Article reSeArcH
the absence of divalent cations (Extended Data Fig. 5k) and the pres-
ence of Na+ only (Extended Data Fig. 5l). Overall, all these features
recapitulate the fundamental electrophysiological properties and the
consensus pharmacological profile of mitoKATP^18 ,^19 , with MITOK
forming the K+-permeant channel and MITOSUR acting as a modu-
latory subunit that carries the ATP-binding site. We next investigated
the membrane topology of MITOSUR by performing a protease pro-
tection assay using an antibody that covers the ATP-binding region
(amino acids 394–693). Extended Data Figure 6a shows that proteinase
K leads to the loss of the MITOSUR-specific band, which indicates
that the ATP-binding cassette is exposed to the intermembrane space.
Overall, the membrane orientations of both MITOK and MITOSUR
are supported by previous bioenergetics studies^5 and independent pro-
teomic approaches^23.
In light of this architecture, we reasoned that the unregulated K+
uptake in cells that overexpress MITOK alone must be the cause of
organelle impairment (Fig. 1 ). If this is true, the combined overex-
pression of human MITOSUR and mouse MITOK should reverse the
mitochondrial dysfunction. To test this, we first verified the physical
interaction between human MITOSUR and mouse MITOK through
co-immunoprecipitation (Fig. 3a). In addition, immunoblot analysis
of digitonin-solubilized mitochondrial complexes in native conditions
revealed an approximately 500-kDa band that reacted with both
MITOK and MITOSUR antibodies (Fig. 3b)—a size that is compati-
ble with an octamer (four MITOK and four MITOSUR components).
Accordingly, immunoprecipitation of endogenous MITOK could
efficiently pull down MITOSUR using both mouse and human mito-
chondria (Fig. 3c, Extended Data Fig. 6b). In terms of function, the
overexpression of mouse MITOK alone caused a decrease of both mito-
chondrial membrane potential and Ca^2 + uptake (Fig. 3d, e), whereas
the overexpression of the MITOSUR subunit alone did not affect these
parameters. Most importantly, the combined overexpression of the two
subunits fully rescued ΔΨm and Ca^2 + dynamics (which were impaired
by mouse MITOK alone), suggesting the recovery of the proper channel
gating. We also generated a MITOSUR mutant (MITOSUR(K513A))
that is unable to bind ATP. This mutant could interact with mouse
MITOK (Extended Data Fig. 6c)—but did not respond to ATP in elec-
trophysiological experiments (Fig. 3f) and did not rescue the loss of
mitochondrial membrane potential and Ca^2 + accumulation that is
caused by the overexpression of mouse MITOK (Fig. 3d, e). This pro-
vides further confirmation that ATP acts as channel inhibitor. Overall,
our data indicate that MITOK and MITOSUR form a complex that
is responsible for the ATP-sensitive mitochondrial K+ transport both
in vitro and in situ.MITOK controls mitochondrial volume
Despite consensus regarding the cytoprotective role of mitoKATP
opening in stress conditions (which is mainly based on pharmacolog-
ical studies)^10 , the constitutive physiological function of this channel
remains obscure. We therefore generated HeLa cells that are knock-
out for MITOK by CRISPR–Cas9 DNA cleavage, using two different
guides (Extended Data Fig. 7a, b). We first confirmed that MITOK is
required for ATP-dependent K+ fluxes in mitochondria, by measur-
ing mitochondrial swelling rates in a K+-based buffer^9 ,^24. In isolated
wild-type mitochondria, ATP decreases, and diazoxide increases, the
swelling rates (through the inhibition and activation of the mitoKATP
channel, respectively) (Fig. 4a). Accordingly, mitochondria isolated
from MITOK-knockout cell lines swell at a constant rate, independently
of either ATP or diazoxide (Fig. 4b).
MITOK-knockout cells were viable and showed a highly intercon-
nected mitochondrial network by optical microscopy. Although the
gross morphology appeared similar, ablation of MITOK led to the
appearance of several doughnut-shaped (toroidal or ring-like) mito-
chondria (Extended Data Fig. 7c), a phenotype that is associated with
impaired organelle K+ homeostasis^25. ΔΨm was intact, but HeLa cells
knockout for MITOK undergo asynchronous, rapid and transient
depolarizations of single mitochondria (Extended Data Fig. 7d, e),
a phenomenon known as mitochondrial ‘flickering’ or ‘flashes’^26 –^29.
Importantly, this phenotype is specific, as shown by the fact that the
reintroduction of human MITOSUR and mouse MITOK restored ΔΨm
stability (Fig. 4c). In terms of oxidative performance, the ablation of
MITOK caused a decrease in the basal and maximal oxygen consump-
tion rates, despite the similar levels of expression of components of the
electron transport chain (Fig. 4d, Extended Data Fig. 7f). This could be
partially rescued by (i) pharmacological treatment with a minimal dose
of valinomycin (which has no effect in control cells) (Extended Data
Fig. 8a) and (ii) the reintroduction of human MITOSUR and mouse
MITOK (Extended Data Fig. 8b). To understand the causes of altered
organelle function, we investigated mitochondrial ultrastructure by
transmission electron microscopy. Although the gross mitochondrial
morphology was preserved, MITOK-knockout cells show enlarged
cristae (Fig. 4e), which is consistent with the effect of the inhibition
of another inner mitochondrial membrane K+ channel^30. Normal
morphology of cristae was readily restored by the re-expression of the
mitoKATP channel (Fig. 4e). Given that K+ fluxes across the inner mito-
chondrial membrane are the main determinants of the water content of
organelles^31 , we speculated that cristae remodelling could be due to a
dysregulation of matrix volume (as suggested by swelling experiments)
(Fig. 4a, b). The inner mitochondrial membrane rapidly rearranges inControl MITOSUR–FlagMITOK–V5FlagMITOKMITOK–V5 +
MITOSUR–FlagInputCo-IP
bait: FlagFlagMITOKab
1,236
1,048
720
48024214666MITOKMITOSUReC
O 1
O 2200 ms
2 pA
MITOK + MITOSUR(K513A)C
O 1
O 2200 ms
2 pA
MITOK + MITOSUR(K513A) + Mg/ATPMITOK +
MITOSUR(K513A)(^05)
5
10
15
20
25
30
35
–15 –10 –5 0 5 –15 –10 –5 0
Current (pA)
Events (
×^10
3 )
MITOK +
MITOSUR(K513A)
- Mg/ATP
f
0
5
10
15
20
Events (
×^10
3 )
Current (pA)
25
0
0.5
1.0
1.5
2.0
2.5 *
NS
Basal TMRM
uorescence (AU)
Control
MITOK
MITOSUR
MITOK +
MITOSUR
0
20
40
60
80
100
120
5 s
Histamine
100 μM
[Ca
2+
]mt
(μ
M)
- NS
Control MITOK MITOSUR MITOK +MITOSUR MITOSUR(K513A) MITOK +MITOSUR(K513A)
100
kDa
75
50
35
50
Input (mitochondria)Co-IP: IgGCo-IP: MITOK
MITOSUR
MITOK
c
d
50
75
75
kDa kDa
50
Fig. 3 | MITOK and MITOSUR form the mitoKATP channel in situ.
a, C o-immunoprecipitation (co-IP) between overexpressed MITOK and
MITOSUR (representative of three independent experiments). b, Blue-
native PAGE of digitonin-permeabilized mitochondria. Representative of
two independent experiments. c, Co-immunoprecipitation of endogenous
MITOK using liver mitochondria. Representative of two independent
experiments. d, ΔΨm measurements in HeLa cells transfected with the
indicated constructs. n ≥ 9 biological replicates from 3 independent
experiments, P ≤ 0.01 using two-way analysis of variance (ANOVA)
with Holm–Sidak correction NS, not significant. AU, arbitrary units;
TMRM, tetramethylrhodamine methyl ester. e, Measurements of Ca^2 +
concentration in mitochondria ([Ca^2 +]mt) (mean ± s.d.) in HeLa cells that
express the indicated constructs; n = 8 biological replicates (representative
of 3 independent experiments), P < 0.001 using two-way ANOVA with
Holm–Sidak correction. f, Current traces (left) and histograms (right) of
MITOK together with MITOSUR(K513A), before and after the addition of
2 mM Mg and ATP. Representative of three independent preparations.
29 AUGUSt 2019 | VOl 572 | NAtUre | 611