Mapping protein dynamics in living cells i fl fl t ti & l tivia fluorescence fluctuation & correlation
analysis in space, time & reciprocal space
Paul W. Wiseman McGill UniversityDept. of Physics, Dept. of Chemistry
SHG I i
Wiseman Lab Biophysical Research McGillWiseman Lab Biophysical Research McGillSHG ImagingHeart Collagen
THG Imaging ofMalaria Hemozoin
Guiding Neurons on Patterned Protein Gradients Guiding Neurons on Patterned Protein Gradients
IKVAV peptide Gradient
Guided
With D S ti C t ti U d M t l
Chick Dorsal RootGanglion onIKVAV peptide gradient
With Dr. Santiago Costantino U de MontrealDr. Tim Kennedy Montreal Neurological InstituteBelisle et al. Lab on a Chip in press
Goal: Measure the Biomolecular DanceGoal: Measure the Biomolecular Dance
170 mWhat are rules of the game withinthe Living Cell?the Living Cell?
Physico-chemical Understanding
New Techniques are needed…
Paxillin dsRed (red) &
i) Spatio-temporal Image Correlation Spectroscopy (STICS) Paxillin-dsRed (red) &
-actinin GFP (green)in CHO CellTIRF Mi
ii) k-space Image Correlation Spectroscopy (kICS) TIRF Microscopy
Total time = 50 min t =15 s
Spectroscopy (kICS)
How do We Observe the Dance? How do We Observe the Dance? FluorescenceFluorescenceG Fl t P t i ti f i t t t tGreen Fluorescent Protein genetic fusion construct to create a Protein with a fluorescent protein attachedExcite GFP with focused laser light…Collect Fluorescence photons
Emitting
g p
g
~ 3 nmPhotobleachedex 488 nm
em 509 nmFrom http://zeiss-campus magnet fsu edu/articles/probes/fpintroduction htmlFrom http://zeiss campus.magnet.fsu.edu/articles/probes/fpintroduction.html
Many Color Variants
Dr R Campbell U of Alberta
See: Fluorescent Proteins By Dr. R. Campbell (Scolarpedia)
Dr. R. Campbell U of Alberta
Focal Adhesion Proteins…Molecular ClutchFocal Adhesion Proteins…Molecular Clutch
CellMembrane5 nm thick
Vicente-Manzanares, M. et al. JCS (2005)
Fluorescence Correlation SpectroscopyFluorescence Correlation Spectroscopy
BeamFocus
4
Intensity FluctuationsLaser Focus Molecular Dynamics
0
2
i>
Intensity Fluctuations~ 1 um3
i-2
<
i(t)=i(t) –<i>Fluorophores excited in focus Number in the
i(t)
f-1 -0.5 0 0.5 1
-4 Fig. 1 Overview of Fluctuation Spectroscopy
i(t) i(t) <i>excited in focus Number in the Focus fluctuates
f
100f 780 920 l 82MH t
LaserLaser--Scanning Microscopy…ImagingScanning Microscopy…Imaging
TiSapph. laserM1
100fs, 780-920nm pulse 82MHz rep-rate
PMT 3 PMT 2 PMT 1Sample
FilterEm. Filters
M4Dichroic i
Filter
Dichroic mirrors
+L1 pinhole+L2M2 M3
mirrorDichroic mirrors
+L1 pinhole+L2M2 M3t=0 t=1 t=2 t=3 t=4
Evanescent Wave Fluorescence ExcitationEvanescent Wave Fluorescence ExcitationTotal Internal Reflection Fluorescence Microscopy (TIRF
Microscopy)
Widefield TIRFZ D th f fi ldZ-Depth of field~ 100 nm at theboundary
http://www.microscopyu.com/articles/fluorescence/tirf/tirfintro.html
Temporal Image Correlation Spectroscopy (TICS)Temporal Image Correlation Spectroscopy (TICS)
Temporal Autocorrelationof i(x,y,t) = i(x,y,t) - <i>Through Time Series Decay
t=0
Temporal ACF
0.16
0.18
DecaySlowTransport
DynamicsDiffusion Coeff.
t=1
r 11(
0,0,)
0.10
0.12
0.14 &Flow Speeds
Off t
t=2
(s)0 100 200 300 400 500
0.06
0.08 OffsetImmobilePopulation
t=3
t t11 i i
)t ,yi(x,t)y,i(x, ,0 ,0r
t=4
Srivastava and Petersen Methods Cell Sci. 18, 47-54 (1996)t=n
Overview for TalkOverview for Talk
Spatio-Temporal Image Correlation Spectroscopy
k Reciprocal Space Image Correlation p p g
Spectroscopy (kICS)
SpatioSpatio--Temporal Image Correlation SpectroscopyTemporal Image Correlation Spectroscopy
Spatial Correlation as a functionOf Time r11(,,)
t=0
t=1
t=2
tt11 ii
) t ,y ,i(x t)y, i(x, , ,r
t=3
t t Hebert et al. Biophys. J. 88-3601 (2005)t=4
r11(,,)Brownian Flow
t=nDiffusionPeak
FlowPeak Flow
Vector
Vector Maps of Vector Maps of --actinin MEF Cellactinin MEF Cell
TIRF Microscopy Time 100 s with Images sampled at 0.1 HzDr. Claire Brown and Ben Hebert
Vector Maps of Vector Maps of --actininactinin
-actinin/EGFP in MEF Cell actinin/EGFP in MEF CellTIRF Microscopy
r()
Accelerated40 ti f t
ImmobileFiltered
40 times faster thanReal-time
3.4 μm
τ = 0 s 200 s
Correcting for Immobile or Slowly Moving SpeciesCorrecting for Immobile or Slowly Moving Species
Fourier Filtering (Full Time Stack)Filter pixel stacks in temporal frequency domainWith Heaviside Function before STICS analysisR I bil F t
t=0
MAV Removes Immobile Features t=1
MAV3
t=2
t=3
Moving Average Filter Odd # Framest=4 Moving Average Filter Odd # FramesApplies Immobile Filtering over ShorterTime Window
t=n
Space Time Correlation for Mobile FractionSpace Time Correlation for Mobile Fraction
• Filter Immobile in Frequency Space → r(,,) Mobile
FractionFraction
2D Grey Scale Contour MapS ti l C l ti F ti ( )
• Simulation: 90% immobile
Spatial Correlation Function()
r(,,) vs. 90% immobile, 10% with flow + diffusion in x and y
(,, )For Mobile Fraction
Vector Maps of Vector Maps of --actininactinin
9 μm/min
0 μm/min0 μm/min5 μm
0.7
0.8
min
)r()Inverse relationship b/w
retrograde flow & protrusion speed
0.3
0.4
0.5
0.6ra
de fl
ow (
m/m
r()Ff Fourier Filtered
protrusion speed(similar to actin)
0.0 0.2 0.4 0.6 0.8 1.0 1.20.0
0.1
0.2
Ret
rog
Protrusion rate (m/min)0 s 2.8 s1.5 s
()Ff Fourier Filtered
CoCo--Transport of Adhesion Molecules & ActinTransport of Adhesion Molecules & Actin
integrinA F actin
D I
GB illi
talin
5 m1 m/min
ti
actin
GB paxillin
vinculinE J
actin
actin
C HFAK actinC HFAK actin
θVact
Compare everyVector in Ch1 & Ch2For magnitude &
Vgfp
gDirection correlation
Correlated Transport PropertiesCorrelated Transport Properties
1.0
1.2 relative magnitude directional correlation
3T3 Fibroblasts
0 6
0.8
b. u
nits
)
0.4
0.6
Scor
e (a
rb
integrin paxillin FAK talin vinculin a-actinin0.0
0.2
αintegrin paxillin FAK talin vinculin α-actining p
αintegrin paxillin FAK talin vinculin α actinin
Intermediate
~ 70% level
very low magnitude &
low directional correlation
high magnitude &
directional correlation
pax-FAK-talin-vinculin are part of a distinct linkage complex
~ 70% levellow directional correlation directional correlation
Emerging picture of linkageEmerging picture of linkage
actinin
Brown et al. JCS (2006) 119: 5204-5214Faculty of 1000 Selection Jan 2007
-actinin
integrin ECM
actin
g ECM
See Also Ke Hu, et al.Science 5 January 2007: 111-115
Microscopic basis for similar flow fieldsMicroscopic basis for similar flow fields
Always co-localized1 Always co localized
Same speed and direction,not co-localized
2
Same direction,3 different speed
Transient binding/unbinding
3
4and co-transport
4
Mapping the Dance Partners…Mapping the Dance Partners…ChallengeTrying to Resolve the Molecular Dance Partners
4
G i B2
Gaussian Beam Focus
-2
0
1 m3
-1 -0.5 0 0.5 1-4
~ 1 m3
Optical Resolution ~ /2 Macromolecules ~ /50
= Wavelength ~ colour of the light
vs.
g gCross-correlation
Leafs Habsvs.
STICCS: two color crossSTICCS: two color cross--correlationcorrelation
ttii
tyxityxir
21
2112
),,(),,(),,(
tt 21
x y = 1Channel 1 Channel 2
x y = 1Channel 1 Channel 2y = 2
Channel 1 Channel 2
y = 2Channel 1 Channel 2
y = 0Channel 1 Channel 2
y = 0Channel 1 Channel 2
x
t
x
t t
x 2
t
x 2x
y
t
xy
t
No CrossNo Cross--Correlation between integrin & actinCorrelation between integrin & actin
5-integrinEGFP
mRFP-actin CrossCorrelationEGFP
0.01 m/min
0.39 m/min
= 0 14 m/min = 0 18 m/min = 0.14 m/min = 0.18 m/min
Positive ControlPositive ControlActin-GFP, actin-mRFP MEF cell
Cross-l ti
“RFP”A t l ti
GFPA t l ti correlationAutocorrelationAutocorrelation
0.0 m/min
0.55m/min
Strong, consistent GFP, RFP autocorrelations,and cross-correlation
STICCS CHO STICCS CHO 6611‐‐gfpgfp + + PaxillinPaxillin‐‐mcherrymcherry
Unfiltered Fourier filtered
48 μm 48 μm
Video length: 100 seconds
STICCS CHO STICCS CHO 6611--gfpgfp + + PaxillinPaxillin--mcherrymcherry
Auto 16611
Auto 2PaxillinPaxillin
Cross1 x 26611 PaxillinPaxillin 1 x 2
Fourier Filter Frames 51-100 Pixel size 0.21 m
70
80 hist.
60 hist.
60 hist.
0.00 m/min
2.61 m/min
Fourier Filter Frames 51 100
Median 0.81 m/min 0.99 m/min 0.90 m/min
t 2 s
30
40
50
60
70
# of
val
ues
median:0.806av:0.9sd:0.56
20
30
40
50
# of
val
ues
median:0.991av:1.06sd:0.555
20
30
40
50
# of
val
ues
median:0.895av:0.974sd:0.583
0 1 2 3 40
10
20
|v| (m/min)
0 1 2 3 4
0
10
|v| (m/min)
0 1 2 3 4
0
10
|v| (m/min)
Averaged over the Cell
Neuroscience Applications: Growth ConesNeuroscience Applications: Growth Cones
Pathfinding of Chick DRG Neuron Axon Growth Cones
Cytoskeletal Dynamics (both Actin and Microtubules)
3
3.5
4
4.5No NGF50ng/ml NGF
e of
F-a
ctin *
1.5
2
2.5
3
rade
flow
rate
0
0.5
1
Ret
rogr
N=16 eachUtrophinmarker of f actin Disrupt Actin with
blebbistatin<v> = 3.5 m/min+ 50 ng/mL NGFOvernight
Collaboration with Dr. Tadayuki ShimadaProf. Alyson Fournier MNI, McGill
Neuroscience Applications: Growth ConesNeuroscience Applications: Growth Cones
Pathfinding of Chick DRG Neuron Axon Growth Cones
Cytoskeletal Dynamics (both Actin and Microtubules)
EB3 (3 4 images/s)EB3 (3.4 images/s)marker of microtubule tips <v> = 11.4 m/min
We are looking at turning gradients as NGF is applied on id f th th
Collaboration with Dr. Tadayuki ShimadaProf. Alyson Fournier MNI, McGill
one side of the growth cone
Overview for TutorialOverview for Tutorial
Spatio-Temporal Image Correlation Spectroscopy
k Reciprocal Space Image Correlation p p g
Spectroscopy (kICS)
Quantum Dots as Biomolecular LabelsQuantum Dots as Biomolecular LabelsSemiconductor Quantum Dots…Luminescent Nanoparticles
Smaller QD…bl i i
ZnS CapCdSe corebluer emission
Larger QD…reder emissionreder emission
Photostable…Different sizes…
Different Colors, Track proteins
Fluctuation Spectroscopy on QDsFluctuation Spectroscopy on QDs
(CdSe)ZnS (CdSe)ZnS –– Streptavidin (QD605) Streptavidin (QD605) TIRF Illumination CCD TIRF Illumination CCD DetectionDetectionDetection Detection 50ms Integration Time 50ms Integration Time 2000 Frames2000 FramesSee Bachir et al. JAP 99 (2006)Perturbs fluctuation measurements
Single Dot i(t) trace
Nirmal et al. Nature(London) (1996)
Fluctuation Spectroscopy on QDsFluctuation Spectroscopy on QDs
Diffusion only
QD Blinking
QD Blinkingand diffusion
kICS: Reciprocal Space Time CorrelationkICS: Reciprocal Space Time CorrelationImage 2D FT timeImageseries of images correlation
kxx kx
20
40
60
20
40
60
kyy ky
20 40 60 80 100 120
80
100
120
20 40 60 80 100 120
80
100
120t t
Include (t)
Spatial Fluctuations Reciprocal Space2D FFT
Include (t)(1or 0)
See Kolin et al. Biophysical Journal 91 3061-3075 (2006)x y kx ky
Some New Things: kICS kSome New Things: kICS k--space ICSspace ICSk iPoint source
fluorescenceemitters
Imageseries
2D Fouriertransformof images
k-space timecorrelationfunction
20
20
20 40 60 80 100 120
40
60
80
100
120
20 40 60 80 100 120
40
60
80
100
120
constantalinstrumentqt),(*)I( q t),i( rrr
N
io
1
2o
2o
8|| - iexp (t)
2qI t),(i~ krkk
densitynumber cmicroscopi t),(Function SpreadPoint )I(
constant alinstrument q
rr
i 1 82
)t,(i~t),(i~ ),r( * kkk
See Kolin et al. Biophysical Journal 91 3061-3075 (2006)
yp),(funtionn correlatio timespace-k
kICS…separtes photophysics & transportkICS…separtes photophysics & transport
200
InterceptPhoto-Physics
SlopeFor a given :
Single -20
0
-5Circularlyaverage
Single
0
20 -15
-10and logtransform
Noise-20 0 20 0 200 400
15
0.4 0.8 1.2-5
5
kICSkICS: Reciprocal Space Time Correlation: Reciprocal Space Time Correlation
kICSkICS: Reciprocal Space Time Correlation: Reciprocal Space Time CorrelationDetermine the slopes for each value of τ, plot them as a function of τ:
Intercept
0 07
-0.06
-0.05
SlopeD independentof o
-0.09
-0.08
-0.07
No non-linearCurve fitting
-0.11
-0.1
g
Do Not need to Calibrate Beam Size!
2 4 6 8
-0.12
0 01
0
2
4
z
2 4 6 8-0.01
00.01
Residuals-1 -0.5 0 0.5 1-4
-2
Measure radii
kICS: Reciprocal Space Time CorrelationkICS: Reciprocal Space Time Correlation
Computer Simulations
Simlulation Parameters128x128 pixels 500 frames Durisic et al. Biophys. J. 128x128 pixels 500 frames0.1 s/frame, 0.1m/pix, 0.4 m PSF, D=0.1 m2/s, mon=2, moff=1.5
p y93-1338 (2007)
QD Labeled CD73 GPI Anchored Protein In CellsQD Labeled CD73 GPI Anchored Protein In Cells
GPI Lipid Anchored Proteinswww.uoguelph.ca/~fsharom/research/gpi.html
IMR-90 Fibroblast CellsCD73 GPI Anchored Proteins Tagged with QDsggImaged at Video Rate (30 fps) (Chris Lagerholm UNC Chapel Hill)
QD Blinking Does Affect Temporal Image QD Blinking Does Affect Temporal Image Correlation…kICS eliminates the problemCorrelation…kICS eliminates the problem
2 1
6 0.00TICS of QD tagged
DkICS = 0.088 ± 0.008 µm2s-1DTICS = 0.109 ± 0.003 µm2s-1
r 11(0
,0,
)
4
[
m2 s-1
]
-0.04
-0.02QD taggedGPI Anchored Receptors
kICS of QD taggedr
2
0 0 0 2 0 4 0 6 0 8
-D
-0.06
QD taggedGPI Anchored Receptors
(s)0 1 2 3
(s)0.0 0.2 0.4 0.6 0.8
Durisic et al Biophys J
Correlation Volume
QD T t
Durisic et al. Biophys. J. 93-1338 (2007)Quantum dot labeledGPI Anchored ReceptorsQD Transport
FluctuationsQD BlinkingFluctuations
GPI Anchored Receptors
kICSkICS separates photoseparates photo--physics & transportphysics & transport
Different particlenumbersnumbers
Different diffusioncoefficientscoefficients
Different blinkingDifferent blinkingor bleaching
Kolin, et al., Biophys. J. 2006 91: 3061-3075.
kICS can measure Changes in BlinkingkICS can measure Changes in Blinking
Th t C l ti F ti
1
Theta Correlation Function
0.6
0.8
(t+)
GFP-adhesion proteinCHO ll
0.4
(t)
( CHO cell
GFP mono-exponential0.2 bleaching
50100
50100 t (s) (s)
Kolin, et al., Biophys. J. 2006 91: 3061-3075.
QD labelling of T cell receptorsQD labelling of T cell receptors
Quantum dot
Collaboration with
•Michael Edidin JHU Quantum dot
Streptavidin• Jon Schneck JHU
MHC monomerwith biotin
T cell receptorCD8 co-receptor
T cell receptor
Drawing by David S. Goodsell
T cell membrane
Qualitative DifferencesQualitative Differences
• Change in blinking of QDs?• Cell size• Change in distribution of QDs?
10
20
30
• Change in distribution of QDs?• Can these be quantified?
5
10
40
50
60
7015
20
25
30
35 20 40 60 80 100
70
80
90 5 µm
Activated cell – Day 4TCRs are more clustered
10 20 30 40 50
40
45
50
55
5 µm
Naïve cell Day 0Epi-fluorescence captured on an EM-CCDCame aCameraResults are for 300-500 image time series averaged over many cells (~20 cells/day)
kICS detects blinking changes in T CellskICS detects blinking changes in T Cells1 01.0 Activated (Day 4)
Theta Ratio = ’
0.8
Activated)>
Theta Ratio ~0.9
0.6
Naive
(t)
(t+
NaiveActivated
0 4
<
Theta Ratio ~0.4
NaiveActivated Clustered QD/TCRsBlink Less Frequently
0.4
kICS detects this(simulations support this)
1 10 (s)Yu & Van Orden PRL 2006 & Scheibner et al. Nat. Phys. 2007
ConclusionsConclusions
Spatio-Temporal Image Correlation Spectroscopy
Transport maps of FP labeled proteins in cellsp p p
k Reciprocal Space Image Correlation
Spectroscopy (kICS)
Separates photo-physics and transport fluctuationsp p p y p
Data mining the images (there’s gold in those hills
(fl t ti ))(fluctuations))
AcknowledgementsAcknowledgements
• Michael Edidin JHU• Allan “Rick” Horwitz UVaSTICS Molecular Clutch kICS
• Jon Schneck JHU
• The Schneck lab
• Claire Brown UVa & McGill University
• Ben Hebert UVa
• David Kolin
• David Ronis
• David Kolin
• Tim ToplakJohns Hopkins
U i it • David Ronis
• Nela Durisic
Ch i L h l UNC
• Tim Toplak
• Tadayuki Shimada (MNI)
l ( )
University
• Chris Lagerholm UNC
•Jeremy
• Alyson Fournier (MNI)
Schwartzentruber
Thanks to the Group! Thanks to the Group!
Ben HebertDavid KolinNela Durisic David KolinNela Durisic
Thanks to Prof Rick Horwitz UVa & Dr. Claire Brown McGill
Thanks to the Group! Thanks to the Group!
Tim Toplak
& Jeremy Schwartzentruber, Dominique GuilletSebastien Cote, Benoit Vallaincourt
Correlated Transport Depends on [ECM]Correlated Transport Depends on [ECM]
1.0Directional Correlation
B
P illiLow = 2 g/ml
EGFP Protein mRFP-Actin
2 g/ml Directional
arb.
uni
ts)
0.6
0.8
Talin
PaxillinLow 2 g/mlHigh = 5 g/mlFibronectin
Talin
5 g/ml
Correlation
Sco
re (a
0.2
0.45 g/ml
FibronectinTalin
0.0
Low High HighLow5 g/ml
Fibronectin [FN]
Paxillin
Quantum dot blinkingQuantum dot blinking
Intensity1
ton
toffy
time
on off 0off
On and off time distributions have “power-law” decays• Fractal• Occur on all timescales• Occur on all timescales• ton and toff have power law distribution
mon depends on:
Log(Probability) slope:mon
mon depends on:• Core• Shell• Surface functionalization
E it ti i t it
Log(ton)
• Excitation intensity• Ligands in solution (e.g., pH)• Manufacturer/Batch of QDs
“ l ” bli k
Three blinking simulationsThree blinking simulations
Intensity
on off
“slow” blinker
0
1
time
on off
“fast” blinker
1Intensity
on off0
1
timetetramer “slow” blinking4
Intensity
0time
0
0
kICS can measure blinking within clusterskICS can measure blinking within clusters
y = -0.1119x - 0.0432R2 = 0.9931
-0.2
-0.10 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
) Computer
0 5
-0.4
-0.3
t)the
ta(t+
tau)
>
slow blinkingtetramers
f t bli ki slow blinking
ComputerSimulations
PSF I t t d
-0.7
-0.6
-0.5
ln(<
thet
a(t fast blinking
monomersslow blinkingmonomers
PSF Integrated
0
2
4
y = -0.2631x - 0.507R2 = 0.9873
-0.9
-0.8
ln(tau)
-1 -0.5 0 0.5 1-4
-2
ln(tau)
Dot-taggedproteins
aggregategg g
Degree of aggregationDegree of aggregationAdd T C ll G th F t D 6 & 10 ti ti hAdd T Cell Growth Factor Days 6 & 10…activation echo
1000
ion
100Aggr
egat
Deg
ree
of
10
D
Naive 1 2 3 4 5 6 7 8 9 10 11 12
Days after activation
Diffusion coefficientDiffusion coefficient
0.01
m2 /s
)oe
ffici
ent (
iffus
ion
co
0 00
Di
Naive 1 2 3 4 5 6 7 8 9 10 11 120.00
Days after activation
QD clusters blink differently than single QDsQD clusters blink differently than single QDs
Single QDSingle QD
QDCluster
As well Scheibner et al. Nature Physics 3:106–110, 2007