LETTERS
Macrophage migration inhibitory factor stimulates
AMP-activated protein kinase in the ischaemic heart
Edward J. Miller1*, Ji Li1*{, Lin Leng2, Courtney McDonald2, Toshiya Atsumi5, Richard Bucala2,3*
& Lawrence H. Young1,4*
Understanding cellular response to environmental stress has
broad implications for human disease. AMP-activated protein
kinase (AMPK) orchestrates the regulation of energy-generating
and -consuming pathways, and protects the heart against isch-
aemic injury and apoptosis1. A role for circulating hormones such
as adiponectin2 and leptin3 in the activation of AMPK has received
recent attention. Whether local autocrine and paracrine factors
within target organs such as the heart modulate AMPK is
unknown. Here we show that macrophage migration inhibitory
factor (MIF), an upstream regulator of inflammation4, is released
in the ischaemic heart, where it stimulates AMPK activation
through CD74, promotes glucose uptake and protects the heart
during ischaemia-reperfusion injury. Germline deletion of theMif
gene impairs ischaemic AMPK signalling in the mouse heart.
Human fibroblasts with a low-activityMIF promoter polymorph-
ism5 have diminished MIF release and AMPK activation during
hypoxia. Thus, MIF modulates the activation of the cardioprotec-
tive AMPK pathway during ischaemia, functionally linking
inflammation and metabolism in the heart. We anticipate that
genetic variation in MIF expression may impact on the response
of the human heart to ischaemia by the AMPK pathway, and that
diagnostic MIF genotyping might predict risk in patients with
coronary artery disease.
Macrophage migration inhibitory factor (MIF) is a pleiotropic
cytokine that controls the inflammatory ‘set point’ by regulating
the release of other pro-inflammatory cytokines6. MIF is expressed
in several cell types, including monocytes/macrophages7, vascular
smooth muscle8 and cardiomyocytes9, and is released on stimulation
from pre-formed storage pools. MIF is involved in the pathogenesis
of inflammatory diseases, such as atherosclerosis8,10, rheumatoid
arthritis5, sepsis4, asthma11 and acute respiratory distress syndrome12.
HumanMIF gene expression is determined by promoter polymorph-
isms, including a tetra-nucleotide CATT repeat at position –794 (ref.
5). MIF signalling is known to activate ERK1/2 MAPK (ref. 13)
through a receptor complex comprising CD74 (ref. 14) and CD44
(ref. 15). In contrast, the chemokine receptors CXCR2 and CXCR4
participate in MIF-mediated migratory function10.
MIF also stimulates glycolysis during sepsis, increasing the syn-
thesis of fructose 2,6-bisphosphate and cellular glucose uptake16. The
signalling pathways by which MIF exerts its metabolic effects are
unknown, but one candidate is the AMP-activated protein kinase
(AMPK)—an important regulator of both glycolysis and glucose
uptake during cellular stress1. AMPK senses the cellular energy state
and affects diverse pathways to increase cellular ATP production
and limit energy consumption. AMPK activity is regulated by AMP
binding to its regulatory c-subunit17 and by phosphorylation of the
catalytic a-subunit by upstream kinases, including LKB1 (ref. 18) and
CaMKKb (ref. 19). In the heart, AMPK stimulates 6-phosphofructo-
2-kinase activity and glycolysis20, induces glucose transporter-4
(GLUT4, encoded by the SLC2A4 gene) translocation21, increases
ischaemic glucose uptake1,22 and limits myocardial injury and
apoptosis1.
AMPK phosphorylation is also modulated by the adipocyte-
derived circulating hormones leptin3 and adiponectin23, raising the
possibility that cytokines might also activate AMPK. We hypothe-
sized that AMPK might be activated in an autocrine/paracrine fash-
ion by MIF in the heart during ischaemia, linking the regulatory
control of inflammation and metabolism.
Initial experiments examined whetherMIF has a role in the stimu-
lation of the AMPK pathway during hypoxia in rat heart muscles.
Hypoxic activation of AMPK (Fig. 1a) was associated with a twofold
increase in muscle MIF release (Fig. 1b), the latter consistent with
previous results in cardiomyocytes24. Pre-treatment with anti-MIF
antibody reduced hypoxic AMPK activation by 67% (Fig. 1c). One of
the important AMPK actions during hypoxia and ischaemia is to
increase glucose transport1,22. Hypoxic glucose transport was inhi-
bited 38% by anti-MIF antibody (Fig. 1d), indicating that secreted
extracellular MIF modulates downstream AMPK action.
To investigate whether MIF modulates AMPK, we added MIF to
normoxic heart muscles. MIF caused time- and dose-dependent
increases in AMPK phosphorylation (Fig. 1e and f), and increased
heart muscle glucose uptake (Fig. 1g). Hypoxia and insulin-
stimulated glucose uptake in the heart are mediated by translocation
of the glucose transporter GLUT4 to the cell surface where it is
physiologically active21. We used a cell-membrane impermeant
photolabel compound and found significant translocation of
GLUT4 to the cell surface (Fig. 1h), elucidating the mechanism
through which MIF increases glucose uptake.
We next examined whether MIF modulates AMPK signalling in
the ischaemic heart.MIF is expressed by cardiomyocytes9,24, endothe-
lial cells, monocytes and macrophages7. We used the isolated mouse
heart perfused with crystalloid buffer, eliminating the potential
contribution of MIF from circulating cells. MIF was highly expressed
in cardiomyocytes, according to immunohistochemical data (Fig.
2a). Ischaemia triggered cardiac MIF release into the coronary
venous effluent and decreased heart MIF content after ischaemia-
reperfusion (Fig. 2b).
To determine whetherMIF plays a part in ischaemic AMPK activa-
tion, we used hearts fromMif2/2mice25 and compared them to wild-
type controls. Mif2/2 mice demonstrated a normal baseline cardiac
phenotype with respect to left ventricular size and function, histology
and the expression of AMPK and glucose transporter proteins
*These authors contributed equally to this work.
1Cardiovascular Medicine Section of the Department of Internal Medicine, 2Rheumatology Section of the Department of Internal Medicine, 3Department of Pathology, and
4Department of Cellular and Molecular Physiology, Yale University School of Medicine, New Haven, Connecticut 06520, USA. 5Department of Medicine II, Hokkaido University,
Sapporo, 060-8638, Japan. {Present address: University of Wyoming School of Pharmacy, Laramie, Wyoming 82072, USA.
Vol 451 | 31 January 2008 |doi:10.1038/nature06504
578
Nature Publishing Group©2008
(Supplementary Fig. 1). When perfused with mixed-substrate buffer
and subjected to 15min of global ischaemia, AMPK activation was
significantly blunted in the Mif2/2 hearts owing to decreased phos-
phorylation of the activating Thr 172 site (Fig. 3a). The tumour-
suppressor kinase LKB1 (also known as SKT11) has an important
role in mediating AMPK phosphorylation during ischaemia18.
However, we observed no change in the expression of LKB1, or
CaMKKb (also known as Camkk2), another potential upstream
kinase (Supplementary Fig. 1).
Because AMPK mediates glucose uptake during ischaemia1, we
examined whether the defect in AMPK signalling in the Mif2/2
hearts also diminished downstream glucose uptake. Although glu-
cose uptake was normal inMif2/2 hearts during control perfusions,
the stimulation of glucose uptake during ischaemia-reperfusion was
significantly blunted compared towild-type hearts (Fig. 3b). This was
associated with impaired glycogen synthesis inMif2/2 hearts during
post-ischaemic reperfusion, despite a comparable amount of gly-
cogen breakdown during ischaemia (Supplementary Fig. 2).
Consistent with prior observations that AMPK deficiency is func-
tionally deleterious to the heart during ischaemia-reperfusion1,
Mif2/2 hearts also demonstrated impaired ischaemic tolerance
(Fig. 3c).Mif2/2 hearts subjected to ex vivo ischaemia had decreased
post-ischaemic left ventricular function (Fig. 3c) as well as increased
ischaemic diastolic pressure and reduced contractility during reper-
fusion (Supplementary Fig. 3).Mif2/2 hearts subjected to in vivo left
coronary occlusion/reperfusion showed 2.3-fold greater infarct size
compared to wild-type controls (Fig. 3d). These results indicate
that MIF promotes early adaptive responses in the heart during
ischaemia-reperfusion.
Human gene mutations influence AMPK signalling and MIF
expression. Individuals with rare mutations in the AMPK c2 subunit
-pAMPK (Thr 172)
-AMPKα
-pAMPK
(Thr 172)
-AMPKα
Hypoxia
*
*
†
†
†
* * *
* *
-pAMPK (Thr 172)
-AMPKα
g
-AMPKα
* *
MIF (ng ml–1) MIF (400 ng ml–1)
* *
*
* *
MIF (ng ml–1)
0.0
2.5
5.0
7.5
10.0
α1 α2 α1
Control
Hypoxia
Control
0
1
2
Control Hypoxiaα2
Hypoxia
A
M
P
K
a
ct
iv
ity
(p
m
ol
m
in
–1
g
–1
)
M
IF
p
ro
d
uc
tio
n
(n
g
p
er
m
in
p
er
m
g
of
m
us
cl
e)
p
A
M
P
K
(r
el
at
iv
e
un
its
)
p
A
M
P
K
(r
el
at
iv
e
un
its
)
p
A
M
P
K
(r
el
at
iv
e
un
its
)
G
lu
co
se
u
p
ta
ke
(n
m
ol
g
–1
m
in
–1
)
G
lu
co
se
u
p
ta
ke
(n
m
ol
g
–1
m
in
–1
)
C
el
l-
su
rf
ac
e
G
LU
T4
(fo
ld
in
cr
ea
se
)
0
20
40
Control 100 200
100 200 400
MIF (400 ng ml–1)
30 60 120
-pAMPK (Thr 172)
(min)
0
1
2
Control 30 60400 (min)120
0
1
2
3
4
Anti-MIF-mAb– + –
IgG– – +
Hypoxia
Anti-MIF-mAb– + –
– – +
0
20
40
60
80
0
20
40
Control 200 400 800
IgG
h
-t-GLUT4
-s-GLUT4
Control MIF
0.0
0.5
1.0
1.5 *
Control MIF
(400 ng ml–1)
dcba
f
MIF (ng ml–1)
e
Figure 1 | Role of MIF in heart muscle AMPK signalling during hypoxia.
a, Immunoblots show phosphorylated and total AMPK, bars show a2 or a1
AMPK activity. *P5 0.001, versus control; {P5 0.012, a1 versus a2.
b, Muscle MIF release. *P5 0.03, versus control. c, d, Inhibition of AMPK
activation and downstream glucose transport by anti-MIF antibody
(100mgml21). *P5 0.02, versus control; {P5 0.04, versus hypoxia alone.
e, AMPK activation by recombinantMIF (60min). *P, 0.02, versus control.
f, AMPK activation by recombinant MIF (400ngml21). *P, 0.05, versus
control. g, Glucose uptake during incubation with recombinant MIF.
*P, 0.05, versus control. h, Immunoblots of cell-surface (s-GLUT4) and
total (t-GLUT4)GLUT4 inmuscles incubatedwithout orwith 400 ngml21 rat
MIF.*P, 0.001versus control,n5 3–6per group.Values aremeans6 s.e.m.
a
b
Anti-MIFIgG
-MIF
4
M
IF
r
el
ea
se
(n
g
p
er
m
in
p
er
g
o
f w
ild
-t
yp
e
he
ar
t)
M
IF
d
en
si
to
m
et
ry
(r
el
at
iv
e
un
its
)
3
2
1
0 0
1
-MIF
-AMPKα
WT
Baseline Reperfusion
Baseline Reperfusion Baseline Reperfusion
Mif–/–
Figure 2 | Heart MIF expression and release triggered by ischaemia.
a, Immunohistochemistry of wild-type (WT) mouse hearts with MIF
antibody or non-immune immunoglobulin G (IgG). Immunoblots of heart
lysates confirm the lack of immunoreactivity of the MIF antibody inMif2/2
hearts. Total AMPK is shown for loading comparison. b, Coronary effluent
MIF production from wild-type hearts during baseline normal perfusion or
during reperfusion after 10min of ischaemia. MIF concentration was
multiplied by the coronary flow to calculate the production rate. *P5 0.01,
versus baseline by unpaired t-test comparingmeans ofMIF concentration at
five baseline and five reperfusion time points. MIF immunoblots of heart
homogenates quantified by densitometry. *P5 0.003 versus control
perfusions, n5 2–3 hearts each. Values are means6 s.e.m.
NATURE |Vol 451 | 31 January 2008 LETTERS
579
Nature Publishing Group©2008
(PRKAG2, GeneID 51422) develop glycogen overload cardiomy-
opathy and Wolf–Parkinson–White syndrome26. A common poly-
morphism in the humanMIF promoter, containing 5, 6, 7 or 8 CATT
tetra-nucleotide repeat units (–794 CATT5–8), also has functional
consequences on MIF expression5. The CATT5 allele demonstrates
low MIF promoter activity compared to the others5 and has been
associated with less severe clinical manifestations of inflammatory
diseases such as asthma11, cystic fibrosis27 and rheumatoid arthritis5,
presumably owing to decreased MIF signalling. The MIF promoter
genotype varies in the population according to ethnicity, but the low
expression genotype is relatively commonwith 6%of Caucasians and
14.5% of African-Americans homozygous for the –794 CATT5
allele28. Despite demonstrable changes in MIF promoter activity,
there are few data demonstrating the influence of the low expression
genotype on the level of cellular MIF release.
Thus, we examined whether polymorphisms in the human MIF
promoter might lead to functional differences in MIF secretion and
cellular AMPK activation, using early passage human dermal fibro-
blasts. Cells from three of seven subjects were homozygous for the
low expression 2794 CATT5 allele (‘5/5’ genotype) and the remain-
der had at least one high expression 6-, 7- or 8-CATT repeat allele
(‘non-5/5’ genotype). The 5/5 cells had significantly less MIF release
into the culturemedia, during both normal and hypoxic incubations,
when compared to non-5/5 cells (Fig. 4a). ReducedMIF release from
the 5/5 cells was associated with less AMPK phosphorylation during
hypoxic stress (Fig. 4b). To determine whether the relative MIF defi-
ciency in the 5/5 cells was responsible for the impaired AMPK activa-
tion during hypoxic stress, MIF (10 ngml21) was added to the media
during hypoxic incubation. Exogenous MIF restored hypoxic AMPK
activation in the 5/5 cells to levels that were equivalent to the non-5/5
cells (Fig. 4b). In contrast, MIF did not augment AMPK activation in
hypoxic non-5/5 fibroblasts (Fig. 4b). Similarly, the addition of exo-
genous MIF to hypoxic rat heart muscles did not augment AMPK
activation (Supplementary Fig. 4). These data indicate that endogen-
ous MIF release maximally modulates AMPK phosphorylation dur-
ing hypoxia in normal heart tissue and cells. However, in relatively
MIF-deficient cells (that is the 5/5 MIF promoter genotype), which
have diminished MIF secretion during hypoxia, exogenous MIF
augmented AMPK activation. The results indicate that recombinant
MIF (or MIF agonists) might have a therapeutic effect by increasing
AMPK activation during ischaemia or hypoxia in selected individuals
with the low-expression 5/5 MIF promoter genotype. Thus, these
experiments demonstrate that a common polymorphism in the
MIF promoter leads to differential MIF release, which has conse-
quences in cellular stress signalling in human cells. They also imply
that exogenous MIF might have a beneficial effect in hypoxic tissues,
specifically in patients with the 5/5 genotype.
Taken together with the results implicating MIF in the activation
of AMPK in the ischaemic heart, these data raise the possibility that a
common polymorphism in theMIF promoter influences the suscep-
tibility of patients with coronary artery disease to ischaemic injury.
AMPK is under current investigation as a potential target molecule
for the treatment of type 2 diabetes, because of its metabolic actions
that increase skeletal muscle glucose uptake and suppress hepatic
glucose production. AMPK is also a potential target in ischaemic
heart disease, because of its cardioprotective effects1 and potential
role in ischaemic preconditioning29. Treatment with MIF or MIF
agonists warrants further study as an adjunctive therapy targeted at
a
0.0
0.1
0.2
0.3
0.4
0.5
0.6
G
lu
co
se
u
p
ta
ke
(µ
m
ol
m
in
–1
g
–1
)
14
4
3
2
1
0
p
A
M
P
K
(r
el
at
iv
e
un
its
)
A
M
P
K
α
2
ac
tiv
ity
(p
m
ol
m
in
–1
m
g–
1 )
12
10
8
6
4
2
0
WT
Mif–/–
WT
Mif–/–
WT
Mif–/–
WT
Mif–/–
WT
pAMPK (Thr 172)
AMPKα
Mif–/–
*†
*†
*†
†
†
b
c d
0 30 45 75
0
10,000
20,000
30,000
40,000
50,000
m
m
H
g
m
in
–1
*
Time (min)
Baseline Ischaemia Reperfusion
-Coronary flow4 ml min–1 4 ml min–1
0
10
20
30
In
fa
rc
t
si
ze
(p
er
c
en
t
of
r
is
k
ar
ea
)
*
Co
nt
ro
l
Isc
ha
em
ia
Co
nt
ro
l
Isc
ha
em
ia
Co
nt
ro
l
Co
nt
ro
l
Isc
ha
em
ia
Isc
ha
em
ia
Co
nt
ro
l
Co
nt
ro
l
Isc
ha
em
ia
Isc
ha
em
ia
Co
nt
ro
l
Co
nt
ro
l
Isc
ha
em
ia
Isc
ha
em
ia
Figure 3 | Genetic MIF deletion impairs ischaemic heart AMPK activation
and glucose uptake, and exacerbates post-ischaemic cardiac dysfunction
and injury. a, AMPKphosphorylation and activity after ischaemic or control
perfusions. *P, 0.05, versusMif2/2 ischaemic hearts; {P, 0.05, ischaemic
versus control, n5 3–4 hearts for each genotype. b, Glucose uptake during
control perfusion and during reperfusion after ischaemia (n5 5 for each
genotype). *P5 0.01, versus wild-type baseline, {P5 0.04, versus Mif2/2
reperfusion. c, Heart-rate–left-ventricular-developed pressure product
during control perfusion and post-ischaemic reperfusion. n5 6–7 hearts for
each genotype. *P5 0.03, by repeated measures ANOVA during
reperfusion. d, Myocardial infarction induced by 15min of left coronary
occlusion in vivo followed by 4 h of reperfusion. Viable myocardium stained
red with TTC; infarcted tissue, white; and normal non-ischaemic tissue,
blue. The infarct area was quantified and expressed as a per cent of the
ischaemic area at risk. *P5 0.04 versus wild type. n5 5–6 hearts per
genotype. Values are means6 s.e.m.
LETTERS NATURE |Vol 451 | 31 January 2008
580
Nature Publishing Group©2008
AMPK activation during acute myocardial ischaemia or infarction.
To the extent that MIF is released from the heart during ischaemic
preconditioning, MIF agonists might also augment preconditioning
by increasing AMPK activation during ischaemia. Therapy directed
at AMPKmight prove most effective in patients with low-expression
MIF promoter polymorphisms. These hypotheses deserve further
investigation and might also be addressed by analysis of gene banks
from large cardiovascular clinical trials.
To define the proximal mechanisms linking MIF and AMPK
activation better, we next examined whether components of MIF
cell-surface receptor complex, which is comprised of the ligand-
binding component, CD74 (ref. 14), and the signal-transducing
component, CD44 (ref. 15), is involved in AMPK activation during
hypoxia. Treatment of human fibroblasts with a CD74-specific short
interfering RNA (siRNA) decreased MIF receptor CD74 protein
expression and blunted hypoxia-stimulated AMPK phosphorylation
(Fig. 4c). We also studied MIF-induced AMPK phosphorylation
in CD74null/CD44null COS-7/M6 cells that were stably transfected
with either CD74 alone, CD44 alone, or CD74 together with CD44
(ref. 15). COS-7/M6 cells that expressedCD74 orCD44 alone showed
no AMPK response to either hypoxia or exogenously added MIF. In
contrast, COS-7/M6 cells expressing both transmembrane proteins
showed significant AMPK phosphorylation responses (Supplemen-
tary Fig. 5). These results support an important role for the two-
component receptor complex, consisting of the MIF binding CD74
protein and the signal-transducing CD44 protein, in MIF-mediated
AMPK signalling during cellular hypoxia in the heart. A CD74-
dependent interaction between MIF and CXCR2 also has been
reported and has a role in inflammatory cell recruitment10.
Whether MIF activation of CXCR2 also has a role in the cellular
response to hypoxic injury beyond its migratory function is worthy
of additional investigation.
In conclusion, these results define new models of both MIF action
and AMPK activation, establishing a link between pathways central
to inflammation and metabolism. MIF release leads to autocrine/
paracrine activation of the AMPK-signalling pathway in the isch-
aemic heart. In other inflammatory disease states, high levels of
MIF signalling, potentially activating additional pathways, might
be deleterious. A common polymorphism in the human MIF pro-
moter influences AMPK activation, andmight predispose susceptible
individuals to ischaemic injury and provide a potential new risk
marker for patients with coronary artery disease.
METHODS SU