Received: 29 March 2020 | Revised: 7 July 2020 | Accepted: 13 July 2020

DOI: 10.1111/aji.13307

ORIGINAL ARTICLE

Correlation of fecal metabolomics and gut microbiota in mice

with endometriosis

Zhexin Ni1 | Shuai Sun1 | Yanli Bi1 | Jie Ding1 | Wen Cheng1 | Jin Yu1 | Ling Zhou1 |
Mingqing Li2 | Chaoqin Yu1

1
Department of Gynecology of Traditional
Chinese Medicine, Changhai Hospital, Naval Abstract
Medical University, Shanghai, China Problem: Endometriosis (EMS) is a chronic inflammatory disease with unclear patho-
2
Shanghai Key Laboratory of Female
genesis. Three studies have uncovered the influence of gut microbiota on mice with
Reproductive Endocrine Related Diseases,
Hospital of Obstetrics and Gynecology, EMS, but no study has investigated the characteristics of fecal metabolomics to de-
Fudan University, Shanghai, China
termine some important clues on EMS. This research aims to uncover the interaction
Correspondence between fecal metabolomics and gut microbiota in EMS mice.
Mingqin Li, Shanghai Key Laboratory of
Method of study: Female C57BL/6J mice were used to construct the EMS model.
Female Reproductive Endocrine Related
Diseases, Hospital of Obstetrics and Non-target metabolomics was applied to detect the fecal metabolites of EMS mice.
Gynecology, Fudan University, Shanghai,
The 16s rRNA sequencing was used for clarifying the composition of the gut mi-
China.
Email: [email protected] crobiota. The functional characteristics of gut microbiota were analyzed using the
Chaoqin Yu, Department of Gynecology PICRUSt. The receiver operator characteristic curve (ROC) analysis was utilized for
of Traditional Chinese Medicine, Changhai determining the potential important differential metabolites, and the Spearman cor-
Hospital, Naval Medical University,
Shanghai, China. relation coefficient was applied for expressing the correlation between the important
Email: [email protected] differential metabolites and gut microbiota.
Funding information Results: A total of 156 named differential metabolites were screened. The diversity
This work was supported by the National and the abundance of gut microbiota in EMS mice decreased. Eleven pathways were
Natural Science Foundation of China (grant
numbers, 81703874, 81774352). involved in the differential metabolites and the functional prediction of gut micro-
biota, among which the second bile acid biosynthesis and alpha-linolenic acid (ALA)
metabolism were the significant enrichment pathways. The increased abundance of
chenodeoxycholic and ursodeoxycholic acids and the decreased abundance of ALA
and 12,13-EOTrE were found in the feces of EMS mice.
Conclusion: The abnormal fecal metabolites, which are influenced by dysbacteriosis,
may be the characteristics of EMS mice and can be the potential important indices to
distinguish the disease.

KEYWORDS

endometriosis, intestines, metabolomics, microbiota

Zhexin Ni and Shuai Sun contributed equally to this work.

© 2020 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd

Am J Reprod Immunol. 2020;00:e13307.  wileyonlinelibrary.com/journal/aji | 1 of 14

https://doi.org/10.1111/aji.13307
2 of 14 | NI et al.

1 | I NTRO D U C TI O N 2 | M E TH O DS A N D M ATE R I A L S

Endometriosis (EMS) is an estrogen-dependent disease character- 2.1 | Experimental animals

ized by endometrial glands and stroma implanted and growing out-
side the uterine, thereby covering the mucosa and the myometrium.1 Six-week-old healthy female C57BL/6J mice were purchased from
The main symptoms are dysmenorrhea, intercourse pain, chronic Beijing Weitong Lihua Experimental Animal Technology Co., Ltd. The
pelvic pain, and infertility. 2 Globally, EMS is a fairly common gyneco- experimental animal production license number is SCXK (Beijing) 2016-
logical disease that affects nearly 10% of women in the childbearing 0006. The estrus period was monitored using a vaginal smear every
age.3 Its pathogenesis has not yet been clarified, and the theory of morning for two weeks, and the following experiments were performed
4
retrograde menstruation is currently widely accepted. However, on mice with normal estrous cycles. Mice were randomly divided into
90% of women have menstrual reflux, and only 6%–10% develop the control (CO) and the EMS model (EM) groups and bred under SPF
EMS,5 which indicates that the occurrence and the development of conditions in Shanghai Changhai Hospital Animal Center with 12 hours
EMS is also related to other factors. light and 12 hours darkness. Sterile dressings were changed once a
The gut microbiota is a collection of microorganisms that exists week, and sterile nutrient pellets and sterile water were freely con-
in the human gastrointestinal tract. Under normal circumstances, sumed for 24 hours. All experimental operations were approved by the
the gut microbiota regulates the stability of the microecological en- Animal Care and Use Committee of Shanghai Changhai Hospital.
vironment, which can inhibit the reproduction of pathogenic bacte-
ria and break down harmful substances, to play an important role in
maintaining the host's homeostasis.6,7 However, the disorder of gut 2.2 | Construction of the EMS mouse model
microbiota and the multiplication of pathogenic bacteria lead to the
destruction of the intestinal mucosal barrier and the translocation The construction of an EMS mouse model was described in previous
of bacteria and endotoxins, causing the development of a variety studies.13,20 Briefly, after 2 weeks of acclimation, mice were injected
8,9
of chronic inflammatory diseases. Local inflammation increases subcutaneously with estrogen solution on days 1, 4, and 7, and endome-
intestinal permeability and further exacerbates intestinal mucosal trial fragments were transplanted on day 8. The donor mice were anes-
damage and flora disorder. EMS is an inflammation-related disease, thetized with ether, sacrificed through neck removal, and then immersed
and the gut microbiota disorder is one of the important causes of in 75% alcohol for 3 minutes. The uterus was obtained by dissection and
chronic inflammation. Previous studies have shown that the gut mi- cut along the longitudinal axis of the uterus. The endometrial layer was
crobiota is affected by estrogen and can significantly affect estro- peeled and cut into 1 mm × 1 mm × 1 mm fragments. Endometrial de-
gen levels.10,11 Therefore, the occurrence and development of EMS bris was aspirated with syringes, and the donor and recipient mice were
may be affected by the gut microbiota. Bailey and Coe have found implanted intraperitoneally at a ratio of 1:2. The estrogen solution was
that the gut microbiota distribution of rhesus monkeys with EMS injected subcutaneously on the day after transplantation and on days 11
is different from that of the healthy group.12 Yuan et al have de- and 14 of modeling. After three weeks of natural growth, the mice were
tected that the composition of the gut microbiota in mice with EMS dissected, and specimens were collected. The endometrial transplanta-
is different from that of mice in the sham-operated group 42 days tion process was performed under sterile conditions. No mice died dur-
after the establishment of the models.13 Chadchan et al have found ing the experiment. The control group was intraperitoneally injected
that broad-spectrum antibiotics or metronidazole reduces the EMS with saline with adipose tissue under the same conditions.
lesions and can be recovered through the transplantation of feces
from EMS mice.14
The gut microbiota can affect the immune environment in the 2.3 | Collection of feces
intestine of the host through metabolites and the host's neural func-
tion.15,16 In addition to helping the host metabolize nutrients, the gut Mouse feces were obtained at the estrus interval three weeks after
microbiota produces a variety of metabolites. Some of these me- modeling. After anesthesia with ether, the mice were sacrificed by neck
tabolites are absorbed by the host's intestinal wall,17 and the rest is removal and immersed in 75% alcohol for 3 minutes. The mice were
mixed in solid feces and excreted by the host.18 Non-targeted me- dissected on a sterilization table to obtain the feces from the cecum
tabolomics can detect substances in the fecal supernatant, thereby segment of the mice. Fecal particles were collected in sterile cryotubes,
providing a new direction for the diagnosis and the treatment evalu- immediately transferred to liquid nitrogen, and stored at −80°C.
ation of diseases non-invasively.19
In this study, non-targeted mass spectrometry and 16s
rRNA sequencing were used to explore the characteristics 2.4 | Liquid chromatography-mass spectrometry
and the correlation of fecal metabolomics and gut microbi- (LC–MS) detection
ota in the EMS mouse model to find potential metabolites
and further provide research basis for the diagnosis and the treat- The feces of six mice (50 mg) were randomly sampled from each
ment of EMS. group. The extraction solution (400 μL, methanol:water = 4:1) was
NI et al. | 3 of 14

added and crushed using a high-throughput tissue disrupter (50 Hz, phase B was acetonitrile/isopropanol (1:1, contains 0.1% formic
6 minutes) at −20°C. The solution was vortexed for 30 s and soni- acid). The flow rate, injection volume, and column temperature were
cated on ice for 30 minutes (5℃, 40 KHz). The sample was stored at 0.40 mL/min, 20 μL, and 40°C, respectively. The sample mass spec-
−20°C for 30 minutes and centrifuged (13 000 g) at 4°C for 15 min- trum signal was collected using positive and negative ion scanning
utes. The supernatant was transferred to the LC–MS injection vial mode and ion spray voltage. The quality control samples mixed with
for analysis (AB SCIEX, UPLC-TripleTOF system). The column was a all test samples in this study were used to evaluate the stability of
BEH C18 column (100 mm × 2.1 mm i.d., 1.7 μm; Waters). The mo- the analytical system during operation to ensure the reliability of the
bile phase A was water (containing 0.1% formic acid), and the mobile results.

F I G U R E 1 Analysis of the differential metabolites between the two groups. A, Principal component analysis (PCA). The distance of each
coordinate point represents the degree of aggregation and dispersion between samples. A close distance indicates high similarity between
the samples. PC1 and PC2 represent the contribution values of the first and second principal components, respectively. B, Orthogonal
partial least squares discrimination analysis (OPLS-DA). Through orthogonal rotation, the information irrelevant to the group is filtered,
which can better distinguish the differences between groups and improve the efficiency of the model. Comp1 represents the first predicted
principal component decomposition degree, and orthogonal Comp1 represents the first orthogonal component decomposition degree. C,
OPLS-DA model validation. The abscissa represents the replacement reservation degree of the replacement test. The ordinate represents
the values of R2 (green dot) and Q2 (blue triangle), and the two dashed lines represent the regression lines of R2 and Q2. D, Volcano map
of differential metabolites. The abscissa is the multiple change value in the expression difference of metabolites between the two groups,
and the ordinate is the statistical test value of the expression difference of metabolites. A high p value indicates significant expression
difference. The abscissa and ordinate values are all logarithmically processed. Each point in the figure represents a specific metabolite
4 of 14 | NI et al.

2.5 | Data processing and analysis correction, and peak alignment. Only the variables with non-zero
values above 80% in any set of samples were retained, and the mini-
The raw data of the mass spectrum signal was imported into the me- mum values in the original matrix filled the missing values. The total
tabolomics processing software Progenesis QI (Waters Corporation) peaks were normalized, whereas the variables with relative stand-
for baseline filtering, peak identification, integration, retention time ard deviation of QC samples ≥30% were deleted. The data were
NI et al. | 5 of 14

F I G U R E 2 Analysis of differential metabolite-related KEGG pathway between the two groups. A, Metabolites with variable important
in projection (VIP) > 2. Each column and row represent a sample and a metabolite, respectively, and the color represents the relative
content of the metabolite in this group of samples. The right side is the VIP bar graph of the metabolite, and the bar length represents the
contribution value of the metabolite to the difference between the two groups. The color of the bar is expressed using the P value. B, Pie
chart of HMDB subclass compounds. The different colors in each pie represent different HMDB classifications, and the area represents the
relative proportion of metabolites in the classification. C, Classification map of the KEGG pathway related to differential metabolites. The
ordinate is the name of the KEGG pathway, and the abscissa is the number of metabolites annotated to the pathway. The color bar on the
right represents that the KEGG types belong to KEGG category. D, Column chart of the KEGG enrichment analysis. The abscissa represents
pathway name, and the ordinate represents enrichment rate. The color gradient of the column indicates the significance of enrichment.
***P < .001, **P < .01, *P < .05. E, Bubble chart of KEGG topological analysis. Each bubble represents a KEGG pathway. The horizontal axis
represents the relative importance of metabolites in the pathway. The vertical axis represents the significance of metabolites in pathway
enrichment. The bubble size represents the impact value, and a large bubble indicates high importance

subjected to log10 transformation to obtain the data matrix that was discriminant analysis (LDA) on the two groups to identify the species
finally used for subsequent analysis. After using the Progenesis QI at the genus level that significantly affected the sample division. The
to search and identify the database, the information from the mass PICRUSt software was used to standardize the OTU abundance.
spectrometry was matched with the metabolic database (http:// According to the OTU abundance and the corresponding greengene
www.hmdb.ca/; https://metlin.scrip​ps.edu/), and the data were ana- id, the KEGG data (http://www.genome.jp/kegg/) library was used
lyzed on the free online platform of the Majorbio Cloud Platform to obtain the pathway level 3 information, and the path abundance
(www.major​bio.com). was calculated in accordance with the OTU abundance.
The principal component analysis (PCA) was used to observe the
overall distribution between samples and the degree of dispersion
between groups. The orthogonal partial least squares discrimina- 2.7 | Statistical methods
tion analysis (OPLS-DA) was used to distinguish the overall differ-
ences in metabolic profiles and the differential metabolites between The Excel table was used to screen for the same pathway be-
groups. In OPLS-DA, metabolites with variable important in projec- tween the differential metabolite-related KEGG pathway and the
tion (VIP) > 1 were set as difference variables. The method of 200 gut microbiota-related KEGG pathway. Differential metabolites
replacement tests was used to investigate the fitting effect of the were found in these pathways on the Majorbio platform (www.
OPLS-DA model and prevent the model from overfitting. The dif- Major​bio.com). The receiver operator characteristic curve (ROC)
ferential metabolites between the groups were selected (VIP > 1, analysis was used to calculate the area under the curve (AUC) and
P < .05) by using Student's t test combined with the multivariate determine the degree of influence of metabolites on the discrimi-
analysis method, and the volcano map of the differential metabolites nation between groups to discover the important metabolites.
was made. Metabolites with AUC values greater than 0.5 were set as poten-
tially important differential metabolites. The R language (pheat-
map package) was used to calculate the Spearman correlation
2.6 | 16s rRNA sequencing and data processing coefficient between the key metabolites and the genus bacteria,
and the resulting numerical matrix was displayed using a heat map.
On the basis of the LC–MS detection, the feces of five mice were The quantitative data of normal distribution were expressed using
randomly selected from each group. The 16s rRNA sequencing mean with standard deviation and analyzed using Student's t-test.
methodology was described in our previous research. 21 Sequencing Non-normally distributed quantitative data were analyzed using
was performed using the Illumina's Miseq PE300 platform (Shanghai the Wilcoxon rank-sum test, and fdr was performed on p-values.
Meiji Biomedical Technology Co., Ltd.). The mothur software (1.30.1) The SPSS 21.0 software was used for analysis. Two-tailed P < .05
was used to analyze the gut microbiota alpha diversity index (Sobs, was considered statistically significant.
Shannon). Beta diversity was reflected by the three-dimensional
principal coordinate analysis (PCoA). The ANOSIM analysis was
used to determine whether the grouping was meaningful, and the 3 | R E S U LT S
Jensen-Shannon distance between samples was calculated on the
basis of the relative abundance of the bacteria at the genus level to 3.1 | Characteristics of feces metabolomics in EMS
complete the grouping analysis. The Wilcoxon rank-sum test and the mice
multiple test corrections (fdr) were used to analyze the differences
in species composition at the phylum and genus levels between The fecal metabolites of the two groups of mice were analyzed using
the two groups. The LEfSe (http://hutte​nhower.sph.harva​rd.edu/ LC–MS. From the results of the PCA analysis, the dispersion degree
galax​y/root?tool_id=lefse_upload) was used to perform the linear of the EM group is significantly greater than that of the CO group,
6 of 14 | NI et al.

and a definite difference exists between the two groups (Figure 1A). A total of 296 differential metabolites (VIP > 1, P < .05; Figure 1D)
Further analysis of OPLS-DA (Figure 1B) and model validation were detected between the two groups, including 156 named me-
(Figure 1C) show that after removing the unnecessary factors, the tabolites. A total of 24 differential metabolites with VIP > 2, such as
prediction ability of the model is better (R2X = 0.517, R2Y = 0.996, dehydrocarpaine I, maslinic acid, goyaglycoside g, are displayed due
Q2 = 0.776), and differences in metabolites exist between the two to the excessive quantity (Figure 2A). Compared with the informa-
groups. tion in the database of HMDB compounds, the four types with the
NI et al. | 7 of 14

F I G U R E 3 Composition of gut microbiota in two groups. A, Rank–abundance curve. The abscissa represents the ranking level of
OTU number, and the ordinate represents the relative percentage content of the OTU number. The abscissa position of the extension
endpoint of the sample curve is the number of OTU. B, 3D-PCoA analysis. The X, Y, and Z axes represent three selected principal axes,
and the percentage represents the explanatory value of the principal axes for the differences of sample composition. Close sample points
indicate similarity of the species composition. C, Column chart of species composition at the phylum level. D, Column chart of top 20
abundance species at the genus level. The abscissa is grouping, and the ordinate is the proportion of species in the sample. The columns
of different colors represent different species, and the length of the columns represents the proportion of the species. E, Venn map of
species at the genus level. In the above figure, different colors represent different groups, the number of overlapping parts represents the
number of common bacteria in the two groups, and the number of non-overlapping parts represents the number of bacteria unique to the
corresponding group. The figure below shows the column chart of the total number of bacteria in each group at the genus level. Pie charts of
the abundance structure of unique bacteria in the (F) CO and the (G) EM groups

highest abundance are amino acids, peptides, and analytes (13.79%), 3.3 | Characteristic bacteria and functional
triterpenoids (6.9%), fatty acids and compounds (5.75%), and glyco- prediction of gut microbiota in EMS mice
phorophosphocholines (5.75%, Figure 2B).
In terms of the related KEGG pathways, 17 types of KEGG path- Ten characteristic bacteria with LDA greater than 2 in the
ways exist under four categories (Figure 2C), including 45 specific EM group, such as Allobaculum (4.8), Akkermansia (4.3), and
KEGG pathways related to differential metabolites. The metabolites Parasutterella (4.1), were screened (Figure 4A). The PICRUSt soft-
are significantly enriched in 10 pathways (P < .05), such as choline ware was further used to predict the function of gut microbiota in
metabolism in cancer; cutin, suberine, and wax biosyntheses; and EM group. Significant differences are observed between the EM
secondary bile acid biosynthesis (Figure 2D). In terms of importance, and the control groups in the functional abundance of 80 path-
alpha-linolenic acid (ALA) metabolism; cutin, suberine, and wax ways at the KEGG pathway level 3. The functional abundance
biosyntheses; glycerophospholipid metabolism, and isoflavonoid of 62 pathways, such as oxidative phosphorylation and alanine,
biosynthesis are significantly affected by enriched differential me- aspartate, and glutamate metabolism, is significantly increased
tabolites (Figure 2E). (P < .05), whereas that of 16 pathways, such as starch and su-
crose metabolism and bacterial motility proteins, is significantly
decreased (P < .05, Figure 4B).
3.2 | Composition of the gut microbiota in The intersection of 80 differential KEGG pathways involved in
EMS mice the functional prediction of gut microbiota and 45 KEGG pathways
involved in fecal metabolites was analyzed. Eleven common path-
The trend of the rank–abundance curve of the OTU level in the EM ways, including primary and secondary bile acid biosyntheses; lysine
group is steeper than that in the CO group, suggesting decreased di- and xylene degradation; histidine, phenylalanine, tryptophan, ara-
versity of the EM group (Figure 3A). The PCoA analysis shows a sig- chidonic acid, ALA, and biotin metabolism; and flavonoid biosynthe-
nificant difference in species composition between the two groups sis, are found (Figure 4B).
(Figure 3B). The alpha diversity analysis shows that the Shannon
(P < .05) and Sobs (P > .05) indices of the EM group are lower than
those of the CO group, suggesting that the diversity and richness of 3.4 | Correlation between differential
bacteria decreased (data not shown). metabolites and gut microbiota in EMS mice
At the phylum level (Figure 3C), the decreased abundance of
Bacteroides and Firmicutes (P > .05) and ratio of Firmicutes/Bacteroides Together with the KEGG enrichment analysis of differential
(2.25 vs 2.01) and the increased abundance of Proteobacteria and metabolites and 11 pathways in part 3, the secondary bile acid
Verrucomicrobia (P < .05) are observed in the EM group. Among biosynthesis (Figure 5A) and ALA metabolism (Figure 5B) pathways
the top 20 abundant species at the genus level, the abundance of play important roles in the interaction of metabolites and gut
Allobaculum, Akkermansia, Parasutterella, and Rikenella in the EM microbiota in EMS mice. The two pathways involve four differential
group has increased significantly (P < .05), whereas the abundance metabolites (Table 1), namely chenodeoxycholic acid (CDCA),
of eight species of bacteria, such as Lachnospiraceae_NK4A136_ ursodeoxycholic acid (UDCA), ALA, and 12,13s-epoxy-9z,11,15z-
group, Lactobacillus, Bacteroides, has decreased significantly octadecatrienoic acid (12,13-EOTrE). The ROC analysis shows that
(P < .05, Figure 3D). A total of 70 species at the genus level are the AUC values of CDCA, UDCA, ALA, and 12,13-EOTrE are 0.94,
found in both groups (Figure 3E). The CO group contains nine 0.89, 0.92, and 0.81, respectively (Figure 6A–D), suggesting their
unique genera (Figure 3F), whereas the EM group contains five potential as important biological indicators to distinguish the two
unique genera, namely, Citrobacter, Ruminococcaceae_NK4A214_ groups.
group, Family_XIII_UCG-001, Rikenellaceae_RC9_gut_group, and We further analyzed the correlation between the four import-
Parabacteroides (Figure 3F). ant metabolites and the top 50 abundance of species in the EM
8 of 14 | NI et al.

F I G U R E 4 Linear discriminant analysis and function prediction of gut microbiota in two groups. A, Histogram of linear discriminant
analysis (LDA). A high LDA score indicates great effect of species abundance on the difference between groups. B, Histogram of the KEGG
pathway level 3 with significant difference. The abscissa shows 80 KEGG pathways with significant difference (P < .05), and the ordinate
shows functional abundance. Yellow-labeled pathways are related to differential metabolites
NI et al. | 9 of 14

group (Figure 6E). The abundance of CDCA is negatively correlated 4 | D I S CU S S I O N

with the abundance of [Eubacterium]_coprostanoligenes_group
and Blautia. A significant negative correlation exists between In this study, we have uncovered the fecal metabolomics of EMS
the abundance of UCDA and Candidatus_Stoquefichus. The abun- mice and the KEGG pathway that may be affected and determined
dance of ALA is positively correlated with that of Helicobacter, the characteristic species within EMS mice. Through cross-analysis
Ruminococcus_1, Ruminococcaceae_UCG-009 but negatively cor- and ROC analysis, we have identified four important differential me-
related with that of norank_f__Bacteroidales_S24-7_group. The tabolites that can be used as potential biological indicators for the
abundance of 12,13-EOTrE is positively correlated with that of mouse models of EMS, providing new evidence for the non-invasive
Candidatus_Stoquefichus. diagnosis and treatment of EMS.
10 of 14 | NI et al.

F I G U R E 5 Map of KEGG pathways that play important roles in the interaction between fecal metabolomics and gut microbiota. A,
Map of secondary bile acid biosynthesis. The red formula is the position of the two differential metabolites chenodeoxycholic (CDCA) and
ursodeoxycholic (UDCA) acids in the pathway. B, Map of alpha-linolenic acid (ALA) metabolism. The red dots indicate the positions of two
different metabolites involved in this experiment, that is, ALA and 12,13s-epoxy-9z,11,15z-octadecatrienoic acid (12,13-EOTrE)

Although mice do not have menstruation like humans, the pro- mystery. At present, three studies have investigated the changes in
cess of endometrial debris flowing back into the abdominal cavity the gut microbiota during the construction of EMS mice.13,14,22 These
with menstrual blood, which can artificially construct a mouse model studies may provide new ideas for the non-invasive assessment of
of EMS, can be simulated. The animal model has provided great con- EMS through gut microbiota disorder in the future. However, these
venience for studying the pathogenesis of EMS, which remains a research results are contradictory in some part due to differences in
NI et al. | 11 of 14

TA B L E 1 Four metabolites involved in KEGG pathways and function prediction of gut microbiota

Abundance (mean with SD)

Retention FC (CO/ P
Metabolites Structure M/Z Mode time (min) VIP EM) CO group EM group value

Chenodeoxycholic acid 375.3 pos 5.8 1.45 0.9259 3.26 ± 0.12 3.52 ± 0.16 0.009

Ursodeoxycholic acid 375.3 pos 6.1 1.22 0.9504 3.85 ± 0.13 4.05 ± 0.13 0.022

Alpha-linolenic acid 261.2 pos 7.9 1.12 1.0704 2.66 ± 0.09 2.49 ± 0.17 0.049

12,13-EOTrE 293.2 pos 6.8 1.1 1.072 2.65 ± 0.15 2.47 ± 0.11 0.043

Abbreviations: CO, control group; EM, endometriosis group; FC, fold change; M/Z, mass-to-charge ratio; Mode, ion detection mode; Retention time,
retention time of charged ions in the chromatogram.

modeling methods and different time points of stool sampling. Similar or taurine to form bound bile acid, which is stored in the gall blad-
to our modeling process, Yuan et al have constructed the model by der and excreted into the intestine under the stimulation of food. 23
13
intraperitoneal injection of endometrial fragments for 42 days and Although most of the bound bile acids are reabsorbed through the
found that the gut microbiota of the EMS mice is abnormal, which is hepatointestinal circulation, a small part remains in the intestine
featured by the increased ratio of Firmicutes/Bacteroidetes and in- and can serve as a matrix for gut microbiota metabolism. 24 UDCA
creased abundance of Bifidobacterium. Chadchan et al have estab- is one of the secondary bile acids formed by gut microbiota through
lished the model by endometrial colonization surgery for 21 days 7α-dehydroxylation. 25 The gut microbiota dissociates bile acids by
and found that the gut microbiota of EMS mice is characterized by bile salt-hydrolyzing enzymes, which are widely distributed in Gram-
increased abundance of Bacteroidetes and decreased abundance of negative and Gram-positive flora, to relieve part of the toxic effect
Firmicutes. Drinking water containing antibiotics for three weeks can of bile acids and help the host to metabolize carbon, nitrogen, and
significantly inhibit endometrial cell proliferation and inflammatory sulfur. 26 In this study, the diversity and the abundance of gut micro-
response in endometriotic lesions and reduce inflammatory factor biota of EMS mice have decreased especially in the Verrucomicrobia
levels in abdominal cavity.14 Although Hantschel et al have consid- and Proteobacteria, and disorder in the three characteristic gen-
ered the time point of fecal sampling, they took a different model- era of Allobaculum (Firmicutes), Akkermansia (Verrucomicrobia), and
ing method from the above two studies to build an EMS mouse (ie, Parasutterella (Proteobacteria).
ovariectomy and endometrial colonization) and found differences in At present, reports about the direct relationship between
the composition of gut microbiota at 3, 7, and 21 days after the es- CDCA and EMS are not available, but many studies show that
tablishment of the model. 22 CDCA is closely related to gut microbiota and contributes to pro-
Although the results of the three studies are different, the discov- moting intestinal homeostasis. CDCA blocks lipopolysaccharide
ery of a close association between EMS and the gut microbiota is not (LPS)-induced activation of the myosin light chain kinase (MLCK)
affected. Considering that the surgical process may have a great ef- pathway in an FXR-dependent manner, thereby protecting the
fect on mice, we have chosen to adopt the method of intraperitoneal intestinal epithelial barrier function. 27,28 Animal studies have
injection of endometrial debris to build a mouse model of EMS and the shown that UDCA can stimulate the migration of intestinal ep-
sampling time on the 21st day after modeling. At this time point, the ithelial cells and protect the intestinal barrier through EGFR and
gut microbiota is significantly different, and the overall time is not too COX-2-dependent mechanisms, 29 thus alleviating the inflamma-
long. Different from previous studies, we have used non-targeted me- tory response and reducing the levels of TNF-α, IL-1β, and IL-6. 30
tabolomics combined with function prediction of the gut microbiota Therefore, the anti-inflammatory effect of CDCA and UDCA is
to analyze the important differential metabolites in the feces and the mainly achieved by protecting intestinal barrier and inhibiting the
relative KEGG pathways. We have found that secondary bile acid bio- leakage of inflammatory substances. The disorder of gut microbi-
synthesis and ALA metabolism pathways play important roles in the ota can affect the immunity and permeability of the host intestinal
interaction of fecal metabolites and gut microbiota in the EMS mice. wall and promote inflammation, 31 whereas EMS is an inflammatory
Two differential metabolites, CDCA and UDCA, are important disease, 32 indicating their close relationship. The increased CDCA
components of the secondary bile acid biosynthesis. CDCA is a pri- and UDCA in the intestine of EMS mice build an effective protec-
mary bile acid synthesized in the liver and combined with glycine tive wall between gut microbiota disorder and EMS.
12 of 14 | NI et al.

F I G U R E 6 Receiver operator characteristic curve (ROC) analysis and bacterial association analysis of fecal metabolites. ROC analysis of
(A) chenodeoxycholic acid, (B) ursodeoxycholic acid, (C) alpha-linolenic acid, and (D) 12,13-EOTrE. The area under the curve (AUC) indicates
the discrimination performance. A large AUC indicates a good model. The CI is the 95% confidence interval of AUC and calculated on the
basis of the non-parametric resampling method. The point on the curve refers to the best threshold, which is determined on the basis of the
ROC curve to distinguish the two groups. (E) Heatmap of correlation between metabolites and top 50 abundance species at the genus level.
The correlation (R) value is displayed in different colors in the figure. The legend on the right is the color range of different R values, and the
legend on the left is the bacteria cluster tree; *.01 < P ≤ .05, **P ≤ .01

In addition, some studies show that EMS is related to abnor- a direct impact on the inflammatory, which is mainly achieved by
mal lipid metabolism, which is manifested by the low body mass inhibiting iNOS and COX-2.
index, decreased body fat in human, 33 decreased number of adi- Endometriosis is widely believed to be an estrogen-dependent
pocyte stem cells, and disorder of lipid metabolism in the animal chronic inflammatory disease,44,45 which may do damage to oo-
model. 34 CDCA and UDCA can improve lipid metabolism and re- cyte development and cause infertility.46 However, ALA can reduce
duce cholesterol saturation in bile, total cholesterol, and low-den- the response to oxidative stress and malondialdehyde (MDA) level
sity cholesterol. 35-37 We have speculated that the change in the in oocytes culture medium, thus promoting the maturation of oo-
species abundance and increased CDCA and UDCA levels may be a cytes,47 which may improve oocyte quality in patients with EMS.48
self-regulation mode of gut microbiota in EMS mice, which can pro- Intriguingly, the increased estrogen level can accelerate ALA metab-
vide the necessary energy to adapt to the inflammatory response olism and weaken its anti-inflammatory effect.49 Therefore, it may
by mobilizing lipid metabolism. An in-depth study on the relation- be an effective method to maintain a certain level of ALA in women
ship between these two substances and EMS should be conducted with EMS for controlling the disease.
in the next study. Our study shows that the content of ALA in the feces of EMS
Another important pathway worth exploring is ALA metabolism, mice is significantly reduced, which is not conducive to the body's
which includes ALA and 12,13-EOTrE. ALA, a beneficial unsaturated response to inflammation and promotes the development of inflam-
fatty acid, can improve lipid distribution in vivo and inhibit inflam- mation in EMS.50 The specific mechanism of ALA's effect on EMS
38,39
mation and antioxidative stress. In vitro studies have shown that will be carried out in our next study. In addition, a study on the cor-
ALA can inhibit the inflammatory characteristics of typical activated relation between 12,13-EOTrE and inflammation or EMS is lacking,
(M1) macrophages in LPS-induced human THP-1 cell line and reduce but we speculate that 12,13-EOTrE is related to inflammation and
the production of IL-1β, IL-6, TNF-α, and prostaglandin D2.40 Several may be a potential important marker of EMS.
studies have indicated that ALA can inhibit the accumulation of ni- In conclusion, we carried out the fecal metabolomics and gut mi-
trite and prostaglandin E2, and inhibit the protein and mRNA expres- crobiota research of the animal model of EMS and provided the orig-
sion of iNOS and COX-2 in a dose-dependent manner, thus reducing inal evidence support for the non-invasive research of EMS from a
the level of inflammatory factors.41-43 These suggest that ALA has new method and perspective. Dysbacteriosis exists in the intestinal
NI et al. | 13 of 14

tract of mice with endometriosis and may further change the overall 14. Chadchan SB, Cheng M, Parnell LA, et al. Antibiotic therapy with
metronidazole reduces endometriosis disease progression in mice: a
state of the host by affecting metabolites, such as CDCA, UDCA,
potential role for gut microbiota. Hum Reprod. 2019;34:1106-1116.
ALA, and 12,13-EOTrE. An in-depth and detailed research on spe- 15. Gao J, Xu K, Liu H, et al. Impact of the gut microbiota on intesti-
cific flora, specific metabolites, and EMS will be carried out in our nal immunity mediated by tryptophan metabolism. Front Cell Infect
next experiment. Microbiol. 2018;8:13.
16. Giau VV, Wu SY, Jamerlan A, An SSA, Kim SY, Hulme J. Gut
Microbiota and their neuroinflammatory implications in alzheimer's
AC K N OW L E D G M E N T S disease. Nutrients. 2018;10(11):1765.
The authors thank laboratory staff for helpful advices. 17. Ottosson F, Brunkwall L, Ericson U, et al. Connection between
BMI-related plasma metabolite profile and gut microbiota. J Clin
Endocrinol Metab. 2018;103:1491-1501.
C O N FL I C T S O F I N T E R E S T
18. Kang DW, Ilhan ZE, Isern NG, et al. Differences in fecal microbial
The authors declare that there is no conflict of interest that could be metabolites and microbiota of children with autism spectrum disor-
perceived as prejudicing the impartiality of the research reported. ders. Anaerobe. 2018;49:121-131.
19. Mizuno H, Ueda K, Kobayashi Y, et al. The great importance of
normalization of LC-MS data for highly-accurate non-targeted
AU T H O R C O N T R I B U T I O N S
metabolomics. Biomed Chromatogr. 2017;31:e3864. https://doi.
CY and ML conceived the study. ZN, SS, YB and JD preformed org/10.1002/bmc.3864
the experiments and analyzed data. ZN and SS drafted the man- 20. Ni ZX, Bi YL, Sun S, Yu CQ. Effect of blood stasis caused by cold on
uscript. WC, JY, and LZ provided experimental guidance. All au- the gut microbiota of endometriosis mice. Chin J Integr Trad West
Med. 2019;39:1214-1218.
thors agreed on the order in which their names were listed in the
21. Zhou L, Ni Z, Cheng W, et al. Characteristic gut microbiota and pre-
manuscript.
dicted metabolic functions in women with PCOS. Endocr Connect.
2020;9:63-73.
ORCID 22. Hantschel J, Weis S, Schäfer KH, et al. Effect of endometriosis on
Mingqing Li https://orcid.org/0000-0002-9276-0722 the fecal bacteriota composition of mice during the acute phase of
lesion formation. PLoS One. 2019;14:e0226835.
Chaoqin Yu https://orcid.org/0000-0003-1614-2655
23. Long SL, Gahan CGM, Joyce SA. Interactions between gut bacteria
and bile in health and disease. Mol Aspects Med. 2017;56:54-65.
REFERENCES 24. Wahlström A, Sayin SI, Marschall HU, Bäckhed F. Intestinal cross-
1. Bulun SE. Endometriosis. N Engl J Med. 2009;360:268-279. talk between bile acids and microbiota and its impact on host me-
2. Ballard KD, Seaman HE, de Vries CS, Wright JT. Can symptomatol- tabolism. Cell Metab. 2016;24:41-50.
ogy help in the diagnosis of endometriosis? Findings from a national 25. Zhang X, Fan D, Hua X, Zhang T. Large-scale production of ur-
case-control study–Part 1. BJOG. 2008;115:1382-1391. sodeoxycholic acid from chenodeoxycholic acid by engineering
3. Giudice LC, Kao LC. Endometriosis. Lancet. 2004;364:1789-1799. 7α- and 7β-hydroxysteroid dehydrogenase. Bioprocess Biosyst Eng.
4. Sampson JA. Metastatic or embolic endometriosis, due to the men- 2019;42:1537-1545.
strual dissemination of endometrial Tissue into the venous circula- 26. Song Z, Cai Y, Lao X, et al. Taxonomic profiling and populational pat-
tion. Am J Pathol. 1927;3:93-110. terns of bacterial bile salt hydrolase (BSH) genes based on world-
5. Czyzyk A, Podfigurna A, Szeliga A, Meczekalski B. Update on endo- wide human gut microbiome. Microbiome. 2019;7:9.
metriosis pathogenesis. Minerva Ginecol. 2017;69:447-461. 27. Song M, Ye J, Zhang F, et al. Chenodeoxycholic Acid (CDCA) pro-
6. Hamasaki H. Exercise and gut microbiota: clinical implications for tects against the lipopolysaccharide-induced impairment of the
the feasibility of Tai Chi. J Integr Med. 2017;15:270-281. intestinal epithelial barrier function via the FXR-MLCK pathway. J
7. Obata Y, Castaño Á, Boeing S, et al. Neuronal programming by mi- Agric Food Chem. 2019;67:8868-8874.
crobiota regulates intestinal physiology. Nature. 2020;578:284-289. 28. Gadaleta RM, van Erpecum KJ, Oldenburg B, et al. Farnesoid X re-
8. Boulangé CL, Neves AL, Chilloux J, Nicholson JK, Dumas ME. ceptor activation inhibits inflammation and preserves the intestinal
Impact of the gut microbiota on inflammation, obesity, and meta- barrier in inflammatory bowel disease. Gut. 2011;60:463-472.
bolic disease. Genome Med. 2016;8:42. 29. Golden JM, Escobar OH, Nguyen MVL, et al. Ursodeoxycholic acid
9. Mulcahy-O'Grady H, Workentine ML. The challenge and poten- protects against intestinal barrier breakdown by promoting entero-
tial of metagenomics in the clinic. Front Immunol. Front Immunol. cyte migration via EGFR- and COX-2-dependent mechanisms. Am J
2016;7:29. Physiol Gastrointest Liver Physiol. 2018;315:G259-G271.
10. Huang G, Xu J, Lefever DE, Glenn TC, Nagy T, Guo TL. Genistein 30. Ko WK, Kim SJ, Jo MJ, et al. Ursodeoxycholic acid inhibits inflam-
prevention of hyperglycemia and improvement of glucose tolerance matory responses and promotes functional recovery after spinal
in adult non-obese diabetic mice are associated with alterations of cord injury in rats. Mol Neurobiol. 2019;56:267-277.
gut microbiome and immune homeostasis. Toxicol Appl Pharmacol. 31. Gil-Cardoso K, Ginés I, Pinent M, Ardévol A, Blay M, Terra X.
2017;332:138-148. Effects of flavonoids on intestinal inflammation, barrier integrity
11. Flores R, Shi J, Fuhrman B, et al. Fecal microbial determinants of and changes in gut microbiota during diet-induced obesity. Nutr Res
fecal and systemic estrogens and estrogen metabolites: a cross-sec- Rev. 2016;29:234-248.
tional study. J Transl Med. 2012;10:253. 32. Jiang L, Yan Y, Liu Z, Wang Y. Inflammation and endometriosis. Front
12. Bailey MT, Coe CL. Endometriosis is associated with an altered pro- Biosci. 2016;21:941-948.
file of intestinal microflora in female rhesus monkeys. Hum Reprod. 33. Ferrero S, Anserini P, Remorgida V, Ragni N. Body mass index in
2002;17:1704-1708. endometriosis. Eur J Obstet Gynecol Reprod Biol. 2005;121:94-98.
13. Yuan M, Li D, Zhang Z, Sun H, An M, Wang G. Endometriosis induces 34. Zolbin MM, Mamillapalli R, Nematian SE, Goetz L, Taylor HS.
gut microbiota alterations in mice. Hum Reprod. 2018;33:607-616. Adipocyte alterations in endometriosis: reduced numbers of stem
14 of 14 | NI et al.

cells and microRNA induced alterations in adipocyte metabolic 44. García-Gómez E, Vázquez-Martínez ER, Reyes-Mayoral C, Cruz-
gene expression. Reprod Biol Endocrinol. 2019;17:36. Orozco OP, Camacho-Arroyo I, Cerbón M. Regulation of inflam-
35. Ghosh Laskar M, Eriksson M, Rudling M, Angelin B. Treatment with mation pathways and inflammasome by sex steroid hormones in
the natural FXR agonist chenodeoxycholic acid reduces clearance endometriosis. Front Endocrinol (Lausanne). 2020;10:935.
of plasma LDL whilst decreasing circulating PCSK9, lipoprotein(a) 45. Zubrzycka A, Zubrzycki M, Perdas E, Zubrzycka M. Genetic, epigen-
and apolipoprotein C-III. J Intern Med. 2017;281:575-585. etic, and steroidogenic modulation mechanisms in endometriosis. J
36. Fiorucci S, Mencarelli A, Palladino G, Cipriani S. Bile-acid-activated Clin Med. 2020;9:1309.
receptors: targeting TGR5 and farnesoid-X-receptor in lipid and glu- 46. Pessoa de Farias Rodrigues M, Lima Vilarino F, de Souza Barbeiro
cose disorders. Trends Pharmacol Sci. 2009;30:570-580. Munhoz A, et al. Clinical aspects and the quality of life among
37. Simental-Mendía LE, Simental-Mendía M, Sánchez-García A, et al. women with endometriosis and infertility: a cross-sectional study.
Impact of ursodeoxycholic acid on circulating lipid concentrations: BMC Womens Health. 2020;20:124.
a systematic review and meta-analysis of randomized placebo-con- 47. Sun LJ, Hu LJ, Guan YC, et al. Effect of alpha-linolenic acid on in
trolled trials. Lipids Health Dis. 2019;18:88. vitro maturation of human immature oocytes. Matern Child Health
38. Boskabady MH, Kaveh M, Shakeri F, Mohammadian Roshan N, Care China. 2015;30:5002-5005.
Rezaee R. Alpha-linolenic acid ameliorates bronchial asthma 48. Nishihara T, Matsumoto K, Hosoi Y, Morimoto Y. Evaluation
features in ovalbumin-sensitized rats. J Pharm Pharmacol. of antioxidant status and oxidative stress markers in follicular
2019;71:1089-1099. fluid for human in vitro fertilization outcome. Reprod Med Biol.
39. Burak C, Wolffram S, Zur B, et al. Effect of alpha-linolenic acid 2018;17:481-486.
in combination with the flavonol quercetin on markers of cardio- 49. Mason JK, Kharotia S, Wiggins AKA, et al. 17β-estradiol increases
vascular disease risk in healthy, non-obese adults: A randomized, liver and serum docosahexaenoic acid in mice fed varying lev-
double-blinded placebo-controlled crossover trial. Nutrition. els of α-linolenic acid. Lipids. 2014;49(8):745-756. https://doi.
2019;58:47-56. org/10.1007/s11745-014-3913-8
40. Pauls SD, Rodway LA, Winter T, Taylor CG, Zahradka P, Aukema 50. Lin YH, Chen YH, Chang HY, Au HK, Tzeng CR, Huang YH. Chronic
HM. Anti-inflammatory effects of α-linolenic acid in M1-like macro- niche inflammation in endometriosis-associated infertility: cur-
phages are associated with enhanced production of oxylipins from rent understanding and future therapeutic strategies. Int J Mol Sci.
α-linolenic and linoleic acid. J Nutr Biochem. 2018;57:121-129. 2018;19:2385.
41. Anand R, Kaithwas G. Anti-inflammatory potential of alpha-linole-
nic acid mediated through selective COX inhibition: computational
and experimental data. Inflammation. 2014;37:1297-1306.
How to cite this article: Ni Z, Sun S, Bi Y, et al. Correlation of
42. Ren J, Han EJ, Chung SH. In vivo and in vitro anti-inflammatory
activities of alpha-linolenic acid isolated from Actinidia polygama
fecal metabolomics and gut microbiota in mice with
fruits. Arch Pharm Res. 2007;30:708-714. endometriosis. Am J Reprod Immunol. 2020;00:e13307.
43. Erdinest N, Shmueli O, Grossman Y, Ovadia H, Solomon A. Anti- https://doi.org/10.1111/aji.13307
inflammatory effects of alpha linolenic acid on human corneal epi-
thelial cells. Invest Ophthalmol Vis Sci. 2012;53:4396-4406.