Antibody's specifity...

Mike Clark mrc7 at
Fri Jul 25 05:36:45 EST 2003

In article <ant2417027a1Pk=+ at>, Mike Clark
<URL:mailto:mrc7 at> wrote:
> [snip] Must rush now. I'll deal with your other points tomorrow, but the
> Tour de France beckons.........
> Mike

> In article <3F1FFC08.5010306 at>, Bryan Heit
> <URL:mailto:bjheit at> wrote:

> >>>Also all antibodies have the potential to bind a large number of
> >>>different antigens and thus specificity can easily be seen to be  a
> >>>"relative" concept.  
> >>>
> >>>      
> >>>
> >>Specificity is not a relative concept, at least not in any of the lab's 
> >>I've worked with.
> >>    
> >>
> >
> >Specificity is a relative concept and it is a pity that the labs ypu have
> >worked in haven't appreciated this fundamental scientific principle. If you
> >increase the sensitivity of an assay procedure you often reduce the
> >specificity because you start to see low affinity cross-reactions which
> >were originally below your threshold of selection. Specificity is entirely
> >dependent on a definition of thresholds at which you regard a result as
> >negative. If you change the threshold then defined negatives and positives
> >in the system change as well.
> >
> See, this is where we're arguing over definitions.  As I mentioned at 
> the beginning most of the interactions I've had with B-cell people is 
> with groups who look at repertoire development.  They have very specific 
> definitions of affinity and specificty, as they need a "ruler" to 
> measure the development of the immune response.  

Yes but unfortunately these definitions of specificity and affinity are not
clearly indicated as a restricted definition but tend to find their way
into standard texts. Specificity of the immune response is one of the most
misunderstood concepts in immunology and this misunderstanding is
perpetuated so long as students learn from standard texts that affinity and
specificity are dependent variables that go hand in hand. In fact there is
only a somewhat loose relationship between affinity and specificity. The
real observation is that some of the most "specific" antibodies are in fact
what would be considered low affinity antibodies. For example the
specificity of IgM antibodies for carbohydrate antigens is a result of low
affinity but high avidity, and it is the avidity that is important in this
respect. Why do you think antibodies have evolved to be multivalent?

> I will concede that specificity can vary depending on the assay you put 
> an antibody into, or the conditions you use it at, but I would argue 
> that these variations in specificity are more a product of the assay's 
> rather then the antibodies.

But that is exactly how specificity should be  defined it is a functional
definition based on an assay and not an intrinsic property of an antibody.

> For example, in a western blot your  proteins will be partially
> denatured, even if you use a "non-denaturing"  gel.  This is simply a
> product of the solvents and gels we use for  blotting.  So an antibody
> designed for use with ELISA, which "never"  sees denatured protein
> (assuming you're better at ELISA's then I am),  may not work in a
> western.  But this variation in "specificity" is due  to environment,
> rather then an intrinsic characteristic of the antibody. 
>  Likewise, differences in temperature, osmolarity, pH, etc can affect 
> protein conformation and thus antibody affinity, but this once again is 
> due to the assay rather then the antibody.

Yes so as you point out the affinity of the interaction can also depend on
the conditions so the specificity of the antibody can also change with the

Yes and two receptors can bind the same ligand differently under different
conditions. Taking another example, both protein A and FcRn interact with
the same site on antibody IgG Fc regions. 


However FcRn binds with higher affinity at low pH and lower affinity at
neutral pH. Protein A is the other way round. So the specificity of Protein
A or FcRn is pH dependent, but both are accepted as being specific for the
same binding site on the IgG.

> > I'll give you another thought experiment. Take an antibody-antigen
> > crystal structure. Now make selected mutations in the contact residues
> > for antigen one at a time. Conservative changes are likely to make
> > small changes in the affinity of interaction, whilst non-conservative
> > changes are likely to have a big effect.
> >
> >How specific is you antibody if you define specificity in terms of ability
> >to descriminate different mutated forms of your antigen? 
> >
> Ahh, but you could easily define the specificity as the affinity for the 
> now mutated antigens.  You could easily determine the Kd, on/off rates, 
> etc for each antigen.  From this data you could state the specificity 
> for each antigen empirically.  But you seem to be arguing my point (or 
> maybe I missed yours).  I was arguing (apparently badly) that a single 
> antibody tends to recognize a single epitope.  You can modify this 
> epitope in minor ways and still have some antibody binding to the 
> epitope,

This I agree with.

> but it is highly unlikely that the antibody will identify an  dissimilar
> eptiope - like the one in the original question which only  had 50%
> homology to the original epitope.

It may be improbable that it reacts with any single antigen chosen at
random, but this is not the same as assuming it has no reactivity with any
other antigen or epitope.

You keep ignoring the fact that one antibody can demonstrably have multiple
antigens to which it can bind. For example returning to "mimitopes" if you
take an antibody against a carboydrate antigen and then screen it against a
large peptide library it is very likely that you will identify different
peptides that bind to the antibody V-region. Because they are peptides and
the original antigen was a carbohydrate they cannot be the same or a
similar "epitope" but they represent true cross reactions of the antibody
on different antigens.

> It is conceivable that you  could "build" a similarly shaped/charged
> epitiope from a completely  different amino acid sequence that would be
> recognized, but I would  argue this would arise naturally very rarely,
> and even if done  deliberately would be difficult.

Experience of screening of libraries for mimitopes, or making of
anti-idiotypes would seem to go against your assertion.

> > Specificity IS defined by the system, it is NOT an absolute concept.
> >
> Once again, depends on how you are defining specificity.  I tend to 
> think in vivo, and in vivo specificity is very well defined.

I think you are wrong about your definition of specificity being applicable
to in-vivo. I would argue that functional definitions such as the ability
of an immune response to "neutralise" a pathogen, to cause complement
activation, or to mediate phagocytosis or ADCC have as much or more to do
with antibody specificity than does any definition based on single site
binding affinity or kinetic parameters.

> But we're  arguing semantics and definitions, which no one will ever win.

Don't you see, in science semantics can be very important. If scientists
cannot communicate their ideas clearly even between themselves because
different scientists have very different definitions for the same words,
then science education is failing.

> > Ah! But return to my numbered points above! Are your monoclonal
> > antibodies to human adhesion and signalling molecules by any chance
> > mouse monoclonal antibodies? If so then were they made by immunising a
> > mouse with human antigen? Have you tried them out on other species such
> > as dog, rat, non-human primate, rabbit etc etc? What then of
> > specificity?
> >
> Many of the antibodies we used origonally were raised in rabbits because 
> we hoped they would cross react with both human and mouse.  In rabbits 
> we could raise both anti-human and anti-mouse antibodies, but they 
> rarely cross reacted.  In hind sight these are bad example, as many 
> adhesion molecules are highly glycosylated and these glycosylation 
> states are variable between species.  
> A better example would be the TLR4 antibody I'm fighting with right now. 
>  It was raised using a peptide, rather then the whole molecule, and the 
> portion used has near perfect a.a. homology between mice and humans 
> (this was done with the hopes of creating a Ab that cross reacts b/w 
> species).  The antibody binds beautifully to human TLR4 and shows little 
> cross reactivity on westerns or FACS (you need to use massive amounts of 
> antibody to get non-specific bands on western).  We can only get random 
> binding to mouse cells.  
> >Are the anti-mouse antibodies also mouse monoclonals? Were these raised by
> >immunising mice with mouse antigen? Are the monoclonals auto-reactive or
> >allo-reactive in specificity?
> >
> Varies, depending on the antibody.  All are monoclonal (nothing else 
> would be publishable), most from rats or rabbits (we have one raised in 
> a donkey, god only knows why).  Since many are purchased I don't know 
> about auto/allo reactivity.
> >Of course you may be working with another species of antibody such as rat
> >monoclonal antibodies. But them of course the immune response in a rat has
> >undergone negative selection on rat antigens, so you need to know how the
> >similar human, mouse and rat antigens might be related by sequence and
> >structure.
> >
> This is often a problem with mice, but it kinda proves the point I was 
> trying to make (apparently badly): antibodies recognize a single 
> epitope, and rarely will identify >1 protein unless the proteins share 
> identical or nearly identical epitopes.

Yes but you are arguing about cross-reactivity in a restricted universe of
antigens such as encountered in your western blotting experiments. Worse
still you are using antibodies which have already been selected as having
the specificity you want them to have. What I am trying to get across is
that there is a world of difference between looking at a couple of highly
chosen and selected monoclonal antibodies in your laboratory, and then
trying to extrapolate from these observations to a concept of specificty
of any randomly selected antibody encountered in a normal immune response.

> I still think we're arguing  about semantic/definitions.  I seem to be
> taking a more biochemical view  of things - defining specificity as
> binding to only one epitope,  irregardless of the location of the
> epitope.  You seem to be arguing  from a techniques point - antibodies
> can and do react between proteins  and species, but going back to my
> point, this usually occurs as a result  of shared epitopes rather then
> specificity for another epitope.

No, my point is that it is wrong to assume that the epitopes are shared. An
antibody can and does have the potential to react with multiple and
different epitopes. So sometimes cross-reactivity is because of a shared
epitope and sometimes it is not.

I'll give you some more examples of the numbers paradox which indicates
that one antibody one epitope is a false concept.

If you take the sequences of many monoclonal antibodies described in the
literature and run them against the sequence databases, you will frequently
find that many of them are almost identical to other antibodies
previously described as specific for a different antigen. [Of course this
is to be expected because the germline repertoire is relatively small and
all antibodies are based initially on the sequences in the germline
repertoire]. In contrast if you make several different monoclonal
antibodies to the same epitope of an antigen and then you sequence the
antibodies you frequently find that they are very different in sequence. So
the apparent paradox is that antibodies with similar sequences can see
different antigens and antibodies with very different sequences can see the
same epitope of an antigen.

Mike                            <URL:>
M.R. Clark, PhD. Division of Immunology
Cambridge University, Dept. Pathology
Tennis Court Rd., Cambridge CB2 1QP
Tel.+44 1223 333705  Fax.+44 1223 333875

More information about the Immuno mailing list