INTRODUCTION

Following entry by phagocytosis/endocytosis, intracellular
pathogens have evolved a variety of strategies for evading
proteolysis by the host (Finlay and Falkow, 1997). Some, such
as Listeria monocytogenes (Portnoy and Jones, 1994), Shigella
flexneri (High et al., 1992), the protozoa Theileria parva
(Dobbelaere and Heussler, 1999) and Trypanosoma cruzi (Hall,
1993) dissolve the phagosomal membrane and reproduce in
the cytoplasm. Others, such as Mycobacterium tuberculosis,
Chlamydia trachomatis (Finlay and Falkow, 1997) and the
protozoa Toxoplasma gondii (Sibley, 1993) inhibit fusion of
the phagosome/endosome by lysosomes and grow within an
expanding permeablized endosome. And others, such as
Coxiella burnetti (Heinzen et al., 1996) and the protozoan
Leishmania survive and grow within the inhospitable confines
of a phagolysosome. Perhaps the most bizarre and interesting
solution is for the pathogen to enter the cell in an endosome
and from there somehow move to the endoplasmic reticulum
(ER) and its interconnected perinuclear space. Here it is safe
from lysosomal degradation.

At least three organisms have followed this last strategy, the

virus SV40 and two bacteria, Brucella abortus (Pizarro-Cerda

et al., 1998) and L. pneumophila (Swanson and Isberg, 1995).
In the case of SV40, this virus is taken up in small uncoated
vesicles or caveoli that become continuous with a complex
tubular network of smooth membranes generated as extensions
of the ER (Pelkmans et al., 2001; Kartenbeck et al., 1989). In
the case of Legionella pneumophila, the causative agent of
Legionnaires’ disease, the bacterium enters the cell in a
phagosome that becomes surrounded by vesicles and
mitochondria (Horwitz, 1983). This vacuole provides an
intracellular sanctuary for L. pneumophila where these bacteria
are protected from lysosomal degradation. This remarkable
behavior of L. pneumophila, as first described by Horwitz
(Horwitz, 1983), severely puzzled us. By what mechanisms
does L. pneumophila induce vesicles to surround the
phagosome? Do they protect it from fusion with lysosomes,
and how does a former plasma membrane become studded with
ribosomes? After all, the lipid composition and thickness of the
plasma membrane and its unusual protein composition is very
different from the ER. Specifically, the plasma membrane is
rich in cholesterol, in amounts roughly equimolar to the sum
of all the phospholipids in the membrane, and sphingolipids
(such sphingomyelin and glycolipids), whereas the ER
membranes lack or have extremely low concentrations of both.

4637

Within five minutes of macrophage infection by Legionella
pneumophila, the bacterium responsible for Legionnaires’
disease, elements of the rough endoplasmic reticulum
(RER) and mitochondria attach to the surface of the
bacteria-enclosed phagosome. Connecting these abutting
membranes are tiny hairs, which are frequently periodic
like the rungs of a ladder. These connections are stable and
of high affinity - phagosomes from infected macrophages
remain connected to the ER and mitochondria (as they
were in situ) even after infected macrophages are
homogenized. Thin sections through the plasma and
phagosomal membranes show that the phagosomal
membrane is thicker (72

±

2 Å) than the ER and

mitochondrial membranes (60

±

2 Å), presumably owing to

the lack of cholesterol, sphingolipids and glycolipids in the

ER. Interestingly, within 15 minutes of infection, the
phagosomal membrane changes thickness to resemble that
of the attached ER vesicles. Only later (e.g. after six hours)
does the ER-phagosome association become less frequent.
Instead ribosomes stud the former phagosomal membrane
and L. pneumophila reside directly in the rough ER.
Examination of phagosomes of various L. pneumophila
mutants suggests that this membrane conversion is a four-
stage process used by L. pneumophila to establish itself in
the RER and to survive intracellularly. But what is
particularly interesting is that L. pneumophila is exploiting
a poorly characterized naturally occuring cellular process.

Key words: L. pneumophila, Macrophage, Endoplasmic reticulum,
Ribosomes, Membrane, Intracellular survival

SUMMARY

How the parasitic bacterium

Legionella pneumophila

modifies its phagosome and transforms it into rough
ER: implications for conversion of plasma membrane
to the ER membrane

Lewis G. Tilney

1

*, Omar S. Harb

1

, Patricia S. Connelly

1

, Camenzind G. Robinson

2

, and Craig R. Roy

2

1

Department of Biology, University of Pennsylvania, Philadelphia, PA 19104, USA

2

Yale University School of Medicine, Section of Microbial Pathogenesis, New Haven, CT 06511, USA

*Author for correspondence (e-mail: [e-mail])

Accepted 13 September 2001
Journal of Cell Science 114, 4637-4650 (2001) © The Company of Biologists Ltd

RESEARCH ARTICLE

4638

This results in the endoplasmic reticulum membrane being
thinner than the plasma membrane, a feature that may influence
differences in accumulation of transmembrane proteins
(Bretscher and Munro, 1993).

In this current study, we amplify and extend the

morphological description of Horwitz (Horwitz, 1983),
revealing several previously undescribed phenomena
associated with the establishment of the L. pneumophila
phagosome. More specifically, we illustrate morphologically
the existence of physical connections between the ER vesicles
and mitochondria, and the L. pneumophila phagosome.
Furthermore, we show that the thickness of the phagosomal
membrane containing L. pneumophila changes to resemble an
ER membrane. This is followed some hours later by the
attachment of ribosomes directly to the ‘newly thinned’
phagosomal membrane. We then describe the behavior of five
L. pneumophila mutants that amplify our morphological
description of the wild-type L. pneumophila by emphasizing
some of the changes that occur as L. pneumophila adapts to
life within the macrophage. Overall, the results presented in
this study support the idea that L. pneumophila subverts normal
cellular processes to protect itself from proteolysis. Thus,
studying the interaction of L. pneumophila with its host cell
gives us the opportunity to understand additional features of
basic cell biological phenomena that occur in eukaryotic cells.
Study of these basic processes, which include the attachment
of ER vesicles to the plasma membrane and changes in
membrane composition, are some of our goals for the future.
Ultimately we must determine why uninfected host
macrophages behave in this fashion and how pathogens
orchestrate this behavior. This report then, we hope, will
stimulate others to investigate what we assume is so far an
undescribed or poorly described pathway in eukaryotic cells in
which the plasma membrane is converted to the ER.

MATERIALS AND METHODS

Cell lines and bacterial strains

U937 cells (ATCC#: CRL-1593.2) are a human monocyte lymphoma
cell line (Sandstrom and Nilsson, 1976) that exhibit macrophage-like
characteristics when stimulated with phorbol myristic acid (PMA).
Two srains of L. pneumophila, (AA100 and CR39), both of which are
serogroup 1 strains, were utilized in this study.

U937 cell maintenance and differentiation

U937 cells were maintained in suspension in RPMI-1640 (Gibco)
supplemented with 10% heat-inactivated fetal bovine/calf serum
(FBS) and glutamine. Cells can be differentiated by the addition of
50 ng/ml PMA (5

µ

l of a 1 mg/ml solution of PMA per 100 ml of

cells). For the purpose of electron microscopy U937 cells were either
differentiated in six-well culture plates or in 12.5 cm

2

flasks. For six-

well culture plates 1

×

10

6

cells/well were seeded and incubated at

37°C with 5% CO

2

for 48 hours in the presence of PMA. For 12.5

cm

2

flasks, U937 cells were treated with PMA in a 175 cm

2

flask for

two days. Adherent macrophages were then removed and replated into
12.5 cm

2

flasks and incubated at 37

°

C with 5% CO

2

for 12-16 hours.

Before infection U937 cells were washed three times in warm culture
media.

L. pneumophila strains and growth conditions

L. pneumophila strain AA100 (graciously provided by Dr Yousef Abu
Kwaik at the University of Kentucky, Lexington, KY) was grown on

buffered charcoal yeast extract (BCYE) agar plates for 48 hours at
37

°

C. L. pneumophila has previously been shown to be most virulent

following logarithmic growth (Byrne and Swanson, 1998). In order to
achieve this growth phase, a loop full of the plate-grown bacteria was
inoculated into 5 ml of pre-warmed buffered yeast extract (BYE)
media in a 50 ml conical and grown at 37

°

C with shaking for ~18

hours as described previously (Harb and Abu Kwaik, 2000). Bacterial
growth was monitored spectophotometrically and bacteria were
harvested once the culture OD

550

reached 2.0-2.2 (Gao et al., 1999).

An OD

550

of 1 is equal ~1

×

10

9

/ml.

L. pneumophila strain CR39 and its isogenic mutants were grown

for 48 hours at 37°C on BCYE agar plates, then resuspended in PBS
to an OD

600

of 10.0. Bacteria from this suspension were then diluted

into 12.5 cm

2

flasks containing U937 cells to achieve the appropriate

multiplicity of infection (MOI).

Infection of U937 cells with

L. pneumophila

Bacteria were suspended to the appropriate MOI in U937 culture
media, in this case MOI of 20. This culture was added to the
differentiated macrophages in six-well plates and spun down at 150 g
for 10 minutes to synchronize the infection. Plates were transferred
to 37

°

C in air supplemented with 5% CO

2

for the appropriate time.

Extracellular bacteria were removed by washing the macrophages
three times with prewarmed culture media. For longer time points,
extracellular bacteria were killed using 50

µ

g/ml gentamicin for one

hour as described previously (Harb and Abu Kwaik, 1998).

Preparation of

L. pneumophila vacuoles

4.5

×

10

7

U937 cells in 50 ml of RPMI 1640 (10% FBS) were

differentiated with PMA for 48 hours. L. pneumophila were grown
to an OD

550

of 2-2.2 (~18 hours) in BYE media, and U937 cells

were infected at an MOI of 5. Infection was allowed to proceed for
two hours following which cells were removed from flasks using a
cell scraper in 6 ml of RPMI. 1 ml of the resuspended cells was not
homogenized. The rest of the cells were homogenized in a dounce
homogenizer (in/out up to five times; after each time 1 ml of cells
was removed into a microfuge tube and left on ice until
homogenization was complete). Microfuge tubes were spun down at
400 g for three minutes to remove large fragments and/or unbroken
cells (4

°

C). The supernatant was removed and spun down at 2000 g

for one minute (4

°

C). The pellet was resuspended in 0.5 ml fixative

and spun down at 12,000 g for three minutes at 4

°

C. Fresh fixative

was added to the pellets, which were prepared for electron
microscopy.

Electron microscope techniques

U937 cells were grown to confluence on either six-well culture plates
or 12.5 cm

2

plastic tissue culture flasks. L. pneumophila that had

reached the post-exponential growth phase were suspended at the
appropriate MOI and added to the culture plates for the allotted time
interval. At the appropriate time, for example, after 0.5, 1 or 1.5 hours,
the extracellular bacteria were washed off three times with warmed
culture media and the plates incubated with warmed culture media
until fixation. The media was removed and the U937 cells fixed in situ
with a freshly made solution of 1% glutaraldehyde (from an 8% stock
from Electron Microscopy Sciences (EMS), Fort Washington, PA) 1%
OsO

4

in 0.05 M phosphate buffer at pH 6.2 for 45 minutes. After

fixation, the cells in petri plates were rinsed three times with cold
distilled water and en bloc stained with uranyl acetate overnight.
The petri plates were then dehydrated in ethanol then placed into
hydroxypropyl methacrylate (EMS), which does not react with the
plastic in the petri dish, and embedded in L 112, an epon substitute
(Ladd, Burlington, VT). Following polymerization of the epon, the
block was cut out and mounted and thin sections were cut through
their exposed surfaces. Thin sections were collected on naked grids
stained with uranyl acetate and lead citrate and examined in a Philips
200 electron microscope.

JOURNAL OF CELL SCIENCE 114 (24)

4639

How

Legionella pneumophila modifies its phagosome

Measurements of membrane thickness and/or membrane
separation from electron micrographs

All our electron micrographs were photographed at 40,000

×

and

printed at a magnification of 100,000

×

. Individual prints were selected

under an illuminated dissecting microscope at a magnification of 10

×

.

The membrane thickness was measured by placing an ocular
micrometer disc on top of the segments of membranes that we wished
to measure. We selected only those portions of the phagosomal
membrane or the bound ER membranes where the membrane was
cut transversely. Thus, after osmication one sees at one million
magnification two clearly defined dense lines separating an
intermediate space. If the dense lines are not clear and their margin
not sharp but fuzzy, then the section is not cut perfectly normal to the
membrane. Obviously these regions were not measured. On clearly
defined transverse sections, we measured the width of the membrane
as defined by the outer edges of the two dense lines that in electron
micrographs define what we know is a membrane bilayer. To eliminate
ambiguities in measurements, only one person (L.G.T.) measured all
the membrane profiles. Each measurement included in Table 2 and
Table 3 is of a separate ER vesicle attached to a phagosome containing
a L. pneumophila bacterium. The number of separate phagosomes
measured is also documented on all the tables.

As shown in Table 2 and Table 3, the number of individual separate

phagosomes measured varied from 7–13, and the total number of ER
vesicles with its associated phagosomal regions measured ranges from
18-48. The range here is due to the selection of only what L.G.T.
considered to be perfect transverse section through the membrane. In
all the tables, the measurements include the mean and standard
deviations. Unpaired t-tests were
performed on all the data in the tables so
that we could compare the width of the
phagosomal membrane and the ER
membrane in the wild-type and mutants
at varying times after infection. t-tests
were also carried out by comparing the
amount of coverage of the phagosome
by ER vesicles at 15 minutes and 6 hours
(Table 1). The P

values for the

comparisons are included in the Results
section.

Measurements of the separation of

ER

vesicles from the phagosomal

membrane in regions where they are
bound together by the osmiophilic
hairs were made in the same way
as measurements of the membrane
thickness, for example prints at
100,000

×

magnification were examined

under a dissecting microscope at 10

×

.

The values of membrane thickness of

the phagosomal membrane and ER
membranes varies, as documented in all

the tables. For example, the mean values for the thickness of the ER
membrane in the icm mutants (Table 3) varies from 57 to 64 Å, and in
the 5 and 15 minute infections, they vary from 60 to 64 Å (Table 2).
These differences are not statistically significant. For simplicity and to
focus the reader’s attention on the relative thickness of the ER and
phagosomal membranes where they contact each other, we have placed
on (Figs 1-3 and 7-9) values for these thicknesses as either 60 Å or 70
Å rather than the measured values here, for example, 61 Å or 57 Å. In
showing our manuscript to other scientists they all felt this was
appropriate labeling because (1) the actual values were not significantly
different from the 60 or 70 Å values, and (2) it simplified the
presentation of data.

RESULTS

L. pneumophila-containing phagosomes are
encased rapidly by ER vesicles and mitochondria
Five minute infection

Using a multiplicity of infection (MOI) of 20 bacteria/
macrophage, a number of L. pneumophila were located within
infoldings of the surface membrane. We have no way of
knowing if these infoldings represent early stages in
endocytosis, as serial sections were not cut. In addition to these
infoldings, a few bacteria in this brief exposure were within
phagosomes. A subset of these phagosomes were clearly

Fig. 1. Thin section cut near the surface
of a U937 macrophage fixed after five
minutes of infection, with L.
pneumophila infected at an MOI of 20.
Within the phagosome is a L.
pneumophila bacterium. Attached to the
basal surface of the phagosomal
membrane are a series of vesicles of the
ER. The bracketed region in (a) is
shown at higher magnification in (b).
The endosomal membrane is 70 Å
thick, whereas the attached ER
membranes are only 60 Å thick.

4640

associated with organelles even at this early time period
following infection, although most of these phagosomes were
not associated with vesicles or mitochondria (Fig. 1a). On the
phagosome membrane opposite the cell surface, five vesicles
were observed that have apparent connections to the
phagosome as they approach within 70 Å of the phagosomal
membrane. At points of attachment, tiny projections were
observed connecting the abutting surfaces. The intimate
association between the phagosomal membrane and two of the
attached vesicles was more easily seen on higher magnification
(Fig. 1b). These membranes were cut transversely and show
the characteristic thin-section image of membranes, which
consist of two dense lines separated by a less dense space. The
thickness of these membranes will be important in what
follows. In this micrograph, and others like it, the plasma
membrane and the phagosomal membrane were approximately
70 Å thick, whereas the associated vesicle membrane was
thinner and in appropriate regions, where measurements were
possible, was 60 Å thick. Exact measurements are presented
on the tables and will be discussed in greater detail

subsequently.

15 and 30 minute infections

Following 15 minutes of continuous exposure to L.
pneumophila it was easy to find numerous examples
of L. pneumophila within phagosomes. Out of 37
phagosomes each containing a single L. pneumophila
bacterium, 15 did not have ER or mitochondria
associated with them, 21 were completely surrounded
by vesicles and/or mitochondria (Fig. 2a) and one was
partially surrounded (similar to that shown in Fig. 1).
Some of these vesicles had ribosomes associated with
them, and thus could be identified as rough ER. The
separation between the phagosomal membrane and the
membrane lining an attached organelle, albeit vesicles
of ER or mitochondria, was remarkably constant. We
measured the distance between 30 phagosomes
containing

L. pneumophila and attached ER vesicles.

The values measured were 67

±

11 Å. Thus in thin

section, these two membranes look like parallel
railroad tracks. Connecting these two membranes were
tiny osmiophilic hairs that were periodic (Fig. 2b
arrows). The phagosomal membrane closely follows
the contours of the enclosed rod-shaped L.
pneumophila bacterium, which in turn makes the
phagosome rod shaped. If the hairs were bars or sheets
they should appear differently, depending on the
orientation of the bacterium. However, regardless
of the plane of section through a phagosome
[longitudinal (Fig. 2a), transverse (Fig. 3a) or an
oblique section], these connections appeared as hairs,
never as bars or sheets.

As mentioned, in these early stages of infection we

encountered L. pneumophila-containing phagosomes
that have only a few ER vesicles connected to their
surfaces, such as those depicted in (Fig. 1). Note that
the region of attachment to the phagosome for most of
these vesicles is small compared to their total surface
area, which does not affect their spherical appearance
significantly. By 15 minutes, however, many of the
vesicles attached to these phagosomes were flattened,

almost pancake shaped, as if these ER vesicles had flattened out
as they zippered progressively to the phagosomal surface by the
tiny hair-like connections (Fig. 3a,b). This leaves little free space
between adjacent vesicles. We measured at 15 minutes, and at
later infection times, the amount of phagosomal surface that was
in direct contact with the cytoplasm or the phagosomal surface
not covered by attached organelles (Table 1). We ignored all

JOURNAL OF CELL SCIENCE 114 (24)

Fig. 2. (a) Longitudinal section through a L. pneumophila bacterium residing
in a phagosome after 15 minutes of infection at an MOI of 20. The phagosome
is surrounded by attached ER vesicles, many of which are the rough ER type
(arrowheads point to attached ribosomes). The surface of the ER vesicles is
flattened and smooth where it is attached to the phagosomal membrane but
irregular in contour on its unattached surface. Even at this low magnification it
is possible to detect tiny osmiophilic hairs at the attachment surface between
the phagosomal and ER membranes. (b) Higher magnification view of the area
outlined in (a). The hairs can be seen better in this micrograph (arrows). Note
that by 15 minutes, the thickness of the phagosomal and ER membranes is the
same, about 60 Å each.

Table 1. Percentage of the phagosomal surface covered by

attached ER and/or mitochondria

Number of

phagosomes measured

Range (%)

Average (%)

15 minutes

10

85-99

92.2±4.5

2 hours

10

88-98

94.6±2.9

6 hours

8

41-70

55.3±13

Largest gaps between attached vesicles and phagosome

15 min

0.14

µ

m

2 hr

0.05

µ

m

6.5 hr

0.57

µ

m

4641

How

Legionella pneumophila modifies its phagosome

those which were either partially enclosed, similar to the
situation in (Fig. 1), or completely lacking associated
mitochondria and vesicles. Out of 10 L. pneumophila-containing
phagosomes, 85 to 99% of the phagosomal surface was
associated with vesicles/mitochondria (Table 1). The largest gap
we found between vesicles on the phagosome was 0.14

µ

m. In

most cases the gap was only 0.03

µ

m.

Continuous exposure to L. pneumophila for an additional 15

minutes presents images similar to those we already described
in the 15 minute sample. Some of the phagosomes were
completely surrounded by organelles, others partly surrounded,
and a few had no associated organelles, as the latter had
probably been endocytosed just before fixation.

Two hour infection

Macrophages were exposed to L. pneumophila for 30 minutes,
extracellular bacteria removed, and the macrophages incubated
for an additional 1.5 hours before fixation. By this time, we
could find no L. pneumophila in phagosomes that lacked
associated ER vesicles and mitochondria. The only change
from the 15 or 30 minute samples was that the enveloping ER
vesicles, initially composed of the small ER vesicles, had fused
to form long flattened structures connected to the phagosomal
membrane by the tiny osmiophilic hairs (Fig. 3c,d).

We measured 10 L. pneumophila-containing phagosomes

and found that between 88 and 98% of the endosomal surface
was covered by organelles (Table 1). The largest gap between
attached vesicles of ER was 0.05

µ

m. As both longitudinal and

transverse sections through the phagosome gave the same
results, this means that the endosomal surface area not covered
by organelles would be at worst a circle not larger than 500 Å
in diameter. Thus, the associated vesicles may provide a
barricade against fusion with tiny acidifying vesicles and/or
lysosomal vesicles larger than 0.05

µ

m.

Six hour infection

Two changes from the 15 minute, 30 minute and 120 minute

samples were apparent at this stage. First, by this time some of
the L. pneumophila within an enclosed phagosome were
elongating and undergoing binary fission (data not shown).
Second, there were now regions on the former phagosomal
membrane that lacked both small spherical ER vesicles and
large flattened ER vesicles, and in their place ribosomes were
attached (Fig. 4a). In other words, six hours after infecton, a
portion of the former phagosomal membrane began to
resemble bona fide RER, albeit with regions of attached
vesicles (both rough and smooth ER).

The percentage of the former phagosomal membrane

containing L. pneumophila that had associated ER vesicles/
mitochondria was determined once again. The values from
eight independent examples ranged from 41 to 70% (Table 1).
In the cytoplasmically exposed regions, ribosomes were now
attached. The largest gap between attached vesicles was 0.57

µ

m. Data in Table 1 indicate that there was significantly less

coverage of the phagosome by ER vesicles after six hours
compared to 15 minutes or 2 hours after infection (P<0.0001).

20 hour infection

By 20 hours, numerous L. pneumophila were now located
within a single vacuole, which indicates that replication had
occurred within these compartments (Fig. 4b). One of the most
significant changes observed at 20 hours was that the vacuolar
membrane no longer had attached RER vesicles but was now
completely covered with ribosomes, a clear indication that it
had become RER.

Host organelles remain attached to phagosomes
containing

L. pneumophila following cell

fractionation

Following a two hour infection with L. pneumophila, crude
phagosomal fractions were prepared as described (see
materials and methods). In all cases, the isolated phagosomes
containing L. pneumophila were connected to vesicles of the
ER, both rough and smooth, and also to mitochondria (Fig. 5a).

Fig. 3. (a) Transverse section through
a phagosome containing L.
pneumophila after a 15 minute
infection period. As in (Fig. 2),
vesicles of ER, both studded with
ribosomes and without, are attached to
the phagosome. The region indicated
is shown at higher magnification in
(b). Both the phagosomal membrane
and the ER membranes have the same
thickness. (c) Longitudinal section
through a phagosome containing L.
pneumophila after a two hour infection
period at an MOI of 1 applied for 30
minutes. The phagosome is
surrounded by attached ER vesicles.
The area boxed in (c) is shown at a
higher magnification in (d).

4642

In no case did we find L. pneumophila-containing phagosomes
that lacked an abutting population of ER vesicles, nor did we

find free L. pneumophila. Of particular importance, was the
fact that although the ER vesicles appeared swollen, the surface
abutting the endosomal membrane was separated by
approximately 70 Å. In favorable regions the tiny hairs that
connect the vesicles to the phagosomal membrane could
be seen (data not shown). There were instances where
phagosomes containing L. pneumophila had the nucleus of the
host cell attached. As shown in Fig. 5b, the phagosomal
membrane was connected to the outer nuclear envelope, and,
as expected, the separation of these membranes was
approximately 70 Å. In retrospect this connection could have
been predicted since the outer nuclear envelope is contiguous
with the ER.

The thickness and thus the lipid composition of the
phagosomal membrane surrounding

L. pneumophila

is altered upon association of host vesicles

The lipid composition of cellular membranes dramatically
affects how they appear in electronmicrographs. For instance,
the fixation and staining procedures we have followed allow
us to differentiate plasma membrane from membranes
surrounding cellular organelles. In our micrographs, plasma
membrane and the newly formed phagosomal membrane
appear about 70 Å in diameter, whereas, the ER and/or
mitochondrial membranes appear thinner (about 60 Å).
Interestingly, we found that at early stages of infection, when
ER vesicles and mitochondria first attach to phagosomes
containing L. pneumophila, disparity in thickness was apparent
in these abutting membranes (Fig. 1). However, 15 minutes
after infection, when ER vesicles surround the phagosome, the
former phagosomal membrane displayed the same thickness as
the ER membrane (Fig. 2b; Fig. 3b). Thus, within 15 minutes
there appears to be a dramatic decrease in thickness of the
phagosomal membrane when the phagosome is surrounded by
organelles. This conclusion is strengthened by the data
presented in Table 2, which shows that the phagosomal
membrane was significantly thicker than the ER membrane
after five minutes of infection (P<0.0001). However, after 15
minutes, the phagosomal membranes had the same thickness.
Since the thickness of a membrane is influenced by the
presence of cholesterol and sphingolipids (Bretscher and
Monro, 1993), by measuring the thickness of phagosomal and

JOURNAL OF CELL SCIENCE 114 (24)

Fig. 4. (a) Transverse section through a L. pneumophila bacterium
enclosed in a vacuole. The U937 cells were exposed to L.
pneumophila for 30 minutes, then washed free of unattached bacteria
and incubated for an additional 5.5 hours before fixation. The surface
of the phagosome has reduced numbers of attached ER vesicles by
this time. In their place are ribosomes that are directly attached to the
phagosome. (b) Thin sections through a vacuole in a U937 cell
containing two L. pneumophila. This U937 cell had been exposed to
L. pneumophila for 30 minutes and after removal of unattached
bacteria incubated for 19.5 hours. By this time, the L. pneumophila
had replicated in the rough ER – identified as such by ribosomes
attached to the membrane of the vacuole.

Fig. 5. U937 cells were incubated with
L. pneumophila for two hours, then
washed, scrapped off the petri plates
and homogenized. A low-speed pellet
of unbroken cells or large cell
fragments was removed by
centrifugation. The supernatant was
then centrifuged on a tabletop
centrifuge and the pellet fixed. It
contains L. pneumophila enclosed in its
phagosome, and still attached to the
surface of the phagosome are
mitochondria and ER vesicles. (a) Since
the ER is continuous with the outer
nuclear membrane (NM), L.
pneumophila-containing phagosomes
can also be found attached to the
nucleus (N) as well as to the outer ER
vesicles (b).

4643

How

Legionella pneumophila modifies its phagosome

ER membranes we conclude that L. pneumophila-containing
phagosomes are altered in their lipid composition. This
alteration occurs rapidly between five and 15 minutes after
infection.

L. pneumophila exploit a natural and as of yet
uncharacterized host cellular process

Through careful examination of the plasma membrane of
infected U937 cells, we found instances where L. pneumophila
were attached to host cells, probably as a prelude to
phagocytosis. Interestingly, in these instances the cytoplasmic

plasma membrane directly beneath the associated extracellular
L. pneumophila had attached ER vesicles (data not shown). We
concluded that these vesicles were attached to the plasma
membrane because they were separated from each other by a
constant distance of approximately 70 Å, and between the two
abutting membranes were tiny hair-like connections.

We also found ER vesicles attached to the plasma membrane

(Fig. 6a) where extracellular L. pneumophila were not
associated with the cell surface. One possible interpretation is
that the L. pneumophila were attached to the surface at these
locations but failed to remain connected during fixation,
dehydration and embedding. Another possibility is that even in
uninfected U937 cells, ER vesicles may be attached to the
plasma membrane for purposes that have not been uncovered.
To rule out the latter, we examined the plasma membrane of
uninfected U937 cells. Much to our surprise we found ER
vesicles (Fig. 6b,c), some complete with attached ribosomes
(Fig. 6c), connected to the plasma membrane of these
uninfected macrophages. These ER vesicles must be attached
to the plasma membrane because 1) unattached vesicles were
separated from the plasma membrane by a cortical layer of
actin, 2) there was consistent spacing of approximately 70 Å
between the two membrane bilayers, and 3) osmiophilic hairs
were detectable at attachment sites (arrows).

Simultaneous infection of U937 macrophages with
two bacteria,

L. pneumophila, a Gram-negative

bacterium and a second Gram-positive bacterium

Although simultaneous infection of macrophages with two
bacteria was done in error and was due to contamination with
a second bacterium, the results proved interesting. The question
we inadvertently addressed was, is the connection of the ER
vesicles and mitochondria around endosomes restricted to
vacuoles containing L. pneumophila or is this a global
phenomenon occurring for other endosomes in L. pneumophila-
infected cells? In other words, do individual endocytosed L.
pneumophila induce the ER association for each endosome or
does L. pneumophila induce, perhaps by transforming the host
cell, a generalized response to all endosomes?

Because the two infecting bacteria are morphologically

distinct, L. pneumophila being Gram-negative and the other
bacteria being Gram-positive, it was easy to identify the fate
of each in thin sections. As was the case for L. pneumophila,
the Gram-positive bacterium entered the cell surrounded by
an endosomal membrane (Fig. 7a). However, unlike L.
pneumophila-containing endosomes, ER vesicles were never
seen connected to the endosomal membrane containing the
Gram-positive bacterium. The thickness of endosomal
membranes was measured for vacuoles containing these Gram-
positive bacteria. In all cases, the thickness of the membrane
surrounding these Gram-positive bacteria matched that of
lysosomes and the plasma membrane, never that of the ER
vesicles situated nearby in the same micrograph (Fig. 7a,b). In
the same macrophage, Gram-negative L. pneumophila were
found in phagosomes whose membrane thickness was the same
as the connected ER vesicles (data not shown).

Recruitment of ER vesicles to phagosomes
containing

L. pneumophila requires the Dot/Icm

transporter

Genetic analysis of intracellular growth mutants has revealed

Fig. 6. Thin sections through a portion of the plasma membrane of
three U937 macrophages. All are printed at the same magnification.
All three sections show ER vesicles attached to the plasma
membrane. In the areas indicated by the arrows we can see the hairs
that attach these membranes together. (a) This macrophage was
exposed to L. pneumophila for 15 minutes at an MOI of 20 before
fixation. Extracellular L. pneumophila were present near this section
as well as internalized L. pneumophila. (b) and (c) Uninfected U937
cells. As in (a). ER vesicles are attached to the plasma membrane.

Table 2

Number of

regions of

Number of

attached ER/

different

phagosome phagosomes

Thickness

measured

measured

5 min infection

Phagosomal membrane

72±7 Å

19

12

ER membrane

60±5 Å

11

12

15 min infection

Phagosomal membrane

64±3 Å

21

13

ER membrane

64±3 Å

21

13

4644

that L. pneumophila requires a specialized transport apparatus

for evading lysosome fusion (Andrews et al., 1998; Berger et
al., 1994; Matthews and Roy, 2000; Roy et al., 1998; Sadosky
et al., 1993; Segal and Shuman, 1997; Wiater et al., 1998). This
transport apparatus is encoded by 24 dot and icm genes located
on the L. pneumophila chromosome (Segal et al., 1998; Vogel
et al., 1998) and is similar to type IV transporters found in a
number of other bacteria (Christie and Vogel, 2000). To
determine whether the Dot/Icm transport apparatus is required
for the attachment of ER vesicles to phagosomes containing L.
pneumophila, we studied the dotA mutant, which fails to
express virulence functions requiring the Dot/Icm transporter
(Coers et al., 2000).

L. pneumophila dotA mutants do not grow and reproduce

inside U937 cells, but instead reside in endosomes that
fuse with lysosomes (Berger et al., 1994). We found that
phagosomes containing L. pneumophila dotA mutants did not
have ER vesicles or mitochondria attached to their surface (Fig.
8a). Furthermore, the phagosomal membrane surrounding dotA
bacteria remained the same thickness as the plasma membrane
and the lysosomal membrane (Fig. 8b). These data indicate that
the attachment of ER vesicles to phagosomes containing L.
pneumophila, as well as the lipid exchange that results in
thinning of the phagosomal membrane, requires a functional
Dot/Icm transport apparatus.

Distinct processes control evasion of lysosome
fusion, recruitment of ER vesicles, lipid exchange
and direct attachment of ribosomes to phagosomes
containing

L. pneumophila

To determine which bacterial products are required for the
attachment of ER vesicles to phagosomes containing L.
pneumophila, we examined icm mutants that have phenotypes
that are different from those mutants lacking a functional
transporter, such as the dotA mutants. The icmR, icmS and
icmW genes are essential for growth of L. pneumophila in
primary macrophages; however, unlike the dotA mutants, loss-
of-function mutations in these genes result in bacteria that have
a limited capacity to survive and grow in U937 cells (Coers
et al., 2000). Accordingly, these mutants may, by their
phenotypes, be able to inform us about what the missing
proteins in each mutant accomplishes for intracellular growth.

icmS mutant

Macrophages were incubated for 45 minutes with L.
pneumophila

mutant CR393 (

∆

icmS). Extracellular L.

pneumophila were washed away and endosomes containing
mutant bacteria were allowed to develop for an additional
45 minutes. Like phagosomes containing wild-type L.
pneumophila, those harboring

∆

icmS mutants had rough and

smooth ER vesicles and mitochondria attached by tiny hairs.
Furthermore, by 90 minutes, the phagosomal membrane
thickness resembled that of the attached ER vesicles (Table 3).
Statistical analysis indicates that there was no significant

JOURNAL OF CELL SCIENCE 114 (24)

Fig. 7. Co-infection of U937 cells with two bacteria; L. pneumophila,
a Gram-negative bacterium and a second Gram-positive bacterium
are depicted here in this thin section of a phagosome. No ER vesicles
attach themselves to the phagosome of the Gram-positive species.
The thickness of the phagosomal membrane remains the same as the
plasma membrane, 70 Å. This bacterium will be killed subsequently
in a lysosome.

Fig. 8. Thin section through a U937 cell with a phagosome
containing a dotA mutant. This U937 cell was infected for 45
minutes prior to fixation. The phagosomal membrane does not have
attached ER vesicles. Furthermore, the thickness of the phagosomal
membrane remains as thick as the plasma membrane (70 Å thick
(b)). As in Fig. 7, phagosomes containing dotA mutants fuse with
lysosomes.

4645

How

Legionella pneumophila modifies its phagosome

difference in the thickness of phagosomal membranes
surrounding icmS mutants and the ER (P=0.486). The only
differences we found between this mutant and wild-type L.
pneumophila was that the ER vesicles did not flatten along
the surface of the phagosome but instead appeared as
nearly spherical vesicles attached to the former phagosomal
membrane (data not shown). Thus, the phagosomal surface was
not covered as extensively by ER vesicles and/or mitochondria.
The range was 62 to 88% with a mean of 75% versus 88%
in the wild-type (Table 4, P=0.0104). In a few cases, there
were no ER vesicles/mitochondria attached to phagosomes
containing icmS mutants. We also found some L. pneumophila
in various stages of breakdown in lysosomes.

In a further group of experiments, U937 cells were infected

for 45 minutes then unattached L. pneumophila were washed
away and the macrophages cultured for a further 7.25 hours for
a total of 8 hours of infection. In these samples we found that
many of the phagosomes contained multiple bacteria indicating
that replication of L. pneumophila had occurred within these
vacuoles. The thickness of the former endosomal membrane
enclosing the now duplicated L. pneumophila was the same
thickness as ER membranes, which was what we expected
from the situation at 90 minutes after infection. Most of the
endosomes still had ER vesicles attached to them by the tiny
hairs and a few had ribosomes attached to them directly.

icmW mutant

As with the wild-type, the phagosomes enclosing L.
pneumophila mutant CR157 (

∆

icmW) had ER vesicles and

mitochondria attached to their cytoplasmic surfaces by tiny
hairs (Fig. 9a,b). The separation of abutting membranes was
approximately 70 Å (Fig. 9b). We measured the percentage of
the phagosome surface covered by ER/mitochondria in 12
different icmW-containing phagosomes in macrophages that
had been exposed to L. pneumophila for 90 minutes (45 minute
infection + 45 minutes of growth). The percentage varied from
a minimum of 41% to a maximum of 87% with a mean of 74%
(Table 4). Thus, these phagosomes were more naked than
phagosomes containing wild-type bacteria (P=0.0203), a fact
that may explain why icmW mutants end up in lysosomes.
Of particular interest to us was the observation that the
phagosomal membrane surrounding the icmW mutants was the
same thickness as the plasma membrane, even though the
phagosomal membrane was partially surrounded with attached
ER and mitochondria (Fig. 9b). We measured 47 phagosomes
(Table 3), and in all cases, the former phagosomal membrane
was approximately 70 Å thick (68.4±5.2 Å), in contrast to the
membrane thickness of attached ER or mitochondria (56.8±5.8
Å). Statistical analysis indicates that this difference is highly
significant (P<0.0001). In short, the membrane surrounding
icmW mutants fails to change thickness rapidly, suggesting that
the icmW gene product is an important determinant for lipid
exchange (Table 3).

At eight hour time points, we were able to find icmW mutants

replicating in U937 cells. Like the wild-type L. pneumophila,
replicating icmW mutants were identified in phagosomes that
contained multiple bacteria (Fig. 9c). Generally, much of the
phagosome surface remained at least partially covered with
bound ER vesicles and mitochondria (Fig. 9c). However, by this
stage there were a few areas in which ribosomes were attached
directly to the former phagosomal membrane. We measured the
thickness of 18 endosomes containing icmW mutants (Table 3).
By eight hours, the membrane thickness was similar to attached
ER vesicles and mitochondria membranes (Fig. 9d; Table 4).
Thus, although the phagosome membrane thickness was
approximately 70 Å at 90 minutes, by eight hours the thickness
of the membrane surrounding these icmW mutants was no longer
statistically different from that of the ER (P=0.2865). In short,
these L. pneumophila were now residing in bona fide RER.

icmW icmS double mutant

Previous studies did not reveal any phenotypic differences when
an icmW mutant was compared to either an icmS mutant or an
icmS icmW double mutant (Coers et al., 2000). In this study,
however, we have observed that icmW mutants reside in
phagosomes that do not change membrane thickness as rapidly
as phagosomes containing icmS mutants. These data suggest
that the icmW mutation should be dominant over the icmS
mutation. Accordingly, one would predict that mutant L.
pneumophila lacking both the IcmW and IcmS proteins would
have a phenotype that resembles the icmW single mutant. To
test this, U937 cells were infected with L. pneumophila CR503
(

∆

icmW

∆

icmS) for 45 minutes and were then incubated for

an additional 45 minutes after extracellular bacteria were
removed. Phagosomes containing the combined mutant had ER
vesicles and mitochondria attached. Of the 18 phagosomes
scored, coverage of the phagosome by mitochondria and ER

Table 3

Number of

regions of

Number of

attached ER/

different

phagosome phagosomes

Thickness

measured

measured

icmW
90 min infection

Phagosomal membrane

68.4±5.2 Å

47

13

ER membrane

56.8±5.8 Å

48

13

icmS
90 min infection

Phagosomal membrane

64.7±3.1 Å

21

9

ER membrane

64±3.1 Å

21

9

icm S icmW
90 min infection

Phagosomal membrane

69±3.9 Å

18

7

ER membrane

60±4.9 Å

18

7

icmR
90 min infection

Phagosomal membrane

73.6±6.6 Å

38

13

ER membrane

None connected

icmW
8 hr infection

Phagosomal membrane

61.7±4.4 Å

18

12

ER membrane

60±5 Å

18

12

Table 4. Percentage of phagosomal surface covered by ER

Number of

phagosomes

measured

Range

Mean

Wild type

9

60-94

87.6±10.6

icmS

10

62-88

75.4±7.8

icmW

12

41-87

74 ±13.2

icmS icmW

13

33-97

78.2±15.5

Analysis of L. pneumophila icm mutants at 90 minutes.

4646

vesicles varied from 33% of the surface to 97% of the surface.

On average, 78% of the phagosomal membrane surface was
covered by ER and/or mitochondria (Table 4). These differences
in phagosomal surface coverage are not statistically significant
(P=0.1306). When we compared the membrane thickness of the
phagosome (Table 3) to the attached ER vesicles, as predicted,
the thickness of the phagosomal membrane was significantly
thicker than the ER membrane (Table 3, P<0.0001). Like the
icmW mutant, by eight hours the thickness of the phagosome
membrane surrounding icmW icmS double mutants was equal
to that of the attached ER vesicles (data not shown). Thus, as
predicted, the

∆

icmW allele of this mutant delays the thinning

of the endosomal membrane that surrounds these bacteria to the
thickness of the ER membrane (Table 3).

icmR mutant

In thin sections, the association of ER vesicles and
mitochondria was not observed for most phagosomes enclosing
the mutant L. pneumophila strain CR343 (

∆

icmR) at time points

taken either 90 minutes or eight hours post infection. We found
two

∆

icmR-containing endosomes that had associated ER

vesicles following eight hours of infection. The thickness of the
phagosomal membrane surrounding the icmR mutant remains
unchanged throughout the infection (Table 3). Interestingly, at
both 90 minutes and eight hours infection times, we found that
the membrane limiting the icmR phagosomes often exhibited
signs of fracture, although the L. pneumophila remain at least
partially enclosed by endosomal membrane. We seldom found
broken phagosomal membranes with any of the other L.
pneumophila mutants or with the wild-type infections.

DISCUSSION

The following new observations have been made by studying

the morphology of phagosomes containing wild-type L.
pneumophila: (1) within five minutes of infection, the
phagosome containing L. pneumophila becomes surrounded by
elements of the rough and smooth endoplasmic reticulum and
mitochondria. The abutting membranes of these organelles
were connected to the phagosome by fine connections or hairs
with dimension of 70

×

50 Å. These connections are stable as

organelles such as the ER, mitochondria and even the cell
nucleus remain connected to the L. pneumophila-containing
phagosomes following phagosome isolation. This constitutes
the first description of a physical connection between the L.
pneumophila phagosome and its associated vesicles. (2) In
cells that had internalized both L. pneumophila and a second
bacterium, ER vesicles and mitochondria are found exclusively
attached to the L. pneumophila-containing phagosomes. These
data are consistent with previous results demonstrating that L.
pneumophila infection does not affect the processing of
heterologous endosomes (Coers et al., 1999). (3) At about 15
minutes after infection, the thickness of the endosomal
membrane, a derivative of the plasma membrane, is reduced
from approximately 70 Å, the thickness of the plasma
membrane, to the thickness of ER or mitochondrial
membranes. (4) 6.5 hours following infection ribosomes are
found attached directly to the former endosomal membrane of
L. pneumophila. As the number of ribosomes increases, the
association of the ER and mitochondria with the former
endosomal membrane decreases and eventually disappears, so
by this stage L. pneumophila is growing inside bona fide RER.

Does the attachment of the ER and mitochondria to
the

L. pneumophila-containing phagosome act as an

effective barricade against lysosomal fusion?

An obvious function for the attached organelles is to prevent
the phagosome containing L. pneumophila from fusion with
primary lysosomes and/or acidifying vesicles. As the attached

JOURNAL OF CELL SCIENCE 114 (24)

Fig. 9. U937 cells were infected with L.
pneumophila icmW mutants for 45
minutes, then unattached bacteria were
washed away and incubation was
continued for an additional 45 min
before fixation. As is the case with wild-
type L. pneumophila, ER vesicles
become attached by tiny hairs (arrows
in b) to the phagosomal membrane
surrounding icmW mutants. Unlike the
membrane surrounding wild-type L.
pneumophila, which decreases from 70
Å to 60 Å within 15 minutes, there is no
observable decrease in the thickness of
the membrane surrounding icmW
mutants at 1.5 hours (b). (c) and (d)
U937 cells were infected with icmW
mutants for 45 minutes, then washed
and incubated for an additional 7.25
hours. Replicating bacteria could be
found at this time and the phagosome
membrane surrounding them had
attached ER vesicles and mitochondria.
However, unlike the situation after 1.5 hours (a), the thickness of the phagosomal membrane surrounding these replicating icmW mutants had
decreased from 70 to 60 Å. (d) The membrane surrounding replicating icmW mutants is the same thickness as the ER membrane.

4647

How

Legionella pneumophila modifies its phagosome

organelles almost completely cover the endosomal surface
(92% of the surface after 15 minutes and 95% of the surface
after two hours), and since lysosomal enzymes, such as LAMP-
1, are not found in early L. pneumophila-containing vacuoles
(Clemens and Horwitz, 1995; Roy et al., 1998; Swanson and
Isberg, 1995), this is indeed an attractive hypothesis.
Consistent with this hypothesis is the fact that dotA mutants of
L. pneumophila, and bacteria other than L. pneumophila, both
of which do not enclose themselves in an ER barricade, are
contained in phagosomes that fuse with lysosomes (Berger et
al., 1994). However, this hypothesis cannot completely explain
why L. pneumophila-containing phagosomes do not fuse with
lysosomes because (1) a related Legionella species, Legionella
micdadei, replicates in endosomes with no attached ER
vesicles and/or ribosomes (Gao et al., 1999; Gerhardt et al.,
2000). Furthermore, when L. pneumophila is phagocytosed by
a protozoan, its natural host (Harb et al., 2000) ER vesicles and
mitochondria take two hours to bind to the phagosome
containing L. pneumophila, yet during this period the L.
pneumophila are not killed by lysosomes (L.G.T and O.S.H,
unpublished). (2) The icmR mutant of L. pneumophila does not
attach ER vesicles and mitochondria to its endosomes, yet this
mutant is still able to grow in U937 cells and has been shown
to evade fusion with lysosomes more effectively than dotA
mutants (Coers et al., 2000). (3) Even though icmS mutants and
icmW mutants recruit ER vesicles, these mutants are defective
in evading fusion with lysosomes (Coers et al., 2000). (4)
Phagosomes containing dotA mutants begin to accumulate the
late endosomal protein LAMP-1 within five minutes of uptake,
whereas phagosomes containing wild-type L. pneumophila
evade this rapid endocytic maturation event (Roy et al., 1998;
Wiater et al., 1998). As maximal coverage of ER vesicles
on the surface of phagosomes containing wild-type L.
pneumophila is not observed until 15-30 minutes after uptake,
it seems unlikely that these vesicles account for the complete
evasion of LAMP-1 acquisition.

Defining a four-stage process used by

L.

pneumophila to establish itself in the RER

On the basis of the data presented here and that of others, we
propose that L. pneumophila orchestrates a four-stage process
that transforms a plasma-membrane-derived phagosome into
an organelle that is very similar, if not identical, to RER. In the
first stage, L. pneumophila inhibits the rapid fusion of host
endosomes with the phagosome in which it resides. This
process requires the Dot/Icm type IV secretion apparatus, as
well as the icmW and icmS gene products, as it has been shown
previously that mutations in these genes severely affect the
ability of L. pneumophila to avoid fusion with lysosomes
(Coers et al., 2000).

In the second stage, L. pneumophila directs the attachment

of ER vesicles and mitochondria to the surface of the
surrounding phagosome. Like the first stage, this second stage
also requires the Dot/Icm secretion apparatus, indicating that
this process is mediated by proteins injected into the
phagosome by L. pneumophila. Our mutant analysis, however,
indicates that the injected factors required at this stage are
probably distinct from those that mediate the first stage. To be
more specific, L. pneumophila icmW and icmS mutants are
defective in first stage events, but they still promote the
attachment of ER vesicles and mitochondria to the phagosome

surface, albeit at an efficiency that is slightly reduced when
compared to wild-type bacteria (Table 4). In contrast,
phagosomes containing icmR mutants can evade immediate
fusion with lysosomes and do not attach ER vesicles or
mitochondria to the phagosome rapidly. IcmR is a chaperone
protein (Coers et al., 2000), which suggests that the L.
pneumophila factor(s) directly responsible for promoting the
attachment of ER vesicles to the phagosome may still be either
low or of reduced quality. This explains why icmR mutations
result in a severe second stage defect. This hypothesis predicts
that, given enough time, some phagosomes containing icmR
mutants should eventually convert to compartments resembling
RER, as we observed, after eight hours.

During the third stage, the thickness of the membrane

surrounding L. pneumophila changes. This stage is probably
concomitant with the second because it takes place between 15
and 30 minutes after phagocytosis. We have not identified a L.
pneumophila mutant that is defective for this third stage
of phagosome conversion. Although our data indicate that
icmW mutations have a significant effect on the third-stage
conversion events, this effect may be non-specific. For
instance, this mutation could have slight pleiotropic effects on
Dot/Icm transporter function that would perturb the normal
flow of all factors secreted by this apparatus. In the fourth and
final stage, ribosomes are found attached directly to the former
phagosomal membrane enclosing L. pneumophila. This stage
begins many hours after the membrane has changed thickness.
In order to attach ribosomes to the former phagosomal
membrane, translocans must be present. These in turn bind to
the signal recognition particle on the ribosomes. Although we
do not yet know precisely how these transmembrane channels
appear in the former phagosomal membrane, it is possible that
this occurs by the fusion of the attached RER vesicles with the
phagosomal membrane at these later time periods. Consistent
with this possibility is that the number of attached RER
vesicles decreases as the number of ribosomes directly attached
to the former phagosomal membrane increases. Furthermore,
by this time the volume of the phagosome, which was constant
from 15 minutes until two hours, increases.

It has been reported recently that the viral pathogen SV40

directs its transport from the cell surface to ER by a two-step
transport pathway (Pelkmans et al., 2001). Viral particles first
enter cells through caveolae and then traffic to a novel sorting
compartment called a caveosome. After several hours, viral
particles exit the caveosome in tubular compartments that
transit along microtubules and deposit SV40 into smooth ER.
It is not known whether caveolae participate in uptake of
L. pneumophila. There are, however, striking morphological
differences in the early compartments in which these two
pathogens reside. Most notably, ER is found to associate
rapidly with phagosomes containing L. pneumophila, whereas
ER markers are not observed on the early organelle in which
SV40 resides. Thus, the process L. pneumophila use for
trafficking to the ER appears to be distinct from that used by
SV40.

The attachment of ER vesicles to

L. pneumophila

endosomes cannot be related to autophagy

Swanson and Isberg hypothesized that the association of the L.
pneumophila with the macrophage and endoplasmic reticulum
might be due to the L. pneumophila exploiting the autophagic

4648

machinery of the macrophages (Swanson and Isberg, 1995).
Although this is an intriguing hypothesis it is incompatible
with our results for the following reasons: (1) within five
minutes of the addition of L. pneumophila to the macrophage
cell line or as soon as endocytosis occurs, elements of the ER
and mitochondria are connected to the L. pneumophila
endosome, presumably by the tiny hairs described here. This
is too rapid for autophagy, which usually takes one hour to
induce (Kim and Klionsky, 2000). (2) Not only are ER vesicles
bound to the endosome but also mitochondria. Mitochondria
are often seen within autophagic vacuoles in cells undergoing
autophagy but they are not connected to the outer surface of
the phagosomal membrane. (3) Besides mitochondria, the L.
pneumophila containing phagosome is also bound to the outer
membrane of the nucleus – a situation that does not occur
during autophagy for obvious reasons, for example, autophagy
of the nucleus would be lethal. (4) The abutting membranes of
the ER and endosomes are connected together by a constant
distance and by tiny hairs in L. pneumophila infections but
neither hairs nor a constant spacing of the membranes occurs
during autophagy (Kim and Klionsky, 2000). (5) There is a
reduction in thickness of the plasma membrane of the
endosome in the first few minutes of phagocytosis. This change
in membrane thickness of engulfed membranous material is not
seen in autophagy. (6) Ribosomes are bound to the six-hour
and older vacuoles containing replicating L. pneumophila,
whereas autophagosomes never have attached ribosomes (Kim
and Klionsky, 2000).

What might be the significance of the change in
thickness of the endosomal membrane when it
associates the ER and mitochondria and how might
this be accomplished?

From the pathogen’s point of view, what it wants to accomplish
is to reside in a membrane-limited compartment that nutrients
enter but that neither lysosomes nor acidifying vesicles fuse
with. One possibility to resist acidification and lysosome fusion
may involve the conversion of the bacterial phagosomal
membrane to an ER-like one. This would certainly involve a
change in the lipid composition of the phagosomal membrane
to match that of the ER. What is different about the lipid
composition of these two membranes? It is now well
established that the plasma membrane, as well as the
endosomal, the lysosomal and trans-Golgi membranes, contain
cholesterol (in roughly equimolar amount to the sum of all the
other lipids) (Bretscher and Munro, 1993) as well as
sphingolipids. Neither of these components is present to any
extent in membranes of the ER.

In vitro studies of Nezil and Bloom (Nezil and Bloom, 1992)

using purified egg phospholipids, such as phosphatidylcholine,
showed that the addition of cholesterol leads to an increase in
membrane thickness. Accordingly, it has been suggested that
cholesterol and other lipids such as the sphingolipids, both of
which are present in the Golgi, might be responsible for the
increase in the thickness of the membrane as they move from
the ER through the Golgi to the plasma and lysosomal
membranes (Bretscher and Munro, 1993). Furthermore, it has
been suggested that the detergent-insoluble lipid rafts present
in the plasma membrane, which are composed of sphingolipids
and cholesterol, are responsible for moving these specific lipids
from the Golgi to the plasma membrane (Simons and Ikonen,

1997; Simons and Ikonen, 2000). Suffice it to say, newly
forming endosomes and caveoli contain both cholesterol
and sphingolipids and so when they enter the cytoplasm
they resemble the lipid composition of lysosomes and
phagolysosomes (Golgi derivatives). In fact, if cholesterol is
depleted from cells, phagocytosis of at least certain species of
bacteria (Ferrari et al., 1999; Gatfield and Pieters, 2000) is
inhibited. It is not known if L. pneumophila phagocytosis
requires cholesterol or not.

The puzzle then is how does the thickness of the endosomal

membrane that surrounds the phagocytosed L. pneumophila
change within minutes? As the change in membrane thickness
only occurs if the ER vesicles and mitochondria bind to
the phagosome, presumably these organelles may help to
orchestrate this result.

Shortly after we started this study, we anticipated that the

change in membrane thickness must be due to the fusion of the
ER membranes with the plasma membrane. However, if this
were to occur it would be novel, as in no other case does the
ER membrane fuse with the plasma membrane, a scenario
which would short circuit the Golgi and lead to the discharge
into the extracellular space of resident ER proteins and proteins
destined for lysosomes. In short, a disaster! Interestingly,
we have shown in this report that the membrane thickness
change, which we presume is symptomatic to changes in lipid
composition, occurs within 15 minutes of infection, yet
ribosomes do not attach directly to the former endosomal
membrane for at least six hours (stage 4). Furthermore, the
volume of the endosomes does not increase in parallel with
these changes in membrane lipids, which one would anticipate
would occur if ER vesicles fuse with the endosomal membrane.
Thus, fusion of the phagosomal membrane with the ER
vesicles cannot account for the rapid change in membrane
thickness observed, but it could account for attachment of the
ribosomes to the former endosomal membrane after six hours
of infection.

It is known that lipids, such as cholesterol, flux in and out

of the plasma membrane (Ferrari et al., 1999; Lange and
D’Alessandro, 1977; Slotte and Bierman, 1987). This flux
could clearly influence membrane thickness as measured in our
electron micrographs, but at least for the red cell, these fluxes
are slow, for example, they occur over hours not minutes. What
we do know from this study is (1) that wild-type L.
pneumophila is somehow orchestrating changes in membrane
thickness, a process that does not occur in the dotA, icmR
mutants or in bacterial infections other than L. pneumophila,
and (2) that the change in membrane thickness is associated
with the attachment of the ER vesicles to the phagosome.

As we have demonstrated in this report that ER vesicles

attach to the plasma membrane in uninfected cells by the hair-
like connections, and since the icmW mutant uses these hairs
to attach ER vesicles to the endosome within minutes but
changes in membrane thickness do not occur for hours, it
is clear that these events (ER attachment and changes in
membrane thickness) are the result of L. pneumophila taking
advantage of a host cell machinery that is already in place.

We should mention for completeness that measurement of

membrane thickness in thin sections is only a first step, and a
rather crude one, in analyzing what is going on in the lipid-
changeover process. Nevertheless, since both membranes are
zippered together and thus parallel to each other, we are

JOURNAL OF CELL SCIENCE 114 (24)

4649

How

Legionella pneumophila modifies its phagosome

confident that we can reliably detect differences in the
thickness of these membranes, although the molecular nature
of these differences cannot be inferred by our techniques.

What might be the general significance of
changeover in lipids of the endosomal membrane?

Pathogens and/or specialized cells have been useful in the past
for directing our attention to basic cell biological phenomena.
Although there are many examples to illustrate this point, one
striking case is the bacterial pathogen Listeria, which has given
us valuable insight in the control of actin assembly. We believe
that L. pneumophila may be another case. What our present
study has shown us is that ER vesicles and mitochondria are
attached to endosomes containing this pathogen but having
seen this, we find that ER vesicles are also attached in
uninfected macrophages to precursors of endosomes,
namely the plasma membrane. Such ER-plasma-membrane
attachments, although common in plant (Staehelin, 1997) and
protozoan cells (Sinai et al., 1997), are infrequently described
in animal cells.

To be more specific, ER vesicles are often found attached to

the plasma membrane in plant cells at distances comparable to
what we have described here (Craig and Staehelin, 1988;
Staehelin, 1997). In addition, these membranes are connected
by tiny hairs similar to those described here. The hairs are
best seen by examining the structure of plasmadesmata,
intercellular bridges that connect the cytoplasm of adjacent
plant cells. In plasmadesmata, a slender ER membranous
tubule or desmotubule spans the narrow bridge and is
continuous in the two adjacent cells with rough ER (Tilney et
al., 1991). The desmotubule in turn is connected to the plasma
membrane by tiny osmiophilic hairs (Ding et al., 1992; Tilney
et al., 1991).

In protozoa, ER-plasma-membrane connections are common

beneath the pellicle of ciliated protozoa and in the parasitic
protozoa that include the apicoplexa. The ER-plasma-
membrane attachment has been called the alveolar system in
ciliates and the internal membrane complex in the apicomplexa.

In uninfected higher animal cells such ER-plasma-

membrane connections are found in skeletal muscle cells
where the sarcoplasmic reticulum is bound to the transverse
tubule system or T system, in neurons (Henkart et al., 1976)
and in the outer hair cell of the cochlea (Brownell et al., 2001;
Forge, 1991; Pollice and Brownell, 1993). In these last three
systems in higher animal cells, subsurface cisterns containing
calcium stores appear to be instrumental in the regulation of
cell function after stimulation from outside the cell.

What L. pneumophila has shown us is that the endosome

becomes transformed into an ER membrane first by a change
in membrane thickness and later by the attachment of
ribosomes. Since cells can flux labelled lipids out of the plasma
membrane into the ER in plant cells (Grabski et al., 1993), and
since bacterial products such as polymyxin B mediates
exchange between membranes in which no membrane fusion
occurs (Cajal et al., 1996a; Cajal et al., 1996b), the possibility
exists that animal cells may have the innate capability of
transforming plasma membrane into an ER membrane.

L.G.T. wishes to thank Paul Edelstein for his enthusiastic help

in the beginning of this study in providing samples from L.
pneumophila-infected guinea pig macrophages. The authors thank

Kelly Vranich for preparing manuscript drafts. L.G.T. is supported by
NIH GM52857, C.R.R. by NIH R29A141699, and O.S.H. by NIH
F32AI10654.

REFERENCES

Andrews, H. L., Vogel, J. P. and Isberg. R. R. (1998). Identification of linked

Legionella pneumophila genes essential for intracellular growth and evasion
of the endocytic pathway. Infect. Immun. 66, 950-958.

Berger, K. H., Merriam, J. J. and Isberg, R. R. (1994). Altered intracellular

targeting properties associated with mutations in the Legionella
pneumophila dotA gene. Mol. Microbiol. 14, 809-822.

Bretscher, M. S. and Munro, S. (1993). Cholesterol and the Golgi apparatus.

Science 261, 1280-1281.

Brownell, W. E., Spector, A. A., Raphael, R. M. and Popel, A. S. (2001).

Micro- and nanomechanics of the cochlear outer hair cell. Annu. Rev.
Biomed. Eng. 3, 169-194.

Byrne, B. and Swanson, M. S. (1998). Expression of Legionella pneumophila

virulence traits in response to growth conditions. Infect. Immun. 66, 3029-
3034.

Cajal, Y., Ghanta, J., Easwaran, K., Surolia, A. and Jain, M. K.

(1996a). Specificity for the exchange of phospholipids through polymyxin
B mediated intermembrane molecular contacts. Biochemistry 35, 5684-
5695.

Cajal, Y., Rogers, J., Berg, O. G. and Jain, M. K. (1996b). Intermembrane

molecular contacts by polymyxin B mediate exchange of phospholipids.
Biochemistry 35, 299-308.

Christie, P. J. and Vogel, J. P. (2000). Bacterial type IV secretion: conjugation

systems adapted to deliver effector molecules to host cells. Trends
Microbiol. 8, 354-360.

Clemens, D. L. and Horwitz, M. A. (1995). Characterization of the

Mycobacterium tuberculosis phagosome and evidence that phagosomal
maturation is inhibited. J. Exp. Med. 181, 257-270.

Coers, J., Kagan, J. C., Matthews, M., Nagai, H., Zuckman, D. M.

and Roy, C. R. (2000). Identification of Icm protein complexes that
play distinct roles in the biogenesis of an organelle permissive for
Legionella pneumophila intracellular growth. Mol. Microbiol. 38, 719-
736.

Craig, S. and Staehelin, L. A. (1988). High pressure freezing of intact plant

tissues. Evaluation and characterization of novel features of the
endoplasmic reticulum and associated membrane systems. Eur. J. Cell Biol.
46, 81-93.

Ding, B., Turgeon, R. and Parthatsarathy, M. V. (1992). Substructure of

freeze substituted plasmadesmata. Parasitology. 199, 28-41.

Dobbelaere, D. and Heussler, V. (1999). Transformation of leukocytes by

Theileria parva and T. annulata. Annu. Rev. Microbiol. 53, 1-42.

Ferrari, G., Langen, H., Naito, M. and Pieters, J. (1999). A coat protein on

phagosomes involved in the intracellular survival of Mycobacteria. Cell 97,
435-447.

Finlay, B. B. and Falkow, S. (1997). Common themes in microbial

pathogenicity revisited. Microbiol. Mol. Biol. Rev. 61, 136-169.

Forge, A. (1991). Structural features of the lateral walls in mammalian

cochlear outer hair cells. Cell Tissue Res. 265, 473-483.

Gao, L. Y., Susa, M., Ticac, B. and Abu Kwaik, Y. (1999). Heterogeneity

in intracellular replication and cytopathogenicity of Legionella pneumophila
and Legionella micdadei in mammalian and protozoan cells. Microb.
Pathog. 27, 273-287.

Gatfield, J. and Pieters, J. (2000). Essential role for cholesterol in entry of

Mycobacteria into macrophages. Science 288, 1647-1650.

Gerhardt, H., Walz, M. J., Faigle, M., Northoff, H., Wolburg, H. and

Neumeister, B. (2000). Localization of Legionella bacteria within
ribosome-studded phagosomes is not restricted to Legionella pneumophila.
FEMS Microbiol. Lett. 192, 145-152.

Grabski, S., de Feijter, A. W. and Schindler, M. (1993). Endoplasmic

reticulum forms a dynamic continuum for lipid diffusion between
contiguous soybean root cells. The Plant Cell. 5, 25-38.

Hall, B. F. (1993). Trypanosoma cruzi: mechanisms for entry into host cells.

Semin. Cell Biol. 4, 323-333.

Harb, O. S. and Abu Kwaik, Y. (1998). Identification of the aspartate-beta-

semialdehyde dehydrogenase gene of Legionella pneumophila and
characterization of a null mutant. Infect. Immun. 66, 1898-1903.

Harb, O. S. and Abu Kwaik, Y. (2000). Essential role for the Legionella

4650

pneumophila rep helicase homologue in intracellular infection of
mammalian cells. Infect. Immun. 68, 6970-6978.

Harb, O. S., Gao, L. Y. and Abu Kwaik, Y. (2000). From protozoa to

mammalian cells: a new paradigm in the life cycle of intracellular bacterial
pathogens. Environ. Microbiol. 2, 251-265.

Heinzen, R. A., Scidmore, M. A., Rockey, D. D. and Hackstadt, T. (1996).

Differential interaction with endocytic and exocytic pathways distinguish
parasitophorous vacuoles of Coxiella burnetii and Chlamydia trachomatis.
Infect. Immun. 64, 796-809.

Henkart, M., Landis, D. M. and Reese, T. S. (1976). Similarity of junctions

between plasma membranes and endoplasmic reticulum in muscle and
neurons. J. Cell Biol. 70, 338-347.

High, N., Mounier, J., Prevost, M. C. and Sansonetti, P. J. (1992). IpaB of

Shigella flexneri causes entry into epithelial cells and escape from the
phagocytic vacuole. EMBO J. 11, 1991-1999.

Horwitz, M. A. (1983). Formation of a novel phagosome by the Legionnaires’

disease bacterium (Legionella pneumophila) in human monocytes. J. Exp.
Med. 158, 1319-1331.

Kartenbeck, J., Stukenbrok, H. and Helenius, A. (1989). Endocytosis of

simian virus 40 into the endoplasmic reticulum. J. Cell Biol. 109, 2721-
2729.

Kim, J. and Klionsky, D. J. (2000). Autophagy, cytoplasm-to-vacuole

targeting pathway, and pexophagy in yeast and mammalian cells. Annu. Rev.
Biochem. 69, 303-342.

Lange, Y. and D’Alessandro, J. S. (1977). Characterization of mechanisms

for transfer of cholesterol between human erythrocytes and plasma.
Biochemistry 16, 4339-4343.

Matthews, M. and Roy, C. R. (2000). Identification and subcellular

localization of the Legionella pneumophila IcmX protein: a factor essential
for establishment of a replicative organelle in eukaryotic host cells. Infect.
Immun. 68, 3971-3982.

Nezil, F. A. and Bloom, M. (1992). Combined influence of cholesterol and

synthetic amphiphillic peptides upon bilayer thickness in model membranes.
Biophys. J. 61, 1176-1183.

Pelkmans, L., Kartenbeck, J. and Helenius, A. (2001). Caveolar endocytosis

of simian virus 40 reveals a new two-step vesicular-transport pathway to the
ER. Nat. Cell Biol. 5, 473-483.

Pizarro-Cerda, J., Meresse, S., Parton, R. G., van der Goot, G., Sola-

Landa, A., Lopez-Goni, I., Moreno, E. and Gorvel, J. P. (1998). Brucella
abortus transits through the autophagic pathway and replicates in the
endoplasmic reticulum of nonprofessional phagocytes. Infect. Immun. 66,
5711-5724.

Pollice, P. A. and Brownell, W. E. (1993). Characterization of the outer hair

cell’s lateral wall membranes. Hear. Res. 70, 187-196.

Portnoy, D. A. and Jones, S. (1994). The cell biology of Listeria

monocytogenes infection (escape from a vacuole). Ann. NY Acad. Sci. 730,
15-25.

Roy, C. R., Berger, K. H. and Isberg, R. R. (1998). Legionella pneumophila

DotA protein is required for early phagosome trafficking decisions
that occur within minutes of bacterial uptake. Mol. Microbiol. 28, 663-
674.

Sadosky, A. B., Wiater, L. A. and Shuman, H. A. (1993). Identification of

Legionella pneumophila genes required for growth within and killing of
human macrophages. Infect. Immun. 61, 5361-5373.

Sandstrom, C. and Nilsson, K. (1976). Establishment and characterization

of a human histocytic lymphoma cell line (U937). Int. J. Cancer. 17, 565-
577.

Segal, G., Purcell, M. and Shuman, H. A. (1998). Host cell killing and

bacterial conjugation require overlapping sets of genes within a 22-kb region
of the Legionella pneumophila genome. Proc. Natl. Acad. Sci. USA. 95,
1669-1674.

Segal, G. and Shuman, H. A. (1997). Characterization of a new region

required for macrophage killing by Legionella pneumophila. Infect. Immun.
65, 5057-5066.

Sibley, L. D. (1993). Interactions between Toxoplasma gondii and its

mammalian host cells. Semin. Cell Biol. 4, 335-344.

Simons, K. and Ikonen, E. (1997). Functional rafts in cell membranes. Nature

387, 569-572.

Simons, K. and Ikonen, E. (2000). How cells handle cholesterol. Science 290,

1721-1726.

Sinai, A. P., Webster, P. and Joiner, K. A. (1997). Association of host cell

endoplasmic reticulum and mitochondria with the Toxoplasma gondii
parasitophorous vacuole membrane: a high affinity interaction. J. Cell Sci.
110, 2117-2128.

Slotte, J. P. and Bierman, E. L. (1987). Movement of plasma-membrane

sterols to the endoplasmic reticulum in cultured cells. Biochem. J. 248, 237-
242.

Staehelin, L. A. (1997). The plant ER: a dynamic organelle composed of a

large number of discrete functional domains. Plant J. 11, 1151-1165.

Swanson, M. S. and Hammer, B. K. (2000). Legionella pneumophila

pathogenesis: a fateful journey from amoebae to macrophages. Annu. Rev.
Microbiol. 54, 567-613.

Swanson, M. S. and Isberg, R. R. (1995). Association of Legionella

pneumophila with the macrophage endoplasmic reticulum. Infect. Immun.
63, 3609-3620.

Tilney, L. G., Cooke, T. J., Connelly, P. S. and Tilney, M. S. (1991). The

structure of plasmodesmata as revealed by plasmolysis, detergent extraction,
and protease digestion. J. Cell Biol. 112, 739-747.

Vogel, J. P., Andrews, H. L., Wong, S. K. and Isberg, R. R. (1998).

Conjugative transfer by the virulence system of Legionella pneumophila.
Science 279, 873-876.

Wiater, L. A., Dunn, K., Maxfield, F. R. and Shuman, H. A. (1998). Early

events in phagosome establishment are required for intracellular survival of
Legionella pneumophila. Infect. Immun. 66, 4450-4460.

JOURNAL OF CELL SCIENCE 114 (24)