Re: CFSE

From: Andrew D. Wells (adwells@mail.MED.UPENN.EDU)
Date: Wed Nov 21 2001 - 13:40:32 EST


> Alice Given wrote:
>
> One additional warning:  we have found that CFSE at slightly high
> concentrations (5
> uM) inhibits antigen specific T-cell proliferation.  The PKH dyes do not
> appear to
> have this effect -- perhaps because they are labeling lipids rather than
> proteins.
> ConA-induced proliferation did not appear to be as sensitive to CFSE as did
> tetanus-induced proliferation.

We have extensive experience using CFSE to visualize T cell division in
vitro and in vivo in response to mitogens (J. Clin. Invest. 100:3173, 1997),
nominal antigens (J. Immunol. 162:5212, 1999) and alloantigens (J. Immunol.
166:973, 2001).  By comparing the function of labeled and unlabeled T cells
in functional assays that do not depend on the dye for a readout (e.g., dye
exclusion, activation marker expression, DNA content, BrdU or H3-TdR
incorporation, Ca2+ flux and kinase activation - see J. Immunol. 165:2432,
2000 and J. Clin. Invest 108:895, 2001), we have become quite confident over
the years that CFSE at concentrations as high as 5 uM (we most commonly use
2 uM during the labeling) does <not> negatively impact on the function of
<naïve> T cells.

We have, however, noted that at least under some circumstances, labeling
<recently activated> T cells with CFSE can inhibit their cell cycle
progression.  This, as Alice points out, is most likely because CFSE
covalently attaches to proteins.  When stimulated, naïve T cells must
synthesize de novo much of the proteins required to progress through the
cell cycle and gain the effector function of activated T cells - the initial
CFSE labeling would therefore not affect the activity of this de novo
machinery.  However, exposing activated T cells to a covalent protein
modifier most likely poisons the new machinery that these cells have
synthesized on now depend on for their function.  This, therefore, is a very
important consideration in planning an experiment using a cell labeling
approach such as CFSE.

As for the relative utility of pkh vs. CFSE dyes, I think there are some
important distinctions between the two that should be made clear to those
who do not use these dyes regularly.  It is my opinion that the pkh dyes
simply do not approach the functionality of CFSE for measuring division in
either in vivo or in vitro assays (as long as one is starting with a naïve
or resting cell population).  While I have no particular personal interest
in which dye investigators choose to use (i.e., I do not hold stock in
Molecular Probes, and I do not hold a patent on any technique based on the
dyes), I do have experience with both dyes, and my opinion is based on three
main aspects:  1) the intensity of fluorescence, 2) the stability of the dye
over time, and 3) the homogeneity of labeling.

CFSE labels more intensely than pkh, and the half-life of CFSE independent
of division is longer - the half-life of pkh26 for me in mouse T cells in
culture is less than a day, although it may be greater in human T cells.
For in vivo tracking and proliferation analysis, CFSE performs much better
than the pkh dyes, especially in any experiment that lasts more than ~5
days.

But by far the most notable difference between the two is in the third
aspect mentioned above, homogeneity of labeling.  Because the starting cell
population can be labeled with CFSE such that the intensity of the vast
majority of cells varies less than ~50%, each divided population,
representing 2^n-fold diminutions in the original intensity, can be
visualized as distinct populations and a logical gating strategy can be used
to mark each population.  The number of events under each peak can be
directly counted, and used to model the response.  Indeed, we first reported
in 1997 on the use of this simple modeling technique to determine precursor
frequencies and proliferative capacities from complex cell mixtures without
the need for limiting dilution (J. Clin. Invest. 100:3173, 1997).

However, pkh-labeled cells do not generate distinct division populations,
and the data must be analyzed using deconvolution software (e.g., ModFit).
This is not necessary when using CFSE, because the data are not convoluted
in the first place.  I have spent some time using CFSE data to test some of
the proliferation algorithms currently available, including the best one,
which has been developed by Verity.  In my opinion, none are particularly
good yet at modeling proliferation dye data.  For instance, in may hands,
roughly 2/3 of the CFSE distributions of labeled T cells stimulated under
various conditions fail to generate a good fit when analyzed using ModFit's
current proliferation module (and FloJo's module is significantly less
successful at generating good fits).  Remember, CFSE shows you the 'real'
distribution of the division populations, so if the model doesn't generate a
fit, then it's the model that is wrong.  The reason for the discrepancy
between the actual and predicted distributions is what the pro's in this
field constantly remind us of in this user group - that the detectors are
not perfectly logarithmic.  All the modeling algorithms are currently stuck
with the assumption that the log amp is perfect, and while this is not a
problem for DNA content data, which spans only 2- to 4-fold along the amp,
it is the reason for the failure to fit a significant proportion of
proliferation data.  According to Verity, this is <the> major hurdle in
designing algorithms to model proliferation data that span several decades
along the logamp.

So, when using a dye that does not show you the actual distribution of cell
divisions within your population of interest, one must realize that
somewhere around two thirds of the time your modeling software is giving you
inaccurate numbers.  I realize that this assertion may be somewhat
inflammatory for those investigators that rely on deconvolution software for
the analysis of their proliferation data, but I think it is true, and it
represents a serious limitation to the interpretation of pkh-determined
precursor frequencies, especially low frequencies.

So I think there are important uses as well as caveats for both dyes, and
further discussion within this forum can be of use to those who are trying
to decide which of the many techniques out their for measuring cell cycle
progression are the most appropriate for their particular needs.

-Andrew


Andrew D. Wells, Ph.D.
Department of Medicine
University of Pennsylvania
728 Clinical Research Building
415 Curie Boulevard
Philadelphia, PA  19104

(215) 573-1840 [office]
(215) 898-1951 [laboratory]
(215) 573-2880 [FAX]

adwells@mail.med.upenn.edu



This archive was generated by hypermail 2b29 : Sun Jan 05 2003 - 19:01:40 EST