results in: 2006=-1, 2007/2010 = 0, 2008/2011 = 1, 2009 = 2, 2013 = 3. Year after treatment was coded as a categorical factor. Finally, the timepoint of sampling denoted if sampling was done before fertilization (2009, 2011, and 2013) or after (2007, 2008 and 2010) In 2006 sampling was done in July and was therefore coded as sampling before as well. Pairwise permanovas were conducted with pairwise.adonis2 (package vegan). For all models see Table 2.

Table 2 Models used to analyze the effect of land management system (= system), soil health treatment (= SHT), number of years after treatment and timepoint if sampling on the Bray-Curtis distance of fungal and bacterial communities
Full size table

In addition, a permanova test was performed with Pp densities as the explanatory variable with the densities grouped as seven categories (defined as P1 to P7 with 0–10, 11–50, 51–100, 101–200, 201–500, 501–900 and > 900 densities of Pp per 100 ml soil, respectively). It was not possible to conduct a similar analysis for yield because of the large differences in yield between the different crops.

For identifying taxa that were related to the observed differences between systems, SHTs and years, the unrarefied sequencing data was additionally filtered by removing all ASVs that did not occur in at least 25% of all samples. Subsequently ASVs were agglomerated per genus using the package phyloseq vs. 1.44.0 (McMurdie and Holmes 2013). Random forest predictions based on the compositions in groups of samples were conducted to find differentiating taxa between system combined with year, treatment combined with year and combined with year after treatment and for timepoint of sampling. Random forest analyses were carried out using the function randomForest (package randomForest vs. 4.7–1.1, Liaw and Wiener (2002) with Blocks as strata. In addition, a random forest analysis was conducted for Pp densities coded as categories (as described above) to find the most important taxa in predicting low Pp densities.

Results

Effect of land management system and soil health treatments on P. penetrans and yield

There was no effect of ‘land management system’ on the number of P. penetrans per 100 ml soil (χ2 = 0.084, p = 0.772). However, there was an effect of ‘year’ (χ2 = 68.97, p = < 0.001) and an interaction between ‘year’ and SHT (χ2 = 65.23, p = < 0.001). The number of Pp significantly decreased in 2007 compared to 2006 and subsequently increased in 2008 and 2009, followed by another decrease in 2010 and increase in the following years (Table 3).

Table 3 Average densities of pp per 100 ml soil with standard deviation in the measured years and the three soil health treatments
Full size table

In both 2006 and 2013, there was no difference in Pp densities between the three treatments, but in all other years the CB treatment contained significantly less Pp than the CT and AD treatments (Table 3). Only in 2010 also the AD treatment contained significantly less Pp than the CT.

In 2010 and 2011, the two years following the second SHT application in 2009, there were significant differences in net yield of potato and carrot, respectively, between the SHTs with and interaction with land management system (2010: χ2 = 15.70, p = < 0.001; 2011: χ2 = 6.27, p = 0.04). Net potato yield was significantly higher in the AD treatment in the organic system and in the AD and CB treatment in the conventional system compared to CT (Table 4). In the conventional system the highest yield was obtained from the CB treatment. Net carrot yield was in both systems higher in the AD and CB treatment compared to CT and in the conventional system, yield was highest in the CB treatment. No visible damage of Pp on carrot roots was observed. Maize yield in 2013 did not differ significantly between the SHTs, but was significantly higher in the conventional system compared to the organic system.

Table 4 Average net yield as ton per hectare with standard deviation of the crops in years 2007–2011 and 2013, for the three SHTs and the two land management systems
Full size table

Effect of land management system on the microbial community

There were no significant differences in beta diversity Bray-Curtis distances between the samples that were replicated in 2 sequencing runs for both 16 S rRNA and ITS data (permanova test analysis results: F = 1.274, p = 0.128 and F = 0.212, p = 0.978, respectively). Subsequently, all replicated samples were pooled after preprocessing steps.

The 16 S rRNA sequencing data were rarefied at 25,000 and the ITS data at 2500 per sample for alpha diversity analysis. Samples with library size below 25,000 for 16 S rRNA data (2 samples) and 2500 for ITS data (3 samples) were removed prior to all analyses to reduce library size differences and sparsity within the data sets.

The rarefied dataset of bacterial 16 S rRNA and of fungal ITS sequences contained 51,080 and 1700 unique ASVs, respectively. The alpha diversity Shannon-index showed a significantly higher bacterial diversity in the organic system in the years 2006, 2009 and 2013 (Fig. S1a), while the inverse Simpson index D2 showed no differences between the two systems (Fig. S1b). Fungal diversity was higher in the organic system in the years 2009, 2010 and 2013 for both indices (Fig. S1c, d).

The unrarefied dataset of bacterial 16 S rRNA and fungal ITS sequences contained 53,976 and 1994 unique ASVs, respectively. Permanova analysis on beta diversity Bray-Curtis distances showed that there was a significant effect of system (F = 6.550, R2 = 0.027, p = 0.001) and year (F = 7.678, R2 = 0.190, p = 0.001) and their interaction (F = 1.436, R2 = 0.035, p = 0.026) on the bacterial community composition (Fig. 1a). The combination of land management system with year revealed that in the years 2006, 2008, 2009 and 2013, the community composition in the organic system was significantly different from the conventional system (Table S3). Also, the timepoint of fertilization had an effect (F = 8.008, R2 = 0.046, p = 0.001), but there was no interaction with land management system (F = 0.874, R2 = 0.005, p = 0.609).

Fig. 1
figure 1

Microbial community composition per year and land management system. Axis 1 and 2 of a PCoA analyses of (A) the bacterial community composition and (B) the fungal community composition per year; color indicates land management system

Full size image

Similarly, there was a significant effect of system (F = 13.838, R2 = 0.041, p = 0.001) and year (F = 16.830, R2 = 0.298, p = 0.001) and their interaction (F = 2.321, R2 = 0.041, p = 0.001) on the fungal community composition (Fig. 1b). The community compositions in the conventional and organic system were significantly different from each other in 2008, 2009, 2010, 2011 and 2013 (Table S4). Furthermore, there was a difference between timepoint of fertilization (F = 6.873, R2 = 0.039, p = 0.001), but no interaction with system (F = 1.111, R2 = 0.006, p = 0.303).

Effect of soil health treatments on the microbial community

Because the SHTs were applied twice, between the growing seasons 2006–2007 and 2009–2010 and not yearly, we also assessed the effect of time after SHTs. Therefore, an additional variable of number of years after treatment was introduced.

For the bacterial communities, a significant effect of SHT (F = 6.930, R2 = 0.114, p = 0.001) was found and an interaction with time past SHT (F = 2.029, R2 = 0.050, p = 0.001). There was no significant interaction between land management system and SHT (F = 0.969, R2 = 0.016, p = 0.448). Calculating the pairwise interaction between SHT for each year past treatment showed that there were significant differences in all years but there was no difference between the plots in 2006, before SHT applications. The amount of variance explained was highest in the years immediately following the SHT, which suggests larger differences between the SHTs in these years (Table S5, Fig. 2a).

Fig. 2
figure 2

Microbial community composition per year and SHT. Axis 1 and 2 of a PCoA analysis of (A) the bacterial community composition and (B) the fungal community composition per year; color indicates SHT; the SHTs noCT, noAD, noCB indicate the plots of CT, AD and CB in 2006 before application of SHTs

Full size image

For the fungal communities as well, there were significant effects of treatment (F = 13.453, R2 = 0.159, p = 0.001) and an interaction between SHT and year past treatment (F = 2.622, R2 = 0.046, p = 0.001), but not between land management system and SHT (F = 0.870, R2 = 0.010, p = 0.655). Differences between treatments were significant in all years past treatment except in 2006 (before the SHTs were applied). The amount of variance explained was again higher in the year after applying the SHTs (Table S6, Fig. 2b).

Differential abundance

Bacteria

Several random forest analyses were conducted to determine which bacterial taxa were most important in differentiating between the land management systems and SHTs. For the top 20 most important taxa the relative abundances in the respective land management system and SHT were calculated. The relative abundances of the bacterial taxa Saccharomonospora, Herbinix and unspecified bacterial taxa belonging to the order Fibrobacterales and the family Ruminococcaceae were significantly higher in the organic system than in the conventional system in the years after the start of the experiment (Fig. 3a, Table S7). In addition, an analysis with timepoint of fertilization (Table S8) showed several taxa that were more abundant in years with sampling after (e.g. Fermentimonas, Gelidibacter) and a few before fertilization (e.g. Nitrosomonadaceae MND1, Isosphaeraceae).

Fig. 3
figure 3

Bacterial differential abundance per year in combination with land management system and SHT. Heatmap of the log10 frequencies of the 20 most important bacterial taxa in a random forest analysis with (A) system and year and (B) treatment and year; the colored bar shows the respective phylum affiliation

Full size image

Bacteria belonging to the phylum Firmicutes, the genera Aneuribacillus, Clostridium and Bacillus were more abundant in the AD treatment as determined by a random forest analysis with SHT and year as an explanatory factor (see Table S9). The taxa Paeniglutamicibacter and Tomitella, Arenibacter, Aquamicrobium and Paenisporosarcina were more abundant in the CB treatment (Fig. 3b). Notably, Arenibacter, Paeniglutamicibacter and Tomitella are more abundant in the years 2007 and 2010 directly following the treatment CB. An additional analysis with SHT and number of years after treatment showed that the taxa Stenotrophomonas and Oerskovia were also among the taxa separating samples from CB treatments in the years directly after treatment (Fig. S2).

Fungi

For fungi, the taxa with the highest importance for predicting differences between system and year included several taxa that were more abundant in the organic system in most years. These were the taxa Articulospora, Holtermanniella, Cutaneotrichosporon, Bipolaris and Paraphaeosphaeria (Fig. 4a, Table S10). In addition, a number of taxa was more abundant when sampling was conducted before fertilization as compared to after fertilization (Table S11), such as the taxon Tetracladium.

Fig. 4
figure 4

Fungal differential abundance per year in combination with land management system and SHT. Heatmap of the log2 frequencies of the 20 most important fungal taxa in a random forest analysis with (A) system and year and (B) treatment and year; the colored bar shows the respective phylum affiliation

Full size image

The taxa with the highest importance for separating the SHTs combined with year (see Table S12) included Clohesyomyces and Pseudeurotium, which were more abundant in the AD treatment compared to the other treatments (Fig. 4b). The taxa Mortierella and an unclassified Microascaceae were more abundant in the CB treatment. Analysis of treatment combined with year after treatment showed that Mortierella and Microascaceae were more abundant in the CB treatment in the year after the treatment, while a number of other taxa were lower abundant following application of this treatment (Fig. S3). Pseudorotium was more abundant in the AD treatment in the year after application.

Association between the microbial community composition and P. penetrans abundance

Bacteria

The bacterial community differed significantly between the Pp density categories (defined as P1 to P7 with 0–10, 11–50, 51–100, 101–200, 201–500, 501–900 and > 900 densities of Pp per 100 ml soil, respectively) (F = 4.114, p = 0.001, Fig. 5a). Especially the lower Pp density P1 and P2 categories were different from the remaining categories (Table S13). A random forest analysis was used to identify the most important taxa contributing to differences in the microbial community between the 7 different categories of Pp infestation. The bacterial taxa more abundant at low Pp densities were Stenotrophomonas, Tomitella, Arenibacter, and a taxon of the family of Rhizobiaceae (Fig. 5b, Table S14). Several taxa, such as Brevundimonas, Candidimonas, Aquamicrobium and Paeniglutamicibacter were more abundant at P1 and partly P2.

Fig. 5
figure 5

Bacterial community composition and differential abundance per Pp density category. (A) PCoA of the bacterial communities in all samples; colors represent the seven Pp density categories (defined as P1 to P7 with 0–10, 11–50, 51–100, 101–200, 201–500, 501–900 and > 900 densities of Pp per 100 ml soil, respectively) and shapes represent the three SHTs and plot before SHT application (noCT, noAD and noCB) in 2006, (B) Heatmap of the log2 frequencies of the 20 most important bacterial taxa in a random forest analysis with Pp density category; the colored bar shows the respective phylum affiliation

Full size image

Fungi

The fungal communities differed significantly between Pp density categories (F = 6.148, p = 0.001, Fig. 6a). The categories P1 and P2 differed most from the other categories (Table S15). The genera Mortierella, Apiotrichum and a member of the Microascaceae were more abundant at low Pratylenchus densities (P1, P2) compared to higher densities, while taxa like Tetracladium and Minimedusa were more abundant at higher Pratylenchus densities (Fig. 6b, Table S16).

Fig. 6
figure 6

Fungal community composition and differential abundance per Pp density category. (A) PCoA of the fungal communities in all samples; colors represent the seven Pp density categories (defined as P1 to P7 with 0–10, 11–50, 51–100, 101–200, 201–500, 501–900 and > 900 densities of Pp per 100 ml soil, respectively) and shapes represent the three SHTs and plot before SHT application (noCT, noAD and noCB) in 2006, (B) Heatmap of the log2 frequencies of the 20 most important fungal taxa in a random forest analysis with Pp density category; the colored bar shows the respective phylum affiliation

Full size image

Discussion

Pratylenchus penetrans and yield

We hypothesized that the organic land management system would reduce the density of Pp in soil. However, in this study we found no significant effect of land management system and thus we reject our first hypothesis. Similar results were found by Quist et al. (2016), who also concluded that crop type had the largest effect on Pp abundance. In this study, all the crops used can be considered as moderate to good hosts for Pp (i.e. suitable for multiplication of Pp). Although Pp densities were lower in the CT plots after growing wheat and barley in 2007 and in 2009, we expect that this was the result of a longer fallow period after harvest (in July), compared to lily and carrot (in November) (Halbrendt and LaMondia 2004). Potato and maize were harvested in September followed by green manure crops until February.

Despite the fact that land management system had no significant effects, we still found significant effects of SHTs on nematode numbers. The CB treatment, a combination of compost, marigold and chitin resulted in significantly lower Pp numbers after the first application until 2011. Marigold has been demonstrated to consistently reduce numbers of Pp by producing nematocidal compounds (Hooks et al. 2010) and chitin and compost are assumed to induce changes in the soil microbial community that can inhibit nematode populations (Akhtar and Malik 2000; Briar et al. 2016; Renčo et al. 2009). Here we demonstrate that this effect can remain for up to 3 years after the treatment. This is in accordance with a study performed by Evenhuis et al. (2004), wherein it was found that Pp was present in lower densities to control for at least 3 years after a crop rotation with marigold. The present design cannot distinguish whether the combination with compost and chitin amendment had an additional effect on nematode numbers. An earlier study in the same long-term experiment reported no significant difference in Pp populations between a marigold-only and the combination treatment in the conventional system (Korthals et al. 2014). In the same study also the chitin treatment alone significantly reduced Pp, while compost alone resulted in higher Pp numbers, although not significant. This indicates that marigold could be the main factor influencing Pp densities with a potentially additional effect of chitin. However, the CB treatment was selected for analysis in this study, since it was assumed that a combination of effective treatments might elicit a more long-term effect. In the present study we find that the positive effect was only lost in 2013. This was both four years after the last treatment and after consecutive planting of moderately (carrot in 2011) to good hosts (potato in 2010 and maize in 2012 and 2013) for Pp.

The AD treatment led to a significant reduction of Pp densities directly after the treatment in 2010. Anaerobic disinfestation has previously been found effective against Pp with studies indicating that both the anaerobic conditions and production of toxic compounds in combination with changes in the microbial community were responsible for reduced nematode numbers (Strauss and Kluepfel 2015). However, the effect on Pp numbers was not as strong as with the CB treatment. Moreover, the effect was already lost after one year, while long-term effects have also been reported, especially in the context of apple replanting disease in tree orchards. A study by Hewavitharana and Mazzola (2016) suggested that the carbon sources used for an AD treatment can play an important role in effectiveness. In the present study a mixture of rye grass species was used as carbon source. Grass has been found to be highly effective when used in AD, but it remains unclear if there are differences between grass species.

In accordance with the Pp abundances, yield was significantly higher in CB and AD treatments in 2010 and 2011 after the second application in 2009. Thus, in support of our second hypothesis, both yield and Pp abundances were affected by the SHTs. After the second application of the SHTs in 2009 in both AD and CB the Pp abundances decreased below the potential damage threshold of 100 Pp per 100 ml soil for potato and carrot. In the case of CB a reduction below this threshold could already be observed after the first treatment application in 2006. This further indicates that the SHTs increased soil suppressiveness to Pp, although this cannot conclusively be proven with the available measurements. The lack of a significant effect in 2007 could have been due to the initial phytotoxic effect on potato of chitin in the CB treatment and due to the loss of the organic crop because of Phytophotora infestans infection.

Microbial community composition

In our third hypothesis we expect an association between Pp numbers and microbial community composition. In support of this hypothesis, the bacterial and fungal community differed from the control both with the CB and AD treatment as did Pp numbers and yield. Anaerobic disinfestation is known to lead to an increase in facultative or obligate anaerobic bacteria belonging to the Firmicutes, such as Clostridium (Streminska et al. 2013). Similarly, the increased fungal taxa, such as Pseudeurotium and Clohesyomyces have previously been found in anaerobic environments (Chen et al. 2020). Which groups thrive or survive after anaerobic disinfestation can be dependent on the source or the incorporated organic material (Poret-Peterson et al. 2020) (see Sect. 4.4). Of the individual constituents of the combination treatment (compost, chitin, and marigold) marigold has been shown to have a lesser and compost and chitin amendment a larger effect on the soil microbiome (see for example Cazzaniga et al. (2023); Kraut-Cohen et al. (2023); Andreo-Jimenez et al. (2021), but there is as of yet no report on their combined effect. In accordance with these studies, we found higher abundances of several bacterial taxa belonging to the Actinobacteria and Bacteroidetes. These taxa were reported to grow on complex substrates, such as compost and chitin (Han et al. 2022). In addition, fungal taxa were found, which were associated with compost and chitin amendments (Li et al. 2018; Ootsuka et al. 2021).

Just as Pp numbers and yield were most strongly affected in the years after SHT application the differences in microbial community composition between the treatments were more pronounced directly after treatment. Especially the CB treatment differed strongly from the other treatments in 2007 and 2010. In the following years, these differences became smaller, but they remained significant throughout the remainder of the study. This contrasts with results by Lupatini et al. (2019) who did not find a strong effect of the combination treatment on the fungal community composition in 2013 within the same long-term experiment. However, their findings were not completely comparable as, besides some differences in sampling procedure and time, the samples were analyzed with a different target region for the PCRs (18 S rRNA versus ITS region), PCR conditions and bioinformatic approach. Our results showed that a strong disturbance, such as elicited by the combination of treatments in this study can have a long-term effect on the soil microbial community. This result is notable, since previous studies mainly found short term effects of commonly applied soil treatments, such as solarization and organic amendments (Kanaan et al. 2018; Kraut-Cohen et al. 2023). Long-term shifts in the microbial community can lead to a more sustainable improvement of soil health in agricultural systems and the provision of important ecosystem services.

We also found that both bacterial and fungal communities differed between the organic and conventional system, although the effect of land management system was less strong than the effect of SHT. This is in accordance with many studies that found differences in community composition (e.g. Bonanomi et al. (2016); Pershina et al. (2015). Often these differences are linked to changes in the amount and type of organic material added and changes in pH (Hartmann et al. 2015; Orr et al. 2015). In the present study, we cannot assess differences in biochemical parameters between SHTs and between the land management systems because these parameters were not measured in all SHTs and both land management systems. For assessment of differences in biochemical parameters between SHTs, we refer to Korthals et al. (2014). Still, it is likely that also in the present experiment the use of cattle farmyard manure and slurry in organic system as compared to cattle slurry and mineral fertilizers in conventional system stimulated a different set of microorganisms able to degrade more recalcitrant organic materials, such as Saccharomonospora sp. and Herbinix sp. These genera are known to include thermophilic microorganisms and might originate from composted material incorporated into the organically managed soil (i.e. farmyard manure in this study). Still, no higher abundances of these thermophilic genera were observed when sampling was done directly after fertilization compared to before fertilization. Also, several fungal taxa were more abundant in the organic system. These included taxa closely associated with and involved in the degradation of plant tissue like Articulospora and Paraphaeosphaeria and several yeasts with as yet poorly documented function. Similar to bacteria, these taxa were not affected by sampling before or after fertilization. As neither yield nor Pp abundances were strongly affected by land management system, these differences in microbial community composition could be unrelated to these observations.

In this study we observe that the effect of the SHTs was stronger than the effect of land management system. Even though the land management practices are applied every year, there is no indication that the land management system effect increased over the years. Extrapolating from the available data between 2006 and 2013 it seems likely that the plots with different SHTs will continue to become more similar again with respect to microbial community composition, yield and Pp abundances until the next treatment is applied. Therefore, in order to maintain Pp suppressiveness, such treatments might have to be applied every two to three years. Differences between the organic and conventional system on the other hand are likely to persist. However, it should be noted that the present study was conducted in one location only and on a sandy soil. Since microbial communities are known to vary strongly between locations and soil types it is possible that the same treatments at another location might have a different effect or differ in the duration of the observed effect.

In agreement with the changes in microbial community composition in the SHTs, we observed that both the bacterial and the fungal community composition significantly differed between Pp abundance categories, giving further credibility to the proposition that the SHTs elicited a microbially mediated soil suppressiveness. However, it is likely that other factors, such as the direct toxic effect of marigold and the repeated cultivation of maize had additional effects on the Pp density in soil.

Disease suppressiveness against plant parasitic nematodes mediated by soil microorganisms has been reported previously. In their review Topalović et al. (2020) summarize current knowledge on bacterial and fungal taxa that have been associated with suppressiveness and their potential mode of action. A number of these taxa were also shown to be associated with suppressiveness in the present study. Especially members of the family Rhizobiaceae and the genus Rhizobium are known for their effectiveness against different genera of plant parasitic nematodes (e.g. Hallmann (2003); Wolfgang et al. (2019). It has been indicated that Rhizobiaceae are involved in inducing systemic resistance or the production of nematocidal compounds. Also, the bacterial taxon Stenotrophomonas was reported to produce nematocidal toxins (Huang et al. 2009) and a member of the fungal genus Mortierella has even been reported to feed on nematodes (DiLegge et al. 2019). Still, a number of taxa that were associated with low Pp numbers have not been reported previously, several of which were more abundant in the combination treatment. It is possible that this treatment stimulated a number of microbial taxa with effectivity against Pp. For example, chitin degrading Mortierella were shown to be stimulated by chitin amendment (Andreo-Jimenez et al. 2021). The observed long-term reduction in Pp densities indicates that besides abiotic factors, soil microbes may play an important role in nematode suppressiveness.

Challenges

It must be noted that the microbiome analyses in this study are based on amplicon sequencing data which are known to be prone to several sources of bias. For one, the primer choice is critical for determining which taxa will be amplified in a given sample prior to sequencing. We mitigated this bias as much as possible by carefully selecting primers that showed a wide coverage (Klindworth et al. 2013; Op De Beeck et al. 2014). Bias in PCR is further increased with increasing number of PCR cycles. Therefore, in this study we chose to subject eight subsamples to a low number of cycles and to pool the PCR products, which will reduce this source of bias. With respect to data-analysis amplicon sequencing data is known to be compositional and sparse. Compositionality implies that shifts in relative abundance might not represent shifts in absolute numbers of certain groups; for more details on underlying patterns that can explain the observed shifts, see Alteio et al. (2021) and Gloor et al. (2017). The analyses performed in this study, such as random forest, are usually found to perform well with compositional and sparse data (Thompson et al. 2019). Still, it must be considered that they are an indication of proportions of taxa and not absolute abundances. Another factor of importance is that within the soil DNA pool that can be amplified by PCR around 40% is estimated to be relic DNA, i.e. extracellular DNA or from non-viable cells (Alteio et al. 2021). With understanding of the nature and limitations of amplicon sequencing data, its use is and will remain a valuable approach in investigating the composition of microbial communities in soil. In addition, it is also a cost-effective tool prior to more targeted and expensive (sequencing and bioinformatics) “omics” approaches like metagenomics and/or metatranscriptomics to study the functional potential, actual activity and gene expression within the microbial communities (Alteio et al. 2021). More targeted approaches will be needed to gain more insights on the mechanisms behind nematode suppressiveness mediated by the microbiome as a result of SHTs.

Conclusions

In this study we showed that a combination of SHTs resulted in shifts in the microbial communities and led to a reduction in the plant parasitic nematode Pp densities in soil. This effect could be detected for several years after the treatment application. Our results provide evidence supporting that soil treatments might promote a nematode suppressive microbiome. Still, more mechanistic studies are needed to confirm this hypothesis.