PREPARATORY EXPERIMENTS
1. Prepare standard curves of fluorophores
Ψ Note: With every qF3 experiment, we imaged standard curves for the fluorophores.
Below we include instructions for the purification of these fluorophores:
A) Protein expression in E. coli
a. Transform BL21-AI™ One Shot™ Chemically Competent E. coli with DNA
construct, “His6TEV-mCerulean3-pBluescript” or “His6TEV-Venus-pBluescript” onto LB-agar.
b. Select a single colony and grow overnight in 50 mL Luria Broth (LB) + 100µg/mL
Ampicillin (AMP) at 37oC, shaking (250rpm).
c. Prepare two 1.5 litre LB-AMP. Add 25 mL of overnight culture to each then grow at
37oC, shaking (250rpm) until OD600 reaches 0.6
d. Induce with 0.2% L-Arabinose
e. Reduce temperature to 30oC
f. Incubate for 3 more hours, shaking.
g. Harvest bacterial pellet via centrifugation. (4500xg, 15min). Discard the supernatant and
freeze pellet (-20oC until purification).
B) Prepare buffers for protein purification
Ψ Note: For purification chill all buffers to 4oC and adjust final pH to 7.5.
a. 100mL Lysis Buffer: 50mM NaH2PO4 pH 8.0, 300mM NaCl, 10mM Imidazole, 1%
CHAPs, 2mM MgCl2.
Ψ Note: Add fresh: 2x PIN and 1x PMSF and 10µg/mL DNAse.
b. 100mL Wash Buffer: 50mM NaH2PO4 pH 8.0, 300mM NaCl, 10mM Imidazole.
c. 20mL Elution Buffer: 50mM NaH2PO4 pH 8.0, 300mM NaCl, 250mM Imidazole.
d. 4L Dialysis Buffer: 50mM NaH2PO4 pH 7.5, 150mM NaCl.
C) Purify fluorophore protein
Ψ Note: Collect small samples throughout each step for SDS PAGE.
a. Resuspend pellet in 5mL Lysis Buffer per gram of pellet.
b. Lyse bacteria (we use a French press)
c. Centrifuge (22000xg, 20min, 4oC)
d. Transfer supernatant to new 50mL falcon tube and discard the pellet.
e. Wash 1mL Ni-NTA Agarose slurry in 10mL Lysis buffer then add beads to
the supernatant
f. Incubate on rotation (4oC, 1h)
g. Pour supernatant + Ni-NTA slurry into a 15mL column, the resin will collect at the
bottom. Collect the supernatant and gently pour over the resin two more times.
h. Wash the Ni-NTA resin with 50-100mL Wash buffer (until the flow-through is colourless)
i. Prepare 10x 1.7mL Microcentrifuge tubes and label: E1 (elution 1), E2, E3… E10.
j. Add 10mL elution buffer. Let the first 1mL pass, then collect fluorophore in 10x 1mL
fractions.
k. Combine five fractions that have the highest intensity.
l. Dialyze protein, stirring at 4oC: (1L for 4h, 2L overnight, 1L for 4h).
m. Remove from dialysis tubing and centrifuge protein at 21000xg, 10min.
Ψ Note: If there is any pellet (precipitation) transfer supernatant to a new tube.
n. Add 20% Glycerol to dialyzed protein. Keep at 4oC, until concentration determined
and standard curves prepared (in our experience these fluorophores are stable at
4oC for > 1 year).
D) Determine the concentration
Ψ Note: Fluorescence of these fluorophores interferes with Bradford and BCA
assays. We ran a Coomassie-stained SDS PAGE gel with a known protein,
approximately the same size (BCL-XL), to estimate concentration in associated text.
a. Once concentration is known, make serial dilutions of mCerulean3 and Venus
proteins in Dialysis buffer + 20% Glycerol.
b. Store 500µL aliquots at -80oC.
2. Determine G factor (optional depending on the instrument)
Ψ Note: This step is required to correct for the overestimate in Venusfree concentration due to FRET stimulated emission. This step was necessary on our INO-FHS instrument because the acceptor channel detection is not synchronized with acceptor excitation. On a system with time-gated acceptor detection, this step is not necessary as the stimulated emission can be temporally gated out. To determine G Factor we based our experiment on the work done by Butz et al., 20167. Comparatively, our analysis is simplified because FRET efficiency is known from the lifetime. The rationale for the G-factor analysis is included in the Supplementary methods of our recent publication. Before this experiment, we had already determined the optimal imaging configuration (filter settings/acquisition time/objective/zoom, etc.) for rapid acquisition of live cell samples6. The following is the method used to determine the G-factor:
A) Prepare sample (3 days)
a. Seed 20 wells of a CellCarrier-384 Ultra microplate well plate with 2000 BMK-DKO
cells per well in a total volume of 25 µL DMEM complete media.
b. Leave plate at room temperature for 15 mins to allow cells to settle before returning
to the incubator (37oC, 5% CO2) for 12-24h
c. Transfect 2 wells each with plasmids encoding:
i) nothing
ii) mCerulean3
iii) Venus
iv) mCerulean3 + Venus
v) mCerulean3-5aa-Venus
vi) mCerulean3-13aa-Venus
vii) mCerulean3-31aa-Venus
viii) mCerulean3-39aa-Venus
Ψ Note: We used TransIT-X2 Transfection reagent and transfected cells as
detailed in Osterlund et al., 20196. We add pSPUTK plasmid (Stratagene Santa
Clara CA, Cat. No: CB4278654) to every transfection reaction to minimize
overexpression while maintaining a constant amount of DNA in all of the
transfection reactions.
All transfection reactions are prepared as follows:
Ψ Note: Reaction for transfection of 2 wells.
1. In a 1.7mL microcentrifuge tube add 50ng DNA of interest, 50ng pSPUTK, 9ul
Opti-MEM and 0.4µL Mirus reagent
2. Vortex transfection reaction
3. Centrifuge transfection reaction briefly (burst)
4. Incubate 15-30 mins (maximum) at room temperature.
! CRITICAL: do not leave standing more than 30 min
5. Add 50ul of complete media and mix gently by pipetting
6. Add 25ul of this transfection reaction on top of seeded cells.
7. Return plate to the incubator for 3-5h
8. Remove media with a pipette and add fresh complete media
Ψ Note: Changing media minimizes toxicity caused by the transfection
reagent.
Ψ Note: Alternatively, treat cells at this time point (Section 7.B).
d. Incubate plate 12-24h (37oC, 5% CO2)
B) Collect FLIM-FRET data on INO-FHS (2h)
a. Add the following standards to empty wells in the sample plate:
i) Venus protein standard curve
ii) mCerulean3 protein standard curve
iii) Quenched fluorescein
iv) 10nM fluorescein in 0.1M NaOH
b. Load imaging configuration (excitation wavelengths/filter settings) matching selected settings
that will be used for screening.
Ψ Note: Detailed method for selection of image acquisition settings and operation
of our INO-FHS instrument is included elsewhere6.
c. Acquire several images per well (we acquired ten 200µm by 200µm images per
sample) with each of three excitation configurations:
i) Acceptor excitation only (510nm)
ii) Donor and acceptor excitation together (510nm + 433nm)
iii) Donor excitation only (433nm)
C) Analyze data to determine G-factor
Ψ Note: Analyze protein standard curve using the method described later in Analysis Section 2:
The analysis will yield two values, the slopes of the mCerulean3 and Venus standard curves
(mc3_slope and ven_slope).
a. For the cell samples, run the region of interest (ROI) segmentation
analysis using INO_Analysis software as described later in Analysis Section 3. This
generates CSV files, which contain FLIM and spectral profiles for each ROI.
Ψ Note: for fitting analysis in the INO-FHS analysis software set ‘A0’ to 100 and
click the checkbox next to A0. This fixes the fitting analysis to a single exponential
because the lifetime of mCerulean3-Venus dimer constructs is homogenous.
b. Copy the generated .CSVs into a single folder (CSV directory) and note the Well
IDs used (ie. A1, A2 etc).
c. Download MATLAB scripts from GitHubLink:
https://github.com/DWALab/Osterlund_et_al_2021.git
subfolder: Gfactor_Estimation_Code
i) GFactorByMultipleCSVs.m
ii) CalcConcentratrionsV2.m
d. Open script ‘GFactorByMultipleCSVs.m’
e. Copy the path to CSV directory
f. input the data_path parameter with the path to CSV directory
g. input mc3_slope
h. input venus_slope
i. input the Well ID for sample 1c: ii. mCerulean3 (mc3_only_well_id)
j. Similarly, input the Well IDs for remaining dimer constructs.
Ψ Note: In our case, we have 4 constructs/samples with varying amino acid lengths
(section 2.A.c: v/vi/vii/viii), which are inputted as “construct_1_well_id”,
“construct_2_well_id”, “construct_3_well_id”, “construct_4_well_id”.
k. Run GFactorByMultipleCSVs.m script.
Ψ Note: The script will first open the mCerulean3 only wells to determine bleedthrough into the
acceptor channel and the lifetime of mCerulean3 expressed in cells. To reduce noise effects on
estimated lifetime only analyze ROI profiles that have, on average, mCerulean3 concentrations equal
to and above 0.5 µM and lower than 5 µM. Optionally, this range can be modified based on user
preference. The script will generate a 3D plot for FRET vs total mCerulean3 concentration and total
Venus concentration. This profile is fit in 3D to estimate the G-factor.
l. The G factor will be displayed. Copy G factor to notes (use later).
3. Generate stable cell lines expressing donor (mCerulean3) fused to a protein of interest
Ψ Note: We used baby mouse kidney (BMK) cells with BAX and BAK genes knocked out (DKO), to remove complications in screen analysis due to induced cell death with BH3 protein expression or BH3 mimetic treatment. Cells were routinely tested for mycoplasma. The mCerulean3 fusion protein is referred to as the donor.
A) Clone gene of interest as a fusion with mCerulean3, joined by a flexible linker
Plasmids encoding N-terminal fusions of mCerulean3 to four anti-apoptotic proteins:
a. mCerulean3-BCL-XL-s2193
b. mCerulean3-BCL-2-s2193
c. mCerulean3-BCL-W-FUGW
d. mCerulean3-MCL-1-pLVX
Ψ Note: Throughout this text, we use superscript “C” to indicate N-terminal fusion of
mCerulean3. ie. mCerulean3-MCL-1 is shortened to CMCL-1.
Ψ Note: We chose N-terminal protein fusion since it did not disrupt protein localization or
function7. Alternatively, C-terminal fluorescent protein fusions are possible as long as
protein function is not affected, and so long as sufficient FRET can be detected with
acceptor-fused interacting proteins.
B) Depending on the type of plasmid, transfect or infect cells with the construct of interest
Ψ Note: For genes in s2193 backbone we transfected BMK-DKO cells, selected with 5
µg/mL Blasticidin for two weeks then sorted for mCerulean3 expression. For genes in
FUGW and pLVX backbone, we transfected HEK-293T cells with viral packaging
components (PAX2 and VSVG), harvested virus after 48-72h, and infected BMK-DKO
cells. For selection, 4ug/mL Puromycin was used to generate the BMK-DKO-CMCL-1 cell
line. No antibiotics were necessary for creating our BMK-DKO-CBCL-W cell line as viral titre
was high enough such that selection/sorting was unnecessary.
C) Select and/or sort for mCerulean3 expression in cells
D) Culture cells for 1-2 weeks.
Ψ Note: We repeated sorting until Canti-apoptotic protein is expressed in >90% of cells.
E) Expand and freeze early passages of the stable cell line
F) Test fusion-protein function and stability
Ψ Note: We tested the anti-apoptotic function of our mCerulean3-fusion proteins in
another cell line, MCF7 breast cancer cells because we need BAX and BAK proteins
present for cells to die. We assume the function is maintained in BMK-DKO cells.
? Troubleshoot: If fusion protein is not functional try changing the location of
mCerulean3 fusion to the C-terminus. Alternatively, try increasing linker length between
the protein of interest and the fluorophore.
4. Assemble constructs to transiently express acceptor (Venus) fused to a protein of interest
Ψ Note: If structure N-terminal region of protein of interest has a critical function, then
design C-terminal fusion instead, adding a stop codon at the end of mCerulean3.
A) Create a plasmid encoding an acceptor fusion protein by inserting the cDNA for the
gene of interest downstream of Venus in pEGFP-C1 vector. Include a flexible
linker between the Venus and target protein. For example, we generated constructs
encoding Venus fused to the N-terminus of each of six BH3-only(VBH3) proteins:
a. Venus-BAD-pEGFP-C1
b. Venus-BIML-pEGFP-C1
c. Venus-tBID-pEGFP-C1
d. Venus-PUMA-pEGFP-C1
e. Venus-BIK-pEGFP-C1
f. Venus-NOXA-pEGFP-C1
B) Assemble plasmids encoding collisional control proteins; ie the protein of interest with
mutation(s) to eliminate binding. For example, for each VBH3 protein, we made four mutations of
critical residues in the BH3 binding regions to Glutamic acid (4E mutation):
a. Venus-BAD-4E-pEGFP-C1
b. Venus-BIML-4E -pEGFP-C1
c. Venus-tBID-4E -pEGFP-C1
d. Venus-PUMA-4E -pEGFP-C1
e. Venus-BIK-4E -pEGFP-C1
f. Venus-NOXA-4E -pEGFP-C1
! Critical: The collisional control is required for screening analysis later. This negative
control is necessary to distinguish true binding from the increased likeliness that two
proteins will collide when colocalized on a membrane surface. If the binding site is
unknown a non-binding Venus-fused protein of interest of approximately the correct size
and localization can be used. We have made some constructs available for this purpose:
a. Venus-7aa-ActA-pEGFP-C1 (Addgene ID # 177405), mitochondrial localization,
facing cytoplasm
b. Venus-7aa-MAO-pEGFP-C3 (Addgene ID # 166765), mitochondrial localization,
facing cytoplasm
c. Venus-7aa-Cb5-pEGFP-C1 (Addgene ID # 166764), endoplasmic reticulum
localization facing cytoplasm
5. Prepare 1000x stocks of compounds of interest
We used DMSO as a solvent to dissolve BH3 mimetic compounds to be tested in a qF3 screen. Whatever compound/solvent is used we suggest keeping dilution consistent throughout the plate. BH3 mimetics were frozen/thawed no more than twice.
A) Dissolve BH3 mimetic in fresh anhydrous DMSO to make 20mM stocks.
B) Store 5µL aliquots at -80oC, desiccated, for future use.
QUANTITATIVE FAST FLIM-FRET (qF3) SAMPLE PREPARATION & DATA ACQUISITION
Ψ Note: In our associated paper, mCerulean3 fused to an anti-apoptotic protein of interest (Canti-apoptotic) was stably expressed in BMK-DKO cells (BMK-DKO-Canti-apoptotic). Stable cells were transfected with Venus-fused to a BH3 protein of interest (VBH3) to measure the binding curve for Canti-apoptotic + VBH3. We treated this interaction with a panel of BH3 mimetic drugs, at several concentrations to measure the efficacy of these compounds to inhibit the interaction between Canti-apoptotic & VBH3. These terms are used in the example below.
6. Plan screen platemap to include the following:
A) One column of empty wells
Purpose: space for fluorescence protein standards
B) Untransfected BMK-DKO- Canti-apoptotic cells treated with DMSO diluted 1000x in
complete media.
Purpose: To determine the lifetime of the donor alone, Canti-apoptotic protein
C) Untransfected BMK-DKO- Canti-apoptotic cells, treated with BH3 mimetic diluted
from stock 1000x in complete media.
Purpose: To measure if the BH3 mimetic changes the lifetime of the donor. Each
compound must be tested separately. Minimally, include the highest concentration
tested.
D) BMK-DKO- Canti-apoptotic cells transiently transfected with plasmid (to express VBH3
protein of interest) and treated with DMSO diluted 1000x in complete media.
Purpose: To measure binding and generate a binding curve of Canti-apoptotic + VBH3.
Include at least one well for each VBH3 protein tested.
E) BMK-DKO- Canti-apoptotic cells transfected with plasmid (to express VBH3
protein of interest) and treated with BH3 mimetic diluted from stock 1000x in complete
media. Include titration of BH3 mimetic in the plate.
Purpose: To measure the efficacy of BH3 mimetic to inhibit the interaction between the
Canti-apoptotic & VBH3 proteins.
F) BMK-DKO- Canti-apoptotic cells transfected with plasmid (to express the control
non-binding mutant of acceptor-fusion (VBH3-4E)) and treated with DMSO diluted 1000x
in complete media.
Purpose: to use as a negative control for collisions.
7. Prepare live cell sample (3 days)
Ψ Note: The sample plate of cells is prepared as we have recently described for traditional FLIM-FRET4,6,8 in a 384 well plate, only scaled up in the number of samples we can acquire per day (Figure 1A). All steps are performed in a biological safety cabinet, using aseptic techniques.
* Hint: Print plate-map (Section 6). Move column or row-wise across the plate and check off each transfection/treatment as they are added.
? Troubleshoot: If any changes occur while transfecting/treating cells (ie. two wells swapped) note the change in the platemap. Our analysis software enables correction of such mistakes by combining replicate data according to what is indicated as in the well on the individual platemaps, rather than using a single master plate-map.
A) Seed cells (Day 1, ~1h)
a. Trypsinize cells stably expressing mCerulean3-fused protein (Donor) of interest in a
10cm dish (70-90% confluent)
b. Resuspend cells in 5mL DMEM complete media
c. Count cells using a hemocytometer
d. Dilute cells to 120 cells/µL, total volume >10 mL
e. While continuously mixing cells by inverting tube, seed 25 µL per well of a Corning
CellCarrier 384 well plate. Final 2000 cells per well. Do not tilt the plate while seeding.
! Critical: Remember to leave one column empty for standards.
f. Let stand at room temp for 15 minutes
g. Return to incubator (37oC, 5% CO2) for 12-24h
B) Transfect cells (Day 2, ~1h)
Ψ Note: See Section 2.A.c where we have already given a step-by-step description for
preparing a transfection reaction for 2 wells. For each construct included in the screen
determine the number of wells transfected and make a master transfection mix. We
also include control reactions with transfection reagent + media without DNA added to
“Untransfected” wells; Resulting in each well treated with transfection reagent. We
always overestimate the number of wells (make enough for 2 extra wells) when making
master mixes to guarantee enough to transfect all wells in platemap.
C) Treat cells (Day 2, ~2-3h)
3-5 hours after the transfection we routinely replace media with fresh DMEM complete
(50µL per well) as shown in Figure 1A. For convenience, we selected this timepoint to
treat cells with protein-protein interaction inhibitors.
a. Thaw 1 aliquot of each compound prepared in Section 5
b. In a 1.7mL tube, dilute each compound (or DMSO only) in complete media.
Ψ Note: Calculate volume needed based on plate design (Section 6). We
treat with 50µL per well in our 384-well plate format.
Ψ Note: Maintain the same concentration of solvent DMSO per well for each
concentration of compound tested. We diluted BH3 mimetics in DMSO, for 1:1000
dilution of each in complete media.
c. Remove transfection reaction media from each well with a pipette tip and replace
media with 50µL of appropriate treatment.
Ψ Note: We removed media from only one row at a time to minimize the time that
cells are left without media.
d. Return to incubator (37oC, 5% CO2) for 12-16h
8. Collect FLIM-FRET data on INO-FHS (1-2h setup + 24h automated data collection)
A) Add the following standards to empty wells in the sample plate (40µL per well):
a. Venus protein standard curve
b. mCerulean3 standard curve
c. Quenched fluorescein
d. 10nM fluorescein in 0.1M NaOH
B) Load imaging configuration (excitation wavelengths/filter settings), set environmental
controls (5% CO2 and 25oC), designate a location to save data, and align 384 well plate in
NIS-Elements.
Ψ Note: Step by step method for operation of our INO-FHS instrument, selection of
imaging settings and acquisition time is described elsewhere6. We summarized steps for
data acquisition in Figure 1B, with a few updates.
C) Go to well with stable untransfected cells. Select ‘fast scanning mode’ (0.5s
Frames Per Second (FPS)), focus on cells and align the pinhole manually by moving
the x/y position of the pinhole by 1000 µm steps. Adjust focus if needed. Find
the position where average counts are at maximum.
Ψ Note: We find this method allows us to monitor changes in the confocal image while
aligning the pinhole, which results in better quality images compared to when we align
the pinhole using fluorescein as we previously described6.
D) Go to well with 10nM fluorescein in 0.1M NaOH. Move focal plane into solution
(100 µm above the focal plane of the cells). Select 0.055 FPS and acquire the image.
Ψ Note: The purpose of this step is to ensure sufficient photon counts. When we
optimized the imaging parameters6 we measured our 10nM fluorescein in 0.1M
NaOH sample. From previous experience, an average of 10-13 counts per
pixel is sufficient for data collection.
? Troubleshoot: If the observed count from standard fluorescein is much less than
expected, go back to the “Untransfected” cells and retry the pinhole alignment. If this
does not work, add fresh fluorescein to the plate; check that there is sufficient water on
the objective and/or increase laser power to meet minimum expected counts.
E) Once settings have been optimized and pinhole aligned save the configuration.
F) Move back to the focal height of the cells.
G) Collect, set, and save “DARK” counts as described6
H) Move to quenched fluorescein sample. Collect, set, and save instrument response
function (IRF) as described6
I) Collect data for fluorophore standard curves (~10 minutes):
a. Move up 100μm (into solution)
b. In NIS Elements ‘WellSelection’: select the wells containing standard gradients
of the donor (mCerulean3) and acceptor (Venus) purified protein.
c. In NIS Elements ‘GeneratedPoints’: select 2 fields of view per well
d. In INO-FHS acquisition software, check “autosave images” and “save to new file”.
Save the configuration file.
e. In NIS Elements click “INO-F-HS initialize” then “INO-F-HS Grab”. Then run
automated data collection.
f. Once completed, move generated, ‘.TIFF’ files to a new subfolder called
“StandardGradient_TIFFs”
J) Move to a well with cells. Move down 100um then turn on autofocus. Set optimal
focal height using fast scanning mode (0.5 FPS).
K) In INO-FHS acquisition software, set frame rate to 0.055 FPS, check “autosave
images” and check “save to new file” then save the configuration file.
L) Collect data for the screen on INO-FHS (24h)
Ψ Note: As we described recently6, we use NIS Elements software to acquire one field
of view (FOV) per well of a 384 well plate sample then rescan the plate for a second FOV
and so on. This minimizes the difference in acquisition time across the plate.
a. In NIS Elements ‘WellSelection’: select wells containing cell samples.
b. In NIS Elements ‘GeneratedPoints’: select 1 field of view per well
c. Repeat a. for WellSelection1, WellSelection2, etc., and input a single GeneratedPoint
for each.
! Critical: Be sure to select a different location for each FOV.
Ψ Note: Analysis required a minimum of four 200μm x 200μm images per well6. In
general, we acquired 8-10 images per well for each screen.
d. Save NIS Elements Job file
e. Run automated data collection (24h)
! Critical: Generated, ‘.TIFFs’ files are large. Make sure there is enough space to
save the amount of data to be collected in your screen! Each file is 405 MB, thus for
8 fields of view, in a 384 well plate the total size is ~1.25TB (one replicate of a
screen).
9. Repeat steps 7-8 to acquire 2 more replicates
QUANTITATIVE FAST FLIM-FRET DATA ANALYSIS
SETUP FOR ANALYSIS
A) Install Software:
a. MATLAB R2020a
! Critical: Data analysis requires MATLAB Add-on Toolboxes:
• Curve Fitting Toolbox
• Signal Processing Toolbox
• Image Processing Toolbox
b. Microsoft Excel
c. INO Client Release_r10357 package:
• INO_F-HS_Acquisition
• INO_F-HS_Analysis
• INO_F-HS_BatchAnalysis
Ψ Note: We used GraphPad Prism8 and CorelDraw X8 to plot heatmaps and figures
displayed in this text. ROI selection done in INO-FHS software can alternatively be done
using MATLAB or Python.
Ψ Note: Contact lead contact for access to INO_F-HS software package. To install the
software you need to first run the vcredist_x64.exe. This installs the C++ compiler used by
the INO FHS. Then copy “DMGraph.dll” to your C:\Windows\SysWOW64 location. Finally,
you need to register the “DMGraph.dll”. This is done by running the "Register
DMGraph.bat". Depending on your Windows 10 security settings, the .bat file may not be
processed correctly. In that case, you will need to use the command line to register it
manually. If there are any issues please the contact lead contact:
[email protected].
B) Download our “Quantitative FAST FLIM-FRET” analysis package from
https://github.com/DWALab/Osterlund_et_al_2021.git
The “Quantitative_Fast_FLIM_FRET_Analysis” package contains 2 folders:
“Replicate_Analysis” and “Combined_Reps Analysis”.
C) Organize data as shown in Figure 2A. The main root folder here is called,
“Experiment”; each replicate subfolder is named, “Rep#_ YYYMMDD”. User
gives the experiment name.
D) Create a “MasterPlatemap” in Microsoft Excel, following the template given
(See example used in our BCL-XL screen: https://doi.org/10.5683/SP2/QN9C9R).
A separate sheet is made for each replicate (Rep1, Rep2, etc.). The well ID is divided into column
A (Row) and B (column B). For each well, the cell line (column C), transfection (column D),
treatment (column E) and concentration of treatment (column F) are indicated clearly.
Ψ Note: If there is a problem with running MATLAB analysis later on, usually it is due
to a problem with the user’s platemap.
? Troubleshoot: Check all these rules have been followed:
a. Use consistent naming conventions for each replicate (ie. if transfection is “V-BAD”
in Rep1 user cannot use, “VBAD” in the second replicate).
b. Include DMSO-treated transfected and untransfected wells (used as controls). In
the platemap column E, must fill treatment ID as “DMSO” and “untreated”
respectively.
c. Be aware of special symbol use. As in the example, report concentration of drug
treatment micromolar using the letter “u” instead of micro “μ”; MATLAB doesn’t
recognize this symbol.
d. Make sure all wells included in the “platemap” are in the screen. Do not include
empty wells or wells used for the fluorescent standards in this platemap.
e. Avoid using spaces when indicating cell lines, transfection, or treatment. A dash or
underscore links words; for example instead of, ‘mC3 MCL-1’ use ‘mC3-MCL-1’.
D) As shown in Figure 2A copy our, “CombinedReps_Analysis” folder to your main
experiment folder and make copies of our “Replicate_Analysis” folder to each
replicate subfolder.
ANALYSIS
Ψ Note: The analysis is broken down into a series of steps (1-13) summarized in Figure 2. We run the “Replicate Analysis” package (steps 1-7) for each replicate (Figure 2B) then combine and analyze results in the “Combined Replicates Analysis” package (Steps 8-13) (Figure 2C). Final data may be plotted as heatmaps in a screen. In our associated manuscript (Science Advances), we used qF3 analysis to compare a panel of BH3 mimetic drugs for disruption of several BCL-2 family protein-protein interactions in cells. As an alternative to following the step-by-step protocol below, follow our tutorial videos publicly available on YouTube:
https://www.youtube.com/playlist?list=PLUiSJrzzg9voe5sjA57oIbfOLAGIrHXRc
The above link includes the Replicate Analysis (Video 1) then Combined Replicate Analysis (Video 2).
REPLICATE ANALYSIS
1. Compress INO TIFF files (time required depends on the quantity of data. Can take 1-2 days)
Ψ Note: The collected INO .TIFF files contain FLIM and hyperspectral cubes collected through raster confocal imaging. The INO-FHS software provides a data compression option that uses the LZW lossless compression to reduce stored image files. However, compression adds significant overhead to the collection time (an additional 2 hours for 4 FOV in a 384 well plate). Consequently, we created a simple graphical user interface GUI in MATLAB to compress data post-acquisition. The data compression reduces the size of collected images 15-30 times; an important consideration in collecting and storing large screen data. The compression of a large data set may take 1-2 days on a single desktop. We recommend starting this step while collecting the following experimental replicates.
A) In the qF3_Analysis > Replicate_Analysis, locate the folder:
1_CompressINO_TIFFs
B) Open and run ‘Compress_INO_DATA_GUI_V2.m’ in MATLAB. A pop-up window will
appear containing three buttons.
a. Click on “Select Input Folder” and navigate to the path where the uncompressed TIFFs
are located.
b. Click on “Select Output Folder” to select a directory where the compressed TIFFs
should be stored.
Ψ Note: The output folder must be different from the input folder to avoid
overwriting original data. For consistency, we suggest compressing data and save
into the INO_TIFFs_compressed folder (Figure 2A).
c. Click “Compress Data”
Ψ Note: When compression is complete the status message will display:
“Data compression is complete”
C) Confirm that all files have been successfully compressed by comparing the number of
files in the input folder to those in the output folder.
Ψ Note: The script generates a log.txt file containing names of files that it could not
compress. This rarely occurs, but if it does find the original .TIFF and copy directly
to the compressed data location.
D) Delete folder of original .TIFF data.
2. Protein Gradient Analysis (takes less than 1 hour)
Ψ Note: The second step in our replicate analysis is performed to extract relationship between photon counts and concentration. As described earlier, donor and acceptor protein gradients are imaged before every experiment.
A) Move the compressed .TIFFs for the mCerulean3 and Venus protein gradient into a
separate folder
B) Open the INO_F-HS_Analysis software and create the analysis configuration file:
a. Click “Load Data” and select any .TIFF file from the protein gradient
Ψ Note: We will use this image to set up analysis parameters, which we can apply to
the remaining protein gradient TIFFs (in Batch Analysis).
b. In the General section, under “Lifetime Fit Algorithm”, select the “MLE-NP_fixed”
algorithm.
Ψ Note: This is an implementation of the Maximum Likelihood Estimation (MLE)
algorithm provided by INO. In our screen we did not use this, rather we extract
lifetime decays in .CSVs and use phasor analysis (later).
c. In the “Fit Parameters” section, change Tau_1 to 4000. Fix the Tau_1 by clicking on
the checkbox next to it.
d. In the “Fit Parameters” section, change A0 to 50 (initial estimate of the contribution of
the unbound lifetime in fitting).
e. In the ROI section, check the box next to “Auto ROI”. In the drop-down menu
select, “NaiveQuadrantBinning”.
Ψ Note: The image will be split into 4 quarters displayed in the scan viewer. TCSPC
decay and the Spectral profile for pixels belonging to each ROI will be combined to
generate 4 global profiles. Each of these profiles will be used to measure average
photon counts and spectral intensity per pixel.
f. In the IRF section, click “Load” and select the ‘IRF.IRF’ instrument response
function profile collected for that experiment.
g. In the calibration section click “Load” and select the “DARK.HSCalib” hyperspectral
calibration profile collected for that experiment.
h. Click “Save Config” to save all selected parameters for analysis. Save as
“QuadrantAnalysis”. File type automatically recognized as .INI = configuration
Settings.
i. Exit the INO INO_F-HS_Analysis application
C) Open the INO_F-HS Batch analysis application and run batch analysis:
a. Click “Load Config” and load the created “QuadrantAnalysis”.
? Troubleshoot: The INO-FHS analysis software looks for a specific folder to save
temporary data. If this folder does not exist in your computer directory, the software
will leave that information empty causing an error. If there is an error loading the
configuration file, open your saved analysis configuration file in a text editor such
as Note Pad. Search for “LocalSaveCacheDirectory” and look for the values after
the “=” sign. If this value is empty add two single quotations ‘ ’ to indicate that it is
an empty string or path.
b. Click “Load Folder” and select the path where the protein gradient .TIFFs are stored.
c. Click “Save Config” and designate a path to save a copy of the configuration file.
Ψ Note: Can save over the original QuadrantAnalysis configuration file.
d. Click “Save Folder” and designate a path to save the generated analysis results.
Ψ Note: Here, results will be saved to a new folder “Results_YYMMDDTime” saving
the date and time (military hour, minutes, seconds) that the analysis was run.
For consistency we suggest saving results within Replicate_Analysis> folder:
2_ProteinGradientAnalysis.
e. Click “Process Batch” to start the batch analysis process.
Ψ Note: A pop-up window should appear showing progress. For each .TIFF
analyzed, two files will be generated in the Results_YYMMDDTime folder: a
TIFF file containing the segmentation maps of the generated ROIs and a .CSV file
containing extracted data for each of interest (ROI) including ROI size, number of
photons, MLE fitting profile estimates, number of iterations the MLE required to
converge to a solution, TCSPC time-bin resolution, IRF profile, spectral wavelengths
detected and spectral profile. Next, we run a MATLAB script to extract the standard
curve data from these .CSV files.
D) In the “Protein Gradient Analysis\ 2_MATLAB_ExtractPixel-wise counts” folder open our
“LoopThroughPurifiedProteinCSV.m” script in MATLAB.
a. Copy the directory path to the “Results_YYMMDDTime” folder generated from the
batch analysis of protein gradient image data. In MATLAB, paste this directory in
the path parameter.
b. Run “LoopThroughPurifiedProteinCSV.m”. In the inputted path directory an excel
file named, “combined_gradient_results.xls” will be generated as results are
exported.
Ψ Note: Combined_gradient_results.xls contains three columns: the well ID, the
average photon count per ROI at Time 0 (T0) in the lifetime decay (FLIM channel) and
the average intensity measured at 530nm in the spectral channel.
E) In Microsoft excel, determine the relationship between intensity and concentration using
mCerulean3 and Venus protein gradients.
a. Input the mCerulean3 and Venus concentrations corresponding to each Well_ID.
b. Plot concentration of mCerulean3 (x-axis) versus photon counts (column B, y-axis)
for the donor (mCerulean3) protein gradient.
c. Plot concentration of Venus (x-axis) versus hyperspectral counts (Column C, y-
axis) the acceptor (Venus) protein gradient.
d. Fit data plotted in b and c to a straight line. The slopes of the line of best fit will be
used later in the analysis.
Ψ Note: We fix the intercept to 0,0 and fit data that falls in a range of intensities where
the relationship is linear. On the INO-FHS system, we determined this range in pilot
experiments: imaging purified protein gradients with configuration settings that we
had selected for screening6. We collected data for mCerulean3 and Venus protein
concentrations up to 50μM. We observed saturation in the TCSPC channel for high
concentrations of mCerulean3 and no saturation of the hyperspectral channel (DNS).
If the image has saturated pixels then the relationship between counts/concentration
cannot be linear. Thus, we selected a range of protein concentrations well below the
point of saturation that we could expect to observe a linear relationship for both
proteins (0-8μM mCerulean3 and 0-50μM Venus) in future experiments. We do not
extrapolate this relationship beyond the concentrations of our purified fluorophores.
3. Select Subcellular ROIs (time required depends on the quantity of data. Can take 1-2 days)
Ψ Note: Collected INO .TIFF data are analyzed similarly to the protein gradient analysis using the INO-FHS analysis and INO-FHS batch processing software. Here, we identify subcellular regions of interest (ROIs), bin pixels within each ROI, and extract data. We selected the segmentation algorithm compiled as a dynamically linked library (DLL) written and compiled using Visual Studio C++ 2015. The DLL is loaded as a plugin (Client Release_r10357>INO_F-HS_Confocal_Microscope>bin>ProcessPlugin). The INO software permits the user to define parameters used in the analysis, stored as a text file. Parameters such as the size of the Laplacian of a Gaussian filtering kernel, segmentation using FLIM or hyperspectral channel, as well as the minimum ROI size allowed. The saved FLIM-Hyperspectral cubes are saved into a multipage tiff format. The DLL has access to loaded FLIM and Hyperspectral cube pointers in memory. The segmentation DLL utilizes Open Computer Vision (OpenCV 3.4.1) library to perform image filtering and multi-seed watershed segmentation. Example segmentation is shown in Figure 3.
A) Open the INO_F-HS_Analysis software and create an analysis configuration file
a. Click “Load Data” and select any .TIFF file from the screen
b. In the General section, under “Lifetime Fit Algorithm”, select the “MLE-NP_fixed”
algorithm.
c. In the “Fit Parameters” section, change Tau_1 to 3800 (initial estimate of the
untransfected lifetime in picoseconds) and check the box next to Tau1. Change A0
to 50 (initial estimate of the contribution of the untransfected lifetime).
Ψ Note: Initial estimate is just a starting point for the fitting. Checking the box fixes
a parameter.
d. In the ROI section, check the box next to “Auto ROI”. In the drop-down menu
select, “NehadWatershedSegmentation”.
Ψ Note: This will segment the cells into ROIs. Clicking on any ROI will display the
TCSPC and the spectral profile. Note that some ROIs selected may appear to
segment background pixels, however, these ROIs will not meet the minimum number of
photons required for fitting and will not be exported in analysis.
e. In the IRF section, click “Load” and select the ‘IRF.IRF’ instrument response
function profile collected for that experiment.
f. In the calibration section, click “Load” and select the “DARK.HSCalib” hyperspectral
calibration profile collected for that experiment.
g. Click “Save Config” to save all selected parameters for analysis. Save as
“WatershedSegmentation”. File type .INI = configuration settings
h. Exit the INO INO_F-HS_Analysis application
B) Open the INO_F-HS Batch analysis application and run batch analysis for the entire screen:
a. Click “Load Config” and load the created “WatershedSegmentation”.
b. Click “Load Folder” and select the location of the INO TIFFs_compressed folder.
c. Click “Save Config” and designate a path to save a copy of the configuration file.
d. Click “Save Folder” and designate a path to save the generated analysis results.
Ψ Note: For consistency, we suggest saving results to Replicate_Analysis> folder:
3_SelectROIs
e. Click “Process Batch” to start the batch analysis process.
4. Combine CSV data belonging to the same well (time required is less than 20 minutes)
Ψ Note: INO_analysis software generates a comma-separated value (CSV) and a .TIFF file (saved ROI segmentation) for each image analyzed. Here, since multiple fields of views are acquired per well, we use command prompt to combine CSV files belonging to the same well into a “combined_csv”. Command prompt/bash scripts are currently most efficient for combining large CSV files. The generated CSV files contain numerical values represented using an ASCII format and they tend to be larger than the typical binary format. Combined CSV files can reach up to 160 Megabytes. For a single screen repetition, the size of the generated combined CSV files can reach up to 50 Gigabytes.
A) In the Results folder of batch analyzed screen data, move all the .CSV files to a
separate subfolder, “CSVs”.
B) Open the “Command Prompt” application
Ψ Note: In Windows, click on Start Menu and type “cmd” to find it.
C) Change to the CSV file directory by typing “cd” followed by a space and directory path
(ie. cd C:\Data), then press ‘ENTER’ to execute the command.
Ψ Note: Right-click to paste in command prompt.
D) type “dir” followed by hitting the ‘ENTER’ key
Ψ Note: command prompt will generate a list of files in the selected directory at this step.
? Troubleshoot: If you do not see a list of filenames, redo 4.C)-D). If again you do not see
a list of filenames, we suggest making a temporary copy of your “CSV” folder to the
desktop. Then redo 4.C)-D), inputting the desktop>CSV folder directory instead. Delete
temporary copies after F).
E) Copy and paste the following line into the command prompt window:
For %i in (A,B,C,D,E,F,G,H,I,J,K,L,M,N,O,P) DO For /L %j in (1,1,24) Do copy /b WellID_%i%j-* combined_%i%j.csv
Ψ Note: This command generates well IDs in order. It identifies the files that have the
same IDs and copies their content to a single file. You will see “Combined_WellID” files
generated in the location of your CSV files ie. Combined_A2.
F) Move all “Combined_CSV” data files to a separate subfolder.
Ψ Note: For consistency, we suggest using the Combined_CSVs folder located:
Replicate_Analysis>4_CombineDataPerWell.
5. Filter CSV data and extract binding curves (time required = usually 2-4h)
Ψ Note: In this step, slopes of our standard curves (Analysis step 2) are used to convert intensities to concentrations and determine the fractional change in angular frequency (Δꞷ): a value that is linearly related to bound fraction, by phasor analysis of lifetime per ROI (see associated manuscript for more details). Figure 4A is an overview of the tasks completed in this step. The MATLAB script requires the MasterPlatemap (created earlier).
! Critical: This script uses the G-Factor determined as described in Preparatory Experiment Step 2: Determine G factor. The G-Factor must be updated if the experiment is being carried out on a different microscope or if the filter settings have been changed on the INO-FHS. To update the G-Factor parameter open the “CurveExtractionAndShift.m” script located in Replicate_Analysis> 5_FilterData_ExtractRawCurves and change the parameter, “G_FACTOR” to the value determined for the instrument/settings used for the screen. As noted earlier, instruments that have a time-gated acceptor channel will not require the G-factor correction at all. Contact author, Nehad Hirmiz for advice on how to adapt this script for such instruments.
A) Copy the full path to the MasterPlatemap, for example, C:\Data\MaterPlateMap.xlsx.
B) In MATLAB, open and run the FilteredDataExportationSonstantTx.m script located in
Replicate_Analysis> 5_FilterData_ExtractRawCurves as follows:
a. Update the ‘excel_file’ parameter by pasting the full path to the MasterPlatemap
and update the ‘excel_sheet’ parameter to indicate which sheet to reference. For
example, in our MasterPlatemap sheets are labelled “Rep1”, “Rep2”, etc.
Ψ Note: Need to match the sheet name in the MasterPlatemap exactly.
b. Indicate which cell line(s) exist in the plate by updating the cells parameter.
For example, cells = {{‘mC3-MCL-1’},{‘mC3-BCL-2’},{‘mC3-BCL-XL’}}
Ψ Note: The name {‘here’}, must exactly match how the name appears in your
MasterPlatemap, column C.
c. Copy the path to your Combined_CSV folder (Analysis step 4G). Then in MATLAB
update the parameter: path_to_combined_csv
Ψ Note: paste path here ‘path\’, leaving the \ at the end.
d. From Analysis Step 2, get slopes of Venus and mCerulean3 gradients and
update the parameters: venus_gradient_slope and mc3_gradient_slope
e. Run the script.
! CRITICAL: Let the script run to completion before opening any exported file in
Microsoft Excel.
Ψ Note: By default, the results are saved into the same directory as the script in a
subfolder, “RawCurves”. Results for each well in the platemap are saved as:
combined_WELL_ID_filtered_raw.csv (example combined_A2_filtered_raw.csv).
Each file contains calculated Venus concentration, mCerulean3 concentration, free
Venus concentration, bound mCerulean3 fraction, and FRET activity per ROI.
Columns A-E data were determined by phasor analysis of lifetime. To generate an unbinned
binding curve with raw data: plot column C versus column D (Δꞷ) or verses column E
(FRET efficiency). We found the phasor lifetime analysis more reliable than the
double exponential fit performed in the INO-Batch Analysis software (see Analysis
Step 3A.b-c) since the phasor analysis does not require user input. However,
the INO fit analysis data is still exported in columns F-J, in case useful. The Free
Venus concentration (column F) recalculated using the fraction of mCerulean3 with
a lifetime shorter than 𝜏1 (where 𝜏2 is unfixed) as determined from the double
exponential fit (column G), corresponding FRET efficiency (column I) and
Chi2 to estimate the goodness of fit (column J). Finally, we export the
acceptor:donor intensity ratio (column H), which may be used to generate
traditional, non-quantitative FLIM-FRET binding curves if desired. All this is
exported for each ROI to provide users with many options to examine/filter data.
Ψ Note: In addition, this script exports the phasor coordinates for each
untransfected, treated well in the platemap (Figure 4A). This data is exported as,
“Untransfected_Phasor_Coordinates.xlsx” in the RawCurves subfolder.
6. Get Lifetime for Untransfected controls (time required= 5 mins)
Ψ Note: In Analysis Step 5, we extracted the phasor coordinates of the unbound lifetime (𝜏1 in phasor plot) for each Untransfected treated control to use in determining Δꞷ for corresponding treated, transfected wells. Some treatments may affect the donor lifetime, which can lead to errors in the 2-component analysis. For example, see the change in 𝜏1 with ABT-199 treatment of BMK-DKO cells expressing mCerulean3-BCL-2 (Figure 4B). In this step, we convert the exported phasor coordinates from Step 5 to a lifetime (ns) to examine the effect of each drug on the donor lifetime.
A) In folder Replicate_Analysis> 6_Get_LifetimeUntransfectedControls> open MATLAB
script: “GetLiftimeFromPhasorCoordinates.m”
a. Copy full path to the “Untransfected_Phasor_Coordinates.xlsx” file generated in the
previous step then in MATLAB, update path_to_phasor_coordinate parameter
b. Run the script
Ψ Note: Results are exported in the same location as the script with filename:
UntransfectedLifetimeFromPhasor.xlsx. This file includes only untransfected wells
from the platemap. The Well ID, Cell line, Treatment, and Concentration, and average
lifetime (nanoseconds) per well are given in columns (A-E, respectively).
Ψ Note: Later in step 13, we combine 3 or more replicates to examine the effect of treatment on donor lifetime. If a compound directly changes donor lifetime > 0.2ns then remove from published screen data. For example, we observed ABT-199 treatment has a direct effect on mCerulean3 lifetime in BMK-DKO cells expressing mCerulean3-BCL-2 in our associated manuscript. As a result, we observed unexpected trends in the data suggesting errors in the 2 component analysis. Hence, we considered the qF3 binding curves for ABT-199 preliminary and removed them from the main paper.
7. Bin binding curve data (time required = 30 mins)
Ψ Note: In this step, we filter, and bin the raw binding curve data generated in Analysis step 5 (Figure 4C). This will allow us to combine binding curves from multiple replicates. This script also generates a sheet, “Donor_Concentration_Table .xlsx” containing mean donor concentration in each well, which will be used in later steps to examine the variation of donor expression across the plate.
A) In the Replicate_Analysis> 7_Bin_CurveData_FreeAcceptor folder, open and run the
MATLAB script: “UnshiftedRawCurveBinning.m script.m” as follows:
a. Update the ‘excel_file’ parameter by pasting the full path to the MasterPlatemap
and update the ‘excel_sheet’ parameter to indicate which sheet to reference. For
example in our example MasterPlatemap sheets = “Rep1”, “Rep2”, etc.
Ψ Note: Need to match the sheet name in the MasterPlatemap exactly.
b. Copy the directory to the “RawCurves” folder containing .CSV files exported in
Analysis Step 5. Input the path in MATLAB.
c. Modify filters if desired. Open MATLAB script, “BinRawUnshiftedData.m” and
update parameters: B (Δꞷ) and FA (free acceptor).
B=B(FA<50 & mC3>1 & mC3<3);
FA=FA(FA<50 & mC3>1 & mC3<3);
These lines specify that we keep ROIs that have free acceptor less than 50 μM and
that have mCerulean3 concentration between 1-3 μM.
Ψ Note: Default script filters data to keep ROI data that have a mCerulean3
concentration from 1-3µM and a VenusFree concentration less than 50µM.
d. Modify Bins if desired. Open MATLAB script, “BinRawUnshiftedData.m” and
update parameter: bins1
Ψ Note: Default bins for quantitative curves are indicated in bins1 parameter:
[-1,1,2,3,4,6,8,10,12,16,20,30,40,50]
Default bins for Acceptor:donor intensity curves are indicated by the bins2 parameter.
e. Run the script. Results will be exported to three subfolders in the same location
as the script:
i) AD_FRET_Binned_Curves subfolder
Ψ Note: In this folder, we extract traditional FLIM-FRET binding curves
for each well. The data is binned by Acceptor:Donor intensity ratio and
for each bin, median FRET efficiency, standard deviation, and number of
points are exported. While the script bins and exports these data we did not
use these results in our associated publication.
ii) INO_Binned_Curves subfolder
Ψ Note: In this folder, we extract the binned binding curves generated from the
INO-FHS fitting analysis. The data is binned by VenusFree concentration and
for each bin, median Δꞷ, standard deviation and the number of points per bin
are exported. While the script bins and exports these data we did not
use these results in our recent publication.
iii) Phasor_Binned_Curves subfolder
Ψ Note: In this folder, we extract the binned binding curves generated from the
Phasor lifetime analysis (step 5). The data is binned by VenusFree
concentration and for each bin, median Δꞷ, standard deviation and
number of points are exported and used in our associated publication.
ANALYSIS OF COMBINED REPLICATES
Ψ Note: For three or more biological replicates of the screen, complete the “Replicate Analysis” (Figure 2B, steps 1-7). The next part of the analysis “Combined Replicates Analysis” package (Steps 8-13) are illustrated in Figure 2C. As an alternative to following the step-by-step protocol below, follow our tutorial videos in Combined Replicate Analysis (Video 2).
8. Combine binned data for multiple replicates
A) In CombinedReps_Analysis> 8_CombineBinnedReps folder, open MATLAB script,
“CombineBinnedFilterDataFor3Reps.m” to combine 3 replicates
OR
“CombineBinnedFilterDataFor4Reps.m” to combine 4 replicates.
! CRITICAL: The bins parameter in this script must be the same as the bins used in the
Replicate Analysis. If bins were changed earlier in analysis Step 7. A) d then update
the bins parameter here.
a. Update the path to the MasterPlatemap (excel_file parameter)
Ψ Note: Plate maps for repeat experiments do not need to be identical. The script
uses the first repetition “Rep1” sheet as a reference and searches for the identical
Cells/transfection/treatment/concentration in the remaining repetitions.
b. Copy each path to binned data generated in Replicate Analysis Step 7 and input
parameters: path_to_binned_data1, path_to_binned_data2, path_to_binned_data3
c. Input the path to two folders where data will be exported:
i) The combined binned data (combined_data_export_path).
Ψ Note: Rep1, Rep2, Rep3, Rep4 binned data are pasted into one excel file.
ii) The average of the binned data (average_binned_export_path).
Ψ Note: For consistency we suggest to save data to CombinedReps_Analysis>
8_CombineBinnedReps subfolder, “CombinedBinned_Reps”,
“AveragedBinned_Reps”.
d. Run the script. For each transfected well in the platemap, a file will be generated
containing the combined binned data from 3+ replicates.
9. Fit combined, binned binding curves (Time required= ~1h)
Ψ Note: Refer to our associated publication for rationale and details behind screen analysis and fitting criteria. In summary, see Figure 2C. In order to identify a true protein-protein interaction, each binding curve must be compared to a collisional (negative) control. The script constructs a table of control transfections (DMSO treated only) and finds the negative control for each. We then set the minimum difference between the test and negative control for continued screen analysis: the difference in median Δꞷ between the untreated positive and negative controls (Δβ) must be larger than 0.05. If Δβ> 0.05 then we calculate % Resistance to drug treatment, as described. However, binding does not mean there is sufficient data to fit for the determination of Kd. So, we established a second criterion for fitting Kd: the shape ratio (sRatio) must be greater than 2. This is a non-parameterized approach to describe the shape of the binding curve, without fitting. If sRatio is >2 then we consider that the binding curve has sufficiently plateaued to fit data with a single-site binding Hill slope equation (fixed Hill Slope=1) and determine Kd.
A) In CombinedReps_Analysis>9_Fit_CombinedCurves open MATLAB script,
“AutomatedKdExtractionRepsAllPoints.m” and run as follows:
a. Update the path to the MasterPlatemap (excel_file parameter)
Ψ Note: The script uses “Rep1” sheet as a reference to identify the positive and
negative controls for each experiment.
b. Input path to the folder containing combined, binned CSVs (parameter=
path_to_binned_csvs) generated analysis step 8: CombinedReps_Analysis>
8_CombineBinnedReps> CombinedBinned_Reps.
c. Input cell line(s) exactly as they appear in the MasterPlatemap. Paste each name
{‘here’} ie. cells ={{‘mC3-BCL-2’},{‘mC3-BCL-XL’}}.
d. Input positive/test transfection(s) exactly as they appear in the MasterPlatemap.
Paste each name {‘here’} ie. = {{‘V-BAD’},{‘V-tBID’},{‘V-BIML’}}.
e. Input corresponding negative/collisional controls exactly as they appear in the
MasterPlatemap. Paste each name {‘here’} ie. = {{‘V-BAD-4E’},{‘V-tBID-4E’},
{‘V-BIML-4E’}}.
! Critical: A negative control must be indicated for each positive control and must be
listed in matching order.
Ψ Note: The same negative control can be used for multiple test wells.
ie. if positive controls are = {{‘V-BIML’},{‘ V-BIML-mutant1’},{‘ V-BIML-mutant2’}}
the corresponding negative control “V-BIML-4E” should be inputted three times.
Negative controls = {{‘V-BIML-4E’},{‘V-BIML-4E’},{‘V-BIML-4E’}}.
f. Run script. At the beginning of the analysis, each positive control binding curve will
displayed in a pop-up window as shown in Figure 5A appears for the user to review.
Ψ Note: The script estimates the point of binding saturation based on the median of
the last 6 points in the binding curve (median Δꞷ of the points with the highest
VenusFree concentration). The estimated Saturation point is plotted as a blue line.
i) If the user approves the estimated saturation parameter, then hit “ENTER” to
continue with the next dataset to review.
ii) If the saturation parameter is not correctly estimated, the user can override
and input the saturation parameter by typing, “-1” in the command window
then manually entering the saturation parameter (ie. 0.2), then hitting ENTER
to continue analysis.
Ψ Note: This script exports the saturation parameters for all of the controls in the file
“SatParameters_CombinedBinnedReps.csv”. This saves the fitting parameters,
including any user input, for future replication of fitting analysis and for plotting the fit
later in step 10. Note that the fitting parameters determined using these positive
control wells are subsequently fixed in fitting all wells with the matching transfection
as well as for fitting the corresponding negative control. The script exports full plate
analysis (“CombinedBinnedReps_Results.xlsx”) to the same location as the
combined, binned data: CombinedReps_Analysis> 8_CombineBinnedReps>
CombinedBinned_Reps folder. This file contains columns A-F copied from the
MasterPlatemap for data navigation. In columns G-P we export:
G: Sum the number of ROIs analyzed in total to make up the curve)
H: Sum the number of binned points in the combined, binned curve)
I: Maximum VenusFree (μM) concentration of the binned points in the combined,
binned curve
J: Cumulative Sum of the area under the curve
K: Median Δꞷ of binned points that fall within 10-20 μM on the combined, binned
curve
L: % Resistance to treatment (drug), calculated as previously published4
M: Shape Ratio, calculated as described earlier
N: Kd (μM VenusFree), calculated from fitting combined, binned data to a one site
specific binding Hill slope equation
O: Lower confidence interval of fitting Kd (μM VenusFree)
N: Upper confidence interval of fitting Kd (μM VenusFree)
10. Plot combined binned data (Time required: ~10-20 mins)
Note: In interpreting heatmap results, sometimes it helps to open the binding curve to examine the data and fit. To save time, in this step we run a script that plots each binding curve in the platemap (combined, binned data exported in step 8). Usually, we want to compare the result to the positive and negative control, so the script is designed to find the corresponding control and plot them on the same graph, as shown in Figure 5B. Exported plots display the upper and lower confidence interval for the fit of each curve using a shaded area. Blue shade for positive control, gray shade for the negative control, and red shade for the binding curve of interest (WellID + cells + transfection + treatment saved in the title of the graph). In the “simple curve plotting” algorithm we apply the same thresholds (Δβ <0.05 or sRatio <2), and use the same saturation parameters (including user-input), so the displayed fit matches what was done in step 8 to extract Kd.
A) Go to the CombinedReps_Analysis> 8_CombineBinnedReps>CombinedBinned_Reps
folder and open the “SatParameters_CombinedBinnedReps.csv”, which was exported in
Analysis Step 9. Save as an excel file, “SatParameters_CombinedBinnedReps.xlsx”.
B) In the CombinedReps_Analysis>10_Plot_CombinedCurves folder, open
“SimpleCurvePlotting.m” in MATLAB
a. Input the full path (“ie. C:\Data\CombinedReps_Analysis\8_CombineBinnedReps
\CombinedBinned_Reps\ SatParameters_CombinedBinnedReps.xlsx”) to the
saturation parameter table to update the satpath parameter.
b. Copy the directory to the “CombinedBinned_Reps” folder (Step 8.A.c.i)
and update the path_to_binned_csv parameter (For example,
C:\Data\Analysis\8_CombinedBinnedReps\CombinedBinnedReps)
Ψ Note: file must be named SatParameters_CombinedBinnedReps.xlsx, otherwise
need to update the “satfile” parameter as well
c. Update the path to the MasterPlatemap (excel_file parameter) and indicate the
name of the sheet used as a reference. By default, this is, “Rep1”
d. Update experimental conditions similar to how we did in Step 9.
i) Input cell line(s) exactly as they appear in the MasterPlatemap. Paste each
name {‘here’} ie. cells ={{‘mC3-BCL-2’},{‘mC3-BCL-XL’}}.
ii) Input transfection_reference parameter (positive/test transfections) exactly as
they appear in the MasterPlatemap.
Paste each name {‘here’} ie. = {{‘V-BAD’},{‘V-tBID’}}.
iii) Input negative_reference parameter (negative/collisional controls) exactly as
they appear in the MasterPlatemap.
Paste name {‘here’} ie. = {{‘V-BAD-4E’},{‘V-tBID-4E’}}
! Critical: A negative control must be indicated for each positive control and
must be listed in matching order.
e. Input a directory to save plotted graphs, update the path_to_final_graphs parameter
Ψ Note: For consistency, we save graphs in the location:
CombinedReps_Analysis>10_Plot_CombinedCurves subfolder, “Combined_Curves”
f. Optional * select format for export (at end of the script)
Ψ Note: Default script exports a .jpg and .epsc file for each graph. If another
format is desired, then modify the line towards the end of this script, ‘saveas’
g. Run script.
11. Determine average donor concentration per well (Time required: 5 mins)
Ψ Note: Detected binding affinity depends on the amount of available donor and acceptor labelled proteins in each ROI. We restrict our analysis to ROIs that have a limited range of donor labelled protein expression to ensure consistency across the plate map and between replicates. We perform this step to check that expression of the donor is consistent across the screen plate and that no treatment has a dramatic effect on donor concentration. Recall that in Replicate_Analysis step 7, we exported a table containing the average donor concentration per well for each replicate in file “Donor_Concentration_Table.xlsx”. Here we combine these data for all replicates: the script identifies what is in the well from Rep1 then it finds the corresponding treatment in Rep2, 3 and 4 to combine the data.
A) In folder CombineReps_Analysis > 11_GetAvgDonorConcentrationPerWell, open
“GetAverageDonorConcentration_3Reps.m” in MATLAB
Ψ Note: If you have 4 replicates, “GetAverageDonorConcentration_4Reps.m”
a. Input path to the folder containing for each Donor_Concentration_Table.xlsx from
replicates 1-4, updating parameters: path_donor_table_1, path_to_donor_table2,
path_to_donor_table3 and path_to_donor_table4
b. Run script.
Ψ Note: The results will be exported in the same folder. Two files will be generated:
the first, “Average_Donor_Concentration.xlsx” using the list format contains data in
list format, where columns A-F were copied from the MasterPlatemap for data
navigation and in columns G-J we export:
G: Mean of replicates of the average mCerulean3 concentration (μM) per well
H: Standard deviation of column G
I: Median of replicates of the average mCerulean3 concentration (μM) per well
J: Mode of replicates of the average mCerulean3 concentration (μM) per well
The second file exported, “Average_Donor_Concentration_Maps.xlsx” takes the
same data and transforms it to a platemap array for plotting heatmaps.
12. Generate Heatmaps for the entire screen (Time required: 5 mins)
Ψ Note: In this step, we convert results (exported in Analysis Step 8) from a list format to a 384 well platemap array. This is ideal for plotting heatmaps, viewing and interpreting results from large screens.
A) Navigate to CombinedReps_Analysis > 12_GenerateHeatmap and open script,
“GenerateParameterMap.m” in MATLAB
a. Copy path to the “CombinedBinnedRepsResults.xlsx” file saved in step 8 then
update excel_path parameter
Ψ Note: If the filename/sheet name has been modified the user must update the script.
b. Run Script. Wait for the message indicating analysis completed.
Ψ Note: the script converts the data from a list to a platemap array format, saved as
“GeneratedHeatMapsFinal.xlsx” in the same location as the excel_path.
B) (Optional) navigate to CombinedReps_Analysis > 12_GenerateHeatmaps and open
“PLOTHeatmap.m” script designed to plot Heatmaps in MATLAB
Ψ Note: Heatmaps displayed in our associated paper were plotted in Graphpad Prism. This
is an alternative script to generate heatmaps.
a. Open resulting data, “GeneratedHeatMapsFinal.xlsx” exported in platemap array
format. For whatever value you wish to plot a heatmap of (ie. Kd, number of points
per well, Venus expression per well, %Resistance etc.) copy data from excel.
Ψ Note: User can copy the entire 384 well plate, a section of it, or rearrange data for
plotting.
b. After kd_map, paste data [ here ] (within the brackets in place of example data)
c. Select color map (myColorMap parameter) and scale (parameter caxis=
([min_value,max_value]))
Ψ Note: The default colormap is “Jet” divided into 100 segments.
d. Run the script. A pop up window should appear with the resulting heatmap.
? Troubleshoot: You may see the following, “Error using vertcat Dimensions of
arrays being concatenated are not consistent”. This error occurs due to the presence
of empty wells in the pasted data. To fix this, make sure each cell copied contains a
numerical value. To get around this, we manually fill in any empty well with some
known outstanding numerical value. Make sure to keep track of changes in data to
avoid confusion. For example, if data was insufficient for fitting, then the value for Kd
will be exported as ‘NaN’ in Analysis step 9. In converting to a platemap array, all
wells marked “NaN” will be empty. We suggest filling empty wells with a nonsensical
value; for example, a Kd of ‘-1000’ is easy to recognize and remove (make appear
blank) before publishing final heatmaps.
13. Investigate the Lifetime for Untransfected Cells
Ψ Note: In Replicate Analysis Step 6b as we exported lifetime data for all untransfected wells in the file: UntransfectedLifetimeFromPhasor.xlsx. In this step of the analysis, we combine the results from multiple replicates based on finding what is in each row in MasterPlatemap Rep1 (Untransfected cell line, treatment, and concentration) then searching for the corresponding data in Replicates 2,3,4. This method of combining data is useful to account for any change in Well locations between replicates. If there were no changes in the platemap design between replicates, the user can easily combine data by hand.
A) Navigate to CombinedReps_Analysis > 13_Combined_UntransfectedLifetimeReps and
open “CombineLifetimeInfofrom_3Reps.m” in MATLAB
Ψ Note: for 4 replicates use “CombineLifetimeInfofrom_4Reps.m”
a. Update the full paths (path_excel1, path_excel2, path_excel3, and path_excel4) to the
“UntransfectedLifetimeFromPhasor.xlsx” file generated for each replicate respectively.
b. Run the script. The results will be saved in the same location as the script.
Ψ Note: By this method, we plot data shown in Supplementary Figure 5 of the main
associated text. We identified that ABT-199 has a significant impact on the mCerulean3 donor lifetime
and all binding curves probed by this compound were subsequently removed after failing to pass
quality control standards.