broad defects in the energy metabolism of leukocytes...
TRANSCRIPT
©20
16N
atu
re A
mer
ica,
Inc.
All
rig
hts
res
erve
d.
nature immunology aDVaNCE ONLINE PUBLICaTION �
A rt i c l e s
The immune response during the acute phase of sepsis is characterized by inflammation and the activation of immunological effector mecha-nisms (for example, phagocytosis and intracellular killing) aimed at eliminating the pathogen. During the late phase, the immune system shifts toward an anti-inflammatory state to diminish inflammation and initiate tissue repair. This chain of events can be distorted with dire consequences for the outcome: an exaggerated systemic release of cytokines in the acute phase (‘cytokine storm’) can cause hypo-tension, cardiovascular dysfunction, tissue damage and multi-organ failure, while during the later phases of the disease, the activation of modulatory pathways can prematurely inhibit host defense (innate immunotolerance or innate immunoparalysis)1. However, studies have demonstrated that these functional states can occur simultane-ously. In agreement with that, a study has shown that phagocytes from patients with sepsis who had survived the infection display both decreased inflammation and increased expression of genes encoding products involved in phagocytosis, microbial killing and tissue repair, found to be mediated by a pathway dependent on the transcription factor HIF-1α2. Nevertheless, exaggerated inhibitory effects on leu-kocytes during the late phase of sepsis can lead to immunoparalysis associated with increased susceptibility to secondary infections and a poor outcome3. Understanding the mechanisms that regulate
inflammatory responses in sepsis is crucial for the identification new therapeutic strategies for reversing dysregulation of the immune system in severely ill septic patients.
Studies have identified a crucial role for the cellular metabolism of glucose in the functional fate of cells of the immune system. In this context, naive or tolerant cells rely mainly on oxidative phosphoryla-tion and β-oxidation as energy sources, while after stimulation, leu-kocytes shift their metabolism toward aerobic glycolysis (the Warburg effect)3,4. Defects in tissue metabolism have been reported in sepsis and seem correlated with outcome. Elevated ATP concentrations in muscle biopsies of patients with sepsis are associated with better survival5, whereas mitochondrial dysfunction and energy shortage have been hypothesized to be an underlying cause of organ dysfunc-tion6,7. In addition, two small studies have reported mitochondrial dysfunction8 and redox imbalance9 in leukocytes of patients with sepsis. However, a comprehensive investigation of cellular metabolism of leukocytes in sepsis and its effect on immunological function and outcome has not been performed.
In this study we aimed to assess energy metabolism of cells of the immune system in two clinically relevant infection models: bacte-rial sepsis caused by the Gram-negative bacterium Escherichia coli, and fungal sepsis caused by the opportunistic fungal pathogen
1Department of Internal Medicine and Radboud Center for Infectious Diseases, Radboud University Medical Center, Nijmegen, the Netherlands. 2Center of Experimental & Molecular Medicine, Division of Infectious Diseases, Amsterdam Medical Center, University of Amsterdam, Amsterdam, the Netherlands. 34th Department of Internal Medicine, University of Athens, Medical School, Athens, Greece. 4Department of Intensive Care and Radboud Center for Infectious Diseases, Radboud University Medical Center, Nijmegen, the Netherlands. 5Department of Biochemistry, Radboud Institute for Molecular Life Sciences, Nijmegen, the Netherlands. 6Institute of Metabolism and Systems Research, University of Birmingham, Birmingham, UK. 7Department of Intensive Care Medicine, University Medical Center Utrecht, Utrecht, the Netherlands. 8Department of Medical Microbiology, University Medical Center Utrecht, Utrecht, the Netherlands. 9Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, Utrecht, the Netherlands. 10Department of Intensive Care Medicine Academic Medical Center, University of Amsterdam, Amsterdam, the Netherlands. 11These authors contributed equally to this work. 12These authors jointly directed this work. Correspondence should be addressed to M.G.N. ([email protected]) or T.v.d.P. ([email protected]).
Received 26 June 2015; accepted 14 January 2016; published online 7 March 2016; doi:10.1038/ni.3398
Broad defects in the energy metabolism of leukocytes underlie immunoparalysis in sepsisShih-Chin Cheng1,11, Brendon P Scicluna2,11, Rob J W Arts1,11, Mark S Gresnigt1, Ekta Lachmandas1, Evangelos J Giamarellos-Bourboulis3, Matthijs Kox4, Ganesh R Manjeri5,6, Jori A L Wagenaars5, Olaf L Cremer7, Jenneke Leentjens1,4, Anne J van der Meer2, Frank L van de Veerdonk1, Marc J Bonten8,9, Marcus J Schultz10, Peter H G M Willems5, Peter Pickkers4, Leo A B Joosten1, Tom van der Poll2,12 & Mihai G Netea1,12
The acute phase of sepsis is characterized by a strong inflammatory reaction. At later stages in some patients, immunoparalysis may be encountered, which is associated with a poor outcome. By transcriptional and metabolic profiling of human patients with sepsis, we found that a shift from oxidative phosphorylation to aerobic glycolysis was an important component of initial activation of host defense. Blocking metabolic pathways with metformin diminished cytokine production and increased mortality in systemic fungal infection in mice. In contrast, in leukocytes rendered tolerant by exposure to lipopolysaccharide or after isolation from patients with sepsis and immunoparalysis, a generalized metabolic defect at the level of both glycolysis and oxidative metabolism was apparent, which was restored after recovery of the patients. Finally, the immunometabolic defects in humans were partially restored by therapy with recombinant interferon-g, which suggested that metabolic processes might represent a therapeutic target in sepsis.
©20
16N
atu
re A
mer
ica,
Inc.
All
rig
hts
res
erve
d.
� aDVaNCE ONLINE PUBLICaTION nature immunology
A rt i c l e s
Figure 1 Immunometabolism is upregulated during the acute inflammatory phase. (a) Production of TNF, IL-6, lactate and NAD+ by human PBMCs left unstimulated (treated with RPMI medium)) or simulated for 24 h with C. albicans or LPS (n = 8 individual donors). (b) Immunoblot analysis of phosphorylated (p-) S6K and 4EBP1, as well as total actin (loading control throughout), in human PBMCs obtained from two donors (A and B) and given no pretreatment or preincubated for 1 h with metformin (below lanes), then left unstimulated (RPMI) or stimulated for 2 h with C. albicans or LPS (above lanes). *P < 0.05 and **P < 0.01 (Wilcoxon signed-rank test). Data are representative of three experiments (a; mean + s.e.m.) or two experiments (b).
Candida albicans. On the one hand, we investigated the role of host defense in the metabolic shift toward aerobic glycolysis during acute infection. On the other hand, we investigated the relationship between the dysregulation of cellular metabolism and innate immunoparalysis (or innate immunotolerance) in patients with severe sepsis.
RESULTSThe shift toward aerobic glycolysis in acute inflammationThe first aim of our study was to investigate cellular metabolism during the stimulation of leukocytes by pathogens. Microarray assess-ment of human peripheral blood mononuclear cells (PBMCs) stimu-lated for 4 h or 24 h with either C. albicans10 (Supplementary Fig. 1a,b) or E. coli–derived lipopolysaccharide (LPS) (Supplementary Fig. 1c) revealed substantial upregulation in the expression of mRNA encoding products involved in glycolysis, whereas the expression of mRNA encoding rate-limiting enzymes in the tricarboxylic acid (Krebs) cycle was downregulated. In addition, genes encoding key products involved in the metabolic checkpoint kinase mTOR–HIF-1α pathway (important for the regulation of glycolysis11) were also upregulated (Supplementary Fig. 1a). Those gene-transcription data were con-firmed by the finding of a substantial increase in the production of lactate and concentration of NAD+ when cells were stimulated with C. albicans or LPS (Fig. 1a), which demonstrated a shift toward aerobic glycolysis and thus illustrated induction of the Warburg effect12. Such a shift toward glycolysis has been shown to be driven by activation of mTOR as an intracellular sensor of glucose13. In line with that, 2 h of stimulation with C. albicans or LPS, induced phosphorylation of S6K and 4EBP1, which are downstream targets of the mTOR complex mTORC1 (ref. 13) (Fig. 1b); this showed that an mTOR-regulated increase in anaerobic glycolysis was a key metabolic pathway in acute inflammation.
The mTOR pathway is essential for survival in sepsisInhibition of the mTOR-pathway in human PBMCs by metformin (an activator of the kinase AMPK14) inhibited the phosphorylation of S6K and 4EBP1 (Fig. 1b); in addition, inhibition of mTOR with metformin, rapamycin or torin-1 (ref. 15) decreased the cytokine production induced by C. albicans after stimulation for 24 h (TNF and IL-1β but not IL-6) or 5 d (IFN-γ, IL-17 and IL-22) (Fig. 2a,b). The in vivo inhibition of metabolic pathways by metformin resulted in significantly diminished survival of mice with sepsis caused by C. albicans, lower ex vivo cytokine production by splenocytes and increased fungal outgrowth in kidneys but not liver, relative to that of mice not treated with metformin (Fig. 2c–e). Together these results attested to the importance of the mTOR pathway in robust cytokine production and an effective host defense.
Regulation of cellular metabolism pathways in sepsisAfter showing a role for mTOR-induced glycolysis in host defense during the acute phase of infection, we aimed to assess glycolysis and oxidative phosphorylation in leukocytes of patients with sepsis. We assessed the genome-wide gene expression of whole-blood leu-kocytes isolated from healthy volunteers undergoing experimental endotoxemia, a human model of the inflammatory response observed during sepsis, induced by intravenous administration of a low dose of E. coli LPS (4 ng per kg body weight)16. As is commonly found in this model16–18, we noted robust alterations in the blood transcrip-tome after the administration of LPS, such as elevated expression of genes encoding typical factors involved in pro-inflammatory signal-ing (signaling via IL-1 and IL-6) and anti-inflammatory signaling (signaling via IL-10), as well as signaling via Toll-like receptors, at 4 h after the administration of LPS (Supplementary Fig. 2a,b). Genes encoding products involved in not only the glycolysis I and mTOR pathways but also oxidative phosphorylation and fatty acid oxidation were under-expressed in blood leukocytes 4 h after the administra-tion of LPS (Fig. 3a), a time point at which the cytokine-production capacity of leukocytes was substantially reduced (Supplementary Fig. 2c,d). The metabolic regulator–encoding genes HIF1A, NAMPT and EPAS1 (HIF2A) were overexpressed after the administration of LPS, whereas no differences were shown for expression of the transcrip-tion factor–encoding genes PPARA, PPARG and HIF-1α inhibitor– encoding gene HIF1AN (Supplementary Fig. 3a–c). Moreover, markers characteristically expressed by M1 macrophages were largely over-expressed after the administration of LPS, in contrast to the predomi-nant under-expression of markers characteristically expressed by M2 macrophages (Fig. 3a).
To determine whether the changes noted above could be identified in actual patients with sepsis as well, we performed a genome-wide microarray analysis of blood from a cohort of 33 patients with bacte-rial sepsis (blood cultures positive for E. coli) and 13 patients with fungal sepsis (blood cultures positive for Candida species) (Fig. 3b–e and Supplementary Table 1). Both bacterial sepsis and fungal sepsis induced substantial changes (a change in expression (upregulation or downregulation) of twofold or more) in the transcriptome profiles that encompassed more than 2,000 genes for each infection (Fig. 3b). Most of the genes modified were altered in both bacterial sepsis and fungal sepsis, although gene sets specific for either infection were also identified (Fig. 3c). Among the common pathways influenced in both types of sepsis, changes in the expression of genes encod-ing products involved in cellular metabolism were very prominent, including oxidative phosphorylation, glycolysis and mTOR signaling pathways (Fig. 3d). Patients with sepsis manifested overexpression of genes encoding products involved in glycolysis and oxidative phosphorylation, coupled with those encoding products involved
a6 60
**
**
*
**
**
**
40
20
0
4
0 0
1
2
3
4
2TN
F (
ng/m
l)
Lact
ate
(mM
)
*
*
0
1
2
3
4
NA
D+ (
nmol
/g p
rote
in)
IL-6
(ng
/ml)
bRPM
I
LPS
Candid
a
RPMI
LPS
Candid
a
RPMI
LPS
Candid
a
RPMI
LPS
Candid
a
Donor A
p-S6K
p-4EBP1
Actin
Donor B
Metformin Metformin
RPMI LPSC. albicans
©20
16N
atu
re A
mer
ica,
Inc.
All
rig
hts
res
erve
d.
nature immunology aDVaNCE ONLINE PUBLICaTION �
A rt i c l e s
in mitochondrial dysfunction, whereas genes encoding products involved in fatty acid oxidation and mTOR signaling were under-expressed (Fig. 3e). Furthermore, patients with sepsis had overexpres-sion of genes characteristically expressed by M2 macrophages and under-expression of genes characteristically expressed by M1 macro-phages (Fig. 3e). Analysis of transcription factor–binding site motifs in the promoter regions of genes altered during sepsis revealed a clear role for the mTOR-dependent HIF-1α axis in the regulation leukocyte genes during both E. coli sepsis and Candida sepsis (Supplementary Fig. 3d). This was in line with the findings of other studies that have shown the importance of HIF-1α in processes other than metabo-lism during the functional reprogramming of human monocytes in sepsis2. These data showed that during sepsis, profound changes in the metabolism of leukocytes were encountered.
Defective energy metabolism of tolerant monocytes In a subsequent set of experiments, we assessed whether the induction of innate immunotolerance in human monocytes modulated cellular metabolism. We used an in vitro model of innate immunotolerance induced by stimulation of monocytes with high concentrations of LPS or C. albicans (in contrast to the low concentrations that induce priming11,19,20). When monocytes were rendered immunotolerant, the production of pro-inflammatory cytokines after restimulation was almost completely abolished (Fig. 4a). In line with that, restimula-tion of tolerant monocytes with LPS resulted in diminished produc-tion of lactate and enhanced production of NAD+ compared with that of non-tolerant cells (Fig. 4a), indicative of a defective ability to mount a Warburg effect. Immunoblot analysis of components of the mTOR pathway showed less phosphorylation of mTORC1 targets in tolerant cells than in non-tolerant cells (Fig. 4b). When glycolysis is impaired, the β-oxidation of fatty acids is often used as an alternative
source of energy21. However, monocytes that were rendered toler-ant by pre-exposure to LPS were also characterized by decreased amounts of the fatty-acid transporters CD36 and CPT1 compared with that of control cells exposed to culture medium alone (Fig. 4b). Finally, immunotolerant monocytes also showed a marked decrease in oxygen consumption compared with that of naive monocytes (Fig. 4c). Therefore, a decrease in all major metabolic pathways (glycolysis, oxidative phosphorylation and β-oxidation) occurred in immunotolerant monocytes.
Severe metabolic defects in leukocytes of patients with sepsisTo determine whether defects in cellular metabolism similar to those noted above characterized leukocytes from patients with sepsis, we assessed the cytokine-producing capacity and glycolysis of cells from a separate set of patients with sepsis that either responded normally to stimulation with LPS (high responders (n = 6)) or lacked the ability to produce cytokines after ex vivo stimulation with LPS (immuno-tolerant (n = 5)). Stimulation of PBMCs with LPS showed normal secretion of cytokines from PBMCs from healthy control subjects and those from high-responder patients with sepsis (Fig. 5a). These cells were also able to mount the release of lactate and NAD+ (Fig. 5a). In contrast, PBMCs from immunotolerant patients with sepsis were unable to produce lactate or NAD+ after stimulation with LPS (Fig. 5a). We also assessed metabolic processes in monocytes from patients during septic shock and recovery. When we monitored immunotolerant patients with sepsis over time (>7 d), we found that the ex vivo cellular production of cytokines and lactate was restored after patients recovered from sepsis (Fig. 5b).
Assessment of the mTOR pathway and fatty-acid transporters (CD36 and CPT1) in monocytes by immunoblot analysis showed diminished activation of these factors in patients during sepsis,
4a
*3
TN
F (
ng/m
l)
IL-6
(ng
/ml)
2
1
0
4 5 2.0 *1.5 RPMI
Metformin1.0
0.5
0
4 **3210
3
TN
F (
ng/m
l)
IL-6
(ng
/ml)
IL-1
β (n
g/m
l)
2
1
0
5 2.0
1.5
1.0
DMSORapamycinTorin
IL-1
β (n
g/m
l)
0.5
0
**
**
43210
b1.5 2.5
2.0
* DMSORapamycinTorin
MetforminRPMI
*1.51.0
IL-2
2 (n
g/m
l)
0.50
1.0
0.5
IL-1
7 (n
g/m
l)
0
600
400
200
0
300
400 * *
**
*1.0 2.0
1.5
1.0
IL-2
2 (n
g/m
l)
0.5
0
0.80.60.4
IL-1
7 (n
g/m
l)
0.20
200
100
0
IFN
-γ (
pg/m
l)IF
N-γ
(pg
/ml)
c PBSMetformin
10080604020S
urvi
val (
%)
00 10 20 30 40
Time (d)
**
e80 800 0.07*
600
400
200
0
60
40
20
lL-1
β (p
g/m
l)
lL-6
(pg
/ml)
0
2.0 500400300200100
0
1.5
1.0
0.5
TN
F (
ng/m
l)
IFN
-γ (
pg/m
l)
0
8001,000 * *
*
600
400
200
0
600400200
IL-1
7 (p
g/m
l)
IL-2
2 (p
g/m
l)
0
d
0
2
543210
4
6 *
6 5* 4
PBSMetformin
321
0
4
2
0
C. a
lbic
ans
(log 1
0 C
FU
)
C. a
lbic
ans
(log 1
0 C
FU
)
C. a
lbic
ans
(lo
g 10
CF
U)
C. a
lbic
ans
(log 1
0 C
FU
)
PBSMetformin
Kidney (day 1)
Kidney (day 3)
Liver (day 1)
Liver (day 3)
Figure 2 Inhibition of the mTOR pathway diminishes cytokine production and increases mortality. (a,b) Production of TNF, IL-6 and IL-1β (a) or IFN-γ, IL-17 and IL-22 (b) by human PBMCs (a) or human lymphocyte (b) treated with vehicle (DMSO), rapamycin or torin-1 (top row) or left untreated (RPMI) or treated with metformin (bottom row) and left unstimulated (left bar of each pair) or simulated for 24 h (a) or 7d (b) with C. albicans (n = 6 individual donors). (c) Survival of mice (n = 10 per group) given a daily intraperitoneal injection of PBS or metformin (key) starting 1 d before infection with C. albicans and continuing until 7 d afterward. (d) Fungal outgrowth in the kidneys (left) and liver (right) 1 and 3 days post infection of the mice in a comparable experiment as in c (n = 5 per group). (e) Production of mouse IL-1β, IL-6, TNF, IFN-γ, IL-17 and IL-22 by splenocytes obtained from mice as in c at day 7 after infection and left unstimulated (left bar of each pair) or restimulated for 24 h in vitro with LPS (IL-1β, IL-6 and TNF) or for 5 d in vitro with C. albicans (IFN-γ, IL-17 and IL-22). *P < 0.05 and **P < 0.001 (Wilcoxon signed-rank test (a,b,d,e) or Kaplan-Meier curve (c)). Data are representative of two experiments (a,b; mean + s.e.m.) or one experiment (c–e; mean + s.e.m. in d,e).
©20
16N
atu
re A
mer
ica,
Inc.
All
rig
hts
res
erve
d.
� aDVaNCE ONLINE PUBLICaTION nature immunology
A rt i c l e s
d e3.5
3.0
Adj
P v
al (
–log
10)
Adj
P v
al (
–log
10)
2.52.01.5
1.00.5
0
14.012.010.0
8.06.04.02.0
0
Mito
chon
drial
dys
func
tion
EIF2
signa
ling
Regula
tion
of e
lF4
and
p70S
6K si
gnali
ng
iCOS-iC
OSL sig
nalin
g in
helpe
r T ce
lls
mTOR si
gnali
ng
CD28 si
gnali
ng in
help
er T
cells
PKCθ sign
aling
in T
lym
phoc
ytes
Natur
al kil
ler ce
ll sign
aling
T cell a
ntige
n re
cept
or si
gnali
ng
Calcium
-indu
ced
T lym
phoc
yte a
popt
osis
Role o
f NFAT in
regu
lation
of t
he im
mun
e re
spon
se
Oxidat
ive p
hosp
hory
lation
IL-1
0 sig
nalin
g
IL-1
sign
aling
iNOS si
gnali
ng
Cyclin
s and
cell c
ycle
regu
lation
Argini
ne d
egra
datio
n VI (
argin
ase
2 pa
thway
)
Glycog
en b
iosyn
thes
is II
(from
udp
-d-g
lucos
e)
IL-6
sign
aling
Argini
ne d
egra
datio
n I (
argin
ase
path
way)
ATP5S EIF4G1TSC2EIF4EBP1AKT1PRKCQEIF3MEIF3HEIF4BEIF3EEIF3DEIF3LRPS19RPS25RPS4XRPS15RPS16RPS5VEGFBEIF3KDGKZEIF3GEIF4A1RPS6KA4NAPEPLDRRAS2RAC1RPS6PPP2R5CPPP2R5EPPP2R1APRKCBPRKCIPRKD3RHOQRPS23RPS28RPS29RPS11RPS14EIF3BAKT3PIK3C2BRPS21EIF3IRPS18EIF3ERPS3PRKCZPIK3C2AEIF4G2MAPK1MAPKAP1RHOT2PDPK1PIK3CDPIK3R5ATG13TSC1EIF4A3PPP2CBPIK3R4PRKCHRPS6KA5FNBP1ATMEIF3AMTORPIK3R1RHOFPRKAG2RHOC
COX7A2LVPS9D1NDUFV2UQCRBATP5ECOX7CCOX7BNDUFA4NDUFA2COX17NDUFS5ATP5LNDUFA6UQCR10ATP5J2COX6A1COX7A2NDUFA1NDUFB1NDUFB3COX6CNDUFS4ATP5C1ATP5J UQCRQ
mTOR signalingOxidative phosphorylation
Fatty acid oxidation
Glycolysis I
M1 markers
ALDH4A1TMLHEACAA1ACSL4ACSL3ACSL1SLC27A1ALDH1A1SLC27A3ALDH9A1ACSL5HSD17B10HADHALDH3A2HSD17B8ECHS1PTGS2ALDH2HSD17B4HADHAHADHB
PGAM1
M2 markersGAPDHPGK1ENO1GPI
FBP1ALDOC
IL1R1TNFFCER2HLA-AIL23ACD86HLA-DRB1
Healthy subjectsScaled expression
–4 –2 0 2 4Patient response to C. albicans infection Patient response to E. coli infection
CLEC7ATGFBIMAFCD36TGFBR2TGM2CCL22MS4A6ACD163MS4A4AIL1R2IL1RNIL10
PTGS2
50 40
30
20
10
0
40
30
20
10
0
TMLHEHADHBALDH2ACAA1ACSL4ACSL3ACSL1ALDH1A1HSD17B4ECHS1HSD17B8HADHHSD17B10SLC27A1ALDH3A2ALDH9A1ACSL5ALDH4A1SLC27A3HADHA
IL23ACD86FCER2CCL4IL1R1CD83IL1BTNF
TGM2IL1R2IL1RNCD36MS4A6ACD302MAFCLEC7ACD163MS4A4A
Fatty acid oxidation
M1 markers
M2 markers
RPS6KA1RPS6KA2PPP2R2BEIF3APRKCZRPS26EIF4A2FNBP1EIF3BRPS6KB2AKT1EIF3IRPS11RHOCDDIT4EIF3LEIF4BPIK3R1PRKCHPPP2R1ARPS27ARPS23RRASRPS6KA4RPS15ARPS17RPS27PPP2R4RPS4Y1EIF3HRPS19EIF3KRPS10RPS5RPS15RPS2RPS14RPS29EIF3ERPS20RPS28VEGFBEIF3M EIF3GEIF3DEIF4A1RPS13RPS25RPS24RPS27L
Oxidative phosphorylationba
c
mTOR signaling
COX8A
UQCRFS1COX4I1SDHB
ATP5LNDUFS7ATP5BSURF1NDUFA4COX7C ATP5ISDHAF1
NDUFA8COX10
CYC1ATP5C1ATP5O
NDUFA12NDUFB5
ATP5A1ATP5G2NDUFB11ATP5G1
NDUFS8NDUFB8
COX5A NDUFA3NDUFS3SDHANDUFA13ATP5DNDUFA7NDUFB10UQCRHLNDUFAB1UQCRHATP5HATP5F1NDUFS5NDUFS4ATPAF1ATP5JNDUFA11NDUFB7NDUFB2NDUFV1UQCRQ
Glycolysis I
Before LPS After LPS
Adj
P v
alue
(–l
og10
)
Adj
P v
alue
(–l
og10
)
PFKPFBP1GPITPI1ENO1ALDOC
Scaled expression
–2 –1 0 1 2
Unique patient response to C. albicansUnique patient response to E. coli
Common patient response
C. albicansE. coli
6761,624 7661,367
Expression (log2 fold)
943775
1,9564,021
492
86420
86420
Common patient response
–2
–2
–4
–4
86420–2–4 86420–2–4
Exp
ress
ion
(log
2 fo
ld)
C. a
lbic
ans
Expression (log2 fold)
Expression (log2 fold)E. coli
338
Figure 3 Blood transcriptome analysis of human endotoxemia and critically ill septic patients with culture-confirmed systemic infection with Candida or E. coli. (a) Genome-wide transcriptional profiles of blood from healthy male subjects (n = 8) before and 4 h after (above plots; gray and green key) the administration of E. coli LPS (4 ng per kg body weight) in a clinically controlled setting; expression of genes (right margins) is presented as centered and ‘scaled’ log2 fluorescence intensity (blue and red key); labels above plots indicate grouping of genes by product function. (b) Genome-wide transcriptome analysis of blood from patients critically ill with clinically well-defined sepsis (Supplementary Table 1) with confirmed infection of the bloodstream with E. coli (left) or C. albicans (right), with expression (log2 values) plotted against the adjusted P value for the difference in expression (Adj P val). Numbers in top left and right corners indicate number of genes with significantly differential expression (adjusted P value, <0.05), upregulated (red) or downregulated (blue) onefold or more in septic patients relative to that in healthy subjects. (c) Unique and overlapping blood transcriptional responses in E. coli– or C. albicans–infected septic patients in b (left); arrows and adjacent numbers indicate genes with significantly differential expression (adjusted P value, <0.05), upregulated (upward red arrow) or downregulated (downward blue arrow) in septic patients. Right, gene expression in E. coli–infected septic patients plotted against that in C. albicans–infected septic patients (Spearman’s rho = 0.91; P < 2.2 × 10−16). (d) Significance of the difference in the expression of over-expressed genes (top row) or under-expressed genes (bottom row) common to Candida sepsis and E. coli sepsis, categorized into the top 10 most over-represented pathways in which the gene products are involved (horizontal axes). (e) Transcriptional profiles of healthy subjects and E. coli– or C. albicans–infected septic patients in b (above plots; gray, gold and green key), presented as in a. Data are representative of one experiment with 8 subjects (endotoxemia) or 46 subjects (sepsis).
©20
16N
atu
re A
mer
ica,
Inc.
All
rig
hts
res
erve
d.
nature immunology aDVaNCE ONLINE PUBLICaTION �
A rt i c l e s
while these effects were reversed once patients recovered (Fig. 5c). Consistent with the data reported above showing severe metabolic disturbances in patients with sepsis, the maximum oxygen consump-tion of PBMCs was also substantially impaired in patients with sepsis, compared with that in healthy control subjects (Fig. 5d). These data demonstrated that both glycolytic pathways and oxidative-metabolism pathways were substantially downregulated in patients with sepsis, which led to what we called ‘immunometabolic paralysis’ here (Supplementary Fig. 4).
Partial restoration of cellular metabolism by IFN-gRestoring energy metabolism might represent a potential therapy for patients with sepsis with immunometabolic paralysis. IFN-γ is a cytokine with the ability to potently stimulate monocyte function, and recombinant IFN-γ therapy in septic patients has been shown to (partially) reverse immunoparalysis both in vitro and in vivo22–24.
Moreover, IFN-γ polarizes macrophage differentiation toward an M1 phenotype, which is associated with increased glycolysis25, and the mTOR–HIF-1α–regulated metabolic switch from oxidative phosphor-ylation to glycolysis in mouse dendritic cells and splenocytes is depend-ent on autocrine IFN-γ production26. We therefore hypothesized
–2 0123456
*
0
100 110 120 130 140 150 160 170
2
4
6
Time (min)
LPSRPMI
RPMIC. albicans
Basal
Leak
Max Rox
O2 fl
ow p
er
106 c
ells
(pm
ol/s
)
–2
0
90 100 120110 130 140 150
2
4
6
Time (min)
O2 fl
ow p
er
106
cells
(pm
ol/s
)
Res
pira
tion
per
106 c
ells
(pm
ol/s
)
0123456 *
Basal
Leak
Max Rox
Res
pira
tion
per
106 c
ells
(pm
ol/s
)
cLPSRPMI
RPMIC. albicans
0 0 0
5
10
15
0.5
1.0
1.5
2.0
1
2
3***
TN
F (
ng/m
l)
Lact
ate
(mM
)
NA
D+ (
nmol
/g p
rote
in)
**a*
**
*
*
*
Control
LPSC. albicans
Prestimulation Prestimulation
p-S6K
p-4EBP1
Actin
Restimulation
Actin
CPT1
CD36
RestimulationRPMI
RPMI
LPS RPMI LPS
LPS
RPMI
LPS
RPMI
LPS
RPMI
LPSb
Figure 4 In vitro innate immunotolerant monocytes have impaired metabolism. (a) Production of TNF, IL-6, lactate and NAD+ by monocytes (n = 6 individual donors) left untreated (Control) or pre-incubated with LPS (10 ng/ml) or C. albicans (1 × 104 colony-forming units per ml) (key), allowed to ‘rest’ for 24 h, then restimulated for 24 h with RPMI (left-most bar of each pair or group) or LPS (10 ng/ml). (b) Immunoblot analysis of phosphorylated S6K and 4EBP1 (left) and CD36 and CPT1 (right) in monocytes (n = 4 individual donors) left untreated (RPMI) or pre-incubated with LPS (above lanes), allowed to ‘rest’ for 24 h, then restimulated for 24 h with RPMI or LPS (below lanes). (c) Oxygen consumption of monocytes (n = 1 donor (left); pooled data from n = 6 individual donors (right)) left untreated (RPMI) or pre-incubated with LPS (top row) or C. albicans (bottom row), then allowed to ‘rest’ for 24 h; respiration results are presented as basal respiration (basal), proton leak after inhibition of ATP synthase (leak), maximal oxygen consumption (max), residual oxygen consumption or non-mitochondrial oxygen consumption (rox). *P < 0.05, **P < 0.01 and ***P < 0.001 (Wilcoxon signed-rank test). Data are representative of two experiments (a,b; mean + s.e.m. in a) or six experiments (c; mean + s.e.m.).
d 86420
–2
86420
–280 90 100
110 120 130 140 120 130Time (min)Time (min)
Time (min)
HealthySeptic
O2
flow
per
106
cells
(pm
ol/s
)
O2
flow
per
106
cells
(pm
ol/s
)
O2
flow
per
106
cells
(pm
ol/s
)
Res
pira
tion
per
106
cells
(pm
ol/s
) HealthySeptic
140
86420
–2
8
6
4
2
0
Basal
Leak
Max Rox
Healthy
Septic(immunotolerant)
Septic(after recovery)
c
Time (d)p-S6K
p-4EBP1
CD36
CPT1
Actin
0 7 13 0 7 7
Patient 4Patient 3Patient 2Patient 1
0 04 14
b
∆ La
ctat
e (m
M)
5
4
3
2
1
0
1.5
1.0
0.5
0
** *
*
TN
F (
ng/m
l)
15
10
**
5
0
400
300
200
100
0
IL-6
(ng
/ml)
IL-1
0 (p
g/m
l)
3
2
1
0
**
IL-1
β (
ng/m
l)
Healthy
Septic(high responders)
Septic(immunotolerant)
1.5
1.0
*
Lact
ate
(mM
)
0.5
0
50
40
30
20
10
0
**
∆ N
AD
+ (
nmol
/g p
rote
in)30
20
***
IL-6
(ng
/ml)
10
0
a 1.5
1.0***
TN
F (
ng/m
l)
0.5
0
40
20
30
*
AT
P (
nmol
/106 c
ells
)
10
0
Patient 1
Patient 3
Patient 2
Figure 5 Tolerant monocytes from septic patients show impaired metabolic pathways. (a) Production of TNF, IL-6 and ATP and synthesis of lactate and NAD+ by PBMCs from healthy control subjects (n = 11) and septic patients with normal cytokine responsiveness (high responders; n = 6) or defined as immunotolerant or displaying immunoparalysis (n = 5). (b) Production of TNF, IL-6, IL-1β and IL-10, and change in lactate production (∆ lactate: RPMI-stimulated samples versus LPS-stimulated samples), in monocytes obtained from healthy control subjects (n = 10) and immunotolerant septic patients (n = 8), before (black) and after (gray) recovery, then stimulated ex vivo with LPS. (c) Immunoblot analysis of phosphorylated S6K and 4EBP1 and total CD36 and CPT1 in cells from septic patients (1–4; above blots) during immunoparalysis (initial time point(s) for each) and after recovery (final time point for each) (above lanes). (d) Oxygen consumption (top row and bottom left) in PBMCs from healthy control subjects (n = 3) and septic patients (1–3); bottom right, pooled data (presented as in Fig. 4c). *P < 0.05, **P < 0.01 and ***P < 0.001 (one-way analysis of variance). Data are representative of eight experiments (a; mean + s.e.m.), six experiments (b; mean + s.e.m.), four experiments (c) or three experiments (d; mean + s.e.m.).
©20
16N
atu
re A
mer
ica,
Inc.
All
rig
hts
res
erve
d.
� aDVaNCE ONLINE PUBLICaTION nature immunology
A rt i c l e s
that IFN-γ might be able to restore glycolysis in tolerant monocytes. First, we showed that treating LPS-tolerant human monocytes in vitro with IFN-γ restored lactate production upon stimulation with LPS (Fig. 6a), while oxygen consumption was not influenced (Fig. 6b). Inhibition of the mTOR pathway with rapamycin abrogated the effect of IFN-γ (Fig. 6a). To further assess the mTOR pathway in IFN-γ therapy, we exposed tolerant and non-tolerant (naive) monocytes to IFN-γ for 2 h, after which we assessed the phosphorylation of the mTOR pathway by immunoblot. IFN-γ was able to induce phos-phorylation of components of the mTOR pathway in both tolerant monocytes and non-tolerant monocytes (Fig. 6c). Furthermore, in a proof-of-concept trial24, we treated patients with fungal sepsis (n = 8) with recombinant IFN-γ three times a week. Ex vivo stimulation of PBMCs isolated after subcutaneous administration of recombinant IFN-γ showed an increase in lactate production relative to lactate production before IFN-γ treatment that accompanied upregula-tion of the cytokine-producing capacity24 (Fig. 6d), indicative of a restoration of the ability to mount glycolytic responses. Therefore, IFN-γ was able to partially restore the metabolic defects of innate immunotolerance by promoting glycolysis, which was impaired in immunotolerant monocytes.
DISCUSSIONCollectively, our study supports three main conclusions. First, during an acute phase of infection, the shift from oxidative phosphorylation to aerobic glycolysis (the Warburg effect) was an important mecha-nism of host defense. Second, leukocytes from patients with severe sepsis displayed severe defects in cellular energy metabolism that were associated with an impaired ability to respond to secondary stimu-lation, a process equivalent to a true ‘immunometabolic paralysis’. Third, therapeutic approaches that reverse this metabolic defect might represent a novel therapeutic approach for the treatment of sepsis.
In our experiments, tolerant leukocytes had broad defects in metabolic pathways, not only failing to increase glycolysis but also
displaying decreased oxygen consumption and less transport of fatty acids. These observations were somewhat unexpected, as they showed that leukocytes from patients with sepsis did not merely reverse glucose metabolism from glycolysis to oxidative phosphorylation but displayed multiple energy-metabolizing defects. These defects cor-related with a loss of cytokine-producing ability, which was reversed upon recovery. Mitochondrial dysfunction has been shown to cor-relate with a poor outcome in sepsis, but such studies have focused mainly on non-hematopoietic cells27, with only one study reporting defects in leukocytes8. Here we found that the defects comprised sev-eral metabolic pathways and correlated with immunological function. A remaining point of discussion is whether the metabolic defects were one of the main causes of the immunological phenotype or whether they were a feature of adaptation of cells of the immune system during infection. However, there are suggestions that infection and metabolic defects are intertwined: for example, on the one hand, we showed that inhibition of metabolic pathways influenced the immunological response, and on the other hand, it has been shown that IL-10 down-regulates glycolysis in a mouse model28, which suggests a relationship between cytokine production and metabolism. Finally, another point is the cell populations in which these metabolic changes occur during sepsis. Our experiments showing the importance of the regulation of metabolic pathways in innate immunotolerance used a whole-blood stimulation assay that mimicked the interplay of all cells of the immune system. Although we performed additional experiments to assess the effects on more specific cell populations (monocytes) in vitro, comprehensive analysis of the in vivo changes in various cell subpopulations in patients with sepsis is warranted.
Another discussion point is the finding that metformin, used as a modulator of metabolic pathways, displays effects beyond simply activation of AMPK and inhibition of the mTOR pathway. We nev-ertheless decided to investigate its effects because of its broad clinical use, which made these experiments relevant from a translational per-spective. The mTOR inhibitor rapamycin could not be used to study
d
*
*
* **
**
**
D1 D2 D3 D7 D14 D28
LPSPHAC. albicans
Lact
ate
(fol
d)
2.5
2.0
1.5
1.0
0.5
0
0.6
Naive
0
***
*
***
00.
01 0.1
IFN-γ (µg/ml)
1
LPS tolerant
0.4
∆ La
ctat
e (m
M)
0.2
0
–0.2
a b c
0.20
0.15
0.10
0.3
0.2
0.1
0
0.05
0
Naive
Naive
Naive
Naive
*** * *** ** *
** 0.06* **
*
PBS
Basal
Leak
Max
IFN-γ
IFN-γ – + – +
Tolerant
LPS to
leran
t
LPS to
leran
t
LPS to
leran
t
Naive
NaiveNaiv
e
LPS to
leran
t
LPS to
leran
t
LPS to
leran
t
Lact
ate
(mM
)
Res
pira
tion
per
106 c
ells
(pm
ol/s
)
Lact
ate
(mM
)
p-m
TO
R/a
ctin
p-4E
BP
1/ac
tin
p-S
6K/a
ctin
4
p-mTOR
p-S6K
p-4EBP1
Actin
3
2
1
0
4
3
2
1
0
4
3
2
1
0
8
6
4
2
0
–IFN-γ (1 µg/ml)IFN-γ (1 µg/ml) + rap
–IFN-γ (100 ng/ml)IFN-γ (100 ng/ml) + rap Naive
Naive + IFN-γ
TolerantTolerant + IFN-γ
Figure 6 Immunotherapy with IFN-γ partially restores the metabolic defects by induction of the mTOR pathway. (a) Change in lactate production in purified monocytes left untreated (Naive) or pre-incubated with LPS (LPS tolerant), allowed to ‘rest’ for 24 h, then treated in vitro with various concentrations (horizontal axis) of IFN-γ, and then restimulated for 24 h with LPS (left), and lactate production in purified monocytes (n = 6 individual donors) left untreated (Naive) or pre-incubated with LPS (LPS tolerant), allowed to ‘rest’ for 24 h, then left untreated (−) or treated with IFN-γ (1 µg/ml (middle) or 100 ng/ml (right)) in the presence (+ rapamycin) or absence of rapamycin (key), and then restimulated for 24 h with LPS (middle and right). (b) Respiration (presented as in Fig. 4c) in tolerant monocytes (n = 4 individual donors) pre-incubated with LPS, allowed to ‘rest’ for 24 h, then treated for 3 h with PBS or IFN-γ (1 µg/ml). (c) Immunoblot analysis (left) of phosphorylated mTOR, S6K and 4EBP1 in monocytes (n = 6 individual donors) pre-incubated with LPS (Tolerant) or left untreated (Naive), allowed to ‘rest’ for 24 h, then left untreated (−) or treated for 2 h with IFN-γ (1 µg/ml) (below lanes). Right, quantification of the results at left, presented as band intensity for phosphorylated mTOR, S6K or 4EBP1 relative to that of actin. (d) Lactate production of PBMCs obtained from septic patients24 (n = 5) before (day 1 (D1)) and after (from day 2 (D2) and onward) treatment with recombinant IFN-γ (100 mg, administered subcutaneously three times a week), followed by restimulation of cells ex vivo for 24 h with LPS, PHA or C. albicans. *P < 0.05, **P < 0.01 and ***P < 0.001 (one-way analysis of variance (a,d) or Wilcoxon signed-rank test (b,c)). Data are representative of two experiments (a–c; mean + s.e.m.) or one experiment (d; mean + s.e.m.).
©20
16N
atu
re A
mer
ica,
Inc.
All
rig
hts
res
erve
d.
nature immunology aDVaNCE ONLINE PUBLICaTION �
A rt i c l e s
cellular metabolism in the live infection model, due to its well-known antifungal effect29. However, our use of more specific inhibitors of mTOR in vitro, such as torin-1 and rapamycin, confirmed the effect of mTOR on inflammation. Furthermore, many studies have reported that rapamycin inhibits TNF production, suggestive of an anti-inflammatory effect, similar to our observations. In contrast, other studies have shown that rapamycin upregulates IL-12 produc-tion and downregulates IL-10 production and thus results in immu-nostimulatory effects, as seen for IL-6 in our in vitro experiments. Several hypotheses have been raised to explain these differences, the most likely being that the kinase Akt–mTOR–HIF-1α pathway has differential effects on the various cytokines, depending on the cell type and stimulus or infection present30.
Our study has identified cellular metabolic processes as a potential therapeutic target in sepsis. So far, only very few agents with known metabolic-regulatory ability have been tested in sepsis models. Among these, activation of the transcription factor PPARγ, which is associ-ated with increased glycolysis and fatty-acid metabolism31–33, has been shown to have a beneficial effect on mortality in mouse sepsis models3,34,35. Furthermore, several therapeutic agents able to restore mitochondrial activity have been studied in rodent models of sepsis, with beneficial outcomes36. In the present study, we investigated IFN-γ, as it is known to upregulate cytokine production in innate immuno-tolerance22–24. Autocrine IFN-γ production is involved in the mTOR–HIF-1α–dependent metabolic switch from oxidative phosphorylation to glycolysis in mouse dendritic cells and splenocytes26. Indeed, our data showed that in vitro exposure of immunotolerant monocytes to IFN-γ activated the mTOR pathway and partially restored the ability of leukocytes to mount a glycolytic response. Those published data26 and our findings differ from those of another published report showing the downregulatory effect of IFN-γ on mTORC1, with methodological dif-ferences and different cell populations as likely causes of this37. Thus, metabolic recovery and immunological recovery seem to be correlated, and improved strategies to correct the metabolic abnormalities should be considered for patients with sepsis and immunoparalysis.
Apart from the effects of IFN-γ on mTOR, other potential mechanisms might have a role as well. First, IFN-γ induces sus-tained occupancy of the transcription factors STAT1 and IRF1 and also induces histone acetylation at promoter and enhancer sites of genes encoding pro-inflammatory cytokines38 and possibly genes encoding metabolic factors. Second, the microRNA miRNA-146 has an important role in disruption of the synthesis of pro-inflammatory cytokines in innate immunotolerance39, and IFN-γ could potentially influence the expression of miRNA-146, as has been shown for other microRNAs40,41. Third, restoring metabolism pathways via IFN-γ could potentially lead to an increase in succinate, an important positive regulator of IL-1β production via HIF-1α42. Finally, sirtuins have been shown to have an important role in the regulation of the immunotolerant phenotype of monocytes, and inhibition of the sir-tuin SIRT1 can restore the immunotolerant (metabolic) features of monocytes21. Promisingly, IFN-γ has also been shown to repress the expression of SIRT1 in muscle cells43.
In conclusion, the switch from oxidative phosphorylation to aero-bic glycolysis is an important component of the cellular response to infection. Broad defects in the energy metabolism of leukocytes comprising both glycolysis and oxidative mechanism characterized a state of immunometabolic paralysis in patients with severe sepsis, and this not only gives new insight into the pathophysiology of the disease but also might represent a novel therapeutic target. Immunotherapy with recombinant IFN-γ is one of the potential approaches to cor-recting these defects.
METhODSMethods and any associated references are available in the online version of the paper.
Accession codes. GEO: microarray data, GSE48119.
Note: Any Supplementary Information and Source Data files are available in the online version of the paper.
ACKNOWLEdGMENTSSupported by the European Research Council (ERC-StG-310372 to M.G.N.) and the Center for Translational Molecular Medicine (Molecular Diagnosis and Risk Stratification of Sepsis project; 04I-201).
AUTHOR CONTRIBUTIONSS.-C.C., B.P.S., R.J.W.A., M.S.G., E.L., E.J.G.-B., M.K., G.R.M., J.A.L.W., J.L. and A.J.v.d.M. performed the experiments; B.P.S., R.J.W.A. and J.A.L.W. performed the analyses; E.J.G.-B., O.L.C., F.L.v.d.V., M.J.B., M.J.S. and P.P. were involved in the clinical studies; P.H.G.M.W. L.A.B.J., T.v.d.P. and M.G.N. designed the studies; and all authors were involved in writing and correcting the manuscript.
COMPETING FINANCIAL INTERESTSThe authors declare no competing financial interests.
reprints and permissions information is available online at http://www.nature.com/reprints/index.html.
1. Angus, D.C. & van der Poll, T. Severe sepsis and septic shock. N. Engl. J. Med. 369, 840–851 (2013).
2. Shalova, I.N. et al. Human monocytes undergo functional re-programming during sepsis mediated by hypoxia-inducible factor-1α. Immunity 42, 484–498 (2015).
3. Vachharajani, V., Liu, T. & McCall, C.E. Epigenetic coordination of acute systemic inflammation: potential therapeutic targets. Expert Rev. Clin. Immunol. 10, 1141–1150 (2014).
4. Cheng, S.C., Joosten, L.A. & Netea, M.G. The interplay between central metabolism and innate immune responses. Cytokine Growth Factor Rev. 25, 707–713 (2014).
5. Carré, J.E. et al. Survival in critical illness is associated with early activation of mitochondrial biogenesis. Am. J. Respir. Crit. Care Med. 182, 745–751 (2010).
6. Singer, M. Cellular dysfunction in sepsis. Clin. Chest Med. 29, 655–660 (2008).7. Carré, J.E. & Singer, M. Cellular energetic metabolism in sepsis: the need for a
systems approach. Biochim. Biophys. Acta 1777, 763–771 (2008).8. Belikova, I. et al. Oxygen consumption of human peripheral blood mononuclear
cells in severe human sepsis. Crit. Care Med. 35, 2702–2708 (2007).9. Chen, T. et al. Redox imbalance provokes deactivation of macrophages in sepsis.
Proteomics Clin. Appl. 3, 1000–1009 (2009).10. Smeekens, S.P. et al. Functional genomics identifies type I interferon pathway as
central for host defense against Candida albicans. Nat. Commun. 4, 1342 (2013).
11. Cheng, S.C. et al. mTOR- and HIF-1α-mediated aerobic glycolysis as metabolic basis for trained immunity. Science 345, 1250684 (2014).
12. Rodríguez-Prados, J.C. et al. Substrate fate in activated macrophages: a comparison between innate, classic, and alternative activation. J. Immunol. 185, 605–614 (2010).
13. Inoki, K., Ouyang, H., Li, Y. & Guan, K.L. Signaling by target of rapamycin proteins in cell growth control. Microbiol. Mol. Biol. Rev. 69, 79–100 (2005).
14. Choi, Y.K. & Park, K.G. Metabolic roles of AMPK and metformin in cancer cells. Mol. Cells 36, 279–287 (2013).
15. Zhou, H., Luo, Y. & Huang, S. Updates of mTOR inhibitors. Anticancer. Agents Med. Chem. 10, 571–581 (2010).
16. van der Meer, A.J. et al. The selective sirtuin 1 activator SRT2104 reduces endotoxin-induced cytokine release and coagulation activation in humans. Crit. Care Med. 43, e199–e202 (2015).
17. Scicluna, B.P. et al. Role of tumor necrosis factor-α in the human systemic endotoxin-induced transcriptome. PLoS One 8, e79051 (2013).
18. Calvano, S.E. et al. Inflamm and Host Response to Injury Large Scale Collab. Res. Program. A network-based analysis of systemic inflammation in humans. Nature 437, 1032–1037 (2005).
19. Saeed, S. et al. Epigenetic programming of monocyte-to-macrophage differentiation and trained innate immunity. Science 345, 1251086 (2014).
20. Morris, M.C., Gilliam, E.A. & Li, L. Innate immune programing by endotoxin and its pathological consequences. Front. Immunol. 5, 680 (2014).
21. Liu, T.F., Vachharajani, V.T., Yoza, B.K. & McCall, C.E. NAD+-dependent sirtuin 1 and 6 proteins coordinate a switch from glucose to fatty acid oxidation during the acute inflammatory response. J. Biol. Chem. 287, 25758–25769 (2012).
©20
16N
atu
re A
mer
ica,
Inc.
All
rig
hts
res
erve
d.
� aDVaNCE ONLINE PUBLICaTION nature immunology
A rt i c l e s
22. Döcke, W.D. et al. Monocyte deactivation in septic patients: restoration by IFN-γ treatment. Nat. Med. 3, 678–681 (1997).
23. Leentjens, J. et al. Reversal of immunoparalysis in humans in vivo: a double-blind, placebo-controlled, randomized pilot study. Am. J. Respir. Crit. Care Med. 186, 838–845 (2012).
24. Delsing, C.E. et al. Interferon-γ as adjunctive immunotherapy for invasive fungal infections: a case series. BMC Infect. Dis. 14, 166 (2014).
25. Zhu, L., Zhao, Q., Yang, T., Ding, W. & Zhao, Y. Cellular metabolism and macrophage functional polarization. Int. Rev. Immunol. 34, 82–100 (2015).
26. Pantel, A. et al. Direct type I IFN but not MDA5/TLR3 activation of dendritic cells is required for maturation and metabolic shift to glycolysis after poly IC stimulation. PLoS Biol. 12, e1001759 (2014).
27. Brealey, D. et al. Association between mitochondrial dysfunction and severity and outcome of septic shock. Lancet 360, 219–223 (2002).
28. Martin, F.P. et al. Metabolic assessment of gradual development of moderate experimental colitis in IL-10 deficient mice. J. Proteome Res. 8, 2376–2387 (2009).
29. Vézina, C., Kudelski, A. & Sehgal, S.N. Rapamycin (AY-22,989), a new antifungal antibiotic. I. Taxonomy of the producing streptomycete and isolation of the active principle. J. Antibiot. (Tokyo) 28, 721–726 (1975).
30. Weichhart, T., Hengstschläger, M. & Linke, M. Regulation of innate immune cell function by mTOR. Nat. Rev. Immunol. 15, 599–614 (2015).
31. Shapiro, H., Lutaty, A. & Ariel, A. Macrophages, meta-inflammation, and immuno-metabolism. ScientificWorldJournal 11, 2509–2529 (2011).
32. Panasyuk, G. et al. PPARγ contributes to PKM2 and HK2 expression in fatty liver. Nat. Commun. 3, 672 (2012).
33. Vats, D. et al. Oxidative metabolism and PGC-1β attenuate macrophage-mediated inflammation. Cell Metab. 4, 13–24 (2006).
34. Drosatos, K. et al. Peroxisome proliferator-activated receptor-γ activation prevents sepsis-related cardiac dysfunction and mortality in mice. Circ Heart Fail 6, 550–562 (2013).
35. Ferreira, A.E. et al. PPAR-γ/IL-10 axis inhibits MyD88 expression and ameliorates murine polymicrobial sepsis. J. Immunol. 192, 2357–2365 (2014).
36. Dare, A.J. et al. A systematic review of experimental treatments for mitochondrial dysfunction in sepsis and multiple organ dysfunction syndrome. Free Radic. Biol. Med. 47, 1517–1525 (2009).
37. Su, X. et al. Interferon-γ regulates cellular metabolism and mRNA translation to potentiate macrophage activation. Nat. Immunol. 16, 838–849 (2015).
38. Qiao, Y. et al. Synergistic activation of inflammatory cytokine genes by interferon-γ-induced chromatin remodeling and toll-like receptor signaling. Immunity 39, 454–469 (2013).
39. Brudecki, L., Ferguson, D.A., McCall, C.E. & El Gazzar, M. MicroRNA-146a and RBM4 form a negative feed-forward loop that disrupts cytokine mRNA translation following TLR4 responses in human THP-1 monocytes. Immunol. Cell Biol. 91, 532–540 (2013).
40. Schmitt, M.J. et al. Interferon-γ-induced activation of Signal Transducer and Activator of Transcription 1 (STAT1) up-regulates the tumor suppressing microRNA-29 family in melanoma cells. Cell Commun. Signal. 10, 41 (2012).
41. Imaizumi, T. et al. IFN-γ and TNF-α synergistically induce microRNA-155 which regulates TAB2/IP-10 expression in human mesangial cells. Am. J. Nephrol. 32, 462–468 (2010).
42. Tannahill, G.M. et al. Succinate is an inflammatory signal that induces IL-1( through HIF-1α. Nature 496, 238–242 (2013).
43. Li, P. et al. Interferon gamma (IFN-γ) disrupts energy expenditure and metabolic homeostasis by suppressing SIRT1 transcription. Nucleic Acids Res. 40, 1609–1620 (2012).
©20
16N
atu
re A
mer
ica,
Inc.
All
rig
hts
res
erve
d.
nature immunologydoi:10.1038/ni.3398
ONLINE METhODSPBMC and monocyte isolation. Buffy coats were obtained from healthy donors after written informed consent was provided (Sanquin blood bank, Nijmegen, The Netherlands). PBMC isolation was performed by dilution of blood in pyrogen-free PBS and differential density centrifugation over Ficoll-Paque (GE Healthcare). Cells were washed twice in PBS and resuspended in RPMI culture medium (Roswell Park Memorial Institute medium; Invitrogen) supplemented with 10 µg/ml gentamicin, 10 mM l-glutamine and 10 mM pyruvate. PBMCs were counted in a Coulter counter (Coulter Electronics) and adjusted to 5 × 106 cells/ml. A 100-µl volume was added to round-bottomed 96-well plates (Greiner) for PBMC experiments. For monocyte experiments, monocytes were separated from other cells by hyper-osmotic density gradi-ent centrifugation over Percoll (Sigma-Aldrich), and afterwards washed once with PBS. Monocytes were counted and adjusted to 1 × 106 cells/ml. A 100-µl volume was added to flat-bottomed 96-well plates (Corning) and cells were let to adhere for 1 h at 37 °C; subsequently, a washing step with 200 µl warm PBS was performed to yield maximal purity. Monocyte isolation from septic patients was performed by MACS isolation with CD14 positive magnetic beads (MACS Miltenyi) after the PBMC isolation step.
Monocyte-PBMC stimulation. Cells were stimulated with RPMI medium, Escherichia coli LPS (serotype 055:B5; 10 ng/ml; Sigma-Aldrich), or 1 × 106/ml Candida albicans ATCC MYA-3573 (UC 820) for 24 h or 7 days (in the pres-ence of 10% serum for lymphocyte-derived cytokines) and supernatants were stored at −20 °C until cytokine and lactate measurements were per-formed. For NAD+ measurements, cells were lysed in lysis buffer and stored at −80 °C. To induce tolerance, monocytes were pre-incubated with LPS (10 ng/ml) or Candida (1 × 104 colony-forming units (CFU)/ml) in the pres-ence of 10% serum. After 24 h, cells were washed once with warm PBS and let to rest for 24 h in 10% serum supplemented medium. Thereafter, cells were res-timulated with RPMI medium or 10 ng/ml LPS for 2 h (for immunoblot analy-sis) or for 24 h (for analysis of the production of cytokines, lactate and NAD+). Supernatants and lysates were stored as described above. In some experiments, cells were pre-incubated for 1 h with the mTOR-pathway inhibitors 10 nM rapamycin (Sigma) or 100 nM torin-1 (Tocris), or the AMPK inducer 3 mM metformin (R&D Systems), after which the stimulus was added. For the IFN-γ experiments, cells were pre-incubated for 1 h with 1 µg/ml of recombinant IFN-γ (Immukine; Boehringer Ingelheim) before restimulation.
Cytokine, lactate, NAD+, and ATP measurements. Cytokine production was determined in supernatants using commercial ELISA kits for TNF-α, IL-17 and IL-22 (R&D Systems) and IL-1β, IL-6 and IFN-γ (Sanquin) following the instructions of the manufacturer. Lactate concentration was measured using a Lactate Fluorometric Assay Kit (Biovision) and NAD+ was measured by a NAD/NADH Fluorometric Quantification Kit (Biovision). Lactate results are depicted as increase (∆) lactate production: the difference of lactate production between RPMI-stimulated samples and LPS-stimulated samples. ATP was measured by a bioluminescence ATP determination kit (Thermo Fisher Scientific).
Immunoblot analysis. Cells were cultured, lysed and stored as described above. Monocytes from septic patients were directly lysed after CD14 posi-tive MACS isolation and stored at −20 °C. Total protein amounts were deter-mined by BCA assay (Thermo Fisher Scientific), and equal amounts of proteins were loaded on pre-cast 4–15% gels (Bio-Rad). The separated proteins were transferred to a nitrocellulose membrane (Bio-Rad), which was blocked in 5% BSA (Sigma). Incubation overnight at 4 °C with rabbit polyclonal antibodies to phosphorylated S6K (9234), phosphorylated 4EBP1 (9459) and phosphor-ylated Akt (4060) (all from Cell Signaling), CPT1 (139480) and CD36 (9154) (both from Santa Cruz), and actin (A2066) (Sigma) were used to determine the protein expression, which was visualized using a polyclonal secondary anti-body (P0217; Dako) and SuperSignal West Femto Substrate (Thermo Fisher Scientific) or ECL (Bio-Rad).
Oxygen consumption. 5 × 106 PBMCs were incubated for 48 h with RPMI medium, LPS or Candida in the presence of 10% serum. Culture medium was collected, after which the cells were washed and re-suspended in 60 µl of the col-lected culture medium. Oxygen consumption was performed on 5 × 106 PBMCs
freshly isolated from septic patients or healthy volunteers, collected in 50 µl RPMI medium. The cell suspensions were then used for cellular O2 consump-tion analysis. Oxygen consumption was measured at 37 °C using polarographic oxygen sensors in a two-chamber Oxygraph (OROBOROS Instruments). First, basal respiration was measured. Next, leak respiration was determined by addition of the specific complex V inhibitor oligomycin A (2.5 µM). Then, maximum oxygen consumption was quantified by application of increasing concentrations of the mitochondrial uncoupler FCCP (p-trifluoromethoxy carbonyl cyanide phenyl hydrazone; final concentration, 0.5 to 14 µM). Finally, non-mitochondrial oxygen consumption was assessed by addition of the specific complex I inhibitor rotenone (30 nM) and the complex III inhibitor antimycin A (2.5 µM).
Animal experimental models. Animals experiments were conducted in accordance at the Center of Animal Use for Research Purposes of ATTIKON University hospital in accordance with the 56/2013 EU law (license 4/2013). Forty C57BL/6 male mice 7–9 weeks old weighting 25–27 g were studied. No power calculation was performed. Mice were housed in cages with a constant-flow air exchange, they were fed standard lab chow and they had access to water ab libitum. Mice were randomly assigned to each of three groups and interventions were performed after ether anesthesia. Researchers were blinded to the interventions: treated intraperitoneally on days −1, 0, 1, 2, 3, 4, 5, 6 and 7 with 200 µl phosphate buffered saline (PBS), pH = 7.2, and challenged with one log-phase 5 × 106 per mouse inoculum of C. albicans intravenously via the tail vein on day 0; or treated intraperitoneally on days −1, 0, 1, 2, 3, 4, 5, 6 and 7 with 250 mg/kg of metformin (Glucophage; Bristol) diluted in PBS, pH = 7.2, and challenged with one log-phase 5× 106 per mouse inoculum of C. albicans intravenously via the tail vein on day 0. Survival was monitored for 28 d for 10 mice per group and animals were sacrificed as 5 mice per group on days 3 and 7. During sacrifice, one midline abdominal incision was per-formed under aseptic conditions and tissue samples were taken from the liver, spleen and right kidney. Once weighed, segments of liver and right kidney were homogenized with 1 ml RPMI-1640 nutrient broth (Biochrom), and 100 µl of the homogenate was serially diluted four times 1:10 in 0.9% NaCl. A 0.1-ml aliquot of each dilution was plated onto Sabouraud agar (Becton Dickinson). Plates were incubated at 37 °C for 48 h, and the number of viable colonies in each dilution was multiplied by the appropriate dilution factor. Results are presented as log10 CFU/g. The lower detection limit was 10 CFU/g. The spleen was gently squeezed and then passed through a sterile filter (250 µm; 12- x13-cm; Alter) for splenocyte collection. After counting of splenocytes with trypan blue exclusion of dead cells, splenocytes were incubated into sterile flat wells at a density of 5 × 106 cells/ml in 1 ml of RPMI-1640 medium supplemented with 2 mM glutamine, 10% FBS, 100 U/ml of penicillin G, and 0.1 mg/ml of streptomycin in the absence or presence of 10 ng/ml LPS, E. coli O55:B5 (24 h), or 1 × 106 CFU/ml of heat-killed yeasts C. albicans (5 d) at 37 °C in 5% CO2. At the end of incubation, the plates were centrifuged and the supernatants were collected. Cytokines were measured in supernatants by an enzyme immunoassay (R&D Systems).
Human endotoxemia. Eight healthy men (age, 22.9 ± 0.5 years (mean ± s.e.m.)) were given intravenous administration of LPS (from E. coli; US standard refer-ence endotoxin; 4 ng/kg body weight; provided by the US National Institutes of Health). No power calculation was performed. Blood was collected before and 4 h after LPS administration in PAXgene tubes (Becton-Dickinson) for microarray analysis of total RNA isolated using the PAXgene blood RNA kit (Qiagen) according to manufacturer’s instructions. Blood was also collected in heparin tubes (Becton-Dickinson) for whole blood and PBMC restimula-tion with 100 ng/ml LPS for 3 h. After restimulation, cytokine release was assessed by Luminex-based multiplex assay. The study was approved by the institutional scientific and ethics committees of Academic Medical Center, Amsterdam (114009), and written informed consent was obtained from all subjects (clinicaltrials.gov: NCT01014117). All study procedures were con-ducted in accordance with the declaration of Helsinki including current revisions and Good Clinical Practice guidelines.
Patient studies. Clinical samples were obtained from patients with blood cul-ture proven E. coli or C. albicans sepsis who had been enrolled in the MARS
©20
16N
atu
re A
mer
ica,
Inc.
All
rig
hts
res
erve
d.
nature immunology doi:10.1038/ni.3398
(Molecular Diagnosis and Risk Stratification of Sepsis) cohort, a large pro-spective observational study in the intensive care units (ICUs) of two tertiary teaching hospitals (Academic Medical Center, Amsterdam and University Medical Center Utrecht, Utrecht) in the Netherlands that took place from 2011 to 2014 (ClinicalTrials.gov identifier NCT01905033). Blood was collected in PAXgene tubes and total RNA was isolated by means of the PAXgene blood RNA kits using the QIAcube system (Qiagen) according to the manufactur-er’s instructions. The Medical Ethical Committees of both study centers gave approval for an opt-out consent method (IRB number 10-056C). PAXgene blood was also obtained from 42 healthy controls (age range, 30–63 years; 24 males, 18 females) after written informed consent was provided. For ex vivo cytokine and lactate production of septic patients, blood was drawn via an arterial catheter from eight newly admitted septic patients within 24 h after admission to the intensive care unit of Radboud University Medical Center Nijmegen. Once a patient had recovered, blood was drawn before release from the ICU and if possible once more at a later time point. This was carried out in accordance with the applicable rules concerning the review of research ethics committees and informed consent. All patients or legal representatives were informed about the study details. For the IFN-γ pilot study, the primary end point data have been published elsewhere24. In short, eight patients with invasive infection with Candida species and/or Aspergillus fumigatus not only received the standard antifungal therapy but also were treated with recom-binant IFN-γ (100 mg thrice a week, for 2 weeks); at indicated time points, blood was drawn, and ex-vivo PBMC stimulations were performed24. From the supernatants of cells cultured for 24 h, lactate was measured. The Arnhem-Nijmegen medical ethical committee approved the study (NL28823.091.10 and Clinicaltrials.gov NCT01270490). All study procedures were conducted in accordance with the declaration of Helsinki, including current revisions and Good Clinical Practice guidelines.
Microarray analysis. Genome-wide gene-expression analysis of blood from the human endotoxemia cohort (4 ng E. coli LPS per kg body weight) was done using the Illumina HumanHT-12 V3.0 expression beadchip (GSE48119), whereas blood from the patients with sepsis from the MARS cohort were analyzed by hybridization to the human genome U219 96-array plate and scanned using the GeneTitan instrument (Affymetrix) (GSE65682). Genome-wide gene expression analysis of in vitro–stimulated PBMCs from healthy volunteers was done by using the Illumina HumanHT-12 V4.0 expression beadchip (GSE42606); further details are available10.
Statistics. Differences in cytokine, lactate and NAD+ were analyzed using a Wilcoxon signed-rank test or one way analysis of variance, where applicable. Survival curves were analyzed by a log-rank test, and for oxygen consumption, a Student’s t-test was used. All these analyses were performed in Graphpad prism 5. Dichotomous and continuous clinical data were assessed by Fisher’s exact test and Wilcoxon’s rank sum test, respectively. Microarray data are presented in the form of volcano plots (integrating log2 fold values and multiple-test adjusted probabilities) and heat map plots, generated in R studio44.
Study design and patient selection. The study was performed within the con-text of the MARS project in the mixed ICUs of two tertiary referral centers in the Netherlands (Academic Medical Center, Amsterdam, and University Medical Center Utrecht, Utrecht; clinicaltrials.gov identifier NCT01905033)45,46. The Medical Ethics committees of both participating centers approved an opt-out consent method (IRB no. 10-056C).
MARS patient cohort. The MARS ICU patient cohort was enrolled between January, 2011 and July, 2012, comprising 46 patients with sepsis with blood cul-ture–proven infection with Escherichia coli (E.coli) (n = 33) or Candida species (Candida albicans (n = 7), Candida glabrata (n = 5), Candida lusitaniae (n = 1) and Candida tropicalis (n = 1)) (n = 13). Severity was assessed on a daily basis from ICU admission by means of the Acute Physiology and Chronic Health Evaluation (APACHE IV) score47. Shock was defined as hypotension requiring the use of noradrenaline in a dose of >0.1 µg/kg/min during at least 50% of the day. Blood was collected in PAXgene blood tubes (Becton-Dickinson, Breda, the Netherlands) for RNA isolation. Total RNA was isolated by means of the
PAXgene blood mRNA kit (Qiagen) and Qiacube automated system (Qiagen). Blood was also obtained from 42 healthy controls (age range, 30–63 years; 24 males, 18 females) after providing written informed consent.
Radboud University Medical Center cohort. Eight newly admitted patients with sepsis (age range, 33–76 years, 3 males, 5 females) were included after informed consent was obtained. Shock was defined as described above. Blood was drawn via an arterial catheter within 24 h after admission to the intensive care unit of the Radboud University Medical Center Nijmegen. Blood cultures were positive for E. coli (n = 2) or C. albicans (n = 1), and in other patients, pathogens were cultured from urine or pleural effusions containing E. coli (n = 4) or C. glabrata (n = 1).
Microarray data processing and normalization. MARS cohort in vivo blood gene-expression data. Quality and integrity of RNA was assessed by Nanodrop spectrophotometry (260 nm:280 nm) and bioanalysis (Agilent). RNA (RIN > 6.0) was processed and hybridized to the Human Genome U219 96-array plate using the GeneTitanR instrument (Affymetrix) as described by the manu-facturer at the Cologne Center for Genomics, Cologne, Germany. Raw data scans (.CEL files) were read into the R language and environment for statistical computing (version 2.15.1; R Foundation for Statistical Computing, Vienna, Austria; http://www.R-project.org/). Pre-processing and quality control was performed by using the Affy package version 1.36.1 (ref. 48). Array data were normalized using the Robust Multi-array Average (RMA) method and summa-rized by medianpolish. The resultant 49,386 log-transformed probe intensities were filtered by means of a 0.5 variance cutoff using the genefilter method49 to recover 24,646 probes in at least one sample. The occurrence of non- experimental chip effects was assessed by means of the Surrogate Variable Analysis50 and corrected using the ComBat method51. Differential gene expression across healthy subjects and E. coli or Candida blood culture– positive patient samples was done by means of the limma method52,53. We adjusted for the potential effect of age in our model (additive). Throughout, Benjamini-Hochberg54 multiple-comparison–adjusted probabilities, correct-ing for 24,646 tests, defined significance. Probes were annotated using the U219 annotation database available through R/Bioconductor55.
Human endotoxemia in vivo blood gene-expression data. PAXgene blood total RNA was isolated by means of the Qiacube automated system (Qiagen) using PAXgene blood RNA kits according to manufacturer’s instructions. Synthesis, amplification and purification of anti-sense RNA was performed by using the Illumina TotalPrep RNA Amplification Kit (AM-IL1791; Ambion) following the Illumina Sentrix Array Matrix expression protocol at ServiceXS. A total of 750 ng biotinylated cRNA was hybridized onto the HumanHT-12v3 Expression BeadChip (Illumina). The raw scan data were read using the beadarray package (version 1.16.0)56, using the R statistical environment (version 2.11.0)24. Illumina’s default pre-processing steps were performed using beadarray. Estimated background was subtracted from the foreground for each bead. For replicate beads, outliers greater than 3 median absolute deviations from the median were removed and the average signal was calculated for the remaining intensities. Log-transformation was applied to summarized data in order to remove mean-variance relationship in intensities. Resulting data were then quantile-normalized57. Quality control was performed both on bead level and on bead summary data using the arrayQualityMetrics package (v2.6.0)58. For each of the 48,804 probes, a detection score was calculated by comparing their aver-age signals with the summarized values for the negative control probes. This step filtered the non-expressed probes, thus yielding 20,700 probes. Probes were re-annotated using the illuminaHumanv3BeadID.db package available from Bioconductor. Assessment of differential gene expression was done by means of the limma method52,53. Throughout, Benjamini-Hochberg54 multiple-comparison–adjusted probabilities, correcting for 20,700 tests, defined significance.
Human in vitro gene expression data. Genes encoding products involved in metabolic signaling pathways were selected and analyzed for differential expression in samples treated with RPMI medium (control) or stimulated with LPS or Candida- (4 and 24 h) as described before10 by using limma52,53.
©20
16N
atu
re A
mer
ica,
Inc.
All
rig
hts
res
erve
d.
nature immunologydoi:10.1038/ni.3398
A multiple-comparison Benjamini-Hochberg–adjusted P value of <0.05 defined significance.
Gene-expression pathway analysis. Ingenuity pathway analysis (http://www.ingenuity.com) was used to identify significant enrichment of canonical signaling pathways. Genes were stratified as over- or under-expressed and were analyzed by selection of the Ingenuity Gene Knowledge Base as reference and specification of human species. All other parameters were kept default. Association significance was measured by Fisher’s exact test Benjamini-Hochberg–adjusted P values. Transcription factor–binding site–motif anal-ysis was performed on all multiple-comparison significant genes (P < 0.05 (Benjamini-Hochberg)) per comparison using oPossum methodology59,60. The initial 2,000 base pairs of DNA sequence upstream and downstream of the gene transcription start site was used. All other parameters were default. Fisher’s scores above median and percent transcription factor–binding site hits of all input DNA sequences >50% were used as cutoffs for transcription factor pathway analysis using the ToppGene bioinformatics suite61,62. Significance was demarcated using Fisher’s exact test Benjamini-Hochberg–adjusted P values (P < 0.05).
49. Bourgon, R., Gentleman, R. & Huber, W. Independent filtering increases detection power for high-throughput experiments. Proc. Natl. Acad. Sci. USA 107, 9546–9551 (2010).
50. Leek, J.T. & Storey, J.D. Capturing heterogeneity in gene expression studies by surrogate variable analysis. PLoS Genet. 3, 1724–1735 (2007).
51. Johnson, W.E., Li, C. & Rabinovic, A. Adjusting batch effects in microarray expression data using empirical Bayes methods. Biostatistics 8, 118–127 (2007).
52. Smyth, G.K. Bioinformatics and Computational Biology Solutions Using R. (eds. Gentleman, R., Carey, V., Huber, W., Irizarry, R. & Dudoit, S.) 397–420 (Springer, 2005).
53. van Lieshout, M.H., Scicluna, B.P., Florquin, S. & van der Poll, T. NLRP3 and ASC differentially affect the lung transcriptome during pneumococcal pneumonia. Am. J. Respir. Cell Mol. Biol. 50, 699–712 (2014).
54. Benjamini, Y. & Hochberg, Y. Controlling the false discovery rate-a practical and powerful approach to multiple testing. J. R. Stat. Soc. B 57, 289–300 (1995).
55. Gentleman, R.C. et al. Bioconductor: open software development for computational biology and bioinformatics. Genome Biol. 5, R80 (2004).
56. Dunning, M.J., Smith, M.L., Ritchie, M.E. & Tavaré, S. Beadarray: R classes and methods for Illumina bead-based data. Bioinformatics 23, 2183–2184 (2007).
57. Bolstad, B.M., Irizarry, R.A., Astrand, M. & Speed, T.P. A comparison of normalization methods for high density oligonucleotide array data based on variance and bias. Bioinformatics 19, 185–193 (2003).
58. Kauffmann, A., Gentleman, R. & Huber, W. arrayQualityMetrics--a bioconductor package for quality assessment of microarray data. Bioinformatics 25, 415–416 (2009).
59. Ho Sui, S.J., Fulton, D.L., Arenillas, D.J., Kwon, A.T. & Wasserman, W.W. oPOSSUM: integrated tools for analysis of regulatory motif over-representation. Nucleic Acids Res. 35, W245–W252 (2007).
60. Ho Sui, S.J. et al. oPOSSUM: identification of over-represented transcription factor binding sites in co-expressed genes. Nucleic Acids Res. 33, 3154–3164 (2005).
61. Chen, J., Bardes, E.E., Aronow, B.J. & Jegga, A.G. ToppGene Suite for gene list enrichment analysis and candidate gene prioritization. Nucleic Acids Res. 37, W305–W311 (2009).
62. Hoeksema, M.A. et al. IFN-γ priming of macrophages represses a part of the inflammatory program and attenuates neutrophil recruitment. J. Immunol. 194, 3909–3916 (2015).
44. R core team. R: A Language and Environment for Statistical Computing (R Foundation for Statistical Computing, Vienna, Austria). http://www.R-project.org/ (2014).
45. Klein Klouwenberg, P.M. et al. Interobserver agreement of Centers for Disease Control and Prevention criteria for classifying infections in critically ill patients. Crit. Care Med. 41, 2373–2378 (2013).
46. Klein Klouwenberg, P.M. et al. MARS Consortium. Electronic implementation of a novel surveillance paradigm for ventilator-associated events. Feasibility and validation. Am. J. Respir. Crit. Care Med. 189, 947–955 (2014).
47. Knaus, W.A., Draper, E.A., Wagner, D.P. & Zimmerman, J.E. APACHE II: a severity of disease classification system. Crit. Care Med. 13, 818–829 (1985).
48. Gautier, L., Cope, L., Bolstad, B.M. & Irizarry, R.A. affy--analysis of Affymetrix GeneChip data at the probe level. Bioinformatics 20, 307–315 (2004).