Isotype controls (my annual 3 cents worth on this topic)

From: Mario Roederer (roederer@drmr.com)
Date: Thu Nov 01 2001 - 19:58:47 EST


Actually, isotype controls are not a particularly good control.  They
are rarely matched: the F/P ratio is not the same, and how do you
know if you are using them at exactly the same concentration as the
reagent of choice?  If you don't know that the F/P ratio is exactly
the same, and if you don't know if you are using it at exactly the
same concentration as your antibody reagent, then it isn't the right
control.

Indeed, since each one of your commercial reagents is titrated by the
manufacturer to give optimal signal to background, each one is sold
at a different concentration of antibody.  Have you contacted the
manufacturers to determine the bottled concentration of each reagent
you use, so that you can use the appropriate concentration of your
isotype control?  And then use a different isotype concentration as
the control for each reagent in your various panels?  If the answer
to either of these questions is "no", then how can you assert that
your isotype control actually gives you the correct amount of
background binding in your experiment?  i.e., your "isotype control"
does no more than let you that there may actually be some background
binding, but doesn't give you the ability to estimate how much.

In fact, I've known people to "titrate" their isotype controls to get
background binding that is less than what they think their positive
should be.  Hmm.

In another way, isotype controls are rarely used properly:  most
people do a single sample that has all isotype controls in all
channels.  This doesn't help!  One must use a control for which cells
are stained with all reagents EXCEPT the one of interest (and if you
insist on using an isotype control for that channel, so be it).  We
term these controls "FMO" or "Fluorescence Minus One" controls.  (For
more discussion of the need of FMO controls, see my paper on
Compensation in the upcoming issue of Cytometry).

Staining controls are very difficult to generate.  In general, the
best control for antibody binding is a cell that is exactly the same
as your cell of interest, but lacking the antigen of interest.  Of
course, this is rarely achievable.  However, one will often find very
similar cells that meet the bill.  In immunophenotyping of peripheral
blood, you can use "nonexpressing" cell types as internal controls
(i.e., naive T cells can serve as a control for measuring activation
markers on memory T cells).  Of course, you need to be careful,
because some "nonexpressing" cells actually express the marker.

Isotype controls have their place.  However, most people don't use
them properly.  In general, I counsel people NOT to use isotype
controls, but rather to use their brains to come up with a set of
appropriate negative controls (which MUST be included in all
experiments, as others have noted).  Blind reliance on isotype
controls is one of the most common mistakes in publications--and
leads to the erroneous placement of gates.

In any case, you are correct that investigators need to be educated
more.  This is one of the discussions that pops up every few years on
the mailing list; perhaps it's time to have a FAQ's page (no pun
intended!) on the Purdue site which includes the various discussion
points, and rather than coming up with a conclusion, this page can
simply serve to put forth various peoples' views so that researchers
can judge for themselves whether or not isotype controls are useful.

mr

(PS, there is no such thing as "bad data", only "bad interpretation of data.").

>It is very important to have an isotype control. I have come across some
>investigators who have not being using isotype controls. I have seen some
>using only unstained cells as negative controls which is also wrong. It is
>like doing any other assay without a negative control. When we have negative
>or non-specific binding control in other assays (ELISA, RIA etc) how can you
>not have isotype controls in flow cytometry. It also relates to the recent
>discussion we have been having on this forum regarding bad data. I think
>investigators need to be educated more.
>
>Indresh  Kaur Ph.D.
>Phone: 281-483-8791



This archive was generated by hypermail 2b29 : Wed Apr 03 2002 - 11:58:00 EST