Metformin Is Associated WithHigher Relative Abundance ofMucin-Degrading Akkermansiamuciniphila and Several Short-Chain Fatty Acid–ProducingMicrobiota in the GutDiabetes Care 2017;40:54–62 | DOI: 10.2337/dc16-1324

OBJECTIVE

Recent studies suggest the beneficial effects of metformin on glucose metabolismmay be microbially mediated. We examined the association of type 2 diabetes,metformin, and gut microbiota in community-dwelling Colombian adults. On thebasis of previous research, we hypothesized that metformin is associated withhigher levels of short-chain fatty acid (SCFA)–producing and mucin-degradingmicrobiota.

RESEARCH DESIGN AND METHODS

Participants were selected from a larger cohort of 459 participants. The presentanalyses focus on the 28 participants diagnosed with diabetesd14 takingmetformind and the 84 participants without diabetes who were matched(3-to-1) to participants with diabetes by sex, age, and BMI. We measured de-mographic information, anthropometry, and blood biochemical parameters andcollected fecal samples from which we performed 16S rRNA gene sequencing toanalyze the composition and structure of the gut microbiota.

RESULTS

We found an association between diabetes and gut microbiota that was modifiedby metformin use. Compared with participants without diabetes, participantswith diabetes taking metformin had higher relative abundance of Akkermansiamuciniphila, a microbiota known for mucin degradation, and several gut micro-biota known for production of SCFAs, including Butyrivibrio, Bifidobacteriumbifidum, Megasphaera, and an operational taxonomic unit of Prevotella. Incontrast, compared with participants without diabetes, participants withdiabetes not taking metformin had higher relative abundance of Clostridiaceae02d06 and a distinct operational taxonomic unit of Prevotella and a lowerabundance of Enterococcus casseliflavus.

CONCLUSIONS

Our results support the hypothesis that metformin shifts gut microbiota compo-sition through the enrichment of mucin-degrading A. muciniphila as well as sev-eral SCFA-producing microbiota. Future studies are needed to determine if theseshifts mediate metformin’s glycemic and anti-inflammatory properties.

1VidariumdNutrition, Health and Wellness Re-search Center, Grupo Empresarial Nutresa, Me-dellin, Colombia2Department of Epidemiology, Johns HopkinsBloomberg School of Public Health, Baltimore,MD3Dinamica I.P.S.dEspecialista en Ayudas Diag-nosticas, Medellin, Colombia4EPS y Medicina Prepagada Suramericana S.A.,Medellin, Colombia

Corresponding author: Juan S. Escobar, jsescobar@serviciosnutresa.com.

Received 20 June 2016 and accepted 27 Septem-ber 2016.

This article contains Supplementary Data onlineat http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc16-1324/-/DC1.

© 2017 by the American Diabetes Association.Readers may use this article as long as the workis properly cited, the use is educational and notfor profit, and thework is notaltered.More infor-mation is available at http://www.diabetesjournals.org/content/license.

Jacobo de la Cuesta-Zuluaga,1

Noel T. Mueller,2

Vanessa Corrales-Agudelo,1

Eliana P. Velasquez-Mejıa,1

Jenny A. Carmona,3 Jose M. Abad,4 and

Juan S. Escobar1

54 Diabetes Care Volume 40, January 2017

EMER

GINGTECHNOLO

GIESANDTH

ERAPEU

TICS

Microbial communities (microbiota) andtheir associated genes (microbiome)constitute the interface of our environsand our cells, and their composition isbelieved to play a deterministic role inhuman health and disease. In particular,the development of type 2 diabetes, adisease rising in prevalence around theglobe, has been linked in nonhuman (1)and human (2–5) studies to imbalancesin microbiota of the intestinal tract(gut). However, the most recent humanstudy on this topic found that the asso-ciation was modulated in a potent-ially beneficial manner by metformintreatment (6).Metformin (1,1-dimethylbiguanide

hydrochloride) is the most frequentmedication used to treat type 2 diabetes(7), and findings from recent studiessuggest it may also prevent cancer (7)and cardiovascular events (8). Metfor-min has pleiotropic effects, yet themajority of mechanistic studies havefocused on changes in liver function(7,9,10). Although metformin certainlyalters hepatic glucose production viaeffects on AMPK, there is growing evi-dence that the genesis of its action is inthe gut (11–15).Metformin is ;50% bioavailable, al-

lowing for near-equal intestinal andplasma exposure, but intestinal accumu-lation of metformin is 300 times that ofthe plasma (16), making the gut the pri-mary reservoir for metformin in humans.Unlike oral administration, intravenousadministration of metformin in humansdoes not improve glycemia (14). More-over, in mice, oral administration of abroad-spectrum antibiotic cocktail withoral metformin abrogates metformin’sglucose-lowering effect (12). Providingyet further evidence that the glucose-lowering effect ofmetforminmayoriginatein the lower bowel, a delayed-release oralmetformin, which targets the ileum,had a similar or greater glucose-loweringeffect than immediate-release or extended-release metformin, despite the delayed-releasemetformin having lower systemicexposure than the other metformin for-mulations (15).Recent studies in animals (11,12,13)

and humans (6,17) provide evidencethat metformin may partially restoregut dysbiosis associated with type 2 di-abetes. In mice fed a high-fat diet, met-formin treatment increased the relativeabundance of Akkermansia muciniphila

(11–13), a mucin-degrading bacteriathat has been shown to reverse meta-bolic disorders (1,12). In humans, partic-ipants with diabetes taking metforminhad similar abundance of Subdoligranu-lum and, to some extent, Akkermansiacompared with control subjects withoutdiabetes, suggesting that metforminmay help ameliorate a type 2 diabetes–associated gut microbiome (6). It has alsobeen shown that people with diabetestaking metformin had a higher relativeabundance of Adlercreutzia (17), andmetagenomic functional analyses dem-onstrated significantly enhancedbutyrateandpropionate production in peoplewithdiabetes using metformin (6). In contrast,people with diabeteswhowere not treat-ed with metformin had a higher abun-dance of Eubacterium and ClostridiaceaeSMB53 (17) and lower levels of short-chain fatty acid (SCFA) producers, suchas Roseburia, Subdoligranulum, and acluster of butyrate-producing Clostri-diales (6). These findings provide evi-dence that gut microbes may contributeto the antidiabetes effects of metforminthrough pathways that include mucindegradation and SCFA production.

In this study we aimed to test the gen-eralizability of previous observations con-cerning the influence ofmetformin on theassociation of type 2 diabetes and gutdysbiosis in a Colombian adult popula-tion. Given the considerable variation inthe microbiota associated with type 2 di-abetes and that the gut microbiota of Co-lombians is different to that of otherpopulations (18), we hypothesized thatthe microbial taxa involved in the type 2diabetes dysbiosis of Colombians are dif-ferent to those observed in Chinese andEuropean populations (2,3) but that theeffect of metformin is similar, i.e., throughenrichment of mucin-degrading and SCFA-producing microbiota.

RESEARCH DESIGN AND METHODS

Study DesignBetween July and November 2014, we en-rolled 459 men and women 18–62 yearsold, with BMI $18.5 kg/m2, living inthe Colombian cities of Medellin, Bogota,Barranquilla, Bucaramanga, and Cali. Allparticipants enrolled in the study were in-sured by the health insurance providerEPS y Medicina Prepagada SuramericanaS.A. (EPS SURA). We excluded pregnantwomen, individuals who consumed anti-biotics or antiparasitics ,3 months prior

to enrollment, and individuals diagnosedwith Alzheimer disease, Parkinson dis-ease, or any other neurodegenerative dis-eases; current or recent cancer (,1 year);and gastrointestinal diseases (Crohn dis-ease, ulcerative colitis, short bowel syn-drome, diverticulosis, or celiac disease).

This study was conducted in accor-dance with the principles of the Decla-ration of Helsinki, as revised in 2008, andhad minimal risk according to theColombianMinistry of Health (Resolution008430 of 1993). All of the partici-pants were thoroughly informed aboutthe study and procedures. Participantswere assured of anonymity and confi-dentiality. Written informed consent wasobtained from all the participants beforebeginning the study. The BioethicsCommittee of Sede de InvestigacionUniversitariadUniversity of Antioquia re-viewed the protocol and the consentforms and approved the procedures de-scribed here (approbation act 14–24–588dated 28 May 2014).

Anthropometric, Clinical, and DietaryEvaluationsWe calculated BMI as weight (kg)/heightsquared (m2) to classify participants aslean (18.5 # BMI , 25.0 kg/m2), over-weight (25.0 # BMI , 30.0 kg/m2), orobese (BMI$30 kg/m2). In addition, val-ues of HDL, LDL, VLDL, total cholesterol,triglycerides, apolipoprotein B, fastingglucose, glycated hemoglobin (HbA1c),fasting insulin, adiponectin, and hs-CRPwere obtained (collection and measure-ment explained in Supplementary Data).Dietary intake was evaluated through24-h dietary recalls (see SupplementaryData).

DNA Extraction and SequenceAnalysisEach participant collected their own fe-cal sample in a hermetically sealed, ster-ile receptacle provided by the researchteam. Samples were immediately refrig-erated in household freezers and broughtto an EPS SURA facility in each city within12 h; receipt of samples occurred exclu-sively in the morning (6 A.M.–12 P.M.). Assuch, stools were collected between theevening of the day before and the morn-ing of the day of sample receipt. Fecalsamples were stored on dry ice andsent to a central laboratory via next-daydelivery. Before DNA extraction, stoolconsistency was evaluated by trained lab-oratory technicians.

care.diabetesjournals.org de la Cuesta-Zuluaga and Associates 55

Total microbial DNA was extractedusing the QIAamp DNA Stool Mini Kit(Qiagen, Hilden, Germany) followingthe manufacturer’s instructions, with aslight modification consisting in a bead-beating stepwith the lysis buffer (20 s at15 Hz using a stainless steel bead with a5-mm diameter). After extraction, wequantified DNA concentration using aNanoDrop spectrophotometer (NyxorBiotech, Paris, France) and sent it to theMicrobial Systems Laboratory, Universityof Michigan Medical School (Ann Arbor,MI). The V4 hypervariable region of the16S rRNA gene was amplified using theF515 (59-CACGGTCGKCGGCGCCATT-39)and R806 (59-GGACTACHVGGGTWTCTAAT-39) primers and sequenced usingthe Illumina MiSeq sequencing platformwith V2 chemistry and the dual-indexsequencing strategy (19). In addition toDNA from fecal samples, we sequencednegative controls (ultrapure water andthe QIAamp DNA Stool Mini Kit’s EB elu-tion buffer), a DNA extraction blank,and a mock community (HM-782D, BEIResources, Manassas, VA) in each instru-ment’s run. Sequences were curatedfollowing the MiSeq standard operatingprocedure implemented by Mothurv.1.36 (20) (see Supplementary Data). Rawsequences were deposited at the NCBI andcan be accessed through the BioProject(accession number PRJNA325931).

Definition of Type 2 Diabetes andSelection of Control SubjectsWe identified 28 participants in our studythat had type 2 diabetes; 26 self-reportedphysician-diagnosed diabetes prior tothe beginning of the study and 2 werediagnosed through laboratory testing(fasting blood glucose $126 mg/dLandHbA1c$6.5%).Of the28participantswith type 2 diabetes, 14 were under met-formin treatment, 14 were not (1 wastreated with insulin alone, 2 with gliben-clamide, and 11were under no pharmaco-logical treatment for type 2 diabetes,including the 2 participants unaware oftheir diabetes status) (SupplementaryTable 1).We matched each participant with di-

abetes with three participants withoutdiabetes in our study based on sex, age(to the closest possible age; maximumdifference between case and controlsubjects 6 years; mean 1.5 years; median1 year), and BMI category (lean, over-weight, or obese). This left us with a total

analytic sample of 112 study participantscomprising 14 with type 2 diabetes usingmetformin (T2D-met+), 14 with type 2 di-abetes not using metformin (T2D-met2),and 84 without diabetes (ND).

Statistical AnalysesAnthropometric and clinical variableswere compared across study groups us-ing ANOVA and t tests after checking forhomoscedasticity and normal distribu-tion of residuals (using Fligner-Killeentests of homogeneity of variances andShapiro-Wilk normality tests). Whennecessary, variables were appropriatelytransformed (natural log for uncon-strained variables or arcsin square rootfor proportions). Sex ratio and stool con-sistency were compared using x2 tests.Statistical analyses were performedwith R v.3.2.2 (21).

Curated DNA sequences ranged from69 to 102,660 sequences per sample(median 28,699). To limit the effects ofuneven sampling, we rarefied the dataset to 4,091 sequences per sample, result-ing in the exclusion of one T2D-met2 par-ticipantwith69 reads. Although rarefactionmay lead tomissing low-abundance data, itis a powerful way to reduce the likelihoodof detecting false positives, especiallyamong those operational taxonomic units(OTUs) with very low abundance.

The gut microbiota structure and com-position was assessed by quantifying andinterpreting similarities based on intra-and intergroup diversity analyses (a andb diversity, respectively). For a diversity,we calculated Good’s coverage and thenumber of OTUs of each sample usingMothur and constructed rarefactioncurves.Wecompared these indices amonggroups of participants using analysis ofsimilarity (ANOSIM) with 1,000 permuta-tions using the Vegan package of R (22).

b Diversity was assessed usingphylogeny-based generalized UniFrac dis-tances (with the a parameter controllingweight on abundant lineages = 0.5) calcu-latedwith theGUniFrac package of R (23).For this, we first reduced the alignmentand the OTU table to one representativesequence per OTU, then obtained a dis-tance matrix from uncorrected pairwisedistances between aligned sequences,and finally constructed a relaxed neighbor-joining phylogenetic tree using Mothurand Clearcut. Comparisons amonggroups of participants were performedusing the adonis function (ANOVA using

distance matrices) of the permutationalmultivariate ANOVA (PERMANOVA) im-plemented in theVegan package of R (22).

We next used linear discriminantanalysis (LDA) effect size (LEfSe) (24) toagnostically identify microbial bio-markers. LEfSe uses the nonparametricfactorial Kruskal-Wallis sum rank test todetect individual OTUs with significantdifferential abundance among groupsof participants, then performs a set ofpairwise tests among groups of partici-pants using the unpaired Wilcoxon ranksum test, and finally uses LDA to esti-mate the effect size of each differen-tially abundant OTU (24). The strengthof LEfSe compared with standard statis-tical approaches is that, in addition toproviding P values, it provides an estima-tion of the magnitude of the associationbetween each OTU and the grouping cat-egories (e.g., metformin, type 2 diabetes)through the LDA score. For stringency,microbial biomarkers in our study wereretained if they had a P , 0.05 and a(log10) LDA score $3, i.e., one order ofmagnitude greater than LEfSe’s default.

Finally, across groups, we tested for dif-ferences in relative abundance of themucin-degrading A. muciniphila andmajor butyrate-producing microbialgenera, including Butyrivibrio, Roseburia,Subdoligranulum, and Faecalibacterium.For this analysis, we pooled all OTUsclassified in each of these phylotypes(4 for A. muciniphila, 11 for Butyrivibrio,4 for Roseburia, 10 for Subdoligranulum,and5 for Faecalibacterium) and tested fordifferences using ANOVA and t tests onarcsin square root transformed relativeabundances. This pooling served toexamine whether differences in relativeabundance of these groups of bacteriaoccurred across all OTUs or only inspecific OTUs.

Results from LEfSe and from the pooledanalysis of phylotypes were corrected formultiple testing using the Bayesian ap-proach implemented in the qvalue pack-age of R (25). Tests were consideredsignificant if they had a P value # 0.05and a q value#0.1.

RESULTS

In Table 1 we present the characteristicsof T2D-met+, T2D-met2, and ND partic-ipants. There were no statistically signif-icant differences (all P values. 0.10) indemographic, anthropometric, or clini-cal parameters between T2D-met+ and

56 Type 2 Diabetes, Metformin, and Gut Microbiota Diabetes Care Volume 40, January 2017

Table

1—Generalch

aracte

risticsofT2D-m

et+,

T2D-m

et2,andND

particip

ants

amongco

mmunity

-dwellin

gColombianadults

Group

P*

T2D-m

et+

T2D-m

et2

ND

T2D-m

et +vs.

T2D-m

et 2T2D

-met +

vs.ND

T2D-m

et2vs.

ND

n14

1484

dd

d

Age

(years)50

610

446

947

69

0.110.33

0.24

Sex(F/M

)0.36

0.500.43

0.700.83

0.84

Anthropometry

BMI(kg/m

2)31.88

64.63

32.156

6.3631.11

64.53

0.980.56

0.62Bodyfat

(%)

0.416

0.050.41

60.03

0.406

0.040.96

0.680.58

Waist

circumferen

ce(cm

)104.6

69.8

102.86

14.2102.0

611.3

0.700.39

0.85

Clinicalp

arameters

Totalch

olestero

l(mg/d

L)178

651

2086

44187

630

0.110.54

0.10HDL(m

g/dL)

406

1138

66

446

120.77

0.210.0136

LDL(m

g/dL)

1056

35128

639

1156

280.11

0.310.26

VLD

L(m

g/dL)

386

4052

652

316

150.19

0.640.0471

Apolipoprotein

B(m

g/dL)

1026

30108

630

976

260.56

0.550.18

Triglycerides

(mg/d

L)176

6202

2446

253154

675

0.150.97

0.07Fastin

gglu

cose

(mg/d

L)127

647

1456

7690

610

0.570.0047

0.0060HbA1c[%

(mmol/m

ol)]

6.96

1.4(52.0

615.3)

7.16

1.7(54.0

618.6)

5.66

0.3(38.0

63.3)

0.780.0026

0.0047Fastin

ginsulin

(mU/m

L)22.24

612.58

24.246

12.8615.20

68.79

0.490.11

0.0029Insulin

resistance

index

2.96

1.53.7

63.0

1.96

1.10.39

0.04200.0018

Leptin

(ng/m

L)7.39

66.09

8.346

5.407.89

66.14

0.380.91

0.25Adiponectin

(mg/m

L)4.59

61.97

4.966

3.226.79

63.90

0.910.0195

0.07hs-C

RP(m

g/L)2.70

62.53

3.356

2.764.04

65.02

0.610.21

0.62

Dietary

intake

Energy

intake

(calories)

1,8436

2681,865

6584

1,9456

5720.75

0.610.57

Carb

ohydrate

(g)250

637

2606

86267

688

0.920.52

0.68Pro

tein(g)

74.76

9.169.7

613.1

74.16

13.80.26

0.850.26

Fat(g)

60.76

13.060.2

621.3

62.96

17.10.64

0.760.51

Cholestero

l(mg)

3366

31330

641

3476

380.67

0.280.18

Dietary

fiber

(g)17.9

64.8

18.66

6.417.5

64.8

0.880.73

0.64

Stoolco

nsisten

cy[n

(%)]

0.380.58

0.50Hard

4(28)

2(14)

13(15)

dd

dNorm

al8(57)

9(64)

54(64)

dd

d

Mushy

2(14)

1(7)

13(15)

dd

d

Diarrh

eic0(0)

2(14)

4(5)

dd

d

Valu

espresen

tedas

mean

6SD

.*AllP

values

from

ttests

exceptin

sexandsto

olco

nsisten

cy(x

2tests).

care.diabetesjournals.org de la Cuesta-Zuluaga and Associates 57

T2D-met2 participants. Compared withND participants, T2D-met+ participantshad higher fasting glucose, HbA1c, and in-sulin resistance than ND participants andlower levels of the insulin-sensitizing hor-mone adiponectin (P , 0.05). No otherdemographic, anthropometric, or clinicalparameters were statistically different.

16S rRNA Gene SequencingGut microbiota communities were spe-cific to each participant with marked in-tersubject differences (Fig. 1A) (overallinterindividual generalized UniFrac dis-tance = 0.720 6 0.009). We found highcoverage across all groups of participants(mean Good’s coverage 6 SD = 0.990 60.001); 99% of OTUs were detected atleast by two DNA reads, demonstratingthorough sampling of the gut micro-biota. We next tested for differencesin the number of observed OTUs acrossthe groups of participants. We found nodifferences between participants withdiabetes and ND participants (ANOSIMstatistic R = 0.005, P value = 0.425) orbetween T2D-met+ and T2D-met2 partic-ipants (ANOSIM statistic R = 20.018, P =0.557). The number of observed OTUstended to be more similar between T2D-met+ and ND than between T2D-met2

and ND participants (Supplementary Fig.1); however, these differences were notstatistically significant (T2D-met+ vs. ND:ANOSIM statistic R = 0.018, P = 0.348;T2D-met2 vs. ND: ANOSIM statistic R =0.009, P = 0.409).We observed no significant differences

in b diversity estimates among the threegroups of participants (PERMANOVA: R2 =0.019, P = 0.335) (Fig. 1A) or betweenparticipants with diabetes and ND partic-ipants (R2 = 0.009, P = 0.416) (Fig. 1B).However, the comparison between met-formin and nonmetformin users reachedsignificance (R2 = 0.013, P = 0.036) (Fig.1C), demonstrating differences in thebacterial community structure associatedwith metformin use. The difference wasalso significant when comparing T2D-met+ and ND participants (R2 = 0.015,P = 0.036) but not when comparingT2D-met2 and ND participants (R2 =0.008, P = 0.943). These results suggestedthe microbial communities of T2D-met+

versus T2D-met2 were modestly phylo-genetically dissimilar.We next used LEfSe to examine differ-

ences in the relative abundance of gutmicrobiota at the OTU level. Note that

we were only interested in OTUs dis-playing strong associations in the LDA(represented by OTUs with [log10] LDA

scores $3); such stringency resulted infewer retained, butmore likely biologicallyrelevant, OTUs (19 displayed in Fig. 2 and

PC

o2 (6

.03%

)PCo1 (10.34%)

ND

R2 = 0.019P = 0.34

T2D-met+

T2D-met-

PC

o2 (6

.03%

)R2 = 0.009

P = 0.42

ND

T2D

PCo1 (10.34%)

PC

o2 (6

.03%

)

met-

R2 = 0.013P = 0.036

T2D-met+

PCo1 (10.34%)

A

B

C

Figure 1—Principal coordinates analysis based on generalized UniFrac. A: Comparison among thethree groups of participants. B: Comparison between participants with diabetes and ND partici-pants. C: Comparison between T2D-met+ and participants not taking metformin (T2D-met2 andND). Ellipses encompass 75%of data distribution in each group of participants. R2 and P values fromPERMANOVA.

58 Type 2 Diabetes, Metformin, and Gut Microbiota Diabetes Care Volume 40, January 2017

Fig. 3 out of 273 statistically significant ifnot taking LDA scores into account).When comparing T2D-met2 and ND par-ticipants, we found that OTUs belongingto Clostridiaceae 02d06 (Firmicutes|Clostridiaceae|OTU074) and Prevotella(Bacteroidetes|Prevotellaceae|OTU236)were overrepresented in T2D-met2 par-ticipants, whereas Enterococcus casse-liflavus (Firmicutes|Enterococcaceae|OTU048) was more abundant in ND par-ticipants (Figs. 2A and 3A and B). Whencomparing T2D-met+ participants to NDparticipants, we found that OTUs ofButyrivibrio (Firmicutes|Lachnospiraceae|OTU062), a different OTU of Prevotella(Bacteroidetes|Prevotellaceae|OTU028),Megasphaera (Firmicutes|Veillonellaceae|OTU069), Bifidobacterium bifidum (Acti-nobacteria|Bifidobacteriaceae|OTU176),

twoOTUsofMollicutesRF39 (Tenericutes|OTU284 and OTU067), and Bulleidiap-1630-c5 (Firmicutes|Erysipelotricha-ceae|OTU077) were more abundant inT2D-met+ than in ND participants. Incontrast, four OTUs of Clostridiales in-cluding Clostridium celatum (Firmicutes|Clostridiaceae|OTU014), ClostridiaceaeSMB53 (Firmicutes|Clostridiaceae|OTU026), Oscillospira (Firmicutes|Ruminococcaceae|OTU024), and Cellu-losibacter alkalithermophilus (Firmicutes|Ruminococcaceae|OTU025) were moreabundant in ND than in T2D-met+ partic-ipants (Figs. 2B and 3C andD). OTUs fromPrevotella (Bacteroidetes|Prevotellaceae|OTU028) andMegasphaera (Firmicutes|Veillonellaceae|OTU069) were enrichedin T2D-met+ compared with T2D-met2

participants, whereas OTUs from

Oscillospira (Firmicutes|Ruminococcaceae|OTU024), Barnesiellaceae (Bacteroidetes|OTU102), and a different OTU of Clostri-diaceae 02d06 (Firmicutes|Clostridiaceae|OTU080) were enriched in T2D-met2

compared with T2D-met+ participants(Figs. 2C and 3E and F).

Finally, when we pooled mucin-degrading and butyrate-producing mi-crobes, we found that A. muciniphilaand Butyrivibrio were more abundant(3.4 and 4.4 times, respectively) in T2D-met+ than in T2D-met2 participants;differences were statistically significantfor A. muciniphila (F1, 109 = 9.46, P =0.003, q value = 0.01) but not for Butyr-ivibrio (F1, 109 = 3.03, P = 0.08, q value =0.21) (Fig. 3G and H). There were no sig-nificant differences in the other groupsof butyrate producers between metfor-min and nonmetformin users (Roseburia:F1, 109 = 1.44, P = 0.23, q value = 0.39;Subdoligranulum: F1, 109 = 0.001, P =0.97, q value = 0.97; Faecalibacterium:F1, 109 = 0.53, P = 0.47, q value = 0.59).There were no significant differences inany of these groups of bacteria betweenparticipants with diabetes and ND partic-ipants (all P . 0.1 and q values .0.2).

CONCLUSIONS

In our community-based sample of Colom-bian adults, we provide evidence consis-tent with previous literature (6,11–13,17)that the association between gut micro-biota and type 2 diabetes is modified bymetformin use. T2D-met+ participants hadhigher relative abundance of purportedlybeneficial mucin-degrading and SCFA-producing bacteria compared with T2D-met2 and ND participants matched onage, sex, and BMI.

Several studies have demonstratedthat type 2 diabetes is associated withgut microbiota composition, but find-ings on the taxa involved have been in-consistent (2–5). Some of the variance inprevious study findings may be ex-plained by confounding factors, includ-ing demographics, body weight, andtreatment with drugs, such as metfor-min. After matching on age, sex, andBMI and stratifying comparisons onmetformin, we found only modest asso-ciations between type 2 diabetes andgut microbiota composition. One OTUrelated to Prevotella was higher amongparticipants with diabetes, whereas an-other OTU related to E. casseliflavuswaslower among participants with diabetes.

Figure 2—LDA scores (log10) of the OTUs displaying differences between pairs of groups ofparticipants. ND vs. T2D-met2 (A), ND vs. T2D-met+ (B), T2D-met+ vs. T2D-met2 (C) participants.

care.diabetesjournals.org de la Cuesta-Zuluaga and Associates 59

Prevotella has been associated withcarbohydrate-baseddiets anddegradationof complex polysaccharides (26), whereasE. casseliflavus is an opportunistic patho-gen that may cause serious infections inimmunosuppressed individuals (27).We found associations between met-

formin use and gut microbiota composi-tion that were largely consistent whetherwe compared T2D-met+ participants toND control subjects or to T2D-met2 par-ticipants, suggesting that metformin mayhave direct microbial effects. Our findingsare congruent with multiple lines of evi-dence indicating the gut-mediated glyce-mic effect of metformin stem fromalterations in the gut microbiota compo-sition. Cabreiro et al. (28) first demon-strated that metformin impacts themetabolism of the microbiota hosted byCaenorhabditis elegans. Three mousestudies have shown thatmetformin treat-ment in mice on high-fat diet shifts themicrobiota composition toward that ofmice fed normal chow by increasingabundance of Akkermansia spp. (11–13).Moreover, a follow-up experiment byShin et al. (12) found that mice fed high-fat diets treated with either culturedA. muciniphila or metformin had similarimprovements in mucin-producing gob-let cells, proinflammatory interleukin-6,and glucose tolerance, suggesting that

A. muciniphila alone may explain thebeneficial effects of metformin.

In human studies, a small nonrandom-ized clinical trial of 12 patientswith type 2diabetes demonstrated that stoppingmetformin treatment for 7 days led toalterations in the gut microbiota andglucagon-like peptide 1 (17). A three-country cross-sectional metagenomicstudy found that metformin use was pos-itively associated with SCFA-producingbacteria (6). Metformin has also beenshown to enhance active and totalglucagon-like peptide 1 (17,29,30), whichis consistent with the hypothesis thatit increases SCFA production throughmodification of the gut microbiota com-position. There is also the possibility thatmetformin alters glucose metabolismthrough effects on bile acid secretion (31).

As we hypothesized, on the basis of theaforementioned nonhuman (11–13) andhuman (6,17) studies, metformin use inour studywas associatedwith greater rel-ative abundance of the mucin-degradingA. muciniphila. Through LEfSe biomarkerdiscovery, we also found metformin waspositively associated with the mucolyticbacterium B. bifidum. The higher abun-dance of the mucin-degrading bacteriaA. muciniphila and B. bifidum in the gutmicrobiota of metformin users suggeststhat metformin’s health benefits may

derive from the strengthening of the in-testinal mucosal barrier. A. muciniphilaplays a crucial role in maintaining the in-tegrity of the mucin layer, thereby reduc-ing translocation of proinflammatorylipopolysaccharides and controlling fatstorage, adipose tissue metabolism, andglucose homeostasis (1,12). Likewise,B. bifidum can grow on gastric mucinas a sole carbon source, and genomeanal-ysis revealed that this bacterium can usehost mucins (32), potentially contributingto gastrointestinal health in the samewayas A. muciniphila.

T2D-met+ participants in our studyalso had higher relative abundance ofsome SCFA-producing bacteria, but notothers (e.g., Roseburia, Subdoligranulum,Faecalibacterium). Those positively as-sociated with metformin use includedB. bifidum, Prevotella (an OTU distinctfrom the OTU enriched in T2D-met2 par-ticipants),Megasphaera (a bacterium re-lated to Megamonas), and Butyrivibrio(although this taxa was no longer signifi-cant after adjusting for multiple compar-isons). These microbiota have beenassociated with production of SCFAs, in-cluding butyrate, propionate, and acetate(26,33–35). SCFAs may be beneficial forhealth. Butyrate, in particular, is one ofthe preferred energy sources of the co-lonic epithelium (36). Recent studies in

0

0.01

0.02

0.03

0.04Re

la�v

e ab

unda

nce

0

0.10

0.20

0.30A B C D E F G H

******

**

*

**

**

****

***

**

****

**

+***

*

*

*

*

*

Figure 3—Relative abundance of the groups of bacteria displaying differences among participants (logLDA.3). Data presented as mean6 SE. Openbars = T2D-met+; gray bars = T2D-met2; black bars = ND. A: OTUs enriched in T2D-met2 compared with ND. B: OTUs enriched in ND compared withT2D-met2. C: OTUs enriched in T2D-met+ compared with ND. D: OTUs enriched in ND compared with T2D-met+. E: OTUs enriched in T2D-met+

compared with T2D-met2. F: OTUs enriched in T2D-met2 comparedwith T2D-met+.G: Eleven pooled OTUs classified as Butyrivibrio enriched in T2D-met+ compared with T2D-met2 and ND. H: Four pooled OTUs classified as A. muciniphila enriched in T2D-met+ compared with T2D-met2 and ND(note the change in scale). +P , 0.1; *P , 0.05; **P , 0.05 and q value ,0.1; ***P , 0.05 and q value ,0.05.

60 Type 2 Diabetes, Metformin, and Gut Microbiota Diabetes Care Volume 40, January 2017

mice showed that an increase in the co-lonic production of SCFAs, especially bu-tyrate and propionate, triggers intestinalgluconeogenesis, benefiting glucose andenergy homeostasis and reducing hepaticglucose production, appetite, and bodyweight (37). Acetate produced by bifido-bacteria improves the intestinal defensemediated by epithelial cells and protectsthehost against lethal infection (34). Also,SCFAs, particularly butyrate, stimulateepithelial metabolism and deplete intra-cellular O2, resulting in stabilization of thetranscription factor HIF-1 and increasingepithelial barrier function (38). In humans,low levels of butyrate-producing bacteriahave been associatedwith colonic disease(e.g., inflammatory bowel disease), high-lighting the role of SCFAs in disease resis-tance (39,40).Our study is not without limitations.

Our findings are based on observationaldata that cannot provide causal infer-ence. We were able to reduce the po-tential for confounding by matching onage, sex, andBMI. Becauseourmetformin–gut microbiota associations werelargely consistent whether we usedas a reference group the T2D-met2 orND participants, we believe our findingswere not due to confounding by indi-cation. However, we cannot rule outunmeasured or residual confounding. An-other limitation to our study was the lackof information on dose and duration ofmetformin treatment. This limitationcould have resulted in a weaker, moreconservative, association (attenuation to-ward the null) between metformin andgut microbiota composition and struc-ture. Future studies are warranted todetermine the dose-response of themetformin-microbiota relationship. Wealso had a small sample size relative to aprevious observational study on this topic(Forslund et al. [6] analyzed 784 gutmetagenomes of Danish, Swedish, andChinese participants, of which, 199 hadtype 2 diabetes) and thusmay have beenunderpowered to detect statistical sig-nificance for measures of a diversityand associations of microbial composi-tion that were smaller inmagnitude. Nev-ertheless, our ability to largely confirmhypotheses generated from previousstudies with our modest sample size andin a different population suggests that theeffect ofmetformin on the gutmicrobiotais robust and replicable across diversepopulations.

In conclusion, our study of Colombianadults provides evidence congruent withthe hypothesis that metformin has directeffects on gut microbiota compositionthroughaugmentationofmucin-degradingA. muciniphila as well as several SCFA-producingbacteria.Randomizedcontrolledtrials areneeded todeterminewhether theantidiabetes and anti-inflammatory effectsofmetformin aremediated by the changesto gut microbiota composition.

Acknowledgments. Foremost, the authors areindebted to all the participants who agreed to takepart in this study. The authors also thank NataliaZuluaga (of Vidarium) for preselection of potentialparticipants; Luz G. Betancur and Natalia E. Guarin(of Vidarium) for help in participant recruitment;ErikaM. Loaiza,Natalia Pareja,D. TatianaGarcia, andYuliana M. Franco (of Vidarium) for their invalu-able help during field work; Amalia Toro, ConnieJ. Arboleda, and Ana C. Ochoa (of EPS SURA) fortheir help authorizing and coordinating activitieswithin EPS SURA; and the administrative andlaboratory staff of EPS SURA and Dinamica I.P.S.The authors also acknowledge Paola A. Rios andGiovanni Torres from the Instituto Colombianode Medicina Tropical for their help with samplehandling and pretreatment and APOLO Scien-tific Computing Center of EAFIT University forhosting the bioinformatics resources for thestudy.Funding. This work was funded by Grupo Em-presarial Nutresa, EPS SURA, and Dinamica I.P.S.The funders havenot had any role in designing

or conducting the study; in the collection, man-agement, analysis, or interpretation of the data;in the preparation, review, or approval of themanuscript; or in the decision to submit themanuscript for publication.Duality of Interest. J.d.l.C.-Z., V.C.-A., E.P.V.-M.,and J.S.E. are employees of Grupo EmpresarialNutresa. J.A.C. is an employee of Dinamica I.P.S.J.M.A. is an employee of EPS SURA. No otherpotential conflicts of interest relevant to this articlewere reported.Author Contributions. J.d.l.C.-Z. processedfecal samples and DNA sequences, performedanalyses, and wrote the manuscript. N.T.M.conceived the study, put forward hypothesesto be tested, and wrote the manuscript. V.C.-A.designed the cohort study, recruited partici-pants, coordinatedfieldactivities, collectedfecaland blood samples, and measured anthropo-metric variables. E.P.V.-M. processed fecal sam-ples andDNAsequences. J.A.C. coordinatedfieldactivities and transport and treatment of sam-ples. J.M.A. coordinated participant recruitmentand field activities. J.S.E. designed the cohortstudy, coordinated participant recruitment, su-pervised field activities and transport and treat-mentof samples, performedanalyses, andwrotethe manuscript. All authors read and approvedthe final version of the manuscript. J.S.E. is theguarantor of this work and, as such, had fullaccess to all the data in the study and takesresponsibility for the integrityof thedataand theaccuracy of the data analysis.

Prior Presentation. Parts of this study werepresented at the XXIII Latin American Congressof Microbiology, Rosario, Argentina, 26–30September 2016.

References1. Everard A, Belzer C, Geurts L, et al. Cross-talkbetween Akkermansia muciniphila and intesti-nal epithelium controls diet-induced obesity.Proc Natl Acad Sci U S A 2013;110:9066–90712. Qin J, Li Y, Cai Z, et al. A metagenome-wideassociation study of gut microbiota in type 2diabetes. Nature 2012;490:55–603. Karlsson FH, Tremaroli V, Nookaew I, et al.Gut metagenome in European women with nor-mal, impaired and diabetic glucose control. Na-ture 2013;498:99–1034. Larsen N, Vogensen FK, van den Berg FWJ,et al. Gut microbiota in human adults withtype 2 diabetes differs from non-diabetic adults.PLoS One 2010;5:e90855. Zhang X, Shen D, Fang Z, et al. Human gutmicrobiota changes reveal the progression ofglucose intolerance. PLoS One 2013;8:e711086. Forslund K, Hildebrand F, Nielsen T, et al.;MetaHIT consortium. Disentangling type 2 dia-betes and metformin treatment signatures inthe human gut microbiota. Nature 2015;528:262–2667. Pernicova I, Korbonits M. Metformin–modeof action and clinical implications for diabetesand cancer. Nat Rev Endocrinol 2014;10:143–1568. Lamanna C, Monami M, Marchionni N,Mannucci E. Effect of metformin on cardiovas-cular events and mortality: a meta-analysis ofrandomized clinical trials. Diabetes Obes Metab2011;13:221–2289. Pryor R, Cabreiro F. Repurposing metformin:an old drug with new tricks in its binding pock-ets. Biochem J 2015;471:307–32210. Rena G, Pearson ER, Sakamoto K.Molecularmechanism of action of metformin: old or newinsights? Diabetologia 2013;56:1898–190611. Lee H, Ko G. Effect of metformin on meta-bolic improvement and gut microbiota. Appl En-viron Microbiol 2014;80:5935–594312. Shin N-RR, Lee J-CC, Lee H-YY, et al. An in-crease in the Akkermansia spp. population in-duced by metformin treatment improvesglucose homeostasis in diet-induced obesemice. Gut 2014;63:727–73513. Zhang X, Zhao Y, Xu J, et al. Modulation ofgut microbiota by berberine and metforminduring the treatment of high-fat diet-inducedobesity in rats. Sci Rep 2015;5:1440514. Bonora E, CigoliniM, Bosello O, et al. Lack ofeffect of intravenous metformin on plasma con-centrations of glucose, insulin, C-peptide, gluca-gon and growth hormone in non-diabeticsubjects. Curr Med Res Opin 1984;9:47–5115. Buse JB, DeFronzo RA, Rosenstock J, et al.The primary glucose-lowering effect of metfor-min resides in the gut, not the circulation: Re-sults from short-term pharmacokinetic and12-week dose-ranging studies. Diabetes Care2016;39:198–20516. Bailey CJ, Wilcock C, Scarpello JH. Metfor-min and the intestine. Diabetologia 2008;51:1552–155317. Napolitano A, Miller S, Nicholls AW, et al.Novel gut-based pharmacology of metformin in

care.diabetesjournals.org de la Cuesta-Zuluaga and Associates 61

patients with type 2 diabetesmellitus. PLoS One2014;9:e10077818. Escobar JS, Klotz B, Valdes BE, Agudelo GM.The gut microbiota of Colombians differs fromthat of Americans, Europeans and Asians. BMCMicrobiol 2014;14:31119. KozichJJ,WestcottSL,BaxterNT,Highlander SK,Schloss PD. Development of a dual-index se-quencing strategy and curation pipeline for an-alyzing amplicon sequence data on the MiSeqIllumina sequencing platform. Appl Environ Mi-crobiol 2013;79:5112–512020. Schloss PD, Westcott SL, Ryabin T, et al.Introducing mothur: open-source, platform-independent, community-supported softwarefor describing and comparing microbial com-munities. Appl Environ Microbiol 2009;75:7537–754121. R Development Core Team. R: A Languageand Environment for Statistical Computing.Vienna, Austria, R Foundation for StatisticalComputing, 201222. Oksanen J, Blanchet FG, Kindt R, et al.Vegan: Community Ecology Package. R packageversion 2.3-1, R Foundation for Statistical Com-puting, 2015. p. 264.23. Chen J, Bittinger K, Charlson ES, et al. Asso-ciating microbiome composition with environ-mental covariates using generalized UniFracdistances. Bioinformatics 2012;28:2106–211324. Segata N, Izard J, Waldron L, et al. Metage-nomic biomarker discovery and explanation.Genome Biol 2011;12:R60

25. Storey JD, Bass AJ, Dabney A, RobinsonD. qvalue: Q-value estimation for false discoveryrate control. R package version 2.2.2 [Internet],2015. Available from http://github.com/jdstorey/qvalue. Accessed 19 January 201626. Wu GD, Chen J, Hoffmann C, et al. Linkinglong-term dietary patterns with gut microbialenterotypes. Science 2011;334:105–10827. Murray BE. Vancomycin-resistant enterococ-cal infections. N Engl J Med 2000;342:710–72128. Cabreiro F, Au C, Leung KY, et al. Metforminretards aging in C. elegans by altering microbialfolate and methionine metabolism. Cell 2013;153:228–23929. Maida A, Lamont BJ, Cao X, Drucker DJ.Metformin regulates the incretin receptor axisvia a pathway dependent on peroxisome prolif-erator-activated receptor-a in mice. Diabetolo-gia 2011;54:339–34930. Mannucci E, Ognibene A, Cremasco F, et al.Effect of metformin on glucagon-like peptide1 (GLP-1) and leptin levels in obese nondiabeticsubjects. Diabetes Care 2001;24:489–49431. Wahlstrom A, Sayin SI, Marschall H-U,Backhed F. Intestinal crosstalk between bileacids and microbiota and its impact on host me-tabolism. Cell Metab 2016;24:41–5032. Turroni F, Bottacini F, Foroni E, et al. GenomeanalysisofBifidobacteriumbifidumPRL2010 revealsmetabolic pathways for host-derived glycan forag-ing. ProcNatlAcadSciUSA2010;107:19514–1951933. Sakon H, Nagai F, Morotomi M, Tanaka R.Sutterella parvirubra sp. nov. and Megamonas

funiformis sp. nov., isolated from human faeces.Int J Syst Evol Microbiol 2008;58:970–97534. Fukuda S, Toh H, Hase K, et al. Bifidobacte-ria can protect from enteropathogenic infectionthrough production of acetate. Nature 2011;469:543–54735. Gargari G, Taverniti V, Balzaretti S, et al.Consumption of a Bifidobacterium bifidumstrain for 4 weeks modulates dominant intesti-nal bacterial taxa and fecal butyrate in healthyadults. Appl Environ Microbiol 2016;82:5850–585936. Hamer HM, Jonkers D, Venema K, VanhoutvinS, Troost FJ, Brummer RJ. Review article: the roleof butyrate on colonic function. Aliment Pharma-col Ther 2008;27:104–11937. De Vadder F, Kovatcheva-Datchary P,Goncalves D, et al. Microbiota-generated me-tabolites promote metabolic benefits via gut-brain neural circuits. Cell 2014;156:84–9638. Kelly CJ, Zheng L, Campbell EL, et al. Cross-talk between microbiota-derived short-chainfatty acids and intestinal epithelial HIF aug-ments tissue barrier function. Cell Host Microbe2015;17:662–67139. Machiels K, Joossens M, Sabino J, et al. Adecrease of the butyrate-producing speciesRoseburia hominis and Faecalibacterium praus-nitzii defines dysbiosis in patients with ulcera-tive colitis. Gut 2014;63:1275–128340. Sokol H, Seksik P, Furet JP, et al. Low countsof Faecalibacterium prausnitzii in colitis micro-biota. Inflamm Bowel Dis 2009;15:1183–1189

62 Type 2 Diabetes, Metformin, and Gut Microbiota Diabetes Care Volume 40, January 2017