Research

System analysis and conceptual models

Objective: To understand the two case study catchments, define their system boundaries and examine their characteristics. To investigate the role of manufacturing plants and AMR hotspots, establish the Source, Pathway, and Receptor (SPR) linkages.

This will involve collation and interpretation of relevant existing large datasets (available online, open access, national agencies).

A systematic geo-database will be constructed that will include, information on:

• Watershed geomorphology.
• Waterbody typology and characteristics (dilution and hydrodynamic features).
• Climatic data.
• Land use.
• Potential sources of AMR such as pharmaceutical companies and antimicrobial production (Figure 1), hospitals, sewerage systems, and landfills.

Watersheds, in the areas of Puducherry and Chennai will be delineated via digital terrain analyses. A preliminary survey and consultation with the local authorities will be used to select eight manufacturing facilities (five in Puducherry and three in Chennai) out of a list of companies operating in these areas already compiled by our Indian Co-Is.

The selection will be based on the microbials produced, manufacturing methods and volumes of effluents discharged. Additional data will be acquired regarding the selected manufacturing facilities including: the size, type, and quantities of antimicrobials produced, water consumption, and data on any existing effluent treatment processes or other systems in place to deal with AMR.

Candidate antimicrobials will be screened for our risk assessment and experimental component, based on methodology developed by the Voulvoulis group.  A minimum of three antimicrobials produced by the three main methods i.e. industrial fermentation, semi-synthetic and synthetic will be selected for study.

Outputs from the spatial analysis using GIS and the screening of antimicrobials will be jointly interpreted by experts of Imperial. Indian partners will identify potential factors controlling the hazard and construct SPR linkages of the identified hazards for the case studies.

In addition, depending on antimicrobial type and on various hydrologic and hydrogeographic factors (e.g. water depth, water width, surface and volume of the water body, degree of mixing, pH-value, oxygen content, water hardness, conductivity, flow rate, character of the riparian zone, sediment particle size), WP1 will finalise the sampling details for analysis “downstream”, and identify specific stretches of the water bodies, distance from the sources of emission and environmental matrices (e.g. water, sediment, soil) to be monitored, the sampling period and duration. This in turn will be used to investigate the fate and transport along each SPR linkage, to finalise the sampling campaign (WP2 & WP3). 

Quantification of antimicrobials

Objective: determine the extent of antimicrobial waste contamination in the two case studies.

A preliminary sampling campaign will take place in Pondicherry will serve as a scoping study for method development. A limited number of all the types of samples we are aiming to analyse will be collected and processed as a test run for our main sampling campaign. This will include analytical, microbiological, and DNA studies

The main sampling campaign will take place over a year at quarterly intervals (winter, summer, monsoon, and autumn) to account for stream flow variability and other seasonal factors. Effluent discharges from the eight manufacturing sites (five located in Pondicherry and three in Chennai) will be sampled, with one site upstream used as control (background levels) at each case study region. Three sites would be located downstream at increasing distances from the five discharge points in Pondicherry, and two downstream of the three discharge points in Chennai.

We aim to collect triplets of composite water samples for the analysis of antimicrobials and physicochemical parameters (i.e. dissolved oxygen, pH, water temperature and conductivity). Each sample will be a mixture of samples from surface waters on the left, centre and right river side of the stream channel. Once mixed, all samples will be transferred into 2L PE bottles. Samples from the 0–5 cm surface layer of the sediment will be collected in glass jars. In the case of solid antimicrobial manufacturing waste (AMW)in landfills, we will be selecting soil samples from the top 15 cm of the soil profile and then homogenize them for further analysis. Control samples (background levels) will be collected from upstream sites selected to represent areas unaffected by AMW pollution sources. Duplicate samples will be randomly collected as part of quality control (QC). Furthermore, a number of samples will also be validated through analysis by a third party independent lab in India as part of QC.

The antibiotic residue separated (fractions) off the samples will be partitioned with an equal volume of chloroform and the MS data acquisition will be performed as per Kushwaha et al. (2019 in Journal of Liquid Chromatography & Related Technologies) with suitable modifications. Briefly, the chloroform layer will be combined, dried and reconstituted in methanol-water (9:2) to inject (2 mL) in LC-MS. The LC separation will be performed at a flow rate of 0.5 ml per min under a gradient of eluent-A composed of 0.1% (v/v) formic acid in water and eluent-B composed of acetonitrile. The HRMS data will be obtained in a LC-MS-MS system and the MS data will be acquired with capillary voltage 3 kV, drying gas temperature 300°C, drying gas flow rate 12 L per min, and a nebulizer pressure 35 psi.

The scan source parameters will be set as per the technical requirements. The data will be acquired over a mass scan range of m/z 100–2000, with ESI in positive mode. MS/MS acquisition of molecular ion will be operated with same parameters using fixed collision energy 30 eV in the positive mode for target antibiotics. Facilities are available at Pondicherry University where these analyses will take place. ICL/CEP with Pondicherry University will perform the data analysis. Data will be cleaned, formatted and processed according to QC established protocols. Statistical analysis will focus on multivariate analysis and exploring relationships (linear regression and mixed model methodology) between antimicrobials; physico-chemicals and other variables. In addition, PCA will be used to investigate stratification.

Penetration of AMR

Objective:  investigate the penetration of AMR bacteria and mobile genetic elements into selected habitats and communities exposed to effluent discharges relative to communities with lower (or no) levels of exposure.

Samples from the environment will be cultured on selective and non-selective chromogenic and other agar to identify cultivable AMR bacteria. Automated antimicrobial testing (AST) will be coupled with spectroscopic methods, e.g. MALDI-TOF-MS and ESI-MS to identify species. All the methods will conform to standards set out in the Clinical Microbiology Procedures Handbook and Performance Standards for Antimicrobial Susceptibility Testing (CLSI) 29th Informational Supplement, 2019 using recommended ATCC standard strains for Kirby Bauer, MRSA, MSSA, VRSA, VRE, VSSA, hVISA, and VISA.

Multidrug resistant (MDR) enterobacteriaceae will be defined by resistance to at least 3 of the following 4 groups- ciprofloxacin, ceftriaxone/ceftazidime, piperacillin tazobactam, gentamicin/tobramycin, and ertapenem/meropenem. Species with > 30 isolates per year will be subject to whole genome sequencing (WGS). MDR organisms will be screened for beta lactamses (MBL, ESBL, and AmpC) enzymes. Phenotypically AMR bacteria will be stored for molecular AMR screening; DNA will be extracted for WGS. MIC and MBC scores will be evaluated following CLSI or EUCAST breakpoints.

Environmental water samples will be filtered to remove large particles, then tprokaryotes enriched onto membranes by filtration prior to DNA extraction in India using bespoke kits and protocols (Rowe 2015, Ma 2019). Environmental DNAs submitted to the UK will undergo 16s ribosomal (r)RNA analysis using protocols developed for Illumina MiSeq. The 16s rRNA reads will be assembled into Operational taxonomix unit (OTU) using QIIME2, with NCBI 16S RefSeq curated database. Analysis will provide a comprehensive genus level overview of bacterial constituents in each environmental sample (n=140), compared with controls.

Based on culture and 16s rRNA, a subset of samples (~8) with optimised DNA and microbial content will be selected for metagenomic analysis, ensuring appropriate controls at all times. Paired-end metagenomic sequencing using Illumina HiSeq4000 will be undertaken ensuring adequate depth of coverage in the sequencing process.
Variability, species abundance and taxonomy will be evaluated using Simka (22) and MetaPhlan2 (23). Characterization at the species level will be performed using metaMLST (24). Confirmation of 16S classification will be performed by extracting reads of interest through HMMER (25). Metagenomes will be de novo assembled using a dedicated assembly tool, such as metaSPAdes (26) and MegaHIT(27). Antibiotic and virulence genes will be detected using BLAST with appropriate databases, such as Abricate.

Blank samples will be added to control for contaminants in handling or sequencing. Data analyses will be performed using containers and workflow managers, to increase the reproducibility of data analysis. We will determine the presence of integrons and plasmids that can mobilise ARGs using INTEGRALL and Plasmid Finder. Metagenomic data will identify AMR genes of particular prominence or relevance. We will then use metagenomic data to inform targeted AMR gene detection by qPCR using enabling testing of all the samples collected in relation to 16s rRNA gene content, greatly increasing the power of our study.

WP3 will also examine gut (faecal) microbiota in downstream and control sampling areas. Samples (n=56 in total) will be obtained from fish or other animals that use exposed waters and compared with those in control (non-exposed) areas. Following institutional ethical approval, human faecal samples (n=85; 10 per site) will be obtained from consenting users of the affected and control water sources for microbiota examination. Samples will be subject to culture on selective media as above. DNA will be extracted immediately and submitted to ICL for 16s ribosomal RNA analysis; metagenomic analysis using Illumina HiSeq4000 will be performed on a subset (~6 animal and ~6 human) selected based on culture and 16s rRNA.

Assemblies will be analysed and screened for AMR gene content as above. Based on metagenomic data, we will then undertake targeted qPCR to detect specific AMR genetic elements on all microbiota samples collected (in relation to 16s rRNA gene content), as for environmental samples. Data will be grouped according to sites of collection, by genus, species, antibiotic non-susceptibility, and AMR gene presence, (recorded as frequencies or percentages). AMR gene abundance will be calculated per 16s rRNA gene content, permitting non-parametric comparison of groups. Data analysis will be undertaken in collaboration with our partner Prof Aanensen (WSTI) to compare our findings with global AMR genomic analyses he is undertaking, to place our results into an international context.

Finally, to determine the extent to which AMR elements have penetrated into the regional healthcare setting we will analyse AMR infections with our partners from IGMCRI and AVMCH who are prospectively collecting AMR pathogens that colonise or cause invasive infections in those living near sampling sites. DNA will be extracted from isolates in India and WGS undertaken in London to determine if the AMR elements associated with disease are similar or linked to the AMR elements identified in our environmental and faecal microbiota studies.

 Integration and risk analysis

Objective: The development of standards, new tools for AMR risk assessment, Substance Flow Analysis (SFA) models for the case studies, and the validation of existing ERA methods.

We will perform SFA on the selected antimicrobials (three scenarios in total, to cover the variation in the production methods) and will calculate concentrations based on production loads, the flows and stock change rates of the system, and outputs from WP2 and WP3. First, we will implement innovative statistical approaches to dissect the associations and correlation between antimicrobial waste, antimicrobial levels, and abundance of AMR genes found in the samples using a multilevel Poisson model; an observation-level random effect will be adopted to model the over dispersion inherent to count data.

Statistical methods, including those implemented in SNPTEST, EPACTS, EMMAX, DAPC package in R, SCOPA and MARV multi-variable multi-omics analytical approaches, together with large scale clustering techniques will be used for antimicrobial waste, AMR gene and external variables. We will evaluate causal relationships between bacterial profiles and AMR genes using bi-directional and multivariable MR instrumental variable (IV) analyses to address the important question of AMR genes and external factors. We will implement a bi-directional multivariable MR and Egger regression, which limit confounding due to pleiotropy, to test antimicrobial waste and direction of the relationship between microbiota profiles and AMR genes. For each of the selected antimicrobials we will model a surrogate PNEC for AMR, adopting Minimal Inhibitory Concentrations (MICs) of the identified bacteria (WP3). This approach will predict upper boundaries for resistance. MIC data will be obtained from the public EUCAST database, which it the most comprehensive dataset available where theoretical PNECs (PNECR(T)) have been calculated for 111 antibiotics.