Anal microbiota profi

les in HIV-positive andHIV-negative MSM

Guoqin Yua, Doug Fadroshb, Bing Mab, Jacques Ravelb and

James J. Goederta

Copyright © L

aInfections and ImBethesda, and bIn

Correspondence t

Tel: +1 240 276 7Received: 12 Sept

DOI:10.1097/QAD

ISS

Objective: Because differences in anal microbial populations (microbiota) could affectacquisition of HIV or other conditions, especially among MSM, we profiled themicrobiota of the anal canal, assessed its stability, and investigated associations withdiversity and composition.

Design: Microbiota profiles in anal swabs collected from 76 MSM (52 in 1989, swab-1;66 1–5 years later, swab-2) were compared by HIV status (25 HIV-positive), T-cellsubsets, and questionnaire data.

Methods: Bacterial 16S rRNA genes were amplified, sequenced (Illumina MiSeq), andclustered into species-level operational taxonomic units (QIIME and Greengenes).Regression models and Wilcoxon tests were used for associations with alpha diversity(unique operational taxonomic units, Shannon’s index). Composition was compared byAdonis (QIIME).

Results: Most anal bacteria were Firmicutes (mean 60.6%, range 21.1–91.1%) orBacteroidetes (29.4%, 4.1–70.8%). Alpha diversity did not change between the twoswabs (N¼42 pairs). In swab-2, HIV-positives had lower alpha diversity (P�0.04) andaltered composition, with fewer Firmicutes and more Fusobacteria taxa (P�0.03), notcompletely attributable to very low CD4þ cell count (median 232 cells/ml), prior AIDSclinical diagnosis (N¼17), or trimethoprim–sulfamethoxazole use (N¼6). Similar butweaker differences were observed in swab-1 (HIV-positive median 580 CD4þ cells/ml;no trimethoprim–sulfamethoxazole). Associations with T-cell subsets, smoking, andsexual practices were null or inconsistent.

Conclusions: The anal microbiota of MSM was relatively stable over 1–5 years.However, with uncontrolled, advanced HIV infection, the microbiota had alteredcomposition and reduced diversity partially attributable to antibiotics. Investigationsof microbial community associations with other immune perturbations and clinicalabnormalities are needed. � 2014 Wolters Kluwer Health | Lippincott Williams & Wilkins

AIDS 2014, 28:753–760

Keywords: anal, HIV, human microbiome, MSM, risk factors, sexual activities

Introduction

The anus is an important body site both for acquisitionof HIV infection, especially among MSM, and for somemanifestations of AIDS such as chronic, ulcerativemucocutaneous herpes simplex [1]. In other bodysites, including those in proximity to the anus such asthe distal gut, skin, and vagina, the resident microbes

ippincott Williams & Wilkins. Unaut

munoepidemiology Branch, Division of Cancerstitute of Genomic Sciences, University of Mary

o James J. Goedert, 9609 Medical Center Drive,

103; fax: +1 240 276 7806; e-mail: goedertj@member 2012; revised: 13 November 2013; acce

.0000000000000154

N 0269-9370 Q 2014 Wolters Kluwer He

can have mutualistic, commensal or pathogenicinteractions with each other and with the host,influencing myriad functions including local andsystemic immunity and ultimately the balance betweenhealth and disease [2–5]. The anal canal’s microbialpopulations (the anal microbiota) have not beenextensively described or studied using moleculartechniques.

horized reproduction of this article is prohibited.

Epidemiology and Genetics, National Cancer Institute,land School of Medicine, Baltimore, Maryland, USA.

Room 6E106 MSC 9767, Bethesda, MD 20892-9704, USA.

ail.nih.govpted: 13 November 2013.

alth | Lippincott Williams & Wilkins 753

Co

754 AIDS 2014, Vol 28 No 5

The current cross-sectional and cohort study of MSMsought to profile the microbiota of the anal canal, to assessits stability over 1–5 years, and to investigate whetherdifferences in the diversity and composition of themicrobiota were associated with HIV infection, CD4þ

and CD8þ lymphocyte subsets in blood, antibiotic use,sexual activities, or smoking. These were considered inthe context of limited studies of HIV and microbiota atother body sites [6,7].

Materials and methods

Cohort background and covariate dataThe Washington, DC and New York City [informally,DC Gay (DCG)] study cohort enrolled and followed acohort of 245 consecutive outpatient MSM of threeprimary care physicians from 1982 through 1999. Duringthose years, all patients provided signed informedconsent, and the study was reviewed and approvedannually by the Special Studies Institutional ReviewBoard (IRB) of the National Cancer Institute. Exemptionfrom IRB review for subsequent analyses of thespecimens and data was granted by the NationalInstitutes of Health Office of Human Subjects ResearchProtection.

Data on clinical conditions, antibiotics [specifically,trimethoprim–sulfamethoxazole (TMP-SMX) and clar-ithromycin], sexual, smoking, and other activities duringthe year before sampling were obtained by self-administered questionnaire. HIV serostatus and viralload, as well as peripheral blood CD4þ and CD8þ cellcounts, were determined by standard methods [1,8–10].No patient seroconverted to HIV-positive betweenspecimens used in the current study.

Specimen collectionAs shown in Supplemental Figure 1 (http://links.lww.-com/QAD/A455), baseline for the current analysis wasthe visit in 1989, when clinicians collected swabs from theanal canal for evaluation of human papillomavirus (HPV)and anal cytology [11]. Following inspection of theexternal anus for lesions, a Dacron swab was dipped insterile saline, gently inserted approximately 4 cm into theanal canal, rotated two times, then withdrawn andsmeared on a glass slide that was sprayed for subsequentcytologic analysis [11]. A second swab was collected withthe same procedure, but was immediately immersed intransport media (Virapap; Life Technologies, Gaithers-burg, Maryland, USA), refrigerated for 2–18 h, and thenstored frozen at or below �208C until testing. The sameprocedures were used in subsequent years. HPV testingresulted in depletion of the specimens with cytologicatypia [11]. Thus, the current study included all menwithout cytologic atypia who had a pair of specimens inthe repository, one from 1989 (swab-1) and a second from

pyright © Lippincott Williams & Wilkins. Unautho

approximately 1–5 years later (swab-2). The mean (andmedian) interval between swab-1 and swab-2 was 3 years(range 0.5–5.6 years) and was longer in the HIV-positivethan the HIV-negative men (mean 3.7 versus 2.5 years).

DNA extraction and sequencingPreliminary experiments with four selected samplesdemonstrated good DNA yield and 16S rRNA ampli-fication. Excluding seven vials (of 162) that were empty,DNA was extracted from the 155 remaining samples withthe following minor modifications [12]. A 500 ml aliquotof each sample was mixed with 350 ml of lysis buffercomposed of 0.05 mol/l potassium phosphate buffercontaining 50 ml lyzosyme (10 mg/ml), 6 ml of mutano-lysin (25 000 U/ml; Sigma–Aldrich, St Louis, Missouri,USA) and 3 ml of lysostaphin (4000 U/ml in sodiumacetate; Sigma–Aldrich). The mixture was incubated for1 h at 378C, following which 10 ml proteinase K (20 mg/ml), 100 ml 10% SDS, and 20 ml RNase A (20 mg/ml)were added. This mixture was incubated for 1 h at 558C.To further lyze microbial cells, Lying Matrix B 2-ml beads(MP Biomedicals, Santa Ana, California, USA) wereadded, following which mechanical disruption (beadbeating) was performed on the mixture using a FastPrepinstrument (MP Biomedicals, Solon, Ohio, USA) set at6.0 m/s for 30 s. The lysate was processed using theQIAsymphony SP protocol Pathogen complex 400(Qiagen, Gaithersburg, Maryland, USA) according tothe manufacturer’s recommendation. The DNA waseluted into 100 ml of storage buffer [QIAsymphonyreagent buffer AVE (0.04% sodium azide), Qiagen], pH8.0. PCR inhibitors were removed from the extractedDNA using the Zymo-Spin intravenous spin filtercolumn according to the manufacturer’s recommen-dations (Irvine, California, USA). DNAwas quantified byQuant-iT PicoGreen (Molecular Probes, Inc., Eugene,Oregon, USA) in a SpectraMax M5 microplate reader(Molecular Devices, Sunnyvale, California, USA).

A region of approximately 469 bp encompassing the V3and V4 hypervariable regions of the 16S rRNA gene wastargeted for sequencing. This region provides ampleinformation for taxonomic classification of microbialcommunities from specimens associated with humanmicrobiome studies and was used by the HumanMicrobiome Project [2]. Fusion dual barcoded primers319F (5’-ACTCCTACGGGAGGCAGCAG-3’) and806R (5’-GGACTACHVGGGTWTCTAAT-3’) wereused to amplify the V3-V4 region of bacterial 16S rRNAgenes. The amplicons were pooled in equimolarconcentration and sequenced on an Illumina MiSeqInstrument using the 250-bp paired-end protocol. Thesequenced reads were processed using the following steps:removal of primer sequence, truncation of reads nothaving an average quality of 20 over a 30-bp slidingwindow based on the phred algorithm [13,14] imple-mented previously [15,16], removal of trimmed readshaving less than 75% of their original length, and removal

rized reproduction of this article is prohibited.

Anal microbiota profiles in HIV-positive and HIV-negative MSM Yu et al. 755

of the mate of reads that were discarded for having lessthan 75% original length. QIIME (version 1.6.0) [17] wasused for all further sequence processing steps, includingquality trimming and demultiplexing. Quality trimmingin QIIME was performed using the following criteria:truncate sequence before three consecutive low-qualitybases and re-evaluate for length, no ambiguous base calls,minimum sequence length of 150 bp after trimming, andremove sequences with less than 60% identity to a prebuiltGreengenes database of 16S rRNA gene sequences(October 2012 version) [18]. Further data processingincluded denoizing by clustering similar sequences withless than 3% dissimilarity using USEARCH [19] and de-novo chimera detection and removal in UCHIME v5.1[20]. Paired reads were stitched together with ‘N’between each sequence and processed as one sequencein the analysis. Of 133 samples successfully sequenced,analysis was restricted to the 118 (89%) that had at least5000 sequences. This included 42 men at both swab-1and swab-2, 10 men at only swab-1, and 24 men at onlyswab-2. Success or failure did not differ by HIV status(Supplemental Figure 1, http://links.lww.com/QAD/A455).

Analysis of 16S rRNA sequence dataQIIME was used to cluster the sequence reads at 97%nucleotide sequence identity. A closed-reference oper-ational taxonomic unit (OTU) picking protocol withUSEARCH against the Greengenes database wasemployed. Reads that did not match the Greengenesdatabase, as well as those representing only a single OTU,were excluded. Of the 1 869 502 amplicon sequencesprocessed, 94.4% (1 765 159) hit a reference sequence atgreater than or equal to 97% sequence identity. In order tocompare microbiota diversity within individuals at thesame sequence depth, a random sample of 5000 sequenceswas drawn without replacement from each sample20 times.

Four estimates of alpha diversity – number of uniqueOTUs, Shannon’s index, Chao1 and phylogeneticdistance_whole tree – were calculated. The total numberof unique OTUs does not take relative abundance ofOTUs into account. Chao1 is bias-corrected forsingleton OTUs [21]. Shannon’s index is defined as(negative) the sum over OTUs of the product of therelative abundance of the OTU times the naturallogarithm of the relative abundance [22]. A measure ofalpha diversity that reflects phylogenetic divergenceamong OTUs within an individual is called phylogeneticdistance_whole tree.

Taxonomy was assigned using RDP classifier 2.2.Summarize_taxa.py script in QIIME was used tocalculate the relative abundance of phylum, class, order,and genus-level taxa. Linear discriminant analysis effectsize (LEfSe) was used to identify taxa for which therelative abundance was different between the HIV-

Copyright © Lippincott Williams & Wilkins. Unaut

positive and HIV-negative groups [23]. Weighted andunweighted UniFrac distances between samples, whichmeasure the pair-wise phylogenetic distance differencebetween microbial communities, were calculated inQIIME by using the existing tree from the Greengenesdatabase. Weighted UniFrac distance accounts for therelative abundance of each taxon in the communities,whereas unweighted UniFrac distance does not.

Statistical analysesAssociations of HIV, unprotected sexual practices,smoking, current or recent use of TMP-SMX, CD4þ

and CD8þ cell counts, and other variables with alphadiversity estimates were tested by Wilcoxon rank-sumtests or fitting logistic regression models for categoricalvariables and fitting linear regression models forcontinuous variables. Wilcoxon signed-rank tests alsowere used to compare differences between swab-1 andswab-2 in alpha diversity, smoking, unprotected sexualpractices, and CD4þ and CD8þ cell counts. Associationswith beta diversity were tested by the Adonis methodin QIIME.

Results

Characteristics of the study participantsHigh-quality anal microbiome data were obtained for 76participants, which included 42 with paired data fromboth swab-1 and swab-2, 10 with data only from swab-1,and 24 with data only from swab-2. In 1989, at the time ofswab-1, the participants had a mean age of 41 years (range31–65 years) and included 46 White non-Hispanic menand six men of other race/ethnicity. Twenty men hadbeen HIV-positive for a median of 6 years, of whom onlyone had been diagnosed with AIDS (Kaposi sarcoma).The other 32 men were persistently HIV-negative at alltime points.

Table 1 presents CD4þ and CD8þ cell counts, sexualactivities, smoking intensity, and anal microbial diversityat swab-1 and swab-2 for all participants. At swab-1 in1989, the HIV-positive men had lower CD4þ cell counts[median and interquartile range (IQR) 580 (432–721)cells/ml] and modestly higher CD8þ cell counts [median(IQR) 635 (485–1089) cells/ml] compared to the HIV-negative men (median CD4þ 962 cells/ml, median CD8þ

546 cells/ml). By the time of swab-2, 1–5 years later,CD4þ cell counts in the HIV-positive men had fallensignificantly [median (IQR) 232 (32–415) cells/ml], butthere were no significant changes in their CD8þ cellcounts or in CD4þ or CD8þ cell counts in the HIV-negative men. By swab-2, 16 more HIV-positive men haddeveloped AIDS, including 11 more with Kaposisarcoma, seven with resolved Pneumocystis jiroveci pneu-monia (PJP), and eight with other AIDS-defining clinical

horized reproduction of this article is prohibited.

Co

756 AIDS 2014, Vol 28 No 5

Table 1. T-cell subsets, sexual activities and smoking in the cohort when the anal swabs were collected.

Swab-1 (n¼52) [median(IQR)] Swab-2 (n¼66) [median(IQR)] P valuea

T cells/mlCD4þ cell count

HIV-positive 580 (432–721) 232 (32–415) 0.0002HIV-negative 962 (723–1138) 914 (605–1162) 0.57

CD8þ cell countHIV-positive 635 (485–1089) 517 (314–973) 0.34HIV-negative 546 (426–764) 593 (436–800) 0.85

Unprotected sexual activityb

No. of male partners 5 (2–10) 2.5 (1–7.8) 0.07No. of new male partners 1 (0–2.25) 1 (0–2) 0.85Fellatio, receptive 0 (0–0) 0 (0–1) 0.15Fellatio, insertive 0 (0–0) 0 (0–1) 0.33Anal intercourse, receptive 0 (0–0) 0 (0–0) 0.07Anal intercourse, insertive 0 (0–0) 0 (0–0) 0.92Anilingus, receptive 0 (0–1) 0 (0–1) 1.00Anilingus, insertive 0 (0–1) 0 (0–0) 0.25Fisting, receptive 0 (0–0) 0 (0–0) 0.89Fisting, insertive 0 (0–0) 0 (0–0) 0.45

SmokingCigarettes per day 0 (0–8.5) 0 (0–3.8) 0.84

Anal microbial alpha diversityNo of unique OTUs

HIV-positive 196 (141–287) 199 (141–263) 0.99HIV-negative 233 (177–338) 313 (209–373) 0.16

Shannon’s indexHIV-positive 5.8 (5.4–6.4) 5.8 (5.3–6.2) 0.88HIV-negative 6.1 (5.6–6.5) 6.3 (5.8–6.7) 0.21

IQR, interquartile range; OTU, operational taxonomic unit.aWilcoxon signed-rank test for difference between swab-1 and swab-2.bAs reported previously,[8] specific activities within the previous 12 months were reported in ordinal categories (0¼none, 1¼1–4 times, 2¼5–9times, . . . 8¼more than 200 times). Even with IQR (0–0), at least three participants reported each activity.

conditions. Twelve HIV-positive and two HIV-negativemen died during 1991–1997.

At swab-1 in 1989, the participants reported a median of 5(IQR 2–10; mean 13) male sex partners per year, whichdecreased to 2.5 (IQR 1–7.8; mean 8) per year by swab-2. Frequencies of sexual activities and smoking did notchange significantly in the interval between swabs, butonly small minorities of the participants reported each ofthe unprotected sexual activities. Six participants, allHIV-positive, reported taking TMP-SMX at or within3 months of swab-2; none had taken clarithromycinduring this interval. No participant reported taking eitherdrug within 3 months of swab-1, but these medicationswere not explicitly included in the swab-1 (1989) ques-tionnaire.

Anal microbiota by HIV status and antibiotic useSupplementary Figure 2 (http://links.lww.com/QAD/A455) presents the relative abundance of genus-levelbacterial taxa that were detected in the anal swabs. Twoanal microbiota clusters were identified, based primarilyon differences in relative abundance of taxa in the generaPrevotella and Bacteroides, both in the phylum Bacter-oidetes. These clusters did not differ by HIV status orswab (Supplementary Figure 2, http://links.lww.com/QAD/A455).

pyright © Lippincott Williams & Wilkins. Unautho

Anal microbial alpha diversity did not change significantlyduring the interval between swabs in either the HIV-positive or HIV-negative men (Table 1). This wasconfirmed by analysis of 42 men with results from pairedswabs (P� 0.17; Supplementary Tables 1 and 2, http://links.lww.com/QAD/A455). In the 42 paired results,from swab-1 to swab-2, the mean number of uniqueOTUs increased from 262 to 294 (paired t-test P¼ 0.27)and mean Shannon’s index increased from 6.0 to 6.1(paired t-test, P¼ 0.52). In contrast, all four measures ofalpha diversity indicated lower microbial diversity in analswabs from HIV-positive men compared with HIV-negative men, and the differences were statisticallysignificant for swab-2, but not for swab-1 (Table 2,Fig. 1). Adjustment by linear regression for age, race/ethnicity, and smoking attenuated the statistical signifi-cance.

In the six participants with current or recent use of TMP-SMX compared to the other HIV-positive participants,three estimates of alpha diversity were 10–17% lower atswab-2, but these differences were not statisticallysignificant (Supplementary Table 3, http://links.lww.-com/QAD/A455). The corresponding phylogeneticdistance_whole tree estimate was only 1% lower. Withadjustment for TMP-SMX use, as well as age, race/ethnicity, and smoking, the phylogenetic distance_wholetree estimate of alpha diversity, but not the other

rized reproduction of this article is prohibited.

Anal microbiota profiles in HIV-positive and HIV-negative MSM Yu et al. 757

Table 2. Association between HIV status and anal microbial alpha diversity.

Swab-1 Swab-2

Mean P value Mean P value

Alpha diversityestimate

HIVþ(n¼20)

HIV�(n¼32) Wilcoxon Adjusteda

HIVþ(n¼25)

HIV�(n¼41) Wilcoxon Adjusteda

No. of unique OTUs 236 276 0.22 0.30 246 306 0.03 0.17Shannon’s index 5.8 6.1 0.33 0.21 5.7 6.2 0.02 0.15Chao1 292 331 0.35 0.44 312 376 0.04 0.34PD_whole tree 5.7 6.2 0.29 0.44 5.4 6.8 0.001 0.02

OTUs, operational taxonomic units; PD, phylogenetic distance.aBy linear model (alpha diversity¼HIV status and age, race/ethnicity, number of cigarettes per day, and use of trimethoprim–sulfamethoxazolewithin 3 months).

estimates, remained statistically significantly lower inHIV-positive compared to HIV-negative individuals(P¼ 0.02; Table 2).

Considering beta diversity, which compares microbialcommunity composition between all pairs of individuals,unweighted UniFrac distance at swab-2 was significantlyassociated with both HIV status and TMP-SMX use(P¼ 0.03 for each; Supplementary Table 3, http://links.lww.com/QAD/A455). Weighted UniFrac distance

Copyright © Lippincott Williams & Wilkins. Unaut

650

HIV No. of OTUs

No.

of O

TU

s

(a) (b

HIV–

HIV+

*

*

450

250

50

swab 1 swab 2

(d(c)

HIV FirmicutesHIV–

HIV+

Rel

ativ

e ab

unda

nce 1.2

1.0

0.8

0.6

0.4

0.2swab 1 swab 2

Fig. 1. Box plots comparing HIV-positive (black boxes) to HIV-nmicrobiota alpha diversity (a, b) and on relative abundance ofsignificantly lower alpha diversity [number of unique operationalabundance of Firmicutes at swab-2 (�). At both swabs, HIV-positivedifference at swab-2 persisted with exclusion of men with antibio(IQR), with central bars for medians and whiskers for 1.5 times th

(which adjusts for relative abundance of each taxon) wasmarginally associated with TMP-SMX use (P¼ 0.06),but not with HIV status.

There were large differences among individuals in thedistributions of taxa across the major phyla. Overall, thesetaxa were predominantly in the phyla Firmicutes (mean60.6%, range 21.1–91.1%) and Bacteroidetes (29.4%,4.1–70.8%), with most of the remainder in the phylaProteobacteria (3.9%, 0.04–45.6%), Actinobacteria

horized reproduction of this article is prohibited.

)

HIV Shannon’s indexHIV–

HIV+ *

*

***

Sha

nnon

’s in

dex

10

9

8

7

6

5

4

3swab 1 swab 2

)

HIV

Fusobacteria

HIV–

HIV+

Rel

ativ

e ab

unda

nce

0.05

0.04

0.03

0.02

0.01

0.00swab 1 swab 2

egative participants (gray boxes) on two estimates of analtaxa in two bacterial phyla (c, d). HIV-positive men hadtaxonomic units (OTUs) and Shannon’s index] and relativemen had higher relative abundance of Fusobacteria, and this

tic exposure (���). Boxes encompass the interquartile rangee IQR. Outlier values are filled circles.

Co

758 AIDS 2014, Vol 28 No 5

Table 3. Differences in relative abundance of taxa at the phylum level by HIV status.

Phyla

Swab-1, means Swab-2, means

HIVþ (n¼20) HIV� (n¼32) P valuea HIVþ (n¼25) HIV� (n¼41) P valuea P valueb

Firmicutes 59.9% 65.5% 0.10 53.4% 61.5% 0.03 0.27Bacteroidetes 28.8% 26.8% 0.39 33.2% 29.6% 0.42 0.82Proteobacteria 4.6% 2.3% 0.006 4.4% 4.4% 0.73 0.93Actinobacteria 5.0% 3.7% 0.98 3.3% 3.2% 0.71 0.73Fusobacteria 1.1% 0.7% 0.02 5.1% 0.4% 0.0005 0.007

Phyla with mean relative abundance below 1% are not presented.aBy Wilcoxon rank-sum test.bBy Wilcoxon rank-sum test, excluding patients with recent antibiotic use.

(3.6%, 0–24.7%), and Fusobacteria (1.6%, 0–35.3%).Table 3 presents mean relative abundance of these taxa byHIV status and swab. Compared to HIV-negative men,HIV-positive men had higher abundance of Proteobac-teria in swab-1, lower abundance of Firmicutes in swab-2,and higher abundance of Fusobacteria in both swabs(Table 3, Fig. 1). The higher abundance of Fusobacteria atswab-2 in HIV-positive men persisted (P¼ 0.007) whenparticipants with TMP-SMX use were excluded. ByLEfSe analysis [23], HIV-positive men had higherabundance of Fusobacteria and Streptococcus taxa and lowerabundance of Clostridium, Ruminococcus, and Odoribacter-aceae taxa.

Anal microbiota by T-cell subsets, sexualactivities, and smokingCD4þ cell counts at the time of swab-2 tended to bepositively associated with Shannon’s index and phyloge-netic distance_whole tree estimates of alpha diversity(P¼ 0.06 and P¼ 0.05, respectively), but not afteradjustment for TMP-SMX use (Table 4). Alpha diversitywas neither associated with CD4þ cell count at swab-1nor with CD8þ cell count at either swab. Thirteen of the25 HIV-positive men had plasma HIV load performed atswab-2 [median 60 780 (range 671–172 735) copies/ml].

pyright © Lippincott Williams & Wilkins. Unautho

Table 4. Association between T-cell subset levels and anal microbial alph

Alpha diversity T cells/ml

Swab-1 (n¼5

Coefficient

Number of unique OTUsCD4þ cell count �0.05927CD8þ cell count �0.08085

Shannon’s indexCD4þ cell count �0.00007CD8þ cell count �0.00027

Chao1CD4þ cell count �0.09326CD8þ cell count �0.10547

PD_whole treeCD4þ cell count �0.00033CD8þ cell count 0.00002

OTUs, operational taxonomic units; PD_whole tree, phylogenetic diversitaUnivariate linear regression models.bLinear regression models adjusted for age, race, recent antibiotic usage a

Plasma HIV load was not associated with alpha diversity(P� 0.47; Supplementary Table 4, http://links.lww.com/QAD/A455). CD8þ cell count was associated withboth unweighted and weighted UniFrac distance atswab-1 (P¼ 0.02 for each; Supplementary Table 5,http://links.lww.com/QAD/A455). These estimates ofbeta diversity were neither associated with CD8þ cellcount at swab-2 nor with CD4þ cell count at either swab.Neither alpha nor beta diversity estimates at swab-2differed between the men previously diagnosed with anAIDS-defining clinical condition and the other HIV-positive men (data not presented).

Of the several sexual activities listed in Table 1, alphadiversity was associated only with unprotected receptiveanal intercourse at the time of swab-1 (P¼ 0.03, data notshown). Only six patients reported this activity during theyear before swab-1. None of the sexual activities wereassociated with UniFrac distances at either swab(Supplementary Table 5, http://links.lww.com/QAD/A455). Weighted (but not unweighted) UniFrac distancewas marginally associated with average number ofcigarettes smoked per day (swab-1 P¼ 0.03, swab-2P¼ 0.06). Alpha diversity was not associated smokingwith at either swab (data not shown).

rized reproduction of this article is prohibited.

a diversity.

2) Swab-2 (n¼66)

P valuea Coefficient P valuea P valueb

0.11 0.0368 0.30 0.500.10 0.0627 0.09 0.12

0.77 0.0004 0.06 0.250.37 0.0003 0.25 0.32

0.07 0.0333 0.52 0.730.13 0.0942 0.08 0.10

0.57 0.0011 0.05 0.101.00 0.0006 0.33 0.30

y whole tree.

nd number of cigarettes per day.

Anal microbiota profiles in HIV-positive and HIV-negative MSM Yu et al. 759

Discussion

The study found that men with uncontrolled HIVinfection and severe immune deficiency had reducedalpha diversity and altered composition of the analmicrobiota in swab-2. Less advanced HIV-relatedimmune deficiency, at swab-1, was associated withsimilar, albeit non-significant, differences in the analmicrobiota. No warts, ulcer, fissure, abscess, or other analabnormality was observed. It must be, however, notedthat we only included men who had normal analcytology. Microbiota associations with T-cell subsetlevels, sexual practices, and smoking were marginal orinconsistent. Thus, confounding of HIV status bymedications must be considered carefully. Zidovudineand other first-generation nucleotide reverse transcriptaseinhibitors (NRTIs) were used by the majority ofimmune-deficient men [combination antiretroviraltherapy (ART) had not yet been developed], but thereis no evidence that NRTIs would affect the microbiota. Incontrast, use of TMP-SMX for PJP prophylaxis couldhave had a profound effect. As expected, we observedlower alpha diversity and altered composition withcurrent or recent TMP-SMX use, but associations of HIVstatus with the anal microbiota persisted followingexclusion of or adjustment for the six TMP-SMX users.

The composition of the anal microbiota was dominatedby taxa in two phyla, Firmicutes and Bacteriodetes, withlower abundances of Proteobacteria, Actinobacteria, andFusobacteria. Two clusters were identified in the analmicrobiota, which differed primarily in relative abun-dance of taxa in the genera Prevotella and Bacteroides, butnot by HIV status or swab. Thus, at the phylum andcluster level, the anal microbiota resembled the micro-biota of feces and of the coronal sulcus of the penis[2,12,24–26], although differences might be found withdeeper taxonomic comparisons. Highly prevalent skinbacteria, such as Staphylococcus epidermidis [4], were notfound in the anal microbiota. Likewise, Lacobacillus spp.were rarely detected (average relative abundance 0.2%),despite common ontogeny and similar squamo-columnarepithelial junctions in the cervix and anus [27].

Comparisons by HIV status of the microbiota at mostbody sites are lacking. In one study, HIV had no effect onvaginal microbial diversity or composition withoutbacterial vaginosis, and only slight increases in diversityand alterations in composition with bacterial vaginosis[6]. Among HIV-infected women, bacterial vaginosisappears to be independent of the level of immunedeficiency [28,29]. The microbiota of saliva (determinedby 16S rDNA fingerprinting rather than sequencing)were described in a review study by Saxena et al. [7] ashaving altered composition and higher alpha diversity in31 HIV-positive compared to 40 HIV-negative individ-uals. The same study indicated that six HIV-positivecompared to six HIV-negative individuals had slight

Copyright © Lippincott Williams & Wilkins. Unaut

differences in composition in the esophageal, gastric,duodenal, and colonic mucosal microbiota [7]. In ourunweighted UniFrac analysis, HIV-positive patients hadsignificantly altered composition in anal swab-2, withreduced abundance of Firmicutes and increased abun-dance of Fusobacteria compared with HIV-negativeindividuals. Validations of these observations, as well asequivalent comparisons at other body sites, are needed.

In our study, anal microbiota profiling was successful in 25HIV-positive and 51 HIV-negative MSM. Our smallsample size, especially with adequate sequences in swab-1, limited our ability to assess changes over time andassociations with earlier stage HIV. Because specimenswith cytologic atypia had been depleted and could not beevaluated, the population evaluated is biased towardMSM who had normal anal cytology. We also lackedimportant covariate data, including diet, BMI, and arange of anal cytologies. Our data on antibiotics may havebeen incomplete except for TMP-SMX at swab-2. Thestrengths of the study included lack of bias by HIV statusand careful, longitudinal assessment of sexual activities,smoking, and T-cell subsets linked to each of the twoswabs that were tested.

In summary, we found that the anal microbiota of MSMresembled fecal and penile microbiota at the phylum level[2,12,24–26], and that MSM with uncontrolled,advanced HIV infection had reduced alpha diversityand altered composition of the anal microbiota. Futureresearch can build on these findings to identify antibiotic-independent effects of HIV and other severe immunedeficiencies, as well as microbial associations with anal orsystemic clinical abnormalities.

Acknowledgements

We are grateful to our clinical collaborators (Drs RichardDiGioia, Ronald Grossman, and William Sanchez) andepidemiologic and statistical collaborators (Drs RobertBiggar, Jianxin Shi, and Mitchell Gail).

G.Y. analyzed and interpreted the data and wrote a draftmanuscript. D.F. conducted the pilot study and laboratoryprocedures. B.M. processed the raw 16S rRNA sequencedata. J.R. supervised the laboratory procedures, gener-ation of the data and revised the manuscript. J.J.G.designed and supervised the overall study, interpreted thedata, and revised the manuscript. All authors contributedto conception and design of the study, and critical revisionof the manuscript, and saw and approved the final version.

Supported in part by the Intramural Research Program ofthe National Cancer Institute.

Conflicts of interestThere are no conflicts of interest.

horized reproduction of this article is prohibited.

Co

760 AIDS 2014, Vol 28 No 5

References

1. Goedert JJ, Sarngadharan MG, Biggar RJ, Weiss SH, Winn DM,Grossman RJ, et al. Determinants of retrovirus (HTLV-III) anti-body and immunodeficiency conditions in homosexual men.Lancet 1984; 2:711–716.

2. Human Microbiome Project Consortium. Structure, functionand diversity of the healthy human microbiome. Nature 2012;486:207–214.

3. Arumugam M, Raes J, Pelletier E, Le Paslier D, Yamada T,Mende DR, et al. Enterotypes of the human gut microbiome.Nature 2011; 473:174–180.

4. Grice EA, Kong HH, Conlan S, Deming CB, Davis J, Young AC,et al. Topographical and temporal diversity of the human skinmicrobiome. Science 2009; 324:1190–1192.

5. Gajer P, Brotman RM, Bai G, Sakamoto J, Schutte UM, Zhong X,et al. Temporal dynamics of the human vaginal microbiota. SciTransl Med 2012; 4:132ra152.

6. Spear GT, Sikaroodi M, Zariffard MR, Landay AL, French AL,Gillevet PM. Comparison of the diversity of the vaginal micro-biota in HIV-infected and HIV-uninfected women with orwithout bacterial vaginosis. J Infect Dis 2008; 198:1131–1140.

7. Saxena D, Li Y, Yang L, Pei Z, Poles M, Abrams WR, et al. Humanmicrobiome and HIV/AIDS. Curr HIV/AIDS Rep 2012; 9:44–51.

8. Goedert JJ, Biggar RJ, Winn DM, Mann DL, Byar DP, Strong DM,et al. Decreased helper T lymphocytes in homosexual men. II.Sexual practices. Am J Epidemiol 1985; 121:637–644.

9. Goedert JJ, Biggar RJ, Weiss SH, Eyster ME, Melbye M, Wilson S,et al. Three-year incidence of AIDS in five cohorts of HTLV-III-infected risk group members. Science 1986; 231:992–995.

10. Goedert JJ, Biggar RJ, Melbye M, Mann DL, Wilson S, Gail MH,et al. Effect of T4 count and cofactors on the incidence of AIDSin homosexual men infected with human immunodeficiencyvirus. J Am Med Assoc 1987; 257:331–334.

11. Caussy D, Goedert JJ, Palefsky J, Gonzales J, Rabkin CS, DiGioiaRA, et al. Interaction of human immunodeficiency and papil-loma viruses: association with anal epithelial abnormality inhomosexual men. Int J Cancer 1990; 46:214–219.

12. Flores R, Shi J, Gail MH, Gajer P, Ravel J, Goedert JJ. Assessmentof the human faecal microbiota: II. Reproducibility and asso-ciations of 16S rRNA pyrosequences. Eur J Clin Invest 2012;42:855–863.

13. Ewing B, Hillier L, Wendl MC, Green P. Base-calling of auto-mated sequencer traces using phred. I. Accuracy assessment.Genome Res 1998; 8:175–185.

14. Ewing B, Green P. Base-calling of automated sequencer tracesusing phred. II. Error probabilities. Genome Res 1998; 8:186–194.

pyright © Lippincott Williams & Wilkins. Unautho

15. Li H, Durbin R. Fast and accurate short read alignment withBurrows-Wheeler transform. Bioinformatics 2009; 25:1754–1760.

16. Li H, Durbin R. Fast and accurate long-read alignment withBurrows-Wheeler transform. Bioinformatics 2010; 26:589–595.

17. Caporaso JG, Kuczynski J, Stombaugh J, Bittinger K, BushmanFD, Costello EK, et al. QIIME allows analysis of high-throughputcommunity sequencing data. Nat Methods 2010; 7:335–336.

18. McDonald D, Price MN, Goodrich J, Nawrocki EP, DeSantis TZ,Probst A, et al. An improved Greengenes taxonomy withexplicit ranks for ecological and evolutionary analyses ofbacteria and archaea. ISME J 2012; 6:610–618.

19. Edgar RC. Search and clustering orders of magnitude fasterthan BLAST. Bioinformatics 2010; 26:2460–2461.

20. Edgar RC, Haas BJ, Clemente JC, Quince C, Knight R. UCHIMEimproves sensitivity and speed of chimera detection. Bioinfor-matics 2011; 27:2194–2200.

21. Chao A. Estimating the population size for capture-recapturedata with unequal catchability. Biometrics 1987; 43:783–791.

22. Shannon CE. A mathematical theory of communication. BellSyst Tech J 1948:379–423and 623–656.

23. Segata N, Izard J, Waldron L, Gevers D, Miropolsky L, GarrettWS, et al. Metagenomic biomarker discovery and explanation.Genome Biol 2011; 12:R60.

24. Turnbaugh PJ, Hamady M, Yatsunenko T, Cantarel BL, DuncanA, Ley RE, et al. A core gut microbiome in obese and lean twins.Nature 2009; 457:480–484.

25. Liu CM, Hungate BA, Tobian AA, Serwadda D, Ravel J, Lester R,et al. Male circumcision significantly reduces prevalence andload of genital anaerobic bacteria. MBio 2013; 4:e00076.

26. Price LB, Liu CM, Johnson KE, Aziz M, Lau MK, Bowers J, et al.The effects of circumcision on the penis microbiome. PLoS One2010; 5:e8422.

27. Herfs M, Hubert P, Moutschen M, Delvenne P. Mucosal junc-tions: open doors to HPV and HIV infections? Trends Microbiol2011; 19:114–120.

28. Sewankambo N, Gray RH, Wawer MJ, Paxton L, McNaim D,Wabwire-Mangen F, et al. HIV-1 infection associated withabnormal vaginal flora morphology and bacterial vaginosis.Lancet 1997; 350:546–550.

29. Mitchell C, Balkus JE, Fredricks D, Liu C, McKernan-Mullin J,Frenkel LM, et al. Interaction between lactobacilli, bacterialvaginosis-associated bacteria, and HIV type 1 RNA and DNAGenital shedding in U.S. and Kenyan women. AIDS Res HumRetroviruses 2013; 29:13–19.

rized reproduction of this article is prohibited.