 |
 |
 |
 |
TD> Don't be bashful folks, join in.
Okay. CAUTION: I found I had a lot more to say than I thought I did, and
really didn't know when to stop. This message may be overly long and boring. :)
CC> [paraphrased] Higher sample concentration leads to a smaller core stream
CC> diameter when threshold rate remains constant, thus improving sort
CC> purity and yield, and allows successful higher speed sorting.
TD> [paraphrased] Dead time (system) aborts degrade sort purity at high
TD> threshold rates. Furthermore, this effect is influenced by the pre sort
TD> purity; lower pre sort purities lead to lower post sort purities.
I agree with both of these points. You certainly don't want to use a sample
concentration of a million cells per ml for higher speed sorting (3000 to
8000 cells/second threshold rate; not to be confused with true high speed
sorting, in which threshold rates start at about 10,000 cells/second and go
up from there). Also, you should be aware of the effects of system aborts
on sort purities at higher speeds; if you can't accept post sort purities
less than 99%, then higher speed sorting isn't for you. (Unless you sort
beads; see below). :)
CC> I have no numbers for recovery [...]
TD> A large number of the [target] cells [went] into the waste tank.
TD> Yield (number of cells in the tube you want divided by the number of
TD> cells you get) also decreases dramatically at high rates.
BH> The number most often missing or under-emphasized in claims of high
BH> speed [sorting] is yield. [...] The yield drops even more dramatically
BH> when more than one [time-separated laser] beam is used.
I use a slightly different definition for recovery (see below), so to me,
recovery and yield are two different animals. I think TD's definition of
yield is the same as mine (it's the reciprical, but I think that's a typo).
If you want to improve yield on a one-way higher speed sort, consider using
the "unsorted right" option. It catches all the target cells that were
system aborts (not sure about sort aborts). You still lose cells to waste,
but I prefer unsorted right to centrifuging the 7.6L waste tank to pellet
the waste cells, then resuspend to a higher concentration. :) (The
concentration in the sort collection tubes is enough for a "slow" 2000
cells/second threshold sort. The waste is a little too dilute.) I've tried
the unsorted right option twice now, and (surprise!) the post sort purity
for unsorted right cells (33%, 35%) is very close to the pre sort purity
(27%, 45%), or it is very close to 1/3rd; I'd be interested to find out
what other people get with different pre sort purities. I haven't tried
dual-laser sorting yet, so I can't comment on Brian's statement from direct
experience; however, it makes sense, since the dead time is increased. In
addition to being missing and under-emphasized, yield is also most often
undefined, and difficult to compare from source to source.
CC> High cell concentrations may cause other problems (clumping/clogging) [...]
I find that they like to stick around in the sample tubing, and to the
walls of the sample test tube. In fact, simply washing the sample test tube
with a small amount of PBS (or whatever) will give you more cells to sort.
Clumping/clogging shouldn't be a big problem if the sample preparation is
clean: filter the sample before putting the tube on the cytometer if
necessary. Also, don't use the inline filter that you can insert at the tip
of the sample pickup tube; I find that the filter clogs up pretty quickly
with more highly concentrated samples (and your threshold rate drops from
7000 cells/second to 10 cells/second really fast). :) The inline filter
seems to work pretty well for other purposes, but pre filtering samples
works even better (with a nylon mesh or gauze or whatever).
Here are some data that I collected on a FACS Vantage, most likely with the
standard sort nozzle (I don't remember when our MacroSORT was installed). All
sorts were done in NORMAL-R mode, with 2 drops, a frequency of about 25 kHz,
and phase gating off. Here are MY definitions for a few key terms:
%Purity: percent target cells, based on a representative post sort
sample
%Recovery: percent of (microscope count)/(cytometer count), based on a
short (20-50 beads) test sort onto a microscope slide. This accounts for
cell loss due to poor sidestream formation (fanning).
%Yield: percent of (hemacytometer post sort count)/(theoretical yield),
based on a longer (>10^6 collected) sort into test tube(s). Theoretical yield
is given by (hemacytometer pre sort count)*(percent pre sort purity)*100.
This accounts for cell loss due to aborted events (either system [dead time]
aborts or sort [sort window coincidence] aborts).
Sort I
Sort population: 23% (murine lymph node; indirect CD8-FITC labeling)
Threshold Rate %Purity
-------------- -------
3000 99.4
4000 99.5
5000 98.3
6000 97.7
Sort II
Sort population: ? (beads)
Threshold Rate %Recovery
-------------- ---------
1000 100
2000 90
3000 69* *probably a fluke; only based on one measurement
4000 82 (average of two measurements)
5000 72 (average of two measurements)
7000 61
Sort III
Sort population: 68% (PI-labeled beads "diluted" with unlabeled beads)
Threshold Rate %Purity
-------------- -------
2000 99.6
4000 99.6
8000 99.1
I haven't done a yield study (yet). The differences between Sort I and Sort
III are probably mostly due to the simple fact (made popular around here by
MAK): cells are cells, and beads are beads. I used the recovery data from
Sort II to determine my optimal threshold rate for sorting. The motivation
is to collect as many cells as possible in as little time as possible,
while maintaining acceptible purity (sound familiar?). The argument goes
like this: If I sort a P% population (P is some number, [0-100]) at a
threshold of 2000 cells/second, then I should recover p*2000*0.9=p*1800
cells/second (p=P/100). If I sort the same P% population at 5000/sec, then
I recover p*5000*0.7=p*3500 cells/second; either double what I can get at
2000/sec, or the same in half the time. Yes, I will lose more cells (not
important in MY situation), but I can make an 8 hour sort take only 4 hours.
I did a least-squares fit of my recovery data to an exponential function
(because it just felt right) :) and came up with the maximum "recovery
speed" of about p*4000/sec at 7300/sec threshold (it would help if I had
more data points, but that takes time!). :) Again, I lose about as many
cells as I sort, but then again I start with on the order of 10^8 cells
(let's see, 10^8 cells, at 1000/sec, takes almost 28 hours; 10^8 cells at
7300/sec takes less than 4 hours). Furthermore, 8 hours(=28800 sec)*p*1000*1=
p*2.9*10^7; and 4 hours(=14400 sec)*p*7300*.5=p*5.3*10^7. Not only have I
halved the time spent sorting (i.e., falling asleep watching the droplet
breakoff and looking out for plugs) :), but I've almost doubled my yield!
(This is misleading, though, because the yield at 1000/sec threshold was
not optimal [i.e., I only sorted 8 out of 28 hours]). Here are some "real
world" results from sorting murine lymph node cells for CD8-FITC positives
(the %Purity before sorting varies due to a variance of tumor present):
%Purity (before) %Purity (after)
---------------- ---------------
45 97
45 97
27 96
30 97
45 98
16 92 (nobody's perfect) :)
I apologize for not posting %Yield; I didn't keep track of those numbers. :(
Some sorts may have been run at less than 7300/sec threshold due to less cells
being available to sort (less than the standard 1*10^8), but I didn't keep
records on this, either. (Can you say INFORMAL results?) :) These sorts
were moved to a MACS system, as soon as we could get our hands on one. :)
Disclaimers: Don't base your sorting decisions on my work (unless you plan
on sorting murine CD8 cells on the exact same FACS Vantage that I used). :)
Find out what YOUR sorter's capabilities are, for the kinds of sorting
experiments that YOU want to run. Use cells instead of beads whenever
possible. Repeat your experiments (like I plan too) more than once.
Sometimes cytometers just have "bad days" when it comes to sorting. If you
can get away with a column, a magnetic bead sorter, or a FACSort, then by
all means, use these other methods and give your cytometer operator a
break. :) As Howard Shapiro once wrote, "...whenever I get the urge to sort
cells, I try to lie down until it passes off."
Correction: In a previous message that I did not send directly to this list
(but managed to get posted here twice by others), I incorrectly stated that
I used an 8000 cells/second threshold rate; the correct "target" rate that
I use is 7300/sec. Also, I incorrectly stated a post sort purity of 98% or
better for this speed; the correct purities are listed above. Finally, any
information related to our EPICS 753 should be taken with a grain of salt;
our particular 753 had some unique problems that I wouldn't expect to find
on other 753s. I overstated the sorting performance of the FACS Vantage in
that previous post because I was in a hurry, and didn't check my notes (can
you ever forgive me?). :)
Curiosity: While digging through some old stacks of paper, I came across
"The Art of Fluorescent Activated Cell Sorting", from the Becton Dickinson
Research Notes series, produced by their Technical Publications Department
(part number 23-2137-00). It includes a study done in 1991 that compared
the purity and yield (defined slightly differently) of the different sort
modes of a FACStarPLUS. The curious part is the footnote that indicates
that a high threshold rate was used; however, the actual rate was not
specified.
Wanted: The high speed sort option that was demonstrated at ISAC XVII. I
called last week and was disappointed to find out that it is not a product
yet. If I recall correctly, it was installed on a Vantage, and sorting at
30,000/sec threshold! (Someone please correct me if I am wrong.)
Acknowledgements: Mark Cameron, did you ever think your simple post would
ever spur such a thread? :) Thanks to Mike Salmon for providing the
appropriate skepticism that lead to more information being shared, and by
echoing the use of magnetic beads over FACS when appropriate. Thanks to
Mario Roederer, "Frank", and others (?) for their discussion, as well. I'm
really enjoying this thread, and look forward to contributions from other
sources (I know you're out there!). :) Doesn't Cytomation deserve a pat on
the back for providing a commercial cytometer capable of true high speed
sorting? Do we have enough lasers and detectors on our commercial sorters
now that we can concentrate on improving the sorting? (I specifically
mention "commercial" so as not to slight the zappers and HiSSers and what
not). My apologies to anyone I may have inadvertently left out.
/\/\/\_ Eric Van Buren, vanburen%flovax.dnet@rocdec.roc.wayne.edu
\ \ \ Immunology & Microbiology
\_^_/ Wayne State University, Detroit, Michigan, USA
|
|
|
|