Plasmidome AMR screening (PAMRS) workflow: a rapid screening workflow for phenotypic characterization of antibiotic resistance in plasmidomes [version 1; peer review: awaiting peer review]

Background: Phenotypic characterization of antimicrobial resistance (AMR) in bacteria has remained the gold standard for investigation and monitoring of what resistance is present in an organism. However, the process is laborious and not attractive for screening multiple plasmids from a microbial community (plasmidomes). Instead, genomic tools are used, but a major bottle neck that presence of genes does not always translate into phenotypes. Methods: We designed the plasmidome AMR screening (PAMRS) workflow to investigate the presence of antibiotic resistant phenotypes in a plasmidome using Escherichia coli as a host organism. Plasmidomes were extracted from the faecal matter of chicken, cattle and humans using commercial plasmid extraction kits. Competent E. coli cells were transformed and evaluated using disk diffusion. Thirteen antibiotic resistant phenotypes were screened. Results: Here, we show that multiple antibiotic resistant phenotypes encoded by plasmids can be rapidly screened simultaneously using the PAMRS workflow. E. coli was able to pick up to 7, 5 or 8 resistant phenotypes from a single plasmidome from chicken, cattle or humans, respectively. Resistance to ceftazidime was the most frequently picked up phenotype in humans (52.6%) and cattle (90.5%), whereas in chickens, the most picked up resistant phenotype was resistance to co-trimoxazole, ceftriaxone and ampicillin (18.4% each). Conclusions: This workflow is a novel tool that could facilitate studies to evaluate the occurrence and expression of plasmid-encoded antibiotic resistance in microbial communities and their associated plasmid-host ranges. It could find application in the screening of Open Peer Review Reviewer Status AWAITING PEER REVIEW Any reports and responses or comments on the article can be found at the end of the article. AAS Open Research Page 1 of 11 AAS Open Research 2021, 4:18 Last updated: 26 APR 2021


Introduction
Antibiotic-resistant bacterial infections are a major public health problem. Annually, some estimated 30,000 deaths are recorded in the EU alone 1 , while the morbidity and mortality in low to middle income countries in South America, Asia and Africa are even higher 2-4 . Generally, bacteria develop mechanisms to either prevent antibiotics from reaching toxics levels inside the cell, modify the targets of the antibiotic or degrade the antibiotic itself 5 . These drug evasion mechanisms are normally developed either through gene mutations or acquisition of antibiotic resistance genes (ARGs) from other bacteria through horizontal gene transfer (HGT), a process that contributes significantly to the spread of ARGs 6 . While HGT can occur in any environment, it is more common in niches within the soil 7 , wastewater treatment plants 8,9 , and the gut microbiome of humans and animals [10][11][12] .
Transformation of bacteria with naked DNA from the extracellular environments is a major contributor to HGT and the spread of ARGs 13 . Interestingly, over 80 bacterial species can be transformed due to the presence of genes involved in DNA uptake, suggesting that, this trait of competency is widespread 14 . While it is not clear what stimuli enhance the transformation of bacteria, the absence of nutrients and the presence of competence-inducing peptides have been identified as triggers 15 . Although DNAse degrades most gut DNA 16 , intact plasmids have been isolated from the gut contents of plasmid-fed rodents, suggesting that environmental or extracellular DNA are taken up by naturally occurring competent bacteria 13 . Data suggest that E. coli can be transformed with extracellular plasmid DNA under natural conditions 17,18 , and these observations suggest that bacteria could be transformed by DNA in the gut and may contribute to the spread of ARGs. Some bacterial plasmids have been shown to carry mobile genetic elements that can confer resistance to a variety of antimicrobials, including last line [19][20][21][22][23] . It has been found that large conjugative resistance plasmids follow the same evolutionary trajectories as their non-conjugative mini-replicons in the same and other species 19 . In addition plasmids harbouring multiple antimicrobial-resistance determinants (R-plasmids) can be transferred in simulated natural microenvironments from various bacterial pathogens of human, animal, or fish origin to susceptible strains isolated from a different ecological niche 24 .
Traditionally, detection of phenotypes of antibiotic resistance has been performed through culture-based techniques. To understand the role of plasmids in spreading antibiotic resistance, efforts have centred around polymerase chain reaction (PCR) amplification of targets or sequencing. Although sequencing and PCR approaches provide a high resolution of AMR investigation, they are limited by their inability to determine which genetic elements are expressed and which ones are not, hence the need for phenotypic tools to complement such metagenomic or plasmidome sequencing approaches. To determine the burden of active or expressed resistance genes in human, chicken, and cattle gut microbiomes, we developed the Plasmidome AMR screening (PAMRS) workflow using a combination of plasmidome isolation, bacterial transformation, and multi-disk diffusion to rapidly investigate the ecology of thirteen (13) antibiotic resistant phenotypes from gut plasmidomes using Escherichia coli as a host.

Development of the workflow
The workflow integrates the well-established protocols for plasmid extraction, bacteria transformation and antimicrobial susceptibility testing by disk diffusion. We adopted E. coli JM109 as host for our experiments. The BioBrick plasmids pSB1C3, pSB1A3, pSB1K3 and pSB1T3 containing the red fluorescent protein (J04450) expression cassette and encoded by resistance genes for chloramphenicol, ampicillin, kanamycin, and tetracycline respectively (http://parts.igem.org/Catalog) were used as plasmids to evaluate the workflow. The plasmid backbones for all the four plasmids are similar with the differences being the type of antibiotic resistance carried ( Figure 1). The replication origin is a pUC19-derived pMB31 which is permissive in E. coli and some other bacteria.

Transformation of cells
Chemically competent E. coli JM109 cells were made using the TSS Buffer protocol as described previously 25 with modifications. Briefly, 2µL of plasmid was added to 100 µL of competent cells and incubated on ice for 45 minutes. The cells were heat shocked at 42°C for 90 seconds and the tubes were returned to ice for 90 seconds. Following incubation on ice, 0.9mL Luria broth was added and the cells were conditioned at 37 °C for 1 hour. Recombinant clones were identified by inoculating 100 µL of the conditioned cells on Luria Agar incorporated with appropriate antibiotics (i.e. 40 µg/µl chloramphenicol; carbenicillin 80 µg/µl; kanamycin 40 µg/µl and tetracycline 5 µg/µl) (Merck KGaA, St. Louis, USA). As an alternative to incorporation of antibiotics into the agar, antibiotic disks with the respective antibiotics were evaluated. To do this, the disks were placed onto the agar plates without antibiotics incorporated into them. The inoculated plates with either antibiotics incorporated within, or antibiotic disks (Mast Group Ltd, Bootle, UK) placed on the surface was incubated at 37°C overnight. Recovery of recombinant clones was compared between plates with antibiotics incorporated and those with antibiotic disks placed on the surface. Following experiments with individual plasmids, combinations of all four were made and evaluated as described above. First, equal concentrations of plasmids (1:1:1:1) was evaluated and then varying ratios (1:3:5:10) for the four antibiotics (interchangeably) was also evaluated.
Evaluation of the workflow with gut plasmidomes from humans, cattle, and chickens We conducted a cross-sectional evaluation of gut plasmidomes from humans, cattle and chickens within the Ho Municipal area in the Volta Region, Ghana. The chicken specimens included broilers, old layers, cockerels, and local free-range hens and roosters purchased from markets, and street and community vendors. The chickens were obtained alive and sacrificed by slicing the jugular veins with a sharp knife. Thirty-eight (38) chickens were used. Faecal material was harvested from the caeca of the chicken. Faecal material was also sampled from the rectum of slaughtered cattle at the Ho abattoir. Twenty one (21) cattle specimen were obtained. We also collected the faecal material from volunteers in 15 randomly selected households within the Ho municipal area. A household was defined as one or several persons who live in the same dwelling and share meals. The households were not selected in a particular format. The study team knocked on doors to introduce the study and those that consented were included. Volunteers were handed sterile stool containers to submit their stool samples, after written inform consent had been obtained. The faecal samples were kept at 4°C and processed within 24 hours.

Plasmidome extraction and transformation
A loopful of each faecal sample was picked using a sterile 10 µL microbiological loop and inoculated into 5 mL of brain heart infusion broth (BHIB) (Oxoid, UK). The broth was cultured at 37°C with shaking at 200 rpm overnight. Plasmidome extraction was performed using the QIAprep Spin Miniprep Kit (Qiagen GmbH, Germany) following the manufacturer's protocol. Briefly, the cells were harvested by centrifuging 1.8 mL of culture in a 2 mL microcentrifuge tube at 7,000xg for 15 minutes at 4°C. The supernatant was discarded, and the pellet was resuspended and lysed. Plasmid DNA (pDNA) was purified by binding to the spin columns. The columns were washed and the pDNA was eluted with 50 µL buffer EB. Competent E. coli cells were transformed with the extracted plasmidomes as described above and screened.
Escherichia coli JM109 was used as host for screening of the plasmidomes. E. coli was chosen because it is an important member of the gut microbiome as well as the antibiotic resistance story and often used an indicator or proxy organism. Chemically competent E. coli JM109 cells were prepared using TSS buffer (pH 6.5) and then transformed 25 with the extracted plasmidome. Briefly, 2 µL of extracted plasmidome was added to 100 µL of competent cells and incubated on ice for 45 minutes. Although the original publication doesn't recommend a heat shock, we included a heat shot based on our observation that helps improve transformation efficiency in a number of cases. Heat shock was performed at 42°C for 90 seconds and then the tubes were returned to ice for 90 seconds before adding 0.9 mL Luria broth. Cells were conditioned by incubating at 37°C for 1 hour.

Antibiotic screening
To screen for susceptibility to multiple antibiotics, we employed the disk diffusion method. Mueller Hinton Agar 2 (MHA) (Merck KGaA, Germany) was prepared and poured according to the manufacturer's instructions and allowed to dry. Plates were then inoculated with 100 µL of the transformed cells using sterile glass balls. The plates were left to dry in a laminar flow cabinet after which up to six antibiotic disks (Mast Group, Bootle, UK; Table 1) were applied using a manual disk stamper. The plates were incubated at 37°C, overnight. About three colonies growing in the zone of clearing was picked and subcultured on a fresh MHA plate, overnight. A single colony from the overnight plate was picked and checked to confirm

Results
Evaluation of the use of antibiotic disks to screen transformants Traditionally, transformants are screened on plates by incorporating the antibiotic into the agar and poured (Figure 2, A1 to A4). However, to enable rapid and cheaper screening of multiple resistant phenotypes we designed and evaluated experiments with antibiotic disks (Figure 2, C1 to C4). Using antibiotic disks was found to be suitable for selecting clones just like incorporating antibiotics into the agar (Figure 2, Rows A and C). When the four plasmids were combined, almost the same number of colonies were recovered irrespective of the antibiotic that was incorporated into the medium for selection (Figure 3, A1 to A4). In contrast to screening for one antibiotic at a time by incorporation of that antibiotic into the agar, all four The BioBrick plasmids pSB1T3, pSB1A3, pSB1C3 and pSB1K4 expressing the red fluorescent protein cassette (J04450) and conferring resistance to tetracycline, ampicillin, chloramphenicol, and kanamycin respectively were used. Competent E. coli JM109 cells were transformed with each plasmid and evaluated in plates incorporated with the respective antibiotics (A1 to A4) and on plates with antibiotic disks (C1 to C4). Plates B1 to B4 are plates with antibiotics inoculated with the untransformed cells demonstrating that the cells used for transformations did not already have resistance to any of the antibiotics used. A1 to A4 shows resistant clones are recovered when selection is done by incorporating antibiotics in the plates and this was similar when antibiotic disks were used (C1 to C4).
antibiotics could be screened at once on a single plate by using discs (Figure 3, B1 and B2). Few transformants picked up all four antibiotics (Figure 3, B3 and B4). When the four plasmids were combined in different ratios, colonies resistant colonies were formed for all plasmids irrespective of the ratios used ( Figure 3, C1 to C4 and D1 to D4).

Demonstration of the PAMRS workflow with human, chicken and cattle gut plasmidomes
Having demonstrated that antibiotic disks could be used to rapidly screen for multiple antibiotic resistant phenotypes at once on a single plate, we further tested the PAMRS workflow ( Figure 4) on plasmidomes extracted from faecal specimens from chickens, cattle and humans. Thirteen antibiotics were screened. The cells used for transformation was not resistant to any of the antibiotics ( Figure 5, A and B). Cells that pick up resistant plasmids were seen to grow in the zone of clearing as single colonies ( Figure 5, C and D). For a plasmidome that did not contain resistant plasmids, clear zones were seen ( Figure 5E). The colonies on initial plates were confirmed successfully when plated on fresh plates with the same antibiotic disk they were initially resistant to ( Figure 5F).
Of all the samples screened, 36.8% and 31.6% of chickens and humans did not have any antibiotic resistant conferring plasmids ( Figure 6B). Furthermore, 21.1%, 19.0% and 15.8% of chickens, cattle and humans had plasmids conferring antibiotic resistance to only one antibiotic. The highest number of resistance conferred in one particular sample was 8 in humans (5.3%), 5 in cattle (9.5%) and 7 in chickens (2.6%). In humans however, resistance to 5 antibiotics was found in 21.1% of the subjects followed by 6 antibiotics (10.5%) and 5.5% each for 2, 4 and 7, respectively. Similarly, in the chickens, resistance to 2, 3, 4, 5, and 6 antibiotics in the same sample was recorded in 15.8%, 7.9%, 7.9%, 5.3%, and 2.6% respectively. Within the cattle, the highest number of antibiotic resistance recorded in a single sample was 3 (28.6%), followed by 4 (23.8%) and two (19.0%) ( Figure 6B).

Discussion
The ecological impacts of plasmids are without doubt. Plasmids carry genes responsible for producing products known to help their prokaryotic hosts compete and survive better, as well as remain flexible and adaptable in the environment. Some of these plasmid-encoded characteristics include antibiotic resistance, production of antimicrobials, degradation of xenobiotics, among others 26,27 . Interrogation of plasmid functions have long been studied [28][29][30] . These studies are based on the ability of plasmids to confer selectable markers to their hosts. With genomic tools, we have come to understand that similar to prokaryotic hosts, plasmids are also defined by ecological niches 31,32 . However, the presence of genes does not always translate to a biological cause of phenotypes 33 . Hence, using associated genetic markers can lead to incorrect predictions of functionality and the associated biological role organisms or communities.
One advantage of genomic approaches is that they give a bigger picture and appears to provide rapid answers to questions. However, since associations between genetic markers and phenotypes can be spurious 34 , we sought to develop protocols that can be used in parallel with genomic tools to interrogate the actual role and functionality of plasmids in microbial communities. In this study, we demonstrated this with antibiotic resistant functions of plasmids. We have shown that with a combination of tools that are not traditionally used together it is feasible to rapidly investigate plasmid-encoded antibiotic resistance phenotypes. The protocol we developed and evaluated comprised extraction of the total plasmid community (the plasmidome) from a sample, transformation of a host bacteria of interest with the extracted plasmidome and then screening for the presence of the phenotypes. Studies on plasmidomes so far have largely been sequencing and bioinformatics based 35,36 and can be complemented by phenotypic studies such as what we have demonstrated to obtain deeper insights into what genes are functional and those that are not. The system we have demonstrated can particularly be targeted to investigate the importance of plasmids in the dissemination of antibiotic resistance in medically important pathogens such as the glass priority pathogens 37 . A targeted evaluation of plasmidomes can reveal unique adaptations employed by microbial communities Figure 6. Prevalence of antibiotic resistant conferring plasmids in chickens, cattle, and humans. The prevalence of resistance to individual antibiotics conferred by plasmids showed a higher diversity in humans, followed by chickens and then cattle (A). The highest number of resistance conferred by plasmids from a single sample in humans, chickens and cattle was 8, 7 and 5 respectively (B).
in medical facilities, human and animal guts and the environment as they respond to gradual exposure to antibiotics.
From the samples we evaluated with this method, we observed that E. coli, a very important pathogen and indicator organism can pick up resistant phenotypes from chicken, cattle and human faecal samples. This has many implications for the control of antibiotic resistance and suggests that the environment (e.g. soil and water) where excreta from animals and humans intersect is an important sink for evolution and spread of antibiotic resistant phenotypes encoded by plasmids. This interaction can be further exacerbated with the formation of biofilms which provide a protected environment for microbes to interact and share mobile genetic elements [38][39][40] . Another key finding from this study is that humans compared to chickens and cattle appear to harbour more antibiotic resistant plasmids. Further studies in this area are required to enable the quantification of the contribution of antibiotic resistant plasmids from humans to the overall antibiotic resistant burden. Overall, the plasmidomes from the three species studies were found to harbour plasmids conferring resistance to multiple antibiotics. Resistance to ceftazidime was highest in humans and cattle. This is particularly worrying as it is a drug for treating the difficult to manage difficult to treat infections by Pseudomonas aeruginosa 41 and other organisms. It has been demonstrated that, mutationdriven evolution of resistance to ceftazidime in Pseudomonas aeruginosa could develop in a span of just 30 days of exposure to increasing concentrations of the drug 42 . With the very high levels of resistance from the cattle and human samples, there is the need to review the use of the drug in these populations as well as other areas. Overall, it was evident that at least one plasmidome contained a plasmid that confers resistant to each of the thirteen antibiotics investigated. This affirms the calls for action to halt the antibiotic resistant menace and the need to investigate the mechanisms of evolution and spread of resistance so as to develop sustainable ways to stop the spread.

Conclusion
This work has demonstrated that, antibiotic resistant phenotypes conferred by plasmidomes can be rapidly screened with phenotypic tools like the PAMRS workflow ( Figure 4). The application of these techniques will enable assessment of expressed drug resistant phenotypes which otherwise cannot be determined by genome sequencing and assembly of plasmids. The tool described here can be applied to the study of other plasmidome functions such as virulence genes, among others. Data from this study suggests that humans may carry plasmids conferring resistance to more antibiotics compared to chickens and cattle. This however needs to be studied within a larger study population.