Soil microbial communities are sensitive to differences in fertilization intensity in organic and conventional farming systems

Abstract Intensive agriculture has increased global food production, but also impaired ecosystem services and soil biodiversity. Organic fertilization, essential to organic and integrated farming, can provide numerous benefits for soil quality but also compromise the environment by polluting soils and producing greenhouse gases through animal husbandry. The need for reduced stocking density is inevitably accompanied by lower FYM inputs, but little research is available on the impact of these effects on the soil microbiome. We collected soil samples from winter wheat plots of a 42-year-old long-term trial comparing different farming systems receiving farmyard manure at two intensities and measured soil quality parameters and microbial community diversity through DNA metabarcoding. High-input fertilization, corresponding to 1.4 livestock units (LU) improved the soil’s nutritional status and increased soil microbial biomass and respiration when compared to low-input at 0.7 LU. Bacterial and fungal α-diversity was largely unaffected by fertilization intensity, whereas their community structure changed consistently, accompanied by an increase in the bacterial copiotroph-to-oligotroph ratio in high-input systems and by more copiotrophic indicator OTUs associated with high than low-input. This study shows that reduced nutrient availability under low-input selects oligotrophic microbes efficiently obtaining nutrients from various carbon sources; a potentially beneficial trait considering future agroecosystems.


Introduction
The gr een r e volution substantiall y incr eased global food pr oduction, but also led to lar ge-scale degr adation of soil with se v er e consequences for human health and the environment (Gomiero et al. 2011, Fowler et al. 2013, Yang et al. 2020. There is an urgent need for more sustainable agroecosystems that do not compromise ecosystem health, but rather restore already degraded soils while providing sufficient yields to meet the growing global food demand. Microbes play a pivotal role in the functioning of a gr oecosystems (Wagg et al. 2014, Bender et al. 2016, Díaz-Rodríguez et al. 2021, Suman et al. 2022 ) and understanding the factors shaping soil microbial diversity is of prime interest to counteract soil degr adation (Cavicc hioli et al. 2019, Zhou et al. 2020.
Agricultur al mana gement factors suc h as tilla ge (Kr aut-Cohen et al. 2020 , Fang et al. 2022 ), fertilization (Wang et al. 2017, Bei et al. 2018, cr op div ersification (Stefan et al. 2021 ), and cov er cr opping (Kim et al. 2021 ) have shown to affect soil microbial communities. Farming systems (FS) implementing these practices into different mana gement str ategies hav e also been shown to pr omote distinct soil micr obial comm unities (Hartmann et al. 2015, Bonanomi et al. 2016, Lupatini et al. 2017, Durrer et al. 2021. Organic fertilization, a k e y element in organic farming, but also incor por ated in con ventional systems , is known to enhance soil quality by increasing soil organic carbon (SOC) content and im-proving soil structure compared to synthetic mineral nitrogen (N) fertilizer (Mäder et al. 2002, Liang et al. 2012, Francioli et al. 2016. Since most soil microbes obtain energy and nutrients for metabolic r eactions fr om or ganic compounds, an incr ease in SOC is connected with stimulating effects on soil microbial biomass and activity (Ren et al. 2019, Ma et al. 2020a. A recent metaanalysis sho w ed that re placing synthetic N with man ure-based fertilizer at equivalent N rates improved crop productivity, reduced r eactiv e N pollution, and incr eased SOC stor a ge (Xia et al. 2017 ). Ren et al. ( 2019 ) identified in their meta-analysis that manur e a pplication incr eased soil micr obial biomass carbon and nitrogen by around 40% and 55%, respectively, compared to mineral fertilized soils . Furthermore , a meta-analysis by Shu et al. ( 2022 ) synthesized that organic amendments significantly increased micr obial div ersity and shifted micr obial comm unity structur e compared to mineral-only fertilization.
Ho w e v er, ther e ar e also concerning r eports about the incr eased abundance of hea vy metals , antibiotics , and pathogenic , e v en potentiall y antibiotic-r esistant micr oor ganisms, in soils r eceiving farmy ar d manure (FYM) (Kumar et al. 2013, Lin et al. 2016. Aside from possible soil pollution, animal husbandry itself str ongl y contributes to global warming through emissions of the greenhouse gases methane and nitrous oxide, calling for an urgent reduction of animal stocking densities (Eisen and Brown 2022 ).
For the above-mentioned reasons, it is crucial to understand how soil microbial communities react to long-term animalmanure-based fertilization at contrasting intensities. For example, a study by Wang et al. ( 2021 ) based on shotgun sequencing sho w ed bacterial community structure as well as functional gene composition to differ across a gradient of organic fertilization in a grassland system whereas fungi and ar chaea, as w ell as microbial α-div ersity, r emained unaffected. A study on arable soils by Ma et al. ( 2020b ) based on shotgun sequencing combined with stable isotope phosphorous (P) and carbon (C) labelling demonstrated that long-term FYM application alters soil or ganic P stoc ks and cycling and that microbial functional gene abundance was coupled with P cycling rates with differences between mean and high input systems.
To provide further knowledge about how animal-based manure fertilization intensity (FI) affects the soil microbiome, the present study aims to analyze ho w lo w-input compared to high-input fertilization affects bacteria and fungi in three FYM-receiving certified and pr actice-r ele v ant FSs of the 42-year-old DOK [bio-Dynamic , bioOrganic , K on ventionell (German for conventional)] arable long-term trial in Switzerland (Krause et al. 2020 ). For this purpose, we collected soil samples from a biodynamic (BIODYN), an organic (BIOORG) and a conventional (CONFYM) system, of whic h all r eceiv ed FYM at two fertilization intensities (FI; 0.7 livestock units (LU) and 1.4 LU). We then analyzed the nutritional status (C and N) of the soil, microbial biomass and respiration, and fungal and bacterial diversity.
Pr e vious work of the DOK trial already identified the five FSs (additionally an unfertilized control (NOFERT) and an exclusively mineral fertilized system (CONMIN)) to harbor distinct microbial communities (Hartmann et al. 2015 ). The main factor distinguishing the systems was whether the y recei ved FYM, whereas the impact of plant protection regimes was of subordinate importance. As the FS effects are already well described (Hartmann et al. 2015. Esperschütz et al. 2007 ), we hereafter mainly focus on FI effects.
We hypothesized that the w ell-kno wn nutritional constraints in the low-input systems of the DOK trial might select for more diverse and rather oligotrophic microbial communities being able to efficiently degrade more complex recalcitrant C compounds and likely retrieve energy and nutrients via different metabolic str ategies compar ed to high-input systems . T her efor e, we hypothesize to observe increased α-diversity under low-input systems accompanied by a shift in microbial community structure to w ar ds oligotrophic taxa based on previously suggested trophic modes (Fierer et al. 2007, Leff et al. 2015, Ho et al. 2017. Conv ersel y, copiotr ophic gr oups that pr efer entiall y metabolize labile C sources ar e expected to be enriched in high-input soils and CONFYM systems receiving the least recalcitrant organic inputs.

I. Study site description
The DOK long-term field experiment was established in 1978 and is located in Therwil, Switzerland (47 • 30 N; 7 • 32 E; 306 m abov e sea le v el). The soil is a haplic luvisol with 15% sand, 70% silt, and 15% clay and the climate is mild with a mean annual temperature of 10.5 • C and mean annual precipitation of 842 mm (Krause et al. 2020 ). The trial compar es four differ ent FSs, of whic h two organic systems (BIOD YN , BIOORG) and one conventional (CON-FYM) system at two FI (0.7 and 1.4 LU) were included in the current study. The DOK experiment is based on a 7-year crop rotation (sixth crop rotation period: maize, so y a, winter wheat, catch crop, potatoes, winter wheat, gr ass clov er, and gr ass clov er) in thr ee tempor all y shifted par allels . T he field plots ar e arr anged as r andomized split-block design with four replicates of eac h tr eatment and cr op. A sc heme of the experimental design layout is presented in Figure S1 (Supporting Information).
The high-input versions of BIOD YN , BIOORG, and CONFYM receiv e or ganic fertilizer corr esponding to 1.4 LU per hectar e, while the low-input versions receive 0.7 LU per hectare. Aside from slurry, CONFYM r eceiv es stac ked manur e, BIOORG r otten manure, and BIODYN composted manure . T he CONFYM system is additionally complemented with mineral fertilizer following Swiss fertilization recommendations ( https://www.agrocontrol.ch/oln ) and is, ther efor e, the system r eceiving the highest N inputs. Detailed information on annual nutrient inputs is provided in Table 1 . During the first two cr op r otation periods the manurer eceiving systems wer e maintained at 0.6 and 1.2 LU, and fr om the third onwards at 0.7 and 1.4 LU, r espectiv el y, due to increasing stocking densities in the regionally applied FSs. Weed management in the BIOORG and BIODYN is done mechanically while in the CONFYM systems weeds are controlled with herbicides. Pests and diseases in the CONFYM system are controlled with synthetic insecticides and fungicides while in the BIODYN and BIOORG systems, only biological pest control agents are applied. In the BIODYN system biod ynamic pre parations are applied to soils , plants , and compost, and plant growth regulators in the CONFYM system. More detailed information about fertilizer treatment and management can be found in Table 1 and Krause et al. ( 2020 ).

II. Soil sampling
We sampled bulk soil betw een ro ws of winter wheat in late February 2019 before the first spring fertilization. For each plot (5 m × 20 m), 12 soil cores were taken with an auger to a depth of 20 cm (plough horizon), pooled, and immediately cooled. A total of 24 samples were taken [six treatment combinations (three FSs and two FI) from four field replicates]. The soil was immediatel y tr ansported to the labor atory and sie v ed to 5 mm. Subsamples for microbial biomass and activity analyses were immediatel y stor ed at 4 • C wher eas subsamples for molecular anal ysis wer e stor ed at −20 • C, and subsamples for soil c hemical analyses were air-dried and stored at 4 • C until further processing.

III. Biogeoc hemical anal ysis
To determine total soil nitrogen (N tot ) and carbon content (C org ), a subsample of the air-dried soil was ground, homogenized, and analyzed via dry combustion on a CN analyzer in duplicates (Elementar Analysensysteme GmbH, Vario MAX Cube, Hau, Germany). Soil pH was determined in an aqueous suspension 1:2.5 (w eight/v olumne). P ermanganate oxidizable carbon (PoxC) was extr acted and anal yzed following the principle of Weil et al. ( 2003 ) and modified as described in Bongiorno et al. ( 2019 ). Microbial biomass C (C mic ) and microbial biomass N (N mic ) were assessed in triplicates using the c hlor oform fumigation method (Vance et al. 1987 ).
To determine soil basal r espir ation, subsamples of field-moist soil were incubated at 25 • C for 7 da ys . Afterwards , soil basal respiration was assessed in triplicates by incubation of soil samples for 28 days in hermetically sealed microcosms, and the capture of CO 2 in alkali acid traps (0.025 M NaOH) as described in von Arb et al. ( 2020 ). Soil basal r espir ation was defined as the av er a ge C-miner alization r ate during the second week of incubation. Table 1. Main c har acteristics and mean annual inputs of the different FSs in the DOK trial. The biodynamic (BIODYN), bioorganic (BIOORG), and conventional (CONFYM) FSs are maintained at two FI: 0.7 and 1.4 LU fertilization equi valents. LU n umbers refer to the organic matter (OM) input before system-specific processing of manure inputs took place. Ntot refers to total nitrogen inputs; Nmin is the sum of NH 4 + -N and NO 3 −N from slurry or mineral fertilizer inputs. P refers to phosphorus. Data about mean annual inputs are r etrie v ed fr om Kr ause et al. ( 2022 ).

IV. Molecular biological analysis
For metabar coding, DN A w as extr acted fr om l yophilized soil samples (400 mg) using the NucleoSpin 96 Soil kit (Mac hery-Na gel, Dürr en, German y) with the SL2 + Enhancer SX lysis buffer according to the manufacturer's instructions. Each sample was extracted in two technical replicates and pooled afterw ar ds. Before PCR, DN A w as diluted (1:10) to reduce inhibitory effects on PCR. A two-step PCR a ppr oac h using CS1/CS2-ta gged (Fluidigm, South San Francisco, CA, USA) primers targeting the V3-V4 region of the 16S rRNA gene (341F and 806R as modified in Frey et al. 2016 ), and primers targeting the fungal internal transcribed spacer region ITS2 (ITS3ngsmix1-5 and ITS4ngsUni; Tedersoo and Lindahl 2016 ) w as deplo y ed. A negativ e contr ol containing double-distilled water instead of DNA and positive controls containing ZymoBIOMICS Micr obial Comm unity DN A Standar d (Zymo Resear c h Cor por ation, Irvine, CA, USA) or a fungal mock community (Bakker 2018 ) wer e included, r espectiv el y. The first PCR with the CS1/CS2-tagged primers (Fluidigm) was performed in technical triplicates (Table  S1, Supporting Information, for primer sequences and PCR cycling conditions) using an SYBR green approach (Kapa SYBR Fast qPCR Kit Master Mix (2 ×) Univ ersal; Ka pa Biosystems, Wilmington, MA, USA) on a CFX96 Touch Real-Time PCR Detection System (Bio-Rad Laboratories , Hercules , C A, USA). Subsequently, PCR triplicates were pooled and purified using a magnetic bead solution ( https://openwetwar e.or g/wiki/SPRI _ bead _ mix ). A subsample of each purified DN A sample w as loaded on an agarose gel (1.25%) for visualization and validation. The second PCR, library preparation, and paired-end sequencing on an Illumina MiSeq sequencing platform (Illumina, San Diego, CA, USA) using the MiSeq v3 chemistry were performed at the Genome Quebec Innovation Center (Montreal, Canada) according to the amplicon guidelines provided by Illumina. Raw sequences and meta-data are deposited in NCBI Sequence Read Arc hiv e (SRA) under the accession number PR-JNA841851.

V. Bioinformatics
Raw sequence data were processed using the Euler Scientific Compute Cluster at ETH Zürich. For most steps, USEARCH v11.0.667 (Edgar 2010 ) commands were used with default settings unless stated otherwise.

Clustering and taxonomy
The pr epr ocessed 16S rRNA and ITS r eads wer e cluster ed using CLUSTER_O TUS (Edgar 2013 ). T he taxonomy for 16S and ITS data was assigned using SINTAX (Edgar 2016 ) (sintax_cutoff 0.85) based on the r efer ences SILVA_128_16S_utax_work.fa (Quast et al. 2013 ) and UNITE_v82_Fungi_04 Befor e downstr eam anal ysis nonbacterial (arc haea, mitoc hondria, and c hlor oplasts) and nonfungal sequences wer e r emov ed, resulting in 833 054 bacterial sequences assigned to 7114 OTUS, and 750 486 fungal sequences assigned to 2088 O TUs . Results of positiv e contr ols containing bacterial and fungal moc k comm unities are shown in Figure S2 (Supporting Information).

VI. Statistical analysis
All analyses were conducted in r v 4.1.2 (R Core Team 2021 ) and r studio v 1.4.1717 (RStudio Team 2020 ) and plots were created using ggplot2 (Wickham 2016 ) unless indicated otherwise.

Univ aria te da tasets
Linear mixed-effect models using the nlme:: lme function (Pinheiro et al. 2020 ) were used to assess the effects of FI, FS, and their interaction on soil geochemical parameters, and microbial α-diversity. Field-plot nested in block was set as a random factor while varIdent modelled the inconsistent variance between FSs . T he nlme:: ano voa_lme function (Pinheiro et al. 2020 ) was used to r etrie v e the statistical significance of the main effects. Estimated marginal means and 95% confidence intervals for each FI le v el within FSs were obtained using the emmeans:: emmeans function (Russell et al. 2020 ) and plotted along with the raw data. Planned contrasts between FI within FSs were again analyzed with the emmeans function if a significant main effect for FI or interaction effect has been found. Planned contrasts between FSs within FI le v els wer e similarl y anal yzed if a significant FS or interaction effect has been found but r esults, her eof, ar e onl y pr ovided in the supplementary information. Data were transformed using log, squar e-r oot, or inv erse functions to satisfy the assumption of normal distribution and variance homogeneity of model residuals.

Sequencing datasets
Data wer e anal yzed using the phyloseq pac ka ge (McMurdie and Holmes 2013 ), and r ar efaction plots ( Figure S3, Supporting Information) wer e cr eated using the function phyloseq.extended:: ggr ar e. Bacterial and fungal data were rarefied to even depth using the phyloseq:: r ar efy_e v en_depth function r esulting in 6651 bacterial and 1896 fungal O TUs , r espectiv el y (25% of bacterial and 40% of fungal sequences were lost). α-diversity estimates (Shannon index and observed richness) were calculated using the phyloseq:: estimate_richness function and Pielou's evenness was obtained by dividing the Shannon index by the natural logarithm of the observ ed ric hness. Experimental tr eatment effects on the r espectiv e α-diversity metrics were assessed using linear mixed-effect models as described for the other univariate data.
For β-div ersity anal yses, we filter ed the dataset to r emov e OTUs with fewer than 10 reads in less than 5% of the samples to reduce the extreme sparsity of microbiome data (Cao et al. 2021 ). After filtering, the dataset contained 611 730 bacterial and 452 784 fungal sequences clustering into 4138 bacterial and 1200 fungal O TUs . O TU tables wer e tr ansferr ed into r elativ e abundances by total sum scaling. Permutational multivariate analysis of variance (PERMANOVA) (Anderson 2001 ) using the vegan:: adonis2 function (Oksanen et al. 2022 ) based on Bray-Curtis distances was used to test treatment effects (FS, FI, and their interaction) at 9999 permutations restricted on the field blocks . T he homogeneity of m ultiv ariate dispersion in the tr eatment gr oups was confirmed using PERMDISP (Anderson et al. 2006 ) implemented through the betadisper function in vegan (Oksanen et al. 2022 ). P airwise differ ences between FI within FSs were obtained using the rv aidemoir e :: pairwise .perm.mano va function (Hervé 2022 ). To gr a phicall y pr esent bacterial and fungal comm unity structur es, nonmetric m ultidimensional scaling (NMDS) w as emplo y ed (Bra y-Curtis distances , k = 3 and trymax = 100) using vegan:: metaMDS (Oksanen et al. 2022 ). A canonical analysis of principal coordinates (Anderson and Willis 2003 ) was performed using the CAPdiscrim function implemented in the BiodiversityR package (Kindt and Coe 2005 ) constraining the function by FS and FI. Joint biplots with the environmental variables correlating with the projections of the or dination w ere calculated using vegan:: envfit (Oksanen et al. 2022 ). P -values of correlations were corrected for multiple testing based on Benjamini-Hochberg corrections using stats:: p.adjust (R Core Team 2021 ). The graphical presentation was restricted to variables that exceeded an a priori threshold of r 2 > 0.6 and adjusted P -values < .01.
Each bacterial OTU was classified into oligotroph or copitroph lifestyle based on its ribosomal RNA ( rrn) copy number (oligotrophy < 0.5, copiotroph ≥5 according to Bledsoe et al. 2020 ) using the RDP classifier (Wang et al. 2007 ) and the NCBI taxonomy integrated within the rrn operon database (Stoddard et al. 2015 ). The classification was performed at the lo w est possible taxonomic r ank, pr ovided in the rrn operon database. The r elativ e abundance of copiotr oph, oligotr oph, and unclassified OTUs was summed for each soil sample to calculate the copiotr oph-to-oligotr oph r atio within a soil bacterial comm unity. To r etrie v e information about fungal trophic modes, the FUNGuild database and its bioinformatics script were used (Nguyen et al. 2016 ) leading to a ppr oximatel y 50% of OTUs being classified (with confidence le v els of "possible ," "probable ," and "highl y pr obable"). The r elativ e abundance of OTUs assigned to a specific trophic mode or a combination of trophic modes was summed per sample. Experimental treatment effects on the copiotr oph-to-oligotr oph r atio as well as on the relative abundance of fungal trophic modes were assessed with linear mixed-effect models as described for the univariate data.
We finally screened the data for OTUs specifically associated with the different FSs and FI. For this test, we r emov ed sequences with fewer than 20 reads in less than 10% of the samples for bacteria and fewer than 50 reads in less than 25% of the samples for fungi. The indicspecies:: multipatt function (De Cáceres and Legendre 2009 ) w as emplo y ed to identify OTUs associated with one or mor e tr eatment combinations (10 4 perm utations, "r.g" function to correct for unequal group sizes) (Dufrêne andLegendre 1997 , De Cáceres et al. 2012 ). Multiple testing correction was performed by calculating q -values using the qvalue package (Storey et al. 2021 ). A bi-partite association netw ork betw een treatment groups and statistically associated bacterial and fungal indicator OTUs was generated using Cytoscape v3.9.1 (Shannon et al. 2003 ) with treatments as source nodes , O TUs as target nodes , and association strength as connecting edges. Only treatment-OTU associations with q < 0.05 were selected and the network was built using the Allegro Fruchterman-Reingold layout with edges weighted by association strength (normalized square-root transformed).

I. Geochemical and biological soil quality indicators
C org , PoxC, and N tot were elevated under 1.4 LU compared to 0.7 LU (Fig. 1 A-C). Planned contrasts between FI within FSs rev ealed significantl y higher v alues for C org , PoxC, and N tot, in highcompared to low-input plots except for C org in BIOORG. FSs dif-  Table  S2, Supporting Information). Higher C mic and N mic values were identified under 1.4 LU compared to 0.7 LU (Fig. 1 D and E) with planned contr asts r e v ealing significant differ ences between FI in BIODYN and BIOORG but not in CONFYM. FSs differed in microbial biomasses with the highest values in BIOD YN , follo w ed b y BIOORG and CONFYM (Fig. 1 F; Table S2, Supporting Information). Higher soil DNA content was found in 0.7 LU compared to 1.4 LU plots with planned contrasts for FI being significant in all FSs (Fig. 1 F). FSs also differed in their DNA content with the highest values in BIOD YN , follo w ed b y BIOORG and CONFYM (Fig. 1 F; Table  S2, Supporting Information). Microbial biomass C:N was slightly, but statistically significant, elevated at 0.7 LU compared to 1.4 LU with planned contrasts revealing significant differences between FI only for BIODYN (Fig. 1 G). Mor eov er, FSs differ ed in their mi-crobial biomass C:N with the highest values in CONFYM follo w ed by BIOORG and BIODYN (Fig. 1 G; Table S2, Supporting Information). Basal r espir ation was higher at 1.4 LU compared to 0.7 LU and planned contrasts revealed significant differences between FI for all FSs (Fig. 1 H). The metabolic quotient remained unaffected by FI but was affected by FSs, with the lo w est values in BIODYN (Fig. 1 I).

Data overview
After r emoving an y nonbacterial sequences and filtering r ar e OTUs (pr e v alent in less than 5% of the samples and with fewer than 10 reads), the bacterial dataset contained 4138 O TUs . T he five most abundant phyla were Proteobacteria , Actinobacteria , Verrucomicrobia , Planctom ycetes , and Acidobacteria (Figur e S4A, Support-ing Information). The filtered fungal dataset contained 1200 OTUs and the four most abundant phyla wer e Ascom ycota , Mortierellom ycota , Basidiom ycota , and Chytridiom ycota (Figur e S4B, Supporting Information).

Microbial αand β-diversity
For bacteria, ɑ -diversity (Shannon and observed richness) was slightly higher at 0.7 LU compared to 1.4 LU, but differences between FI within FSs wer e statisticall y supported only for BIODYN ( Fig. 2 A and B) with compar able v alues between FSs (Fig. 2 B; Table  S2, Supporting Information). Bacterial community evenness was higher in 0.7 LU compared to 1.4 LU, while planned contrasts between FI within FSs r e v ealed no significant differences (Fig. 2 B). Bacterial comm unities wer e mor e e v en under BIODYN compar ed to BIOORG and CONFYM ( The PERMANOVA test identified distinct bacterial and fungal comm unity structur es in 0.7 LU and 1.4 LU plots and different FSs ( Table 2 ). The FS explained around twice as m uc h data v ariation (bacteria: 24%, fungi: 27%) compared to FI (bacteria: 12%, fungi: 13%) ( Table 2 ). For bacteria, the pairwise PERMANOVA r e v ealed trends for differences between FI in BIOORG ( P = .085), and BIO-DYN ( P = .085) but not in CONFYM ( P = .259) whereas, for fungi, comm unities differ ed between FI in BIOORG ( P = .029), BIODYN ( P = .028), and CONFYM ( P = .030).
The NMDS of bacterial and fungal communities confirmed effects identified by the PERMANOVA, with FSs separating along the first axis, and FI separating along the second axis ( Fig. 3 A  and B). Microbial community structure differed between FSs along the gradient of BIODYN-BIOORG-CONFYM, whereas, in the fungal dataset, BIODYN and BIOORG r epr esent one cluster and CONFYM another one (Fig. 3 A and B). These patterns were also confirmed by the CAP ordinations with reclassification values of 79% and 92%, r espectiv el y (Fig. 3 C and D). In the bacterial and fungal dataset C mic , N mic , C org , PoxC, N tot , and pH correlated with the projections of communities from BIODYN and BIOORG at 1.4 LU fertilization (Fig. 3 C; Table S3, Supporting Information). For bacteria, DNA content correlated with projections of 0.7 LU systems and grain yield with projections of 1.4 LU CONFYM (Fig. 3 C and D).

Micr obial tr ophic lifestyles and indicative OTUs
Ov er all, bacterial comm unities wer e oligotr oph dominated under all FSs and fertilization le v els (Figur e S5A, Supporting Information). The copiotr oph-to-oligotr oph r atio was significantl y higher under 1.4 LU compared to 0.7 LU with planned contrasts revealing significant differences between FI only for CONFYM (Fig. 4 ). BIODYN and BIOORG sho w ed a lo w er copiotr oph-to-oligotr oph r atio compared to CONFYM (Table S2, Supporting Information). The distribution of fungal trophic modes sho w ed a more complex picture . T he cumulative relative abundance of OTUs associated with a single trophic mode was lowest for the symbiotrophs, followed by pathortophs and sa pr otr ophs . T he majority of O TUs , ho w e v er, were assigned to a combination of se v er al tr ophic modes or r e-mained unclassified and no clear treatment dependent patterns wer e observ ed (Figur e S5B, Supporting Information).
A total of 110 bacterial and 67 fungal O TUs (bO TU and fO TU, r espectiv el y) wer e associated ( q < 0.05) with specific treatments or treatment combinations (Table S4, Supporting Information). The bipartite network constructed based on these indicator OTUs sho w ed 0.7 LU and 1.4 LU to be separated from each other as well as FSs (Fig. 5 ). FI of eac h FS gr ouped closel y except for BIOORG, wher e a bidir ectional formation was observ ed with 0.7 LU BIOORG closely associated with the CONFYM systems and 1.4 LU BIOORG associated with the BIODYN systems (Fig. 5 ).
When exclusiv el y looking at OTUs indicative of FI and simultaneously associated with all three FSs of the respective FI, four indicator bOTUs and zero fOTU were identified for 1.4 LU and three bOTUs and two fOTUs for 0.7 LU systems (Table 3 , Fig. 5 ). Two out of the four bOTUs indicative of 1.4 LU were attributed to a copiotrophic lifestyle, while none of the bOTUs indicative of 0.7 LU systems was classified as copiotroph ( Table 3 ). The two fOTUs indicative of 0.7 LU could not be assigned to a trophic mode.
The cluster of nodes indicative for the two 1.4 LU organic systems was built exclusiv el y of bacterial OTUs while the cluster of nodes indicative for 0.7 LU organic systems was built of fungal OTUs only ( Fig. 5 ; Table S4, Supporting Information). Overall, fungal indicator OTUs wer e mor e often associated with 0.7 LU systems and CONFYM compared to 1.4 LU BIODYN and 1.4 LU BIOORG ( Fig. 5 ; Table S4, Supporting Information).

Discussion
The need for sustainable a gr oecosystems combined with the demand for decreasing livestock density (Köninger et al. 2021 , Eisen andBrown 2022 ) calls to further identify how a reduction of animal manure-based FI affects soil microbial communities. By using soil samples obtained from winter wheat plots of a 42-year-old long-term trial, we found that high inputs (1.4 LU) of animal manur es in differ ent FSs (i) incr eased soil C and soil N contents, (ii) tended to decrease soil bacterial alpha-diversity, and (iii) altered soil microbial community structure as compared to low-input systems (0.7 LU) across all FSs.

I. High-input systems increase soil C and N content compared to low-input FSs
Chemical soil properties provide information on the nutritional status of the soil r epr esenting the micr obial habitat (Vor oney 2007 ). The amount of C introduced via organic fertilization dir ectl y, and indir ectl y via plant r oots and rhizodeposition, determines the soil C contents with higher values under high-compared to low-input. This finding is consistent with our expectation, and what has been pr e viousl y shown at other experimental sites (e.g. Francioli et al. 2016, Ma et al. 2020a ). The higher soil C contents in BIODYN may be related to the quality of introduced C as this system r eceiv es composted manure with higher recalcitr ance compar ed to BIOORG r eceiving r otten and CONFYM r ecei ving stack ed man ure. Gi ven the strong link between soil C and other geoc hemical par ameters, similar patterns wer e observ ed for PoxC and N content.
Ov er all, micr obial biomass C and N were elevated under highcompared to low-input systems, which was also observed in another long-term trial (Ma et al. 2020b ). This is partially in line with a r ecent meta-anal ysis (Ren et al. 2019 ) showing that ov er all organic fertilizer incr eases micr obial biomass, while more specific the effect of FYM was highly variable and remained insignificant.  It is known that the majority of soil microbes are heterotrophs that obtain energy and nutrients from organic matter and that their growth and activity depend on the concentration and suppl y r ate of bioav ailable or ganic material (Coleman et al. 2004 ). We indeed observed increased basal respiration rates in highinput compared to low-input systems. As the metabolic quotient, a measure of microbial substrate use efficiency (Wardle and Ghani 1995 ), remained unaffected by FI, we assume that the increased soil basal r espir ation under high-input systems is more strongly driven by the observed increased abundance of microbes than by distinct C-mineralization machinery.
Usually, soil DNA content and microbial biomass are highly correlated (Semenov et al. 2018 ) but sur prisingl y, we found higher DNA contents in soil sampled from low-compared to high-input systems. A possible explanation might be the c hr onic P-depletion of the low-input systems in the DOK trial (Jar osc h et al., submitted) likely resulting in freer and more positively charged soil particles acting as putative binding places for passiv el y or activ el y r eleased extr acellular DNA (Na gler et al. 2018 ).

II. Distinct microbial communities under low-compared to the high-input FS
The importance of soil microbial diversity for maintaining ecosystem functioning (Delgado-Baquerizo et al. 2016 ) and stability (Wagg et al. 2021 ) depicts the need to better understand factors sha ping micr obial comm unities in a gr oecosystems to counter act soil degradation (Cavicchioli et al. 2019, Zhou et al. 2020 ) and to pr eserv e or r estor e a gr oecosystem functioning in the futur e.
Gener all y, the effect of fertilization type (or ganic, miner al, or mixed) on soil microbes is complex and contrasting trends were reported in the literature depending on the fertilizer origin and amount, and the edaphic factors. A recent meta-analysis by Shu et al. ( 2022 ) synthesized that organic amendments significantly incr eased micr obial div ersity and shifted micr obial comm unity structur e compar ed to miner al-onl y fertilization. Ho w e v er, the performance of microbial α-diversity varied substantially with organic amendment types, microbial groups, and changes in soil pH. Exemplary more in detail, Francioli et al. ( 2016 ) observed that soils Table 3. List of indicator OTUs associated with contrasting FI. Given are fungal and bacterial indicator OTUs that ar e positiv el y associated ( q < 0.05) with a FI le v el and also associated with all three FSs of the r espectiv e FI. The taxonomic assignment is provided at the lo w est possible le v el and r elativ e abundance across the entire dataset is given. Attributed lifestyles are determined based on rnn copy numbers (oligotrophy < 0.5, copiotroph ≥ 5) to the lo w est possible taxonomic rank provided in the rnnDB. The fungal lifestyles of the two fOTUs remained unclassified using FUNGuild. Relative abundance is given in %.

Treatment association OTU
Relati  . Treatment effects on the bacterial copiotroph-to-oligotroph r atio. The copiotr oph-to-oligotr oph r atio was calculated based on the cum ulativ e r elativ e abundance of either oligotroph or copiotroph classified bacterial O TUs . T he plot shows raw data (semitransparent dots) and estimated marginal means with 95% confidence intervals of linear mixed-effect models assessing treatment effects on the copiotr oph-to-oligotr oph r atio (based on rrn copy numbers amended with FYM and/or a mixture of FYM and mineral fertilizer ar e c har acterized by an incr eased nutritional status, higher microbial biomass, higher bacterial α-diversity, and distinct micr obial comm unity structur e as compar ed to exclusiv el y miner al fertilized soils. Hartmann et al. ( 2015 ) sho w ed in the DOK trial, that or ganicall y fertilized FSs incr ease the micr obial ric hness and shift comm unity structur e when compar ed to exclusiv el y miner al fertilized systems. Based on these observations, we hypothesized to find a higher α-diversity in low-compared to high-input systems because a diminished nutrient availability under low-input likely requires a taxonomically and functionally more versatile community to degr ade complex r ecalcitr ant C and r etrie v e ener gy fr om other sources than the organic fertilizer. Howe v er, we found bacterial and fungal α-diversity to be onl y mar ginall y affected by contrasting fertilizer intensity. The small effect of FI on α-diversity is lar gel y in line with a study by Wang et al. ( 2021 ) showing bacterial, fungal, and arc haeal α-div ersity to be unaffected ov er an organic FI gradient in a grassland soil. Whether the herein observed weak increase in α-diversity under low-compared to high-input BIODYN translates into distinct functioning has not been studied and remains speculative.
As hypothesized, we found distinct fungal and bacterial comm unity structur es under high-and low-input in all FSs . T he correlations of C org , PoxC, and total N content with the projections of the ordination of bacterial and fungal community structures under high-input underpin the possible shift of microbial lifestyles due to contrasting nutrient availabilities (Kim et al. 2021 ). Char acterizing bacterial comm unities based on their putative ecological lifestyles, we identified oligotroph-dominated communities in all FSs and FI. T his ma y be related to the fact that copiotr oph micr obes thriv e under nutrient-ric h conditions; ho w e v er, we collected the samples before the spring fertilization. We presume that the putative copiotroph-classified OTUs would increase in relative abundance after spring fertilization e v ents. Mor eov er, the ov er all low abundance of copiotrophic OTUs might also result from the rather conservative threshold set for rrn copy numbers to classify copio-vs. oligotrophic OTUs (Bledsoe et al. 2020 ).

Figure 5.
Bi-partite association network of fungal and bacterial indicator O TUs . T he network is showing significant ( q < 0.05) positive associations between treatment groups (FSs × FI) and fungal (diamonds) or bacterial (circles) operational taxonomic units (bO TUs , fO TU) based on indicator species analysis. Node sizes reflect the relative abundance of a given OTUs across all samples in the fungal or bacterial dataset, respectively. The Allegr o Fruc hterman-Reingold algorithm was a pplied to construct the netw ork with edges w eighted accor ding to the association strength. Nodes outlined in bold r epr esent indicator OTUs specifically associated with high-or low-input and independent of the FS (see Table 3 ). Node colors r epr esent differ ent phyla. BIODYN = biodynamic (gr een), BIOORG = bioor ganic (blue), and CONFYM = conv entional (or ange). 0.7 = 0.7 LU, 1.4 = 1.4 LUs. Nodes without annotation on the phylum le v el wer e excluded. A full list of indicator species, their taxonomic information, and attributed lifestyles are given in Table S4 (Supporting Information).
The ratio of putative copiotroph-to oligotroph-classified bacteria was slightly higher under high-compared to low-input as well as in conventional compared to organic FSs, pointing out a shift tow ar ds a more copiotroph community under high-input and CON-FYM systems. Fungi cannot be classified into ecological lifestyles based on their rrn copy numbers but they usually show more oligotr ophic featur es than bacteria (Ho et al. 2017 ) and are attributed with a strong capability of degrading recalcitrant polymers to gain nutrients and energy (Van der Wal et al. 2013 ). The distribution of fungal trophic modes based on FUNGuild annotations did not yield clear results as half of the OTUs remained unclassified and many OTUs could not be categorized into one but rather a combination of serval trophic modes. Ho w ever, the slightly higher C mic to N mic ratio in CONFYM and low-input systems as well as the more frequent association of indicative fungal OTUs with low-input and CONFYM in the bi-partite network support the idea of a fungal dominance in the more oligotrophic habitats.
Looking at indicator OTUs associated with either low-or highinput systems, but compellingly also associated with all three FSs of the r espectiv e FI le v el, w e found tw o fungal indicators with described oligotr ophic featur es and thr ee oligotr oph classified bacterial OTUs indicative of low-input systems. In contrast, we found no fungal and four bacterial OTUs to be indicative of high-input systems of which two were classified as putative copiotrophs . T he copiotrophic indicators associated with high-input systems are both assigned to the endospor e-forming Gr am-positiv e bacterial genus Bacillus (Nicholson et al. 2000, Nicholson 2002, of which man y members ar e known for their plant gr owth pr omoting features (Goswami et al. 2016, Saxena et al. 2020. The bacterial indicator OTUs associated with low-input systems were all classified as oligotrophs and were attributed to diverse ways of energy metabolism. For example , bO TU473 is assigned to Geobacter , which is known for establishing electrical contacts with extracellular electron acceptors and other organisms, and acts as the primary agent for coupling the oxidation of organic compounds to the reduction of insoluble Fe(III) and Mn(IV) oxides in many soils and sediments (Lovley et al. 2011, Yi et al. 2013, Holmes et al. 2017. bOTU972 was assigned to Roseiflexus , whic h can gr ow photoheter otr ophicall y under light and chemoheter otr ophicall y in the dark (Hanada 2014 ).
In summary, indicator OTUs associated with high-input systems sho w ed mor e copiotr ophic featur es than indicators associated with low-input systems, which is in line with the enhanced copiotr oph-to-oligotr oph r atio under high-input systems. The higher availability of nutrients and the more easily available C in high-input systems may have depleted some oligotrophic functional members and enriched a few copiotrophic members potentially leading to a higher nutrient dependency. We found indications that the diminished nutrient availability in the low-input systems selects for a functionally versatile microbial community being able to degrade complex recalcitrant C and retrieve energy and nutrients from other sources under nutritional constraints.
Nonetheless, we can only speculate about the possible ecological functions of the identified indicator O TUs , and the classification of bacterial lifestyles is an a ppr oximation onl y. Extended anal yses suc h as the pr ofiling of micr obial metabolic ca pacity (Creamer et al. 2016 ), proteomics (Qian and Hettich 2017 ), shotgun metagenomics (Vogel et al. 2009 ), and combinations thereof (Martinez-Alonso et al. 2019 ) might further r e v eal distinct effects of FI on soil multifunctionality influencing agronomic variables.

Conclusion
We sho w ed that a reduction of FYM a pplication fr om 1.4 LU to 0.7 LU translates into lo w er soil C and N content selecting for distinct microbial communities in bioorganic , biodynamic , and conv entionall y farmed winter wheat plots . T hus , a decrease in livestock density and herewith-associated reduced levels of FYM in-puts in organic and integrated FSs change the habitat quality of a gricultur al soils and affect soil microbial communities . T he reduced nutrient availability under low-input selects for more oligotr ophic micr obiomes likel y mor e efficientl y obtaining nutrients fr om v arious carbon sources; a potentially beneficial trait considering future agroecosystems.