A direct role for GRASP65 as a mitotically regulated Golgi stacking factor

Abstract

Cell-free assays that mimic the disassembly and reassembly cycle of the Golgi apparatus during mitosis implicated GRASP65 as a mitotically regulated stacking factor. We now present evidence that GRASP65 is directly involved in stacking Golgi cisternae. GRASP65 is the major phosphorylation target in rat liver Golgi membranes of two mitotic kinases, cdc2-cyclin B and polo-like kinases, which alone will unstack Golgi membranes, generating single cisternae. Mitotic cells microinjected with antibodies to GRASP65 fail to form proper Golgi stacks after cell division. Beads coated with GRASP65 homodimers form extensive aggregates consistent with the formation of trans oligomers. These can be disaggregated using purified cdc2-cyclin B1 and polo-like kinases, and re-aggregated after dephosphorylation of GRASP65. Together, these data demonstrate that GRASP65 has the properties required to bind surfaces together in a mitotically regulated manner.

Full text

Yanzhuang Wang, Joachim Seemann
1
,
Marc Pypaert, James Shorter
2
and
Graham Warren
3
Department of Cell Biology, Ludwig Institute for Cancer Research,
Yale University School of Medicine, 333 Cedar Street, New Haven,
CT 06520, USA
1
Present address: Department of Cell Biology, University of Texas
Southwestern Medical Center, Dallas, TX 75390, USA
2
Present address: Whitehead Institute, Cambridge, MA 02142, USA
3
Corresponding author
e-mail: graham.warren@yale.edu
Y.Wang and J.Seemann contributed equally to this work
Cell-free assays that mimic the disassembly and re-
assembly cycle of the Golgi apparatus during mitosis
implicated GRASP65 as a mitotically regulated stack-
ing factor. We now present evidence that GRASP65
is directly involved in stacking Golgi cisternae.
GRASP65 is the major phosphorylation target in
rat liver Golgi membranes of two mitotic kinases,
cdc2±cyclin B and polo-like kinases, which alone will
unstack Golgi membranes, generating single cisternae.
Mitotic cells microinjected with antibodies to
GRASP65 fail to form proper Golgi stacks after cell
division. Beads coated with GRASP65 homodimers
form extensive aggregates consistent with the forma-
tion of trans oligomers. These can be disaggregated
using puri®ed cdc2±cyclin B1 and polo-like kinases,
and re-aggregated after dephosphorylation of
GRASP65. Together, these data demonstrate that
GRASP65 has the properties required to bind surfaces
together in a mitotically regulated manner.
Keywords: cdc2 kinase/Golgi/phosphorylation/polo-like
kinase/stacking
Introduction
The central and unique feature of the Golgi apparatus in
most eukaryotic cells is the stack of ¯attened cisternal
membranes with dilated rims (Rambourg and Clermont,
1997). The stack carries out post-translational processing
of newly synthesized proteins as they pass through this
organelle after assembly in the endoplasmic reticulum
(ER) (Farquhar and Palade, 1981). Processing enzymes
include those involved in modifying bound oligosacchar-
ides, and these enzymes are arranged across the stack (in
the cis to trans direction) in the order in which they
function (Kornfeld and Kornfeld, 1985). The stack prob-
ably improves the ef®ciency of post-translational modi®-
cation by concentrating subsets of these enzymes in
individual cisternal compartments so that they can work
more optimally, and it may improve the rate of transport
through the stack by facilitating the speed of transfers
between cisternae (Shorter and Warren, 1999). The
mechanism that organizes Golgi cisternae into stacks is,
however, poorly understood and is of central importance to
studies of Golgi biogenesis.
Proteins involved in stacking Golgi cisternae were ®rst
identi®ed using cell-free systems that exploited the mitotic
fragmentation of Golgi membranes (Rabouille et al.,
1995b). Fragmentation of the Golgi apparatus in animal
cells occurs at the onset of mitosis as part of a process that
is thought to aid the partitioning of these membranes
between daughter cells. At the end of mitosis, reassembly
of stacked cisternal structures occurs in each daughter cell
(Souter et al., 1993). Mimicking this process in the test-
tube led to the identi®cation of GRASP65, a protein that
was exposed and accessible to the alkylating reagent
N-ethylmaleimide (NEM) only after unstacking of Golgi
membranes in the presence of mitotic cytosol (MC) from
HeLa cells (Barr et al., 1997). Antibodies to GRASP65
inhibited re-stacking of mitotic Golgi fragments (MGFs),
as did a soluble form of GRASP65, though not after pre-
treatment with NEM (Shorter et al., 1999). This soluble
form was generated by removing the N-terminal signal for
addition of myristic acid which helps anchor GRASP65 to
Golgi membranes, as do certain members of the p24
family of putative cargo receptors and perhaps other
proteins (Kuo et al., 2000; Barr et al., 2001). GRASP65
has also been implicated in cisternal stacking in vivo by
recent studies on apoptosis (Lane et al., 2002). Cleavage of
GRASP65 by caspase-3 correlates with Golgi fragmenta-
tion, and this is inhibited, at least partially, by expression
of a caspase-resistant form of GRASP65.
GRASP65 is located in cis Golgi membranes, whereas a
second member of the GRASP family, GRASP55, is
located more towards the middle of the Golgi stack
(Shorter et al., 1999). These locations argue that the
GRASP family of proteins help determine stacking of
different cisternal layers (Pfeffer, 2001). Both GRASPs
are bound to members of the golgin family of coiled-coil
proteins. GRASP65 is bound to the C-terminal domain of
GM130 via a PDZ-like domain, whereas GRASP55 is
bound to golgin-45 (Barr et al., 1998; Short et al., 2001).
GM130 is thought to provide the base of a tether that
anchors COPI vesicles to Golgi membranes. Giantin
(another golgin) in the vesicle is linked to GM130 by a
bridging molecule, p115 (Lowe et al., 1998; Sonnichsen
et al., 1998). This tether is also involved in Golgi stacking,
at least in vitro, helping to bring cisternal membranes
together for subsequent binding by GRASP65 (Shorter and
Warren, 1999).
GRASP proteins are targets of mitotic kinases.
GRASP65 is phosphorylated both in vivo and in vitro by
cdc2/B1 (cdc2 complexed with cyclin B1) and polo-like
(plk) kinases (Lin et al., 2000). GRASP55 is phosphoryl-
A direct role for GRASP65 as a mitotically regulated
Golgi stacking factor
The EMBO Journal Vol. 22 No. 13 pp. 3279±3290, 2003
ã European Molecular Biology Organization
3279

ated in vitro by ERK2 (Jesch et al., 2001). What remains
unclear is the consequence of these phosphorylation
events. Mitotic fragmentation involves vesiculation and
tubulation of cisternae as well as cisternal unstacking
(Misteli and Warren, 1995). It is not clear whether GRASP
phosphorylation leads primarily to cisternal unstacking or
affects one or more of the other processes. It is also unclear
whether GRASPs play a direct role in cisternal stacking or
an indirect role via other, as yet uncharacterized, stacking
factors.
To address these issues, we have used our cell-free
assay and recombinant mitotic kinases to dissect out the
role played by GRASP65 phosphorylation. We then used
mitotic cells microinjected with antibodies to GRASP65 to
test the role of GRASP65 in the reassembly of Golgi stacks
after cell division. We have also used recombinant
GRASP65 alone to determine whether it has the capacity
to bind surfaces together. Our data show that GRASP65
has the properties required of a mitotically regulated
stacking protein.
Results
GRASP65 is the major target of cdc2/B1 and
plk kinases on Golgi membranes
Our earlier work had shown that GRASP65 is a prominent
phosphorylation target on rat liver Golgi (RLG) mem-
branes for mitotic kinases from HeLa cell cytosol (Barr
et al., 1997). Subsequent work by Erikson and colleagues
using cells transfected with cDNA encoding GRASP65
identi®ed cdc2/B1 and plk as the two mitotic kinases
responsible for almost all of this phosphorylation (Lin
et al., 2000). The question then was whether endogenous
GRASP65 in Golgi membranes was the main target of
these two kinases.
Puri®ed RLG membranes were incubated with MC from
HeLa cells in the presence of [g-
32
P]ATP, and labeled
proteins were visualized by SDS±PAGE and autoradio-
graphy. As shown in Figure 1A (lane 5), 8±10 proteins
were clearly labeled that were not labeled using interphase
cytosol (IC, lane 4). A similar number were detected when
RLG membranes were treated with puri®ed, recombinant
cdc2/B1 and plk kinases, and most were identical to those
phosphorylated by MC (Figure 1A, compare lanes 3 and
5). Importantly, the most strongly labeled protein had a
mol. wt of 65±70 kDa, the same as that for GRASP65.
Essentially all of this phosphorylated protein could be
removed by antibodies to GRASP65, which con®rmed its
identity (Figure 1B; Supplementary ®gure 1 available at
The EMBO Journal Online) and showed that it was the
major phosphorylated protein in Golgi membranes. Note
that GM130 co-immunoprecipitated with GRASP65,
consistent with their known interaction with each other
(Barr et al., 1997, 1998).
GRASP65 was phosphorylated by recombinant cdc2/B1
(Figure 1A and B, lane 1) and even more strongly by
recombinant plk (Figure 1A and B, lane 2). Both kinases
together labeled GRASP65 to a similar extent to that
obtained using MC, which had the same kinase activity
(Figure 1A and B, compare lanes 3 and 5). GM130 was
also labeled by both kinases (Figure 1A and B, lanes 1±3),
although the labeling was much weaker than for
GRASP65, consistent with it having far fewer phosphoryl-
ation sites. GM130 has a single phosphorylation site for
cdc2/B1 (Lowe et al., 1998) and a different one (or ones)
Fig. 1. GRASP65 is the major target of cdc2/B1 and plk kinases on
Golgi membranes. (A) Membranes were incubated in the presence of
[g-
32
P]ATP with puri®ed cdc2/B1 (lane 1), plk (lane 2), cdc2/B1 and
plk (lane 3), interphase cytosol (IC, lane 4) or mitotic cytosol (MC,
lane 5). Kinases and MC were adjusted to equivalent levels of kinase
activity. The membranes were isolated and analyzed by SDS±PAGE
followed by autoradiography. Arrows indicate phosphorylated GM130
and GRASP65. The strongly labeled protein in lanes 1 and 3 (asterisks)
is cyclin B1. (B) Membrane samples from (A) were solubilized and
GRASP65 immunoprecipitated followed by SDS±PAGE and autoradio-
graphy. Phosphorylated GM130 co-precipitated with phosphorylated
GRASP65. The additional band in lanes 2 and 3 is GST±plk (arrow-
heads) which co-precipitated with GRASP65. (C) Detection of phos-
phorylated GRASP65 by a bandshift assay. RLG membranes were
treated with either buffer (lane 1) or MC (lanes 2±4). Membranes were
isolated and treated with alkaline phosphatase (CIP) in the absence
(lane 3) or presence (lane 4) of b-glycerophosphate (b-GP), and ana-
lyzed by immunoblotting. (D) RLG membranes were incubated with
the indicated kinases and immunoblotted for GRASP65. Note that the
combination of cdc2/B1 and plk kinases phosphorylated GRASP65 to a
similar extent as MC. (E) Phosphorylation and dephosphorylation of
GRASP65 by MC and IC. RLG membranes were ®rst incubated with
MC (lanes 2±6) and then IC (lanes 3±6), in the absence or presence of
microcystin (M.C., lane 4), Inhibitor-2 (In-2, lane 5) or okadaic acid
(O.A., lane 6), followed by immunoblotting for GRASP65. Note that
microcystin and okadaic acid inhibited dephosphorylation by IC,
implicating PP2A in this process.
Y.Wang et al.
3280

for plk (not shown). Kinase-dead forms of cdc2 and plk
showed no activity towards GRASP65 or GM130, and
neither did active kinases pre-treated with kinase inhibi-
tors such as olomoucine, roscovitine and staurosporine
(not shown). Cyclin B2 in association with cdc2 labeled
the same Golgi proteins as cyclin B1, with the same
ef®ciency (not shown).
Phosphorylation of GRASP65 was also analyzed using
a bandshift assay. Treatment of RLG membranes with MC
led to a sharp increase in molecular weight on SDS±PAGE
(Figure 1C, lane 2). This shift was caused by phosphoryl-
ation, since treatment of MGFs with calf intestine alkaline
phosphatase (CIP), a non-speci®c protein phosphatase,
restored the molecular weight almost back to its original
value (lane 3). Restoration was abolished when b-glycero-
phosphate, a general phosphatase inhibitor, was included
in the reaction (lane 4, b-GP). This result validates the
bandshift assay as a means of studying the phosphoryl-
ation pattern of GRASP65.
We then tested the ef®ciency of GRASP65 phosphoryl-
ation by puri®ed kinases and MC using the bandshift
assay. Incubation of RLG membranes with puri®ed cdc2/
B1 had a modest effect (Figure 1D, lane 3) compared with
plk (lane 4), consistent with the relative contributions of
these kinases in phosphorylating GRASP65 (Lin et al.,
2000). Neither kinase alone, however, could increase the
molecular weight to the extent observed with MC (lanes 2
and 6) even when higher levels were used (not shown).
Only when the two kinases were added together was the
increase similar to that obtained using MC (Figure 1D,
compare lanes 5 and 6). Together, these data argue that
cdc2/B1 and plk kinases can largely mimic MC in its effect
on the phosphorylation pattern of Golgi proteins and that
endogenous GRASP65 is the major phosphorylation
target.
Phosphorylated GRASP65 could be dephosphorylated
by IC. The increase in molecular weight observed after
treatment of RLG membranes with MC (Figure 1E, lane 2)
could be reversed by subsequent treatment of the isolated
membranes with IC (lane 3), suggesting the presence of a
phosphatase or phosphatases that normally complete the
mitotic Golgi cycle. Microcystin inhibited dephosphoryla-
tion by IC (lane 4), suggesting that PP1 and/or PP2A was
the responsible phosphatase. Inhibitor-2 (In-2), a speci®c
peptide inhibitor of PP1, showed no effect (lane 5),
eliminating PP1 as the phosphatase. This was con®rmed
using okadaic acid, which inhibits PP2A at much lower
concentrations than PP1. Low concentrations of okadaic
acid inhibited dephosphorylation of GRASP65 (lane 6).
Microinjection of puri®ed cdc2/B1 leads to
fragmentation of the Golgi apparatus and
phosphorylation of GM130
To test the effect in vivo of the recombinant cdc2/B1 and
plk kinases we had prepared, they were microinjected into
normal rat kidney (NRK) cells. The cells were ®xed 30 min
after injection and double labeled for phosphorylated
GM130 (Figure 2, left panels, PS-25) and for total GM130
(Figure 2, right panels, GM130). Co-injected biotinylated
bovine serum albumin (BSA) was detected by ¯uores-
cently labeled NeutrAvidin to identify the injected cells
(indicated by asterisks).
Microinjection of cdc2/B1 led to GM130 phosphoryl-
ation and Golgi disassembly. More than 98% of the
injected cells showed phosphorylation of GM130, and
87% showed fragmentation of the Golgi manifested as
breakdown of the juxta-nuclear ribbon into punctate
structures (Figure 2A, cdc2/B1, Table I). Microinjection
of cdc2 without cyclin B had no effect (Table I). These
results are consistent with those obtained by Pines and
colleagues using cells transfected with cDNAs encoding
cdc2 and cyclin B (Draviam et al., 2001).
When cells were microinjected with plk, neither phos-
phorylation of GM130 at Ser25 nor breakdown of the
Golgi ribbon was detected at the resolution afforded by
light microscopy (Figure 2B, plk), consistent with work by
Malhotra and colleagues using permeabilized cells
(Sutterlin et al., 2001). Plk does phosphorylate GM130
in vitro (Figure 1A and B) but not at Ser25, the site
phosphorylated by cdc2/B1, and the one recognized by the
anti-phosphoserine antibodies, PS-25.
When both cdc2/B1 and plk kinases were microinjected,
the effect was very similar to injection of cdc2/B1 alone.
GM130 was phosphorylated in 94% of the injected cells,
and the Golgi ribbon broke down in 78% of them
(Figure 2C, cdc2/B1 + plk, Table I). This slightly lower
percentage was probably due to the lower concentration of
the cdc2/B1 protein injected. Other mitotic phenomena
were also detected, including condensation of the chromo-
somal DNA, breakdown of the nuclear envelope, re-
arrangement of the microtubule cytoskeleton and rounding
up of the cells (not shown).
We also eliminated the possibility that any of these
effects were the consequence of triggering apoptosis.
Injected cells were not labeled by antibodies to cleaved
PARP, an apoptotic marker (Lane et al., 2002)
(Supplementary ®gure 2). Apoptotic cells also degrade
GRASP65 so that this protein is no longer recognized by a
monoclonal anti-GRASP65 antibody (Lane et al., 2002).
In all cases, injected cells could be labeled with this
antibody (Supplementary ®gure 2).
Regulation of cisternal stacking by protein
phosphorylation
The consequence of phosphorylation by cdc2/B1 and plk
kinases on Golgi structure was tested using isolated rat
liver Golgi stacks. These were treated with puri®ed cdc2/
B1, plk or both kinases. Using the intersection method for
quantitation (Rabouille et al., 1995b), 55 6 7% of the
cisternae were found in stacks in starting Golgi mem-
branes (Figure 3A and F) and this was not signi®cantly
changed when they were treated with buffer alone (not
shown). After treatment with cdc2/B1, many single
cisternae were generated (Figure 3B) and the percentage
of stacked cisternae dropped to 25 6 3% (Figure 3F),
suggesting that partial unstacking had occurred. When the
cdc2/B1 inhibitor roscovitine was added, or when Golgi
stacks were treated with the kinase-dead form of cdc2/B1,
cyclin B1 or cdc2 alone, the Golgi cisternae remained
stacked (not shown).
A similar effect on unstacking was detected when Golgi
membranes were treated with plk (Figure 3C), the
percentage of stacked cisternae dropping from 55 6 7to
27 6 3% (Figure 3F). The kinase-dead form of plk had no
effect (not shown). When the Golgi stacks were treated
GRASP65 and stacking Golgi cisternae
3281

with both cdc2/B1 and plk (Figure 3D), the percentage of
stacked cisternae dropped even lower, to 11 6 4%
(Figure 3F), effectively converting most of the stacks
into single cisternae. About 70% of total membranes were
present in cisternae irrespective of the treatment, showing
that the effect of these two kinases was restricted to
unstacking.
We then tested whether single cisternae could re-stack,
after dephosphorylation of GRASP65 using IC. When
single cisternae generated by treatment with both recom-
binant kinases were incubated with IC, Golgi stacks were
rebuilt (Figure 3E). Quantitation showed that the percent-
age of stacked cisternae rose from 11 6 4to426 3%,
similar to the 55 6 7% found in starting membranes
(Figure 3F). These results suggest that the phosphorylation
cycle of GRASP65 correlates with the cisternal stacking
cycle.
Microinjection of mitotic cells with antibodies to
GRASP65
To test the role of GRASP65 in vivo, af®nity-puri®ed
antibodies, or non-speci®c IgG were microinjected into
NRK cells in metaphase. After further incubation to allow
the cells to complete cell division, the injected cells were
processed for electron microscopy (EM) (Figure 4).
Microinjection of IgG had little, if any, discernible
effect on the reassembly of Golgi stacks. Representative
images in Figure 4B and F show closely apposed and
¯attened cisternal membranes, with 4±6 cisternae/stack,
and ribbons of linked stacks (Figure 4F). In marked
contrast, microinjection of anti-GRASP65 antibodies
severely disrupted the reassembly of individual stacks.
Table I. Quantitation of the kinase microinjected NRK cells
Protein injected Cells
counted
Cells with
phosphorylated
GM130
Cells with
fragmented
Golgi
Cdc2 409 0% 0%
Cdc2/B1 218 98% 87%
plk 330 0% 0%
Cdc2/B1 + plk 215 94% 78%
The percentage of cells with phosphorylated GM130 (PS-25 positive)
and fragmented Golgi 30 min after kinase microinjection. Results are
representative of three independent experiments.
Fig. 2. Microinjection of cdc2/B1 but not plk kinases leads to GM130 phosphorylation and Golgi fragmentation. NRK cells were injected with
cdc2/B1 (A), plk (B) or cdc2/B1 and plk (C). Biotinylated BSA was used as an injection marker (asterisks). After 30 min at 37°C, cells were ®xed
and double labeled with polyclonal antibodies to phosphorylated GM130 (PS-25, left panels) and a monoclonal antibody to GM130 (GM130, right
panels). Bar, 10 mm.
Y.Wang et al.
3282

Structures ranged from tubulo-reticular networks
(Figure 4A) to highly fenestrated, cisternal-like structures
(Figure 4C). Stacked cisternae were observed, but they
were often not properly aligned (Figure 4D and E) and the
cross-sectional pro®les were more reminiscent of swollen
tubules, not ¯attened cisternae (Figure 4E). Quantitation
showed that the percentage of stacked Golgi membranes
was nearly 4-fold lower, 9.9 6 2.7% (n = 26) in cells
injected with anti-GRASP65 antibodies compared with
37 6 7.0% (n = 25) in IgG-injected cells. This dramatic
effect of anti-GRASP65 antibodies on the morphology of
individual stacks did not extend to the overall organization
of the Golgi ribbon, which appeared normal when
visualized using antibodies to GM130, mannosidase II
and TGN38 (data not shown). This implicates GRASP65
in the stacking of cisternae in individual stacks and not the
overall ribbon-like structure.
GRASP65 forms homodimers
One possible mechanism for cisternal stacking is for
GRASP65 in adjacent membranes to interact, forming
oligomers that hold the cisternae together (Barr et al.,
1998). Earlier work using gel ®ltration suggested that
GRASP65 can form dimers or trimers, but the precise
number of molecules in the complex was unclear (Barr
et al., 1998). To determine the oligomeric state, GRASP65
tagged with maltose-binding protein (MBP±GRASP65)
and His-tagged GRASP65 (His-GRASP65) were co-
expressed in bacteria and the proteins puri®ed using
af®nity columns.
As shown in Figure 5A, the ratio of MBP±GRASP65 to
His-GRASP65 in the lysate was ~1:3 (lane 1). After
puri®cation utilizing the His tag and nickel af®nity
chromatography, the ratio of the two proteins changed to
~1:8 (Figure 5A, lane 2). When this sample was puri®ed
further on an amylose column utilizing the MBP tag, this
ratio of the two tagged proteins in the complex was ~1:1
(Figure 5A, lane 3). Quantitation of the blots from three
independent experiments gave a mean ratio 6 SEM of
1.06 6 0.15.
The puri®cation was then carried out using the same two
columns in the reverse order. Once again, the ®nal ratio
was ~1:1 (Figure 5B, lane 3). This ratio was also
unaffected by differences in the expression level of the
two proteins, which ranged from 1:1 to 1:5 for MBP±
GRASP65:His-GRASP65. This eliminates the possibility
Fig. 3. Puri®ed cdc2/B1 and plk kinases unstack Golgi cisternae. (A±E) RLG stacks were either left untreated (A), or treated with cdc2/B1 (B), plk
(C) or both (D) at 37°C for 20 min. Membranes were re-isolated after kinase treatment (D) and further treated with IC for 60 min at 37°C (E).
Membranes were ®xed and processed for EM. Bar, 0.5 mm. (F) Quantitation of (A±E), by the intersection method, to estimate the percentage of single
or stacked cisternal membranes. Results represent the mean of three independent experiments 6SEM. Note the increased number of single cisternae
after kinase treatments.
GRASP65 and stacking Golgi cisternae
3283

that the complex is a tetramer or higher, even-numbered
oligomer. Such oligomers would contain the tagged
GRASP65 proteins at a ratio re¯ecting their relative
expression levels. The oligomeric state was also not the
consequence of interactions between the tags since co-
expressed MBP and His-GRASP65 did not co-purify.
Taken together, these results strongly suggest that
recombinant GRASP65 is a homodimer.
GRASP65 homodimers are resistant to treatment
with mitotic kinases
If GRASP65 homodimers link adjacent cisternae, then we
might expect them to be sensitive to treatment with mitotic
kinases. To test this possibility, puri®ed heterodimers
containing MBP±GRASP65 and His-GRASP65 were
bound to amylose beads and treated with either MC or
IC from HeLa cells, or puri®ed cdc2/B1 and/or plk kinases
(Figure 5C). Both proteins shifted in molecular weight
when treated with MC (Figure 5C, lane2), or puri®ed cdc2/
B1 and/or plk kinases (lane 5±7), showing that both
proteins were phosphorylated. No shift was detected when
the proteins were treated with IC (lane 4) or with MC
treated with the general kinase inhibitor staurosporine
(lane 3). Incubation of the beads under these conditions
released some protein from the amylose beads, but the
ratio of the two proteins, either bound to the beads or in the
Fig. 4. Antibodies to GRASP65 prevent proper re-stacking of Golgi membranes in post-mitotic cells. NRK cells in metaphase were microinjected with
af®nity-puri®ed, anti-GRASP65 antibodies (A and C±E) or non-speci®c rabbit IgGs as a control (B and F). After completion of cell division, the
injected cells were processed for EM. Results of two individual experiments are shown in (A and B) and (C±F), respectively. Note the properly
assembled stacks in cells injected with IgGs (B and F) and the disorganized Golgi structures in cells injected with antibodies to GRASP65 (A and
C±E). Bar, 0.5 mm.
Y.Wang et al.
3284

supernatant, remained unchanged irrespective of their
phosphorylation state (Figure 5C). Further binding of the
supernatant proteins to nickel or amylose beads showed
that the released proteins were still heterodimers of the two
tagged GRASP65s (not shown). These results show that
mitotic phosphorylation of recombinant GRASP65 dimers
does not break them down into monomeric units.
GRASP65 homodimers form higher order
oligomers under interphase but not mitotic
conditions
We then tested whether GRASP65 homodimers can
oligomerize further to form complexes that might help
stack cisternae. Puri®ed His-GRASP65 was incubated
with or without cdc2/B1 kinase and loaded onto glycerol
gradients. After centrifugation and fractionation,
GRASP65 was visualized by western blotting. In the
absence of kinase, GRASP65 distributed throughout
the gradient, with a peak near the top and smaller peaks
in the long tail (Figure 6A, upper panel). After treatment
with cdc2/B1 kinase, the pro®le was much simpler, with a
single peak near the top of the gradient (fraction 3,
Figure 6A, lower panel). These data show that GRASP65
forms a mixture of differently sized oligomers that
apparently can be reduced to a single oligomeric complex
in the presence of mitotic kinase.
We also tested the oligomerization of GRASP65 using
differently tagged dimers. MBP±GRASP65 and His-
GRASP65 were expressed and puri®ed separately in
bacteria, so that each should form homodimers. The two
proteins were then mixed and incubated with either buffer
alone, MC or IC. Protein complexes were isolated using
nickel beads, and bound proteins were analyzed by
immunoblotting for GRASP65.
Under control conditions, almost all MBP±GRASP65
was brought down by nickel beads, showing that it had
formed a complex with His-GRASP65 (Figure 6B, com-
pare lanes 2 and 3). Similar results were obtained in the
presence of IC, though some of the MBP±GRASP65 was
present in the unbound fraction (Figure 6B, compare
lanes 8 and 9). Treatment with MC shifted the molecular
weight of both proteins, showing that phosphorylation had
occurred (Figure 6B, compare lanes 1 and 4). Furthermore,
almost all of the MBP±GRASP65 was present in the
unbound fraction, leaving only His-GRASP65 bound to
the nickel beads (Figure 6B, compare lanes 5 and 6). These
data argue that mitotic phosphorylation of GRASP65
dimers prevents the formation of higher order oligomers.
These data were con®rmed using the recombinant
kinases. Separately puri®ed MBP±GRASP65 and His-
GRASP65 proteins were mixed and incubated together
Fig. 6. GRASP65 dimers form higher order oligomers in a cell cycle-
dependent manner. (A) His-GRASP65 was incubated in the absence or
presence of cdc2/B1, sedimented in glycerol gradients and fractions
analyzed by western blotting. Note that the mixture of GRASP65 oligo-
mers is reduced to a single oligomeric complex after treatment with
mitotic kinase. (B) Separately expressed and puri®ed MBP±GRASP65
and His-GRASP65 proteins were mixed and incubated in the presence
of buffer (control, lanes 1±3), MC (lanes 4±6) or IC (lanes 7±9). The
protein complex was then isolated using nickel beads. Equal portions of
the input (I), unbound (U) or bound (B) fraction were analyzed by
immunoblotting for GRASP65. Note the absence of MBP±GRASP65 in
the bound complexes treated with MC (lane 6). (C) As in (B) except
that puri®ed dimers were incubated with cdc2/B1 and plk alone or in
combination. Complexes were isolated on nickel beads (upper panel) or
amylose beads (lower panel) and processed as in (B). Note that bound
complexes under control conditions contained both tagged GRASP65
proteins (lane 3) whereas kinase treatment separated them (lanes 6, 9
and 12).
Fig. 5. Recombinant GRASP65 forms stable dimers. (A and B) MBP-
tagged GRASP65 and His-tagged GRASP65 were co-expressed in
E.coli and puri®ed sequentially on nickel followed by amylose columns
(A), or the reverse (B). Samples were fractionated by SDS±PAGE.
Note that the ratio of the two tagged proteins in the complex was ~1:1,
irrespective of the order of puri®cation. (C) Puri®ed heterodimers of
MBP±GRASP65/His-GRASP65 bound to amylose beads were treated
with buffer alone (lane 1), MC ( lane 2), MC and the general kinase in-
hibitor staurosporine (lane 3), IC (lane 4), cdc2/B1 (lane 5), plk (lane 6)
or both (lane 7). Beads (bound, upper panel) or supernatant (unbound,
lower panel) were analyzed by immunoblotting for GRASP65. The
extra bands in lanes 6 and 7 are cGST±plk (arrowhead) and its frag-
ments (asterisk). Note that none of the treatments changed the ratio of
the two proteins in the bound or unbound fractions.
GRASP65 and stacking Golgi cisternae
3285

with puri®ed cdc2/B1 and/or plk kinases. Protein com-
plexes were isolated using either nickel (Figure 6C, upper
panel) or amylose (lower panel) beads. Nickel beads
brought down His-GRASP65 and associated MBP±
GRASP65 under control conditions (lanes 1±3), but
MBP±GRASP65 was almost all in the unbound fraction
after treatment with any of the kinases, alone or in
combination (lanes 4±12). The ef®cacy of the kinases was
judged by the shift in molecular weight compared with
control conditions. Similar results were obtained using
amylose beads, though the ef®ciency of binding MBP±
GRASP65 was lower. Nevertheless, His-GRASP65 only
bound to the MBP±GRASP65 on the amylose beads under
control conditions. After kinase treatment, His-GRASP65
was found almost exclusively in the unbound fraction.
Together, these data argue that phosphorylation of
GRASP65 dimers by cdc2/B1 and/or plk prevents the
association of dimers into higher order oligomers.
Higher order oligomers are suf®cient to cause
aggregation of GRASP65-coated beads in a
mitotically regulated manner
The next step was to test whether the higher order
oligomers could act to link adjacent cisternal membranes.
As a model system, we chose to coat beads with GRASP65
proteins and ask whether these would then aggregate in a
manner dependent on GRASP65 oligomerization. We
chose Dynal magnetic beads since coupled proteins are
located exclusively on the surface. Puri®ed His-GRASP65
or BSA (as a control) was coupled by direct cross-linking.
The coated beads were incubated with either IC or MC,
or puri®ed cdc2/B1 and plk kinases, or BSA (as a control).
The beads were then transferred to glass slides and
observed under bright-®eld illumination. As shown in
Figure 7A, in the presence of BSA, beads coated with
His-GRASP65 formed aggregates, whereas control beads
formed none. Quantitation con®rmed these observations,
Fig. 7. GRASP65 can form trans oligomers. (A) Puri®ed His-GRASP65 (upper panel), or BSA (as control, lower panel) was covalently coupled to
Dynal beads and incubated with BSA, IC, MC or cdc2/B1 and plk (kinases). After incubation, the beads were placed on glass slides and random ®elds
photographed. A representative image of each condition is shown. Bar, 500 mm. (B) As in (A) except that the GRASP65 beads were ®rst aggregated
using IC and then treated with either BSA, MC or a combination of cdc2/B1 and plk (kinases). Kinase-treated beads were treated further with IC
(IC®kinases®IC). Bar, 500 mm. (C and D) Quantitation of (A) and (B) and MBP±GRASP65 beads. Note that aggregates only formed in the presence
of buffer and IC and that these aggregates were disassembled reversibly by MC and recombinant kinases.
Y.Wang et al.
3286

showing that 24% of His-GRASP65 beads were found in
aggregates compared with 0.7% of control beads
(Figure 7C). Aggregates were de®ned as clusters of ®ve
or more beads.
Aggregation was even more dramatic when the beads
were incubated with IC. More than 91% of His-GRASP65-
coated beads were found in the aggregates compared with
1% of the control beads (Figure 7C). When the beads were
incubated in the presence of MC (Figure 7A, MC), the
percentage of beads in aggregates was found to be very
low, <3% for the His-GRASP65 beads compared with
0.9% for the control beads (Figure 7C, column 3). Similar
results were also obtained using the puri®ed kinases
(Figure 7A, kinases). About 8% of the beads were found in
the aggregates for His-GRASP65 beads, whereas no
aggregates were detected for the control beads
(Figure 7C, column 4).
Other controls were also carried out, treating His-
GRASP65 beads with reagents known to inhibit cisternal
stacking. NEM alkylates GRASP65 and prevents cisternal
stacking (Barr et al., 1997). It also prevented the
aggregation of the beads in the presence of IC. Only
28 6 3% of NEM-treated beads aggregated in the presence
of IC compared with 84 6 5% in the absence of such
treatment.
Anti-GRASP65 antibodies and soluble GRASP65 also
inhibit cisternal stacking (Barr et al., 1997), and they
inhibited the aggregation of His-GRASP65 beads.
Aggregation in the presence of anti-GRASP65 antibodies
was 28 6 5% compared with 77 6 4% for control
antibodies. Only 9 6 2% aggregated in the presence of
soluble His-GRASP65 compared with 43 6 6% in the
presence of a control His-tagged protein, the tail and acidic
domain of the tethering protein p115 (Dirac-Svejstrup
et al., 2000).
The pattern of results was not affected by the nature of
the tag attached to the N-terminus of GRASP65. MBP±
GRASP65 beads gave results very similar to those
obtained using His-GRASP65 beads (Figure 7C and D).
The manner in which the proteins were coupled to the
beads also had no effect. Antibodies to the tags gave the
same pattern of results as cross-linking (not shown). The
size of the beads was also unimportant, and differently
sized beads gave rise to mixed aggregates (not shown).
Taken all together, these data show that immobilized
GRASP65 dimers can link structures by forming trans
oligomers and that this is prevented by all treatments
known to affect cisternal stacking in vitro.
Aggregates of GRASP65-coated beads can be
disaggregated by mitotic kinases and then
re-aggregated
Since mitotic kinases act on pre-existing Golgi stacks to
unstack them (Figure 3), we asked whether the GRASP65
bead aggregates could be disaggregated by mitotic
kinases. To test this possibility, His-GRASP65 or control
(BSA-coated) beads were ®rst incubated with IC
(Figure 7B, IC) then washed and incubated with buffer
either containing BSA (IC®BSA), MC (IC®MC) or
puri®ed cdc2/B1 and plk (IC®kinases).
When treated with BSA, most of the aggregates
remained, the percentage of His-GRASP65 aggregates
falling from 91 to 73% (Figure 7B and D). In marked
contrast, treatment with MC lowered the percentage of
His-GRASP65 aggregates to only 12%. The recombinant
kinases were even more effective, lowering the percentage
of His-GRASP65 aggregates to a mere 1%. Less than 5%
of the control beads were found in aggregates irrespective
of the treatment (Figure 7B and D, BSA beads). A similar
pattern of results was obtained using MBP±GRASP65
beads (Figure 7C and D).
Finally, we tested the possibility that the beads
disaggregated by the two recombinant kinases could be
re-aggregated after dephosphorylation. As shown in
Figure 7B, treatment of beads with IC to dephosphorylate
GRASP65 (Figure 1E) caused the beads to aggregate. The
percentage aggregation rose from 1 6 1% (after treatment
with kinases) to 93 6 5% (after treatment with IC). These
data provide further corroboration that the system is
mimicking the mitotic stacking cycle of Golgi cisternae
mediated by the phosphorylation state of GRASP65.
Discussion
GRASP65 was ®rst identi®ed as a Golgi protein, whose
function was sensitive to the alkylating agent, NEM, but
only after disassembly of Golgi stacks in the presence of
MC (Barr et al., 1997). Since NEM treatment of Golgi
membranes after disassembly (but not before) was known
to inhibit subsequent reassembly of stacked membranes
(Rabouille et al., 1995a), this suggested that GRASP65
might be involved in the stacking process. This was
con®rmed by subsequent experiments using soluble,
recombinant GRASP65 or GRASP65 antibodies, which
inhibited cisternal stacking but not cisternal regrowth
(Barr et al., 1997). Further study also showed that cleavage
of GRASP65 during apoptosis facilitates Golgi fragmen-
tation (Lane et al., 2002). These experiments did not,
however, determine whether GRASP65 played a direct
role in cisternal stacking, or aided the process indirectly
through as yet unidenti®ed stacking factors. The possibil-
ity of an indirect role was strengthened recently by work
suggesting that GRASP proteins might act as a checkpoint
control in budding yeast (Norman et al., 1999), which in
animal cells might help prevent entry of cells into mitosis
until after the Golgi ribbon is fragmented (Sutterlin et al.,
2002).
We ®rst studied the effect of cdc2/B1 and plk on RLG
membranes. These two kinases are responsible for most of
the mitotic phosphorylation of GRASP65 in cells trans-
fected with cDNA encoding this protein (Lin et al., 2000).
Using
32
P labeling and bandshift assays, we showed that
endogenous GRASP65 in Golgi membranes is phos-
phorylated by these two kinases almost as well as by using
MC. Furthermore, GRASP65 is the major target of these
two kinases in isolated RLG membranes. Most of the shift
in molecular weight of endogenous GRASP65 was caused
by plk, not cdc2/B1 (Figure 1D), in agreement with the
experiments by Erikson and colleagues, mapping the
different phosphorylation site (Lin et al., 2000). In
contrast, recombinant GRASP65 was shifted more by
cdc2/B1, not plk (Figures 5C and 6C). We do not yet
understand the reason for this, though it might have to do
with the absence of GM130, which is tightly bound to
endogenous GRASP65, via a PDZ-like motif at the
C-terminus (Barr et al., 1998).
GRASP65 and stacking Golgi cisternae
3287

We then studied the effect of these kinases on cisternal
stacking. Puri®ed RLG stacks treated with recombinant
kinases released single cisternae. Either cdc2/B1 or plk
were competent, but the two together had the greatest
effect, lowering the percentage of stacked cisternae from
55 to 11%, a 5-fold decrease. Since these conditions of
incubation did not lead to vesiculation or tubulation of
cisternae (the percentage of membrane in cisternae
remained unchanged at ~70% irrespective of the treat-
ment), these data show that treatment with cdc2/B1 and
plk suf®ces for separation of stacked cisternae. The key
role played by phosphorylation was demonstrated by
the fact that kinase-dead forms of cdc2 and a range
of kinase inhibitors prevented cisternal unstacking.
Dephosphorylation by IC also led to re-stacking of single
cisternae. The fact that GRASP65 is the major phos-
phorylation target in RLG membranes further argues that
this protein mediates the effect triggered by kinases and
phosphatases.
We next tested the role of GRASP65 in vivo, since its
role as a stacking factor has relied solely on experiments
using our cell-free reassembly assay (Barr et al., 1997).
Antibodies were microinjected into NRK cells in the early
phases of mitosis and the cells allowed to complete mitosis
and cell division before analysis by electron microscopy.
Microinjection of non-speci®c IgGs had no discernible
effect on reassembly, whereas antibodies to GRASP65
severely disrupted this process. Qualitatively, the Golgi
membranes were more swollen, fenestrated and improp-
erly aligned. Quantitatively, the percentage of stacked
membrane was nearly 4-fold lower than control cells. The
fact that some stacking does occur might re¯ect the action
of GRASP55, which stacks medial rather than cis cisternae
(Shorter et al., 1999). The absence of any discernible
effect on the Golgi ribbon at the light level suggests that
the role of GRASP65 is restricted to the reconstruction of
individual stacks.
We next tried to mimic stacking using puri®ed
GRASP65 so as to determine whether this protein alone
could mediate this process. We ®rst determined the
oligomeric state of GRASP65 using differently tagged
proteins. Co-expression showed that recombinant
GRASP65 is dimeric, but these dimers could not be
reduced to the monomeric state using either MC or
recombinant kinases, suggesting that dimerization alone is
unlikely to be the mechanism of Golgi stacking as
previously speculated (Barr et al., 1998). We further
showed, however, that these dimers could form higher
order oligomers whose formation was sensitive to phos-
phorylation by MC and kinases. This suggested that
GRASP65 dimers on one membrane might be able to
interact with GRASP65 dimers on another and so stack
cisternae.
We tested this idea using a model system comprising
Dynal beads coated with recombinant GRASP65. This was
attached to the surface of the beads either by cross-linking
or by using antibodies to the tags. We used both MBP- and
His-tagged GRASP65 to eliminate effects that might be
due to the nature of the tags. In all cases, the results were
the same. Beads coated with tagged GRASP65 formed
extensive aggregates. Beads coated with BSA did not.
Interestingly, the size of the aggregates formed with
GRASP65 beads was much larger in the presence of IC
than BSA. This suggests that there are components in IC
that enhance the formation of aggregates, and this assay
might provide the means of identifying them.
GRASP65 beads did not aggregate after treatments that
were known to prevent cisternal stacking in vitro.
Aggregation was inhibited after treatment with NEM or
antibodies to GRASP65. Aggregation was also inhibited in
the presence of soluble, recombinant GRASP65. Most
importantly, aggregation of GRASP65 beads did not occur
in the presence of MC or the two mitotic kinases.
Aggregates formed in the presence of IC could even be
disaggregated by treatment with either MC or the two
kinases. It also proved possible to disaggregate the beads
with the two kinases and then re-aggregate them using IC
(Figure 7B). This latter experiment exactly follows the
conditions we used to unstack and re-stack RLG mem-
branes (Figure 3), arguing that the aggregation/disaggre-
gation cycle of the beads is faithfully recapitulating the
stacking/unstacking cycle of Golgi membranes.
In conclusion, these data provide the best evidence to
date that GRASP65 plays a direct role in cisternal
stacking. We envisage dimers on adjacent cisternal
membranes interacting to form higher order, trans
oligomers. Direct phosphorylation by cdc2/B1 and plk
kinases would break up the trans oligomers, leading to
cisternal unstacking. Subsequent dephosphorylation
would re-stack the cisternae. GRASP65 is restricted to
cis cisternae, so this does not explain why unstacking is
nearly complete (Barr et al., 1997). GRASP55, however, is
located more towards the middle of the stack (Shorter
et al., 1999) and can also be mitotically phosphorylated by
several kinases including cdc2/B1 (Y.Wang and
G.Warren, unpublished results). Unstacking might there-
fore be the consequence of phosphorylation of both
GRASP family members. Further work will determine
whether they are suf®cient to stack Golgi cisternae, or just
necessary.
Materials and methods
Reagents
All reagents were from Sigma Co., Roche or Calbiochem, unless
otherwise stated. The following antibodies were used: polyclonal
antibodies against GM130, phosphorylated GM130 (PS-25, M.Lowe)
and MBP (NEB); monoclonal antibodies against GM130 (Transduction
Laboratories), GRASP65 (F.Barr), a-tubulin (K.Gull) and RGS-His
(Qiagen). Secondary antibodies were from Molecular Probes. Antibodies
against GST±GRASP65 (amino acids 202±451) were raised in rabbits and
puri®ed using His-GRASP65.
Puri®cation of kinases
Baculoviruses encoding human cyclin B1, cyclin B2, cyclin A, cdc2 wild-
type (cdc2WT), cdc2 kinase-dead form (cdc2KD, K33R) and cdc2
constitutively active form (cdc2AF, T14A/Y15F) were kindly provided
by M.Jackman (Cambridge, UK). His-tagged human cdc2±cyclin
complexes, and single cyclins or cdks were expressed in Hi-5 cells
using a baculoviral system (Krude et al., 1997) and puri®ed on nickel (Ni-
NTA) columns. The constitutively active form of cdc2±cyclin B1complex
(cdc2/B1) was used in most experiments. Histone H1 kinase activity was
measured as described (Lowe et al., 1998).
Baculoviruses and plasmids encoding GST-tagged human plkWT
(wild-type), plkKD (kinase-dead form, K82M) or plkDC (amino acids 1±
356) were kindly provided by Q.Y.Lin (Harvard). Proteins were
expressed in Hi-5 insect cells or in BL21(DE3)CodonPlus-RIL bacteria
(Stratagene) and puri®ed using glutathione±Sepharose (Amersham-
Phamacia) (Lin et al., 2000).
Y.Wang et al.
3288

The speci®c activity of plk was measured using His-GRASP65 as the
substrate. Plk in TBMD buffer (Lin et al., 2000) was mixed with 0.5 mCi/
ml[g-
32
P]ATP, 0.2 mg/ml His-GRASP65 and 25 mM ATP. After 30 min at
37°C, His-GRASP65 was pulled down by nickel beads and analyzed by
SDS±PAGE and phosphoimaging.
Phosphorylation and dephosphorylation of Golgi proteins
A5mg aliquot of RLG was mixed with 500 mg of HeLa cytosol, or
equivalent kinase activity as described (Lowe et al., 1998). For [g-
32
P]ATP labeling, 0.5 mCi/mlof[g-
32
P]ATP and 5 mM ATP were used.
After 60 min at 37°C, the membranes were pelleted through 0.4 M
sucrose at 55 000 r.p.m. for 30 min in a TLA55 rotor (Beckman). Samples
were analyzed by immunoprecipitation, SDS±PAGE, autoradiography
and/or phosphoimaging.
Phosphorylation of GRASP65 was also analyzed by a bandshift assay.
A5mg aliquot of MGFs was treated with 20 U of CIP or 100 mgofICat
37°C for 60 min and analyzed by immunoblotting. In some reactions,
phosphatase inhibitor b-glycerophosphate (50 mM), microcystin LR
(0.5 mM), In-2 (0.5 mM) or okadaic acid (0.1 mM) was included.
Microinjection experiments
NRK cells were injected with: (i) 4.8 mg/ml plk; (ii) 0.7 mg/ml cdc2
(without cyclin); (iii) 2.5 mg/ml cdc2/B1; and (iv) 1.2 mg/ml cdc2/
B1 + 2.4 mg/ml plk. Biotinylated BSA followed by Alexa 350
NeutrAvidin was used as the co-injection marker. After 30 min at
37°C, cells were processed for immuno¯uorescence (Seemann et al.,
2000).
Cells in metaphase were injected with af®nity±puri®ed GRASP65
antibodies (4 mg/ml) or non-speci®c rabbit IgGs, and incubation
continued for 3 h. Fluorescein isothiocyanate (FITC)±dextran was co-
injected so that the divided cells could be identi®ed under the
¯uorescence microscope and the uninjected cells were removed by
scraping with a microinjection needle. These divided cells were then
processed for EM (Seemann et al., 2000).
Stacked Golgi membranes were quantitated using the intersection
method (Rabouille et al., 1995b). Pro®les of Golgi cisternae, vesicles and
tubules were identi®ed using morphological criteria (Souter et al., 1993)
and were readily distinguishable from ER and organelles on the endocytic
pathway. Stacked regions were de®ned as Golgi membranes <15 nm
apart and >150 nm in length. This length was chosen to exclude tethered
vesicles from the calculation.
Cell-free Golgi disassembly and reassembly assay
Reactions were performed with cytosol (Rabouille et al., 1995b) or
equivalent kinase activities at 37°C for 20 min. For reassembly, MGFs
were collected through 0.4 M sucrose in KHM buffer (Shorter and
Warren, 1999), treated with IC at 37°C for 60 min and processed for EM.
Membrane pro®les were quantitated by the intersection method
(Rabouille et al., 1995b).
GRASP65 oligomerization
GRASP65 cDNA in pMAL-2CX (NEB) and pET30a (Novagen) were co-
transformed into BL21(DE3)Gold bacteria. The proteins were puri®ed on
nickel (Qiagen) or amylose (NEB) agarose. This was followed by a
second puri®cation on the alternative resin. Puri®ed MBP±GRASP65 and
His-GRASP65 dimers on nickel or amylose beads were incubated with
MC, IC or kinases for 60 min at 37°C.
To test oligomerization of GRASP65, separately expressed and
puri®ed MBP±GRASP65 and His-GRASP65 were treated with IC, MC
or kinases. The protein complex was isolated using amylose beads and
analyzed by immunoblotting.
For sedimentation analysis, His-GRASP65 puri®ed on a nickel column
followed by gel ®ltration (Superose-6 HR10/30, Amersham-Phamacia)
was incubated at 37°C for 2 h in gradient buffer [25 mM HEPES-KOH
pH 7.4, 150 mM KCl, 5 mM MgCl
2
, 1 mM dithiothreitol (DTT)] with or
without cdc2/B1. Glycerol gradients [10±35% (w/v)] were centrifuged
(2 h, 65 000 r.p.m., VTI65.1 rotor) and fractions analyzed by western
blotting.
Aggregation of GRASP65-coated beads
His- and MBP±GRASP65, or BSA were clari®ed by ultracentrifugation
and cross-linked to Dynalbeads M500. The beads were incubated for
60 min at 37°C with IC (+staurosporine), MC or kinases (+microcystine
LR) in KHM with nocodazole and observed under a microscope. NEM
treatments were carried out as described (Barr et al., 1997). In some
experiments, bead aggregates were washed with KHM buffer, incubated
with IC at 37°C for 60 min, washed again and treated further with MC or
kinases at 37°C for 60 min.
For quantitation, single beads or aggregated beads from 12 random
pictures of each sample were counted. Aggregates were de®ned as those
with >5 beads since all controls had <4 beads. For large aggregates, only
visible, surface beads were counted; therefore, the number of beads in
these aggregates was underestimated.
Supplementary data
Supplementary data are available at The EMBO Journal Online.
Acknowledgements
We thank M.Jackman for the baculoviruses, cyclins and cdc2, Q.Y.Lin
for the baculoviruses and cDNAs for plk, M.Lowe, K.Gull and T.Hunt for
antibodies, J.Lee for constructing the MBP-GRASP65 plasmid, M.Beard
and H.Meyer for critical reading of the manuscript, C.Horensavitz for
help with the EM, and the entire Warren/Mellman group for discussions
and support. This work was funded by the NIH.
References
Barr,F.A., Puype,M., Vandekerckhove,J. and Warren,G. (1997)
GRASP65, a protein involved in the stacking of Golgi cisternae.
Cell, 91, 253±262.
Barr,F.A., Nakamura,N. and Warren,G. (1998) Mapping the interaction
between GRASP65 and GM130, components of a protein complex
involved in the stacking of Golgi cisternae. EMBO J., 17, 3258±3268.
Barr,F.A., Preisinger,C., Kopajtich,R. and Korner,R. (2001) Golgi matrix
proteins interact with p24 cargo receptors and aid their ef®cient
retention in the Golgi apparatus. J. Cell Biol., 155, 885±891.
Dirac-Svejstrup,A.B., Shorter,J., Waters,M.G. and Warren,G. (2000)
Phosphorylation of the vesicle-tethering protein p115 by a casein
kinase II-like enzyme is required for Golgi reassembly from isolated
mitotic fragments. J. Cell Biol., 150, 475±488.
Draviam,V.M., Orrechia,S., Lowe,M., Pardi,R. and Pines,J. (2001) The
localization of human cyclins B1 and B2 determines CDK1 substrate
speci®city and neither enzyme requires MEK to disassemble the Golgi
apparatus. J. Cell Biol., 152, 945±958.
Farquhar,M.G. and Palade,G.E. (1981) The Golgi apparatus (complex)Ð
(1954±1981)Ðfrom artifact to center stage. J. Cell Biol., 91, 77s±
103s.
Jesch,S.A., Lewis,T.S., Ahn,N.G. and Linstedt,A.D. (2001) Mitotic
phosphorylation of Golgi reassembly stacking protein 55 by mitogen-
activated protein kinase ERK2. Mol. Biol. Cell, 12, 1811±1817.
Kornfeld,R. and Kornfeld,S. (1985) Assembly of asparagine-linked
oligosaccharides. Annu. Rev. Biochem., 54, 631±664.
Krude,T., Jackman,M., Pines,J. and Laskey,R.A. (1997) Cyclin/Cdk-
dependent initiation of DNA replication in a human cell-free system.
Cell, 88, 109±119.
Kuo,A., Zhong,C., Lane,W.S. and Derynck,R. (2000) Transmembrane
transforming growth factor-a tethers to the PDZ domain-containing,
Golgi membrane-associated protein p59/GRASP55. EMBO J., 19,
6427±6439.
Lane,J.D., Lucocq,J., Pryde,J., Barr,F.A., Woodman,P.G., Allan,V.J. and
Lowe,M. (2002) Caspase-mediated cleavage of the stacking protein
GRASP65 is required for Golgi fragmentation during apoptosis. J.
Cell Biol., 156, 495±509.
Lin,C.Y., Madsen,M.L., Yarm,F.R., Jang,Y.J., Liu,X. and Erikson,R.L.
(2000) Peripheral Golgi protein GRASP65 is a target of mitotic polo-
like kinase (Plk) and Cdc2. Proc. Natl Acad. Sci. USA, 97, 12589±
12594.
Lowe,M., Rabouille,C., Nakamura,N., Watson,R., Jackman,M.,
Jamsa,E., Rahman,D., Pappin,D.J. and Warren,G. (1998) Cdc2
kinase directly phosphorylates the cis-Golgi matrix protein GM130
and is required for Golgi fragmentation in mitosis. Cell, 94, 783±793.
Misteli,T. and Warren,G. (1995) A role for tubular networks and a COP
I-independent pathway in the mitotic fragmentation of Golgi stacks in
a cell-free system. J. Cell Biol., 130, 1027±1039.
Norman,T.C. et al. (1999) Genetic selection of peptide inhibitors of
biological pathways. Science, 285, 591±595.
Pfeffer,S.R. (2001) Constructing a Golgi complex. J. Cell Biol., 155,
873±875.
Rabouille,C., Levine,T.P., Peters,J.M. and Warren,G. (1995a) An NSF-
GRASP65 and stacking Golgi cisternae
3289

like ATPase, p97 and NSF mediate cisternal regrowth from mitotic
Golgi fragments. Cell, 82, 905±914.
Rabouille,C., Misteli,T., Watson,R. and Warren,G. (1995b) Reassembly
of Golgi stacks from mitotic Golgi fragments in a cell-free system. J.
Cell Biol., 129, 605±618.
Rambourg,A. and Clermont,Y. (1997) Three-dimentional structure of the
Golgi apparatus in mammalian cells. In Berger,E.G. and Roth,J. (eds),
The Golgi Apparatus. Birkhauser Verlag, Basel, Switzerland, pp. 37±
61.
Seemann,J., Jokitalo,E.J. and Warren,G. (2000) The role of the tethering
proteins p115 and GM130 in transport through the Golgi apparatus
in vivo. Mol. Biol. Cell, 11, 635±645.
Short,B., Preisinger,C., Korner,R., Kopajtich,R., Byron,O. and Barr,F.A.
(2001) A GRASP55±rab2 effector complex linking Golgi structure to
membrane traf®c. J. Cell Biol., 155, 877±883.
Shorter,J. and Warren,G. (1999) A role for the vesicle tethering protein,
p115, in the post-mitotic stacking of reassembling Golgi cisternae in a
cell-free system. J. Cell Biol., 146, 57±70.
Shorter,J., Watson,R., Giannakou,M.E., Clarke,M., Warren,G. and
Barr,F.A. (1999) GRASP55, a second mammalian GRASP protein
involved in the stacking of Golgi cisternae in a cell-free system.
EMBO J., 18, 4949±4960.
Sonnichsen,B., Lowe,M., Levine,T., Jamsa,E., Dirac-Svejstrup,B. and
Warren,G. (1998) A role for giantin in docking COPI vesicles to Golgi
membranes. J. Cell Biol., 140, 1013±1021.
Souter,E., Pypaert,M. and Warren,G. (1993) The Golgi stack
reassembles during telophase before arrival of proteins transported
from the endoplasmic reticulum. J. Cell Biol., 122, 533±540.
Sutterlin,C., Lin,C.Y., Feng,Y., Ferris,D.K., Erikson,R.L. and
Malhotra,V. (2001) Polo-like kinase is required for the
fragmentation of pericentriolar Golgi stacks during mitosis. Proc.
Natl Acad. Sci. USA, 98, 9128±9132.
Sutterlin,C., Hsu,P., Mallabiabarrena,A. and Malhotra,V. (2002)
Fragmentation and dispersal of the pericentriolar Golgi complex is
required for entry into mitosis in mammalian cells. Cell, 109,
359±369.
Received November 6, 2002; revised April 28, 2003;
accepted May 12, 2003
Y.Wang et al.
3290