Rhizoctonia solani disease suppression: addition of keratin-rich soil amendment leads to functional shifts in soil microbial communities

Abstract Promoting soil suppressiveness against soil borne pathogens could be a promising strategy to manage crop diseases. One way to increase the suppression potential in agricultural soils is via the addition of organic amendments. This microbe-mediated phenomenon, although not fully understood, prompted our study to explore the microbial taxa and functional properties associated with Rhizoctonia solani disease suppression in sugar beet seedlings after amending soil with a keratin-rich waste stream. Soil samples were analyzed using shotgun metagenomics sequencing. Results showed that both amended soils were enriched in bacterial families found in disease suppressive soils before, indicating that the amendment of keratin-rich material can support the transformation into a suppressive soil. On a functional level, genes encoding keratinolytic enzymes were found to be abundant in the keratin-amended samples. Proteins enriched in amended soils were those potentially involved in the production of secondary metabolites/antibiotics, motility, keratin-degradation, and contractile secretion system proteins. We hypothesize these taxa contribute to the amendment-induced suppression effect due to their genomic potential to produce antibiotics, secrete effectors via the contractile secretion system, and degrade oxalate—a potential virulence factor of R. solani—while simultaneously possessing the ability to metabolize keratin.


Introduction
Food security is threatened by a combination of increasing disease incidence in crops and the target to reduce the application of pesticides in 2030 by 50% following the new European Farm to Fork and Biodiv ersity str ategies (Pr oposal on the sustainable use of plant protection products and amending Regulation (EU) 2021/2115).Enhancing soil suppr essiv eness a gainst soil borne pathogens is a promising strategy to control diseases and crop losses caused by those pathogens.Rhizoctonia solani is the main causal agent of black scurf disease in potato and cr own r ot in sugar beet, among other arable crops.
Disease suppression of Rhizoctonia solani in soils can be induced b y tw o entir el y differ ent str ategies .T he first a ppr oac h r equir es the successive planting of host crops in the presence of their pathogen resulting in reduced disease incidence over time .T his phenomenon of disease decline by monocropping has first been described for take-all ( Gaeumannomyces graminis var.tritici) in barley and wheat (Gerlagh 1968, Sarniguet et al. 1992, Raaijmakers and Weller 1998 ).Take-all decline is defined as the spontaneous reduction in the incidence and severity of the disease and increase in yield occurring with continuous monoculture of the host crop following a se v er e attac k of the disease (Schlatter et al. 2017 ).Also, for R. solani -induced diseases a decline has been reported in both field and pot experiments for se v er al cr ops , i.e .wheat (Lucas et al. 1993, Roget 1995, Wiseman et al. 1996, Mazzola and Gu 2002 ), sugar beet (Hyakumachi et al. 1990, Sayama et al. 2001, Gómez Ex-pósito 2017 ), radish (Henis et al. 1978, Chet and Baker 1980, Chern and Ko 1989 ), potato (Velvis et al. 1989 , Jager andVelvis 1995 ) and cauliflo w er (Davik andSundheim 1984 , Postma et al. 2010 ).
The second a ppr oac h to stim ulate disease suppr ession r equir es the addition of organic amendments into soils .T hese amendments can be side streams from industrial food processing, farming, and a gricultur al activities .T he r e-use of suc h materials in a gricultur e aligns with the incr easing inter est in circularity, reducing the environmental impact and promoting valorization of waste pr oducts (Alv ar enga et al. 2017, Abbott et al. 2018, De Corato 2020 ).A lar ge v ariety of or ganic pr oducts, r esidual str eams fr om plant and animal production, food industry or society have been tested for their potential to suppress soilborne diseases .T he effects are product-and disease dependent and ther efor e difficult to tr anslate into pr actical a pplications so far (Bonanomi et al. 2010 ).The addition of compost (Tuitert et al. 1998, Pér ez-Piquer es et al. 2006, Termorshuizen et al. 2006 ) and cellulose-containing products (Kundu andNandi 1985 , Clocchiatti et al. 2021 ) have been studied extensiv el y, but effects wer e often unpr edictable.
Against R. solani particularly, the addition of chitin-and keratinric h pr oducts hav e been shown to decr ease disease (Postma andSc hilder 2015 , Andr eo-Jimenez et al. 2021 ).Disr egarding the a ppr oac h to induce Rhizoctonia suppr essiv eness in soils, se v er al organisms or combinations thereof were commonly found to correlate with lo w er disease incidence in differ ent cr ops and soils.These include members of the bacterial families Oxalobacteriaceae , Comamonadaceae and Burkholderiaceae , Pseudomonadaceae as well as members of the orders Hyphomicrobiales and Sphingobacteriales (notably Flavobacterium , Chryseobacteria and Chitinophaga ) and the fungal family Mortierellaceae (Bonanomi et al. 2010, Chapelle et al. 2016, Gómez Expósito 2017, Gómez Expósito et al. 2017, Carrión et al. 2018, Carrión et al. 2019, Andreo-Jimenez et al. 2021, Yin et al. 2021 ).
A universal pattern of the responsible microorganisms and especially the underlying mechanisms for Rhizoctonia suppressive soils is still lacking, although several potential mechanisms of disease suppression have been proposed.One model attributes disease suppression to the expression of a non-ribosomal peptide synthetase of Pseudomonadaceae family members (Mendes et al. 2011 ).Another mechanism points to the role of oxalic and phenylacetic acid as the main driver of suppression.R. solani , extending hyphae to w ar ds the plant r oot, r eleases oxalic and phenylacetic acid, activating certain rhizobacterial families .T his induces o xidati v e str ess, triggering surviv al r esponses via the ppGpp pathway, including enhanced motility, biofilm formation, and secondary metabolite production (Chapelle et al. 2016 ).These hypotheses have been proposed based on experimental set up in natur al or a gricultur al soils without amendments .T he disease suppr ession mec hanisms via or ganic amendments r emain unexplored.
The objective of the present work is to gain a deeper insight on how micr obial comm unities and their molecular functions can be linked to Rhizoctonia disease suppression after the amendment with a ker atin-ric h side str eam fr om the farming industry.We propose that the addition of keratinaceous compounds leads to an enrichment of specific genes in the microbial community that eac h by themselv es hav e been shown to play a r ole in disease suppression.

Selected samples and sequencing
Soil samples were collected as large batches from the top 20 cm of two different experimental fields in the Netherlands in 2017: The alluvial sandy soil with a low organic matter content (1.5%) and neutral pH (7.2) was collected from Lisse (N 52.2552, E 4.5477) and the sandy soil with a higher organic matter content (4.0 to 4.3%) and a slightly acidic pH (5.5) was obtained from Vredepeel (N 51.5417, E 5.8730).Additional par ameters measur ed can be found in the publication of Andreo-Jimenez et al. 2021 .A volume of 1.3l soil per replicate and soil was used.Half of them was amended with a pig hair meal product (1.4 g/kg soil) (Darling Ingr edients), her eafter r eferr ed to as ker atin-ric h amendment.As contr ol tr eatment the same amount of soil was amended with 1.2 g calcium nitrate (Ca(NO 3 ) 2 /kg soil to ensure the same nitrogen equivalents added as in the ker atin-ric h tr eatment (i.e.0.2 g N/kg soil).All treatments were performed with 4 replicates and incubated in open plastic bags for three weeks at room temperature ( ∼20ºC).Soils in combination with the tr eatment wer e then tested for their disease suppression potential in bioassays with sugar beet seedlings by e v aluating the disease spread (plants with damping off or br own-gr ey lesions on the stem) pr e viousl y.The ker atin-ric h amendment with pig hair (r eferr ed to as K er atin-3) sho w ed a significant ( P < 0.01) decrease in disease spread compared to the control (Andreo-Jimenez et al. 2021 ).
The total of 16 samples was then sampled for DNA extraction to perform shotgun metagenomics sequencing and stored at −20ºC until use.DN A w as extracted using the MoBio Po w erMag soil DN A isolation KF kit (MoBio Laboratories, Inc., Carlsbad, CA, USA), with the manufactur er's pr otocol adjusted for a 4-fold input weight of 1 g soil per sample.Lysis occurred in 5-ml MoBio Po w erWater DN A bead tubes supplemented with 1 g of 0.1-mm glass beads.For King Fisher DNA processing, a 96-well format and two technical replicates per sample w ere emplo y ed, incorporating a double binding step in the protocol to utilize all available lysate per sample .T he resulting DNA eluates were combined per sample and stored at −20 • C. DN A concentration w as determined using a Pico Green assay on a Tecan Infinite M200Pro.
Shear ed DNA extr acts wer e used for libr ary pr epar ation (Next Generation Sequencing Facilities, Wageningen University & Researc h, Wa geningen, The Netherlands) and those were paired-end sequenced (2 ×150 nt) on a Illumina NovaSeq 6000 platform (BaseClear B.V., Leiden, the Netherlands).The raw sequencing data used for this study are available on the NCBI sequence r ead arc hiv e (SRA) under BioPr oject number PRJNA966095 (r e vie wer link: https://datavie w.ncbi.nlm.nih.gov/object/PRJNA966095?r e vie wer=seaasdqnmti48epii65f2qfrm4 .) Both soils were tested for damping off disease suppression caused by Rhizoctonia solani AG2-2IIIB in a sugar beet assay and were found to lead to a significant decrease in disease incidence (Andreo-Jimenez et al. 2021 ).

Bioinforma tics w orkflo w
Raw reads were subjected to an all-in-one preprocessing using FASTP (Chen et al. 2018 ) with default settings for pairedend data.The remaining reads were passed to Kraken2 (Wood et al. 2019, Lu et al. 2022 ) for detection of taxa using a custom database using the following kraken standard databases as basis: complete bacteria (70813 genomes), complete archaea (739 genomes), complete fungi (1678 genomes), complete viral (14744 genomes) and complete protozoan (11151 genomes) databases ( https:// benlangmead.github.io/aws-indexes/ k2 ).After building the database, a collection of soil microbes that have been identified as important players in this ecosystem pr e viousl y (Andr eo-Jimenez et al. 2021 ) and were either absent or underr epr esented in the built database, were added to the library ( Supplementary Table S1 ).Kraken2 was run in -paired mode on individual samples using the kraken2 standard output to be visualized with PAVIAN (Breitwieser and Salzberg 2019 ).The absolute number of hits were extracted per family and genus level entry.These r esults wer e then used to inv estigate differ ential abundance distributions of microbial families and genera per soil with DESeq2 (Love et al. 2014 ) separately, using treatment (keratin vs control) as design input.Results were then visualized using EN-HANCEDVOLCANO (Blighe et al. 2019 ).In addition br ac ken was run for abundance estimation (Lu et al. 2017 ).Using the r elativ e abundances the impact of ker atin-ric h soil amendments on taxonomic composition was assessed using PERMANOVA tests, appl ying Br ay-Curtis dissimilarity on famil y and genus le v el.The Wilcoxon test was applied to identify significant differences in taxon abundance between treatment groups.Multiple testing corr ection (Benjamini-Hoc hber g) was performed, and onl y taxa with adjusted p-values < 0.01 were considered significant.Differences in variability were calculated between keratin-amended and control treatments and the Top 50 taxa on genus and famil y le v el wer e plotted using ggplot2 ( Supplementary Figs S1  and S2 ).

Functional annotation and statistical analysis
Reads were assembled into proteins directly using the proteinle v el assembl y tool PLASS (Steinegger et al. 2019 ) in paired-end mode to assemble QC-passed reads directly into proteins.

Pfam enrichment analysis
For the enrichment analysis of protein domains in the dataset, the most recent Pfam annotations of Pfam-A and Pfam-N (15-11-2021 v ersion) wer e downloaded fr om the EMBL-EBI FTP server ( http:// ftp.ebi.ac.uk/ pub/ databases/ Pfam/ ) and converted into a mmseqs profile database (Steinegger and Söding 2017 ).Using the protein assemblies of each sample separately as query, mmseqs easy-search was used to find Pfam profile hits for each sequence, choosing the best hits greedily ( -greedy-best-hits ).The hits per Pfam domain were counted and the resulting count matrix per sample was analyzed for differential abundance with DeSeq2 using "treatment" and "soil" as parameters in the design after prefiltering out low counts ( < 50).Acquired p-values were corrected for multiple testing using the Benjamini and Hoc hber g method.PCA plots wer e cr eated using log-r atio tr ansformed output fr om the DESeq analysis.Pfam domains with a Log2Fold change of > 1 and P adj < 0.01 were subset ( Supplementary Table S2 ).Differentially abundant domains were plotted using EnhancedVolcano (FCCutoff = 1.0,P-Cutoff = 10e-6).To e v aluate the significance of Pfam composition differences among keratin-rich amendments, we generated a Bray-Curtis dissimilarity matrix.Subsequently, an adonis test was conducted to assess the influence of the keratin amendment on dissimilarity.Follo wing this, Tw o-Way ANOVA w as performed for each numeric response variable in the dataset, extr acting P -v alues.In cases wher e v ariables sho w ed significance in ANOVA ( P < 0.05), post-hoc tests (Tuk e y's HSD) were carried out.T he outcomes , comprising P -values , significant variables , and ANOVA summaries, were consolidated into Tables S3 and S4 .

Detection and classification of keratinases
The Peptidase Full-length Sequences were downloaded from the MEROPS peptidase database version 12.1 (Rawlings et al. 2017 ) and converted into a DIAMOND database (Buchfink et al. 2015, Buchfink et al. 2021 ).Assembled proteins of keratin-amended soils from Lisse and Vredepeel w ere sear ched against the peptidase database with a DIAMOND blastp run with a block size value of 12 and only showing the single best hit and further default settings.Protein sequences of the positives hits were extracted, and taxonomic information was added by running mmseqs taxonomy using the UniRef100 database as a r efer ence with default settings.Information on the MEROPS family, the taxonomic lineage, the soil, and the sum of hits of each family per MEROPS family was combined using custom scripts in python and R and visualized using ggplot2.For the sake of visibility, a subset was taken containing known keratinases (Qiu et al. 2020 ) and microbial families that w ere sho wn to be more abundant in amended soils .T hese data were also visualized using ggplot2.

Untargeted approach using protein clustering
Due to the extr emel y high number of reads that could not be assigned to any taxonomic group, we also performed an untar geted a ppr oac h to be able to classify pr oteins unique to the tr eatment independent fr om homologies with entries in av ailable databases.To this end, protein assemblies were combined into tr eated and contr ol samples per soil.The data load w as do wnsized by clustering proteins and only k ee ping the re presentati ve sequences: We first converted the protein assemblies into mm-seqs databases and used linclust of the mmseqs2 pac ka ge on the pooled assemblies of the control and the treated samples in bidir ectional cov er a ge mode with sequence identity threshold and an alignment cov er a ge of 80% (-min-seq-id 0.8 -cov-mode 0 -c 0.8) to pr eserv e the m ulti domain structur e of pr oteins.Repr esentativ e pr oteins per cluster wer e then extr acted fr om the clustering results ( mmseqs createsubdb inDB_clu inDB inDB_clu_rep) and converted into a fasta flat file ( mmseqs convert2fasta inDB_clu_rep inDB_clu_r ep.fasta).The r epr esentativ e pr oteins of the contr ol per soil wer e conv erted into a DIAMOND database and used as query a gainst the r epr esentativ e pr oteins of the ker atin-tr eated samples in a diamond blastp run with a block size value of 12 and only showing the single best hit and further default settings.Unaligned pr oteins wer e stor ed and used to cr eate an mmseqs2 database to identify taxonomy and functions of proteins present using the mmseqs taxonom y a gainst the Unir ef100 database .T he result of the functional c har acterization of pr oteins was exported into a .tsvfile also containing the taxonomic lineage.Next to that we performed a functional annotation using the standalone version of KofamScan (version 1.3.0)with the standard database (KO profiles (release 24-Apr-2023) KO list (release 26-Apr-2023) (Aramaki et al. 2019 ).The data sets were then merged into a single file per soil containing functions and taxonomy in different columns .T he table was parsed and families with a total of more than 200 protein hits were extracted.From those we used the top 100 most abundant KO identifiers and plotted the data using ggplot2.

Sequencing results, annotation, and community profile
A summary of the samples and sequenced reads per sample is available in Table 1 .Samples from Lisse soil were sequenced much deeper than Vredepeel soil samples .T he amount of reads that could be classified using the customized Kraken database was low and depended on the soil and the treatment it received (Table 1 ).Inter estingl y, the percenta ge of classified r eads in Lisse soils that had been supplemented with the ker atin-ric h amendment incr eased fr om 24.70 ±0.46 to 36.92 ±0.89, wher eas classified r eads in Vredepeel soil dropped after having r eceiv ed ker atin-ric h tr eatment (from 26.98 ±0.40 to 21.01 ±0.35).Most of all classified reads disregarding soil and treatment could be attributed to bacteria (96.81%-98.59%),follo w ed b y fungi (0.16%-0.31%).Viral and protozoan contribution to the metagenome was less than 0.01% of the total classified reads.

Taxonomic shifts after ker a tin-rich amendment in soil
As bacterial and fungal taxa r epr esented the lar gest gr oups in the identifiable fraction, we focused our further analysis on these.The PERMANOVA analyses revealed significant differences in micr obial comm unity dissimilarity based on both the 'treatment' (ker atin-ric h amendment or control) and 'soil' (Vredepeel or Lisse) factors.At the genus le v el, the addition of a ker atin-ric h pr oduct had a substantial influence, explaining 63.21% of the dissimilarity (F = 24.055,P < 0.001), while the soil type had a marginal effect (R 2 = 20.08%,P = 0.067).This pattern persisted at the famil y le v el, with treatment significantly contributing to 67.13% of dissimilarity ( P < 0.001), and soil type showing a less robust impact (R 2 = 18.35%,P = 0.063).
Differ ential abundance anal ysis of classified taxa on famil y le v el sho w ed a similar shift in both soils upon the addition of the ker atin-ric h pr oduct.Both soils incr eased in Flavobacteriaceae and Sphingobacteriaceae (Bacteroidota), Boseaceae , Phyllobacteriaceae and Caulobacteraceae (Alpha pr oteobacteria), Oxalobacteraceae and Comamonadaceae (Beta pr oteobacteria), Rhodanobacteraceae and Steroidobacteraceae (Gamma pr oteobacteria), as well as one (Vredepeel; Bacteriovoraceae ) or even three (Lisse; Bacteriovoraceae, Bdellovibrionceae, Halobacteriovoraceae ) families of the bacterial phylum Bdellovibrionota, containing obligate predatory bacteria (Fig. 1 A and B; Fig. S1 ).The only fungal group that incr eased significantl y was the zygote fungal famil y of Mortierellaceae (Fig. 1 B).Differences between the two soils were represented by Nitrosomanadaceae , Rhizobiaceae and Bdellovibrionaceae which did show a significant increase after keratin-rich amendment in Lisse soil, but not in Vredepeel soil.Furthermore, Micrococcaceae and Caulobacteraceae wer e mor e abundant in Lisse soil and did show a steeper increase after keratin amendment (Fig. 1 A).

Ker a tinol ytic potential of abundant taxa
Amending both soils with a ker atin-ric h compound compar ed to an inorganic source of nitrogen (Ca(NO 3 ) 2 ) led to similar changes in the taxonomic composition of the identifiable microbial fraction after an incubation time of three w eeks.It w as ther efor e expected that the potential to degr ade ker atinaceous compounds could play a role in the enrichment of certain taxa.The taxonomic families shown to be more abundant after the incubation with keratin sho w ed a high occurrence of proteases belonging to MEROPS families of known keratinases (Fig. 2 ).Lisse soil did contain a m uc h higher number of pr otease matc hes.Ho w e v er, dif-ferences between Lisse and Vredepeel soil response are difficult to inter pr et due to the une v en r ead depth of both samples .T he absence of a certain protein family could also be caused by insufficient depth, especially of organisms with a lo w er abundance.In Lisse soil proteases of all known keratinase families could be r ecov er ed for Steroidobacter aceae , Solirubrobacter aceae , Rhodanobacteraceae , Pseudomonadaceae , Oxalobacteriaceae , Nocardioidaceae , Nitrosomonadaceae , Comamonadaceae and Chitinophagaceae .The MEROPS families with the highest number of hits were S09 (24.4% of subset hits), M04 (15.9%),M38 (14.1%), and S01 (10.4%).Much less common in frequency as well as in taxonomic distribution were the metalloprotease families M36 (0.14%) and M32 (0.13%).

Specific protein domains were enriched after ker a tin-rich amendment
Next to the taxonomic classification of reads and k eratinase-lik e protease families, a classification on Pfam domains of assembled proteins was used as a proxy for functional changes in the microbiome upon the soil amendment with a ker atin-ric h pr oduct.The fraction of proteins carrying a Pfam domain found in the database differed between the control and the keratin-amended soils.If soils had r eceiv ed a ker atin-ric h amendment the number of proteins that could be assigned to a Pfam domain dropped in both soils (Table 1 ) (Lisse 39.88 ±5.69% with ker atin-ric h amendment, 48.65 ±10.22% in contr ol; Vr ede peel 24.70 ±8.46% with k er atin-ric h amendment, 38.07 ±0.48% in control) although standard deviation was high.The biological replicates of each treatment category per soil did show similar Pfam patterns and formed distinct clusters in a Principal component analysis (PC A) (Fig. 3 ).T he keratin-rich amendment did have an influence on the protein domain abundances, resulting in a shift along PC1 that was visible in both soils (Fig. 3 ).The analysis indicates that the keratin amendment did not have a significant explanatory effect on the changes observed in the Pfam pattern for either soil.In Vr edepeel soil, the R 2 v alue is 0.1619 with a non-significant P -value of 0.842, while in Lisse soil, the R 2 value is 3.2785 with a P -value of 0.104.The majority of the dissimilarity in both cases is attributed to r esidual v aria- tion within the dataset.Ther efor e, based on this analysis, the impact of the keratin amendment on Pfam pattern changes does not appear to be statistically significant in either soil.
Ho w e v er, when ANOVA is performed for each Pfam domain individually and reveals significant differences between the keratinamended and control groups, it suggests that the presence of keratin amendment has a statistically significant impact on the abundance or distribution of those specific Pfam domains.In Vredepeel soil, 9350 Pfam domains were detected that had a nonzero total read count, 2209 of them having a P -adjusted value of < 0.01.In Lisse soil, 10 133 different protein domains could be detected, of which 4248 had a P -adjusted < 0.01.To identify which domains wer e enric hed if the ker atin-ric h pr oduct was added to the soil, a cutoff of Log 2 FC > 1.0 was chosen in addition to a P -Figure 2. Counts of hits on MEROPS protein families with known ker atinol ytic r epr esentativ es within micr obial families mor e abundant in soils having r eceiv ed a ker atin-ric h amendment.adjusted < 0.01 ( Table S2 ).Results sho w ed that 480 protein domains were enriched in both Lisse and Vredepeel soils after having r eceiv ed an amendment with ker atin.An additional 1227 and 179 Pfam domains were exclusively enriched in Lisse and Vrede peel soil, respecti vely.Of the 480 Pfam domains enriched in both types of soil after ker atin-ric h amendment, 168 were associated with domains of unknown functions (DUFs) and could not be c har acterized further.Potential domains of interest included a domain related to chitinase C (PF06483), several components of the type VI secretion system (T6SS) (PF18443, PF18426, PF13296, PF06744), domains related to the proteins involved in the biogenesis of bacterial cell a ppenda ges (PF00419, PF02753, PF06864, PF09977, PF10671, PF13681, PF15976, PF16823, PF16970) and flagella (PF08345, PF02108, PF03646, PF03961, PF05247, PF05400, PF06490, PF07196, PF07317, PF10768), peptidases with a domain similar to the M9 family (PF01752, PF08453), enhancin-like metallopeptidase domains of family M60 (PF17291, PF13402), microbial collagenases (PF01752, PF08453) and a number of phage(-like) proteins (PF00959, PF03374, PF04233, PF04466, PF05065, PF05133, PF05367, PF05954, PF06761, PF06892, PF09306, PF09669, PF13876, PF14395, PF16083, PF16510, PF18013, PF18352).

Functional shifts in disease suppressi v e soils
To identify proteins unique to keratin-enriched amendments irr espectiv e of the soil, we conducted protein clustering for each soil and treatment.This resulted in 64109105 and 78028651 clusters for Lisse and Vredepeel soils under the control treatment, and 112892232 and 72610580 clusters for soils amended with a ker atin-ric h substr ate, r espectiv el y (Table 1 ).Subsequently, aligning the clustered protein re presentati ves from the control to the ker atin-ric h tr eatment using DIAMOND and r etaining onl y the single best hit, 554610 pr oteins r emained unaligned.These wer e consider ed unique to ker atin-ric h amendment.As downstr eam pr ocessing r equir ed the functional annotation of those proteins , K ofamScan was used to assign KEGG orthologies.Of all unique proteins only 53435 (3.44%) could be assigned.Overall, there was a strong overlap in functional groups with Pfam enric hment anal ysis despite the differ ent a ppr oac h (Fig. 4 ).This included proteins potentially involved in the production of secondary metabolites/antibiotics such as 4 -demethylrebeccamycin synthase (K19888) found in r epr esentativ es of the Burkholderiales , Sphingomonadaceae and se v er al Actinom ycetes , as well as gramicidin S synthase 2 (K16097) predominantly found in Bacteroidota .Proteins that could play a role in keratin degradation, such as serine protease (K14645) present in many organisms and the collagenase kumamolisin (K08677) found in Oxalobacter aceae , Rhodanobacter aceae , Bacillaceae , Micrococcaceae and Intersporangiaceae as well as se v er al tr ansporters potentiall y involved in transport of k eratin deri vati ves such as a basic amino acid/pol yamine antiporter (K03294).Inter esting wer e also se v er al enzyme families that degrade more complex substrates such as hyalur onate l yase (K011727), alpha-L-rhamnosidase (K05989) and ar abidan endo-1,5-alpha-L-ar abinosidase (K06113), 2,6-dioxo-6phen ylhexa-3-enoate/TCOA hydr olase (K22677) or exo-acting protein-alpha-N-acetylgalactosaminidase (K25767) and corresponding transporters like cholesterol transport system auxiliary component (K18480).In addition, se v er al major facilitator superfamil y tr ansporters and m ultidrug r esistance pr oteins wer e r ecov er ed fr om soils amended with ker atin-ric h amendments (K03762, K05519, K08166, K18353, K18555, K18904, K18926), a trait shared among a wider range of micr oor ganisms .T he presence of a hemoglobin/tr ansferrin/lactoferrin r eceptor pr otein (K16087) seems to be a trait of Pseudomonadota .Several genes involved in the acquisition of iron could be identified such as a ferric enterobactin receptor (K19611) with the highest number of hits found in Bacteriodota , Caulobacteraceae , Sphingomondaceae and Oxalobacteraceae and an ir on −sider ophor e tr ansport system substr ate −binding pr otein (K25109), which seems to be more widely distributed.An interesting candidate protein is an unc har acterized pr otein (K07126) especiall y enric hed among Oxalobacteraceae , Phyllobacteriaceae , Comanomadaceae , Pseudomonadaceae and the onl y pr otein found in the fungal family of Mortierellaceae at the set cut-off.Other interesting features include proteins involved in motility (K02416, K02397, K10564) and the type VI secretion system secreted protein VgrG (K11904) and type VI secretion system protein ImpH (K11895).

Discussion
It has been shown pr e viousl y that the ker atin-ric h amendment in the form of pig hair to both Vredepeel and Lisse soils studied here was able to induce disease suppression against R. solani and affected the abundance of specific bacterial and fungal families as was shown by metataxonomic profiling (Andreo-Jimenez et al. 2021 ).In the presented study, we aimed to identify functional traits of the microbial community in disease-suppressive soils that could potentially contribute to this mechanism.We hypothesized that amendment-induced disease suppression should lead to similar functional changes in different soil types.
To explore this hypothesis, we employed shotgun metagenomics, comparing taxonomic and functional profiles between soil types through comprehensive data mining.Our findings underscore the substantial role of the ker atin-ric h amendment in sha ping micr obial comm unity composition, particularly at the family taxonomic level, while acknowledging a discernible influence of soil type.Microbial families that changed most significantl y thr ee weeks after the amendment of a ker atin-ric h substr ate wer e Flavobacteriaceae and Sphingobacteriaceae (Bacteriodetes), Boseaceae , Phyllobacteriaceae and Caulobacteraceae (Alpha pr oteobacteria), Oxalobacteraceae and Comamonadaceae (Beta pr oteobacteria), Rhodanobacteraceae and Steroidobacteraceae (Gamma pr oteobacteria), as well the famil y of obligate bacterial predators Bacteriovoraceae .Within the fungal kingdom only the Mortierellaceae increased upon the addition of ker atin.Pr evious work using metabarcoding analysis on the same soil samples r e v ealed also Oxalobacter aceae , Flavobacter aceae and Mortierellaceae as the microbial families that were associated with the most pathogen-suppr essiv e tr eatments i.e. ker atin and c hitin-ric h amendments (Andreo-Jimenez et al. 2021 ).
The initial enrichment of taxa and functions is attributed to the addition of the ker atin-ric h amendment, with micr oor ganisms capable of degrading keratin or metabolizing its degradation products thriving.Notably, the increased abundance of potential ker atinol ytic enzymes in taxa r esponding to the amendment supports this hypothesis.Despite the differ ent natur e of ker atin (pol ypeptide) and c hitin (pol ysacc haride) molecules, their amendment to soil seems to result in a comparable shift in the soil microbiome as was shown by Andreo-Jimenez et al. ( 2021 ).Wieczorek et al. ( 2019 ) confirmed that members of the Oxalobacteraceae and se v er al Bacteroidetes families wer e the initial c hitin degr aders in a gricultur al soils using stable isotope pr obing .Inter estingl y, they could also show an increased labelling in bacterial preda-tors, suggesting that micr oor ganisms that degr ade c hitin ar e pr ey of Bacteriovoraceae and Bdellovibrionaceae .The increase of these two families in our experiment could have the same reason.Keratinrich and chitin-rich amendment stimulate similar degraders and ther efor e also similar predators.
As the complete degradation of keratin, similar to chitin and cellulose, r equir es the synergistic action of several enzymes, it is impossible to deduce how the breakdown of keratin is orc hestr ated in the group of microorganisms (Qiu et al. 2020 ).Questions such as the succession pattern or whether some community members can independently execute the complete degr adation pathway r emain unanswer ed with the av ailable data.
It seems that selected organic amendments can stimulate the same groups that are associated with disease suppression without the addition of an amendment.This implies that the ability of a micr oor ganism to degr ade ker atin, c hitin and/or their deriv ates indir ectl y contributes to the transformation of a conducive into a Rhizoctonia-suppr essiv e soil.Knowledge on the functional mechanism behind this phenomenon ho w e v er r emains patc hy, mostl y restricted to certain single taxa.Carrión et al. ( 2018 ) could show that the production of sulfurous volatile compounds with antifungal activity by Burkholderiaceae lead to disease suppression of R. solani in situ.They could also show that members of the Chitinophagaceae and Flavobacteriaceae in the root endosphere were enriched in disease-tolerant plants (Carrión et al. 2019 ).These families encoded enzymes with enhanced enzymatic activities associated with fungal cell-wall degradation and secondary metabolite biosynthesis being a direct or indirect protective trait.
Mendes et al .attributed disease suppression to the expression of a nonribosomal peptide synthetase of Pseudomonadaceae (Mendes et al. 2011 ).Chapelle et al. ( 2016 ) propose a vital role for oxalic and phenylacetic acid which, during hyphal growth of R. solani, induce a stress response of specific rhizobacterial families leading to the onset of survival strategies including motility, biofilm formation and the production of secondary metabolites.
We hypothesize that certain groups of microorganisms that ar e incr easing in abundance after the addition of the ker atin-ric h amendment, due to their ability to metabolize the substrate (as w as sho wn on their r epertoir e of r ele v ant pr oteases), also encode versatile functions that are commonly associated with disease suppression.
Among these are the contractile injection systems, such as the type VI secretion system (T6SS) or the extracellular contractile injection system (eCIS).Both are nanomachines resembling the bacteriopha ge puncturing structur e, to secr ete a variety of effectors that play a significant role in competition.These effectors can be injected into neighboring bacterial or eukaryotic cells or the environment causing arrest of their growth or even cell death, scavenging of nutrients or are being used for cell to cell signaling (Coulthurst 2019, Gallegos-Monterrosa and Coulthurst 2021, Geller et al. 2021 ).Se v er al effectors secr eted via this system have e v en been described to have a direct antifungal effect (Trunk et al. 2018, Trunk et al. 2019 ).It is a common feature in Pseudomonadota, wher e mor e than 25% of sequenced genomes ar e estimated to carry at least one T6SS (Bo y er et al. 2009 ) and it is especially abundant in plant-associated microbes (Bernal et al. 2018 ).Our findings sho w ed that proteins belonging to contractile injection systems ar e enric hed in the meta genome after ker atin-ric h amendment, especially in βand γ -Proteobacteria.
The ability to produce antibiotics or bacteriocins is a feature often associated with disease suppression.Although it was not possible to identify complete biosynthetic gene clusters from the data a vailable , due to the lack of genetic context because of proteinle v el assembl y, we could identify a fe w pr oteins that might play a role in antibiosis, such as 4 -demethylrebeccamycin synthase or gramicidin S synthetase 2. From literature it is known that many r epr esentativ es of enric hed taxa hav e been described for their ability to produce antimicrobial compounds: Several members of the Oxalobacteraceae , among which Massilia , Duganella , Pseudoduganella and Janthinobacterium are able to produce several different antimicrobial compounds .T he best studied being the purple pigment violacein, a bisindole with antimicrobial activity among other biological functions (Choi et al. 2015a,b , Dahal et al. 2021 ).The phylum of Actinomycetota is known as a group harboring a wealth of gene clusters encoding antimicr obial a gents.Most antibiotics in clinical use ar e originall y isolated from these microbes (Genilloud 2017, van der Meij et al. 2017 ).And within the Bacteroidota a r ecentl y published study has identified a high density of biosynthetic gene clusters per genome, with the genus Chitinophaga as the most interesting in terms of abundance and diversity (Brinkmann et al. 2022 ).This aligns with the discovery of Carrion et al. who found a consortium of endophytic Chitinophaga and Flavobacterium to consistently suppress fungal root disease, by the expression of chitinases, nonribosomal peptide synthases and polyketide synthases (Carrión et al. 2019 ).
Oxalotrophy, the ability to use oxalic acid as a carbon source, is a r ar e tr ait in bacteria, r estricted to a fe w r epr esentativ es of the phyla Actinobacteria , Firmicutes and Pseudomonadota .It is ho w e v er often found in micr oor ganisms that are associated with plants (Hervé et al. 2016 ).A study on Burkholderia strains in the rhizosphere of white lupin sho w ed that 98% of the strains were able to grow on plant-secreted oxalate as a carbon source compared to only 2% of other strains isolated from the same environment.Oxalic acid is ther efor e suggested to stimulate the recruitment of plant-beneficial members from the soil microbiome (Kost et al. 2014 ).Rhizoctonia solani , like many other pathogenic fungi, is also able to produce oxalic acid to acidify host tissue and sequester calcium from host cell walls (Yang et al. 1993, Dutton and Evans 1996, Palmieri et al. 2019 ).Increased virulence of R. solani has been shown to coincide with ele v ated le v els of oxalic acid being excr eted (Na gar ajkumar et al. 2005 ).This leads Chapelle et al. ( 2016 ) to propose that the stress response activated under attack by the oxalic and phenylacetic acid produced either by R. solani itself or r eleased fr om plant r oots shifts leads to a Rhizoctonia -suppr essiv e microbiome.Taxa that are also found to be enriched after keratinric h amendment, especiall y members of the Oxalobacteraceae ar e known to metabolize oxalate and could thereby influence the invasion success of R. solani or detoxify oxalate for other community members.
We realize that many questions remain, such as the role of the major fraction of unknown proteins and/or organisms that were increased upon keratin-rich amendment.That it is currently not possible to consider that 'microbial dark matter' is a problem, because it could have significant implications on the mechanism of disease suppression of soils.Also, the pathogen was not quantified in the soil or plant tissue.Pathogen molecular quantification needs to be considered in future experiments where the disease suppr essiv e mec hanism of suc h or ganic materials will be tested.In addition, the role of Mortierella and its relatives remains elusive for now.It has been shown for se v er al fungal species that the ability to degrade keratin often co-occurs with the ability to break down other r ecalcitr ant substr ates suc h as cellulose and c hitin via l ytic pol ysacc haride monoxygenases (Lange et al. 2016 ).That this fungal group is enriched after keratin-rich amendment is due to its ability to degrade complex compounds therefore suggests itself.Inter estingl y, it has been shown r ecentl y that Mortierella species harbor bacterial endophytes of the Burkholderiaceae family (Takashima et al. 2018 ), some of which even protect the fungal host from nematode attack by the production of a biosurfactant (symbiosin) with antibiotic properties (Büttner et al. 2021, Büttner et al. 2023 ).
Taken together we propose that the amendment of a keratinric h side str eam pr oduct fr om the farming industry (but probabl y also c hitin and cellulose) stim ulates the enric hment of taxa that have been associated with Rhizoctonia disease suppression pr e viousl y.Man y of these taxa can metabolize the substrate or its deri vati ves and are well equipped for a life in the rhizosphere that could explain their contribution to disease suppression.This includes , among others , the ability to produce a wide array of secondary metabolites and effectors that can be secreted via contractile injection systems and the ability to use oxalic acid.

Figure 1 .
Figure 1.Differ entiall y abundant micr obial families after ker atin-ric h amendment.Volcano plots r epr esent bacteria and fungi, whic h ar e differ entiall y abundant within ker atin-ric h pr oduct tr eatment compar ed with the calcium nitr ate contr ol, in Lisse (A) and Vr edepeel (B) soils (FDR = 1 ×10 −6 , Log 2 FC = 0.75).Based on normalized reads obtained after the Kraken2 w orkflo w.Bubble size depicts the abundance of the family as the total number of reads.

Figure 3 .
Figure 3. Principal component analysis (PCA) of Pfam domain composition of ker atin-ric h and control treatment in Lisse and Vredepeel soil.

Figure 4 .
Figure 4. Abundance of the top 100 most abundant KEGG orthologies in microbial families with a total of ≥200 protein hits .T he number of protein hits was Log-transformed Log 10 (n + 1) and families grouped by a higher taxonomic order (Phylum; Pseudomonadota split on the le v el of class).

Table 1 .
Sample description of Vredepeel and Lisse soils as well as sequencing, assembly, classification, and clustering statistics per treatment and soil.