Introduction
Actin polymerization is thought to provide the motive force for crawling
cells by driving the spread of lamellipodia 1. It may also be
involved when membrane ruffles engulf external fluid during macropinocytosis 2. Here we show that, in cells transfected with a fusion protein consisting
of green fluorescent protein (GFP) and 
-actin, pinosomes (vesicles used
for fluid uptake) ignite a burst of actin polymerization when they are pinched
off from the plasma membrane. Pinosomes then move into the cytosol at the
tips of short-lived actin 'comet tails' that are similar to those
that propel Listeria3 and other microorganisms 4, 5 through infected cells. Like Listeria, pinosomes appear
to carry the machinery required for propulsive actin polymerization. The direction
of pinosome movement indicates that they may acquire this machinery from the
crests of membrane ruffles. We suggest that actin polymerization moves the
leading edge of ruffles. Endocytic vesicles may also use actin polymerization
to move into the cytosol after being pinched off from the plasma membrane.
We transfected rat basophilic leukaemia (RBL) cells with GFP-labelled -actin 6 and allowed them to ruffle and undergo macropinocytosis in a mildly
hyperosmolar medium 7. Ruffles faintly aglow with GFP–actin
engulfed fluid and formed vesicular structures near the plasma membrane (not
shown). The structures fluoresced (Fig. 1a, arrowhead)
as they ignited a burst of actin polymerization that caused a streak of actin
to move through the cell, much like actin 'comet tails'. The tips
of the actin tails were cup-shaped as if to accommodate vesicles, and appeared
to push spherical objects that were visible as dark shadows. Large tails sometimes
appeared hollow (not shown).
Figure 1: Tails of F-actin move through live RBL cells.
a, Formation of GFP-labelled actin tails during pinocytosis. Arrowhead
points to polymerized actin partly surrounding a dark structure that probably
represents a pinocytic vesicle; arrows show the growth of an actin tail whose
cup-shaped end probably propels the pinocytic vesicle. Results were recorded
by EFM 8. Times are relative to the beginning of the sequence.
Video sequences showing actin comet tails in cells expressing GFP–actin
are available as Supplementary Information
. b, The speed of motile actin tails
is plotted. The abscissa shows speed, measured as the distance between the
tip of an actin tail in successive video frames divided by 5 s (the interval
between frames). Mean speed was 0.24 
0.10 
m s-1
(s.d.). Actin tails persist for some time after they have stopped
moving; the speed of such 'spent' actin tails (0–0.05 m
s-1) was not included in the histogram or the mean, as it would
result in a large bar at the origin. We made 699 measurements of 126 actin
tails in 9 cells; results were recorded as in a. c, Average
fluorescence in a region of interest (ROI, see inset) measured in successive
frames and plotted against time. Curves from 25 tails in 3 cells were normalized,
with 100% representing peak fluorescence and 0% the local background, and
then averaged; bars give s.e.m. In a–c the external solution
was mildly hyperosmolar and in b, c it also contained 10 nM
PMA to enhance the frequency of pinocytic events 18.
The results shown in Fig. 1a were recorded by evanescent field fluorescence microscopy (EFM) 8, a technique that selectively images the bottom of cells where they adhere to a glass coverslip. Some actin tails seemed to originate in the middle of the cell in such recordings; such tails were also seen when the bottom of chemically fixed cells was viewed by laser-scanning confocal microscopy (LSCM). However, three-dimensional images of such tails reconstructed from serial confocal sections showed them to extend to the plasma membrane above (data not shown). In four cells analysed in this way, the appearance of all tails was consistent with the tails having begun at the plasma membrane.
To characterize the motion of actin tails, we recorded video clips by EFM
and plotted the speed of actin tails as a histogram (Fig. 1b
). The speed at the peak of the histogram (0.2 m s-1
; mean 0.24 m s-1) was comparable to results
obtained with Listeria9 (0.23 m s-1)
and Shigella4 (0.1 m s-1). We
also measured the fluorescence at a fixed point on an actin tail and watched
it dim as the tail moved on and the actin depolymerized. Tails dimmed with
half-lives of 20 s (Fig. 1c), values that are again
comparable with those obtained for Listeria3 (33 s)
and Shigella10 (38 s). Actin tails thus appear to exhibit
the same speed and persistence regardless of whether they propel pinosomes
or bacteria. However, whereas Listeria tails are active for up to 30
min 11, pinosome tails stopped and faded after only 1–2
min under our experimental conditions (Fig. 1a). Under
more physiological conditions, tails may be shorter and even less persistent,
making them harder to see and causing endosomes to move more slowly 2 or for less time. The limited lifetime of actin tails indicates
that actin polymerization may be regulated during pinocytosis.
To identify the cargo carried by actin tails, we added rhodamine-labelled
dextran to the external medium and studied live cells in equatorial sections
by LSCM. Cells contained red, dextran-containing pinosomes (42.0
2.9 per section and cell), of which 20.4 6.2% were associated with
a green tail (consisting of GFP–actin) during 250 s of observation (Fig. 2a). At least 90% of the tails (62 of 69 in 7 cells)
had red pinosomes at their tips. The actual percentage of tails carrying pinosomes
is probably higher, as very small vesicles must be well in focus to be seen
and it is possible that the tips of some actin tails were never in focus during
the observation time. Fig. 2b shows the plasma membrane
invaginating to engulf fluid in a pinosome. An actin tail was initiated between
the plasma membrane and the pinosome, and moved inward and out of the focal
plane. The pinosome beneath the plasma membrane vanished, presumably carried
away by the actin tail. Actin tails thus appear to begin at plasma-membrane
sites of pinosome formation. For one cell, speeds of pinosomes were measured
over 250 s and plotted as a histogram (Fig. 2c). Pinosomes
moved rapidly only when associated with an actin tail.
Figure 2: Actin tails propel pinosomes.
a, A GFP–actin-expressing cell in external medium containing 4 mg ml-1 Texas-Red-conjugated dextran (Molecular Probes). Dextran generated a uniform red background and filled pinosomes so they became visible as red spots. Two (arrows) were associated with an actin tail. b, Magnified images of the same cell as shown in a, taken at the indicated intervals. Arrow, site of a plasma-membrane invagination. Red, Texas-Red-labelled dextran; green, GFP–actin. c, Speed of pinosomes in association with actin (right) or unassociated with actin (left). Ten of the forty-five pinosomes in this cell were temporarily associated with actin, most often with an actin tail. As actin tails do not vanish immediately after pinosomes have stopped moving, even actin-associated pinosomes experience periods of immobility. d, Angle of actin tails initiated at the plasma membrane. Fluorescent phalloidin marks an F-actin tail (green); fluorescent IgE (red) is taken up at the plasma membrane in a pinosome (arrow) in a fixed cell. e, Histogram showing the angles of actin tails with the plasma membrane. Abscissa, angles with the plasma membrane (dashed lines in d); these angles are positive when actin tails point their pinosomes into the cytosol and negative when they point towards the plasma membrane. LSCM was used throughout. External medium contained PMA and was isosmolar in a, b and mildly hyperosmolar in c–e. f–h, A model of how pinosomes gain polarity and direction. f, Ruffles carry actin-polymerization sites (black) at their crests. Polymerizing actin moves the tips in the direction shown by arrows; lines illustrate the direction of polymerization. g, The ruffle deposits the actin-polymerizing machinery at the point at which the pinocytic cup pinches off. h, Continued actin polymerization at this site forms a tail that propels the pinosome into the cytosol.
Full size image (34 KB)To determine whether actin tails were an artefact of transfection or GFP
expression, we studied fixed, untransfected cells. Pinosomes were labelled
by letting cells take up fluorescent immunoglobulin E bound to Fc
RI
receptors on the plasma membrane, and F-actin was labelled with fluorescent
phalloidin. Phalloidin easily stained actin tails. Cells were viewed by LSCM
in equatorial sections that were focused as much as possible on the tips of
tails (Fig. 2d). Of 84 actin tails in 50 cells, 78 clearly
carried pinosomes and only 2 clearly did not. Thus actin tails also occur
in untransfected cells, and most or all carry pinosomes at their tips.
To determine whether pinosomes depart from the plasma membrane in a preferred direction, we studied actin tails that touched the plasma membrane (29 of 84). We measured the angles of these tails in relation to the plasma membrane (Fig. 2d) and plotted them as a histogram ( Fig. 2e). Most tails (26 of 29) pointed towards the cytosol, and 23 of these were within 45° of being perpendicular to the membrane. Thus, actin tails do not form at random angles with the plasma membrane (P<0.001, Komolgoroff–Smirnov test).
We have shown that pinosomes ignite transient bursts of actin polymerization after they pinch off from the plasma membrane, and that they then move into the cytosol at the tip of the resulting actin 'comet tails'. Comet tails have previously been seen in association with 'rocketing endosomes' in intact, uninfected, La3+- and Ni2+-treated cells 12 and, most recently, with Golgi-derived clathrin-coated vesicles in intact, uninfected HeLa cells 13. We have shown now that actin tails in RBL cells are associated specifically with moving pinosomes. These tails are similar or identical to those that propel Listeria and Shigella, in both their speed and their persistence. In Listeria, movement is thought to result from propulsive actin polymerization 3 and the same may apply to pinocyte movement.
Our results show that the machinery for propulsive actin polymerization is concentrated in ruffles. Ruffles must carry this machinery (which may include the complex Arp2/3 (ref. 14) because pinosomes do, and they are formed from ruffles 2. Most pinosomes have a region in which this machinery is more active than elsewhere, because uniform actin polymerization over the entire pinosome surface would prohibit movement. (Indeed, rare pinosomes that show GFP–actin fluorescence around their entire circumference do not move; data not shown.) This more active region must usually be the 'navel' (the point at which the pinosome pinched off from the plasma membrane), because most pinosomes move nearly vertically into the cytosol. The consistent localization of actin polymerization on the pinosome can be explained most simply if the machinery for actin polymerization is most active at the crests of ruffles (Fig. 2f–h). Its presence there strengthens the idea that ruffles extend by a mechanism involving actin polymerization.
After budding from the plasma membrane, pinosomes and endocytic vesicles must generally move into the cytosol for processing. Actin is widely thought to participate in endocytosis 7, 15, 16, 17. We propose that one of actin's functions in both endocytosis and pinocytosis may be to help endocytic vesicles and pinosomes move into the cytosol, by means of a brief burst of actin polymerization.
Methods
Cells.
Human -actin was fused to the carboxy terminus
of EGFP (Clontech) and expressed under control of the long -actin promoter
as described 6. RBL cells were grown to confluence at 37 °C
and in 5% CO2 in RPMI medium (Sigma). For transfection, they were
collected from 90-mm dishes by scraping, washed three times in HBS (150 mM
NaCl, 5 mM KCl, 1.8 mM CaCl2, 0.8 mM MgCl2, 20 mM HEPES,
1 mM glucose, pH 7.4), suspended in 0.4 ml HBS and electroporated with 30
g vector. Cells were then plated onto flamed glass coverslips and used
24–48 h later. For imaging, glass coverslips with adherent cells were
transferred to the experimental chamber containing HBS supplemented with either
10 nM PMA or 150 mM sucrose or both, as indicated. They were imaged 15–30
min later and kept at 37 °C throughout.
Imaging.
Cells were observed with an epi-illumination
evanescent-field microscope that selectively illuminated a thin layer of cytosol
where cells adhered to the glass coverslip 8. Even in thick
cells the method avoids exciting significant out-of-focus fluorescence, so
that all fluorescent light returning through the objective can be used for
imaging and none is lost at the pinhole of a confocal microscope. Up to 200
images were taken with a back-illuminated cooled charge-coupled-device camera
(Princeton Instruments) at 0.2 Hz from single cells without significant bleaching
or photodamage. Movements observed are projections of three-dimensional trajectories
into the plane of the coverslip. For simultaneous viewing of two fluorophores
in living cells, we used a Leica TCS-NT confocal microscope (
100, 1.3
NA oil objective). Closed experimental chambers were made of a microscope
slide, a piece of parafilm and a coverslip with adherent cells, and sealed
with silicone grease. Up to 50 images were taken at 0.2 Hz with the confocal
pinhole opened to increase light collection. Throughout, the microscope or
microscope stage, specimen and objective were placed in an enclosure and maintained
at 37 °C. Images were analysed with MetaMorph (Universal Imaging); the
resulting numerical data were analysed in Origin (MicroCal).
Immunofluorescence.
For immunofluorescence, untransfected
RBL cells were washed at room temperature to remove medium and then incubated
for 30 min (4 °C) in hyperosmolar HBS with 10 nM PMA and 2 g ml
-1 anti-DNP-IgE (Sigma) conjugated to Alexa-568 (Molecular Probes).
Alexa-568–IgE binds to the FcRI receptor and is internalized
in pinosomes. Cells were washed thrice in HBS (4 °C), placed in stimulation
buffer (15 min at 37 °C), then fixed in PBS (Sigma) plus 3.7% paraformaldehyde
(15 min), permeabilized with PBS + 0.1% Triton-X100 (2 min) and stained for
F-actin in PBS + 20 U ml-1 Oregon-green–phalloidin (Molecular
Probes) for 60 min (all at 20 °C). Cells were then viewed by LSCM with
the pinhole stopped down to maximize resolution. Serial confocal sections
of phalloidin-stained cells were deconvoluted using the Huygens-2 software
package (Scientific Volume Imaging, Hilversum, Netherlands); the point spread
function was determined by optical serial-sectioning of a fluorescent bead.
Vertical sections were calculated using the program Imaris (Bitplane, Zurich,
Switzerland).
