Letters

https://doi.org/10.1038/s41564-021-00880-5

Akkermansia muciniphila secretes a glucagon-like

peptide-1-inducing protein that improves glucose
homeostasis and ameliorates metabolic disease
in mice
Hyo Shin Yoon1,9, Chung Hwan Cho1,9, Myeong Sik Yun1, Sung Jae Jang1, Hyun Ju You1,2,3,4,
Jun-hyeong Kim5, Dohyun Han 6, Kwang Hyun Cha 7, Sung Hyun Moon1,5, Kiuk Lee5, Yeon-Ji Kim8,
Sung-Joon Lee 8, Tae-Wook Nam5 and GwangPyo Ko 1,3,4,5 ✉

The gut microbiota, which includes Akkermansia muciniph- A. muciniphila has direct effects on the host energy expenditure18.
ila, is known to modulate energy metabolism, glucose toler- Little evidence for the mechanisms involved, including specific host
ance, immune system maturation and function in humans1–4. cellular components or bacterial proteins, in the beneficial effects
Although A. muciniphila is correlated with metabolic dis- of A. muciniphila has been reported. In particular, investigation
eases and its beneficial causal effects were reported on host of the relationship between A. muciniphila and host adipose tis-
metabolism5–8, the molecular mechanisms involved have not sue depots—including interscapular brown adipose tissue (iBAT),
been identified. Here, we report that A. muciniphila increases which mediates non-shivering thermogenesis—is lacking.
thermogenesis and glucagon-like peptide-1 (GLP-1) secretion The gut microbiota is known to influence the enteroendocrine
in high-fat-diet (HFD)-induced C57BL/6J mice by induction system19,20 and modulate the host immune system through micro-
of uncoupling protein 1 in brown adipose tissue and systemic bially derived metabolites or cellular membrane components21. In
GLP-1 secretion. We apply fast protein liquid chromatogra- particular, bacterial metabolites, such as short-chain fatty acids
phy and liquid chromatography coupled to mass spectro- (SCFAs)22–26, secondary bile acids27–29, indoles30 and lipopolysaccha-
photometry analysis to identify an 84 kDa protein, named ride31, regulate appetite by stimulating the release of gut hormones,
P9, that is secreted by A. muciniphila. Using L cells and mice such as GLP-1, and by activating enteric neuronal signalling32,
fed on an HFD, we show that purified P9 alone is sufficient to which contribute to energy homeostasis. The bacterial chaperone
induce GLP-1 secretion and brown adipose tissue thermogen- protein ClpB, which is present in commensal and pathogenic bacte-
esis. Using ligand–receptor capture analysis, we find that P9 ria, has been shown to regulate appetite33. Although it is known that
interacts with intercellular adhesion molecule 2 (ICAM-2). A. muciniphila affects the gut-hormone-releasing L cells in the gut34,
Interleukin-6 deficiency abrogates the effects of P9 in glucose evidence of the bioactive molecule of A. muciniphila that is involved
homeostasis and downregulates ICAM-2 expression. Our in GLP-1 secretion is lacking.
results show that the interactions between P9 and ICAM-2 Gut-microbiota-associated proteins also activate the host
could be targeted by therapeutics for metabolic diseases. immune system35. Recent research suggests that, in contrast to
The prevalence of metabolic disease has reached epidemic pro- other commensal bacteria36, A. muciniphila induces a unique type
portions during the past three decades9. The human gut microbiota of immune response that involves homeostatic antigen-specific-IgG
regulates various metabolic functions, including intestinal barrier antibody and T-cell responses, and a cytotoxic T-cell response
homeostasis, glucose homeostasis and energy absorption1–4. Several in a colorectal cancer model21,37. Furthermore, in high-fat-diet-
studies have demonstrated that an abundance of A. muciniphila is (HFD)-fed mice, this bacterium activates adipose-tissue-resident
correlated with metabolic disorders, such as obesity and type-2 dia- regulatory T-cells, and its component, Amuc_1100, has been
betes, in both preclinical and clinical studies10–15, and that supple- reported to interact with toll-like receptor 2 (refs. 7,38).
mentation with viable or pasteurized A. muciniphila ameliorates As reported previously5,7, oral administration of A. muciniphila
metabolic endotoxemia and improves gut-barrier function, thereby (ATCC-BAA-835) significantly reduced body mass, improved glu-
improving the systemic metabolic profile5–8. Furthermore, the pres- cose tolerance, and increased the serum concentrations of insu-
ence of A. muciniphila reduces the intestinal energy absorptive lin and β-oxidation gene markers compared with HFD-fed mice
capacity under cold conditions2 and the presence of A. muciniph- (Extended Data Fig. 1a–g). Interestingly, the mass and size of the
ila is correlated with fat browning16,17. Moreover, pasteurized iBAT were lower, whereas the mass and size of the epididymal

1
Department of Environmental Health Sciences, Graduate School of Public Health, Seoul National University, Seoul, Korea. 2Institute of Health and
Environment, Seoul National University, Seoul, Korea. 3Bio-MAX/N-Bio, Seoul National University, Seoul, Korea. 4Center for Human and Environmental
Microbiome, Seoul National University, Seoul, Korea. 5KoBioLabs Inc., Seoul, Korea. 6Proteomics Core Facility, Biomedical Research Institute, Seoul
National University Hospital, Seoul, Korea. 7Natural Product Informatics Research Center, Korea Institute of Science and Technology, Gangneung, Korea.
8
Department of Biotechnology, College of Life Sciences and Biotechnology, School of Life Sciences and Biotechnology, Korea University, Seoul, Korea.
9
These authors contributed equally: Hyo Shin Yoon, Chung Hwan Cho. ✉e-mail: [email protected]

Nature Microbiology | VOL 6 | May 2021 | 563–573 | www.nature.com/naturemicrobiology 563

Letters Nature Microbiology

HFD HFD + Am

UCP1
a HFD HFD + Am
b c HFD
30 HFD + Am
40
*** 4 6

Number of epWAT cells per fraction

HFD

epWAT
Number of iBAT cells per fraction

P = 0.0043

P = 0.0215

P = 0.0245
HFD + Am
30
20 3

Relative expression
P = 0.0113

Ucp1 expression
4
iBAT

20
2
****** 10
10 *** *** 1
2

0
0 0
0

2,500

7,500

12,500

17,500

22,500

27,500
0
200

600

1,000

1,400

1,800

2,200

2,600

AT

AT

ea

16

lin
c1
iB

e
id

dm

Ap
rg
C
ig
Lipid droplet size (µm2) Lipid droplet size (µm2)

Pr

a
Pp
d HFD e f HFD g
HFD + Am HFD + Am

Overlay
40 32.0
80 15
P = 0.0485
P = 0.0180
P = 0.0276 100

Plasma GLP-1 (pg ml–1)

CD11b+CD206+ cells (%)

101 102 103 104 105

HFD

60

Relative expression
P = 0.0077
Temperature (°C)

10
35

HFD
P = 0.0052 22.0 21.3%
32.0
40 50

5
HFD + Am

10 10 10 10 10
P = 0.0269
1 2 3 4 5

20

HFD + Am
30 0
59.5%
0

Am
0
Events

FD
22.0

+
cg

FD
2
sk
Am

sk
FD

H
Pc

Pc
in
l

101 102 103 104 105

ta

+
Sk
ec

CD206
FD
R

Fig. 1 | A. muciniphila activates brown adipocytes and induces GLP-1 expression. a, Frequency distribution of the surface area of the iBAT (left) and
epWAT (right) after 14 weeks of consumption of an HFD, or an HFD supplemented with A. muciniphila (HFD + A. muciniphila (Am)). n = 50 cells were
measured from three mice per group. Representative haematoxylin and eosin (H&E)-stained sections of epWAT (top) and iBAT (bottom). Scale bars,
100 μm. n = 3 per group. b, Representative immunohistochemical images of UCP1 expression (brown) in iBAT (n = 3 per group) and quantitative PCR with
reverse transcription (RT–qPCR) analysis of Ucp1 expression in iBAT and igWAT. Scale bars, 100 μm. c, Thermogenic iBAT-specific gene expression after
A. muciniphila treatment. d, Rectal and dorsal skin temperatures (left). Right, dorsal temperature, measured using infrared thermography. e, Flow cytometry
analysis of M2 macrophages in the stromal vascular fraction of the iBAT as a dot plot (left) and histogram (right). f, mRNA expression of Gcg, Pcsk1 and
Pcsk2 in the ileum. g, Plasma GLP-1 concentration 10 min after oral glucose challenge. For a–g, data are mean ± s.e.m. n = 7–10 (HFD) and n = 7–10 (HFD +
A. muciniphila) mice per group. Statistical significance was calculated using two-tailed unpaired t-tests; ***P < 0.001.

white adipose tissue (epWAT) in HFD-fed mice were unaffected by of mice administered with A. muciniphila (Extended Data Fig. 2f–h).
A. muciniphila treatment (Fig. 1a and Extended Data Fig. 1c). These These results show that A. muciniphila influences the thermogenic
results are consistent with the recent findings that the adminis- activity of iBAT and the abundance of iBAT-specific M2-like mac-
tration of viable A. muciniphila does not affect epWAT adipocyte rophages in mice.
diameter or fat mass7, which suggests that there is a specific loss To evaluate the underlying mechanism of thermogenic gene
of iBAT after A. muciniphila treatment. Next, to assess the func- upregulation in the iBAT, we compared the caecal metabolites
tional metabolic capacity of iBAT, we analysed thermogenic gene between mice that were orally administered with viable A. muciniph-
transcription. A significant induction of the mitochondrial-specific ila and control HFD-fed mice. Although creatine and glycine were
gene encoding uncoupling protein 1 (Ucp1) and the related ther- present at higher concentrations in the caecal contents of treated
mogenic differentiation marker transcripts was observed in the mice compared with the controls, there were few differences. We
iBAT (Fig. 1b,c) but not in the inguinal WAT (igWAT) of HFD-fed therefore concluded that A. muciniphila treatment did not have a
mice administered with A. muciniphila (Fig. 1b), which implies major impact on caecal metabolites (Extended Data Fig. 1h).
that the anti-obesogenic effects of A. muciniphila act through the GLP-1 regulates BAT thermogenesis39–41, so we next investi-
iBAT, rather than the igWAT. Skin temperature over the iBAT as gated whether oral administration of A. muciniphila affects GLP-1
well as rectal temperature were significantly higher in HFD-fed expression in cultured L cells and in HFD-fed mice. As most
mice administered with A. muciniphila compared with the controls GLP-1-secreting L cells are located in the distal ileum, we mea-
(Fig. 1d). Furthermore, the number of anti-inflammatory M2 mac- sured GLP-1 expression in the ileum after oral administration of
rophages (CD11b+CD206+) in the iBAT was significantly higher in A. muciniphila in HFD-fed mice. Transcription of Gcg (which
mice administered with A. muciniphila (Fig. 1e). encodes GLP-1 or glucagon), Pcsk1 and Pcsk2 was measured in each
A. muciniphila and S24-7 (also known as Muribaculaceae) fam- group of mice. A much higher expression of Gcg and Pcsk1 was found
ily spp. were more abundant in mice that were orally administered in HFD-fed mice that were orally administered with A. muciniphila
with A. muciniphila and, similarly, the 16S rRNA gene count for A. compared with the controls (Fig. 1f). Similarly, the plasma GLP-1
muciniphila was positively correlated with both iBAT temperature concentration was higher in HFD-fed mice that were orally admin-
and plasma GLP-1 concentration, and negatively correlated with the istered with A. muciniphila compared with the controls (Fig. 1g).
level of glucose intolerance (Extended Data Fig. 2a–e). Moreover, Next, to identify how A. muciniphila induces GLP-1 secretion,
transcripts of M2 macrophage markers were increased in the iBAT we treated human enteroendocrine L cells (NCI-H716) with live

564 Nature Microbiology | VOL 6 | May 2021 | 563–573 | www.nature.com/naturemicrobiology

Nature Microbiology Letters
A. muciniphila (SNUG-61027, isolated from human faeces with integrity and increasing the secretion of endocannabinoid-like
ethical approval or ATCC-BAA-835) pellets or the culture super- lipid molecules7—was also expressed as a control, as in the previ-
natants of the bacteria, and then measured GLP-1 concentra- ous study. We confirmed the production of a single protein of the
tion. We found that the cell-free supernatant (CFS) substantially correct size from each plasmid using SDS–polyacrylamide gel elec-
increased GLP-1 whereas the bacterial pellet did not (Extended trophoresis (Extended Data Fig. 4g). A list of the proteins identified
Data Fig. 3a). To determine whether this effect was dose depen- in sample 3 is provided in Extended Data Fig. 4h and the sequence
dent, we treated NCI-H716 cells with A. muciniphila CFS at con- of each protein had a high similarity (>98%) to that of equivalent
centrations of 10–100% (v/v) and found that GLP-1 was induced proteins produced by the type strain (ATCC-BAA-835).
in a dose-dependent manner (Extended Data Fig. 3b). We tested We treated NCI-H716 cells with each of the nine proteins
the effects of 23 Lactobacillus and 24 Bifidobacterium strains iso- expressed, and P1 (UniProt: B2UKW8), P5 (UniProt: B2URM2)
lated from human faeces with ethical approval on GLP-1 secretion and P9 (UniProt: B2UM07) were shown to induce GLP-1 in a
by NCI-H716 cells. The CFS (10% v/v) of ATCC-BAA-835 and of dose-dependent manner in NCI-H716 cells. However, the effect of
a newly isolated SNUG-61027 strongly induced GLP-1 secretion, Amuc_1100 in inducing GLP-1 secretion was negligible compared
whereas the CFS of all 47 strains tested had no effect (Extended with the effect of P9 (Fig. 2b). GLP-1 induction by P9 was also
Data Fig. 3c and Supplementary Table 1b). confirmed in human primary intestinal epithelial cells (Fig. 2b).
As enteroendocrine cells can respond to SCFAs by increasing To determine whether these effects would also be present in mice
their secretion of GLP-1, acetate and propionate, which are present that were fed a normal chow diet, mice were intraperitoneally (i.p.)
in the A. muciniphila CFS, were administered to enteroendocrine injected daily with each protein for 2 weeks, and their weight gain
cells. However, although GLP-1 secretion was increased, it was not and glucose tolerance were compared. Significant weight loss and a
by as much as the increase that was observed in the CFS, suggesting decrease in glucose intolerance were shown in mice that were treated
that SCFAs are not the only bacterial products that induce GLP-1 with P9, in contrast to the effects of the other proteins (Extended
secretion (Extended Data Fig. 3d,e). Data Fig. 5a,b). Furthermore, to verify whether cloned proteins
To identify other A. muciniphila-derived molecule(s) that are homologous to P9 that were derived from other gut bacteria have
responsible for the GLP-1 secretion, CFS of A. muciniphila (SNUG- similar effects to P9, and because P9 belongs to the peptidase S41A
61027) that was cultured in 5% fetal bovine serum (FBS)-based family, we cloned the equivalent S41 member to P9 from Escherichia
brain–heart infusion broth medium was filtered as described pre- coli (EcPrc) and compared its ability to induce GLP-1 with that of
viously42 to produce a series of fractions (Extended Data Fig. 4a). P9 from A. muciniphila in NCI-H716 cells (Extended Data Fig. 5c).
NCI-H716 cells were treated with 5 mg of each fraction and GLP-1 Interestingly, P9 from A. muciniphila induced GLP-1 but EcPrc did
secretion was measured. Interestingly, the 100–300 kDa filtrate sub- not in L cells, which indicates that A. muciniphila-derived P9 has a
stantially increased GLP-1 secretion, and the 30–100 kDa filtrate specific effect on GLP-1 secretion.
increased GLP-1 secretion, but to a lesser extent (Extended Data We also compared the effects of P9 on GLP-1 secretion with ace-
Fig. 4b). Treatment with proteinase K abolished the effects of both tate and propionate in mice. A single i.p. injection of P9 induced a
fractions, indicating that the molecule or molecules responsible for high level of GLP-1 secretion, whereas i.p. injection with SCFAs or
the induction of GLP-1 secretion is probably a protein(s) (Extended EcPrc in mice did not, demonstrating that A. muciniphila-derived
Data Fig. 4c). To isolate the candidate protein(s) produced by A. P9 induces GLP-1 through a different mechanism compared with
muciniphila, the 100–300 kDa filtrate of SNUG-61027 and control the other GLP-1 inducers, such as SCFAs in vivo (Extended Data
medium, both of which contained approximately 25 mg ml−1 pro- Fig. 5d).
tein, were passed through anion-exchange Mono Q columns, the Consistent results were also obtained in HFD-fed mice that were
collected fractions were used to treat NCI-H716 cells and the GLP-1 orally administered P9 daily for 8 weeks—mice that were fed P9
concentrations were measured. The m2–m4 fractions induced showed significantly lower weight gain (Fig. 2c) and food intake
GLP-1 secretion (Extended Data Fig. 4d); these fractions were (Fig. 2d) compared with the controls. The adipose tissue volume
therefore concentrated and applied to size-exclusion columns to (Fig. 2e,f and Extended Data Fig. 6a) and the glucose intolerance
obtain further fractions, which were tested for their ability to induce (Fig. 2g) of the mice were also significantly lower, whereas the
GLP-1 secretion. Of these, the G17–G20 fractions induced the most ileal Gcg and Pcsk1 expression was significantly higher (Fig. 2h).
GLP-1 secretion by L cells (Extended Data Fig. 4e). Furthermore, P9 induced thermogenesis, as shown by significant
Further analysis of these fractions was performed using liquid increases in the expression of BAT-specific genes (Fig. 2i) and in
chromatography coupled with tandem mass spectrometry (LC–MS/ body temperature (Fig. 2j and Extended Data Fig. 6b) at room tem-
MS) to identify the proteins that are responsible for GLP-1 secre- perature and after a cold shock. Body composition analysis revealed
tion. Approximately 90 µg of protein from the 100–300 kDa filtrate significantly higher lean mass and lower fat mass (Extended Data
obtained from the size cut-off filtration (sample 1), the products Fig. 7a). Moreover, indirect calorimetry showed that mice that were
of ion-exchange chromatography (sample 2) and the products administered with P9 had a lower respiratory quotient and higher
of the size-exclusion chromatography (sample 3) were analysed. fatty acid oxidation, although without any difference in energy
Bovine-specific proteins that were present in the basal medium expenditure (Extended Data Fig. 7b–d). Collectively, these results
were excluded from the list of candidate proteins. Ninety-eight indicate that the administration of P9 prevents obesity by regulating
A. muciniphila-derived proteins and 130 bovine-specific proteins glucose homeostasis and inducing thermogenesis by iBAT.
were identified: 95 in sample 1, 36 in sample 2 and 10 in sample 3 Next, to determine how P9 induces higher GLP-1 expression,
(Extended Data Fig. 4f), a list of which is provided in Supplementary RNA-sequencing (RNA-seq) analysis was performed in P9-treated
Table 3. An overview of the proteins identified under each filtra- NCI-H716 and control cells. Notably, P9 induced the expression
tion condition, before and after filtering for proteins predicted to be of genes encoding calcium-related signalling proteins (Fig. 3a and
secreted, is shown (Fig. 2a). Supplementary Table 7). Furthermore, using a phosphokinase
To investigate the effects of A. muciniphila-derived proteins, array, we found that P9 increased the expression of phosphorylated
we cloned the cDNAs corresponding to the proteins identified in cAMP-response-element-binding protein (p-CREB) and phos-
sample 3 (Methods), and each protein was tested for its ability to phorylated heat shock protein 27 (p-HSP27; Extended Data Fig. 8a).
induce GLP-1 secretion in L cells. The outer membrane protein Calcium influx was significantly higher (Extended Data Fig. 8b),
Amuc_1100—which is derived from ATCC-BAA-835 and has and pretreatment with calcium inhibitors reduced the P9-induced
previously been shown to prevent obesity by restoring gut-barrier expression of GLP-1 (Fig. 3b and Extended Data Fig. 8c).

Nature Microbiology | VOL 6 | May 2021 | 563–573 | www.nature.com/naturemicrobiology 565

Letters Nature Microbiology

a b f LFD HFD HFD + P9

4,000 300
P < 0.0001

GLP-1 expression in NCI-H716

Secretomics –1 P < 0.0001
3,500 10 µg ml

GLP-1 expression in human

127 3,000 100 µg ml–1

InEpCs (pg ml–1)

Centrifugal filter unit

cells (pg ml–1)

2,500 200
2,000
39 100
FPLC: MonoQ
80
100
60
10
FPLC: GPC 40
20
0 0
P9 Control P9

l
P1
P2

P3

P4
P5
P6
P7

P8

uc P9

0
tro

10
on
LFD HFD HFD + P9

_1
C

Am

epWAT
c 200 d 5 e

igWAT
*ϕ versus HFD HFD
Food intake (g per mouse per day)

LFD

P < 0.0001
P = 0.0047
ϕϕϕϕ

Tissue mass (g per g body mass)

180 HFD
versus LFD HFD + P9 0.15
P = 0.0115

ϕϕϕϕ 4

P < 0.0001
P = 0.0207
HFD + P9
ϕϕϕ
P = 0.0226

160
Weight gain (%)

ϕϕϕ
ϕϕ 0.10

iBAT
140 3
*****
ϕϕ

P = 0.0265
P = 0.0281
* * **
ϕϕϕϕ
120 ϕϕϕ
2 0.05
100

Liver
80 1 0
0 1 2 3 4 5 6 7 8 1 2 3 4 5 6 epWAT igWAT iBAT
Time (weeks) Week

HFD
g h i j
HFD + P9
500 60,000 5 5

P = 0.0450
*ϕ versus

P = 0.0008
HFD P = 0.0354 P = 0.0017 40

Temperature (°C) after cold shock

ϕϕ P < 0.0001 P = 0.0103
Glucose AUC (mg dl–1 min–1)

versus LFD

P = 0.0014
ϕϕ

P = 0.0265
400 4 4
Blood glucose (mg dl–1)

ϕ
Relative expression
Relative expression

P = 0.0067
* 40,000

P = 0.0139

P = 0.0160
300 ϕ 3 3 35
ϕϕ
200 2 2
20,000
ϕ 30
100 1 1

0 0 0 0
0 30 60 90 120 Dorsal Ventral Eye Rectal
ea

1a
1
D

FD

P9

cg

cp
LF

sk

id

gc
G
H

C
Pc

ar
FD

Pp
H

Fig. 2 | A purified GLP-1-inducing protein, P9, from A. muciniphila ameliorates obesity and influences glucose homeostasis by promoting thermogenesis.
a, Overview of the number of proteins identified following the various types of filtration, before and after filtering for proteins that were predicted to be
secreted. FPLC, fast protein liquid chromatography; MonoQ, anion-exchange column chromatography; GPC, gel permeation chromatography. b, Expression
of GLP-1 after treating NCI-H716 cells with purified proteins produced by A. muciniphila (data represent three independent experiments performed in
duplicate, n = 6 per group; left) and treating human primary intestinal epithelial cells (InEpC) with P9 (50 μg ml−1; n = 7 per group; right). c, Weight gain
in HFD-fed mice that were orally administered with P9 (100 μg per mouse) for 8 weeks. d, Food intake (g per mouse per day) (red, HFD; green, HFD +
P9). e, Tissue masses (g per g body mass). f, The appearance of the mice (top) and H&E-stained sections of each adipose depot (bottom). Scale bars,
100 μm. n = 3 per group. g, Oral glucose tolerance testing (left) was conducted, and the areas under the curves were measured (right). h, Expression of
GLP-1-specific genes in the ileum. i, Expression of thermogenesis-specific genes in the iBAT. j, Temperatures of several parts of the body (dorsal, ventral,
eye, rectal) after a cold shock at 5 °C for 4 h. Data are mean ± s.e.m. Number of mice per group: for c–e, g–h and j, low-fat diet: n = 8 (c, e and g), HFD: n = 8,
HFD + P9: n = 8; for i, HFD: n = 7, HFD + P9: n = 7. Data were analysed using the Kruskal–Wallis test followed by Dunn’s post hoc test, with comparison
to the control group for b (left). For c and d, data were analysed using two-way analysis of variance (ANOVA), followed by the Bonferroni post hoc test;
for g (left), two-way ANOVA, followed by Tukey’s post hoc test; for g (right), one-way ANOVA, followed by Tukey’s post hoc test. For b (right), e and h–j,
statistical analysis was performed using two-tailed unpaired t-tests. The asterisks (*) and phi (Φ) indicate statistical significance; Φ or *P < 0.05, ΦΦ or
**P < 0.01, ΦΦΦ or ***P < 0.001, ΦΦΦΦ or ****P < 0.0001.

A GLP-1 receptor β-arrestin assay confirmed that P9 does not abundantly enriched protein was ICAM-2. We therefore pretreated
bind to the GLP-1 receptor (Extended Data Fig. 8d). Moreover, cells with ICAM-2 peptide or anti-ICAM-2 or anti-KTN1 antibod-
pretreatment with inhibitors of G-protein-coupled receptor ies, and measured the P9-induced GLP-1 secretion. Interestingly,
(GPCR) histamine H1 receptor, alpha-1β adrenergic receptor ICAM-2 peptide significantly reduced P9-induced GLP-1 secre-
and muscarinic acetylcholine receptor M1 did not alter GLP-1 tion in a dose-dependent manner (Fig. 3d). Next, to determine
secretion by L cells after P9 treatment (Extended Data Fig. 8e). the importance of the interaction with ICAM-2, we pretreated
To identify the receptor for P9, we performed a protein-binding cells with anti-ICAM-2 antibodies and measured the effects of P9.
assay using ligand–receptor capture (LRC)-TriCEPS technology. GLP-1 secretion was significantly reduced by treating the cells in
The cellular proteins that bound to P9 in NCI-H716 cells were this manner; however, the level of suppression was insufficient to
identified using LC–MS/MS (Fig. 3c). The most significantly prevent the entire effect of ICAM-2 (Fig. 3e). These results indicate
enriched P9-binding protein was kinectin 1 (KTN1), and the most that P9 directly binds to ICAM-2.

566 Nature Microbiology | VOL 6 | May 2021 | 563–573 | www.nature.com/naturemicrobiology

Nature Microbiology Letters
a Biological process b
Top 20 terms of GO functional analysis
Response to corticosterone
Response to mineralocorticoid 150 P = 0.0077
Response to progesterone P = 0.0002

Relative expression of GLP-1

Negative regulation of G-protein-coupled receptor P < 0.0001 P = 0.0001

Cellular response to calcium ion

100
Response to cAMP Intersection size
Response to organophosphorus 1.00
Regulation of G-protein-coupled receptor signalling 1.25
1.50 50
Response to glucocorticoid 1.75
Response to rurine-containing compound 2.00

Response to calcium ion Adjusted P

Response to corticosteroid 0.020 0
0.025
Negative regulation of locomotion involved in locomotory behaviour 0.030 P9 – + + + + + + + +
0.035
Light adaption 0.040
666-15 (CREB inhibitor) – – + – – – – – –
0.045
Transport, plasma membrane to Golgi
U-73122 (PLC inhibitor) – – – + – – – – –
Adaptation of rhodopsin-mediated signalling
EGTA-Ringer (ion chelator) – – – – + – – – –
Negative regulation of heat generation
Diltiazem (L-type inhibitor) – – – – – + – – –
Conditioned taste aversion
Mibefradil (T-type inhibitor) – – – – – – + – –
Negative regulation of adenylate-cyclase-activating signalling
H-89 (PKA inhibitor) – – – – – – – + –
Regulation of locomotion involved in locomotory behaviour
Gallein (G protein βγ inhibitor) – – – – – – – – +
0 0.1 0.2 0.3
Gene ratio
(intersection size/query size)

c d e
4 3,000 P < 0.0001 2,000
P < 0.0001 P = 0.0210
P < 0.0001 P < 0.0001

GLP-1 secretion in NCI-H716

KTN1
GLP-1 secretion in GLUTag

P9 P < 0.0001 P < 0.0001

3 1,500
–log10(adjusted P)

2,000

cells (pg ml–1)

cells (pg ml–1)

2 PFKAP 1,000
DJC10
1,000
1 500
ICAM2

0 0 0
–7 –6 –5 –4 –3 –2 –1 0 1 2 3 4 5 6 7 8 P9 – + + + + – + – + P9 – + + +
log2-transformed fold change ICAM-2 peptide (1:5) – – + – – – – – – Negative antibodies – – + –
ICAM-2 peptide (1:10) – – – + – – – – –
Anti-ICAM-2 – – – +
ICAM-2 peptide (1:50) – – – – + – – – – antibodies
Negative antibodies (1:10) – – – – – + + – –
Anti-KTN1 antibodies (1:5) – – – – – – – + +

Fig. 3 | P9 induces GLP-1 secretion through a Ca2+-dependent pathway and ICAM-2. a, RNA-seq analysis of NCI-H716 cells treated with P9 (50 μg ml−1)
for 30 min. n = 3 per group. b, GLP-1 secretion by GLUTag cells that were pretreated with calcium inhibitors (10 μM) for 15 min, and then with P9
(100 μg ml−1) for 2 h. n = 6 per group. c, An LRC-TriCEPS experiment (Dualsystems Biotech) using P9 was performed in NCI-H716 cells. n = 3 per group.
The enriched proteins were purified and processed for MS-based proteomics analysis (LC–MS/MS) and the P values obtained were adjusted for multiple
testing using the Benjamini–Hochberg method. d, GLP-1 secretion was measured after treatment of GLUTag cells with ICAM-2 peptide (concentration of
1:5, 1:10 or 1:50) or anti-KTN1 antibodies (1:5) for 1 h and then with P9 (50 µg ml–1) for 1 h. e, GLP-1 secretion was measured after treatment of NCI-H716
cells with anti-ICAM-2 antibodies (concentration of 1:10) for 1 h and then with P9 (50 µg ml–1) for 1 h. d, e, n = 6 per group. For b, d and e, data are
mean ± s.e.m. For a, correction for multiple testing was performed using the Benjamini–Hochberg method. For b, d and e, data represent the results of
three independent experiments performed in duplicate. For b and e, data were analysed using the Kruskal–Wallis test followed by Dunn’s post hoc test. For
d, data were analysed using two-tailed unpaired t-tests.

The increase in interleukin-6 (IL-6) secretion that is induced Ucp1 expression was significantly increased by IL-6 treatment in
by exercise has been reported to stimulate GLP-1 secretion by primary preadipocytes that were obtained from HFD-fed mice
intestinal L cells and pancreatic α-cells, thereby promoting insulin (Extended Data Fig. 9c). IL-6 induces GLP-1 in GLUTag cells,
secretion43. More recently, proinflammatory stimuli, such as endo- although to a lesser extent compared with the P9-treated group, and
toxin and IL-1β, have been shown to induce GLP-1 secretion in an the additive effects of GLP-1 secretion were observed after treat-
IL-6-dependent manner, causing a reduction in blood glucose44. ment with both P9 and IL-6 compared with the P9-treated group
We therefore hypothesized that P9 may not only induce GLP- (Extended Data Fig. 9d,e).
1, but also IL-6, which regulates glucose homeostasis. We found To determine whether the effects of P9 on glucose tolerance and
that P9 strongly induced IL-6 expression in macrophage cell lines obesity require the GLP-1 receptor (GLP-1R) signalling pathway
(Extended Data Fig. 9a). We also found that A. muciniphila induced and whether these effects are IL-6 dependent, we i.p. injected exen-
specific cytokine expression patterns, characterized by an upregu- din(9–39) (a GLP-1R antagonist) and orally administered P9 into
lation of IL-6 but not tumour necrosis factor-α in the ileum and HFD-fed mice, and compared the effects with the injection of P9
colon of HFD-fed mice, and in macrophages and colonic epithelial alone. Interestingly, exendin(9–39) treatment suppressed the effects
cells in vitro (Extended Data Fig. 1i–n and Supplementary Table of P9 on glucose tolerance (Fig.4a), insulin tolerance (Extended Data
1a). IL-6 increased thermogenic gene expression in immortalized Fig. 10c), weight gain and food intake (Extended Data Fig. 10d–e),
brown preadipocytes (BACs; Extended Data Fig. 9b). Similarly, which suggests that P9 activates the GLP-1 receptor pathway and

Nature Microbiology | VOL 6 | May 2021 | 563–573 | www.nature.com/naturemicrobiology 567

Letters Nature Microbiology

a 500 500
WT 60,000 lL-6 KO 60,000
P = 0.0253 WT

Glucose AUC (mg dl–1 min–1)

lL-6 KO

Glucose AUC (mg dl–1 min–1)

400
Blood glucose (mg dl–1)

Blood glucose (mg dl–1)

400

40,000 300 40,000

300

200
* 200
20,000 20,000
HFD ** HFD
100 100
HFD + P9 HFD + P9
HFD + P9 + Ex9–39 HFD + P9 + Ex9–39
0 0 0 0
0 30 60 90 120 0 30 60 90 120

9– +
FD

P9

9– +
FD

P9
Ex P9
39
Time (min)

H
Time (min)

Ex P9
+

39
H

+
+
FD

+
FD
FD
H

FD
H
H

H
b c e
40 40 P = 0.0007
WT lL-6 KO P = 0.0016
Rectal temperature (°C)

Rectal temperature (°C)

Plasma GLP-1 (pg ml–1)

P < 0.0001
P = 0.0007 P = 0.0014
38 38 100 P < 0.0187 P < 0.0001
Akkermansia muciniphila
P9
ICAM-2
36 36
50
34 34

32 32 0 PLC
9– +

9– +
FD

P9
FD

P9

FD

P9

FD

P9
Ex P9
39

Ex P9
39
H
H

Ca2+
H

H
+

+
+

+
FD
FD

FD

FD

IL-6
FD

FD
H
H

H
H

WT lL-6 KO GLP-1
CREB
d HFD HFD
HFD + P9 HFD + P9
HFD + P9 + Ex9–39 HFD + P9 + Ex9–39
2.0 2.5
WT lL-6 KO Il6 –/– Il6 +/+
P = 0.0147
2.0 GLP-1 × GLP-1
1.5 P = 0.0250
Thermogenesis × Thermogenesis
1.5 Body weight × Body weight
Icam2
Icam2

1.0
Glucose homeostasis × Glucose homeostasis
1.0

0.5
0.5

0 0

Fig. 4 | P9 regulates glucose homeostasis and promotes thermogenesis through the GLP-1R signalling pathway and IL-6. a, Oral glucose tolerance
testing and area under the curve data for WT and IL-6-KO HFD-fed mice that were orally administered with P9 (100 μg per mouse) with or without i.p.
injection of exendin(9–39) (Ex9–39; GLP-1R antagonist; 0.8 μg per mouse) for 8 weeks. b, Rectal temperatures of WT (left) and IL-6-KO (right) mice.
c, Plasma GLP-1 concentration in WT and IL-6-KO mice. d, Icam2 mRNA expression in the ileum of WT (left) and IL-6-KO (right) mice. e, Summary of
the identified mechanism. The figure was created using BioRender.com. For a–d, data are mean ± s.e.m. Number of mice per group: for a and b, WT: n = 7
(HFD), n = 7 (HFD + P9), n = 7 (HFD + P9 + exendin(9–39)); IL-6 KO: n = 6 (HFD), n = 6 (HFD + P9), n = 6 (HFD + P9 + exendin(9–39)); for c and d, WT:
n = 7 (HFD), n = 8 (HFD + P9), n = 7 (HFD + P9 + exendin(9–39)); IL-6 KO: n = 6 (HFD), n = 6 (HFD + P9), n = 6 (HFD + P9 + exendin(9–39)). Data were
analysed using two-way ANOVA followed by Tukey’s post hoc test (a); one-way ANOVA followed by Tukey’s post hoc test (b–c); and two-tailed unpaired
t-tests (d).

that this mediates its effects on glucose homeostasis and body mass. that normal ICAM-2 expression may require IL-6. An overview for
Interestingly, the effects of P9 on glucose homeostasis were com- graphical summary is shown in Fig. 4e.
pletely abrogated in IL-6-knockout (KO) mice, which implies that Previous studies have focused on the outer membrane of
IL-6 is essential for the induction of GLP-1 secretion by P9 and its A. muciniphila and have characterized 79 proteins that might repre-
effects on glucose homeostasis. Although P9 alone induced ther- sent candidates that mediate an interaction with the host and have
mogenesis, the coinjection of exendin(9–39) was associated with a metabolic effects45. A comparison between the proteins of the outer
lower rectal temperature, both at room temperature and after cold membrane of A. muciniphila and those of our proteomic analysis of
shock. Moreover, these effects of P9 were not observed in IL-6-KO the CFS had only one protein in common (Amuc_0576), suggest-
mice (Fig. 4b and Extended Data Fig. 10f), which suggests that ing that the proteins secreted by A. muciniphila differ from those
both GLP-1R signalling and IL-6 are involved in iBAT activation that are present in the outer membrane. The similarity between the
and thermogenesis. The circulating GLP-1 concentration was sub- nine proteins derived from SNUG-61027 and ATCC-BAA-835 was
stantially increased by the P9 treatment of wild-type (WT) mice, >98%, and the P9 sequence was 99.41% similar. To identify homo-
but not of IL-6-KO mice (Fig. 4c). Interestingly, the expression of logues of P9 or functionally similar peptides in human gut micro-
Icam2 mRNA was lower in IL-6-KO mice (Fig. 4d), which suggests biota, we compared the entire human gut metagenome (GenBank:

568 Nature Microbiology | VOL 6 | May 2021 | 563–573 | www.nature.com/naturemicrobiology

Nature Microbiology Letters
PPYE01311095.1) with the P9 sequence and found 99.78% query together, cultured in a bioreactor for 2 d and then inoculated into BHI medium
coverage and 100% sequence similarity, indicating that this protein supplemented with 5% human serum. After streaking twice more, a single colony
was picked. All of the procedures were performed under anaerobic conditions. A
is also present in human gut microbiota. list of the characteristics of each volunteer from which A. muciniphila was isolated
Receptors involved in GPCR signalling mediate the secretion is provided in Supplementary Table 2a. Whole-genome sequencing showed that
of GLP-1 from enteroendocrine cells46–48. However, ICAM-2 (also the mean nucleotide identity between two bacterial colonies was 99.08%, which
known as CD102) is principally known as an immune cell integrin implies that they were identical. Furthermore, 16S-rRNA sequence Basic Local
Alignment Search Tool analysis showed that this isolate was 100% similar to the
that is involved in the immune cell interactions that are required for
ATCC-BAA-835.
cell barrier penetration49. Here, we propose a function of ICAM- By direct streaking, A. muciniphila (SNUG-61027) was inoculated onto BHI
2, whereby it acts as a GPCR-like signalling molecule for P9 and agar (BD Bioscience) supplemented with 0.5% porcine mucin (Sigma-Aldrich)
mediates the release of GLP-1 from enteroendocrine cells. Although and 0.05% cysteine (Sigma-Aldrich). After incubating for 48 h, a single colony
SCFAs are well known as GLP-1 inducers and SCFAs act on was picked and cultured in BHI broth medium containing 5% FBS for 36 h, and
then subcultured in fresh medium. SNUG-61027 grown in FBS-based BHI broth
GPCR signalling pathway, acetate and propionate produced by A. medium was used for supernatant fractionation. To maintain anaerobic conditions,
muciniphila could not induce GLP-1 as robustly as P9 in our study. all of the procedures were performed in an anaerobic chamber (Coy Laboratory
These results indicate that P9 acts on ICAM-2 through a different Products). Lactobacillus spp. and Bifidobacterium spp. were obtained from KCTC
signal cascade distinctly with SCFAs to induce GLP-1. Furthermore, (Korean Collection for Type Culture) or isolated from either healthy fresh infant
our findings are consistent with the notion that glucose-induced faeces or frozen adult faeces using BL agar (Kisan Biotech), Bifidus Selective
Medium (BSM; Sigma-Aldrich) and TOS-propionate agar (Sigma-Aldrich) in an
GLP-1 secretion is mediated by the activation of intracellular Ca2+ anaerobic chamber. The characteristics of the donors from whom faecal samples
signalling50,51 and CREB- and PLC-mediated pathways52–54. Further were provided are shown in Supplementary Table 2b. Bacteria were cultured in
studies must determine how ICAM-2 might activate GPCR-like MRS broth for 24 h and then subcultured for another 24 h before being used in cell
downstream signalling and induce GLP-1 secretion. culture experiments.
IL-6 has multiple tissue-specific effects that are achieved in a
paracrine or endocrine manner. Hepatic disruption of IL-6 signal- Experimental animals. C57BL/6J mice (SLC) were housed (no more than four
per cage) in a pathogen-free animal facility under a 12 h–12 h light–dark cycle and
ling causes insulin resistance55, and BAT-derived IL-6 is required were given free access to food and water. Male mice (aged 6 weeks) were fed either
for the effects of BAT transplantation on glucose homeostasis56. a low-fat diet (D12450K, Research Diets) or a HFD (D12492, Research Diets),
More recently, it was shown that an acute exercise-induced increase or were fed an HFD and were also administered daily with A. muciniphila for 14
in IL-6 delays gastric emptying, thereby reducing postprandial weeks. At the end of the treatment period, the mice were euthanized and analysed.
The group sizes were n = 7–10 mice per condition, based on a previous study7.
insulin secretion in humans57. In accordance with the findings of
The mice were matched for body mass and randomized to groups before each
previous studies that A. muciniphila induces higher IL-6 expres- experiment and the variance in each dataset was compared between the groups
sion in human-derived peripheral blood mononuclear cells42, using Barlett’s test. Experiments were not performed in a blinded manner. Food
and that A. muciniphila-derived extracellular vesicles induce IL-6 intake and body mass were measured once each week.
in a dose-dependent manner58, here A. muciniphila induced the The purified protein from A. muciniphila, P9 (100 μg per mouse) was orally (or
intraperitoneally) administered to a second cohort of mice during HFD feeding
secretion of IL-6 from the gut through P9, which has beneficial and the results were compared with those of the HFD-fed group. The mice were
effects on metabolic homeostasis. Furthermore, P9 is produced by euthanized at 8 weeks and then analysed (n = 8 (low-fat diet), n = 8 (HFD), n = 8
A. muciniphila and its metabolic effects act through IL-6. Moreover, (HFD + P9)). The effects of P9 were also analysed in IL-6-KO HFD-fed mice, and
host thermogenesis is prevented when either GLP-1R or IL-6 the results were compared with those of WT mice that were administered P9.
signalling is inhibited. These results demonstrate that P9 is IL-6-KO mice (Jackson Laboratory) from the C57BL/6J background were used and
age-matched (6 weeks) and sex-matched (male) WT mice of the same background
directly involved in the effects of A. muciniphila on IL-6 and were used as controls (WT: n = 8 (HFD), n = 8 (HFD + P9); IL-6-KO: n = 6 (HFD),
GLP-1 signalling. n = 6 (HFD + P9)). All of the mice were initially acclimated to their environment
Despite the positive effects of P9, it should be considered that for 1 week while consuming a normal chow diet and were raised in the same
sustaining P9 in an ecologically competitive harsh condition under facility under the same conditions. The effects of the HFD on the metabolism—
an HFD could be challenging if A. muciniphila could not outcom- including blood glucose, weight gain and food intake—were similar in the IL-6-KO
and WT mice (Extended Data Fig. 10a,b). At the end of the experiments, the
pete other bacteria in vivo. In humans, GLP-1 is actively degraded animals were anaesthetized with isoflurane and blood samples were collected by
within 2 min (ref. 59) and, to circumvent this limitation, many stud- retro-orbital sinus puncture, then the mice were killed by cervical dislocation.
ies have focused on developing synthetic GLP-1 agonists that are
resistant to degradation by dipeptidyl peptidase 4 inhibitors. GLP-1 Glucose and insulin tolerance testing. Glucose tolerance testing was performed
receptor agonists are currently recommended as an add-on therapy using oral gavage of 2 g kg−1 body mass or i.p. administration of 1 g kg−1 body
mass glucose after 16 h of fasting (2200–1400). Blood samples were collected
to metformin for patients with type-2 diabetes and may be used as a
by tail-tip bleeding 0, 15, 30, 60, 90 and 120 min later, and the blood glucose
monotherapy in patients who cannot tolerate metformin60. However, concentrations were measured using a glucometer (Accu-Check Performa; Roche
daily subcutaneous injection of a GLP-1 agonist is associated with Diagnostics). Intraperitoneal insulin tolerance testing was performed 1 week after
some adverse effects, including nausea, vomiting, diarrhoea and the intraperitoneal glucose tolerance testing. Mice were fasted for 5 h (0900–1400),
other gastrointestinal symptoms61. We found that oral administra- insulin (Humulin R, Eli Lilly) was next injected at a concentration of 0.5 U kg−1
body mass, and then their blood glucose concentrations were measured, as
tion of P9 in mice substantially improved glucose homeostasis. The described above.
structure of the protein and its safety with regard to its potential
therapeutic use are presently under investigation. Serum insulin measurements. To measure serum insulin concentration, blood
samples were allowed to clot for 30 min at room temperature, then serum was
Methods separated by centrifugation at 4,000g for 10 min and stored at −80 °C. Serum
Isolation and cultivation of A. muciniphila strains. A. muciniphila insulin concentration was measured using the Ultra-Sensitive Mouse Insulin
(ATCC-BAA-835), purchased from the American Type Culture Collection ELISA kit (Crystal Chem).
(ATCC), was streaked onto brain–heart infusion (BHI) agar supplemented
with 0.5% porcine mucin and 0.05% cysteine in an anaerobic workstation (Coy Plasma GLP-1 measurements. For the measurement of plasma GLP-1, mice
Laboratory Products). After incubation for 48 h at 37 °C in an anaerobic jar using were fasted for 5 h in the morning. Blood samples were obtained by retro-orbital
the GasPack 100 system (BD Bioscience), bacteria were collected from the plates, sinus puncture 0 min and 10 min after oral gavage of 2 g kg−1 body mass glucose
suspended in anaerobic phosphate-buffered saline (PBS) containing glycerol, and collected into prechilled EDTA-coated tubes containing 1 μg ml−1 diprotin
and then aliquoted and stored at −80 °C. A. muciniphila suspended in 200 μl A (6019; Tocris Bioscience), centrifuged without delay, and the plasma was then
anaerobic PBS (4.0 × 108 colony-forming units per mouse) was orally administered separated and stored at −80 °C. To measure GLP-1 after a single administration,
to mice. A. muciniphila (SNUG-61027) was isolated from freshly collected faecal male mice (aged 8 weeks) were fasted for 5 h in the morning. A single dose of
samples that were obtained from three healthy individuals who had not been each respective group (PBS, P9, EcPrc, sodium acetate (S2889; Sigma-Aldrich)
treated with antibiotics during the preceding year. Each stool sample was mixed and sodium propionate (P1880; Sigma-Aldrich)) was administered i.p. at 300 μg

Nature Microbiology | VOL 6 | May 2021 | 563–573 | www.nature.com/naturemicrobiology 569

Letters Nature Microbiology
per mouse and blood samples were obtained by cardiac puncture 10 min after added and the mixture was passed through a cell strainer (100 µm). The cells were
oral gavage of 2 g kg−1 body mass glucose. The collected blood was collected into then washed twice with PBS, and the red blood cells were lysed. The cells obtained
prechilled EDTA-coated tubes containing 1 μg ml−1 diprotin A (Tocris Bioscience) were seeded into 12-well plates with prewarmed DMEM containing 4.5 g l−1 glucose
and 500 Kallikrein inhibitor units of aprotinin (A6279; Sigma-Aldrich) and plasma (HyClone Laboratories), 20% FBS (GenDEPOT) and 1% penicillin–streptomycin.
was separated by centrifugation at 4,000g for 10 min then stored at −80 °C. GLP-1 Floating cells were removed the next day, fresh prewarmed culture medium was
concentration was determined using the GLP-1 ELISA kit (RayBiotech). added, and the samples were prepared for FACS analysis or qPCR analysis was
performed after IL-6 treatment.
RNA isolation and RT–qPCR analysis. RNA was extracted from tissues using
the Easy-spin Total RNA extraction kit (iNtRON Biotechnology). RNA was Bacterial supernatant filtrate preparation. The supernatant was collected from
reverse-transcribed using the LeGene cDNA synthesis master mix (LeGene) a 150 ml culture of A. muciniphila (SNUG-61027) and filtered using filters with
according to the manufacturer’s instructions. The relative mRNA expression levels a number of differing pore sizes and molecular-mass cut-offs. Swing-bucket
were determined by qPCR using the Rotor-Gene Q (Qiagen) or QuantStudio 6 flex centrifugation (Eppendorf 5415R centrifuge; Eppendorf) was used for all of
Real-Time PCR system (Thermo Fisher Scientific). A list of the sequences of the the procedures. The supernatants were centrifuged at 7,000g for 10 min at 4 °C
primers used for RT–qPCR is provided in Supplementary Table 4. and then passed through polyethersulfone filters (0.22 μm; Merck Millipore) to
remove the residual bacterial cells. The derived filtrates were then passed through
Histological analysis. The epididymal, interscapular and inguinal fat pads were 1,000 kDa filters (Vivaspin 20 polyethersulfone ultrafiltration unit; Sartorius)
fixed in 4% formaldehyde, processed into paraffin blocks, sectioned and stained at 6,000g for 30 s. The 1,000 kDa filtrates were then loaded onto 300 kDa filters
with H&E. The slides were scanned under a light microscope using Imaging Sys (Sartorius) and centrifuged at 6,000g for a further 30 s. The 300 kDa filtrates were
(Nikon). The mean adipocyte size was measured using Image Scope (LEICA then passed through 100 kDa, 30 kDa and 10 kDa filters (Pall Microsep Centrifugal
Biosystems). Device; Pall) at 3,200g for 10 min. Each filtrate (500 μl) was frozen at −80 °C until
assayed.
Temperature measurements. The skin temperature over the iBAT was recorded The BCA Protein Assay kit (Thermo Fisher Scientific) was used to determine
using an infrared camera (T420 Compact Infrared Thermal Imaging Camera; the protein content of each filtrate, according to the manufacturer’s instructions.
FLIR) and analysed using the FLIR tools software. Eight mice per group were Proteinase K (iNtRON Biotechnology) was also added to a portion of the 100 kDa
briefly anaesthetized using isoflurane, and the mean temperature of the area filtrate at 100 µg ml−1, which was incubated at 55 °C for 1 h, and then the enzyme
surrounding the iBAT was recorded from each picture and analysed. Rectal activity was inactivated at 90 °C for 10 min.
temperature was measured using a digital thermometer (Testo 925, Testo)
according to the manufacturer’s protocol. To cold-stress the mice, they were FPLC analysis. To further separate the proteins of interest in the fractions from
individually placed in a humidity-controlled chamber at 5 °C and 40–60% bulk proteins, chromatography was performed using an AKTA Explorer fast
humidity. protein LC (FPLC) system (GE Healthcare Bio-Sciences). The FPLC system was
equipped with a sample loop of 2.0 ml and a Mono Q anion-exchange column of
Flow cytometric analysis. To isolate adipocytes from the iBAT of each mouse 5 × 50 mm (GE Healthcare), with a column volume of 1 ml column volume. Buffer
(n = 8 per group), the interscapular adipose tissue was dissected, minced finely A (10 mM potassium phosphate and 50 mM NaCl) and buffer B (10 mM potassium
with scissors and digested in Hank’s balanced salt solution (HBSS) containing phosphate and 1 M NaCl) were prepared. The 100 kDa filtrate and control medium,
collagenase (1 mg ml−1 collagenase type II, 0.5% bovine serum albumin (BSA)) each containing 25 mg protein, were applied to the FPLC system at a flow rate of
at 37 °C for 30 min. The cells were then passed through a cell strainer (100 µm), 1 ml per min, with continuous monitoring of the absorbance at 280 nm. Twenty-six
washed twice with PBS and the red blood cells were lysed. The cells were next fractions were collected. After identifying the functional fractions using a GLP-
washed with cold PBS and incubated with staining cocktail (CD45-APC-Cy7 1-secretion assay, they were selected and concentrated using a 30 kDa Microsep
(BD557659), 7AAD-PercP-Cy5.5 (BD559925), CD11b-PE-Cy7 (BD561098) and centrifugal device (Pall). For size-exclusion chromatography (gel permeation
CD206-Alexa Fluor 647 (BD565250) (BD Bioscience)) diluted at 1:200 in FACS chromatography (GPC)/SEC), the AKTA Explorer FPLC system was equipped
buffer (2% FBS in PBS) for 20 min on ice in the dark. The cells were then washed with a GPC column with a column volume of 120 ml. Buffer A (10 mM potassium
and analysed using the FACS Verse flow cytometer (BD Bioscience) and the data phosphate and 50 mM NaCl) was prepared. The samples were loaded at 3 ml min−1
were analysed using FlowJo v.10.4.2. The gating strategy for detecting M2-like with continuous monitoring, and 40 fractions were collected and stored at −80 °C
macrophages (CD11b+CD206+) in the iBAT is described in Extended Data Fig. 2i. until analysis.

Cell culture and measurement of GLP-1 secretion. The NCI-H716 human LC–MS/MS analysis. The fractions of the A. muciniphila culture supernatant were
intestinal cell line was obtained from the ATCC (ATCC-CCL-251) and cultured sent to the Proteomics Core Facility at Seoul National University Hospital in Seoul,
in RPMI 1640 supplemented with 10% FBS (GenDEPOT) and 1% penicillin– Korea for protein identification and analysis. The samples were analysed using an
streptomycin (Life Technologies) at 37 °C under 5% CO2. The GLUTag cell line was LC–MS instrument consisting of an Ultimate 3000 RSLC system (Dionex), coupled
a gift from D. J. Drucker, University of Toronto, and was cultured in Dulbecco’s via a nanoelectrospray ion source (Thermo Fisher Scientific) to a Q Exactive
modified Eagle’s medium (DMEM) containing 10% FBS and 1% penicillin– plus Orbitrap (Thermo Fisher Scientific), according to previously described
streptomycin at 37 °C under 5% CO2. For the GLP-1-secretion assay, the cells were procedures63. MS raw files were searched using Maxquant software (v.1.6.1.0)64 and
cultured in collagen-coated 96-well plates (3 × 105 cells per ml) overnight. On the the Andromeda search engine65 against the UniProt reference proteome database
day of treatment, the cells were starved in HBSS containing 0.2% BSA for 2 h, and (A. muciniphila; UP000001031) and the UniProt reference proteome database (Bos
then treated with bacterial pellets or supernatants for 2 h. The cell supernatants Taurus; UP000009136). The false-discovery rate was set to 0.01 for both proteins
were then collected for the measurement of GLP-1 concentration. The RAW 264.7 and peptides with a minimum length of six amino acids, and was determined by
and CT26 cell lines were purchased from the Korean Cell Line Bank and cultured searching a reverse database.
in DMEM containing 10% FBS and 1% penicillin–streptomycin. Immortalized
BACs were provided by Kai Ge62 (NIDDK, National Institutes of Health (NIH)) Plasmid constructs and protein expression. The nucleotide sequences of the
and maintained in DMEM containing 10% FBS and 1% penicillin–streptomycin. following genes were obtained by whole-genome sequencing (Supplementary
Each cell line was authenticated by morphology and growth characteristics as well Table 5) conducted at Macrogen using the isolated A. muciniphila strain SNUG-
as manufacturing companies, and was assessed for mycoplasma contamination 61027: pgk (UniProt: B2UKW8), Amuc_0017 (UniProt: B2UL75), fucI (UniProt:
regularly. E. coli lipopolysaccharide (L2630) was obtained from Sigma-Aldrich. B2UN39), Amuc_0405 (UniProt: B2UND1), Amuc_1282 (UniProt: B2URM2),
glsA (UniProt: B2UL96), Amuc_1829 (UniProt: B2UN36), Amuc_0576 (UniProt:
Human intestinal epithelial cell culture and treatment. Primary human intestinal B2UPD6) and Amuc_1631 (UniProt: B2UM07). The homologous protein to P9
epithelial cells (CC-2931, frozen vial of 800,000 cells) were purchased from Lonza was identified in E. coli DH5α (EcPrc) in the UniProt database (https://www.
and authenticated by Lonza. Cells were grown using the SmGM-2 BulletKit (CC- uniprot.org/). The Amuc_1100-expressing plasmid was previously described7. Each
3182) with supplements (CC-4149) provided by Lonza. Cells were seeded into a of the corresponding cDNAs was cloned into the pET-26b plasmid (Novagen),
96-well flat-bottomed collagen-coated plate at 5 × 104 cells per well and incubated which contains an isopropyl β-d-1-thiogalactopyranoside (IPTG)-inducible
overnight. On the day of treatment, the cells were starved in HBSS containing 0.2% promoter. A list of the primer sequences used to generate the construct is provided
BSA for 1 h and then treated with P9 (50 μg ml−1) for 1 h. The cell supernatants in Supplementary Table 6. Sequences in bold are restriction sites for the Nde1
were then collected for the measurement of GLP-1 concentration and the cell and XhoI enzymes (Thermo Fisher Scientific). A His-tag was added to the C
pellets were obtained for qPCR analysis. terminus of each protein for subsequent purification. The correct sequences
of the resulting plasmids were then confirmed. These vectors were transfected
Isolation of primary mouse preadipocytes. To isolate primary BACs, into BL21 Escherichia coli (RRID: WBHT115(DE3)) and grown in LB broth
interscapular adipose tissue was dissected from male C57BL/6J mice. Six depots containing kanamycin (50 μg ml−1), into which 1.0 mM IPTG was added during the
were pooled, minced finely with scissors and digested for 30 min in HBSS mid-exponential growth stage, after which a further 4 h was provided for protein
containing collagenase (1 mg ml−1 collagenase type II, 2% BSA, 25 mM HEPES) at expression. The cells were pelleted by centrifuging for 10 min at 10,733g and the
37 °C for 30 min. To stop the enzymatic activity, 10% FBS-containing medium was cell pellets were stored at −80 °C until lysis. The cell pellets were resuspended

570 Nature Microbiology | VOL 6 | May 2021 | 563–573 | www.nature.com/naturemicrobiology

Nature Microbiology Letters
and lysed by sonication (Vibracell VCX500, Sonics & Materials). For protein and sequenced on the MiSeq platform using a paired-end 2 × 300 bp reagent kit
purification, TALON Metal Affinity resin and the HisTALON buffer kit (Takara (Illumina). Raw reads were demultiplexed using CASAVA (v.1.8) and imported
Bio USA) were used according to the manufacturer’s instructions, with minor into the QIIME2 pipeline (v.2020.2). Low-quality sequences, duplicated sequences
modifications. The purified proteins were then dialysed against endotoxin-free and chimeric sequences were eliminated using Divisive Amplicon Denoising
distilled water. Removal of endotoxin from the purified protein solution was Algorithm 2 (DADA2). Sequences were classified into operational taxonomic
accomplished by treatment with 1% Triton X-114 (Sigma-Aldrich) for 30 min units at the 99% similarity level against the Greengenes 13.8 database using the
at 4 °C. The phase containing the endotoxin was separated by centrifugation at closed-reference method. The ALDEx2 package in R was used to compare the
12,000 r.p.m. for 10 min. This procedure was repeated three times. The purified differential abundance of taxonomic units in the HFD and HFD + A. muciniphila
proteins were then incubated three times with SM-2 beads (Bio-Rad Laboratories) groups.
for 1 h at 4 °C to remove residual Triton X-114. The purification of each protein
was confirmed using SDS–polyacrylamide gel electrophoresis and Coomassie blue Indirect calorimetry and body composition measurements. C57BL/6J mice
staining by the presence of single band of the expected size. were fed a 60% HFD and were orally administered 200 μl of P9 (100 μg per mouse)
or the same volume of endotoxin-free distilled water for 10 d. The mice were
RNA-seq analysis. RNA was extracted from NCI-H716 cells after treatment with individually placed into metabolic chambers (Oxylet system, Panlab-Harvard
P9 (50 μg ml−1) for 30 min and then processed for RNA-seq analysis at Macrogen. Apparatus) and acclimated for 2 d before measurements were made. Oxygen
Macrogen’s application, which is based on the Illumina sequencing protocol, was consumption (VO2) and carbon dioxide production (VCO2) were calculated every
used to analyse the transcriptome. Three control and three treatment samples 3 min using METABOLISM (v.2.2.01, Panlab-Harvard Apparatus). The respiratory
were analysed. Differentially expressed genes were defined as showing >1.5-fold quotient (RQ) was calculated as VCO2/VO2 and energy expenditure (EE) was
changes in expression, with P < 0.05. The use of these criteria resulted in the calculated using the following formula: EE = VO2 × 1.44 × (3.815 + 1.232 × RQ).
selection of 25 differentially expressed genes. Gene set enrichment analysis was Fatty acid oxidation was calculated using the following formula:
performed using Gene Ontology and the KEGG database. Of the 27,685 genes (1.6946 × VO2) − (1.7012 × VCO2). The body composition of the animals
identified, 12,019 with at least one zero fragments per kilobase of transcript per (percentage of fat mass and lean mass) was also measured after the indirect
million mapped reads were excluded, which left 15,666 genes to be processed for calorimetry measurements using the Minispec LF90 Analyzer (Bruker Optics),
Gene Ontology analysis. according to the manufacturer’s instructions. Body mass was measured before these
measurements to enable the calculation of the fat and lean-mass percentages.
Calcium measurement. A Fluo-4 NW Assay kit (Thermo Fisher Scientific) was
used to measure intracellular calcium influx on a Flexstation 3 microplate reader GLP-1R β-arrestin assay. For the β-arrestin assay, HTLA cells (HEK293 cell line
(Molecular Devices). Cells were seeded onto poly-l-lysine-coated (Sigma-Aldrich) stably expressing a tTA-dependent luciferase reporter and a β-arrestin2-TEV fusion
96-well black-walled plates at 2 × 104 cells per well and incubated overnight. The gene)69 were a gift from B. L. Roth and were maintained in DMEM with 10% FBS,
next day, the growth medium was replaced with 100 μl probenecid per well to penicillin (100 U ml−1), streptomycin (100 μg ml−1), puromycin (2 μg ml−1) and
prevent the extrusion of the dye from the cells, and the cells were then incubated hygromycin B (100 μg ml−1) in a humidified atmosphere at 37 °C in 5% CO2. The
for 30 min at 37 °C and at room temperature for a further 30 min. Flexstation 3 was confluent HTLA cells were transfected with 5 μg of GLP-1R-Tango and 500 ng
used to analyse the calcium influx (excitation/emission, 485 nm/535 nm). of pRL-SV40P construct (66295, 27163; Addgene) using Lipofectamine 3000
(L3000015; Invitrogen). The next day, transfected cells were transferred at 40,000
LRC analysis based on TriCEPS. TriCEPS (Dualsystems Biotech) was used for cells per well into poly-l-lysine-coated 96-well plates. After overnight incubation,
LRC analysis, as previously described66. TriCEPS was left to bind to the ligand exendin-4 (1933; Tocris Bioscience) or P9 was prepared and diluted in HBSS
in pH 8.2, 25 mM HEPES solution at room temperature with gentle shaking for with 20 mM HEPES (pH 7.4), then exendin-4 (50 μM) or P9 (50 μg ml–1) was
90 min, and then this was added to oxidized NCI-H716 cells, which were incubated administered to HTLA cells for 24 h. Next, the cells were washed, lysed and analysed
at 4 °C in the dark for 90 min on a rotator. After centrifugation, the cell pellets were by luminescence using the Infinite 200 Pro microplate reader (Tecan Group) and
collected and, under ice-cold conditions, were sent for LC–MS/MS analysis at the Dual-Glo Luciferase Assay System (2920; Promega). The luciferase activities
Dualsystems Biotech. LRC data were analysed using a statistical ANOVA model67 were normalized to Renilla luminescence and relative activity were calculated.
and the adjusted P value for the differential abundance of each protein was plotted
against the magnitude of the fold-enrichment. Ethics statement. The experimental procedures were reviewed and approved by
the Institutional Animal Care and Use Committee of Seoul National University
Antibody inhibition test. GLUTag and NCI-H716 cells were incubated with an (SNU-180723-1-1). All of the experiments were performed in accordance with the
ICAM-2-blocking peptide (MBS823225; MyBioSource), anti-ICAM-2 antibodies relevant guidelines and regulations and were approved by the Institutional Review
(BD Pharmigen, 3C4 (mIC2/4); BD553326), anti-total KTN1 antibodies Board of Seoul National University (IRB No 1405/002-008) and Samsung Medical
(A44072; Antibody Solutions) or a negative control antibody (mouse anti-β Center (SMC 2014-11-023-001).
tubulin (32-2600); Thermo Fisher Scientific) for 1 h, before P9 (50 µg ml−1)
treatment for 1 h, and the supernatants were then collected and stored at −80 °C Statistical analysis. All data are expressed as the mean ± s.e.m. Mouse groups were
until they were assayed. compared using one-way or two-way ANOVA, followed by Tukey’s post hoc test
or the unpaired t-test. For in vitro data, the Kruskal–Wallis, followed by Dunn’s
Caecal metabolite analysis. For nuclear magnetic resonance (NMR)-based post hoc test or the Mann–Whitney U-test was performed. Statistical analysis was
metabolomic analysis, caecal contents (50–100 mg) were mixed with 600 μl performed using GraphPad Prism v.7.04. Statistical significance is indicated by
deionized distilled water, vortexed and homogenized using a tissue homogenizer. asterisks (*) or phi (Φ); Φ or *P < 0.05, ΦΦ or **P < 0.01, ΦΦΦ or ***P < 0.001, ΦΦΦΦ or
After centrifugation (14,000g at 4°C) for 10 min, 60 μl of deuterium oxide ****P < 0.0001.
(containing 0.025 mg ml−1 3-(trimethylsilyl) propionic acid-d4 sodium salt, 60 μl
of 1 mM imidazole, 60 μl of 2 mM NaN3 and 120 μl of 0.5 M KH2PO4 were added Reporting Summary. Further information on research design is available in the
to 300 μl of the supernatant. The mixtures were vortexed for 1 min and centrifuged Nature Research Reporting Summary linked to this article.
at 14,000g for 10 min. The clear supernatants were then transferred to NMR tubes
(5 mm; Wilmad-Lab Glass) for NMR analysis. The 1H-NMR spectra were acquired Data availability
using a Varian 500 MHz NMR system (Varian) spectrometer equipped with a cold RNA-seq data and 16S rRNA gene sequencing data that support the findings of
flow-probe. 1H-NMR spectra were collected at 25 °C using the water presaturation this study have been deposited at the European Nucleotide Archive (ENA) and
pulse sequence. Spectra were collected with 64 transients using a 4 s acquisition are publicly available under the accession numbers PRJEB36198 (RNA-seq) and
time and a 2 s recycle delay. Tentative assignments of 1H NMR signals were PRJEB36225 (16S rRNA gene sequencing). Whole-genome sequencing data of
performed using Chenomx NMR Suite v.8.3 (Chenomx), according to the Human SNUG-61027 is deposited at the ENA under accession number PRJEB42664 and
Metabolome Database and a previous publication68. is publicly available. Similarly, metabolomic data were deposited at MetaboLights
under accession number MTBLS1824 and are publicly available. Source data are
Analysis of and bioinformatics pipeline for the caecal microbiota. The caecum provided with this paper.
of each mouse was collected in a sterile tube, immediately frozen in liquid nitrogen
and stored at −80 °C until further use. DNA was extracted using the QiaAmp Received: 16 March 2020; Accepted: 16 February 2021;
Fast DNA Stool Mini Kit (Qiagen), with slight modification of the manufacturer’s
protocol. Caecal samples were also homogenized with stainless steel beads (5 mm;
Published online: 5 April 2021
Qiagen) after homogenization with glass beads (0.1 mm). Illumina-adapted
universal primers 515F/806R, which target the V4 region of the 16S rRNA, were References
used for DNA amplification, and the DNA was then purified using the QIAquick 1. Geach, T. Gut microbiota: mucin-munching bacteria modulate glucose
PCR purification Kit (Qiagen). Purified amplicons were quantified using the metabolism. Nat. Rev. Endocrinol. 13, 66 (2017).
KAPA Library Quantification Kit (KAPA Biosystems). The concentration of each 2. Chevalier, C. et al. Gut microbiota orchestrates energy homeostasis during
sample was normalized to 4 nM, and they were then pooled into a single tube cold. Cell 163, 1360–1374 (2015).

Nature Microbiology | VOL 6 | May 2021 | 563–573 | www.nature.com/naturemicrobiology 571

Letters Nature Microbiology
3. Parekh, P. J., Balart, L. A. & Johnson, D. A. The influence of the gut 31. Lebrun, L. J. et al. Enteroendocrine L cells sense LPS after gut barrier injury
microbiome on obesity, metabolic syndrome and gastrointestinal disease. to enhance GLP-1 secretion. Cell Rep. 21, 1160–1168 (2017).
Clin. Transl. Gastroenterol. 6, e91 (2015). 32. Gribble, F. M. & Reimann, F. Enteroendocrine cells: chemosensors in the
4. Sonnenburg, J. L. & Backhed, F. Diet–microbiota interactions as moderators intestinal epithelium. Annu. Rev. Physiol. 78, 277–299 (2016).
of human metabolism. Nature 535, 56–64 (2016). 33. Tennoune, N. et al. Bacterial ClpB heat-shock protein, an antigen-mimetic of
5. Everard, A. et al. Cross-talk between Akkermansia muciniphila and intestinal the anorexigenic peptide α-MSH, at the origin of eating disorders. Transl.
epithelium controls diet-induced obesity. Proc. Natl Acad. Sci. USA 110, Psychiatry 4, e458 (2014).
9066–9071 (2013). 34. Everard, A. et al. Responses of gut microbiota and glucose and lipid
6. Cani, P. D. et al. Endocannabinoids—at the crossroads between the gut metabolism to prebiotics in genetic obese and diet-induced leptin-resistant
microbiota and host metabolism. Nat. Rev. Endocrinol. 12, 133–143 (2016). mice. Diabetes 60, 2775–2786 (2011).
7. Plovier, H. et al. A purified membrane protein from Akkermansia muciniphila 35. Rooks, M. G. & Garrett, W. S. Gut microbiota, metabolites and host
or the pasteurized bacterium improves metabolism in obese and diabetic immunity. Nat. Rev. Immunol. 16, 341–352 (2016).
mice. Nat. Med. 23, 107–113 (2017). 36. Donaldson, G. P. et al. Gut microbiota utilize immunoglobulin A for mucosal
8. Depommier, C. et al. Supplementation with Akkermansia muciniphila in colonization. Science 360, 795–800 (2018).
overweight and obese human volunteers: a proof-of-concept exploratory 37. Ansaldo, E. et al. Akkermansia muciniphila induces intestinal adaptive
study. Nat. Med. 25, 1096–1103 (2019). immune responses during homeostasis. Science 364, 1179–1184 (2019).
9. The GBD 2015 Obesity Collaborators. Health effects of overweight and 38. Shin, N. R. et al. An increase in the Akkermansia spp. population induced by
obesity in 195 countries over 25 years. N. Engl. J. Med. 377, 13–27 (2017). metformin treatment improves glucose homeostasis in diet-induced obese
10. Lim, M. Y. et al. The effect of heritability and host genetics on the gut mice. Gut 63, 727–735 (2014).
microbiota and metabolic syndrome. Gut 66, 1031–1038 (2017). 39. Krieger, J. P. et al. Glucagon-like peptide-1 regulates brown adipose tissue
11. Dao, M. C. et al. Akkermansia muciniphila and improved metabolic health thermogenesis via the gut-brain axis in rats. Am. J. Physiol. Regul. Integr.
during a dietary intervention in obesity: relationship with gut microbiome Comp. Physiol. 315, R708–R720 (2018).
richness and ecology. Gut 65, 426–436 (2016). 40. Beiroa, D. et al. GLP-1 agonism stimulates brown adipose tissue
12. Karlsson, C. L. et al. The microbiota of the gut in preschool children with thermogenesis and browning through hypothalamic AMPK. Diabetes 63,
normal and excessive body weight. Obesity 20, 2257–2261 (2012). 3346–3358 (2014).
13. Santacruz, A. et al. Gut microbiota composition is associated with body 41. Lynch, L. et al. iNKT cells induce FGF21 for thermogenesis and are required
weight, weight gain and biochemical parameters in pregnant women. Br. J. for maximal weight loss in GLP1 therapy. Cell Metab. 24, 510–519 (2016).
Nutr. 104, 83–92 (2010). 42. Ottman, N. et al. Pili-like proteins of Akkermansia muciniphila modulate host
14. Dao, M. C. et al. Akkermansia muciniphila abundance is lower in severe immune responses and gut barrier function. PLoS ONE 12, e0173004 (2017).
obesity, but its increased level after bariatric surgery is not associated with 43. Ellingsgaard, H. et al. Interleukin-6 enhances insulin secretion by increasing
metabolic health improvement. Am. J. Physiol. Endocrinol. Metab. 317, glucagon-like peptide-1 secretion from L cells and alpha cells. Nat. Med. 17,
E446–E459 (2019). 1481–1489 (2011).
15. Zhang, L. et al. Akkermansia muciniphila can reduce the damage of gluco/ 44. Kahles, F. et al. GLP-1 secretion is increased by inflammatory stimuli in an
lipotoxicity, oxidative stress and inflammation, and normalize intestine IL-6-dependent manner, leading to hyperinsulinemia and blood glucose
microbiota in streptozotocin-induced diabetic rats. Pathog. Dis. https://doi. lowering. Diabetes 63, 3221–3229 (2014).
org/10.1093/femspd/fty028 (2018). 45. Ottman, N. et al. Characterization of outer membrane proteome of
16. Liu, J. et al. Gypenosides reduced the risk of overweight and insulin Akkermansia muciniphila reveals sets of novel proteins exposed to the human
resistance in C57BL/6J mice through modulating adipose thermogenesis and intestine. Front. Microbiol. 7, 1157 (2016).
gut microbiota. J. Agric. Food Chem. 65, 9237–9246 (2017). 46. Hauge, M. et al. Gq and Gs signaling acting in synergy to control GLP-1
17. Gao, X. et al. Polyphenol- and caffeine-rich postfermented pu-erh tea improves secretion. Mol. Cell. Endocrinol. 449, 64–73 (2017).
diet-induced metabolic syndrome by remodeling intestinal homeostasis in 47. Ulven, T. Short-chain free fatty acid receptors FFA2/GPR43 and FFA3/GPR41
mice. Infect. Immun. https://doi.org/10.1128/IAI.00601-17 (2018). as new potential therapeutic targets. Front. Endocrinol. 3, 111 (2012).
18. Depommier, C. et al. Pasteurized Akkermansia muciniphila increases 48. Husted, A. S., Trauelsen, M., Rudenko, O., Hjorth, S. A. & Schwartz, T. W.
whole-body energy expenditure and fecal energy excretion in diet-induced GPCR-mediated signaling of metabolites. Cell Metab. 25, 777–796 (2017).
obese mice. Gut Microbes https://doi.org/10.1080/19490976.2020.1737307 49. Gorina, R., Lyck, R., Vestweber, D. & Engelhardt, B. β2 integrin-mediated
(2020). crawling on endothelial ICAM-1 and ICAM-2 is a prerequisite for
19. Gribble, F. M. & Reimann, F. Function and mechanisms of enteroendocrine transcellular neutrophil diapedesis across the inflamed blood-brain barrier. J.
cells and gut hormones in metabolism. Nat. Rev. Endocrinol. 15, 226–237 Immunol. 192, 324–337 (2014).
(2019). 50. Kuhre, R. E., Frost, C. R., Svendsen, B. & Holst, J. J. Molecular mechanisms
20. Greiner, T. U. & Backhed, F. Microbial regulation of GLP-1 and L-cell biology. of glucose-stimulated GLP-1 secretion from perfused rat small intestine.
Mol. Metab. 5, 753–758 (2016). Diabetes 64, 370–382 (2015).
21. Wang, L. et al. A purified membrane protein from Akkermansia muciniphila 51. Flamez, D. et al. Altered cAMP and Ca2+ signaling in mouse pancreatic islets
or the pasteurised bacterium blunts colitis associated tumourigenesis by with glucagon-like peptide-1 receptor null phenotype. Diabetes 48, 1979–1986
modulation of CD8+ T cells in mice. Gut https://doi.org/10.1136/ (1999).
gutjnl-2019-320105 (2020). 52. Lee, J. H., Wen, X., Cho, H. & Koo, S. H. CREB/CRTC2 controls GLP-
22. Tolhurst, G. et al. Short-chain fatty acids stimulate glucagon-like peptide-1 1-dependent regulation of glucose homeostasis. FASEB J. 32, 1566–1578
secretion via the G-protein-coupled receptor FFAR2. Diabetes 61, 364–371 (2018).
(2012). 53. Shin, S. et al. CREB mediates the insulinotropic and anti-apoptotic effects of
23. Psichas, A. et al. The short chain fatty acid propionate stimulates GLP-1 and GLP-1 signaling in adult mouse β-cells. Mol. Metab. 3, 803–812 (2014).
PYY secretion via free fatty acid receptor 2 in rodents. Int. J. Obes. 39, 54. Bala, V. et al. Release of GLP-1 and PYY in response to the activation of G
424–429 (2015). protein-coupled bile acid receptor TGR5 is mediated by Epac/PLC-epsilon
24. Li, Z. et al. Butyrate reduces appetite and activates brown adipose tissue via pathway and modulated by endogenous H2S. Front. Physiol. 5, 420 (2014).
the gut-brain neural circuit. Gut 67, 1269–1279 (2018). 55. Wunderlich, F. T. et al. Interleukin-6 signaling in liver-parenchymal cells
25. Yadav, H., Lee, J. H., Lloyd, J., Walter, P. & Rane, S. G. Beneficial metabolic suppresses hepatic inflammation and improves systemic insulin action. Cell
effects of a probiotic via butyrate-induced GLP-1 hormone secretion. J. Biol. Metab. 12, 237–249 (2010).
Chem. 288, 25088–25097 (2013). 56. Stanford, K. I. et al. Brown adipose tissue regulates glucose homeostasis and
26. Chambers, E. S. et al. Effects of targeted delivery of propionate to the human insulin sensitivity. J. Clin. Invest. 123, 215–223 (2013).
colon on appetite regulation, body weight maintenance and adiposity in 57. Lang Lehrskov, L. et al. Interleukin-6 delays gastric emptying in humans with
overweight adults. Gut 64, 1744–1754 (2015). direct effects on glycemic control. Cell Metab. https://doi.org/10.1016/j.
27. Lund, M. L. et al. L-cell differentiation is induced by bile acids through cmet.2018.04.008 (2018).
GPBAR1 and Paracrine GLP-1 and serotonin signaling. Diabetes 69, 614–623 58. Kang, C. S. et al. Extracellular vesicles derived from gut microbiota, especially
(2020). Akkermansia muciniphila, protect the progression of dextran sulfate
28. Albaugh, V. L. et al. Role of bile acids and GLP-1 in mediating the metabolic sodium-induced colitis. PLoS ONE 8, e76520 (2013).
improvements of bariatric surgery. Gastroenterology 156, 1041–1051 (2019). 59. Gupta, V. Glucagon-like peptide-1 analogues: an overview. Indian J.
29. Brighton, C. A. et al. Bile acids trigger GLP-1 release predominantly by Endocrinol. Metab. 17, 413–421 (2013).
accessing basolaterally located G protein-coupled bile acid receptors. 60. John B. Buse, J. B. et al. 2019 update to: management of hyperglycemia in
Endocrinology 156, 3961–3970 (2015). type 2 diabetes, 2018. A consensus report by the American Diabetes
30. Chimerel, C. et al. Bacterial metabolite indole modulates incretin secretion Association (ADA) and the European Association for the Study of Diabetes
from intestinal enteroendocrine L cells. Cell Rep. 9, 1202–1208 (2014). (EASD). Diabetes Care 43, 487–493 (2020).

572 Nature Microbiology | VOL 6 | May 2021 | 563–573 | www.nature.com/naturemicrobiology

Nature Microbiology Letters
61. Filippatos, T. D., Panagiotopoulou, T. V. & Elisaf, M. S. Adverse effects of Foundation of Korea (NRF) (no. NRF-2018R1A2A1A05078258). C.H.C. was supported
GLP-1 receptor agonists. Rev. Diabet. Stud. 11, 202–230 (2014). by the Global Ph.D. Fellowship program and a NRF grant funded by the Korean
62. Park, Y. K. et al. Distinct roles of transcription factors KLF4, Krox20, and government (no. NRF-2018H1A2A1061914).
peroxisome proliferator-activated receptor γ in adipogenesis. Mol. Cell. Biol.
https://doi.org/10.1128/MCB.00554-16 (2017). Author contributions
63. Lee, H. et al. Quantitative proteomic analysis identifies AHNAK (neuroblast H.S.Y., C.H.C. and G.K. contributed to the experiment design and interpreted all of
differentiation-associated protein AHNAK) as a novel candidate biomarker the results. H.S.Y. narrowed down to the target protein using the in vitro L cell system,
for bladder urothelial carcinoma diagnosis by liquid-based cytology. Mol. performed BAT-related experiments and isolated bacterial strains from infant faecal
Cell. Proteomics 17, 1788–1802 (2018). samples. C.H.C. contributed to daily administration of treatment in mouse experiments.
64. Tyanova, S., Temu, T. & Cox, J. The MaxQuant computational platform for C.H.C. performed cloning and, with assistance from M.S.Y., carried out protein
mass spectrometry-based shotgun proteomics. Nat. Protoc. 11, 2301–2319 expression. M.S.Y. and S.J.J. provided technical assistance with tissue sampling. S.J.J.
(2016). performed QIIME II with correlation analysis and the GLP-1R β-arrestin assay. H.J.Y.
65. Cox, J. et al. Andromeda: a peptide search engine integrated into the participated in discussions and provided all of the assistance. J.-h.K. performed a single
MaxQuant environment. J. Proteome Res. 10, 1794–1805 (2011). injection of P9 in mice and analysed systemic GLP-1. D.H. performed the proteomic
66. Frei, A. P. et al. Direct identification of ligand-receptor interactions on living analysis. Metabolomic analysis and isolation of the A. muciniphila strain, with assistance
cells and tissues. Nat. Biotechnol. 30, 997–1001 (2012). from K.H.C., and the culture conditions of A. muciniphila were optimized by S.H.M.;
67. Frei, A. P., Moest, H., Novy, K. & Wollscheid, B. Ligand-based receptor T.-W.N. and K.L. performed the FPLC analysis. S.-J.L. and Y.-J.K. performed the
identification on living cells and tissues using TRICEPS. Nat. Protoc. 8, metabolic chamber analysis. H.S.Y. wrote the manuscript with C.H.C. and edited the
1321–1336 (2013). manuscript. All of the authors discussed the results and approved the final text of the
68. Le Gall, G. et al. Metabolomics of fecal extracts detects altered metabolic manuscript.
activity of gut microbiota in ulcerative colitis and irritable bowel syndrome. J.
Proteome Res. 10, 4208–4218 (2011).
69. Kroeze, W. K. et al. PRESTO-Tango as an open-source resource for Competing interests
interrogation of the druggable human GPCRome. Nat. Struct. Mol. Biol. 22, G.K. is a founder of KoBioLabs Inc., a company that aims to characterize the role
362–369 (2015). of host–microbiome interactions in chronic diseases. The other authors declare no
competing interests.
Acknowledgements
We thank P. Helbling and M. Pavlou (Dualsystems Biotech) for their discussion Additional information
regarding the LRC assay and proteomics analysis; K. Ge (National Institutes of Health, Extended data is available for this paper at https://doi.org/10.1038/s41564-021-00880-5.
USA), J. B. Kim (Seoul National University, Korea) and O. J. Kwon (Catholic University,
Supplementary information The online version contains supplementary material
Korea) for providing the immortalized BACs; B. L. Roth (The University of North
available at https://doi.org/10.1038/s41564-021-00880-5.
Carolina at Chapel Hill, USA) for providing HTLA cells; S. J. Lee (Korea University,
Korea) for providing the calcium inhibitors; W. K. Huh (Seoul National University, Correspondence and requests for materials should be addressed to G.K.
Korea) for providing the GPCR inhibitors; J. S. Han (Seoul National University, Korea) Peer review information Nature Microbiology thanks Marc Donath, Lesley Hoyles and
for FPLC usage; Y. K. Oh (Seoul National University, Korea) for infrared camera usage; Liping Zhao for their contribution to the peer review of this work. Peer reviewer reports
Y. Chung (Seoul National University, Korea) for temperature-controlled chamber usage; are available.
S. H. Yoon, L. Song, J. W. Kim and S. E. Choi (Seoul National University, Korea) for their
Reprints and permissions information is available at www.nature.com/reprints.
technical assistance with the tissue sampling and preparation for FACS analysis; T. J.
Ahn (KoBioLabs, Inc.) for cytokine analysis of bacterial isolates in cell lines; I. S. Cho Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
(KoBioLabs, Inc.) for producing EcPrc; and the staff at the Samsung Medical Center published maps and institutional affiliations.
for providing infant faecal samples. This work was supported by the National Research © The Author(s), under exclusive licence to Springer Nature Limited 2021

Nature Microbiology | VOL 6 | May 2021 | 563–573 | www.nature.com/naturemicrobiology 573

Letters Nature Microbiology

Extended Data Fig. 1 | See next page for caption.

Nature Microbiology | www.nature.com/naturemicrobiology

Nature Microbiology Letters
Extended Data Fig. 1 | Effects of administration of viable A. muciniphila for 14 weeks on insulin tolerance, adipocyte size, liver steatosis markers and
intestinal cytokines levels as well as immune profiling of A. muciniphila effects on macrophage and colon cells. a, Body mass after administration of a
low-fat (LF), high-fat (HF), or a high-fat diet supplemented with A. muciniphila (HF + Akk). b, Intraperitoneal glucose tolerance testing (IPGTT) data are
presented in dot plots and as the areas under the curves (AUCs). c, Fat mass (g g body weight (BW)−1), d, plasma insulin concentration, e, liver mass (g g
BW−1), and f, H&E staining of liver sections (Scale bar = 100 μm, n = 3 mice per group). g, mRNA expression of fatty acid and ß-oxidation genes in liver. h,
Relative abundance of cecal metabolites, including short-chain fatty acids (SCFA), branched-chain amino acids (BCAA), tricarboxylic acid intermediates
(TCAi), and metabolites involved in creatine synthesis, quantified by nuclear magnetic resonance (NMR). mRNA expression of cytokines in (i) colon
and (j) ileum in HF and HF + Akk group. IL-6 secretion was measured after the stimulation of k, CT26 and l, Raw264.7 cell lines with Lactobacillus spp.,
Bifidobacterium spp., or Akk (ATCC BAA-835) (cell to bacteria: 1:10). TNF-α expression in response to Akk, Lactobacillus spp., and Bifidobacterium spp in m,
CT26 cell lines and n, Raw 264.7 cell lines. The experiment was repeated twice. E. coli lipopolysaccharide (LPS) was used as a positive control. Data are
presented as the means ± SEMs. Number of mice per group for a–b: LF: 10, HF: 10, HF + Akk: 10. Number of mice per group for c: LF: 6, HF: 10, HF + Akk:
10. Number of mice per group for d–e and g–j: HF: 10, HF + Akk: 10. Data were analyzed using one-way ANOVA, followed by Tukey’s test for a and b (right
panel) and two-way ANOVA, followed by Tukey’s test for b (left panel). For c–e and g–j, the two-tailed unpaired t-test was used to analyze the data.
k–n represent the results of three independent experiments performed in duplicate and were analyzed using the Kruskal-Wallis test, followed by Dunn’s
post-hoc test. **, *** and φφφφ indicate significant differences (P < 0.01, < 0.001 and < 0.0001, respectively).

Nature Microbiology | www.nature.com/naturemicrobiology

Letters Nature Microbiology

Extended Data Fig. 2 | The relative abundance of A. muciniphila correlates with iBAT temperature and GLP-1 secretion and A. muciniphila administration
to HF-diet-fed mice increases the M2 macrophage count in adipose tissue. a, Differences in the relative abundance of bacterial species between the
HF and HF + Akk groups are represented by ALDEx2. The differences in abundance between and within each group for individual species were analyzed.
Organisms (at the OTU and nearest neighbor species levels) with significant p values are shown as pink circles (Welch’s t-statistic, corrected using the
Benjamini-Hochberg method). b, Data showing the 16S rRNA gene count in the HF and HF + Akk groups. c–e, Scatter plots illustrating the statistical
relationship (Spearman’s correlation) between the relative abundance of A. muciniphila and metabolic phenotypes: OGTT AUC (oral glucose tolerance test
area under the curve), iBAT temp (interscapular brown adipose tissue temperature), and GLP-1 (plasma glucagon-like peptide-1 concentration). f–h, mRNA
expression of M1 and M2 macrophage markers in iBAT, igWAT, and epiWAT. i, Gating strategy for detection of M2-like macrophages (CD11b+ CD206+) in
iBAT of HF-fed mice. Data are presented as the means ± SEMs. Number of mice per group for a–h: HF: 10, HF + Akk: 10. Data in f–h were analyzed using
the two-tailed unpaired t-test.

Nature Microbiology | www.nature.com/naturemicrobiology

Nature Microbiology Letters

Extended Data Fig. 3 | Short-chain fatty acids produced by A. muciniphila are not the only factors responsible for the induction of glucagon-like
peptide-1 (GLP-1). a, Glucagon-like peptide 1 (GLP-1) secretion after treatment of NCI-H716 cells with viable A. muciniphila (cell to bacteria ratio, 1:20) or
cell-free supernatant (CFS) (10% v/v) (ATCC BAA-835 or SNUG-61027). b, GLP-1 secretion by NCI-H716 cells after treatment with various amounts of
A. muciniphila (10–100% v/v). c, GLP-1 secretion after the treatment of NCI-H716 cells with CFS (10% v/v) from A. muciniphila (ATCC BAA-835, SNUG-
61027), Korean fecal strains, Lactobacillus spp., or Bifidobacterium spp. d, Short-chain fatty acids (SCFA) derived from the CFS of A. muciniphila, measured
by GC-MS. e, GLP-1 secretion by NCI-H716 cells induced by the CFS of A. muciniphila (100% v/v), acetate (1 mM, 10 mM), or propionate (1 mM or 10 mM).
Data are presented as the means ± SEMs. The data in a and b represent the results of two independent experiments performed in duplicate. The data
in c–e are the results of triplicate analyses. The data in a–e were analyzed using the Kruskal-Wallis test, followed by Dunn’s post-hoc test. *, **, and
*** indicate significant differences (P < 0.05, < 0.01, and < 0.001, respectively).

Nature Microbiology | www.nature.com/naturemicrobiology

Letters Nature Microbiology

Extended Data Fig. 4 | See next page for caption.

Nature Microbiology | www.nature.com/naturemicrobiology

Nature Microbiology Letters
Extended Data Fig. 4 | Profiling of A. muciniphila supernatant fractions that induced glucagon-like peptide-1 (GLP-1) secretion. a, Schematic workflow
used to identify GLP-1-inducing fractions derived from A. muciniphila. b, The cell-free supernatant (CFS) of A. muciniphila was fractionated using molecular
size cut-off filters, as indicated. A volume of each fraction containing 5 mg of protein was used to treat NCI-H716 cells. c, 100–300 kDa filtrate and the
30–100 kDa filtrate of A. muciniphila were treated with proteinase K (100 µg ml−1) as indicated (+PK). A control sample was used that comprised BHI
broth containing 5% FBS and was treated with proteinase K. The effects of each on GLP-1 expression was then analyzed. d, 100 K filtrates of control media
and A. muciniphila were separated by ion exchange fast protein liquid chromatography (FPLC) and 26 fractions (m1–m26) were obtained. Each fraction
was used to treat NCI-H716 cells and GLP-1 secretion was analyzed by ELISA. e, Concentrates of the m2–m4 fractions of Control and SNUG-61027 were
separated by size columns (to generate fractions G1–G34) using the FPLC system, then the procedures listed above were followed. f, Venn diagram of the
identified proteins. Sample 1, Sample 2, and Sample 3 are the 100 K concentrates obtained from size cut-off filtration, concentrates from the ion exchange
chromatography, and concentrates from the size-exclusion chromatography, respectively. g, SDS-PAGE gel of the expressed proteins were processed in
parallel in the same gel. h, Nine proteins identified by LC-MS/MS from Sample 3. Nine candidate proteins were cloned using an E.coli expression system
according to the protein sequences for the SNUG-61027 strain. The sequence similarity for each protein between SNUG-61027 and ATCC-BAA-835
was analyzed. Data are presented as the means ± SEMs. The data in b–e represent the results of triplicate experiments and were analyzed using the
Kruskal-Wallis test, followed by Dunn’s post-hoc test.

Nature Microbiology | www.nature.com/naturemicrobiology

Letters Nature Microbiology

Extended Data Fig. 5 | Intraperitoneal administration of P9 improves glucose homeostasis and body mass in normal chow (NC) diet-fed mice. a, Weight
gain was measured in NC-fed mice administered candidate proteins intraperitoneally (i.p.) (P1, P5, P9, or Amuc_1100) at 100 μg per mouse for 2 weeks. b,
Oral glucose tolerance testing (OGTT) was performed and the AUCs were calculated. The data in c represent the effects of P9 derived from A. muciniphila
and EcPrc (S41 family derived from E. coli) on GLP-1 secretion by NCI-H716 cells. d, Plasma GLP-1 levels were measured after the single i.p. injection
(300 μg per mouse) of each respected group. Data are presented as the means ± SEMs. Number of mice per group for a and b: NC: 8, P1: 8, P5: 8, P9:
8, Amuc_1100: 8 and for d, NC: 8, ErPrc: 8, Acetate: 8, Propionate: 8. Data were analyzed using one-way ANOVA, followed by Tukey’s post-hoc test for
a, d. For b, data were analyzed using two-way ANOVA, followed by Tukey’s post-hoc test. The data in c are the results of duplicate experiments and were
analyzed using the Kruskal-Wallis test, followed by Dunn’s post-hoc test. * or φ indicate significant differences, P < 0.05.

Nature Microbiology | www.nature.com/naturemicrobiology

Nature Microbiology Letters

Extended Data Fig. 6 | Oral administration of P9 reduces adipose tissue mass and induces thermogenesis in high-fat diet-fed mice. a, Gross appearance
of the adipose tissue depots. b, Representative infrared thermographic images of the temperatures of the mice are presented at room temperature (RT)
(top) and after a cold shock at 5 °C for 4 h (bottom) (n = 3 per group).

Nature Microbiology | www.nature.com/naturemicrobiology

Letters Nature Microbiology

Extended Data Fig. 7 | Indirect calorimetry and body composition analysis following acute treatment with P9. Mice were fed a high-fat diet and were
orally administered P9 protein 100 µg or vehicle for 10 days (n = 8 mice per group). Oxygen consumption (VO2), carbon dioxide production (VCO2),
respiratory quotient (RQ), and energy expenditure were calculated every 3 min using METABOLISM software (V2.2.01, Panlab-Harvard Apparatus). Fatty
acid oxidation (FAO) was calculated using the following formula: (1.6946×VO2)-(1.7012×VCO2). a, Body composition of the mice (% fat mass, lean mass),
measured using a Minispec LF90 analyzer. b, Energy expenditure, c, FAO, and d, RQ. Data are presented as the means ± SEMs. Number of mice per group:
HF: 8, HF + P9: 8. Data were analyzed using the two-tailed unpaired t-test. *, ** and *** indicate significant differences (P < 0.05, < 0.01 and < 0.001,
respectively).

Nature Microbiology | www.nature.com/naturemicrobiology

Nature Microbiology Letters

Extended Data Fig. 8 | The induction of GLP-1 secretion, by P9 involves activation of the CREB signaling pathway, but was attenuated by GPCR
antagonists. a, Kinase phospho-profiles after P9 or vehicle treatment of NCI-H716 cells for 10 min, performed using a Proteome profile phospho-kinase
array kit (Red square: p-CREB, Blue square: p-HSP27) (experiments were performed in duplicate). b, Ca2+ influx into NCI-H716 (left) and GLUTag (right)
cells after treatment with 50 μg ml−1 or 100 μg ml−1 of P9 (n = 6 per group). c, NCI-H716 cells were treated with calcium inhibitors (10 μM) 15 min before the
P9 treatment (100 μg ml−1) for 2 h, then GLP-1 secretion was quantified (n = 6 per group). d, GLP-1 receptor (GLP-1R) beta-arrestin activity test data (n = 3
per group) (exendin-4: GLP-1R agonist). e, NCI-H716 cells were treated with GPCR antagonists (10 μM) or vehicle (DMSO) for 15 min before P9 treatment
(50 μg ml−1), then incubated for an additional 2 h, after which GLP-1 secretion was analyzed (n = 6 per group). Data are presented as the means ± SEMs.
The data in c and e represent the results of three independent experiments. The data in a and c–e were analyzed using the Kruskal-Wallis test, followed by
Dunn’s post-hoc test.

Nature Microbiology | www.nature.com/naturemicrobiology

Letters Nature Microbiology

Extended Data Fig. 9 | P9 induces IL-6 secretion by macrophage cell lines and IL-6 directly regulates thermogenic gene expression in brown adipocytes.
a, IL-6 expression in Raw264.7 cells after P1, P5, or P9 treatment (10 μg ml−1) overnight. b, Thermogenic gene expression in immortalized brown
preadipocytes (BAC) and c, Ucp1 gene expression in primary preadipocytes derived from interscapular brown adipose depots treated with recombinant
mouse IL-6 (200 ng ml−1) for 6 h. d, GLP-1 secretion by GLUTag cells after treatment of IL-6 in glucose condition (0.1 mM). e, GLP-1 secretion by GLUTag
cells after treatment of IL-6 or P9 or IL-6 together with P9 in glucose condition (0.1 mM). Data are presented as the means ± SEMs. Data were analyzed
using the Kruskal-Wallis test, followed by Dunn’s post-hoc test for a, d, and e; and by the Mann-Whitney test for b and c. The data represent the results
of three independent experiments performed in duplicate, except in b, where they represent the results of two independent experiments performed in
duplicate.

Nature Microbiology | www.nature.com/naturemicrobiology

Nature Microbiology Letters

Extended Data Fig. 10 | P9 promotes thermogenesis via the GLP-1R signaling pathway and IL-6. a, Weight gain and food intake, and b, OGTT and ITT
data after 8 weeks of HF-diet-feeding (WT: n = 8, IL-6KO: n = 6). c, Insulin tolerance testing (ITT) and AUC data for wild type (WT) and IL-6 knockout
(IL-6 KO) HF-fed mice that were orally administered P9 (100 μg per mouse) ± i.p. injected with exendin 9–39 (a GLP-1R antagonist, 0.8 μg per mouse)
for 8 weeks. n = 8 mice per group for WT (left) and n = 6 mice per group for IL-6 KO (right) mice. d, Weight gain, e, food intake (g per mouse per day),
and f, temperatures of the iBAT and eye, measured after a cold shock at 5 °C for 4 h by infrared thermography (n = 8 mice per group). Data are presented
as the means ± SEMs. Number of mice per group for c–e: WT HF: 7, HF + P9: 8, HF + P9 + Ex9–39: 7; IL-6 KO mice HF: 6, HF + P9: 6, HF + P9 + Ex9–39:
6. Number of mice per group for f: WT HF: 8, HF + P9: 8, HF + P9 + Ex9–39: 8; IL-6 KO mice: HF: 6, HF + P9: 6, HF + P9 + Ex9–39: 6. Data were analyzed
using two-way ANOVA, followed by Tukey’s post-hoc test for c (left panel), d, and e. For the data in c (right panel), one-way ANOVA, followed by Tukey’s
post-hoc test was used. The unpaired t-test was used to analyze the data in f. *, ***, and **** indicate significant differences (P < 0.05, < 0.001, and
< 0.0001, respectively).

Nature Microbiology | www.nature.com/naturemicrobiology

 nature research | reporting summary
Corresponding author(s): GwangPyo Ko
Last updated by author(s): Jan 28, 2021

Reporting Summary
Nature Research wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency
in reporting. For further information on Nature Research policies, see Authors & Referees and the Editorial Policy Checklist.

Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.
n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement
A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly
The statistical test(s) used AND whether they are one- or two-sided
Only common tests should be described solely by name; describe more complex techniques in the Methods section.

A description of all covariates tested

A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons
A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient)
AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals)

For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted
Give P values as exact values whenever suitable.

For Bayesian analysis, information on the choice of priors and Markov chain Monte Carlo settings
For hierarchical and complex designs, identification of the appropriate level for tests and full reporting of outcomes
Estimates of effect sizes (e.g. Cohen's d, Pearson's r), indicating how they were calculated
Our web collection on statistics for biologists contains articles on many of the points above.

Software and code

Policy information about availability of computer code
Data collection Excel (Versions 16.42 and 16.43; Microsoft) was used for most data collection. Reverse transcription qPCR data was collected using an
Applied Biosystems StepOnePlus PCR system using the StepOnePlus software (Version 2.1; Applied Biosystems).

Data analysis QIIME2 software package (QIIME2 version 2020.2 ; http://qiime.org/), Greengenes 13.8 database, CASAVA, METABOLISM, FlowJo
ver.10.4.2 software, GraphPad Prism 7.04
For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors/reviewers.
We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Research guidelines for submitting code & software for further information.

Data
Policy information about availability of data
All manuscripts must include a data availability statement. This statement should provide the following information, where applicable:
- Accession codes, unique identifiers, or web links for publicly available datasets
- A list of figures that have associated raw data
- A description of any restrictions on data availability
October 2018

RNA sequencing data and 16S rRNA gene sequencing data that support the findings of this study have been deposited in the European Nucleotide Archive (ENA)
with the accession number, PRJEB36198 for RNA sequencing addressed in Figure 3. and PRJEB36225 for 16S rRNA gene sequencing addressed in Extended Data Fig.
2. Whole genome sequencing data of SNUG-61027 is deposited in ENA under accession number, PRJEB42664 and publicly available.

1
 nature research | reporting summary
Field-specific reporting
Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.
Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf

Life sciences study design

All studies must disclose on these points even when the disclosure is negative.
Sample size Most of the obese mice experiments, 8 mice are used. Also the availability of animal facility limited us to do 8 mice per group. Although, for
the knockout mice we did 6 mice per group due to the availability of the facility and capacity of the in-house breeding mice.

Data exclusions We did not exclude or include any data to the experiment.

Replication All the results in the manuscripts were repeated more than two times and was all replicated.

Randomization We randomized the mice group to have the same average starting weight group as close as possible. Also other samples were randomly
sorted into experimental groups.

Blinding Our investigators were not blinded to group allocation for data collection and analysis for any experiment performed in this study due to size
of the research group. All the mice used in this experiment was handled mostly by a single investigator. Because of this issue, we did not have
additional personal that was required for blinding.

Reporting for specific materials, systems and methods

We require information from authors about some types of materials, experimental systems and methods used in many studies. Here, indicate whether each material,
system or method listed is relevant to your study. If you are not sure if a list item applies to your research, read the appropriate section before selecting a response.

Materials & experimental systems Methods

n/a Involved in the study n/a Involved in the study
Antibodies ChIP-seq
Eukaryotic cell lines Flow cytometry
Palaeontology MRI-based neuroimaging
Animals and other organisms
Human research participants
Clinical data

Antibodies
Antibodies used ICAM-2 blocking peptide (#MBS823225) (MyBioSource Inc., San Diego, CA), anti-ICAM-2 (BD Pharmigen, 3C4 (mIC2/4)
(#BD553326) (BD Bioscience San Jose, CA), anti-total KTN1 antibody (#A44072) (Antibodies Solutions, Cambridge, UK), mouse
anti-β tubulin [Thermo Fisher Scientific]), CD45-APC-Cy7 (#BD557659), CD11b-PE-Cy7 (#BD561098), and CD206-Alexa Fluor 647
(#BD565250)

Validation 1. ICAM-2 blocking peptide (#MBS823225) (MyBioSource Inc., San Diego, CA)
Host: Synthetic
Species Reactivity: Human, Mouse, Rat
Application: Blocking (BL)
The peptide is used to block Anti-CD102 Antibody reactivity.
https://www.mybiosource.com/icam2-human-mouse-rat-blocking-peptide/cd102/823225
October 2018

2. anti-ICAM-2 (BD Pharmigen, 3C4 (mIC2/4) (#553326) (BD Bioscience San Jose, CA)
Purified Rat Anti-Mouse CD102
Host: Rat
Type: Primary
Monoclonal (Clone 3C4 (mlC2/4))
Application: Blocking experiments, Flow cytometry, Immunohistochemitry (IHC), Immunoprecipitation (IP).
https://www.citeab.com/antibodies/2408256-553326-purified-rat-anti-mouse-cd102

3. anti-total KTN1 antibody (#A44072) (Antibodies Solutions, Cambridge, UK)

2
 Host: Rabbit
Reactivity: Human

nature research | reporting summary

Clonality: Polyclonal
Applications: Western Blot
https://www.antibodies.com/ktn1-antibody-a44072

4. mouse anti-β tubulin (#32-2600) [Thermo Fisher Scientific]

Host: Mouse
Reactivity: Dog, C. elegans, Human, Mouse, Non-human primate, Rat
Clonality: Monoclonal
Applications: Western Blot and Blocking experiments
https://www.thermofisher.com/antibody/product/beta-Tubulin-Antibody-clone-2-28-33-Monoclonal/32-2600

5. APC-Cy™7 Rat Anti-Mouse CD45 (BD Bioscience San Jose, CA)

Host: Rat
Reactivity: Mouse
Clonality: Monoclonal
Applications: Flow cytometry
https://www.bdbiosciences.com/eu/applications/research/stem-cell-research/cancer-research/mouse/apc-cy7-rat-anti-
mouse-cd45-30-f11/p/557659

6. PE-Cy™7 Rat Anti-CD11b (BD Bioscience San Jose, CA)

Host: Rat
Reactivity: Mouse
Clonality: Monoclonal
Applications: Flow cytometry
https://www.bdbiosciences.com/us/applications/research/stem-cell-research/mesenchymal-stem-cell-markers-bone-marrow/
mouse/negative-markers/pe-cy7-rat-anti-cd11b-m170/p/561098

7. Alexa Fluor® 647 Rat Anti-Mouse CD206 (BD Bioscience San Jose, CA)
Host: Rat
Reactivity: Mouse
Clonality: Monoclonal
Applications: Flow cytometry
https://www.bdbiosciences.com/us/reagents/research/antibodies-buffers/immunology-reagents/anti-human-antibodies/cell-
surface-antigens/alexa-fluor-647-rat-anti-mouse-cd206-mr5d3/p/565250

Eukaryotic cell lines

Policy information about cell lines
Cell line source(s) NCI-H716 human intestinal cell line (ATCC® CCL-251™)
HEK293 human embryonic kidney [HEK-293] (ATCC® CRL-1573)
GLUTag cell line was provided by Dr. D. J. Drucker (University of Toronto, Toronto, Canada)
Raw 264.7 (Korean Cell Line Bank)
Primary human intestine epithelial cells (InEpC) (CC-2931)
Immortalized brown preadipocytes were provided by Dr. Kai Ge (NIDDK, NIH, Bethesda, MD)

Authentication Cells were authenticated by morphological appearance and growth characteristics.

Mycoplasma contamination All cell lines used in this manuscripts were negative in mycoplasma contamination test.

Commonly misidentified lines The immortalized brown preadipocytes cell line was created in Kai Ge's lab (NIDDK, NIH, Bethesda, MD) and we have data in
(See ICLAC register) this manuscript (Extended Data Fig.9b). This was necessary because we needed to understand the brown preadipocyte cell
and its response to IL-6. The reference (PMID: 27777310) of this cell line is provided in the manuscript.

Animals and other organisms

Policy information about studies involving animals; ARRIVE guidelines recommended for reporting animal research
Laboratory animals The mouse we used in manuscript were C57BL/6J strain, male and mostly 8 weeks in age.

Wild animals The study did not involve wild animals.

October 2018

Field-collected samples The study did not involve samples from the Field.

Ethics oversight The animals were reviewed and approved by the Institutional Animal Care and Usage Committee (IACUC) of Seoul National
University.
Note that full information on the approval of the study protocol must also be provided in the manuscript.

3
Human research participants

nature research | reporting summary

Policy information about studies involving human research participants
Population characteristics The human subjects provided their fecal samples for research reasons the age of the participants were under 4 months in age.
The participants was n=20 in total and the gender was 7 males and 13 females. The weight was 2.7-4.0 kg and had no disease
status.

Recruitment The participants were randomly selected that were under 4 months of age. The fecal sample was collected from each human
participant.

Ethics oversight All experiments were performed in accordance with the relevant guidelines and regulations and were approved by the
Institutional Review Board of Seoul National University (IRB No 1405/002-008) and Samsung Medical Center (SMC
2014-11-023-001).
Note that full information on the approval of the study protocol must also be provided in the manuscript.

Flow Cytometry
Plots
Confirm that:
The axis labels state the marker and fluorochrome used (e.g. CD4-FITC).
The axis scales are clearly visible. Include numbers along axes only for bottom left plot of group (a 'group' is an analysis of identical markers).
All plots are contour plots with outliers or pseudocolor plots.
A numerical value for number of cells or percentage (with statistics) is provided.

Methodology
Sample preparation To isolate adipocytes from the iBAT of each mouse (n = 8/group), the interscapular adipose tissue was dissected, minced finely
with scissors, and digested in HBSS buffer containing collagenase (1 mg/ml collagenase type II, 0.5 % BSA) at 37 °C for 30 min.
The cells were then passed through a 100 μm cell strainer and washed twice with PBS, and the RBCs were lysed. The cells were
then washed with cold PBS and incubated with staining cocktail (CD45-APC-Cy7 (#BD557659), 7AAD-PercP-Cy5.5 (#BD559925),
CD11b-PE-Cy7 (#BD561098), and CD206-Alexa Fluor 647 (#565250) (BD; BD Bioscience) in FACS buffer (2 % FBS in PBS) for 20
min on ice in the dark. The cells were then washed and analyzed using a FACS Verse flow cytometer (BD Bioscience). The Flow
cytometry data were analyzed using FlowJo ver.10.4.2 software. All the data were primarily gated for CD45+, 7AAD−.

Instrument FACS Verse flow cytometer (BD Bioscience)

Software FlowJo version 10.4.2 (BD Bioscience) was used to analyze the data.

Cell population abundance CD45+, 7AAD- populations were counted until 10,000 cells to analyze further population.

Gating strategy Cells were gated for CD45 positive cells then for live cells, 7AAD negative populations were gated. CD11b positive, CD206
positive populations were gated for final analysis. The gating strategy is showed in Extended data fig 2i.

Tick this box to confirm that a figure exemplifying the gating strategy is provided in the Supplementary Information.

October 2018