Loss of Microbial Diversity and Body Site Heterogeneity in Individuals with Hidradenitis

Home , Body odor

bioRxiv preprint doi: https://doi.org/10.1101/612457; this version posted April 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Loss of microbial diversity and body site heterogeneity in individuals with Hidradenitis

Suppurativa

Andrea M. Schneider1*, Lauren C. Cook1*, Xiang Zhan2, Kalins Banerjee2, Zhaoyuan Cong1,

Yuka Imamura-Kawasawa3, Samantha L. Gettle1, Amy L. Longenecker1, Joslyn S. Kirby1, and

Amanda M. Nelson1

1) Department of Dermatology, Penn State Milton S. Hershey Medical Center, Hershey, PA,

USA

2) Department of Public Health Sciences, Penn State College of Medicine, Hershey, PA, USA

3) Institute for Personalized Medicine, Departments of Biochemistry and Molecular Biology

and Pharmacology, Penn State College of Medicine, Hershey, PA, USA

*These authors contributed equally to this work.

RUNNING TITLE Microbe dysbiosis in Hidradenitis Suppurativa

AUTHOR CONTACT AND ORCID

Andrea M. Schneider ([email protected]); ORCID: 0000-0002-9725-4634

Lauren C. Cook ([email protected]); ORCID: 0000-0002-8987-4286; Current address: 2201

Forest Hills Drive, Suite 7, Harrisburg, PA, 17112

Xiang Zhan ([email protected]); ORCID: 0000-0001-9650-143X

Kalins Banerjee ([email protected]); ORCID: 0000-0003-4314-5114

Zhaoyuan Cong ([email protected]); ORCID: 0000-0001-9895-2912

Yuka Immamura-Kawasawa ([email protected]); ORCID: 0000-0002-8638-

6738

Samantha L. Gettle ([email protected]); ORCID: 0000-0002-5471-8670

Amy L. Longenecker ([email protected]); ORCID: 0000-0002-3804-2789

Joslyn S. Kirby ([email protected]); ORCID: 0000-0002-6741-3796

Amanda M. Nelson ([email protected]); ORCID: 0000-0001-6375-8706

AUTHOR CONTRIBUTIONS

LEAD AUTHORS

*AMS: conceptualization, data curation, formal analyses, investigation, software, visualization,

and writing.

*LCC: conceptualization, data curation, funding acquisition, methodology, resources, and

writing.

AMN: conceptualization, data curation, formal analyses, funding acquisition, methodology,

project administration, resources, supervision, visualization, and writing.

*equal contributions

SUPPORTING AUTHORS

XZ: formal analyses, software, visualization, and writing.

KB: formal analyses, software, and visualization.

ZC: investigation and methodology.

YIK: methods, investigation, resources, and writing.

SLG: project administration, data curation, and resources.

ALL: project administration, data curation, and resources.

JSK: resources, formal analyses, and writing.

All authors edited the manuscript, gave final approval for publication, and agreed to be

accountable for the work.

CORRESPONDING AUTHOR

Amanda M. Nelson, PhD

Assistant Professor

Department of Dermatology

Penn State University College of Medicine

500 University Drive

C7801 BMR

Hershey, PA, 17033

[email protected]

Phone: 717-531-0003 x284512

ABBREVIATIONS

FWER: family-wise error rate

HS: Hidradenitis Suppurativa

HSN: HS non-lesional

HSL: HS lesional

KEGG: Kyoto Encyclopedia of Genes and Genomes

MCC: mock-community control

MiRKAT: Microbiome Regression Kernel Association Test

OTUs: operational taxonomic units

PICRUSt: Phylogenetic Investigation of Communities by Reconstruction of Unobserved States

PSU: pilosebaceous unit

STAMP: Statistical Analysis of Metagenomic Profile

ABSTRACT

Hidradenitis Suppurativa (HS) is a chronic, scarring, inflammatory skin disease affecting

hair follicles in axillae, inguinal, and anogenital regions. Dysbiosis in HS patients compared to

healthy subjects is documented. However, whether dysbiosis is specific to particular body sites

or skin niches is unknown. We investigated the follicular and skin surface microbiome of the

axilla and groin of HS patients (n=11) and healthy individuals (n=10) using 16S rRNA gene

sequencing (V3-V4). We sampled non-lesional (HSN) and lesional skin (HSL) of HS patients. -

diversity was significantly decreased (p<0.05) in HSN and HSL skin compared to normal skin

with loss of body site and skin niche heterogeneity in HS samples. The relative bacterial

abundance of specific microbes was also significantly different between normal and HSN (15

genera) or HSL (21 genera) skin. Smoking and alcohol use influenced the -diversity (p<0.08) in

HS skin. We investigated metabolic profiles of bacterial communities in HS and normal skin

using a computational approach. Metabolism, Genetic Information Processing, and

Environmental Information Processing were significantly different between normal and HS

samples. Altered metabolic pathways associated with dysbiosis of HS skin suggest mechanisms

underlying the disease pathology and information about treatment with drugs targeting those

pathways.

INTRODUCTION

Hidradenitis Suppurativa (HS) is a chronic scarring inflammatory skin disease affecting

the pilosebaceous units (PSU) of the axilla, inframammary folds, groin, and buttocks. Clinical

features include painful nodules, abscesses, draining sinus tracts, and fistulas. Antibiotics are

commonly prescribed to treat HS, however, lesions are often sterile. With the increased

prevalence of antibiotic resistance, understanding the role of bacteria in HS is critical.

The skin microbiome varies by individual, body site, age, and disease status (Grice et al.

2009; Kong et al. 2012; Oh et al. 2014). Dysbiosis, or microbial imbalance, of the skin

microbiome occurs in atopic dermatitis, epidermolysis bullosa, chronic non-healing skin wounds,

and others (Fuentes et al. 2018; Kalan et al. 2016; Kong et al. 2012).

Dysbiosis has also been reported in HS (Guet-Revillet et al. 2017; Nikolakis et al. 2017;

Ring et al. 2017b; Ring et al. 2017a). Collectively, the HS microbiome has been studied by

invasive and non-invasive sampling of non-lesional and lesional skin using culture and next

generation sequencing analyses. The only consistent finding between these studies was that

bacterial dysbiosis exists in HS skin compared to normal skin. However, where in the skin the

dysbiosis is occurring and whether or not this dysbiosis alters microbiome functions remains

unexplored.

In this study, we investigated whether dysbiosis is specific to a particular body site or

skin niche and applied a computational approach to understand the biological impact of dysbiotic

communities in HS patients. We sampled both the follicular and skin surface microbiome of the

axilla and groin of HS patients (n=11) and healthy individuals (n=10) followed by 16S rRNA

gene sequencing (V3-V4). This study reveals that loss of bacterial diversity in HS patients

extends to both skin niches. Alterations in the metabolic profile due to dysbiosis may contribute

to HS symptoms and pathology.

RESULTS

Study subjects and sequencing methods

Ten normal subjects and 11 subjects with HS were enrolled; including males and females,

ages 18-60 years after providing informed written consent under an IRB approved protocol. The

diagnosis of HS was confirmed by a dermatologist. For subjects with HS, the Hurley Stage,

disease duration, and family history of HS and other inflammatory conditions were documented.

Use of medications, alcohol, hygiene products and smoking history were documented for all

subjects (Table S1).

Cyanoacrylate follicular biopsy (glue) and swab samples were collected from the axilla

vault and inguinal crease (groin) for each individual. Four samples were collected from each

normal individual, while both non-lesional (HSN) and lesional (HSL) skin was sampled in each

HS patient (8 samples) (Figure 1a). HSN skin appeared clinically normal and was ~5 cm away

from an active lesion; lesional skin was defined by an inflammatory nodule. We specifically

chose this sampling distance to keep within the same anatomical body area (ie: axilla vault). A

total of 128 patient samples, four negative sampling controls, and one positive mock community

control (MCC) (Figure S1) were collected.

16S rRNA gene sequencing V1-V3 is preferred in skin microbiome studies, while V4 is

more common for gut microbiome research (Meisel et al. 2016). We chose the V3-V4 region

(Table S2; Supplemental Methods M3) in an effort to maximize identification of both skin and

gut communities, as common gut microbiota have been reported within HS lesions (Ring et al.

2017b). Only samples with a minimum of 75,000 sequences were included in subsequent

analyses (n = 117) (Table S3).

In total, we identified 994 operational taxonomic units (OTUs) following the convention

of 97% similarity for OTU assignment. OTUs that did not reach an abundance of 0.5% or higher

in at least one sample were excluded from further analysis as these were considered potential

sequencing errors (Li 2015). After filtering, 272 OTUs became the focus of downstream

statistical analysis, belonging to 146 distinct genera and 1 unclassified group. When body site or

skin niche were compared, samples were paired by body site, skin niche, and non-lesional or

lesional skin to account for intrapersonal variability.

Bacterial -diversity is significantly decreased on HS compared to normal skin

First, we compared bacterial composition between normal, HSN, and HSL regardless of

sampling method and body site. There was no significant difference in -diversity (Shannon

diversity index) between normal, HSN, and HSL skin (Figure 1b). However, a significant loss of

-diversity (Bray-Curtis dissimilarity index) was observed in both HSN and HSL compared to

normal skin (Figure 1c). Overall differences in diversity of the bacterial communities were

assessed with the Microbiome Regression Kernel Association Test (MiRKAT) (Zhao et al.

2015). The community level composition in normal skin was distinctly different from the

composition of both HSN (p=0.036) and HSL (p=0.010) skin (Figure 1c).

We investigated community level diversity between HSN and HSL skin. No statistical

difference in bacterial composition between HSN and HSL skin (p=0.4404) was found, in

contrast to prior reports (Guet-Revillet et al. 2017; Ring et al. 2017b). When comparing

individual taxa between HSN and HSL skin for each individual, no single genus was

significantly different under a family-wise error rate (FWER) of 0.05, indicating homogeneous

bacterial compositions. Overall, the bacterial communities between HSN and HSL samples were

highly similar with a loss of -diversity in HS skin compared to normal skin.

The relative abundance of individual taxa were compared between HS and normal skin.

Fifteen (9 of the top 20) genera were significantly different between normal and HSN samples

and 21 (14 of the top 20) genera were significantly different between normal and HSL samples

(Figure 1d, Table S4, S5). With the high degree of similarity between HSN and HSL skin, we

averaged these samples together for each HS patient and compared the top 10 most abundant

genera between HS skin and healthy skin (Figure 1e). Staphylococcus and Corynebacterium

comprised ~35% of the bacteria in both normal and HS skin; however, Corynebacterium was ~2

fold more abundant in HS skin. Like previous studies, Cutibacterium (formally

Propionibacterium (Scholz and Kilian 2016)) was more abundant in normal skin (18.8% vs 1%,

p<0.05), and Peptoniphilus and Porphyromonas were more abundant in HS skin (Ring et al.

2017b) (Figure 1e, Table S4, S5). Interestingly, Arcanobacterium was detected in samples from

9 of 11 HS patients (mean relative abundance 0.07%) but not detected in any sample from

normal subjects. Overall, the HS disease state affects the relative abundance of a number of skin

commensals and opportunistic pathogens.

Body site and skin niche heterogeneity is lost in HS skin compared to normal skin

Skin microbiota vary by anatomical body site in healthy individuals with the

microenvironment influencing bacterial colonization; skin sites with similar moisture, lipids, and

pH have a more similar composition than disparate sites (Grice et al. 2008). Moreover, the skin’s

surface and follicular microbiomes are not 100% identical (Grice et al. 2008; Hall et al. 2018).

Because HS is a disease of the PSU that commonly affects the axilla and groin, we investigated

whether body site or skin niche are different in HS compared to normal subjects. For these

analyses, we stratified our data by body site or skin niche and included only ‘paired’ samples to

account for intrapersonal variability (Figures 2-3, S2-3). Within HS skin, whether HSN or HSL,

we observed no difference in - or -diversity when comparing between axilla vs. groin and skin

surface vs. follicle, highlighting the relative homogeneous bacterial populations at both

anatomical body locations and within skin niches. However, in healthy individuals, there are

differences between the axilla and groin as well as between skin’s surface and follicle. In

general, -diversity is increased in the groin compared to the axilla (Figure 2). Moreover, the

skin’s surface is more diverse (-diversity) than the follicle, especially within the axilla (Figure

3). These data highlight the importance of both sampling method and body site sampling to the

outcome of microbiome studies. Taken together, these data demonstrate that there is a loss of

heterogeneity between body sites and skin niches in HS patients compared to healthy controls.

Individual lifestyle factors are associated with altered bacterial diversity in HS skin

We investigated correlations between bacterial diversity and subject demographics

(Table S1). For each individual, we calculated the average relative abundance of each bacterial

genus, including all samples regardless of location or sampling method. We used the MiRKAT

association test (-diversity) to investigate associations with sex, age, race, and use of hygiene,

alcohol, or tobacco products (Figure 4a). This study was not sufficiently powered to detect

significant changes associated with sex, race, or medication use (Table S1).

HS disease severity is classified by Hurley Stage (I, II, III). Previous studies

demonstrated increased bacterial diversity with increasing Hurley Stage (Guet-Revillet et al.

2017; Nikolakis et al. 2017), however, due to limited sample size, no significant difference in -

diversity was found between disease stage in our study (Figure 4a-b, Table S6). The relative

abundance of each genus was similar between Hurley Stage II and III except for Prevotella and

Atopobium, which were significantly more abundant in stage III compared to stage II (Table S7).

Lifestyle choices may impact the microbiome. Both tobacco and alcohol use are linked to

altered gut microbiomes (Capurso and Lahner 2017). However, the impact on the skin

microbiome is not clear. Tobacco use is common in HS patients and is associated with increased

disease severity (Akdogan et al. 2018; Sartorius et al. 2009). The number of HS patients who

smoked (5 of 11) significantly differed from normal subjects (0 of 10) (Table S6). In HS

patients, a trend in altered -diversity (p = 0.063) between smoking and non-smoking was

observed (Figure 4c). Alcohol use is also common among HS patients (Garg et al. 2018).

Alcohol use also influenced bacterial diversity (p=0.078) between HS patients who were alcohol

users vs. those who were not (6 and 5 patients, respectively) (Figure 4d). Together, in our HS

patient population, these results suggest that lifestyle factors can impact skin microbiome

diversity.

KEGG Ontology indicates functional differences between microbiota of HS and normal

skin

Like previous studies, our data illustrated decreased-diversity between normal and HS

skin. However, the biological impact of this altered diversity remains unexplored. Using a

computational approach with PICRUSt, we investigated the metabolic profiles of HS and normal

bacterial communities (Langille et al. 2013). We used KEGG pathway analysis to identify

significant functional differences between the microbiota of normal, HSN, and HSL skin (Figure

5a, Supplemental Methods M4). Several KEGG Orthology groups were different between

normal, HSN, and HSL samples: Level 1 (n=3), Level 2 (n=15), and Level 3 (n=52) (Table S8).

At the highest categorical level, Metabolism, Genetic Information Processing, and

Environmental Information Processing were significantly different between normal and HSN or

HSL skin (Table S8). Several Metabolism pathways were enriched in normal compared to HS

skin including: Pyruvate, Arginine and proline; Propanoate; Retinol; Valine, leucine, and

isoleucine biosynthesis; and Valine, leucine, and isoleucine degradation. Purine and pyrimidine

metabolism were significantly enriched in both HSN and HSL compared to normal skin (Figure

5b). Within Genetic Information Processing, all KEGG Ortholog terms were significantly

enriched in HSN and HSL compared to the normal skin (Figure 5c); this combined with

increased purine and pyrimidine metabolism suggests increased microbial turnover in HS

compared to normal skin.

Metabolic pathways are influenced by different genera in normal and HS skin.

We next identified the specific genera associated with KEGG Orthologs involved in key

metabolic pathways. To do this, we combined all significant KEGG Orthologs (K number)

affiliated with each pathway and determined the top 10 contributing genera to each pathway

using PICRUSt (Figure 6, Table S9). The genera contributing to each pathway differed between

HS and normal skin. For example, Cutibacterium contributed significantly to propanoate

metabolism in normal skin (p=0.004), while Corynebacterium was the dominant contributor in

HS skin (p=0.062) (Figure 6a, Table S10). Cutibacterium was the dominant contributor to

retinol metabolism in normal skin, while no single genera dominated in HS skin (p=0.0003)

(Figure 6b, Table S10). Corynebacterium is a major contributor to multiple amino acid

metabolism pathways in HS skin including valine, leucine, and isoleucine biosynthesis and

degradation as well as arginine and proline metabolism (Figure 6c-e, Table S10). Overall,

dysbiosis not only changes which bacteria are present, but also impacts the biological functions

of the skin microbial community.

DISCUSSION

Dysbiosis of the skin microbiota is common in many inflammatory skin diseases. Our

data are in accordance with previous studies documenting dysbiosis within HS skin compared to

normal skin (Guet-Revillet et al. 2017; Nikolakis et al. 2017; Ring et al. 2017b; Ring et al.

2017a). Unlike normal subjects that demonstrate heterogeneity of bacterial communities across

body sites, within the follicle and at the skin surface; we demonstrate that this heterogeneity is

lost in HS skin. Moreover, we identified metagenomic functional differences between HS and

normal skin. Dysbiosis in HS skin may be the result of the disease process, as suggested by

increased metabolic turnover, or may contribute to the disease process as suggested by altered

amino acid metabolism pathways.

Smoking impacts HS severity and cessation is often recommended to ease symptoms

(Sartorius et al. 2009). While differences in β-diversity based on lifestyle choices did not reach

statistical significance in our study, smoking and alcohol likely impact the skin microbiome in

HS patients. Both smoking and alcohol use increase inflammation (Rom et al. 2013; Szabo and

Saha 2015), which may alter skin bacteria composition through host-microbiome crosstalk.

While smoking and alcohol use impact β-diversity of the gut microbiome (Capurso and Lahner

2017), to our knowledge ours is the first study to link smoking and alcohol use to an altered skin

microbiome.

We uncovered changes in specific bacterial compositions comparing HS and normal skin.

Cutibacterium, a commensal bacterium found within the PSU, was significantly decreased in

HSL and HSN compared with normal skin, consistent with other published work (Ring et al.

2017b). Decreased abundance of Cutibacterium in HS is likely due to destruction of the PSU

during disease progression (Kamp et al. 2011). Loss of Cutibacterium in non-lesional skin of HS

subjects suggests that disruption of PSU architecture may precede the development of clinically

detectable HS lesions. Alternatively, antimicrobial peptides and inflammatory mediators initiated

by follicular keratinocytes in the early stages of HS may form a microenvironment inhospitable

to Cutibacterium colonization (Hotz et al. 2016). Short-chain fatty acids, including propionic

acid, lactic acid, and acetic acid, produced by Cutibacterium have antimicrobial activity against

Staphylococcus aureus and other opportunistic pathogens (Shu et al. 2013). Loss of

Cutibacterium together with a decrease in propanoate metabolism may relate to the observed

increased abundance of opportunistic pathogens including Peptoniphilus, Porphyromonas,

Agrobacterium, Pseudomonas, and Arcanobacterium in HS skin. Peptoniphilus and

Porphyromonas have been previously identified in HS lesions (Ring et al. 2017b), and all these

pathogens have been isolated from healthy and immunocompromised patients with skin

infections, chronic wounds, abscesses, and pressure ulcers (Gellatly and Hancock 2013; Hulse et

al. 1993; Kim et al. 2016; Murphy and Frick 2013; Tan et al. 2006). Methods that stabilize the

Cutibacterium community to maintain homeostasis and keep pathogenic communities at bay

may help improve symptoms of HS.

Using metagenomics, we identified several metabolic pathways that are different between

HS and normal skin and these differences may impact HS pathology. Amino acid metabolism

may impact body odor and skin pH. Malodor significantly impairs the quality of life in HS

patients (Alavi et al. 2018). The breakdown of isoleucine to isovaleric acid contributes to “sour”

odor (Lam et al. 2018), and we identified altered isoleucine biosynthesis and degradation

pathways in HS skin. Corynebacterium significantly contributed to these pathways. In other

studies, Corynebacterium produced thioalcohols (3M3SH, 3M2H, HMHA) which are linked

with “sour” malodor in adults (Bawdon et al. 2015; Fredrich et al. 2013; Taylor et al. 2003).

Interestingly, in individuals who do not use underarm hygiene products, as some HS patients do

not, Corynebacterium was the most abundant genera (Urban et al. 2016). Even though

Corynebacterium is a common skin commensal, in the context of HS, it may trigger insidious

inflammatory and odor changes. Altered arginine and proline metabolism in HS skin may incite

an acidic skin pH, which may further contribute to the state of dysbiosis.

Finally, our data suggest that rather than targeting the bacterial dysbiosis of HS with

antibiotics, there is an opportunity to target the metabolic profiles of bacteria, with the goal of

restoring a “normal metabolome.” For example, we observed a significant decrease in retinol

metabolism among bacteria in HS skin. Apart from pharmacologic intervention with systemic

retinoids, alternative approaches to restore a normal metabolome using prebiotics, probiotics or

microbiome transplants may be of interest in HS, similar to the use of skin microbiome

transplants in atopic dermatitis (Blok et al. 2013; Nakatsuji et al. 2017). In acne individuals,

vitamin B12 supplementation impacted C. acnes B12 metabolic pathways suggesting that host

interventions can positively or negatively impact the bacterial metabolome (Kang et al. 2015).

Influencing the bacterial metabolome through vitamin supplementation in HS skin may be

possible as results from this study indicated that several amino acid and vitamin metabolism

pathways are altered in HS skin. Computational analyses provide initial clues to understanding

the functions of individual bacteria within the microbial community; however future studies

confirming these hypotheses are warranted.

In conclusion, bacterial dysbiosis is present in HS compared to normal skin. This

dysbiosis is apparent at the skin’s surface, in the follicle, and at distinct body sites. Since

dysbiosis can impact the biological function of the skin microbiota, understanding the bacterial

metabolome and how it influences the host may be critical to determining the contribution of

bacteria to HS pathology and may suggest novel therapeutic options.

MATERIALS AND METHODS

Detailed methods are provided in Supplementary Materials.

Study participants

Twenty one volunteers (10 normal and 11 HS) were recruited under an approved IRB protocol

(Penn State College of Medicine) and provided written informed consent prior to the study.

Subjects included males and females, ages 18-65. Subjects were excluded if they had known

allergies to cyanoacrylates or had used topical antimicrobials in the previous 2 weeks or systemic

antibiotics in the previous month.

Microbiome sampling and DNA extraction

Two methods were used to sample the skin microbiome of the axilla and groin—cyanoacrylate

glue to sample the pore/follicle and swabs to sample the skin surface. Non-lesional and lesional

skin were sampled in HS patients, while healthy skin was sampled in normal subjects. Mock

community control (MCC; ATCC) (Figure S3) and negative sample controls (i.e. no skin

contact) were processed alongside samples from human subjects.

16S rRNA gene sequencing (V3-V4)

16S rRNA gene sequencing was performed at the Genome Sciences and Bioinformatics Core the

Penn State College of Medicine. Sequencing was performed using Illumina MiSeq (paired 300

bp reads) for the V3-V4 region using a two-step, tailed PCR approach that has been validated by

Illumina (Illumina, Inc, San Diego, CA).

Sequence processing and taxonomic mapping

Fastq samples were processed using Pear (version 0.9.6) to join paired end reads, and

subsampling was performed at a depth of 75,000 reads using seqtk (version 1.0-r82). Samples

that did not meet minimum quality scores or subsampling requirements were removed from

analysis (Table S2). Operational Taxonomic Units (OTUs) were determined using QIIME

(version 1.9.1_py2.7.11) (Caporaso et al. 2010). Reads were mapped with GreenGenes 16S

rRNA gene reference database (version 13.8) at 97% sequence similarity (McDonald et al. 2012)

using the default OTU picking method, uclust (Edgar 2010). Data was converted to biom format

(genus level) and prepared for downstream statistical analyses in R. Genera that had ≥ 2%

relative abundance in negative controls were removed from all downstream analyses.

Diversity analyses and statistics

All data analyses were conducted using the statistical software R. Diversity between samples

were compared by calculating both -diversity (Shannon diversity index – within sample

diversity) and -diversity (Bray-Curtis dissimilarity index – between-sample diversity; 0 =

identical communities, 1 = dissimilar communities). Differences in Shannon index between

groups were evaluated by the Wilcoxon rank-sum test. Differences in Bray-Curtis dissimilarity

index were assessed by MiRKAT (Zhao et al. 2015) or GHT (Zhao et al. 2018) for HSN-HSL

paired samples. Both MiRKAT and GHT evaluate the overall shift in microbiome composition.

Differential abundances of individual taxon were tested using the Wilcoxon rank sum test

followed by the Bonferroni correction to preserve the family-wise error rate (FWER) of 0.05.

Functional analyses

Functional differences of OTUs were determined using PICRUSt (Langille et al. 2013).

Functional data were generated to level 3 functional categories of the Kyoto Encyclopedia of

Genes and Genomes (KEGG) orthologs (Kanehisa and Goto 2000). Statistical Analysis of

Metagenomic Profile (STAMP) package (Parks et al. 2014) was used to make comparisons

between multiple groups (normal, HSN, and HSL) using the Kruskal-Wallis H-test with

Bonferonni correction.

CONFLICT OF INTEREST

The authors state no conflict of interest.

ACKNOWLEDGEMENTS

This work was supported by a research grant from the American Acne and Rosacea Society to

LCC and the Department of Dermatology Research Endowment at the Penn State College of

Medicine. The authors thank Dr. Dave Adams (PSU Dermatology) for assistance with patient

recruitment and clinical expertise and Jacob B. Hall, PhD for advice and guidance with data

analyses. We appreciate language editing services and writing instruction provided by DerMEDit

(www.dermedit.com).

REFERENCES

Akdogan N, Alli N, Uysal PI, Topcuoglu C, Candar T, Turhan T. Visfatin and insulin levels and

cigarette smoking are independent risk factors for hidradenitis suppurativa: a case–control study.

Arch. Dermatol. Res. Springer Berlin Heidelberg; 2018;310(10):785–93

Alavi A, Farzanfar D, Lee RK, Almutairi D. The Contribution of Malodour in Quality of Life of

Patients With Hidradenitis Suppurativa. J. Cutan. Med. Surg. 2018;22(2):166–74

Bawdon D, Cox DS, Ashford D, James AG, Thomas GH. Identification of axillary

Staphylococcus sp. involved in the production of the malodorous thioalcohol 3-methyl-3-

sufanylhexan-1-ol. Enright M, editor. FEMS Microbiol. Lett. Oxford University Press;

2015;362(16):fnv111

Blok JL, van Hattem S, Jonkman MF, Horváth B. Systemic therapy with immunosuppressive

agents and retinoids in hidradenitis suppurativa: a systematic review. Br. J. Dermatol.

Wiley/Blackwell (10.1111); 2013;168(2):243–52

Caporaso JG, Kuczynski J, Stombaugh J, Bittinger K, Bushman FD, Costello EK, et al. QIIME

allows analysis of high-throughput community sequencing data. Nat. Methods. NIH Public

Access; 2010;7(5):335–6

Capurso G, Lahner E. The interaction between smoking, alcohol and the gut microbiome. Best

Pract. Res. Clin. Gastroenterol. Baillière Tindall; 2017;31(5):579–88

Edgar RC. Search and clustering orders of magnitude faster than BLAST. Bioinformatics.

Oxford University Press; 2010;26(19):2460–1

Fredrich E, Barzantny H, Brune I, Tauch A. Daily battle against body odor: towards the activity

of the axillary microbiota. Trends Microbiol. Elsevier Current Trends; 2013;21(6):305–12

Fuentes I, Guttmann-Gruber C, Tay ASL, Piñón Hofbauer J, Denil SLIJ, Reichelt J, et al.

Reduced Microbial Diversity Is a Feature of Recessive Dystrophic Epidermolysis Bullosa-

Involved Skin and Wounds. J. Invest. Dermatol. Elsevier; 2018;

Garg A, Papagermanos V, Midura M, Strunk A, Merson J. Opioid, alcohol, and cannabis misuse

among patients with hidradenitis suppurativa: A population-based analysis in the United States.

J. Am. Acad. Dermatol. 2018;79(3):495–500.e1

Gellatly SL, Hancock REW. Pseudomonas aeruginosa : new insights into pathogenesis and host

defenses. Pathog. Dis. Oxford University Press; 2013;67(3):159–73

Grice EA, Kong HH, Conlan S, Deming CB, Davis J, Young AC, et al. Topographical and

temporal diversity of the human skin microbiome. Science. NIH Public Access;

2009;324(5931):1190–2

Grice EA, Kong HH, Renaud G, Young AC, Program NCS, Bouffard GG, et al. A diversity

profile of the human skin microbiota. Genome Res. Genetics and Molecular Biology Branch,

National Human Genome Research Institute, National Institutes of Health, Bethesda, Maryland

20892, USA.; 2008;18(7):1043–50

Guet-Revillet H, Jais J-P, Ungeheuer M-N, Coignard-Biehler H, Duchatelet S, Delage M, et al.

The Microbiological Landscape of Anaerobic Infections in Hidradenitis Suppurativa: A

Prospective Metagenomic Study. Clin. Infect. Dis. Oxford University Press; 2017;65(2):282–91

Hall JB, Cong Z, Imamura-Kawasawa Y, Kidd BA, Dudley JT, Thiboutot DM, et al. Isolation

and Identification of the Follicular Microbiome: Implications for Acne Research. J. Invest.

Dermatol. Elsevier; 2018;138(9):2033–40

Hotz C, Boniotto M, Guguin A, Surenaud M, Jean-Louis F, Tisserand P, et al. Intrinsic Defect in

Keratinocyte Function Leads to Inflammation in Hidradenitis Suppurativa. J. Invest. Dermatol.

Elsevier; 2016;136(9):1768–80

Hulse M, Johnson S, Ferrieri P. Agrobacterium Infections in Humans: Experience at One

Hospital and Review. Clin. Infect. Dis. Oxford University Press; 1993;16(1):112–7

Kalan L, Loesche M, Hodkinson BP, Heilmann K, Ruthel G, Gardner SE, et al. Redefining the

Chronic-Wound Microbiome: Fungal Communities Are Prevalent, Dynamic, and Associated

with Delayed Healing. MBio. American Society for Microbiology (ASM); 2016;7(5)

Kamp S, Fiehn AM, Stenderup K, Rosada C, Pakkenberg B, Kemp K, et al. Hidradenitis

suppurativa: a disease of the absent sebaceous gland? Sebaceous gland number and volume are

significantly reduced in uninvolved hair follicles from patients with hidradenitis suppurativa. Br.

J. Dermatol. Wiley/Blackwell (10.1111); 2011;164(5):1017–22

Kanehisa M, Goto S. KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res.

Oxford University Press; 2000;28(1):27–30

Kang D, Shi B, Erfe MC, Craft N, Li H. Vitamin B12 modulates the transcriptome of the skin

microbiota in acne pathogenesis. Sci. Transl. Med. NIH Public Access; 2015;7(293):293ra103

Kim B, Pai H, Hwang KT, Lee Y. Porphyromonas pogonae identification from a soft tissue

infection: The first human case. Anaerobe. Academic Press; 2016;42:37–9

Kong HH, Oh J, Deming C, Conlan S, Grice EA, Beatson MA, et al. Temporal shifts in the skin

microbiome associated with disease flares and treatment in children with atopic dermatitis.

Genome Res. Cold Spring Harbor Laboratory Press; 2012;22(5):850–9

Lam TH, Verzotto D, Brahma P, Ng AHQ, Hu P, Schnell D, et al. Understanding the microbial

basis of body odor in pre-pubescent children and teenagers. Microbiome. BioMed Central;

2018;6(1):213

Langille MGI, Zaneveld J, Caporaso JG, McDonald D, Knights D, Reyes JA, et al. Predictive

functional profiling of microbial communities using 16S rRNA marker gene sequences. Nat.

Biotechnol. Nature Publishing Group; 2013;31(9):814–21

Li H. Microbiome, Metagenomics, and High-Dimensional Compositional Data Analysis. Annu.

Rev. Stat. Its Appl. Annual Reviews; 2015;2(1):73–94

McDonald D, Price MN, Goodrich J, Nawrocki EP, DeSantis TZ, Probst A, et al. An improved

Greengenes taxonomy with explicit ranks for ecological and evolutionary analyses of bacteria

and archaea. ISME J. Nature Publishing Group; 2012;6(3):610–8

Meisel JS, Hannigan GD, Tyldsley AS, SanMiguel AJ, Hodkinson BP, Zheng Q, et al. Skin

Microbiome Surveys Are Strongly Influenced by Experimental Design. J. Invest. Dermatol.

Elsevier; 2016;136(5):947–56

Murphy EC, Frick I-M. Gram-positive anaerobic cocci – commensals and opportunistic

pathogens. FEMS Microbiol. Rev. Oxford University Press; 2013;37(4):520–53

Nakatsuji T, Chen TH, Narala S, Chun KA, Two AM, Yun T, et al. Antimicrobials from human

skin commensal bacteria protect against Staphylococcus aureus and are deficient in atopic

dermatitis. Sci. Transl. Med. 2017;9(378)

Nikolakis G, Liakou A, Bonovas S, Seltmann H, Bonitsis N, Join-Lambert O, et al. Bacterial

Colonization in Hidradenitis Suppurativa/Acne Inversa: A Cross-sectional Study of 50 Patients

and Review of the Literature. Acta Derm. Venereol. 2017;97(4):493–8

Oh J, Byrd AL, Deming C, Conlan S, Kong HH, Segre JA, et al. Biogeography and individuality

shape function in the human skin metagenome. Nature. Nature Publishing Group;

2014;514(7520):59–64

Parks DH, Tyson GW, Hugenholtz P, Beiko RG. STAMP: statistical analysis of taxonomic and

functional profiles. Bioinformatics. Oxford University Press; 2014;30(21):3123–4

Ring H, Bay L, Kallenbach K, Miller I, Prens E, Saunte D, et al. Normal Skin Microbiota is

Altered in Pre-clinical Hidradenitis Suppurativa. Acta Derm. Venereol. 2017a;97(2):208–13

Ring HC, Thorsen J, Saunte DM, Lilje B, Bay L, Riis PT, et al. The Follicular Skin Microbiome

in Patients With Hidradenitis Suppurativa and Healthy Controls. JAMA Dermatology. American

Medical Association; 2017b;153(9):897

Rom O, Avezov K, Aizenbud D, Reznick AZ. Cigarette smoking and inflammation revisited.

Respir. Physiol. Neurobiol. Elsevier; 2013;187(1):5–10

Sartorius K, Emtestam L, Jemec GBE, Lapins J. Objective scoring of hidradenitis suppurativa

reflecting the role of tobacco smoking and obesity. Br. J. Dermatol. Wiley/Blackwell (10.1111);

2009;161(4):831–9

Scholz CFP, Kilian M. The natural history of cutaneous propionibacteria, and reclassification of

selected species within the genus Propionibacterium to the proposed novel genera

Acidipropionibacterium gen. nov., Cutibacterium gen. nov. and Pseudopropionibacterium gen.

nov. Int. J. Syst. Evol. Microbiol. Microbiology Society; 2016;66(11):4422–32

Shu M, Wang Y, Yu J, Kuo S, Coda A, Jiang Y, et al. Fermentation of Propionibacterium acnes,

a Commensal Bacterium in the Human Skin Microbiome, as Skin Probiotics against Methicillin-

Resistant Staphylococcus aureus. Manganelli R, editor. PLoS One. Public Library of Science;

2013;8(2):e55380

Szabo G, Saha B. Alcohol’s Effect on Host Defense. Alcohol Res. National Institute on Alcohol

Abuse and Alcoholism; 2015;37(2):159–70

Tan TY, Ng SY, Thomas H, Chan BK. Arcanobacterium haemolyticum bacteraemia and soft-

tissue infections: Case report and review of the literature. J. Infect. W.B. Saunders;

2006;53(2):e69–74

Taylor D, Daulby A, Grimshaw S, James G, Mercer J, Vaziri S. Characterization of the

microflora of the human axilla. Int. J. Cosmet. Sci. John Wiley & Sons, Ltd (10.1111);

2003;25(3):137–45

Urban J, Fergus DJ, Savage AM, Ehlers M, Menninger HL, Dunn RR, et al. The effect of

habitual and experimental antiperspirant and deodorant product use on the armpit microbiome.

PeerJ. PeerJ Inc.; 2016;4:e1605

Zhao N, Chen J, Carroll IM, Ringel-Kulka T, Epstein MP, Zhou H, et al. Testing in Microbiome-

Profiling Studies with MiRKAT, the Microbiome Regression-Based Kernel Association Test.

Am. J. Hum. Genet. Elsevier; 2015;96(5):797–807

Zhao N, Zhan X, Guthrie KA, Mitchell CM, Larson J. Generalized Hotelling’s test for paired

compositional data with application to human microbiome studies. Genet. Epidemiol. Wiley-

Blackwell; 2018;42(5):459–69

FIGURE LEGENDS

Figure 1: Bacterial -diversity is significantly decreased on HS compared to normal skin. a) Schematic of experimental design b) Shannon diversity index (-diversity) for normal (n=34), HSN (n=36), and HSL (n=36) skin samples; Wilcoxon rank sum test, p < 0.05 considered significant. HSN = HS Non-lesional, HSL = HS Lesional. c) Bray-Curtis dissimilarity index for -diversity comparison, Wilcoxon rank sum test, p < 0.05 significant. The Bray-Curtis dissimilarity index is a scale from 0 to 1: 0 indicates identical communities and 1 indicates completely dissimilar communities. d) Top 20 significantly different genera between normal, HSL, and HSN skin. Mean relative abundance ± SD: * significantly different from normal. e) Relative bacterial abundance of the top 10 genera of each subject.

Figure 2: Body site heterogeneity is lost in HS compared to normal skin. a-d) Shannon diversity index (-diversity) and Bray-Curtis dissimilarity index (-diversity) comparison between axilla and groin stratified by normal and HSL skin and sampling method as indicated. The Wilcoxon rank sum test was applied to determine statistical differences in Shannon index between groups with p < 0.05 considered significant. a) Normal swab, n=10 from each body site, b) normal glue, n=5 from each body site, c) HS lesional swab, n=11 from each body site, and d) HS lesional glue, n=8 from each body site.

Figure 3: The follicle and surface microbiome signatures are different in normal skin, but not in HS skin. a-d) Shannon diversity index (-diversity) and Bray-Curtis dissimilarity index (-diversity) comparison between sampling methods (glue vs. swab) stratified by skin type and body site as indicated. The Wilcoxon rank sum test was applied to determine statistical differences in Shannon index between groups, with p < 0.05 considered significant. a) Normal axilla, n=7 collected by each method, b) normal groin, n=7 collected by each method, c) HS lesion axilla, n=9 collected by each method, and d) HS lesion groin, n=9 collected by each method.

Figure 4: Individual lifestyle factors are associated with altered bacterial diversity in HS skin. a) MiRKAT analyses were applied to determine -diversity differences (p-values) between

normal subjects or HS patients associated with each variable as listed in the table. b-d) Bray- Curtis dissimilarity index (-diversity) comparison in HS patients by Hurley stage, stage I (n=1, red), stage II (n=4, blue), stage III (n=6, grey) (b) smoking (n=5, red) vs. non-smoking (n=6, blue) (c) and between alcohol (n=6, red) vs. no alcohol (n=5, blue) (d).

Figure 5: KEGG Ontology indicates functional differences between microbiota of HS and normal skin. a) PCA plot of the overall functional differences of Level 3 KEGG pathways between normal (n=34), HSN (n=36), and HSL (n=36) skin. Multiple group statistical analysis was performed in STAMP (Kruskal-Wallis H-test with Bonferroni correction). b-c) Selected Level 2 (bold) and Level 3 (non-bold) KEGG Ortholog pathways enriched in HS (HSN + HSL) and normal skin. Val = valine, leu = leucine, ile = isoleucine. Multiple group statistical analysis performed in STAMP (Kruskal-Wallis H-test with Bonferroni correction). * indicates significance compared to normal (p-value < 0.05).

Figure 6: Metabolic pathways are influenced by different genera in normal and HS skin. Top 10 genera contributing to metabolic pathways with each column representing the mean bacterial OTU abundance from all samples collected from one subject. a) Propanoate Metabolism, b) Retinol Metabolism, c) Valine, Leucine, Isoleucine Biosynthesis, d) Valine, Leucine, and Isoleucine Degradation, e) Arginine and Proline Metabolism.

bioRxiv preprint doi: https://doi.org/10.1101/612457; this version posted April 18, 2019. The copyright holder for this preprint (which was not a certified by peer review) is the author/funder,b who has grantedα-Diversity bioRxiv a license toc display the preprint in perpetuity. It is made available underβ -Diversity Normal HS Patients aCC-BY-NC-ND 4.0 International license. (n = 10) (n = 11) Shannon Index Average β-Diversity BC Dissimilarity = 0.655 Skin Type 4 p = 0.225 *p = 0.010  Non-Lesion 0.6 p = 0.244 *p = 0.036  Lesion p = 0.987 1.0 p = 0.564  0.4     3 DNA  0.2   0.8  2 PC2 16S rRNA 0.0 gene sequencing (V3-V4) 0.6 1 -0.2

Swab Epidermis Curtis Dissimilarity Bray Average 0.4 -0.4 Glue Dermis Normal HSLHSN Normal HSLHSN 0.40.0-0.4 Sampling Depth & Method PC1 d * e 1.0 Gordonia * HSL Rothia * HSN

Neisseria * Normal 0.8 * Enhydrobacter *

Pseudomonas * 0.6

Acinetobacter * * Cutibacterium * 0.4 * WAL_1855D *

* Relative Abundance 1-68 * 0.2 * Dialister * * Porphyromonas * 0.0 Finegoldia * Normal HS * Anaerococcus * Staphylococcus Peptoniphilus Enhydrobacter * Peptoniphilus * Corynebacterium Finegoldia Agrobacterium 0 5 10 15 20 Cutibacterium Prevotella Other % Relative Abundance + SD Anaerococcus Porphyromonas bioRxiv preprint doi: https://doi.org/10.1101/612457; this version posted April 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. a Surface/Swab b Follicle/Glue α-Diversity β-Diversity α-Diversity β-Diversity Shannon Index BC Dissimilarity=0.523 Shannon Index BC Dissimilarity=0.620 4 0.4 4 0.6 p=0.248 *p=0.032 0.2 3 3 0.4 PC2

PC2 0.0 2 0.2 Normal 2 -0.2 0.0 1 -0.2 -0.4 1 Axilla Groin -0.4 0.0 0.4 Axilla Groin -0.4 0.40.0 PC1 PC1 c α-Diversity β-Diversity d α-Diversity β-Diversity Shannon Index BC Dissimilarity=0.544 Shannon Index BC Dissimilarity=0.632 0.4 4 0.4 4 p=0.797 p=0.195 0.2 0.2 3 3 0.0 PC2 PC2

HS Lesion 0.0 2 -0.2 2

-0.2 1 -0.4

1 Axilla Groin -0.4 0.40.0 Axilla Groin -0.4 0.40.0 PC1 PC1 bioRxiv preprint doi: https://doi.org/10.1101/612457; this version posted April 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. a Axilla b Groin α-Diversity β-Diversity α-Diversity β-Diversity Shannon Index BC Dissimilarity=0.595 Shannon Index BC Dissimilarity=0.593 4 0.6 4 0.4 *p=0.007 p=0.318

3 0.4 3 0.2 PC2 0.2 PC2 2 Normal 0.0

0.0 2 1 -0.2

-0.2 1 Follicle/ Surface/ -0.4 0.40.0 Follicle/ Surface/ -0.4 0.40.0 Glue Swab PC1 Glue Swab PC1 c α-Diversity β-Diversity d α-Diversity β-Diversity Shannon Index BC Dissimilarity=0.605 Shannon Index BC Dissimilarity=0.570 4 4 0.4 0.2 p=0.222 p=0.863 3 0.2 0.0 3 PC2 PC2 0.0 2 -0.2 HS Lesion 2 -0.4 -0.2 1

-0.6 1 Follicle/ Surface/ -0.4 0.40.0 Follicle/ Surface/ -0.4 0.40.0 Glue Swab PC1 Glue Swab PC1 bioRxiv preprint doi: https://doi.org/10.1101/612457; this version posted April 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

a Association test p-values between b Hurley Stage: β-Diversity microbiome composition and meta-variables HS Skin MiRKAT: p = 0.323 Variable HS Normal 0.3 Hurley Stage 0.323 NA 0.2 Age 0.058 0.932 Stage I Sex 0.042 0.052 0.1 Stage II

Race (white/non-white) 0.613 0.728 PC2 0.0 Alcohol use (yes/no) 0.078 0.934 Stage III Smoking (yes/no) 0.063 NA -0.1 Deodorant use (yes/no) 0.517 NA -0.2 Medications use (yes/no) 0.504 0.425 -0.2 0.0 0.2 0.4 PC1 c Smoking: β-Diversity d Alcohol: β-Diversity HS Skin MiRKAT: p = 0.063 HS Skin MiRKAT: p = 0.078

0.3 0.3

0.2 0.2 Smoking Alcohol 0.1 0.1 PC2 PC2 No Smoking 0.0 No Alcohol 0.0

-0.1 -0.1

-0.2 -0.2

-0.2 0.0 0.2 0.4 -0.2 0.0 0.2 0.4 PC1 PC1 a Normal HSN HSL 0.02

bioRxiv0.01 preprint doi: https://doi.org/10.1101/612457; this version posted April 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. 0.00

-0.01 PC2 (19.8%)

-0.02

-0.03

-0.03 -0.02 -0.01 0.00 0.01 0.02 0.03 PC1 (48.5%)

b Metabolism p-value Increased Amino Acid Metabolism Arginine and proline metabolism 0.00011 ↑ Normal Val, leu and ile biosynthesis 0.03707 ↑ Normal Val, leu and ile degradation 0.05750 ↑ Normal Metabolism of Cofactors and Vitamins Retinol metabolism 0.00301 ↑ Normal Thiamine metabolism 0.01193 ↑ HS Carbohydrate Metabolism Propanoate metabolism 0.06790 ↑ Normal Nucleotide Metabolism Purine metabolism 0.00002 ↑ HS Pyrimidine metabolism 0.00120 ↑ HS Energy Metabolism Methane metabolism 0.04660 ↑ HS Lipid Metabolism Glycerolipid metabolism 0.07962 ↑ Normal Lipid biosynthesis proteins 0.11170 ↑ Normal

c Genetic Information Processing p-value Increased Replication and Repair DNA replication 0.00002 ↑ HS DNA replication proteins 0.00005 ↑ HS Mismatch repair 0.00000 ↑ HS Homologous recombination 0.00022 ↑ HS Nucleotide excision repair 0.01022 ↑ HS Transcription RNA polymerase 0.00033 ↑ HS Transcription machinery 0.00330 ↑ HS Translation Aminoacyl-tRNA biosynthesis 0.00028 ↑ HS Ribosome biogenesis in eukaryotes 0.00051 ↑ HS Folding, Sorting and Degradation Protein export 0.00732 ↑ HS RNA degradation 0.01833 ↑ HS bioRxiv preprint doi: https://doi.org/10.1101/612457; this version posted April 18, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. a Propanoate Metabolism b Retinol Metabolism 125 16

100 12

75 8

4 25 OTU Abundance in Sample x1,000 Abundance OTU in Sample x1,000 Abundance OTU

0 0 Normal HS Normal HS c Valine, Leucine, Isoleucine Biosynthesis d Valine, Leucine, Isoleucine Degradation 600 450

400 300

200 150 OTU Abundance in Sample x1,000 Abundance OTU OTU Abundance in Sample x1,000 Abundance OTU

0 0 Normal HS Normal HS e Arginine and Proline Metabolism 1-68 Gordonia 300 Agrobacterium Klebsiella

Anaerococcus Kocuria

200 Campylobacter Paracoccus

Corynebacterium Peptoniphilus

Cutibacterium Porphyromonas 100 Dermacoccus Prevotella

OTU Abundance in Sample x1,000 Abundance OTU Enhydrobacter Pseudomonas

0 Finegoldia Normal HS