Soil samples were collected from four different locations previously utilized for ametoctradin environmental fate testing and regularly used for agricultural studies: California (CA), New Jersey (NJ), and Germany (LUFA 2.2 and LUFA 2.3) (Table 1). Roughly 1 kg of each soil sample was placed in two replicate containers linked to a CO₂-scrubbed air source. Each sample was then either treated with 2.4 μg mL^–1 of ametoctradin or remained untreated. After 14 days, the soil samples were harvested and analyzed in triplicate using HPLC-MS/MS. In each case, degradation of ametoctradin was confirmed along with the identification of four major metabolites: M650F01, M650F02, M650F03, and M650F04 (Figure 1). These compounds appeared to be the product of degradation by soil microorganisms capable of metabolizing the long aliphatic chain on the ametoctradin molecule while the amine ring structures remain largely unaltered. The NJ soil sample displayed the highest and most complete level of ametoctradin degradation, with only 13.6% of the original amount of agrochemical remaining after the 14-day period. The M650F03 metabolite appeared to be the main degradation product in every soil type.

Microbiome analysis of treated and untreated soils by 16S (Figure 2 and Supplementary Table S1) and 16S/18S (Supplementary Figure S2 and Supplementary Table S2) rRNA amplicon sequencing showed that Proteobacteria were the dominant bacterial phylum present in both the treated and untreated soils, representing up to 54% of the reads in untreated soils and 78% of the reads in treated soils. The Acidobacteria were also commonly identified, comprising up to 39% of the reads in untreated soils and 28% in treated soils. Within Eukaryotes, ametoctradin treatment only revealed significant reduction in the Oomycetes, particularly Phytophthora infestans, which are the target organisms for ametoctradin (Supplementary Table S2).

Species richness varied between the four soil samples, while the most diverse soils were NJ and LUFA 2.3, with Shannon-Wiener Index (SWI) values around 8. The CA and LUFA 2.2 soils had SWI values of 7.75 and 7.3, respectively (Table 1). All soils showed a significant decrease in diversity when treated with ametoctradin (CA, NJ, LUFA 2.2: p < 0.0001, LUFA 2.3 p = 0.053). Moreover, as indicated in Supplementary Table S1, several OTUs increased in relative abundance after treatment. Those OTUs that increased in relative abundance include known degraders of xenobiotic compounds such as Mycobacterium, Rhodococcus, Comamonas, Burkholderia, Variovorax, Methylibium, and Pseudomonas. In fact, many of these genera were below the limit of detection in untreated soils. This increase in relative abundance after soil treatment suggested the presence of resistant organisms and may be connected to microbially mediated degradation (Berg et al., 2012; Jia et al., 2015).

Soil microbial community similarities and differences have been highlighted by beta diversity analysis (Figure 6A). Although the number of samples did not allow for statistical analysis, the data showed that untreated soils tended to cluster together while treated soil sample communities tended to diverge from each other, suggesting that ametoctradin treatment on soil microbial communities varied according to the location of the soil.

Analysis of ex vivo systems showed that M650F01, M650F02, M650F03, and M650F04 were the main products of degradation of BASF650F and identified several species including Mycobacterium, Rhodococcus, Comamonas, Burkholderia, Variovorax, Methylibium, and Pseudomonas as relevant participants in the degradation process. We next implemented an in vitro continuous up-flow PBBR system to culture the complex microbial communities present in the soils in an attempt to confirm and reproduce data from ex vivo systems. The PBBR system consisted of a water-jacketed glass column packed with expanded glass beads. Mixing was achieved using a recirculation loop and a peristaltic pump operated at a recycling ratio of 50:1. ametoctradin was added to the medium at a maximum soluble concentration of 148 μg L^–1. Finally, continuous addition of fresh medium containing ametoctradin was performed at five dilution rates over the course of the experiment (see section “Materials and Methods”).

Since NJ soil samples contained the highest number of taxa and the highest diversity values, the ametoctradin-treated NJ soil was chosen to inoculate the PBBR. After inoculation, the reactor was maintained under continuous culture conditions for 1 month to permit enrichment and selection of taxa. After this maturation period, the reactor was subjected to five different dilution rates (0.03, 0.05, 0.06, 0.07, and 0.08 h^–1) over the course of the experiment. Sequencing confirmed that no eukaryotic taxa remained in the reactor after maturation, indicating that any ametoctradin degradation observed in the reactor was entirely due to prokaryotic organisms.

Once the culture reached steady-state conditions, liquid samples were taken at retention time intervals and residual levels of ametoctradin, and its degradation products, were monitored by LC-MS/MS. The community structure was also analyzed at those times by 16S rRNA amplicon sequencing. Following this period of maturation, the reactor was sequentially exposed to five sequentially increasing dilution rates (0.03, 0.05, 0.06, 0.07, and 0.08 h^–1) of ametoctradin-containing medium over the course of the experiment. Sequencing of rRNA amplicons from samples obtained at each dilution rate confirmed the washout and absence of eukaryotic OTUs in the reactor after maturation. Nearly complete degradation of ametoctradin was observed in the reactor due to prokaryotic activity. At the lower dilution rates (0.03, 0.05, and 0.06 h^–1), residual levels of ametoctradin were low (4.3 ± 0.89 μg L^–1). When the dilution rate was increased to 0.07 and 0.08 h^–1, the residual levels of ametoctradin also increased, reaching a maximum concentration of 15.2 μg L^–1 (Figure 3). These variations in residual levels of ametoctradin matched shifts in the reactor microbial community, with the Burkholderiales, particularly Comamonas, dominating the community at the lower dilution rates (0.03, 0.04, and 0.05 h^–1). A shift towards members of the Clostridiales, specifically Clostridium, was observed at the higher dilution rates (0.06 h^–1 and above) (Figure 3 and Supplementary Table S3). These results indicated that Burkholderiales may be key to the ametoctradin biodegradation process.

To confirm that the low residual levels of ametoctradin found in the reactor were due to microbial degradation, we performed a pulse experiment in which we quantified the levels of ametoctradin and its metabolites using LC-MS/MS at various sampling points over a period of 83 h. This data was then compared against a model for theoretical washout in which the compound would slowly be removed from the system at predictable intervals based on the dilution rate. It is important to note that the pulse of ametoctradin into the reactor was well above its limit of solubility in water, and the rapid decrease in ametoctradin levels in the first 8 h is an artifact of this limitation (Figure 4). However, increased levels of metabolites M650F01, M650F03, and M650F04 were detected in the samples, indicating active biodegradation. Additionally, the levels of ametoctradin remained relatively stable around 0.148 μg mL^–1 (limit of solubility in water) for a period of 42 h, suggesting that the additional insoluble portion from the pulse was being solubilized as degradation of the parent compound was occurring.

Microbiome analysis of the soil-derived community showed that abundance of members of the Betaproteobacteria continued to increase up to a dilution rate of 0.06 h^–1 after which they begin to wash out of the system (Figure 5 and Supplementary Table S3). A Comamonas sp. was particularly abundant at these dilution rates, where it comprised nearly 60% of the reads. At higher dilution rates (0.07 and 0.08 h^–1), it is apparent that members of the Firmicutes, particularly Clostridium spp., begin to dominate the microbial community in the bioreactor. Beta diversity analysis of the soil-derived communities sampled at different dilution rates showed clear differences between soil and soil-derived populations, with treated and untreated soil samples clustering separately from the PBBR samples. Likewise, samples collected from low dilution rate conditions were grouped separately from high dilution rate condition samples illustrating the dynamics of the microbial community (Figure 6). Community shifts were likely due to the changes in relative abundance in Betaproteobacteria and Clostridia detailed above. We also observed that the disappearance of Acidobacteria, a highly abundant phylum in the soil samples, in the PBBR did not hinder the ability of the consortium to degrade ametoctradin, supporting our hypothesis that Burkholderiales were the major drivers of ametoctradin degradation.

Using the predictive software PICRUSt, which permits the determination of metabolic potential from 16S rRNA amplicon sequencing data, we analyzed the KEGG gene pathways associated with degradation of xenobiotic compounds in ex vivo and in vitro systems (Figure 7). The only pathway that showed a significant difference in relative abundance between treatments in all three soils was caprolactam degradation. This pathway contains several enzymes that may participate in the oxidation of the aliphatic side chain present on ametoctradin, including alkane 1-monoxygenase (EC 1.14.15.3), NADP-dependent aldehyde dehydrogenase (EC 1.2.1.4), and NADP-alcohol dehydrogenase (EC 1.1.1.2). The LUFA 2.2 soil showed the largest number of changes in metabolic potential between non-treated and treated samples, with 15 out of 18 pathways examined showing significantly higher gene counts in the treated soil. Specifically, gene predictions related to bisphenol, chloroalkane, fluorobenzene, naphthalene, and styrene degradation only showed significant increases in the LUFA 2.2 soil. Several of these pathways include decyclizing enzymes such as catechol dioxygenase (ECs 1.13.11.1 and 1.13.11.2) which might participate in the degradation of the ring structure in ametoctradin.

In vitro analysis of changes in metabolic potential showed that the dilution rates impacted a number of genes for each category that were directly correlated with changes in the relative abundance of Betaproteobacteria, specifically a Comamonas species (Figure 8). Upon examination of the Comamonas genomes in the KEGG database, we found that all five annotated genomes contained decyclizing dioxygenases such as catechol dioxygenase (ECs 1.13.11.1 and 1.13.11.2) and protocatechuate dioxygenase (EC 1.13.11.8) (Kanehisa and Goto, 2000). This abundance of ring-opening enzymes may help to explain the low amounts of ametoctradin and its metabolites in the reactor effluent.

Two soil samples from the United States and two samples from Germany were collected for ex vivo degradation analysis (Table 1). Each of the four soil samples were treated with 2.4 mg kg^–1 soil of ametoctradin and allowed to react for a total of 14 days. After this incubation period, the soils were analyzed following established methods for residual ametoctradin and its associated metabolites (M650F01, M650F02, M650F03, M650F04) using HPLC-UV and HPLC-MS/MS, respectively (European Food Safety Authority, 2012).

Amplification of the V1–V2 and V3–V4 regions of the bacterial 16S rRNA gene were performed on total DNA from soil samples. Master mixes for these reactions used the Qiagen HotStar Hi-Fidelity Polymerase Kit (Qiagen, Valencia, CA, United States) with a forward primer composed of the Roche Titanium Fusion Primer A (5′-CCATCTCATCCCTGCG TGTCTCCGACTCAG-3′), a 10 bp Multiplex Identifier (MID) sequence (Roche, Indianapolis, IN, United States) unique to each of the samples. The gene-specific part of the primer used the universal bacterial primer 8F (5′-AGAGTTTGATCCTGGC TCAG-3′) to amplify V1–V2 region or 515 F bacterial primer (5′-GTGCCAGCMGCCGCGGTAA-3′) to amplify V3–V4 region (Edwards et al., 1989; Fierer et al., 2008; Caporaso et al., 2011; Walters et al., 2016). The reverse primer was composed of the Roche Titanium Primer B (5′-CCTATCCCCTGT GTGCCTTGGCAGTCTCAG-3′), the identical 10 bp MID sequence as the forward primer and the reverse bacterial primer 338R (5′-GCTGCCTCCCGTAGGAGT-3′), which span the V1–V2 hyper variable region of the bacterial 16S or 806R bacterial primer (5′-GGACTACHVGGGTWTCTAAT-3′) for the amplification of V3–V4 region. The thermal profile for the amplification of both V1–V2 and V3–V4 regions had an initial denaturing step at 94°C for 5 min, followed by a cycling of denaturing of 94°C for 45 s, annealing at 50°C for 30 s and a 1 min 30 s extension at 72°C (35 cycles), a 10 min extension at 72°C and a final hold at 4°C. Each sample was individually gel purified using the Qiaquick Gel Extraction Kit (Qiagen, Valencia, CA, United States) and E-Gel Electrophoresis System (Life Technologies, Invitrogen division), and standardized prior to pooling. Pooled barcoded libraries were sequenced on a Roche 454 Genome Sequencer FLX Titanium instrument (Microbiome Core Facility, Chapel Hill, NC, United States) using the GS FLX Titanium XLR70 sequencing reagents and protocols. Initial data analysis, base pair calling and trimming of each sequence to yield high quality reads were performed by Research Computing at the University of North Carolina at Chapel Hill (Chapel Hill, NC, United States). Sequence data were cleaned and analyzed using Version 1.9 of QIIME software (Caporaso et al., 2010). Sequences are available on the NCBI GenBank under SRA accession number PRJNA625701.

Redundant sequences were removed from the set of representative sequences using the PhyloToAST (Dabdoub et al., 2016) plugin for QIIME. The trimmed set of representative sequences was aligned using the MUSCLE algorithm in Geneious 5.5.9 (Biomatters Inc., Newark, NJ, United States). Maximum likelihood phylogenetic trees were generated using the PHYML plugin for Geneious. Sequence ID numbers on leaves were replaced with taxon information using PhyloToAST (Dabdoub et al., 2016). Trees were annotated with relative abundance information using the Interactive Tree of Life (iToL) software (Letunic and Bork, 2011). Relative abundance data were generated by QIIME and normalized using the cumulative sum scaling (CSS) method and the normalize_table.py command in QIIME for each project (Paulson et al., 2013). Mean relative abundance was calculated from the mean of the triplicate data points for each sample and repeated for each of the different treatments using PhyloToAST.

In order to obtain a rough metagenome for each soil treatment and bioreactor condition, we ran PICRUSt analysis (Langille et al., 2013). The input file used in PICRUSt was a closed-reference OTU table generated for both the soil and bioreactor experiments in QIIME. Predicted metagenomes were statistically analyzed in STAMP (Parks et al., 2014).

The bioreactor enrichment experiment was performed in a packed-bed reactor consisting of a single water-jacketed glass column (6 cm diameter by 13 cm height) filled with porous foam glass particles having a mean diameter of 4–8 mm with a porosity of 55%, and a density of ∼0.2 g cm^–3 (A generous gift from Poraver GmbH). The column had an external loop that consisted of an input for the growth medium, a peristaltic pump controlling the dilution rate, a port from which to take samples, and a harvest tank and associated peristaltic pump. A schematic of the PBBR configuration can be found in the Supplementary Figure S3. The reactor was fed with medium containing 148 μg L^–1 ametoctradin (maximum solubility in water); a full description of medium composition can be found in the Cultivation Medium section “Materials and Methods” below. This medium was recycled within the reactor at ratio of 50:1. The final active volume for the duration of the experiment was 220 mL. The temperature during the experiment was controlled via water jacket at 24°C. The pH was monitored but not controlled and varied between 4.8 and 6.7. Inoculum for the experiment consisted of 1 g of ametoctradin treated NJ soil. After inoculation, the in vitro microbial community was allowed to mature and stabilize for 1 month at a dilution rate of 0.03 h^–1 before adjustments were made to the dilution rate. Over the next 3 weeks the reactor’s microbial community was subjected to five different sequential dilution rates (0.03, 0.05, 0.06, 0.07, and 0.08 h^–1). For each dilution rate, performed for at least a three retention times (RT) (period of adaptation), three samples were taken and stored for sequencing and HPLC-UV or HPLC-MS/MS analysis. The samples represent the mature microbial community that has been influenced by the presence of ametoctradin.

To determine if residual levels of ametoctradin and products of its degradation in the bioreactor were microbially mediated degradation and not washout, an experiment injecting 0.41 μg mL^–1 ametoctradin was performed while maintaining the reactor at a dilution rate of 0.03 h^–1. Samples were taken at 0.2, 6.5, 24, 48, 54, 73, and 83 h to monitor levels of ametoctradin and its metabolites. Actual data analyzed by HPLC-UV or HPLC-MS/MS were compared to theoretical values obtained from a mathematical model for washout at a given dilution rate. The formula for theoretical substrate concentration at time t [(Y)_t] is as follows:

where C₀ and V₀ represent the initial pulse concentration and volume, V_R is the total reactor volume, and D is the reactor dilution rate.

The medium used during these experiments contained (g L^–1) yeast extract 1, ammonium acetate 2, carboxymethyl cellulose 2, dextrose 1, cellobiose 0.1, xylose 0.771, K₂HPO₄ 0.8, KH₂PO₄ 0.2, CaCl₂⋅H₂O 0.09, FeSO₄⋅7H₂O 0.0028, MgSO₄⋅7H₂O 0.2, MnSO₄⋅H₂O 0.003, NaCl 0.1, NaMoO₄⋅2H₂O 0.00025, ZnSO₄⋅7H₂O 0.00025, Biotin 0.0001, Pyridoxol⋅HCl 0.0002, and 0.01 mM EDTA.