Content uploaded by Arne N Linlokken
Author content
All content in this area was uploaded by Arne N Linlokken on Sep 03, 2018
Content may be subject to copyright.
Correlation between the human fecal microbiota and
depression
A. NASERIBAFROUEI,* K. HESTAD,†,‡E. AVERSHINA,§M. SEKELJA,§A. LINLØKKEN,* R. WILSON*&K. RUDI*,§
*Faculty of Education and Science, Hedmark University College, Hamar, Norway
†Division of Mental Health, Innlandet Hospital Trust, Hamar, Norway
‡Department of Psychology, Norwegian University of Science and Technology, NTNU, Trondheim, Norway
§Department of Chemistry, Biotechnology and Food Science, Norwegian University of Life Sciences,
As, Norway
Key Messages
•We found a significant correlation between the gut microbiota and depression in humans. The correlations,
however, were in opposite directions for closely related Operational Taxonomic Units (OTU’s).
•The aim of our work was to investigate the correlation between the human fecal microbiota and depression.
•16S rRNA gene Illumina deep sequencing from 37 depressed and 18 non-depressed individuals was conducted.
The deep sequencing data were analyzed using an in-house generated computer program, enabling the
resolution of closely related OTU’s.
•The order Bacteroidales was overrepresented (p= 0.05), while the family Lachnospiraceae was underrepre-
sented (p= 0.02) with respect to OTU’s associated with depression.
Abstract
Background Depression is a chronic syndrome with a
pathogenesis linked to various genetic, biological, and
environmental factors. Several links between gut
microbiota and depression have been established in
animal models. In humans, however, few correlations
have yet been demonstrated. The aim of our work was
therefore to identify potential correlations between
human fecal microbiota (as a proxy for gut microbiota)
and depression. Methods We analyzed fecal samples
from 55 people, 37 patients, and 18 non-depressed
controls. Our analyses were based on data generated
by Illumina deep sequencing of 16S rRNA gene
amplicons. Key Results We found several correlations
between depression and fecal microbiota. The corre-
lations, however, showed opposite directions even for
closely related Operational Taxonomic Units
(OTU’s), but were still associated with certain higher
order phylogroups. The order Bacteroidales showed an
overrepresentation (p =0.05), while the family Lach-
nospiraceae showed an underrepresentation (p =0.02)
of OTU’s associated with depression. At low taxonomic
levels, there was one clade consisting of five OTU’s
within the genus Oscillibacter, and one clade within
Alistipes (consisting of four OTU’s) that showed a
significant association with depression (p =0.03 and
0.01, respectively). Conclusions & Inferences The
Oscillibacter type strain has valeric acid as its main
metabolic end product, a homolog of neurotransmitter
GABA, while Alistipes has previously been shown to be
associated with induced stress in mice. In conclusion,
the taxonomic correlations detected here may therefore
correspond to mechanistic models.
Keywords 16SrRNA gene, depression, gut microbiota,
Illumina deep sequencing.
INTRODUCTION
Major depressive disorder is a mental disorder which
presents itself with low mood, low self-esteem, and
loss of interest in normally enjoyable activities.
1
Address for Correspondence
Knut Rudi, Norwegian University of Life Sciences,
Department of Chemistry, Biotechnology and Food Science,
PO Box 5003, NO-1432
Aas, Norway.
Tel: +47 64 96 58 73; fax: +47 64 96 59 01;
e-mail: knut.rudi@nmbu.no
Received: 6 September 2013
Accepted for publication: 7 May 2014
©2014 John Wiley & Sons Ltd 1155
Neurogastroenterol Motil (2014) 26, 1155–1162 doi: 10.1111/nmo.12378
Neurogastroenterology & Motility
Depression is a multi-factorial disease being caused by
biological, psychological, and social factors. The diath-
esis–stress model proposes that depression is caused
when stressful life events superimpose on a pre-
existing vulnerable condition.
2
There has been recent major interest in the gut–
brain axis in relation to depression. Three general
mechanisms have been suggested to describe how the
gut microbiota influences depression, namely through
inflammation, the Hypothalamic–Pituitary–Adrenal
axis (HPA), or interference with neurotransmitter
signaling.
3
IgA- and IgM-mediated inflammatory responses to
lipopolysaccharide have been shown to be elevated in
depressed patients.
4
It has also been suggested that
depression can result from maladaptation due to
abnormalities in circulating cytokines.
5,6
A meta-
analysis of the clinical literature implicates higher
inflammatory IL-6 and TNF-ain depressed patients
compared to controls.
7
Furthermore, in mice it has
been shown that gastrointestinal inflammation both
induces anxiety behavior and alters central nervous
system biochemistry.
8
HPA is a neuro-endocrine stress response system,
being important in both mood disorders and functional
diseases. Alterations of the HPA system have been
diagnosed in patients harboring different mental states
including posttraumatic stress disorder,
9
schizophre-
nia,
10
social anxiety
11
, and depression.
12
In a rat model,
it has recently been shown that treatment with probiotic
bacteria can interfere with the HPA response to acute
physiological stress,
13
suggesting a mechanistic con-
nection between the gut microbiota, HPA, and stress.
Direct interference with neurotransmitter signaling
may also be involved in depressive disorders. It has
been shown that the neurotransmitter GABA can be
produced by intestinal bacteria.
14
Furthermore, probi-
otic bacteria can modulate depressive behavior through
GABA signaling in a mouse model.
15
The other
signaling pathway that has been linked to depression
is serotonergic signaling, where it has been shown that
the serotonergic turnover is higher in the striatum in
germ-free mice compared to conventional animals.
16
Despite several indirect lines of evidence correlating
gut microbiota to depression through inflammatory,
stress, or signaling pathways, we still lack knowledge
about the direct correlation patterns between gut
microbiota and depression in humans.
3,17
Therefore,
our study aims at identifying direct correlations
between human fecal microbiota (as a proxy for gut
microbiota) and depressive disorder.
To elucidate fecal microbiota/depression correla-
tions, we used an explorative approach involving deep
Illumina sequencing of 16S rRNA gene amplicons. We
analyzed a relatively large cohort of clinically
depressed patients (n=37) and matched non-depressed
controls (n=18). Due to the complexity of the data, we
chose a combination of multivariate statistical meth-
ods in determining correlations between the fecal
microbiota and depression.
MATERIALS AND METHODS
Patients and design
Thirty-seven depressed patients were recruited from an inward
and outpatient mental health clinic of the Innlandet Hospital in
Norway. They represented a group of patients with mild to severe
depression. All participants had a diagnosis of depression accord-
ing to the research criteria of ICD-10, and using the Montgomery-
Asberg Depression Rating Scale (MADRS), which is a diagnostic
questionnaire with 10 items to evaluate the severity of the
disorder.
18
On the MADRS, a score from 0 to 7 is normal with no
indication of depression. Scores of 7–19 indicate depressive
symptoms of milder forms, 20–34 moderate depression, and
scores above 34 indicate severe depression. A control group of
18 patients with the same age and gender distribution were
recruited from an outpatient neurological unit at the same
hospital. These patients had diffuse symptoms, which possibly
could be related to cerebral disorders, but no disorders could be
found. They all had a careful neurological examination and CT/
MRI scans. Descriptive statistics for the cohort are shown in
Table 1.
The study was designed as a partially blinded observational
study in which clinical information including diagnostic status
remained unknown until after fecal microbial analyses were
finalized.
Stool samples and DNA extraction
Stool samples were immediately frozen at 20 °C in the patient’s
home freezer after defecation. The samples were then transported
at below zero to centralized 70 °C storage. For the patients who
were hospitalized, the samples were taken directly to the
centralized freezer. Within 1 month, the stool specimens were
weighed, and S.T.A.R. (Stool Transport and Recovery; Roche,
Basel, Switzerland) buffer solution was added to each sample at a
ratio of ~1 (stool) to 3 (S.T.A.R. buffer). Samples were vortexed to
achieve homogenous suspension and then stored at 80 °C before
DNA extraction.
The DNA extraction protocol used has previously been exten-
sively validated with respect to CE marking of an irritable bowel
syndrome (IBS) dysbiosis test.
19
For DNA extraction, frozen stool
samples were first allowed to thaw on ice. Microcentrifuge tubes
(2 mL) containing 250-mg glass beads (<106 lm) were filled with a
suspension volume of 0.5 mL of the stool sample. To achieve
bacterial cell lysis, homogenization was performed using a
MagNaLyser (Roche) twice at 2000 rpm for 40 s, with 40 s
cooling between runs. The samples were kept cold during the
rest phase to avoid DNA degradation due to overheating. This step
was followed by centrifugation at 12 300 gfor 5 min. The
supernatant lysate solution was then transferred to a new
microcentrifuge tube in two replicates (designated parallel A and
B) for each of the samples. Fifty microlitres supernatant from the
tubes were transferred to a KingFisher 96-well plate and DNA was
©2014 John Wiley & Sons Ltd1156
A. Naseribafrouei et al. Neurogastroenterology and Motility
extracted using the Mag
TM
mini kit (LGC, Middlesex, UK),
following the manufacturer’s recommendations.
Illumina sequencing
For Illumina sequencing, we used different combinations of
forward and reverse primers for each sample to generate 116
libraries. These combinations are presented in the Supplementary
text. The PCR was conducted in a 25 lL reaction volume with the
following composition: 1.25 U HOT FIREPol
â
DNA polymerase,
19HOT FIREPol
â
buffer B2, 2.5 mM Magnesium dichloride,
0.2 mM dNTPs, 0.2 lM forward primer, 0.2 lM reverse primer,
and 1 lL template. We used 30 cycles of denaturation (95 °C for
30 s), annealing (50 °C for 1 min), and elongation (72 °C for 45 s).
Because of primer-dimer formation, the PCR products were
semiquantified by agarose gel electrophoresis, and all the ampli-
cons were mixed in equimolar concentrations. Finally, the pooled
products were purified using E.Z.N.A PCR product purification
kit (Omega Bio-Tech, Norcross, GA, USA), and submitted for
paired-end 250 bp sequencing on the MiSeq Illumina platform at
the Norwegian High Throughput Sequencing Center (UiO, Oslo,
Norway).
Data analysis
Due to computational speed and resolution, we used a previously
developed word-based approach in OTU identification.
20,21
For
these analyses, we only included sequences with an average Phred
score >31 (per base error rate <0.001) to avoid the influence of
sequence errors. For each sample, 4000 sequences were selected
(2000 from each of the two parallels). Parallels with <2000
sequences satisfying the filtering criteria were discharged from
the analyses. This number was chosen as a trade-off between
amount of information and the number of samples lost. Our
alignment-independent, multimer-based approach was used to
determine the relatedness between the sequences through prin-
cipal component analysis. OTU binning was based on a 0.5 90.5
interval from the score for the two first principal components. The
low interval size was chosen to obtain a high resolution for closely
related OTU’s. The 100 most dominant OTU’s were selected for
phylogenetic reconstruction (CLC Genomics Workbench) and
taxonomic assignment (RDP database with default settings).
We used false discovery rate corrected permutation testing in
determining the significance for the univariate differences
detected. Briefly, this involves determining whether the differ-
ences detected are larger than those expected by chance. For each
OTU, we tested if the difference between the average levels (in
percentage) in depressed vs non-depressed individuals signifi-
cantly deviated from 0.
To reveal potential complex multivariate interactions between
bacteria for the microbiota correlation with depression, we used
partial least square discriminant analysis (PLS-DA). This is a
multivariate statistical method for relating a binary response
variable (in our case depression) to a large number of explanatory
variables (in our case OTU’s). The models were cross-validated
using a Venetian blind approach splitting the dataset into two,
using one half to build the model, and the other half for validation.
For verification, we used the Quantitative Insights Into
Microbial Ecology (QIIME) pipeline following the analytical
recommendations given in the user homepage (http://qiime.org/
1.6.0/tutorials/illumina_overview_tutorial.html) with OTU bin-
ning at the 1% level and closed reference OTU searches with
modified uclust parameters (https://gist.github.com/gregcapor-
aso/5952785). The QIIME analyses were run on Amazon Cloud
Service using StarCluster to load eight m2.49large instances and
start 80 parallel jobs. For each sample, 6000 sequences were
picked (3000 from each parallel). Parallels with <3000 sequence
reads were discharged from the analyses.
RESULTS
Library size and characteristics
We obtained a total of 4 910 922 reads that could be
assigned to the 116 libraries generated. The average
number of reads per library was 42 336. The number of
reads per sample is summarized in Fig. S1. The tech-
nical quality was evaluated by regression analyses from
parallel analyses (replicates A and B) with respect to
OTU distribution. These analyses showed very good
correspondence between the technical replicates, with
an average R
2
value of 0.96 0.04. Comparing samples
from different individuals, on the other hand, generally
gave low R
2
values (<0.3).
Microbiota composition
We identified a total of 1593 OTU’s with a pairwise
sequence distance between the 100 most dominant
Table 1 Clinical characteristics
NMean SD
Age
Control 18 46.1 13.9
Depressed 37 49.2 13.9
Blood pressure medication
Control 18 0.22 0.43
Depressed 37 0.24 0.43
Gender (female, vs male)
Control 18 11 7
Depressed 37 20 17
Systolic blood pressure
Control 18 131.2 14.0
Depressed 37 133.6 23.8
Diastolic blood pressure
Control 18 80.4 12.4
Depressed 37 82.1 15.0
Education (year)
Control 18 13.5 2.4
Depressed 37 12.7 2.8
Mini Mental state
Control 18 29.1 1.0
Depressed 37 28.4 1.6
MADRS depression score
Control 18 7.2 4.8
Depressed 37 26.3 7.6
Depression medication
Control 18 0.06 0.24
Depressed 37 0.73 0.45
Body mass index (BMI)
Control 18 24.7 3.3
Depressed 36 25.9 4.2
©2014 John Wiley & Sons Ltd 1157
Volume 26, Number 8, August 2014 Human fecal microbiota and depression
OTU’s (representing 61.6% of the sequence reads) in
the range 0.5–40% (Fig. S2). The OTU’s were classified
using the RDP classifier (Table S1), and related using
phylogenetic reconstruction (Fig. 1). The most domi-
nant OTU (OTU1) in our dataset was classified as
B. ovatus. This OTU represented 4% of all the
sequence reads (Fig. 1B), and was identified in 54 of
55 individuals. In accordance with a range of other
human fecal microbiota studies, we also found a clear
dominance of the phyla Bacteroidetes and Firmicutes
in our cohort (Table S1).
Correlation between microbiota and depression
There were no significant differences between
depressed and non-depressed patients with respect to
species richness, although there was a slight tendency
(p=0.09, t-test) of a higher number of OTU’s among
the depressed (374 56) compared to the non-
depressed (351 42) individuals. Neither did we iden-
tify any significant differences in a-diversity (Simp-
son’s D index of 39.5 15.9 for depressed vs
34.4 19.6 for non-depressed patients).
At a high taxonomic level (Domain), there was an
underrepresentation of Bacteroidales associated with
depression (p=0.05, t-test for average difference in
OTU ratios in Table S1). At a low taxonomic level
(OTU), however, we could not identify any single OTU
showing significant correlation with depression after
false discovery correction.
As no single OTU showed significant correlation
with depression, we used multivariate PLS-DA analy-
ses to investigate whether we could detect correlation
patterns for the complete microbiota (Fig. 2A). The
multivariate model showed good sensitivity and spec-
ificity, correctly predicting 100% of the classified
depressed patients and 97% of the classified non-
depressed patients (misclassifying only one depressed
patient as non-depressed). Furthermore, the confound-
ing factor of depression medication did not seem to
influence the model, as all 10 depressed patients that
did not receive medication, in addition to the control
that received medication, were all correctly classified.
The redundancy of data was tested using cross-valida-
tion. The cross-validated model showed a sensitivity of
0.86 and a specificity of 0.47 (Fig. 2B). To evaluate the
robustness of the model, we also performed separate
analyses for the A and B parallels, with subsequent
correlations of the loadings. These analyses showed
good correspondence (Fig. 2C).
The OTU’s determined to be correlating with
depression according to the regression model (empiri-
cally defined as loadings >0.05 or <0.05) were not
evenly distributed among the different taxa (Fig. 2A).
There was a higher number of correlating OTU’s than
expected by chance in the Bacteroidetes phylum
(p=0.05, binominal test), while there was underrep-
resentation in the Lachnospiraceae family (p=0.003,
binominal test). At lower taxonomic levels, there were
clades within the genus Alistipes (p=0.007, binominal
test), and Oscillibacter (p=0.03 binominal test) that
showed overrepresentation of correlating OTU’s. How-
ever, no clades showed uniform positive or negative
correlations (Fig. 2A).
In support of the OTU-level correlation, binning the
data at the genus level (using QIIME) gave poor PLS-DA
classification, with misclassification of 22 of the 55
individuals. Furthermore, the cross-validation revealed
no redundancy in the data for this model. Due to the
low correlation at the genus level, we also tested the
direct correlation between the identified OTU’s at the
99% identity level. These analyses showed a cross-
validated sensitivity of 0.83 and a specificity of 0.37,
with none of the individuals being misclassified.
DISCUSSION
The most pronounced high-level correlation detected
in our dataset was a general underrepresentation of
Bacteroidetes related with depression. In both human
and animal studies, low Bacteroidetes levels have
previously been shown to be associated with obesity.
22
It has been previously suggested that there is a link
between obesity and depression through low grade
inflammation,
23
while we have recently established a
correlation between bacteria and low grade inflamma-
tion.
24
In our cohort, however, the BMI was only
slightly higher for the depressed patients compared to
the controls, so it is unlikely that obesity is a
confounding factor in our study.
For the low level taxonomic associations detected, a
recent study in which mice were subjected to stress
over an extended time period, the genus Alistipes was
one of the bacterial groups that showed the highest
increase in the stressed group.
25
Furthermore, Alistipes
has been found to be elevated in chronic fatigue
syndrome
26
and IBS.
27
It has been suggested that
Alistipes is associated with inflammation,
26
and there-
fore potentially linked to depression through inflam-
matory pathways.
3
For Oscillibacter, the type strain of
this genus has valeric acid as its main metabolic end
product.
28
Valeric acid structurally resembles GABA,
and has been shown to bind the GABAa receptor.
Therefore, it is possible that bacteria involved in
valeric acid production and/or metabolism could also
be associated with depression.
29
Knowledge about the
©2014 John Wiley & Sons Ltd1158
A. Naseribafrouei et al. Neurogastroenterology and Motility
role of valeric acid in the gut, however, is very sparse,
despite constituting one of the main short-chain fatty
acids.
The opposite correlations for closely related OTU’s
in our dataset were both striking and surprising. This
may indicate that certain phylogroups interfere with
particular mode-modulating pathways, but that the
effect of the interference represents a fine tuned
balance.
3
Furthermore, the lack of statistically signif-
icant univariate correlations may also indicate that
A
B
Figure 1 Phylogenetic tree of the 100 most dominant OTU’s. (A) The tree was constructed by the neighbor-joining algorithm using bootstrap for
testing the support of the branches. The numbers at the nodes show how many out of 1000 bootstrap trees that supported the particular branch (only
values >500 are shown). The coloring indicates if the OTU is positively (red) or negatively (green) associated with depression. (B) The ranking of the
OTU’s by the percentage of the complete dataset is shown.
©2014 John Wiley & Sons Ltd 1159
Volume 26, Number 8, August 2014 Human fecal microbiota and depression
interaction networks are important. These findings,
however, must be verified in larger cohorts due to the
risk of overseeing effects in small cohorts.
It has previously been shown that both Alistipes and
Oscillibacter levels can be modified through dietary
intervention. A diet high in easily fermentable oligo- or
mono-saccharides with a low healthy food diversity
index promoted the level of Alistipes.
30
In a mouse
feeding trial, the development of insulin resistance was
correlated with reduced levels of Oscillibacter follow-
ing a diet high in fat.
31
Assuming that the correlations
detected here contribute to depression, then the
potential of modulating depressive disorders through
dietary intervention may exist. However, given
that bacterial interaction networks are important in
shaping the gut microbiota, then the response to
interventions may be highly individual, and difficult
to predict.
For the factors recorded, medication is confounding
with depression. However, as all 10 non-medicated
patients, in addition to the medicated control, were
correctly classified in both the in-house and QIIME
A
BC
Figure 2 Partial least square discriminant analysis (PLS-DA) classification of depression based on the 100 most dominant OTU’s. (A) Loading plot for
the correlation of OTU’s on depression. The absolute value for correlations >0.5 are marked with a stippled line. (B) ROC plot for the estimated (black)
and the cross-validated classification (green). Red circles represent the optimal threshold between specificity and sensitivity. (C) Correlation for the
loadings derived from a model built separately from the A and B parallels.
©2014 John Wiley & Sons Ltd1160
A. Naseribafrouei et al. Neurogastroenterology and Motility
99% level OTU multivariate models, we believe that it
is unlikely that medication in itself can explain the
correlation patterns detected. Diet was not logged in this
study, but all the patients and controls were ethnic
Norwegians most likely consuming a traditional Nor-
wegian diet. The depressed patients were also tended to
by medical personnel to ensure proper diet.
Recent evidence has revealed the importance of
OTU binning and definitions in discovering biological
correlation patterns.
32,33
The results presented here
also support the importance of OTU binning at the
right taxonomic level, as binning at the genus level did
not reveal any correlations, while binning at the
99–99.5% identity level revealed relatively strong
correlations between gut microbiota and depression.
In conclusion, we found significant correlations
between gut microbiota and depression. The correla-
tions, however, were complex with opposite directions
for closely related OTU’s.
ACKNOWLEDGMENTS
The work was supported by internal funds from Hedmark
University College, Lillehammer University College, Innlandet
Hospital, and the Norwegian University for Life Sciences.
FUNDING
The study was supported with funding from Lillehammer
University College, Hedmark University College and Innlandet
Hospital trust.
CONFLICTS OF INTEREST
The authors have no competing interests.
AUTHOR CONTRIBUTION
AN performed the research; KH designed the study; EA, MS, AL,
and KR analyzed the data; KR and RW wrote the article with
contribution from all the authors.
REFERENCES
1 Kessler RC, Nelson CB, McGonagle
KA, Liu J, Swartz M, Blazer DG.
Comorbidity of DSM-III-R major
depressive disorder in the general
population: results from the US
National Comorbidity Survey. Br J
Psychiatry Suppl 1996; 30:17–30.
2 Uher R, McGuffin P. The moderation
by the serotonin transporter gene of
environmental adversity in the etiol-
ogy of depression: 2009 update. Mol
Psychiatry 2010; 15:18–22.
3 Foster JA, McVey Neufeld KA. Gut-
brain axis: how the microbiome influ-
ences anxiety and depression. Trends
Neurosci 2013; 36: 305–12.
4 Maes M. An intriguing and hitherto
unexplained co-occurrence: depres-
sion and chronic fatigue syndrome
are manifestations of shared inflam-
matory, oxidative and nitrosative
(IO&NS) pathways. Prog Neuropsy-
chopharmacol Biol Psychiatry 2011;
35: 784–94.
5 Dantzer R, O’Connor JC, Freund GG,
Johnson RW, Kelley KW. From
inflammation to sickness and depres-
sion: when the immune system sub-
jugates the brain. Nat Rev Neurosci
2008; 9:46–56.
6 Hestad KA, Aukrust P, Tonseth S,
Reitan SK. Depression has a strong
relationship to alterations in the
immune, endocrine and neural system.
Curr Psychiatry Rev 2009; 5:287–97.
7 Dowlati Y, Herrmann N, Swardfager
W, Liu H, Sham L, Reim EK, Lancto
ˆt
KL. A meta-analysis of cytokines in
major depression. Biol Psychiatry
2010; 67: 446–57.
8 Bercik P, Verdu EF, Foster JA, Macri
J, Potter M, Huang X, Malinowski P,
Jackson W et al. Chronic gastroin-
testinal inflammation induces anxi-
ety-like behavior and alters central
nervous system biochemistry in
mice. Gastroenterology 2010; 139:
2102–12 e2101.
9 de Kloet ER, Joels M, Holsboer F.
Stress and the brain: from adaptation
to disease. Nat Rev Neurosci 2005; 6:
463–75.
10 Walker EF, Diforio D. Schizophrenia:
a neural diathesis-stress model. Psy-
chol Rev 1997; 104: 667–85.
11 Beaton EA, Schmidt LA, Ashbaugh
AR, Santesso DL, Antony MM,
McCabe RE, Segalowitz SJ, Schulkin
J. Low salivary cortisol levels among
socially anxious young adults: preli-
minary evidence from a selected and
a non-selected sample. Pers Individ
Dif 2006; 41: 1217–28.
12 Holtzheimer PE 3rd, Nemeroff CB.
Future prospects in depression
research. Dialogues Clin Neurosci
2006; 8: 175–89.
13 Ait-Belgnaoui A, Durand H, Cartier
C, Chaumaz G, Eutamene H, Ferrier
L, Houdeau E, Fioramonti J et al.
Prevention of gut leakiness by a pro-
biotic treatment leads to attenuated
HPA response to an acute psycho-
logical stress in rats. Psychoneuroen-
docrinology 2012; 37: 1885–95.
14 Barrett E, Ross RP, O’Toole PW,
Fitzgerald GF, Stanton C. gamma-
Aminobutyric acid production by cul-
turable bacteria from the human
intestine. J Appl Microbiol 2012;
113: 411–7.
15 Bravo JA, Forsythe P, Chew MV,
Escaravage E, Savignac HM, Dinan
TG, Bienenstock J, Cryan JF. Ingestion
of Lactobacillus strain regulates emo-
tional behavior and central GABA
receptor expression in a mouse via
the vagus nerve. Proc Natl Acad Sci U
SA2011; 108: 16050–5.
16 Heijtz RD, Wang S, Anuar F, Qian
Y, Bjorkholm B, Samuelsson A,
Hibberd ML, Forssberg H et al. Nor-
mal gut microbiota modulates brain
development and behavior. Proc
Natl Acad Sci U S A 2011; 108:
3047–52.
17 Dinan TG, Cryan JF. Melancholic
microbes: a link between gut micro-
biota and depression? Neurogastroen-
terol Motil 2013; 25: 713–9.
18 Montgomery SA, Asberg M. A new
depression scale designed to be sensi-
tive to change. Br J Psychiatry 1979;
134: 382–9.
19 Vebø H, Case
´n C, Sekelja M, Nestes-
tog R, Ciemniejewska E, Perminov G,
Ricanec P, Vatn M. Microbiota analy-
sis in IBS and IBD/non-IBD patients
and normal subjects. UEG Week 2013
©2014 John Wiley & Sons Ltd 1161
Volume 26, Number 8, August 2014 Human fecal microbiota and depression
(UEGWEEK2013), Berlin, Germany,
2013, contribution P592 2013.
20 Rudi K, Zimonja M, Kvenshagen B,
Rugtveit J, Midtvedt T, Eggesbo M.
Alignment-independent comparisons
of human gastrointestinal tract
microbial communities in a multidi-
mensional 16S rRNA gene evolution-
ary space. Appl Environ Microbiol
2007; 73: 2727–34.
21 Rudi K, Zimonja M, Naes T. Align-
ment-independent bilinear multivari-
ate modelling (AIBIMM) for global
analyses of 16S rRNA gene phylog-
eny. Int J Syst Evol Microbiol 2006;
56: 1565–75.
22 Ley RE, Turnbaugh PJ, Klein S, Gor-
don JI. Microbial ecology: human gut
microbes associated with obesity.
Nature 2006; 444: 1022–3.
23 Stunkard AJ, Faith MS, Allison KC.
Depression and obesity. Biol Psychi-
atry 2003; 54: 330–7.
24 Troseid M, Nestvold TK, Rudi K,
Thoresen H, Nielsen EW, Lappegard
KT. Plasma lipopolysaccharide is clo-
sely associated with glycemic control
and abdominal obesity: evidence from
bariatric surgery. Diabetes Care 2013;
36: 3627–32.
25 Bangsgaard Bendtsen KM, Krych L,
Sorensen DB, Pang W, Nielsen DS,
Josefsen K, Hansen LH, Sorensen SJ
et al. Gut microbiota composition is
correlated to grid floor induced stress
and behavior in the BALB/c mouse.
PLoS ONE 2012; 7: e46231.
26 Fremont M, Coomans D, Massart S,
De Meirleir K. High-throughput 16S
rRNA gene sequencing reveals alter-
ations of intestinal microbiota in
myalgic encephalomyelitis/chronic
fatigue syndrome patients. Anaerobe
2013; 22:50–6.
27 Saulnier DM, Riehle K, Mistretta TA,
Diaz MA, Mandal D, Raza S, Weidler
EM, Qin X et al. Gastrointestinal
microbiome signatures of pediatric
patients with irritable bowel syn-
drome. Gastroenterology 2011; 141:
1782–91.
28 Katano Y, Fujinami S, Kawakoshi A,
Nakazawa H, Oji S, Iino T, Oguchi A,
Ankai A et al. Complete genome
sequence of Oscillibacter valericig-
enes Sjm18-20(T) (=NBRC 101213
(T)). Stand Genomic Sci 2012; 6:
406–14.
29 H€
olzl J, Godau P. Receptor bindings
studies with valeriana officinalis on
the benzodiazepine receptor. Planta
Med 1989; 55: 642–642.
30 Drescher LS, Thiele S, Mensink GB.
A new index to measure healthy food
diversity better reflects a healthy diet
than traditional measures. J Nutr
2007; 137: 647–51.
31 Qin J, Li Y, Cai Z, Li S, Zhu J, Zhang
F, Liang S, Zhang W et al. A metage-
nome-wide association study of gut
microbiota in type 2 diabetes. Nature
2012; 490:55–60.
32 Sekelja M, Berget I, Naes T, Rudi K.
Unveiling an abundant core microbi-
ota in the human adult colon by a
phylogroup-independent searching
approach. ISME J 2011; 5: 519–31.
33 Edgar RC. UPARSE: highly accurate
OTU sequences from microbial
amplicon reads. Nat Methods 2013;
10: 996–8.
SUPPORTING INFORMATION
Additional Supporting Information may be found in the online version of this article:
Figure S1. Distribution of number of reads per sample.
Figure S2. Heat map of the pairwise distances for the 100 most dominant OTU’s.
Table S1. RDP classification and distribution of the 100 most dominant OTU’s.
A. Naseribafrouei et al. Neurogastroenterology and Motility
©2014 John Wiley & Sons Ltd
1162
