OBJECTIVES: The long-term goal of this project is to determine if fertilization of agricultural soil with manure promotes the establishment of resistance plasmids in the soil microbiome to form a resistance gene reservoir that can persist in the environment.The specific objectives are:1) Determine and improve the limits of the Hi-C method to detect horizontal transfer of antibiotic resistance plasmids in agricultural soils.2) Determine the effect of manure treatment on the spread of a multi-drug resistance plasmid in the rhizosphere.3) Determine the long-term effect of dairy manure fertilization on the bacterial reservoirs of antimicrobial resistance genes and plasmids in an agricultural soil.
APPROACH: i) Hi-C, a new approach to track and monitor the fate and transport of resistance plasmids The Hi-C method allows studying chromosome in a cell's natural state by cross-linking DNA molecules in close physical proximity. It reflects the spatial arrangement of DNA within cells at the time of cross-linking. In communities of prokaryotes, the physical contacts provided by Hi-C provide quantitative and objective information on whether two pieces of DNA are shared by the same cell, without requiring any prior knowledge of the content of the sample.When applied to a bacterial community we can identify fragments of DNA that are in part derived from a resistance gene or a plasmid and a bacterial chromosome. The chromosomal fragments can identify hosts by means of parallel shotgun sequencing of metagenomic DNA. This allows clustering the chromosomal and plasmid fragments based on which genome they belong to.Here, we will apply this method for the first time to determine if resistance plasmids and their antibiotic resistance genes (ARG) inoculated in soil with manure are spreading to other bacteria, as detected by new contacts between the plasmid and indigenous bacteria.ii) Assessing the limits of the Hi-C approach in detecting transfer of a known resistance plasmid in soil We will determine the potential and limits of the Hi-C technology as a novel method to monitor plasmid transfer in soil. This will be determined by assessing the lowest amount of a self-transmissible plasmid (pB10) which would have transferred from a known donor bacterium (D) to a different known recipient bacterium (R), resulting in new transconjugants (T) (Table 1 in proposal). We will apply the Hi-C method to soil samples that contain serially diluted known amounts of T in the presence of D and R at a constant population size. By comparing the success of Hi-C for each scenario we will be able to evaluate its detection limit and other limitations in real environments such as soil.iii) Target capture to enrich plasmid-containing Hi-C reads: The Hi-C+ methodTo improve the sensitivity of the Hi-C method we will couple the Hi-C method with a plasmid-specific capture approach to enrich the Hi-C libraries with DNA fragments containing plasmid-specific contacts. We will use a set of probes derived from the sequence of plasmid pB10 to enrich Hi-C libraries with fragments that contain pB10. This method consists of an in-solution hybridization capture of target DNA using biotinylated custom RNA baits that are complementary to the target sequences (Fig. 8 in proposal). The sensitivity of the Hi-C+ method will be assessed on various scenarios where the plasmid would have transferred at low frequency (Table 1 in proposal). As above, by comparing the success of Hi-C+ for each scenario we will be able to evaluate its detection limit and other limitations in real environments such as soil.iv) Mathematical model to predict plasmid transfer frequencies using Hi-C+To assess the degree to which a plasmid is transmitted to the soil bacterial community, we will create a quantitative model using Hi-C and Hi-C+ data obtained in i) and ii). We will use multiple modeling strategies to find the best method to estimate the frequency of transconjugants for a given plasmid-host pair.First, generalized linear modeling techniques will be tested. These models can accommodate simple non-linearity in the data and various error structures (e.g., Poisson distributed errors). Models that use Bayesian techniques will also be also assessed. These techniques can model bias that might occur due the method and control for it, and latent (unmeasured) variables can also be added, and should allow us to leverage known quantitatively-problematic issues with the method in the estimation process. More complex models will include a dynamic modeling component in combination with Bayesian inference. Evaluation and comparison of these models and their fit to the data will tell us which model to use for Objective 2.v) Experimental design of Objective 2.The experimental setup of our proposed laboratory-scale microcosm experiment is depicted in Figure 9 of the proposal. Briefly, a set of pots (n=100) of soil will be amended with the dairy-cow manure or inorganic fertilizer at agricultural rates matching Objective 3. A donor bacteria (E. coli K12?gfp) containing the plasmid pB10 will be added in a realistic amount for each treatment, while some pots will received the plasmid-free strain to control for the possible occurrence of natural plasmids with similar DNA fragments to pB10. Two sterile barley seeds will be planted 30 days after soil fertilization to be in line with the USDA recommendations for both organic and GAP certification. The containers will be incubated in a growth chamber using standard protocol for plant growth. Sampling of the rhizosphere of three plants at a time will be approximately done every 5-10 days up to the plant is mature.We will first assess the transfer of the plasmid using a qPCR approach targeting both the donor and plasmid genomes. If plasmid/donor ratios gradually increase, it suggests plasmid transfer, and thus we will use the Hi-C+ method. Hi-C+ paired-end reads matching pB10 on one end and a DNA sequence different from E. coli K12::gfp on the other end will be filtered, and will be evidence of plasmid transferred to a new host. The quantitative model built in Objective 1 will be used to estimate the relative frequency of plasmid transfer in soils in between treatments.To evaluate the method, the results of this relative quantification will be compared with the qPCR data, and if needed models will be further optimized as described in Objective 1.We will determine if the new hosts of pB10 came from manure or soil through direct shotgun sequencing of the DNA from the original manure and soil samples, and aligning all the Hi-C reads with putative transconjugant DNA against these two metagenome assemblies (read assemblies will be done using metaSPAdes).We will try to identify the new pB10 hosts by direct shotgun sequencing of DNA for five samples with the highest level of plasmid transfer. Identification of the host DNA fragments from the Hi-C+ reads, the metagenome assemblies, and the putative bacterial genome clusters will be carried out by aligning the DNA sequences against curated databases grouping all the current known microbial genomes such as OneCodex. The identification of the clusters will be run by ProxiMeta™ in close collaboration with Phase Genomics who owns the proprietary software. With the help of Phase Genomics we will further evaluate and improve the software's ability to identify clusters at the genus or species level.vi) Experimental design of Objective 3. We will use the Hi-C+ to characterize the reservoirs of ARG and mobile genetic elements in the USDA research fields in Kimberly, ID (Fig. 10 of the proposal). The approach will be as described in the Objective 1 method, with two main differences:The target capture will be designed to target all known genetic determinants involved in antimicrobial resistance using nucleic acid baits designed based on the recently published reference available database ResCap. This database compiled all both well-known and hypothetical genes encoding resistance to antimicrobials and coding for plasmid family markers. It includes 7,963 ARG, 47,806 putative ARG, 704 metal and biocides resistance genes, 30,794 putative metal and biocide resistance genes, and 2,517 relaxases of the ConjDB database.We will include shotgun libraries in an attempt to identify the most abundant reservoirs of these ARG and mobile genetic elements.Using the same approach at the end of Objective 2 and described in Fig. 5 of the proposal, metagenomic shotgun sequencing of the DNA from each sample will be carried out in parallel of the HiC+. This should allow reconstruction of host genomes using the software ProxiMetaTM.