Vantage vs Calibur 'sensitivity' and MNR counts?

Howard Shapiro hms at shapirolab.com
Thu Jul 14 22:59:22 EST 2005


Ray Hester wrote:

>If you were to compare the capability of a FACSCalibur (cuvette flow cell 
>instrument) vs a FACSVantage (stream-in-air) to distinguish between dim 
>and bright fluorescence peaks, would the peaks actually be farther apart 
>with the cuvette flow cell instrument, i.e., if the Vantage has unstained 
>cells at channel 2 (log scale) and PI stained cells (this isn't a question 
>of ploidy - we are looking at micronucleated retics) at channel 10 (log 
>scale), will the cuvette-style instruments, using the same sample, have 
>these two peaks farther apart?  For example, at channels 2 and 25 on the 
>same log scale?

The ability of a cytometer to discriminate dim peaks stained with a given 
fluorochrome is a function of both the detection quantum efficiency (Q) and 
the background signal level (B). Q is determined by a number of things, and 
is strongly influenced by the fluorochrome spectrum, the PMT quantum 
efficiency, the fluorescence light collection efficiency, and the laser 
power used (bleaching and saturation may alter the dependence on laser 
power). Higher Q decreases the CVs of peaks, and even when the peak 
locations are in the same places, it is easier to discriminate peaks when 
they are narrower, i.e., the brightest cells in the dimmer peak are farther 
away from the dimmest cells in the brighter peak. Higher B pushes the dim 
peak up toward the bright one, making discrimination more difficult. The 
light collection efficiency of a FACSCalibur may be as much as six times as 
high as that of a stream-in-air system, but that advantage is overcome if 
the laser power in the stream-in-air system is six times that in the 
Calibur. You'd basically have to run the same cells on both instruments, 
with each optimally aligned, to find which was better.

Having said that, I should add that PI is not the best choice of dye for 
measuring micronuclei; it is typically used at a relatively high 
concentration, and only enhances fluorescence by a factor of about 30 on 
binding to double-stranded DNA or RNA. Since PI is not DNA-selective, you 
will have to use RNAse on your cells if you want to detect only DNA. You 
would be much better off with something like a SYTO dye from Molecular 
Probes, or with thiazole orange (which is structurally similar to at least 
some SYTO dyes). These dyes enhance fluorescence on the order of a thousand 
times on binding to nucleic acid, and are almost nonfluorescent in 
solution. Therefore, you should get lower background fluorescence from dye 
in the stream, meaning B should be lower for the SYTO dyes, etc., all 
things being equal, and you should therefore see better separation of 
stained and unstained peaks.


>Also, we have an investigator quantitating micronucleated retics (MNR) 
>from blood samples using a three color assay: FITC conjugated anti CD71 
>(positive for retics), r-PE conjugated anti CD 61 (to permit platelets to 
>be gated out), and propidium iodide.  In healthy individuals MNR are 
>present at about 200 per 20,000,000 rbc.  Any thoughts on how easy or 
>difficult it might be to detect changes (due to some particular 
>experimental condition) of appox. 10 per cent in the numbers of these 
>MNR?  That is, if you divide a blood sample from a normal individual into 
>three parts and analyze 20,000,000 rbc from each sample, how much 
>variation can you expect?  Could the variation in sampling be as high as ± 
>10%, i.e., the results might be 180, 200, and 220 MNR for the three 
>aliquots from the same sample?
>
>They are examing MNR-enumeration stains and protocols other than the one 
>described, but for the moment they need answers to these questions.

If the cells of interest (MNR) are present at a frequency of 200 per 
20,000,000 rbc, and you analyze 20,000,000 rbc, you can expect to count 200 
MNR; Poisson statistics dictate that the minimum CV of the measurement is 
(100*14.14)/200, or about 7.1%. If you wanted to reliably detect a 10% 
difference in frequency of occurrence of MNR, you'd need to lower the 
measurement CV, probably to 3% or better. This would require counting over 
1,000 MNR per sample, meaning that you'd have to analyze 100,000,000 (10^8) 
rbc in each sample. At 10,000 cells a second, that means between two and 
three hours per tube.

The above assumes you are triggering on scatter, and therefore must process 
signals from each RBC. If you use a nucleic acid stain that clearly 
discriminates MNR from rbc (see discussion above), you can trigger on the 
fluorescence of this dye, essentially ignoring the rbc, while still 
measuring the antibody fluorescence, and use counting beads to count the 
MNR per unit volume of blood. The rbc count per unit volume can be 
determined in a separate aliquot using an optical or impedance counter, and 
the frequency of MNR among rbc can be calculated from the two counts. RBC 
counts in blood are typically on the order of 5,000,000/uL, and a cytometer 
can usually process 1 uL of stained specimen/second. If you trigger on 
fluorescence, assuming the staining procedure dilutes by a factor of 20, 
you still have 250,000 rbc/uL in the stained specimen, reducing the time 
needed to count 1,000 MNR to minutes instead of hours.

-Howard








More information about the Cytometry mailing list