Re: Validation of Enumeration Techniques

From: Howard Shapiro (hms@shapirolab.com)
Date: Tue Dec 11 2001 - 19:01:30 EST


R. Thomas Leonard wrote:


>Question from a novice:
>
>There are several papers which describe ways in which cell concentration
>can be determined using FCM. When quantifying cells, one must identify
>"events" which are assumed to be the cells of interest.      How does one
>validate, quantitatively, that the events are in fact the cells of
>interest. For example, if we count 1x10E6 bacteria, how do we know that
>there aren't  9x10E5 bacteria and 1x10E5 small fungi that scatter and
>fluoresce in a similar fashion?
>
>My assumption is that one would need to program the cytometer to sort X
>events believed to be cells of interest. We would then need to quantify the
>cells using fluorescence microscopy.  Any suggestions?

Why hold the flow cytometer to a higher standard than the impedance
(Coulter) counter?  The latter will count as cells anything with a signal
set above threshold; if you try to do a "white blood cell count" by putting
in diluted whole blood treated with an agent that nominally lyses red cells
but not white cells, you will get a falsely high count if some of the red
cells are not lysed and a falsely low count if all of the red cells and
some of the white cells are lysed.  You may have better luck discriminating
red from white cells by eye using a hemocytometer, but you are then faced
with imprecision due to dilution errors and counting statistics.  And, if
you had a lot of giant platelets in the blood, you might count those as
white cells.

So, the first principle of cell counting is to have a way to tell the cells
you want to count from everything else.  Or almost everything else; even if
you count 10,000 cells, your precision is limited to 1% by counting
statistics, which means it doesn't matter that much if you count ten things
that aren't the cells you want in with the other 9,990 cells you do want.

While sorting is a good way to confirm that you are counting what you think
you are counting, it isn't absolutely necessary.  If you see clusters in
flow data that represent roughly the same percentages of a mixed population
as you can count by eye, you can generally assume that each cluster
represents one of the visually distinct elements of the population.

With both eukaryotic and bacterial cells, you do have the problem that an
"event" may represent either one cell or an aggregate of two or more.  If
you plate bacteria on agar, to count "colony forming units", you have the
same problem; unless you scan the whole plate at high magnification to
start with (and nobody does), you're not sure whether a colony arose from
one organism or several stuck together.

 From a philosophical point of view, you can only be sure that your cluster
of bacteria is not contaminated by "stealth fungi" if you sort everything
every time.  Nobody does that, either.

What you do need to worry about when trying to do absolute counts with most
fluorescence flow cytometers is figuring out the volume of sample you have
analyzed in the process of counting however many cells you count; most of
these instruments do not deliver the sample with a volumetric pump.  The
commonest way to calculate volume is to add beads at a known concentration
to the sample; you first need to count the beads with an impedance counter
or hemocytometer, etc. etc. - or pay for known numbers pre-aliquotted by
the manufacturer.
But it does work.

-Howard



This archive was generated by hypermail 2b29 : Sun Jan 05 2003 - 19:01:43 EST