I disagree that: "The only argument is loss of fluorescence during the sort." If you have two populations that are close enough together the populations
overlap. If there is, for instance, 50% overlap then the best you can
hope for is a 50% pure population after sorting. I think this is very
likely what is happening here. The real problem isn't the sorter, it's
the prep. You need to get better separation between the neg and pos
populations. There are amplification kits available. Here's another
example to make my point. In another life I had a client that liked to
sort a narrow band of a very broad GFP+ peak. Say we sorted a band that
was between 800 and 1000 MFI. He expected a population upon reanalysis to
show up "right between the goal posts" on a single color plot. It doesn't
work that way. Upon reanalysis you get a normal Gaussian distribution and
most of the cells are between the sort markers but you will get spillover
on BOTH sides of the gate. Look at it this way: If you ran the SAME
exact cell through the cytometer 10,000 times, you wouldn't get a line on
the histogram with all events stacked in the same channel. You would get
a normal Gaussian distribution predictable by the (if) known uncertainties
in the measurement. Think of your calibration particles. They are very
uniform in their spectral properties, but they don't all give the same
exact fluorescence, although the width of the peak (CV) is much smaller
because of this uniformity. One more thing and then I'll desist. Try
seeing what happens if you sort half of a normal distribution. You might
even compare the high half and the low half. Upon reanalysis these should
look very much like what you started with before sorting. Oh, when I say
normal, I mean that it can't be a broad, flat, 3 decade GFP-like peak
which is really a string of overlapping peaks.
David McFarland
GlaxoSmithKline
----- Forwarded by David C McFarland/DEV/PHRD/SB_PLC on 02-Nov-2001 09:00
-----
Volker_Eckstein@med.uni-heidelberg.de
31-Oct-2001 11:10
To: cytometry
cc:
Subject: AW: Sort purity
Hi Simon,
I have had the same problem you described and I discussed it with BD's
sorter specialists.
Various points have to be taken into account:
Weak fluorescence can result from low antigen density, bad
protein/fluorochrom ratio or insufficient amount of used antibody. Even
the
antigen distribution or the fluorochromes distribution (patches) when
labelling cell membranes can give slightly different signals when passing
the analyzing point. The fluorochrome itself (tandemconjugates or not) can
give weaker or stronger signals.
If a sort gate is set all cells or particles which fit these criteria will
be sorted if they can be sorted (no coincidence). When passing the laser
light, weak fluorecent cells may loose fluorescence by bleaching.
Therefore
reanalyses often show the sorted cells not fit the sort gate again. And if
the desired population was close to the negative unwanted ones it is
possible that during reanalyses the fluorescence of the sorted cells fall
into the negatives.
The only argument is loss of fluorescence during the sort.
RegardsVolker
Volker Eckstein PhD
Dept. of Internal Medicine V
Medical School of the University
University of Heidelberg
Heidelberg - GERMANY
volker_eckstein@med.uni-heidelberg.de
-----Ursprüngliche Nachricht-----
Von: Simon Monard [ mailto:smonard@trudeauinstitute.org
<mailto:smonard@trudeauinstitute.org> ]
Gesendet am: Montag, 29. Oktober 2001 20:22
An: Cytometry Mailing List
Betreff: Sort purity
Hi folks.
Can anyone point me to a discussion of ways to estimate the purity of a
sorted population
of cells. This is clearly easy when you are sorting very bright cells from
negatives
but more problematic when sorting cells from a "shoulder". I remember
seeing
a nice
discussion on this subject but cannot remember where. I've been sorting
some
very
weakly positive cells from a population, these cells post sort overlap the
negative
cells quite a bit. My customer seems unhappy about that.
Thanks
Simon Monard
FACS Lab Manager
Trudeau Institute
Saranac Lake
NY12983
Ph 518 891 3080 X352
This archive was generated by hypermail 2b29 : Wed Apr 03 2002 - 11:58:00 EST