Article
1 Drives Dynamics of
onses in
tDNA Deletions
Highlights
d Mitochondrial integrated stress response (ISRmt) progresses
in temporal stages
d ISRmt is conserved in human and mouse diseases with
mtDNA stress
ulfuration
Authors
Saara Forsstro¨m,
Christopher B. Jackson,
Christopher J. Carroll, ...,
Joni Nikkanen, Liliya Euro,
Anu Suomalainen
CorrespondenceForsstro¨m et al., 2019, Cell Metabolism 30, 1–15
December 3, 2019 ª 2019 Elsevier Inc.
https://doi.org/10.1016/j.cmet.2019.08.019d Endocrine FGF21 drives disease-related glucose uptake in
dorsal hippocampusd Autocrine FGF21 drives serine synthesis and transs
in muscleLocal and Systemic Stress Resp
Mitochondrial Myopathy with m
Graphical AbstractFibroblast Growth Factor 2anu.wartiovaara@helsinki.fi
In Brief
Forsstro¨m, Jackson et al. report
mechanisms of orchestrated, stage-wise
progression of mammalian integrated
mitochondrial stress response (ISRmt) in
mitochondrial muscle disease. They
identify acute and chronic ISRmt, with
crucial dependence on auto- and
endocrine actions of FGF21 hormone,
driving local serine biosynthesis,
transsulfuration, and signaling from
muscle to brain.
Cell Metabolism
Article
Fibroblast Growth Factor 21 Drives Dynamics
of Local and Systemic Stress Responses
in Mitochondrial Myopathy with mtDNA Deletions
Saara Forsstro¨m,1,11 Christopher B. Jackson,1,11 Christopher J. Carroll,1,2 Mervi Kuronen,1 Eija Pirinen,3
Swagat Pradhan,1 Anastasiia Marmyleva,1 Mari Auranen,4 Iida-Marja Kleine,1 Nahid A. Khan,1 Anne Roivainen,5,6
P€aivi Marjam€aki,5 Heidi Liljenb€ack,5,6 Liya Wang,7 Brendan J. Battersby,8 Uwe Richter,8 Vidya Velagapudi,9
Joni Nikkanen,1 Liliya Euro,1 and Anu Suomalainen1,4,10,12,*
1Stem Cells and Metabolism Research Program, Faculty of Medicine, University of Helsinki, 00290 Helsinki, Finland
2Molecular and Clinical Sciences Research Institute, St. George’s University of London, London SW170RE, UK
3Clinical and Molecular Metabolism Research Program, Faculty of Medicine, University of Helsinki, 00290 Helsinki, Finland
4Department of Neurosciences, Helsinki University Central Hospital, 00290 Helsinki, Finland
5Turku PET Centre, University of Turku, 20520 Turku, Finland
6Turku Center for Disease Modeling, Institute of Biomedicine, University of Turku, 20520 Turku, Finland
7Department of Anatomy, Physiology, and Biochemistry, Swedish University of Agricultural Sciences, 75007 Uppsala, Sweden
8Institute of Biotechnology, University of Helsinki, 00790 Helsinki, Finland
9Metabolomics Unit, Institute for Molecular Medicine Finland FIMM, HiLIFE, University of Helsinki, 00290 Helsinki, Finland
10Neuroscience Center, University of Helsinki, 00290 Helsinki, Finland
11These authors contributed equally
12Lead Contact
*Correspondence: anu.wartiovaara@helsinki.fi
https://doi.org/10.1016/j.cmet.2019.08.019
SUMMARY
Mitochondrial dysfunction elicits stress responses
that safeguard cellular homeostasis against meta-
bolic insults. Mitochondrial integrated stress
response (ISRmt) is a major response to mitochon-
drial (mt)DNA expression stress (mtDNA mainte-
nance, translation defects), but the knowledge of
dynamics or interdependence of components is
lacking. We report that in mitochondrial myopathy,
ISRmt progresses in temporal stages and develop-
ment from early to chronic and is regulated by auto-
crine and endocrine effects of FGF21, a metabolic
hormonewith pleiotropic effects. Initial disease signs
induce transcriptional ISRmt (ATF5, mitochondrial
one-carbon cycle, FGF21, and GDF15). The local
progression to 2nd metabolic ISRmt stage (ATF3,
ATF4, glucose uptake, serine biosynthesis, and
transsulfuration) is FGF21 dependent. Mitochondrial
unfolded protein response marks the 3rd ISRmt stage
of failing tissue. Systemically, FGF21 drives weight
loss and glucose preference, and modifies meta-
bolism and respiratory chain deficiency in a specific
hippocampal brain region. Our evidence indicates
that FGF21 is a local and systemic messenger of
mtDNA stress in mice and humans with mitochon-
drial disease.
INTRODUCTION
Mitochondrial diseases show an unprecedented clinical vari-
ability, but the molecular mechanisms underlying their tissue-
specific manifestations are poorly understood (Gorman et al.,
2016; Nunnari and Suomalainen, 2012). Recent evidence
suggests that metabolic responses elicited by different mito-
chondrial insults are specific for tissue type and dysfunction
(Suomalainen and Battersby, 2018), raising an exciting question
of whether the responses contribute to tissue-specific manifes-
tations of the disease.
Context and Significance
Mitochondrial disorders happen when the cell’s powerplants (mitochondria) fail to produce enough energy for the body to
function properly and exhibit a diverse array of clinical manifestations. Progressivemuscle weakness and dysfunction in the
respiration apparatus underlie mitochondrial myopathy (MM). The group of Anu Suomalainen from Helsinki University
shows that the metabolic regulator Fibroblast growth factor 21 (FGF21) coordinates step-wise changes within the muscle
cell and in other tissues of the body in the mouse and in humans to progress MM, including changes in glucose and lipid
metabolism, weight loss, and brain defects. The work highlights how amuscle disease can affect brain metabolism and em-
phasizes the importance of considering the whole body holistically when assessing diseases and treatments.
Cell Metabolism 30, 1–15, December 3, 2019 ª 2019 Elsevier Inc. 1
Please cite this article in press as: Forsstro¨m et al., Fibroblast Growth Factor 21 Drives Dynamics of Local and Systemic Stress Responses in Mito-
chondrial Myopathy with mtDNA Deletions, Cell Metabolism (2019), https://doi.org/10.1016/j.cmet.2019.08.019
The first mitochondrial stress response, mitochondrial
unfolded protein response (UPRmt), was characterized in
mammalian cultured cells and in the worm C. elegans (Durieux
et al., 2011; Haynes et al., 2007, 2013; Merkwirth et al., 2016).
Proteotoxic stress activates UPRmt through ATFS-1 (stress-
activated transcription factor S1) that relocates from mitochon-
dria to nucleus, binds to promoters with amino acid response
elements (AAREs), and induces transcription of mitochondrial
heat shock proteins (HSPs), chaperones, and proteases (Hay-
nes et al., 2013; Nargund et al., 2012). In the worm, neuronal
inactivation of mitochondrial cytochrome c oxidase elicited
UPRmt locally, but also in the gut, pointing to a secreted factor
mediating stress signals from the affected neurons to periphery
(Durieux et al., 2011). The worm ‘‘mitokine’’ was recently
reported to be a Wnt-ligand/EGL-20 (Zhang et al., 2018). How-
ever, in mice with mitochondrial dysfunction, UPRmt/HSP in-
duction has been mild if present (Pulliam et al., 2014; Seiferling
et al., 2016), leaving the role of UPRmt in primary mitochondrial
diseases open.
MM manifests as progressive muscle weakness and respira-
tory chain (RC) dysfunction, typically caused by mtDNA
deletions or mtDNA expression defects (mtDNA maintenance,
translation; MM denoting diseases associated with these de-
fects from now on). MM mice present a robust transcriptional
stress response with induction of AARE-regulated genes and
major remodeling of whole-cellular anabolic pathways, including
serine, glutathione, and purine biosynthesis (Bao et al., 2016;
K€uhl et al., 2017; Nikkanen et al., 2016). This mitochondrial inte-
grated stress response (ISRmt), with metabolic and transcrip-
tional components, is regulated by mTORC1 (mechanistic target
of rapamycin C1), a major signal and growth integrator kinase
(Khan et al., 2017; Zhang et al., 2013). In mammals, ISRmt is a
robust response to mtDNA stress, integrating mitochondrial
and cytoplasmic metabolism to pathways of growth and repair.
One of the most upregulated ISRmt components in humans
and mice with MM is fibroblast growth factor 21 (FGF21).
FGF21 hasmanifold roles in regulation of lipid and glucosemeta-
bolism, getting activated by low blood lipids and glucose;
inducing ketone body synthesis, lipolysis, and browning in white
adipose tissue (WAT); improving glucose tolerance; and regu-
lating food intake (Badman et al., 2007; Inagaki et al., 2007; Khar-
itonenkov and Shanafelt, 2009; Kharitonenkov et al., 2005;
Potthoff, 2017). So far, however, MMdiseases are the only group
of inherited diseases that consistently induces FGF21 in a geno-
type- and phenotype-specific manner. In mice and humans with
MM, FGF21 is chronically secreted from the affected muscle
fibers and/or heart independent of the feeding state
(Agnew et al., 2018; Dogan et al., 2014; Keipert et al., 2014;
Lehtonen et al., 2016; Ost et al., 2016; Pereira et al., 2017; Suo-
malainen et al., 2011; Tyynismaa et al., 2010). Interestingly, how-
ever, the response is specific to mtDNA expression diseases,
whereas structural defects of the RC enzymes do not elicit a
similar response (Lehtonen et al., 2016). These data indicate
that different types of mitochondrial dysfunction induce different
stress responses.
Here, we report that mammalian ISRmt progresses in temporal
stages, and its components vary in different tissue types and dis-
ease stages, in post-mitotic or proliferating cell types. We show
that FGF21 orchestrates progression of ISRmt and MM into
systemic disease, with major effects for glucose and lipid meta-
bolism, weight loss, and brain manifestation of MM. Our data
indicate that mitochondrial stress responses are (1) only partially
conserved in species and (2) not generalizable between organ-
isms or even mammalian cell types, and (3) that FGF21 is a key
mediator of metabolic remodeling and progression of ISRmt
locally and systemically.
RESULTS
ISRmt Progresses in Temporal Stages, Involves Different
Activating Transcription Factors (ATFs) at Different
Disease Stages, and Is Partly Distinct from UPRmt
and UPREr
The various arms of ISRmt (UPRmt, ATF/AARE, endocrine factors,
and metabolic remodeling) prompted us to ask whether the
different components are induced in all-or-none, parallel, or
sequential manner in MM mice, in vivo. We utilized the Deletor
mice that show robust ISRmt (Khan et al., 2017; Nikkanen
et al., 2016; Tyynismaa et al., 2010) despite mild MM findings
and normal lifespan (Tyynismaa et al., 2005). The Deletors accu-
mulate mtDNA deletions and show progressive RC deficiency in
skeletal muscle and heart, as a consequence of ubiquitous
expression of a dominant patient mutation in Twinkle, nuclear-
encoded replicative helicase of mtDNA. These mice offer an
excellent opportunity to dissect the long-term stress response
dynamics upon MM progression. The first ISRmt signs in Deletor
muscle coincide with the initial histological signs of RC defi-
ciency (mosaic partial or complete cytochrome c oxidase
[COX] deficiency) in <12-month-old mice and involve highly acti-
vated transcription of Fgf21 and Mthfd2 as well as Gdf15 (Fig-
ures 1A and 1B). At 14–16 months, some COX-deficient fibers
start to show increased mitochondrial mass and activity of
succinate dehydrogenase (SDH, RC complex II, exclusively nu-
clear-encoded and, therefore, not affected by mtDNA deletions)
(Figure 1B). At this time, the second phase of ISRmt involves acti-
vation of Atf5 and the regulatory enzymes of de novo serine
biosynthesis (phosphoglycerate dehydrogenase [Phgdh] and
phosphoserine aminotransferase 1 [Psat1]) (Figure 1A), as well
as mTORC1 (elevated phosphorylation of ribosomal S6 protein)
(Figure 1B). These results indicate that ISRmt induction pro-
gresses in orderly temporal stages.
We then asked whether UPRmt is part of ISRmt in mammals. In
Deletors of >21 months of age, mitochondrial HSP70 transcripts
increase slightly, while mitochondrial HSP10 and HSP60 remain
unchanged (Figure 1A). In >24-month-old Deletor muscle LonP1
(Lon peptidase 1), HSP60 and HSP10 protein amounts are
increased in mitochondrial extracts, while HSP70 and CLPP (ca-
seinolytic mitochondrial matrix peptidase) are wild-type (WT)-
like (Figures 1C and 1D). At this late stage, Atf3 gets induced
(Figure 1A). These data suggest that the affected skeletal muscle
manifests a sequential ATF response, Atf5 being progressively
induced from the 1st to 2nd ISRmt stage and leading the response,
and Atf3 participating in terminal stage along with UPRmt.
Despite non-significant marginal changes in Atf4 transcripts,
we cannot exclude its post-translational induction in the muscle.
In conclusion, ISRmt in mammalian muscle proceeds in three
stages that follow disease severity with initial activation of
MTHFD2, FGF21, and GDF15; secondary induction of mTORC1,
2 Cell Metabolism 30, 1–15, December 3, 2019
Please cite this article in press as: Forsstro¨m et al., Fibroblast Growth Factor 21 Drives Dynamics of Local and Systemic Stress Responses in Mito-
chondrial Myopathy with mtDNA Deletions, Cell Metabolism (2019), https://doi.org/10.1016/j.cmet.2019.08.019
A B
C D
E
F
G
(legend on next page)
Cell Metabolism 30, 1–15, December 3, 2019 3
Please cite this article in press as: Forsstro¨m et al., Fibroblast Growth Factor 21 Drives Dynamics of Local and Systemic Stress Responses in Mito-
chondrial Myopathy with mtDNA Deletions, Cell Metabolism (2019), https://doi.org/10.1016/j.cmet.2019.08.019
transsulfuration, and serine biogenesis; followed by terminal-
stage mild UPRmt (Figure 1E).
Optic atrophy 1 (OPA1)-dependent membrane remodeling is
commonly described as part of mitochondrial stress responses
in cell culture and mitochondrial protein knockout mice
(MacVicar and Langer, 2016; Olichon et al., 2003; Richter et al.,
2013, 2015, 2019). In 24-month-old Deletors with advanced
MM, the long isoform of OPA1 remains unprocessed (Figure 1C),
showing that OPA1 processing is not an integral part of MM-
related ISRmt. ISRmt is also distinct from endoplasmic reticulum
(ER) stress (Gardner et al., 2013; Sidrauski andWalter, 1997; Tay-
lor and Dillin, 2013), based on WT-like binding-immunoglobulin
protein (BiP), X-box binding protein 1 (XBP1) splicing, and Atf6
expression in the aged Deletor muscle (Figures S1A–S1C).
We then asked whether ISRmt induction in the muscle sends
stress signals to non-affected tissues, similar to UPRmt in
C. elegans (Durieux et al., 2011), but found no evidence for
such an event. The key ISRmt components are upregulated
only in the primarily affected tissues with mtDNA deletions (mus-
cle, heart, and BAT; Figures 1A and 1F), but not, for example, in
the liver, a key metabolic organ. Therefore, unlike in the worm,
ISRmt does not spread per se in mammalian tissues via a
secreted factor.
ISRmt Is Conserved in Human Patients with mtDNA
Deletions
FGF21 shows high concentrations in the serum and muscle of
MMpatients, compared to othermitochondrial or non-mitochon-
drial muscle diseases (Lehtonen et al., 2016; Suomalainen et al.,
2011). To examine which parts of ISRmt are conserved frommice
tomen,weanalyzedmuscle biopsy samples of patientswithMM,
caused either by a single heteroplasmic mtDNA deletion or mul-
tiple mtDNA deletions, the latter caused by a dominant Twinkle
mutation, homologous to that inDeletormice (Table 1; clinical de-
tails in Spelbrink et al., 2001; Suomalainen et al., 1992, 1997). The
full ISRmt found in Deletor mice is also activated in human pa-
tients, including MTHFD2, FGF21, GDF15, PHGDH, PSAT1,
and SHMT2 (mitochondrial serine hydroxymethyl-transferase-2)
as well as ATF-regulated ASNS (asparagine synthetase) and
TRIB3 (tribbles pseudokinase 3) (Figure 1G) and UPRmt-linked
transcription factor DDIT3 (DNA-damage-inducible transcript
3). However, all mitochondrial HSPs or proteases remain unin-
duced (Figure 1G). These results indicate robust conservation
of the full ISRmt in mammals, from mice to human patients.
ISRmt Progresses in Temporal Stages in Proliferating
Cells after Inhibition of Mitochondrial Translation
To ask whether the temporal induction of ISRmt is induced in
proliferating cells similar to muscle, we reversibly inhibited
mitochondrial (mt)-translation in mammalian diploid myoblasts.
Multiple mtDNA deletions do not accumulate in proliferating
cells, inhibiting us to study exactly the same mtDNA insult in
cultured cells as in Deletor muscle. However, FGF21 and
GDF15 are induced in mtDNA translation diseases (Lehtonen
et al., 2016), and we chose to mimick ISRmt of muscle by inhib-
iting mt-translation using actinonin. This drug leads to acute
reversible downregulation of mt-translation and an ISRmt-like
response (Richter et al., 2013, 2015). Figure S1 We show that
ISRmt components are robustly induced in mouse myoblasts,
following the decrease of mt-translation from the early time
points (1.5–2.5 h of actinonin exposure) (Figure 2A). This
response is led by >40-fold induction of Fgf21 transcription,
accompanied byGdf15,Mthfd2, and cystathionine gamma lyase
(CTH; both mRNA and protein), a regulatory enzyme of
Figure 1. ISRmt Progresses in Defined Temporal Stages in Muscle and Is Conserved in Mouse and Man
(A) Chronological progression of ISRmt in MM muscle. 1st stage: metabolic hormones (Fgf21 and Gdf15) and mitochondrial folate cycle (Mthfd2). 2nd stage: de
novo serine biosynthesis (Phgdh and Psat1). 3rd stage: mitochondrial HSPs. ATFs 3–5 activate at different time points. FC: mRNA against beta-actin mRNA. For
details and statistical analysis, see Table S1.
(B) Progression of RC enzyme deficiency andmTORC1 activation inMMmuscle. Top: histochemical in situ enzyme activity of COX (brown) and SDH (blue). White
fibers, COX-negative; blue fibers, COX-negative, SDH positive. Bottom: phosphorylated ribosomal S6 protein (pS6 immunohistochemistry). Subsequent frozen
sections; lines and arrows indicate same fibers. Scale bar, 50 mm.
(C and D) UPRmt-associated HSPs and proteases in skeletal muscle mitochondrial extracts; western blot. Porin and TOM20: loading controls, quantification of
signals against porin in (D), n = 5 per group.
(E) Schematic representation of the temporal ISRmt stages of MM.
(F) ISRmt or UPRmt transcript induction in different tissues compared to beta-actin mRNA, n = 6 per group.
(G) ISRmt induction in human patients with MM. Tables 1 and S2 summarize subjects and details of transcript levels results, respectively.
Animal ages: 22–24 months. Bars: average and SD. Box: 25th and 75th percentiles with median and range. Dots: individual subjects. Pairwise two-tailed t test,
statistical significance: *p% 0.05, **p% 0.01, ***p% 0.001, ****p% 0.0001. ASNS, asparagine synthetase; ATF3–5, activating transcription factors; BAT, brown
adipose tissue; CLPP, caseinolytic mitochondrial matrix peptidase; COX, cytochrome c oxidase; DDIT3, DNA damage inducible transcript-3; FC, fold change;
FGF21, fibroblast growth factor 21; GDF15, growth and differentiation factor 15; HSP10/60/70, mitochondrial heat shock proteins 10/60/70; LONP1, lon
peptidase 1; MTHFD2, methylene-tetrahydrofolate-dehydrogenase-2; mTORC1, mechanistic target of rapamycin C1; OPA1, optic atrophy 1; PHGDH, phos-
phoglycerate dehydrogenase; PSAT1, phosphoserine aminotransferase-1; SHMT1(2), serine hydroxymethyltransferase 1 (2, mitochondrial); SDH, succinate
dehydrogenase; TOM20, translocase of outer mitochondrial membrane 20; TRIB3, tribbles pseudokinase 3; WATip/sc, white adipose tissue intraperitoneal or
subcutaneous.
Table 1. Clinical Characteristics of MM Patients
Sex
Age at
Sampling
Age of
Disease
Onset Disease Cause
Patient 1 F 57 33 single mtDNA
deletion
Patient 2 M 55 21 multiple mtDNA
deletionsa
Patient 3 F 54 30 multiple mtDNA
deletionsa
Patient 4 F 70 31 single small mtDNA
deletion
Controls n = 8 6 F/2 M 48–64 – –
aTwinkle mutation leading to 13-amino acid insertion in linker domain =
homologous mutation to that in Deletor mouse.
4 Cell Metabolism 30, 1–15, December 3, 2019
Please cite this article in press as: Forsstro¨m et al., Fibroblast Growth Factor 21 Drives Dynamics of Local and Systemic Stress Responses in Mito-
chondrial Myopathy with mtDNA Deletions, Cell Metabolism (2019), https://doi.org/10.1016/j.cmet.2019.08.019
0 1.5 2.5 4.5 9 16
FC
2
4
50
8
6
0 hours
25
B
Fgf21 
Gdf15
Mthfd2
Cth
Phgdh
Psat1
ISRmt
Ac
t +
 
AT
F4
si
R
N
A
F ATF4 relocalization
C
on
tro
l
Ac
t
ATF4/TOM20/DAPI ATF4
8
6
0
2
4
FC
ATFsE
Atf1
Atf2
Atf3
Atf4
Atf5
Atf6
Atf7
0 1.5 2.5 4.5 9 16
hours0
2
4
6
8
Hsp60
Hsp10
Hsp70
FC
D UPRmt
0 1.5 2.5 4.5 9 16
hours
G ATF 3-5 inhibition 
ATF3 siRNA
ATF4 siRNA
ATF5 siRNA
Ctrl siRNA
10µm
FGF21
0
2
4
8
6
MTHFD2
0 10 20
1
2
4
3
FC
GDF15
10
20
40
30
0
0 10 20 0 10 20
2
6
8
CTH
4
10
PHGDH
1
3
5
4
2
0 10 20 0 10 20 hou
rs
1
2
4
3
ATF4
0
1
2
3
ATF5
0
2
4
8
6
0
ATF3
FC
0 10 20 hours0 10 200 10 20
m
R
N
A
m
R
N
A
C
100
75
hours
Act (h)
C
CTH
OPA1
Porin
Et
O
H
0
KDa
45
37
1.5 2.5 4.5 9 16 16
H ISRmt in cultured cells
A Mitochondrial translation inhibition
000
Subunits
Total
protein
Ac
t (h
)
0 1.5 2.5 4.5 9 16
35S-Met/Cys incorporation
Mt-Nd5
Mt-Co1
Mt-Cyb
Mt-Atp6
Mt-Nd6
Mt-Nd3
Mt-Nd4l
Mt-Atp8
Mt-Nd2
Mt-Nd1
Mt-Co2/3
160 1.5 2.5 4.5 9
1.0
0.8
0.6
0.4
0.2
0.0
35S-Met/Cys signal 
vs. total protein
hours
FC
Figure 2. Temporal Progression of ISRmt in Proliferating Cells after Acutely Inhibited Mitochondrial Translation
(A) mt-translation inhibition upon actinonin exposure; 35S-methionine/cysteine incorporation assay and quantification of total signal versus total protein.
mt-Nd1–6, NADH dehydrogenase subunits 1–6 (complex I); mt-Cytb, cytochrome b (complex III); mt-Co1–3, cytochrome c oxidase subunits 1–3 (complex IV);
mt-Atp6, mt-Atp8, ATP synthase subunits 6 and 8 (complex V).
(B) Temporal transcriptional activation of ISRmt components upon mt-translation inhibition. Actinonin exposure in mouse myoblasts, mRNA quantified against
beta-actin mRNA.
(C) OPA1 cleavage and progressive CTH induction upon actinonin exposure.Western blot analysis of protein extracts frommousemyoblasts, mitochondrial porin
as a loading control.
(D) Temporal activation of mitochondrial HSP expression upon mt-translation inhibition; actinonin exposure, mouse myoblasts.
(E) Temporal activation of ATF3, 4, and 5 expression upon mt-translation inhibition; actinonin exposure, mouse myoblasts.
(F) Nuclear localization of ATF4 upon mt-translation inhibition. Immunofluorescence analysis of untreated (top), actinonin-treated (middle), and actinonin+ATF4
siRNA-treated (bottom) human fibroblasts; scale bar, 5 mm.
(G) ISRmt induction upon mt-translation inhibition; actinonin exposure after siRNA-mediated silencing of ATF3, 4, or 5 in human fibroblasts.
(H) Schematic representation of ISRmt progression in growing cells, upon mt-translation inhibition.
Gene expression analyses in (B), (C), (E), and (G) represent consensus of a minimum of three separate actinonin treatment experiments. Transcripts presented in
(B), (C), and (E) are measured from the same cells. Abbreviations (also see legend of Figure 1): Act, actinonin; CTH, cystathionine gamma-lyase;
DAPI, 40,6-diamidino-2-phenylindole (DNA dye); EtOH, ethanol.
Cell Metabolism 30, 1–15, December 3, 2019 5
Please cite this article in press as: Forsstro¨m et al., Fibroblast Growth Factor 21 Drives Dynamics of Local and Systemic Stress Responses in Mito-
chondrial Myopathy with mtDNA Deletions, Cell Metabolism (2019), https://doi.org/10.1016/j.cmet.2019.08.019
Long-range PCR
M
3 kb
10 kb
500 bp
700 bp
DELWT DEL-FKOFKO
Δ
m
tD
N
A
mtDNA deletions 
K
6
8
10
%
 
o
f f
ib
er
s
COX- COX-/SDH+L
2
4
WT DE
L
FK
O
DE
L-F
KOWT DE
L
FK
O
DE
L-F
KO
0
Respiratory chain activity in skeletal muscle
Light
RER 
(thermoneutral)
Dark
0.7
0.9
1.0
v
C
O
2/v
O
2
1.1
0.8
WT
FKO DEL-FKO
DEL
I
WT DE
L
FK
O
DE
L-F
KO
RER dark 
(thermoneutral)
0.7
0.8
1.0
v
C
O
2/v
O
2
J
0.9
1.1
Whole body metabolism
0
20
40
60
FC
****
ND ND
A
80
0
500
1,000
1,500
pg
/m
l
***
ND ND
WT DE
L
FK
O
DE
L-F
KO
B
2,000
DE
L
DE
L-F
KO
0
0.5
1.0
1.5
ND NDND ND
Heart Liver Brain
Muscle
C FC
DE
L
DE
L-F
KODE
L
DE
L-F
KODE
L
DE
L-F
KO
2.0
ND
Fgf21 Fgf21S-FGF21
WT DE
L
FK
O
DE
L-F
KO
FGF21 expression in mouse lines
E WT FKO DEL
H
E
F
O
il-
R
ed
-
O
DEL-FKO
Adiposity in mouse lines
N
D
4 
vs
 
D
-
Lo
op
 
0
0.5
Intact mtDNA
WT DE
L
FK
O
DE
L-F
KO
1.5 ***
****
1.0
N
FC
18F-FDG
M
u
sc
le
Br
a
in
H
ea
rt
Li
v
e
r
Se
ru
m
BA
T
W
AT
ip
W
AT
sc
Pa
nc
re
a
s
Sp
le
e
n
Lu
ng
Ki
dn
ey
0
1
2
3
SU
V 
(F
C
)
*
*
* * *
*
*
*
*
*
*
* *
18F-FDG
G
0
1
2
3
SU
V 
(F
C
)
*
*
**
0
1
2
3
SU
V 
(F
C
)
18F-FTHA
H
1
2
3
SU
V 
(F
C
)
18F-FTHA
M
us
cl
e
Br
a
in
H
ea
rt
Li
v
e
r
Se
ru
m
BA
T
W
AT
ip
W
AT
sc
Pa
nc
re
a
s
Sp
le
e
n
Lu
ng
Ki
dn
ey
M
us
cl
e
Br
a
in
H
ea
rt
Li
v
e
r
Se
ru
m
BA
T
W
AT
ip
W
AT
sc
Pa
nc
re
a
s
Sp
le
e
n
Lu
ng
Ki
dn
ey
M
us
cl
e
Br
a
in
H
ea
rt
Li
v
e
r
Se
ru
m
BA
T
W
AT
ip
W
AT
sc
Pa
nc
re
a
s
Sp
le
e
n
Lu
ng
Ki
dn
ey
In vivo glucose and fatty acid uptake
0
D
-10%
+10%
0
+20%
+30%
-20%
11 14 16 18 20 22 24 Months
WT
DEL
FKO
DEL-FKO* * *
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
* *
Body weight
DEL vs WT DEL-FKO vs FKO
DEL vs WT DEL-FKO vs FKO
Figure 3. Muscle-Derived FGF21 Inhibits Fat Storage and Induces Glucose Uptake without Affecting Pathological Hallmarks
(A–C) FGF21 expression in Deletor and FGF21 knockout lines; mRNA FC against beta-actin.
(A) Fgf21 in skeletal muscle, n = 5 per group.
(B) Serum FGF21; WT, n = 7; Deletor, n = 6; FKO and DEL-FKO, n = 3.
(C) Fgf21 expression in different tissues, compared to Deletor muscle, n = 3 per group.
(D) Body weight of mice at the age of Deletor phenotype manifestation (11–24 months). WT, FKO, n = 8; Deletor and DEL-FKO, n = 7. Significance calculated
against WT mice.
(legend continued on next page)
6 Cell Metabolism 30, 1–15, December 3, 2019
Please cite this article in press as: Forsstro¨m et al., Fibroblast Growth Factor 21 Drives Dynamics of Local and Systemic Stress Responses in Mito-
chondrial Myopathy with mtDNA Deletions, Cell Metabolism (2019), https://doi.org/10.1016/j.cmet.2019.08.019
transsulfuration pathway (Figures 2B and 2C). Extended transla-
tional stress after time point 4.5 h, the 2nd ISRmt stage, upregu-
lates de novo serine biosynthesis enzymes, together with
mitochondrial HSPs (Figures 2B and 2D). These data indicate
that the acute response of mammalian diploid proliferating cells
to mt-translation stress is orderly and sequential, similar to the
post-mitotic tissues. However, different from MM muscle, mt-
translation defect ISRmt involves activation of OPA1 cleavage
(Figure 2C) and induction of HSPs, indicating cell-type specificity
in responses to mtDNA expression defects.
ATF4 Is Required for ISRmt in Dividing Cells
We then examined the contribution of the different ATFs in
mouse myoblast ISRmt. Out of ATFs 1–7, the expression of
ATF3, ATF4, and ATF5 gets progressively induced in mouse
myoblasts upon decreasing mt-translation, whereas other
ATFs remain uninduced (Figure 2E). Mt-translation inhibition in-
creases nuclear localization of ATF4 (Figure 2F) as a sign of acti-
vation of the transcription factor. To clarify the hierarchy of the
ATFs, we silenced ATF3, ATF4, and ATF5 one by one. We
show that ATF4 is an upstream of regulator of CTH, PHGDH,
MTHFD2,GDF15, and FGF21, as well as ATF3 (Figure 2G). How-
ever, ATF5 activation is independent of ATF4, which is consis-
tent considering the early and progressive induction of ATF5
also in the skeletal muscle, without considerable activation of
ATF4 at the time (Figure 1A). Also, mitochondrial fragmentation
is not affected by ATF4 silencing (Figure 2F), suggesting its regu-
lation by an ISRmt-independent mechanism. ATF3 or ATF5
silencing indicates that they are not essential for ISRmt induction
in proliferating cells. The combined evidence from muscle and
proliferating cells suggests that ATF5, ATF4, and ATF3 all take
part in ISRmt, with cell-type and stress specificity, fine-tuning
metabolism in acute and chronic mtDNA expression stress.
FGF21Modulates Local andSystemicMetabolism inMM
Our findings above show that FGF21 is an early responder to
mtDNA replication and translation stress both in post-mitotic tis-
sues and proliferating cells. To assess FGF21 contributions to
disease progression and ISRmt, we generated Deletor mice lack-
ing FGF21. In short, Deletors with floxed Fgf21 (FGF21LoxP/LoxP)
(Potthoff et al., 2009) were crossed with mice constitutively ex-
pressing cre-transgene (PGK-promoter) (Lallemand et al.,
1998) to generate full-body FGF21 knockout Deletors (DEL-
FKO) (Figure S3A). DEL-FKO mice are unable to induce Fgf21
expression in the affectedmuscle or heart or from the liver during
fasting (Figures 3A–3C and S3B), but their viability is WT-like in
24 months of follow-up.
Body weight of Deletors declines progressively from 16 to
18 months of age (Figure 3D), coinciding with the time when
FGF21 becomes detectable in the blood (Figure S3C). However,
Fgf21 transcript level is high in Deletor muscle already at
12 months of age, progressively increasing from the early mani-
festation of RC deficiency to 25- to 50-fold compared to WT
(Figure S3C). FGF21 knockout prevents weight loss, normalizes
subcutaneous adipocyte size, and prevents WAT browning of
Deletor mice (Figures 3D and 3E). Deletor liver fat is low, the
small fat droplets typically residing intracellularly, whereas in
WT and FKOmice, liver fat is mostly extracellular. DEL-FKO liver
shows WT-like extracellular fat content, whereas FKO in healthy
background leads to excessive fat accumulation (Figure 3F).
These data suggest that systemic FGF21 has manifold conse-
quences to liver and adipose tissue fat content in MM and ex-
plains the low bodyweight—a symptom typical for mitochondrial
disease patients.
To study the roles of FGF21 in systemic metabolism, we per-
formed in vivo uptake assays of glucose and fatty acid analogs
(18F-fluorodeoxyglucose, 18F-FDG; 18F-Fluoro-6-thia-heptade-
canoic acid, 18F-FTHA). The affected muscle, heart, and brown
subcutaneous WAT of Deletors shows significantly increased
18F-FDG uptake (Figure 3G), normalized in DEL-FKO, indicating
that FGF21 drives glucose uptake in the affected tissues in MM
(Figure 3G). Fatty acid uptake (palmitic acid; saturated 16-car-
bon fatty acid) is not changed in Deletor or DEL-FKO mice (Fig-
ure 3H). Metabolic cage examination of systemic metabolism in
Deletors (ambient temperature and thermoneutral +30C) shows
high oxygen consumption and CO2 production, and slightly
increased respiratory exchange ratio compared to WT mice,
consistent with glucose usage as main fuel (Figures 3I, 3J, and
S3D). In DEL-FKO mice, however, metabolism was WT-like.
These results indicate that glucose is the major metabolic fuel
in MM, and the disease-driven high-glucose uptake into the
affected tissues is FGF21 dependent.
FGF21 Has No Effect on MM Disease Signs
The tight connection between FGF21 and advancing disease
made us ask whether loss of FGF21 affected the pathologic hall-
marks of MM: the RC-deficient muscle fibers, mtDNA deletions,
mitochondrial mass, and RC protein amounts (Figures 3K–3N,
S3E, and S3F). None of these disease signs were affected by
FGF21 inactivation, demonstrating that FGF21 does not
(E) Subcutaneous adipocyte size and browning. UCP1 immunohistochemistry (scale bar, 25 mm) and H&E staining (scale bar, 100 mm).
(F) Lipid droplets in liver, Oil Red O staining for lipids and triglycerides; scale bar, 40 mm.
(G and H) In vivo uptake of glucose (18F-FDG) and fatty acid (18F-FTHA, C16:0) analogs in tissues. Statistical analysis and details in Table S3.
(I and J) Respiratory exchange ratio (RER, vCO2/vO2) in thermoneutral condition (+30
C). RER ~ 1, preferred fuel carbohydrates; RER ~ 0.7, lipids as fuel. Light,
sleep; dark, active. (J) Average of RER when mice are active. WT, Deletor, and DEL-FKO, n = 3 per group; FKO, n = 6.
(K–N) Hallmarks of MM in skeletal muscle.
(K) Histochemical in situ RC enzyme activities (brown precipitate, COX-positive; white, COX-negative; blue, SDH positive, COX-negative). Scale bar, 100 mm.
(L) Quantification of (K). WT, FKO, n = 6; Deletor, n = 5; DEL-FKO, n = 8.
(M) Long-range PCR of mtDNA. Smear, multiple deletions (DmtDNA); arrow, full-length 16.6 kb mtDNA. Short-range PCR of mtDNA: non-deleted D-loop region
(573 bp). WT and FKO, n = 3; Deletor and DEL-FKO, n = 5.
(N) qPCR analysis of full-length mtDNA (ratio of ND4 gene region and non-deleted D-loop; n = 5 per group).
Animals: 22–24 months old. Bars: average and SD. Box: 25th and 75th percentiles with median and range. Dots: individual subjects. Pairwise two-tailed t test,
statistical significance: *p% 0.05, **p% 0.01, ***p% 0.001, ****p% 0.0001. Abbreviations (also see legend of Figure 1): UCP1, mitochondrial uncoupling protein
1; 18F-FDG, 18F-fluorodeoxyglucose; 18F-FTHA, 18F-fluoro-6-thia-heptadecanoic acid.
Cell Metabolism 30, 1–15, December 3, 2019 7
Please cite this article in press as: Forsstro¨m et al., Fibroblast Growth Factor 21 Drives Dynamics of Local and Systemic Stress Responses in Mito-
chondrial Myopathy with mtDNA Deletions, Cell Metabolism (2019), https://doi.org/10.1016/j.cmet.2019.08.019
Serine de novo
0
FC
40
30
6
PhgdhPsat1
20
8
10
2
4
***
p=0.056
* **
***** ****** *** ******
Mthfd2 Gdf15 Trib3 Asns
ISRmt before and after FGF21-KO
Atf3 Atf4 Atf5
5
A
*
*
***
FC
4
3
2
1
0
G
dNTP pools
WT DELFKO DEL-FKO
3
2
4
1
0
FC
A TGCA TGCA TGCA TGC
Methyl cycle metabolitesC
3
2
1
4
0
FC
5-
m
-
TH
F
Be
ta
in
e
C
ho
lin
e
G
AA
C
re
a
tin
e
5-
m
-
TH
F
Be
ta
in
e
C
ho
lin
e
G
AA
C
re
a
tin
e
5-
m
-
TH
F
Be
ta
in
e
C
ho
lin
e
G
AA
C
re
a
tin
e
5-
m
-
TH
F
Be
ta
in
e
C
ho
lin
e
G
AA
C
re
a
tin
e
WT DELFKO DEL-FKO
*
*
*** *
AL
A
AR
G
AS
N
AS
P
CY
S
G
LU
G
LN HI
S
IL
E
LE
U
VA
L
M
ET
PH
E
PR
O
SE
R
TH
R
TR
P
TY
R
3
2
1
0
FC
*
*
* * *
E
Amino acids 
DEL vs. WT DEL-FKO vs. FKO
Amino acid and acylcarnitine pools
G
LYAL
A
AR
G
AS
N
AS
P
CY
S
G
LU
G
LN HI
S
IL
E
LE
U
VA
L
M
ET
PH
E
PR
O
SE
R
TH
R
TR
P
TY
R
G
LY
C
ar
n
iti
n
e
 (C
)
Ac
e
ty
l-C
Pr
o
pi
o
n
yl
-
C
Is
o
va
le
ry
l-C
H
ex
a
n
o
yl
-C
O
ct
a
n
o
yl
-
C
D
ec
a
n
o
yl
-
C
D
od
ec
a
n
o
yl
-
C
Te
tra
de
ca
n
o
yl
-
C
Pa
lm
ito
yl
-
C
Is
o
bu
ty
ry
l-L
-
C
St
e
a
ro
yl
-
C
Ar
a
ch
id
yl
-
C
6
4
2
0
8
F
Acylcarnitines
* * * * ** *
FC DEL vs. WT DEL-FKO vs. FKO
C
ar
n
iti
n
e
 (C
)
Ac
e
ty
l-C
Pr
o
pi
o
n
yl
-
C
Is
o
va
le
ry
l-C
H
ex
a
n
o
yl
-C
O
ct
a
n
o
yl
-
C
D
ec
a
n
o
yl
-
C
D
od
ec
a
n
o
yl
-
C
Te
tra
de
ca
n
o
yl
-
C
Pa
lm
ito
yl
-
C
Is
o
bu
ty
ry
l-L
-
C
St
e
a
ro
yl
-
C
Ar
a
ch
id
yl
-
C
Transsulfuration,
Mitochondrial one-carbon cycle
B
MTHFD2105 
MTHFD1L38 
CTH45 
Total 
protein
Total 
protein
Total 
protein
10
15
0
5
FC CTH
4
2
FC
50
100
0
FC
kDa 
0
kDa 
kDa 
MTHFD1L
MTHFD2
0
6
4
2
FC
D FGF21 and transsulfuration pathway
One-carbon metabolism and FGF21H
(legend on next page)
8 Cell Metabolism 30, 1–15, December 3, 2019
Please cite this article in press as: Forsstro¨m et al., Fibroblast Growth Factor 21 Drives Dynamics of Local and Systemic Stress Responses in Mito-
chondrial Myopathy with mtDNA Deletions, Cell Metabolism (2019), https://doi.org/10.1016/j.cmet.2019.08.019
contribute to the development of mtDNA deletions, mitochon-
drial biogenesis, or RC deficiency in skeletal muscle.
FGF21 Regulates Serine Biosynthesis and
Transsulfuration in MM
Fgf21 is a major component in the 1st stage of ISRmt induction in
the skeletal muscle of Deletor mice and also one of the most up-
regulated transcripts in the human patient muscle. Therefore, we
investigated the role of FGF21 in the progression of local muscle
ISRmt in MM. Intriguingly, induction of the 2nd ISRmt stage is
completely dependent on the presence of FGF21: Atf3 and
Atf4, serine de novo biosynthesis (Phgdh and Psat1), and trans-
sulfuration (CTH) all are WT-like in DEL-FKO mice (Figures 4A
and 4B). mTORC1 activation, however, is FGF21 independent,
remaining active in RC-deficient fibers (Figure S4A). The compo-
nents of the 1st ISRmt stage, activated at the same time as
FGF21, are unaffected by FGF21KO (Mthfd2, Atf5; ATF-regu-
lated genes Trib3 and Asns; GDF15) (Figures 4A and 4B). The
data imply differential roles for different ATFs: Atf3 and Atf4 up-
regulation in the 2nd ISRmt stage is dependent on FGF21,
whereas progressive induction of Atf5, starting at the 1st ISRmt
stage, is independently regulated in MM muscle. We also report
uncoupling of regulation of mitochondrial folate cycle: MTHFD1L
induction in Deletors is FGF21 dependent, whereas MTHFD2 is
not (Figure 4B).
In cultured cells with mt-translation inhibition (Figure 2B),
FGF21 was an early 1st ISRmt responder, raising the question
of whether it also regulates downstream ISRmt in these cells.
However, FGF21KO fibroblasts respond to actinonin similar to
WT cells: both transsulfuration and de novo serine biosynthesis
remain induced (Figure S4B). This suggests that the role of
FGF21 as a regulator of ISRmt progression is either stress-signal
or cell-type specific. Altogether, these results show that FGF21
is essential for ISRmt progression and metabolic remodeling
from 1st to 2nd stage, from early to chronic, in post-mitotic skel-
etal muscle in MM.
As the metabolic effects of FGF21 involved biosynthetic path-
ways via folate-driven one-carbon cycle, we analyzed folate
forms and targeted metabolome of 100 metabolites
representing the pathways of interest (Figure S4C). 5-methyl-tet-
rahydrofolate (THF) is rate-limiting for methyl cycle and provides
one-carbon units and precursors for transsulfuration andmethyl-
ation reactions. In Deletor muscle, 5-methyl-THF, betaine,
choline, and glutamate required for methyl cycle and transsulfu-
ration show high concentrations, but in DEL-FKO, their levels are
WT-like (Figures 4C and S4C). Figure 4D summarizes the key
changes in transsulfuration between Deletors and DEL-FKO.
These data suggest that (1) FGF21 drives cell-autonomous
metabolic remodeling in MM, involving major biosynthetic path-
ways, and (2) FGF21-independent MTHFD2 (mitochondrial one-
carbon cycle) does not contribute to cytoplasmic one-carbon
cycle changes in MM muscle.
Amino acids show a general increase in Deletor muscle, and
these levels shift to WT direction in DEL-FKO (Figure 4E). Aspar-
agine is an exception, remaining high in DEL-FKO, consistent
with increased expression of its synthesizing enzyme Asns in
both Deletors and DEL-FKO, and suggesting glutamine usage
of Deletor muscle (Zhang et al., 2014). In DEL-FKO, long-chain
acylcarnitines accumulated (Figures 4F and S4C) despite the
fact that their uptake was not increased (Figure 3H), suggesting
inability to oxidize fatty acids in MM.
Imbalance of nucleotidemetabolism intermediates and dNTPs
is a metabolic hallmark of MM (Nikkanen et al., 2016), purine
biosynthesis being one of the key pathways dependent on mito-
chondrial folate cycle (French et al., 2016). However, dNTP pools
in Deletors are not improved by the absence of FGF21 (Fig-
ure 4G); actually, we show a trend toward elevated dNTPs and
their precursors also in FKO, including deoxyuridine (Figure S4C).
The data suggest that FGF21 has an overall inhibitory effect on
dNTP homeostasis, even in normal muscle.
Altogether, our evidence indicates that FGF21 remodels the
major anabolic biosynthesis pathways, the cytoplasmic one-car-
bon cycle, in MM muscle in an autocrine or paracrine manner
(summarized in Figure 4H).
FGF21 Drives Glucose Uptake and Mitochondrial
Proliferation in Dorsal Hippocampus in MM
FGF21 mediates signals of nutrient metabolism from peripheral
organs to the brain, via binding to b-Klotho receptors in the
choroid plexus and hypothalamus in WT mice (Kurosu et al.,
2007; Ogawa et al., 2007; Potthoff, 2017). However, little infor-
mation exists of the effects of periphery-derived FGF21 on
CNS metabolism in disease. Remarkably, Deletor brain shows
increased in vivo uptake of 18F-FDG specifically in dorsal
Figure 4. FGF21 Regulates Induction of De Novo Serine Biosynthesis and Transsulfuration
(A) Differential FGF21 dependence of ATFs and ISRmt expression in mouse muscle. Transcripts versus beta-actin mRNA (fold change).
(B) Differential FGF21 dependence of mitochondrial folate cycle and transsulfuration. Representative western blot analysis of protein extracts from skeletal
muscle, total protein signal as loading and quantification control. n = 3/group.
(C) Dependence of methyl cycle linked metabolites on FGF21. Targeted metabolomics, significance calculated against WT. WT, FKO, n = 8; Deletor, DEL-FKO,
n = 6.
(D) Schematic representation of transsulfuration pathway and linked metabolites, highlighting the FGF21-dependent changes in Deletor versus WT; DEL-FKO
versus FKO. Red, increased; black, no change; gray, NA.
(E and F) Muscle amino acids and acylcarnitines. Targeted metabolomics, FC normalized to WT. Acylcarnitines shown in the order of increasing chain length (C,
carnitine). Significance calculated against WT. Statistical analysis and details in Tables S4 and S5.
(G) Tissue dNTP pools, FC normalized to WT, n = 3/group.
(H) Schematic representation of folate-driven one-carbon cycle and associated biosynthetic pathways (nucleotide metabolism, methyl cycle, transsulfuration,
and de novo serine biosynthesis) altered in Deletor. Arrows point to FGF21 effects.
Animals: 22–24 months old. Bars: average and SD. Box: 25th and 75th percentiles with median and range. Dots: individual subjects. Pairwise two-tailed t test,
statistical significance: *p% 0.05, **p% 0.01, ***p% 0.001, ****p% 0.0001. Abbreviations (see also Figure 1): dNTP, deoxynucleotide; dTMP, deoxythymidine
monophosphate; GCL, glutamate-cysteine ligase; Hcy, homocysteine; GS, glutathione synthetase; THF, tetrahydrofolate; SAH, S-adenosylhomocysteine; SAM,
S-adenosylmethionine.
Cell Metabolism 30, 1–15, December 3, 2019 9
Please cite this article in press as: Forsstro¨m et al., Fibroblast Growth Factor 21 Drives Dynamics of Local and Systemic Stress Responses in Mito-
chondrial Myopathy with mtDNA Deletions, Cell Metabolism (2019), https://doi.org/10.1016/j.cmet.2019.08.019
AD
I
J
K
E F
G
H
B C
(legend on next page)
10 Cell Metabolism 30, 1–15, December 3, 2019
Please cite this article in press as: Forsstro¨m et al., Fibroblast Growth Factor 21 Drives Dynamics of Local and Systemic Stress Responses in Mito-
chondrial Myopathy with mtDNA Deletions, Cell Metabolism (2019), https://doi.org/10.1016/j.cmet.2019.08.019
hippocampus (DHC; CA2 region). This glucose uptake is fully
FGF21 dependent (Figures 5A, 5B, and S5A). Fgf21 transcription
is not locally elevated in DHC or any region of Deletor brain (Fig-
ure 5C), suggesting that periphery-derived endocrine FGF21
mediates glucose uptake in DHC in MM. To clarify whether
DHC glucose uptake reacts to FGF21 also in WT mice, we
induced FGF21 expression by fasting (24 h) or by intravenous in-
jection of recombinant FGF21, raising FGF21 serum levels 4- to
16-fold higher than in WT, respectively (Figure S5B). However,
these interventions did not affect WT DHC glucose uptake (Fig-
ure S5B). These data indicate that high glucose uptake in DHC is
a manifestation of MM.
Despite the lack of apparent nervous system symptoms, MM
patients and Deletor mice both show mtDNA deletions in their
brain (Suomalainen et al., 1992; Tyynismaa et al., 2005). DEL-
FKO mice harbor similar amounts of mutant mtDNA in their
DHC as the Deletors. Unlike in the muscle, however, mTORC1
activity (S6-phosphorylation) is not increased in the hippocam-
pus of aged Deletors or DEL-FKO (Figure 5E). This conclusion
is further supported by the fact that inhibition of mTORC1 activity
by rapamycin does not affect in vivo glucose uptake in DHC (Fig-
ure S5A), indicating cell-specific roles of mTORC1 in MM.
Unlike in the Deletor muscle, mitochondrial RC activities and
mitochondrial proliferation in DHC remarkably depend on pres-
ence of FGF21. SDH activity is increased in Deletor CA2 of
DHC, especially in the neuronal projections spanning all the layers
of CA2, while the neighboring CA1, CA3, and dentate gyrus are
WT-like (Figures 5F and 5G). CA2 also shows low COX amount
(Figure 5H), consistent with typical RC deficiency. Surprisingly,
however, both the upregulated SDH activity and COX protein
were WT-like in DEL-FKO mice, suggesting that FGF21-depen-
dent mechanism promotes RC deficiency in CA2 of DHC (Figures
5F–5H). Immunohistological analysis of SDHA and mitochondrial
outer membrane protein TOM20 confirms the increase of mito-
chondrialmass inCA2neuronsofDeletors andWT-likemitochon-
drial mass in DEL-FKO (Figures 5I and 5J). Similar appearance of
pyramidal CA2 neurons and lack of astrocytosis or inflammation
suggest no major neuronal damage or cell loss in Deletors or
DEL-FKO (Figures 5K, S5D, and S5E). These data suggest an
intriguing possibility that muscle-derived FGF21 affects RC func-
tion and mitochondrial biogenesis in the brain in MM.
DISCUSSION
Knowledge of the dynamics, regulation, and tissue-specific
characteristics of mammalian stress responses in mitochondrial
disease has been insufficient. Here, we report that ISRmt is tem-
poral and stage-wise orchestrated in mice and humans with
mtDNA stress in skeletal muscle. The early ISRmt is cell autono-
mous with activated expression of FGF21, GDF15, and mito-
chondrial one-carbon cycle, and increasing expression of
ATF5. The 2nd disease stage activates serine and glutathione
synthesis, and glucose uptake into affected tissues and sys-
temicmetabolic changes, and the 3rd terminal stage shows slight
activation of ATF3 and mitochondrial HSPs. Intriguingly, we
show that progression of ISRmt to 2nd stage is dependent on
the metabolic hormone FGF21, which has major local and sys-
temic consequences on metabolism, including in the brain.
These results emphasize crosstalk of the affected and distant
organ systems in mitochondrial disease and indicate that a pri-
mary mitochondrial muscle disease has major systemic
consequences.
Mitochondrial stress responses have been predominantly
characterized in the worm C. elegans, in which RNAi ablation
of RC subunits induced a UPRmt pathway as a consequence of
nuclear relocation of ATFS-1, activating expression of mitochon-
drial HSPs and proteases (Durieux et al., 2011; Haynes et al.,
2007; Nargund et al., 2012). In mammalian cancer cells, mito-
chondrial expression of an aggregation-prone protein elicits a
UPRmt-like response (Zhao et al., 2002), as also occurs in
knockout mice for essential mitochondrial proteins (Bao et al.,
2016; Ost et al., 2015; Pulliam et al., 2014; Seiferling et al.,
2016). We previously described ISRmt as the major stress
response in mammalian MM. The response involved a wide tran-
scriptional andmetabolic remodeling beyondUPRmt (Khan et al.,
2017; Nikkanen et al., 2016), also found in other models with
mtDNA maintenance and expression defects (Twnk, Tfam,
Polrmt, Lrpprc, and Mterf4) (K€uhl et al., 2017). Here, we show
that already the first subclinical disease signs—single COX-
negative muscle fibers in 10- to 12-month-old Deletor mice—
activate local expression of endocrine factors FGF21 and
GDF15 as well as the mitochondrial one-carbon cycle. Fgf21
transcription is first highly upregulated in the muscle, but the
Figure 5. FGF21 Drives Glucose Uptake, Increased Mitochondrial Mass, and Respiratory Activity in CA2 of Hippocampus
(A–B) In vivo glucose (18F-FDG) uptake in coronal brain sections; ex vivo digital autoradiography (ARG); photostimulated luminescence per unit area, PSL/mm2.
(B) Quantification of (A), normalized to cerebellar cortex (similar cerebellar 18F-FDG uptake between the groups confirmed by manual gamma ray emission
measurement; data not shown). WT, FKO, DEL-FKO, n = 6; Deletor, n = 5.
(C) Fgf21 expression in different brain areas of our mouse strains. mRNA versus beta-actin mRNA. Positive controls: deletor muscle and WT mouse liver.
(D) Schematic illustration of hippocampus and CA2 in sagittal view of mouse brain. Hippocampal layers: so, stratum oriens; sp, stratum pyramidale; sr, stratum
radiatum; slm, stratum lacunosum-moleculare.
(E) Phosphorylated S6, a downstream target of mTORC1. Immunohistochemical detection at CA1/CA2 border (dotted line); scale bar, 100 mm.
(F and G) Histochemical in situ enzyme activity of COX (brown, active) and SDH (blue, SDH positive and COX-negative). CA2 in Deletor: increased SDH activity;
scale bar, 200 mm. (G) Enlarged image of CA2 (box) of (F); arrows point to increased SDH activity in long projections (sr); scale bar, 30 mm.
(H) COX protein amount in hippocampus. Immunohistochemical detection of COX (MTCO1 antibody) in CA2 and CA1 border (dotted line). Encircled: pyramidal
cell layer; Deletor, low COX-protein intensity in CA2, when compared to DEL-FKO and WT. Scale bar, 100 mm.
(I and J) Mitochondrial mass in CA2 and CA1 (border: dotted line). (I) SDH amount (SDHA antibody; increased staining in sr with dendritic extensions of pyramidal
neurons). Scale bar, 100 mm. (J) Mitochondrial amount (TOM20 antibody) at CA2/CA1 border increased in extensions of pyramidal neurons (arrows, sr). Scale
bar, 50 mm.
(K) Dendritic network in hippocampal CA2/CA1; MAP2 immunofluorescence. Scale bar, 50 mm.
Animals: 22–24 months old. Bars: average and SD. Dots: individual subjects. Pairwise two-tailed t test, statistical significance: *p% 0.05, **p % 0.01, ***p %
0.001, ****p% 0.0001. In histology, n = 3 per group. Abbreviations (see also Figure 1 legend): cpu, caudoputamen of striatum; CTX, cortex; DG, dentate gyrus;
DHC, dorsal hippocampus; HT, hypothalamus; MAP2, microtubule-associated protein 2; VHC, ventral hippocampus; 18F-FDG, 18F-fluorodeoxyglucose.
Cell Metabolism 30, 1–15, December 3, 2019 11
Please cite this article in press as: Forsstro¨m et al., Fibroblast Growth Factor 21 Drives Dynamics of Local and Systemic Stress Responses in Mito-
chondrial Myopathy with mtDNA Deletions, Cell Metabolism (2019), https://doi.org/10.1016/j.cmet.2019.08.019
protein is increased in the blood only after 4–6 months, at the
same time when ragged-red fibers—a hallmark of MM—emerge
in the muscle. FGF21 presence in the blood induces weight loss
and WAT browning, known endocrine consequences of liver-
derived FGF21 in fasting rodents (Badman et al., 2007; Xu
et al., 2009). In MM muscle, the 2nd ISRmt stage activates only
if FGF21 is present, with upregulated glucose uptake, de novo
serine biogenesis, and transsulfuration. Therefore, the local
and systemic 2nd ISRmt stage is crucially dependent on auto-
or paracrine and endocrine effects of FGF21, which together
mark the progression of the muscle disease from local to sys-
temic. Which of the FGF21-dependent metabolic events are
direct consequences of the hormone remains to be studied.
FGF21-independent components of the second ISRmt stage in
the muscle include mTORC1 activation, nucleotide pool imbal-
ance, and mitochondrial mass increase, agreeing with the previ-
ous data of mTORC1 contribution in RRF formation (Khan et al.,
2017) and the regulatory role of mTORC1 for purine synthesis in
cancer cells (French et al., 2016). In the hippocampus of MM
mice, mitochondrial mass increase or RC deficiency is depen-
dent on periphery-derived FGF21, but mTORC1 is not active,
emphasizing tissue specificity of the responses. Overall, these
data indicate that mtDNA expression defects induce an early,
local subclinical response, while a second systemic, chronic
stage modifies metabolism in the whole organism. These data
underline the complexity of ISRmt and the essential roles of
FGF21, both in the dynamics of ISRmt and progression of MM
to a systemic condition.
Subtle induction of UPRmt-related mitochondrial HSPs occurs
in 2-year-old Deletor mice, forming the third ISRmt stage. In
mammals, this HSP response does not spread beyond the
affected tissue, unlike in C. elegans (Durieux et al., 2011; Zhang
et al., 2018). Still, the endocrine effects of muscle-derived FGF21
on hepatic lipolysis, WAT browning, brain glucose uptake, and
mitochondrial mass and function mimic the ‘‘mitokine’’ concept
of the worm: FGF21 spreads a signal of mitochondrial dysfunc-
tion from the affected organ to distant ones. The exact mecha-
nism of how the FGF21-related muscle-to-brain signaling occurs
in our mice, specifically affecting the dorsal hippocampus, re-
quires future attention, as this region has not previously been
associated with metabolic regulation or mitochondrial disease.
We conclude that endocrine signaling induced bymtDNA stress,
from the affected to distant tissues, is conserved from worms to
mammals but in mammals is importantly contributed by FGF21,
signaling from the periphery to the brain.
FGF21 had differential effects of RC deficiency and mitochon-
drial mass in the muscle and DHC. The dorsal hippocampus of
Deletor mice showed mitochondrial mass increase and RC defi-
ciency only in the presence of FGF21 but no changes in mtDNA
deletion load. In the muscle, however, FGF21 did not affect RC
activities or mitochondrial mass. mTORC1 is induced in muscle
fibers with mitochondrial proliferation, but in the brain of the
same animals, mTORC1 is not activated. These data underscore
the remarkable tissue specificity in the upstream regulation of
the mitochondrial stress responses. Importantly, our evidence
indicates that a ‘‘general mitochondrial stress response’’ does
not exist, but the specific components of ISRmt need to be
characterized separately for different organisms, tissues, and
insults.
ATF transcription factor family regulates the AARE response,
but which homologs of the worm ATFS-1 are important in mam-
mals has been controversial. Previously, toxin-mediated oxida-
tive stress (paraquat) in human embryonic kidney cells was
reported to induce mitochondrial HSPs in an ATF5-dependent
manner (Fiorese et al., 2016). Thereafter, amulti-omics approach
identified ATF4 to be the main driver of ISRmt in cervical cancer
cells treated with variable inhibitors of mtDNA expression, mito-
chondrial protein import, or membrane potential (Quiro´s et al.,
2017). Our data from MM mice show involvement of ATF5 and
ATF3 in post-mitotic muscle, with different temporal dynamics.
ATF5 expression is a robust early responder to mtDNA replica-
tion stress in both mice and human patients. In mice, ATF3
(and 4) get induced in the later stages. Interestingly, upregulation
of ATF3 and ATF4 depends completely on the presence of
FGF21, whereas ATF5 is FGF21 independent, further empha-
sizing the chronology of induction. Their roles are, however,
different in proliferating cells where ATF4 gets activated early,
is FGF21 independent, and is required for the activation of other
ISRmt components. These data indicate that the ATF family is
well conserved as a mitochondrial stress sensor from worms
to mammals, but specific ATFs have different roles in different
cell types and in acute or chronic stress.
In Deletors, the overall health, muscle morphology, or lifespan
is not significantly affected by the presence or absence of
FGF21, and they show no signs of inflammation; the major ef-
fects of FGF21 are metabolic. These findings differ from those
in muscle-specific OPA1-KO mice, in which FGF21 induction
was deleterious, promoting inflammation (Tezze et al., 2017),
which may be explained by the difference in severity of the mito-
chondrial insult in OPA1-KO versus Deletor mice. Embryonic
OPA1-KO inmuscle is lethal at early postnatal life, and even con-
ditional OPA1-KO in muscle results in severe muscle atrophy,
rapid body weight loss, and death of themice <100 days after in-
duction, reflecting the vital nature of the protein (Tezze et al.,
2017). Remarkably, FGF21KO in OPA1-KO animals reverted
the systemic inflammation and prolonged the lifespan for >100
additional days. The results from OPA1-KO and Deletors
together suggest concentration and/or context dependence of
FGF21 effects, which may have clinical relevance: children
with severe, often early lethal disorders of mt-translation or
mtDNA maintenance have very high FGF21 concentration (up
to 200-fold), whereas adults with MM show more modest induc-
tion (2- to 10-fold) (Lehtonen et al., 2016)—the former mimic
FGF21 in OPA1-KO and the latter FGF21 in Deletors. These
data suggest that upon severe mitochondrial dysfunction, inhibi-
tion of FGF21 might be beneficial, whereas in late-onset disor-
ders, the mildly elevated FGF21 driving, for example, glutathione
synthesis, might be protective.
Importantly, our evidence indicates that mammalian ISRmt is a
complex response, with coordinated dynamics, auto/paracrine
feedback loops, and endocrine signaling, affecting the whole or-
ganism, with FGF21 as its circulating messenger. The initial sig-
nal(s) that trigger the early ISRmt remain to be studied. Previous
studies in cultured cells suggested the signal to involve accumu-
lation of aberrant mt-translation products in the inner mitochon-
drial membrane (Richter et al., 2015). Our results highlight the
systemic metabolic consequences of mitochondrial dysfunction
that extend far beyond the primary RC deficiency or the affected
12 Cell Metabolism 30, 1–15, December 3, 2019
Please cite this article in press as: Forsstro¨m et al., Fibroblast Growth Factor 21 Drives Dynamics of Local and Systemic Stress Responses in Mito-
chondrial Myopathy with mtDNA Deletions, Cell Metabolism (2019), https://doi.org/10.1016/j.cmet.2019.08.019
organ. The usefulness of FGF21 as a therapy target remains to
be studied: despite the robust morphologic and metabolic
changes, the functional defects of the Deletor mice with and
without FGF21 are subtle, and their lifespan normal. However,
the FGF21 contribution to weight loss might be relevant for pa-
tients as well, as this is one of their most common and socially
challenging symptoms. Furthermore, our results indicate that a
muscle disease can affect brain metabolism, highlighting the
importance of zooming out from specific tissues when consid-
ering disease pathogenesis and considering the whole organism
when assessing treatment effects.
Limitations of Study
Twinkle dysfunction in the Deletor mouse and patients with MM
causes progressive accumulation of mtDNA deletions in post-
mitotic tissues. mtDNA deletions are typically not produced
and/or maintained in replicating cell types, and the mechanistic
overlap of the mitochondrial stress response in post-mitotic tis-
sues and our cultured cell model for mt-translation stress may
only be partial.
The Deletor and Deletor-FGF21KO mice have a normal life-
span and show no motor deficiency. Therefore, the effects of
FGF21 or absence for the performance of the MM mice could
not be tested. The evidence provided by the full-body Deletor-
FGF21KO mice undisputedly shows that FGF21 is required for
the metabolic phenotypes in the Deletors, especially in the mus-
cle and brain. However, whether these phenotypes are directly
FGF21 dependent, or mediated by a signaling cascade, remains
to be studied.
STAR+METHODS
Detailed methods are provided in the online version of this paper
and include the following:
d KEY RESOURCES TABLE
d LEAD CONTACT AND MATERIALS AVAILABILITY
B Lead Contact
d EXPERIMENTAL MODEL AND SUBJECT DETAILS
B Ethical Approval
B Animal Models
B Cell Lines
B MEF Isolation
B Cell Culture
B Respiratory Function Testing in Animals
B Tracer Assays
d METHOD DETAILS
B Isolation of Mitochondria
B MtDNA Deletion Load and Copy Number Determi-
nation
B Gene Expression Analysis
B Gene Silencing
B Western Blotting
B Histology
B Quantification of FGF21
B Metabolomics
B dNTP Pools
B RNA Sequencing of Patient Muscle
B Production of Recombinant Mouse FGF21
d QUANTIFICATION AND STATISTICAL ANALYSIS
d DATA AND CODE AVAILABILITY
SUPPLEMENTAL INFORMATION
Supplemental Information can be found online at https://doi.org/10.1016/j.
cmet.2019.08.019.
ACKNOWLEDGMENTS
The authors wish to thank A. Harju, M. Innil€a, T. Manninen, B. Hollmann, and A.
Honkaniemi for technical contributions and expertise; the Functional Geno-
mics Unit for virus production; the University of Helsinki Electron Microscopy
Core Facility for their expertise in ultrastructural experiments; and the Univer-
sity of Helsinki Laboratory Animal Center. J. Palmio is thanked for patient
recruitment. We wish to acknowledge funding support from the Academy of
Finland (A.S. and B.J.B.), European Research Council, and Sigrid Juselius
Foundation (A.S.); Swiss National Science Foundation and Novartis Founda-
tion for Medical-Biological Research (C.J.); and Helsinki Biomedical Graduate
Program, Maud Kuistila Memorial Foundation, Oskar O¨flund Foundation, Wal-
demar von Frenkell’s Foundation, and Biomedicum Helsinki Foundation (S.F.).
AUTHOR CONTRIBUTIONS
S.F. designed and conducted the in vivo study, and analyzed and interpreted
data; C.J. designed and conducted the in vitro study, and analyzed and
interpreted data. C.J.C., M.K., J.N., and L.E. participated in study design
and supervision; I.-M.K., S.P., A.M., N.K., P.M., and H.L. conducted
experiments, and analyzed and interpreted data. L.W. has expertise in nucle-
otide analysis and interpretation. E.P. andM.A. collected and analyzed patient
data. U.R. and B.J.B. participated in study design and interpretation. A.R. su-
pervised PET analyses, and V.V. supervised metabolomics analysis and inter-
pretation. A.S. conceived, designed, and supervised the study. S.F., C.J., and
A.S. wrote the manuscript, which was edited and approved by all co-authors.
DECLARATION OF INTERESTS
The authors declare no competing interests.
Received: March 8, 2019
Revised: July 9, 2019
Accepted: August 20, 2019
Published: September 12, 2019
REFERENCES
Agnew, T., Goldsworthy, M., Aguilar, C., Morgan, A., Simon, M., Hilton, H.,
Esapa, C., Wu, Y., Cater, H., Bentley, L., et al. (2018). A Wars2 mutant mouse
model displays OXPHOS deficiencies and activation of tissue-specific stress
response pathways. Cell Rep 25, 3315–3328.
Ahola, S., Auranen, M., Isohanni, P., Niemisalo, S., Urho, N., Buzkova, J.,
Velagapudi, V., Lundbom, N., Hakkarainen, A., Muurinen, T., et al. (2016).
Modified Atkins diet induces subacute selective ragged-red-fiber lysis in mito-
chondrial myopathy patients. EMBO Mol. Med 8, 1234–1247.
Ahola-Erkkil€a, S., Carroll, C.J., Peltola-Mjo¨sund, K., Tulkki, V., Mattila, I.,
Sepp€anen-Laakso, T., Oresic, M., Tyynismaa, H., and Suomalainen, A.
(2010). Ketogenic diet slows down mitochondrial myopathy progression in
mice. Hum. Mol. Genet. 19, 1974–1984.
Badman, M.K., Pissios, P., Kennedy, A.R., Koukos, G., Flier, J.S., and
Maratos-Flier, E. (2007). Hepatic fibroblast growth factor 21 is regulated by
PPARalpha and is a key mediator of hepatic lipid metabolism in ketotic states.
Cell Metab 5, 426–437.
Bao, X.R., Ong, S.E., Goldberger, O., Peng, J., Sharma, R., Thompson, D.A.,
Vafai, S.B., Cox, A.G., Marutani, E., Ichinose, F., et al. (2016). Mitochondrial
dysfunction remodels one-carbon metabolism in human cells. Elife 5, e10575.
Dogan, S.A., Pujol, C., Maiti, P., Kukat, A., Wang, S., Hermans, S., Senft, K.,
Wibom, R., Rugarli, E.I., and Trifunovic, A. (2014). Tissue-specific loss of
Cell Metabolism 30, 1–15, December 3, 2019 13
Please cite this article in press as: Forsstro¨m et al., Fibroblast Growth Factor 21 Drives Dynamics of Local and Systemic Stress Responses in Mito-
chondrial Myopathy with mtDNA Deletions, Cell Metabolism (2019), https://doi.org/10.1016/j.cmet.2019.08.019
DARS2 activates stress responses independently of respiratory chain defi-
ciency in the heart. Cell Metab 19, 458–469.
Durieux, J., Wolff, S., and Dillin, A. (2011). The cell-non-autonomous nature of
electron transport chain-mediated longevity. Cell 144, 79–91.
Fiorese, C.J., Schulz, A.M., Lin, Y.F., Rosin, N., Pellegrino, M.W., and Haynes,
C.M. (2016). The transcription factor ATF5 mediates a mammalian mitochon-
drial UPR. Curr. Biol. 26, 2037–2043.
French, J.B., Jones, S.A., Deng, H., Pedley, A.M., Kim, D., Chan, C.Y., Hu, H.,
Pugh, R.J., Zhao, H., Zhang, Y., et al. (2016). Spatial colocalization and func-
tional link of purinosomes with mitochondria. Science 351, 733–737.
Gardner, B.M., Pincus, D., Gotthardt, K., Gallagher, C.M., and Walter, P.
(2013). Endoplasmic reticulum stress sensing in the unfolded protein
response. Cold Spring Harb. Perspect. Biol. 5, a013169.
Gorman, G.S., Chinnery, P.F., DiMauro, S., Hirano, M., Koga, Y., McFarland,
R., Suomalainen, A., Thorburn, D.R., Zeviani, M., and Turnbull, D.M. (2016).
Mitochondrial diseases. Nat. Rev. Dis. Prim 2, 16080.
Haynes, C.M., Petrova, K., Benedetti, C., Yang, Y., and Ron, D. (2007). ClpP
mediates activation of a mitochondrial unfolded protein response in C. ele-
gans. Dev. Cell 13, 467–480.
Haynes, C.M., Fiorese, C.J., and Lin, Y.F. (2013). Evaluating and responding to
mitochondrial dysfunction: the mitochondrial unfolded-protein response and
beyond. Trends Cell Biol. 23, 311–318.
Inagaki, T., Dutchak, P., Zhao, G., Ding, X., Gautron, L., Parameswara, V., Li,
Y., Goetz, R., Mohammadi, M., Esser, V., et al. (2007). Endocrine regulation of
the fasting response by PPARalpha-mediated induction of fibroblast growth
factor 21. Cell Metab 5, 415–425.
Keipert, S., Ost, M., Johann, K., Imber, F., Jastroch, M., van Schothorst, E.M.,
Keijer, J., and Klaus, S. (2014). Skeletal muscle mitochondrial uncoupling
drives endocrine cross-talk through the induction of FGF21 as a myokine.
Am. J. Physiol. Endocrinol. Metab. 306, E469–E482.
Khan, N.A., Nikkanen, J., Yatsuga, S., Jackson, C., Wang, L., Pradhan, S.,
Kivel€a, R., Pessia, A., Velagapudi, V., and Suomalainen, A. (2017). mTORC1
regulates mitochondrial integrated stress response and mitochondrial myop-
athy progression. Cell Metab.26 26, 419–428.e5.
Kharitonenkov, A., and Shanafelt, A.B. (2009). FGF21: a novel prospect for the
treatment of metabolic diseases. Curr. Opin. Investig. Drugs 10, 359–364.
Kharitonenkov, A., Shiyanova, T.L., Koester, A., Ford, A.M., Micanovic, R.,
Galbreath, E.J., Sandusky, G.E., Hammond, L.J., Moyers, J.S., Owens, R.A.,
et al. (2005). FGF-21 as a novel metabolic regulator. J. Clin. Invest 115,
1627–1635.
K€uhl, I., Miranda, M., Atanassov, I., Kuznetsova, I., Hinze, Y., Mourier, A.,
Filipovska, A., and Larsson, N.G. (2017). Transcriptomic and proteomic land-
scape of mitochondrial dysfunction reveals secondary coenzyme Q deficiency
in mammals. Elife 6, e30952.
Kurosu, H., Choi, M., Ogawa, Y., Dickson, A.S., Goetz, R., Eliseenkova, A.V.,
Mohammadi, M., Rosenblatt, K.P., Kliewer, S.A., and Kuro-o, M. (2007).
Tissue-specific expression of betaKlotho and fibroblast growth factor (FGF) re-
ceptor isoforms determines metabolic activity of FGF19 and FGF21. J. Biol.
Chem. 282, 26687–26695.
Lallemand, Y., Luria, V., Haffner-Krausz, R., and Lonai, P. (1998). Maternally
expressed PGK-Cre transgene as a tool for early and uniform activation of
the Cre site-specific recombinase. Transgenic Res. 7, 105–112.
Lehtonen, J.M., Forsstro¨m, S., Bottani, E., Viscomi, C., Baris, O.R., Isoniemi,
H., Ho¨ckerstedt, K., O¨sterlund, P., Hurme, M., Jylh€av€a, J., et al. (2016).
FGF21 is a biomarker for mitochondrial translation and mtDNA maintenance
disorders. Neurology 87, 2290–2299.
MacVicar, T., and Langer, T. (2016). OPA1 processing in cell death and disease
- the long and short of it. J. Cell Sci. 129, 2297–2306.
Martı´, R., Dorado, B., and Hirano, M. (2012). Measurement of mitochondrial
dNTP pools. Methods Mol. Biol. 837, 135–148.
Merkwirth, C., Jovaisaite, V., Durieux, J., Matilainen, O., Jordan, S.D., Quiros,
P.M., Steffen, K.K., Williams, E.G., Mouchiroud, L., Tronnes, S.U., et al. (2016).
Two conserved histone demethylases regulate mitochondrial stress-induced
longevity. Cell 165, 1209–1223.
Nandania, J., Kokkonen, M., Euro, L., and Velagapudi, V. (2018a).
Simultaneous measurement of folate cycle intermediates in different biological
matrices using liquid chromatography-tandem mass spectrometry.
J. Chromatogr. B Analyt. Technol. Biomed. Life Sci. 1092, 168–178.
Nandania, J., Peddinti, G., Pessia, A., Kokkonen, M., and Velagapudi, V.
(2018b). Validation and automation of a high-throughput multitargetedmethod
for semiquantification of endogenous metabolites from different biological
matrices using tandem mass spectrometry. Metabolites 8.
Nargund, A.M., Pellegrino, M.W., Fiorese, C.J., Baker, B.M., and Haynes, C.M.
(2012). Mitochondrial import efficiency of ATFS-1 regulatesmitochondrial UPR
activation. Science 337, 587–590.
Nikkanen, J., Forsstro¨m, S., Euro, L., Paetau, I., Kohnz, R.A., Wang, L., Chilov,
D., Viinam€aki, J., Roivainen, A., Marjam€aki, P., et al. (2016). Mitochondrial DNA
replication defects disturb cellular dNTP pools and remodel one-carbonmeta-
bolism. Cell Metab 23, 635–648.
Nunnari, J., and Suomalainen, A. (2012). Mitochondria: in sickness and in
health. Cell 148, 1145–1159.
Ogawa, Y., Kurosu, H., Yamamoto, M., Nandi, A., Rosenblatt, K.P., Goetz, R.,
Eliseenkova, A.V., Mohammadi, M., and Kuro-o, M. (2007). BetaKlotho is
required for metabolic activity of fibroblast growth factor 21. Proc. Natl.
Acad. Sci. USA 104, 7432–7437.
Olichon, A., Baricault, L., Gas, N., Guillou, E., Valette, A., Belenguer, P., and
Lenaers, G. (2003). Loss of OPA1 perturbates the mitochondrial inner mem-
brane structure and integrity, leading to cytochrome c release and apoptosis.
J. Biol. Chem. 278, 7743–7746.
Ost, M., Keipert, S., van Schothorst, E.M., Donner, V., van der Stelt, I., Kipp,
A.P., Petzke, K.J., Jove, M., Pamplona, R., Portero-Otin, M., et al. (2015).
Muscle mitohormesis promotes cellular survival via serine/glycine pathway
flux. FASEB J 29, 1314–1328.
Ost, M., Coleman, V., Voigt, A., van Schothorst, E.M., Keipert, S., van der Stelt,
I., Ringel, S., Graja, A., Ambrosi, T., Kipp, A.P., et al. (2016). Muscle mitochon-
drial stress adaptation operates independently of endogenous FGF21 action.
Mol. Metab 5, 79–90.
Pereira, R.O., Tadinada, S.M., Zasadny, F.M., Oliveira, K.J., Pires, K.M.P.,
Olvera, A., Jeffers, J., Souvenir, R., McGlauflin, R., Seei, A., et al. (2017).
OPA1 deficiency promotes secretion of FGF21 from muscle that prevents
obesity and insulin resistance. EMBO J 36, 2126–2145.
Potthoff, M.J. (2017). FGF21 and metabolic disease in 2016: a new frontier in
FGF21 biology. Nat. Rev. Endocrinol 13, 74–76.
Potthoff, M.J., Inagaki, T., Satapati, S., Ding, X., He, T., Goetz, R.,
Mohammadi, M., Finck, B.N., Mangelsdorf, D.J., Kliewer, S.A., et al. (2009).
FGF21 induces PGC-1alpha and regulates carbohydrate and fatty acid meta-
bolism during the adaptive starvation response. Proc. Natl. Acad. Sci. USA
106, 10853–10858.
Pulliam, D.A., Deepa, S.S., Liu, Y., Hill, S., Lin, A.L., Bhattacharya, A., Shi, Y.,
Sloane, L., Viscomi, C., Zeviani, M., et al. (2014). Complex IV-deficient
Surf1(-/-) mice initiate mitochondrial stress responses. Biochem. J. 462,
359–371.
Quiro´s, P.M., Prado, M.A., Zamboni, N., D’Amico, D., Williams, R.W., Finley,
D., Gygi, S.P., and Auwerx, J. (2017). Multi-omics analysis identifies ATF4 as
a key regulator of the mitochondrial stress response in mammals. J. Cell
Biol. 216, 2027–2045.
Richter, U., Lahtinen, T., Marttinen, P., Myo¨h€anen, M., Greco, D., Cannino, G.,
Jacobs, H.T., Lietze´n, N., Nyman, T.A., andBattersby, B.J. (2013). Amitochon-
drial ribosomal andRNAdecay pathway blocks cell proliferation. Curr. Biol. 23,
535–541.
Richter, U., Lahtinen, T., Marttinen, P., Suomi, F., and Battersby, B.J. (2015).
Quality control of mitochondrial protein synthesis is required for membrane
integrity and cell fitness. J. Cell Biol. 211, 373–389.
Richter, U., Ng, K.Y., Suomi, F., Marttinen, P., Turunen, T., Jackson, C.,
Suomalainen, A., Vihinen, H., Jokitalo, E., Nyman, T.A., et al. (2019).
Mitochondrial stress response triggered by defects in protein synthesis quality
control. Life Sci. Alliance 2.
14 Cell Metabolism 30, 1–15, December 3, 2019
Please cite this article in press as: Forsstro¨m et al., Fibroblast Growth Factor 21 Drives Dynamics of Local and Systemic Stress Responses in Mito-
chondrial Myopathy with mtDNA Deletions, Cell Metabolism (2019), https://doi.org/10.1016/j.cmet.2019.08.019
Rygiel, K.A., Grady, J.P., Taylor, R.W., Tuppen, H.A.L., and Turnbull, D.M.
(2015). Triplex real-time PCR - an improvedmethod to detect a wide spectrum
of mitochondrial DNA deletions in single cells. Sci. Rep 5, 9906.
Seiferling, D., Szczepanowska, K., Becker, C., Senft, K., Hermans, S., Maiti, P.,
Ko¨nig, T., Kukat, A., and Trifunovic, A. (2016). Loss of CLPP alleviates mito-
chondrial cardiomyopathy without affecting the mammalian UPRmt. EMBO
Rep 17, 953–964.
Sidrauski, C., and Walter, P. (1997). The transmembrane kinase Ire1p is a site-
specific endonuclease that initiates mRNA splicing in the unfolded protein
response. Cell 90, 1031–1039.
Spelbrink, J.N., Li, F.Y., Tiranti, V., Nikali, K., Yuan, Q.P., Tariq, M., Wanrooij,
S., Garrido, N., Comi, G., Morandi, L., et al. (2001). Human mitochondrial
DNA deletions associated with mutations in the gene encoding Twinkle, a
phage T7 gene 4-like protein localized in mitochondria. Nat. Genet. 28,
223–231.
Suomalainen, A., and Battersby, B.J. (2018). Mitochondrial diseases: the
contribution of organelle stress responses to pathology. Nat. Rev. Mol. Cell
Biol. 19, 77–92.
Suomalainen, A., Majander, A., Haltia, M., Somer, H., Lo¨nnqvist, J.,
Savontaus, M.L., and Peltonen, L. (1992). Multiple deletions of mitochondrial
DNA in several tissues of a patient with severe retarded depression and familial
progressive external ophthalmoplegia. J. Clin. Invest 90, 61–66.
Suomalainen, A., Majander, A.,Wallin, M., Set€al€a, K., Kontula, K., Leinonen, H.,
Salmi, T., Paetau, A., Haltia, M., Valanne, L., et al. (1997). Autosomal dominant
progressive external ophthalmoplegia with multiple deletions of mtDNA: clin-
ical, biochemical, and molecular genetic features of the 10q-linked disease.
Neurology 48, 1244–1253.
Suomalainen, A., Elo, J.M., Pietil€ainen, K.H., Hakonen, A.H., Sevastianova, K.,
Korpela, M., Isohanni, P., Marjavaara, S.K., Tyni, T., Kiuru-Enari, S., et al.
(2011). FGF-21 as a biomarker for muscle-manifesting mitochondrial respira-
tory chain deficiencies: a diagnostic study. Lancet Neurol. 10, 806–818.
Taylor, R.C., and Dillin, A. (2013). XBP-1 is a cell-nonautonomous regulator of
stress resistance and longevity. Cell 153, 1435–1447.
Tezze, C., Romanello, V., Desbats, M.A., Fadini, G.P., Albiero, M., Favaro, G.,
Ciciliot, S., Soriano, M.E., Morbidoni, V., Cerqua, C., et al. (2017). Age-associ-
ated loss of OPA1 in muscle impacts muscle mass, metabolic homeostasis,
systemic inflammation, and epithelial senescence. Cell Metab. 25, 1374–1389.
Tyynismaa, H., Mjosund, K.P., Wanrooij, S., Lappalainen, I., Ylikallio, E.,
Jalanko, A., Spelbrink, J.N., Paetau, A., and Suomalainen, A. (2005). Mutant
mitochondrial helicase Twinkle causes multiple mtDNA deletions and a late-
onset mitochondrial disease in mice. Proc. Natl. Acad. Sci. USA 102,
17687–17692.
Tyynismaa, H., Carroll, C.J., Raimundo, N., Ahola-Erkkil€a, S., Wenz, T.,
Ruhanen, H., Guse, K., Hemminki, A., Peltola-Mjøsund, K.E., Tulkki, V., et al.
(2010). Mitochondrial myopathy induces a starvation-like response. Hum.
Mol. Genet. 19, 3948–3958.
Xu, J., Lloyd, D.J., Hale, C., Stanislaus, S., Chen, M., Sivits, G., Vonderfecht,
S., Hecht, R., Li, Y.S., Lindberg, R.A., et al. (2009). Fibroblast growth factor
21 reverses hepatic steatosis, increases energy expenditure, and improves in-
sulin sensitivity in diet-induced obese mice. Diabetes 58, 250–259.
Zhang, Z., Tsukikawa, M., Peng, M., Polyak, E., Nakamaru-Ogiso, E.,
Ostrovsky, J., McCormack, S., Place, E., Clarke, C., Reiner, G., et al. (2013).
Primary respiratory chain disease causes tissue-specific dysregulation of the
global transcriptome and nutrient-sensing signaling network. PLoS One 8,
e69282.
Zhang, J., Fan, J., Venneti, S., Cross, J.R., Takagi, T., Bhinder, B., Djaballah,
H., Kanai, M., Cheng, E.H., Judkins, A.R., et al. (2014). Asparagine plays a crit-
ical role in regulating cellular adaptation to glutamine depletion. Mol. Cell 56,
205–218.
Zhang, Q., Wu, X., Chen, P., Liu, L., Xin, N., Tian, Y., and Dillin, A. (2018). The
mitochondrial unfolded protein response is mediated cell-non-autonomously
by Retromer-dependent Wnt signaling. Cell 174, 870–883.
Zhao, Q., Wang, J., Levichkin, I.V., Stasinopoulos, S., Ryan, M.T., and
Hoogenraad, N.J. (2002). Amitochondrial specific stress response in mamma-
lian cells. EMBO J. 21, 4411–4419.
Cell Metabolism 30, 1–15, December 3, 2019 15
Please cite this article in press as: Forsstro¨m et al., Fibroblast Growth Factor 21 Drives Dynamics of Local and Systemic Stress Responses in Mito-
chondrial Myopathy with mtDNA Deletions, Cell Metabolism (2019), https://doi.org/10.1016/j.cmet.2019.08.019
STAR+METHODS
KEY RESOURCES TABLE
REAGENT or RESOURCE SOURCE IDENTIFIER
Antibodies
Mouse-anti-ATF4 Santa Cruz Cat# sc-390063; RRID:AB_2810998
Goat-anti-b-ACTIN Santa Cruz Cat# sc-1616; RRID:AB_630836
Rabbit anti-BiP Cell Signaling Cat# 3177; RRID:AB_2119845
Rabbit anti-CLPP Proteintech Cat# 15698-1-AP; RRID:AB_2245115
Rabbit anti-CTH/CSE Proteintech Cat# 12217-1-AP; RRID:AB_2087497
Rabbit anti-HSP10 Abcam Cat# ab108600; RRID:AB_10864371
Rabbit anti-HSP60 Abcam Cat# ab46798; RRID:AB_881444
Rabbit anti-HSP70 Abcam Cat# ab53098; RRID:AB_880311
Rabbit anti-LONP1 Sigma Cat# HPA002192; RRID:AB_1079695
Rabbit anti-MTHFD2 Abcam Cat# ab37840, RRID:AB_776544
Rabbit anti-MTHFD1L Proteintech Cat# 16113-1-AP, RRID:AB_2250974
Rabbit anti-OPA1 BD Biosciences Cat# 612606; RRID:AB_399888
Rabbit anti-PORIN Abcam Cat# ab15895; RRID:AB_2214787
Mouse anti-SDHA Abcam Cat# ab14715; RRID:AB_301433
Rabbit anti-TOM20 Santa Cruz Cat# sc-11415; RRID:AB_2207533
Rabbit anti-Vinculin Abcam Cat# ab129002; RRID:AB_11144129
Rabbit anti-UCP1 Abcam Cat# ab10983; RRID:AB_2241462
Rabbit anti-phospho-S6 Cell Signaling Technology Cat# 4858; RRID:AB_916156
Rabbit anti-GFAP Millipore Cat# AB5804; RRID:AB_2109645
Rabbit anti-IBA1 Wako Cat# 019-19741; RRID:AB_839504
Chicken anti-MAP2 Abcam Cat# ab5392; RRID:AB_2138153
Goat anti-mouse IgG Jackson ImmunoResearch Cat# 115-035-146; RRID:AB_2307392
Goat anti-rabbit IgG Jackson ImmunoResearch Cat# 111-035-144; RRID:AB_2307391
Rabbit anti-goat IgG Millipore Cat# 401504-2ML; RRID:AB_437813
Goat anti-rabbit IgG Alexa Fluor488 Abcam Cat# ab150077
Goat anti-chicken IgG Alexa Fluor405 Abcam Cat# ab175674
Donkey anti-rabbit IgG Alexa Fluor594 Abcam Cat# ab150068
Bacterial and Virus Strains
BL21(DE3) E. coli cells (with pET-SUMO-FGF21) Suomalainen Lab N/A
Chemicals, Peptides, and Recombinant Proteins
18F-FTHA (18F-Fluoro-6-thia-heptadecanoic acid) Suomalainen Lab N/A
18F-FDG (18F-Fludeoxyglucose) Suomalainen Lab N/A
Recombinant FGF21 Suomalainen Lab N/A
Rapamycin LC Laboratories Cat#R-5000
DMSO Sigma Aldrich Cat#D8418
PEG-400 Sigma Aldrich Cat#202398
Phosphatase inhibitor cocktail Thermo Scientific Cat#78420
SYBR Green Supermix Bio-Rad Cat#1725006CUST
Protein assay kit Bio-Rad Cat#500-0114
Pierce Protease Inhibitor Mini Tablets, EDTA-Free Thermo Scientific Cat#88666
Actinonin Sigma Aldrich Cat#A6671
JET Prime transfection reagent Polypus Cat#114-07
Uridine Calbiochem Cat#6680
Dexamethasone Sigma Aldrich Cat#D1756
(Continued on next page)
e1 Cell Metabolism 30, 1–15.e1–e7, December 3, 2019
Please cite this article in press as: Forsstro¨m et al., Fibroblast Growth Factor 21 Drives Dynamics of Local and Systemic Stress Responses in Mito-
chondrial Myopathy with mtDNA Deletions, Cell Metabolism (2019), https://doi.org/10.1016/j.cmet.2019.08.019
Continued
REAGENT or RESOURCE SOURCE IDENTIFIER
Insulin Sigma Aldrich Cat#I-1882
Fetuin Sigma Aldrich Cat#F-2379
Phusion High-Fidelity DNA Polymerase Thermo Scientific Cat#530L
Oil Red O Sigma Aldrich Cat#O0625
Glutaraldehyde solution Sigma Aldrich Cat#G7651
TRIzol Reagent Thermo Scientific Cat#15596026
RNeasy Mini Kit Qiagen Cat# 74104
Formic Acid VWR CHEMICALS Cat#84865.290
Citric Acid Sigma Aldrich Cat#251275
Acetonitrile VWR CHEMICALS Cat#1.59014.2500
Phenylmethylsulfonyl fluoride Sigma Aldrich Cat# 10837091001
3, 3 –diaminobenzidine (DAB) Sigma Aldrich Cat#D8001
Cytochrome c Sigma Aldrich Cat#C2506
Sodium succinate Sigma Aldrich Cat#S2378
Catalase Sigma Aldrich Cat#C9322
Nitro Blue Tetrazolium Sigma Aldrich Cat#N5514
MitoTracker Red CMXRos Thermo Scientific Cat# M7512
VECTASTAIN Elite ABC-Peroxidase Kit Vector Laboratories Cat# PK-6101; RRID:AB_2336820
Antibody Diluent, Background Reducing Agilent, Dako Cat# S3022
VECTASHIELD Antifade Mounting Medium with DAPI Vector Laboratories Cat# H-1200
Champion pET SUMO Expression System Thermo Scientific Cat# K30001
HisTrap FF Crude, 5 GE Healthcare Cat# 11000458
HisTrap FF 1ml GE Healthcare Cat# 17531901
HiTrap Desalting columns with Sephadex G-25 resin GE Healthcare Cat# 17140801
HiTrap DEAE Sepharose FF GE Healthcare Cat# 17515401
Superdex 75 10/300 GL Ge healthcare Cat# 17517401
SUMO protease Thermo Scientific Cat# 12588018
Bio-Beads SM-2 resin Bio-Rad Cat# 1523920
Critical Commercial Assays
Mouse/Rat FGF-21 Quantikine ELISA Kit R&D Systems Cat#MF2100
Maxima First Strand cDNA Synthesis Kit for RT-qPCR Thermo Scientific Cat#K1671
Deposited Data
Human muscle biopsy RNA-Seq data Gene Expression Omnibus
data repository
GEO: GSE129811
Experimental Models: Cell Lines
Mouse myoblasts C2C12 ATCC CRL-1772
Human diploid myoblast male control line (established in-lab) Suomalainen Lab #2572
Human diploid fibroblasts male control line (established in-lab) This study #3532
Experimental Models: Organisms / Strains
Tg/ACTB/twnk-p.353-365-dup/BL6 Suomalainen Lab N/A
Fgf21LoxP/LoxP Steven A. Kliewer (Potthoff et al., 2009)
Tg(Pgk1-cre)1Lni The Jackson Laboratory B6.C-Tg(Pgk1-cre)1Lni/CrsJ, Stock No:020811
Oligonucleotides
See Table S6 for primer sequences This paper N/A
siRNA Control 1/2 ThermoFisher Cat#s4390843/46
siRNA ATF3 ThermoFisher Cat# s1699/s1701
siRNA ATF4 ThermoFisher Cat#s1703/04
siRNA ATF5 ThermoFisher Cat# s22424/25
(Continued on next page)
Cell Metabolism 30, 1–15.e1–e7, December 3, 2019 e2
Please cite this article in press as: Forsstro¨m et al., Fibroblast Growth Factor 21 Drives Dynamics of Local and Systemic Stress Responses in Mito-
chondrial Myopathy with mtDNA Deletions, Cell Metabolism (2019), https://doi.org/10.1016/j.cmet.2019.08.019
LEAD CONTACT AND MATERIALS AVAILABILITY
Lead Contact
Requests and detailed information of reagents should be directed to and will be made available by the Lead Contact, Anu Suoma-
lainen (anu.wartiovaara@helsinki.fi).
EXPERIMENTAL MODEL AND SUBJECT DETAILS
Ethical Approval
The National Animal Experiment Review Board and Regional State Administrative Agency for Southern Finland approved the animal
experimentation, following the European Union Directive. The human materials were collected and used with informed consent ac-
cording to the Helsinki Declaration and approved by the Ethical Review Board of Helsinki University Central Hospital.
Animal Models
The Deletor mice expressed a dominant in-frame duplication, homologous to a human mitochondrial myopathy mutation. Two pre-
viously reported transgene lines were used in this study. We generated the Deletor mouse with Fgf21 knockout by utilizing the Cre-
Lox recombination system (Figure S3A). The heterozygous Twinkledup 353-365 (Deletor) males and PGK-Cre females were first crossed
for two generations with Fgf21LoxP/LoxP. The Deletor x Fgf21LoxP/LoxP males were then crossed with PGK-Cre x Fgf21LoxP/LoxP fe-
males. PGK-Cre has ubiquitous expression. All genotypes were born in the expected Mendelian ratios and were viable. An equal
percentage of animals (80-86%) in all study groups survived until euthanizing at 22-24months of age. The population wasmaintained
with crosses between Deletor-FKO males and FKO females. Male mice were used for all analyses because of slightly more pro-
nounced findings of mitochondrial myopathy in Deletor males than females.
Cell Lines
The human fibroblasts cell line was established from a healthy 45-years-old male Caucasian donor via skin biopsy using routine pro-
tocols. Primary low passaged human myoblasts were from a healthy voluntary 45-years-old male Caucasian donor (Ahola et al.,
2016). Murine C2C12 myoblasts were obtained from ATCC.
MEF Isolation
Embryonic fibroblasts were cultured from E13.5 WT and FGF21KOmouse embryos. Embryos were collected in phosphate-buffered
saline (1xPBS, without calcium and magnesium ions) and the head and liver were separated. Remaining tissue material was cut into
small pieces and homogenized in 5% trypsin in 1xPBS (1ml/embryo). This suspension was incubated at +37C and mixed by pipet-
ting every 5 minutes. 1ml of this suspension was added to 9 ml of DMEM media supplemented with GlutaMax (Gibco), 10% FBS,
amino acids (MEM Non-Essential Amino Acids Solution, Thermo Fisher) and penicillin/streptomycin. Cells were plated on gelatin-
coated dishes. Confluent dishes were split the following day and harvested on day 3-4 by trypsinization for further analyses.
Cell Culture
Human myoblasts from a healthy 45-year-old male Caucasian donor (Khan et al., 2017) were grown in 420 ml Ham‘s F10 with
GlutaMax (Gibco 41550-021) supplemented with 75 ml of heat-inactivated fetal bovine serum (FBS, Sigma, F9665), 250 mg bovine
serum albumin (BSA, Sigma #A-4503), 250 mg fetuin (Sigma #F-2379), 90 mg insulin (Sigma #I-1882), 5 ml of 39 mg/ml dexameth-
asone (Sigma, #D1756), 500 ml epidermal growth factor (10 mg/ml, BD Biosciences, 354052), 10 mg uridine (Calbiochem, 6680).
All other cells were grown in Dulbecco’s Modified Eagle’s Medium DMEM (Sigma-Aldrich), with 4.5 g/l glucose supplemented
with 10% fetal bovine serum, 20 mM pyruvate, 100 units/ml of penicillin/streptomycin and 50 mM uridine. All cells were maintained
in a humidified chamber at 37C at 5% CO2.
Continued
REAGENT or RESOURCE SOURCE IDENTIFIER
Software and Algorithms
AIDA image analysis software Elysia Raytest N/A
Fiji / ImageJ software NIH N/A
STAR 2.5.0a GenSoft N/A
HTSeq 0.10.0 Simon Anders N/A
DESeq2 1.18.1 package Bioconductor N/A
Prism v.7 GraphPad Software N/A
Other
Mouse diet Altromin Spezialfutter Cat#C1000
e3 Cell Metabolism 30, 1–15.e1–e7, December 3, 2019
Please cite this article in press as: Forsstro¨m et al., Fibroblast Growth Factor 21 Drives Dynamics of Local and Systemic Stress Responses in Mito-
chondrial Myopathy with mtDNA Deletions, Cell Metabolism (2019), https://doi.org/10.1016/j.cmet.2019.08.019
Respiratory Function Testing in Animals
Oxygen consumption and carbon dioxide production were measured in Oxymax Lab Animal Monitoring System (CLAMS; Columbus
Instruments). In CLAMS the mice were housed in individual cages in temperature-regulated chamber, settled for 24 hours and were
recorded for 2 days. The settling and the first day of recording were in room temperature (+22C) followed by 24 hours at thermo-
neutral conditions (+30C). Respiratory exchange ratio (RER) was calculated to indicate a rough estimate of the preferred fuel (car-
bohydrate breakdown vs. fatty acid oxidation) of the organism.
Tracer Assays
Pre-conditioning
For fasting experiments food was removed after initial feeding period at 7-9 pm (beginning of dark period) and tracer uptake took
place in the following morning after 24 hours of fasting.
For FGF21 injection experiments, animals were intraperitoneally injected 4 hours prior to tracer injection with either saline placebo
or recombinant mouse FGF21 (see details for production in method details). The injected dose for recombinant FGF21 was
0.06 mg/kg.
Rapamycin treatment wasmade according to Khan et al. (2017). Rapamycin (LC Laboratories #R-5000) was dissolved in DMSO to
yield stock of 100mg/ml, further diluted in 5%PEG-400 to final concentration of 1.2 mg/ml. Solution was sterile filtered and stored at
80C prior to use. Before administration, the mice were weighed and the dose was adjusted to 8 mg/kg/day. Mice were injected
intraperitoneally for 6 weeks.
Assay Protocol
Prior to all tracer uptake assays, food was removed from cages for 3 hours. The animals were kept in anesthesia with small dose of
inhaled isoflurane (1-1,5% in oxygen or room air carrier) on temperature-controlled mats throughout the procedure. Glucose analog
18F-FDG (18F-Fludeoxyglucose; total activity injected 5MBq) or fatty acid analog 18F-FTHA (18F-Fluoro-6-thia-heptadecanoic acid;
total activity 5MBq) was administered intravenously. The uptake periods for 18F-FDG and 18F-FTHAwere 60 and 30minutes, respec-
tively. Biodistribution of tracers was assayed by detecting gamma rays emitted from excised tissues with manual gamma counter.
Standardized uptake values (SUV) in different tissues were calculated for each sample. Quantitative phosphor imaging autoradiog-
raphy of 18F-FDG in brain was performed after 60 minutes of in vivo uptake period. Coronal brain cryosections (20 mm) were cut on
objective glass and exposed on imaging plate (plate (Fuji BAS-TR2025) for approximately 3.5 hours. The plates were scanned with
Fuji BAS-5000 Phosphor imager with 25 mm resolution (digital autoradiography, ARG). The ARG images were analyzed for count den-
sities (photostimulated luminescence per unit area, PSL/mm2) with a computerized image analysis program (AIDA image analyzer).
Cerebellar cortex was selected to be the reference brain region for signal normalization as it showed equal uptake in all study groups
in ex vivo tissue-panel assay.
METHOD DETAILS
Isolation of Mitochondria
Mitochondria were isolated from quadriceps femoris (QF) fromDeletors to enrich for mitochondrial protein. QFwas homogenized in a
Potter-Elvejhemwith a rotor-driven pestle in HIM buffer (10 mMHEPES-KOH, pH7.6, 100mMKCl, 3 mMMgCl2, 0.1 mMEDTA, 10%
glycerol, 0.1% fatty-acid free BSA) supplemented with 1 mM PMSF and protease inhibitors (Sigma). One QF was lysed in 5 ml HIM
buffer, strained with a 100 mm cell strainer and sieve rewashed with 5 ml homogenization buffer. Enriched mitochondrial fractions
were obtained by differential centrifugation at a low step at 700xg for 20 min and resulting supernatant at 10‘000xg for 10 min
and was repeated. All steps were performed at 4C.
MtDNA Deletion Load and Copy Number Determination
Total cellular DNA was isolated from frozen quadriceps femoris (QF) by standard proteinase K and phenol–chloroform method. For
determination mtDNA deletion load we used long-range PCR reactions for deleted and undeleted fractions of mtDNA (Ahola-Erkkila
et al., 2010) using primers in non-deleted region between 16S rRNA andND1 to enable partly deletedmtDNAmolecules. 10 ng of total
genomic DNA was amplified with primers for long template PCR using Phusion high-fidelity DNA polymerase (Thermo Fisher) and
supplied GC buffer PCR reaction was initial denaturation for 30 s at 98C, 22 cycles of 10 s at 98C and 12 min extension at 72C
and a final extension step for 10 min at 72C. For total mtDNA amount determination a non-deleted mtDNA fragment was amplified
using Phusion high fidelity DNA polymerase and GC buffer: initial denaturation of 30 s at 98C; 21 cycles of 10 s at 98C and 2 min at
72C; final extension step of 10min at 72C. All PCR products were agarose gel (1%) electrophoresed and imaged using ChemiDoc
XRS+ System (Bio-Rad). Amount of the total mtDNA product was compared to the deleted mtDNA products.
Alternatively, mtDNA deletion load was determined by a modified a triplex assay (Rygiel et al., 2015) adapted for mice. The mouse
triplex assay was used for single-well detection using 5‘-modified primers with ROX (D-Loop probe), HEX (ND1 probe) or FAM (ND4
probe) for single-well detection. Primer efficiency and linear range of amplification was determined by serial dilution. A standard curve
was included in every run. All primer sequences can be found from Table S6.
Cell Metabolism 30, 1–15.e1–e7, December 3, 2019 e4
Please cite this article in press as: Forsstro¨m et al., Fibroblast Growth Factor 21 Drives Dynamics of Local and Systemic Stress Responses in Mito-
chondrial Myopathy with mtDNA Deletions, Cell Metabolism (2019), https://doi.org/10.1016/j.cmet.2019.08.019
Gene Expression Analysis
RNAwas extracted by standard Trizol-chloroform precipitation for tissues and by using the Qiagen miRNA kit for cells. DNase diges-
tion was performed and total RNA reverse transcribed using theMaxima first strand RT (ThermoFisher). 10-20mg of tissuewere used
for homogenization. Total cellular RNA was extracted from snap frozen tissues in TRIzol reagent (Invitrogen) and homogenized with
Fast-Prep w-24 LysingMatrix D (MPBiomedical) and Precellys w-24 (Bertin Technologies). RNA of cultured cells was directly lysed in
TRIzol and extracted. RNA concentration was measured with NanoDrop1000 Spectrophotometer and integrity checked in agarose
gel electrophoresis. Twomicrograms of RNA was DNAse treated (Ambion). Onemicrogram of total RNA was used to generate cDNA
using Maxima first-strand cDNA synthesis kit (Thermo Scientific). Quantitative real-time PCR amplification of cDNA was performed
with IQ SybrGreen kit (Bio-Rad) on CFX96 Touch qPCR system (Bio-Rad). Relative expressions were normalized to b-actin or 18S
rRNA as indicated. All primer sequences can be found from Table S6.
Gene Silencing
Gene silencing was performed using SilencerSelect (ThermoFisher) siRNAs for control siRNA 1 (s4390843) & 2 (s4390846) and
against ATF4 (s1703, s1704), ATF3 (s1699, s1701), ATF5 (s22424, s22425). Briefly, two consecutive rounds of 25 picomoles of
siRNAs per 6 cm2 dish were used to transfect human fibroblasts on day 1 at 50% confluency and day 4 in order to achieve maximum
knockdown efficiency using Polyplus jetPRIME transfection kit. Transfection efficiency was confirmed by RT-qPCR.
Western Blotting
Total protein from cells was extracted using RIPA buffer supplemented with phosphatase inhibitors (ThermoFisher Scientific) and
sodium orthovanadate in order to preserve phosphorylation sites. Briefly, tissues (20 mg) were homogenized in 20-30 volumes of
1XTBSwith bead homogenizer (5000 beats for 20s) (Precellys) and Triton-X was added to yield 1% volume of the extract. Themixture
was incubated on ice for 30min and collected by centrifugation at 14,000xg for 20min. Protein concentrations were determined using
Bradford assay. Proteins were separated in appropriate gels by SDS-PAGE usually on 4-20% gradient gels, transferred to polyviny-
lidene difluoride (PVDF) membranes and blocked with 5% milk or 3% BSA in 1xTBS-T (Tris Buffered Saline with 0.1% Tween20) for
1 h and primary antibodies incubated overnight at 4C, Ab concentration range according to manufacturer’s instructions (1:250-
1:1000) in 3% BSA in TBS-T buffer. Antibodies used are listed in key resources table. Secondary HRP antibodies were used in
1:10000 dilution in 5%milk in TBS-T and signal detected by Clarity Western ECL Substrate (Biorad) with ChemiDoc XRS+ System
(Biorad). Signal intensities were quantified by ImageLab 5.2.1 (Biorad).
Histology
Respiratory Chain Enzyme Activity and Oil-red-O, Fresh-frozen Tissues
Tissues were freshly embedded in OCT medium (Tissue-Tek) and frozen in 2-methylbutane bath under liquid nitrogen cooling.
Simultaneous histochemical activity analysis of cytochrome-c-oxidase (COX) and succinate dehydrogenase (SDH) activities was
performed on QF and brain slices (12 mm), incubation times being 30 minutes for COX (RT) and 40 minutes for SDH (+37C). Buffer
contents for COX: 0.05 M phosphate buffer (pH 7.4) with 3, 3 –diaminobenzidine (DAB), catalase, cytochrome c and sucrose and for
SDH: 0.05Mphosphate buffer (pH 7.4) with nitro-blue tetrazolium and sodium succinate. After stainings the sliceswere dehydrated in
ascending alcohols, xylene-treated andmounted. Frommuscle sections, COX-negative/SDH-negative, COX-negative/SDH-positive
and normal fibers were quantified.
For Oil-Red-O staining 8 mm sections of frozen liver were fixed in formic calcium and incubated in Oil-Red-O solution for 10 min
followed by brief hematoxylin staining for nuclei.
Immunohistochemistry in Tissues
Tissues were fixed in 10% buffered formalin and embedded in paraffin. Citric acid antigen retrieval and H2O2 treatment were
performed. The samples were blocked for non-specific staining by incubating for 30 min in 2% horse serum. Primary antibody
incubations were performed overnight at 4C. Detectionwas donewith Vectastain Elite ABCHRPKit according to themanufacturer’s
instructions followed by chromogen DAB staining. The nuclei were briefly counterstained with hematoxylin. Antibodies: SDHA
(Complex II) Abcam, ab14715 (1:500); UCP1 Abcam, ab10983 (1:500).
Immunofluorescence in Tissues
Tissues were fixed in 10%buffered formalin and embedded in paraffin. Citric acid antigen retrieval was performed. The samples were
blocked for non-specific staining by incubating for 30 min in 2% horse serum. Incubation with antibodies was done in Antibody
diluent (Dako), primary antibody incubation for overnight at 4C and secondary in room temperature with appropriate Alexa Fluor
antibodies (from Abcam). Antibodies: GFAP from Millipore, AB5804 (1:500); IBA1 from Wako, 019-19741 (1:1000); MAP2 from
Abcam, ab5392 (1:500) and TOM20 from Santa Cruz, sc-11415 (1:500).
Adipocytes
Formalin-fixed paraffin-embedded sections were stained with hematoxylin and eosin.
Staining in Cells
Living cells grown on cover slips were stained for 15 min in 50 nM Mitotracker CMXRos in full media, rinsed with 1xPBS and fixed in
ice-cold acetone or directly fixed in ice-cold acetone for immunofluorescence stainings. Fixed samples were blocked (5% horse
serum, 1% BSA, 0.1% Triton-X-100 in 1xPBS) and incubated with the primary antibodies against TOM20 (SantaCruz, #11415,
1:250) and ATF4 (SantaCruz B-3, sc390063) overnight. Secondary anti-rabbit Alexa488 was used for TOM20 and anti-mouse
e5 Cell Metabolism 30, 1–15.e1–e7, December 3, 2019
Please cite this article in press as: Forsstro¨m et al., Fibroblast Growth Factor 21 Drives Dynamics of Local and Systemic Stress Responses in Mito-
chondrial Myopathy with mtDNA Deletions, Cell Metabolism (2019), https://doi.org/10.1016/j.cmet.2019.08.019
Alexa594 for ATF4. All samplesweremounted in VectaShield (Vector Laboratories) containing DAPI to counterstain for nuclei. Images
were recorded on a Zeiss AxioImager 1.2 equipped with an ApoTome.
Electron Microscopy
Attached cells were directly fixed in 2% glutaraldehyde in phosphate buffer, post-fixed in 1% osmium tetroxide and dehydrated
through ascending concentrations of alcohol and embedded in Epon812. 60nm ultrathin sections were obtained on a Reichert-
Jung Ultracut ultramicrotome equipped with Diatome (Switzerland) diamond knife, transferred to copper grids, stained with uranyl
acetate and lead citrate and observed in a Philips CM12 electron microscope at 80kV. Images were recorded with a Morada digital
camera and analyzed using iTEM software.
Quantification of FGF21
Terminal blood samples were collected via heart puncture, were let to coagulate for 15 minutes at room temperature and centrifuged
at 4C at 3,000x g for 15 minutes. Separated serum was frozen and stored in -80C prior to assay. Mouse/Rat FGF21 Qusntikine
ELISA (R&D Systems) was used according to manufacturer’s instructions. Absorbance measurement was made with SpectraMax
190 absorbance microtiter plate reader.
Metabolomics
Metabolomic analyses were performed from 20mg of muscle extract usingWaters Acquity ultra performance liquid chromatography
(UPLC) and triple-quadrupole mass spectrometry analysis in FIMM (Institute for Molecular Medicine Finland). All metabolite stan-
dards, ammonium formate, ammonium acetate and ammonium hydroxide were obtained from Sigma-Aldrich (Helsinki, Finland). For-
mic acid (FA), 2-proponol, acetonitrile (ACN), and methanol (all HiPerSolv CHROMANORM, HPLC grade, BDH prolabo) were
purchased from VWR International (Helsinki, Finland). Isotopically labelled internal standards were purchased from Cambridge
Isotope Laboratory. The exact procedure is described and validated in Nandania et al. (2018a), (2018b). The polar metabolome
was extracted from frozen 20 mgmouse muscle and lysed with a Precellys homogenizer in 1.4 mm ceramic (Zirconium oxide) beads
containing tubes including 10 ml of labeled internal standard mix and 1:20 sample: extraction solvent in a two-step extraction proto-
col. The first step contained 10 parts of 100% ACN + 1% FA for 3 cycles at 5500 rpm with 30s pausing in-between. Sample was
centrifuged at -20C at 2 min in an Eppendorf 5404R centrifuge at 5000 rpm and supernatant collected. The remaining pellet was
re-lysed with 10 parts of 80/20%ACN/H2O + 1% FA as described above. Calibration solutions were serially diluted in a 96-well plate.
Samples were analyzed on an ACQUITY UPLC-MS/MS system (Waters Corporation, Milford, MA, USA) with specific settings re-
ported in Nandania et al. (2018b). Detection was performed on a Xevo TQ-S tandem triple quadrupole mass spectrometer (Waters,
Milford, MA, USA), was operated in both positive and negative polarity. Samples were ionized by electro spray ionization (ESI) and
dwell time and subsequent data acquisition were automatically calculated by MassLynx 4.1 software and processed using
TargetLynx software. Quantification was performed by internal standards and external calibration curves.
dNTP Pools
Total dNTP pool measurements were performed and analyzed as described (Martı´ et al., 2012; Nikkanen et al., 2016). Approximately
50-100 mg of frozen muscle tissue was homogenized in 10 volumes (v/w) homogenization buffer (10 mM Tris/HCl, pH 7.5, 0.2 mM
ethylene glycol tetra-acetic acid (EGTA), 0.5 % bovine serum albumin (BSA), 210 mMmannitol, 70 mM sucrose) using a polytron ho-
mogenizer. Obtained homogenate was centrifuged for 10min at 4C. Coldmethanol (-20C) was added to the resulting supernatants
for a final concentration of 60% (v/v) and stored at -80C for 3h. Samples were then heated to 95C for 3 min, cooled on ice and
centrifuged at 16000 x g at 4C for 20 min. Obtained supernatants were collected and dried down. Appropriate amounts of re-hy-
drated extracts were added to the reaction mix containing 40 mM Tris/HCl, pH 7.5, 10 mM MgCl2, 5 mM dithiothreitol (DTT) and
0.25 mm specific primed oligonucleotide, 0.75 mm 3H-dTTP or 3H-dATP, 0.30-unit Taq DNA polymerase for a total of 20 ml and incu-
bated for 60 min at 48C. 15 ml of reaction mix was spotted onto DE-81 filter paper and dried. Spotted samples were then washed 3x
with 5% Na2HPO4, 1x water and 1x with 95% ethanol. Reaction products were analyzed using a liquid scintillation counter with 3-6
measurements per individual sample and quantified compared to standard curves.
RNA Sequencing of Patient Muscle
ABergstro¨m needle biopsy from vastus lateraliswas collected from eight controls and from four PEO patients. Total muscle RNAwas
extracted with standard TRIzol and chloroform method and purified with RNA purification kit (RNeasy; Qiagen). A total of one micro-
gram of total RNA was used for global transcriptomics analysis performed by the Beijing Genomic Institute (BGI) using their standard
protocols. In brief, transcriptome sequencing was performed using 50bp paired-end sequencing on an Illumina sequencing platform.
The resulting raw readswere preprocessed by the sequencing provider to filter out reads failing criteria for quality, unknown base and
adaptor sequence content. Clean reads, specified in FASTQ format, amounted to approximately 26 million reads per sample. Sub-
sequent analysis was performed in-house and reads were aligned against the GRCh38 Human genome assembly (DNA primary as-
sembly in FASTA format and gene annotation in GTF format from Ensembl release 92) using STAR 2.5.0a, giving read alignments in
BAM format. Counting of readsmapping to genes was performed with htseq-count, HTSeq 0.10.0, in union overlap resolution mode.
The obtained read counts were then used for differential gene expression analysis with the Bioconductor DESeq2 1.18.1 package.
Fold changes in binary logarithmic scale and respective p-values were extracted for pairwise comparisons.
Cell Metabolism 30, 1–15.e1–e7, December 3, 2019 e6
Please cite this article in press as: Forsstro¨m et al., Fibroblast Growth Factor 21 Drives Dynamics of Local and Systemic Stress Responses in Mito-
chondrial Myopathy with mtDNA Deletions, Cell Metabolism (2019), https://doi.org/10.1016/j.cmet.2019.08.019
Production of Recombinant Mouse FGF21
Cloning and Expression
Champion pET SUMO Protein Expression system was used. BL21(DE3) E. coli cells with pET-SUMO-FGF21 expression vector were
grown at +32C in Luria-Bertani media until the OD600 was 0.6-0.8. Then, 0.3 mM of isopropyl-D-thiogalactoside (IPTG) was added
to induce expression of the fusion protein and cells were grown for another 4 hours. Cell cultures were pelleted at 8000 rpm for 10min
at +4C, washed with 25mM Tris-HCl, pH7.5, 200mM NaCl, centrifuged for 15 min at 5000rpm, +4C and frozen at -20C.
Purification
Whole purification procedure was performed at +4C. All chromatography steps were done on AktaPurifier 10 (GE Healthcare). A cell
pellet harvested from 1 L of culture was melted, suspended in 50ml of 20mM NaH2PO4, pH7.6, 150 mM NaCl, 0.5mM PMSF and
disrupted by sonication. Cell lysate was ultra-centrifuged at 40,000 rpm for 30 min and supernatant was collected. NaCl and imid-
azole were added to yield final concentrations of 500mMand 20mM, respectively. Protein solutionwas filtered through 0.45mmPVDF
filter (Millipore) and loaded at 3 ml/min flow rate onto HisTrap Crude FF 5ml column (GE Healthcare) equilibrated with 20mM
NaH2PO4, pH 7.6, 500 mM NaCl containing 20 mM imidazole. After washing with 10 x column volume of 20 mM imidazole and
5 x column volume of 50 mM imidazole, the his-tagged sumo-FGF21 fusion was eluted with 200 mM imidazole in the equilibration
buffer. Eluate was concentrated with 10 kDa cut-off Vivaspin concentrators (Millipore), protein was desalted against 20mM
NaH2PO4, pH 7.6 using HiTrap Desalting column 5ml (GE Healthcare) at 3 ml/min flow rate. Then the protein was loaded at 3ml/min
flow rate onto HiTrap DEAE FF 5ml column (GE Healthcare), equilibrated with 20mM NaH2PO4, pH 7.6. After washing with 120mM
NaCl, the protein was eluted with 200 mM NaCl in the equilibration buffer. Pooled fractions with eluted protein were concentrated to
1ml in Vivaspin concentrators with 10kDa cut-off (Millipore) and desalted on HiTrap Desalting column 5ml (GE Healthcare) against
5 mM TrisHCl, pH 8.0, 155 mM NaCl. Cleavage of sumo tag was performed using SUMO protease (Thermo Scientific) according
to manufacturer’s instructions. After removal of NP-40 with BioBeads (Bio-Rad) followed by adjustment of NaCl concentration to
500mMand imidazole concentration to 20mM the protein solution was loaded at 0.5ml/min onto HisTrap FF 1ml column (GEHealth-
care) equilibrated with 20mMNaH2PO4, pH 7.6, 500 mMNaCl, 20 mM imidazole. Flow through containing FGF21 was concentrated
to 400ml in Vivaspin concentrators and loaded onto Superdex 75 10/300 GL column equilibrated with PBS, pH7.4. Gel filtration was
run at 0.2 ml/min. Fractions containing FGF21 were pooled and concentrated. Protein concentration was determined using Bradford
method with bovine serum albumin as a standard.
QUANTIFICATION AND STATISTICAL ANALYSIS
Statistical analyses and their graphical representation were performed in GraphPad Prism v.7.0 software (GraphPad Software, USA).
Statistical test used in each experiment is indicated in figure legend. P-values of less than 0.05 were considered statistically signif-
icant. Outlier analysis for all analyses was done using GraphPad Prism 7.0 ROUT method (Q=1%) and false positive analysis for me-
tabolomics with Benjamin-Hochberg method, critical value of 0.2.
DATA AND CODE AVAILABILITY
All data generated or analyzed during this study are included in this published article and its supplementary information files. Mate-
rials used in this study are available from the corresponding author on reasonable request. Human muscle biopsy RNA-Seq data
deposited to Gene Expression Omnibus data repository with accession number GEO: GSE129811.
e7 Cell Metabolism 30, 1–15.e1–e7, December 3, 2019
Please cite this article in press as: Forsstro¨m et al., Fibroblast Growth Factor 21 Drives Dynamics of Local and Systemic Stress Responses in Mito-
chondrial Myopathy with mtDNA Deletions, Cell Metabolism (2019), https://doi.org/10.1016/j.cmet.2019.08.019