Nature - 2019.08.29

reSeArcH Article

single band can be detected in western blots using HeLa cells (Extended

Data Fig. 1g). In most human tissues, isoform 1 shows moderate levels
of expression, whereas isoform 2 is barely detectable. This pattern is

reversed in the case of the spleen, which shows substantial expression
of isoform 2 and marginal expression of isoform 1 (Extended Data

Fig. 1f). In terms of their roles, isoform 1 localizes to mitochondria
and—similar to mouse MITOK—its overexpression induced morpho-

logical and functional organelle impairment (Extended Data Fig. 2a, b),
which indicates that both mouse and human MITOK genes encode

similar proteins (although some human cells express a shorter and
less-active splicing variant). Overall, overexpression of MITOK causes a

severe perturbation of mitochondrial structure and function. Although
several mechanisms could account for these effects, we reasoned that

MITOK could act as a cation channel, because valinomycin (a K+ ion -
ophore, a molecule that mediates K+ influx into the matrix) closely

mimics the phenotype observed on MITOK overexpression.
The unambiguous demonstration of channel activity necessarily

requires a simplified reconstitution approach that uses recombinant
proteins. We thus measured the channel activity of mouse MITOK in

the planar lipid bilayer using MITOK from two systems (Escherichia
coli and the wheat-germ cell-free transcription and translation tool),

both of which express MITOK at high levels (Extended Data Fig. 3a, b).
We observed channel activity in a medium that contained only K+

as cation (Fig. 1g–h). Burst-like, flickering activity and cooperative
transitions between dual- or multi-states were observed^4 ,^17 (Extended

Data Fig. 3c, d). The channel showed an ohmic behaviour (Extended
Data Fig. 3e), was voltage-independent (Extended Data Fig. 3f) and

was selective for K+ over chloride (PK:PCl = 1:0.02) (Extended Data

Fig. 3g). Channel conductance was 57  ±  11  pS in both 100  mM KCl

and K-gluconate media^18 ,^19 (Fig. 1h, Extended Data Fig. 3e). Channel

activity could be blocked by addition of barium, an inhibitor of K+

channels (Extended Data Fig. 3h), but not by paxilline, an inhibitor of

the BKCa channels (Extended Data Fig. 3i).

MITOK and MITOSUR form the mitoKAT P channel

On the basis of these data, we pursued the hypothesis that MITOK

could be mitoKATP—despite the fact that (i), similar to the lysosomal K+

channel TMEM175^20 , MITOK does not contain the typical K+ selectiv-

ity filter (and, accordingly, allows permeation of Na+) (Extended Data

Fig. 4a) and (ii) the purified protein per se did not respond to ATP

(Extended Data Fig. 4b) or 5-HD (Extended Data Fig. 4c). However,

we reasoned that ATP sensitivity could be conferred by a regulatory

sulfonylurea-receptor (SUR)-like subunit. Ten ATP-binding cassette

(ABC) proteins can be detected in mitochondria^15 , most of which belong

to ABCB subfamily^21. We focused on ABCB8 (which we hereafter

name MITOSUR) because (i), of the ten ABC proteins in mitochondria,

the tissue expression of this protein best correlates with MITOK, and (ii)

it has previously been suggested to be part of mitoKATP^22. We expressed

in vitro mouse MITOK together with MITOSUR; these were folded

and incorporated into liposomes, as indicated by thermal stability

assay (Extended Data Fig. 5a–c) and membrane extraction (Extended

Data Fig. 5d), with a membrane orientation that resembled that in

mitoplasts (Extended Data Figs. 5e, f, 6a). MITOK and MITOSUR

were able to form a K+ permeable channel (Fig. 2a, Extended Data

Fig. 5g–i) that was (i) inhibited by millimolar concentrations of

ATP, (ii) activated by diazoxide (Fig. 2a, Extended Data Fig. 5g, h)

and (iii) blocked by both the sulfonylurea glibenclamide (Fig. 2b) and

5-HD (Fig. 2c). The channel conductance slightly decreased upon addi-

tion of 1  mM Mg^2 + (Extended Data Fig. 5j). Activity was observed in

c

TM1 TM2

NH 3

COOH

IMM

Matrix

IMS

Anti-MITOKC-term

Anti-MITOKN-term

d

e Control MITOK

MITOK HSP60 Merge

f

200 ms

5 pA

C

O 1

O 2

O 3

+20 mV

MITOK in E. coli

200 ms

5 pA

C

O 1

+20 mV

MITOK in vitro

g

HomogenateMitochondria Mitoplasts OMM + IMS

MITOK

MCU

VDAC2

GAPDH

35

50

35

35

kDa

b

50 aa

a CC TM mmMITOK TM

CC TM hsMITOK-iso1

CC TM hsMITOK-iso2 TM

TM

Fig. 1 | Biochemical and functional characterization of MITOK.
a, Representation of human (hs) and mouse (mm) MITOK proteins.
Transmembrane (TM) and coiled-coil (CC) domains are indicated. iso1,
isoform 1; iso2, isoform 2. b, Immunolocalization of MITOK (green)
and the mitochondrial marker HSP60 (red). Representative of four
independent experiments. Scale bar, 10  μm. c, Subcellular fractionation
of mouse liver (replicated twice). IMS, inter-membrane space; OMM,
outer mitochondrial membrane. d, Representation of MITOK membrane
topology. C-term, C terminus; IMM, inner mitochondrial membrane;
N-term, N terminus; anti-MITOK, antibody raised against the indicated
region of MITOK. e, Transmission electron microscopy images of control
and MITOK-overexpressing HeLa cells (replicated three times).
f, g, R epresentative current traces with MITOK purified from E. coli
(f, n = 5 biological replicates from 2 independent preparations) or expressed
in vitro (g, n = 23 biological replicates from 10 independent preparations).

a

500 ms

5 pA C

O 1

O 2

O 3

C

O 1

O 2

O 3

500 ms

5 pA

C

O 1

O 2

O 3

500 ms

5 pA

–60 mV

–60 mV

–60 mV

C

O 1

O 2

O 3

500 ms

5 pA

–60 mV

Control + Mg/ATP (0.5 mM)

+ Mg/ATP (1 mM) + Mg/ATP (1 mM)+ Diazoxide (80 μM)

C

O 1

(^) O 2
200 ms
5 pA
–40 mV
Control
C
O 1
O 2
200 ms
5 pA
–40 mV

• Glibenclamide
  C
  O 1
  O 2
  200 ms
  5 pA
  –40 mV


• Glibenclamide


• Diazoxide
  –20 –10 0
  0
  2
  4
  6
  Events (
  ×^10
  3 )
  Current (pA)
  Control
  –20 –10 0
  0
  4
  8
  12
  16
  20
  Events (
  ×^10
  3 )
  Current (pA)


• 5-HD
  bc
  Fig. 2 | Electrophysiological characterization of recombinant MITOK
  co-expressed with MITOSUR. a, Current traces before (control) and
  after the first addition of 500  μM Mg and ATP, the second addition of
  500  μM Mg and ATP and the third addition of 80  μM diazoxide. All
  traces were obtained from the same experiment, representative of four
  independent experiments. b, Current recordings before and after addition
  of 30  μM glibenclamide. The channel was re-activated by subsequent
  addition of 100  μM diazoxide (n = 4 for inhibition by glibenclamide,
  n = 2 for reactivation by diazoxide). c, Representative histograms before
  (top) and after (bottom) addition of 5-HD (100 μM, n = 5 independent
  experiments).
  610 | NAtUre | VOl 572 | 29 AUGUSt 2019