Nature - USA (2019-07-18)

Letter reSeArCH

Fig. 4d, g). Notably, yeast Mgm1(Y520A)—containing a mutation in the
BSE–stalk contact in the tetramer interface—exerted a strong dominant
negative effect on respiratory yeast growth when co-expressed with
endogenous Mgm1 (Extended Data Fig. 4h). The corresponding yeast
strain retained mitochondrial DNA (Extended Data Fig. 4i), enabling
us to examine Mgm1-specific deficits on mitochondrial morphology
and ultrastructure. The expression of yeast Mgm1(Y520A) induced
fragmentation of the mitochondrial network (Extended Data Fig. 4j, k),
a reduction in the number and length of cristae and an increase in crista
diameter (Extended Data Fig. 4l, m).
We used electron cryo-tomography (cryo-ET) and subtomogram
averaging to determine the structure of membrane-bound Mgm1
(Fig. 4a, b, Extended Data Fig. 5a, b). In the absence of nucleotide, or
upon the addition of GTPγS, Mgm1 remodelled Folch liposomes into
membrane tubes of varying diameters ranging from around 18 nm
to 140 nm. The Mgm1 coat decorated membrane tubes in a regular
lattice. For subtomogram averaging, preformed tubes with diameters
of around 20 nm were used in order to increase the number of par-
ticles for averaging. These tubes also stimulated the GTPase activity
of Mgm1, although less strongly than did Folch liposomes (Extended
Data Fig. 5c). The final resolution of the subtomogram average volume
was 14.7 Å for the nucleotide-free and the nucleotide-bound forms
(Extended Data Table 1). No substantial differences were apparent
between the two volumes. Notably, the Mgm1 tetramer fits the sub-
tomogram average volume with only minor positional domain rear-
rangements (Fig. 4b, Extended Data Fig. 6a, c, e). The G domain was
in a closed conformation relative to the BSE domain and was located
furthest from the membrane, the stalk was in the middle, and the

paddle was next to the membrane. The Mgm1 coat in Fig. 4b, c can

be viewed as a left-handed four-start helix, consisting of four parallel

helical filaments (Extended Data Fig. 7a, b). Similar filaments were

observed on Folch lipid tubes of different diameters, although their

helical parameters varied (Extended Data Fig. 5d). The filament back-

bone was formed by stalks oligomerizing in an alternating fashion via

interface-1 and interface-2. This is in contrast to dynamin filaments,

in which the stalks oligomerize via three interfaces^22 ,^23 (Extended Data

Fig. 2e). Another difference compared with dynamin^22 is that we did

not observe interactions between the G domains of adjacent helix turns.

Instead, contact was established through the paddle domains (Fig. 4b, c,

Extended Data Fig. 6a). Mutation of the conserved residues F779

and S780 in the paddle contact site affected membrane binding and

stimulated GTPase activity only mildly (Extended Data Fig. 3a, d).

Expression of the corresponding mutant protein in yeast complemented

the loss of endogenous Mgm1 with respect to respiratory growth, but

the cells exhibited moderate alterations of mitochondrial morphology

and mitochondrial genome maintenance (Extended Data Fig. 4d, f, g).

The tendency for Mgm1 to form a left-handed helix on the convex

exterior of membrane tubes is consistent with the curvature of the crys-

tallographic tetramer that arises from the interaction between inter-

faces-1 of two dimers. A model in which several dimers are connected

via identical interfaces-1 results in a continuous filament with dimen-

sions and helix parameters (radius, pitch) similar to those observed

by cryo-ET (Extended Data Fig. 7b, c). Microsecond-scale molecular

dynamics simulations starting from the crystallographic tetramer

provide further evidence for the curvature preference of the Mgm1

interface-1 (Extended Data Fig. 7d–k). The most likely curvature and

twist in the simulated interface-1 was the same as that in the crystal

lattice. The simulation results also suggested that there is sufficient

flexibility in interface-1 to account for the observed variable radii of

Mgm1 helical filaments.

We followed the dynamics of Mgm1 assembly on the membrane by

live fluorescence confocal imaging. By manipulating streptavidin beads

adhering to giant unilamellar vesicles (GUVs) with optical tweezers^24 ,

a

90°

BSE

G domain

Stalk

Paddle

Dimer 1 Dimer 2

Dimer 1

Dimer 2

BSE

G domain

Stalk

Paddle

BSE

G domain

Stalk

Paddle

G domain

Stalk

Paddle

Interface-1

Interface-1

E534

α 2 B

α2CS α 4 S

α 4 S

α1NS

α1NS

R855

b

OD

600 nm

(AU)

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

OD

600 nm

(AU)

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0510 15 20

Time (h)

0510 15 20

Time (h)

EV

WT

D542A

K545A

Interface-1

EV

WT

Y520A

R621A

BSE–stalk contact

A555D559K562E858

A556S

R855

E858K562D559A555A556

Y537

R646

Fig. 3 | Assembly mechanism of Mgm1. a, The tetramer, as seen in the
crystal. Two dimers (grey or colour-coded by domain) interact via stalk
interface-1 and a small BSE–stalk contact. See insets for details. b, Ye a s t
respiratory growth complementation assays with Mgm1 containing
mutations in interface-1 residues (left; D542 and K545 in yeast Mgm1
correspond to D559 and K562 in the C. thermophilum protein) and
residues of the BSE–stalk contact (right; Y520 and R621 in yeast Mgm1
correspond to Y537 and R646 in the C. thermophilum protein). A
representative growth curve from n = 3 independent experiments is
shown. EV, empty vector control; OD600nm, optical density at 600 nm; AU,
arbitrary units. See also Extended Data Fig. 4.

a

GUV

Bead

Force Tubule

Injection

pipette Micro

pipette

Optical

d tweezers

0 s60 s

Mgm1 lament

25-nm radius

BSE

Paddle

Stalk

Mgm1

Paddle

BSEStalk

G domain

Paddle

Stalk

BSE

G domain

90°

54-nm pitch

b

c

Extra

Outer density

leaet

Inner

leaet

G domain

Fig. 4 | Mgm1 forms a helical lattice on the outside of lipid

tubes. a, Mgm1 forms a regular protein coat on the outer surface of

galactocerebroside-containing lipid tubes, enabling analysis by cryo-ET.

b, The subtomogram average shows the protein lattice with Mgm1 flexibly

fitted into the cryo-ET volume of the apo form. The outer leaflet of the

membrane is not well defined in this reconstruction. Arrows indicate

density not attributable to the protein, which was assigned to the outer

membrane leaflet. c, Four filaments of Mgm1 dimers wrap around a

membrane tube in a left-handed surface lattice. Stalks assemble via

interface-1 and interface-2 in an alternating manner (Extended Data

Fig. 7). d, Tube-pulling assay for the generation of a tube surface that is

accessible from the outside. Mgm1 (green fluorescence) binds to the tube

and the GUV surface (n = 8 independent experiments).

18 JULY 2019 | VOL 571 | NAtUre | 431