Lecture given (remotely) at Swasnea University, October 2020 on using GWAS in bacteria.

1. Challenges of using genome-wide
association studies in bacteria
Ben Pascoe
b.pascoe@bath.ac.uk | Twitter: @benizao
www.sheppardlab.com

2. The Sheppard Lab Group - Who we are, what we do
Sam Sheppard
Sion Bayliss
Ben Pascoe
Jessica Calland Emily Rudolph
Evangelos Mourkas
Billy Monteith
Grant Futcher

3. Population genomics and evolution of bacteria
Pathogenicity – Bayliss et al. In prep.
Host adaptation – Mourkas et al. (2020) PNAS
– Sheppard et al. (2013) PNAS
Biofilm – Pascoe et al. (2015) Env Micro
Survival – Yahara et al. (2017) Env Micro
Pathogenicity – Monteil et al. (2016) MGen. Host/pathogen – Muñoz-Ramirez et al. (2020) The ISME
Journal
– Berthenet et al. (2018) BMC Biology
Pathogenicity – Meric et al. (2018) Nature Comms
Antimicrobial resistance
– Mourkas et al. (2019) Env Micro
– Sproston et al. (2018) MGen

4. What are Genome-wide association studies?

5. Genome-wide association studies in humans

6. Challenges of using GWAS in bacteria: different size genomes
Gorlicz et al 2020; Trends in Genetics
https://doi.org/10.1016/j.tig.2019.11.006

7. Wordsfoundincattle
Expected by clonal decent
Observed
Challenges of using GWAS in bacteria: clonal reproduction
Earle et al. Nature Micro 2016 (1) 16041
Falush & Bowden Trends Micro 2006 14(8): 353
Sheppard et al Nature Rev Gen 2018 19:549

8. Weight associations according to phylogeny

9. Is my data suitable for GWAS? – Population structure
Earle et al. Nature Micro 2016 (1) 16041
Falush & Bowden Trends Micro 2006 14(8): 353
Sheppard et al Nature Rev Gen 2018 19:549

10. ‘early’ GWAS
papers: S.
aureus (2014-
2017)
• Can we identify genes
involved in AMR?
Laabei et al Genome Research 2014 24(5):839; Laabei & Massey 2016 Curr Gen 62(3):523; Recker et al Nature Micro 2017 2(10):1381

11. Is my data suitable for GWAS? – Phenotype clustering
Earle et al. Nature Micro 2016 (1) 16041
Falush & Bowden Trends Micro 2006 14(8): 353
Sheppard et al Nature Rev Gen 2018 19:549

12. Are Campylobacter from chicken and cows the same?
952
22
42
45
177
682
48
1275
661692
61
206
354
257
1034
574
21
Old signals: host associated
clonal complexes
New signals: host associated
mobile elements

13. ‘early’ GWAS papers: Campylobacter (2013-2017)
Sheppard et al, PNAS 2013; Pascoe et al, Environmental Microbiology 2015; Monteil et al, Microbial Genomics 2016, Yahara & Meric et al,
Environmental Microbiology 2017 REVIEW: Sheppard et al, Nature Genetics Rev 2018
 K-mer based approach
 Limited to single clonal complex
 Small numbers of selected isolates

14. 0
200
400
600
800
1000
1200
1400
1600
1800
0 500 1000 1500
Map associated kmers to to 99 genes
76% of words map to 10 genes
IpxB
surE
Cj0294
Cj0295
panD
panC
panB
Cj0299
Cj0309
I II III
Sheppard et al. (2013) Genome-wide association study identifies vitamin B5 biosynthesis as a host specificity factor in Campylobacter. PNAS (2013)

15. Gene function and ecology: pantothenic acid (Vitamin
B5) synthesis
Sheppard et al. (2013) Genome-wide association study identifies vitamin B5 biosynthesis as a host specificity factor in Campylobacter. PNAS (2013)
Cellcount

16. Long term adaptation of Campylobacter to cattle
Mourkas et al 2020 PNAS 117 (20) 11018-11028; doi: 10.1073/pnas.1917168117

17. Campylobacter: Human infection in the UK – predominantly by eating
undercooked contaminated poultry products
Nichols et al 2012: BMJ Open e001179.
952
22
42
45
177
682
48
1275
661692
61
206
354
257
1034
574
21

18. Survival of Campylobacter through poultry processing
Yahara & Méric et al (2017) Genome-wide association of functional traits linked with Campylobacter jejuni survival from farm to fork.
Environmental Microbiology 19(1):361–380

19. Gene name Alias Predicted function
Associate
d in
cj0143c -
Putative periplasmic solute binding protein for ABC transport
system
ST-21
cj0576 lpxD UDP-3-O-acylglucosamine N-acyltransferase (EC 2.3.1.-) ST-21
cj0625 hypD Hydrogenase isoenzymes formation protein ST-21
cj0694 ppiD Putative periplasmic protein ST-21
cj1048c dapE
Succinyl-diaminopimelate desuccinylase (SDAP desuccinylase)
(EC 3.5.1.18) (N-succinyl-LL-2,6-diaminoheptanedioate
amidohydrolase)
ST-21
cj1049c - Putative LysE family transporter protein ST-21
cj1051c cjeI Restriction modification enzyme ST-21
cj1161c - Putative cation-transporting ATPase ST-21
cj1414c kpsC Capsule polysaccharide modification protein ST-21
cj1444c kpsD Capsule polysaccharide export system periplasmic protein ST-21
cj1569c nuoK
NADH-quinone oxidoreductase subunit K (EC 1.6.99.5) (NADH
dehydrogenase I subunit K) (NDH-1 subunit K)
ST-21
cj1309c - Uncharacterized protein ST-45
cj1364c fumC Fumarate hydratase class II (Fumarase C) (EC 4.2.1.2) ST-45
cj1366c glmS
Glutamine--fructose-6-phosphate aminotransferase [isomerizing]
(EC 2.6.1.16)
ST-45
cj1367c - Putative nucleotidyltransferase ST-45
cj1368 - Putative radical SAM domain protein ST-45
cj1373 - Putative ntegral membrane protein ST-45
cj1375 - Putative multidrug efflux transporter ST-45
cj1377c - Fumarate dehydrogenase ST-45
cj1378 selA L-seryl-tRNA(Sec) selenium transferase (EC 2.9.1.1) ST-45
Genes associated with survival throughout the poultry
processing chain
Capsule and cell
wall
Oxygen sensing
and tolerance
Same associated region
in ST-45 complex
ST-21
complex
ST-45
complex
Yahara & Méric et al (2017) Genome-wide association of functional traits linked with Campylobacter jejuni survival from farm to fork.
Environmental Microbiology 19(1):361–380

20. Back to the lab: clinical isolates grow better in an oxygen
enriched environment
Yahara & Méric et al (2017) Genome-wide association of functional traits linked with Campylobacter jejuni survival from farm to fork.
Environmental Microbiology 19(1):361–380

21. Testing function with isogenic mutants
Growth in oxygen-limited conditions Formate dehydrogenase activity
Gene name Alias Predicted function
Associated
in
cj1569c nuoK
NADH-quinone oxidoreductase subunit K (EC 1.6.99.5) (NADH dehydrogenase I subunit
K) (NDH-1 subunit K)
ST-21
cj1377c - Fumarate dehydrogenase ST-45
- nuoK is involved in NADH activity and switching from anaerobic to oxygen rich
environment. Mutants grow better in enhanced O2
- FDH is an excreted product of the resident microbiota activity is reduced in ∆cj1377
and clinical strains have reduced FDH activity.
Yahara & Méric et al (2017) Genome-wide association of functional traits linked with Campylobacter jejuni survival from farm to fork.
Environmental Microbiology 19(1):361–380

22. Campylobacter survival: biofilm formation
Pascoe et al (2015) Enhanced biofilm formation evolves from divergent genetic backgrounds in host generalist Campylobacter jejuni.
Environmental Microbiology 17(11):4779-89.

23. High biofilm formation also increases oxygen tolerance
Pascoe et al (2015) Enhanced biofilm formation evolves from divergent genetic backgrounds in host generalist Campylobacter jejuni.
Environmental Microbiology 17(11):4779-89.

24. Autogenous vaccine target:
“survivor strains”
Vaccinate breeders
with “survivor
strains”
Chicks not colonised with
Campylobacter in first 2 weeks
Protection from MABs Natural competition
in gut: discourage
colonisation of
“survivor strains”
Aim: less Campylobacter found on meat, cross protection from multiple lineages
“survivor”
“non-survivor”
(Sahin et al, 2003)
Real world applications: vaccine design
Jessica
Calland

25. Choose Campylobacter strains that
contain the most survival-
associated GWAS elements
 4 used to produce vaccine
Real world applications: vaccine design

26. New methods – specifically designed for bacteria
treeWAS
o Uses tree to weight associations
o 3 different association scores
o Variety of inputs
Collins & Didelot (2018)
PLoS Comput Biol 14(2): e1005958
pyseer
o Linear mixed-models to correct for
population structure
o Additional heritability scores
o Variety of inputs
Lees et al (2018) Bioinformatics 34 (24)
bugWAS
o Gene presence and SNVs
o Uses genetic background to correct for
population structure
o Targets strongly selected phenotypes
(like AMR)
Earle et al (2016) Nature Micro
Young et al (2019) eLife 22(8): e42486
dbGWAS
o Flexible inputs – so, can incorporate
accessory genome
o Corrects for population structure using
de Bruijn graphs
o Alignment and reference free method
Jaillard et al (2018) PloS Genetics 14(11):
e1007758

27. More detailed GWAS
outputs:
Campylobacter IBS
https://microreact.org/project/HySB
s-3Qz
Peters & Pascoe et al (unpublished)
treeWAS GitHub: https://github.com/caitiecollins/treeWAS
Collins & Didelot (2017): PloS Computational Biology 14(2): e1005958

28. Application:
GWAS-informed risk
prediction
https://microreact.org/project/HySB
s-3Qz
Peters & Pascoe et al (unpublished)

29. Can we predict virulence from GWAS outputs?
Meric and Mageiros et al. (2018) Nature Communications 9:5034; doi: 10.1038/s41467-018-07368-7

30.  >22,000 C. jejuni and C. coli isolates
 3 C.coli clades (including ancestral clades 2 & 3)
 Globally disseminated C. jejuni clonal complexes that contribute
towards a high % of human disease
Biggest challenge >> greater statistical power

31. Summary: how to design a GWAS study
To consider:
o Is it appropriate for my phenotype/dataset?
o Population structure - how will I correct for lineage effects?
o Which method shall I use?
Summary: what can we do with GWAS?
o Understand simple and complex phenotypes
o Identify novel AMR elements
o Host-associated genetic elements
o Link functional phenotypes
o Fishing experiments >> back to the lab
o Real world applications > vaccines?
o Predict virulence?

32. SHEPPARD
LABORATORY
http://www.sheppardlab.com
Guillaume
Méric
Leonardos
Mageiros
Susan
Murray
Elvire
Berthenet
Sion
Bayliss
Evangelos
Mourkas
Jessica
Calland
Campy GWAS:
David Kelly (Sheffield), Martin Maiden & Keith
Jolley (Oxford), Madhusudan Grover (Mayo
Clinic), Margaret Kosek (John Hopkins)
Stapyhlococcus GWAS:
Jukka Corander (Cambridge/Oslo), Ruth Massey
(Bristol), Dietrich Mack (Bioscientia), Stefan
Schwarz (Freie, Berlin), Herminia de Lencastre
(ITQB, Lisbon), Fintan Moriarty (Davos), Ed Feil
(Bath)
Helicobacter GWAS:
Koji Yahara (Tokyo), Kaisa Thorell (Stockholm),
Daniel Falush (Bath)