colonic microbiome is altered in alcoholism

Colonic microbiome is altered in alcoholism

Ece A. Mutlu,1* Patrick M. Gillevet,3* Huzefa Rangwala,3,4 Masoumeh Sikaroodi,3 Ammar Naqvi,3

Phillip A. Engen,1 Mary Kwasny,2 Cynthia K. Lau,1 and Ali Keshavarzian1

1Department of Medicine, Division of Digestive Diseases and Nutrition, Section of Gastroenterology, Rush University MedicalCenter; 2Department of Preventive Medicine, Northwestern University, Chicago, Illinois; 3Microbiome Analysis Center,George Mason University, Prince William Campus, Manassas; and 4Department of Computer Science, George MasonUniversity, Fairfax, Virginia

Submitted 20 September 2011; accepted in final form 6 January 2012

Mutlu EA, Gillevet PM, Rangwala H, Sikaroodi M, Naqvi A,Engen PA, Kwasny M, Lau CK, Keshavarzian A. Colonic micro-biome is altered in alcoholism. Am J Physiol Gastrointest LiverPhysiol 302: G966–G978, 2012. First published January 12, 2012;doi:10.1152/ajpgi.00380.2011.—Several studies indicate the impor-tance of colonic microbiota in metabolic and inflammatory disordersand importance of diet on microbiota composition. The effects ofalcohol, one of the prominent components of diet, on colonic bacterialcomposition is largely unknown. Mounting evidence suggests thatgut-derived bacterial endotoxins are cofactors for alcohol-inducedtissue injury and organ failure like alcoholic liver disease (ALD) thatonly occur in a subset of alcoholics. We hypothesized that chronicalcohol consumption results in alterations of the gut microbiome in asubgroup of alcoholics, and this may be responsible for the observedinflammatory state and endotoxemia in alcoholics. Thus we interro-gated the mucosa-associated colonic microbiome in 48 alcoholicswith and without ALD as well as 18 healthy subjects. Colonic biopsysamples from subjects were analyzed for microbiota compositionusing length heterogeneity PCR fingerprinting and multitag pyrose-quencing. A subgroup of alcoholics have an altered colonic micro-biome (dysbiosis). The alcoholics with dysbiosis had lower medianabundances of Bacteroidetes and higher ones of Proteobacteria. Theobserved alterations appear to correlate with high levels of serumendotoxin in a subset of the samples. Network topology analysisindicated that alcohol use is correlated with decreased connectivity ofthe microbial network, and this alteration is seen even after anextended period of sobriety. We show that the colonic mucosa-associated bacterial microbiome is altered in a subset of alcoholics.The altered microbiota composition is persistent and correlates withendotoxemia in a subgroup of alcoholics.

alcohol; alcoholic liver disease; pyrosequencing; length heterogeneitypolymerase chain reaction; colon; colonic microbiota

ALCOHOLISM IS ASSOCIATED WITH tissue injury and organ dysfunc-tion in a subgroup of alcoholics, and such injury may lead tomultiple complications including alcoholic liver disease (ALD)and neurological complications in 20–30% of them (14). Theobservation that only some, but not all, alcoholics developtissue injury indicates that chronic alcohol abuse is necessarybut not sufficient to cause organ dysfunction. Thus othercofactors besides direct toxicity of alcohol may be involved inthe development of complications from alcoholism. Severalanimal experiments and human observational studies suggestthat proinflammatory gut-derived bacterial products like endo-toxin may be cofactors for the development of tissue injury

associated with alcohol abuse: First, serum endotoxin levelsare elevated in both humans and rats with ALD, and theselevels correlate with ALD severity (3, 27). Second, monocytesfrom alcoholics with ALD appear to be primed for producingcytokines and oxidants after exposure to endotoxin (18). Fi-nally, lowering serum endotoxin levels by oral administrationof nonabsorbable antibiotics (1) or probiotics such as lactoba-cillus (26) and prebiotic oats (19) attenuates EtOH-inducedliver damage in rats.

One possible cause for increased levels of gut bacterial-derived proinflammatory products in alcoholics could be alter-ations in the gut microbiome composition or function. In fact,diet (of which alcohol is a major component in many societies)has been shown to have a significant effect on the gut micro-biome. The effect of chronic alcohol consumption on gutmicrobiome composition has not been well studied in humans,despite the new advances in molecular biology that have madeit possible to extensively interrogate microbiota in complexbiological environments like the gut (8). Thus the primary aimof this study was to characterize the gut microbiome compo-sition in alcoholics using nonculture, next generation sequenc-ing technologies to interrogate the 16S ribosomal RNA (16SrRNA) and validated computational techniques to taxonomi-cally classify and compare gut bacteria.

MATERIALS AND METHODS

Subjects

Sixty-six subjects were recruited at a tertiary medical center afterInstitutional Review Board approval of the studies by the RushUniversity Institutional Review Board and verbal and written in-formed consent of each subject was obtained. The following groups ofsubjects were recruited.

ALD (n � 19): inclusion criteria. Criteria for the ALD group wereas follows: 1) fulfill the National Institute on Alcohol Abuse andAlcoholism (NIAAA) (29) and DSM-IV criteria (2) for alcoholism;2) have a regular drinking history of at least 10 years (the minimumtime thought to be required for development of liver disease);3) presence of clinically significant liver disease as defined by at leastone of the following: elevated alanine aminotransferase (ALT) oraspartate aminotransferase (AST) that is �1.5� normal, and eitherlow platelets, low albumin, or elevated bilirubin; clinical evidence ofliver disease on the physical exam; when available, radiological[computerized tomography (CT) or ultrasound] or histological evi-dence of liver disease. To avoid confounding effects of advancedcirrhosis on bacterial composition, we chose to study only patientswith mild liver disease. In fact, the majority of our subjects with ALDhad a Child-Pugh class of A (see Table 1). Exclusion criteria were asfollows: 1) positive for Hepatitis C antibody, hepatitis C RNA, orhepatitis B surface antigen; 2) evidence of liver disease of anotheretiology such as autoimmune disease.

* E. A. Mutlu and P. M. Gillevet contributed equally to this work.Address for reprint requests and other correspondence: A. Keshavarzian,

Rush Univ. Medical Center, 1725 W. Harrison, Suite 206, Chicago, IL 60612(e-mail: [email protected]).

Am J Physiol Gastrointest Liver Physiol 302: G966–G978, 2012.First published January 12, 2012; doi:10.1152/ajpgi.00380.2011.

0193-1857/12 Copyright © 2012 the American Physiological Society http://www.ajpgi.orgG966

There are two subgroups within the ALD group. For active alco-holics with liver disease (AA � ALD, n � 8), criteria were all criteriafor ALD plus actively drinking up to 7 days before sample collectionper subject report or other evidence such as clinical records or exam.However, none were drinking 3 days before giving consent andsigning the consent form to assure that they fully understood the

study. For sober alcoholics with liver disease (SA � ALD; n � 11),criteria were all criteria for ALD plus no alcohol consumption for atleast 1 mo before sample collection per subject report or otherevidence such as clinical records or exam.

Alcoholics without liver disease (n � 29). Inclusion criteria foralcoholism and minimum duration of alcohol consumption were

Table 1. Subject characteristics

ALD ALC HC P Value

n 19 28 18Age 50 (31, 71) 41 (23, 71) 49 (30, 62) 0.216Sex, male, % 12 (63) 23 (79) 12 (67) 0.425Race, Caucasian, % 14 (82) 20 (71) 13 (72) 0.757Amount of drinking* 57,060 (4,320, 587,520) 35,280 (5,760, 287,280) 1,224 (0, 8,640) � 0.001Duration of drinking, years 28 (7, 51) 19 (5, 41) 24 (0, 43) 0.120Child-Pugh, A,B,C, % 74, 16, 10 — — —AST 47 (21, 163) 26 (17, 67) 20 (15, 30) � 0.001ALT 31 (18, 133) 30 (10, 95) 20 (10, 49) 0.082Total bilirubin 0.9 (0.1, 13.1) 0.5 (0.1, 2.0) 0.4 (0.3, 1.1) 0.005Endotoxin 2.516 (0.860, 3.530) 2.461 (1.000, 6.125) 0.675 (0.389, 1.458) 0.001At least college education, % 5 (28) 12 (43) 9 (50) 0.378Binge drinking, % 5 (45) 9 (53) 0 (0) 0.006Need for increasing drinks, % 9 (53) 19 (66) 1 (6) � 0.001Usual drinks, %

Beer 8 (44) 17 (59) 9 (53) 0.639Wine 6 (33) 11 (38) 11 (65) 0.121Hard liquor 12 (67) 23 (79) 3 (18) � 0.001

History, %Ulcers 4 (22) 3 (10) 0 (0) 0.095IBS 1 (6) 0 (0) 1 (6) 0.303Transfusion 9 (50) 2 (7) 1 (6) � 0.001Jaundice 8 (44) 4 (14) 0 (0) 0.002Gastrointestinal bleed 12 (67) 3 (10) 1 (6) � 0.001

Family history of ALD, % 2 (11) 8 (28) 3 (17) 0.383Family history of drinking, % 7 (41) 19 (79) 8 (47) 0.027Sodium 140 (126, 145) 140 (135, 145) 140 (138, 143) 0.771Potassium 4.0 (3.5, 5.9) 4.1 (3.4, 5.0) 4.0 (3.5, 5.0) 0.350Chloride 104 (95, 109) 104 (99, 108) 105 (101, 108) 0.416Bicarbonate 25 (21, 28) 26 (22, 33) 25 (21, 30) 0.044Blood urea nitrogen 13 (2, 49) 14 (8, 26) 13 (7, 19) 0.349Creatinine 0.8 (0.6, 1.6) 0.9 (0.6, 1.6) 1.0 (0.7, 1.2) 0.032Glucose 99 (57, 176) 81 (51, 115) 88 (71, 159) 0.026Total protein 6.9 (4.7, 8.6) 7.3 (6.0, 9.3) 7.5 (6.9, 8.2) 0.034Albumin 3.7 (1.9, 4.6) 4.1 (3.6, 4.5) 4.2 (3.6, 4.7) 0.017Calcium 9.1 (7.6, 9.7) 9.1 (8.0, 10.4) 9.4 (8.6, 10.2) 0.043Alkaline phosphatase 112 (60, 720) 72 (45, 123) 67 (43, 107) 0.008Hemoglobin A1c 5.6 (4.7, 6.1) 5.6 (5.3, 7.3) 5.6 (5.0, 7.9) 0.613Ferritin 169 (33, 336) 74 (19, 735) 80 (6, 170) 0.069Hemoglobin 14.0 (6.4, 18.4) 14.7 (12.5, 16.8) 14.8 (11.3, 16.4) 0.548MCV 91.8 (79.4, 105.0) 91.7 (75.6, 105.0) 86.7 (66.2, 93.1) 0.004Platelet 130 (54, 427) 218 (124, 603) 235 (177, 342) �0.001WBC 6.2 (2.4, 9.9) 6.2 (3.5, 15.8) 6.3 (3.3, 11.4) 0.573CRP 5.8 (1.0, 50.9) 5.0 (1.0, 57.0) 5.0 (5.0, 98.4) 0.171ASMA, % 1 (11) 2 (17) 1 (9) � 0.999ANA, % 3 (27) 1 (8) 1 (8) 0.405PT 13.7 (11.2, 22.6) 12.2 (10.9, 14.0) 12.2 (11.0, 13.2) 0.020INR 1.14 (0.91, 1.80) 1.03 (0.91, 1.24) 1.02 (0.92, 1.13) 0.036Fiber, adjusted, g/day 11.2 (8.7, 30.9) 17.7 (10.7, 21.7) 14.0 (8.6, 50.6) 0.524%Fruits/vegetables, adjusted 3.6 (1.5, 11.6) 4.7 (2.9, 8.9) 3.9 (1.4, 7.0) 0.473%Fat, adjusted 34.5 (28.5, 45.9) 33.4 (26.2, 39.6) 32.2 (27.6, 59.6) 0.728Smoking, %

Never 5 (26) 4 (14) 11 (61) 0.007Current 7 (37) 16 (55) 2 (11) 0.007Quit 7(37) 9 (31) 5 (28) 0.007

Illegal drug use, % 8 (44) 18 (64) 2 (11) 0.002

For subject characteristics, n (%) or median (minimum, maximum) by group. *Estimated total drinks � (average drinks/day) � (days/month reporteddrinking) � 12 � (years drinking); **Kruskal-Wallis used for nonparametric continuous or ordinal data; �2 test or Fisher exact test used for categorical data.ALD, alcoholics with liver disease; ALC, alcoholics without liver disease; HC, healthy controls; ALT, alanine aminotransferase; AST, aspartate aminotrans-ferase; IBS, irritable bowel syndrome; MCV, mean corpuscular volume; WBC, white blood count; CRP, C-reactive protein; ASMA, anti-smooth muscleantibody; ANA, antinuclear antibody; PT, prothrombin time; INR, international normalized ratio.

G967COLONIC BACTERIA IN ALCOHOLISM

AJP-Gastrointest Liver Physiol • doi:10.1152/ajpgi.00380.2011 • www.ajpgi.org

identical to the ALD group. Alcoholics were excluded for this groupif they had any evidence of liver disease; specifically, they wereexcluded if they had ALD as defined in the inclusion criteria for theALD group. Alcoholics were also excluded if they had any viral orautoimmune liver disease as defined in the exclusion criteria for theALD group.

There are two subgroups within the alcoholics without liver disease(ALC) group: active alcoholics without liver disease (AA; n � 14)and sober alcoholics without liver disease (SA; n � 15). Criteria todefine actively drinking and sobriety were identical to the ALD group.

Healthy control group (n � 18): inclusion criteria. Criteria for thehealthy control group (HC) were as follows: 1) normal physical exam,no digestive complaints, no known liver disease, normal liver functiontests (ALT, AST, bilirubin, alkaline phosphatase, serum albumin) and2) consumption of no more than a moderate amount of alcohol[NIAAA definition (29)]. Exclusion criteria were as follows: 1) dailydrinkers (�3� per wk) and 2) drinking (�3 drinks per occasion).

Additional exclusion criteria for all groups. Additional exclusioncriteria for all groups were as follows: 1) use of antibiotics for at least4 wk before sample collection; 2) unreliable drinking history (to ruleout closet drinkers or pretenders); 3) significant renal impairment(creatinine �1.2 mg/dL); 4) diseases that affect gastrointestinal mo-tility such as scleroderma, insulin-dependent diabetes, and/or uncon-trolled diabetes (Hgb-A1c �8%); 5) clinically significant dehydration,clinically detectable ascites, or significant peripheral edema, sepsis; 6)clinically significant cardiac failure; 7) regular daily use of medica-tions that may affect intestinal permeability such as NSAIDs orintestinal motility (e.g., metoclopramide); 8) subjects positive forother markers of liver disease such as smooth muscle antibody,hepatitis B surface antigen, hepatitis C antibody, or hemochromatosismarkers; 9) subjects with very low platelet count (�80 k), uncorrect-able prolonged PT (�15 s), or history of bleeding that precludebiopsies; 10) Asian descent due to the possible confounding effect ofa different polymorphism of enzymes involved in alcohol metabolism.Demographic characteristics of the study subjects enrolled in each ofthe groups are given in Table 1. Severity of liver disease was gradedby the Child-Pugh score (33).

From these subjects, the microbiota from the biofilm associatedwith the gut mucosa (mucosa-associated microbiome) was chosen tobe analyzed because we have shown previously that the mucosa-associated microbiome can be very different from the luminal micro-biome (12, 20).

Tissue Procurement

A limited and unprepped sigmoidoscopy was performed usingOlympus video scopes (Olympus America, Center Valley, PA) forresearch purposes. During biopsy procurement, we inflated the rectumwith air. All subjects had solid stool; therefore, there was littlecovering of the mucosa with mucoid stool itself, and solid chunks ofstool were seen in the rectum. Care was taken not to use any suctionduring advancement of the scope to 20–25 cm from the anal verge.The sterile biopsy forceps were not taken out of the channel of thescope until an area that is completely clear of stool was seen with clearpink mucosa. Biopsies were taken from the pink mucosa that is notcovered with any stool, at the sigmoid colon at about 20–25 cm fromthe anal verge using a 2.2 mm sterile standard biopsy forceps. Allsamples were immediately snap frozen at the time of collection inliquid nitrogen and were stored in a �80°C freezer until analysis.

Interrogation of Intestinal Bacteria

We used molecular methods to interrogate and characterize gutmicrobiome composition in alcoholics. First, we used Length Heter-ogeneity PCR (LH-PCR) fingerprinting to rapidly survey our samplesand standardize the community amplification. We then interrogatedthe microbial taxa associated with the gut mucosal microbiome usingmultitag pyrosequencing (MTPS) on a subset of the samples (51 of 66

samples) (13). We used MTPS to interrogate gut mucosal microbiomeof all patients with ALD (n � 19), 22 of 28 subjects with ALC, and10 of 18 healthy subjects. We elected to interrogate all subjects withALD because, according to our original hypothesis, the alcoholicswith liver disease group was our experimental group, whereas thealcoholics without liver disease group was our control group foralcoholism. We randomly selected samples from HC and ALC groupswith �2:1 favoring ALC group over the healthy subject group. TheMTPS latter technique allows the rapid sequencing of multiple sam-ples at one time yielding thousands of sequence reads per sample. Wechose to interrogate the mucosa-associated microbiome rather thanstool because of potentially higher relevance of this to mucosalepithelial function in contrast with the luminal fecal microbiome,which has been postulated to be transient and could be related todietary factors (30).

LH-PCR fingerprint analysis. LH-PCR fingerprinting was done aspublished previously (20). Fingerprints were obtained in duplicate ortriplicate for each sample. Briefly, total genomic DNA was extractedfrom tissue using Bio101 kit from MP Biomedicals, Montreal, Que-bec, as per the manufacturer’s instructions. About 10 ng of extractedDNA was amplified by PCR using a fluorescently labeled forwardprimer 27F [5=-(6FAM) AGAGTTTGATCCTGGCTCA G-3=] andunlabeled reverse primer 355R= (5=-GCTGCCTCCCGTAGGAGT-3=) that are universal primers for bacteria (21). The LH-PCR productswere diluted according to their intensity on agarose gel electrophore-sis and mixed with ILS-600 size standards (Promega) and HiDiFormamide (Applied Biosystems, Foster City, CA). The diluted sam-ples were then separated on the SCE9610 fluorescent capillary se-quencer (Spectrumedix, State College, PA) and processed using theGenoSpectrum software package (Spectrumedix LLC, State College,PA). The GenoSpectrum software package deconvolves the fluores-cence data and converts it into electropherograms where the peaks ofthe electropherograms represent PCR amplicons representing differ-ent species or Operational Taxonomic Units (OTU). The LH-PCRfingerprinting data were then analyzed using a custom PERL scriptthat combines data from several runs, interleaves the various profiles,and normalizes the data. The normalized peak areas were calculatedby dividing an individual peak area by the total peak area in thatprofile. Hence each normalized peak area corresponded to the relativeabundance of a specific OTU within the sample. Duplicate or triplicateLH-PCRs were run on each sample, and the most consistent profilewas selected for further analysis. Peaks constituting less than 1% ofthe total community from each sample were eliminated from theanalysis to remove the variable low abundance components within thecommunities. We chose threshold of 1% because this value corre-sponds to the detection limit of the LH-PCR technology, and anypeaks less than 1% may not be reproducible. Additionally, an under-lying a priori assumption for this filtering is that the low abundancecomponents of the community vary between individual subjects andwill not contribute significantly to the functionality of the gut mucosalmicrobiome (11). We have also used the LH-PCR fingerprintingmethodology as a quality control measure to assure that we arelinearly amplifying the community so that the resulting sequenceanalysis accurately represents the community composition as de-scribed below.

MTPS. We employed a MTPS process (13) to characterize themicrobiome from a subset of the mucosal samples that were used inthe LH-PCR analysis. Specifically, we have generated a set of 48emulsion PCR fusion primers that contain the 454 emulsion PCRlinkers and different 7 base barcode on either of the 27F or 355Runiversal 16S rRNA primers. Thus each mucosal sample was ampli-fied with a uniquely barcoded set of forward and reverse 16S rRNAprimers, and then up to 48 samples were pooled and subjected toemulsion PCR and pyrosequenced using a GS-FLX pyrosequencer(Roche). Data from each pooled sample were deconvoluted by sortingthe sequences into bins based on the barcodes using custom PERLscripts. Thus we were able to normalize each sample by the total

G968 COLONIC BACTERIA IN ALCOHOLISM


number of reads from each barcode. We have noted that ligatingtagged primers to PCR amplicons distort the abundances of thecommunities, and thus it is critical to incorporate the tags during theoriginal amplification step. We therefore used fusion primers duringthe pyrosequencing reaction and eliminated the ligation step that hasbeen used by others (4). Several groups have employed variousbarcoding strategies to analyze multiple samples, and this strategy isnow well accepted (34).

Analysis of MTPS Data

Quantitative Insights Into Microbial Ecology (QIIME) softwarepipeline (VirtualBox Version 1.1.0) was used to analyze the MTPSdata (6). Low quality sequences and sequences less than 100 bp wereeliminated from the analysis by filtering using custom PERL scripts(28). OTUs were picked using uclust (22, 23) at a 97% similarity.Sequences were aligned with PyNAST (5) and identified using theRDP database and a naïve Bayesian classifier (36) using a 75%bootstrap value threshold.

We tabulated results for each taxa in each sample. We visuallyexamined the ordination of cases (i.e., clustering of the cases) byprincipal coordinates (PCO) analysis for the presence of dysbiosis.The PCO and canonical correspondence analysis (CCA) were per-formed using the Multivariate Statistical Package (Kovach, Wales,UK). A Bray Curtis distance metric was used for the PCO analysis.Environmental variables were normalized for the CCA.

With Qiime, weighted and unweighted Unifrac distances (15) werealso used to generate -diversity graphs. The FastUnifrac was used togenerate the overall Unifrac p-test (16), which is generated by ana-lyzing the clustering of the taxa for each sample in a phylogram andthen comparing the topology of the phylograms for the samples foreach class. -Diversity was assessed using the Chao estimator.

Measurement of Endotoxin

Gram negative bacterial endotoxin in human serum specimens wasquantitated using the QCL-1000 kit manufactured by BioWhittaker/Cambrex in compliance with the U.S. FDA Guideline Validation ofthe LAL test as an end-product endotoxin test for human and animalparenteral drugs, biological products, and medical devices. Serumblanks were used in addition to the kit standards.

Diet Analysis

National Institutes for Health, Eating at America’s Table All dayFruit and Vegetable Screener (riskfactor.cancer.gov/diet/screeners/fruitveg/instrument.html), and Percent Energy from Fat Screener(http://riskfactor. cancer.gov/diet/screeners/fat/) were employed to es-timate fruit, vegetable, fiber, and fat intake.

Statistics

SAS (Version 9.1; Cary, NC) statistical package was used toanalyze clinical metadata and differences in clinical variables betweenthe dysbiotic and nondysbiotic groups. SPSS (Version 17.0.0; Chi-cago, IL) was used to perform nonparametric Kruskal-Wallis orMann-Whitney tests in the clinical study groups and to performmedian tests, as appropriate. �2-Test or t-tests were used to detectdifferences in proportions between groups as appropriate in SAS orSPSS. Metastats was used to compare bacterial groups in the dysbioticand nondysbiotic analysis, with a nonparametric t-test as describedpreviously (37). R-project packages rgl and car were used to generatescatterplots (R-project.org).

RESULTS

Study Subjects

There were no statistically significant differences in terms ofage, sex, and race among the three study groups, namely ALC,

ALD, and HC (Table 1). As expected, the estimated cumula-tive lifetime amount of alcohol intake was significantly higherin both alcoholic subject groups (i.e., ALC and ALD) com-pared with HC (P � 0.001). The duration of drinking wassimilar among the two groups of alcoholics (P � 0.12). Bingedrinking, need for increasing amounts of drinking, and hardliquor drinking were more frequent in the alcoholic subjectgroups compared with HC (Table 1). Most of the subjects withALD had a Child-Pugh class of A, compatible with mildcirrhosis. The total bilirubin, AST, PT, and INR were signifi-cantly higher in the ALD group compared with the ALC andHC groups as expected (Table 1). Albumin, calcium, RBC, andplatelet counts were lower in the ALD group (Table 1). TheALT was not significantly different between the groups despitea higher numeric value in the ALD group (P � 0.082). Asexpected, history of GI bleed and blood transfusion and jaun-dice were more frequent in the ALD group compared with theALC and HC groups (Table 1). Smoking and history of pastdrug use was reported more often in both of the alcoholicgroups, compared with HCs (Table 1). Also as expected,alcoholic groups had more diabetic cases, who had mildincreases in serum glucose without a significantly elevatedhemoglobin A1c (Table 1). Additionally, serum endotoxinlevels were significantly higher in both alcoholic groups com-pared with HC (P � 0.001) (Fig. 1). The endotoxin values forall HC subjects were in the first 25% quartile of all theendotoxin values. There was no difference between serumendotoxin levels among the alcoholics with and without liverdisease (P � 0.419).

Analysis of LH-PCR Fingerprint Data

Total DNA was extracted from each of the 66 biopsysamples. The V1 to V2 hypervariable regions of the 16S rRNAwere amplified with PCR using universal bacterial primers.This generated PCR products (i.e., amplicons) from all of thebacterial taxa in each sample, which vary in length based onthe size of the hypervariable region within the bacteria in agiven sample. Each PCR product from each of the samples was

Fig. 1. Endotoxin values by study group. Endotoxin values are in endotoxinunits per milliliter. When the 3 study groups [healthy controls (HC) vs.alcoholics without liver disease (ALC) vs. alcoholics with liver disease (ALD)]are compared, endotoxin values were statistically significantly different (P �0.001; Kruskal-Wallis). Results of post hoc comparisons are given as bars attop part of graph.



http://riskfactor.cancer.gov/diet/screeners/fruitveg/instrument.html

http://riskfactor.cancer.gov/diet/screeners/fruitveg/instrument.html

http://cancer.gov/diet/screeners/fat/

http://R-project.org

then separated using fluorescent capillary electrophoresis, toproduce an electropherogram consisting of peaks of variablelengths, representing OTUs or taxa in the sample. The height ofeach peak on the electropherogram corresponds to the abun-dance of a particular OTU within the sample. Thus eachelectropherogram is a fingerprint of the bacterial communitywithin a given sample. Each electropherogram was used to doan initial analysis of the quality of the amplification processand to estimate the community diversity.

Fingerprints were analyzed to visualize clustering of the 66samples using PCO with a Bray Curtis distance measure. PCOis an ordination method similar to principal components anal-ysis (PCA) except it uses a distance metric instead of covari-ances. The method performs a matrix analysis (i.e., an Eigenanalysis) to plot the variance of the data along orthogonal axesor principle components. The first principle component repre-sents the largest amount of variation in the dataset, whereas thesecond principle component represents the next largest mea-sure of the variance.

Figure 2 shows the distribution of each case along the firstthree axes of the PCO in a three-dimensional scatterplot.The locations for each of the ellipsoids that contain 70% ofthe cases in the ALC group, the ALD group, and the HCgroup along the first three PCO axes appeared markedlydifferent among the groups. Specifically, a proportion ofalcoholics from both the ALC and ALD groups was locatedfar away from the HC cluster. This finding suggests that asubgroup of subjects with alcoholism (both with and withoutliver disease) may have altered colonic microbiota compo-sition in comparison with HC.

Analysis of MTPS Data

To identify the specific bacterial taxa that were implicated inthe dysbiotic bacterial communities in alcoholics, we per-

formed MTPS on 51 of the 66 sigmoid mucosa samples fromHC (n � 10), ALC (n � 22), and ALD (n � 19). We obtained111,174 raw reads from two GS FLX pyrosequencing runs andidentified the appropriate tags in 105,207 of these reads.Low-quality sequences below read lengths of 100 bp werefiltered out, leaving 80,121 total reads that were analyzed. Thefiltered reads had an average read of 1,571 per sample and anaverage read length of 243 bps. Negative controls did notdemonstrate contamination during the pyrosequencing process.

�-Diversity analyses. -Diversity is the measure of changein diversity between samples across environmental gradients(38). In our case, it reflects the changes in bacterial composi-tion between different levels of alcohol exposure and disease inthe clinical study groups, i.e., it reflects shifts in the microbialcommunity composition with exposure to alcohol. Variousmetrics can be used to determine differences in bacterialcomposition between the clinical study groups. These metricscan be based on mathematical distances such as a Euclidiandistance or a Bray Curtis distance. Alternatively, these metricscould measure the topology of a phylogenetic tree (Unifracdistance) constructed using the samples. We have used bothmethods to analyze our data.

TAXA ABUNDANCE ANALYSIS. Taxa present in each samplewere tabulated according to the RDP10 bacterial sequencedatabase using a naive Bayesian classifier. The samples werethen ordinated using PCO for clustering, i.e., for the presenceof dysbiosis. Figure 3 depicts the PCO analysis of the relativeabundance of each bacterial taxa in a sample at the class level.A cluster of subjects with a similar microbiome compositionhas been denoted by a circle in Fig. 3 and is referred to as thenondysbiotic group. The rest of the samples located outside ofthis circle represent the cases that have a dysbiotic microbiomecomposition and have been denoted as the dysbiotic group. Asshown in the PCO graph, using this definition, not all subjectswith alcoholism or liver disease were dysbiotic, but there were13 cases identified as dysbiotic. Furthermore, most of the HCswere clustered closer to each other compared with the alcohol-

Fig. 3. Ordination by PCO plots of the taxa abundance at the class level in 3dimensions. The axes represent the first 3 highest discriminating axes using aBray Curtis distance measure. HC are depicted as blue. ALC are depicted asgreen. ALD are depicted as magenta. The core microbiome cluster is denotedby a manually inserted circle. Cases outside of the circle are classified asdysbiotic.

Fig. 2. Principal coordinate analysis (PCO) plots of the length heterogeneity(LH)-PCR fingerprint abundance data in 3 dimensions. The axes represent thefirst highest discriminating axes using a Bray Curtis distance measure. HC aredepicted as blue. ALC are depicted as green. ALD are depicted as magenta.Each dot corresponds to 1 case. Circles denote the 70% ellipsoid for eachgroup.



ics. It also appears that the microbiome composition (i.e., taxaand their abundance) was primarily altered in alcoholics whenthey were compared with HC: None of the healthy subjects(blue dots) (0 of 10) were outside the core cluster (i.e., are notdysbiotic), whereas 8 of 22 (36.7%) ALD subjects (magentadots) and 5 of 19 (26.3%) of ALC subjects (green dots)were dysbiotic. However, PCO analysis did not show a visualdifferential clustering between subjects with liver disease com-pared with those without liver disease, even though more of thedysbiotic cases were from the ALD group. Similarly, the soberand active alcoholics did not differentially cluster in the graph.

UNIFRAC BASED ANALYSIS. Second, differences between thestudy groups in the sequence data were analyzed usingweighted Unifrac analysis. Unifrac analysis examines the re-lationships within the studied cases based on their distancesfrom each other in a phylogenetic tree. The sequences werefirst clustered into OTUs and then assigned taxonomic ID anda neighbor joining tree was generated as described in theMATERIALS AND METHODS. Overall Unifrac p-test for the entiresample set was performed on FastUnifrac (16), indicatingsignificant clustering of the sample class within the phyloge-netic tree (P � 0.001). A PCO analysis using weighted Unifracdistances between cases (Fig. 4) demonstrates that the 70%ellipsoid for the HC group appears different than the corre-sponding ellipsoids for both the ALC and ALD groups, al-though there is some overlap. When individual cases areexamined in Fig. 5 (shown at a different 3-dimensional anglefor further clarity), about 25% of the alcoholic cases (11 of 41)(that are denoted by 2 separate circles) lie away from the maincluster of cases.

Whether these eleven cases show a continuum of changeaway from HC needs to be further evaluated. The dysbioticcases were almost equally distributed among the ALC andALD groups, indicating no apparent difference by liver diseasestatus. In secondary comparisons, sobriety status did not

clearly differentiate the dysbiotic cases from the rest (Fig. 6),neither did serum endotoxin value quartiles (Fig. 7).

In summary, both the taxa abundance and Unifrac-based-diversity analyses show that a subset of alcoholics is dysbi-otic. Below, we study the differences in individual bacterialtaxa in the clinically defined groups, as well as groups definedby the above analyses.

Fig. 4. Ordination by weighted Unifrac distances by subject group. PCO plotsof the subjects by subject group using weighted Unifrac distances are shown.The axes represent the first 3 highest discriminating axes. HC are depicted asblue. ALC are depicted as green. ALD are depicted as magenta. Circlesrepresent 70% ellipsoids for each group.

Fig. 5. Ordination by weighted Unifrac distances identifying dysbiotic cases.PCO plot by study group at a different angle denoting the dysbiotic cases isshown. HC are depicted as blue. ALC are depicted as green. ALD are depictedas magenta. Cases that lie away from the core HC group ellipsoid are given in2 circles and denote the dysbiotic cases identified by weighted Unifracanalysis. Dysbiotic cases belong to both the ALC and ALD groups.

Fig. 6. Ordination by weighted Unifrac distances by sobriety status. PCO plotof the subjects by sobriety status using weighted Unifrac distances is shown.HC are depicted as blue. ALC are depicted as green: sober alcoholics withALC are depicted as light green and active alcoholics with ALC are depictedas dark green. ALD are depicted as magenta: sober alcoholics with ALD aredepicted as light magenta and active alcoholics with ALD are depicted as darkmagenta. The dysbiotic cases were not discriminated by sobriety status.



Differences between individual bacterial taxa. DIFFERENCES

BETWEEN TAXA IN THE CLINICALLY DEFINED STUDY GROUPS. Themajority (99.9%) of the sequences found in the dataset wereclassified as bacteria, with �0.01% of the sequences classifiedas archaea or other. The minor differences among the studygroups (HC vs. ALC vs. ALD) in archaea or unclassifiedsequences were not statistically significant. At the phylumlevel, the sequences seen were typical of gut microbiota(Fig. 8) and there were no differences between the clinicallydefined study groups at this taxonomic level. At the family

level, as shown in Fig. 9, the mean abundance of Bacte-roidaceae from Bacteroidetes was decreased in the alcoholicgroups compared with the HC and the groups were statisticallysignificantly different (P � 0.035; Kruskal-Wallis). When taxathat had an abundance �1% were examined only, there wereno other major differences between the study groups.

DIFFERENCES BETWEEN THE BACTERIAL TAXA OBSERVED IN DYS-

BIOTIC AND NONDYSBIOTIC CASES. The community compositionof the dysbiotic and nondysbiotic groups appeared differentregardless of the methodology used to identify dysbiotic cases.In fact, eleven cases were identified as dysbiotic in both of the-diversity analysis methods (namely, visual examination ofthe ordination by PCO using Bray-Curtis or weighted Unifracdistances). The eleven cases denoted as dysbiotic by both ofthese ordination methods were then used to identify the differ-ences between bacterial taxa in the dysbiotic and nondysbioticgroups. We compared the 11 dysbiotic cases to the nondysbi-otic ones at the class level using the Metastats statisticalanalysis (37). Results are shown in Table 2. At the class level,there was a uniform reduction of Bacteroidetes in the dysbioticcases. Other major differences included decreases in Clostridiaand increases in Bacilli and Gammaprotoebacteria in the dys-biotic group compared with the nondysbiotic group (Fig. 10).In fact, when the cases were ordered at the phyla level in termsof their Bacteroidetes abundance, all of the eleven dysbioticcases were seen to cluster at the lower Bacteroidetes-abun-dance-end of the graph (Fig. 11). Therefore, the communitycomposition of the dysbiotic cases is very different from thosewithin the nondysbiotic cases, suggesting an overall disarray ofthe gut bacterial microbiome in the dysbiotic cases.

Network analysis. Network analysis analyzes the asymmet-ric relationships between discrete entities in complex models.This form of analysis has recently been used to investigate theecological relationships between bacterial components in thevaginal microbiome (10). We have modeled previously undi-rected unweighted networks for each of the five patient classesto represent the potential correlations between the different

Fig. 7. Ordination by weighted Unifrac distances by endotoxin quartile. PCOplot of the subjects by endotoxin quartile using weighted Unifrac distances isshown. Endotoxin quartile increases as colors get darker: white, light yellow,orange, and red represent 1st (lowest), 2nd, 3rd, and 4th (highest) endotoxinquartiles, respectively. The dysbiotic cases were not discriminated by endo-toxin quartile. Cases in which endotoxin levels were not available weredepicted in green.

Fig. 8. Pie chart of multitag pyrosequencing data ana-lyzed at the phylum level. Uncommon phyla that are avery small fraction of the total may not be visible in thechart even though they are present in the legend.



phylotypes within patient groups (28). In this network model,the identified bacterial phylotypes are represented by a set ofvertices V or nodes. An edge Ei,j exists between two givenvertices Vi and Vj if both phylotypes are found together in thesample above the mean for that phylotype in all samples of aclass. These network models are visualized in the Cytoscape(5) software package. The number of edges, i.e., the number ofconnections for each node, is considered the degree of a node.The degree of the node is then rank ordered and plotted aseither a connectivity plot (Fig. 12) or a cumulative distributionfunction (CDF). In Fig. 15, we observed different connectivitybetween the dysbiotic and the nondysbiotic groups, both of

which were defined in Fig. 5. The dysbiotic group (gray line)had lower connectivity compared with the nondysbiotic (blackline).

There was also a different connectivity between the healthy,sober alcoholic, and active alcoholic classes (Fig. 13). Cumu-lative distribution function plots of the subclasses indicate thatHC (black dot) have the highest connectivity, whereas theactively drinking alcoholics (stars) have the lowest (Fig. 14).The sober alcoholics (open circle) have intermediate connec-tivity. Interestingly, the presence of ALD is not correlated withconnectivity in this dataset as described in detail previously(28). Therefore, the effects of alcohol alone (in shaping thetypes of organisms that are associated with the colonic mucosa)appear to be more predominant than the effects seen with liverdisease.

Correlation of clinical features with dysbiosis. Becausethere was a significant difference between the composition ofthe microbiome in our dysbiotic cases, we then investigatedwhether this difference could be correlated with clinical fea-tures. Specifically, we looked for clinical differences that couldpossibly explain the dysbiotic subset of cases; when the dys-biotic cases were compared with the nondysbiotic cases, therewas no difference in age, sex, ethnicity, BMI, or recruitmentsite. The dysbiotic cases had a higher frequency of diabetes(45% vs. 3% in the dysbiotic vs. nondysbiotics, respectively;P � 0.001). There was also a higher mean hemoglobin A1cvalue in the dysbiotics compatible with mild diabetes [HgbA1c � 6.0(range � 5.3–6.7) in dysbiotics vs. HgbA1c � 5.5 (range �4.7–6.0) in nondysbiotics (P � 0.009)]. This result also cor-

Fig. 9. Bar graph of mean percent abundance of Bacteroidaceae at family levelin the study groups � 2 SE. Mean abundance of Bacteroidaceae was decreasedin the alcoholic groups (P � 0.035; Kruskal-Wallis).

Table 2. Differing taxa between dysbiotic and nondysbiotic subject groups

Name of Bacterial Taxa at Class Level

Percent Abundance of Bacterial Taxa

Q Value

Dysbiotic Nondysbiotic

Mean Variance Mean Variance P Value

Archaea; Other 0.0000 0.0000 0.0000 0.0000 1.000 1.000Acidobacteria; Acidobacteria 0.0000 0.0000 0.0000 0.0000 1.000 1.000Actinobacteria; Actinobacteria 0.0283 0.0008 0.0242 0.0003 0.634 1.000Bacteroidetes; Bacteroidetes 0.0911 0.0100 0.4007 0.0061 0.001 0.016Bacteroidetes; Flavobacteria 0.0002 0.0000 0.0001 0.0000 0.359 1.000Bacteroidetes; Other 0.0001 0.0000 0.0025 0.0000 0.009 0.123Bacteroidetes; Sphingobacteria 0.0005 0.0000 0.0001 0.0000 0.022 0.256Chloroflexi; Anaerolineae 0.0000 0.0000 0.0001 0.0000 1.000 1.000Cyanobacteria; Cyanobacteria 0.0011 0.0000 0.0003 0.0000 0.148 1.000Firmicutes; “Bacilli” 0.1651 0.0356 0.0074 0.0001 0.001 0.016Firmicutes; “Clostridia” 0.1626 0.0330 0.4105 0.0068 0.001 0.016Firmicutes; “Erysipelotrichi” 0.0195 0.0018 0.0386 0.0024 0.256 1.000Firmicutes; Other 0.0002 0.0000 0.0022 0.0000 0.001 0.016Fusobacteria; Fusobacteria 0.0017 0.0000 0.0120 0.0023 0.337 1.000Lentisphaerae; Lentisphaerae 0.0000 0.0000 0.0000 0.0000 1.000 1.000Root; Bacteria; Other 0.0092 0.0001 0.0144 0.0003 0.244 1.000Proteobacteria; Alphaproteobacteria 0.0104 0.0007 0.0033 0.0000 0.513 1.000Proteobacteria; Betaproteobacteria 0.0563 0.0062 0.0287 0.0009 0.354 1.000Proteobacteria; Deltaproteobacteria 0.0022 0.0000 0.0034 0.0001 0.643 1.000Proteobacteria; Epsilonproteobacteria 0.0019 0.0000 0.0020 0.0000 0.984 1.000Proteobacteria; Gammaproteobacteria 0.4480 0.1116 0.0358 0.0029 0.001 0.016Proteobacteria; Other 0.0008 0.0000 0.0004 0.0000 0.612 1.000SR1; SR1_genera_incertae_sedis 0.0000 0.0000 0.0000 0.0000 1.000 1.000Spirochaetes; Spirochaetes 0.0008 0.0000 0.0129 0.0063 0.837 1.000TM7; TM7_genera_incertae_sedis 0.0000 0.0000 0.0000 0.0000 1.000 1.000Verrucomicrobia; Verrucomicrobiae 0.0001 0.0000 0.0005 0.0000 0.032 0.330Root; Other 0.0001 0.0000 0.0001 0.0000 1.000 1.000

The bacterial taxa associated with the nondysbiotic and dysbiotic groups were compared using Metastat at the class level. Mean is the mean percent abundanceof the listed taxa.



responds to the observation that there is a higher incidence ofmild diabetes in subjects who are heavy drinkers in otherstudies. The bowel movement frequency in the dysbiotic andnondysbiotics was similar (P � 0.05). The symptom scales fordiarrhea, gas, and bloating were not significantly different inthe dysbiotic and nondysbiotic cases (P � 0.05), suggestingthat it is unlikely that the cause of bacterial compositionchanges in colonic biopsies of these subjects could be exocrinepancreatic insufficiency or small intestinal bacterial over-growth. However, reflux symptoms were more common amongthe dysbiotic subjects (36% vs. 8% in the dysbiotic vs. non-dysbiotics, respectively; P � 0.034). There was a higherfrequency of diuretic use among the dysbiotics versus thenondysbiotics (50% vs. 17% in the dysbiotic vs. nondysbiotics,respectively; P � 0.043; 50%). Dysbiotic cases also had alower mean serum chloride level (102 mmol/L vs. 104 mmol/Lin the dysbiotic vs. nondysbiotics respectively; P � 0.048) anda higher red cell distribution width (15% vs 13.8% in thedysbiotic vs. nondysbiotics respectively; P � 0.015). Other-wise, there were no differences in comorbitidies, medicationuse, gastrointestinal symptoms, smoking history, family his-tory including the family history of alcoholism, use of illicitdrugs, smoking, employment, education level, type of drinkingbehavior (binge drinking vs. weekend drinking vs. daily drink-ing), type of EtOH consumed (beer vs. wine vs. hard liquor), orparameters typically measured in a complete blood count andmetabolic panel such as hemoglobin, platelet count, renalfunction tests, or liver AST, ALT, total bilirubin, and APlevels. In a limited group of subjects, dietary data were avail-able (n � 19 for total group; n � 6 in HC; n � 6 for ALC; n �7 for ALD groups). Adjusted percent energy from fiber andadjusted percent energy from fat in the diet and adjustedpercent energy from fruits and vegetables were not differentbetween the dysbiotic and nondysbiotic cases (all P � 0.05).

Correlation of serum endotoxin with dysbiosis. CCA is amultivariate direct gradient analysis method used in ecology.In CCA, the bacterial data are directly related to one or moreenvironmental variables. The CCA is similar to the PCOanalysis in that it first clusters the data based on the taxacomposition but performs a constrained ordination using envi-ronmental factors to determine how the microbiome data aredistributed along the environmental gradients. In our case, wewanted to visualize whether the cases distributed along low tohigh endotoxin values (the environmental gradient). CCA wasperformed on a subset of the MTPS samples (41 subjects) thathad endotoxin level measurements on the day of mucosalbiopsy. The endotoxin gradient is given by a linear arrow onthe x axis. Higher endotoxin values are seen in cases to the leftof the graph and lower values are seen in cases to the right ofthe graph. Specifically, one can see that a portion of alcoholicsshown in gray and open triangles cluster to the left of thecontrols (shown in black squares). The microbiome composi-tion in the samples distributed along an endotoxin gradientfrom low endotoxin (in HC) to high endotoxin (in alcoholicgroups) but did not correlate with state of alcoholism (sober oractive alcoholic states).

�-Diversity. -Diversity as assessed by the Chao1 index wasnot significantly different within the study groups (Fig. 15) inthis dataset. The rarefaction curves suggest that an increasedamount of reads may further identify additional OTUs ofinterest.

DISCUSSION

Alcohol consumption could be a major factor influencing thegut microbiome composition and function, and, in turn, the gutmicrobiome can have a profound impact on alcohol metabo-lism as well as the metabolic and biological effects of alcoholin the body. However, the data on the effect of alcohol on thegut microbiome in humans are very limited. To the best of ourknowledge, this study represents a first attempt at showingchanges in the gut microbiome in human alcoholics usingnonculture methodologies. Using a variety of state of the artanalysis methods, we show that chronic alcohol consumption isassociated with altered dysbiotic microbiota composition in asubset of alcoholics. We report that the alcoholics with dys-biosis had lower median abundances of Bacteroidetes andhigher ones of Proteobacteria. When the study subjects areexamined according to study group, the alcoholic groups had areduction in abundance of Bacteroidaceae.

In this dataset, there was no correlation between the durationof sobriety and the presence of dysbiosis, suggesting that theeffects of chronic alcohol consumption are not temporary butrather long-lasting; a subset of both actively drinking and soberalcoholics had dysbiotic mucosal-associated microbiota. Therewas a higher frequency of mild diabetes in our dysbioticsubjects. This could be due to the well-described higher fre-quency of mild diabetes in subjects with heavy alcohol use andmay simply be reflective of the subject population (35). Alter-natively, it is also possible that mild diabetes could alter thecolonic microbiota composition directly. There were no clini-cal symptoms of pancreatic exocrine deficiency or small intes-tinal bacterial overgrowth in these subjects to suggest anindirect effect of diabetes on bacterial composition. We sug-gest that future cross-sectional studies in alcoholics should

Fig. 10. Bar graphs of the differences in major taxa at the class level in dysbioticvs. nondysbiotic cases by all analysis methods. Eleven cases were found to bedysbiotic by all ordination methods employed in the study. Dysbiotic cases hadlower percent mean abundances of Bacteroidetes (P � 0.016; Metastats) andBacilli and Clostridia (P � 0.016 both; Metastats) and higher percent meanabundances of Gammaproteobacteria (P � 0.016; Metastats).



match for the presence of mild diabetes between cases andcontrols or make adjustments to the study design as appropri-ate. Between the dysbiotic and nondysbiotic subjects, there washigher frequency of diuretic use, suggesting more cirrhotic

cases were among the dysbiotics, although none of our patientshad decompensated cirrhosis with ascites. Reflux symptomswere more common among dysbiotics; this suggests largeramounts of alcohol consumption immediately before samplecollection among the dysbiotics leading to upper gastrointes-tinal tract symptoms, even though cumulative lifetime alcoholintake and the type of alcohol consumed were not different.Red cell distribution width was also slightly higher in thedysbiotics, which perhaps may suggest acute alcohol intakeamong the dysbiotics. There were no significant differences inother clinical metadata such as BMI (which has been shown tocorrelate with microbiota composition in other studies) ordietary fat and fiber intake, suggesting the microbiota changesobserved in the dysbiotic group are not simply due to theeffects of a confounder. However, dietary assessments wereperformed in a very limited subset of our cases and should beperformed in all subjects going forward.

Our inability to detect clearer differences between alcoholicswith and without liver disease might be due to several reasons.First, the differences seen in mucosa-associated bacterial com-position in our dataset appear to be most evident when healthysubjects are compared with those with alcoholism. This sug-gests that chronic alcohol consumption, rather than liver dis-ease, is the most important event that appears to alter micro-biota composition. Second, the number of subjects in the

Fig. 11. Rank order by abundance of the Bacte-roidetes phylum. In the stacked histogram, the yaxis shows the percent abundance of the 4 mostabundant phyla for each study subject and the xaxis labels show the group for the study subject. SAdenotes a subject who was a sober alcoholic with-out liver disease; SA � ALD denotes a subject whowas a sober alcoholic with liver disease; AA de-notes a subject who was an active alcoholic withoutliver disease; AA � ALD denotes a subject whowas an active alcoholic with liver disease. Theabundance of the Bacteroidetes phylum in eachsubject was rank ordered and graphed in order ofrising percent abundance. Bacteroidetes is denotedby the yellow portion of the bars for each subject.In this stacked histogram, the other most abundanttaxa in each subject are color coded as follow:Actinobacteria phylum (green); Firmicutes phylum(red); Proteobacteria phylum (blue); Archea (pink);and all other sequences (brown). A rise in theBacteroidetes phylum abundance is seen at aboutthe 30% level. The 13 samples that had the lowestabundance in the Bacteroidetes phylum have beenmarked at the left lower corner of the graph.

Fig. 12. Connectivity plot of dysbiotic and nondysbiotic groups from networkanalysis. Each taxa is represented as a node in complex graph, and an edge ismade between 2 nodes if they are present in the same class and above a definedthreshold. We then compared network topologies. We present the connectivityplot by node (taxa) for the 2 defined categories.



current study may have been sufficient to identify the grossmicrobiome disruptions that are related to alcoholism itselfcompared with the healthy state, but the microbiome changesin subjects with liver disease may be more subtle and requirea larger sample size, more in depth reads, and/or alternativeanalytical methods. Third, there may have been a clear corre-lation between dysbiotic microbiota and the specific phase ofliver disease (steatohepatitis, liver fibrosis, cirrhosis), whichwould have been missed by our current analysis, because of thelack of liver biopsies to identify different phases of liverdisease and relatively small number of cases expected for eachliver disease phase in this study. Finally, we acknowledge thatour method of classifying ALD is based on clinical/biochem-ical and radiological criteria and not based on the histology.Therefore, despite all the measures that we have taken toensure accurate classification of subjects using validated in-struments and definitions, it is certainly possible that somesubjects that are denoted as merely having alcoholism could

have cryptic presence of histological fibrosis without anyclinical or biochemical abnormalities. However, this confound-ing factor cannot be eliminated fully, because routine liverbiopsy in alcoholics is not the standard of care and in mostalcoholic cases liver biopsy is not clinically indicated and thuswas not ethical to perform.

Other limitations of our study included the use mucosalbiopsy samples; in this first study, we hypothesized that mu-cosal bacterial communities would be the most relevant tostudy due to their spatial proximity to the epithelial cells in thegut mucosa. However, future studies could also consider theuse of fecal specimens, where a significant portion of alcoholmetabolism may also be taking place. Furthermore, it is im-portant to note that none of our subjects were actively drinkingthe day of sample procurement, because we needed to obtainvalid informed consents. We did not measure alcohol levelsbefore sample acquisition, and alcohol intake was based onself-report of drinking. However, there was no incentive for thesubjects to intentionally alter their reporting of alcohol intakesince we were recruiting both healthy subjects and alcoholics.

A recent study of the fecal microbiome of subjects withcirrhosis demonstrated similar findings to our study, showing areduction in the Bacteroidetes and an increase in Proteobacteria(especially Gammaproteobacteria class) compared with healthy con-trols (7). Additionally in this study, Fusobacteria were alsoenriched in the cirrhotic group. The etiology of cirrhosis wasmostly hepatitis B related, although a limited number ofcirrhotics with alcohol-related cirrhosis were also included. Inthis study, alcoholic cirrhotics had more Prevotellaceae at thefamily level (7). However, alcohol consumption of all thesubjects was not reported. Therefore, it is unclear whetherthe effects observed are due to cirrhosis or alcoholism. Fur-thermore, a subset of cirrhotic patients did overlap with con-trols similarly, suggesting not all cirrhotics are dysbiotic. Thesefindings in conjunction with ours suggest that there are bothphyla-level and family-level differences in subjects with alco-holism.

Dysbiosis of the gut microbiome that results in more proin-flammatory/pathogenic bacteria could be clinically relevant inalcoholics in general and in patients with alcoholic liver dis-ease because: 1) gut-derived bacteremia and sepsis are com-mon in alcoholics and in particular among those with ALD,2) gut leakiness and consequently increased translocation of

Fig. 15. Rarefaction curve using chao1 index by study group. Curves suggestno differences in -diversity.

Fig. 13. Cumulative distribution function (CDF) plot of subject classes fromnetwork analysis. Each taxa is represented as a node in complex graph, and aconnection is made between 2 nodes if they are present in the same class andabove a defined threshold. We then compared network topologies. We presentthe CDF of the degree distributions per node (taxa) for the 3 defined categories.

Fig. 14. Canonical correspondence analysis (CCA) using endotoxin as theenvironmental variable and bacterial taxa at the class level. HC are depicted asblack squares. ALC are depicted as gray upward triangles. ALD are depictedas open downward triangles.



the microbiota and bacterial products from the gut lumen to thecirculation has been well described in alcoholics (31) and in ananimal model of alcoholic steatohepatitis (25), and 3) manip-ulation of the gut microbiome in an effort to increase theintestinal content of lactic acid-type bacteria at the expense ofother potentially more pathogenic species may ameliorate liverdysfunction in cirrhotics (9, 24). Because intestinal microbiotacan affect intestinal epithelial cell function and intestinal bar-rier integrity (17), it is certainly plausible that mucosa-associ-ated dysbiotic microbiota can contribute to gut leakiness inalcoholics and in particular in sober alcoholics where there areno longer direct toxic effects of alcohol on epithelial cells.Therefore, it is not surprising that many investigators haveproposed that the gut microbiome, in addition to its role in thepathogenesis of overt infective episodes and sepsis, can alsocontribute to the proinflammatory state of cirrhosis even in theabsence of infection (32).

Changes in microbial function, rather than abundance, mayalso lead to increased levels of gut-derived proinflammatoryfactors such as endotoxin. For example, there may be corre-lated sets of bacterial groups that differ taxonomically but arefunctionally equivalent, and such a set may differ from indi-vidual to individual. One can infer that the network analysisessentially reflects the biological relevance of such patterns inbacterial groups (or patterns of metabolic capacity) that coexisttogether in each particular ecological environment (i.e., diseaseclass). Network analysis discriminates patterns of co-occurringbacterial groups, not just simple microbiome composition (di-versity and relative abundance). In our study, the networkanalysis of the disease subclasses indicates that healthy sub-jects have the highest connectivity, suggesting that in thehealthy state, the microbiome probably has the highest meta-bolic capacity to adapt to changing environmental conditions.In contrast, the active alcoholics had the lower connectivity,suggesting that the metabolic robustness of the microbiome inthe alcoholic state has been reduced and focused on copingwith the environmental perturbation of alcohol. Hence, ournetwork analysis could be considered a surrogate for studyingglobal metabolic pathways based on microbiome compositionand indicates that alcoholism can lead to long-term changes inthe connectivity of the mucosa-associated bacteria and thatpresence or absence of alcohol in the colonic environment maybe an important factor affecting the stability and metaboliccapacity of the colonic microbiome. Nevertheless, it should benoted that our study did not directly examine the functionalityof the bacterial components, and there probably are functionalchanges not reflected in the bacterial taxa composition. In fact,our CCA shows a correlation between endotoxemia and gutmicrobiome composition and suggests that the dysbiotic mi-crobiome could contribute to endotoxemia in alcoholics. Fur-ther studies using direct measurement of microbiota functionsuch as metagenomic, transcriptomic, and metabolomic assays(i.e., the metabiome) are needed to determine whether changesin bacterial function rather than composition are better indifferentiating alcoholics from healthy subjects and alcoholicswith and without liver disease, including those with apparentlynormal microbiome composition. Finally, from an ecologicalpoint of view, this study examines the taxa, which make up thebulk of the microbiome with the assumption that the bulkshould contribute to the most of the metabolic capacity,whereas the effects of those minor taxa may not contribute

much to the overall metabolic activity of the microbiome.However, minor taxa may in fact have a significant impact onthe immune system through classic amplification cascades.Thus it is not surprising that not all alcoholics have an alteredmicrobiome composition.

Although associations found in our dataset may be importantfor alcoholism and its complications such as ALD, and areexpected to open up new avenues for research, they are notnecessarily causal. In fact, cross-sectional studies cannot es-tablish causality but are essential to design more exhaustivelongitudinal or interventional human studies to demonstratesuch causality. Our study now provides a scientific rationalefor investing in such studies. Furthermore, our study can openup the opportunity for animal studies that can directly evaluatecausal relationships between alcohol-induced changes in mi-crobiota and tissue injury. Indeed, we recently showed thatalcohol-fed rats that developed gut leakiness, endotoxemia, andsteatohepatitis after 8–10 wk of daily alcohol consumptionalso had dysbiotic colonic microbiomes (25). This animalmodel now provides an opportunity to determine any potentialcausal role for dysbiosis in alcohol-induced endotoxemia, gutleakiness, and steatohepatitis. Furthermore, use of germ-freeanimals with inoculation with single or groups of bacteria canfurther be utilized to determine the role of bacteria in alcohol-induced organ dysfunction.

In conclusion, chronic alcohol use is associated with changes inthe mucosa-associated colonic bacterial composition in a subset ofalcoholics and thus may contribute to the pathogenesis of com-plications of alcoholism. Future studies are required to confirmthese findings and to determine the biological, functional, andclinical significance of shifts in the microbiome composition andconnectivity in alcoholism.

ACKNOWLEDGMENTS

We thank Megan Bakaitis, Nancy Licciardi, and Erica Morset and Drs.Anezi Bakken and Ashkan Farhadi for assistance with subject recruitment. Wealso thank the Unifrac and FastUnifrac team for assistance in working with theonline system.

GRANTS

The study was supported by National Institutes of Health Grants RO-1AA-013745 (to A. Keshavarzian); R 21 DK-071838 (to E. Mutlu); R21AT-001628 (to E. Mutlu); SBIR 1R43DK-074275 (to P. M. Gillevet), andNIAAA 1RC2AA-019405–01 (to A. Keshavarzian and P. M. Gillevet) and agift from Mrs. and Mr. Larry Field.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

Author contributions: E.A.M., P.M.G., M.S., P.A.E., C.K.L., and A.K.performed experiments; E.A.M., P.M.G., H.R., M.S., A.N., M.K., and C.K.L.analyzed data; E.A.M., P.M.G., and A.K. interpreted results of experiments;E.A.M., P.M.G., H.R., and A.N. prepared figures; E.A.M., P.M.G., and A.K.drafted manuscript; E.A.M., P.M.G., H.R., M.S., and A.K. edited and revisedmanuscript; E.A.M., P.M.G., H.R., M.S., and A.K. approved final version ofmanuscript; A.K. conception and design of research.

REFERENCES

1. Adachi Y, Moore LE, Bradford BU, Gao W, Thurman RG. Antibioticsprevent liver injury in rats following long-term exposure to ethanol.Gastroenterology 108: 218–224, 1995.



2. Ball SA, Tennen H, Poling JC, Kranzler HR, Rounsaville BJ. Person-ality, temperament, and character dimensions and the DSM-IV personalitydisorders in substance abusers. J Abnorm Psychol 106: 545–553, 1997.

3. Bigatello LM, Broitman SA, Fattori L, Di Paoli M, Pontello M,Bevilacqua G, Nespoli A. Endotoxemia, encephalopathy, and mortality incirrhotic patients. Am J Gastroenterol 82: 11–15, 1987.

4. Binladen J, Gilbert MT, Bollback JP, Panitz F, Bendixen C, Nielsen R,Willerslev E. The use of coded PCR primers enables high-throughputsequencing of multiple homolog amplification products by 454 parallelsequencing. PloS One 2: e197, 2007.

5. Caporaso JG, Bittinger K, Bushman FD, DeSantis TZ, Andersen GL,Knight R. PyNAST: a flexible tool for aligning sequences to a templatealignment. Bioinformatics (Oxford, England) 26: 266–267, 2010.

6. Caporaso JG, Kuczynski J, Stombaugh J, Bittinger K, Bushman FD,Costello EK, Fierer N, Pena AG, Goodrich JK, Gordon JI, HuttleyGA, Kelley ST, Knights D, Koenig JE, Ley RE, Lozupone CA,McDonald D, Muegge BD, Pirrung M, Reeder J, Sevinsky JR, Turn-baugh PJ, Walters WA, Widmann J, Yatsunenko T, Zaneveld J,Knight R. QIIME allows analysis of high-throughput community se-quencing data. Nature Methods 7: 335–336, 2010.

7. Chen Y, Yang F, Lu H, Wang B, Chen Y, Lei D, Wang Y, Zhu B, LiL. Characterization of fecal microbial communities in patients with livercirrhosis. Hepatology 54: 562–572, 2011.

8. Eckburg PB, Bik EM, Bernstein CN, Purdom E, Dethlefsen L, Sar-gent M, Gill SR, Nelson KE, Relman DA. Diversity of the humanintestinal microbial flora. Science 308: 1635–1638, 2005.

9. Forsyth CB, Farhadi A, Jakate SM, Tang Y, Shaikh M, KeshavarzianA. Lactobacillus GG treatment ameliorates alcohol-induced intestinaloxidative stress, gut leakiness, and liver injury in a rat model of alcoholicsteatohepatitis. Alcohol 43: 163–172, 2009.

10. Foster JA, Krone SM, Forney LJ. Application of ecological networktheory to the human microbiome. Interdiscip Perspect Infect Dis 2008:839501, 2008.

11. Gillevet P, Sikaroodi M, Keshavarzian A, Mutlu EA. Quantitativeassessment of the human gut microbiome using multitag pyrosequencing.Chem Biodivers 7: 1065–1075, 2010.

12. Gillevet PM. Compositions and Methods for Diagnosing Colon Disor-ders, PCT/US05/39887. PCT/USPTO: George Mason University, 2005.

13. Gillevet PM. Multitag Sequencing and Ecogenomic Analysis, EPO07871488.8, PCT/US2007/084840. WPO: BioSpherex, 2006.

14. Grant BF, Dufour MC, Harford TC. Epidemiology of alcoholic liverdisease. Semin Liver Dis 8: 12–25, 1988.

15. Hamady M, Knight R. Microbial community profiling for human micro-biome projects: tools, techniques, and challenges. Genome research 19:1141–1152, 2009.

16. Hamady M, Lozupone C, Knight R. Fast UniFrac: facilitating high-throughput phylogenetic analyses of microbial communities includinganalysis of pyrosequencing and PhyloChip data. ISME J 4: 17–27, 2010.

17. Hooper LV, Wong MH, Thelin A, Hansson L, Falk PG, Gordon JI.Molecular analysis of commensal host-microbial relationships in theintestine. Science 291: 881–884, 2001.

18. Hunt NC, Goldin RD. Nitric oxide production by monocytes in alcoholicliver disease. J Hepatol 14: 146–150, 1992.

19. Keshavarzian A, Choudhary S, Holmes EW, Yong S, Banan A, JakateS, Fields JZ. Preventing gut leakiness by oats supplementation amelio-rates alcohol-induced liver damage in rats. J Pharmacol Exp Ther 299:442–448, 2001.

20. Komanduri S, Gillevet PM, Sikaroodi M, Mutlu E, Keshavarzian A.Dysbiosis in pouchitis: evidence of unique microfloral patterns in pouchinflammation. Clin Gastroenterol Hepatol 5: 352–360, 2007.

21. Lane D. 16s/23s rRNA sequencing. In: Nucleic Acid Techniques inBacterial Systematics, edited by Goodfellow M. West Sussex, England:John Wiley & Sons, 1991, p. 115–175.

22. Li W, Godzik A. Cd-hit: a fast program for clustering and comparinglarge sets of protein or nucleotide sequences. Bioinformatics (Oxford,England) 22: 1658–1659, 2006.

23. Li W, Jaroszewski L, Godzik A. Clustering of highly homologoussequences to reduce the size of large protein databases. Bioinformatics(Oxford, England) 17: 282–283, 2001.

24. Liu Q, Duan ZP, Ha DK, Bengmark S, Kurtovic J, Riordan SM.Synbiotic modulation of gut flora: effect on minimal hepatic encephalop-athy in patients with cirrhosis. Hepatology 39: 1441–1449, 2004.

25. Mutlu E, Keshavarzian A, Engen P, Forsyth CB, Sikaroodi M, Gil-levet P. Intestinal dysbiosis: a possible mechanism of alcohol-inducedendotoxemia and alcoholic steatohepatitis in rats. Alcohol Clin Exp Res33: 1836–1846, 2009.

26. Nanji AA, Khettry U, Sadrzadeh SM. Lactobacillus feeding reducesendotoxemia and severity of experimental alcoholic liver (disease). ProcSoc Exp Biol Med 205: 243–247, 1994.

27. Nanji AA, Khettry U, Sadrzadeh SM, Yamanaka T. Severity of liverinjury in experimental alcoholic liver disease. Correlation with plasmaendotoxin, prostaglandin E2, leukotriene B4, and thromboxane B2. Am JPathol 142: 367–373, 1993.

28. Naqvi A, Rangwala H, Keshavarzian A, Gillevet P. Network-basedmodeling of the human gut microbiome. Chem Biodivers 7: 1040–1050,2010.

29. O’Connor PG, Schottenfeld RS. Patients with alcohol problems. N EnglJ Med 338: 592–602, 1998.

30. Probert HM, Gibson GR. Bacterial biofilms in the human gastrointesti-nal tract. Current issues in intestinal microbiology 3: 23–27, 2002.

31. Rimola A. Infections in liver disease. Oxford, England: Oxford UniversityPress, 1991.

32. Riordan SM, Skinner N, Nagree A, McCallum H, McIver CJ, Kur-tovic J, Hamilton JA, Bengmark S, Williams R, Visvanathan K.Peripheral blood mononuclear cell expression of toll-like receptors andrelation to cytokine levels in cirrhosis. Hepatology 37: 1154–1164, 2003.

33. Sharara AI, Rockey DC. Gastroesophageal variceal hemorrhage. TheNew England journal of medicine 345: 669–681, 2001.

34. Turnbaugh PJ, Hamady M, Yatsunenko T, Cantarel BL, Duncan A,Ley RE, Sogin ML, Jones WJ, Roe BA, Affourtit JP, Egholm M,Henrissat B, Heath AC, Knight R, Gordon JI. A core gut microbiomein obese and lean twins. Nature 457: 480–484, 2009.

35. van de Wiel A. Diabetes mellitus and alcohol. Diabetes/metabolismresearch and reviews 20: 263–267, 2004.

36. Wang Q, Garrity GM, Tiedje JM, Cole JR. Naive Bayesian classifierfor rapid assignment of rRNA sequences into the new bacterial taxonomy.Applied and environmental microbiology 73: 5261–5267, 2007.

37. White JR, Nagarajan N, Pop M. Statistical methods for detectingdifferentially abundant features in clinical metagenomic samples. PLoScomputational biology 5: e1000352, 2009.

38. Whittaker RH. Evolution and measurement of species diversity. Taxon21: 213–251, 1972.



colonic microbiome is altered in alcoholism

Documents