################################################################
Dataset Information
################################################################

Dataset Name:
Live-cell imaging of cell-to-cell variability in caspase-8 FRET reporter activity and cell fate trajectory modeling

Dataset Description:
This analysis dataset presents calculated parameters fit to a phenomenological model that defines an initiator caspase threshold that underlies cell-to-cell variability in cell killing by TRAIL (TNF10) and therapeutic antibody TRAIL receptor agonists. The analysis is based on live-cell, microscopy-based fate tracking of cells following perturbation with TRAIL, therapeutic antibody TRAIL receptor agonists, and other agents that modulate the cellular fate decision.

--Data in Package:
20232.txt

--Metadata in Package:
Small_Molecule_Metadata.txt
Protein_Metadata.txt
Cell_Line_Metadata.txt

################################################################
Center-specific Information
################################################################

Center-specific Name:
HMS_LINCS

Center-specific Dataset ID:
20232

Center-specific Dataset Link:
http://lincs.hms.harvard.edu/db/datasets/20232/

################################################################
Assay Information
################################################################

Assay Protocol:
1. Costar 96-well plates were coated with 50 uL/well of rat-tail collagen I at 40 ug/ml for 2 hours at 37C, washed 3 times with 200 uL PBS, and allowed to dry in a tissue culture hood.<br />
2.HeLa cells stably expressing a FRET-based initiator caspase reporter (ICRP), or derivative cell lines expressing additional transgenes where indicated, were plated (2500 cells in 100 uL per well of DMEM media with 10% FBS). Only the center wells of each plate were used, and 200 uL of media were added to all surrounding wells. Plates were allowed to sit in the tissue culture hood for 20 min before being moved to an incubator for 24 hr.<br />
3. The cell media was refreshed with media lacking phenol red in preparation for imaging.<br />
4. Cells were pre-imaged for 30 min before drug addition (in media lacking phenol red).<br />
5. Perturbagens were added to the media as specified in the dataset.<br />
6. Cells were imaged every 5 min for 24 hr using a PerkinElmer Operetta robotic microscope with a live-cell chamber (at 37C with 5% CO<sub>2</sub>) using a 10x objective (NA = 0.4) and filter configurations for CFP (Excitation: 425-450 nm / Emission: 460-500 nm) and CFP-YFP FRET (Excitation: 425-450 nm / Emission: 520-560 nm). For cells stably expressing proteins of interest tagged with mCherry and for cells stably expressing a mitochondrial inter-membrance space reporter (IMS-RP, MOMP reporter), which was used for methods validation only (see point 11 below), one additional filter configuration (Excitation: 550-600 nm / Emission: 610-660 nm) was used throughout the experiment when the MOMP reporter was present or at the beginning of the experiment when a protein of interest tagged with mCherry was present.<br />
7. Image segmentation was performed as follows: For background subtraction, images I<sub>raw</sub> were tiled into 64 blocks sized 170 by 128 pixels. The Gaussian-filtered intensity distribution in each tile was used as a measure of local background for reconstruction of a full-resolution background image I<sub>bg</sub> by bilinear interpolation over the tile modes. Given zero-mean background intensity in I<sub>CFP</sub> = I<sub>CFP,raw</sub>  I<sub>CFP,bg</sub>, the lower half of background intensity was wrapped over the upper half in the distribution of absolute intensities |I<sub>CFP</sub>|, such that the 20th percentile p<sub>20</sub> of |I| could be used as a robust estimate of background variability. At a threshold t = 3 p<sub>20</sub>, false-positive mask elements in M<sub>t</sub> = (I<sub>CFP</sub> > t) were sparse enough to be reliably suppressed by morphological erosion. Before erosion, contacting cells were separated by watershed segmentation of I<sub>CFP</sub> after applying a wide Gaussian filter. The Gaussian filter was parameterized such that oversegmentation was rare. The combined final mask M for single-cell readouts was obtained from the threshold mask M<sub>t</sub> and the watersheds W by eroding M<sub>t</sub>'W.<br />
8. Intensity readouts in the form of the FRET ratio I<sub>CFP</sub> / I<sub>FRET</sub> were extracted from each cell in each segmented frame. To reduce the influence of chromatic aberration and between-channel jitter, the background-subtracted FRET image I<sub>FRET</sub> was aligned to the background-subtracted CFP image I<sub>CFP</sub> with subpixel resolution by bilinear interpolation. The required shift was determined for every frame by quadratic peak interpolation of the cross-correlation coefficients between I<sub>FRET</sub> and I<sub>CFP</sub>. The mask M determined from I<sub>CFP</sub> was applied per cell to both I<sub>CFP</sub> and the aligned I<sub>FRET</sub>, such that the FRET ratio could be calculated as the median of the pixel-wise ratio of intensities in the mask area. Due to erosion of M<sub>t</sub>'W, thin protrusions of cells with larger relative error in I<sub>FRET</sub> and I<sub>CFP</sub> were excluded from the median of pixel-wise ratios. As a measure of expression level of mCherry-tagged proteins of interest (FLIP-S/FLIP-L, Bcl-2/Bcl-XL), the 80<sup>th</sup> percentile in the mask area of the background-subtracted, aligned mCherry image was used.<br />
9. Cell tracking was performed as follows: In addition to intensity readouts, the mask centroid (x, y)<sub>i,f</sub> was determined for each cell i in every frame f. To overcome between-frame jitter, between-frame shifts x and y were determined by cross-correlation analysis of consecutive CFP images that were downsampled 4-fold. Those shifts, x and y, were applied to centroids (x, y)<sub>i,f</sub> for establishing tracks between frame f and f+1. Cells i and j in two consecutive frames f and f+1 were connected if (x+x, y+y)<sub>i,f</sub>  is the nearest position to (x, y)<sub>j,f+1</sub> among all (x+x, y+y)<sub>f</sub>  and (x, y)<sub>j,f + 1</sub> is the nearest position to (x+x, y+y)<sub>i,f</sub> among all (x, y)<sub>f+1</sub>.<br />
10. Evaluation of cell death times was based on changes in the cell morphology or, when the IMS-RP MOMP reporter was co-expressed, on its intracellular distribution. Morphological variables indicative of cell rounding that were extracted for each cell in each frame were the area of the single-cell mask and an edge metric quantifying the contrast at the cell boundary: 4 line-scans of  I<sub>CFP</sub> rotated by 45 around the mask centroid were Gaussian-filtered, and the median of the corresponding 8 maximal intensity slopes in the 3-pixel proximity of the mask boundary were determined. Spread-out cells yielded shallow maximal slopes for this readout, whereas cell rounding resulted in a sharp increase in intensity slope at the mask boundary.<br /> 
11. To validate the use of ICRP for fate classification, IMS-RP was co-expressed with the initiator caspase FRET probe ICRP. To detect MOMP, the position of the nucleus was estimated by finding within a single-cell mask the position of peak intensity in CFP fluorescence after filtering I<sub>CFP</sub> with a wide Gaussian. This approach made use of the observation that ICRP is not excluded from the nucleus and that epifluorescence is integrated over the widest z-range in that place. With intact mitochondrial outer membranes, IMS-RP is excluded from the nuclear region. After MOMP, IMS-RP becomes ubiquitously distributed inside the cell, including the nuclear area. Sharp variations in all of these three variables were indicative of either cell division or apoptosis. If cells regained their original values in edge and area metrics within less than 4 hr, the event was classified as a division; otherwise it was classified as apoptosis. Loss of tracking within these 4 hr or incomplete recovery was treated as an undermined fate. All cells in the fields of view were analyzed, and only some were discarded when tracking was lost before the end of the experiment or before cell death or when an event was not properly classified as either a cell division or a cell death (see above). With this stringent approach, adjusting the cutoffs for fate classification did not affect the results qualitatively.<br />
12. The tracking approach was validated by two tests. First, in the conditions that had IMS-RP co-expressed, the agreement between the morphological assessment and calls based solely on IMS-RP was 80  2%, and the times of MOMP had a Pearsons correlation of r = 98.8  0.4 across 4 conditions with different doses of TRAIL. Second, among the cells that were classified as dying, we located the inflection point in the FRET ratio that indicates the sudden activation of effector caspases. To pinpoint this event, we monitored the discrepancy between the finite derivative of the FRET ratio and the derivative of the FRET ratio after smoothing over 7 frames. We defined the time of MOMP as 4 frames before this discrepancy exceeded twice its standard deviation.<br />
13. Trajectory fitting was conducted as follows: The average FRET ratio trajectory of all untreated cells was subtracted from the FRET ratio trajectory of each treated cell. Noise in the trajectories was then filtered using the MATLAB function <em>filtfilt</em> with a window size corresponding to 55 min (11 frames). For each trajectory, the minimal value of the FRET ratio was subtracted from the trajectory. The derivative of the FRET ratio was computed using finite differences and the trajectories filtered using the MATLAB function <em>filtfilt</em> with a window size corresponding to 55 min (11 frames).<br />
The fitted model is based on the equation:<br />
 <img src="/_static/dataset_images/20232_equation-1.PNG"/>
 <br />
The time of maximal value for the derivative of the FRET ratio () was used as the end point to fit the model to the FRET ratio using the MATLAB <em>fit</em>. The parameter <em>t<sub>0</sub></em> was constrained to be in the range <em>[30min,   30min]</em> and <em>k</em> to <em>[0, 0.01]</em>. In cases in which the fit to a trajectory was bad (r&sup2; <0.5), the fit was improved by ending the fitting on a secondary maximum; the fit with the best r&sup2; was kept. All fits were tested for significance using an F-test against a flat model <em>(FR(t)=cste)</em> with <em>p=0.05</em> as a cutoff. In cases in which the fit was not significantly better than the flat model, <em>k</em> was set to the value 10&#8315;&#8311; (minimum value observed for a fit). Cells that died early (at times less than 70 min.) and whose trajectory could not be fitted by the equation above were discarded from subsequent analysis on the assumption that they represented other forms of cell death or loss.<br />
14. The value  was determined by minimizing the error function:<br />
 <img src="/_static/dataset_images/20232_equation-2.PNG"/>
 <br />
where  is the Heavyside function: <img src="/_static/dataset_images/20232_equation-3.PNG"/><br />
15. In a twodimensional landscape of  and k computed from singlecell trajectories,  corresponds to a line defined as  = /2<em>k</em> + t<sub>0</sub>. The line separates cells by fate, with surviving cells falling to the left of the fate boundary (low k and/or short ) and dead cells to the right of the boundary (higher k and/or longer ). The accuracy of the boundary was evaluated for all experiments based on the average value <sub>T</sub> = 2.63  10&#8315;&sup3 trained on the data with doses of TRAIL equal to and above 10 ng/ml.

Date Updated:
2016-03-22

Date Retrieved from Center:
11/13/2015

################################################################
Metadata Information
################################################################

Metadata information regarding the entities used in the experiments is included in the accompanied metadata. A metadata file per entity category is included in the package. For example, the metadata for all the cell lines that were used in the dataset are included in the Cell_Lines_Metadata.txt file.
Descriptions for each metadata field can be found here: http://www.lincsproject.org/data/data-standards/
[/generic/datapointFile]
[/generic/reagents_studied]
