Cellular stress promotes NOD1/2-dependent inflammation via the endogenous metabolite sphingosine-1-phosphate
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Article
Cellular stress promotes NOD1/2-dependent
inflammation via the endogenous metabolite
sphingosine-1-phosphate
Gang Pei1,*,† , Joanna Zyla1,2 , Lichun He3,4, Pedro Moura-Alves1,5 , Heidrun Steinle6,
Philippe Saikali1, Laura Lozza1, Natalie Nieuwenhuizen1, January Weiner1, Hans-Joachim Mollenkopf7,
Kornelia Ellwanger6, Christine Arnold6 , Mojie Duan3,4, Yulia Dagil8, Mikhail Pashenkov8,
Ivo Gomperts Boneca9,10,11, Thomas A Kufer6 , Anca Dorhoi12,13,** & Stefan HE Kaufmann1,14,***
Abstract Subject Categories Immunology; Metabolism
DOI 10.15252/embj.2020106272 | Received 17 July 2020 | Revised 25 March
Cellular stress has been associated with inflammation, yet precise 2021 | Accepted 29 March 2021
underlying mechanisms remain elusive. In this study, various unre- The EMBO Journal (2021) e106272
lated stress inducers were employed to screen for sensors linking
altered cellular homeostasis and inflammation. We identified the
intracellular pattern recognition receptors NOD1/2, which sense Introduction
bacterial peptidoglycans, as general stress sensors detecting
perturbations of cellular homeostasis. NOD1/2 activation upon Nucleotide-binding oligomerization domain-containing protein 1/2
such perturbations required generation of the endogenous (NOD1/2) are intracellular pattern recognition receptors that acti-
metabolite sphingosine-1-phosphate (S1P). Unlike peptidoglycan vate innate immune responses by sensing bacterial peptidoglycans
sensing via the leucine-rich repeats domain, cytosolic S1P directly (Caruso et al, 2014; Philpott et al, 2014). In addition to well-estab-
bound to the nucleotide binding domains of NOD1/2, triggering lished roles in bacterial sensing, both NOD1 and NOD2 modulate
NF-jB activation and inflammatory responses. In sum, we unveiled immune responses to a broad range of peptidoglycan-free microbes,
a hitherto unknown role of NOD1/2 in surveillance of cellular such as respiratory syncytial virus (Sabbah et al, 2009), vesicular
homeostasis through sensing of the cytosolic metabolite S1P. We stomatitis virus (Sabbah et al, 2009), influenza A virus (Lupfer et al,
propose S1P, an endogenous metabolite, as a novel NOD1/2 activa- 2014), human cytomegalovirus (Fan et al, 2016), hepatitis C virus
tor and NOD1/2 as molecular hubs integrating bacterial and meta- (Vegna et al, 2016), Plasmodium berghei (Finney et al, 2009), Plas-
bolic cues. modium falciparum (Corbett et al, 2015), and Trypanosoma cruzi
(Silva et al, 2010). Disruption of actin dynamics by chemicals
Keywords NOD-like receptors; inflammation; sphingolipid metabolism; (Legrand-Poels et al, 2007; Bielig et al, 2014) and manipulation of
cellular homeostasis; NOD1/2 Rho GTPases by pathogens (Keestra et al, 2013) also induce NOD1/
1 Department of Immunology, Max Planck Institute for Infection Biology, Berlin, Germany
2 Department of Data Science and Engineering, Silesian University of Technology, Gliwice, Poland
3 State Key Laboratory of Magnetic Resonance and Atomic Molecular Physics, Key Laboratory of Magnetic Resonance in Biological Systems, National Center for Magnetic
Resonance in Wuhan, Wuhan Institute of Physics and Mathematics, Innovation Academy for Precision Measurement Science and Technology, Chinese Academy of
Sciences, Wuhan, China
4 University of Chinese Academy of Sciences, Beijing, China
5 Nuffield Department of Medicine, Ludwig Institute for Cancer Research, University of Oxford, Oxford, UK
6 Department of Immunology, Institute of Nutritional Medicine, University of Hohenheim, Stuttgart, Germany
7 Microarray Core Facility, Max Planck Institute for Infection Biology, Berlin, Germany
8 Institute of Immunology of the Federal Medical-Biological Agency of Russia, Moscow, Russia
9 Institut Pasteur, Department of Microbiology, Biology and Genetics of the Bacterial Cell Wall, Paris, France
10 CNRS UMR2001, Integrative and Molecular Microbiology, Paris, France
11 INSERM, Équipe AVENIR, Paris, France
12 Institute of Immunology, Friedrich-Loeffler-Institut, Greifswald-Insel Riems, Germany
13 Faculty of Mathematics and Natural Sciences, University of Greifswald, Greifswald, Germany
14 Hagler Institute for Advanced Study at Texas A&M University, College Station, TX, USA
*Corresponding author. Tel: +49 383517 4978; E-mail: gang.pei@fli.de
**Corresponding author. Tel: +49 383517 1624; E-mail: anca.dorhoi@fli.de
***Corresponding author. Tel: +49 3028460 500/502; E-mail: kaufmann@mpiib-berlin.mpg.de
†
Present address: Institute of Immunology, Friedrich-Loeffler-Institut, Greifswald-Insel Riems, Germany
ª 2021 The Authors. Published under the terms of the CC BY 4.0 license The EMBO Journal e106272 | 2021 1 of 19
The EMBO Journal Gang Pei et al
2-dependent inflammatory responses. Additionally, chemical- or We observed that induction of IL6 and IL8 expression and produc-
infection-induced endoplasmic reticulum (ER) stress triggers tion correlated with NOD1/2 expression, without considerable
NOD1/2-mediated inflammation, as well (Keestra-Gounder et al, differences of cell death in association with NOD1 or NOD2 expres-
2016). Whether NOD1/2 detect other types of cellular stress and sion (Fig 1A–D and Appendix Fig S1D–G). Transcription profiling
what enables these receptors to sense such diverse stimuli remain revealed that expression of genes involved in NF-jB activation and
unknown. inflammatory responses was commonly induced by these stimula-
Sphingolipids are a structurally diverse class of lipids character- tions (Fig 1E and F). To exclude effects of doxycycline itself, we
ized by a long-chain amino backbone. They are essential structural generated HeLa doxycycline-inducible GFP cells and stimulated
components of plasma membrane and important mediators involved them with the same chemicals. Compared to HeLa NOD1 or NOD2
in diverse biological pathways (Hannun & Obeid, 2008; Bartke & cells, IL6 production stimulated by these compounds was markedly
Hannun, 2009). Sphingosine, ceramide, ceramide-1-phosphate diminished and doxycycline-dependent IL6 production was not
(C1P), sphingosine-1-phosphate (S1P), and other sphingolipid observed (Fig EV1A). A recent study showed that ER stress triggers
derivatives regulate apoptosis (Birbes et al, 2001; Ganesan et al, endocytosis of the trace peptidoglycan contaminants in the serum,
2010), cell proliferation (Zhang et al, 1990), inflammation (Maceyka thereby inducing pro-inflammatory signaling (Molinaro et al, 2019).
& Spiegel, 2014; Hannun & Obeid, 2018), and autophagy (Young & By using serum-free medium Opti-MEM (Fig EV1B and C) and the
Wang, 2018). Extracellular S1P binds to the cell surface receptors inhibitor of endocytosis-Dynasore (Fig EV1D and E), we revealed
S1PR1-5 and regulates immune cell trafficking (Spiegel & Milstien, that the high IL6 production in response to most of the stimulations
2003, 2011; Rosen & Goetzl, 2005; Maceyka & Spiegel, 2014). S1P above was not blocked by serum depletion or endocytosis inhibi-
also functions as an intracellular signaling molecule and interferes tion, thus excluding the possibility of peptidoglycan contaminants in
with histone acetylation (Hait et al, 2009), calcium mobilization the serum. We conclude that perturbation of cellular homeostasis
from the ER (Ghosh et al, 1990, 1994; Mattie et al, 1994), and TNFa- induces specific NOD1/2-dependent pro-inflammatory signaling.
induced NF-jB activation (Alvarez et al, 2010), which indicates
pleiotropic targets of S1P within the cell. Cellular stress triggers NOD1/2-RIP2-dependent NF-jB activation
Using knockouts and inducible NOD1/2 expression systems, we and MAPK activation
report that diverse stress stimuli induce NOD1/2-RIP2-dependent
inflammatory responses. Gene expression analysis revealed that key To decipher mechanisms underlying NOD1/2-dependent signaling,
enzymes involved in the sphingolipid pathway were induced by NF-jB reporter cells were established on the background of
various stressors. Further, lipidomic profiling demonstrated that HEK293T wild type (WT) and NOD1/2 double knockout (dKO). NF-
cytosolic production of S1P, among other lipid metabolites, was jB luciferase assays confirmed that perturbation of cellular home-
elevated in such conditions. This lipid mediator was essential for ostasis induced NF-jB activation, while NF-jB activation and IL8
NOD1/2 activation. The cytosolic delivery of S1P triggered NOD1/ expression were significantly impaired by NOD1/2 dKO (Fig 2A and
2-RIP2-dependent NF-jB activation and pro-inflammatory responses. B). Consistently, Cxcl2 expression and production of CXCL2, CCL2,
Mechanistic studies elucidated that S1P specifically and directly and IL16 in bone marrow-derived macrophages (BMDMs) upon
bound to the nucleotide binding domains (NBDs) of NOD1/2, inde- perturbation of cellular homeostasis were compromised in both
pendent of the leucine-rich repeats (LRR) domains. Altogether, we Nod1/2 dKO and Rip2 KO cells (Figs 2C and D, and EV2A and B).
established that sphingolipid metabolism governs inflammation trig- Accordingly, NF-jB activation and MAPK activation upon ER stress
gered by cellular stress and identified a central role for S1P in trigger- in BMDMs were also impaired in both Nod1/2 dKO and Rip2 KO
ing NOD1/2-dependent inflammation upon stress induction. BMDMs (Fig EV2C and D). We conclude that disruption of cellular
homeostasis triggers NF-jB activation and MAPK activation in a
NOD1/2-RIP2-dependent manner.
Results
S1P generation is required for inflammation induced by
Various kinds of stress induce NOD1/2-dependent perturbation of cellular homeostasis or by intracellular
inflammatory responses Shigella infection
We postulated that NOD1/2 function as sensors of perturbation of To uncover changes in sphingolipid metabolism upon perturbation
cellular homeostasis. To address this hypothesis, we generated HeLa of cellular homeostasis, primary human dermal fibroblasts (HDFs)
doxycycline-inducible NOD1 or NOD2 cells specific for NOD1 or were employed for two reasons: (i) Sphingolipid metabolism in
NOD2 agonists, respectively (Appendix Fig S1A–C). We stimulated cancer cells is dysregulated (Ryland et al, 2011; Ogretmen, 2018)
these cells with various chemicals that interfere with cellular home- and (ii) NOD1 and NOD2 are strongly expressed in different sources
ostasis through distinct pathways, including perturbation of micro- of human primary fibroblasts (Uehara & Takada, 2007; Hirao et al,
tubule (paclitaxel, vinblastine, nocodazole, colchicine) or actin 2009; Jeon et al, 2012). Numerous stimuli induce generation of
(jasplakinolide, cytochalasin D) dynamics; induction of Golgi (niger- bioactive sphingolipids, including ceramide, C1P, sphingosine, and
icin, monensin, brefeldin A), mitochondrial (actinonin, CCCP, lovas- S1P (Hannun & Obeid, 2008, 2018; Bartke & Hannun, 2009). Indeed,
tatin), or ER stress (tunicamycin, thapsigargin); and protein in HDFs the expression of ACER1/2, SMPD3, SPHK1 and other key
translation inhibition (cycloheximide, anisomycin) and DNA enzymes involved in sphingolipid metabolism (Fig EV3A) were
damage (etoposide). In the rest of this study, we chose 1 or 2 repre- induced by various stressors (Fig 3A). Consistently, lipidomic profil-
sentative chemicals from each class for the stimulation experiments. ing unveiled elevated production of cellular sphingosine (d18:1) and
2 of 19 The EMBO Journal e106272 | 2021 ª 2021 The Authors
Gang Pei et al The EMBO Journal
A B
HeLa NOD1 HeLa NOD2 +Doxycyclin
107 p ≤ 0.001, *** 10 5 p ≤ 0.001, ***
6
10 -Doxycyclin
IL6 mRNA expression (fold) 10 4
IL6 mRNA expression (fold)
105
10 3
104
103 10 2
102
10 1
101
10 0
100
10-1 10-1
Ac ide
co sti l
Ja o daz e
Cy pla chic le
ch o e
N asi e
M ig e n D
iso im A
o n
n s P
ap my n
Cy Bre nencin
ld n
Lo C nin
Et igar in
os n
iE ide
AP
co sti l
Ja o daz e
Cy pla chic le
ch o e
N asi e
M ige n D
iso im A
Ac ide
o n
n s P
ap my n
Et igar in
os n
Cy Bre nencin
ld n
Lo C nin
M e
DP
No bla axe
No bla axe
n
to kin in
al lid
n
to kin in
al lid
id
Tu va CC
Tu va CC
clo fe si
tin yci
Th ica tati
op gi
clo fe si
tin yci
Th ica tati
op gi
s l o
s l o
s c
s c
An ex in
An ex in
-D
o ri
o ri
n t
n t
m
m
Vi acli
Vi acli
h
h
P
P
C
C
Microtubule Actin Golgi Protein Mitochondrial ER DNA Microtubule Actin Golgi Protein Mitochondrial ER DNA
synthesis stress stress damage synthesis stress stress damage
C HeLa NOD1
D HeLa NOD2
2 3 +Doxycyclin
**** ***
* * ** **** -Doxycyclin
**
****
**** **** 2 * **
IL6 (ng/ml)
IL6 (ng/ml)
*** ****
1 **** **** ** ** ****
* **
*
** 1
*
*
0 0
Cy la ic e
ch o e
Cy Br one icin
h el in
is im A
Acom ide
onin
os in
Nonbl tax r
co ast el
Ja Co daz ne
Ni asin e
M ger D
TuLov CC in
Th ca ta P
ap m tin
Et iga cin
o in
iE side
AP
Nonbl tax r
co ast el
Cy la ic e
ch o e
is im A
Acom ide
Ja Co daz ne
Ni asin e
M ger D
Cy Br ne icin
h el in
onin
TuLov CC in
Th ica staCP
ap my tin
Et iga cin
M e
DP
Vi acli Ct
Vi acli Ct
sp lch ol
to kin in
al lid
sp lch ol
to kin in
al lid
id
ni as C
op rg
op rg
clo ef ns
tin yc
clo ef ns
tin yc
An ex din
An ex din
i
i
-D
s y
n a
o
s
P
P
Microtubule Actin Golgi Protein Mitochondrial ER DNA Microtubule Actin Golgi Protein Mitochondrial ER DNA
synthesis stress stress damage synthesis stress stress damage
E F
Mean of Eigengene Vector Mean of Eigengene Vector
TNFα signaling via NFKB (M5890) HeLa NOD1
HeLa NOD2
Control Control
Etoposide 25.00 Vinblastine Etoposide 15.00 Vinblastine
Thapsigargin Nocodazole Thapsigargin 10.00 Nocodazole
15.00
5.00
5.00
Tunicamycin Cytochalasin D Tunicamycin Cytochalasin D
-5.00 0.00
-15.00
Lovastatin Jasplakinolide Lovastatin Jasplakinolide
CCCP Nigericin CCCP Nigericin
Actinonin Brefeldin A Actinonin Brefeldin A
Anisomycin Cycloheximide Anisomycin Cycloheximide
Figure 1. NOD1/2 sense perturbation of cellular homeostasis.
A, B qRT–PCR analysis of IL6 expression upon indicated stimulations in the presence or absence of NOD1 (A) or NOD2 (B). HeLa inducible NOD1 cells (A) or NOD2 cells
(B) were induced in the absence or presence of doxycycline overnight and afterward stimulated with various stimuli for 4 h.
C, D ELISA analysis of IL6 production in supernatants of HeLa NOD1 cells (C) or NOD2 cells (D) upon different stimulations. HeLa NOD1 cells (C) or NOD2 cells (D) were
induced in the absence or presence of doxycycline overnight and stimulated with various stimuli for 20 h, and afterward, supernatants were collected for ELISA.
E, F Sample mean value of eigengene analysis of TNFa signaling via NF-jB pathway (E) and inflammatory response pathway (F) of HeLa NOD1 or NOD2 cells upon
indicated stimulations.
Data information: (A, B) Means SD of three technical replicates from one representative experiment out of four independent experiments. P values were calculated
with ANOVA test of linear mixed model for fix effect of main factors. (C, D) Means SEM of three independent experiments. Each dot represents one independent
experiment. P values were calculated using two-way ANOVA. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, and ****P ≤ 0.0001.
S1P (d16:1, d17:1, d18:1) in HDFs upon various stressors (Fig 3B), to stress-induced inflammation, THP-1 cells, BMDMs, and human
whereas the abundance of other lipid classes remained unaffected CD14+ monocytes were treated with inhibitors targeting distinct
(Fig EV3B). S1P ELISA further confirmed that the abundance of S1P steps of the sphingolipid pathway and IL6 expression and produc-
upon stimulation with various stressors was increased up to 3 µM tion were evaluated upon tunicamycin-induced ER stress
(Fig 3C). To interrogate whether sphingolipid metabolism contributes (Fig EV3C–G). In all types of cells, IL6 expression and production
ª 2021 The Authors The EMBO Journal e106272 | 2021 3 of 19
The EMBO Journal Gang Pei et al
A B
HEK293T HEK293T WT
800
*** 100 **** **** NOD1/2 dKO
****
400 ****
NF-kB luciferase (fold)
****
IL8 mRNA expression
****
10
* *** ***
*
15
* * * ** * *
10 1
5
0 0.1
Ni ide
No las tr
d e
Cy olc ole
Ja hal ine
Cy laki n D
he lide
iE e
on n
ef in
Tu C A
ca P
op in
AP
No las tr
d e
Cy lc le
Ja hal ine
Cy laki n D
Ni ide
he lide
Tu C A
ca P
on n
ef in
op in
iE e
AP
co tin
id
co tin
id
ni CC
ni CC
M rici
M r ici
nb C
nb C
Co azo
Br ens
Et yc
Br ens
Et yc
n
n
-D
-D
C az
c c
c c
xim
os
xim
os
di
di
i
i
clo no
clo no
sp as
sp as
m
m
to hi
to hi
ge
ge
el
el
Vi
Vi
C D
BMDMs **** BMDMs B6J WT
100 ****
8.0 ** Nod1/2 dKO
Cxcl2 mRNA expression (fold)
**** **** **** Rip2 KO
**** **** **** **** **** ****
** **** ****
**** 4.0
CXCL2 (ng/ml)
10 ** ** ** **
**
* **
1.0 ** * *
1
0.5
**
0.1 0.0
e
l tr
Cy cod e
Ja ala e
ki D
Cy Nig ide
he icin
ca P
Et ycin
e
DP
e
e
la n D
e
N de
CP
he in
Et cin
e
DP
id
n
ch ol
id
ni CC
nb C
sp sin
No stin
l
id
id
No asti
to azo
M
clo ric
to az
xim
l
os
clo er
li
Tu CC
no
M
y
Tu C
m
xim
os
i
no
Ja alas
m
Cy ige
op
la
Cy od
op
ca
ki
nb
la
ch
c
Vi
ni
Vi
sp
Figure 2. NOD1/2 signaling is required for NF-jB activation and inflammatory response upon perturbation of cellular homeostasis.
A NF-jB luciferase assay in HEK293T wild-type (WT) and NOD1/2 knockout (KO) NF-jB reporter cells upon various stimulations.
B qRT–PCR analysis of IL8 expression upon indicated stimulations in HEK293T WT and NOD1/2 KO cells.
C qRT–PCR analysis of Cxcl2 expression upon indicated stimulations in WT, Nod1/2 KO, and Rip2 KO bone marrow-derived macrophages (BMDMs).
D Multiplex analysis of CXCL2 production in supernatants of WT, Nod1/2, and Rip2 KO BMDMs upon indicated stimulations for 20 h.
Data information: (A-C) Means SD of three technical replicates from one representative experiment out of three independent experiments. (D) Means SEM of three
independent experiments. Each dot represents one independent experiment. P values were calculated using two-way ANOVA. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, and
****P ≤ 0.0001.
were abolished by the ceramidase inhibitor ceranib-2 and by the infection, suggesting the involvement of S1P in Shigella-induced NF-
sphingosine kinase inhibitor SKI-II (Fig EV3C–G), suggesting that jB activation (Figs 3J and EV3J). Together, we conclude that S1P
S1P production is essential for ER stress-induced IL6 production. generation is required for inflammation induced by perturbation of
Moreover, IL6 expression and production in response to various cellular homeostasis and by intracellular Shigella infection.
stressors in HDFs and HeLa inducible NOD1 or NOD2 cells were
markedly diminished by inhibition of sphingosine kinases (Fig 3D– S1P is produced through the hydrolysis pathway upon ER stress
I). In humans and mice, two genes (Sphk1 and Sphk2) encode the
sphingosine kinase (Adams et al, 2016). Due to relatively low IL6 S1P is mainly generated through the de novo synthesis or by the
production by BMDMs upon other stress stimuli as observed before hydrolysis pathway (Spiegel & Milstien, 2003; Hannun & Obeid,
(Keestra-Gounder et al, 2016), we employed Sphk1 and Sphk2 KO 2018). Serine palmitoyltransferase (SPTLC1) and sphingomyelinase
BMDMs and validated whether Sphk1 and Sphk2 contributed to (SMPD) are the first enzymes involved in the de novo synthesis or
stress-induced CXCL2 expression and production. Consistent with a the hydrolysis pathway, respectively. To interrogate through which
more pronounced decrease of S1P in Sphk2 KO cells (Canlas et al, pathway S1P is produced upon stress induction, the two enzymes
2015; Zhang et al, 2015), Sphk2 KO caused more profound reduction essential for de novo synthesis or the hydrolysis pathway were
in CXCL2 expression and production (Fig 3H and I). MAPK activa- blocked by specific chemical inhibitors (Fig EV3A). IL6 expression
tion induced by ER stress was also reduced by Sphk2 KO (Fig EV3H and production in THP-1 cells upon ER stress were abrogated by
and I). To investigate the role of S1P production in bacterial-induced acidic sphingomyelinase inhibitor imipramine, but not by inhibition
activation of NOD1/2, we infected HeLa cells with the Gram-nega- of serine palmitoyltransferase (L-cycloserine) or neutral sphin-
tive enteropathogenic bacterium Shigella flexneri. SPHK1/2 double gomyelinase (GW4869), uncovering that S1P is mainly produced via
knockdown (dKD) resulted in lower IL8 production upon Shigella the hydrolysis pathway upon ER stress (Appendix Fig S2).
4 of 19 The EMBO Journal e106272 | 2021 ª 2021 The AuthorsIL6 (pg/ml)
J
IL8 (pg/ml)
G
Cy
D
to IL6 mRNA expression (fold)
A
Cy c
0
500
1000
1500
2000
0
50
200
400
600
800
ha
0.1
1
10
100
to la
c
Figure 3.
CERK
SPHK2
SPHK1
SMPD4
SMPD3
SMPD2
SMPD1
ASAH2
ASAH1
ACER3
ACER2
ACER1
SPTLC1
ha Ct sin
la r Vi
nb D
Vi sin
nb D Cy la Ctr
l
Mock
Gang Pei et al
clo stin
e Cytochalasin D
2
No ast
in
****
he
***
ª 2021 The Authors
Cy cod e xim Vinblastine
clo az id
e
****
he ole
xim Et
op Cycloheximide
***
***
os
Et ide id Etoposide
e
3 h M90T
op
4
os CCCP
***
id
****
e CC
HDFs
HeLa NOD2
CC Ni
CP Nigericin
***
ge
Ni CP
g ric Brefeldin A
0.0594
Tu eri in
6 h M90T
c Tunicamycin
6
ni Br
ca in ef
m e
**
yc Tu ldin iE-DAP
in ni
ca A MDP
*
M m
yc
6 h TNF
DP in
***
-SPHKi
+SPHKi
-SPHKi
+SPHKi
B
Cxcl2 mRNA expression (fold)
H
Cy
0.1
1
10
100
1000
to
c Ct IL6 (pg/ml)
E
ha
la r
Ja
0
100
300
600
sp sin Cy
la D
SPHK1/2 dKD
to
n
ki c
ha
Ct
r
Vi oli la
Scrambled shRNA
nb de
la Vi sin
nb D
No stin la
*
e Cy
Cy cod clo stin
Sphinganine-1-phosphate (d18:0)
Sphingadienine-1-phosphate (d18:2)
Sphinganine (d18:0)
Sphingadienine (d18:2)
S1P (d18:1)
S1P (d17:1)
S1P (d16:1)
Sphingosine (d18:1)
Sphingosine (d16:1)
clo az he e
*
he ole xim
Ctr
1
xim
Et ide
Et ide op Cytochalasin D
op os
BMDMs
id
****
os e
id Vinblastine
e
HDFs
*
CC
CC
NiC Nocodazole
*
Ni CP ge P
g
2
Tu eri ric
in Cycloheximide
ni c Br
ef
ca in e Anisomycin
m
** *** *** *** *** *** *** ***
*** **** *** *** *** *** *** ***
yc Tu ldin
in ni Etoposide
***
ca A
***
m
yc CCCP
CXCL2 (pg/ml) in
I
3
Nigericin
0
100
200
500
1000
1500
Cy
-SPHKi
+SPHKi
to
ch Ct Brefeldin A
r
Ja alas
sp i IL6 (pg/ml) Tunicamycin
la n D
ki
F
n
Thapsigargin
4
Vi oli
0
500
1000
1500
2000
nb de Cy
* ****
la to
c
ha
Ct
r
No stin
e la
S1P (μM)
C
Cy cod Vi sin
0
1
2
3
4
****
clo az nb D Cy
he ole la to
***
ch Ct
xim No stin
e a r
Cy cod Vi lasi
Et ide
**
nb n
D
*
op clo az
o
****
BMDMs
os he le No last
**** ****
id
*
e xim Cy co ine
clo da
CC Et ide
*
****
op he zol
N i CP
g os An xim e
****
id
*
iso id
Tu eri
c e
HeLa NOD1
ni m e
ca in CC
m Et yci
****
****
*
yc Ni CP op n
in ge os
Tu r id
e
ni icin
ca C
****
m
**
Ni CCP
yc
in Br g e r
The EMBO Journal e106272 | 2021
*
*
iE Tu efe icin
-D n ld
AP Th ica in A
ap my
B6N WT
-SPHKi
sig cin
+SPHKi
Sphk2 KO
Sphk1 KO
ar
gi
n
*
p=0.054
5 of 19
The EMBO JournalThe EMBO Journal Gang Pei et al
◀ Figure 3. S1P generation is essential for pro-inflammatory responses induced by stressors and by Shigella infection.
A Fluidigm analysis of expression of genes involved in sphingolipid metabolism in human dermal fibroblasts (HDFs) upon indicated stimulations for 4 h. Heatmap
shows means of 4 independent experiments. The color scale represents the relative gene expression compared to control from lower (blue) to higher levels (red).
B Lipidomic profiling of sphingolipid metabolites in HDFs upon indicated stimulations for 2 h. Fold changes of each metabolite were calculated against
corresponding controls. Heatmap shows means of three independent experiments. The color scale represents the relative levels of various lipids from lower (blue)
to higher abundance (red).
C ELISA analysis of cytosolic S1P production in HDFs upon indicated stimulation for 4 h.
D qRT–PCR analysis of IL6 expression in HDFs upon indicated stimulations for 4 h in the absence or presence of the inhibitor of sphingosine kinases-SKI-II (SPHKi,
50 µM).
E–G ELISA analysis of IL6 in supernatants of HDFs (E), HeLa NOD1 cells (F), or HeLa NOD2 cells (G) upon indicated stimulations for 20 h in the absence or presence of
the inhibitor of sphingosine kinases.
H qRT–PCR analysis of Cxcl2 expression in WT, Sphk1 KO, or Sphk2 KO BMDMs upon indicated stimulations for 4 h.
I Multiplex analysis of CXCL2 production in supernatants of WT, Sphk1, and Sphk2 KO BMDMs upon indicated stimulations for 20 h.
J ELISA analysis of IL8 production upon Shigella infection. HeLa scrambled control cells and SPHK1/2 double KD cells were infected with Shigella flexneri M90T for 3
and 6 h or stimulated with TNFa for 6 h.
Data information: (C, E-G, I, J) Means SEM of three independent experiments. Each dot represents one independent experiment. (D, H) Means SD of three technical
replicates from one representative experiment out of three independent experiments. P values were calculated using one-way or two-way ANOVA. *P ≤ 0.05, **P ≤ 0.01,
***P ≤ 0.001, and ****P ≤ 0.0001.
Cytosolic S1P induces NOD1/2-RIP2-mediated NF-jB activation interaction between NOD1/2 and RIP2 by the pharmacological
and inflammation inhibitor-GSK583 abolished IL6 production by cytosolic S1P
(Appendix Fig S3A and B). Consistently, cytosolic S1P triggered
To investigate whether S1P can activate NOD1/2, cells were stimu- MAPK activation and NF-jB activation (Figs 4D and Appendix Fig
lated with extracellular and cytosolic S1P. Extracellular S1P stimula- S3G). Furthermore, cytosolic S1P induced NOD1/2-dependent NF-
tion did not induce marked IL6 or IL8 production (Fig EV4A–D). In jB activation and IL8 expression in HEK293T cells (Fig 4E and F).
contrast, cytosolic delivery of S1P with digitonin induced abundant Finally, Il6 expression and production of IL6, CXCL2, CCL5, and
IL6 or IL8 production in a NOD1/2-dependent manner (Fig 4A and IL16 induced by cytosolic S1P were markedly impaired in Nod1/2
B). Compared to the known NOD1/2 activators iE-DAP or MDP, KO and Rip2 KO BMDMs (Fig 4G–K). We conclude that cytoso-
cytosolic S1P induced higher IL6 production at the same concentra- lic S1P induces NOD1/2-RIP2-mediated NF-jB activation and
tion (Fig EV4E and F). Similar to iE-DAP or MDP, S1P stimulation inflammation.
did not induce NOD1 or NOD2 degradation (Fig EV4G).
The biological activities of extracellular S1P are mediated by S1P directly binds to NOD1/2 via NBDs
S1PRs, a group of G protein-coupled receptors which regulate
immune cell trafficking, cytoskeleton reorganization, and cell prolif- We hypothesized that S1P may directly interact with NOD1/2. To
eration upon S1P binding (Rosen & Goetzl, 2005). To address the this end, immunoprecipitations were performed with different lipid-
involvement of S1PRs in S1P-induced NOD1/2 activation, specific coated beads. NOD1 and NOD2 were specifically immunoprecipi-
inhibitors against S1PR1 (W146), S1PR2 (JTE013), S1PR1 and S1PR3 tated with S1P-coated beads, demonstrating interactions between
(VPC23019), and S1PR4 (CYM50358), as well as the general S1PR S1P and NOD1/2 (Fig 5A and B). Immunoprecipitation with THP-1
inhibitor FTY720 or blocking GPCR signaling by pertussis toxin, cells further confirmed the interaction between S1P and endogenous
were employed. All inhibitors failed to block IL6 production by NOD1 (Fig 5C). Direct stochastic optical reconstruction microscopy
cytosolic S1P (Appendix Fig S3A and B). Cytosolic S1P is irre- (dSTORM) demonstrated the oligomerization of NOD1/2 induced by
versibly degraded by S1P lyase, yielding ethanolamine phosphate S1P (Fig EV5A). Dual-color dSTORM demonstrated that cytosolic
and the long-chain aldehyde trans-2-hexadecenal (Bourquin et al, S1P colocalized with NOD1/2 (Fig 5D). Single particle tracking in
2010). Inhibiting S1P lyase by its specific inhibitor, THI resulted in live cells further demonstrated that S1P colocalized with NOD1/2
higher IL6 production by S1P, but not by iE-DAP or MDP, further and moved together with NOD1/2, confirming its interaction with
confirming the roles of cytosolic S1P in NOD1/2 activation NOD1/2 inside living cells (Fig 5E and F).
(Appendix Fig S3C and D). Notably, other metabolites involved in To delineate potential S1P binding sites, we performed in silico
sphingolipid metabolism, such as C2 ceramide (10 µM), C16 cera- docking on the rabbit NOD2 crystal structure (PDB code: 5IRN) with
mide (10 µM), and sphingosine (10 µM), did not induce IL6 produc- the S1P molecule using AutoDock (Morris et al, 2009). The docking
tion, indicating specificity of S1P as lipid species able to trigger result revealed the accommodation of S1P in the pocket of NOD2,
innate activation (Appendix Fig S3E and F). Thus, cytosolic S1P corresponding to the NACHT/NBD. Residues in the NBD, such as
specifically induces NOD1/2-dependent inflammatory response in a K285, S286, R314, D359, T404, E580, and H583, were predicted to
S1PR independent manner. interact with S1P (Fig 6A). Furthermore, molecular dynamics simu-
To interrogate the downstream signaling of NOD1/2 activation lation of 10 ns was carried out using the docked complex and
by cytosolic S1P, NOD1 and NOD2 were immunoprecipitated and showed almost no movement of the S1P, validating the docking
Western blotting was employed for detecting RIP2. Cytosolic S1P result. To determine whether NOD1/2 directly interacts with S1P,
stimulated the interaction between NOD1/2 and RIP2, indicating NOD1/2 proteins were purified (Fig 6B) and microscale ther-
activation of NOD1/2 signaling via RIP2 (Fig 4C). Blocking the mophoresis (MST) assays were performed with purified proteins
6 of 19 The EMBO Journal e106272 | 2021 ª 2021 The AuthorsGang Pei et al The EMBO Journal
(Fig 6C–E). MST experiments revealed direct interactions of S1P These findings demonstrate the specificity of S1P interaction with
with NOD1 or NOD2, which was not observed with GFP or the NOD1 or NOD2. Intriguingly, the Kd value of S1P with NOD1/2 was
NOD-like receptor NLRP3 (Fig 6C). The purified NOD1 or NOD2 lower than the values of the canonical NOD1/2 agonists iE-DAP or
was specific for their bacterial ligands iE-DAP and MDP, respectively MDP in our system (Fig 6D and E). To substantiate our docking
(Fig EV5B and C). The Kd value of S1P interacting with NOD1 or results regarding binding of S1P to the NACHT/NBDs of NOD1/2,
NOD2 was around 2 or 5 µM, respectively. An interaction of C16 HEK293T NOD1/2 KO cells were transfected with different truncated
ceramide with NOD1 or NOD2 was not detected (Fig EV5B and C). NOD1/2 mutants and the IL8 release was evaluated upon S1P
A B
HeLa NOD1 HeLa NOD2
****
800 800 -Doxycycline
****
+Doxycycline
****
****
600 600
****
****
****
IL6 (pg/ml)
IL6 (pg/ml)
400 **** 400
200 200
0 0
Ctr S1P iE-DAP Ctr S1P MDP
C D E
HeLa NOD1 30’ HeLa NOD1 60’ HeLa NOD2 30’ HeLa NOD2 60’
HeLa NOD1 HeLa NOD2
- S1P iE-DAP - S1P iE-DAP - S1P MDP - S1P MDP
- S1P iE-DAP - S1P MDP
p-JNK
GFP IP HEK293T WT
IB: GFP 20
NOD1/2 dKO
NF-kB luciferase (fold)
p-P38
GFP IP 15
IB: RIP2
p-ERK
10 ***
Lysate
IB: GFP 5
p-P65
Lysate
IB: RIP2 0
β-Actin Ctr S1P
F G H
***
B6J WT
HEK293T BMDMs
30 100 *** Nod1/2 dKO
50 *** Rip2 KO
80
IL8 mRNA expression (fold)
WT
Il6 mRNA expression
40 20 ***
NOD1/2 dKO
IL6 (pg/ml)
60
30
40
20 *** 10 ** *** ** ***
10 20
0 0 0
S1P Ctr S1P MDP Ctr S1P MDP
I J K
3 5 400
**** B6J WT
***
*** ** Nod1/2 dKO
4 300
*** Rip2 KO
CCL5 (ng/ml)
CXCL2 (ng/ml)
2
IL16 (pg/ml)
3
200 *** **
*** **** ****
2
1
** ** *** *** ** * 100
1
0 0 0
Ctr S1P MDP Ctr S1P MDP Ctr S1P MDP
Figure 4.
ª 2021 The Authors The EMBO Journal e106272 | 2021 7 of 19The EMBO Journal Gang Pei et al
◀ Figure 4. Cytosolic S1P-induced NF-jB activation and inflammation is mediated by NOD1/2.
A, BELISA analysis of IL6 in HeLa NOD1 cells (A) or HeLa NOD2 cells (B) upon indicated stimulations for 20 h. For cytosolic delivery of S1P, iE-DAP, or MDP, cells were
treated together with digitonin (5 µg/ml).
C GFP immunoprecipitation (IP) of NOD1/2 upon S1P, iE-DAP, or MDP stimulation. S1P (20 µM), iE-DAP (20 µM), or MDP (20 µM) was delivered into the cytosol of
HeLa NOD1 or HeLa NOD2 cells with digitonin (5 µg/ml) for 1 h. Afterward, cell lysates were collected for IP.
D Western blot analysis of MAPK and NF-jB activation upon S1P stimulation. HeLa NOD1 or NOD2 cells were treated with cytosolic S1P (20 µM), iE-DAP (20 µM), or
MDP (20 µM) for indicated times.
E NF-jB luciferase assay in HEK293T WT and NOD1/2 KO NF-jB reporter cells upon cytosolic S1P stimulation.
F qRT–PCR analysis of IL8 expression upon cytosolic S1P stimulation in HEK293T WT and NOD1/2 KO cells.
G qRT–PCR analysis of Il6 expression in WT, Nod1/2, and Rip2 KO BMDMs upon S1P (10 µM) or MDP (10 µM) stimulations together with digitonin (2.5 µg/ml) for 4 h.
H–K Multiplex analysis of IL6 (H), CXCL2 (I), CCL5 (J), and IL16 (K) production in supernatants of WT, Nod1/2, and Rip2 KO BMDMs upon S1P (10 µM) or MDP (10 µM)
stimulations together with digitonin (2.5 µg/ml) for 20 h.
Data information: (A, B, E-G) Means SD of three independent experiments. (H-K) Means SEM of three independent experiments. Each dot represents one
independent experiment. P values were calculated using two-way ANOVA (A, B, E, G-K) or Student’s t-test (F). *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, and ****P ≤ 0.0001.
Source data are available online for this figure.
stimulation. IL8 production was induced by S1P in cells expressing interacts with the b-phosphate of ADP (Maekawa et al, 2016), we
NOD1/2 WT. IL8 production in cells expressing LRR-deficient interrogated whether S1P and ADP compete with each other for
NOD1/2 (NOD1/2ΔLRR) was also elevated upon S1P stimulation, binding to NOD1/2. To this end, the binding affinities of ADP to
but not upon iE-DAP/MDP stimulation, suggesting that the LRR NOD1/2 were evaluated in the presence or absence of S1P. In line
domains of NOD1/2 are not essential for the interaction with S1P with a previous report (Askari et al, 2012), the Kd values of ADP
(Fig 6F). Consistently, MST assays demonstrated that S1P still inter- alone with NOD1 or NOD2 were around 100 or 150 nM, respec-
acted with NOD1ΔLRR and NOD1ΔCARD, confirming that the NBD tively. However, the Kd values of ADP with NOD1 or NOD2 were
was responsible for the interaction with S1P (Fig EV5D). To address increased to 1.5 µM by S1P (Fig 7A and B). Similarly, the binding
whether S1P competes with bacterial agonists for the binding to affinities of S1P for NOD1/2 were also compromised by the presence
NOD1/2, we performed MST assays using purified NOD1/2 to of ADP (Fig 7C and D). Further, MST assays demonstrated that the
measure their binding affinities with iE-DAP or MDP in the presence Kd value of NOD1 H517A with S1P was increased to around 15 µM,
or absence of S1P. The binding affinities of NOD1/2 for iE-DAP or suggesting impaired binding of S1P to NOD1 H517A. The ADP bind-
MDP were not affected by S1P, further confirming that binding sites ing was also compromised by H517A mutation (Figs 7E and F, and
of S1P and bacterial agonists are sterically independent (Fig 6G and EV5I). Thus, S1P directly binds to NOD1/2 via interaction with
H). This further strengthens our observations that S1P binds to the H517 and subsequently impairs ADP binding. Hence, S1P likely
NBD, but not the LRR domains of NOD1/2. To investigate whether replaces ADP from the NBD to trigger NOD1/2 activation.
co-delivery of S1P with iE-DAP or MDP had synergistic or additive
effects on NOD1/2 activation, HeLa NOD1 or NOD2 cells were co-
stimulated with S1P and various concentration of iE-DAP or MDP. Discussion
Indeed, co-stimulation of S1P with iE-DAP or MDP induced signifi-
cantly higher IL6 production compared to single stimulation alone, In this study, we demonstrate that perturbation of cellular home-
indicating independent activating mechanisms of NOD1/2 upon S1P ostasis induces production of the cellular metabolite S1P, which
and iE-DAP or MDP stimulation (Fig EV5E and F). Crohn’s disease- binds to NOD1/2 and triggers NOD1/2-dependent inflammation.
associated mutation NOD2 1007fs results in a partial deletion in the Hence, NOD1/2 not only sense microbial ligands but also function
LRR region and hence defective detection of MDP (Hugot et al, as general stress sensors via monitoring of cytosolic generation of
2001; Ogura et al, 2001). Based on our data, S1P should still be able the endogenous metabolite S1P. Our study establishes a novel acti-
to induce activation of NOD2 1007fs. Indeed, IL8 production in vation mechanism of mammalian NOD1/2 in which these pattern
HEK293T cells expressing NOD2 1007fs was increased upon S1P recognition receptors recognize cytosolic S1P upon various stress
stimulation, demonstrating the activation of NOD2 1007fs by S1P stimuli. Recent findings have established the concept that the meta-
(Fig EV5G). bolic status of immune cells controls their responses. Most of these
To identify the specific amino acids involved in the interaction, studies focused on glycolysis and TCA cycle, amino acid metabo-
various conserved amino acids in the NBD region of human NOD1 lism, and fatty acid synthesis (O’Neill & Pearce, 2016; O’Neill et al,
were mutated (Fig EV5H). Cells expressing these mutants were 2016). Our findings further extend this concept by establishing a
stimulated by S1P or iE-DAP to identify mutations that specifically novel link between sphingolipid metabolism and NOD1/2 activation
abolish S1P activation. ELISA analysis showed that K328A or H517A upon stress induction. Previous studies reported that sphingosine
could be activated by iE-DAP but not by S1P. Intriguingly, iE-DAP kinases/S1P are required for ER and mitochondrial unfolded protein
induces higher IL8 production in cells expressing K328A or H517A responses and contribute to protection against ER, mitochondrial,
than WT, further confirming the different activation mechanisms by and oxidative stress (Lee et al, 2015; Qi et al, 2015; Kim & Sieburth,
S1P and iE-DAP stimulation (Fig EV5H). Consistent with the in 2018; Kim & Sieburth, 2019). SPHK2 is required for inflammatory
silico docking, this result suggests that H517 in the winged-helix responses in macrophages upon ureteral obstruction and titanium
(WH) domain of human NOD1 is critical for the interaction with particle stimulation (Ghosh et al, 2018; Yang et al, 2018). It inhibits
S1P (Fig EV5H). Assuming that the conserved H517 residue IL6 production at late time points upon LPS stimulation (Weigert
8 of 19 The EMBO Journal e106272 | 2021 ª 2021 The AuthorsGang Pei et al The EMBO Journal
A B C
Lipid beads IP Lipid beads IP
Mock S1P
Cardiolipin
Cardiolipin
Ceramide
Ceramide
LBPA
LBPA
S1P
S1P
LPA
LPA
ctr
ctr
IP NOD1
IP IP
IB: GFP-NOD1 IB: GFP-NOD2
IP IP
IB: RIP2 IB: RIP2 NOD1
Lysates Lysates Lysates
IB: GFP-NOD1 IB: GFP-NOD2
Lysates Lysates
IB: RIP2 IB: RIP2
HeLa NOD1 HeLa NOD2
D
NOD1-AF647 NOD2-AF647
S1P-TAMRA S1P-TAMRA
E
0s 0.2s 0.4s 0.6s 0.8s
NOD1-GFP
S1P-TAMRA
F
0s 0.2s 0.4s 0.6s 0.8s
NOD2-GFP
S1P-TAMRA
Figure 5.
ª 2021 The Authors The EMBO Journal e106272 | 2021 9 of 19The EMBO Journal Gang Pei et al
◀ Figure 5. S1P interacts with NOD1/2.
A, B Immunoprecipitation with various lipid-coated beads using HeLa NOD1-GFP (A) and NOD2-GFP cells (B). Cell lysates of HeLa NOD1 or HeLa NOD2 cells were
incubated with lipid-coated beads at room temperature for 2 h.
C Immunoprecipitation with S1P-coated beads using THP-1 lysates.
D Dual-color dSTORM imaging of HeLa NOD1 or NOD2 upon S1P stimulation. After stimulation with S1P-TAMRA (20 µM) for 2 h, HeLa NOD1 or NOD2 cells were
fixed, permeabilized, and stained with anti-GFP antibody and Alex Fluor 647-conjugated secondary antibody. Images were reconstructed from 10,000 raw frames.
Scale bar: 200 nm.
E, F Single particle tracking of NOD1-GFP (E) or NOD2-GFP (F) with S1P-TAMRA. After doxycycline induction overnight, HeLa NOD1 or NOD2 cells were stimulated with
S1P-TAMRA (20 µM) for 1 h together with digitonin. Scale bar: 5 µm.
Source data are available online for this figure.
et al, 2019). However, SPHK2 does not affect inflammatory overall levels that we measured. Thus, it is conceivable that endoge-
responses upon short-time stimulation with LPS (Xiong et al, 2013). nous S1P is sufficient to induce NOD1/2 activation. How S1P gener-
Our studies complement the current knowledge about immune roles ation is regulated at the spatial level and whether these subcellular
of SPHKs by showing that S1P, a product of SPHKs, modulates cyto- sites of S1P serve as signaling hubs for cell stress responses and
kine responses during stress-induced inflammation. We propose inflammation remain to be established. Moreover, the cellular
that S1P may function as a central player in coordinating cellular origin, e.g., hematopoietic versus non-hematopoietic, and the cellu-
stress responses and inflammation upon stress induction. Interac- lar activation, e.g., by cytokines and microenvironment, alter the
tion partners and downstream effects of S1P may be context-depen- expression levels of NOD1/2 and their downstream signaling mole-
dent and deserve further investigations. cules. Accordingly, S1P-triggered inflammatory responses may differ
As a pleiotropic second messenger, S1P acts both extracellularly in distinct tissues and at particular stages of a disease.
and intracellularly to regulate diverse processes, including immune NOD1/2 trigger immune responses to diverse peptidoglycan-free
cell trafficking, inflammation, and apoptosis (Spiegel & Milstien, microbes and underlying mechanisms are ill defined. We discovered
2011; Maceyka et al, 2012). Our study identifies NOD1/2 as novel that various stress stimuli induced production of the cellular
receptors for cytosolic S1P, in addition to HDAC1/2 and TRAF2, metabolite S1P and subsequent NOD1/2-dependent inflammation.
which have been previously identified as cytosolic interactors (Hait Since bacteria and viruses induce ER stress (Celli & Tsolis, 2015),
et al, 2009; Alvarez et al, 2010), and establishes S1P as the missing DNA damage (Zgur-Bertok, 2013), and protein translation block
link between cellular stress and NOD1/2-mediated inflammation. (Fontana et al, 2011; Dunbar et al, 2012), it is tempting to assume
Considering that S1P is structurally and metabolically conserved that NOD1/2 sensing of host cytosolic S1P represents an ancient
through evolution (Hannun & Obeid, 2008), we propose that cytoso- alert mechanism for infectious insult. Many intracellular pathogens,
lic S1P generated upon perturbation of cellular homeostasis repre- such as Legionella pneumophila and Burkholderia pseudomallei,
sents an endogenous stress-associated molecular pattern (SAMP). secrete S1P lyases as virulence factors facilitating intracellular
Contrary to canonical damage- or danger-associated molecular survival (Rolando et al, 2016; McLean et al, 2017), supporting the
patterns (DAMPs) released after cell lysis (Matzinger, 1994), S1P is importance of S1P sensing by NOD1/2 during intracellular infection.
generated in the cytosol upon cellular stress induction without S1P binds to the NBDs and activates RIP2-mediated signaling, which
considerable cell death. The enzymes generating S1P show distinct differs from peptidoglycan sensing via LRR domains of NOD1/2 and
subcellular compartmentalization during stress responses. SPHK1 is indicates a different activation mechanism of NOD1/2 by S1P. We
activated and targeted to mitochondria upon mitochondrial stress discovered that S1P interacts with the amino acid (H517 of NOD1)
(Kim & Sieburth, 2018), and SPHK2 is translocated into the ER upon that stabilizes the b-phosphate of ADP, which is critical for main-
serum starvation (Maceyka et al, 2005). In this study, we demon- taining the inactive conformation of NOD1/2 (Zurek et al, 2012;
strate that the total abundance of S1P increases up to 3 µM upon Maekawa et al, 2016). Therefore, S1P could induce NOD1/2 activa-
delivery of various stressors. Considering that SPHKs are enriched tion by promoting their active conformation through replacing of
in intracellular compartments, we postulate that the intracellular ADP in the NBD region. This bears similarities with ATP and MDP
concentration of S1P at various subcellular sites may exceed the binding to NOD2 (Mo et al, 2012). Our finding that S1P impairs the
Figure 6. S1P directly binds to NOD1/2 via the NBD and triggers their activation.
A
B
Molecular docking of S1P with rabbit NOD2.
SDS–PAGE analysis of purified NOD1-GFP and NOD2-GFP.
▸
C Microscale thermophoresis (MST) analysis of direct binding of S1P to NOD1, NOD2, NLRP3, or GFP.
D MST analysis of direct binding of NOD1 to S1P and iE-DAP.
E MST analysis of direct binding of NOD2 with S1P and MDP.
F ELISA analysis of IL8 in supernatants of HEK293T NOD1/2 KO cells transfected with indicated NOD1 or NOD2 mutants. After 24-h transfection, cells were stimulated
with S1P (20 µM), iE-DAP (20 µM), or MDP (20 µM) together with digitonin for 20 h.
G, H MST analysis of direct binding of NOD1 (G) or NOD2 (H) with iE-DAP or MDP in the presence of S1P (2 µM).
Data information: (C) Means SD from three independent experiments. (D, E, G, H) One representative experiment out of three independent experiments. (F)
Means SD of three technical replicates from one representative experiment out of three independent experiments. P values were calculated using two-way ANOVA.
*P ≤ 0.05 and ****P ≤ 0.0001.
10 of 19 The EMBO Journal e106272 | 2021 ª 2021 The AuthorsGang Pei et al The EMBO Journal
A B C
Marker (kDa) NOD1-GFP NOD2-GFP 1.2
S1P : NOD1
180
F581 130 S1P : NOD2
0.8
E580 110
Fraction Bound
L582 95 S1P : GFP
R314
W470 H583 70
55 0.4 S1P : NLRP3
C313
V467 40
K285 D359 35
P466
S286 T404 0.0
F427 25
15
Y435 L228 -0.4
Y232
10 -9 10 -8 10 -7 10 -6 10 -5 10 -4 10 -3
Ligand Concentration
D E
Kd (NOD1/S1P)=2.09±0.77 μM
1.2 Kd (NOD1/iE-DAP)=27.12±7.87 μM 1.2 Kd (NOD2/S1P)=5.36±1.91 μM
NOD1 : S1P Kd (NOD2/MDP)=31.11±18.07 μM NOD2 : S1P
0.8 NOD1 : iE-DAP 0.8 NOD2 : MDP
Fraction Bound
Fraction Bound
0.4 0.4
0.0 0.0
-0.4 -0.4
10 -9
10 -8
10 -7
10 -6
10 -5
10 -4
10 -3 10 -9 10 -8 10 -7 10 -6 10 -5 10 -4 10 -3 10 -2
Ligand Concentration Ligand Concentration
F G
500 Ctr
1.5
**** S1P NOD1 : iE-DAP
400
NOD1 : iE-DAP/S1P
iE-DAP/MDP 1.0
Fraction Bound
300
IL8 (pg/ml)
* 0.5
200 * ****
****
0.0
*
100
*
-0.5
0 10 -10 10 -9 10 -8 10 -7 10 -6 10 -5 10 -4 10 -3
D1 pty
D D
NO D1 D
NO D1 RR
NO 1 K T
CA R
NO NOD RDs
NO CA D
T
'C R
NO D1 D
D2 RDs
D2 R
Ligand Concentration
D W
W
D2 208
D1 'LR
NO LR
NO CAR
B
' NB
NO AR
L
NO 1 N
NO Em
'
D2 2
H
1.5
NOD2 : MDP
1.0 NOD2 : MDP/S1P
Fraction Bound
0.5
0.0
-0.5
-1.0
10 -10 10 -9 10 -8 10 -7 10 -6 10 -5 10 -4 10 -3
Ligand Concentration
Figure 6.
binding of ADP to NOD1/2 supports this possibility. Further, ADP In context of infection, the assumption of an evolutionary conserved
replacement stabilizes the active conformation resulting in the acti- role of NOD1/2 in detecting pathogens or pathogen-induced cellular
vation of NOD1/2. The detailed structural requirements for S1P and stress via NBDs is supported by the finding that the 290 NOD-like
iE-DAP/MDP driven NOD1/2 signaling deserve further elucidation. receptors (NLRs) identified in Hydra magnipapillata lack LRRs, but
ª 2021 The Authors The EMBO Journal e106272 | 2021 11 of 19The EMBO Journal Gang Pei et al
A NOD1 WT B NOD2 WT
Kd(NOD1: ADP)=109±54.5 nM Kd(NOD2: ADP)=154±77 nM
Kd(NOD1: ADP/S1P)=1300±86.6 nM NOD1 : ADP Kd(NOD2: ADP/S1P)=1500±81.4 nM
1.5 1.5 NOD2 : ADP
NOD1 : ADP/S1P NOD2 : ADP/S1P
1.0
1.0
Fraction Bound
Fraction Bound
0.5
0.5
0.0
0.0
-0.5
-0.5 -1.0
10 -10 10 -9 10 -8 10 -7 10 -6 10 -5 10 -4 10 -3 10 -10 10 -9 10 -8 10 -7 10 -6 10 -5 10 -4 10 -3
Ligand Concentration Ligand Concentration
C NOD1 WT D NOD2 WT
Kd(NOD1: S1P)=1.74±0.43 μM Kd(NOD2: S1P)=2.4±0.38 μM
NOD1 : S1P NOD2 : S1P
1.5 Kd(NOD1: S1P/ADP)=16.1±9.4 μM 1.5 Kd(NOD2: S1P/ADP)=16.5±3.85 μM
NOD1 : S1P/ADP NOD2 : S1P/ADP
1.0 1.0
Fraction Bound
Fraction Bound
0.5 0.5
0.0 0.0
-0.5 -0.5
10 -10 10 -9 10 -8 10 -7 10 -6 10 -5 10 -4 10 -3 10 -10 10 -9 10 -8 10 -7 10 -6 10 -5 10 -4 10 -3
Ligand Concentration Ligand Concentration
E NOD1 WT/H517A F NOD1 WT/H517A
Kd(NOD1 WT: S1P)=2.2±0.51 μM Kd(NOD1 WT: ADP)=187.1±74.8 nM
NOD1 WT : S1P NOD1 WT : ADP
1.5 Kd(NOD1 H517A: S1P)=13.7±6.1 μM 1.5 Kd(NOD1 H517A: ADP)=998.1±189.2 nM
NOD1 H517A : S1P NOD1 H517A : ADP
1.0 1.0
Fraction Bound
Fraction Bound
0.5 0.5
0.0 0.0
-0.5 -0.5
-1.0 -1.0
10 -9 10 -8 10 -7 10 -6 10 -5 10 -4 10 -10 10 -9 10 -8 10 -7 10 -6 10 -5 10 -4
Ligand Concentration Ligand Concentration
Figure 7. S1P binds to NOD1 via H517 and impairs ADP binding.
A, B MST analysis of direct binding of NOD1 (A) or NOD2 (B) with ADP in the presence of S1P (2 µM).
C, D MST analysis of direct binding of NOD1 (C) or NOD2 (D) with S1P in the presence of ADP (2 µM).
E, F MST analysis of direct binding of S1P (E) or ADP (F) with NOD1 WT or NOD1 H517A.
Data information: (A-F) One representative experiment out of three independent experiments.
still interact with the death domain (DD) or caspase recruitment abundance of host sphingolipids represents the most significant
domain (CARD)-containing proteins, resembling inflammasomes in metabolite signature in inflammatory bowel disease (IBD) patients
vertebrates (Lange et al, 2011). This together with our evidence on (Brown et al, 2019). Moreover, numerous NOD2 mutations in the
NOD1/2 activation by S1P via the NBDs substantiates the potential NBD are associated with IBD, Blau syndrome, and early-onset
relevance of metabolites as primordial signals for infection. sarcoidosis (McGovern et al, 2001; Caso et al, 2015). It has been
Metabolites have recently gained attention for their potential shown that the association of plasma membrane of NOD2 is essen-
roles in the host–microbiome cross talk. Most microbes belonging to tial for MDP-induced NF-kB activation (Barnich et al, 2005; Lecine
the Bacteroidetes phylum, a dominant proportion of human gut et al, 2007). Intriguingly, our data demonstrate that Crohn’s disease-
microbiome, have been found to produce sphingolipids (Heaver associated mutation NOD2 1007fs which is unresponsive to MDP is
et al, 2018). Recently, it has been established that the increased still activated by S1P. Consistently, NOD1 H517A that mainly
12 of 19 The EMBO Journal e106272 | 2021 ª 2021 The AuthorsGang Pei et al The EMBO Journal
resides in the cytosol (Zurek et al, 2012) is activated by cytosolic dissolved in water with sonication (stock concentration 2 mM). S1P-
delivery of iE-DAP with digitonin. Thus, it is plausible that cytosolic TAMRA was dissolved in methanol (stock concentration 200 µM).
NOD1/2 can detect their ligands in the cytosol. Our findings encour- All other chemicals were dissolved in DMSO if not specified.
age further investigations into the S1P-NOD1/2 axis in sphingolipid- Antibodies used in this study include anti-ACTB (Sigma-Aldrich,
mediated host–microbiome interactions under physiologic and A2228), anti-GFP (Proteintech, 66002-1-Ig), anti-GFP (ChromoTek,
pathologic conditions. PABG1), anti-RIP2 (Cell Signaling Technology, #4142), anti-SPHK2
Non-resolving chronic inflammation resulting from the failure to (ECM Biosciences, SP4621), anti-phospho-p38 (Cell Signaling Tech-
reestablish homeostasis underlies the pathogenesis of many stress- nology, #4511), anti-phospho-JNK (Cell Signaling Technology,
related diseases, such as obesity, diabetes, and cancer (Chovatiya & #4668), anti-phospho-ERK1/2 (Cell Signaling Technology, #4370),
Medzhitov, 2014; Kotas & Medzhitov, 2015). Hence, our findings anti-phospho-p65 (Cell Signaling Technology, #3033), and anti-
shed light on novel molecular mechanisms of the inflammation NOD1 (Cell Signaling Technology, #3545).
associated with chronic diseases caused by disturbed homeostasis
and establish the SPHKs-S1P-NOD1/2 axis as a potential target for Cell culture
novel therapeutic strategies for chronic inflammatory diseases.
The human monocytic cell line THP-1 was obtained from the Ameri-
can Type Culture Collection (ATCC, TIB-202) and maintained in
Materials and Methods RPMI 1640 (Gibco, 31870) with 10% (v:v) heat-inactivated fetal
bovine serum (FBS, Sigma-Aldrich, F0804), 1 mM sodium pyruvate
Mice (Gibco, 11360070), 2 mM L-glutamine (Gibco, 25030081), 10 mM
HEPES buffer (Gibco, 15630080), pH 7.2-7.5, and 50 µM 2-mercap-
B6N.129S6-Sphk1tm1Rlp/J (Sphk1/) and B6N.129S6-Sphk2tm1Rlp/J toethanol (Gibco, 31350010). To differentiate THP-1 into macro-
(Sphk2/) mice were originally obtained from the Jackson Labora- phage-like cells, THP-1 cells were stimulated with 50 ng/ml phorbol
tory (Bar Harbor, ME, USA). These mice were maintained under 12-myristate 13-acetate (Sigma-Aldrich, P8139) for 24 h and then
specific pathogen-free (SPF) conditions at the Max Planck Institute incubated with fresh medium for another 48 h. HEK293T WT and
for Infection Biology in Berlin, Germany. 6- to 15-week-old female HEK293T NOD1/2 dKO cells were generated as described (Dagil
mice were employed for the experiments. et al, 2016). HeLa inducible NOD1-GFP and NOD2-GFP cells were
generated by co-transfection of pcDNA5/FRT/TO-EGFP-NOD1 or
Plasmids, antibodies, and reagents pcDNA5/FRT/TO-EGFP-NOD2 (Addgene #131206 and #131207)
together with pOG44 in a 9:1 ratio into HeLa FlpIN T-REx cells
NLRP3-GFP (Addgene, 73955) was obtained from Addgene. NOD1- (kindly provided by the Hentze Lab, EMBL, Heidelberg) using Lipo-
or NOD2-related plasmids were constructed as described (Kufer fectamine 2000 (Thermo Fisher Scientific) and selected with 10 µg/
et al, 2008). Chemicals used in this study include NVP231 (10 µM, ml blasticidin and 500 µg/ml hygromycin B. HeLa inducible NOD1-
Tocris, 3960), N-oleoylethanolamine (NOE, 30 µg/ml, Sigma, GFP and NOD2-GFP cells, human neonatal dermal fibroblasts, and
O0383), SKI-II (25 µM, Sigma, S5696), PF543 (20 µM, Sigma, HEK293T cells (DSMZ, ACC305) were maintained in complete
PZ0234), ceranib-2 (10 µM, Sigma, SML0607), fumonisin B1 Dulbecco’s modified Eagle’s medium (DMEM) 4.5 g/l glucose
(10 µM, Sigma, F1147), deoxynojirimycin (20 µM, Sigma, D9305), (Gibco, 10938) with 10% (v:v) heat-inactivated FBS, 1 mM sodium
GW4869 (20 µM, Sigma, D1692), imipramine (50 µM, Sigma, pyruvate, and 2 mM L-glutamine. To induce NOD1 or NOD2 expres-
I0899), L-Cycloserine (100 µM, Sigma, C1159), KIRA6 (5 µM, Merck sion, HeLa NOD1 or NOD2 cells were treated with 0.5 µg/ml or
Millipore, 532281), 4µ8C (1 µM, Merck Millipore, 412512), PERK 0.25 µg/ml doxycycline for at least 16 h, respectively. All cells have
inhibitor I (10 µM, Merck Millipore, 516535), thapsigargin (5 µM, been checked for mycoplasma contamination regularly. Bone
Sigma, SML1845), tunicamycin (5 µg/ml, Sigma, T7765), nigericin marrow-derived macrophages (BMDMs) were obtained from tibial
(5 µM, Sigma, N7143), paclitaxel (5 µM, Sigma, N7191), cytocha- and femural bones and generated with DMEM containing 20% L929
lasin D (5 µM, Sigma, C8273), jasplakinolide (0.2 µM, Sigma, cell supernatant, 10% FBS, 5% heat-inactivated horse serum, 1 mM
J4580), nocodazole (10 µM, Sigma, M1404), colchicine (10 µM, sodium pyruvate, and 2 mM L-glutamine. For BMDM stimulation
Sigma, C9754), vinblastine (10 µM, Sigma, V1377), monensin experiments, BMDMs were treated with IFN-c (20 ng/ml) overnight
(Sigma, 30552), brefeldin A (0.5 µg/ml, Sigma, B6542), cyclohex- before stimulation.
imide (50 µM, Sigma, C1988), anisomycin (1 µM, Sigma, A9789),
actinonin (50 µM, Sigma, A6671), lovastatin (30 µM, Sigma, Cell viability analysis
1370600), carbonyl cyanide m-chlorophenyl hydrazine (CCCP,
5 µM, Sigma, C2759), etoposide (10 µM, Sigma, E1383), iE-DAP Cell viability assays were performed with CellTiter 96 AQueous one
(Invivogen, tlrl-dap), MDP (Invivogen, tlrl-mdp), C2 ceramide (Enzo solution reagent (Promega) according to the manufacturer’s instruc-
life sciences, BML-SL100-0005), C16 ceramide (Enzo life sciences, tions. Cell viability upon various stimulations was normalized to
BML-SL115-0005), sphingosine (Enzo life sciences, BML-EI155- corresponding controls.
0025), sphingosine-1-phosphate (S1P, Enzo life sciences, BML-
SL140-0001), and S1P-TAMRA (Echelon Biosciences, S-200T). S1P- Shigella infection
coated beads (Echelon Biosciences, S-6110-2) and sphingolipid-
coated bead pack (Echelon Biosciences, P-B00SS), C2 ceramide, C16 Shigella flexneri M90T afaE (Clerc & Sansonetti, 1987) was kindly
ceramide, and sphingosine, were dissolved in ethanol. S1P was provided by Philippe Sansonetti (Institute Pasteur) and were grown
ª 2021 The Authors The EMBO Journal e106272 | 2021 13 of 19You can also read
