Drinking Water Source and Intake Are Associated with Distinct Gut Microbiota Signatures in US and UK Populations

Abstract

Background:
The microbiome of the digestive tract exerts fundamental roles in host physiology. Extrinsic factors including lifestyle and diet are widely recognized as key drivers of gut and oral microbiome compositions. Although drinking water is among the food items consumed in the largest amount, little is known about its potential impact on the microbiome.

Objectives:
We explored the associations of plain drinking water source and intake with gut and oral microbiota compositions in a population-based cohort.

Methods:
Microbiota, health, lifestyle, and food intake data were extracted from the American Gut Project public database. Associations of drinking water source (bottled, tap, filtered, or well water) and intake with global microbiota composition were evaluated using linear and logistic models adjusted for anthropometric, diet, and lifestyle factors in 3413 and 3794 individuals, respectively (fecal samples; 56% female, median [IQR] age: 48 [36-59] y; median [IQR] BMI: 23.3 [20.9-26.3] kg/m2), and in 283 and 309 individuals, respectively (oral samples).

Results:
Drinking water source ranked among the key contributing factors explaining the gut microbiota variation, accounting for 13% [Faith's phylogenetic diversity (Faith's PD)] and 47% (Bray-Curtis dissimilarity) of the age effect size. Drinking water source was associated with differences in gut microbiota signatures, as revealed by β diversity analyses (P < 0.05; Bray-Curtis dissimilarity, weighted UniFrac distance). Subjects drinking mostly well water had higher fecal α diversity (P < 0.05; Faith's PD, observed amplicon sequence variants), higher Dorea, and lower Bacteroides, Odoribacter, and Streptococcus than the other groups. Low water drinkers also exhibited gut microbiota differences compared with high water drinkers (P < 0.05; Bray-Curtis dissimilarity, unweighted UniFrac distance) and a higher abundance of Campylobacter. No associations were found between oral microbiota composition and drinking water consumption.

Conclusions:
Our results indicate that drinking water may be an important factor in shaping the human gut microbiome and that integrating drinking water source and intake as covariates in future microbiome analyses is warranted.

Full text

The Journal of Nutrition
Nutritional Epidemiology
Drinking Water Source and Intake Are
Associated with Distinct Gut Microbiota
Signatures in US and UK Populations
Tiphaine Vanhaecke,1Oriane Bretin,1Marion Poirel,2andJulienTap
1
1Danone Research, Palaiseau, France; and 2IT&M Innovation on behalf of Danone Research, Neuilly-sur-Seine, France
ABSTRACT
Background: The microbiome of the digestive tract exerts fundamental roles in host physiology. Extrinsic factors
including lifestyle and diet are widely recognized as key drivers of gut and oral microbiome compositions. Although
drinking water is among the food items consumed in the largest amount, little is known about its potential impact on
the microbiome.
Objectives: We explored the associations of plain drinking water source and intake with gut and oral microbiota
compositions in a population-based cohort.
Methods: Microbiota, health, lifestyle, and food intake data were extracted from the American Gut Project public
database. Associations of drinking water source (bottled, tap, ltered, or well water) and intake with global microbiota
composition were evaluated using linear and logistic models adjusted for anthropometric, diet, and lifestyle factors in
3413 and 3794 individuals, respectively (fecal samples; 56% female, median [IQR] age: 48 [36–59] y; median [IQR] BMI:
23.3 [20.9–26.3] kg/m2), and in 283 and 309 individuals, respectively (oral samples).
Results: Drinking water source ranked among the key contributing factors explaining the gut microbiota variation,
accounting for 13% [Faith’s phylogenetic diversity (Faith’s PD)] and 47% (Bray–Curtis dissimilarity) of the age effect size.
Drinking water source was associated with differences in gut microbiota signatures, as revealed by βdiversity analyses
(P<0.05; Bray–Curtis dissimilarity, weighted UniFrac distance). Subjects drinking mostly well water had higher fecal α
diversity (P<0.05; Faith’s PD, observed amplicon sequence variants), higher Dorea,andlowerBacteroides,Odoribacter,
and Streptococcus than the other groups. Low water drinkers also exhibited gut microbiota differences compared
with high water drinkers (P<0.05; Bray–Curtis dissimilarity, unweighted UniFrac distance) and a higher abundance
of Campylobacter. No associations were found between oral microbiota composition and drinking water consumption.
Conclusions: Our results indicate that drinking water may be an important factor in shaping the human gut microbiome
and that integrating drinking water source and intake as covariates in future microbiome analyses is warranted. J Nutr
2022;152:171–182.
Keywords: drinking water, water source, water intake, human microbiome, gut microbiota diversity, oral
microbiota, American Gut Project
Introduction
Environmental exposure and lifestyle factors strongly inuence
the composition of the human microbiome. Both short- and
long-term dietary patterns are among the main modiable
drivers shaping the digestive tract microbiome structure (1,2).
From the oral cavity to the large intestine, microbial commu-
nities are in close interaction with the external environment
and the epithelial barriers that delineate the inner self (3).
These communities exert a marked inuence on the host
during homeostasis and disease episodes through a range of
physiological functions (4).
Whereas the role of dietary patterns on oral and gut
microbial communities has been extensively explored, drinking
water, a vital source of uids to replace daily body water losses
and maintain homeostasis, has rarely been a consideration in
human microbiome research. The US Institute of Medicine
has set a daily Adequate Intake (AI) for total water for the
adult population of 2.7 L for women and 3.7 L for men, of
which 70%–80% comes from plain drinking water and other
beverages (5). In Europe, the daily AI for total water for the
adult population set by the European Food Safety Authority is
2.0 L for women and 2.5 L for men, of which 80% is estimated
to come from uids (6).
C
The Author(s) 2021. Published by Oxford University Press on behalf of the American Society for Nutrition. This is an Open Access article distributed under
the terms of the Creative Commons Attribution-NonCommercial License (https://creativecommons.org/licenses/by-nc/4.0/), which permits non-commercial re-use,
distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact journals.permissions@oup.com
Manuscript received June 16, 2021. Initial review completed July 28, 2021. Revision accepted August 25, 2021.
First published online October 12, 2021; doi: https://doi.org/10.1093/jn/nxab312. 171

Drinking water may come from various origins and be
subjected to different treatments before its consumption.
Whereas natural mineral water, spring water, and well water
solely come from groundwater, tap water originates either
from surface water or groundwater, depending on community
water systems’ location and setting. Drinking water may be
subjected to various treatments, depending on the quality of
raw water and applicable regulations. Common treatments
consist of occulation, sedimentation, and ltration to remove
debris and control the chemical and mineral composition.
Some drinking waters may further be disinfected to remove or
control disease-causing microorganisms (7,8). The combination
of environmental conditions underground and at the source,
and along the processing, distribution, and storage, results in
multiple and distinct physicochemical (9), mineral (10), and
microbial signatures (11–14) of drinking waters. Consumed
in large amounts on a daily basis, drinking water may be
considered a potential source of microbial gut diversity (15).
The effect of drinking water on the gut microbiome is
poorly understood. A limited number of studies have explored
microbiota composition after the ingestion of different types of
drinking water and observed that water type led to differences
in gut microbiota compositions (16). Consumption of tap water
resulted in an increased amount of bacteria associated with
antibiotic resistance in the mouse gut microbiota compared with
sterilized water (16). Today, the effect of drinking water on
human gut and oral microbiota is even less explored. During
infant microbiome development, tap water and boiled water
intake correlated with gut microbiota signatures, indicating that
drinking water may be a determinant of microbiome acquisition
(17). In adults, limited evidence suggests that the source of
drinking water consumed is associated with oral (9)orgut
microbiota compositions (18,19). The mechanisms by which
drinking water may interact with microbial communities, if
any, remain to be elucidated but could involve water pH (20,
21), solute and mineral composition (9,16,18,19), natural
intrinsic microbial communities (16,18), or residual chlorine
and disinfection byproducts that remain in most tap waters (22,
23).
To our knowledge, most population-based gut microbiota
cohorts do not consider plain drinking water as part of the diet
(24,25). Despite its primary role for health (26), water is under-
researched and often referred to as the forgotten nutrient (27).
A recent citizen-science microbiota-based cohort, the American
Gut Project (AGP), provided a unique opportunity to analyze
multiple environmental and lifestyle factors, including diet and
the consumption of plain drinking water. In this exploratory
analysis, we investigated whether plain drinking water source
and intake were associated with fecal and oral microbiota
compositions as measured by 16S ribosomal RNA (rRNA) gene
sequencing in multivariate-adjusted models.
The authors reported no funding received for this study.
Author disclosures: TV and JT are full-time employees of Danone Research. OB
was an intern at Danone Research at the time of analysis. MP is a consultant at
IT&M Innovation and performs statistical analyses for Danone Research.
Supplemental Methods, Supplemental Figures 1–8, and Supplemental Table 1
are available from the “Supplementary data” link in the online posting of the
article and from the same link in the online table of contents at https://academ
ic.oup.com/jn/.
Address correspondence to TV (e-mail: tiphaine.vanhaecke@danone.com)and
JT (e-mail: julien.tap@danone.com).
Abbreviations used: AGP, American Gut Project; AI, Adequate Intake; ASV, ampli-
con sequence variant; Faith’s PD, Faith’s phylogenetic diversity; PERMANOVA,
permutational multivariate analysis of variance; rRNA, ribosomal RNA.
Methods
Participant recruitment and sample processing
Data for this research originate from the AGP database, a self-selected
citizen-scientist cohort initiated by the University of California San
Diego (28). It contains analyses of fecal, oral, and skin samples sent
by >24,000 volunteers worldwide. After contributing $99, participants
receive a kit to collect their biological samples and mail them for 16S
rRNA gene sequencing to establish their microbial composition. Self-
reported metadata are collected through a web portal (https://micros
etta.ucsd.edu/). Surveys, methodology, sampling, and laboratory test
procedures have been described elsewhere (28).
Ethical committee approval for the collection of AGP data was
obtained either from the University of Colorado Boulder Review Board
(protocol no. 12-0582; December 2012–March 2015) or from the
University of California Review Board, San Diego (protocol no. 141853;
February 2015–present) in accordance with the Declaration of Helsinki
and all participants provided informed consent. Return of results to
participants, public deposition of de-identied data, and subsequent
analyses are allowed by the IRB-approved protocol (28). To investigate
the associations between drinking water consumption and gut and
oral microbiota composition, Redbiom (29) was used to fetch in Qiita
(30) the 100-nt-long sequences associated with 20,454 stool sample
identiers available in the database on 5 December, 2019 using the
Deblur-Illumina-16S-V4-100nt-fbc5b2 context (28).
Participant and sample selection
Participants who completed both the standard AGP questionnaire
(general health status, disease history, and lifestyle data) and a validated
picture-based FFQ (VioScreen™; http://www.viocare.com/vioscreen.ht
ml)(n=3022 participants out of 15,254) were screened to exclude
participants with aberrant total water intake, based on the variable
“Water in g” available in Vioscreen™ (see Supplemental Methods for
more details). When available, coherence of uid variables between the
general questionnaire and Vioscreen™ was evaluated. Fecal samples
were then screened for unusual dominant microbiota (see Supplemental
Methods). This rst set of exclusion criteria resulted in the exclusion
of 2045 and 4564 fecal samples for the drinking water source and
the drinking water intake analysis, respectively (Figure 1,Supplemental
Figure 1). Individuals with missing information on confounding factors
were ltered out (see Supplemental Methods). Participants were further
ltered based on their age, geographical location, and antibiotics
intake to reduce variability (see Supplemental Methods). The resulting
study population consisted of 3413 subjects for the drinking water
source analysis and 3794 subjects for the drinking water intake
analysis.
For oral microbiota analyses, data from 1383 oral samples
coming from 1140 participants were extracted. Supplemental Figure 2
details exclusion criteria. Complete data were available for n=283
participants for the drinking water source analysis and n=309
participants for the water intake analysis.
Numerical ecology and statistical analyses
All statistical analyses were performed using R Statistical Software
version 3.6 (R Core Team). Characteristics of participants were
compared between groups of drinking water source and groups of
low and high water drinkers using ANOVA with post hoc tests for
continuous variables and chi-square tests for categorical variables (31).
All Pvalues reported were adjusted for multiplicity with Benjamini–
Hochberg’s correction procedure.
Microbiota αdiversity
Observed amplicon sequence variants (ASVs), Chao1 (32), Shannon
index (33), and Faith’s phylogenetic diversity (Faith’s PD) (34)were
used as response variables in multivariate-adjusted linear models (see
Supplemental Methods). For all analyses performed in this study,
α=0.05 was considered the signicance threshold for adjusted P
values. In the water intake analysis, participants were separated into
different age categories to reduce α-diversity variability induced by age
172 Vanhaecke et al.

2045 parcipants without
answer to drinking water
source or “Not sure,” with
extreme TWI,1incoherent
answers on alcohol
consumpon,2or with
unusual dominant
microbiota3
15,254 stool samples fetched from Redbiom
13,209 stool samples
3413 stool samples
1331 parcipants without
answer to age or aged ≤3 y,
without answer to sex or
BMI
3589 parcipants who
took anbiocs in the past
year
1373 parcipants without
answer to diet type, plant
diversity, alcohol
frequency, SSB frequency
1180 parcipants not
residing in the USA or
United Kingdom
2323 parcipants without
answer to infant feeding,
exercise frequency, level of
educaon
FIGURE 1 American Gut Project participant selection and data ltering process for gut microbiota analysis according to their drinking water
source. 1Based on age- and sex-specic Adequate Intake as dened by the European Food Safety Authority and TWI expressed as mL/kg.
Participants with either TWI <25% AI or aged <14 y with TWI >300% AI, or TWI >200 mL/kg of body weight were excluded. 2Participants
were excluded if Alcohol_frequency =“Never” AND Alcohol_consumption =“Yes.” 3The top 10 genera list in adults and the top 10 genera list
in nonadults were merged, resulting in 12 genera. Samples were agged as outliers when their respective read mass had <25% of those 12
genera. SSB, sugar-sweetened beverage; TWI, total water intake.
(35). Adjusted means of αdiversity were computed, then compared
between all pairs of groups using pairwise comparisons of estimated
marginal means. Pvalues were corrected with Benjamini–Hochberg’s
procedure (31) to account for multiple testing. To measure variables’
effect sizes, the proportion of variance (R2) captured by a given variable
was calculated as the ratio of the sum of squares explained by the
variable to the total sum of squares in type II ANOVAs when applicable
(see Supplemental Methods).
Drinking water consumption and gut microbiota 173

TAB L E 1 Characteristics of the American Gut Project participants according to their drinking water source1
All Bottled City Filtered Well Pvalue
(n=3413) 2(n=284) (n=1784) (n=1147) (n=198)
Age, y 48 [36–59] 49 [39–62] 47 [35–59] 47 [36–58] 52 [42–61] 0.001
Sex 0.014
Female 1909 (56) 164 (58) 951 (53) 678 (59) 116 (59)
Male 1504 (44) 120 (42) 833 (47) 469 (41) 82 (41)
BMI 0.070
Normal (BMI =18.5–24.9) 2073 (61) 154 (54) 1079 (61) 723 (63) 117 (59)
Underweight (BMI ≤18.4) 351 (10) 32 (11) 200 (11) 97 (9) 22 (11)
Overweight (BMI =25.0–29.9) 824 (24) 79 (28) 432 (24) 265 (23) 48 (24)
Obese (BMI ≥30.0) 165 (5) 19 (7) 73 (4) 62 (5) 11 (6)
Diabetes3,488 (3) 8 (3) 44 (3) 28 (2) 8 (4) 0.392
Cardiovascular disease3,585 (3) 7 (3) 46 (3) 30 (3) 2 (1) 0.786
Kidney disease3,644 (1) 8 (3) 21 (1) 15 (1) 0 (0) 0.020
IBS70.001
Diagnosed by a medical professional 406 (12) 25 (9) 224 (13) 132 (12) 25 (13)
Diagnosed by an alternative medicine
practitioner
33 (1) 3 (1) 12 (1) 16 (1) 2 (1)
Self-diagnosed 213 (6) 13 (5) 119 (7) 73 (6) 8 (4)
IBD3,8105 (3) 8 (3) 54 (3) 39 (3) 4 (2) 0.161
Smoking frequency90.112
Never 3190 (94) 260 (92) 1656 (93) 1089 (95) 185 (93)
Rarely 111 (3) 8 (3) 62 (4) 35 (3) 6 (3)
Occasionally 37 (1) 3 (1) 25 (1) 7 (1) 2 (1)
Regularly 20 (1) 2 (1) 13 (1) 4 (0.4) 1 (1)
Daily 50 (1) 10 (4) 27 (2) 9 (1) 4 (2)
Exercise frequency <0.001
Never 96 (3) 17 (6) 52 (3) 24 (2) 3 (2)
Rarely 377 (11) 44 (16) 201 (11) 115 (10) 17 (9)
Occasionally 832 (24) 77 (27) 434 (24) 270 (24) 51 (26)
Regularly 1393 (41) 104 (37) 749 (42) 467 (41) 73 (37)
Daily 715 (21) 42 (15) 348 (20) 271 (24) 54 (27)
Country <0.001
USA 1887 (55) 195 (69) 792 (44) 740 (65) 160 (81)
United Kingdom 1526 (45) 89 (31) 992 (56) 407 (36) 38 (19)
Collection season 0.008
Spring 887 (26) 59 (21) 454 (25) 320 (28) 54 (27)
Summer 699 (20) 49 (17) 377 (21) 228 (20) 45 (23)
Fall 803 (24) 73 (26) 451 (25) 244 (21) 35 (18)
Winter 1024 (30) 103 (36) 502 (28) 355 (31) 64 (32)
Level of education <0.001
Did not complete high school 96 (2) 10 (4) 44 (3) 35 (3) 7 (4)
High school or GED equivalent 126 (4) 14 (5) 65 (4) 43 (4) 4 (2)
Some college or technical school 335 (10) 44 (16) 150 (8) 120 (11) 21 (11)
Associate’s degree 78 (2) 12 (4) 21 (1) 36 (3) 9 (5)
Bachelor’s degree 906 (27) 74 (26) 455 (26) 317 (28) 60 (30)
Some graduate school or professional 248 (7) 20 (7) 112 (6) 98 (9) 18 (9)
Graduate or Professional degree 1624 (48) 110 (39) 937 (53) 498 (43) 79 (40)
Fed as infant <0.001
Primarily breast milk 1873 (55) 143 (50) 1029 (58) 601 (52) 100 (51)
Primarily infant formula 930 (27) 90 (32) 429 (24) 339 (30) 72 (36)
Both 610 (18) 51 (18) 326 (18) 207 (18) 26 (13)
Diet type 0.007
Omnivore 2681 (79) 231 (81) 1425 (80) 866 (76) 159 (80)
Omnivore but no red meat 238 (7) 20 (7) 114 (6) 93 (8) 11 (6)
Vegetarian 174 (5) 15 (5) 101 (6) 52 (5) 6 (3)
Vegetarian but eat seafood 227 (7) 14 (5) 108 (6) 89 (8) 16 (8)
Vegan 93 (3) 4 (1) 36 (2) 47 (4) 6 (3)
(Continued)
174 Vanhaecke et al.

TAB L E 1 (Continued)
All Bottled City Filtered Well Pvalue
(n=3413) 2(n=284) (n=1784) (n=1147) (n=198)
Types of plants per week, n<0.001
<5 205 (6) 43 (15) 97 (5) 53 (5) 12 (6)
6–10 792 (23) 87 (31) 416 (23) 252 (22) 37 (19)
11–20 1243 (36) 90 (32) 672 (38) 410 (36) 71 (36)
21–30 752 (22) 35 (12) 397 (22) 270 (24) 50 (25)
>30 421 (12) 29 (10) 202 (11) 162 (14) 28 (14)
Sugar-sweetened beverage frequency <0.001
Never 2503 (73) 200 (70) 1266 (71) 886 (77) 151 (76)
Rarely 663 (19) 45 (16) 401 (23) 182 (16) 35 (18)
Occasionally 143 (4) 19 (7) 72 (4) 43 (4) 9 (5)
Regularly 60 (2) 10 (4) 21 (1) 27 (2) 2 (1)
Daily 44 (1) 10 (4) 24 (1) 9 (1) 1 (0.5)
Alcohol frequency <0.001
Never 667 (20) 79 (28) 293 (16) 249 (22) 46 (23)
Rarely 937 (27) 89 (31) 460 (26) 332 (29) 56 (28)
Occasionally 765 (22) 56 (20) 428 (24) 254 (22) 27 (14)
Regularly 757 (22) 42 (15) 449 (25) 223 (19) 43 (22)
Daily 287 (8) 18 (6) 154 (9) 89 (8) 26 (13)
One liter of water a day frequency10 <0.001
Never 71 (2) 11 (4) 42 (2) 16 (1) 2 (1)
Rarely 239 (7) 17 (6) 152 (9) 57 (5) 13 (7)
Occasionally 379 (11) 28 (10) 220 (12) 114 (10) 17 (9)
Regularly 852 (25) 68 (24) 476 (27) 263 (23) 45 (23)
Daily 1865 (55) 160 (56) 887 (50) 697 (61) 121 (61)
1Values are medians [IQRs] for continuous variables and n(%) for categorical variables. Adjusted Pvalues are calculated from global chi-square test. Occasionally means 1–2
times/wk; rarely means a few times per month); regularly means 3–5 times/wk. GED, General Educational Development; IBD, Inammatory Bowel Disease; IBS, Irritable Bowel
Syndrome.
2Unless stated otherwise.
3Diagnosed by a medical professional (doctor, physician assistant).
4n=3391, missing data from n=9(city),n=11 (ltered), n=2 (well).
5n=3385, missing data from n=1 (bottled), n=15 (city), n=9 (ltered), n=3 (well).
6n=3389, missing data from n=1 (bottled), n=10 (city), n=10 (ltered), n=3 (well).
7n=3380, missing data from n=2 (bottled), n=18 (city), n=10 (ltered), n=3 (well).
8n=3354, missing data from n=3 (bottled), n=27 (city), n=27 (ltered), n=2 (well).
9n=3408, missing data from n=1 (bottled), n=1(city),n=3 (ltered).
10 n=3406, missing data from n=7(city),n=3 (ltered).
Microbiota βdiversity
βDiversity was assessed using unweighted and weighted UniFrac
distances as well as Bray–Curtis dissimilarity in QIIME at ASV level,
with a rarefaction at 1000 sequences (see Supplemental Methods).
To investigate interindividual differences in microbiota βdiversity in
relation to consumption of drinking water, permutational multivariate
analysis of variance (PERMANOVA) was performed using the adonis
function in the vegan R package with full adjustment. Only confounding
factors with equivalent dispersion for all βdiversity metrics were
retained, i.e., age, BMI, and diet type.
A permutation test was applied to assess whether dispersion of
groups of interest was equivalent (see Supplemental Methods). To
measure variables’ effect sizes, the proportion of variance (R2)captured
by a given variable was calculated as the ratio of the sum of squares
explained by the variable to the total sum of squares in PERMANOVAs
when applicable.
Microbiota taxonomy
Differences in the abundance of bacterial genera between drinking water
groups were assessed in adjusted models using the DESeq2 R package
(36) (see Supplemental Methods). DESeq2 is a differential abundance
detection method for 16S gene sequencing but with limited performance
on low-abundant ASVs with relatively high variance. Therefore, ASVs
were trimmed to those with a mean relative abundance ≥0.01% and
prevalence ≥10%.
Results
Participants’ characteristics
Participants were classied into groups of drinking water
source or intake based on their answers to the AGP standard
questionnaire. Participants’ characteristics according to their
drinking water source and intake are shown in Table 1 and
Supplemental Table 1, respectively. “Drinking water source”
means the source of drinking water usually consumed at home
by participants, i.e., bottled, city (tap water), ltered, or well.
A total of 3413 participants were included in this analysis. In
a separate analysis, 3794 participants were classied as either
high water drinkers (if they reported “daily” or “regularly”)
or low water drinkers (if they reported “never,” “rarely,”
“occasionally”) based on the variable “One liter of water a day
frequency.”
Participants included in these analyses were 56% women,
the median [IQR] age was 48 [36–59] y, and the majority
of participants had a normal BMI (61%) or were overweight
Drinking water consumption and gut microbiota 175

0%
25%
50%
75%
100%
Bottled
City
Filtered
Well
Proportion of
participants, %
Omnivore
Omnivore but do not eat red meat
Vegetarian but eat seafood
Vegetarian
Veg an
Diet type
A
0%
25%
50%
75%
100%
Bottled
City
Filtered
Well
Proportion of
participants, %
Fruit
B
0%
25%
50%
75%
100%
Bottled
City
Filtered
Well
Proportion of
participants, %
Vegetable
C
0%
25%
50%
75%
100%
Bottled
City
Filtered
Well
Proportion of
participants, %
Daily
Regularly
Occasionally
Rarely
Never
Whole grain
D
0%
25%
50%
75%
100%
Bottled
City
Filtered
Well
Proportion of
participants, %
>30
21–30
11–20
6–10
<5
Plant diversity
E
0%
25%
50%
75%
100%
Bottled
City
Filtered
Well
Proportion of
participants, %
Fermented plant
F
0%
25%
50%
75%
100%
Bottled
City
Filtered
Well
Proportion of
participants, %
Red meat
G
0%
25%
50%
75%
100%
Bottled
City
Filtered
Well
Proportion of
participants, %
Daily
Regularly
Occasionally
Rarely
Never
One liter of water
H
FIGURE 2 Distribution of dietary proles of American Gut Project participants according to the drinking water source (n=3413). (A) Diet type,
(B) fruit intake, (C) vegetable intake, (D) whole grain intake, (E) plant diversity of intake, (F) fermented plant intake, (G) red meat intake, (H) intake
of 1 L water/d. Occasionally meant 1–2 times/wk; rarely meant a few times per month; regularly meant 3–5 times/wk.
(24%). The vast majority of participants were nonsmokers
(94%), and only a small proportion reported having a chronic
disease (3% diabetes, 3% cardiovascular disease, 1% chronic
kidney disease, 3% inammatory bowel disease).
In the drinking water source subsample, 55% of participants
were American, whereas 45% were from the United Kingdom
(Tabl e 1 ). The proportion of subjects coming from the United
Kingdom and the average education level were higher in the
city group than in other groups. Subjects drinking bottled
water tended to consume more sugar-sweetened beverages, less
alcohol, and fewer ber-containing foods (fruits, vegetables, and
plants) (Tabl e 1 ,Figure 2). Fewer omnivores and red meat eaters
were in the ltered water group.
In the water intake analysis, the proportion of subjects
exercising regularly or daily was higher among the high water
drinkers than among the low water drinkers (Supplemental
Table 1). High water drinkers were younger than low water
drinkers. High water drinkers also had healthier dietary habits
because they drank less sugar-sweetened beverages and alcohol
and consumed more plants than low water drinkers.
Subjects’ demographic, dietary, and lifestyle variables associ-
ated with drinking water source and intake and that are known
to be associated with gut microbiota composition were regarded
as confounding factors and included in the statistical models
(Supplemental Figure 3, Supplemental Methods).
Plain drinking water and microbiota αdiversity
αDiversity, as evaluated by Faith’s PD and observed ASVs, was
associated with drinking water source in fully adjusted models
(P<0.05, type II ANOVA) (Figure 3A, Supplemental Figure
4A). Post hoc pairwise comparisons of the 4 groups of drinking
water source revealed that the αdiversity in the participants
drinking mostly well water was higher than in the other groups
(bottled compared with well, P<0.05; city compared with
well, P=0.05; ltered compared with well, P<0.05; pairwise
comparisons of estimated marginal means; Faith’s PD).
Results obtained for Shannon and Chao1 also showed
a trend for higher αdiversity in participants drinking well
water, although not reaching statistical signicance (Shannon,
P=0.14; Chao1, P=0.38; type II ANOVA) (Supplemental
Figure 4C, D). Whereas the majority of bottled water drinkers
came from California, well water drinkers were more evenly
distributed across the United States and United Kingdom
(Figure 3E). Figure 3B presents effect sizes, calculated as
the proportions of α-diversity variability explained by each
variable. The amounts of variance (R2) explained by age,
country, and BMI were 2.1 ×10−2,2.0×10−2, and 9.6 ×10−3,
respectively. Among all diet-related variables, plant diversity
(number of types of plants consumed per week) and diet
type were those explaining the most variance in αdiversity
with R2of 6.0 ×10−3and 3.7 ×10−3, respectively. The
inuence of drinking water source was 13% of that of age,
73% of that of diet type, and of the same magnitude as
those of collection season and alcohol frequency (Faith’s PD)
(Figure 3B). In addition, drinking water source variance was of
the same magnitude as those of alcohol frequency and exercise
frequency for observed ASVs (Supplemental Figure 4E). No
associations between drinking water intake and any of the α
diversity indexes for gut microbiota were observed (data not
shown). There were no differences in oral microbiota αdiversity
between drinking water sources or between high and low water
drinkers (Supplemental Figure 5A, Supplemental Figure 6A).
Plain drinking water and microbiota βdiversity
Analysis of microbiota variability revealed differences in β
diversity between groups of drinking water source, as measured
by Bray–Curtis dissimilarity and weighted UniFrac distance
(P<0.05; PERMANOVAs) in models adjusted for age,
BMI, and diet type (Figure 3C, Supplemental Figure 5B).
All pairs comparing city water and other water sources
were signicant ( P<0.05) (Figure 3C), with R2ranging
between 8.2 ×10−4and 1.6 ×10−3. The highest variance
explained was observed when comparing bottled with well
176 Vanhaecke et al.

10
12
14
Bottled City Filtered Well
Drinking water source
Adjusted D diversity
(Faith's PD)
A
0.0027
0.0207
0.0011
0.0096
0.0197
0.0031 0.0037
0.006
0.0032
0.000
0.005
0.010
0.015
0.020
Sex Drinking
water
source
Collection
season
Alcohol
frequency
Diet
type
Plant
diversity
BMI Country Age
D Diversity (Faith's PD)
variance explained (R2)
B
0.00155
0.00095
0.00304
0.00102
0.00082
0.00099
0.000
0.001
0.002
0.003
Bottled vs.
city
Bottled vs.
filtered
Bottled vs.
well
City vs.
filtered
City vs.
well
Filtered vs.
well
Drinking water source
E Diversity (Bray−Curtis)
variance explained (R2)
P > 0.05
P < 0.05
C
0.00199
0.00424
0.00265
0.00205
0.000
0.001
0.002
0.003
0.004
Drinking
water
source
Diet
type
BMI Age
E diversity (Bray−Curtis)
variance explained (R2)
D
Bottled
City
Filtered
Well
0% 25% 50% 75% 100%
Proportion of participants, %
States of residence
California
England
Other US states
and UK countries
E
FIGURE 3 Diversity analysis of fecal microbiota in American Gut Project participants according to their drinking water source (n=3413). (A)
αDiversity measured by Faith’s PD (type II ANOVA, P<0.05). Data are presented as adjusted means and 95% CIs. Adjusted for age, sex, BMI,
infant feeding, level of education, country, collection season, exercise frequency, diet type, plant diversity (number of types of plants consumed
per week), alcohol, and sugar-sweetened beverage consumption. Groups are bottled (n=284), city (n=1784), ltered (n=1147), and well
(n=198). (B) Faith’s PD effect sizes. Proportions of variance captured by signicant variables in fully adjusted models. (C) βDiversity measured
by Bray–Curtis dissimilarity. Proportions of variance captured by pairwise comparisons of drinking water source groups; adjusted for age, BMI,
and diet type. (D) Bray–Curtis dissimilarity effect sizes. (E) State or countr y of residence of participants according to their drinking water source.
Faith’s PD, Faith’s phylogenetic diversity.
water sources (R2=3.0 ×10−3). Several pairwise comparisons
were signicant across the 3 diversity metrics, with differences
between bottled and city consistently detected (Supplemental
Figure 5B). The amounts of variance ( R2) explained by age
and BMI were 4.2 ×10−3and 2.7 ×10−3, respectively.
The amounts of variance captured by drinking water source
were 47% and 50% of those captured by age for Bray–
Curtis dissimilarity and weighted UniFrac distance, respectively
(Figure 3D, Supplemental Figure 5C). In addition, low and
high water drinkers were detected to have different global fecal
microbiota compositions as measured by unweighted UniFrac
distance and Bray–Curtis dissimilarity in adjusted models (P
<0.05, unweighted UniFrac distance; P=0.05, Bray–Curtis
dissimilarity; PERMANOVAs) (Supplemental Figure 6B). “One
liter of water a day frequency” captured 5.6% and 10.6%
of the amounts of variance captured by age for unweighted
UniFrac distance (Supplemental Figure 6C) and Bray–Curtis
dissimilarity (data not shown), respectively. Differences in β
diversity of oral microbiota samples were detected between
low and high water drinkers for Bray–Curtis dissimilarity
and weighted UniFrac distance only when the model was
not adjusted. Adjustments for age, BMI, diet type, collection
season, and alcohol frequency attenuated these differences
(Supplemental Figure 6D). Analysis of the effect sizes revealed
that collection season was the factor explaining the most oral
microbiota diversity (Supplemental Figure 6E).
Plain drinking water and microbiota taxonomy
Differential abundance analyses revealed consistent differences
in participant fecal microbiota between drinking water source
Drinking water consumption and gut microbiota 177

groups after adjustment for confounding factors (see Supple-
mental Methods). Bottled and city water drinkers were notably
enriched in Bacteroides, Odoribacter,andStreptococcus genera
compared with well water drinkers (Figure 4). The same
trend was observed for other genera, including Veillonella and
Fusobacterium. Well water drinkers had higher Dorea genus
than city and ltered water drinkers. Bottled water drinkers’
fecal microbiota was enriched in different genera from the
Lachnospiraceae family compared with the city water drinkers
group. Comparing low and high water drinkers revealed a
differential pattern, with Campylobacter abundance enriched
in the low drinkers group. No differences were detected in
participants’ oral microbiota (Supplemental Figure 7).
Discussion
In this study, we explored the relations of plain drinking water
source and intake with the composition of the gut and oral
microbiota in a large cohort of self-selected participants. This
work is, to the best of our knowledge, the rst to specically
explore the links between drinking water consumption, in terms
of both origin and amount consumed, and the microbiota of the
digestive tract independently of a wide range of confounding
factors (demographics, lifestyle, and diet) in a large sample of
the general population. We observed that drinking water source
was among the major contributing factors to the variation in
gut microbiota composition. We showed that drinking water
source was associated with differences in gut microbiota as
assessed by Bray–Curtis dissimilarity and weighted Unifrac
distance. Subjects drinking mostly well water had higher fecal
microbiota αdiversity as measured by Faith’s PD and observed
ASVs than the other groups drinking either bottled, tap, or
ltered water. We further showed that drinking water intake
was associated with differences in gut microbiota composition
between participants who consumed low and high amounts
of drinking water. Finally, we failed to nd any links between
drinking water consumption and oral microbiota composition
in this cohort.
Lower gut bacterial diversity, as well as gene diversity,
has consistently been associated with impaired gut health
and diseases (37–39), indicating that diverse gut microbial
communities and functions may be more resilient and robust
against environmental inuences (40,41). Hence, higher
diversity appears a hallmark of healthy gut microbiomes.
Although the link between drinking water source and gut
microbiota composition deserves further investigation to es-
tablish whether it is causal, it could be hypothesized that
either the physicochemical, mineral, or microbial composition
of water, or a combination of the 3, may inuence the gut
microbiome. Results from a UK cohort suggest an association
between the composition of fecal microbiota and the mineral
composition of tap water (19). Specically, the intraindividual
fecal microbiota diversity was associated with the average daily
dose of sodium, i.e., a higher dose of sodium was associated
with a lower microbiota diversity. In addition, preclinical studies
have reported differences in the gut microbiota composition
of mice drinking acidic and neutral water, thereby suggesting
that water pH may affect gut microbial communities (20,21).
High intakes of mineral substances, including magnesium and
sodium sulfates, may alleviate constipation and improve stool
consistency in functionally constipated individuals (42,43).
These functional effects are thought to derive either from a
change in the intraluminal osmotic pressure, the stimulation
of NO synthase, or stimulation of intestinal water channels
regulating the fecal water content, which all affect intestinal
motility (43). Although there is a link between intestinal
motility, stool consistency, and fecal microbiota composition
(44–46), it is unlikely that a difference in gut motility may
explain the observed differences in αdiversity in this study. Our
study population neither consisted of functionally constipated
participants nor revealed any differences in bowel movement
quality and frequency between groups (Supplemental Figure 8).
Although we could not investigate the mineral content of the
diet of participants nor the mineral composition of drinking
waters because of the absence of such data, we can hypothesize
that it is unlikely that the mineral content of the drinking
water may explain the differences observed in gut microbiota
composition. Indeed, the variety of geographical locations of
the participants likely introduced various water sources and
mineral and physicochemical signatures within each group of
drinking water sources. Alternatively, the microbial community
in the drinking water, which varies between different water
sources, may also explain the differences observed in gut
microbiota composition. Drinking water contains an intrinsic
community of microorganisms that may differ between natural
sources, along the processing and distribution system, and
under different storage conditions (11–14). Whereas tap and
processed waters are subjected to systematic treatment to con-
trol microbial communities, some others, like natural mineral
waters, are not subjected to any disinfection treatment (47,48).
In addition, it has been recently described that the drinking
water microbiome is structurally and functionally less diverse
and variable across disinfected than across nondisinfected
systems (8). Therefore, the higher αdiversity observed in
the well water group could be due to more diverse natural
microbial communities in drinking water, because water coming
from wells, especially private wells, may not be systematically
subjected to disinfection treatment either (44,45). However,
we cannot fully ascertain the correctness of this hypothesis
owing to the absence of information about the operation of
wells in this database. Previous research reported that exposure
to different types of water with distinct endogenous microbial
communities affects the gut microbiota composition of mice.
In particular, mice drinking autoclaved tap water exhibited the
smallest intragroup variation in gut microbial diversity, and the
largest distance from other groups receiving either tap water,
disinfected water, or natural mineral water (16).
In this study, we observed fecal microbiota taxonomic
differences between individuals drinking mostly well water and
those drinking other sources of water. The Odoribacter genus,
which is positively associated with symptoms of functional
gastrointestinal disorders (46) and negatively correlated with
stool consistency (49), was depleted in well water drinkers’
fecal microbiota. Bacteroides genus relative abundance, which
is negatively associated with diversity in the human gut
microbiome (50), was also lower in well water drinkers than
in the other groups. This is in line with our ndings showing
that well water drinkers have higher αdiversity. Interestingly,
Streptoccocus, together with Veillonella and Fusobacterium,
known to be dominant in the upper digestive tract (51,52),
was found in higher abundance in fecal samples of bottled,
tap, and ltered water drinkers than in the well water drinkers
group. Recent studies report an enrichment of oral and small
intestinal species in the lower digestive tract of patients with
colorectal cancer (53) and liver cirrhosis (54). These results
might indicate some ecological shifts in bottled, tap, and ltered
water drinkers, with upper digestive tract–related species found
178 Vanhaecke et al.

FIGURE 4 Taxonomic analysis of fecal microbiota samples of American Gut Project participants according to their drinking water source
(n=3413) or intake (n=3794). Heatmap of log2 fold changes of pairs. Water source analysis adjusted for age, sex, BMI, infant feeding, level
of education, country, collection season, exercise frequency, diet type, plant diversity (number of types of plants consumed per week), alcohol,
and sugar-sweetened beverage consumption. Water intake analysis adjusted for age, sex, BMI, infant feeding, level of education, continent,
collection season, exercise frequency, diet type, plant diversity, alcohol, and sugar-sweetened beverage consumption. All genera that had a log2
fold change >0.5 or an FDR <0.1 were selected. Colors account for log2 fold change between groups and dots account for FDR signicance. P
values for Wald’s test. FDR, false discovery rate.
in higher abundances in the lower gut, although the differences
observed are small. In our study, well water was the only
source of water coming solely from groundwater, i.e., water that
exists underground beneath the land surface after percolation
through earth layers (44,45), a distinction that may contribute
to explaining the taxonomic differences observed. Although
we found that the increased diversity in the gut microbiota of
well water drinkers was independent of the country or state
of residence, the geographical locations of well water drinkers
were evenly distributed across each country, which may have
confounded the results. In addition, it may be hypothesized
that well water drinkers may more frequently live in rural
than in urban areas and that their exposure to environmental
microbial diversity could be increased, resulting in increased gut
microbiota diversity (55,56).
In our study, drinking water source ranked among the key
contributing factors explaining gut microbiota variation in the
α-andβ-diversity analyses, with effect sizes comparable with
those of alcohol or diet type, thereby adding water as a novel
variable to report in gut microbiota studies. This nding is
consistent with the general observation that environmental
factors, including diet, inuence the gut microbiome (1,2). The
identication of alcohol as a contributor to gut microbiota
composition is also consistent with a recent investigation (57).
In our exploration, the amount of variance of βdiversity
captured by drinking water sources was half of the variance
captured by age, a factor highly associated with microbiota
compositions (35,58). The ranking of effect sizes reported
in our study between age and BMI is similar to that
previously reported in a different cohort (24). Our ndings
are also in line with recent evidence showing differences in
gut microbiota composition in native Himalayan populations
associated with drinking water source (stream compared
with underground well water) (18). The source of drinking
water was identied as the factor most strongly associated
with the gut microbiota composition of these populations,
as assessed by Bray–Curtis dissimilarity, unweighted UniFrac,
and weighted UniFrac distances. This nding was replicated
in another native East-African population of hunter-gatherers
(18).
Drinking water consumption and gut microbiota 179

Surprisingly, we found that drinking water intake was
associated with βdiversity, as measured by unweighted
UniFrac distance and Bray–Curtis dissimilarity, suggesting
that water intake is associated with global fecal microbiota
compositions. Because this observational study cannot prove
causality, we cannot assess whether these drinking water
variables directly affect fecal microbiota compositions. Still, our
ndings emphasize the need to use them as proxies for lifestyle
habits to control in future fecal microbiota analyses. These
differences in gut microbiota compositions were also reected
in taxonomic analyses where we found that high water drinkers
had a lower abundance of the Campylobacter genus known
to cause gastrointestinal infection (59). Although it is known
that increased water intake can alleviate functional constipation
(43,60), little is known about whether the volume of water
consumed could alter gut microbiota composition in the general
population.
When considering the oral microbiota composition, drinking
water source was not associated with differences in αdiversity,
βdiversity, or taxonomy. Previous research reported that the
chemical composition of tap water was the environmental
factor with the highest impact on the composition of the
oral microbiota in a Spanish cohort (9), with some microbial
abundance following a geographical pattern similar to that
of public water quality parameters, e.g., water alkalinity and
water hardness. Our study is likely underpowered to detect
any changes in oral microbiota composition owing to the small
sample size and the wide geographical diversity of samples.
The association observed between drinking water intake and
βdiversity in unadjusted models may reect differences in
lifestyle habits that may confound the association. Further
research with a larger sample size is warranted to conrm these
conclusions.
The present study comes with limitations. Although a wide
range of known confounders was either considered in the
exclusion criteria or adjusted for in the statistical models, we
neither excluded nor adjusted for pregnancy, cesarean deliveries,
or some chronic diseases, which accounted for 1%, 9%, and
1%–3% of the study population, respectively. The reasons for
this include that we wanted to limit the collinearity deriving
from multiple adjustments and maintain a fair representation
of the subjects’ population. It is recalled that AGP is a self-
selected cohort that is not representative of the US or the
UK population. In addition, we excluded participants with
missing data on all confounding factors, which may have
introduced an additional selection bias. A key limitation of
any citizen-based approach is that it relies on self-reported,
subjective measures as well as variables with low granularity to
reduce participants’ burden, which may have introduced some
variability in our analyses. For example, whereas drinking water
source referred to the main source of water available at home,
subjects may have diverse drinking water sources outside of
their home. In addition, bottled water could be either natural
mineral water, spring, or treated water. Finally, in this study,
we used 16S rRNA gene data to explore the overall gut micro-
biota composition, which limits taxonomy resolution. Further
studies using shotgun metagenomics sequencing would allow
functional characterization and improve analysis resolution at
species level. Beyond compositional and functional diversity, gut
microbiome biomarkers such as SCFAs should be proled to
increase knowledge of host–microbe interactions. Nonetheless,
such a citizen-science-based approach enables the collection of
a large number and a diversity of samples. This allowed us to
analyze data from both oral and fecal samples, although AGP
sampling guidelines for the collection of oral samples should
be standardized to gain precision on the sample location (e.g.,
saliva, tongue).
This study showed that drinking water source and intake
were associated with fecal microbiota composition and found
limited associations with oral microbiota composition. The
work performed here contributed to formulating some hypothe-
ses about the links between water consumption and fecal and
oral microbiota compositions. It may guide future microbiota
analyses, because it pointed out the importance of reporting the
source and intake of drinking water consumed by participants
in studies or population-based cohorts investigating the human
microbiome.
Acknowledgments
We are indebted to Aurélie Cotillard for helpful discussion
on statistical analysis, and to Erica T Perrier, Muriel Derrien,
Hana Koutnikova, Lodovico Di Gioia, Matthieu Pichaud,
and Patrick Veiga (all of Danone Research) for the critical
reading of the manuscript. The authors’ responsibilities were as
follows—TV: designed the research; TV and JT: wrote the paper
and had primary responsibility for the nal content; and all
authors: conducted the research, analyzed the data, performed
the statistical analyses, and read and approved the nal
manuscript.
References
1. Wu GD, Chen J, Hoffmann C, Bittinger K, Chen YY, Keilbaugh
SA, Bewtra M, Knights D, Walters WA, Knight R, et al. Linking
long-term dietary patterns with gut microbial enterotypes. Science
2011;334(6052):105–8.
2. Ding T, Schloss PD. Dynamics and associations of microbial community
types across the human body. Nature 2014;509(7500):357–60.
3. Kundu P, Blacher E, Elinav E, Pettersson S. Our gut microbiome: the
evolving inner self. Cell 2017;171(7):1481–93.
4. Relman DA. The human microbiome: ecosystem resilience and health.
Nutr Rev 2012;70(Suppl 1):S2–S9.
5. Institute of Medicine. Dietary Reference Intakes for water, potassium,
sodium, chloride, and sulfate. Washington (DC): National Academies
Press; 2005.
6. EFSA Panel on Dietetic Products, Nutrition, and Allergies (NDA).
Scientic opinion on dietary reference values for water. EFSA J
2010;8(3):1459.
7. CDC. Community water treatment [Internet]. Atlanta, GA: Centers
for Disease Control and Prevention; 2015 [cited May 2021].
Available from: https://www.cdc.gov/healthywater/drinking/public/wat
er_treatment.html.
8. Dai Z, Sevillano-Rivera MC, Calus ST, Bautista-de los Santos QM, Eren
AM, van der Wielen PWJJ, Ijaz UZ, Pinto AJ. Disinfection exhibits
systematic impacts on the drinking water microbiome. Microbiome
2020;8(1):42.
9. Willis JR, González-Torres P, Pittis AA, Bejarano LA, Cozzuto L,
Andreu-Somavilla N, Alloza-Trabado M, Valentín A, Ksiezopolska E,
Company C, et al. Citizen science charts two major “stomatotypes” in
the oral microbiome of adolescents and reveals links with habits and
drinking water composition. Microbiome 2018;6(1):218.
10. Patterson KY, Pehrsson PR, Perry CR. The mineral content of tap
water in United States households. J Food Compos Anal 2013;31(1):
46–50.
11. Leclerc H, Moreau A. Microbiological safety of natural mineral water.
FEMS Microbiol Rev 2002;26(2):207–22.
12. De Roy K, Clement L, Thas O, Wang Y, Boon N. Flow cytometry
for fast microbial community ngerprinting. Water Res 2012;46(3):
907–19.
13. Zhang Y, Oh S, Liu W-T. Impact of drinking water treatment and
distribution on the microbiome continuum: an ecological disturbance’s
perspective. Environ Microbiol 2017;19(8):3163–74.
180 Vanhaecke et al.

14. Sacchetti R, De Luca G, Dormi A, Guberti E, Zanetti F. Microbial
quality of drinking water from microltered water dispensers. Int J Hyg
Environ Health 2014;217(2–3):255–9.
15. Rechenburg A. WHO CC Newsletter Microbiomes: an undervalued
impact of water treatment on human gut health? 2019. December
29. https://www.ukbonn.de/site/assets/les/16453/water_and_risk_vol
29_low.pdf
16. Dias MF, Reis MP, Acurcio LB, Carmo AO, Diamantino CF, Motta
AM, Kalapothakis E, Nicoli JR, Nascimento AMA. Changes in mouse
gut bacterial community in response to different types of drinking water.
Water Res 2018;132:79–89.
17. Baumann-Dudenhoeffer AM, D’Souza AW, Tarr PI, Warner BB, Dantas
G. Infant diet and maternal gestational weight gain predict early
metabolic maturation of gut microbiomes. Nat Med 2018;24(12):
1822–9.
18. Jha AR, Davenport ER, Gautam Y, Bhandari D, Tandukar S, Ng
KM, Fragiadakis GK, Holmes S, Gautam GP, Leach J, et al. Gut
microbiome transition across a lifestyle gradient in Himalaya. PLoS Biol
2018;16(11):e2005396.
19. Bowyer RCE, Schillereff DN, Jackson MA, Le Roy C, Wells PM,
Spector TD, Steves CJ. Associations between UK tap water and gut
microbiota composition suggest the gut microbiome as a potential
mediator of health differences linked to water quality.Sci Total Environ
2020;739:139697.
20. So MH, Gudi R, Karumuthil-Melethil S, Perez N, Johnson BM, Vasu
C. pH of drinking water inuences the composition of gut microbiome
and type 1 diabetes incidence. Diabetes 2014;63(2):632–44.
21. Wolf KJ, Daft JG, Tanner SM, Hartmann R, Khapour E, Lorenz
RG. Consumption of acidic water alters the gut microbiome and
decreases the risk of diabetes in NOD mice. J Histochem Cytochem
2014;62(4):237–50.
22. Boorman GA. Drinking water disinfection byproducts: review
and approach to toxicity evaluation. Environ Health Perspect
1999;107(suppl 1):207–17.
23. Sasada T, Hinoi T, Saito Y, Adachi T, Takakura Y, Kawaguchi Y,
Sotomaru Y, Sentani K, Oue N, Yasui W, et al. Chlorinated water
modulates the development of colorectal tumors with chromosomal
instability and gut microbiota in Apc-decient mice. PLoS One
2015;10(7):e0132435.
24. Falony G, Joossens M, Vieira-Silva S, Wang J, Darzi Y, Faust
K, Kurilshikov A, Bonder MJ, Valles-Colomer M, Vandeputte D,
et al. Population-level analysis of gut microbiome variation. Science
2016;352(6285):560–4.
25. Zhernakova A, Kurilshikov A, Bonder MJ, Tigchelaar EF, Schirmer
M, Vatanen T, Mujagic Z, Vila AV, Falony G, Vieira-Silva S, et al.
Population-based metagenomics analysis reveals markers for gut
microbiome composition and diversity.Science 2016;352(6285):565–9.
26. Perrier ET, Armstrong LE, Bottin JH, Clark WF, Dolci A, Guelinckx
I, Iroz A, Kavouras SA, Lang F, Lieberman HR, et al. Hydration for
health hypothesis: a narrative review of supporting evidence. Eur J Nutr
2021;60(3):1167–80.
27. Benton D, Young HA. Water: the Cinderella nutrient. J Nutr
2019;149(12):2081–2.
28. McDonald D, Hyde E, Debelius JW, Morton JT, Gonzalez A,
Ackermann G, Aksenov AA, Behsaz B, Brennan C, Chen Y, et al.
American gut: an open platform for citizen science microbiome research.
mSystems 2018;3(3):e00031–18.
29. McDonald D, Kaehler B, Gonzalez A, DeReus J, Ackermann G, Marotz
C, Huttley G, Knight R. redbiom: a rapid sample discovery and feature
characterization system. mSystems 2019;4(4):e00215–19.
30. Gonzalez A, Navas-Molina JA, Kosciolek T, McDonald D, Vázquez-
Baeza Y, Ackermann G, DeReus J, Janssen S, Swafford AD, Orchanian
SB, et al. Qiita: rapid, web-enabled microbiome meta-analysis. Nat
Methods 2018;15(10):796–8.
31. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a
practical and powerful approach to multiple testing. J R Statist Soc B
1995;57(1):289–300.
32. Chao A. Nonparametric estimation of the number of classes in a
population. Scand J Statist 1984;11(4):265–70.
33. Shannon CE. A mathematical theory of communication. Bell Syst Tech
J 1948;27(3):379–423.
34. Faith DP. Conservation evaluation and phylogenetic diversity. Biol
Conserv 1992;61(1):1–10.
35. Yatsunenko T, Rey FE, Manary MJ, Trehan I, Dominguez-Bello MG,
Contreras M, Magris M, Hidalgo G, Baldassano RN, Anokhin AP,
et al. Human gut microbiome viewed across age and geography. Nature
2012;486(7402):222–7.
36. Love MI, Huber W, Anders S. Moderated estimation of fold
change and dispersion for RNA-seq data with DESeq2. Genome Biol
2014;15(12):550.
37. Lozupone CA, Stombaugh JI, Gordon JI, Jansson JK, Knight R.
Diversity, stability and resilience of the human gut microbiota. Nature
2012;489(7415):220–30.
38. Sommer F, Anderson JM, Bharti R, Raes J, Rosenstiel P. The resilience
of the intestinal microbiota inuences health and disease. Nat Rev
Microbiol 2017;15(10):630–8.
39. Le Chatelier E, Nielsen T, Qin J, Prifti E, Hildebrand F, Falony G,
Almeida M, Arumugam M, Batto J-M, Kennedy S, et al. Richness
of human gut microbiome correlates with metabolic markers. Nature
2013;500(7464):541–6.
40. Valdes AM, Walter J, Segal E, Spector TD. Role of the gut microbiota
in nutrition and health. BMJ 2018;361:k2179.
41. Cotillard A, Kennedy SP, Kong LC, Prifti E, Pons N, Le Chatelier
E, Almeida M, Quinquis B, Levenez F, Galleron N, et al. Dietary
intervention impact on gut microbial gene richness. Nature
2013;500(7464):585–8.
42. Dupont C, Campagne A, Constant F. Efcacy and safety of a
magnesium sulfate–rich natural mineral water for patients with
functional constipation. Clin Gastroenterol Hepatol 2014;12(8):
1280–7.
43. Bothe G, Coh A, Auinger A. Efcacy and safety of a natural
mineral water rich in magnesium and sulphate for bowel function:
a double-blind, randomized, placebo-controlled study. Eur J Nutr
2017;56(2):491–9.
44. US Environmental Protection Agency (EPA). Learn about private well
water [Internet]. Washington (DC): US EPA [cited May 2021]. Available
from: https://www.epa.gov/privatewells/learn-about-private-water-wel
ls.
45. CDC. Private ground water wells [Internet]. Atlanta, GA: Centers for
Disease Control and Prevention; 2014 [cited May 2021]. Available
from: https://www.cdc.gov/healthywater/drinking/private/wells/.
46. Jeffery IB, O’Toole PW, Öhman L, Claesson MJ, Deane J, Quigley EM,
Simrén M. An irritable bowel syndrome subtype dened by species-
specic alterations in faecal microbiota. Gut 2012;61(7):997–1006.
47. European Commission Directorate-General for Health and Food Safety.
Natural mineral waters and spring water 2020 [Internet]. Brussels,
Belgium: European Commission; [cited May 2021]. Available from:
https://ec.europa.eu/food/safety/labelling_nutrition/mineral_waters_en.
48. European Parliament. Directive 2009/54/EC of the European
Parliament and of the Council of 18 June 2009 on the exploitation
and marketing of natural mineral waters. Brussels, Belgium: European
Parliament; 2009.
49. Vandeputte D, Falony G, Vieira-Silva S, Tito RY, Joossens M,
Raes J. Stool consistency is strongly associated with gut microbiota
richness and composition, enterotypes and bacterial growth rates. Gut
2016;65(1):57–62.
50. Tap J, Derrien M, Törnblom H, Brazeilles R, Cools-Portier S, Doré
J, Störsrud S, Le Nevé B, Öhman L, Simrén M. Identication of an
intestinal microbiota signature associated with severity of irritable
bowel syndrome. Gastroenterology 2017;152(1):111–23.e8.
51. van den Bogert B, Erkus O, Boekhorst J, de Goffau M, Smid EJ,
Zoetendal EG, Kleerebezem M. Diversity of human small intestinal
Streptococcus and Veillonella populations. FEMS Microbiol Ecol
2013;85(2):376–88.
52. The Human Microbiome Project Consortium. Structure, function
and diversity of the healthy human microbiome. Nature
2012;486(7402):207–14.
53. Zeller G, Tap J, Voigt AY, Sunagawa S, Kultima JR, Costea PI,
Amiot A, Böhm J, Brunetti F, Habermann N, et al. Potential of fecal
microbiota for early-stage detection of colorectal cancer. Mol Syst Biol
2014;10(11):766.
54. Qin N, Yang F, Li A, Prifti E, Chen Y, Shao L, Guo J, Le Chatelier E,
Yao J, Wu L, et al. Alterations of the human gut microbiome in liver
cirrhosis. Nature 2014;513(7516):59–64.
55. Ayeni FA, Biagi E, Rampelli S, Fiori J, Soverini M, Audu HJ, Cristino S,
Caporali L, Schnorr SL, Carelli V, et al. Infant and adult gut microbiome
Drinking water consumption and gut microbiota 181

and metabolome in rural Bassa and urban settlers from Nigeria. Cell
Rep 2018;23(10):3056–67.
56. Tyakht AV, Kostryukova ES, Popenko AS, Belenikin MS, Pavlenko AV,
Larin AK, Karpova IY, Selezneva OV, Semashko TA, Ospanova EA,
et al. Human gut microbiota community structures in urban and rural
populations in Russia. Nat Commun 2013;4(1):2469.
57. Vujkovic-Cvijin I, Sklar J, Jiang L, Natarajan L, Knight R, Belkaid
Y. Host variables confound gut microbiota studies of human disease.
Nature 2020;587(7834):448–54.
58. Derrien M, Alvarez AS, de Vos WM. The gut microbiota in the rst
decade of life. Trends Microbiol 2019;27(12):997–1010.
59. Igwaran A, Okoh AI. Human campylobacteriosis: a public health
concern of global importance. Heliyon 2019;5(11):e02814.
60. Anti M, Pignataro G, Armuzzi A, Valenti A, Iascone E, Marmo R,
Lamazza A, Pretaroli AR, Pace V, Leo P, et al. Water supplementation
enhances the effect of high-ber diet on stool frequency and
laxative consumption in adult patients with functional constipation.
Hepatogastroenterology 1998;45(21):727–32.
182 Vanhaecke et al.