Li et al.

Translational Psychiatry (2018)8:34

DOI 10.1038/s41398-017-0078-2 Translational Psychiatry

ARTICLE Open Access

Metabolite identification in fecal

microbiota transplantation mouse livers
and combined proteomics with chronic
unpredictive mild stress mouse livers
Bo Li1,2,3, Kenan Guo4, Li Zeng1,2, Benhua Zeng4, Ran Huo1,2,3, Yuanyuan Luo1,2,5, Haiyang Wang1,2, Meixue Dong1,2,
Peng Zheng1,2,6, Chanjuan Zhou1,2, Jianjun Chen1,2, Yiyun Liu1,2, Zhao Liu1,2, Liang Fang5, Hong Wei2 and
Peng Xie2,3,5,6,7

Abstract
Major depressive disorder (MDD) is a common mood disorder. Gut microbiota may be involved in the pathogenesis of
depression via the microbe–gut–brain axis. Liver is vulnerable to exposure of bacterial products translocated from the
gut via the portal vein and may be involved in the axis. In this study, germ-free mice underwent fecal microbiota
transplantation from MDD patients and healthy controls. Behavioral tests verified the depression model. Metabolomics
using gas chromatography–mass spectrometry, nuclear magnetic resonance, and liquid chromatography–mass
spectrometry determined the influence of microbes on liver metabolism. With multivariate statistical analysis, 191
1234567890():,;
1234567890():,;

metabolites were distinguishable in MDD mice from control (CON) mice. Compared with CON mice, MDD mice
showed lower levels for 106 metabolites and higher levels for 85 metabolites. These metabolites are associated with
lipid and energy metabolism and oxidative stress. Combined analyses of significantly changed proteins in livers from
another depression model induced by chronic unpredictive mild stress returned a high score for the Lipid Metabolism,
Free Radical Scavenging, and Molecule Transports network, and canonical pathways were involved in energy
metabolism and tryptophan degradation. The two mouse models of depression suggest that changes in liver
metabolism might be involved in the pathogenesis of MDD. Conjoint analyses of fecal, serum, liver, and hippocampal
metabolites from fecal microbiota transplantation mice suggested that aminoacyl-tRNA biosynthesis significantly
changed and fecal metabolites showed a close relationship with the liver. These findings may help determine the
biological mechanisms of depression and provide evidence about “depression microbes” impacting on liver
metabolism.

Introduction of disease and affects up to 15% of the general popula-

Major depressive disorder (MDD) is a debilitating tion1. Recent studies suggest that MDD is associated
mental disorder accounting for 12.3% of the global burden with neurotrophic alterations2, an imbalance in the
hypothalamic–pituitary–adrenal axis3, and glutamine
neurotransmitter system dysfunction4. Most research on
Correspondence: Hong Wei (weihong63528@163.com) or
Peng Xie (xiepeng@cqmu.edu.cn) depression focuses on changes to the brain, and few stu-
1
Institute of Neuroscience and the Collaborative Innovation Center for Brain dies have examined the effects of liver metabolism on
Science, Chongqing Medical University, Chongqing, China
2 depression.
Chongqing Key Laboratory of Neurobiology, Chongqing, China
Full list of author information is available at the end of the article
Bo Li, Kenan Guo, Li Zeng, and Benhua Zeng contributed equally to this work.

© The Author(s) 2018

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction
in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if
changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If
material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain
permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
Li et al. Translational Psychiatry (2018)8:34 Page 2 of 12

Accumulating evidence suggests that disruption of (Chongqing, China). Kunming (germ-free) mice were
gastrointestinal microbes is associated with the develop- obtained from the Experimental Animal Research Center
ment or exacerbation of mental disorders in humans, and at the Third Military Medical University and were kept in
the microbe–gut–brain axis may play a key role in main- flexible film gnotobiotic isolators until 6 weeks old and
taining brain health and stress responses5–7. weighing 30–40 g. Mice were housed in standard auto-
A mouse model of depression has been established claved polypropylene cages with access to food and water
using fecal microbiota transplantation (FMT), in which ad libitum under a 12-h dark–light cycle (light from 08:00
fecal matter from MDD patients is implanted into germ- AM until 08:00 PM), and at a constant temperature (23 ±
free mice8,9. The germ-free mouse, which is free from 1 °C) and relative humidity (50 ± 5%). Mice acclimatized
bacterial contamination, has been widely used to investi- for 2 weeks to standard experimental conditions prior to
gate interactions between the microbiota and host10. the commencement of experiments.
The liver is the main organ for substrate and energy
metabolism and plays an important role in oxidative FMT and sample collection
stress, glycogen storage, and the synthesis of secretory MDD patients were diagnosed following a structured
proteins. A previous study reported that liver diseases psychiatric interview using DSM-IV-TR criteria and 17
were associated with depression and suicide attempts11. items from the Hamilton depression rating scale. The
Recently, metabolomics has been used to define meta- FMT was carried out via randomly selecting 0.5 g of
bolic disorders and specific biomarkers of various disease feces from five MDD or healthy individuals under an
states12,13. We have applied metabolomics to the serum14, oxygen-free environment, and all 0.5 g samples were
urine15, and peripheral blood mononuclear cells16 of mixed in 7.5 mL of 0.9% saline to obtain suspension. Then
MDD patients, and the prefrontal cortex of a lipopoly- the microbiota was transplanted into germ-free mice in a
saccharide (LPS)-induced mouse model of depression17. flexible film gnotobiotic isolator. After 2 weeks, behavioral
During these studies, we identified five metabolites in tests were performed. On completion of behavioral tests,
urine from MDD patients uniquely produced by bacteria the livers of mice were immediately collected and stored
in the intestinal tract that were significantly decreased at −80 °C until metabolism analysis.
relative to healthy controls (CONs). Of note, some clinical
studies have reported that MDD induced a relatively high Gas chromatography–mass spectrometry
incidence of irritable bowel syndrome, a disease involving Twenty-four mouse livers (12 MDD and 12 CON) were
gut microbe disorders15,18. It is possible that MDD has an prepared for GC–MS metabolomics analysis. Briefly, a
association with intestinal microbe dysbiosis. Further, we 100-mg liver sample was derivatized with pyridine
analyzed fecal, serum, and hippocampal samples from a hydrochloride solution and N,O-bis(trimethylsilyl)tri-
depression microbe-induced mouse model, and found fluoroacetamide (including 1% trimethylchlorosilane)
that the metabolites might be associated with carbohy- before undergoing GC–MS analysis. Full details for
drate, amino acid, and nucleotide metabolism9. Based on derivatization and GC–MS conditions have been repor-
these findings, we considered “depression microbes” can ted19. Analysis used the exported NetCDF file format
affects metabolism in the gut tract and serum. Whether and TagFinder. As an internal standard, L-2-
microbiota dysbiosis impacts on the liver metabolism chlorophenylalanine (0.03 mg/mL; methanol configura-
profile and whether the liver plays an important role in tion) was used to normalize peak areas of extracted ions.
the microbe–gut–brain axis are not clear.
With diverse methodologies, the combination of dif- Nuclear magnetic resonance
ferent metabolomics platforms can identify more synth- Liver tissue (~ 100 mg) was mixed with 800 µL
esis metabolites than a single method19–21. In the current methanol–water (4:1, v/v) before being homogenized,
study, three metabolomics approaches using nuclear ultrasonic extraction for 5 min, stewing for 20 min at 4 °C,
magnetic resonance (NMR), gas chromatography–mass and centrifugation at 14,000 g for 10 min at 4 °C. Super-
spectrometry (GC–MS), and liquid chromatography–mass natant (200 µL) was transferred to a glass bottle for rapid
spectrometry (LC–MS) were combined to determine centrifugal concentrator volatile drying before being dis-
metabolomics profile alterations in the livers of a mouse solved in 500 µL heavy water in a NMR tube prior to
model of MDD. detection.
Proton spectra were detected using a Varian 600 spec-
Materials and methods trometer at an operating power of 599.925 MHz in 1H. A
Ethical considerations Carr–Purcell–Meiboom–Gill (recycle delay−90°−(τ−180°
Animal experiments were approved by the Third −τ)n−acquisition) pulse sequencer with a relaxation
Military Medical University (Chongqing, China) and the delay of 2.5 s, a mixing time of 100 m/s, a spectral width of
Ethical Committee of Chongqing Medical University 10 kHz, and a data point of 16 K was used cumulatively
Li et al. Translational Psychiatry (2018)8:34 Page 3 of 12

128 times. Discrimination between MDD and CON mice the network. The score is calculated using the right-tailed
was visualized using partial least-squares discriminant Fisher’s exact test and is based on hypergeometric
analysis. Coefficient loading plots of the model were used distribution.
to identify the spectral variables responsible for the
sample differentiation on the score plot. A correlation Results
coefficient of │r│ > 0.553 based on a p-level < 0.05 or Behavioral tests
variable importance in projection (VIP) value > 1.000 was Results of the behavioral tests are reported in our pre-
used as the cutoff value for statistical significance. vious research9. Briefly, immobility times for the forced
swimming test and the tail suspension test significantly
Liquid chromatography–mass spectrometry increased and the center motion distance for the open-
LC–MS preparation was performed as previously field test significantly decreased in “depression microbes”
described9. Briefly, 100 mg of liver sample was homo- mice compared with “healthy microbes” mice, only
genized with 20 µL internal standard (l-2-chloro-L-phe- behavioral significant changes mice used for further
nylalanine, 0.03 mg/mL; methanol configuration) and 800 analysis. Using FMT, we constructed a mouse model of
µL methanol–water solution (4/1, v/v) before ultrasonic depression.
extraction for 5 min, incubation for 20 min at 4 °C, and
centrifugation for 10 min at 14,000 g at 4 °C. Supernatant Metabolites showing a significant difference between
(200 µL) was transferred into a glass bottle for LC–MS MDD and CON mice
metabolomics analysis. Supernatant underwent ultra- Metabolites from 12 MDD and 12 CON mice were used
performance liquid chromatography–tandem mass spec- for OPLS-DA analysis. OPLS-DA score plots showed
trometry (UPLC-Q-TOF/MS). Mass spectrometric data distinct separation between MDD mice and CON mice
were collected using a Waters VION IMS Q-TOF mass using the three metabolomics approaches (LC–MS_pos:
spectrometer equipped with an electrospray ionization R2Y = 0.868, Q2 = 0.683; LC–MS_neg:R2Y = 0.798, Q2
source operating in either positive or negative ion mode. = 0.618; GC–MS: R2Y = 0.549, Q2 = −0.694; NMR: R2Y
Full details are provided in a previous study9. Orthogonal cum = 0.936, Q2 = 0.917) (Fig. 1). R2Y is the cumulative
partial least-squares discriminant analysis (OPLS-DA) model variation in Y, and Q2 is the cumulative predicted
was used to identify differential metabolites in MDD mice variation. Values for these parameters approaching
compared with CON mice. 1.0 indicate a stable model with predictive reliability.
In the current study, R2Y and Q2 values indicated
Metabolomics function and pathway analyses significant metabolic differences between MDD and
For significantly altered metabolites (p < 0.05 and VIP CON mice. The original total ion chromatograms,
> 1.0), pathway analyses was performed using MetaboA- typical base peak intensity chromatograms, and 1H
nalyst 3.0 (http://www.metaboanalyst.ca/) and Ingenuity Carr–Purcell–Meiboom–Gill NMR spectra are shown in
pathway analysis (IPA) software. MetaboAnalyst is a Supplementary Fig. 1. A 199-iteration permutation test
comprehensive web application for metabolomics confirmed that OPLS-DA models were not over-fitted and
data analysis and interpretation. Metabolomics pathway were valid (Supplementary Fig. 2).
analysis used several databases, including the Human From OPLS-DA analysis, a total of 191 significantly
Metabolome Database (HMDB; http://www.hmdb.ca/), different metabolites were identified between the MDD
Metlin (https://metlin.scripps.edu/), and the Kyoto and CON mice using the three metabolomics approaches
Encyclopedia of Genes and Genomes (KEGG; http://www. (106 decreased and 85 increased in the livers of
genome.jp/kegg/), and the MetaboAnalyst tool, which can MDD mice compared with CON mice). Details of the
identify the most significantly changed metabolism metabolites are shown in Table 1 and Supplementary
pathways. Table 1.

Molecular network analysis Classification of the significantly changed metabolites

Previous studies applying proteomics and metabo- Using HMDB and MID for classification of metabolites
nomics to the livers of chronic unpredictive mild stress according to their super class revealed that many belon-
(CUMS) mouse models of depression reported disturbed ged to the Lipid super class (linolenic acid, oxoproline,
lipid metabolism and immune regulation22,23. In the maltotriitol, arachidonic acid, 13-hydroxy-docosanoic
current study, we cross-analyzed the two mouse models acid) and the Amino acids, peptides and analogs super
of depression using IPA. IPA is an advanced bioinfor- class (alanine, isoleucine, glutamine, valine, iminodiacetic
matics software program used to analyze biological acid), as well as Carbohydrates and Carbohydrate Con-
pathways and functions of biomolecules of interest. The jugates (glycogen, glutathione, D-arabitol), Aliphatic
higher the score, the more relevant the molecules are to Acyclic Compounds (ethanolamine, trimethylamine
Li et al. Translational Psychiatry (2018)8:34 Page 4 of 12

Fig. 1 Orthogonal partial least-squares discriminant analysis (OPLS-DA) score plots and 1H nuclear magnetic resonance (NMR)
corresponding coefficient loading plots. a–c OPLS-DA score plots derived from ultra-performance liquid chromatography–tandem mass
spectrometry (UPLC-Q-TOF/MS) electrospray ionization (ESI) (+), UPLC-Q-TOF/MS ESI (−), and gas chromatography–mass spectrometry (GC–MS)
spectra of the major depressive disorder (MDD) group and control (CON) group. d OPLS-DA score plots derived from 1H Carr–Purcell–Meiboom–Gill
NMR spectra of liver extracts and corresponding coefficient loading plots e, f obtained from the CON group and the MDD group. e, f Show the
significance of metabolite variations between the two classes. Peaks in the positive direction indicate metabolites that are more abundant in MDD.
Metabolites more abundant in the CON group are shown as peaks in the negative direction. The key to assignment is shown in Supplementary Fig. 1

N-oxide, phosphocholine, urea) among others. A part of Metabolites analyzed between FMT and CON mice
metabolites were clustering analyzed and emerged sig- The differential metabolites and their respective fold-
nificantly different trends, especially in the Lipid super change were analyzed using MetaboAnalyst, KEGG, and
class (Fig. 2a). Lipid proportion >65% and Amino acids IPA to explore the potential effects of depression
nearly 10% in all metabolites, the number of each class microbes. We identified nine pathways with a p-value <
showed in Fig. 2b. 0.05 that were different in MDD mice compared with
Li et al. Translational Psychiatry (2018)8:34 Page 5 of 12

Table 1 Metabolites identified in livers extracts

Metabolite/super class VIPa FC rb HMDB Platform P-valuec Trendd

Aliphatic acyclic compounds

Ethanolamine 2.60 – – HMDB00149 NMR – ↓
Trimetlylamine oxide 2.63 – – HMDB00925 NMR – ↓
Phosphocholine 1.89 – – HMDB01565 NMR – ↑
Urea 3.85 1.76 – HMDB00294 GC–MS 0.04 ↑
Putrescine 1.16 1.25 – HMDB01414 GC–MS 0.03 ↑
Amino acids, peptides, and analogs
Alanine 1.46 – – HMDB00161 NMR – ↑
Glycine 1.08 – – HMDB00123 NMR – ↑
Glycerol 1.79 – 0.64 HMDB00125 NMR – ↑
Hypoxanthine – – 0.68 HMDB00157 NMR – ↑
Histidine – – −0.90 HMDB00177 NMR – ↓
Lysine 1.56 – – HMDB00182 NMR – ↑
Phosphocreatine – – 0.59 HMDB01511 NMR – ↑
Iminodiacetate 36.22 – −0.98 HMDB11753 NMR – ↓
Alanine 5.44 1.24 – HMDB00161 GC–MS 0.00 ↑
Proline 3.20 1.25 – HMDB00162 GC–MS 0.01 ↑
Isoleucine 2.93 1.45 – HMDB00172 GC–MS 0.01 ↑
Glutamine 3.01 0.46 – HMDB00641 GC–MS 0.00 ↓
Valine 3.80 1.39 – HMDB00883 GC–MS 0.02 ↑
3-Aminoisobutyric acid 4.22 3.21 – HMDB03911 GC–MS 0.00 ↑
Carbohydrates and carbohydrate conjugates
Glycogen 2.74 – – HMDB00131 NMR – ↑
β-Glucose 1.92 – – HMDB00516 NMR – ↓
Glutathione – – 0.58 HMDB00757 NMR – ↑
α-Glucose 3.55 – −0.68 HMDB03345 NMR – ↓
D-Arabitol 1.27 2.19 – HMDB00568 GC–MS 0.00 ↑
Galactinol 14.68 0.59 – HMDB05826 GC–MS 0.00 ↓
Nucleosides, Nucleotides, and Analogs
Inosine – – 0.58 HMDB00195 NMR – ↑
Uridine diphosphate–glucose – – 0.71 HMDB00286 NMR – ↑
Uridine – – 0.61 HMDB00296 NMR – ↑
Lipids
Linolenic acid 1.29 1.44 – HMDB01388 GC–MS 0.03 ↑
Oxoproline 6.14 1.35 – HMDB08177 GC–MS 0.00 ↑
Maltotriitol 6.15 0.64 – HMDB15224 GC–MS 0.00 ↓
Organic acids and derivatives
Lactate 2.42 – – HMDB62492 NMR – ↑
3-Hydroxybutyrate – – 0.63 HMDB00357 NMR – ↑
Li et al. Translational Psychiatry (2018)8:34 Page 6 of 12

Table 1 continued

Metabolite/super class VIPa FC rb HMDB Platform P-valuec Trendd

Lactic acid 3.00 1.14 – HMDB00190 GC–MS 0.01 ↑

Succinic acid 1.56 1.68 – HMDB00254 GC–MS 0.00 ↑
Organophosphorus compounds
O-phosphorylethanolamine 1.09 0.31 – HMDB00224 GC–MS 0.01 ↓
Phosphomycin 1.85 0.49 – HMDB14966 GC–MS 0.00 ↓
Others/unknown
Uracil – – −0.56 HMDB00300 NMR – ↓
Hypoxanthine 3.58 1.29 – HMDB00157 GC–MS 0.00 ↑
Xanthine 2.79 1.27 HMDB00292 GC–MS 0.00 ↑
D-(glycerol 1-phosphate) 1.81 0.38 – HMDB00126 GC–MS 0.01 ↓

VIP variable importance in projection, FC fold change, HMDB Human Metabolome Database. All identified metabolites were grouped by super class (based on HMDB
website information)
1
A VIP value > 1.000 was used as the cutoff value for statistical significance. “-” means the correlation coefficient of │r│ < 0.553 or VIP value of <1.000
2
Correlation coefficients: positive and negative signs indicate a positive or negative correlation in the concentrations. A correlation coefficient of │r│ > 0.553 was
used as the cutoff value for statistical significance based on discrimination significance at a p-level of 0.05 and 11 degrees of freedom (df)
3
p-Value was derived from two-tailed Student’s t-test
d
“↑” indicates higher levels in major depressive disorder (MDD), and “↓” indicates lower levels in MDD

CON mice (Supplementary Table 2). After p-values were Glycerophospholipid is usually subdivided into phos-
adjusted using Holm–Bonferroni corrections and the false phatidylethanolamine (PE), phosphatidylcholine (PC),
discovery rate, only glycerophospholipid metabolism sig- phosphatidic acid (PA), and phosphoinositides (PS). PC
nificantly changed. Canonical pathway overlapping ana- species not only protects cells and their organelles from
lyzed using IPA (Supplementary Fig. 3) included tRNA oxidative stress, but also is an essential component of
charging, glutamate receptor signaling. A total of 16 biomembranes. PE species has been identified as mod-
metabolites were identified as being significantly asso- ulator of inflammation24. The disturbance of glycer-
ciated with glycolysis and the tricarboxylic acid cycle (Fig. ophospholipid metabolism indicated oxidative stress,
2c). inflammatory cell membrane damage, and even apoptosis
in the liver during FMT and CUMS. The common path-
System integrated analysis in FMT mice way disturbance in liver may play an important role in
From system integrated analysis of significant metabo- depression.
lites in the feces, serum and hippocampal samples iden-
tified in a previous study9, the amino acids asparagine, Molecular network analysis of FMT mice using IPA
glutamine, isoleucine, proline, leucine, and glycine, which A total of 191 metabolites and their respective fold-
are involved in aminoacyl-tRNA biosynthesis, were most changes were subjected to molecular interaction network
significantly altered. Details of the KEGG pathways are analysis using IPA software. Lipid Metabolism, Small
shown in Fig. 3a. Molecule Biochemistry, and Cellular Compromise were
We also compared overlapping metabolites in different the most significantly changed network. A total of 21
regions of FMT mice (Fig. 3b), and found that nine metabolites, including L-glutamine, linolenic acid, lysine,
metabolites in the feces had a close association with the phosphorylcholine, urea, L-proline, and glycogen, were
liver, and are mainly involved in energy and lipid meta- associated with the network (Fig. 4).
bolism (Supplementary Fig. 4).
Molecular network analysis of FMT and CUMS mice by IPA
Combined analysis of FMT and CUMS mice In a study that undertook quantitative proteomics
Metabolites showing significant changes detected by analysis of livers from the CUMS mouse model of
LC–MS in livers from FMT and CUMS mice with mini- depression, a total of 66 proteins were reported to exhibit
mum overlapping are listed in Fig. 3c. In super class, significantly different expression22.
mainly of the metabolites belong to lipid. We analyzed We combined proteomics and metabolomics using IPA
metabolic pathways and glycerophospholipid metabolism to examine the two models of depression, which showed
was disturbed in FMT mice. Interesting to note that behavioral changes stemming from different mechanisms.
CUMS mice had the same change. The Lipid Metabolism, Free Radical Scavenging and
Li et al. Translational Psychiatry (2018)8:34 Page 7 of 12

Fig. 2 Data on significant metabolites and energy metabolism. a Clustering analysis different metabolites in the liver (major depressive disorder
(MDD) group vs. the control (CON) group). b Number of metabolites identified using the three complementary approaches in each super class. A
total of 191 metabolites were identified using gas chromatography–mass spectrometry (GC–MS) (blue), nuclear magnetic resonance (NMR) (red),
liquid chromatography–mass spectrometry (LC–MS) (green), or combined approaches (purple), and super classification was performed. c Summary of
the differential metabolites associated with glycolysis and the tricarboxylic acid (TCA) cycle

Molecule Transports network had a high score of 99, and acid, organic acids, tryptophan, and propionic acid, have
included high-density lipoprotein, low-density lipopro- been shown to influence behavior in animals26, con-
tein, the nuclear factor κB signaling pathway (Supple- jugated fatty acids, LPS, peptidoglycan, acylglycerols,
mentary Fig. 5). From canonical pathway analysis, we sphingomyelin, and cholesterol can affect intestinal per-
found that the overlap centered on Glycogen degradation meability and activate the intestine–brain–liver–neural
and Tryptophan Degradation (Supplementary Fig. 6). axis to regulate glucose homeostasis27.
Research on depression usually focuses on the central
Discussion nervous system and peripheral nervous system, rather
Depression is a widespread and debilitating mental than the liver. To the best of the authors’ knowledge, this
disorder that contributes to increased suicide rates and is the first study to apply untargeted metabolomics
has a heavy socioeconomic burden; however, little is approaches to the liver of a FMT mouse model of
known about its pathogenesis. The gut microbiota is the depression.
largest ecosystem in the body and affects numerous Liver has a unique vascular system and its blood mainly
physiological functions. The microbe–gut–brain axis is a comes from intestine through the portal vein28. Due to
communication system that integrates neural, hormonal, the system, liver is vulnerable to exposure to bacterial
and immunological signals and metabolites between the products. Despite profound interindividual variability,
gut and the brain25. Bacterial products, including lactic Gram-negative bacteria, such as Bacteroidetes,
Li et al. Translational Psychiatry (2018)8:34 Page 8 of 12

Fig. 3 Metabolite cross-talk in different regions and chronic unpredictive mild stress (CUMS) mouse model of depression. a Construction of
the aminoacyl-tRNA biosynthesis metabolism pathway in mice. The map was generated using the reference map from Kyoto Encyclopedia of Genes
and Genomes (KEGG) (http://www.genome.jp/kegg/). Green boxes show enzymatic activities. b Venn diagram indicating the number of significant
metabolites in different parts of major depressive disorder (MDD) mice. c Venn diagram indicating the number of significant metabolites in the livers
of the fecal microbiota transplantation (FMT) and CUMS mice models of depression. A common metabolite was hypoxanthine

Enterobacteriaceae, Alistipes, and Proteobacteria were metabolites identified included: (1) lipids (5-oxoproline
strongly increased in MDD patients compared with the and linolenic acid) and lipid metabolism-related mole-
healthy individuals29. Furthermore, increased LPS from cules (α-glucose, β-glucose, and glycerol); (2) amino acids
Gram-negative bacteria induced endothelial hyperper- (glycine, proline, valine, isoleucine, lysine, and histidine);
meability and “leaky gut” in MDD patients30,31. The “leaky and (3) other metabolites (O-phosphorylethanolamine,
gut” increased translocation of gut-derived bacterial pro- putrescine, and trimethylamine N-oxide).
ducts and then stimulated innate immune system, which The amino acids, including glutamine, glycine, lysine,
may involve in the pathophysiology of MDD32. In this valine, and isoleucine, which had significantly changed in
study, FMT from MDD patients may change gut perme- the mouse model of depression compared with CON
ability, hence, liver was exposed to bacterial products and mice. Some of these amino acids have an important role
presented disturbed metabolism profile. in brain function. Glutamine is a neurotransmitter that
In this study, we used three complementary techniques plays a crucial role in glutamatergic neurotransmission
in our untargeted metabolomics approach, these being through contact with astrocytes and neurons in the
GC–MS, NMR, and LC–MS. Using these techniques, 191 glutamine–glutamate cycle33. Preclinical research has
differential metabolites were distinguished between mice reported that glutamine deficiency in the prefrontal cortex
livers treated with “depression microbes” and “healthy and cerebellum increased depressive-like behavior34,35.
microbes”. Importantly, just one metabolite (alanine) was We found that glutamine significantly decreased in per-
identified by two approaches. The application of com- ipheral blood mononuclear cells and the cerebellum of
plementary approaches to metabolomics for the char- the mouse model of depression, suggesting that it could
acterization of liver metabolism is of value. The be a potential biomarker of depression35–37. We also
Li et al. Translational Psychiatry (2018)8:34 Page 9 of 12

Fig. 4 The most significantly changed network between major depressive disorder (MDD) and control (CON) groups. Metabolites in red
were upregulated while those in green were downregulated in MDD mice. Solid lines show direct physical interactions (such as binding) between
the two parties. Dotted lines show indirect interactions or regulations between the two parties

conducted metagenomics using murine cecum feces from level of blood cortisol and the microbe–gut–brain stress
the same batch samples in this study, and relative abun- response in pigs40. In previous studies, we found that
dance of the glutamate biosynthesis enzyme commission lysine levels decreased in the serum of MDD patients41,
numbers showed a contrary trend9. The decrease in glu- and in the cerebellum and prefrontal cortex of CUMS
tamine may suggest that microbes can affect liver gluta- mice35,42. Accordingly, increasing lysine may combat
mate levels and may modulate depressive-like behavior. depressive behavior and improve a person’s emotional
The lysine level increased in MDD mice compared with status.
CON mice. Recent studies report that lysine may affect Metabolism research reports that valine levels decrease
neurotransmitters associated with anxiety and stress in in peripheral blood mononuclear cells and the serum of
the rat38, whereas lysine fortification reduces anxiety and drug-naïve MDD patients14,43. Similarly, alanine was
lessens stress in humans39. Lysine is a constituent of the reported to decrease in the serum of MDD patients, and
serotonin receptor 4 antagonist, which can reduce the may be a potential urine biomarker for MDD patients44.
Li et al. Translational Psychiatry (2018)8:34 Page 10 of 12

In the current study, valine, isoleucine, and alanine Disturbances to oxidative stress are reported to be
increased in MDD mice liver. Levels of the three amino associated with the pathogenesis of MDD46. In the current
acids showed no difference among the serum, hippo- study, we found that the metabolites glycine and glu-
campus, and feces9. These results suggest that MDD tathione were significantly upregulated in MDD mice.
patients and FMT mice accompanied amino-acid meta- These metabolites are involved in oxidative stress. As
bolism disturbed and the trends were not exactly same in glutathione is the primary free radical scavenger in the
different parts. The liver is the center for substrate and brain, lower glutathione levels compromise central ner-
energy metabolism. Glucose is an important energy pro- vous system anti-oxidative activity47. Additionally, lower
vider, and in the current study, it was found to be levels of glutathione and oxidative damage could con-
decreased in MDD mice compared with CON mice. stitute early signaling events in cell apoptosis48. Glycine
Other metabolites involved in glycolysis and the tri- has been suggested to be a member of glutathione bio-
carboxylic acid cycle were markedly downregulated, such synthesis. In our previous metabolism research using the
as lactic acid (Fig. 4). Combined IPA and canonical same FMT mouse model of depression, we found that
pathway analysis revealed glycogen degradation. These glutathione levels decreased in the prefrontal cortex and
changes indicate a disturbance in energy metabolism. In the cecum, and glycine decreased in the hippocampus9.
agreement with previous research, a deficiency in circu- The metabolites in different parts may show different
lating glucose was observed with serum and urine meta- trends in diseased conditions. The elevated levels of glu-
bolomics studies of MDD patients14,44. Also, levels of tathione in the liver suggest increased anti-oxidation
glucose were markedly decreased in the prefrontal cortex activity, which may provide protection from oxidative
of LPS-induced mice17. Blass et al. reported two clinical stress in MDD mice.
cases in which patients with symptoms of depression Disturbances in lipid metabolism have been reported to
showed significant improvement in mood after 2 weeks be associated with geriatric depression in elderly
administration of supplemental malic acid and glucose45. patients49 and in rodent models of depression50. In the
Zheng et al. conducted a metabolomics analysis of feces current study, lipid-related molecules (a total of 126; 66%
and serum using the same FMT mouse model of of all metabolites) showed a tendency to change in the
depression and found increased levels of carbohydrate livers of MDD mice. These molecules included O-phos-
metabolites in MDD mice9. Combining previous results phorylethanolamine, glycerol, and arachidonic acid.
with the results of the current study suggest that Glycerol is the final product of triglyceride metabolism.
“depression microbes” may lead to a glucose disorder in These findings suggest that MDD mice may have lipid
liver. As the brain consumes 25% of the total glucose metabolism dysregulation. Combined proteomics of
available in the body, the decrease in liver glucose may CUMS mice livers showed a high score for the Lipid
result in depressive behavior. Metabolism, Free Radical Scavenging and Molecule
Phosphocreatine is a high-energy phosphate compound Transports network. Furthermore, common fecal and
abundant in the central nervous system and functions as a liver metabolites appeared to show disturbed lipid
transporter in cell energy exchange. It can transfer high- metabolism. Disturbances in the metabolism of the three
energy phosphate to ADP to provide ATP, generating major nutrients in the MDD liver may account for
creatine. Creatinine is a non-enzymatic by-product of the high comorbidity between MDD and metabolic
creatine and phosphocreatine. In the current study, syndrome51.
phosphocreatine showed a significant increase in the This study has some limitations. The findings and
livers of MDD mice compared with CON mice. Zheng conclusions drawn need to be treated cautiously because
et al. reported that creatine in the serum of MDD patients of the risk for overestimation with the relatively small
was significantly decreased14. We did not detect the sec- sample size. Second, we did not validate quantities and
ondary metabolite creatinine in the current study, but species of microbes. The potential mechanism behind the
metabolism research has reported that creatinine in the association between microbes and liver metabolism dis-
urine of MDD patients and in the cerebellum of CUMS turbance is unclear. Further, data were obtained from
mice decreased29,44. Upregulated phosphocreatine in the naive mice, additional studies should use an adult animal
liver may be a compensatory energy source, providing model of depression and conduct in vitro studies. Finally,
beneficial help to improve depressive behavior. we integrated information about metabolites trends from
The findings suggest that disturbances to glycolysis and different depressive models, organs, and regions as far as
the tricarboxylic acid cycle and the phosphocreatine–ATP possible. However, the potential relationships are not
pathway support previous research suggesting that a dis- clear and need further research. In future study, we will
turbance in energy metabolism may participate in the focus on the mechanism of single strains inducing liver
pathophysiology of depression, and the liver may play an metabolism disturbance and the relationship between gut
important role. microbiota and depression.
Li et al. Translational Psychiatry (2018)8:34 Page 11 of 12

Conclusion References
In this study, we developed and analyzed a FMT mouse 1. Reynolds, E. H. Brain and mind: a challenge for WHO. Lancet 361, 1924–1925
(2003).
model of depression, and the findings contribute to pre- 2. Guilloux, J. P. et al. Molecular evidence for BDNF- and GABA-related dys-
vious studies on the involvement of the gut microbiota in functions in the amygdala of female subjects with major depression. Mol.
psychiatric disorders. Using GC–MS, NMR, and LC–MS, Psychiatry 17, 1130–1142 (2012).
3. Haj-Mirzaian, A. et al. Lithium attenuated the depressant and anxiogenic effect
we identified 191 changed metabolites in the MDD mouse of juvenile social stress through mitigating the negative impact of interlukin-
liver compared with CON mice. The study mainly focused 1beta and nitric oxide on hypothalamic-pituitary-adrenal axis function. Neu-
on three major disturbances in metabolism, these being roscience 315, 271–285 (2016).
4. Mitchell, N. D. & Baker, G. B. An update on the role of glutamate in the
associated with lipid, amino acid, and energy metabolism. pathophysiology of depression. Acta Psychiatr. Scand. 122, 192–210 (2010).
Combined analyses with proteomics of CUMS mice livers 5. Rogers, G. B. et al. From gut dysbiosis to altered brain function and mental
showed a changed lipid network, suggestive of lipid dis- illness: mechanisms and pathways. Mol. Psychiatry 21, 738–748 (2016).
6. Hoban, A. E. et al. Regulation of prefrontal cortex myelination by the micro-
orders in the livers of mouse models of depression. biota. Transl. Psychiatry 6, e774 (2016).
Conjoint analyses of metabolites in different parts of the 7. Moloney, R. D., Desbonnet, L., Clarke, G., Dinan, T. G. & Cryan, J. F. The
FMT model suggested that aminoacyl-tRNA biosynthesis microbiome: stress, health and disease. Mamm. Genome 25, 49–74 (2014).
8. Zeng, L. et al. Microbiota modulates behavior and protein kinase C mediated
significantly changed and metabolites in feces were cAMP response element-binding protein signaling. Sci. Rep. 6, 29998 (2016).
most closely associated with the liver. The findings sug- 9. Zheng, P. et al. Gut microbiome remodeling induces depressive-like behaviors
gest that “depression microbes” can disturb the liver and through a pathway mediated by the host’s metabolism. Mol. Psychiatry 21,
786–796 (2016).
different parts of the body carrying out metabolic func- 10. Cryan, J. F. & Dinan, T. G. Mind-altering microorganisms: the impact of the gut
tions, and help determine the relationship between the microbiota on brain and behaviour. Nat. Rev. Neurosci. 13, 701–712 (2012).
liver, microbes, and depression. It is possible that treat- 11. Le Strat, Y., Le Foll, B. & Dubertret, C. Major depression and suicide attempts in
patients with liver disease in the United States. Liver. Int. 35, 1910–1916 (2015).
ment targeting microbes may be potential therapies for 12. Nicholson, J. K. & Lindon, J. C. Systems biology: metabonomics. Nature 455,
MDD. 1054–1056 (2008).
13. Qi Y. et al. Urinary metabolite markers of precocious puberty. Mol Cell Pro-
Acknowledgements teomics; 11, M111.011072 (2012).
This work was supported by the National Key R&D program of China (grant no. 14. Zheng, P. et al. Plasma metabonomics as a novel diagnostic approach for
2017YFA0505700), the National Key R&D program of China (grant no. major depressive disorder. J. Prote. Res. 11, 1741–1748 (2012).
2016YFC1307200), the National Youth Science Foundation (grant no. 15. Chen, J. J. et al. Divergent urinary metabolic phenotypes between major
81601207), the China Postdoctoral Science Foundation (grant no. depressive disorder and bipolar disorder identified by a combined GC-MS and
2017M612923), the Natural Science Foundation Project of China (grant no. NMR spectroscopic metabonomic approach. J. Prote. Res. 14, 3382–3389
81701360), the Chongqing Science & Technology Commission (CSTC2014kjrc- (2015).
qnrc10004), and the Chongqing Science & Technology Commission 16. Li, J. et al. Peripheral blood mononuclear cell-based metabolomic profiling of
(CSTC2017JCYJA0207). We thank Alexander Pishief, LLB, BBmedSc, from Liwen a chronic unpredictable mild stress rat model of depression. Mol. Biosyst. 10,
Bianji, Edanz Group China (www.liwenbianji.cn/ac), for editing the English text 2994–3001 (2014).
of a draft of this manuscript. 17. Wu, Y. et al. Metabolomic analysis reveals metabolic disturbances in the
prefrontal cortex of the lipopolysaccharide-induced mouse model of
Author details depression. Behav. Brain. Res. 308, 115–127 (2016).
1
Institute of Neuroscience and the Collaborative Innovation Center for Brain 18. Ladep, N. G., Obindo, T. J., Audu, M. D., Okeke, E. N. & Malu, A. O. Depression in
Science, Chongqing Medical University, Chongqing, China. 2Chongqing Key patients with irritable bowel syndrome in Jos, Nigeria. World J. Gastroenterol.
Laboratory of Neurobiology, Chongqing, China. 3Key Laboratory of Clinical 12, 7844–7847 (2007).
Laboratory Diagnostics (Ministry of Education), Chongqing, China. 19. Chen, J. J. et al. Combined application of NMR- and GC-MS-based metabo-
4
Department of Laboratory Animal Science, College of Basic Medical Sciences, nomics yields a superior urinary biomarker panel for bipolar disorder. Sci. Rep.
Third Military Medical University, Chongqing, China. 5Department of 4, 5855 (2014).
Neurology, Yongchuan Hospital, Chongqing Medical University, Chongqing, 20. Bouatra, S. et al. The human urine metabolome. PLoS. ONE 8, e73076 (2013).
China. 6Department of Neurology, The First Affiliated Hospital of Chongqing 21. Li, Y. et al. Metabonomics study of essential hypertension and its chinese
Medical University, Chongqing, China. 7South Australian Health and Medical medicine subtypes by using gas chromatography-mass spectrometry and
Research Institute, Mind and Brain Theme, Flinders University, Adelaide, SA, nuclear magnetic resonance spectroscopy. Evid. Based Complement. Altern.
Australia Med. 2013, 625906 (2013).
22. Wu, Y. et al. Quantitative proteomics analysis of the liver reveals immune
regulation and lipid metabolism dysregulation in a mouse model of
Conflict of interest
depression. Behav. Brain. Res. 311, 330–339 (2016).
The authors declare that they have no conflict of interest.
23. Jia, H. M. et al. Chronic unpredictive mild stress leads to altered hepatic
metabolic profile and gene expression. Sci. Rep. 6, 23441 (2016).
24. Devaux, P. F. Static and dynamic lipid asymmetry in cell membranes. Bio-
Publisher's note chemistry 30, 1163–1173 (1991).
Springer Nature remains neutral with regard to jurisdictional claims in 25. Mayer, E. A. Gut feelings: the emerging biology of gut-brain communication.
published maps and institutional affiliations. Nat. Rev. Neurosci. 12, 453 (2011).
26. Hanstock, T. L., Mallet, P. E. & Clayton, E. H. Increased plasma D-lactic acid
Supplementary Information accompanies this paper at https://doi.org/ associated with impaired memory in rats. Physiol. Behav. 101, 653–659 (2010).
10.1038/s41398-017-0078-2. 27. Cabou, C. et al. Central insulin regulates heart rate and arterial blood flow: an
endothelial nitric oxide synthase-dependent mechanism altered during dia-
Received: 8 August 2017 Revised: 1 November 2017 Accepted: 13 betes. Diabetes 56, 2872–2877 (2007).
November 2017 28. Seki, E. & Schnabl, B. Role of innate immunity and the microbiota in liver
fibrosis: crosstalk between the liver and gut. J. Physiol. 590, 447–458 (2012).
Li et al. Translational Psychiatry (2018)8:34 Page 12 of 12

29. Caballero, S. & Pamer, E. G. Microbiota-mediated inflammation and anti- 40. Srinongkote, S., Smriga, M., Nakagawa, K. & Toride, Y. A diet fortified with L-
microbial defense in the intestine. Annu. Rev. Immunol. 33, 227–256 (2015). lysine and L-arginine reduces plasma cortisol and blocks anxiogenic response
30. Kiecolt-Glaser, J. K., Derry, H. M. & Fagundes, C. P. Inflammation: depression to transportation in pigs. Nutr. Neurosci. 6, 283–289 (2003).
fans the flames and feasts on the heat. Am. J. Psychiatry 172, 1075–1091 41. Xu, H. B. et al. Potential clinical utility of plasma amino acid profiling in
(2015). the detection of major depressive disorder. Psychiatry Res. 200, 1054–1057
31. Chatterjee, A. et al. Heat shock protein 90 inhibitors attenuate LPS-induced (2012).
endothelial hyperpermeability. Am. J. Physiol. Lung. Cell. Mol. Physiol. 294, 42. Liu, L. et al. The identification of metabolic disturbances in the prefrontal
755–763 (2008). cortex of the chronic restraint stress rat model of depression. Behav. Brain. Res.
32. Maes, M., Kubera, M. & Leunis, J. C. The gut-brain barrier in major depression: 305, 148–156 (2016).
intestinal mucosal dysfunction with an increased translocation of LPS from 43. Liu, M. L. et al. Severe disturbance of glucose metabolism in peripheral blood
gram negative enterobacteria (leaky gut) plays a role in the inflammatory mononuclear cells of schizophrenia patients: a targeted metabolomic study. J.
pathophysiology of depression. Neuro. Endocrinol. Lett. 29, 117–124 (2008). Transl. Med. 13, 226 (2015).
33. Eid, T., Tu, N., Lee, T. S. & Lai, J. C. Regulation of astrocyte glutamine synthetase 44. Zheng, P. et al. Identification and validation of urinary metabolite biomarkers
in epilepsy. Neurochem. Int. 63, 670–681 (2013). for major depressive disorder. Mol. Cell. Proteom. 12, 207–214 (2013).
34. Lee, Y. et al. Glutamine deficiency in the prefrontal cortex increases depressive- 45. Blass, J. P. Nutritional supplement for cerebral metabolic insufficiencies. US
like behaviours in male mice. J. Psychiatry Neurosci. 38, 183–191 (2013). (2003).
35. Shao, W. H. et al. Combined metabolomics and proteomics analysis of major 46. Michel, T. M., Pulschen, D. & Thome, J. The role of oxidative stress in depressive
depression in an animal model: perturbed energy metabolism in the chronic disorders. Curr. Pharm. Des. 18, 5890–5899 (2012).
mild stressed rat cerebellum. OMICS 19, 383–392 (2015). 47. Gawryluk, J. W. et al. Prefrontal cortex glutathione S-transferase levels in
36. Li, J. et al. Potential antidepressant and resilience mechanism revealed by patients with bipolar disorder, major depression and schizophrenia. Int. J.
metabolomic study on peripheral blood mononuclear cells of stress resilient Neuropsychopharmacol. 14, 1069–1074 (2011).
rats. Behav. Brain. Res. 320, 12–20 (2016). 48. Schulz, J. B., Lindenau, J., Seyfried, J. & Dichgans, J. Glutathione, oxidative stress
37. Yang, P. et al. Alterations of amino acid level in depressed rat brain. Korean J. and neurodegeneration. Eur. J. Biochem. 267, 4904–4911 (2000).
Physiol. Pharmacol. 18, 371–376 (2014). 49. Paige, L. A., Mitchell, M. W., Krishnan, K. R., Kaddurah-Daouk, R. & Steffens, D. C.
38. Smriga, M. & Torii, K. L-lysine acts like a partial serotonin receptor 4 antagonist A preliminary metabolomic analysis of older adults with and without
and inhibits serotonin-mediated intestinal pathologies and anxiety in rats. depression. Int. J. Geriatr. Psychiatry 22, 418–423 (2007).
Proc. Natl. Acad. Sci. USA 100, 15370–15375 (2003). 50. Chen, G. et al. Amino acid metabolic dysfunction revealed in the
39. Smriga, M., Ghosh, S., Mouneimne, Y., Pellett, P. L. & Scrimshaw, N. S. Lysine prefrontal cortex of a rat model of depression. Behav. Brain. Res. 278, 286–292
fortification reduces anxiety and lessens stress in family members in eco- (2015).
nomically weak communities in Northwest Syria. Proc. Natl. Acad. Sci. USA 101, 51. Vaccarino, V. et al. Depression, the metabolic syndrome and cardiovascular risk.
8285–8288 (2004). Psychosom. Med. 70, 40–48 (2008).