G Model
ARTICLE IN PRESS
YSCDB-2046; No. of Pages 7
Seminars in Cell & Developmental Biology xxx (2016) xxx–xxx
Contents lists available at ScienceDirect
Seminars in Cell & Developmental Biology
journal homepage: www.elsevier.com/locate/semcdb
Fusion of the endoplasmic reticulum by membrane-bound GTPases
Junjie Hu a,∗ , Tom A. Rapoport b,∗
a
b
National Laboratory of Biomacromolecules, Institute of Biophysics, Chinese Academy of Sciences, Beijing 100101, China
Howard Hughes Medical Institute and Department of Cell Biology, Harvard Medical School, Boston, MA 02115, United States
a r t i c l e
i n f o
Article history:
Received 8 April 2016
Received in revised form 31 May 2016
Accepted 2 June 2016
Available online xxx
Keywords:
Endoplasmic reticulum
Membrane fusion
Atlastin
a b s t r a c t
The endoplasmic reticulum (ER) membrane forms an elaborate network of tubules and sheets that is
continuously remodeled. This dynamic behavior requires membrane fusion that is mediated by dynaminlike GTPases: the atlastins in metazoans and Sey1p and related proteins in yeast and plants. Crystal
structures of the cytosolic domains of these membrane proteins and biochemical experiments can now
be integrated into a model that explains many aspects of the molecular mechanism by which these
membrane-bound GTPases mediate membrane fusion.
© 2016 Elsevier Ltd. All rights reserved.
Contents
1.
2.
3.
4.
5.
6.
7.
8.
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00
Crystal structures of cytosolic domains of the GTPases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00
A comprehensive model for ATL function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00
Futile cycles of ATL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00
The roles of the TM segments and CT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00
Comparison of Sey1p and ATL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00
Connection and comparison with SNARE-mediated membrane fusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00
Conclusions and perspective. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .00
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 00
1. Introduction
Membrane fusion between viral and cellular membranes, as
well as between intracellular transport vesicles and target membranes, has been studied extensively. These studies demonstrated
that a conformational change of a viral protein or the assembly of
SNARE proteins pulls two membranes together so that they can
fuse [1–5]. A viral fusion protein generally undergoes the conformational change only once, whereas SNAREs participate in multiple
rounds of fusion, as their assembly is counteracted by the dissociation of the complex, a process mediated by the ATPase NSF [6,7].
How endoplasmic reticulum (ER) membranes fuse has been a
long-standing question. Light microscopy has shown that the ER
∗ Corresponding authors.
E-mail addresses: [email protected] (J. Hu), tom [email protected]
(T.A. Rapoport).
forms a dynamic network, in which membrane tubules continuously form and fuse with one another or with edges of membrane
sheets [8]. In contrast to viral and SNARE-mediated fusion reactions, the fusing membranes are identical (homotypic, rather than
heterotypic fusion). Work over the past seven years has shown that
ER fusion is mediated by the atlastins (ATLs) in metazoans, by Sey1p
in yeast, and by Sey1p-related proteins in plants (the Arabidopsis
thaliana homolog is called RHD3 for Root Hair Defective3) [9–11].
The ATLs and Sey1p-like proteins are membrane-bound GTPases
localized to the ER [9]. They belong to the dynamin family, the
founding member of which is dynamin-1, a GTPase involved in
the fission of membrane vesicles from the plasma membrane during endocytosis [12]. ATLs and Sey1p-like proteins do not share
sequence homology, but they all contain a cytosolic N-terminal
GTPase (G) domain, followed by a helical bundle (HB) domain, two
closely spaced transmembrane (TM) segments, and a cytosolic Cterminal tail (CT) that includes an amphipathic helix (Fig. 1). The
http://dx.doi.org/10.1016/j.semcdb.2016.06.001
1084-9521/© 2016 Elsevier Ltd. All rights reserved.
Please cite this article in press as: J. Hu, T.A. Rapoport, Fusion of the endoplasmic reticulum by membrane-bound GTPases, Semin Cell
Dev Biol (2016), http://dx.doi.org/10.1016/j.semcdb.2016.06.001
G Model
YSCDB-2046; No. of Pages 7
2
ARTICLE IN PRESS
J. Hu, T.A. Rapoport / Seminars in Cell & Developmental Biology xxx (2016) xxx–xxx
Fig. 1. Domain structure of membrane-bound GTPases mediating ER fusion. (A) Atlastins are membrane-bound GTPases in metazoans. They consist of a GTPase (G) domain,
a helical bundle (HB) domain, two close spaced transmembrane segments (TM1 and TM2), and a C-terminal tail (CT) that contains an amphipathic helix (AH). Although
not sequence-related, Sey1p in yeast and its relatives in plants (such as the Arabidosis homolog RHD3) have the same domain organization and function similarly. (B) The
membrane topology of these proteins is shown.
HB domain is significantly longer in the Sey1p proteins. All eukaryotes have either ATL or a Sey1p-related protein, but none appears to
have both. Mammals have three isoforms of ATL, with ATL1 being
prominently expressed in neuronal cells [13,14]. Mutations in ATL1
can cause hereditary spastic paraplegia, a neurodegenerative disease that is characterized by the shortening of the axons in the
longest corticospinal motor neurons in the body [15]. This leads to
progressive spasticity and weakness of the lower limbs. Presumably, the lack of ATL fusion activity causes ER morphology defects
in the long axons. Neuronal disorders have also been reported for
mutations in ATL3 in humans and for deletions of ATL in flies and
zebrafish [16–18]. The Sey1p GTPase also has a physiological role, as
its absence in Candida albicans reduces the virulence of the organism [19].
A role for GTP hydrolysis in ER fusion was first shown by the GTPdependent “coalescence” of ribosome-stripped rough microsomes
[20]. Later experiments used a visual assay to demonstrate that
the formation of a tubular ER network from small Xenopus laevis
egg membrane vesicles is dependent on GTP hydrolysis [21]. However, the identity of the GTPase remained unclear. Several reports
suggested that Rab GTPases might be involved [22–24], but these
studies did not agree on the specific Rab protein involved. Fusion of
the Xenopus membranes could be reduced by extracting Rab proteins with GDI [22,23,25], but re-addition of Rab proteins was not
reported, leaving the possibility that GDI may have had a nonspecific effect. As we discuss below, it is now likely that the GTP
requirement in Xenopus egg extracts is not due to Rab proteins.
Whether Rabs have a role in other cells requires further investigation.
A role for ATL in ER fusion was initially shown in Drosophila
melanogaster and tissue culture cells [9,26]. The depletion or mutational inactivation of ATL led to long, unbranched tubules or
fragmented ER. Non-branched ER tubules were also observed when
dominant-negative ATL mutants were expressed [9,14], which presumably form inactive complexes with the endogenous ATL. In
addition, the fusion of ER vesicles in Xenopus laevis egg extracts
was prevented by the addition of an ATL antibody or a cytosolic fragment of ATL [9,27]. A cytosolic fragment carrying a point
mutation that prevented the interaction with the endogenous ATL
protein was unable to block ER fusion [27]. Thus, in the Xenopus
system, most ER fusion appears to be mediated by ATL. The most
convincing evidence for ATL being a fusogen came from the observation that proteoliposomes containing purified Drosophila ATL
undergo GTP-dependent fusion in vitro [26,28]. GTPase-competent
ATL molecules had to be present on both membranes to allow their
fusion [26,28].
A role for Sey1p in ER fusion in Sarcromyces cerevisiae was initially based on (1) the similarity to ATL in its GTP-binding motifs
and membrane topology, (2) its genetic interaction with an ER
tubule-shaping protein (Yop1p), and (3) on ER morphology changes
observed in deletion mutants [9]. Additional experiments showed
that ER fusion is delayed in SEY1-deletion mutants [10], that ATL
and Sey1p can replace each other in yeast and mammalian cells
[10,29], and that reconstituted Sey1p-containing proteoliposomes
undergo GTP-dependent fusion [10,29], similar to those containing ATL. Deletion of the Sey1p-related RHD3 protein in Arabidopsis
causes the shortening of root hairs and ER morphology defects
resembling those of ATL depletion in mammalian systems [30,31].
Finally, RHD3 can replace Sey1p in yeast, and reconstituted purified
RHD3 mediates GTP-dependent membrane fusion in vitro [11]. All
these experiments indicate that Sey1p and its relatives are functional orthologs of ATL.
2. Crystal structures of cytosolic domains of the GTPases
Structural studies of the cytosolic domains of ATL have sought
to explain GTP hydrolysis-dependent membrane fusion. Initially,
two different conformations of the cytosolic domain of human
ATL1 were reported (Fig. 2). In both structures, ATL forms a dimer
with the G domains facing one another [28,32]. The nucleotides are
bound at the interface between the G domains. In one of the structures, called crystal form 1, the following HB domains cross and
then run parallel to one another. The C-termini are only ∼10 Å apart,
suggesting that the following TM segments in the two full-length
proteins would sit in the same membrane. In the other structure,
called crystal form 2, the G domains are arranged in a similar
way as in crystal form 1, but the two HB domains point in different directions. The corresponding full-length proteins would thus
localize to different membranes and tether them together. Based
on these considerations, it was proposed that crystal forms 2 and
1 correspond to pre- and post-fusion states, respectively; the conformational change would bring the membranes close together for
fusion. However, both structures contained only bound GDP, even
though crystal form 2 was obtained in the presence of GDP plus
Pi . It therefore remained uncertain whether the conformational
change was connected to the GTPase cycle. Biochemical experiments demonstrated that ATL forms much stronger dimers in the
presence of GTP than GDP [28], suggesting that the high protein
concentration during crystallization favored dimerization in the
GDP-bound state.
Another human ATL1 structure (crystal form 3) was later
obtained in the presence of GMPPNP or GDP plus AlF4 − , conditions corresponding to a state before Pi release in the GTPase cycle
[33]. However, this structure only caused more confusion over how
conformational changes of ATL induce fusion, as it resembled the
post-fusion state, rather than the expected pre-fusion state. In fact,
the HB domains of the two ATL molecules were even closer to one
another than in crystal form 1. Similarly puzzling was a structure of
GDP-bound Drosophila ATL [34], which had a pre-fusion-like conformation. In summary, the crystal structures did not lead to a
Please cite this article in press as: J. Hu, T.A. Rapoport, Fusion of the endoplasmic reticulum by membrane-bound GTPases, Semin Cell
Dev Biol (2016), http://dx.doi.org/10.1016/j.semcdb.2016.06.001
G Model
YSCDB-2046; No. of Pages 7
ARTICLE IN PRESS
J. Hu, T.A. Rapoport / Seminars in Cell & Developmental Biology xxx (2016) xxx–xxx
3
Fig. 2. Crystal structures of cytosolic domains of ER-fusion GTPases. Shown are space-filling models of different crystal structures determined for the cytosolic domains of
human ATL1 (HsATL1), Drosophila ATL (DmATL), and C. albicans Sey1p (CaSey1p) in the presence of different nucleotides (the PDB accession numbers are indicated). The TM
segments and CT that follow the cytoplasmic domain and are lacking in the crystal structures are shown schematically. The conformations corresponding to crystal forms 1
and 3 of ATL1 and the Sey1p-dimer are likely generated from monomers sitting in the same membrane (cis-dimers). The ATL conformation corresponding to crystal form 2
can probably form with monomers sitting in either different or the same membranes (trans- or cis-dimers). Bar, 5 nm.
coherent picture of how the GTPase cycle of ATL relates to conformational changes.
Because of these confusing results, it is not surprising that conflicting models have been proposed for ATL-mediated membrane
fusion. In one model, GTP hydrolysis would happen after fusion
[35], but this is difficult to reconcile with the observation that
fusion in vitro requires GTP hydrolysis. A second model posits that
GTP hydrolysis occurs in ATL monomers sitting in different membranes [33], which would subsequently form a dimer (trans-dimer).
Because there is little dimerization in the presence of GDP, the interaction would have to occur between ATL molecules that are in the
short-lived transition state of GTP hydrolysis. Not only is this state
poorly populated, but it also forms dimers very slowly [33,36]. It
has also been proposed that ATL molecules interact only in the
same membrane (cis-interaction), rather than in trans across the
two apposed membranes; the cis-dimers would induce high membrane curvature in both lipid bilayers, which would then induce
their fusion [37]. This model implies that there is no membranetethering step involving ATL, which is contradicted by experiments
demonstrating ATL-mediated vesicle tethering using several different techniques [36,38].
3. A comprehensive model for ATL function
An important clue for the reconciliation of the crystal structures with a fusion mechanism came from experiments showing
that nonhydrolyzable GTP analogs or the presence of GDP and
AlF4 − , which induces a transition state during GTP hydrolysis, cause
dimerization of the cytosolic domain of ATL, but full cycles of GTP
hydrolysis are required for membrane tethering [33,36,38]. This
suggested the possibility that dimers of ATL form on the same
membrane (cis-interaction) and compete with the trans-dimers
that mediate tethering [36]. This idea was tested by experiments
in which ATL molecules were attached to a supported bilayer, so
that they could only interact in cis and not with ATL molecules
sitting in a different membrane (trans-interactions) [36]. Indeed, a
nucleotide-dependent interaction of ATL molecules was observed.
These cis-interactions were reduced when a cytosolic ATL fragment
was added to the bulk solution, suggesting that trans- and cisinteractions can compete with one another. The existence of cisinteractions explains the puzzling observation that the presence of
GDP and AlF4 − leads to the formation of tight ATL dimers, but does
not allow vesicle tethering or fusion [36]. The cis-dimers formed
under these conditions cannot dissociate, leaving no monomers for
trans-interactions. GTP hydrolysis is thus required to regenerate a
pool of monomeric ATL molecules in each membrane, which can
then mediate the tethering and fusion of opposing membranes.
The existence of both cis- and trans- dimers can explain the crystal structures. Crystal form 3 corresponds to the transition state
conformation formed during both cis- and trans- interactions. Crystal form 2 corresponds to a state following GTP binding. However,
the GTP-bound state can probably adopt several different conformations, because GTP binding causes the G domains to undock
from the HBs [33], allowing the two domains to move relative to
one another. Although crystal form 2 was initially thought to form
only in trans, the two ATL molecules may actually sit in the same
membrane, with the HBs nearly parallel to the plane of the membrane (Fig. 2). Finally, crystal form 1 probably represents one of the
intermediate states that the ATL dimer can adopt during the transition from a conformation corresponding to crystal form 2 to one
corresponding to crystal form 3.
A plausible model for ATL-mediated membrane fusion begins
with GTP binding to ATL monomers (Fig. 3). ATL likely needs to be in
a conformation corresponding to crystal form 2, because GTP loading requires the first helix of the HB to interact with the base of the G
domain [33]. In the next step, the monomers rapidly form a dimer,
with the bound GTP molecules buried at the interface. This process is likely reversible (not shown in Fig. 3). The dimer is relatively
unstable and can adopt different conformations once the G domains
are released from the HBs. When GTP is hydrolyzed by the dimer,
the two HBs associate with one another, generating a tight dimer
in the transition state. Finally, Pi and GDP are released in a sequen-
Please cite this article in press as: J. Hu, T.A. Rapoport, Fusion of the endoplasmic reticulum by membrane-bound GTPases, Semin Cell
Dev Biol (2016), http://dx.doi.org/10.1016/j.semcdb.2016.06.001
G Model
YSCDB-2046; No. of Pages 7
ARTICLE IN PRESS
J. Hu, T.A. Rapoport / Seminars in Cell & Developmental Biology xxx (2016) xxx–xxx
4
Fig. 3. A model of ATL function. The scheme indicates different stages during the GTPase cycle of ATL. Stage A: ground state, in which ATL molecules (in red and yellow)
are in the nucleotide-free, monomeric state. Stage B: ATL molecules bind GTP and form dimers either in the same membrane (cis-interaction; red–red) or across different
membranes (trans-interaction; red–yellow). Stage C: following GTP-binding, the G-domains of ATL are dislodged from the HB-domains and can rotate (indicated by the
arrows). Stage D: ATL is in the transition state of GTP hydrolysis (GDP–Pi ), in which the HBs of the two molecules interact tightly with one another. These trans-interaction
lead to membrane fusion. Stage E: ATL molecules have released Pi and are in the GDP-bound state. Dissociation of the dimer may occur, but is probably completed after
GDP release, i.e. transition back to stage A. Stages F and G: trans-interaction of ATL molecules leads to tethering, but not fusion. This futile tethering cycle (purple arrows) is
completed when ATL converts back to the original state (stage A). Note that cis-interacting ATL molecules also undergo a nucleotide-dependent futile interaction that does
not lead to membrane fusion.
tial manner, causing the dissociation of the dimer into monomers.
It remains unclear whether dissociation occurs in the GDP-bound
state, in which the monomers have a weak affinity for each other,
or in the nucleotide-free state, in which they have no measurable
affinity [36]. All reactions can occur with ATL molecules sitting
in different or the same membranes (trans- and cis-interactions,
respectively), but membrane tethering and fusion can only occur
with trans-interactions. The tight interaction of the monomers in
the transition state of GTP hydrolysis likely pulls the two membranes together so that they can fuse.
4. Futile cycles of ATL
ATL not only undergoes an apparent futile GTPase cycle involving the formation and dissociation of cis-dimers. It also forms
trans-dimers in a futile manner. Experiments with reconstituted
proteoliposomes showed that multiple GTPase cycles are required
for a successful fusion event [36,38]; the transition state of trans-
dimers often converts back into monomers without having caused
fusion. Other experiments indicated that the efficiency of tethering
and fusion increases with the density of ATL molecules in the two
membranes [36]. Taken together, it seems that a successful fusion
event requires the cooperation of multiple trans-dimerization
events: as one dimer reaches the transition state, other dimers
forming nearby would help to maintain the tethered state and add
to the pulling force exerted on the opposing membranes. The futile
formation of trans dimers suggests that a fraction of three-way
junctions in cells may actually consist of tethered, rather than fused,
tubules.
The futile assembly and disassembly of cis-dimers may simply
be a consequence of using a single protein to mediate multiple
rounds of membrane fusion. ATL molecules have to cycle between
being in the same and different membranes, and during the actual
fusion process they must interact with one another in a conformation that is close to either of these situations. Thus, one would
expect the formation of at least some cis-dimers, which would have
Please cite this article in press as: J. Hu, T.A. Rapoport, Fusion of the endoplasmic reticulum by membrane-bound GTPases, Semin Cell
Dev Biol (2016), http://dx.doi.org/10.1016/j.semcdb.2016.06.001
G Model
ARTICLE IN PRESS
YSCDB-2046; No. of Pages 7
J. Hu, T.A. Rapoport / Seminars in Cell & Developmental Biology xxx (2016) xxx–xxx
to dissociate to regenerate monomers. cis-dimers may generally
be formed at the site of fusion both before and after trans-dimers
caused bilayer merging. Multiple futile cycles of trans-dimerization
that lead to tethering may cause the localized enrichment of
monomers that could then engage in cis-dimerization in each membrane. cis-dimerization at the site of membrane tethering would
effectively lower the energy barrier for the fusion event. cis-dimers
might also form after fusion and transiently keep ATL and Sey1p
in newly formed three-way junctions, which might explain the
enrichment of some isoforms in junctions [23,29]. Whether such
cis-dimers stabilize fused three-way junctions [39] remains to be
tested.
5. The roles of the TM segments and CT
The TM segments of ATL are not mere membrane-anchors. They
cannot be replaced by the TM domains from other proteins, and
some point mutations in the TMs severely affect the efficiency of
fusion [40]. Thus, the TMs might contribute to the formation of a
fusion intermediate, such as a hemifusion stalk or fusion pore. In
addition, they mediate the formation of nucleotide-independent
ATL oligomers, which could increase the local concentration of ATL
molecules at the site of fusion, thus increasing the fusion efficiency.
Finally, the hairpin structure formed by the closely spaced TMs
probably targets ATL to high-membrane curvature regions of the
ER, such as tubules and sheet edges, as similar hairpin structures
are found in other proteins localized to these regions [41,42].
The C-terminal tail (CT) contains an amphipathic helix that promotes membrane fusion [40,43]. ATL lacking the CT has only low
fusion activity, but it can be re-activated by addition of a synthetic
peptide corresponding to the amphipathic helix [40]. The helix
acts by directly binding to and perturbing the phospholipid bilayer,
which could facilitate the merging of the two membranes during
fusion. Binding of the helix to the membrane probably occurs at all
stages of the fusion process. Phosphorylation of the CT of RHD3 has
been reported to stimulate tubule formation, but this modification
seems to be confined to this particular GTPase [44].
6. Comparison of Sey1p and ATL
Recent experiments show that Sey1p functions similarly to ATL
[29]. Crystal structures of the cytosolic domain of C. albicans Sey1p
show that it forms a dimer in GMPPNP or GDP plus AlF4 − , and a
monomer in GDP. In the dimeric state, the HB segments of the two
Sey1p molecules crossover twice, such that the C-termini are on the
same side as the G domains (Fig. 2). Surprisingly, when the reconstituted full-length proteins were tested in a fusion assay, some lipid
mixing was observed even with the non-hydrolyzable GTP analog GMPPNP, although fusion was much more efficient with GTP
and was not observed with GTP␥S [10]. It is possible that Sey1p
monomers form the tightest dimer in their GTP-bound state, rather
than in the transition state as ATL monomers do.
7. Connection and comparison with SNARE-mediated
membrane fusion
In Xenopus extracts and Drosophila cells, most if not all ER fusion
seems to be mediated by ATL, as its inhibition or absence causes the
ER to fragment [26,27]. Similarly, the Sey1p-homolog RHD3 may be
the only fusogen in A. thaliana, as its deletion causes drastic ER morphology defects [31]. In contrast, the deletion of SEY1 in S. cerevisiae
does not grossly disturb the cortical ER network [9], and membrane
fusion is delayed, but not abolished [10]. The alternative fusion
pathway in S. cerevisiae is likely mediated by ER SNAREs. Mutants
of the ER SNAREs Ufe1p, Sec22p, or Use1p have a drastic effect on
5
ER morphology when combined with a deletion of SEY1 [10]. In
addition, combining a temperature-sensitive mutant in Ufe1p with
a SEY1 deletion results in a much slower fusion rate than with each
mutant individually [10]. The SNARE-mediated ER fusion process
also requires the tethering protein Dsl and is distinct from that of
Golgi-derived vesicles with the ER [45]. These data suggest that
Sey1p and ER SNAREs function in alternative fusion pathways in S.
cerevisiae, although it is unclear whether both pathways operate in
parallel in normal cells.
A different model has been proposed in which Sey1p and ER
SNAREs function together in the same pathway [46]. The authors
suggested that the Sey1p GTPase causes only the tethering of membranes, like the Rab GTPases, whereas the SNAREs and NSF ATPase
perform actual fusion. However, this model is difficult to reconcile with the fact that Sey1p alone can mediate membrane fusion
[10,29], and that double mutants in SEY1 and SNAREs have a much
stronger effect on fusion or cell growth than the individual SEY1
or SNARE mutants [10]. These results strongly argue that these
proteins act in alternative pathways, rather than the same.
It is interesting to compare ATL/Sey1p-mediated membrane
fusion with that involving the SNAREs. In both cases, nucleotide
hydrolysis is required, but for ATL/Sey1p this energy is spent during the actual fusion reaction, whereas for SNAREs it is used for the
recycling of the complex after fusion. In both systems, cis- and transcomplexes form, but in the case of SNAREs, it is uncertain whether
they compete. SNAREs interact with several factors, including SM
proteins, which could prevent cis-assembly and catalyze the formation of trans- complexes [47]. In contrast, no factors have been
identified that affect the fusion activity of ATL or Sey1p-like proteins. Futile tethering cycles may occur for SNAREs as well, as
trans-complexes were observed to disassemble before they caused
fusion [48]. As in the case of ATL, fusion is more efficient when
multiple SNARE complexes assemble simultaneously [49]. Thus, in
both systems, not every event of trans-interaction seems to result
in fusion.
8. Conclusions and perspective
Over the past years, significant progress has been made in our
understanding of the mechanism by which ATL and Sey1p mediate
the fusion of ER membranes. Because fusion is mediated by identical molecules in both membranes and because no other proteins
seem to be involved, the reaction is significantly simpler than that
mediated by SNAREs. In addition, crystal structures are available
and the fusion reaction can be recapitulated with purified proteins
in reconstituted proteoliposomes, triggered simply by the addition
of GTP. These features make membrane-bound GTPases particularly attractive to analyze the molecular mechanism of membrane
fusion. Indeed, the combination of structural information and biochemical data has led to a comprehensive model of ATL-mediated
fusion (Fig. 3). Perhaps the most surprising aspect of this model is
the existence of two types of futile cycles. Several aspects of the
fusion mechanism may also apply to the related GTPases that fuse
the membranes of mitochondria. Interestingly, mitofusin (called
Fzo1p in yeast), which mediates outer membrane fusion, only forms
dimers when actively hydrolyzing GTP [50], suggesting that it also
undergoes futile cycling.
Despite the recent progress, we do not understand how exactly
ATL and Sey1p catalyze the merging of phospholipid bilayers. Elucidation of the actual fusion mechanism will require the generation of
intermediates and the determination of the membrane structure in
each of them. The role of lipids needs to be studied in more detail,
as these likely play an important role, demonstrated by observations with SNARE-mediated fusion [51] and preliminary results on
Sey1p [52]. Future studies will likely require the use of biophysi-
Please cite this article in press as: J. Hu, T.A. Rapoport, Fusion of the endoplasmic reticulum by membrane-bound GTPases, Semin Cell
Dev Biol (2016), http://dx.doi.org/10.1016/j.semcdb.2016.06.001
G Model
YSCDB-2046; No. of Pages 7
ARTICLE IN PRESS
J. Hu, T.A. Rapoport / Seminars in Cell & Developmental Biology xxx (2016) xxx–xxx
6
cal methods to understand the details of the actual fusion process,
and structural characterization of full-length ATL or Sey1p. Understanding the precise configurations of the TMs and CT, domains
that interact with lipids during fusion, will be crucial to determining how fusion is catalyzed. Comparison of the membrane domains
of SNAREs [53] and ATLs could give new insights into how a completely different set of proteins evolved to perform the same task.
The biological function of ATL and Sey1p-like proteins is also not
entirely understood. While it is clear that these proteins mediate ER
fusion, it seems possible that they have additional functions, such
as in stabilizing newly formed tubular junctions. ATL and Sey1plike proteins also interact through their TM segments with other
membrane proteins that determine ER morphology. Such interactions have been reported for the reticulons and DP1 (Yop1p in
yeast) [9], which are implicated in generating and stabilizing ER
tubules, and for the lunapark protein [54], which might stabilize
tubular junctions [39,55]. However, how exactly ATL and Sey1plike proteins cooperate with these proteins is unknown, and it is
unclear what specific roles these components play in generating
the tubular ER network. Thus, probably the most important goal in
the field is to reconstitute a tubular membrane network with purified proteins, a system that would allow in-depth studies on the
relationship between membrane -fusion and -shaping and on the
molecular mechanism by which ER morphology is achieved.
Acknowledgments
We thank Tina Liu, Robert Powers, Hanna Tukachinsky, and
Songyu Wang for critical reading of the manuscript. J.H. is supported by the NSF of China (Grant 31225006) and an International
Early Career Scientist grant from Howard Hughes Medical Institute.
T.A.R. is a Howard Hughes Medical Institute Investigator.
References
[1] S.C. Harrison, Viral membrane fusion, Nat. Struct. Mol. Biol. 15 (7) (2008)
690–698.
[2] R. Jahn, R.H. Scheller, SNAREs—engines for membrane fusion, Nat. Rev. Mol.
Cell Biol. 7 (9) (2006) 631–643.
[3] W. Wickner, R. Schekman, Membrane fusion, Nat. Struct. Molecular Biol. 15
(7) (2008) 658–664.
[4] S. Martens, H.T. McMahon, Mechanisms of membrane fusion: disparate
players and common principles, Nat. Rev. Mol. Cell Biol. 9 (7) (2008) 543–556.
[5] T.C. Sudhof, J.E. Rothman, Membrane fusion: grappling with SNARE and SM
proteins, Science 323 (5913) (2009) 474–477.
[6] J. Rizo, J. Xu, The synaptic vesicle release machinery, Ann. Rev. Biophys. 44
(2015) 339–367.
[7] M. Zhao, A.T. Brunger, Recent advances in deciphering the structure and
molecular mechanism of the AAA+ ATPase N-Ethylmaleimide-Sensitive factor
(NSF), J. Mol. Biol. (2015).
[8] C. Lee, L.B. Chen, Dynamic behavior of endoplasmic reticulum in living cells,
Cell 54 (1) (1988) 37–46.
[9] J. Hu, Y. Shibata, P.P. Zhu, C. Voss, N. Rismanchi, W.A. Prinz, T.A. Rapoport, C.
Blackstone, A class of dynamin-like GTPases involved in the generation of the
tubular ER network, Cell 138 (3) (2009) 549–561.
[10] K. Anwar, R.W. Klemm, A. Condon, K.N. Severin, M. Zhang, R. Ghirlando, J. Hu,
T.A. Rapoport, W.A. Prinz, The dynamin-like GTPase Sey1p mediates
homotypic ER fusion in S. cerevisiae, J. Cell Biol. 197 (2) (2012) 209–217.
[11] M. Zhang, F. Wu, J. Shi, Y. Zhu, Z. Zhu, Q. Gong, J. Hu, ROOT HAIR DEFECTIVE3
family of dynamin-like GTPases mediates homotypic endoplasmic reticulum
fusion and is essential for arabidopsis development, Plant Physiol. 163 (2)
(2013) 713–720.
[12] G.J. Praefcke, H.T. McMahon, The dynamin superfamily: universal membrane
tubulation and fission molecules? Nat. Rev. Mol. Cell Biol. 5 (2) (2004)
133–147.
[13] P.P. Zhu, A. Patterson, B. Lavoie, J. Stadler, M. Shoeb, R. Patel, C. Blackstone,
Cellular localization, oligomerization, and membrane association of the
hereditary spastic paraplegia 3A (SPG3A) protein atlastin, J. Biol. Chem. 278
(49) (2003) 49063–49071.
[14] N. Rismanchi, C. Soderblom, J. Stadler, P.P. Zhu, C. Blackstone, Atlastin GTPases
are required for Golgi apparatus and ER morphogenesis, Hum. Mol. Genet. 17
(11) (2008) 1591–1604.
[15] S. Salinas, C. Proukakis, A. Crosby, T.T. Warner, Hereditary spastic paraplegia:
clinical features and pathogenetic mechanisms, Lancet Neurol. 7 (12) (2008)
1127–1138.
[16] D. Fischer, M. Schabhuttl, T. Wieland, R. Windhager, T.M. Strom, M.
Auer-Grumbach, A novel missense mutation confirms ATL3 as a gene for
hereditary sensory neuropathy type 1, Brain: J. Neurol. 137 (2014) e286, Pt. 7.
[17] J. Summerville, J. Faust, E. Fan, D. Pendin, A. Daga, J. Formella, M. Stern, J.A.
McNew, The effects of ER morphology on synaptic structure and function in
Drosophila melanogaster, J. Cell Sci. (2016).
[18] C. Fassier, J.A. Hutt, S. Scholpp, A. Lumsden, B. Giros, F. Nothias, S.
Schneider-Maunoury, C. Houart, J. Hazan, Zebrafish atlastin controls motility
and spinal motor axon architecture via inhibition of the BMP pathway, Nat.
Neurosci. 13 (11) (2010) 1380–1387.
[19] T. Yamada-Okabe, H. Yamada-Okabe, Characterization of the CaNAG3,
CaNAG4, and CaNAG6 genes of the pathogenic fungus Candida albicans:
possible involvement of these genes in the susceptibilities of cytotoxic agents,
FEMS Microbiol. Lett. 212 (1) (2002) 15–21.
[20] J. Paiement, H. Beaufay, D. Godelaine, Coalescence of microsomal vesicles
from rat liver: a phenomenon occurring in parallel with enhancement of the
glycosylation activity during incubation of stripped rough microsomes with
GTP, J. Cell Biol. 86 (1) (1980) 29–37.
[21] L. Dreier, T.A. Rapoport, In vitro formation of the endoplasmic reticulum
occurs independently of microtubules by a controlled fusion reaction, J. Cell
Biol. 148 (5) (2000) 883–898.
[22] A. Audhya, A. Desai, K. Oegema, A role for Rab5 in structuring the endoplasmic
reticulum, J. Cell Biol. 178 (1) (2007) 43–56.
[23] A.R. English, G.K. Voeltz, Rab10 GTPase regulates ER dynamics and
morphology, Nat. Cell Biol. 15 (2) (2013) 169–178.
[24] A. Gerondopoulos, R.N. Bastos, S. Yoshimura, R. Anderson, S. Carpanini, I.
Aligianis, M.T. Handley, F.A. Barr, Rab18 and a Rab18 GEF complex are
required for normal ER structure, J. Cell Biol. 205 (5) (2014) 707–720.
[25] M.D. Turner, H. Plutner, W.E. Balch, A Rab GTPase is required for homotypic
assembly of the endoplasmic reticulum, J. Biol. Chem. 272 (21) (1997)
13479–13483.
[26] G. Orso, D. Pendin, S. Liu, J. Tosetto, T.J. Moss, J.E. Faust, M. Micaroni, A.
Egorova, A. Martinuzzi, J.A. McNew, A. Daga, Homotypic fusion of ER
membranes requires the dynamin-like GTPase atlastin, Nature 460 (7258)
(2009) 978–983.
[27] S. Wang, F.B. Romano, C.M. Field, T.J. Mitchison, T.A. Rapoport, Multiple
mechanisms determine ER network morphology during the cell cycle in
Xenopus egg extracts, J. Cell Biol. 203 (5) (2013) 801–814.
[28] X. Bian, R.W. Klemm, T.Y. Liu, M. Zhang, S. Sun, X. Sui, X. Liu, T.A. Rapoport, J.
Hu, Structures of the atlastin GTPase provide insight into homotypic fusion of
endoplasmic reticulum membranes, Proc. Natl. Acad. Sci. U. S. A. 108 (10)
(2011) 3976–3981.
[29] L. Yan, S. Sun, W. Wang, J. Shi, X. Hu, S. Wang, D. Su, Z. Rao, J. Hu, Z. Lou,
Structures of the yeast dynamin-like GTPase Sey1p provide insight into
homotypic ER fusion, J. Cell Biol. 210 (6) (2015) 961–972.
[30] J.W. Schiefelbein, C. Somerville, Genetic control of root hair development in
arabidopsis thaliana, Plant Cell 2 (3) (1990) 235–243.
[31] H. Zheng, L. Kunst, C. Hawes, I. Moore, A GFP-based assay reveals a role for
RHD3 in transport between the endoplasmic reticulum and Golgi apparatus,
Plant J. 37 (3) (2004) 398–414.
[32] L.J. Byrnes, H. Sondermann, Structural basis for the nucleotide-dependent
dimerization of the large G protein atlastin-1/SPG3A, Proc. Natl. Acad. Sci. U.
S. A. 108 (6) (2011) 2216–2221.
[33] L.J. Byrnes, A. Singh, K. Szeto, N.M. Benvin, J.P. O’Donnell, W.R. Zipfel, H.
Sondermann, Structural basis for conformational switching and GTP loading
of the large G protein atlastin, EMBO J. 32 (3) (2013) 369–384.
[34] F. Wu, X. Hu, X. Bian, X. Liu, J. Hu, Comparison of human and drosophila
atlastin GTPases, Protein Cell 6 (2) (2015) 139–146.
[35] J. Morin-Leisk, S.G. Saini, X. Meng, A.M. Makhov, P. Zhang, T.H. Lee, An
intramolecular salt bridge drives the soluble domain of GTP-bound atlastin
into the postfusion conformation, J. Cell Biol. 195 (4) (2011) 605–615.
[36] T.Y. Liu, X. Bian, F.B. Romano, T. Shemesh, T.A. Rapoport, J. Hu, Cis and trans
interactions between atlastin molecules during membrane fusion, Proc. Natl.
Acad. Sci. U. S. A. 112 (15) (2015) E1851–E1860.
[37] O. Daumke, G.J. Praefcke, Structural insights into membrane fusion at the
endoplasmic reticulum, Proc. Natl. Acad. Sci. U. S. A. 108 (6) (2011)
2175–2176.
[38] S.G. Saini, C. Liu, P. Zhang, T.H. Lee, Membrane tethering by the atlastin
GTPase depends on GTP hydrolysis but not on forming the cross-over
configuration, Mol. Biol. Cell 25 (24) (2014) 3942–3953.
[39] T. Shemesh, R.W. Klemm, F.B. Romano, S. Wang, J. Vaughan, X. Zhuang, H.
Tukachinsky, M.M. Kozlov, T.A. Rapoport, A model for the generation and
interconversion of ER morphologies, Proc. Natl. Acad. Sci. U. S. A. 111 (49)
(2014) E5243–E5251.
[40] T.Y. Liu, X. Bian, S. Sun, X. Hu, R.W. Klemm, W.A. Prinz, T.A. Rapoport, J. Hu,
Lipid interaction of the C terminus and association of the transmembrane
segments facilitate atlastin-mediated homotypic endoplasmic reticulum
fusion, Proc. Natl. Acad. Sci. U. S. A. 109 (32) (2012) E2146–E2154.
[41] J. Hu, W.A. Prinz, T.A. Rapoport, Weaving the web of ER tubules, Cell 147 (6)
(2011) 1226–1231.
[42] J. Chang, S. Lee, C. Blackstone, Protrudin binds atlastins and endoplasmic
reticulum-shaping proteins and regulates network formation, Proc. Natl.
Acad. Sci. U. S. A. 110 (37) (2013) 14954–14959.
[43] J.E. Faust, T. Desai, A. Verma, I. Ulengin, T.L. Sun, T.J. Moss, M.A.
Betancourt-Solis, H.W. Huang, T. Lee, J.A. McNew, The atlastin C-terminal tail
is an amphipathic helix that perturbs the bilayer structure during
Please cite this article in press as: J. Hu, T.A. Rapoport, Fusion of the endoplasmic reticulum by membrane-bound GTPases, Semin Cell
Dev Biol (2016), http://dx.doi.org/10.1016/j.semcdb.2016.06.001
G Model
YSCDB-2046; No. of Pages 7
ARTICLE IN PRESS
J. Hu, T.A. Rapoport / Seminars in Cell & Developmental Biology xxx (2016) xxx–xxx
[44]
[45]
[46]
[47]
[48]
[49]
endoplasmic reticulum homotypic fusion, J. Biol. Chem. 290 (8) (2015)
4772–4783.
H. Ueda, E. Yokota, K. Kuwata, N. Kutsuna, S. Mano, T. Shimada, K. Tamura, G.
Stefano, Y. Fukao, F. Brandizzi, T. Shimmen, M. Nishimura, I. Hara-Nishimura,
Phosphorylation of the C terminus of RHD3 has a critical role in homotypic ER
membrane fusion in arabidopsis, Plant Physiol. 170 (2) (2016) 867–880.
J.V. Rogers, C. McMahon, A. Baryshnikova, F.M. Hughson, M.D. Rose,
ER-associated retrograde SNAREs and the Dsl1 complex mediate an
alternative, Sey1p-independent homotypic ER fusion pathway, Mol. Biol. Cell
25 (21) (2014) 3401–3412.
M. Lee, Y.J. Ko, Y. Moon, M. Han, H.W. Kim, S.H. Lee, K. Kang, Y. Jun, SNAREs
support atlastin-mediated homotypic ER fusion in Saccharomyces cerevisiae, J.
Cell Biol. 210 (3) (2015) 451–470.
R.W. Baker, P.D. Jeffrey, M. Zick, B.P. Phillips, W.T. Wickner, F.M. Hughson, A
direct role for the Sec1/Munc18-family protein Vps33 as a template for
SNARE assembly, Science 349 (6252) (2015) 1111–1114.
H. Xu, Y. Jun, J. Thompson, J. Yates, W. Wickner, HOPS prevents the
disassembly of trans-SNARE complexes by Sec17p/Sec18p during membrane
fusion, EMBO J. 29 (12) (2010) 1948–1960.
J.M. Hernandez, A.J. Kreutzberger, V. Kiessling, L.K. Tamm, R. Jahn, Variable
cooperativity in SNARE-mediated membrane fusion, Proc. Natl. Acad. Sci. U. S.
A. 111 (33) (2014) 12037–12042.
7
[50] N. Ishihara, Y. Eura, K. Mihara, Mitofusin 1 and 2 play distinct roles in
mitochondrial fusion reactions via GTPase activity, J. Cell Sci. 117 (2004)
6535–6546, Pt 26.
[51] M. Zick, C. Stroupe, A. Orr, D. Douville, W.T. Wickner, Membranes linked by
trans-SNARE complexes require lipids prone to non-bilayer structure for
progression to fusion, eLife 3 (2014) e01879.
[52] S. Sugiura, J. Mima, Physiological lipid composition is vital for homotypic ER
membrane fusion mediated by the dynamin-related GTPase Sey1p, Sci. Rep. 6
(2016) 20407.
[53] A. Stein, G. Weber, M.C. Wahl, R. Jahn, Helical extension of the neuronal
SNARE complex into the membrane, Nature 460 (7254) (2009) 525–528.
[54] S. Chen, P. Novick, S. Ferro-Novick, ER network formation requires a balance
of the dynamin-like GTPase Sey1p and the Lunapark family member Lnp1p,
Nat. Cell Biol. 14 (7) (2012) 707–716.
[55] S. Chen, T. Desai, J.A. McNew, P. Gerard, P.J. Novick, S. Ferro-Novick, Lunapark
stabilizes nascent three-way junctions in the endoplasmic reticulum, Proc.
Natl. Acad. Sci. U. S. A. 112 (2) (2015) 418–423.
Please cite this article in press as: J. Hu, T.A. Rapoport, Fusion of the endoplasmic reticulum by membrane-bound GTPases, Semin Cell
Dev Biol (2016), http://dx.doi.org/10.1016/j.semcdb.2016.06.001