Antimicrobial Resistance Genomes, Big Data and Emerging ......09:00-10:30 Session 2: Genomic...

1

Name:

Antimicrobial Resistance –

Genomes, Big Data and Emerging

Technologies

Wellcome Genome Campus Conference Centre, Hinxton, Cambridge, UK

27-29 November 2018

Scientific Programme Committee:

Till Bachmann

University of Edinburgh, UK

Susanna Dunachie

Nuffield Department of Medicine, UK

Iruka Okeke

University of Ibadan, Nigeria

Julian Parkhill

Wellcome Sanger Institute, UK

Sharon Peacock

London School of Hygiene and Tropical Medicine, UK

Tweet about it: #AMR18

@ACSCevents /ACSCevents /c/WellcomeGenomeCampusCoursesandConferences

2

Wellcome Genome Campus Scientific Conferences Team:

Rebecca Twells

Head of Advanced Courses and

Scientific Conferences

Treasa Creavin

Scientific Programme

Manager

Nicole Schatlowski

Scientific Programme

Officer

Jemma Beard

Conference & Events

Organiser

Lucy Criddle

Conference & Events

Organiser

Beccy Jones

Conference & Events

Assistant

Laura Hubbard

Conference & Events Manager

Sarah Offord

Conference & Events Office

Administrator

Zoey Willard

Conference & Events

Organiser

3

Dear colleague,

I would like to offer you a warm welcome to the Wellcome Genome Campus Advanced Courses and

Scientific Conferences: Antimicrobial Resistance. I hope you will find the talks interesting and

stimulating, and find opportunities for networking throughout the schedule.

The Wellcome Genome Campus Advanced Courses and Scientific Conferences programme is run on a

not-for-profit basis, heavily subsidised by the Wellcome Trust.

We organise around 50 events a year on the latest biomedical science for research, diagnostics and

therapeutic applications for human and animal health, with world-renowned scientists and clinicians

involved as scientific programme committees, speakers and instructors.

We offer a range of conferences and laboratory-, IT- and discussion-based courses, which enable the

dissemination of knowledge and discussion in an intimate setting. We also organise invitation-only

retreats for high-level discussion on emerging science, technologies and strategic direction for select

groups and policy makers. If you have any suggestions for events, please contact me at the email

address below.

The Wellcome Genome Campus Scientific Conferences team are here to help this meeting run

smoothly, and at least one member will be at the registration desk between sessions, so please do

come and ask us if you have any queries. We also appreciate your feedback and look forward to your

comments to continually improve the programme.

Best wishes,

Dr Rebecca Twells Head of Advanced Courses and Scientific Conferences [email protected]

mailto:[email protected]

4

General Information

Conference Badges

Please wear your name badge at all times to promote networking and to assist staff in identifying you.

Scientific Session Protocol

Photography, audio or video recording of the scientific sessions, including poster session is not

permitted.

Social Media Policy

To encourage the open communication of science, we would like to support the use of social media at

this year’s conference. Please use the conference hashtag #AMR18. You will be notified at the start

of a talk if a speaker does not wish their talk to be open. For posters, please check with the presenter

to obtain permission.

Internet Access

Wifi access instructions:

Join the ‘ConferenceGuest’ network

Enter your name and email address to register

Click ‘continue’ to send an email to the registered email address

Open the registration email, follow the link ‘click here’ and confirm the address is valid

Enjoy seven days’ free internet access!

Repeat these steps on up to 5 devices to link them to your registered email address

Presentations

Please provide an electronic copy of your talk to a member of the AV team who will be based in the

meeting room.

Poster Sessions

Posters will be displayed throughout the conference. Please display your poster in the Conference

Centre on arrival. There will be two poster sessions during the conference.

Odd number poster assignments will be presenting in poster session 1, which takes place on

Tuesday, 27 November at 17:35 – 19:00.

Even number poster assignments will be presenting in poster session 2, which takes place on

Wednesday, 28 November at 18:05 – 19:30.

The abstract page number indicates your assigned poster board number. An index of poster

numbers appears in the back of this book.

Conference Meals & Social Events

Lunch and dinner will be served in the Hall restaurant. Please refer to the conference programme in

this book as times will vary based on the daily scientific presentations.

All conference meals and social events are for registered delegates.Please inform the conference

organiser if you are unable to attend the conference dinner.

The Hall Bar (cash bar) will be open from 19:00 – 23:00 each day.

Please note there are no lunch or dinner facilities available outside of the conference times.

5

Dietary Requirements

If you have advised us of any dietary requirements, you will find a coloured dot on your badge.

Please make yourself known to the catering team and they will assist you with your meal request.

If you have a gluten allergy, we are unable to guarantee the non-presence of gluten in dishes even if

they are not used as a direct ingredient. This is due to gluten ingredients being used in the kitchen.

For Wellcome Genome Campus Conference Centre Guests

Check in

If you are staying on site at the Wellcome Genome Campus Conference Centre, you may check into

your room from 14:00. The Conference Centre reception is open 24 hours.

Breakfast

Your breakfast will be served in the Hall restaurant from 07:30 – 09:00

Telephone

If you are staying on-site and would like to use the telephone in your room, you will need to

contact the Reception desk (Ext. 5000) to have your phone line activated - they will require your

credit card number and expiry date to do so.

Departures

You must vacate your room by 10:00 on the day of your departure. Please ask at reception for

assistance with luggage storage in the Conference Centre.

Taxis

Please find a list of local taxi numbers on our website. The conference centre reception will also be

happy to book a taxi on your behalf.

Return Ground Transport

Complimentary return transport has been arranged for 13:30 on Thursday, 29 November to

Cambridge station and city centre (Downing Street), and Stansted and Heathrow airports.

A sign-up sheet will be available at the conference registration desk from 15:30 on Tuesday, 27

November. Places are limited so you are advised to book early.

Please allow a 30 minute journey time to both Cambridge and Stansted Airport, and two and a half

hours to Heathrow.

Messages and Miscellaneous

Lockers are located outside the conference centre toilets and are free of charge.

All messages will be posted on the registration desk in the Conference Centre.

A number of toiletry and stationery items are available for purchase at the conference centre

reception. Cards for our self-service laundry are also available.

Certificate of Attendance

A certificate of attendance can be provided. Please request one from the conference organiser

based at the registration desk.

Contact numbers

Wellcome Genome Campus Conference Centre – 01223 495000 (or Ext. 5000)

Wellcome Genome Campus Conference Organiser (Laura) – 07733 338878

If you have any queries or comments, please do not hesitate to contact a member of staff who will

be pleased to help you.

2018-19

wel

lcom

egen

omec

ampu

s.or

g/co

urse

sand

conf

eren

ces

@AC

SCev

ents

CONFERENCESGenome Informatics17–20 September 2018Plant Genomes in a Changing Environment24–26 October 2018 NEWAntimicrobial Resistance - Genomes, Big Data and Emerging Technologies27-29 November 2018 NEWExploring Human Host-Microbiome Interactions in Health and Disease5–7 December 2018Immunogenomics of Disease: Accelerating to Patient Benefit5–7 February 2019Innate Immune Memory6–8 March 2019The Challenge of Chronic Pain20–22 March 2019Genomics of Rare Disease27–29 March 2019Animal Genetics and Diseases8–10 May 2019Applied Bioinformatics and Public Health Microbiology22–24 May 2019

COURSES LABORATORY COURSES

Chromatin Structure and Function30 October–9 November 2018Molecular Pathology and Diagnosis of Cancer18–23 November 2018Derivation and Culture of Human Induced Pluripotent Stem Cells (hiPSCs)10–14 December 2018 Genomics and Clinical Microbiology20–25 January 2019

Genomics and Clinical Virology10–15 February 2019Genetic Engineering of Mammalian Stem Cells10–22 March 2019Next Generation Sequencing5–12 April 2019Malaria Experimental Genetics12–18 May 2019RNA Transcriptomics19–28 June 2019

COMPUTATIONAL COURSESGenetic Analysis of Population-based Association Studies24–28 September 2018Next Generation Sequencing Bioinformatics7–13 October 2018Mathematical Models for Infectious Disease Dynamics18 February–1 March 2019Working with Pathogen Genomes24–29 March 2019Fungal Pathogen Genomics7–12 May 2019Computational Molecular Evolution13–24 May 2019Summer School in Bioinformatics24–28 June 2019Systems Biology: From Large Datasets to Biological Insight8–12 July 2019Proteomics Bioinformatics21–26 July 2019

LECTURE/DISCUSSION COURSESScience Policy: Improving the Uptake of Research into UK Policy 20–21 August 2018 NEWTranslating and Commercialising Genomic Research1–3 October 2018Genomics for Dermatology19–21 November 2018

Molecular Neurodegeneration14–18 January 2019Genomic Practice for Genetic Counsellors28–30 January 2019

OVERSEAS COURSESMolecular Approaches to Clinical Microbiology in Africa8–14 September 2018 (Kenya)Antimicrobial Resistance in Bacterial Pathogens16–21 September 2018 (Kenya) NEWNGS Analysis for Monogenic Disease in African Populations22-23 September 2018 (Rwanda) NEWWorking with Parasite Database Resources28 October–2 November 2018 (Malaysia)Working with Pathogen Genomes11–16 November 2018 (Uruguay)Next Generation Sequencing Bioinformatics20–25 January 2019 (Chile)Next Generation Sequencing Bioinformatics27 January–1 February 2019 (South Africa)Genomics and Epidemiological Surveillance of Bacterial Pathogens3–8 March 2019 (Peru)Practical Aspects of Drug Discovery: At the Interface of Biology, Chemistry and Pharmacology7–12 April 2019 (South Africa)

ONLINE COURSES NEW

Bacterial Genomes: Disease Outbreaks and Antimicrobial ResistanceBacterial Genomes: From DNA to Protein Function Using BioinformaticsBacterial Genomes: Accessing and Analysing Microbial Genome Data

Please see our website for scheduling of online courses

7

Conference Summary

Tuesday, 27 November

12:00-13:20 Registration with lunch

13:20-13:30 Welcome and Introductions

13:30-14:30 Keynote Lecture by Mark Woolhouse, University of Edinburgh, UK

14:30-15:30 Session 1: Machine learning and prediction of antimicrobial resistance

15:30-16:00 Afternoon Tea

16:00-17:30 Session 1 continued

17:15-17:35 Lightning talks

17:35-19:00 Poster Session 1 (odd numbers) with drinks reception

19:00 Dinner

Wednesday, 28 November

09:00-10:30 Session 2: Genomic surveillance and epidemiology: its role in detection, tracking and

control of antimicrobial resistance

10:30-11:00 Morning Coffee

11:00-12:30 Session 3: Genomic evidence that informs the debate on ‘farm to fork’ transmission of

resistant pathogens

12:30-14:00 Lunch

14:00-15:00 Keynote Lecture by Ramanan Laxminarayan CDDEP, USA / India

15:00-15:30 Session 4: Global burden of disease from drug-resistant infections: monitoring and

evaluation


16:00-17:45 Session 4 continued


18:05-19:30 Poster Session 2 (even numbers) with drinks reception

19:30 Conference Dinner

Thursday, 29 November

09:00-10:30 Session 5: Translating bacterial genomics into routine clinical practice


11:00-12:30 Session 6: Emerging technologies and analysis approaches

12.30-12:45 Closing remarks

12:45-13:30 Lunch

13:30 Coaches depart for Cambridge, Stansted Airport & Heathrow Airport

8

Conference Sponsors

9

Antimicrobial Resistance – Genomes, Big Data and

Emerging Technologies

Wellcome Genome Campus Conference Centre,

Hinxton, Cambridge

27 - 29 November 2018

Lectures to be held in the Francis Crick Auditorium

Lunch and dinner to be held in the Hall Restaurant

Poster sessions to be held in the Conference Centre

Spoken presentations - If you are an invited speaker, or your abstract has been selected for a

spoken presentation, please give an electronic version of your talk to the AV technician.

Poster presentations – If your abstract has been selected for a poster, please display this in the

Conference Centre on arrival.

Conference programme

Tuesday, 27 November

12:00-13:20 Registration with lunch

13:20-13:30 Welcome and Introductions

Sharon Peacock, London School of Hygiene and Tropical Medicine, UK

13:30-14:30 Keynote Lecture

Session Chair: Sharon Peacock, London School of Hygiene and Tropical Medicine, UK

Big gaps in our knowledge about AMR

Mark Woolhouse,


14:30-15:30 Session 1: Machine learning and prediction of antimicrobial

resistance

Chair: Julian Parkhill, Wellcome Sanger Institute, UK

14:30 Global surveillance of AMR and the challenges for machine learning

Zamin Iqbal

EMBL-EBI, UK

15:00 Considerations for coding evolution: a meta-analysis of machine-

learning based prediction of antibiotic resistance

Allison Hicks

Harvard University, USA

15:15 Analysis of Machine Learning Methods in Predicting Drug Resistance of

Mycobacterium Tuberculosis

Leonid Chindelevitch

Simon Fraser University, USA

10


16:00-17:30 Session 1 continued: Machine learning and prediction of

antimicrobial resistance

16:00 Using machine learning to predict antimicrobial minimum inhibitory

concentrations and associated genomic features for nontyphoidal

Salmonella

James Davis

University of Chicago, USA

16:30 Machine Learning Approaches for Improving Antimicrobial Resistance

Prediction in M. tuberculosis using PointFinder

Camilla Hundahl Johnsen

Technical University of Denmark, Denmark

16:45 A k-mer-based method for the identification of phenotype-associated

genomic biomarkers and predicting phenotypes of sequenced bacteria.

Erki Aun

University of Tartu, Estonia

17:00 Using machine learning to guide targeted and locally-tailored empiric

antibiotic prescribing in a children’s hospital in Cambodia

Mathupanee Oonsivilai

Mahidol-Oxford Tropical Medicine Research Unit, Thailand


Chair: Julian Parkhill, Wellcome Sanger Institute, UK

17:35-19:00 Poster Session 1 (odd numbers) with drinks reception

19:00 Dinner

Wednesday, 28 November

09:00-10:30 Session 2: Genomic surveillance and epidemiology: its role in

detection, tracking and control of antimicrobial resistance

Chair: Iruka Okeke, University of Ibadan, Nigeria

09:00 Distribution of microbial communities and antimicrobial resistances in

urban environments and in space

Daniela Bezdan

Weill Cornell Medicine, USA

09:30 Changing patterns of antimicrobial resistance acquisition and spread in

a single hospital for priority pathogens

Nick Thomson


11

10:00 Epidemic of carbapenem-resistant Klebsiella pneumoniae in Europe is

driven by nosocomial spread: Inference from a continent-wide

population analysis

Sophia David


10:15 Changing pattern of Carbapenemase-Producing Enterobacteriaceae

(CPE) in Hong Kong hospitals, a 7-year overview.

Margaret Ip

Chinese University of Hong Kong, Hong Kong


11:00-12:30 Session 3: Genomic evidence that informs the debate on ‘farm to

fork’ transmission of resistant pathogens

Chair: Sharon Peacock, London School of Hygiene and Tropical Medicine, UK

11:00 Meat, urine and time machines

Lance Price

George Washington University, USA

11:30 The challenges of leadership

Gwyn Jones

RUMA (Responsible Use of Medicines in Agriculture), UK

12:00 One Health genomic surveillance of E. coli reveals separate

populations and mobile genetic in humans and livestock

Catherine Ludden


12:15 Superbugs in the supermarket: High prevalence of antibiotic

resistances in chicken meat-associated enterococci

Daria Van Tyne

University of Pittsburgh, USA

12:30-14:00 Lunch

14:00-15:00 Keynote Lecture

Chair: Susanna Dunachie, Nuffield Department of Medicine, UK

State of the World’s Antibiotics in 2018

Ramanan Laxminarayan

The Center for Disease Dynamics, Economics & Policy (CDDEP), USA / India

15:00-15:30 Session 4: Global burden of disease from drug-resistant infections:

monitoring and evaluation

Chair: Susanna Dunachie, Nuffield Department of Medicine, UK

15:00 The Challenges Of Estimating the Global Burden of AMR

Susanna Dunachie

Nuffield Department of Medicine, UK


12

16:00-17:45 Session 4 continued: Global burden of disease from drug-resistant

infections: monitoring and evaluation

16:00 Attributable deaths and disability-adjusted life-years caused by

infections with antibiotic-resistant bacteria in the European Union and

the European Economic Area in 2015: a population-level health

estimate

Alessandro Cassini

ECDC, Sweden

16:30 Developing a geospatial modelling framework to estimate the burden

of antimicrobial resistance

Annie Browne

University of Oxford, UK

17:00 Global surveillance of antimicrobial resistance through sewage samples

and functional resistance

Pimlapas Leekitcharoenphon

Technical University of Denmark, Denmark

17:15 See and Sequence: Whole Genome Sequencing for the Surveillance of

Antimicrobial Resistance in the Philippines

Silvia Argimon

Centre for Genomic Pathogen Surveillance, UK

17:30 Transmission dynamics and between-species interactions of multidrug-

resistant Enterobacteriaceae

Thomas Crellen

Mahidol-Oxford Research Unit, Thailand



18:05-19:30 Poster Session 2 (even numbers) with drinks reception

19:30 Conference Dinner

Thursday, 29 November

09:00-10:30 Session 5: Translating bacterial genomics into routine clinical

practice


09:00 Public health surveillance of antimicrobial resistance in gastrointestinal

pathogens

Claire Jenkins

Public Health England, UK

09:30 Time to move on from 150 years of shoe leather epidemiology for

outbreak detection

Sharon Peacock


13

10:00 Defining a relatedness cut-off between MRSA isolate genomes that

predicts the probability of an outbreak

Francesc Coll

London School of Hygiene & Tropical Medicine, UK

10:15 Implementing genomic-based AMR surveillance from the ground up

Anthony Underwood

Centre for Genomic Pathogen Surveillance, WSI, UK


11:00-12:30 Session 6: Emerging technologies and analysis approaches

Chair: Till Bachmann, University of Edinburgh, UK

11:00 Training and validation of a novel 30-mRNA panel, HostDx Sepsis, for

diagnosing and prognosing acute infections and sepsis

Tim Sweeney

Inflammatix, USA

11:30 Antibiotic stewardship in the context of the human microbiome

Debby Bogaert


12:00 Genome-wide association uncovers novel candidate resistance and

compensatory mutations in antibiotic-resistant Neisseria gonorrhoeae

Kevin Ma

Harvard TH Chan School of Public Health, USA

12:15 Characterising beta-lactam antibiotic resistance in healthy human gut

microbiota

Lindsay Pike


12.30-12:45 Closing remarks

Sharon Peacock, London School of Hygiene and Tropical Medicine, UK

12:45-13:30 Lunch

13:30 Coaches depart to Cambridge City Centre and Train Station,

Stansted and Heathrow airports

14

These abstracts should not be cited in bibliographies. Materials contained herein should be

treated as personal communication and should be cited as such only with consent of the

author.

S1

Spoken Presentations Big gaps in our knowledge about AMR Mark Woolhouse University of Edinburgh Though forward projections of the AMR burden are difficult and controversial there is little doubt that the problem is set to get worse. A better evidence base on AMR burden, risk factors and interventions should enable better policy making to avert a crisis. AMR surveillance and monitoring at a global scale remains a significant challenge but is critically important. One innovative approach is to quantify the abundance of resistance genes in untreated sewage samples. The Global Sewage project conducts metagenomics sequencing from close to 100 sites worldwide. AMR gene profiles vary over space and time and are potential indicator of trends in levels of resistance, as well as presenting opportunities for identifying risk factors. Notably, countries that rank low on the human development index (HDI) tend to have high levels of AMR genes, consistent with multifactorial drivers of AMR at the national level. It is harder to identify a direct association with antimicrobial usage based on these data, however. The relationship between usage and resistance at the patient level may also be difficult to discern. This is illustrated by a systematic review of over 200 studies of risk factors for carbapenem resistance that suggests that other factors are at least as strongly associated with the presence of carbapenemase producing organisms as is direct exposure to carbapenems. While there is an understandable interest in resistance to last line antibiotics such as carbapenems, far less attention is paid to front line antibiotics. Conversely, modelling studies of the relationship between usage and the dissemination of resistance indicate that, in practice, protecting the efficacy of front line antibiotics has a considerably greater benefit for public health. This reflects both the very different volumes of usage and the resistance cascade whereby increased resistance to one antibiotic is rapidly followed by increased resistance to others, a pattern visible using data from the European CDC. These three case studies illustrate the complex relationship between evidence and policy. Nonetheless, the AMR problem is urgent and data and knowledge gaps must not be excuses for lack of action.

S2

Notes

S3

Global surveillance of AMR and the challenges for machine learning Zamin Iqbal EMBL-EBI, Wellcome Genome Campus, Hinxton, Cambridge, UK The global rise of antimicrobial resistance poses a huge threat to modern medicine and public health. We already know that bacterial evolution is outstripping our current ability to mount and manage a response. This places huge pressure on scientists and the healthcare sector to innovate and to update processes. I want to talk about recent innovations in genomic data sharing, which allow us to keep up in real time with the entire global corpus of sequenced DNA, running instant searches and even setting up alerts. It is now possible to give a complete global picture of the distribution of mutations, strains and mobile elements across all global data. While this is a very exciting and even transformative development, I will spend much of the talk discussing what these (and other) tools do not enable. In particular, there are some fundamental theoretical challenges to the use of global heterogeneous and historical datasets for machine learning and inference in general. This leads inevitably to discussion of what drug resistance means, and how we can correctly model and conceive of it.

S4

Notes

S5

Considerations for coding evolution: a meta-analysis of machine-learning based prediction of antibiotic resistance

Allison Hicks, Yonatan Grad

Department of Immunology and Infectious Diseases, Harvard T. H. Chan School of Public Health, Boston, Massachusetts

Using machine learning to predict antibiotic resistance phenotypes from whole genome sequencing data has been proposed as a useful supplement to phenotypic testing that can enhance resistance surveillance. However, while algorithmic prediction of antibiotic resistance has often been further presented as a promising diagnostic tool, there has been no systematic evaluation of factors that may influence performance of such models, how they might apply to and vary across clinical populations, and what the implications might be in the clinical setting. Here, we performed a meta-analysis of seven Neisseria gonorrhoeae datasets, as well as Klebsiella pneumoniae, Escherichia coli, and Staphylococcus aureus datasets, using set covering machine and random forest classifiers, as well as random forest regression, to predict ciprofloxacin and, for N. gonorrhoeae, azithromycin resistance phenotypes from whole genome sequencing data. We demonstrate how model performance varies by drug, resistance breakpoint, dataset, and species. Our findings underscore the importance of incorporating relevant biological knowledge, as well as statistical expertise, into design and assessment of such models. We suggest that while doing so can inform tailored modeling for individual drugs and species, the selective pressures applied by molecular diagnostics may impact the bacterial population and the prevalence of resistance mechanisms. Thus, for any such machine learning-based molecular diagnostics, it is important not only to regularly update training sets, but also to monitor the distributions of genetic features and resistance phenotypes routinely and as comprehensively as possible.

S6

Notes

S7

Analysis of Machine Learning Methods in Predicting Drug Resistance of Mycobacterium Tuberculosis

Leonid Chindelevitch, Guo Liang Gan, Matthew Nguyen, Elijah Willie, Brian Lee, Cedric Chauve, Maxwell Libbrecht

Simon Fraser University

The efficacy of antibiotic drug treatment in tuberculosis (TB) is significantly impeded by the development of drug resistance. There is a need for a robust diagnostic system which can accurately predict drug resistance in patients and identify key drug resistance markers. In recent years, researchers have been using whole-genome sequencing (WGS) data to infer antibiotic resistance. We analyze two datasets of different diversity: isolates from the Relational Sequencing TB Data Platform (ReSeqTB), a consortium dedicated to next-generation TB diagnostics, with global isolates from 32 dierent countries, and isolates collected by the British Columbia Centre for Disease Control (BCCDC) in Canada. We predict drug resistance to the first-line drugs: isoniazid, rifampicin, ethambutol, pyrazinamide, and streptomycin, where we focus on using mutations found within 23 genes known to be involved in first-line TB drug resistance. We consider models including logistic regression, support vector machines and random forests, and evaluate the robustness of each model by cross-testing: training on ReSeqTB and testing on BCCDC and vice versa. In general, all methods achieve at least 71% area under the ROC curve (AUC) and are consistently robust when predicting on both datasets. However, training on a diversied dataset, ReSeqTB, does not necessarily improve performance when testing on a homogeneous dataset, BCCDC, except for multidrug resistance. Next, we determine which among the 23 genes are important for resistance to each drug. We explore two methods for this investigation, permutation testing and PLS-Lasso classifier, and find that both methods yield a highly promising consensus for most drugs. Both methods are able to identy katG as important for isoniazid resistance and rpoB as important for rifampicin resistance. We also investigate mutation resistance markers by studying two trained models: logistic regression with L1 penalty and random forests. In both datasets, the mutation G944C in katG is highly ranked in both models, and this mutation is frequently reported in several studies to be resistance-related. For rifampicin resistance, the C1592T mutation in rpoB is also highly ranked in both models. Our work is an exploration of machine learning models on two major datasets of di fferent diversity and a detailed study of the usefulness of the key features in these models for identifying drug resistance markers. We conclude that, given enough high-quality data, standard machine learning techniques can obtain good classication performance and identify those genes and variants that cause drug resistance.

S8

Notes

S9

Using machine learning to predict antimicrobial minimum inhibitory concentrations and associated genomic features for nontyphoidal Salmonella James J. Davis1 Marcus Nguyen1, S. Wesley Long2, Patrick F. McDermott3, Randall J. Olsen2, Robert Olson1, Rick L. Stevens1, Gregory H. Tyson3, Shaohua Zhao3 1University of Chicago and Argonne National Laboratory, Chicago, IL, USA 2Houston Methodist Research Institute and Houston Methodist Hospital, Houston, Texas, USA 3Food and Drug Administration, Center for Veterinary Medicine, Office of Research, Laurel, MD, USA Nontyphoidal Salmonella species are the leading bacterial cause of food-borne disease in the United States. Whole genome sequences and paired antimicrobial susceptibility data are available for Salmonella strains because of surveillance efforts from public health agencies. In this study, a collection of 5,278 nontyphoidal Salmonella genomes, collected over 15 years in the United States, were used to generate XGBoost-based machine learning models for predicting minimum inhibitory concentrations (MICs) for 15 antibiotics. The MIC prediction models have an overall average accuracy of 95% within ± 1 two-fold dilution step (confidence interval of 95-95%), an average very major error rate of 2.7% (confidence interval of 2.4-3.0%) and an average major error rate of 0.1% (confidence interval of 0.1-0.2%). The model predicts MICs with no a priori information about the underlying gene content of the strains. By selecting diverse genomes for training sets, we show that highly accurate MIC prediction models can be generated with fewer than 500 genomes. We also show that our approach for predicting MICs is stable over time despite annual fluctuations in antimicrobial resistance gene content in the sampled genomes. Finally, using feature selection, we explore the important genomic regions identified by the models for predicting MICs. Our strategy for developing whole genome sequence-based models for surveillance and clinical diagnostics can be readily applied to other important human pathogens.

S10

Notes

S11

Machine Learning Approaches for Improving Antimicrobial Resistance Prediction in M. tuberculosis using PointFinder

Camilla Hundahl Johnsen, Derya Aytan, Frank M. Aarestrup, and Ole Lund

Research Group for Genomic Epidemiology, National Food Institute, Technical University of Denmark, Kongens Lyngby, Denmark.

Extensive use of antimicrobials has driven the progressive emergence of resistant bacteria and other microbes. Surveillance of antimicrobial resistance and rapid detection of resistance is important to facilitate an effective treatment, and hence reduce drug mis-usage. Additional rapid detection is essential as resistance call for extra attention and precaution. Next-generation sequencing technologies rapidly yield accurate whole-genome sequencing (WGS) data, from which antimicrobial resistance might be directly determined. At Center for Genomic Epidemiology we aim to provide accurate bioinformatic solutions for determining resistance phenotypes. We have developed both ResFinder and PointFinder, in silico methods for predicting resistance based on acquired resistance genes or chromosomal mutations from WGS data. In this study we focus on improving our PointFinder tool. PointFinder is a rapid, assemblyfree program that detects point mutations, insertions and deletions. Resistance is predicted when any detected mutation is present in our curated database of resistance associated mutations. PointFinder is available for several species, including M. Tuberculosis (Mtb). We optimized the Mtb resistance prediction by feature selection methods, such as forward selection, ReliefF and L1 Norm Support Vector Machine methods. We also found that the literature on Mtb resistance mutations does contain misclassified resistance markers, and by elimination those from our model we improved the prediction performance. Traditionally, the occurrence of a resistance mutation has been treated as a direct indicator for resistance. However, it is now suspected that some mutations need the existence of other mutations to cause drug resistance, and more complex mutation patterns might explain the resistance emergence. Therefore, we have initiated a machine learning study to predict resistance by random forest and neural network prediction models. Pending results using these approaches will show how linked mutation might account for resistance that could not have been explained before.

S12

Notes

S13

A k-mer-based method for the identification of phenotype-associated genomic biomarkers and predicting phenotypes of sequenced bacteria.

Erki Aun(1), Age Brauer(1), Veljo Kisand(2), Tanel Tenson(2), Maido Remm(1)

(1)Department of Bioinformatics, Institute of Molecular and Cell Biology, University of Tartu, Tartu, Estonia (2)Institute of Technology, University of Tartu, Tartu, Estonia

We have developed an easy-to-use and memory-efficient method called PhenotypeSeeker that (a) identifies phenotype-specific k-mers, (b) generates a k-mer-based statistical model for predicting a given phenotype and (c) predicts the phenotype from the sequencing data of a given bacterial isolate. The method was validated on 167 Klebsiella pneumoniae isolates (virulence), 200 Pseudomonas aeruginosa isolates (ciprofloxacin resistance) and 459 Clostridium difficile isolates (azithromycin resistance). The phenotype prediction models trained from these datasets obtained the F1-measure of 0.88 on the K. pneumoniae test set, 0.88 on the P. aeruginosa test set and 0.97 on the C. difficile test set. The F1-measures were the same for assembled sequences and raw sequencing data; however, building the model from assembled genomes is significantly faster. On these datasets, the model building on a mid-range Linux server takes approximately 3 to 5 hours per phenotype if assembled genomes are used and 10 hours per phenotype if raw sequencing data are used. The phenotype prediction from assembled genomes takes less than one second per isolate. Thus, PhenotypeSeeker should be well-suited for predicting phenotypes from large sequencing datasets. PhenotypeSeeker is implemented in Python programming language, is open-source software and is available at GitHub (https://github.com/bioinfo-ut/PhenotypeSeeker/).

S14

Notes

S15

Using machine learning to guide targeted and locally-tailored empiric antibiotic prescribing in a children’s hospital in Cambodia

Mathupanee Oonsivilai, Mo Yin, Nantasit Luangasanatip, Yoel Lubell, Thyl Miliya, Pisey Tan, Lorn Loeuk, Paul Turner, Ben Cooper

Mahidol-Oxford Tropical Medicine Research Unit, Faculty of Tropical Medicine, Mahidol University, Bangkok, Thailand. Cambodia-Oxford Medical Research Unit, Angkor Hospital for Children, Siem Reap, Cambodia. Center for Tropical Medicine and Global Health, Nuffield Department of Medicine, University of Oxford, Oxford, United Kingdom. Division of Infectious Disease, University Medicine Cluster, National University Hospital, Singapore.

Early and appropriate empiric antibiotic treatment of patients suspected of having sepsis is associated with reduced mortality. The increasing prevalence of antimicrobial resistance reduces the efficacy of empiric therapy guidelines derived from population data. This problem is particularly severe for children in developing country settings. We hypothesized that by applying machine learning approaches to readily collect patient data, it would be possible to obtain individualized predictions for targeted empiric antibiotic choices.We analysed blood culture data collected from a 100- bed children's hospital in North-West Cambodia between February 2013 and January 2016. Clinical, demographic and living condition information for each child was captured with 35 independent variables. Using these variables, we used a suite of machine learning algorithms to predict Gram stains and whether bacterial pathogens could be treated with common empiric antibiotic regimens: i) ampicillin and gentamicin; ii) ceftriaxone; iii) none of the above. 243 patients with bloodstream infections were available for analysis. We found that the random forest method had the best predictive performance overall as assessed by the area under the receiver operating characteristic curve (AUC).The random forest method gave an AUC of 0.80 (95%CI 0.66-0.94) for predicting susceptibility to ceftriaxone, 0.74 (0.59-0.89) for susceptibility to ampicillin and gentamicin, 0.85 (0.70-1.00) for susceptibility to neither, and 0.71 (0.57-0.86) for Gram stain result. The most important variables for predicting susceptibility were time from admission to blood culture, patient age, hospital versus community-acquired infection, and age-adjusted weight score.Applying machine learning algorithms to patient data that are readily available even in resource-limited hospital settings can provide highly informative predictions on antibiotic susceptibilities to guide appropriate empiric antibiotic therapy. When used as a decision support tool, such approaches have the potential to improve targeting of empiric therapy, patient outcomes and reduce the burden of antimicrobial resistance.

S16

Notes

S17

Distribution of microbial communities and antimicrobial resistances in urban environments and in space Daniela Bezdan 1,2, David Danko 1,2, Cem Meyden 1,2, Chou Chou 1,2, Christopher E. Mason 1,2 1Department of Physiology and Biophysics, Weill Cornell Medicine, New York, NY, 2The HRH Prince Alwaleed Bin Talal Bin Abdulaziz Alsaud Institute for Computational Biomedicine, Weill Cornell Medicine, New York, NY, USA The majority of the world’s population lives in cities, yet our understanding of the dynamics and distribution of bacteria, viruses, and antimicrobial resistance (AMR) genes in urban built environments is limited. Cities and urban transport hubs like metro stations and airports represent the sites of highest human population density, and are thus hotspot areas for microbial and genetic exchange in both developing and developed countries. Moreover, most inhabitants, visitors and passenger carry a mobile phone, which serves as a “molecular echo” of the travel history of a person as well as a representation of his or her personal microbiome. Through leveraging next-generation sequencing (NGS) and metagenome sequencing methods, combined with rapid computational analysis, we can reveal the microbial and genetic profiles, AMR gene content, localization, and movement patterns ‘imprinted’ on phones and city surfaces, as readily as for clinical samples. We hypothesized that each person reveals a distinct AMR profile that is representative of their city of origin and personal interactions with the environment, and that each person can harbor divergent microbial and viral populations harboring various AMR types. Moreover, the microbiome of urban environments is in constant flux due to the constant exchange of microbes between individuals and surfaces in public spaces, transport hubs, phones, and between each other. Some of these interactions involve “high mobility” AMR genes, often encoded on plasmids. We developed a framework for AMR mapping and modeling AMR distribution that can help to improve the understanding of the epidemiology and dynamics of global AMR movements - especially during large events such as the Olympics in Brazil. Building on our experience with sequencing the subways of New York we established the International Consortium for Metagenomics of Subways and Urban Biomes (MetaSUB) at Weill Cornell Medicine, which is funded by the NIH, WorldQuant, and the Bill and Melinda Gates Foundation. We are currently analyzing >20.000 whole-genome shotgun metagenomes using samples collected in the subways and other transport hubs of more than 85 cities in 21 different countries. To our knowledge this is the first study investigating a comprehensive and worldwide view of urban microbiomes and AMR exchange. In parallel we have developed protocols for interrogating microbiomes obtained from mobile phone surfaces, which we showcased during several international conferences. To complement our knowledge, we are comparing the generated metagenomes with data collected by the Extreme Microbiome Project and microbial isolates collected at the ISS by NASA during the last 11 years. Furthermore, to evaluate our findings and improve the accuracy of microbiome analyses, we developed novel protocols for metagenomic linked-DNA genomic sequencing (using 10X Genomics Chromium) and long-read sequencing (using Oxford NanoporeMinION and PromethION sequencers).

S18

Notes

S19

Changing patterns of antimicrobial resistance acquisition and spread in a single hospital for priority pathogens Nick Thomson Wellcome Sanger Institute, Wellcome Genome Campus, Hinxton, Cambridge, CB10 1SA, UK

Presented will be a seven year retrospective study aiming to understand the differences in the flux in the population structures of two invasive bacteria recognised by the UK Government, WHO and US CDC as priority pathogens due to their ability to become resistant to front line antibiotics: Klebsiella pneumoniae and Enterobacter cloacae. We looked at the relative flux of these bacteria, alongside the antimicrobial resistance genes that they carry, revealing that K. pneumoniae and E. cloacae populations followed markedly different temporal patterns of change in the same hospital. K. pneumoniae showed repeated epidemiological cycles of new highly drug-resistant globally successful lineages sequentially replacing the former dominant lineage. In contrast E. cloacae showed no such temporal pattern, but a continuous representation of the local diversity. For the resistance genes, whilst in K. pneumoniae the resistances could be explained by horizontal acquisition of mobile drug resistance genes, the intrinsic chromosomally encoded beta-lactamase of E. cloacae represented its key resistance mechanism during the study period. Our data emphasizes the importance of ‘bespoke’, species-specific strategies to control multidrug resistant pathogens and the urgent need for a basic scientific understanding of these organisms and the genes they harbour and exchange.

S20

Notes

S21

Epidemic of carbapenem-resistant Klebsiella pneumoniae in Europe is driven by nosocomial spread: Inference from a continent-wide population analysis

Sophia David (1), Sandra Reuter (2), Simon R. Harris (3), Corinna Glasner (4), Silvia Argimon (1), Theresa Feltwell (3), Tommaso Giani (5), Marianne Aspbury (6), Sara Sjunnebo (7), Edward J. Feil (8), Gian Maria Rossolini (5,9), Hajo Grundmann (2,4), David M. Aanensen (1,10)

1. Centre for Genomic Pathogen Surveillance, Wellcome Genome Campus, Hinxton, Cambridge, CB10 1SA, United Kingdom; 2. Institute for Infection Prevention and Hospital Epidemiology, Medical Centre - University of Freiburg, Faculty of Medicine, University of Freiburg, Breisacherstr 115b, 79106 Freiburg, Germany; 3. Pathogen Genomics, Wellcome Sanger Institute, Hinxton, Cambridge, CB10 1SA, United Kingdom; 4. Department of Medical Microbiology, University Medical Centre Groningen, University of Groningen, Hanzeplein 1, 9713 GZ Groningen, The Netherlands; 5. Department of Experimental and Clinical Medicine, University of Florence, Largo Brambilla 3, 50134 Florence, Italy; 6. Natural Sciences, University of Bath, Claverton Down, BA2 7AW, Bath, United Kingdom; 7. Pathogen Informatics, Wellcome Sanger Institute, Hinxton, Cambridge, CB10 1SA, United Kingdom; 8. The Milner Centre for Evolution, Department of Biology and Biochemistry, University of Bath, Claverton Down, BA2 7AW, Bath, United Kingdom; 9. Clinical Microbiology and Virology Unit, Florence Careggi University Hospital, Largo Brambilla 3, 50134 Florence, Italy; 10. Big Data Institute, Li Ka Shing Centre for Health Information and Discovery, Nuffield Department of Medicine, University of Oxford, Oxford, United Kingdom

Public health interventions to control the current epidemic of multidrug-resistant Enterobacterales (Enterobacteriaceae) are reliant upon a comprehensive understanding of the emergence and spread over a wide range of geographical scales. We analysed the genome sequences and epidemiological data of >1700 Klebsiella pneumoniae, isolated from patients treated in 244 hospitals in 32 countries, during the European survey of carbapenemase-producing Enterobacteriaceae (EuSCAPE). We demonstrate that carbapenemase acquisition is the main cause of carbapenem resistance and has occurred across diverse phylogenetic backgrounds including 69 sequence types (ST) of K. pneumoniae sensu stricto. However, carbapenemase-positive isolates are concentrated in four major clonal lineages, STs 11, 15, 101, 258/512, and their derivatives, which together account for 477/682 (69.9%) isolates. We demonstrate a correlation between genetic and geographic distance among carbapenemase-positive isolates, which weakens among isolates with other beta-lactam resistance determinants. More than half of sampled hospitals that contributed carbapenemase-positive isolates likely experienced within-hospital transmission, and inter-hospital spread is far more frequent within a single country than across national borders. We propose a cut-off for the number of single nucleotide polymorphisms (SNPs) that discriminates hospital clusters, and detail the international spread of the most successful epidemic lineage, ST258/512. We here provide population-level evidence that the epidemic of carbapenemase-producing K. pneumoniae in Europe is driven by nosocomial spread of a core of hospital-adapted clones.

S22

Notes

S23

Changing pattern of Carbapenemase-Producing Enterobacteriaceae (CPE) in Hong Kong hospitals, a 7-year overview.

Margaret IP1,5, Chendi Zhu1, Chris K. Lai2, Rita Ng1, Carmen Li1, Sandy K.Y. Chau3, Veranja Liyanapathirana4, Yun Kit Yeoh1, Norman Lo1, Kam Tak Wong1, Nilakshi Barua1, Kitty S.C. Fung3, Mamie Hui1, Raymond Lai1, Dominic N. Tsang2

Dept of Microbiology, Prince of Wales Hospital, Chinese University of Hong Kong1, Dept of Pathology, Queen Elizabeth Hospital2, Dept of Pathology, United Christian Hospital3, Hong Kong, Dept of Microbiology, University of Peradeniya, Sri Lanka, Chinese University Shenzhen Research Institute5, China. Since 2011, a surveillance programme on screening of high-risk patients for carbapenemase-producing Enterobacteriaceae (CPE) has been implemented in major public hospitals in Hong Kong. Our increasing isolation of these strains calls for a comprehensive review on the molecular epidemiology of CPEs in Hong Kong. All Enterobacteriaceae with reduced susceptibility to carbapenem and confirmed with known carbapenemase gene(s) were included. Strains were isolated from both surveillance screening and clinical specimens of patients admitted from three of the seven clusters of hospitals under the Hong Kong Hospital Authority. A total of 571 non-duplicate CPE isolates were saved from 2011 to 2017. Whole genome sequencing was performed on 121 representative strains and the sequence types, plasmids, resistance determinants, virulence profiles and pan-genome trees were examined. Isolation of CPEs increased from one strain in 2011 to 161 and 233 in 2016 and 2017 respectively. 90% of isolates were obtained from surveillance screening or through contact tracing. One-third of cases gave a history of hospitalization abroad, mostly from mainland China, followed by India, Cambodia, Nepal, and Pakistan. The commonest CPEs were E. coli (47%), Klebsiella spp. (39%) and Enterobacter spp. (10%). The most prevalent carbapenemase genes were NDM (53%), IMP (29%), KPC (9%) and were present in all hospitals. The NDM and OXA-48 genes were carried in IncX3 (~50 kb) plasmids in both E. coli and Klebsiella species of multiple ST types. The NDM gene was associated to hospitalization in mainland China. The IMP gene was carried by IncN plasmid (~50 kb) co-harboring qnrS gene, and was present in multiple ST types found predominantly locally in one cluster of hospitals. OXA-48 gene was only detected during 2016/7 while strains co-harboring NDM and OXA-48 in K. pneumoniae were also identified. KPCs predominated in 2014/5 and was associated with ST258 and its variant ST11, carried by IncF plasmid within transposon Tn4401. Of the strains genome sequenced, 83% possessed ESBL genes while all possessed resistance determinants including Aac3-Ib, Qnr, Oqx. The yersiniabactin gene was found in KPC- and NDM-producing K. pneumoniae. The spread of CPEs are mainly due to IncX3 plasmid-mediated transfer of both NDM and OXA-48 genes while IncN plasmid-mediated dissemination of IMP gene in Hong Kong hospitals. The increase in CPE isolation poses great challenge to infection control and management of these infections.Early detection and aggressive implementation of infection control strategies are necessary to prevent spread of CP-CRE in hospitals.

S24

Notes

S25

Meat, urine and time machines

Lance B Price, Cindy M Liu, Maliha Aziz, Marc Stegger

Antibiotic Resistance Action Center, Milken Institute School of Public Health, George Washington University, Washington DC USA & Statens Serum Institut, Copenhagen DK.

Escherichia coli causes >80% of urinary tract infections (UTIs) in the United States (US). E. coli colonizes the gastrointestinal tracts of humans and food animals and frequently contaminates retail meat. Epidemiologic investigations suggest that some UTIs are caused by food-animal E. coli, but directionality is difficult to establish and no studies have quantified the incidence of these infections. We used a combination of core-genome phylogenetics and poultry-associated ColV plasmids to quantify the proportion of UTIs caused by food-animal E. coli in the US. These analyses indicated that 17% of asymptomatic bacteriuria, 12% of cystitis, 9% of pyelonephritis, and 11% of urosepsis cases were caused by food-animal E. coli strains. The antimicrobial susceptibility profiles of food-animal strains from clinical samples shared characteristics with both meat and human strains. Some antimicrobial resistance traits appear to have been lost in the jump from food animals to humans, while others were maintained or gained. Assuming our data are generalizable, we estimate that food-animal E. coli strains could cause 847,000 cases of uncomplicated cystitis, 23,500 cases of pyelonephritis, and 3,538 cases of urosepsis in the US each year. Expanding our knowledge of host-associated accessory elements and combining these with core-genome phylogenetic analyses may allow us to further refine these estimates and approximate the timing of transmission events.

S26

Notes

S27

The challenges of leadership Gwyn Jones Responsible Use of Medicines in Agriculture What has RUMA done in its 21 years and what effect did that have on British livestock sectors during that time? A brief history of the last few years and how things have moved on and the dominance of AMR. RUMA raised its game and changed the way it operates in the last three years or so, why, what effect has it had and where is it today? Where is the UK in terms of responsible use, where are we when compared to other countries in our quest to cut down the amount of antibiotics used in agriculture and in particular, are the various sectors all performing? How do we narrow the gap between the best and the worst? This is not a farmer only issue and everyone needs to play their part. Health and welfare are very important in he UK which adds complications and is not as simple as it is for trading nations who have a clear focus on trade (especially export). For the future, how do we move away from simple number targets which play into the hands of pressure groups? How do we progress to a 'holistic' solution which takes care of all this as a result of doing everything better? How do we move away from AMR dominating our world whilst anthelmintic resistance and endemic disease pose a real threats to our livestock sector.

S28

Notes

S29

One Health genomic surveillance of E. coli reveals separate populations and mobile genetic in humans and livestock

Catherine Ludden, Kathy Raven, Dorota Jamrozy, Theodore Gouliouris, Beth Blane, Francesc Coll, Marcus de Goffau, Plamena Naydenova, Juan Hernandez, Paul Wood, Nazreen Hadjirin, Milorad Radakovic, Nicholas M. Brown, Mark Holmes, Julian Parkhill, Sharon J. Peacock

London School of Hygiene & Tropical Medicine, Keppel Street, London, WC1E 7HT, UK Wellcome Trust Sanger Institute, Wellcome Trust Genome Campus, Hinxton, Cambridge, CB10 1SA, UK Department of Medicine, University of Cambridge, Box 157 Addenbrooke’s Hospital, Hills Road, Cambridge, CB2 0QQ, UK Department of Veterinary Medicine, University of Cambridge, Madingley Rd, Cambridge CB3 0E, UK Royal (Dick) School of Veterinary Studies, University of Edinburgh, Edinburgh, EH25 9RG, UK Clinical Microbiology and Public Health Laboratory, Public Health England, Cambridge CB2 0QQ, UK Cambridge University Hospitals NHS Foundation Trust, Hills Road, Cambridge CB2 0QQ, UK

Background: The increasing prevalence of multi-drug resistant E. coli infections is a serious public health problem. Livestock has been proposed as a source of multidrug-resistant pathogens that infect humans. We conducted an integrated epidemiological and genomic survey in which we compared E. coli isolated from livestock and humans with bloodstream infection. Materials/methods: We sequenced 431 E. coli isolates from livestock (411 from farms, 20 from retail meat) and obtained genomes for 1517 isolates associated with bloodstream infection in the United Kingdom (2001 to 2012). Phylogenetic analyses were used to compare the core (conserved) genome and accessory genes encoding antibiotic resistance and virulence. Mobile elements were detected by extracting contigs carrying antibiotic resistance genes (blaCTX-M-1, blaCTX-M-15, blaTEM-1, sul1, strA, strB, sul2, tetA and tetB) from the assemblies and creating a database from these data. All genomes were then mapped against this database and hierarchical clustering performed on the output. Results: Core genome comparisons demonstrated that livestock and patient isolates were genetically distinct. Screening all 1948 isolates for antimicrobial resistance genes detected 41 genes that were present in variable proportions of human and livestock isolates. The seven most frequently shared genes (each present in >300 isolates) conferred resistance to beta-lactams (blaTEM-1=882), sulphonamides (sul2=530, sul1=522), aminoglycosides (strA=509, strB=478), and tetracyclines (tetA=423, tetB=335). The predominant genes conferring resistance to extended-spectrum cephalosporins were blaCTX-M-15 (human=87, livestock=32) and blaCTX-M-1 (human=1, livestock=82, meat=13). Hierarchical cluster analysis of carriage of contigs encoding these nine genes and their surrounding mobile element revealed that isolates from human and livestock largely resided in distinct clusters. A 10,683bp virulence cassette associated previously with cystitis and neonatal meningitis was identified in 28% and 0% of human and livestock, respectively. Conclusions: Isolates from livestock and patients with bloodstream infection were genetically distinct populations, suggesting that livestock are not the source of E. coli causing severe human infection in our setting. Many antibiotic resistance genes were common to E. coli from livestock and patients, but there was little overlap in the mobile elements that carried these. A known virulence cassette was only detected in disease-associated isolates.

S30

Notes

S31

Superbugs in the supermarket: High prevalence of antibiotic resistances in chicken meat-associated enterococci

Daria Van Tyne, Abigail L. Manson, Tim Straub, Sarah Clock, Michael Crupain, Urvashi Rangan, Michael S. Gilmore, Ashlee M. Earl

University of Pittsburgh School of Medicine; Broad Institute of MIT and Harvard; Department of Ophthalmology, Department of Microbiology and Immunobiology, Harvard Medical School; Department of Ophthalmology, Massachusetts Eye and Ear Infirmary; Roar Biomedical Inc; Johns Hopkins Bloomberg School of Public Health; Rangan Consulting LLC

Factory farms are unique, human-created ecosystems that provide the perfect setting for the development of antibiotic resistance. Agricultural bacterial strains, which are routinely exposed to antibiotics, can serve as a conduit for movement of resistance genes from soil ecologies, where they exist naturally, to hospital-endemic bacteria. To better understand the role of enterococci in the movement of antibiotic resistance from farm to clinic, we surveyed over 300 strains of Enterococcus isolated from raw chicken breast purchased at US supermarkets by Consumer Reports in 2013. E. faecium and E. faecalis were the dominant species detected, and antimicrobial susceptibility testing uncovered striking levels of resistance to medically important classes of antibiotics, particularly those in drug classes approved by the FDA for use as growth promoters in animal and chicken production. Meat labeled as "organic" or "antibiotic-free" had bacteria with fewer phenotypic resistances; however the difference was not statistically significant. We used whole genome sequencing to further study 92 strains of chicken meat-associated enterococci, by analyzing their gene content and relatedness to previously sequenced strains. We observed four instances of E. faecalis chicken meat-associated strains that were closely related to other clinical or commensal E. faecalis strains. We also observed three pairs of closely related chicken meat-associated strains (15 or fewer SNPs in the core genome); in two cases the meat from which the strains were isolated was purchased in different states but had been processed in the same state. Overall we find that both commensal-like and clinical-like strains of E. faecium and E. faecalis are associated with chicken meat, and that some of these strains bear important resistance-conferring genetic elements. The ability of enterococci to persist in the food system positions them as potential vehicles to move resistance genes from the factory farm ecosystem into more human-proximal ecologies.

S32

Notes

S33

The State of the World’s Antibiotics in 2018 Ramanan Laxminarayan The Center for Disease Dynamics, Economics & Policy Declining antibiotic effectiveness is a global health concern affecting both the rich and the poor. Drug resistant bacteria can emerge in any country and spread rapidly everywhere, putting at risk the large population of newborns who suffer from sepsis, mothers who need antibiotics to recover from c-sections and the elderly who need antibiotics to be able to successfully undergo transplants, surgeries and chemotherapy. This talk will cover the current status of antibiotic use and effectiveness worldwide, in human, animal and environmental sectors, and efforts around the world to conserve the antibiotics we have and create the ones that we need.

S34

Notes

S35

The Challenges Of Estimating the Global Burden of AMR Susanna Dunachie Clinical Microbiologist to the GBD-AMR (GRAM) study, Centre for Tropical Medicine and Global Health University of Oxford Estimating the global burden of disease (GBD) caused by antimicrobial resistance (AMR) is highly desirable to establish the current extent of the problem, highlight geographical hotspots, and form a baseline for ongoing monitoring of trends. This information will be used to inform policy makers, funders and governments seeking to prepare to address the problem, as well as scientists researching new approaches to the prevention and management of infections resistant to standard antibiotics. However, creating an estimate of the GBD of AMR with sufficient accuracy to be useful is not without its challenges, and this presentation seeks to outline the key issues alongside potential strategies to mitigate them. The greatest challenge is lack of high quality microbiology data linked to patient information including clinical outcomes. This is especially true for low and middle income countries, where support for microbiology culture laboratories and data collection may be very low or non-existent. The development of methodology to calculate the burden, incorporating key co-variates such as age and comorbidities is crucial, and there are pros and cons of different approaches to calculating the attributable burden of AMR. Other challenges include lack of information for several key areas including population catchment areas for each healthcare facility and antibiotic usage, changing definitions for sepsis and antimicrobial breakpoints, and sampling biases. A systematic approach to estimating the GBD of AMR is needed, and the experience of microbiologists, genomicists, modellers, global health researchers, clinicians and policy makers is essential given the complexity of AMR. Working together as an international research community is critical to build estimates of AMR that are as accurate as they can be, and identification of gaps in knowledge and skills will allow setting of future priorities in research and capacity building.

S36

Notes

S37

Attributable deaths and disability-adjusted life-years caused by infections with antibiotic-resistant bacteria in the European Union and the European Economic Area in 2015: a population-level health estimate

Alessandro Cassini1,2 MSc Liselotte Diaz Högberg1 PhD, Diamantis Plachouras1 PhD, Annalisa Quattrocchi1 PhD, Ana Hoxha1 MSc, Gunnar Skov Simonsen3PhD, Mélanie Colomb-Cotinat4 PhD, Mirjam E. Kretzschmar2, 5 PhD, Brecht Devleesschauwer6, 7 PhD, Michele Cecchini8# PhD, Driss Ait Ouakrim8# PhD, Tiago Cravo Oliveira8# PhD, Marc J. Struelens1 PhD, Carl Suetens1 MD, Dominique L. Monnet1 PhD, and the Burden of AMR collaborative group* 1. European Centre for Disease Prevention and Control (ECDC), Stockholm, Sweden 2. Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, Utrecht, The Netherlands 3. University Hospital of North Norway, and Research Group for Host-Microbe Interaction, Faculty of Health Sciences, UiT – The Arctic University of Norway, Tromsø, Norway 4. Santé publique France, Saint-Maurice, France 5. Centre for Infectious Disease Control, National Institute for Public Health and the Environment (RIVM), Bilthoven, The Netherlands. 6. Department of Epidemiology and Public Health, Sciensano, Brussels, Belgium 7. Department of Veterinary Public Health and Food Safety, Faculty of Veterinary Medicine, Ghent University, Merelbeke, Belgium 8. Organisation for Economic Co-operation and Development (OECD), Paris, France

Background Infections due to antibiotic-resistant bacteria are threatening modern healthcare. Estimating their incidence, complications and attributable mortality is challenging. We estimated the burden of infections caused by antibiotic-resistant bacteria of public health concern in European Union and European Economic Area (EU/EEA) countries in 2015, measured in number of cases, attributable deaths and disability-adjusted life years (DALYs). Methods The incidence of infections with 16 antibiotic resistance-bacterium combinations was estimated from European Antimicrobial Resistance Surveillance Network 2015 data, country-corrected for population coverage. The number of bloodstream infections (BSIs) was multiplied by a conversion factor derived from the European Centre for Disease Prevention and Control point prevalence survey of healthcare-associated infections in European acute care hospitals 2011–2012 to estimate the number of non-BSIs. Disease outcome models for five types of infection were developed based on systematic reviews of the literature. Findings We estimated 671,689 (95% Uncertainty Interval [UI] 583,148-763,966) infections with antibiotic-resistant bacteria of which 63.5% (426,277/671,689) were healthcare-associated. These infections accounted for estimated 33,110 (95% UI 28,480-38,430) attributable deaths and 874,541 (95% UI 768,837-989,068) DALYs. The burden for the EU/EEA was the highest in infants and the elderly, increased since 2007 and was highest in Italy and Greece. Interpretation This study presents the health burden of five types of infection with antibiotic-resistant bacteria, for the first time expressed in DALYs. The EU/EEA burden was comparable to that of influenza, tuberculosis and HIV cumulatively, and heterogeneous across countries. These burden estimates represent useful information for public health decision-makers for prioritizing infectious diseases and interventions.

S38

Notes

S39

Developing a geospatial modelling framework to estimate the burden of antimicrobial resistance Annie J. Browne Big Data Institute, Li Ka Shing Centre for Health Information and Discovery, University of Oxford. Antimicrobial resistance (AMR) has been identified as one of the major threats to human health in the 21st century, with common infections potentially rendered untreatable. However, very little is known regarding the geographic distribution and prevalence of resistance in key pathogens in the present and recent past. The Global Research on Antimicrobial Resistance (GRAM) project, a collaborative effort between the University of Oxford and the Institute of Health Metrics and Evaluations (IHME), aims to exploit techniques in machine learning and geospatial modelling to produce spatially and temporally explicit estimates of the prevalence of AMR in a selection of high priority microbial pathogens. Starting with Salmonella Typhi, we are compiling a comprehensive database of the prevalence of resistance to key antimicrobials. Building on methods successfully used to model a range of health outcomes; we will use this data to predict the prevalence of AMR at a high spatial resolution. Briefly, input data will be linked to a suite of socio-demographic and health related covariates and a stacked generalisation ensemble model will be applied to capture associations between covariates and data. A Bayesian model based geostatistics (MBG) model will then be fit to the data to account for the remaining spatial and temporal correlation and produce pixel level estimates of the prevalence of resistance in each pathogen. AMR provides additional challenges to be addressed prior to implementing these models. Most notably, any measure of antibiotic consumption is often missing from the covariates available to inform the model. As this is thought to be an important driver resistance, we are producing fine scale spatial-temporal estimates of antibiotic consumption in humans in lower and middle income countries (LMICs). Data on antibiotic use have been leveraged from household surveys and will be modelled using the above framework. National data on antibiotic sales will be incorporated into this model to produce estimates of key antibiotic consumption. This will then be used as a covariate to inform the model on the prevalence of resistance. This project is in its infancy but accurate estimations of the prevalence of resistance are vital to enable predictions of the burden of disease attributable to AMR. Understanding the scale and distribution of the issue can greatly aid allocation of resources to combat this threat.

S40

Notes

S41

Global surveillance of antimicrobial resistance through sewage samples and functional resistance

Pimlapas Leekitcharoenphon, Frank M. Aarestrup

National Food Institute, Technical University of Denmark, Denmark

Antimicrobial resistance (AMR) is a serious global public health threats and it threatens to undermine decades of progress in the treatment of infectious diseases. AMR is a complex problem with multiple and interconnected drivers. Current surveillance of AMR is patchy and mainly based on passive reporting of phenotypic laboratory results for specific pathogens isolated from human clinical infections. From a surveillance point of view, urban sewage is attractive because it provides sampling material from a large and mostly healthy population, which otherwise would not be feasible to monitor. We characterized the bacterial resistome from untreated sewage from 79 sites in 60 countries across the globe in 2016. The sewage samples were sequenced on the HiSeq3000 platform (Illumina). The metagenomics sequences were mapped against two resistance genes databases using MGmapper pipeline. The two databases are ResFinder version 3 (3026 genes) and antimicrobial resistance determinants (ARDs, 2514 genes) identified from four functional metagenomics studies. Abundance of AMR genes was normalized into fragments per kilobase reference per million bacterial fragments (FPKM). We found strong correlation of abundance of AMR genes from ResFinder and ARDs databases. There were systematic differences in abundance and diversity of AMR genes between Europe/North-America/Oceania and Africa/Asia/South-America. The similar differences were observed in abundance of bacteria between Europe/North-America/Oceania and Africa/Asia/South-America. The association between antimicrobial use (AMU) and the abundance of AMR genes in the sewage samples was determined using national level of AMU data in 2015 from Europe (www.ecdc.dk) and IQVIA. The result showed that AMU only correlated with abundance of AMR genes from both databases in a specific antimicrobial class indicating that antimicrobial use data explained only a minor part of the AMR abundance and there are still other unexplainable factors that drive the abundance of AMR genes. This study showed that sewage is a good source for an ethically acceptable and economically feasible continuous global surveillance and prediction of antimicrobial resistance and infectious diseases.

S42

Notes

S43

See and Sequence: Whole Genome Sequencing for the Surveillance of Antimicrobial Resistance in the Philippines

Silvia Argimón1, Melissa A. Masim2, June Gayeta2, Victoria Cohen1, Marietta Lagrada2, Agnetta Olorosa2, Polle Macaranas2, Benjamin Jeffries1, Lara Fides Hernandez2, Karis L. Del Castillo2, Manuel Jamoralín2, Marilyn Limas2, Jeremiah Chilam2, Charmian Hufano2, Sonia Sia2, John Stelling3, Celia Carlos2, David M. Aanensen1,4, and the Philippines Antimicrobial Resistance Surveillance Program

1 Centre for Genomic Pathogen Surveillance, Wellcome Genome Campus, Hinxton, UK. 2 Antimicrobial Resistance Surveillance Reference Laboratory, Research Institute for Tropical Medicine, Muntilupa, Philippines. 3 School of Medicine, Brigham and Women's Hospital, WHO Collaborating Centre for Surveillance of Antimicrobial Resistance, Boston, MA, USA. 4 Big Data Institute, University of Oxford, Oxford, UK.

The Philippines Antimicrobial Resistance Surveillance Program (ARSP) consists of 25 sentinel laboratories that routinely capture antimicrobial susceptibility data for a number of bacterial pathogens and front line antibiotics via WHONET. Notifiable resistant strains are sent to the national reference laboratory for confirmation and storage, where we implemented whole genome sequencing (WGS) for prospective surveillance and built capacity for the analysis of genomic data. We characterized the population structure of priority pathogens (Pseudomonas aeruginosa, Acinetobacter baumannii, Klebsiella pneumoniae, Escherichia coli, Salmonella spp, Staphylococcus aureus, Neisseria gonorrhoeae) at the national level, thus establishing a genomic background for prospective surveillance. Contextualization with global collections of public genomes revealed interesting examples of international phylogeographic patterns for some of the pathogens. The New Delhi Metallo-beta-lactamase (NDM) is the main driver of resistance to carbapenems in K. pneumoniae and E. coli, shuttled within variations of the ISAba125 context in diverse plasmids that, together with the core SNP phylogenies, revealed local and regional patterns of circulation of lineages that molecular typing methods alone could not show. The extended-spectrum beta-lactamase (ESBL) CTX-M, was virtually ubiquitous in genomes of the K. pneumoniae and E. coli isolates referred as ESBL-producing. We also identified examples of inter-species transfer of ESBL and carbapenemase genes. Furthermore, the provision of the contextual background datasets through web applications that link genomic and epidemiological data led to the identification of a previously undetected high-risk clone persisting in a neonatal intensive care unit, which prompted further investigation and intervention. In conclusion, we provide examples of the utility of WGS to understand the dynamics of high-risk clones in the Philippines, and of intervention measures ensuing from this analysis to prevent the spread of lineages of public health concern.

S44

Notes

S45

Transmission dynamics and between-species interactions of multidrug-resistant Enterobacteriaceae

Thomas Crellen (1,2), Paul Turner (2,3), Sreymom Pol (2,3), Stephen Baker (2,4), To Nguyen Thi Nguyen (4), Nicole Stoesser (2), Nicholas P.J. Day (1,2), Ben S. Cooper (1,2)

1) Mahidol-Oxford Tropical Medicine Research Unit, Faculty of Tropical Medicine, Mahidol University, 60th Anniversary Chalermprakiat Building, 420/6 Rajvithi Road, Bangkok 10400, Thailand 2) Nuffield Department of Medicine, University of Oxford Henry Wellcome Building for Molecular Physiology, Old Road Campus, Headington, Oxford OX3 7BN 3) Cambodia-Oxford Medical Research Unit, Angkor Hospital for Children, Siem Reap, Cambodia 4) Oxford University Clinical Research Unit, Centre for Tropical Medicine, 764 Vo Van Kiet, Quan 5, Ho Chi Minh City, Viet Nam

Widespread resistance to antibiotics is among the gravest threats to modern medicine, and controlling the spread of multi-drug resistant Enterobacteriaceae has been given priority status by the WHO. Interventions to reduce transmission within hospital wards may be informed by modifiable patient-level risk factors for becoming colonised, however understanding of factors that influence a patient's risk of acquisition is limited. We analyse data from a one year prospective carriage study in a neonatal intensive care unit in Cambodia using Bayesian hierarchical models to estimate the daily probability of acquiring multi-drug resistant organisms, while accounting for patient-level time-varying covariates, including interactions between species, and interval-censoring of transmission events. We estimate the baseline daily probability for becoming colonised with third generation cephalosporin resistant (3GC-R) Klebsiella pneumoniae as 0.142 (95% credible interval [CI] 0.066, 0.27), nearly ten times higher than the daily probability of acquiring 3GC-R Escherichia coli (0.016 [95% CI 0.0038, 0.049]). Prior colonization with 3GC-R K. pneumoniae was associated with a greatly increased risk of a patient acquiring 3GC-R E. coli (odds ratio [OR] 6.4 [95% CI 2.8, 20.9]). Breast feeding was associated with a reduced risk of colonisation with 3GC-R K. pneumoniae (OR 0.73 [95% CI 0.38, 1.5]) and E. coli (OR 0.62 [95% CI 0.28, 1.6]). The use of an oral probiotic (Lactobacillus acidophilus) did not show clear evidence of protection against colonization with neither 3GC-R K. pneumoniae (OR 0.83 [95% CI 0.51, 1.3]) nor 3GC-R E. coli (OR 1.3 [95% CI 0.77, 2.1]). Antibiotic consumption within the past 48 hours did not strongly influence the risk of acquiring 3GC-R K. pneumoniae. For 3GC-R E. coli, ceftriaxone showed the strongest effect for increasing the risk of acquisition (OR 2.2 [95% CI 0.66, 6.2]) and imipenem was associated with a decreased risk (OR 0.31 [95% CI 0.099, 0.76). Using 317 whole-genome assemblies of K. pneumoniae, we determined putatively related clusters and used a range of models to infer transmission rates. Model comparison strongly favored models with a time-varying force of infection term that increased in proportion with the number of colonized patients, providing evidence of patient-to-patient transmission including among a cluster of Klebsiella quasipneumoniae. Our findings provide support for the hypothesis that K. pneumoniae can be spread person-to-person within ward settings. Subsequent horizontal gene transfer within patients from K. pneumoniae provides the most parsimonious explanation for the strong association between colonization with 3GC-R K. pneumoniae and acquisition of 3GC-R E. coli.

S46

Notes

S47

Public health surveillance of antimicrobial resistance in gastrointestinal pathogens

Claire Jenkins

Public Health England, UK

The accessibility of whole genome sequencing (WGS) presents the opportunity for state-of-the-art public health surveillance of antimicrobial resistance (AMR) in gastrointestinal (GI) pathogens. Since 2015, all GI pathogens submitted to Public Health England are sequenced and genome-derived AMR profiles are determined using Genefinder which maps the sequencing reads to a set of reference sequences. Comparisons between phenotypic antimicrobial sensitivity testing (AST) and genome-derived AMR profiles for diarrhoeagenic Escherichia coli, Campylobacter, Salmonella and Shigella species showed >95% concordance between the two methods. Multidrug resistance, and resistance to 3rd generation cephalosporins and the fluroquinolones, are mostly detected in GI pathogens isolated from patients reporting recent travel abroad, and most report travel to the Indian sub-continent. Since 2015, a dramatic increase in resistance to ciprofloxacin has been observed, and MDR is increasingly detected in strains associated with domestically-acquired infection. Although phenotypic AST continues to be performed on a sub-set of isolates to monitor for emerging novel resistance determinant, genome-derived AMR profiling provides a comprehensive, cost effective, robust and real-time approach to monitoring trends in MDR in GI pathogens.

S48

Notes

S49

Time to move on from 150 years of shoe leather epidemiology for outbreak detection Sharon Peacock London School of Hygiene and Tropical Medicine The current paradigm for hospital outbreak detection and investigation is based on methodology first developed over 150 years ago. Daily surveillance to detect patients positive for pathogens of particular importance for nosocomial infection is supported by epidemiological investigation of their relationship in time and place, and to identify any other factor that could link them. The antibiotic resistance pattern is commonly used as a surrogate for bacterial relatedness, although this lacks sensitivity and specificity. Typing may be used to define bacterial relatedness although routine methods lack sufficient discriminatory power to distinguish relatedness beyond the level of bacterial clones. Ultimately, the identification of an outbreak remains a predominately subjective process reliant on the intuition of experienced infection control professionals. In this talk, I will propose a fundamental redesign of hospital outbreak detection and investigation in which bacterial species associated with nosocomial transmission and infection undergo routine prospective whole genome sequencing. Further investigation is triggered by the probability that isolates are associated with an outbreak, as calculated from the degree of genetic relatedness between isolates. Evidence will be provided that supports this model based on studies of MRSA (methicillin-resistant Staphylococcus aureus), together with the benefits of a ‘Sequence First’ approach. The feasibility of implementation is discussed, together with residual barriers that need to be overcome prior to implementation.

S50

Notes

S51

Defining a relatedness cut-off between MRSA isolate genomes that predicts the probability of an outbreak

Francesc Coll, Francesc Coll1*#, Kathy E. Raven2*, Beth Blane2, Danielle Leek2, David Enoch3, Nicholas M. Brown3, Julian Parkhill4, Sharon J. Peacock1,2,4

1 London School of Hygiene and Tropical Medicine, UK.2 University of Cambridge, UK. 3Public Health England, UK. 4Wellcome Sanger Institute, UK. *Contributed equally #Presenting author

Whole-genome sequencing (WGS) of MRSA has emerged as a potentially transformative tool for the detection of outbreaks in hospitals and the community. WGS can be applied to confirm suspected outbreaks or to rule them out, allowing infection control interventions to be targeted or reduced, respectively. The genetic distance between any two MRSA isolates obtained from different patients during an outbreak will be the combined result of single nucleotide polymorphisms (SNPs) accumulated over time (mutation rate) and individual cloud of diversity (genetic variation of the same S. aureus clone in the same person), which itself is a function of carriage/infection duration. Here, we report the findings of a study that proposes a genetic relatedness cut-off (SNPs in the core-genome) between any two MRSA isolates, which could be used to support future outbreak investigations. The results were derived from applying two different approaches to independent cohorts. First, we used a cohort of 340 patients with multiple MRSA isolates sequenced per patient (total 1,152 isolates) processed by the Clinical Microbiology and Public Health Laboratory at the Cambridge University Hospitals NHS Foundation Trust (CUH) from April 2012 to April 2013. A linear mixed regression model estimated a substitution rate of 5 SNPs per genome per year, equivalent to published studies, and a cloud of diversity of 18 SNPs (95 percentile), which led to a SNP cut-off of 23 SNPs to capture MRSA transmission within 6 months. We used a second cohort of consecutive individuals with MRSA-positive samples processed at CUH between February 2018 and July 2018, and investigated strong epidemiological links, that is patients who had shared the same ward, general practice or postcode. We next investigated the relationship between the strength of epidemiological links between patients and the genetic relatedness (number of SNPs) of their MRSA isolates and found strong epidemiological links between patients with isolates differing by no more than 24 SNPs. In summary, the modelled SNP cut-off calculated using MRSA isolates from the same patient and that calculated from integrating genomic and epidemiological data between patients in an independent cohort were comparable.

S52

Notes

S53

Implementing Genomic-based AMR Surveillance From The Ground Up

Dr Anthony Underwood, Silvia Argimon 1, Monica Abrudan 1, Celia Carlos 2, Pilar Donado-Godoy 3, Harry Harste 1, Mihir Kekre 1, Ravi Kumar 4, Dawn Muddyman 1, Iruka N. Okeke 5, John Stelling 6, David M Aanensen 1,7

1. Centre for Genomic Pathogen Surveillance, Wellcome Sanger Institute, Cambridge, UK 2. Research Institute for Tropical Medicine, Manilla, Philippines 3. Agrosavia, Bogata, Colmbia 4. Kempegowda Institute of Medical Sciences (KIMS), Bangalore, India 5. University of Ibadan, Nigeria 6. Brigham & Women's Hospital, Boston, USA 7. Big Data Institute, University of Oxford, UK

There are demonstrable advantages to supplementing phenotypic-based antimicrobial resistance (AMR) surveillance within a country with genomic sequencing of selected isolates. The extra data provided by sequencing allows:

Determination of the origin of AMR isolates found in a country

Investigation into the dynamics of gene flow related to AMR in order to test hypotheses about possible inter lineage and inter species horizontal gene transfer.

Provision of contextual data for ongoing genomic surveillance However, in order to take advantage of this emerging field, there are considerable barriers to implementation of genomics for institutes or organisations, and national networks.. The NIHR Global Health Research Unit (GHRU) for Genomic Surveillance of Antimicrobial Resistance (http://ghru.pathogensurveillance.net) is working with partners in four countries to develop templates and tool kits that are implementing a straightforward pathway to genomics capacity focussed, initially, on AMR. This capability will include:

Collation of metadata pertaining to AMR in electronic form with a minimum required set of fields and centralised collation and aggregation of the data.

Setup and maintenance of laboratories suitable for local pathogen genomics and high-throughput susceptibility testing including - Procurement of relevant equipment and reagents - Methods for tracking samples through a simple LIMS system - Production of SOPs that define standardised methodologies whilst building in flexibility to allow for inter-laboratory variability - IQA methods to check the success of the execution of the SOPs

Accessible technologies that permit bioinformatics analysis of sequence data including - Deployment of computing infrastructure - Simple installation of software required for analyses - Batch processing of multiple samples - Development of local bioinformatics expertise

Provision of large-scale genomic surveys of WHO priority pathogens to contextualise local prospective sequencing.

Web-based tools to deliver bioinformatic outputs for laboratory scientists, clinical staff and public health officials to interpret the genomic data linked to epidemiology.

Financial and project management guidelines to achieve Good Financial Grant Practice (GFGP) We will present challenges encountered, considerations necessary and initial solutions implemented in the early phases of the GHRU.

S54

Notes

S55

Training and validation of a novel 30-mRNA panel, HostDx Sepsis, for diagnosing and prognosing acute infections and sepsis Timothy E Sweeney Inflammatix, Inc

In 2017, sepsis was re-defined as “life-threatening organ dysfunction caused by a dysregulated host response to infection.” Thus, a full clinical diagnostic for sepsis must contain at least two pieces of information, namely, (1) the presence of infection that induces a systemic host immune response, and (2) dysregulation of host protective immunity leading to organ injury or dysfunction. Unfortunately, current diagnosis of infection relies on high-cost testing of multiple sites (blood, sputum, BAL, urine), and still fails to find pathogens in 40-50% of cases of sepsis. While new pathogen-sensing diagnostics are growing in use, they are limited to detecting a discrete range of pathogens and cannot discriminate colonization from infection. They are also all hampered by technical limitations that increase turnaround time.

Inflammatix has a novel solution, an assay we call HostDx Sepsis, which can ‘read’ patterns in the immune response to detect infection and sepsis early, without the need for a pathogen-specific identification test, e.g. culture. The test interprets 30 host mRNAs from whole blood, the patterns of which reflect the body’s first response to infection. HostDx Sepsis has three separate outputs that are reported simultaneously to the physician: (1) the likelihood of a bacterial infection; (2) the likelihood of a viral infection, and (3) the risk of 30-day mortality. HostDx Sepsis uses a neural network to interpret its 30 component genes into the separate likelihoods of infection and mortality. We trained this model across 18 co-normalized datasets of acute infections and sepsis (N=1,092), and showed AUROCs of 0.91 and 0.90, respectively, for the identification of clinician-adjudicated bacterial and viral infections. Next, we used the NanoString quantitative mRNA platform to measure HostDx Sepsis in prospective pilot data from two cohorts (patients with one SIRS criteria in ED, and patients at admission to ICU for sepsis). We applied our pre-fixed algorithm and cutoffs of HostDx Sepsis in these prospective cohorts and found excellent stability of AUROCs, with calibrated outputs at individual probability bands. Our third output that predicts 30-day mortality was shown, when combined with SOFA score, to boost AUROC for 30-day mortality prediction by 0.08, consistent with prior findings. While more study is warranted, this suggests that HostDx Sepsis could allow for more judicious antibiotic prescribing in hospital settings of suspected infection and sepsis.

S56

Notes

S57

Antibiotic stewardship in the context of the human microbiome

Debby Bogaert

The University of Edinburgh, Centre for Inflammation Research

Antibiotics are key to the battle against infectious diseases by their capacity to completely

block or attenuate microbial growth. Their use is widely accepted and has saved many lives.

Still, antibiotic treatment has obvious downsides, like allergic reactions and eradication of

more species than the pathogen targeted, thereby affecting the so-called microbiota,

including potentially its beneficial health effects. In addition to the downsides listed above,

the use of antimicrobial drugs results in selection for less susceptible and even resistant

microorganisms. Antibiotic treatment, selection for drug resistance, and negative effects on

microbiota are, however, interdependent and therefore deserve further integrated

considerations. Currently, research into antibiotic resistance and antibiotic stewardship

rarely takes aspects of microbiota characterization and (antibiotic-induced) microbiota

alterations into account, although this would be a highly rational way forward: first of all

because of the natural pool of antimicrobial resistance genes (i.e., the resistome) within the

microbiome. Second, the microbiome harbors and promotes intrinsic mechanisms to

exchange resistance genes towards (potentially) pathogenic bacteria when under ecological

pressure. Third, microbiota provide indirect resistance against acquisition of, and

overgrowth and invasion by drug-resistant bacteria. This suggests that development and

adaptation of antibiotic treatment towards being less detrimental to the microbiota is

desirable. During this talk, I will provide examples of fundamental and clinical study results

that may help to broaden our vision regarding antibiotic stewardship.

S58

Notes

S59

Genome-wide association uncovers novel candidate resistance and compensatory mutations in antibiotic-resistant Neisseria gonorrhoeae

Kevin C Ma (1), Nicole E Wheeler (2), Leonor Sánchez-Busó (2), Yi Wang (1), Xihong Lin (3), Simon R Harris (2), Yonatan H Grad (1,4)

(1) Department of Immunology and Infectious Diseases, Harvard T.H. Chan School of Public Health, Boston, USA (2) Infection Genomics, Wellcome Sanger Institute, Hinxton, UK (3) Department of Biostatistics, Harvard T.H. Chan School of Public Health, Boston, USA (4) Division of Infectious Diseases, Brigham and Women’s Hospital and Harvard Medical School, Boston, USA

Introduction: The emergence of resistance to the front-line antibiotic combination of azithromycin and ceftriaxone complicates treatment of N. gonorrhoeae, the etiologic agent of gonorrhea. Genome sequencing of clinical isolates has identified genetic markers, such as recombination at the penA locus for ceftriaxone and mutations in 23s rRNA for azithromycin, as highly associated with increased minimum inhibitory concentrations (MICs). However, a subset of resistance remains unexplained, and the emergence of resistance across the phylogeny is not uniform, indicating that additional loci may affect either the acquisition of or compensation for resistance mutations. To systematically identify novel resistance and compensatory mutations in an unbiased manner, we conduct a bacterial genome-wide association study (GWAS) that incorporates both pan-genomic diversity and stringent control for population structure. Methods: Our dataset comprises 1102 previously published isolates collected through the CDC's Gonococcal Isolate Surveillance Project (GISP) (Grad et al., 2016). Bacterial GWAS was conducted using kmers generated from de novo assembled genomes and a linear mixed model in GEMMA. Results: Kmers highly associated with increased azithromycin MICs mapped to mutations in 23s rRNA and to horizontally acquired alleles in the mtr efflux pump, both of which are known macrolide resistance determinants. Conducting GWAS conditional on known loci identified a novel G70D mutation in the 50S ribosomal protein L4 (rplD) as associated with increased MICs. Structural analysis indicates that this residue contacts the azithromycin binding pocket and laboratory validation provides support for its role in conferring resistance. Kmers highly associated with increased ceftriaxone MICs mapped to the known mosaic penA allele XXXIV, with additional kmers mapping to residues in the porin porB. Lineage effects analysis of clades enriched in ceftriaxone reduced susceptibility identified a Q371K mutation in a conserved aconitase hydratase gene (acnB) associated exclusively with the dominant gonococcal lineage that has acquired mosaic penA XXXIV and persisted. Conclusions: Bacterial GWAS recovers known resistance mechanisms for azithromycin and ceftriaxone and identifies a novel resistance mutation in rplD for azithromycin. Lineage effects analysis identifies a candidate mosaic penA compensatory mutation in acnB, a gene previously implicated in gonococcal mouse models as having a compensatory role for laboratory strains transformed with mosaic penA alleles (Vincent et al., 2018). Further work to expand the dataset and to conduct in vivo characterization of these mutations is underway. We conclude that genome-wide association offers a powerful method for interrogating the biological basis of emerging resistance in N. gonorrhoeae.

S60

Notes

S61

Characterising beta-lactam antibiotic resistance in healthy human gut microbiota

Lindsay J. Pike(1),, Samuel C. Forster(1,2,3), Mark Stares(1), Nitin Kumar(1), Stephen Baker(4,5,6), Trevor D. Lawley(1)

1 Host-Microbiota Interactions Laboratory, Wellcome Sanger Institute, Hinxton, CB10 1SA, UK, 2 Centre for Innate Immunity and Infectious Diseases, Hudson Institute of Medical Research, Clayton, Victoria, Australia, 3 Department of Molecular and Translational Sciences, Monash University, Clayton, Victoria, 3800, Australia 4 The Hospital for Tropical Diseases, Wellcome Trust Major Overseas Programme, Oxford University Clinical Research Unit, Ho Chi Minh City, Vietnam. 5 Centre for Tropical Medicine and Global Health, Nuffield Department of Clinical Medicine, Oxford University, Oxford, UK. 6 The Department of Medicine, University of Cambridge, Cambridge, United Kingdom

Antibiotics are a cornerstone of modern medicine and used extensively throughout developed and developing countries in both human and animal healthcare. They are essential to combatting pathogenic infection and have saved millions of lives. However, consumption of antibiotics creates selective pressures on bacteria not directly targeted, resulting in the development of undesirable antimicrobial resistance. This is a major global issue and extensive research is required to mediate its effects. Recent studies indicate the human intestinal microbiome as a reservoir of antibiotic resistance as there are multiple opportunities for antibiotic resistance genes to transfer between the intestinal microbiota and any pathogens present in the intestinal tract. Research into antibiotic resistance and publicly available databases of known antibiotic resistance mechanisms typically focus on clinically relevant bacteria. However, understanding the extent of antibiotic resistance in the human gut microbiota, particularly that of beta-lactam antibiotics, some of the most important and widely-prescribed drugs in human medicine, will assist antibiotic-resistance surveillance programmes and provide a broader understanding of the development of antibiotic resistance. In this study, we aimed to assess the antibiotic resistance potential of the commensal human gut microbiota by leveraging a diverse bacterial culture collection with 737 isolates combined with genomic screens. In vitro screening using antibiotic gradient Etests of 48 bacteria representing phylogenetically diverse human intestinal microbiota against nine antibiotics identified 30 'false-positives'. This means resistance was predicted using a typical whole-genome sequencing antibiotic resistance prediction method (such as ARIBA with the Comprehensive Antibiotic Resistance Database) but the isolates were phenotypically susceptible. Conversely, 59 'false-negatives' - instances of 'unpredicted' phenotypic resistance in the absence of any known genetic determinant of antibiotic resistance were identified. We focused on beta-lactams due to the importance of these drugs and identified two isolates with unpredicted high-level ceftriaxone resistance (growth at the highest concentration on the Etest). These two isolates were among the five most resistant organisms and demonstrated other unpredicted antibiotic resistance, highlighting them as candidates for further analysis. We show that combining in vitro screening and genomics reveals extensive phenotypic resistance in commensal gut bacteria that is missed through genomic or metagenomic prediction-based methods alone. This suggests genomic surveillance of antibiotic resistance in faecal samples or other environmental samples such as wastewater may under-predict the full AMR potential of that microbial community. Further analysis to identify genomic explanations of unpredicted resistance may find novel antibiotic resistance mechanisms that could help improve antibiotic resistance surveillance programmes.

S62

Notes

P1

Poster Presentations

Klebsiella pneumoniae ST22: Molecular characterisation of antimicrobial resistance using whole genome sequencing

Shamsudin Aliyu¹, Olanrewaju Jimoh¹, Abdulrasul Ibrahim¹, Yaqub Yahaya², Mukhtar Shehu Idris², Adebola Tolulope Olayinka¹.

Department of Medical Microbiology, Ahmadu Bello University, Zaria, Kaduna State, Nigeria¹. Department of Medical Microbiology, Ahmadu Bello University Teaching Hospital, Zaria, Kaduna State, Nigeria².

The emergence of carbapenem-resistant enterobacteriaceae (CRE) has become a serious threat to global public health. In particular, infections caused by carbapenem-resistant Klebsiella pneumoniae have become a major source of concern, because of limited treatment options. Whole genome sequencing has become a very important tool in the characterisation of antimicrobial resistance in bacterial pathogens. In this study, we aimed to detect and identify molecular determinants of antimicrobial resistance, through whole genome sequencing of a rare multidrug resistant strain of Klebsiella pneumoniae. The pathogen was isolated from the blood of a neonate that had been managed for sepsis, in Ahmadu Bello University Teaching Hospital, Kaduna State, Nigeria. Paired-end sequencing libraries were prepared and sequenced on an Illumina MiSeq® platform. Sequence assembly was performed using the CLC Genomics workbench. Bioinformatic pipelines from the Center for Genomic Epidemiology (CGE) were used to identify acquired antimicrobial resistance genes and plasmids, as well as to carry out Multilocus Sequence Typing (MLST). Additionally, Comprehensive Antibiotic Resistance Database (CARD) was also used to identify other resistance determinants. The final draft genome sequence consists of a combined 5,313,367 bases, with a G+C content of 54.23 %. Multilocus sequence typing indicated that the isolate belonged to an unusual sequence type (ST) 22. Whole genome sequencing analysis revealed that the isolate harboured different β-lactamase genes, including blaOXA−1, blaCTX−M−15, blaNDM−5, and blaTEM−1b. The isolate was also characterized by the concomitant presence of other resistance determinants, most notably acc(6′)-lb-cr, acc(3′)-lb-5, aph(6)-ld_1, qnrB1_1, tet(A)_6, oqxA, dfrA14_1, CRP, fosA_3, and Klebsiella pneumoniae ompK37. The bacterium was also found to carry four plasmids that were homologous to IncX3, IncFIB(K), IncFIB(pKPHS1), IncFII(K) plasmids. This Klebsiella pneumoniae ST22 strain harboured many antimicrobial resistance genes, some on plasmids; that made it resistant to many antibacterial agents from various drug classes. The ability of this pathogen to easily acquire and spread antimicrobial resistance genes, is a clear cause of concern. Whole genome sequencing will continue to play a vital and expanding role in identifying and characterising antimicrobial resistance in bacterial pathogens.

P2


Derya Aytan Camilla Hundahl Johnsen, Frank M. Aarestrup, and Ole Lund


Extensive use of antimicrobials has driven the progressive emergence of resistant bacteria and other microbes. Surveillance of antimicrobial resistance and rapid detection of resistance is important to facilitate an effective treatment, and hence reduce drug mis-usage. Additional rapid detection is essential as resistance call for extra attention and precaution. Next-generation sequencing technologies rapidly yield accurate whole-genome sequencing (WGS) data, from which antimicrobial resistance might be directly determined. At Center for Genomic Epidemiology we aim to provide accurate bioinformatic solutions for determining resistance phenotypes. We have developed both ResFinder and PointFinder, in silico methods for predicting resistance based on acquired resistance genes or chromosomal mutations from WGS data. In this study, we focus on improving our PointFinder tool. PointFinder is a rapid, assembly-free program that detects point mutations, insertions and deletions. Resistance is predicted when any detected mutation is present in our curated database of resistance associated mutations. PointFinder is available for several species, including M. Tuberculosis (Mtb). We optimized the Mtb resistance prediction by feature selection methods, such as forward selection, ReliefF and L1 Norm Support Vector Machine methods. We also found that the literature on Mtb resistance mutations does contain misclassified resistance markers, and by elimination those from our model we improved the prediction performance. Traditionally, the occurrence of a resistance mutation has been treated as a direct indicator for resistance. However, it is now suspected that some mutations need the existence of other mutations to cause drug resistance, and more complex mutation patterns might explain the resistance emergence. Therefore, we have initiated a machine learning study to predict resistance by random forest and neural network prediction models. Pending results using these approaches will show how linked mutation might account for resistance that could not have been explained before.

P3

Exploring the link between carbapenemase-producing Pseudomonas spp. and ICE

João Botelho, João Botelho1, Adam P. Roberts2, Ricardo León-Sampedro3, Filipa Grosso1, Luísa Peixe1

1UCIBIO/REQUIMTE, Laboratory of Microbiology, Porto, Portugal; 2Liverpool School of Tropical Medicine, Liverpool, United Kingdom; 3Ramon y Cajal Health Research Institute (IRYCIS); CIBER-ESP, Madrid, Spain

The evolution and spread of antibiotic resistance is often mediated by mobile genetic elements. Integrative and conjugative elements (ICE) are the most abundant conjugative elements among prokaryotes. However, the contribution of ICE to horizontal gene transfer of antibiotic resistance has been largely unexplored. A worldwide collection of Pseudomonas genomes (n=4565) was retrieved from NCBI and blasted against a local database of carbapenemases. Hits with carbapenemase-encoding genes (CEG) were further inspected for the presence of ICE. Here we report that ICE belonging to mating-pair formation (MPF) classes G and T are highly prevalent among the opportunistic pathogen Pseudomonas aeruginosa, contributing to the spread of CEG. Most CEG of the MPFG class were encoded within class I integrons, which co-harbour genes conferring resistance to other antibiotics. The majority of the integrons were located within Tn3-like and composite transposons. A conserved attachment site could be predicted for the MPFG class ICE. MPFT class ICE carried the CEG within composite transposons which were not associated with integrons. The data presented here provides a global snapshot of the different CEG-harbouring ICE and sheds light on the underappreciated contribution of these elements for the evolution and dissemination of antibiotic resistance on P. aeruginosa.

P4

Metagenomic insights into the resistome of the human oral cavity

Victoria Carr, Elizabeth Witherden, David Moyes, David Gomez-Cabrero

Centre for Host-Microbiome Interactions

The microbiota in the human oral cavity is more frequently exposed to environmental stressors than any other human microbiota niches. In order for microbes to proliferate in the oral cavity, these microbes and their interactions are required to adapt to these stressors. These adaptations are facilitated by frequent horizontal gene transfer (HGT) of mobile genetic elements (MGEs) containing genes that may express potentially advantageous phenotypes, such as resistance to an antimicrobial drug or product. Thus, the human oral cavity is an ideal body site for profiling a range of acquired, as well as intrinsic, antimicrobial resistance genes (ARGs). Here, I perform the first large-scale study into the resistome of the human oral cavity using shotgun metagenomic sequencing data from hundreds of human oral cavities (and comparative stool samples) from China, the US, the Philippines and the UK. ARGs were identified by aligning and annotating assembled metagenomes against the Comprehensive Antibiotic Resistance Database (CARD), and were counted by mapping metagenomic reads against these annotated sequences. Analysis has shown China and the Philippines have the highest proportion of ARGs in saliva samples than the UK, supporting the impact of antibiotic over-use in these countries. Oral samples (including saliva and dental) contain a higher abundance of specific ARGs, but a lower diversity of total ARGs than stool samples, across all geographic regions investigated. In addition, co-occurrence analysis of oral and stool resistome profiles with taxonomic composition profiles has revealed novel links between ARGs and species that have not been previously identified to contain these ARGs, as well as existing and reported species containing these ARGs.

P5

PathOGiST: Calibrated multi-criterion genomic analysis for public health microbiology

Leonid Chindelevitch Pedro Feijao, Sean La, Matthew Nguyen, Johnathan Wong, Mohsen Katebi, Alex Sweeten, Tony Nguyen, Margaryta Vityaz, Julius Booth, William Hsiao, Cedric Chauve,

Simon Fraser University [William Hsiao is also at the University of British Columbia]

During an infectious disease epidemic, it is essential to cluster pathogen samples into strains so as to allow quick identification of outbreak sources. This is a difficult task, as pathogens evolve quickly and in a variety of ways. Genomics has been revolutionized in the past two decades by the advent of next-generation sequencing (NGS) technologies, which allow analysis of entire genomes. Access to whole-genome sequencing (WGS) data of pathogens is plenty, though efficient tools to analyze this data in the context of public health remain to be developed. In particular, two main issues emerge from the use of WGS data for genotyping. First, methods for differentiating outbreak-related strains from sporadic strains are often based on a single type of genomic variation. This approach captures only a limited amount of the genomic variability and tells only a partial story of the organism's evolutionary history. Second, WGS-based sample clustering algorithms are often not calibrated, meaning that the determination of clustering thresholds or subtyping cutoffs is still mostly arbitrary. There are many forces driving pathogen evolution and as a result, using the wrong set of variants or the wrong cutoffs may mislead the investigation of a pathogen outbreak. We address these issues by developing PathOGiST, a platform that implements and integrates existing and novel genomic variant calling algorithms from WGS data (such as SNPs, Multilocus sequence typing, and copy number variations), together with clustering algorithms based on a multi-criterion genome dissimilarity measure using various kinds of genomic variants and correlation clustering. The final steps include the calibration of the statistical models and algorithms using large reference sets of selected pathogen genomes from epidemiologically confirmed outbreak strains. PathOGiST is an open-source, robust computational framework for the classication of pathogens into epidemiologically related groups based on genomic data. We cluster three sets of bacterial samples taken from real outbreaks, and find that PathOGiST yields clusterings which are in near-perfect agreement with true clusterings derived from wet lab experiments while allowing us to estimate the reliability of these clusterings. PathOGiST will revolutionize how disease outbreaks are managed (in particular helping track the emergence and transmission of drug resistance), ensuring faster responses that will reduce the impact of these outbreaks on both health and the economy.

P6

Rapid and precise alignment of raw reads against redundant databases with KMA

Philip Clausen, Frank M. Aarestrup & Ole Lund

Research Group for Genomic Epidemiology, National Food Institute, Technical University of Denmark, 2800 Kgs Lyngby, Denmark

Background: As the cost of sequencing has declined, clinical diagnostics based on next generation sequencing (NGS) have become reality. Diagnostics based on sequencing will require rapid and precise mapping against redundant databases because some of the most important determinants, such as antimicrobial resistance and core genome multilocus

sequence typing (MLST) alleles, are highly similar to one another.In order to facilitate this, a

novel mapping method, KMA (k-mer alignment), was designed. KMA is able to map raw reads directly against redundant databases, it also scales well for large redundant databases. KMA uses k-mer seeding to speed up mapping and the Needleman-Wunsch algorithm to accurately align extensions from k-mer seeds. Multi-mapping reads are resolved using a novel sorting scheme (ConClave scheme), ensuring an accurate selection of

templates.Results: The functionality of KMA was compared with SRST2, MGmapper, BWA-

MEM, Bowtie2, Minimap2 and Salmon, using both simulated data and a dataset of Escherichia coli mapped against resistance genes and core genome MLST alleles. KMA outperforms current methods with respect to both accuracy and speed, while using a comparable amount of memory. Conclusion: With KMA, it was possible map raw reads directly against redundant databases with high accuracy, speed and memory efficiency. Availability: KMA is implemented in C, and is freely available at: https://bitbucket.org/genomicepidemiology/kma and as web-service at: https://cge.cbs.dtu.dk/services/KMA/.

P7

Carbapenemase-Producing Escherichia coli ST131 are Phylogenetically Diverse and are not limited to ESBL associated Clades. Nicholas Ellaby, Michel Doumith, Katie L Hopkins, Neil Woodford, Matthew J Ellington. Antimicrobial Resistance and Healthcare Associated Infections (AMRHAI) Reference Unit, National Infection Service, Public Health England, London, UK. Introduction Increased antimicrobial resistance (AMR) rates and the emergence multidrug-resistant (MDR) E. coli threaten public health. MDR extra-intestinal E. coli sequence type (ST)131 have been associated with a pandemic of urinary tract and bloodstream infections. Resistance in ST131 is typically due to chromosomal mutations and acquired genes (e.g. CTX-M ESBLs) that cluster with the fimH30 allotype ST131 clades C1 and C2. We assessed the onwards development of ST131 clades, as carbapenemases were acquired, and examined the geographical and temporal distribution in order to evaluate clinical and public health risks. Methods Carbapenemase-producing E. coli (CPE. coli) ST131 isolates, sequenced by PHE’s AMRHAI Reference Unit between January 2014 and June 2016 (illumina HiSeq 2500, WGS), were incorporated into a published phylogenetic framework and dataset. Using the EC958 reference sequence, a RAxML SNP-based phylogeny was calculated and compared with against the previous phylogenetic framework and with results from PHE’s Snapper DB pipeline for cluster analysis. Genes/loci were detected (AMRHAI’s GeneFinder). Results Thirty-nine colistin susceptible ST131 CPE. coli isolates (representing 4.5% of CPE. coli sequenced by AMRHAI) were from 8/9 English regions. These isolates were genetically diverse; over half (23/38) belonged to FQ-R ST131 clades C1 or C2 (8 and 15, respectively), 10 and 6 ST131 CPE. coli belonged to FQ-S clades A and B, respectively. The acquired carbapenemases in the isolates were diverse, KPC-2 occurred most often (N=21 isolates), associated mainly with clade C isolates (N= 14), followed by B (N=4) and A (N=3); KPC producers occurred in 6 of 9 regions. OXA-48-producers (N=10) were predominantly clade A (N=5), two isolates in clades C1 and C2 and one in clades B; they were also geographically widespread (3 regions). The eight remaining isolates produced NDM-1, -5, VIM-1, -4 or OXA-181 and were primarily clade C (N=5). Conclusions Previously, ESBL producing E. coli have been shown to be predominated by ST131 isolates. This pattern was not apparent amongst the carbapenemase-producing E. coli submitted to PHE’s national AMRHAI reference laboratory in 2014-16. The overall diversity of ST131 CPE. coli indicates that they have emerged on multiple independent occasions and emphasises need for ongoing monitoring of this ‘high-risk’ lineage, which has known pandemic potential.

P8

Molecular epidemiology and short term evolution of antimicrobial resistance in Klebsiella pneumoniae isolates

Neris García-González (1), Paula Ruiz-Hueso (1), Susana Sabater (2), Bárbara Gomila (2), Rosario Moreno (2), Fernando González-Candelas (1,3)

(1) FISABIO-CSISP / Univ. Valencia. Joint Research Unit Infection and Public Health, Valencia, Spain. (2) Hospital General Universitario de Castellón, Castellón, Spain. (3) CIBER in Epidemiology and Public Health, Genomics and Health, Valencia, Spain.

Antimicrobial resistance (AMR) is a major threat to public health. Next Generation Sequencing (NGS) technologies open new possibilities to prevent and control the dissemination of AMR genes but their implementation at the clinical level needs a detailed evaluation in terms of sensitivity and specificity but also of investments in time, training, and interpretation of the results. We have compared the results of phenotypic tests with the detection of AMR genes for betalactams, aminoglycosides, quinolones and cotrimoxazol from NGS data. 1038 K. pneumoniae strains from 657 patients were isolated between March 2014 and February 2015 at HGUC. 111 ESBL-producing strains were retained for further analysis. NGS was performed with Illumina NextSeq (2x150 paired-ends). AMR genes were identified in the sequencing reads using SRST2 with the ARG-annot database. The concatenated sequences of the MLST genes for K. pneumoniae along with those AMR genes present in each isolate were used to construct a maximum likelihood phylogenetic tree. A similar tree was constructed with the core genome of the sequenced strains. Dated phylogenies and short term phylodynamics and evolutionary parameters were estimated with BEAST. Most strains (>90%) harbour at least two ESBL-producing genes, usually blaSHV and blaCTXM. We obtained a high but not perfect match between phenotypically determined resistance and the inferred presence of AMR genes. We failed to explain the observed phenotype in 46 of the 1018 (4.5%) tests for beta-lactams, in 16 for the 263 (6.1%) tests performed for aminoglycosides, in 3 of the 151 (2.1%) for fluoroquinolones, and in 11 of the 110 (10%) for cotrimoxazol. The phylogenetic tree from MLST and AMR genes revealed the presence of 74 different haplotypes. Estimates of the short-term evolutionary parameters were obtained for the most abundant ST, namely ST11 (n=53, 47.7% of the sequenced isolates). The genome-wide evolutionary rate was estimated at 2.59E-6 subst/site/year, which leads to an estimate for the date of the common ancestor of these isolates on May 2012.

P9

Comparing antibiotic resistance phenotypes and genotypes on a large-scale international collection

Rebecca A Gladstone1, Stephanie W Lo1, Theresa J Ochoa2, Pak Leung Ho2, Mignon du Plessis2, Jennifer E Cornick2, Brenda Kwambana-Adams2, Rachel Benisty2, Susan A Nzenze2, Shabir A Madhi2, Paulina A Hawkins2, Benjamin J Metcalf3, Yuan Li3, Bernard Beall3, Dean B Everett2, Martin Antonio2, Ron Dagan2, Keith P Klugman2, Anne von Gottberg2, Lesley McGee3, Robert F Breiman4, Stephen D Bentley1, The Global Pneumococcal Sequencing Consortium3.

1Parasites and Microbes, Wellcome Sanger Institute, UK. 2http://www.pneumogen.net/gps/ 3Centers for Disease Control and Prevention, Atlanta, USA *The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the CDC. 4Emory Global Health Institute, USA.

Tools for the prediction of antibiotic resistance from genomic data have published impressive results with phenotypic validation. It is unclear how well this inference will perform on datasets from other areas of the world. We inferred resistance from 13,454 genomes from the Global Pneumococcal Sequencing study representing over 30 countries. We used a pneumococcal pipeline published by the CDC (BJ Metcalf et al 2016) predicting penicillin minimum inhibitory concentration (MIC) using the sequence of the penicillin binding proteins (pbp1a,2b 2x), and detecting known resistance conferring genes for erythromycin (erm/mef), tetracycline (tet) and chloramphenicol (cat) or known amino acid changes conferring resistance to co-trimoxazole (FolA and FolP). Where phenotypic antibiotic susceptibility data were available we compared genotypic and phenotypic resistance status using CLSI interpretation for each antibiotic to determine if the genotype was a useful proxy for resistance in a global research context. Antibiotic Sensitivity 95% CI Specificity 95% CI Penicillin 95.33% 94.46-96.10% 90.14% 89.03-91.16% Erythromycin 70.86% 69.00-72.68% 99.32% 99.02-99.55% Tetracycline 87.36% 85.58-88.99% 97.69% 97.08-98.19% Chloramphenicol 25.19% 22.40-28.15% 99.82% 99.65-99.92% Co-trimoxazole 73.95% 72.10-75.73% 86.82% 86.39-87.25% Positive predictive values were 90% or greater whilst negative predictive values were generally lower reflecting phenotypic resistance not explained by known genotypes. Such comparisons provide a rich resource to identify novel mechanisms but they will likely only represent the minority of the discrepancies. Phenotypic investigation of the discrepancies is needed as methods vary in their accuracy per antibiotic meaning phenotype may be unreliable. Using genotype to infer resistance of isolates from multiple contributing labs had the advantage of being a single standardised method. This study generates new data for resistance in countries where surveillance is limited and allowed acquisition, spread and distribution of resistance to be characterised.

P10

Genomics-based insights into Klebsiella pneumoniae antimicrobial resistance patterns in nation-wide invasive infections

Claire Gorrie 1, Norelle Sherry 1,2,3, Jan Bell 4, Peter Pham 1, Glen Carter 5, Takehiro Tomita 1, Anders Gonçalves Da Silva 1, Susan Ballard 1, Ben Howden 1,2,3

1 Microbiological Diagnostic Unit Public Health Laboratory, Department of Microbiology and Immunology, at Peter Doherty Institute, University of Melbourne, Melbourne, VIC, Australia 2 Antimicrobial Research and Reference Unit, Microbiological Diagnostic Unit Public Health Laboratory, Department of Microbiology and Immunology, at Peter Doherty Institute, University of Melbourne, Melbourne, VIC, Australia 3 Austin Health Dept of Infectious Diseases, Melbourne, VIC, Australia 4 Australian Centre for Antimicrobial Resistance Ecology, University of Adelaide, Adelaide, SA, Australia 5 Doherty Applied Microbial Genomics, Department of Microbiology and Immunology, at Peter Doherty Institute, University of Melbourne, Melbourne, VIC, Australia

Objectives: Klebsiella pneumoniae is a well-recognised global pathogen, frequently implicated in healthcare-associated infections, with a predilection for acquisition of antimicrobial resistance (AMR) determinants. This study aims to better understand the population structure of invasive K. pneumoniae isolates in Australia and assess trends in the distribution of AMR genes by time, geographic location, and patient factors. Methods: 420 bloodstream K. pneumoniae isolates were selected from the Australian Group for Antimicrobial Resistance (AGAR) AMR surveillance program, collected from 30 hospital laboratories between 2013 and 2015. Isolates were chosen to represent the most prevalent AMR phenotypes, with a random selection of susceptible isolates for comparison, and stratified according to year of isolation and geographic location. Stored isolates were sequenced at MDU PHL (Illumina NextSeq) and analysis performed using Nullarbor (https://github.com/tseemann/nullarbor). Genomic data was correlated with antibiograms and clinical information. Results: Phylogenetic analyses revealed the presence of a number of internationally recognised high-risk clones associated with multidrug resistant (MDR) phenotypes, as well novel sequence types circulating locally. While some lineages were found only within a single state, many others were identified throughout all Australian states, including CC14/15, CC17, and CC147. Genetic bases for resistant phenotypes were also interrogated to identify known mechanisms and assess the accuracy of genotype in predicting phenotype, and to investigate AMR gene cooccurrence. Conclusions: Invasive K. pneumoniae isolates in Australia are largely polyclonal, including a range of known and novel sequence types. However, several high-risk MDR clones have also emerged, as has been seen elsewhere internationally. Active AMR surveillance, including the use of high-resolution microbial genomics, will be essential to characterise the current status of AMR in Australia, and direct future interventions to limit its impact.

P11

Diversity of Carbapenemase-Producing Enterobacteriaceae in England As Revealed By Whole Genome Sequencing

Katie L Hopkins, Michel Doumith, Nazim Mustafa, Daniele Meunier, Matthew Ellington, Neil Woodford

National Infection Service, Public Health England, London, UK

Increasing numbers of carbapenemase-producing Enterobacteriaceae (CPE) have been referred to the national reference laboratory since screening started in the early 2000s. Whole genome sequencing (WGS) was applied over 30 months to inform our understanding of CPE epidemiology in England. WGS was applied to the first confirmed CPE from each new patient referred by an English laboratory between 1st Jan '14 and 30th June '16. Multiple isolates from the same patient were included if of different species, or the same species with a different carbapenemase. Illumina HiSeq 2500 WGS data was analysed using an in-house bioinformatics pipeline that determines species identification, MLST profile and antimicrobial resistance gene content. WGS data were obtained for 2658 CPE, representing 60% of CPE received from an English laboratory over this period. CPE were referred from all regions but North West England (NWE, n=941) and London (n=927) were particular foci. OXA-48-like (n=1119, variants OXA-48, -181, 204, -232, -244 and -484), NDM (n=691, NDM-1, -3, -4, -5 and -7 ), KPC (n=570, KPC-2, -3, -4 and -23), VIM (n=100, VIM-1, -4 and -19) and IMP (n=33, IMP-1, -4 and -14) predominated with 1 - 21 isolates harbouring FRI, IMI, NMC-A, SME, GES-5 or various two carbapenemase gene combinations. Klebsiella pneumoniae (Kpn, n=1380), Escherichia coli (Esc, n=723) and Enterobacter cloacae (Ent, n=294) accounted for most isolates. Host species showed significant diversity with 151, 115 and 63 STs amongst K. pneumoniae, E. coli and E. cloacae, respectively. A further 126/1380 K pneumoniae, 45/723 E. coli and 49/294 E. cloacae could not be assigned to a previously defined ST. The top three STs per species were 14 (n=168), 11 (n=160) and 147 (n=105) amongst K. pneumoniae, 38 (n=158), 410 (n=63) and 167 (n=50) amongst E. coli and 108 (n=42), 104 (n=29) and 66 (n=22) amongst E. cloacae. Some clones represented local outbreaks (e.g. all Ent-ST108 with OXA-48 from NWE, 37/42 from one laboratory) whilst others were associated with multiple carbapenemases and referred from all nine regions (e.g. 155/158 Esc-ST38 with OXA-48, -181 and -244 from 66 laboratories). Global high-risk clones Kpn-ST258 (n=39, carrying KPC-2, -3 and -23) and Esc-ST131 (n=39, carrying KPC-2, NDM-1, OXA-48, -181, VIM-1 and -4) were each referred by 8/9 regions. WGS has provided us with unprecedented data on the CPE clones circulating within England and their carbapenemase genes. The diversity suggests a role for clonal spread and carbapenemase gene transfer between STs.

P12

Molecular detection of metallo-ß-lactamase acquired genes, blaIMP and blaVIM in Carbapenems susceptible and resistant Gram-negative clinical isolates

Mudathir Abdallah A Ismail, Hadeel I Ismail, Wafa I Elhag.

Mudathir Abdallah A Ismail, M.Sc. student. Microbiology Department. Faculty of Medical Laboratory Science, Al Neelain University. Khartoum, Sudan. [email protected]. Hadeel I Ismail, Lecturer-Microbiology Department, Al Neelain University. Khartoum, Sudan. Wafa I Elhag, Associate professor of Microbiology Department, Faculty of Medical Laboratory Science, Al Neelain University. Khartoum, Sudan. [email protected] Background: Emergence of Metallo-β-lactamase (MBL) is of great concern in the clinical settings worldwide. The death rate associated with MBL producers is ranging from 18% to 67%. The main purpose of this study is to determine the prevalence of metallo-β lactamase genes among some Gram-negative clinical isolates (carbapenems susceptible and resistant). Methods: This was a prospective descriptive cross-sectional study, which carried out to detect genes responsible for MBL enzymes such as blaVIM and blaIMP by conventional and multiplex PCR, among 200 Gram-negative clinical isolates (Citrobacter spp, E.coli, Enterobacter spp, K.pneumoniae, P.aeruginosa, P.mirabilis, P.valgaris) at Khartoum hospitals during 2015 to 2016. Limitation: The study organisms were not evaluated for non-MBL carbapenemases. Ethical approval was obtained from Al Neelain University Ethical Review Board. Results: The general prevalence of MBL genes by multiplex PCR assays among 200 Gram-negative clinical isolates was 69(34.5%). 27(27%) and 42 (42%) were positive for both blaIMP and blaVIM MBLs genes among carbapenems sensitive and resistant isolates respectively. Verona integron metallo-beta-lactamase (VIM) was the most frequent genes (53.6 %) out of 69 MBLs detected genes, while it was (36.2 %) for imipenemase (IMP). Conclusion: This the first report of (blaIMP and blaVIM) MBLs genes prevalence in some Gram-negative isolates from Khartoum Hospitals, Sudan. Keywords: Metallo β-lactamase, VIM, IMP, Carbapenem, PCR, Gram-negative bacteria, Khartoum.

P13

Eco-evolutionary factors affecting antibiotic resistance

Matti Jalasvuori, Johannes Cairns, Jens Frickel, Katariina Koskinen, Reetta Penttinen, Tommi Patinen, Anna Hartikainen, Roosa Jokela, Liisa Ruusulehto, Sirja Viitamäki, Sari Mattila, Jouni Laakso, Veijo Kaitala, Sven Künzel, Emre Karakoc, Lutz Becks and Teppo Hiltunen

Department of Biological and Environmental Science, University of Jyvaskyla, Finland

Evolution and maintenance of bacterial resistance to antibiotics is a complex issue where various factors and actors, such as mobile genetic elements, leakiness of resistance mechanisms, presence of predators and types of antibiotics and their concentrations, play a prominent role. Therefore, in order to accurately predict when and how resistance evolves, all potential facets in the process needs to be at least acknowledged if not directly taken into account. As such, we have investigated in differing experimental setups the effects that (multi-)stressor selection have on antibiotic resistance. Predation and sublethal antibiotic concentrations slows down the evolution of anti-predator defence and antibiotic resistance compared to the presence of only one of the two stressors (1). The selected mutations are also different, suggesting that in real life communities (where the presence of multiple stressors is likely) the evolutionary trajectories may diverge from controlled laboratory experiments. Further, disruption of conjugative transfer of resistance plasmids was observed to eventually re-sensitizes bacterial communities to antibiotics, but interestingly this effect is nullified in the presence of a predator (2). Predation appears to increase per cell activity of bacteria, which hence amplifies the conjugation frequency of remaining plasmid-harboring bacteria. Different types and concentrations of antibiotics also cause bacterial communities/populations to vary in their evolutionary responses (2-4). Overall, holistic approach to bacterial systems is necessary for understanding resistance. (1) Hiltunen et al., 2018. Multi-stressor selection alters eco-evolutionary dynamics in experimental communities. Nature Ecology & Evolution. In press. (2) Cairns et al., 2018. Black Queen evolution and trophic interactions determine plasmid survival after the disruption of conjugation network. mSystems. In press. (3) Mattila et al., 2017. Conjugative ESBL-plasmids differ in their potential to rescue susceptible bacteria via horizontal gene transfer in lethal antibiotic concentrations. Journal of Antibiotics. 70:805-808. (4) Cairns et al., 2017. Genomic evolution of bacterial populations under co-selection by antibiotics and phage. Molecular Ecology. 26:1848-1859.

P14


Camilla Hundahl Johnsen, Derya Aytan, Frank M. Aarestrup, and Ole Lund


Extensive use of antimicrobials has driven the progressive emergence of resistant bacteria and other microbes. Surveillance of antimicrobial resistance and rapid detection of resistance is important to facilitate an effective treatment, and hence reduce drug mis-usage. Additional rapid detection is essential as resistance call for extra attention and precaution. Next-generation sequencing technologies rapidly yield accurate whole-genome sequencing (WGS) data, from which antimicrobial resistance might be directly determined. At Center for Genomic Epidemiology we aim to provide accurate bioinformatic solutions for determining resistance phenotypes. We have developed both ResFinder and PointFinder, in silico methods for predicting resistance based on acquired resistance genes or chromosomal mutations from WGS data. In this study we focus on improving our PointFinder tool. PointFinder is a rapid, assemblyfree program that detects point mutations, insertions and deletions. Resistance is predicted when any detected mutation is present in our curated database of resistance associated mutations. PointFinder is available for several species, including M. Tuberculosis (Mtb). We optimized the Mtb resistance prediction by feature selection methods, such as forward selection, ReliefF and L1 Norm Support Vector Machine methods. We also found that the literature on Mtb resistance mutations does contain misclassified resistance markers, and by elimination those from our model we improved the prediction performance. Traditionally, the occurrence of a resistance mutation has been treated as a direct indicator for resistance. However, it is now suspected that some mutations need the existence of other mutations to cause drug resistance, and more complex mutation patterns might explain the resistance emergence. Therefore, we have initiated a machine learning study to predict resistance by random forest and neural network prediction models. Pending results using these approaches will show how linked mutation might account for resistance that could not have been explained before.

P15

Carbapenem Resistant Klebsiella – Linking Phenotypes to Epidemical Successfulness

Katariina Koskinen1, Tarmo Ketola1, Reetta Penttinen1, Christian Giske2 and Matti Jalasvuori1

1 Department of Biological and Environmental Science, Nanoscience Center, University of Jyväskylä, Finland, 2 Department of Laboratory Medicine, Division of Clinical Microbiology, Karolinska Institutet, Sweden

Carbapenemase producing Klebsiella pneumoniae strains are considered as a notable threat in the modern-day healthcare. These strains are found to be abundant in hospital environments and capable to transfer between hospitals worldwide. Our study focuses on seven strains classified as epidemically successful and seven strains classified as epidemically non-successful depending on their global distribution patterns in hospital environments. These 14 patient isolates, encoding either KPC, NDM, or VIM carbapenem resistance genes, were analyzed as both individual strains and also grouped by their epidemical status. Epidemical successfulness has earlier noted to associate with certain multi-locus sequence types in some extend. However, sequence typing provides only limited data on specific strains. We expanded our characterization to a broad range of phenotypic features. Multiple environmental and cellular properties were measured to define differences in growth and behavior of epidemically successful and non-successful strains. Further linking these phenotypical properties and multi-locus sequence types with statistical analyses we investigated if epidemical successfulness of the strain can be predicted from existing phenotypic features. In future similar detection and clustering of phenotypic properties can be considered when assessing the probability of certain strain to cause larger epidemics rather than relying only on the genomic based procedures.

P16

Complete genome sequences of a blaKPC-2-positive Klebsiella pneumoniae and Aeromonas spp. isolated from the effluent of urban wastewater treatment plants in Japan

Makoto Kuroda, Tsuyoshi Sekizuka, Koji Yatsu, Takaya Segawa, Yuba Inamine

Pathogen Genomics Center, National Institute of Infectious Diseases, Tokyo, Japan. [email protected]

Background: Antimicrobial resistance genes (ARGs) and the bacteria that harbor them are widely distributed in the environment, especially in surface water, sewage treatment plant effluent, soils and animal wastes. The widespread detection of carbapenemase-producing Enterobacteriaceae (CPE) in the environment is an emerging issue with potentially serious public health implications. Materials/methods: The upper effluent flow of an urban wastewater treatment plant (WWTP) was collected on August, 2017 and August, 2018 in Tokyo Bay, Japan. The effluent was filtered through a PES filter membrane (0.22 µm), and the membrane was incubated with LB-broth supplemented with 1 µg/mL meropenem at 37°C for 14 h, followed by spreading on CHROMagar ESBL plates. A complete genome sequence of the isolate was determined using a NextSeq 500 and Sequel sequencers. The cgMLST was generated using the maximum likelihood phylogenetic method with FastTree. Antimicrobial susceptibility was determined by the Etest, disk diffusion method and Carba NP test. Results: We isolated a KPC-2-producing Klebsiella pneumoniae strain (GSU10-3). GSU10-3 is resistant to most β-lactam antibiotics and other antimicrobial agents (quinolones and aminoglycosides), and GSU10-3 is closely related to KPC-2-positive Chinese clinical isolates (sequence type: ST11) from 2011 to 2017. GSU10-3 harbors a blaKPC-2-positive plasmid pGSU10-3-3 (66.2 kb), which notably carries dual replicons [IncFII (pHN7A8) and IncN]. In addition, we isolated a KPC-2-producing Aeromonas hydrophila (GSH8-1) and Aeromonas caviae (GSH8-5) from two sampling sites in Tokyo Bay, Japan. GSH8-1 and GSH8-5 share a similar blaKPC-2-positive plasmid [53.6 kb, IncP(6) replicon] between two strains. The plasmid sequence exhibits similar backbone to seven plasmids of Aeromonas spp. [A. hydrophila (WCHAH045096 and WCHAH01); A. taiwanensis (L186, L198, and L1713); two other strains], and the plasmids in Pseudomonas aeruginosa 10265, K. oxytoca pKOX3 and Citrobacter freundii CF121SC21. Conclusions: In Japan, the KPC-type has been very rarely detected, while IMP is the most predominant type of carbapenemase in clinical CPE isolates. Although a laboratory testing thus far suggested that Japan may be virtually free of KPC-producing Enterobacteriaceae, we have detected a clinically high impact of CPE from WWTP. AMR monitoring of WWTP effluent may contribute to the early detection of future AMR bacterial dissemination in clinical settings and communities.

P17

Dissemination of K. pneumoniae and mobile elements revealed by whole genome sequencing: A One Health Approach

Catherine Ludden, Danesh Moradigaravand, Theodore Gouliouris, Dorota Jamrozy, Beth Blane, Plamena Naydenova, Juan Hernandez-Garcia, Paul Wood, Nazreen Hadjirin, Milorad Radakovic, Charles Crawley, Nicholas M. Brown, Mark Holmes, Julian Parkhill, Sharon J. Peacock

London School of Hygiene & Tropical Medicine, Keppel Street, London, WC1E 7HT, UK Wellcome Trust Sanger Institute, Wellcome Trust Genome Campus, Hinxton, Cambridge, CB10 1SA, UK Department of Medicine, University of Cambridge, Box 157 Addenbrooke’s Hospital, Hills Road, Cambridge, CB2 0QQ, UK Department of Veterinary Medicine, University of Cambridge, Madingley Rd, Cambridge CB3 0E, UK Royal (Dick) School of Veterinary Studies, University of Edinburgh, Edinburgh, EH25 9RG, UK Cambridge University Hospitals NHS Foundation Trust, Hills Road, Cambridge CB2 0QQ, UK Clinical Microbiology and Public Health Laboratory, Public Health England, Cambridge CB2 0QQ, UK

Background: Klebsiella pneumoniae is a leading cause of nosocomial infections. Genome sequencing has confirmed nosocomial transmission of K. pneumoniae during outbreaks, but studies to determine source of acquisition are lacking. We explored this question using a One Health framework. Materials/methods: We conducted a six-month study on two adult haematology wards at a hospital in the East of England, United Kingdom in 2015. Stools were requested on admission, weekly and at discharge and the environment was repeatedly swabbed. K. pneumoniae isolated from blood cultures from patients on the two wards 12-month prior, during and 6-months after the study were obtained from the diagnostic laboratory. Hospital sewage was sampled on four occasions in 2014-2015. A cross-sectional survey was conducted to isolate K. pneumoniae from 29 livestock farms, 97 meat products and 20 municipal wastewater treatment plants across the East of England in 2014-2015. The resulting K. pneumoniae were sequenced and analyses performed of phylogenetic relatedness and antimicrobial resistance genes. Results: K. pneumoniae was isolated from the following: 23/376 stools from 17/149 (11%) patients; 18/922 environmental swabs; 4/29 livestock farms; 12/20 water treatment plants; 3/4 hospital sewer samples; and 0/97 meat products. K. pneumoniae was obtained from positive blood cultures from 5 patients. Multiple colonies were selected from each positive sample for sequencing. We sequenced 249 K. pneumoniae from these different sources. Phylogenetic analysis revealed a highly diverse population. Each patient carried one or more lineages that was different from any other carriage isolate. Two lineages carried by patients were associated with transient contamination of the ward environment. Each patient with bloodstream infection was infected with a lineage that was different from other cases. This high diversity was replicated in the non-clinical isolate collection, with no evidence for relatedness of isolates obtained from humans versus other sources (farms or wastewater). The same genes encoding cephalosporin resistance were detected in isolates from different reservoirs, but these were carried on different plasmids in human versus livestock isolates. Conclusions: We found no evidence for patient-to-patient transmission, limited evidence for environmental contamination in a hospital setting, and no evidence for livestock as a source of K. pneumoniae that infect humans.

P18

PROMAST: a probiotic formulation for mastitis, reveals comparable efficacy to antibiotic treatment in a field trial.

Harsh Mathur1,2, Michael Kitching1,2, James Flynn3, Noel Byrne3, Pat Dillon3, Riona Sayers3, Mary C. Rea1,2, Colin Hill2,4, R. Paul Ross2,4.

1Teagasc Food Research Centre, Moorepark, Fermoy, County Cork, Ireland; 2APC Microbiome Ireland, University College Cork, Cork, Ireland; 3Teagasc Animal and Grassland Research and Innovation Centre, Moorepark, Fermoy, Co. Cork, Ireland, 4School of Microbiology, University College Cork, Cork, Ireland.

Bovine mastitis is an ongoing significant concern in the dairy and agricultural industry resulting in substantial losses in milk production and revenue. Among the predominant etiological agents of bovine mastitis are Staphylococcus aureus, Streptococcus uberis, Streptococcus dysgalactiae and Escherichia coli. Currently, the treatment of choice for bovine mastitis involves the use of commercial therapeutic antibiotic formulations such as TerrexineTM, which consists of both kanamycin and cephalexin. Such antibiotics are regularly administered in more than one dose resulting in the withholding of milk for processing for a number of days. Furthermore, the above-mentioned pathogens can display resistance to the antibiotics, resulting in treatment failure and persistent mastitis. Here, we describe the optimisation of a formulation of Lactococcus lactis DPC3147 cells, which produce the two-component bacteriocin lacticin 3147 (formulation hereafter designated 'PROMAST'), with a view to treating cows with clinical/sub-clinical mastitis. Critically, in a field trial described here, this "ready-to-use" emulsion containing live L. lactis DPC3147 cells exhibited comparable efficacy to a commercial antibiotic when used to treat mastitic cows. In addition, we found that the L. lactis cells within this emulsion were relatively stable for up to 5 weeks, when stored at 4°C, 22°C or 37°C. The relative ease and cost-effective nature of producing this PROMAST formulation, in addition to its enhanced shelf life compared to previous aqueous-based formulations, indicate that PROMAST could be a viable alternative therapeutic option for bovine mastitis. Moreover, the single-dose administration of PROMAST is a further advantage, as it can expedite the return of the milk to the milk pool, in comparison to some commercial antibiotics. Overall, in this field trial, we show that the probiotic formulation displayed a 47% cure rate compared to a 50% cure rate for a commercial antibiotic control, with respect to curing cows with clinical/sub-clinical mastitis.

P19

Estimating effect of antibiotic use on individual and household transmission of fluoroquinolone resistant Enterobacteriaceae using Markov models

Patrick Musicha1,2,, Andrew Stewardson3, Stephen Habath3 and Ben S. Cooper1,2

1.Mahidol Oxford Tropical Medicine Research Unit, Mahidol University, Bangkok Thailand; 2.Centre for Tropical Medicine and Global Health, Nuffield Department of Medicine, University of Oxford, Oxford, United Kingdom. 3.University of Geneva Hospitals and Faculty of Medicine, Geneva, Switzerland

Background: Antimicrobial resistance continues to be a major threat to human health globally. However, the rates at which antimicrobial resistant bacteria spread between household members and the effects of antibiotic use and other risk factors on this transmission are not well-quantified. We investigated transmission and carriage dynamics of fluoroquinolone resistant Enterobacteriaceae at individual and household levels and estimated the effect of fluoroquinolone and nitrofurantoin usage on the probability of becoming colonized with fluoroquinolone resistant Enterobacteriaceae. Methods: We analysed longitudinal data collected between January 2011 and August 2013 in a prospective case-control study at three European sites of Antwerp (Belgium), Geneva (Switzerland) and Lodz (Poland). Cases were patients with suspected urinary tract infection requiring antibiotic treatment, whereas controls were patients presenting for indications not requiring antibiotic treatment. For each index patient, one to three household members were recruited. Selected participants provided feacal samples at baseline, completion of antibiotic therapy (or 7-10 days after first sample for non-exposed) and at 28 days after the second sample. We used Markov models to estimate rate of colonization with fluoroquinolone resistant bacteria and carriage duration and quantified effects of covariates including antibiotic use, location, travel, and use of antibiotics within 12 months before study participation. Preliminary results: The probability of becoming colonized with fluoroquinolone resistant bacteria was 0.008 per day and colonized individuals remained so for a mean duration of 26.6 days. The daily probability of clearing colonization with fluoroquinolone resistant bacteria was 0.033. Individuals were more likely to become colonized with fluoroquinolone resistant Enterobacteriaceae if they were exposed to fluoroquinolones (hazard ratio 1.96, 95% CI= [1.05, 3.68]) or were from Lodz (hazard ratio 7.36, 95% CI = [3.17,17.09]). The effect of fluoroquinolone use on colonization was greatest during exposure (hazard ratio 2.95; 95% CI = [1.26, 6.91]) but waned post-exposure (hazard ratio 1.44; 95% CI= [0.69, 3.02]). Conclusion: Fluoroquinolone exposure increases the probability of being colonized by fluoroquinolone resistant bacteria. However, the risk of colonization due to fluoroquinolone exposure diminishes post exposure.

P20

Genome-wide analysis of Mycobacterium tuberculosis polymorphisms reveals lineage-specific associations with drug resistance

Yaa Oppong(1,§), Jody Phelan(1), João Perdigão(2), Diana Machado(3), Anabela Miranda(4), Isabel Portugal(2), Miguel Viveiros(3), Taane G Clark(1,5,*), Martin L. Hibberd(1,*)

1 Faculty of Infectious and Tropical Diseases, London School of Hygiene and Tropical Medicine, London, United Kingdom 2 iMed.ULisboa – Research Institute for Medicines, Faculdade de Farmácia, Universidade de Lisboa, Portugal 3 Global Health and Tropical Medicine, Instituto de Higiene e Medicina Tropical, Universidade Nova de Lisboa, UNL, Lisboa, Portugal 4 National Mycobacterium Reference Laboratory, Porto, Portugal 5 Faculty of Epidemiology and Population Health, LSHTM, London, United Kingdom * joint authors

Background: Continuing evolution of the Mycobacterium tuberculosis (Mtb) complex genomes associated with resistance to anti-tuberculosis drugs is threatening tuberculosis disease control efforts. Both multi- and extensively drug resistant Mtb (MDR and XDR, respectively) are increasing in prevalence, but the full set of Mtb genes involved are not known. There is a need for increased sensitivity of genome-wide approaches in order to elucidate the genetic basis of missing resistance and gain a more holistic understanding of Mtb genome evolution in a context of widespread antimicrobial therapy. Population structure within the Mtb complex, due to clonal expansion, lack of lateral gene transfer and low levels of recombination between lineages, may be reducing statistical power to detect drug resistance associated variants. Methods: To investigate the effect of lineage-specific effects on the identification of drug resistance associations, we applied genome-wide association study (GWAS) and convergence-based (PhyC) methods to multiple drug resistance phenotypes of a global dataset of Mtb lineages 2 and 4, using both lineage-wise and combined approaches. Results: We identify both well-established drug resistance variants and novel associations; uniquely identifying associations for both lineage-specific and -combined GWAS analyses. We report 17 potential novel associations between antimicrobial resistance phenotypes and Mtb genomic variants. Conclusions: For GWAS, both lineage-specific and -combined analyses are useful, whereas PhyC may perform better in contexts of greater diversity. Unique associations with XDR in lineage-specific analyses provide evidence of diverging evolutionary trajectories between lineages 2 and 4 in response to antimicrobial drug therapy.

P21

Supporting the National Action Plan on AMR in Tanzania – SNAP-AMR

Katarina Oravcova (2), Stephen Mshana (1), Stephen Mshana (1), Emma Laurie (2), Emma McIntosh (2), Gabriel Shirima (3), Alicia Davis (2), Tiziana Lembo (2), Nicholas Hanley (2), Blandina T Mmbaga (4), Louise Matthews (2), Siana Mapunjo (5), Shona Hilton (2), Frank Van der Meer (6), Douglas R Call (7), Ruth N Zadoks (2)

(1) CUHAS/BMC, Mwanza, Tanzania; (2) University of Glasgow, UK; (3)NM-AIST, Arusha, Tanzania; (4) KCMC/KCRI, Moshi, Tanzania; (5) AMR National Focal Point, Tanzania; (6) University of Calgary, Canada; (7) Washington State University, USA

Antimicrobial resistance (AMR) is a global problem with national solutions and, in Tanzania, -- considering the heterogeneity of human communities and livestock keeping systems, and their access to health care and antimicrobials -- a national problem with local solutions. Our interdisciplinary consortium aims to provide novel insights into biological, socio-economic and cultural drivers of AMR to prioritise use of limited human and financial resources in targeting evidence-based levers of behavioural change that will reduce the risk and clinical and economic burden of AMR. Using socio-anthropological methods, we will elicit information on knowledges, attitudes, practices and access to antimicrobials across tiers of the health care system and in distinct livestock keeping communities in northern Tanzania, where preliminary data has already demonstrated significant differences in antimicrobial use (AMU) and AMR. This will be complemented with measurement of AMU in the health system using the WHO defined daily dose system, and behavioural and health economic approaches to investigate to explore costs, benefits and feasibility of potential interventions and to inform bespoke communication campaigns in health systems and communities. Preliminary data show that the prevalence of AMR, as exemplified by extended spectrum beta lactamase resistance in coliforms, increases from lower (7% of E. coli) to higher (24% of E. coli) tiers in the health system. To quantify transmission of AMR into, across and within tiers of the health system, we will use extensive archives of E. coli carriage isolates from humans and animals in the study area as well as prospectively collected carriage and clinical isolates from participating neonatal wards and hospitals for genomic sequencing and phylogenetic and diversity analyses. The combined outcomes of our research will help priority setting in AMR control by identifying the setting (health system or community) and context (human or veterinary) where behavioural change is practicable and cost-effective. It will inform implementation of the National Action Plan in Tanzania and aims to serve as a generalisable transdisciplinary model of AMR control in low and middle income settings.

P22

Rapid detection of resistance to anti-tuberculous drugs by integrating informatics tools and portable sequencing technology

Jody Phelan1, Denise O’Sullivan2,*, Sonal Shah1,*, Diana Machado3, Jorge Ramos3, Su-sana Campino1, Justin O’Grady4, Miguel Viveiros3, Martin L. Hibberd1, Jim F. Huggett2,5, Taane G. Clark

1Faculty of Infectious and Tropical Diseases, London School of Hygiene & Tropical Medicine, Keppel Street, London, UK, 2 LGC Ltd, Queens Road, Teddington, Middlesex, UK, 3Global Health and Tropical Medicine, Instituto de Higiene e Medic-ina Tropical, Universidade NOVA de Lisboa, Lisbon, Portugal, 4Norwich Medical School, University of East Anglia, Norwich, UK, 5Faculty of Health & Medical Science, University of Surrey, Guildford, UK, 6Faculty of Epidemiology & Population Health, LSHTM, London, UK

Increasing resistance to anti-tuberculosis drugs is a growing threat to public health. Whole genome sequencing is rapidly gaining traction as a tool for investigating drug resistance in Mycobacterium tuberculosis to aid treatment decisions. We rewrote the TBProfiler software tool from scratch integrating the latest software and analysis methods to provide a command-line tool that is easy to use and provides a detailed results, reporting the in-silico drug resistance profile and strain type from raw next generation sequencing data, without requiring internet access. In parallel we have assembled a large database of strains (n=~10,000) for which sequence data and drug susceptibility testing data is available to calculate the sensitivity and specificity of the mutation library. We see high sensitivity and specificity for first-line antituberculosis drugs, whilst the these metrics drop for some second-line drugs., indicating that there are possible resistance mechanisms which still need to be determined.

P23

Use of Hi-C metagenomic sequencing to identify drug resistance genes and their taxonomic origin directly from the gut microbiome of non-human primates

Gregory J Phillips, Christopher Coe, Paul Plummer

College of Veterinary Medicine, Iowa State University Harlow Center for Biological Psychology, University of Wisconsin-Madison

Metagenomic DNA sequencing can reveal antibiotic resistance genes within mixed microbial communities. However, it can be difficult to unambiguously identify the taxonomic source of resistance genes since precise linkage relationships are often lost with short read sequencing technology typically used for metagenomic DNA analysis. To overcome this limitation, we have used Hi-C metagenomic sequencing, which includes a DNA crosslinking step prior to DNA extraction and sequencing, to analyze metagenomic samples. As proof of concept, we have used this approach to identify antimicrobial resistance genes encoded by specific bacterial taxa within the gut microbiome of captive Rhesus macaques. Several different classes of drug resistance determinants were found within the most prevalent taxa in the mammalian gut. In particular, genes encoding resistance to vancomycin and tetracycline were the most commonly found, as well as genes encoding multidrug efflux systems. Genes encoding resistance to nitroimidazole, fosmidomycin, quinolone and polymyxin were also identified in multiple bacterial taxa. These results suggest that the Hi-C sequencing approach holds promise as a means to identify the taxonomic origin of specific resistance determinants within complex microbial communities, as well as to track transmission of drug resistance genes by horizontal gene transfer.

P24

Genomic surveillance of Enterococcus faecium reveals limited sharing of strains and resistance genes between livestock and humans in the United Kingdom

Dr Kathy Raven, Theodore Gouliouris*(1,2), Catherine Ludden(3), Beth Blane(1), Jukka Corander(4,5), Carolyne S. Horner(6), Juan Hernandez-Garcia(7), Paul Wood(8), Nazreen F. Hadjirin(7), Milorad Radakovic(7), Mark A. Holmes(7), Marcus de Goffau(5), Nicholas M. Brown(2), Julian Parkhill(5), Sharon J. Peacock(1,3,5). *Contributed equally

1 Department of Medicine, University of Cambridge, UK 2 Public Health England, Clinical Microbiology and Public Health Laboratory, Cambridge, UK 3 London School of Hygiene and Tropical Medicine, London, UK 4 Department of Biostatistics, University of Oslo, Norway 5 Wellcome Sanger Institute, Cambridge, UK 6 British Society for Antimicrobial Chemotherapy, Birmingham, UK 7 Department of Veterinary Medicine, University of Cambridge, UK 8 Royal (Dick) School of Veterinary Studies, University of Edinburgh, UK

Enterococcus faecium is an important cause of nosocomial infection, most of which is caused by a hospital-adapted lineage. This may have evolved from an animal-adapted lineage, and is genetically distinct from the human commensal lineage. We performed a large genome sequence-based 'One Health' study to determine whether this major hospital lineage is carried by livestock and define the genetic relatedness between isolates from different reservoirs. A cross-sectional surveillance study was conducted to isolate E. faecium from 29 livestock farms and 20 municipal wastewater treatment plants across East Anglia, United Kingdom between 2014 and 2015. We sequenced 636 E. faecium from livestock (256 isolates from cattle, pigs, chickens and turkey) and wastewater (n=383) on an Illumina HiSeq2000 instrument. We obtained additional whole genome data for 782 E. faecium associated with bloodstream infection in the British Isles (47% from East Anglia, 2001-2012), 11 historical strains, and 10 reference strains. Multilocus sequence types (ST) were derived from genome data. Phylogenetic and bioinformatics analyses were performed using open-access tools. E. faecium was isolated from 28/29 farms, none of which were vancomycin-resistant E. faecium (VREfm), suggesting a decrease in VREfm prevalence since the last UK livestock survey in 2003. However, VREfm was isolated from 1-2% of retail meat products, and was ubiquitous in wastewater treatment plants.The 1442 E. faecium isolates were assigned to 218 STs. Bayesian clustering divided the collection into 10 phylogenetic clusters (BAPS groups). Enterococcus faecium from humans and livestock were largely genetically segregated, with 97% human invasive isolates residing in the hospital-adapted lineage and 85% of animal isolates residing basal to this lineage. The four exceptions to this were: (i) BAPS group 5, containing 94 isolates from humans, pigs, cattle and chicken that was ancestral to the hospital-adapted lineage; (ii) 22 isolates from pigs that were nested within the hospital-adapted lineage; (iii) six human isolates in the animal-associated basal lineage; and (iv) human commensal lineage, containing 17 isolates from human infection and all four livestock species. Contrasting with this segregation between human and animal isolates, we observed that highly related E. faecium were shared by different animal species (turkeys and chickens), and by different farms. An analysis of acquired antibiotic resistance genes and their variants revealed limited sharing between humans and livestock. Our findings indicate that the majority of E. faecium infecting patients are largely distinct from those from livestock in this setting, with limited sharing of strains and resistance genes.

P25

Comparative behavioral genomics reveals potential diagnostic genomic markers associated with Pseudomonas aeruginosa sensitivity to aminoglycosides and quinolones group of antibiotics

Wedad Saleh, Arshad M Khan and Nigel J Saunders

Systems and Synthetic Biology Group, Biosciences, Life Sciences. College of Health and Life Sciences, Brunel University London Understanding intrinsic resistance to antibiotics in bacterial pathogens from environmental origin is highly important. This is of special significance in some opportunistic infections such as those caused by Pseudomonas aeruginosa which is currently responsible for adverse health outcomes in vulnerable patient populations and hospitalized patients with critical conditions. Although different research approaches were used to investigate for intrinsic resistance to antibiotics in this organism including; expression profiling, mutant-library screening, functional metagenomics and in-vitro experimental evolutionary studies, this research uses comparative behavioral genomic approach to investigate for resistance determinants at the global level. The current research approach can be considered superior to other previously used approaches in its ability to identify different variants and mutational events in addition to essential and non-essential genes forming specific molecular signatures. It can also show combination of genomic markers associated with specific phenotypes with the potential value to be used as clinical diagnostic markers. This approach is more practical in reflecting real differences in studied isolates which makes it of higher clinical relevance. On testing a group of 164 completely sequenced clinical isolates of Pseudomonas aeruginosa, results showed that although single genes and nucleotide changes can be predictive of sensitivity to antibiotics, combined genomic elements and some observed molecular patterns showed different predictive values. Specially promising is the observation that some markers showed high negative predictive values with the possibility of being used as antibiotic sensitivity determinants. This would help in selecting for proper treatment and thus avoiding the cost of antibiotic misuse and over-prescription. Using whole genome sequence data for prediction of antibiotic sensitivity should not be limited to identification of well-known resistance determinants because multiple genomic markers and unknown molecular signatures appear to contribute to determining sensitivity or resistance in an additive way.

P26

Evergreen: a platform for surveillance of bacterial outbreaks

Judit Szarvas, Johanne Ahrenfeldt, Jose Luis Bellod Cisneros, Martin Christen Frølund Thomsen, Frank M. Aarestrup, Ole Lund

Research Group for Genomic Epidemiology, National Food Institute, Technical University of Denmark, Kongens Lyngby, Denmark

Whole-genome sequencing has an expanding role in public health with the technological advancements in high-throughput sequencing. Public health authorities sequence thousands of pathogenic isolates each month for microbial diagnostics and surveillance of pathogenic bacteria. Whole genome sequencing data could be used to infer the phylogeny of the isolates and uploading the data to public repositories facilitates the surveillance of pathogenic bacteria on an international scale. We have built a platform for the surveillance of pathogenic bacteria, which incorporates publicly available sequencing data. The Evergreen pipeline downloads the newly available whole genome sequencing data daily from the public repositories. To decrease the computational burden, the data is divided into sets by matching the isolates to a closely related reference genome. The reads are mapped to the reference to gain a consensus sequence and the SNP based genetic distance is calculated between all the sequences that were mapped to the same reference. Sequences are clustered together with a threshold of 10 SNPs to reduce the redundancy in each set. Finally, phylogenetic trees are inferred from the non-redundant sequences and the clustered sequences are placed on a clade with the cluster representative sequence. The pilot run so far has placed more than 100,000 E. coli, C. jejuni, L. monocytogenes, S. enterica and Shigella samples into phylogenetic trees. The changes in the sizes of the clusters are monitored, and we have observed several clusters that correspond to known outbreaks. The results from the ongoing analysis are browsable and searchable on https://cge.cbs.dtu.dk/services/Evergreen.

P27

Appearance of synthetic vectors, and associated ARGs, on all branches of the tree of life

George Taiaroa, Paul Kitts, Gregory M. Cook, Deborah A. Williamson

Doherty Institute, University of Melbourne, Australia (GT, DW). Department of Microbiology and Immunology, University of Otago, New Zealand (GT, GC). National Centre for Biotechnology Information, NIH, USA (PK)

Synthetic vectors are necessary for the production of reagents used throughout molecular biology, and in the application of molecular methods. These reagents include polymerases and ligases needed to generate various sequence data. Given that reagent preparation methods vary, there remains a possibility that synthetic vectors contaminate sequencing reagents, and appear in generated sequence data. Here we explore this possibility, aiming to identify synthetic vectors, and associated antibiotic resistance genes (ARGs), in a broad range of next-generation sequence data. A systematic search of public sequence databases (NCBI, ENA, SRA), totalling 750,000 datasets, showed synthetic vectors appear on all branches of the tree of life. Evidence is provided to show that this occurs on a global scale, in varied biological and environmental samples, and at rates sufficient to impact accurate prediction of antimicrobial resistance in clinical settings. Taken together, synthetic vector-associated ARGs are identified in 400 microbial species, including all high priority pathogens. ARGs are also attributed to higher organisms as diverse as humans, brassicas and butterflies. Reagents containing synthetic vector contamination are highlighted, and solutions put forward to address this challenge.

P28

Global surveillance of antimicrobial resistance.

Bram van Bunnik2, Rene Hendriksen1, Patrick Munk1, Patrick Njage1, Luke McNally2, Oksana Lukjancenko1, Timo Röder1, David Nieuwenhuijse3, Susanne Karlsmose Pedersen1, Jette Kjeldgaard1, Rolf S. Kaas1, Philip Thomas Lanken Conradsen Clausen1, Josef Korbinian Vogt1, Pimlapas Leekitcharoenphon1, Milou G. M. van de Schans4, Tina Zuidema4, Ana Maria de Roda Husman5, Simon Rasmussen6, Bent Petersen6, Global sewage consortium, Guy Cochrane7, Thomas Sicheritz-Ponten8, Heike Schmitt5, Jorge Raul Matheu Alvarez9, Awa Aidara-Kane9, Sünje J. Pamp1, Ole Lund6, Tine Hald1, Mark Woolhouse2, Marion Koopmans3, Håkan Vigre1, Thomas Nordahl Petersen1, Frank M. Aarestrup1*

1 National Food Institute, Technical University of Denmark, Denmark 2 Usher Institute, University of Edinburgh, United Kingdom 3 Viroscience, Erasmus Medical Center, The Netherlands 4 RIKILT Wageningen University and Research, The Netherlands 5 National Institute for Public Health and the Environment (RIVM), The Netherlands 6 Department of Bio and Health Informatics, Technical University of Denmark, Denmark 7 European Bioinformatics Institute, UK 8 Centre of Excellence for Omics-Driven Computational Biodiscovery, AIMST University, Malaysia 9 World Health Organization, Switzerland Antimicrobial resistance (AMR) is one of the most serious global public health threats, however, obtaining representative data on AMR for healthy human populations is difficult. In this study we characterized the bacterial resistome from untreated sewage from 79 sites in 60 countries using short read sequencing and mapped the reads to the ResFinder database to identify resistance genes. We found systematic differences in abundance and diversity of AMR genes between Europe/North-America/Oceania and Africa/Asia/South-America. Antimicrobial usage data only explained a minor part of the AMR variation and no evidence for cross-selection between antimicrobial classes nor effect of travel by flight between sites were found. However, AMR abundance was strongly correlated with socio-economic, health and environmental factors, which we used to predict AMR abundances in all countries in the world. Our findings suggest that the global AMR gene diversity and abundance varies by region and are caused by national circumstances. Improving sanitation and health could potentially limit the global burden of AMR. Based on these results we propose to use sewage for an ethically acceptable and economically feasible continuous global surveillance and prediction of AMR.

P29

Analysis of drug-resistance mutations in DNA Gyrase A of Mycobacterium leprae: a structural bioinformatics study

Vaishali P. Waman1, Sundeep Chaitanya Vedithi1, Asma Munir1 and Tom L. Blundell1

1Department of Biochemistry, University of Cambridge, Tennis Court Rd., CB2 1GA, UK

Mycobacterium leprae (M. leprae) is the causative agent of leprosy in humans. Although drugs have been successfully developed to treat leprosy, emergence of drug-resistance is a growing health challenge. Multidrug therapy is a useful means to treat leprosy, however, mutations in respective drug targets (in M. leprae) lead to the emergence of drug-resistant strains. Fluoroquinolones, important second-line drugs to combat drug-resistance in M. leprae, target DNA gyrase, an enzyme involved in negative DNA supercoiling. DNA gyrase encodes 2 subunits, A (gyrA) and B (gyrB) and exists as a hetero-tetramer (gyrA2-gyrB2). Mutations in the quinolone-resistance-determining region of gyrA (G89C, A91V, A91T, S92A and R107L) are known to cause fluoroquinolone-resistance. The molecular mechanisms for fluoroquinolone-resistance in DNA gyrase of M. leprae are largely unknown due to the absence of experimental structures of the catalytic core of DNA gyrase, formed by the N-terminal domain of gyrA and C-terminal domain of gyrB, and complexes with DNA and fluoroquinolones. Hence, the aim of this study was first to build model of the catalytic core of DNA gyrase in M. leprae. Secondly, the model was used to analyze the impact of drug-resistant mutations on protein structure/stability. The following structural bioinformatics approaches were used: 1. Homology modeling using ModSuite server, developed in-house. The experimental structures of the catalytic core of the DNA gyrase of M. tuberculosis (PDB IDs: 5BS8, IIFZ; ~90% identity) were used as a templates for prediction of the catalytic core of M. leprae DNA gyrase. 2. Prediction of the effects of drug-resistance mutations on protein stability, using in-house developed methods namely mutation Cutoff Scanning Matrix (mCSM) and Site-Directed Mutator (SDM). mCSM uses graph-based structural signatures, where each mutation is represented as a signature vector, to train and test predictive machine learning methods. SDM uses Environment-specific substitution tables to calculate the stability difference score between the wildtype and mutant protein structures. 3. Comparative analysis of wild-type and mutant models to analyze effect of mutations on atomic interactions with DNA or fluoroquinolones. This is the first study to report the model of catalytic core of DNA gyrase of M. leprae. The analyses using mCSM and SDM methods indicate that drug-resistance mutations in gyrA are destabilizing (eg. mCSM ΔΔG scores (Kcal/mol): G89C: -1.4, A91V: -0.4, A91T: -0.7, S92A: -0.6, R107L: -0.7). It suggests that fluoroquinolone-resistance in M. leprae is likely to arise from mutations that affect the stability of catalytic core or interaction with that of DNA.

P30

Exploring the use of machine learning for prediction of antimicrobial resistance phenotypes using whole genome sequence data

Nicole E Wheeler, Kevin C Ma, Laura Jenniches, Yonatan Grad, Lars Barquist, David Aanensen

Centre for Genomic Pathogen Surveillance, Wellcome Sanger Institute, Hinxton, UK; Department of Immunology and Infectious Diseases, Harvard T.H. Chan School of Public Health, Boston, USA; Helmholtz Institute for RNA-based Infection Research, Würzburg, Germany; Department of Immunology and Infectious Diseases, Harvard T.H. Chan School of Public Health, Boston, USA; Helmholtz Institute for RNA-based Infection Research, Würzburg, Germany; Centre for Genomic Pathogen Surveillance, Wellcome Sanger Institute, Hinxton, UK & Big Data Institute, University of Oxford, Oxford, UK

Antibiotic resistance is a growing problem worldwide. Uncovering mechanisms through which resistance evolves can help us to identify emerging resistant strains in a timely manner and limit their spread. Genome-wide association studies (GWASs) are an effective way of searching for genetic variants associated with resistance in an unbiased manner while correcting for spurious correlations introduced by population structure. However, it is unclear how individual associations should be used collectively to predict resistance phenotypes in new samples. Moving from the detection of correlations to the prediction of resistance phenotypes can be aided by the incorporation of machine learning methodologies. Standard machine learning algorithms do not correct for population structure, potentially leading to models that incorporate a combination of causative mechanisms and lineage markers. We explore this issue and potential solutions using data on antibiotic resistance from a diverse collection of Neisseria gonorrhoea strains.

P31

Capsule-switching is associated with the recent global expansion of the fluoroquinolone-resistant Escherichia coli ST1193 clone. Rhys White1,2, Brian Forde1,2, Leah Roberts1,2, Melinda Ashcroft1,2, Minh-Duy Phan1, Kate Peters1, Darren Trott4, Justine Gibson5, Joanne Mollinger6, Benjamin Rogers7, Nouri Ben Zakour3, Amanda Kidsley4, Jan Bell4, John Turnidge5, Mark Schembri1, Scott Beatson1,2

1School of Chemistry and Molecular Biosciences and Australian Infectious Disease Research Centre, 2Australian Centre for Ecogenomics, The University of Queensland, Brisbane, Australia; 3Westmead Institute for Medical Research and The University of Sydney, Sydney, Australia; 4School of Animal and Veterinary Sciences, The University of Adelaide, Roseworthy, Australia; 5School of Veterinary Science, The University of Queensland, Gatton, Australia; 6Biosecurity Sciences Laboratory, Biosecurity Queensland, Department of Agriculture and Fisheries, Coopers Plains, Australia; 7Centre for Inflammatory Diseases, Monash University, and Monash Infectious Diseases, Monash Health, Melbourne, Australia Uropathogenic Escherichia coli (UPEC) are a leading cause of urinary and systemic infections. Sequence type (ST)1193 is a globally-disseminated, fluoroquinolone-resistant UPEC clone, second only to ST131 in clinical prevalence. We characterised the emergence of this important lineage with a comprehensive genomics approach. The genomes of 55 Australian ST1193 strains, isolated between 2007-2017, were sequenced and analysed together with 54 publicly available genomes of strains from the same clonal complex (CC14). Long-read sequencing was used to assemble the complete reference genome MS10858 and determine the genomic context of genes encoding antimicrobial resistance and virulence factors. ST1193 acquired chromosomal mutations in gyrA, parC and parE resulting in fluoroquinolone resistance, distinguishing them from other lineages within CC14. Bayesian analysis predicted that ST1193 emerged in 1989, coinciding with an increase in the human use of fluoroquinolones in Australia, after their inclusion in government-subsidised medications in 1988. The ST1193 phylogeny consists of two major clades that are highly similar at the chromosomal level. Whereas Clade 1 is comprised of strains from all over the world including Australia, Clade 2 is almost exclusively Australian. Remarkably, Clade 1 is distinguished by recombination of a 30.4 kb region encompassing the capsular biosynthesis genes causing a switch from the K5 to K1 capsular antigen, which is associated with increased serum survivability. Assembly of MS10858 revealed a novel 11.5 kb composite transposon Tn6623 containing five resistance genes on an F-type plasmid, and an ISEcp1-blaCTX-M-15 element on an IncI1 plasmid. This work describes a comprehensive genomic characterisation of ST1193 and identifies a single recombination event in the capsule locus, associated with the global dissemination of this fluoroquinolone-resistant UPEC clone.

P32

National early warning system for exceptional resistance phenotypes using Big Data

Julie Wilson1, Eddie McArdle1, Annette Little1, Michael Lockhart1.

1Health Protection Scotland Health Protection Scotland (HPS) receives electronic positive microbiology reports from all Scottish microbiology laboratories. This equates to approximately one million records per year. It is not possible to analyse all of these data and so an automated IT algorithm was developed to interrogate instances of exceptional resistance phenotypes. This serves as an early warning system (EWS) that identifies isolates that are of potential public health concern. This enables prompt action to be taken by the submitting laboratory and facilitates timely investigation to confirm veracity and further characterisation if appropriate. It can also impact on the clinical management of cases and allow appropriate infection control procedures to be implemented. In the latter part of 2017, HPS and the Scottish Microbiology and Virology Network AST Subgroup agreed a revised list of exceptional resistance phenotypes based on the EUCAST exceptional phenotype list (http://www.eucast.org/expert_rules_and_intrinsic_resistance/). The

revised list underpinned the development of the EWS in line with the Strategic Aims and Approach section of the UK Five Year Antimicrobial Resistance Strategy, 2013 to 2018 recommendation to “improve the knowledge and understanding of AMR through better information, intelligence, supporting data and developing more effective early warning systems to improve health security”. HPS has several AMR surveillance systems however it is important to monitor AMR in a timely manner in order to serve as an EWS for emerging resistant pathogens. This facilitates rapid distribution of information about emerging multi-drug resistant pathogens to hospitals and public health authorities, and contributes and expedites the implementation of infection control interventions to prevent the emergence or transmission of antimicrobial drug resistant pathogens in health-care facilities. The advantage of a national EWS is that emerging resistance including regional trends which may not be apparent at a local NHS board level can be detected.

http://www.eucast.org/expert_rules_and_intrinsic_resistance/

P33

Genomic surveillance of Neisseria gonorrhoeae and comprehensive surveillance of bacterial cultures and drug susceptibilities in Japan

Koji Yahara, Shu-ichi Nakayama, Ken Shimuta, Ken-ichi Lee, Masatomo Morita, Takuya Kawahata, Toshiro Kuroki, Yuko Watanabe, Hitomi Ohya, Mitsuru Yasuda, Takashi Deguchi, Xavier Didelot, Makoto Ohnishi, (on behalf of Antibiotic-Resistant Gonorrhea Study Group), Keigo Sibayama, Motoyuki Sugai

Antimicrobial Resistance Research Center, National Institute of Infectious Diseases (NIID), Tokyo, Japan

We will present two aspects of AMR surveillance in Japan. The first is genomic surveillance of Neisseria gonorrhoeae. The first extensively drug resistant (XDR) Neisseria gonorrhoeae strain with high resistance to the extended-spectrum cephalosporins (ESC) ceftriaxone was identified in 2009 in Japan, but no other strain with this antimicrobial resistance profile has been reported since. However, surveillance so far has been based on phenotypic methods and sequence typing, not genome sequencing. Therefore, little is known about the local population structure at the genomic level, and how resistance determinants and lineages are distributed and evolve. We analyzed whole-genome sequence data and antimicrobial susceptibility testing results of 204 strains sampled in a region where the first XDR ceftriaxone-resistant N. gonorrhoeae was isolated, complemented with 67 additional genomes from other timeframes and locations within Japan. Strains resistant to ceftriaxone were not found, but we discovered a ST7363 sub-lineage susceptible to ESC in which the mosaic penA allele responsible for reduced susceptibility had reverted to a susceptible allele by recombination. Approximately 85% of isolates showed resistance to fluoroquinolones (ciprofloxacin) explained by linked amino acid substitutions of GyrA with 99% sensitivity and 100% specificity. Approximately 10% showed resistance to macrolides (azithromycin), for which genetic determinants are less clear. Furthermore, we revealed different evolutionary paths of the two major lineages spreading worldwide (published in Microbial Genomics, 2018). The second is one of the largest AMR surveillances in the world, Japan Nosocomial Infections Surveillance (JANIS), in which there are 2,000 hospitals voluntarily participating and collects comprehensive routine bacteriological test results online. It generates not only national and regional AMR reports but also benchmarking data for each member hospital to guide and evaluate infection control practices.

P34

Notes

Speaker & Delegate List

Fahad Alenezi

NHL

[email protected]

Aroob Alhumaidy

National Health Laboratory

[email protected]

Shamsudin Aliyu

Ahmadu Bello University

[email protected]

Ali AlShahib

Public Health England

[email protected]

Silvia Argimon

Centre for Genomic Pathogen Surveillance

[email protected]

Anna Aryee

UCL

[email protected]

Erki Aun

Tartu Ülikool

[email protected]

Derya Aytan

Technical University of Denmark

[email protected]

Maliha Aziz

George Washington University

[email protected]

Till Bachmann

University of Edinburgh

[email protected]

Eva Bernhoff

Stavanger University Hospital

[email protected]

Daniela Bezdan

MetaSUB / WCM

[email protected]

Supriya Bhat

University of Regina

[email protected]

Laura Blackburn

PHG Foundation

[email protected]

Grace Blackwell

EMBL-EBI

[email protected]

Debby Bogaert


[email protected]

Ferdinando Bonfiglio

University of Basel

[email protected]

Joao Botelho

UCIBIO/REQUIMTE

[email protected]

Age Brauer

University of Tartu

[email protected]

Annie Browne

University of Oxford

[email protected]

Josie Bryant

University of Cambridge

[email protected]

David Burstein

Tel Aviv University

[email protected]

Victoria Carr

King’s College London

[email protected]

Alessandro Cassini

ECDC

[email protected]

Leonid Chindelevitch

Simon Fraser University

[email protected]

Michael Chipeta


[email protected]

Philip Clausen

Center for genomic epidemiology

[email protected]

Francesc Coll

London School of Hygiene & Tropical

Medicine

[email protected]

Thomas Crellen

Mahidol-Oxford Research Unit

[email protected]

Joshua DAeth

Imperial College London

[email protected]

Tom Darton

University of Sheffield

[email protected]

Sophia David

Wellcome Trust Sanger Institute

[email protected]

James Davis

University of Chicago

[email protected]

Emma Doughty

Quadram Institute Bioscience

[email protected]

Susanna Dunachie


[email protected]

Fredrik Dyrkell

1928 Diagnostics

[email protected]

Joan Ejembi

Ahmadu Bello University

[email protected]

Nicholas Ellaby


[email protected]

Dimitrios Evangelopoulos

The Francis Crick Institute

[email protected]

Aasmund Fostervold


[email protected]

Louise Fraser

Illumina

[email protected]

Amal Gadalla

Cardiff University

[email protected]

Neris Garcia Gonzalez

University of Valencia

[email protected]

Fawaz Ghali

MMU

[email protected]

Rod Givney

NSW Health Pathology

[email protected]

Rebecca Gladstone

Wellcome Sanger Institute

[email protected]

Claire Gorrie

Microbiological Diagnostic Unit Public Health

Laboratory

[email protected]

Michihiko Goto

University of Iowa Carver College of

Medicine

[email protected]

Olivia Grabmaier

University Erlangen-Nuremberg

[email protected]

Jonathan Green


[email protected]

Natalie Groves


[email protected]

Laura Marcela Guzman Rincon

University of Warwick

[email protected]

Georgina Haines Woodhouse

Oxford University

[email protected]

Axel Hamprecht

University of Cologne

[email protected]

Josh Harling Lee


[email protected]

Henrik Hasman

Statens Serum Institut

[email protected]

Eva Heinz


[email protected]

Andrew Henderson

The University of Queensland

[email protected]

Marit Andrea Klokkhammer Hetland


[email protected]

Allison Hicks

Harvard University

[email protected]

Katie Hopkins


[email protected]

A S Md Mukarram Hossain


[email protected]

Alasdair Hubbard

LSTM

[email protected]

Karl Hultin

1928 Diagnostics

[email protected]

Bogdan Iorga

CNRS-ICSN

[email protected]

Margaret Ip

Chinese University of Hong Kong

[email protected]

Zamin Iqbal

EMBL-EBI

[email protected]

Mudathir Ismail

Al Neelain University

[email protected]

Matti Jalasvuori

University of Jyvaskyla

[email protected]

Nicol Janecko

Quadram Institute Bioscience

[email protected]

Benjamin Jeffrey

Imperial College London

[email protected]

Claire Jenkins


[email protected]

Camilla Hundahl Johnsen


[email protected]

Gwyn Jones

RUMA

[email protected]

Teemu Kallonen

University of Oslo

[email protected]

Bahar Kashef


[email protected]

Håkon Kaspersen

Norwegian Veterinary Institute

[email protected]

Katariina Koskinen

University of Jyväskylä

[email protected]

Karl Kristinsson

Landspitali University Hospital

[email protected]

Emmanuelle Kumaran


[email protected]

Makoto Kuroda

Nati. Instit. of Infect. Dis.

[email protected]

Rachel Kwiatkowska


[email protected]

Bruno Lacroix

bioMérieux

[email protected]

Anders Rhod Larsen


[email protected]

Jesper Larsen


[email protected]

Helena Larsson

kalmar county hospital

[email protected]

Ramanan Laxminarayan

CDDEP/Princeton

[email protected]

Pimlapas Leekitcharoenphon


[email protected]

Hannah Lepper


[email protected]

Joe Lewis

Liverpool School of Tropical Medicine

[email protected]

Junyan Liu


[email protected]

Iren Loehr


[email protected]

Erik Lorén

1928 Diagnostics

[email protected]

Catherine Ludden

LSHTM

[email protected]

Ole Lund


[email protected]

Kevin Ma

Harvard TH Chan School of Public Health

[email protected]

Keith MacKenzie

University of Regina

[email protected]

Charis Marwick

University of Dundee

[email protected]

Harsh Mathur

APC Microbiome Ireland

[email protected]

Louise Matthews

University of Glasgow

[email protected]

Noel Mccarthy

University of Warwick

[email protected]

Catrin Moore


[email protected]

Patrick Musicha

Mahidol-Oxford Research Unit

[email protected]

Shunsuke Numata


[email protected]

Iruka Okeke

University of Ibadan

[email protected]

Mathupanee Oonsivilai

Mahidol-Oxford Tropical Medicine Research

Unit

[email protected]

Yaa Oppong

LSHTM

[email protected]

Katarina Oravcova

University of Glasgow

[email protected]

Nonia Pariente

Nature

[email protected]

Julian Parkhill


[email protected]

Sharon Peacock

London School of Hygiene and Tropical

Medicine

[email protected]

Andreas Petersen


[email protected]

Jody Phelan

London School of Hygiene and Tropical

Medicine

[email protected]

Gregory Phillips

Iowa State University

[email protected]

Bruno Pichon


[email protected]

Lindsay Pike


[email protected]

Padmaja Polubothu

Medical Microbiolog GG&C

[email protected]

Lance Price

George Washington University

[email protected]

Michaela Projahn

German Federal Institute for Risk Assessment

[email protected]

Dhamayanthi Pugazhendhi

Illumina

[email protected]

Santeri Puranen

Aalto University

[email protected]

Kathy Raven


[email protected]

Gisela Robles Aguilar

Oxford AMR GBD Group

[email protected]

John WA Rossen

University of Groningen (UMCG)

[email protected]

Wedad Saleh

brunel university London

[email protected]

Jeremiah Seni

CUHAS - Bugando & UCalgary

[email protected]

Helena SethSmith

University Hospital Basel

[email protected]

Motoyuki Sugai

National Institute of Infectious Diseases

[email protected]

Timothy Sweeney

Inflammatix, Inc

[email protected]

Judit Szarvas


[email protected]

George Taiaroa

University of Melbourne

[email protected]

Nick Thomson


[email protected]

Kin Ming Tsui

Sidra Medicine

[email protected]

Juan Ugalde

uBiome

[email protected]

Anthony Underwood

Centre for Genomic Pathogen Surveillance,

WSI

[email protected]

Bram van Bunnik


[email protected]

Daria Van Tyne

University of Pittsburgh

[email protected]

Hanna Voksepp

kalmar county hospital

[email protected]

Vaishali Waman


[email protected]

Bryan Wee


[email protected]

Nicole Wheeler


[email protected]

Rhys White

The University of Queensland

[email protected]

Julie Wilson

Health Protection Scotland

[email protected]

Rhiannon Wood


[email protected]

Mark Woolhouse


[email protected]

Daniel Wuethrich

University Hospital Basel

[email protected]

Koji Yahara

National Institute of Infectious Diseases

[email protected]

Chendi ZHU

The Chinese University of Hong Kong

[email protected]

Index Koskinen P15

Kuroda P16

Aliyu P1

Argimon S43 Laxminarayan S33

Aun S13 Leekitcharoenphon S41

Aytan P2 Ludden S29, P17

Bezdan S17 Ma S59

Bogaert S57 Mathur P18

Botelho P3 Musicha P19

Browne S39

Oonsivilai S15

Carr P4 Oppong P20

Cassini S37 Oravcova P21

Chindelevitch S7, P5

Clausen P6 Peacock S49

Coll S51 Phelan P22

Crellen S45 Phillips P23

Pike S61

David S21 Price S25

Davis S9

Dunachie S35 Raven P24

Ellaby P7 Saleh P25

Sweeny S55

Garcia Gonzalez P8 Szarvas P26

Gladstone P9

Gorrie P10 Taiaroa P27

Thomson S19

Hicks S5

Hopkins P11 Underwood S53

Ip S23 van Bunnik P28

Iqbal S3 Van Tyne S31

Ismail P12

Waman P29

Jalasvuori P13 Wheeler P30

Jenkins S47 White P31

Johnsen S11, P14 Wilson P32

Jones S27 Woolhouse S1

Yahara P33

Antimicrobial Resistance Genomes, Big Data and Emerging ......09:00-10:30 Session 2: Genomic...

Documents

Transcript of Antimicrobial Resistance Genomes, Big Data and Emerging ......09:00-10:30 Session 2: Genomic...