Quantitative analysis of lymphocyte membrane protein redistribution from fluorescence microscopy.

Full text

Page 1

2004 International Conference on image Processing (ICIP)
QUANTITATIVE ANALYSIS OF LYMPHOCYTE MEMBRANE PROTEIN
REDISTRIBUTION FROM FLUORESCENCE MICROSCOPY
Peter M. Kasson'.2.', Johannes B. Huppa".", Mark M Davis".", and Axel T. Brur~ger',~.'.."
'Biophysics Program, 2Medical Scientist Training Program, 3Department of Molecular and Cellular
Physiology, 4Department of Microbiology and Immunology, 'Stanford Synchrotron Radiation
Laboratory, Stanford University and 6Howard Hughes Medical Institute, Stanford CA 94305
ABSTRACT
The relocalization of plasma membrane proteins is critical
for establishing cellular polarity and regulating cell
signaling. Three-dimensional
microscopy allows the dynamic visualization of proteins
in living cells. We have developed a robust and
automated method to employ fluorescence data acquired
in this manner for quantitative analysis of membrane
protein movements across the cell surface. Our method
utilizes level-set-based surface reconstruction followed by
a maximum likelihood surface registration algorithm for
rigid-body alignment of noisy images. A surtace-walking
technique yields distance maps for the cell surface, which
are then used to measure changes in protein surface
distribution over time. Applying this method to signaling
in T lymphocytes, we have used it to monitor receptor
movements and have validated these results against
previously reported single-particle tracking data.
fluorescence video
1. INTRODUCTION
Relocalization and clustering of signaling proteins in the
plasma membrane drive a diverse array of biological
processes including lymphocyte activation, neuronal
synapse formation, apoptosis, and cell motility. In all of
these cases, redistribution of membrane-associated
signaling proteins has been associated with the initiation
of cellular signaling cascades and the establishment of
cellular polarity. Protein localization is commonly
monitored using fluorescent probes, which can be
visualized in living cells using four-dimensional (x,y,z,t)
microscopy. Pattems of protein localization yield
information about the regulation of these signaling
processes and their failure in disease.
methods for protein relocalization analysis are thus
needed to assess these patterns systematically and to
enable mechanistic analysis of the underlying signaling
networks.
Quantitative
T lymphocyte activation provides a well-studied
example of regulation via membrane protein localization.
During stimulation with antigen, a number of signaling
and adhesion proteins cluster at the interface between the
T cell and the antigen-presenting cell in an organized
fashion. Among the proteins that cluster at this interface
are the T-cell receptor,
histocompatibility complex protein, and the T-cell-
receptor-associated signaling protein CD36.
extensive qualitative study, the precise functions of these
phenomena have yet to he determined, and more
quantitative data will assist in the development and testing
of models for signaling mechanisms.
Previous work on analysis of protein distributions has
included 2-dimensional analyses and those based on
spherical models [l, 21. Three-dimensional work has
been performed on a related problem of object tracking
(labeled spots corresponding to either single or multiple
particles) [3, 41, although these have not explicitly
included surface analyses.
approaches have been used to measure organelle volume
changes [S, 61. Our method utilizes a model-free surface-
based approach to measure
distribution. This allows the capture of global as well as
local distribution changes and increases the accuracy of
measurements on often-irregular cellular surfaces.
Our approach utilizes a segmentation filter previously
reported [2] for identification of membrane voxels. We
subsequently perform level-set surface reconstruction at
each time point [7] and surface registration using a
maximum likelihood approach.
biologically relevant origin for distance measurement is
performed via clustering analysis.
algorithm is then applied to calculate the shortest surface
distance from the origin to each surface point.
Combination of this distance information with membrane
voxel intensities yields a time series of distance-intensity
distributions. In this report, we perform this analysis on
clustering of T-cell-receptor-associated CD35 molecules
and validate the results with reference to T-cell-receptor
single-particle tracking data.
its ligand the major
Despite
Recently, level-set-based
changes in protein
Identification of a
A graph labeling
0-7803-8SS4-3/04/$20.00 02004 IEEE.
2933

Page 2

image, M is the model image, and T is the transform.
P(D I M,T) = p V * D I M )
By conditional probability,
P((T * D)n M )
P(M)
C(T * D), AND M,
4D indeo microscopy data
i
Figure 1. Outline of the analytic process
Shown are the sequential stages in our analytic system
accompanied by volume renderings from progressive analytic
stages of a single timc point from a CD3<-GFP dataset.
2. SURFACE RECONSTRUCTION AND
REGISTRATION
Our segmentation filter, used courtesy of William Moss
[2], offers increased specificity for membrane structures
when compared to a simple edge-detection filter. For
accurate distance mapping, however, we desire a smooth,
continuous surface corresponding to
membrane geometry at each time point. To that end, we
perform level-set-based surface
described by Zhao et al. [7] using the membrane points
defined by our segmentation filter as the target dataset.
We then align the surface thus derived at each time
point against each adjacent time point using a pairwise
registration scheme. We
transformations and perform a global optimization
technique across the 7-dimensional quaternion rotation
and translation space using iterative line searches with
descending step size. Our search metric is a maximum
likelihood criterion: p (D I M,T), where D is the data
the plasma
reconstruction
as
consider rigid-body
p(T* D j M ) =
"OIelsX
We then maximize this probability across transform
space as described above. Using this process, we obtain a
set of rigid-body transformations describing the motion of
the cell membrane over the period of observation.
3. CLUSTERING ANALYSIS
For surfaces isomorphic to a sphere, a continuous two-
dimensional parameterization that preserves surface
distance is impossible. We therefore identify the
biologically relevant reference point
measurement (here the center of the cell-cell interface
where receptor clustering is occurring)
parameterization. We then perform a one-dimensional
surface parameterization to yield a radial distance-
intensity distribution for each time point.
Because the interface
consideration develops very dynamically, ofien changing
with respect to cell shape, accurate identification of the
interface center, and thus the reference point for
parameterization, is challenging. We determine the center
based on k-nearest-neighbors clustering analysis at a
single time point. For each point, we find the 15 nearest
neighbors and assign
a
zintensity(i)G(d, (i)),
ieneipburs
of pol", p
distance from p to i and G is a Parren window function, in
this case a Gaussian with ~ 2 . 5
with the maximal clustering metric is designated the
reference point.
The surface registration information obtained above
is then used to map the reference point to all time points.
We do not repeat the clustering analysis at each time point
because such consideration would bias subsequent
assessment of clustering behavior. However, our analyses
have shown that the use of intensity data for clustering
analysis at a single time point does not impair our ability
to measure clustering phenomena in our lymphocyte
activation system (data not shown).
for distance
prior to
between cells under
clustering
where dp(i) is the surface
metric of
mm. The surface point
2934

Page 3

4. DISTANCE MEASUREMENT AND VELOCITY
ANALYSIS
We perform surface distance parameterization using a
graph labeling strategy for surface walking based on an
upwinding scheme. This algorithm will visit all
connected nodes on the surface and will label them with a
good approximation of the surface distance.
distance-intensity distributions are then obtained by
mapping the surface distance information to the original
membrane points identified by the segmentation filter and
defining the fractional intensity F at each distance x and
Radial
C In/ensity(p,t)
p.dlp)=.t
for time t to be F(x,I) = C Intensity(p,/)
QP
distance at time t is thus d ( t ) = z F ( x , / ) . x . We
assess mean velocity via two measures: AdAt and a S-
point linear fit that has lower noise but is temporally
smoothed.
Q.T
5. EXPERIMENTAL RESULTS
Using the method described above, we have analyzed
receptor movements during T lymphocyte stimulation.
Cell culture and tnicroscopy were performed as previously
described [S, 91. The fluorescently labeled protein under
study was CD3-<, a membrane signaling protein
associated with the T-cell receptor.
distance-intensity
Representative
Figure 2: Radial distribution of intensity.
Plotted are distributions showing the fraction of total intensity present at each surface distance increment from the refcrence point. The
distribution platted in a. is at 2.5 minutes prior tu activation, thc distribution in b. is at the time of activation, and the distribution in c. is
3.5 minutes after activation. Mean intensity-distance values are 10.9.9.9, and 6.1 pm respectively. The blue line shows for reference a
distribution corresponding to uniform intensity across thc surface of a sphere.
a.
b.
Figure 3: CD3zeta receptor velocity in response to T lymphocyte stimulation.
a. shows velocity measurement derived from bulk fluorescence measurements via our method. b. shows a single-particle trace from the
published data [I]. We have calculated velocity for each trace both using the AxIAt velocity dcterminatiun method and using 5-point
moving window linear tits.
2935

Page 4

distributions are shown in Figure 1, and a trace
corresponding to the mean velocity of all receptors across
the surface of a single cell is plotted in Figure 2. As
shown in these figures, our data confirm the qualitatively
observed receplor clustering behavior upon T-cell
activation. To further validate our method, we compared
our mean velocity measurements to previously reported
single-particle tracking data
movements (Figure 2). CD3-6 motions are thought to
approximate those of T-cell receptor, a hypothesis
consistent with our observations. We measure a peak
receptor velocity within 16% of the single-particle
tracking data. The single-particle trace shows a slower
decay in the initial velocity spike than our mean velocity
observations, but the overall behavior is quite consistent.
As our measurements represent a population mean, it is
expected that some of the receptors will have a more
extended velocity profile, particularly those that start
farther away from the cluster center and subsequently
incorporate. The minor discrepancies between our
observations and those from single-particle tracking thus
illustrate how analyses of bulk movements provide
complementary information to single-particle tracking
experiments. Bulk analyses can be performed on a
broader range of experimental systems and yield
population statistics much more readily than laborious
single-particle analyses, although the latter provide a
"gold standard" for movements of individual molecules.
for T-cell receptor
6. CONCLUSIONS
We have presented a novel system for measurement of
membrane protein movements on the cell surface. Our
methods are fully three-dimensional and combine a level-
set surface fitting technique with maximum likelihood-
based surface registration. Applied to membrane receptor
signaling during T lymphocyte activation, this analytic
system measures mean CD3-< velocities, yielding results
that are consistent with previously reported single-particle
tracking experiments. Since our analytic framework is
generally applicable to membrane protein movements, we
anticipate it to be of use to investigations of cell signaling
in other systems such as neuronal synapse formation and
cellular direction sensing.
7. REFERENCES
[I] W. C. Moss, D. I. Irvinc, M. M. Davis. and M. F.
Krummel. "Quantifying signaling-induced reorientation of T cell
receptors during immunological
Proceedings ofthe Nafionul Acudemy of Sciences ofthe United
sf ole.^ ofdmen'co., vol. 99(23), pp. 15024-9,2002,
synapse formation,"
[2] J. Yang, U. Nagavarapu, K. Relloma, M. D. Sjaastad, W. C.
Moss, A. Passaniti, and G. S. Herron, "Telomerized human
microvasculature is functional in vivo," Nutrrre hiorechndoLy .,
vol. 19(3), pp. 219-24,2001.
[3] A. Genovesio, B. Zhang, and J.-C. Olivo-Marin, "Tracking
of multiple tluorescent biological objects in three dimensional
video microscopy," presented at 2003 International Conference
on Image Processing, 2003.
[4] H. Bomfleth, P. Edelmann, D. Zink, T. Cremer, and C.
Cremer, "Quantitative motion analysis of subchromosomal foci
in living cells using four-dimensional microscopy," Biopltys J.
vol. 77. pp. 2871-86, 1999.
[5] W. Tvarusko, M. Bentele, T. Misteli, R. Rudolf, C.
Kaether, D. L. Spector, H. H. Gerdes, and R. Eils, "Time-
resolved analysis and visualization of dynamic processes in
living cells," Proc Null Acud Sci U S
1999.
A , vol. 96, pp. 7950-5,
[6] D. Gerlich, J. Beaudouin, M. Gebherd, J. Ellenberg. R. Eils,
and G. C. R. C. ~H.
G. Intelligent Biainformatics Systems
Department, "Four-dimensional imaging and quantitative
reconstruction to analyse complex spatiotemponl processes in
live cells," Nufirre ce// biologv., vol. 3(9), pp. 852-5,2001
[7] H. K. Zhdo, S. Osher, B. Merriman, and M. Kang, "Implicit
and nonparametric shape reconstmction from unorganized data
using a variational levcl set method," Compirter Vkion and
huge Under.sfunding, vol. 80, pp. 295-3 14,2000.
[8] L. I. Ehrlich, P. I. Ebert, M. F. Kmmmel. A. Weiss, and M.
M. Davis, '"Dynamics of p561ck translocation to the T cell
immunological synapse following agonist and antagonist
stimulation," Immenity, vol. 17, pp. 809-22, 2002.
[9] J. B. Huppa, M. Gleimer, C. Sumen, M. M. Davis, D.
Stanford University School of Medicine, Immunology, and B. C.
B. C. D. S. C. U. S. A. Howard Hughes Medical Institute,
"Continuous T cell receptor signaling required for synapse
maintenance and full effector potential," Nulure immanolop.,
vol. 4(8), pp. 749-55,2003.
0. M.
2936