Chitin- and Keratin-Rich Soil Amendments Suppress Rhizoctonia solani Disease via Changes to the Soil Microbial Community

Bioassays with 10 different organic products (Table 1) added to two sandy soils were performed in 2016 and 2017 to assess the effect of these organic amendments on disease suppressiveness against R. solani in sugar beet seedlings. Two control treatments (i.e., soil with only nitrogen fertilizer [control+N] and nonamended soil [control]), were included. All data from both years and both soils were analyzed together to explore treatment effects on the target disease. Data from 2016 from Vredepeel soil were excluded from the data set as explained in the Materials and Methods. The amendments with plant-2, manure-chitin, and all four keratin products significantly reduced the disease spread compared to the control with an equal dosage of nitrogen (control+N) (Fig. 1). This demonstrates a disease-suppressive effect of these products independent of nitrogen concentration.

Organic amendments had different effects on the measured biological soil properties. Potentially mineralizable nitrogen (PMN), an indicator for labile nitrogen potentially available for plants, showed the most distinct changes across different organic amendments. PMN was increased 2 to 7 times by all organic amendments, except for plant-2, whereas the highest increase occurred with the seed treatment, a product made from defatted brassica seeds (see Fig. S1 in the supplemental material). On the other hand, hot water extractable carbon (HWC), an indicator of labile carbon, did not differ among most treatments: only in the case of manure-chitin (Fig. S1). Soil bacterial biomass was higher in all organic amendment treatments, with the exception of plant-1 and plant-2. Fungal biomass and ergosterol were highest in soil with seed and manure-chitin amendments (Fig. S1). Ergosterol is used as an indicator to quantify living fungal biomass because this lipid is almost exclusively found in fungal cell membranes. However, despite being related parameters, ergosterol and fungal biomass (based on hyphal length) presented different responses to some of the other amendments. For example, plant-3 and manure amendments presented higher ergosterol content, whereas the fungal biomass was not different from that of the control.

Many other physicochemical characteristics were quantified in the amended soils (see Fig. S2 and Table S1 in the supplemental material). All treatments showed a higher availability of copper (Cu.a) and total manganese (Mn.t) content compared with the control. Remarkably, potassium concentration (K.a and K.t) was higher in all amended soils, except in the case of keratin-rich amended soils, whose concentrations were the same as the control. The manure-chitin treatment had very high levels of available calcium and selenium (Ca.a and Se.a), along with higher molybdenum (Mo.a), which was also high in plant-based amendments (Fig. S2). Phosphate concentrations (P.a and P.t) were higher in manure treatments. The rest of the soil variables measured were not different among the treatments.

One of the objectives of this study was to find soil properties that correlated with disease suppression. To achieve this, we performed correlation analyses between all soil properties, concentrations, and R. solani disease spread (number of infected plants). The disease spread had a significant (P < 0.001) negative correlation with the biological soil properties PMN (Pearson, R = −0.36) and HWC (Pearson, R = −0.28), demonstrating that amended soil was more suppressive of R. solani at higher PMN and HWC values (Fig. 2). Other extremely negative correlations (P < 0.001) were between suppression of R. solani and chemical products such as total manganese (Mn.t) and selenium (Se.a) and available calcium (Ca.a), zinc (Zn.a), and copper (Cu.a). Furthermore, available potassium (K.a) and sulfur (S.a) (P < 0.01) were also negatively correlated with disease levels, but to a lesser degree. On the other hand, available phosphate (P.a) and calcium carbonate (CaCO₃) showed a slight positive correlation with disease levels (P < 0.05) (Fig. 2).

Since the organic products had been selected for their variation in C/N ratios and respiration rates (Table 1; see supporting information at https://doi.org/10.4121/12971528), it was relevant to test if these properties can predict disease suppression after the soil was amended with these products. Interestingly, R. solani disease levels in amended soils correlated negatively with the respiration rate of the organic products (R = −0.8, P = 0.0045), whereas the C/N ratio of products showed no significant correlation with the disease (R = 0.5, P = 0.14).

A total of 10,577,545 internal transcribed spacer (ITS) sequences (fungi) and 23,776,295 16S sequences (bacteria) were analyzed. These sequences were assigned to 1,383 and 15,750 fungal and bacterial amplicon sequence variants (ASVs), respectively (see supporting information at https://doi.org/10.4121/12971528). Three weeks after mixing organic products with the soil, microbial community structure was altered and community diversity was diminished (Shannon and inverse Simpson indices) compared to the nonamended and nitrogen fertilizer-amended controls (control and control+N) (Fig. 3). This was due to a lower number of species (richness) (Fig. 3). Bacterial diversity (Shannon diversity index [H′]) values ranged from around 7.5 in the controls, to 6.5 to 6.8 in keratin-amended soil, to 5.5 to 6.0 in seed-amended soil. Fungal diversity (Shannon diversity index [H′]) values were 4.0 in the controls, 2.5 to 3.0 in keratin-amended soil, and less than 0.5 in manure-chitin-amended soil. This extreme low fungal diversity in soil with manure-chitin is due to dominance of the inoculated Agaricus bisporus, which was added as active ingredient in the product amendments with plant-1, plant-2, and plant-3, and manure product had the least effect on bacterial diversity.

In order to study how the entire microbial community composition varied between soil amendments and years, a multivariate analysis (nonmetric multidimensional scaling [NMDS]) was performed. The soil microbial communities differed among types of soil and experimental years (see Fig. S4 in the supplemental material). Microbial community compositions were different between Lisse and Vredepeel soils for fungal (adonis R² = 0.16, P < 0.001) and bacterial (adonis R² = 0.26, P < 0.001) communities (see Fig. S3 in the supplemental material). Both microbial communities varied between the experiments in 2016 and 2017 as well. The type of soil had stronger effects than the treatments, and the year slightly affected microbial communities as well. However, it was possible to focus solely on the effects of the treatment on the microbial community composition and its correlation with disease suppression in a canonical correspondence analysis (CCA). Both fungal and bacterial microbial community compositions in the soils with control, plant, and manure treatments clustered together (Fig. 4 and 5). Furthermore, the four keratin treatments clustered together and partially overlapped with the soil microbial communities from seed-amended soil. The manure-chitin-amended samples presented the most distinct microbial community composition. Both the bacterial (adonis R² = 0.46, P < 0.001) and fungal (adonis R² = 0.51, P < 0.001) community compositions in the four keratin treatments and the manure-chitin treatment all together positively correlated with higher R. solani suppression (Fig. 4 and 5).

From the CCA, it was possible to select the fungal and bacterial taxa (ASVs) that were associated together with the most suppressive treatments (i.e., keratin and manure-chitin amendments) (see supporting information at https://doi.org/10.4121/12971528). The bacterial families Oxalobacteraceae (Massilia spp. and Cupriavidus spp.), Flavobacteraceae (Flavobacterium spp.), and Paenibacillaceae are the most represented taxa among the ASVs correlating strongest with R. solani disease suppression in comparison to the whole microbial community, and all three families were most abundant in chitin- and keratin-rich treatments (Fig. 4; Fig. S4). The manure-chitin amendment, which is composed of very different components compared to the other organic amendments, showed a unique bacterial community composition. Many of the ASVs that strongly associated with disease suppression (see the rank list of species in Data File 12 in the supporting information at https://doi.org/10.4121/12971528) were the most represented in manure-chitin-treated soil samples (i.e., Massilia [Oxalobacteraceae], Microbacterium [Microbacteriaceae], Rhizobium and Devosia [Rhizobiaceae]) (Fig. 4). On the other hand, Flavobacterium spp. (Flavobacteraceae) were more abundant or uniquely represented in keratin-amended soil samples (Fig. 4).

Among all selected fungal ASVs that correlated with R. solani disease suppression, Mortierellaceae was by far the most represented group, and it occurred almost exclusively in keratin-amended soils (Fig. 5; Fig. S4). A phylogenetic cluster of Cutaneotrichosporon spp. (Trichosporonaceae) and Filobasidium spp. (Filobasidiaceae) were only detected in the manure-chitin samples (Fig. 5). Furthermore, one ASV belonging to Agaricus spp. (Agaricaceae) was more abundant in the manure-chitin treatment (Fig. 5c), which is due to the inoculation of the product containing Agaricus bisporus.

To study what types of relationships were established between fungal and bacterial populations in R. solani disease-suppressive treatments, a network analysis was performed, where only the most abundant ASVs were included. Results showed three main fungal network hubs (i.e., Panaeolus spp., Mortierella spp., and an undescribed fungus from the order Chaetothyriales), which are connected to several bacterial ASVs belonging to different taxa (i.e., Variovorax [Comamonadaceae] and Oxalobacteraceae) (see Fig. S5 in the supplemental material). It is remarkable that the relationships are mainly negative, meaning that when the relative abundances of the fungi were reduced, the bacteria were relatively more abundant or vice versa.

Associations between the response of the soil microbial community and specific soil chemical properties to the organic treatments were detected. In manure-chitin samples, the bacterial community was associated with higher levels of copper (Cu.a) and selenium (Se.a) and marginally with available cobalt (Co.a) (Fig. 6; Table S1). On the other hand, in the keratin-rich-amended soils, fungal and bacterial microbial communities were linked to available zinc (Zn.a) and copper (Cu.a). Furthermore, for both treatments, microbial communities were associated with lower phosphate content (P.a) (Fig. 6; Table S1). Moreover, the microbial communities in treatments that did not reduce R. solani disease, such as manure and plant-3 amendments, were associated with higher carbon-to-nitrogen (C/N) ratios and available phosphorus (P.a).

We showed that soil amendments rich in keratin and chitin are able to reduce R. solani disease in sugar beet. Furthermore, the microbial community is distinct in soils treated with different types of amendments (plant, seed, manure, keratin, and manure-chitin) and nonamended soils. Available forms of zinc and copper were related to soil microbial communities and disease suppression. We demonstrated that soil microorganisms likely play an important role in disease suppression in soils. In the case of arable soils, it is possible to steer the soil microbial community by applying various organic amendments that can boost specific biological and chemical soil properties that increase plant and soil resilience and/or disease suppression.

Ten different organic amendments were selected as treatments. They were plant- and animal-based and presented a broad range of C/N ratios (from 3.2 to 25.0) and respiration rates (from 3.7 to 38.7 mmol O₂/kg organic matter [OM]/h) (Table 1). Physicochemical characteristics were measured by Eurofins Agro (Wageningen, The Netherlands) and can be found in supporting information at https://doi.org/10.4121/12971528.

To test the effect of these organic amendments on disease suppression, as well as on several (biological) soil properties, these products were added to two sandy soils. Large batches of bulk soil were sampled from the upper 20 cm of the field, transported to the greenhouse location, stored at room temperature until use. An alluvial sandy soil with a low organic matter content (1.5%) and neutral pH (7.2) was obtained from the experimental fields at Lisse (coordinates: N 52.2552, E 4.5477). The other sandy soil, obtained from the experimental fields at Vredepeel (coordinates: N 51.5417, E 5.8730), had a higher organic matter content (4.0 to 4.3%) and a slightly acidic pH (5.5). More soil characteristics are included in the supporting information at https://doi.org/10.4121/12971528.

The organic amendments were mixed with the soils at 0.2 g N/kg soil to eliminate a possible cofounding effect of nitrogen since N is the main limiting nutrient for (short-term) plant growth. Consequently, the types of organic matter added to the soils differed in their input volumes (Table 1). Granulated products (e.g., manure and seeds) were pulverized, and product with fibers (e.g., plant-3) were chopped into 4-cm fractions prior to application to create a better distribution in the soil. To control for the effect of additional nitrogen on soil disease suppression, two control treatments were included: one with the same amount of nitrogen as the organic amendments was added as calcium nitrate [Ca(NO₃)₂] fertilizer (control+N) and another without any additional nitrogen (control). All treatments were prepared in four replicates, and the entire experiment was performed twice (in 2016 and 2017). Soils with the amendments were mixed, put into open plastic bags, and incubated at room temperature (∼20°C) for 3 weeks. This mixed soil was then used for plant disease bioassays and soil biological and physicochemical parameter measurements. A subsample of each amended soil was stored at −80°C for DNA analysis, and the remaining soil was stored in 10% KOH dissolved in methanol at −20°C for ergosterol quantification.

Disease suppressiveness of the amended soils was tested in a bioassay in which the pathogen was artificially inoculated into the amended soils. The fungal pathogen Rhizoctonia solani (AG 2-2IIIB isolate 12-194a; IRS, Bergen op Zoom, The Netherlands) was inoculated into the treated soils, and after disease development was evaluated in susceptible sugar beet seedlings under growth chamber conditions, as described by Postma and Schilder (21). Tanks with an internal size of 4 by 25 by 30 cm were used with prerinsed florist’s foam blocks (water holding capacity ≈ 55%; Van Dillewijn Verpakkingen BV, Aalsmeer, The Netherlands) of 4 by 25 by 17 cm placed at the bottom of each tank. In each tank, 12 cm of treated soil was packed on top of the water-saturated foam blocks. Sugar beet seeds (F763156 with standard seed treatment hymexazol, thiram, and the insecticide Gaucho1; SESVanderHave, Rilland, The Netherlands) were planted in two rows 2 cm deep with a 2-cm distance between seeds at a density of 22 seeds per tank. The soil surface in each tank was watered and covered with plastic foil for 1 week. Then the soil in each tank was inoculated with five oat kernels colonized with R. solani. The inoculation point was set in one of the sides of the tank, 2 cm in front of the first seedling, just below the soil surface. The disease spread, defined as centimeters from the inoculation point to the last seedling in the row with symptoms, was measured twice a week by counting the centimeter to the last seedling displaying damping off or brown-gray lesions on the stem at the soil level. Disease spread data from 21 days after inoculation with the pathogen were used for our study, since a maximum disease spread of 22 cm was reached in the susceptible treatments at this time point. The experiments were carried out with four replicates in a randomized block design, with each replicate soil amendment treatment present in each block. The bioassay was performed in a growth chamber under the following conditions: 23/18°C (day/night), 60% humidity, and a day/night regimen of 8 h of dark and 16 h of light (230 μmol m⁻² s⁻¹ photoactive light; TL280HF) and the soil water potential was automatically regulated at −5 × 10³ Pa (soil-water metric potential [pF], 1.7).

Bioassay results from the Vredepeel 2016 soil were excluded in the follow-up data analysis because reduced sugar beet growth was observed in all treatments—possibly due to the application of the herbicide MCPA (2-methyl-4-chlorophenoxyacetic acid) in the field 1 month before the start of the bioassays.

Bacterial biomass was determined by confocal laser scanning microscopy (56) and fungal biomass by microscopic counting after staining with fluorescent dyes (57). The concentration of ergosterol, an important sterol in the membrane of most fungi, was measured in the soil samples as an indicator of fungal biomass (58). Hot water extractable carbon (HWC) was measured as the organic carbon present in solution after a 16-h extraction of 4 g of soil in 30 ml water at 80°C. HWC is a labile fraction of organic matter, which originates mainly from microbial exudates and has been proposed as an easily applicable indicator of soil quality (56). Potentially mineralizable nitrogen (PMN)—as an indicator of labile nitrogen that is potentially available for plants—was measured as the increase in ammonium (NH₄⁺) during 1 week of anaerobic (waterlogged) incubation of 16 g of soil in 40 ml water at 40°C (59, 60). Other physicochemical properties of the amended soil samples were analyzed by the commercial laboratory of Eurofins Agro (Wageningen, The Netherlands). More than 20 nutrient levels in the soils were determined: 13 nutrients (P, K, Mg, Na, B, Cu, Mn, Co, Zn, Se, Si, Mo, and Fe) were measured via the CaCl₂ extraction method. Total N, S, C, and P were extracted by ammonium lactate, and K, Ca, Mg and Na were extracted via the Cohex method (61).

DNA was extracted using the MoBio PowerMag soil DNA isolation KF kit (MoBio Laboratories, Inc., Carlsbad, CA, USA). The manufacturer’s protocol was adapted to fit a 4-fold input weight of 1 g soil for each sample. All steps up to the processing of the samples with the King Fisher platform were performed in 5-ml tubes, and volumes of all buffers were proportionally adapted. Lysis was performed in 5-ml MoBio PowerWater DNA bead tubes supplemented with 1 g of 0.1-mm glass beads (MoBio Laboratories, Inc., Carlsbad, CA, USA). In the setup for the King Fisher DNA processing, two technical replicates per sample were used in a 96-well format, and a double binding step was included in the protocol to accommodate all available lysate per sample. The resulting DNA eluates were pooled per sample and stored at −20°C. The DNA concentration was measured with a Pico Green assay (Quant-IT Pico Green dsDNA assay kit; Invitrogen) on a Tecan Infinite M200Pro (Tecan Group, Ltd.). Based on the concentrations, DNA was diluted to 4 ng μl⁻¹ with the elution buffer of the DNA extraction kit used as the template for the PCRs. Universal primers E341F (5′-CCTACGGGNGGCWGCAG-3′) and 805R (5′-GACTACHVGGGTATCTAATCC-3′) were used to target the bacterial V3-V4 16S rRNA gene region (62, 63), and primers ITS86F (5′-GTGAATCATCGAATCTTTGAA-3′) and ITS4 (5′-TCCTCCGCTTATTGATATGC-3′) were used for the fungal ITS2 region (64–66). All primers were synthesized (Integrated DNA Technologies, BVBA, Leuven, Belgium) with the universal Illumina MiSeq adapters. Each PCR mixture contained 1× Q5 reaction buffer, 200 μM deoxynucleoside triphosphates (dNTPs), 0.5 μM 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 the target regions 16S V3-V4 and ITS2 were 98°C for 30 s, followed by 15 or 25 cycles of 98°C for 10 s, 55 or 51°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). Four replicate PCRs were performed in separate runs for each sample for both primer sets, and 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 for 16S V3-V4 libraries and 2 × 300 cycles for ITS2 libraries on an Illumina MiSeq platform, and all reads were demultiplexed (Next Generation Sequencing Facilities, Wageningen University & Research, Wageningen, The Netherlands).

Preprocessing of the demultiplexed reads was performed with QIIME2 (67) v.2018.4. For the ITS2 reads, cutadapt (68) was used as a plugin in QIIME2, to remove primers and possible read-through in the reads. All raw sequence data were denoised with DADA2 q2 plugin (69). Run-specific quality control and filtering, primer removal for 16S data, merging of paired-end reads, and chimera filtering, resulting in tables with the abundances of exact amplicon sequence variants (ASVs) (70), were performed on each sample and set of unique representative sequences. The resulting DADA2 output was merged for further analysis into a single abundance table. Replicates of sample reads were summed. The ASVs with low abundance (frequency of <0.0005%) were removed. One sample from the 16S data set and two from the ITS data set were removed due to their extremely low library size (see supporting information at https://doi.org/10.4121/12971528). Taxonomy was assigned to the ASVs with the Naive Bayes classifiers plugin in QIIME2 (71, 72). The 16S classifier was pretrained on the extracted 16S V3-V4 region of the Silva 16S/18S database release 132 (73), and the ITS2 classifier was pretrained on the extracted ITS2 region of the Unite database (QIIME release for all eukaryotes v.8.0) with dynamic use of clustering thresholds (74). For the fungal ITS2 database, only ASVs with assignments to the kingdom Fungi were kept. For the bacterial 16S database, ASVs that remained unassigned at the kingdom level or were assigned as mitochondria or chloroplasts were removed.

In order to test the differences in R. solani disease scores (rs.dis) between the different treatments, a two-way analysis of variance (ANOVA) was performed with the Anova function from the CAR R package (75). To indicate the significance between treatments and the reference treatment, a t test comparison with Holm correction was applied. The treatment control+N was used as reference in all the comparisons tested. Furthermore, a Spearman correlation analysis was carried out with the biological parameters (PMN, HWC, fungal and bacterial biomass, and ergosterol) and other soil physicochemical data, together with the disease scores, using the ggscatter function and Pearson correlation method from the GGPUBR R package (76). Permutation-based P values were calculated using the permcor function from the RFAST package (77). For the C/N ratio and respiration rate of the products, the average values per treatment were used, as the C/N ratio and respiration rate were only measured once per treatment.

For α diversity, the rarefied abundance data were used. Rarefaction was set at 8,117 sequences for fungi and 6,236 sequences for bacteria. Then richness (observed number of species), the Shannon index, and the inverse of the Simpson index (which takes into account the number of species and how evenly distributed they are) were calculated with the diversity function from the VEGAN package (78) and represented with their means in box plots. All data are included, except for Vredepeel 2016, because these data were influenced by herbicide artifact effects (described above).

Normalized data with the decostand function (Hellinger normalization) were used for the follow-up analysis, including all data sets, except Vredepeel 2016. In order to see how microbial communities differed between soil type and years, a nonmetric multidimensional scaling (NMDS) analysis with the metaMDS function, and a permutational multivariate ANOVA (PERMANOVA) analysis with the adonis function, both from the VEGAN package, were performed with a Bray-Curtis dissimilarity index.

To test how microbial communities in the treated soils were related to disease, a canonical correspondence analysis (CCA) using the cca function was performed using square root normalized data. The disease spread was used as a constrained variable, and a condition to correct by the factors “year” and “soil” was applied. The top 100 ASVs less associated with disease and more represented in keratin- and chitin-rich treatments were selected from the cca object with the function scores from the VEGAN package and included in the phylogenetic tree. Loading the abundance data together with sample taxonomy information and metadata, a complex data set object was created with the function phyloseq from the PHYLOSEQ R package (79). Subsequently, a phylogenetic tree was created with the function ggtree from the GGTREE R package. To study how soil microbial communities are correlated with other soil factors, another CCA was performed. Soil parameters were defined as constrained variables, and they were selected based on their correlation with disease from both the linear model (lm) and correlation (corr) analyses. The factors “soil” and “year” were included in the model as conditions. An ANOVA of the CCA object per treatment type (keratin and chitin rich) was performed to test for significant relationships with the different soil parameters. From the CCA object, the top 260 ASVs were selected from the scores function output to estimate the number of ASVs in each family category for the suppressive treatments and presented in a bar plot.

To study how the fungal and bacterial abundances related to each other in the disease-suppressive amended soils, a Spearman correlation analysis between bacteria and fungi for proportional abundance was performed. Only the keratin- and chitin-rich amendments were included in the analysis. Both bacteria and fungi were filtered, and taxa without at least 50 occurrences were removed. A Spearman rank correlation with an associated P value was calculated between each combination of bacteria and fungi. The P values were adjusted for multiple comparison with a false-discovery rate (FDR), which was calculated with the Benjamini-Hochberg adjustment. Edges with an FDR of <0.05 and their associated nodes were included in the network diagram.

All sequences have been submitted to the NCBI database under BioProject no. PRJNA657959.