Passage through the Golgi is necessary for Shiga toxin
B subunit to reach the endoplasmic reticulum
Jenna McKenzie
1
, Ludger Johannes
2,3
, Tomohiko Taguchi
4
and David Sheff
1
1 Department of Pharmacology, Carver College of Medicine, University of Iowa, Iowa City, IA, USA
2 Institut Curie, Centre de Recherche, Laboratoire Trafic, Signalisation et Ciblage Intracellulaires, Paris, France
3 CNRS UMR144, Paris, France
4 Department of Biochemistry, Osaka University Graduate School of Medicine, Japan
Shiga toxin (Stx) is a bacterial exotoxin responsible for
an estimated 165 million annual cases of severe dysen-
tery worldwide [1]. The toxin attacks cytosolic targets
in mammalian cells. To reach these targets, the toxin
navigates a retrograde pathway that passes sequentially
through the plasma membrane, endosomes, Golgi and
endoplasmic reticulum (ER) [2–5]. Passage through the
Golgi appears to be rate limiting on this pathway,
resulting in prominent labeling of this organelle. How-
ever, such prominent labeling may be misleading.
Recycled transferrin was assumed to pass sequentially
through the early endosomes (EEs) and recycling endo-
somes (REs) based on prominent labeling of the RE at
later time points [5,6]. It later became evident that the
majority of transferrin actually bypasses the RE. The
same may be true for Golgi passage of Stx. Empirical
data supporting a requirement for passage through the
Golgi is lacking. Indeed, treatment with brefeldin A
provides protection against the holotoxin, suggesting
involvement of the Golgi. However, that protection is
incomplete, suggesting that Golgi passage may be
favored but not required [7,8]. Furthermore, other
toxins, such as diphtheria toxin, bypass the Golgi and
ER by escaping the endosomal compartment directly
into the cytosol [9]. SV40 virus is internalized into a spe-
cialized compartment which can communicate directly
Keywords
endosomes; Golgi; membrane traffic;
retrograde traffic; Shiga toxin
Correspondence
D. Sheff, Department of Pharmacology,
Carver College of Medicine, University of
Iowa, Iowa City, IA 52242-2600, USA
Fax: +1 319 335 8930
Tel: +1 319 335 7705
E-mail: david-sheff@uiowa.edu
(Received 11 November 2008, revised 4
January 2009, accepted 7 January 2009)
doi:10.1111/j.1742-4658.2009.06890.x
Both Shiga holotoxin and the isolated B subunit, navigate a retrograde
pathway from the plasma membrane to the endoplasmic reticulum (ER) of
mammalian cells to deliver catalytic A subunits into the cytosol. This route
passes through early ⁄recycling endosomes and then through the Golgi.
Although passage through the endosomes takes only 30 min, passage
through the Golgi is much slower, taking hours. This suggests that Golgi
passage is a key step in retrograde traffic. However, there is no empirical
data demonstrating that Golgi passage is required for the toxins to enter
the ER. In fact, an alternate pathway bypassing the Golgi is utilized by
SV40 virus. Here we find that blocking Shiga toxin B access to the entire
Golgi with AlF
4
)
treatment, temperature block or subcellular surgery
prevented Shiga toxin B from reaching the ER. This suggests that there is
no direct endosome to ER route available for retrograde traffic. Curiously,
when Shiga toxin B was trapped in endosomes, it entered the cytosol
directly from the endosomal compartment. Our results suggest that traffick-
ing through the Golgi apparatus is required for Shiga toxin B to reach the
ER and that diversion into the Golgi may prevent toxin escape from endo-
somes into the cytosol.
Abbreviations
BFA, brefeldin A; EE, early endosome; ER, endoplasmic reticulum; MEM, minimal Eagle’s medium; PDI, protein disulfide isomerase; RE,
recycling endosome; Stx, Shiga toxin; StxB, Shiga toxin B; Tfn, transferrin; TfnR, transferrin receptor; TGN, trans-Golgi network; WGA,
wheatgerm agglutinin.
FEBS Journal 276 (2009) 1581–1595 Journal compilation ª2009 FEBS. No claim to original US government works 1581
with the ER, bypassing endosomes and Golgi [10].
There may even be alternative retrograde pathways
between the endosomes and ER that either include or
bypass the Golgi, where the majority of traffic nor-
mally passes through the Golgi. To investigate these
possibilities, we examined the fate of Stx where access
to the Golgi was blocked.
Stx is secreted by Shigella dysenteriae. It is highly
homologous to the Shiga-like toxins (also termed vero-
toxins) secreted by enterohemorrhagic strains of Esc-
herichia coli. Stx is a member of the A-B
5
family of
toxins, which are composed of one enzymatic A sub-
unit, noncovalently bound to a B subunit composed of
a homopentamer of B fragments [11]. The Stx A sub-
unit is an rRNA N-glycosidase, which stops protein
synthesis and causes cell death [12]. The A subunit
must be delivered to the host-cell cytosol to encounter
its ribosomal substrate. To reach this destination, it is
carried by a homopentameric B subunit (StxB) along a
retrograde pathway from the plasma membrane
through the EE ⁄RE to the Golgi and the ER. Stx
takes advantage of trafficking through the Golgi to
facilitate cleavage and activation of the catalytic
A subunit by trans-Golgi network (TGN) resident
furin protease [13]. The catalytic domain remains
attached to the anchor domain by a disulfide bridge
that is cleaved when the complex enters the cytosol.
Entry of the catalytic A subunit into the cytosol is via
retrotranslocation [14–16]. The B subunit initially gains
entry to cells by binding the neutral glycosphingolipid,
globotriaosyl ceramide (Gb3 or CD77) at the cell
surface [17]. Bound toxin is endocytosed via both
clathrin-dependent and -independent mechanisms and
is delivered to EEs [18–20]. There is no known protein
receptor for Stx B subunit (StxB), and the mechanism
by which it is recruited into clathrin-coated pits
remains unknown. StxB binding to Gb3 at the cell sur-
face induces changes in plasma membrane topology
resulting in the formation of tubular invaginations that
facilitate internalization [21]. It remains to be deter-
mined whether this toxin-induced pathway or clathrin-
mediated endocytosis is predominant in normal cells.
In both cases, newly internalized StxB appears to be
delivered to EEs. StxB binding is not a passive process.
Binding and endocytosis of the toxin is accompanied
by activation of Syk kinase and activation of microtu-
bule networks, which facilitate transport into the cell
[22,23].
Passage of Stx through the EEs ⁄REs is well docu-
mented and involves many proteins that are now being
identified [3,9,24]. Two Rab GTPases, Rab11a and
Rab6A¢, regulate retrograde traffic of Stx from the
EEs ⁄REs to the Golgi, suggesting that this is a
regulated vesicular trafficking process [25,26]. In addi-
tion, components of the retromer complex, specifically
sorting nexins 1 and 2 and Vps26 are required for traf-
fic of Stx through the endosomes, but it is still unclear
if this mediates an intra-endosomal step or if they are
required for delivery to the TGN [27–29]. Delivery to
the TGN does appear to involve the GARP complex,
first identified in yeast as mediator of retrograde traffic
into the Golgi [30]. It is clear that Stx does not pass
through the late endosomes [24]. Instead, direct trans-
port to the TGN is mediated by syntaxin 5, syntaxin 6,
and syntaxin 16, a pathway that is shared by the
endogenous protein TGN38, which cycles between
Golgi and plasma membrane via REs [31,32]. Unlike
TGN38, traffic of StxB from endosomes to Golgi is
dependent upon the Golgin, GCC185 [31,33–35].
Here we examine the traffic of StxB which follows
the same route as the holotoxin through the retrograde
trafficking pathway [15,36]. We perturbed access to the
Golgi by an AlF
4
)
treatment, temperature block and
subcellular surgery to examine whether there exit
routes for StxB to bypass the Golgi while trafficking
from endosomes to ER. Using these systems, we deter-
mined that Golgi transit is required for trafficking to
the ER.
Results
StxB co-localizes with transferrin-positive
endosomes
We first sought to establish a time-line for retrograde
traffic of StxB in green monkey kidney BSC-1 cells.
These cells were selected due to their distinct endo-
somal and Golgi morphologies that allow ready visual
identification. Like HeLa cells, different strains of
BSC-1 cells show different affinities for StxB. Our lab-
oratory strain (a gift from I. Mellman) binds StxB
readily. Another strain reported by Spooner et al. [37]
does not. For co-localization studies, cells were
infected with adenovirus containing human transferrin
receptor (TfnR), a well-studied marker of the endo-
cytic recycling pathway [5,38]. This infection did not
alter the morphology of internalized StxB observed in
uninfected cells (not shown). Cells were labeled on ice
with both Cy3–StxB and Alexa 488–transferrin (Tfn)
for 30 min. Internalization of both labels was per-
formed at 37 C in label-free medium for the indicated
times (Fig. 1A). After 5 min, Tfn was in peripheral
puncta representing EEs (Fig. 1A; 5 min) [5]. StxB
co-localized with Tfn throughout the EEs. This
suggested that although internalization of StxB may be
through clathrin-dependent or -independent mechanisms,
Shiga toxin in the Golgi J. McKenzie et al.
1582 FEBS Journal 276 (2009) 1581–1595 Journal compilation ª2009 FEBS. No claim to original US government works
they converge on the EEs [39,40]. After 10 min, StxB
and Tfn co-localized in both the peripheral EEs and a
perinuclear organelle, identified by Tfn pulses as the
RE (Fig. 1A; 10 min) [41]. After 30 min, Tfn primarily
labeled the endosomes, although the signal was weaker
due to recycling of Tfn into the media, whereas StxB
had entered a separate perinuclear structure (Fig. 1A;
30 min). This structure had the appearance of a Golgi
ribbon in these cells. The difference in localization was
more obvious after 45 min (Fig. 1A; 45 min arrow
indicates transferrin-containing endosomes). At the
later times, Tfn had recycled out of the endosomes and
was no longer clearly visible although StxB remained
in Golgi morphology (Fig. 1A; 60, 120 min) [38,42].
At 180 min, the internalized StxB took on a lacy
appearance typical of the ER (Fig. 1A), suggesting
that a substantial amount of the toxin had been deliv-
ered to the ER [43]. Thus, endocytosed StxB was deliv-
ered into the endocytic recycling pathway within
5 min, was transferred to perinuclear endosomes
within 10–20 min, and then was delivered to the Golgi
within 30–45 min of internalization.
A
B
C
Fig. 1. Trafficking of StxB in BSC-1 cells. Cy-3 StxB was bound to BSC-1 cells on ice and internalized at 37 C for the times shown. (A) StxB
passes through Tfn-positive endosomes. Alexa 488 Tfn and StxB bound to BSC-1 cells expressing human Tfn receptor on ice and then
warmed for times shown. Both co-localized up to 20 min. By 45 min Tfn (green) and StxB (red) had separated. Arrow indicates perinuclear
endosome. StxB remained in a Golgi-like ribbon for the remainder of the Tfn ⁄StxB time-course 60–180 min. (B) Internalized StxB (red) with
cells fixed and immunolabeled for Golgi marker GM130 (green). Note co-localization (yellow). (C) Internalized StxB (red) with cells fixed and
immunolabeled for ER marker PDI (green). Note co-localization at 240 min. Inset is indicated area magnified. Bars = 10 lM.
J. McKenzie et al. Shiga toxin in the Golgi
FEBS Journal 276 (2009) 1581–1595 Journal compilation ª2009 FEBS. No claim to original US government works 1583
StxB is delayed in the Golgi before entering the ER
We next characterized the passage of StxB through the
Golgi of BSC-1 cells under normal cell culture condi-
tions (Fig. 1B,C). The distribution of StxB at various
time points was compared with that of the cis ⁄medial
Golgi marker GM130, or the ER marker protein disul-
fide isomerase (PDI) [44,45]. Cells were labeled with
StxB as before and fixed for immunofluorescence. StxB
initially partly co-localized with the cis ⁄medial Golgi
marker GM130 after 20 min (Fig. 1B), and co-localiza-
tion increased up to 120 min (Fig. 1B). This confirmed
that StxB passes from the transferrin-positive endo-
somes to the Golgi rather than to another compart-
ment such as late endosomes [24]. Passage through the
Golgi was slow, as observed elsewhere [2]. To deter-
mine how long it took for StxB to enter the ER, we
internalized StxB for up to 4 h and labeled the cells
for the ER marker, PDI (Fig. 1C). StxB remained in a
perinuclear ribbon (Golgi, as shown by co-localization
in Fig 1B) up to 120 min. StxB began to co-localize
with PDI at 150 min (not shown) and 180 min (not
shown). By 240 min (Fig 1C), StxB was localized to
the ER as shown by co-localization with the ER resi-
dent, PDI. These data support the observation that
passage through the Golgi is the slowest step in the
retrograde pathway, requiring up to 120 min [46].
Taken together, Fig. 1A–C established a normal time-
course of StxB traffic in BSC-1 cells. We used this
time-course as a basis for our further experiments.
Passage through the Golgi is required for StxB to
reach the ER in cytoplasts
We wished to test directly if passage through the Gol-
gi ⁄TGN was required for StxB entry into the ER. To
accomplish this, we made use of subcellular surgery to
create cytoplasts lacking a Golgi apparatus [47].
Peripheral extensions of adherent BSC-1 cells were
cleaved using a glass micro-pipette to create cytoplasts
(peripheral areas lacking a nucleus) and karyoplasts,
containing the nucleus, the Golgi apparatus and the
REs [48]. Cytoplasts generated in this manner lack a
Golgi apparatus, and importantly, cannot regenerate
one [47]. By contrast, cytoplasts can regenerate func-
tional REs from peripheral EEs, as we have previously
demonstrated. Recycling of Tfn in cytoplasts is com-
plete and follows the same kinetics in cytoplasts as in
whole cells [6]. Cytoplasts and karyoplasts were labeled
with StxB and Tfn for 5 min at 37 C (rather than on
ice to avoid releasing the cytoplast from the coverslip)
and both ligands were chased into the cytoplasts for
various times (Fig. 2A). After 10 min, StxB co-local-
ized with Tfn in endosomal structures (Fig. 2A; 10 min
yellow arrow). After 30 min, Tfn and StxB continued to
co-localize with Tfn in endosomes (compare Fig. 2A;
30 min to Fig. 1A). Because Tfn recycles out of
cytoplasts at longer StxB internalization times
(120 min), it was necessary to add Tfn to the media
for 5 min and chase in unlabeled media for the final
25 min of the assay before fixation to illuminate the
endocytic pathway. Surprisingly, after 120 min,
although the majority StxB (red arrows) remained
inside the cytoplasts, it did not co-localize with endo-
somal structures (labeled with Tfn, green arrow).
Rather, it appeared in a diffuse cytosolic-like pattern
(red arrows, Fig. 2A; 120 min). To ensure that Golgi
was not inadvertently included in the cytoplasts, we
immunolabeled cytoplasts for GM130 and found it to
be absent from the cytoplast, but readily visible in the
karyoplast (Fig. 2B). To identify which compartment
the StxB had entered, we chased StxB into cytoplasts
for 120 min and labeled the plasma membrane (wheat-
germ agglutinin, WGA; Fig. 2C), ER (PDI; Fig. 2D),
and cytosol (Rho GDI; Fig. 2E). StxB (red arrows)
did not co-localize with WGA (green arrows; Fig. 2C)
and thus had not recycled to the plasma membrane.
Nor did it co-localize with the ER marker, even when
allowing 240 min for co-localization with PDI (green
arrows; Fig. 2D). However, StxB did co-localize with
cytosolic GDI (yellow arrow; Fig. 2E), suggesting that
StxB was in the cytosol. The GDI immunolabel
required methanol fixation, which causes a grainy cast
to cytosolic proteins. Cytosolic depletion using SLO or
saponin proved unfeasible as treated cytoplasts
detached from the coverslip. Together, these results
suggest that when the Golgi was absent, StxB did not
enter the ER. Furthermore, under these conditions the
toxin was able to escape the endosomes directly into
the cytosol. At no time was an ER morphology or
co-localization with PDI of StxB observed. While it is
possible that some remnant Golgi, below the threshold
of visualization, was present in the cytoplast, it was
clearly insufficient to mediate StxB traffic to the ER.
This phenomenon may occur to some extent during
normal transit of the endosomes, although the amount
of toxin available for escape may be minimal as the
toxin passes rapidly through the endosomes to the
Golgi. It may, however, correspond to the brefel-
din A-resistant toxicity reported elsewhere [8].
Aluminum fluoride traps StxB and Tfn in
perinuclear endosomes
We wished to confirm the requirement for Golgi pas-
sage and to quantify escape of StxB from endosomes
Shiga toxin in the Golgi J. McKenzie et al.
1584 FEBS Journal 276 (2009) 1581–1595 Journal compilation ª2009 FEBS. No claim to original US government works
into the cytosol. However, cytoplasts are extremely
small, and must be made individually, making frac-
tionation impossible. Therefore, we used a pharmaco-
logical approach. Aluminum fluoride (AlF
4
)
)isan
activator of small GTPases and is well-documented to
block recycling of Tfn from the RE [5,49]. Although
the precise target of AlF
4
)
at the RE is not known,
the effect of this drug on Tfn recycling is immediate
and remarkably specific to recycling out of the RE in
nonpolar cells and to basolateral recycling from the
RE in polarized cells [5]. Treatment for > 1 h results
in dispersal of the Golgi although both the TGN and
the ER remain functional [50,51]. Because StxB
co-localized extensively with Tfn in perinuclear REs,
we suspected that AlF
4
)
might also block retrograde
StxB from the endosomes to the TGN just as it did
for recycling traffic to the plasma membrane. Fortu-
itously, both StxB and Tfn were trapped together in
the REs following AlF
4
)
treatment (Fig. 3). This was
especially apparent after 60 min, when Tfn would
normally have recycled out of the cell, and StxB would
normally have moved to the Golgi. Both remained in
the endosomes of treated cells after 60 and even
120 min (Fig. 3; 60 min, 120 min, yellow arrows).
Although this result is based on the fortunate effects
of AlF
4
)
treatment on these two pathways, it does not
necessarily imply that the same drug target is involved
in both retrograde and recycling pathways. It does,
however, present a unique opportunity. As in the
cytoplast, StxB is prevented from reaching the Golgi,
and it is trapped inside of the endosomes. This allowed
us to quantify the consequences of trapping StxB in
endosomes.
StxB leaks into the cytosol when trapped at
endosomes
StxB trapped in the endosomes of AlF
4
)
-treated cells
took on a diffuse cytosolic appearance at later time
points following internalization (Fig. 3; 120 min, red
A B
E D C
Fig. 2. StxB cannot access the ER in BSC-1 cytoplasts. BSC-1 cells were manually cut with a glass needle to create karyoplasts (k) contain-
ing both the nucleus and Golgi and cytoplasts (c). All cytoplasts and karyoplasts were labeled with Cy-3 StxB (red) that was internalized for
times shown. (A) Shiga and Tfn (green) internalized together for 10 min then chased for 10 or 30 min. For 120 min, Tfn was internalized for
the final 25 min. (B) Cytoplast with StxB (red) immunolabeled for Golgi marker GM130 (green). (C) Cytoplast stained for plasma membrane
with wheat germ agglutinin (green). Note that the cytoplast has moved next to the karyoplast but the two remain separate. (D) Cytoplast
labeled for ER marker PDI (green), at various times of StxB (red) internalization. Note exclusion of StxB from ER. (E) Cytoplast labeled
for cytosolic marker GDI (green) note co-localization (yellow) with StxB (red). Insets are cytoplasts presented in single channels with larger
inset showing a magnified view of the combined channels. Red arrows indicate StxB, green arrows indicate other compartment markers as
indicated. Bars = 10 lM.
J. McKenzie et al. Shiga toxin in the Golgi
FEBS Journal 276 (2009) 1581–1595 Journal compilation ª2009 FEBS. No claim to original US government works 1585
