I too have been frustrated for years with "saturating reagents". This
has been compounded significantly by the fact that many reagents
bottled by manufacturers are NOT saturating. Most FITC and biotin
reagents are saturating; about half of the PE reagents are not
saturating, and most Cy5PE tandems are not. (I have not tested PerCP
reagents). There is one main reason that these are not bottled at
saturation: they would be too bright at saturation. i.e., if a Cy5PE
reagent is bottled at saturation, then it is off-scale on the FACS.
(And no, you can NOT TURN DOWN THE PMT VOLTAGE--because then
compensation becomes a problem. The negative cells must be onscale!).
So, to keep these reagents on-scale, the manufacturers dilute them such
that they reagent becomes antibody-limited in its binding. Of course,
the "correct" way to make these reagents is to add unlabeled antibody
(cold competitor) such that the combination of labeled and unlabeled
antibody is both saturating and on-scale. (Yes, this is much more
difficult than diluting a reagent, and that's probably why they don't
do it).
In preparation of our reagents nowadays, we titre the reagent and
record two separate values: the saturating concentration, and the
"separating" concentration. The latter is the minimum amount of
reagent needed to achieve good separation of the populations of
interest, and can be considerably less than saturation.
It is crucial to note that at subsaturating concentrations, the
resulting fluorescence is dependent on all of the factors noted by Dr.
Milford and Dr. Martz. However, practically speaking, the cell number
is rarely a factor: usually, the reagent (even at substantially
subsaturating concentrations) is in vast excess over antigen. In fact,
using bottled reagent from manufacturers, I can typically stain 100
million cells with almost as much fluorescence as 1 million cells.
(However, this does NOT mean that you can stain 1 million cells with
1/100th of the reagent--unless you are willing to wait for the 100+ hour
incubation time). It is well-worth doing a nonlinear least squares fit
of the fluorescence to the one-site binding equation, as suggested by
Dr. Martz. We use Kaleidagraph (Mac) or JMP (MAC/PC) to do this.
BTW, we use a slightly modified version of the function that Dr. Martz
suggests, which does not require us to subtract the background
(autofluorescence) but rather fits this value as well:
F = A + Fmax / (1 + k / Ab)
The three parameters to be fit are:
Ab = antibody concentration at which the fluorescence F was measured;
k = concentration of half-maximal binding;
Fmax = theoretical maximal fluorescence (at infinite [Ab]).
Note that it is important to watch the fluorescence of "negative" cells
so that you can assess the amount of nonspecific staining. As noted by
others, additional washes may help reduce this latter aspect.
We too find most antibodies saturate around 5 ug/ml concentrations.
(The reason for this is that the "on-rate" for protein-protein
interactions does not vary much at all for antibody binding reactions.
The "off-rate" is essentially zero, since Abs are bivalent. This is
why we can wash our cells! Note that experiments performed with single
binding site reagents (Fc), the affinity is driven mostly by varying off-rates).
When we titre, we typically start at 20-50 ug/ml and do two-fold
dilutions from there. Some reagents require very high concentrations
(for instance, every antibody against human CD20 I have tested requires
at least 50 ug/ml, and some don't saturate even at that concentration);
some saturate at very low concentrations (e.g., PharMingen's anti-CD62L
saturates at well below 100 ng/ml). Paradoxically, high-affinity
antibodies are not necessarily better to use at just above saturation:
since they are at lower concentration, they will have a higher dependence
on the cell number that is being stained (the Ab/Ag ratio is no longer
in "vast" excess).
The primary advantage of saturating concentrations of antibody is that
the final fluorescence does not depend on incubation time, volume
(i.e., pipetting errors are irrelevant), cell number, etc., within very
large error margins. The only dependent will be the number of antigens
per cell.
mr