Summary - Sorting Kills

Ray Hester rhester at jaguar1.usouthal.edu
Thu Jun 17 13:45:27 EST 2004


Hi,

Below is a summary of responses to my question whether the sort process could have been
responsible for the apparent dead cells we observed in the post-sort analysis.

One obvious (though not to me at the time) aid would have been the addition of P.I. to
the post-sort sample to see if the cells that appeared dead by light scatter were, in
fact, dead.

Thanks for all the responses.

Ray Hester
Univ. of South Alabama
Mobile, AL

 ............................................
My guess is, since those small particles won't contribute to the signal of the events in
the sort gate, they're going along for the ride with sorted events.  If they had another
color, they could be eliminated (DNA dye?).

..........................................
Yes, sorting can kill cells.  90% viability is pretty good.

Something else to consider:  There is literature stating that GFP 
expression can also kill certain types of cells.  I think EGFP 
doesn't  kill cells (if memory serves me correctly).

..........................................
Based on your data you may be sorting dead cells to start with.  I
would recommend using propidium iodide to stain the dead cells and
sorting based on live, intact cells plus GFP positives.  This should
also help you with post sort analysis.	Also depending on the cell type
the sorter can actually be damaging the cells.	Some cells are more
fragile than others and can be lysed going through the sorter.	Other
cells are fine and sorting can actually improve the purity of the cells
with regard to dead cell contamination.

.................................................
It is possible you sorted dead cells and in your post sort you validated
this result. we have done many high speed sorts (HSS) and have tested for
dead cells using a variety of methods, annexin V +, PI, EB/AO etc. In all
cases there were always less then 1% of the cell population dead. In short
we did not find that HSS causes excessive cell death. Hence, I would
recommend you sort using a viability marker such as PI or EMA to effectively
gate out the dead cells. Also, I would not rely on the scatter gate to
determine dead cells but would rely on a post sort microscopic examine of
the cells stained with EB/AO in a hemocytometer.

....................................................
I often sort GFP cells on a DiVa at 35psi and 8000 cells per sec, mostly this is fine but
on a couple of occassions we have generated distinct populations of positive dead cells
not in the original sample. I find this does correlates with how soon after transfection
we sort the cells i.e. how fragile they are. Cells sorted a week after transfection are
fine, but one day after transfection can result in a lot more cell death.

............................................
That death percent is reasonable. I didn't look at your plots, but over
time cells will die both in culture and out as well as through the
sorting process. There are things you can do to minimize cell loss due
to viability, but given your percentage it seems like your process is
fine. If you let me or us all know the reason this has become an issue
perhaps we could help. 

...................................................
If its dead cells you should increase the FSC gain to get the whole dead  population
above the theshold.  Using a combination of forward and side scatter would demonstrate
better whether its debris or dead cells. You would do better to increase the FSC gain and
set a fairly tight gate on FSC and SSC. The  "dead cells" you see in the post sort sample
could be stuff that was just below theshold in your pre-sort sample and not de novo dead
cells at all. The rate at which you pass cells through the instrument isn't going to make
any difference to viability. Its the sheath pressure and nozzle size which will
contribute to shear forces that break cells. Increasing the sample rate may increase the
incidence of cells bumping into each other but thats all.

........................................................

The question is:
is the debris you see in the sorted sample leftover debris from the 
original sample? That is what I am inclined to think as from experience 
I never seem to be able to get rid 100% of the original debris/dead 
cells in a sample, the higher the original debris content the more you 
get in the sorted sample.

........................................................................................

Your images didn't transfer, so I can't comment on these.  However, on
reanalysis, I will often see a shift to lower channel numbers usually across
all parameters.  What was the pressure you used?  Second question - what are
you using for shealth fluid?  Third - are you collecting these cells in a
tube that contains media or are you dropping them directly into a tube?  I
find growth media with a little protein seems to aid with recovery.

You can get some shear if the nozzle is too small for the cell type.  I
sometimes use a 100 u tip and have went as large as 200 u.  I guess the
bottom line is - they did they did grow after sorting - so you must be doing
some things right.  

...................................................
I think this percentage of dead cells seems normal. Dead cells will lose 
GFP if it is cytoplasmicly expressed, so they may have been positive and 
then squirted lethally against the side of the tube. Also, aggregates of 
positive/negative can be mistakenly sorted as single positive, even with 
stringent clump detection/elimination.

For re-culture, the cell death is trivial, as the survivors usually grow.

........................................................
I think your gate for viable may not be what it seems.. Did you look at PI exclusion and
verify that these are viable pre and post sort in the gate you show FSC vs FL1?

Personally if I don't want to sort dead, I use PI exclusion.  IF I am not concerned  (
i.e. g=dead cells don't grow) then I use a scatter gate.  Let me caution that I have
found that traditional live dead by light scatter (SSC vs FSC ) does not pan out for
transfected or altered cell lines.  When I sort one of these I always use PI exclusion. I
have found that scatter gating is unusable on these samples.

Looking at the sort it looks pretty good as to giving you recovery that is localized
within the sort gate on reanalysis.

Did you bleach and rinse between you sorted and post sort reanalysis??
.................................................................
Just a thought here,  when I ran the Flow Core lab at _________
University, I found that people often did not filter their collection media
containing Fetal Calf Serum  The proteins in FCS tend to precipitate out
and form a very distinct pattern on FS vs. SSC.  I would suggest running
one of the collection tubes before even any acquisition/sorting to see if
there is "stuff" in that tube.	It took me a while to figure that out when
I first started and on PI wanted to blame me for not knowing what I was
doing when in fact it wasn't me at all.
Also, in my experience, I found that GFP tends to photobleach fairly
quickly in some cells.	I typically used 50mW of 488 power which gave fine
results.

................................................................
Your viability certainly seems reasonable for the conditions you 
mentioned.  We usually use 100 mW of UV.  I'm also guessing that your 
nozzle tip is rather small.  We almost never (possibly never, ever) 
use 70 uM tips anymore.  If you are looking for GFP I assume you have 
a cell line.  We never run cell lines with a nozzle below 80 uM and 
almost always find 90 uM or even 100 (sometimes 130) is much more 
appropriate.  If you have a FACSVantage SE w/ accudrop you can watch 
the side-streams (through the accudrop camera) and see whether your 
nozzle size is appropriate or not.  If you see fanning of the 
side-streams, your nozzle is not large enough.	You should get one 
nice distinct side-stream during the sort.

In addition, we like to add PI to sort checks.	This way you should 
know for sure whether the extraneous cells are dead (or debris, which 
is what it looks like in your case).

(It really doesn't look bad, you just have a lot of debris.)

I hope this helps.







More information about the Cytometry mailing list