Suliana Manley1, Jennifer M Gillette1,
George H Patterson1, Hari Shroff2, Harald F Hess2,
Eric Betzig2& Jennifer Lippincott-Schwartz1
We combined photoactivated localization microscopy (PALM)
with live-cell single-particle tracking to create a new method
termed sptPALM. We created spatially resolved maps of
single-molecule motions by imaging the membrane proteins
Gag and VSVG, and obtained several orders of magnitude more
trajectories per cell than traditional single-particle tracking
enables. By probing distinct subsets of molecules, sptPALM
can provide insight into the origins of spatial and temporal
heterogeneities in membranes.
Cell membranes are characteristically heterogeneous both structu-
rally anddynamically.Byperformingsingle-particletrackingin live
cells, one can access information on the heterogeneities in the
motions of individual proteins to provide insights into many
cellular events1–3. For single-molecule detection, proteins of inter-
est are conjugated to probes such as gold beads4or fluorescent-
protein chimeras3. However, the number and density of single
molecules that can be tracked in an individual cell is limited in
traditional single-particle tracking studies because only a single
tion requires that molecular separations be greater than their
diffraction-limited size. Thus, single-particle tracking has been
unable to probe cell membranes in the spatially and temporally
resolved fashion required to determine structural signatures of
Super-resolution techniques such as PALM5enable the imaging
of fluorescent-protein chimeras to reveal the organization of
genetically expressed proteins on the nanoscale with a density of
molecules high enough to provide structural context. Recently, the
use of photoactivatable markers for subdiffraction localization has
been applied to imaging in living cells, but the dynamics of
individual molecules were not resolved6. Here we combined the
techniques of PALM and single-particle tracking to resolve the
dynamics of individual molecules by tracking them in live cells; a
method we call sptPALM. We obtained information on the posi-
tions of single molecules by activating, localizing and bleaching
many subsets of photoactivatable fluorescent-protein chimeras
(Supplementary Video 1 online). To enable sptPALM on live
cells, we optimized the data acquisition rate and cell viability by
(COS7), operating in a total internal reflection geometry and using
ahigh numerical aperture objective (Olympus APO100XO-HR-SP;
1.65 NA). In addition, we used proteins tagged with EosFP7, a
photoconvertible protein with a high photon count and a high
contrast ratio between converted and unconverted forms. Using
these parameters, we imaged membrane proteins at a rate of 20
frames per second without inducing detrimental effects to living
cells (Supplementary Fig. 1 and Supplementary Video 2 online).
This permitted the construction of maps of single molecule diffu-
sion for up to thousands of molecules in the plasma
membrane, providing a means of obtaining spatially resolved
information on cellular dynamics and local environments on the
We transfected COS-7 cells with either of two membrane
proteins known to have distinctly different spatial distributions
and mobilities: the tsO45 vesicular stomatitis virus G protein
(VSVG) and the human immunodeficiency virus type 1 (HIV-1)
structural protein Gag. VSVG is distributed relatively homoge-
neously across the plasma membrane, where a large fraction freely
diffuses8; in contrast, Gag multimerizes into immobile virus-like
particles (VLPs) that can bud from the plasma membrane9. Time-
integrated PALM images (Fig. 1a) reflect the distribution of all
molecular peaks localized to o25 nm. As expected, Gag tagged
with tandem dimer EosFP (Gag-Eos) formed bright puncta of
100–200 nm size (Fig. 1b), consistent with electron microscopy
data of VLPs. In addition, several larger bright regions (possibly
corresponding to clusters of VLPs) and areas of lower than average
density (voids) were visible5. Unlike Gag-Eos, VSVG tagged with
dimeric EosFP (VSVG-Eos) displayed gradual variations in
domains (Fig. 1b).
In addition to examining the spatial distribution of proteins, use
of sptPALM provided dynamic information on individual mole-
cules. As in traditional single-particle tracking, we created trajec-
tories by linking molecular peaks in consecutive frames according
to their proximity (Supplementary Methods, Supplementary
for both Gag-Eos and VSVG-Eos (see example trajectories in
Fig. 1c). Unlike traditional single-particle tracking in which a
single ensemble of molecules is imaged and tracked until they are
RECEIVED 25 OCTOBER 2007; ACCEPTED 19 DECEMBER 2007; PUBLISHED ONLINE 13 JANUARY 2008; DOI:10.1038/NMETH.1176
1National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, Maryland 20892, USA.2Howard Hughes Medical Institute, Janelia
Farm Research Campus, Ashburn, Virginia 20147, USA. Correspondence should be addressed to J.L.-S. (firstname.lastname@example.org).
NATURE METHODS | VOL.5 NO.2 | FEBRUARY 2008 | 155
© 2008 Nature Publishing Group http://www.nature.com/naturemethods
bleached3, photoactivatable fluorophores enabled multiple ensem-
bles of molecules to be activated, imaged and bleached. This
allowed us to obtain high-density single-molecule trajectories, up
to B50 per mm2(Fig. 1d). Gag-Eos trajectories varied from
apparently immobile populations to mobile populations with
pathlengths up to B1 mm. In contrast, VSVG-Eos appeared highly
mobile, exploring large regions of the plasma membrane. The use
of sptPALM allowed localization and tracking of many overlapping
trajectories because the distance between fluorescent molecules at
any time was greater than several times the width of their point
A notable improvement afforded by sptPALM is the ability
to probe the dynamics of many molecules in a single cell. By
comparison, 10–100 cells are typically used to obtain statistically
significant single-particle tracking information on individual
proteins tagged with conventional labels because each cell yields
less than 10 tracks on average10–12. We achieved similar statistics
with sptPALM on a single cell (Fig. 2a); moreover, the large
number of molecular trajectories we obtained allowed us to
define molecular environments within individual cells. We
calculated the mean-squared displacement (MSD) as a function
of time lag, Dt, for trajectories longer than 15 frames (B3%
of all molecules). For diffusive behavior, we expected the MSD to
increase linearly with Dt for both VSVG and Gag (Fig. 2b).
To determine the short-time diffusion coefficients, D, for these
longer-livedmolecules,weusedtherelationshipMSD¼ C+ 4DDt,
where C is a constant offset. A substantial fraction of the
trajectories resulted in MSDs that increase in a sublinear
fashion (Fig. 2c). This behavior is known as anomalous diffusion:
it has been demonstrated for several membrane proteins and
lipids13, and is hypothesized to be a result of local barriers
Figure 1 | sptPALM imaging of Gag and VSVG expressed in live COS7 cells.
(a) PALM images of Gag-Eos and VSVG-Eos, integrated over 500 s (10,000
images). Molecules are rendered as Gaussian normalized peaks, with a width
corresponding to the uncertainty in their position. Arrows indicate large
Gag-enriched regions. Scale bars, 2 mm. (b) Magnified regions, corresponding
to boxed regions in a. Arrowheads indicate enriched puncta. Scale bars,
200 nm. (c) Two representative magnified Gag and VSVG single molecule
trajectories from the cells shown in a. The tracks represent diffusive (black)
and confined (red) movement. Scale bar, 100 nm. (d) Complete sptPALM
trajectories of localized Gag and VSVG molecules that are longer than
15 frames (750 ms). Each color indicates a different track. Scale bar, 2 mm.
Number of tracks
Fraction of tracks
0.3 0.4 0.5
Figure 2 | Analysis of single molecule trajectories in live COS7 cells. (a) Distribution of trajectory durations of Gag-Eos
and VSVG-Eos single molecules from the images in Figure 1. The histogram represents all the trajectories detected in a
17 ? 17 mm field of the cell membrane with 10,000 images and 50-ms exposures. (b) Calculated MSD for the two
Gag and VSVG trajectories shown in Figure 1c. The tracks represent trajectories with diffusive and confined movement.
(c) Calculated MSD for Gag and VSVG trajectories that display Brownian diffusion and anomalous diffusion. Short lines are
drawn to guide the eye, with slopes 1 (black) and 0.6 (green). (d) Diffusion maps of the Gag- and VSVG-expressing cells
illustrated in Figure 1. Each point represents the starting position of one trajectory with a minimum length of 15 frames
(750 ms). Molecules with D o 0.01 mm2/s are plotted as dark blue points, while the most mobile molecules are plotted
as red points. Representative errors for diffusion coefficients are indicated left of the color bar for the shortest (maximum
error) and longest (minimum error) tracks. Scale bars, 2 mm. (e) Histogram of the distribution of diffusion coefficients of
single Gag and VSVG molecules with a minimum track length of 15 frames (750 ms). For fixed VSVG (4% paraformaldehyde), the distribution of diffusion
coefficients was obtained from two cells. For live Gag and VSVG, the distribution of diffusion coefficients was obtained from three cells each.
156 | VOL.5 NO.2 | FEBRUARY 2008 | NATURE METHODS
© 2008 Nature Publishing Group http://www.nature.com/naturemethods
The high density of dynamic information available from Download full-text
sptPALM allowed us to create a spatially resolved map of single
molecule diffusion coefficients (Fig. 2d) beyond the error owing
to finite run length (Supplementary Methods). This map empha-
sized the large clusters of immobile Gag-Eos molecules, which
may correspond to regions where VLPs were concentrated. No-
tably, although VSVG-Eos is characterized as a highly mobile
molecule, we detected less mobile regions that may represent
filipodial structures where molecular mobility was slowed by
folds in the membrane.
These diffusion maps of VSVG-Eos and Gag-Eos reflect the
dynamics of molecules in single cells. By combining data from
three cells, we constructed a histogram of diffusion coefficients
(Fig. 2e), which confirmed that a larger fraction of VSVG-Eos
molecules were mobile than of Gag-Eos. A control experiment on
fixed cells transfected with VSVG-Eos revealed greater than 95% of
these molecules had D o 2.5 ? 10–2mm2/s (Supplementary Fig. 4
online). The average diffusion coefficient measured for the mobile
fraction in live cells for Gag-Eos was 0.11 ± 0.08 mm2/s, and for
recovery after photobleaching (FRAP)8measurements. Diffusion
maps such as these reveal dynamic heterogeneities in cell mem-
branes unlike diffusion measurements obtained by traditional
single-particle tracking or ensemble measurements such as FRAP.
The heterogeneities, in turn, can be characterized further by their
size, morphology and number of molecules.
The immobile fraction of Gag may represent molecules trapped
in VLPs, whichare known to cluster in plasmamembranedomains
enriched in tetraspanins14. These domains appeared as large bright
spots when imaged with diffraction-limited fluorescence micro-
scopy (Fig. 3a). We used sptPALM to identify and map the
immobile fraction of Gag-Eos (Fig. 3b), revealing structures at
scales larger than the dimensions of individual VLPs. To further
characterize these structures, we performed a clustering analysis
by identifying molecules with neighbors within a 300-nm radius,
then grouped molecules with shared neighbors into the same
cluster. We displayed all clusters containing a minimum of
5 molecules, which demonstrated that diffraction-limited
regions of similar size and intensity may contain dramatically
different numbers of molecules (Fig. 3c). Cluster analysis from
sptPALM revealed these differences and provided quantitative
information inaccessible by total internal reflection fluorescence
microscopy. We found that in this cell, 1,074 molecules, repre-
senting 67% of all trajectories shown in the diffusion maps,
belong to 54 clusters. Although this information by itself
does not allow us to distinguish between different scenarios
for cluster formation, a future study of the characteristics of
Gag-Eos dynamics in clustered regions could provide insight by
revealing whether molecules exhibit reduced mobility or direc-
ted motion as a result of interactions with cellular factors that
aid VLP budding.
The analyses applied here are only the beginning for this new
method; data can be further mined to explore different subsets of
molecules based on their spatial organization or dynamics.
Although development of probes and instrumentation will cer-
tainly lead to improvements in these techniques, presently available
reagents and imaging tools are capable of providing new informa-
tion, as demonstrated in this work. sptPALM can be used to reveal
spatially resolved information about membrane protein dynamics.
This provides a local context for individual molecules, an impor-
tant clue for understanding the mechanisms that drive behaviors
including clustering and anomalous diffusion, both phenomena
ubiquitous to membrane proteins.
Note: Supplementary information is available on the Nature Methods website.
This project was supported by the Intramural Research Program of the US National
Institute of Child Health and Human Development, National Institutes of Health,
and performed while S.M. held a National Research Council Research Associateship
Award at the National Institutes of Health. We thank D. Blair and A.D. Douglass
for providing MATLAB code and helpful discussions.
Published online at http://www.nature.com/naturemethods/
Reprints and permissions information is available online at
1. Teramura, Y. et al. EMBO J. 25, 4215–4222 (2006).
2. Yildiz, A. et al. Science 300, 2061–2065 (2003).
3. Douglass, A.D. & Vale, R.D. Cell 121, 937–950 (2005).
4. Kusumi, A., Sako, Y. & Yamamoto, M. Biophys. J. 65, 2021–2040 (1993).
5. Betzig, E. et al. Science 313, 1642–1645 (2006).
6. Hess, S.T. et al. Proc. Natl. Acad. Sci. USA 104, 17370–17375 (2007).
7. Wiedenmann, J. et al. Proc. Natl. Acad. Sci. USA 101, 15905–15910 (2004).
8. Kenworthy, A.K. et al. J. Cell Biol. 165, 735–746 (2004).
9. Jouvenet, N. et al. PLoS Biol. 4, 2296–2310 (2006).
10. Wieser, S., Moertelmaier,M., Fuertbauer, E., Stockinger, H.& Schutz,G.J.Biophys.
J. 92, 3719–3728 (2007).
11. Vrljic, M., Nishimura,S.Y.,Brasselet,S.,Moerner, W.E. & McConnell, H.M.Biophys.
J. 83, 2681–2692 (2002).
12. Ewers, H. et al. Proc. Natl. Acad. Sci. USA 102, 15110–15115 (2005).
13. Feder, T.J.,Brust-Mascher, I., Slattery, J.P.,Baird, B. & Webb, W.W. Biophys. J.70,
14. Nydegger, S., Khurana, S., Krementsov, D.N., Foti, M. & Thali, M. J. Cell Biol. 173,
Figure 3 | Cluster analysis on the immobile fraction of Gag. (a) Total internal
reflection fluorescence microscopy image of the unconverted Gag-Eos taken
under 488 nm excitation before photoconversion and data collection for
sptPALM. Circles indicate corresponding regions in each image. (b) A map of
the positions of molecules with D o 0.05. Each point represents the starting
position of one trajectory with a minimum length of 15 frames (750 ms).
(c) Immobile Gag molecules within a 300 nm radius were grouped as
clusters. Each color indicates a separate cluster of at least five molecules.
Scale bar, 2 mm.
NATURE METHODS | VOL.5 NO.2 | FEBRUARY 2008 | 157
© 2008 Nature Publishing Group http://www.nature.com/naturemethods