The focus of the present study was a medium-sized, high performance dairy farm in England, UK which housed ~200 milking Holstein Friesian cattle at the time of sampling. Comprehensive descriptions of the farm layout, antibiotic usage and waste-handling practice are provided by Baker et al. (2022) and Todman et al. (2024). Cattle are housed indoors and resulting waste streams (feces, urine, antimicrobial footbath washings, rainwater run-off and antibiotic-contaminated milk) are directed into a 3,000 m³ slurry tank following liquid–solid separation. Thus, the slurry tank contains a cocktail of microorganisms and chemical compounds receiving regular input.

Slurry is periodically used to fertilize nearby grassland and arable fields. However, as the farm lies within a nitrate vulnerable zone (NVZ), slurry application is prohibited between October and January. An NVZ is defined as an area deemed at risk from agricultural nitrate pollution, and this designation is currently applied to over half of land in England (GOV.UK, 2025). Excess slurry produced during the ‘closed period’ is accommodated in an additional 8,000 m³ lagoon. An adjoining field with a long-term history of multiple, sub-annual slurry applications was selected to represent a slurry-treated grassland site (52^o50’23.93”N, 1^o14’38.82”W). The site is characterized by heavy clay loam topsoil and was maintained as a permanent grassland for the duration of the study. Grass was occasionally harvested prior to slurry application for silage production. Surface application of slurry at this location was carried out using an umbilical system connected to slurry tanker-mounted dribble-bars. Slurry was drawn from either the primary tank or overflow lagoon.

A nearby site with no documented history of systematic agricultural waste exposure was chosen to represent an untreated grassland site (52^o50’1.64”N, 1^o15’14.65”W). The untreated site is characterized by silty clay loam topsoil and has been maintained as a permanent grassland for more than 50-years. The two field sites share the same local climate; however, they are separated by both abiotic and biotic barriers including hedges, several rows of trees and a road, which we assume limits cross-exposure for the purposes of this study.

Soil samples were collected over an entire season of slurry application from the treated site (seven occasions between May 2017 and May 2018). For the May 2017 application, treated soil samples were collected 5 days before, within 24 h after, and 56 days following slurry application to quantify short-term temporal responses. These intervals were chosen to represent (i) baseline pre-application conditions, (ii) the immediate introduction of slurry-borne microorganisms and resistance genes, and (iii) the subsequent recovery period after short-term slurry effects typically diminish. Subsequent sampling dates focused on the longer-term effects of slurry amendment.

Sampling of the untreated site was carried out twice, once in January and May 2018 (corresponding samples were collected from the treated site on the same day). To maximize capture of local heterogeneity, three soil cores were taken from five locations arranged 20 m apart in a ‘W’ transect. All soil cores were taken using a 10 cm stainless steel auger sterilized with ethanol between each use. Cores were deposited immediately in sterile plastic resalable bags and transported to the laboratory within 2 hours. Upon arrival, each triplicate set of cores was combined and thoroughly mixed with a sterile spatula. Fifty g of soil from each composite sample was stored at −80 °C pending DNA extraction; the remainder was stored at 4 °C for physiochemical analyses (Supplementary File 1). Samples sizes were n = 35 (slurry fertilized site soil), n = 10 (untreated site soil).

Slurry tank samples were collected monthly between June and October 2017 (× 5 sampling occasions). Samples were collected using a clean stainless-steel bucket and aliquoted into two sterile glass bottles (n = 10) as described by Baker et al. (2022). A visual timeline of the sampling strategy is summarized in Supplementary Figure 1.

Environmental DNA was extracted from triplicate composite soil samples using DNeasy PowerSoil Kits (Qiagen) according to the manufacturer’s instructions. DNA quality and yield were determined by spectrophotometry (NanoDrop 1000, Thermo Scientific) and fluorometry (Qubit, Invitrogen), respectively and extraction replicates subsequently pooled.

High sensitivity DNA assays (Qubit, Invitrogen) for negative control DNA extractions performed on UltraPure DNase/RNase-Free Distilled Water (Invitrogen) were below the limit of detection. Sequencing and demultiplexing of soil sample extractions (n = 45) was performed by Edinburgh Genomics on an Illumina NovaSeq platform (150 bp paired end libraries). The current work considers a subset of 10 slurry tank metagenomic libraries reported by Baker et al. (2022). Slurry tank samples were extracted with QIAamp Powerfecal Kits (Qiagen) and sequenced by Edinburgh Genomics on an Illumina HiSeq platform. After sequencing, adapter sequences and low-quality reads were filtered with TrimGalore (v 0.4.4) using default settings. On average, 12 Gbytes of data were produced per sample. Metagenomic sequences were uploaded to the European Nucleotide Archive (ENA) under the following accession: PRJEB38990. Individual accessions and descriptions of samples analyzed in this study are detailed in Supplementary File 2. Note, sample S86 was excluded from all analyses due to suspected contamination. Therefore, final sample sizes included in analyses were n = 10 (slurry), n = 34 (slurry fertilized soil), and n = 10 (untreated soil).

Short-read metagenomic libraries were subsequently assembled as individual samples using Megahit (v1.1.3) with ‘meta-large’ settings applied (k-mer range: 27–87, k-step = 10) and a minimum multiplicity for filtering of two. Three separate co-assemblies were also generated from slurry-treated soil, untreated soil and slurry metagenomes, respectively. For co-assemblies, a k-mer range of 29–89 (k-step = 10) and a minimum multiplicity for filtering of two was used.

Statistical analyses based on CAP vectors r > 0.4 indicated that tetracycline (Mann–Whitney p_adjust < 0.05; = 2.39 × 10⁻²), MLS (Mann–Whitney, p_adjust < 0.01; = 6.8 × 10⁻³), mupirocin (Mann–Whitney, p_adjust < 0.001; = 3 × 10⁻⁷) and rifamycin (Mann–Whitney, p_adjust < 0.001; = 5 × 10⁻⁵) ARGs distinguished treated and untreated soils. After correction for multiple testing, multidrug ARGs (Mann-Witney, p_adjust > 0.05; = 8.5 × 10⁻²) were not significantly different between the two soils (Figure 2A).

The average normalized abundance of total detected ARGs was significantly different between sites (Kruskall Wallis χ² = 27.4, p < 0.001; = 1 × 10⁻⁶). Specifically, total ARG abundance was greater in cattle slurry (0.60 ± 0.01 cpbg) relative to both treated (Pairwise Wilcoxon, p_adjust < 0.001; 2 × 10⁻⁹) and untreated (Pairwise Wilcoxon, p_adjust < 0.001; = 2 × 10⁻⁵) soil. There was a significant difference between the soil of the treated (0.25 ± 0.001 cpbg) and untreated site (0.23 ± 0.007 cpbg) (Pairwise Wilcoxon, p_adjust < 0.05; = 2.25 × 10⁻²).

PERMANOVA applied to Hellinger distances of the ARG category data suggested sample group had a significant influence on ARG compositions (pseudo-F = 582.3, p_perm < 0.001; = 1 × 10⁻⁵), contributing to approximately 19% of overall resistome variation. Furthermore, pairwise comparisons showed that the slurry resistome was significantly different from the soil samples, irrespective of site (treated site, t = 37.1, p_{perm (adjust)} < 0.001; = 1 × 10⁻⁵; untreated site, t = 24.0, p_{perm (adjust)} < 0.001; = 1 × 10⁻⁵), and that despite the high-level similarities hitherto described, soil samples from the treated field were also significantly different from those collected from the site with no history of slurry application (t = 4.0, p_{perm (adjust)} < 0.001; = 1 × 10⁻⁵). Nonetheless, CAP shows that both field site soil resistomes cluster together on axis 1, with separation primarily on axis 2 (Figure 1A). This was mirrored by bacterial taxa at the phylum level (pseudo-F = 3183.5, p_perm < 0.001; = 1 × 10⁻⁵) (for pairwise comparisons see Supplementary Table 1).

To understand the immediate impact of slurry application on the resistome of soil with a history of regular exposure, PERMANOVA was applied to a subset of treated site soil samples collected before and after the first application of slurry in May 2017 when 60 m³ ha⁻¹ of slurry was applied. These samples correspond to a soil time series collected 5 days before, within 24 h of, and 56 days after, slurry application. Time had a significant influence on the resistome of these treated soil samples (PERMANOVA, pseudo-F = 2.4, p_perm < 0.001; = 7 × 10⁻⁴). At the ARG category level, pairwise comparisons indicated that the resistome was altered significantly on the day of slurry application relative to 5 days before (t = 1.7, p_{perm (adjust)} < 0.05; = 1.2 × 10⁻²). Additionally, the resistome composition within 24 h of slurry application was significantly altered compared to samples obtained 56 days later (t = 1.8, p_{perm (adjust)} < 0.05; = 1.2 × 10⁻²). However, the resistome of samples collected 56 days after slurry application was not significantly different from the pre-application samples (t = 1.2, p_{perm (adjust)} > 0.05; = 2.05 × 10⁻¹), suggesting a transitory change in state. Differences were also identified in bacterial phyla compositions over the same period (pseudo-F = 4.7117, p_perm < 0.01; = 2.5 × 10⁻³). However, according to pairwise comparisons, bacterial phyla compositions remained distinct from pre-application soil assemblages 56 days after slurry application (Figure 1B; Supplementary Table 1).

To investigate the enrichment of slurry-associated ARGs in soil, we conducted targeted analysis of MLS and tetracycline resistance genes, as these were shown to dominate the slurry resistome (Figure 2A; Supplementary Figure 2A). Furthermore, these ARG categories were among CAP vectors (r > 0.4) associated with soil samples < 24 h and 56 days after the first application of slurry in 2017, respectively (Supplementary Figure 4). Kruskall-Wallis tests showed that the normalized abundance of tetracycline ARGs (χ² = 4.17, p_adjust > 0.05; = 1.24 × 10⁻¹) did not change significantly, whereas the abundance of MLS ARGs did change over the same period (χ² = 7.43, p_adjust < 0.05; = 4.88 × 10⁻²). Therefore, only MLS ARGs were considered in post hoc tests.

Pairwise Wilcoxon tests show that the relative abundance of MLS ARGs was significantly different < 24 h after the first slurry application of 2017 (p_adjust < 0.05; = 4.76 × 10⁻², Figure 3). However, pairwise comparisons of soil samples show that MLS ARGs declined to pre-application levels (pre-application soil comparison to 56 days after slurry application: p_adjust > 0.05; = 5.48 × 10⁻¹; Figure 3).

Analysis of MLS ARGs in January and May 2018 showed that the relative abundance of MLS ARGs was significantly different in the untreated soil compared to the slurry-treated soil in both January and May (ART- ANOVA, F = 10.31, p < 0.01; = 5.5 × 10⁻³). The relative abundance of MLS ARGs was also significantly different in January compared to May, regardless of treatment (ART- ANOVA, F = 10.25, p < 0.01; = 5.6 × 10⁻³), however, no significant interaction effect was observed between site and month (ART- ANOVA, F = 3.03, p > 0.05; = 1.01 × 10⁻¹), as MLS ARGs increased in relative abundance from January to May in both locations. However, these genes were more abundant in the treated site overall (Figure 3).

We detected significant shifts in the resistome and bacterial phyla immediately after the first application of slurry in the 2017 season. Although ARG compositions returned to their original states within 8 weeks of this application event, the composition of bacterial phyla remained altered for longer. Nonetheless, slurry-associated taxa and ARGs were not significantly enriched by the end of the 2017 application period, suggesting that the two subsequent slurry applications of that season did not result in any cumulative elevation of slurry-associated bacteria or ARGs. This is consistent with a previous metagenomic study which found limited evidence of persistent ARG transfer from dairy cattle manure to field soils (Sukhum et al., 2021).

The dominant ARG subtypes detected in the slurry-treated field site also dominated the untreated soil and showed comparable relative abundances, indicating repeated slurry applications had not altered the core resistome structure (Supplementary Figures 2B,C).

Nonetheless, CAP highlights that the overall beta-diversity of ARG categories and bacterial phyla of soil receiving long-term slurry amendment were distinct from the control field site (Figures 1A, 2A). In terms of ARGs, CAP vector overlay (r > 0.4) and statistical tests indicate this was primarily due to the elevation of rifamycin and mupirocin ARGs in the treated site (Figure 2A).

Rifamycin and mupirocin resistance gene categories were not prevalent in slurry samples. No increase in these ARGs in soil was shown to coincide directly with slurry application events, suggesting their enrichment does not arise from direct introduction of genetic material via slurry. Rifamycin resistance genes such as rphB and rbpA were among the most abundant ARGs in soil samples, irrespective of amendment history and are most likely encoded by indigenous soil bacteria. Rifamycin B was first described in a soil actinomycete (Verma et al., 2011), and rbpA is known to occur frequently in Actinobacteria (Dey et al., 2012). The rbpA gene was also the only ARG to be recovered from soil metagenome assembly contigs; given that the soil was not sequenced to saturation, their recovery demonstrates their inherent abundance in soil. Nucleotide megaBLAST of the longest representative contig (S121_k87_335727, 1.2 kbp) indicated they were borne chromosomally within environmental Mycobacterium spp. (closest match Mycobacterium sp. ITM-2016-00318 accession CP134400.1, 97% query cover, nucleotide identity similarity 88.1%). Considering this, the greater abundance of rifamycin ARGs in the slurry-treated field site is most likely the result of long-term application which caused shifts in the resident bacterial community, leading to the enrichment of intrinsically resistant taxa (e.g., Actinobacteria; see Figure 1A). In contrast, the most abundant mupirocin ARG in both field sites was the ileS gene intrinsic to Bifidobacteria, a genus which typically inhabits gastro-intestinal tracts of animals. This ARG was an order of magnitude less abundant in cattle slurry relative to grassland soils, which was unexpected as it was shown to associate with cattle waste rather than soil in a previous study (Kang et al., 2022). The apparently high recovery of this ARG in field soil with no history of manure amendment may instead be explained by the presence of wild mammals.

We found that MLS ARGs dominated slurry tank resistomes together with tetracycline resistance genes. MLS genes are identified routinely among the core resistome of cattle waste and associated field run-off (Hu Y. et al., 2016; Noyes et al., 2016; Gou et al., 2018; Kang et al., 2022). There are no farm records of MLS antibiotics being administered to the milking herd (Baker et al., 2022), but tulathromycin (a macrolide antibiotic) was occasionally administered to calves, whose waste does not enter the tank. The occurrence of MLS ARGs is therefore unlikely to be connected to antibiotic use. Tetracycline ARGs are also known to be abundant in cattle waste (Kang et al., 2022; Shen et al., 2024) and were among the most commonly used class of antibiotics on the farm alongside beta-lactams. However, MLS and tetracycline ARGs may be enriched in cattle during early life due to dietary influences on microbial recruitment in the gut (Liu et al., 2019). Of these two categories, only MLS ARGs were significantly increased in soil immediately after the first slurry application of 2017 and they rapidly declined toward pre-application levels.

The transient enrichment of select MLS resistance genes in cattle-waste amended soils has been documented in previous field (Fahrenfeld et al., 2014; Muurinen et al., 2017) and microcosm (Chen et al., 2019) studies. Competitive action of indigenous soil microbiota is proposed to constrain the distribution and persistence of ARGs introduced to soil. Several laboratory-scale studies appear to support this (Peng et al., 2016; Chen et al., 2017; Klümper et al., 2019; Pérez-Valera et al., 2019). In contrast, at least one qPCR-based study has shown that MLS ARGs not only increase following decades of annual cattle waste application, but also remain elevated for over a decade after manure application has ceased (Zhang et al., 2023). Furthermore, in the study by Zhang et al. (2023), manure was applied only once per year. In another qPCR study by Macedo et al. (2020), soil was collected from experimental field sites of contrasting texture before and after three successive applications of cattle manure. Macedo et al. (2020) used the data to infer that select ARGs would decline to pre-application levels within 29–42 days. This implies that while a year-long interval could lead to an enriched MLS resistome in one study, select MLS ARGs (ermB) have also been shown to decline over much shorter time-periods.

Environmental factors, both abiotic and biotic, are therefore likely to contribute to the variation observed in the literature. Solid and liquid cattle waste have distinct chemical and biological properties which can affect their influence on soil microbial communities (Neufeld et al., 2017); environmental conditions, such as exposure to solar radiation, and rainfall may dramatically reduce or promote the survival of introduced bacteria (Hodgson et al., 2016; Jang et al., 2017). The rate and frequency of application on farms can be variable and the waste material itself may have variable bacterial loads, even if their community structures remain relatively stable. Hurst et al. (2019) found general seasonal trends in cattle manure-associated ARGs and demonstrated site-specific differences. In the present work, the relative abundance of MLS ARGs declined in the slurry tank from May to October 2017 (Supplementary Figure 5), which may explain why soil collected only 2 weeks after slurry application in October showed no apparent elevation in slurry ARGs. However, we note that previous analysis of the slurry indicated that the core resistome (including MLS ARGs) did not change significantly over the course of sampling (Baker et al., 2022). Nonetheless, in samples collected 84 days after the first slurry application of 2018, a significant elevation in MLS ARGs was observed compared to pre-application abundance in January (Figure 3). If this were due to slurry application, it would suggest that changes in either the slurry (unfortunately not sampled at this time) or the prevailing conditions in the field environment were able to promote the enrichment of these ARGs over a much longer period than shown after preceding applications, indicating considerable temporal variability in resistome behavior can exist within the same farming system.

Accordingly, it could be argued that this finding merely represents seasonal shifts in macrolide ARGs due to factors unrelated to slurry application. In a watershed-scale study the abundance of ARGs in soil significantly increased in spring relative to all other seasons (Xiang et al., 2021). This is also the time at which organic fertilizer is generally applied to enhance crop growth. We note that although samples collected from the treated site during May show an increase in MLS ARGs, a significant increase was also observed in the site with no history of cattle slurry amendment. Nevertheless, specific MLS subtypes associated with slurry such as lnu(H) were detected in treated field soil during May 2018, but were undetectable in all untreated site samples collected. Although our data indicate that the survival of slurry ARGs may be extended during different times of the year, especially spring, the small untreated soil sample size and lack of concurrent slurry samples require that we are cautious in our interpretation. Further multi-year, multi-farm studies are needed to confirm this finding.

A further observation from this study is that ARG alpha-diversity exhibited greater temporal variability than bacterial taxonomic alpha-diversity, particularly in untreated soil. This decoupling between resistome and microbiome dynamics suggests that ARG composition can fluctuate independently of major shifts in community structure. One possible explanation is that many ARGs, particularly those occurring at low abundance, are subject to rapid ‘input-decay’ cycles driven by transient introduction via environmental sources and subsequent loss through degradation or dilution. In contrast, taxonomic diversity tends to be stabilized by dominant and resilient soil taxa that maintain community structure across seasons. The higher responsiveness of ARG diversity to short-term environmental changes implies that the resistome may be a more sensitive indicator of AMR risk than overall microbial diversity. For on-farm management, this finding highlights the importance of temporal monitoring: infrequent or single-time sampling may underestimate the true dynamism of ARG pools, while management practices that reduce seasonal inputs or promote conditions favoring microbial stability could help moderate resistome variability.

To our knowledge, we provide one of the few examples where MAGs have been used to demonstrate the presence of a slurry-borne organism in slurry-amended field soil. A key finding of this study was the assembly of a Proteiniphilum sp. MAG from the slurry-receiving soil. This shared high sequence homology (> 99.5% ANI) with another MAG retrieved independently from slurry metagenomes. We suggest that MAG-based analyses have unique advantages over relying solely on unassembled metagenomic reads, culturing fecal indictor organisms or 16S rRNA amplicon sequencing. Members of the identified genus are unlikely candidates for culture-based studies investigating animal waste contamination as this role is dominated by more readily cultivated fecal indicator organisms such as Escherichia coli (Lau and Ingham, 2001; Rufete et al., 2006; Anjum et al., 2021). However, analysis of our slurry metagenomes showed that reads assigned to Proteiniphilum sp. were between 15- and 41-times more prevalent than those classified as Escherichia sp., indicating that while they lack the clinical status of E. coli as a potential pathogen, they are more representative of the cattle slurry microbiome and therefore more suitable indictors of contamination. Proteiniphilum spp. are hydrolytic facultative anaerobes, first isolated from anaerobic sludge (Chen and Dong, 2005; Hahnke et al., 2016). The genus is known to be abundant in cattle waste (Pandey et al., 2018), has been isolated from cattle rumen (Kim et al., 2020), and is prevalent in mesophilic digestate produced from cattle waste (Hahnke et al., 2016; Garbini et al., 2022). Furthermore, although 16S rRNA amplicon datasets can enable inference of bacterial taxa transmitted from slurry to soil, comparing MAGs generated from slurry (source) and field soil (sink) allows more robust investigation into whether the populations of a given taxon are likely to be closely related at the genomic level.

It is important to acknowledge, however, that MAGs can often represent an aggregate of closely related genomes, rather than a discrete genome (Setubal, 2021). It is therefore important to interpret their significance with care. Application of long-read or hybrid sequencing, together with improved assembly and binning methods may minimize this issue in time. Nonetheless, we suggest the recovery of these MAGs provides sufficient evidence of bacterial transfer which goes beyond differential abundance testing based on unassembled reads. This is because these MAGs recruited many more reads in the source than the sink (Figures 6A,B) and, critically, are genetically ‘conserved’ in both. In addition, read recruitment in the control field soil remained low in 2018, meanwhile, over the same period, their relative abundance increased in the treated site following slurry application in May 2018 (Figure 6B). Additionally, we show that there is good breadth of coverage of bin 127 in treated soil immediately after slurry application and in May 2018, whereas coverage of bin 127 was poor across both January and May 2018 in untreated soil samples (Supplementary File 4). We do not believe this MAG is an artifact of the binning process since it is unlikely that such similar MAGs would have arisen independently in both soil and slurry metagenomes; environments which are clearly distinct from one another.

To investigate the introduction and persistence of this slurry-associated biomarker, we used the Proteiniphilum slurry MAG (bin 127) to estimate relative abundance over time in the slurry (Figure 6A), treated soil, and untreated soil samples (Figure 6A). We observed an increase in the abundance of the Proteiniphilum MAG immediately after the first 2017 slurry application, however the bacterial genome-normalized abundance declined to pre-application levels within 56 days. Enrichment of Proteiniphilum-associated 16S rRNA amplicons in bulk soil following swine manure-amendment has been reported (Wolters et al., 2018). Wolters et al. (2018) found that the relative abundance of Proteiniphilum returned to pre-application levels within 156 days. However, the study did not incorporate multiple fertilization events and did not collect further soil samples during the intervening period; it is therefore possible that Proteiniphilum had declined much earlier than 156 days. Likewise, Begum et al. (2024) found that a single swine slurry application produced a transient increase in Proteinipilum in soil using 16S rRNA amplicon data. We show subsequent slurry application events in 2017 did not appear to coincide with further increases in the abundance of the slurry-associated Proteiniphilum MAG. Despite successive slurry applications, the relative abundance of bin 127 declined and then remained relatively stable in soil throughout the remainder of the year (Figure 6B). At the beginning and end of the closed period, the estimated abundance closely resembled the distribution in soil prior to the first application of the season.

A possible explanation for the absence of a progressive increase in Proteiniphilum with repeated slurry applications could be that input declined. Proteiniphilum abundance showed a contemporaneous decline in the slurry tank (Figure 6A). Although previously reported analysis of the slurry tank taxonomy shows the stability of dominant phyla over time (Baker et al., 2022), this does not extend to lower taxonomic ranks. This suggests, that even though the slurry tank continues to receive fresh material throughout the year, environmental factors and farm practice may influence the relative abundance of slurry tank bacterial taxa. Alternatively, environmental factors affecting the field soil may be responsible for the declining persistence of slurry borne Proteiniphilum. For example, Sun et al. (2022) demonstrated that the prevalence of Proteiniphilum was positively associated with organic acids which increased following cattle amendments applied to field soil in spring and early summer but not late summer. We show that read recruitment of the Proteiniphilum MAG increased from January to May 2018 in the slurry-treated site, while it remained stable over the same period in the untreated site (Figure 3B). Given the similar composition and location of the two field sites, the divergent recruitment of reads suggests that the addition of slurry rather than the local environment was responsible for the observed increase in the slurry-treated site.

Therefore, while both Proteiniphilum and macrolide-lincosamide-streptogramin (MLS) resistance genes were associated with the slurry source, and exhibited transient enrichment following slurry application, their temporal dynamics were not linked directly. The relative abundance of the Proteiniphilum MAG increased immediately after slurry application and returned to baseline within approximately 8 weeks, whereas MLS ARGs displayed enrichment patterns that varied seasonally and were sometimes evident even when Proteiniphilum abundance had already declined. Furthermore, no ARGs were detected within the Proteiniphilum MAG itself, suggesting that the transient increase in this bacterium was not a direct vector for MLS gene dissemination. These observations imply that bacterial transfer and ARG transfer can occur as partly independent processes: while slurry-borne bacteria such as Proteiniphilum may transiently establish in soil, ARG enrichment may also result from horizontal gene transfer among native soil taxa or from selective pressures acting on the existing resistome. This distinction highlights the complexity of slurry-soil interactions and underscores that changes in ARG abundance cannot be assumed to mirror the persistence of specific slurry-borne organisms.

Although only a limited number of high-quality metagenome-assembled genomes (MAGs) were recovered, these genomes provide valuable insight into the fate of specific slurry-borne microorganisms within the soil environment. The Proteiniphilum MAG, for example, illustrates the transient establishment and subsequent decline of a slurry-associated bacterium following application. However, these MAG-based observations are used here as representative case studies rather than as the sole basis for community-wide interpretation. Broader inferences about soil microbial and resistome dynamics are supported by compositional and diversity analyses derived from the complete metagenomic dataset. Together, these complementary lines of evidence provide confidence that the observed temporal changes reflect true community-level responses, even though individual genome reconstructions remain limited.

It is important to note that the slurry-treated and untreated field sites differed slightly in soil texture (heavy clay loam versus silty clay loam). Variations in clay and silt content can influence soil moisture retention, aeration, and organic matter dynamics, thereby shaping microbial community composition and the persistence of extracellular DNA or antibiotic resistance genes (e.g., Carson et al., 2010). Such physicochemical differences may therefore contribute to some of the baseline variation observed between the two field sites. However, both fields are geographically proximate, share the same climatic conditions, and have been maintained as permanent grassland for many years. Moreover, the key temporal patterns observed in this study (such as the transient enrichment of macrolide-lincosamide-streptogramin (MLS) resistance genes following slurry application and the short-term detection of a slurry-associated Proteiniphilum MAG) were closely linked to the periodic timing of slurry inputs rather than static site characteristics. These findings suggest that while soil texture may partly modulate microbial and resistome profiles, it is unlikely to account for the principal treatment-related effects observed here.