Transcript Multiplexed Fluorescence Lifetime Imaging of Cell

Development and application of an
automated FLIM multiwell plate reader for
high content analysis
Douglas Kelly1,2,3, Anca Margineanu2, Sean Warren1,2, Mesayamas
Kongsema3, Jia Chan4, Eric W.-F. Lam3, Matilda Katan4, Chris
Dunsby2, Paul M. W. French2
1Institute
of Chemical Biology, Imperial College London, UK
2Photonics Group, Imperial College London, UK
3Department of Surgery and Cancer, Imperial College London, UK
4Department of Structural and Molecular Biology, University College London
Outline
• Introduction
– Fluorescence Lifetime Imaging
– Förster Resonance Energy Transfer
• Data acquisition
– Schematic of instrument
– Overview of software
• Applications
– Test experiments: dyes, FRET constructs
– FOXM1-SUMOylation in response to doxorubicin
– Rac1 biosensor response to CdGAP overexpression
– Rassf-family protein interaction screen
Fluorescence Lifetime Imaging Microscopy
Fluorescence lifetime imaging may be considered to be the construction of images
with contrast originating from differences in the characteristic decay time of
fluorescence following excitation.
•
This characteristic lifetime depends on radiative and non-radiative decay rates kr
and knr as the fluorophore returns to the ground state.
intensity
•

intensity
t
t
1

knr  kr
Fluorescence Lifetime Imaging Microscopy
•
FLIM is most commonly performed
using time correlated single photon
counting on a laser scanning confocal
microscope.
•
In TCSPC, the relative delay between
photon arrival and excitation pulses is
used to build up a histogram of photon
arrival times, which can be fitted to a
decay model to yield fluorescence
lifetimes.
•
This method gives best S/N per photon,
but is slow. This is linked to a number
of factors: point scan speed, detector
dead time, limits on excitation power
related to sample photodamage, and the
inherent limit that the probability of
detecting a photon following an
excitation pulse must be << 1 to avoid
miscounting.
Fluorescence Lifetime Imaging Microscopy
•
Increasing throughput of FLIM microscopy mitigates for biological noise and permits
screening/dose response experiments to be carried out easily.
•
Employing a gated optical intensifier (GOI) to implement wide field time gated FLIM,
all pixels are interrogated in parallel. This gives a higher S/N per unit acquisition
time[1], and makes increased throughput viable.
•
The GOI amplifies signal when in the ‘on’ state, occluding signal in the ‘off’ state.
•
Acting like an “ultrafast shutter”, each gate provides a snapshot of the decay at a
given delay after excitation, allowing reconstruction of the decay and fitting of
fluorescence lifetimes.
[1] Talbot, C. B. et al. J. Biophoton. 1, 514–521 (2008).
Förster Resonance Energy Transfer
1
E
, R0 ~ nm
6
1  R R0 
FRET efficiency E
1
•
FRET is a phenomenon whereby energy is
transferred non-radiatively between donor
and acceptor fluorophores when certain
spectral, orientation and separation criteria
are met.
•
FRET typically occurs over nanometres,
making it a useful readout for interactions on
biologically relevant scales.
•
Since FRETting donors have an additional
non-radiative pathway accessible from the
excited state, the fluorescence lifetime is
shorter than for non-FRETting counterparts:
FRET can be read out by FLIM.
0.8
0.6
0.4
0.2
0
0
0.5
1
1.5
R/R0
 DA
E  1
D
2
2.5
Förster Resonance Energy Transfer
•
In biological contexts, FLIM-FRET can read
out signalling events or environment
changes reported by genetically encoded
FRET biosensors. The presence of a
signalling molecule prompts a
conformational change, and hence a
change in FRET.
•
Alternatively, two different proteins of
interest can be modified to be expressed
fused to fluorescent proteins. Here, an
increase in FRET reflects an interaction
between the proteins of interest.
•
In either case, it can be informative to fit a
biexponential model, which allows both
FRET efficiency and FRETting donor to be
extracted from data.
Intramolecular FRET:
Single chain biosensors
Intermolecular FRET:
Separate labelled proteins


I (t )  I 0  non  FRETting exp  t


non  FRETting 



...  I 0  FRETting exp  t


FRETting 

FLIM plate reader: instrument schematic
Spinning diffuser
Fianium SC400-6
supercontinuum
source
Olympus IX-81 chassis:
automated nosepiece, AF,
filter selection and stage
ND/Spectral filter wheels
Sample
Hamamatsu ORCA
ER-II CCD
Kentech
HRI GOI
FLIM plate reader: control software
• Software written in-house for flexibility and
expandability
• User can construct FLIM experiments with
sequences of arbitrarily-ordered functions,
including:
•
Movement between FOVs
•
Z-stacks
•
Time course imaging
•
Phase contrast imaging
•
Change of filters for FLIM in different spectral
windows
• Software implements hardware autofocus to
ensure high image quality across multiwell plates
• Simple liquid handling is supported
• Cell pre-find reduces experiment time and
sample exposure...
FLIM plate reader: control software
• Cell pre-find eliminates time wasted imaging non-fluorescent cells when fluorescent
probes are introduced by transient transfection, or empty regions when stably transfected
cell lines are seeded sparsely.
• Low magnification fluorescence or phase contrast images are used to find objects of
interest using internal functions or by calling CellProfiler.
• Positions are mapped such that cells are located in the centre of eventual high
magnification FLIM fields of view.
(a)
(b)
(c)
(a) Set up for low magnification
phase contrast prefind showing
raw image, variance image,
and eventual cell identification
following variance, area and
intensity thresholding.
(b) Search pattern showing low
magnification search FOVs in
relation to the well.
(c) High magnification FLIM
camera FOV resulting from
prefind.
FLIM plate reader: dye calibration experiment
• Simplest possible experiment
1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
0.95 0.25 0.75 0.05 0.50 0.95 0.25 0.75 0.05 0.50
• Fit globally to return lifetime values of pure
dyes and ratio in each well
0.75 0.05 0.50 0.95 0.25 0.75 0.05 0.50 0.95 0.25
0.90 0.10 0.60 0.00 0.40 0.90 0.10 0.60 0.00 0.40
0.60 0.00 0.40 0.90 0.10 0.60 0.00 0.40 0.90 0.10
• Results match TCPSC-measured values
0.50 0.95 0.25 0.75 0.05 0.50 0.95 0.25 0.75 0.05
• Acquisition in 700s
0.40 0.90 0.10 0.60 0.00 0.40 0.90 0.10 0.60 0.00
Mix ratio R6G:RB
• Two spectrally indistinguishable dyes mixed in
different ratios
1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
Dye
Globally
fitted
lifetime (ns)
TCSPC
measured
lifetime (ns)
Rhodamine B
1.7
1.7±0.05
Rhodamine 6G
3.9
3.9±0.1
Contribution of long lifetime
  I  exp  t

I (t )  I 0  R 6G exp  t



0 RB
  RB 
  R 6G 
Typical experiment pipeline: FRET plate
• To illustrate the experiment pipeline and the performance of the plate reader with more
typical samples, we prepare a plate to simulate intermolecular FRET using a donor
fluorophore (mCerulean) and a donor-acceptor construct (mCerulean-5-mVenus)
developed in the Vogel lab[2]
• Cells are sufficiently sparsely seeded that cell prefind is required, c.f. typical samples. We
seek 8 “acceptable” fields of view per well.
1.00
0.95 0.90 0.75 0.60 0.50 0.40 0.25 0.10 0.05 0.00
1.00
0.95 0.90 0.75 0.60 0.50 0.40 0.25 0.10 0.05 0.00
1.00
0.95 0.90 0.75 0.60 0.50 0.40 0.25 0.10 0.05 0.00
1.00
0.95 0.90 0.75 0.60 0.50 0.40 0.25 0.10 0.05 0.00
1.00
0.95 0.90 0.75 0.60 0.50 0.40 0.25 0.10 0.05 0.00
1.00
0.95 0.90 0.75 0.60 0.50 0.40 0.25 0.10 0.05 0.00
1.00
0.95 0.90 0.75 0.60 0.50 0.40 0.25 0.10 0.05 0.00
Ratio mCerulean to mCerulean-5-mVenus
by plasmid weight
[2] Koushik, S. V. et al., Biophysical Journal 91, L99–L101 (2006).
Typical experiment pipeline: FRET plate
• Large areas of the plate are screened with 10x magnification objective
• Low magnification fields are interrogated for suitable cells
• An image of the entire plate is accumulated during prefind – particularly useful as an
overview if multiple transfection conditions are being trialled in a single plate.
Prefind FOV
Eventual FLIM FOV
Entire plate prefind image
Single well
Final, centred FLIM FOV
Typical experiment pipeline: FRET plate
• Following prefind, an acquisition sequence is set up to reflect a typical experiment
• We acquire FLIM images at donor-wavelength, a single acceptor fluorescence image and
a phase contrast image for each saved field of view.
FLIM sequence
Donor integrated
intensity
Acceptor intensity
Phase contrast
Typical experiment pipeline: analysis and results
• Apply global double exponential fit with
lifetimes fixed to values obtained from each
construct alone.
• Fitted results show mean lifetime
decreasing with increasing FRET
700
Acceptor intensity, DN
• Acceptor intensity imaging shows that
expression of FRETting construct follows
expected pattern
600
500
400
300
200
100
0
• FRETting contribution (biexponential fit)
follows expected trend
0.8
0.6
0.4
0.2
0
1
0.8
0.6
0.4
0.2
0
1.2
Short lifetime contribution
1500 ps
Mean fitted lifetime, ps
3000 ps
1
1
0.8
0.6
0.4
0.2
0
-0.2
Ratio mCer:C5V
Typical experiment pipeline: FRET plate
• Excluding sample preparation time which varies depending on application, users can
expect to have quantitative FRET results within three hours of starting the experiment.
• In this case, 8 FOV per well = 56 FOV per condition > 56 cells per condition.
?
25
min
120
min
Imaging:
Sample
preparation
Cell prefind
• FLIM
• Acceptor
channel
• Phase contrast
>1
min
Controls:
• Background
• IRF/reference
dye
>1
min
Analysis:
• Segmentation
• Fitting
• Interpretation
Unsupervised
~150
min
FOXM1-SUMOylation: doxorubicin response
• FOXM1 is a transcription factor, aberrant expression of which is linked to tumorigenesis,
angiogensis and metastasis across a range of malignancies.
• Biochemical experiments undertaken in parallel to our FLIM-FRET studies show that small
ubiquitin-like modifier (SUMO1) is covalently attached to FOXM1 in response to DNA
damage in breast cancer cell lines.
• Doxorubicin is an anthracycline commonly used to treat breast cancer. Its mode of action
is not completely understood; however it has been shown to intercalate into DNA and
promote double strand breaks (DSB), interrupting cell cycle progression and eventually
contributing to cell death.
(a)
(b)
S
FOXM1
WT
S
FOXM1
mut
(a) Cartoon representation of fluorescent fusion proteins FOXM1WT-eGFP, SUMO1-tRFP and
negative control, non-SUMOylatable mutant FOXM1mut-eGFP. (b) Plate layout.
• We seek to probe interaction of FOXM1 and SUMO1 in response to doxorubicin treatment
of MCF7 cells using fluorescent fusion proteins, reading out the noisy FRET data on the
FLIM plate reader.
FOXM1-SUMOylation: doxorubicin response
• Of all conditions, only FOXM1-eGFP + SUMO1-tRFP shows a response to doxorubicin
treatment. FOXM1-mut is a mutated form of FOXM1 lacking SUMOylation sites; the basal
decrease in lifetime exhibited by this mutant and the offset in lifetime between conditions
at zero treatment time suggests an additional interaction mechanism contributing to the
FRET signal – we hypothesise that this may be linked to the presence of SUMOinteracting motifs (SIMs) on FOXM1.
(a) Map showing plating strategy. (b) Gallery view illustrating heterogeneity in cell response.
(c) Map showing single fitted lifetime. (d) Plot showing fitted lifetime response of FOXM1eGFP + SUMO1-tRFP cells to 0.1 µM doxorubicin treatment time course (Dunnett’s test).
Rac1 activation by cdGAP
• In keratinocytes, the dysregulation of cell-cell junctions has implications in the study of
tumour metastasis.
• Activity of small GTPases such as Rac1 is modulated by GTPase activating proteins
(GAPs) and guanine-nucleotide exchange factors (GEFs).
• Our collaborators are interested in spatio-temporal activation of Rac1 in response to
overexpression of cdGAP, with a view to examining the role played by cdGAP in
maintenance of adherens junctions.
• Rac1 activation is probed using a single chain biosensor based on those developed in the
Matsuda lab[3] – upon activation, Rac1 can bind PAK, causing a conformational change
and an increase in FRET.
GEF
mRFP
PAK
Rac1
GDP
Rac1
GFP
GTP
GAP
PAK
[3] Itoh, R. E. et al. Molecular and cellular biology 22, 6582–6591 (2002).
Rac1 activation by cdGAP
• Data acquired on the FLIM plate reader show a decrease in Rac1 activation upon
overexpression of cdGAP, supporting findings of bulk biochemical experiments.
• The ability to image multiple spectral channels allows confirmation that cdGAP is
overexpressed: immunofluorescence techniques are used such that the presence of the
near-IR dye Cy5 indicates expression of exogeneous cdGAP. .
• Furthermore, the imaging capabilities of FLIM allows segmentation to distinguish between
membrane-associated activation and bulk-cell activation.
(a) Exemplar fields in the absence and presence of exogeneous cdGAP. (b) Lifetime results
at the cell membrane. (c) Lifetime results for whole cells.
Rassf-family interaction screen
• RASSF family proteins are implicated in regulation of apoptosis, microtubule stabilisation
and DNA damage response: they have been shown to play a role in tumour suppression.
• Classical RASSF proteins (1-6) contain a SARAH domain; N-terminal RASSF proteins lack
this domain.
• MST1 protein has also been shown to play a role in cancer signalling, and has been
shown to interact with RASSF proteins via heterodimerisation of the SARAH domain.
• We coexpress eGFP-RASSF family proteins MST1-mCherry, SARAH-deficient MST1ΔmCherry and a truncated MST1 consisting of the SARAH-domain only in order to
determine which RASSF proteins interact with MST1.
(a)
(b)
GFP
mCh
SARAH
R
R
eGFP-N-terminal RASSF
SARAH
mCherry-MST1 (full length)
eGFP-Classical RASSF
GFP
MST1
mCh
MST1
R
mCherry-MST1Δ
mCh
SARAH
mCherry-SARAH
(a) Cartoon showing FRET donor proteins. (b) Cartoon showing acceptor proteins: full length
mCh-MST1, non-interacting mCh-MST1Δ and truncated mCh-SARAH.
Rassf-family interaction screen
• As expected, classical RASSF proteins show shortened lifetime (increased FRET) when
coexpressed with both mCherry-SARAH and mCherry-MST1 compared to donor-only
wells. N-terminal RASSFs, lacking the SARAH domain, do not show any evidence of
binding.
FLIM plate reader: flexibility
• The ability to acquire large amounts of data quickly on the FLIM plate reader makes it a
useful tool for assay characterisation and optimisation.
• The FLIM plate reader has been applied to many FRET-based experiments as well as UVwavelength autofluorescence studies:
Ras/Raf interaction
in response to
gefitinib
treatement
FOXM1
SUMOylation in
response to DNA
damage
Autofluorescence
as readout for
cytotoxicity
Gag protein
aggregation and
response to NMT
inhibitor
Autofluorescence
of structural
proteins in
osteoarthritis
Rho activation at
cell-cell junctions
Screening for PPIs
during mitosis
Rac activation at
cell-cell junctions
RASSF-family
interaction screen
Summary
•
Our multiwell plate reading instrument allows high content FLIM
experiments to be performed in a practical timescale – as fast as 1 plate
per 15 minutes.
•
The instrument and acquisition software are flexible, allowing a large
range of different FLIM experiments to be automated.
•
FLIM-FRET can provide robust, information-rich high-content data.
•
We have illustrated application of FLIM-FRET in drug response studies
and interaction screening.
Thanks
Imperial College Photonics:
Collaborators:
•
Paul French
•
Eric Lam
•
Chris Dunsby
•
Mesayamas Kongsema
•
Anca Margineanu
•
Matilda Katan
•
Sean Warren
•
Jia Chan
•
Dominic Alibhai
•
Jess McCormack
•
Sunil Kumar
•
Vania Braga
Funding/commercial partners:
Kentech Instruments Ltd