ImageStream: Ten Technical Questions and Answers
1. Even the best CCD camera is orders of magnitude less sensitive than a PMT. How does the ImageStream manage to collect enough photons to approach the sensitivity of a flow cytometer?
Several factors are responsible for the remarkable sensitivity and dynamic range. The most important is time delay integration (TDI). TDI was originally designed to increase the sensitivity of cameras used for scanning x-ray imaging applications. The application to image moving cells is novel and depends on a remarkable velocity-sensing technology. Briefly, rather than taking a single image of a cell as it passes before the camera, photocharges are successively shifted across the length of the chip as the cell traverses the 512 pixels of the cameras field. Conceptually, TDI is equivalent to panning on the moving cell to hold it in the cameras field of view. Since each pixel maps to 0.5 microns of the imaged cell and the flow rate is approximately 25 mm/second, each image is captured over a period of about 10msec. Successful application of TDI to cytometry requires knowing the location of a cell at any given time with submicron resolution. This is accomplished with a separate infrared laser, which, in conjunction with 2 PMTs, measures object velocity and focus in real time for closed-loop process control.
Other factors that increase the amount of collected light include:
- Slow speed. The flow rate, 30mm/sec, is approximately 1/30 that of a conventional flow cytometer, and 1/300 that of a high-speed instrument;
- High laser power. The ImageStream uses a 200 mw 488nm solid state laser for excitation (Cf. 15-20 mw in most cuvette-based flow cytometers);
- A sensitive 6-channel CCD TDI camera specifically designed for this application.
2. Doesn't the cell rotate or wobble as it crosses the camera field?
No. We were fully convinced of the wonders of hydrodynamic focusing when we collected razor sharp images of red blood cells.
3. Why doesn't the brightfield lamp interfere with the measurement of emitted light?
A filter wheel allows a selectable filter to be interposed. Violet (450 nm) light is used when all 4 fluorescent probes are used. Near red (630 nm) gives the best brightfield resolution.
4. How does the resolution compare to conventional fluorescence microscopy?
The numeric aperture of the objective is 0.75. When perfectly focused, the images are comparable to that of conventional fluorescence microscopy using a 40X objective. In our hands, focus is acceptable for the great majority of events (sharp enough for calculation of cellular features), and really tight about 20% of the time.
5. What about the depth of field?
Comparable to conventional microscopy with a 40X objective and a long working distance condenser.
6. Has Amnis done its homework and taken advantage of recent advances in flow cytometry automated color compensation?
IDEAS analysis software provides automated calculation of compensation matrices using the inverted matrix method. Unlike any flow cytometry software that I have seen, the user can visually gate out autofluorescent events from bivariate scatter plots prior to calculation of spillover coefficients, improving compensation.
7. What about data display?
All parameters can be displayed on linear or "hyperlog" scales, similar to those used in flow cytometry. The hyperlog function is linear as it passes through zero and approaches logarithmic within the first decade (or within a user-defined range). Boolean gates of a variety of shapes can be placed on populations. Gated regions can be "color-evented." IDEAS has a long way to go before it has all of the functionality of the best commercial flow cytometry software, but it has tackled the most important issues first.
8. Can flow plots be correlated with photographic images of cells?
Yes! Clicking on individual points in a bivariate scatter plot brings up the actual image of the cell. When a gate is created around a population, a gallery of all of the images within that gate can be displayed, the equivalent of "virtual cell sorting."
9. Six digital images are collected and stored for each cell. How long does it take to acquire and display the data?
A typical blood mononuclear cell maps to about 400 pixels. There are 1024 channels of resolution per pixel and 6 images per cell. A great deal of processing is performed by the INSPIRE acquisition software in real time by the built-in dual Xeon processor server running 4 simultaneous threads. Instrument control is managed by a separate internal Linux box. Since color compensation must be performed on each individual pixel, this is done post-acquisition using IDEAS software. This computation-intensive task is best done in batch mode on a separate workstation. Both the size of the data files and the complexity of the data guarantee that exploratory analysis will be time consuming. Once the analytical strategy has been developed for a particular test, analysis templates and batch processing take some of the pain out of analysis. We have requested a state-of-the-art Windows-based workstation to free the ImageStream's dedicated computer from the task of data analysis.
10. How large are the data files?
This imaging system creates 100-200 composite images per second. A 50,000-event raw image file (RIF) occupies just over 1 GB of file space, and this requirement is doubled when a compensated image file (CIF) is created for analysis. We are currently establishing a separate server with 3.6TB of RAID5 disk space on a gigabit network to support the ImageStream.
HOW LONG WILL IT TAKE ME TO SORT?
Calculation for the 70 micron nozzle (lymphs, PBMC, bone marrow, yeast):
Nominal maximum throughput (70µm) = 20-25,000 events/second ~72 - 90(10)6 events/hour
Net throughput (corrected for normal recovery/yield estimates) = 70% ~50–63(10)6 events/hour
Estimated throughput = 50(10)6cells/hour (50 million cells per hour that can be run through the machine)
For most cell lines, a 100 micron nozzle is needed, which reduces throughput by at least 70% – adjust calculation accordingly.
Collection rate, 70µm (maximum) = (viability) x (expression) x 50(10)6 events/hour
Attempting to sort out a sub-population (5%) from your cell line.
The number of cells you can provide is unlimited, and the viability is 90%.
A pure, sorted population of at least 5(10)6 cells needs to be recovered. Given the estimated sorting capabilities (see the calculation above), estimate sorting time as follows:
Collection rate (maximum) = 0.9 (viability 90%) x 0.05 (expression 5%) x 50(10)6 events/hour
= (0.9) x (0.05) x 50 million cells/hour = (.045) x 50(10)6 = 2.25(10)6 = 2,250,000 per hour.
100 micron nozzle reduces by at least 70% - adjust calculation accordingly.
Recovery of 5(10)6 total =[5(10)6 Ã· 2.25(10)6] = 2.22 hours of sorting, at minimum. Add an additional 45 minutes to sorting time to set up the machine.
However, please assume that something might go wrong, and you will need twice the number of cells, and it will take twice as long.
REMEMBER: 4 SEPARATE POPULATIONS CAN BE COLLECTED AT THE SAME TIME.
FOR ANY SORT, YOU WILL NEED TO BRING THE FOLLOWING:
- Negative control
- Compensation controls if needed
- Sort sample(s):
Bring controls and samples in 12 x 75 mm sterile; snap-cap; POLYPROPYLENE tubes, with no more than 5% serum.
The control(s) only needs to be 300 uL final volume, with at least 300,000 cells. Sample should be re-suspended at a concentration on 30-40(10)6 cells/ml.
- Collection tubes (4 separate populations can be collected at the same time):
For each tube for collecting sorted cells, fill the tube ¼ full with 50% serum/media. You may use either 12x75mm tubes or 15 ml tubes.
(12x75mm tubes are needed for 4-way sorting - fill the tube ¼ full with 50% serum/media.)
Cells to be collected exit the machine at a concentration of ~1,000,000 cells/ml using the 70 micron nozzle; 500,000 cell/ml using the 100 micron nozzle. You can now calculate how many collection tubes you will need for each population collected. (Example: For 70µm nozzle, you place 4ml media/serum in the bottom of a 15ml tube, leaving 11 ml empty volume for collecting cells = 11ml x 1,000,000 cells/ml = 11(10)6that may be collected in that tube.)
FOLLOWING THE SORT
For maximum recovery, fill collection tubes to the top and let sit 30 minutes before spinning them down.