What Influences The Male Urogenital Tract Microbiome? Kirsty Lee - - PowerPoint PPT Presentation

Kirsty Lee Garson Supervisor: Prof Nicola Mulder

What Influences The Male Urogenital Tract Microbiome?

Testosterone Levels Age Circumcision Status Mucosal Immune Responses Male Urogenital Tract Microbiome STIs Sexual Behaviour

What influences the male urogenital tract microbiome?

Collection of Penile Swabs, Samples and Particpant Data

Visit 4 24 weeks after circumcision Visit 1 Circumcision Visit 3 12 weeks after circumcision Visit 2 2 weeks after circumcision Visit 4 24 weeks after circumcision

Edendale Hospital, KZN (n=48) Chris Hani Baragwanath Hospital, Gauteng (n=33) 81 participants 13–24 years old

Overview

90 samples Circumcision Status STI Status Sexual History Ethnicity

Overview

Edendale Hospital, KZN (n=48) Chris Hani Baragwanath Hospital, Gauteng (n=33) 81 participants 13–24 years old

Circumcision Status STI Status Sexual History Ethnicity

HPV: Human papillomavirus (n=37) MG: Mycoplasma genitalium (n=4) TV: Trichomonas vaginalis (n=2) CT: Chlamydia trachomatis (n=8) NG: Neisseria gonorrhoeae (n=1) HSV-2: Herpes simplex virus 2 (n=1) Neg: None of the above (n=13)

Overview

Coloured (n=5) Sotho (n=4) Swati (n=1) Tsonga (n=1) Tswana (n=4) Xhosa (n=3) Zulu (n=36) CT+HPV+ CT+MG+HPV+ CT+MG+NG+HPV+ CT+TV+HPV HPV+ HSV2+HPV+ TV+HPV+ CT+ CT+MG+ Sexually-inexperienced (n = 17) Sexually-experienced (n = 45) Post-circumcision (n=20) Pre-circumcision (n=70)

81 participants, 90 samples

Aims What is the composition of these bacterial communities? Which factors influence its composition? Can we train a model to detect patterns in the microbiome?

Aim One: Derive sequence data from clinical samples

Collect samples Extract DNA Amplify and sequence a marker gene Group sequences by 97% similarity Assign taxonomy using reference databases

Group 1 Group 2 Group 3

GATACAGAGATGCAT GTATACAGAGATGCAT GGATACAGAGATGCAT GATCACAGAGATGCAT ATAGATACAGAGATCAT ATAGTATACAGAGACAT TATGATACAGAGACAT TATGTATACAGAACAT TAGGGATACAGACAT

Align sequences and create a phylogeny

TATGATACAGAGAC Bacteria_Firmicutes_...

Kingdom Phylum Class Order Family Genus Species "Bacteria" "Firmicutes" "Bacilli" "Lactobacillales" "Lactobacillaceae" "Lactobacillus" "iners"

Aim One: Derive sequence data from clinical samples

Aim Two: Analyse the microbial composition of samples

What is the composition of these bacterial communities? How does it differ between healthy individuals and those with disease? (marker gene analysis)

Exploratory Analysis (alpha & beta diversity, heatmaps, etc.) (phyloseq, labdsv, vegan, ape, ggplots) Marker Discovery/Statistical Testing (differential abundance testing/indicator species analysis) (metagenomeSeq, labdsv) Microbiome Pattern Recognition (random forest classifier) (scikit-learn)

2.1 Measure diversity within samples

Before

After

Circumcision Simpson Index (inverted) Observed Richness Chao1 Shannon Index SChao1=Sobs+

n1

2

2n 2

2

PCoA.1(21.3%) PCoA.2(8.7%)

Before Circumcision After Circumcision

2.2 Measure diversity across samples

After Circumcision Before Circumcision Name of sample Species

2.2 Measure diversity across samples

2.2 Measure diversity across samples

After Circumcision Before Circumcision Name of sample Species

2.3 Assess sample composition

Relative abundance Name of sample Family

2.4 Compare sample composition across groups

Before Circumcision After Circumcision Genus

Pre-circumcision samples Post-circumcision samples

• No. of reads

2.5 Compare the number of times microbes of interest occur across groups

Genus

2.6 Pattern recognition

Explore potential underlying structures within these bacterial communities by grouping them into subcommunities:

• certain OTUs vary together
• functional redundancy exists
• many zero values in microbiome data makes analysis more challenging
• random forest classifier assumes independence of features

Before Circumcision After Circumcision Before Circumcision 15 1 After Circumcision 2 3

2.6 Pattern recognition: predicting group membership using a random forest classifier

Actual Predictions

2.6 Pattern recognition: predicting group membership using a random forest classifier

OTU 011 Peptoniphilus species OTU 065 Escherichia coli OTU 150 Lactobacillus species OTU 064 Lactobacilus iners OTU 039 Clostridiales order [Mogibacteriaceae] family OTU 159 Enterococcus species Relative Importance

Which species are most influential in making predictions?

Acknowledgements

Division of Immunology, University of Cape Town Department of Medicine, Edendale Hospital, Kwa-Zulu Natal Computational Biology Division, University of Cape Town

• Dr. Pierre-Yves Lablanche (AIMS)
• Dr. Arun Aniyan (SKA)
• Dr. Michelle Lochner (AIMS)

Perinatal HIV Research Unit, Soweto, Gauteng Study participants Professor Nicola Mulder Professor Heather Jaspan Clive Gray, Jo-Ann Passmore, Rushil Harrayparsad, Hoyam Gamieldien, Selena Ferrian, Abraham Olivier, Heather Jaspan, Katie Viljoen, Dirk Lang, Susan Cooper, Nyari Chigorimbo, Lungile Mayiza, Doug Wilson (Edendale hospital), Marcus McGilvray (WizzKids), Rusha Govender (WizzKids), Hillary Mukudu, Janan Dietrich, Neil Martinson (PHRU); Nono Mkhize, Raveshni Durgiah, Lynn Morris (NICD), Tom Hope, Minh Dinh, Gianguido Cianci, Francesca Chiodi, Sylvie Amu