Antibody's specifity...

Mike Clark mrc7 at
Thu Jul 24 12:45:02 EST 2003

In article <3F1FFC08.5010306 at>, Bryan Heit
<URL:mailto:bjheit at> wrote:
> Sorry this took so long, I ran some of this by one of the B-cell people 
> here, and I think our "conflict" is more a matter of semantics and our 
> background.  Most of the B-cell/Ab material I am familiar with are from 
> the labs here who study the development of antibody repertoire, which 
> apparently has it's own definitions for things like affinity and 
> sepcificty, and these are not widely used outside of their field.  Also, 
> I snipped a lot, not trying to hide anything, my server doesn't let me 
> post more old text then new...
> Mike Clark wrote:
> >Over the years several monoclonal antibodies have been reported in the
> >literature with biological cross reactions and many more have gone
> >unreported because it is not considered worthy of repeated publication of
> >a similar trivial observation. One of the first I can recall was an
> >antibody to human CD4 which cross reacted with an antigen in the brain of
> >sheep.
> >
> Which, assuming we heard the same story, was cross reacting with CD4 on 
> the microgleal cells.  If this is the case then you're talking about a 
> cross-species reaction reaction between the same gene product.

That is possibly a true explanation but the same antibody did not react
with human brain. I don't know if the levels of CD4 expression are
different in human and sheep.

As pointed out in other posts in this thread, if you have ever made
monoclonal antibodies from hybridomas, you will be struck by the number of
antibodies that react with the controls and thus you end up in rejecting
them as "non-specific".

> > The point is that most antibodies are only defined as "specific" when
> > they fail to show detectable cross reactions in the assay in which they
> > are put.
> >
> Depends on your definition of specific.

But that is exactly my point. Specificity is not based on a universal
definition but requires some extra information and understanding of the

> This of course depends on where  you are coming from, but the labs here
> have some very specific  definitions of specificity, including ranges of
> on/off rate, Kd's, etc,  that they use to define specificity for an
> antigen.

Yes but you need to elaborate more on even this point. If you have two
antigens called A and B and your antibody is meant to be specific for A,
and not B, what are the acceptable on and off rates or Kds? What if you
then decrease the Kd for both antigens by the same amount, has your
specificity changed?

If you go to my website you will find some lecture notes of mine on
antibody affinity


Included in there is analysis of a paper on antibodies to different species
of lysozyme in which I point out that "specificity" and "affinity" are not
easily related concepts.

> Obviously this  doesn't have much impact on doing a western blot, but
> when you are  looking an repertoire development, etc, you need a way to
> quantify "specificity" and apparently there way is commonly accepted.

No I don't think there is a "commonly accepted" or universal definition of
specificity in terms of purely kinetic parameters. 

For example we isolated the antibody Campath-1 based on detection
of specific lysis of human lymphocytes using human complement. The
definition of specificity requires a measurement of lysis not just of
binding. There are many other antibodies which bind to human lymphocytes
but not many of them give specificity in lysis. So you can define the
specificity of an antibody based on a functional assay such as killing or
neutralisation rather than on a concept of Kd. It is just as valid to
consider these as measurements of antibody specificity for antigen as any
other measurement of how an antibody interacts with antigen.

> > > But this process has almost nothing to do with the  maturation of an
> > > antibody you would buy or isolate.
> >>
> Bad wording on my part.  What I was trying to say is that the initial 
> stages of B-cell development (i.e. positive and negative selection) will 
> generate antibodies that are generally of: low affinity for any 
> particular antigen, and will bear little resemblance to the antibodies 
> generated after an immune response.

Yes but that is also partly because of a process of selection. The
frequency of B-cells making an antibody of any given specifity might be low
initially but it can be enhanced through selection and clonal expansion. 

What of T-independent immune responses, are they not also specific?

Blood groups A, B and O are defined by the specificity of such antibodies!

> > 
> > Multiple steps of selection have everything to do with what you buy and
> > or isolate. I will mention a few of the many that there might be.
> >
> >[1] What was the nature of the immunogen? Was it a pure antigen or was it a
> >mixture of antigens? Was it in adjuvant? How closely related is the antigen
> >to any similar host antigen in the animal immunised?
> >
> But irregardless of the vaccination method all of the antibodies will 
> have some general characteristics:
> 1) as a group they will identify a limited range of antigen epitopes.

That isn't only a function of antibodies and antigens, it is a consequence
of any ligand-receptor interaction. The affinity of interaction will vary
for different molecular pairs such that only a limited range of pairings
from all the universe of possibilities will give binding above a selected

> 2) The recognized epitiopes will be derived from the antigen(s) 
> administered.

There I must disagree with you. I have frequently  found antibodies which
recognise SRBC from spleens of mice or rats that have never been immunised
with SRBC.

Also naturally occuring A and B blood group antibodies don't depend on
blood transfusion for their origin.

Anti-NP, TNP or oxazalone "specific" antibodies can be found in the sera of
some strains of rats and mice without any immunisation with such antigens.

> This may be only one epitope (i.e. in the case of an  injected peptide),
> or multiple epitopes (in the case of an injected  protein or group of
> proteins) 3) each individual altibody will liekly only recognise 1
> epitope in the  origonal antigen 4) There are exceptions for 1&2, for
> example if you were to "vaccinate"  for anergy you will not
> expand/develop the B-cell repertoire, but lets  just assume you are
> vaccinating to develop an immune response.

Look I'll give another example.

I inject a mouse with human IgG as an antigen, and get back some
monoclonals which react with human IgG1, IgG2, IgG3 and IgG4 equally. Other
monoclonals react only with IgG1 but not with IgG2 3 or 4, and other
monoclonals react only with the allotype G1m(1,17) but not G1m(3).

So which of these monoclonal antibodies do you consider  specific and which
are not specific? 

I have examples of all of the above antibodies as reagents in my

> > [2] How many times was the animal immunised? Was the same form of
> > immunogen or antigen used every time?
> >
> Just one point, if you're using a different immunogen you are going to 
> generate a new response that will have nothing to do with the previous 
> response.  A different group of naive cells will be expanded and 
> selected upon, whereas the cells from your original will be unaffected.

Not true about them being independent. For example I could give one
immunisation with human T-cells, and then follow up with purified CD3. Both
forms of immunogen contain CD3 and so both immunisations might expand a CD3
specific population of B-cells.

> > [3] Am I using a polyclonal antisera from the animal? If so has it been
> > purified in any way?
> >
> But in the case of the original question a polyclonal serum isn't really 
> applicable here.  The question was about 1 AB recognizing >1 Ag, with 
> low homology between the antigens.  After all, you could take serum from 
> you or me and find antibodies what react with pretty much anything under 
> the sun - but it isn't one antibody doing these interactions, it's 
> thousands of different antibodies, each reacting with one (or a few) 
> epitopes.

In that last sentence you at last start to see the point. Each single
antibody can react with a few epitopes, but these will vary from antibody
to antibody. So each different monoclonal antibody, even to the same
epitope has its own unique possible family of cross reacting epitopes.

> > 
> > But you forget to point out that at each stage of somatic mutation
> > there is the potential for a different affinity on a range of different
> > antigens including the immunogen, but also including other
> > self-antigens. Thus a process of negative selection is required
> > following each cycle of somatic mutation.
> >
> I've seen a few people who would disagree with this point - I've heard a 
> few people argue that self mimicry may work because there is a lack of 
> negative selection during affinity maturation.  I don't know if I buy 
> this argument, but epitope spreading does occur during affinity 
> maturation, a process driven by forming "new" antibodies that react with 
> things other then then original antigen.  And from what I understand of 
> the process the selection that does occur during this process is diven 
> by affinity - B-cells with low affinity for the antigens present in the 
> GC get outcompeted and die.  I am unaware of any proof that negative 
> selection against self antigens occurs at this phase (not to say there 
> isn't, B-cell development is not my area of specialty).

If you follow the literature on autoimmunity you will see what happens in
models where you deliberatly generate auto-antibodies by immunising with a
xeno- or allo- antigen. The animal frequently starts to make detectable 
levesl of auto-antibodies but this response becomes self-limiting over time
and eventually the allo or xeno-specific response predominates and the
auto-reactivity declines.

> > In the early days of making monoclonal antibodies scientists frequently
> > wanted to isolate the genes encoding the antigen recognised by the
> > antibody. They did this by selection of random shot-gun clones using
> > monoclonal antibodies. Many, many scientists ended up cloning and
> > expressing the "wrong" genes based on reactivity with their antibody!
> > Few of these scientists wrote up their failure to clone and select the
> > right genes!
> >
> Not to be rude to those scientists, but that suggests an error on their 
> part.


> You mention that this was a problem a while ago - and we all know  that
> even today antibody purification from sera is plagued with all  kinds of
> problems, including pulling out other antibodies that aren't  specific to
> your antigen.

I was speaking of monoclonal antibodies such as the early workshops on the
CD antibodies.

> Before I would accept that there results  represent cross reactivity of a
> single antibody I would ask:
> 1) were they truly working with a monoclonal?


> 2) how did they select/purify their antibody?

They were defined as specific in the leukosite antigen workshops that
defined the whole CD nomenclature we use today.

> 3) how pure was their original antigen used for immunization?

That is irrelevant because we are discussing reactivity of monoclonal
antibodies and you claim that they are specific for only one antigen.

> 4) how similar were the proteins that they identified in comparison to 
> the protein the antibody actually identified.

Sometimes no primary sequence homology at all.

> I would argue that their results were more likely a result of poor 
> antibody purity, rather then cross-reactivity between dissimilar
> epitopes.

Your arguments are wrong.

> > Finally have you never read the extensive literature on "mimitopes"?
> > Mimitopes are selected antigens, often from peptide libraries, which
> > show reactivity with a selected antibody. However they are called
> > mimitopes because they mimmick an epitope of the antigen, rather than
> > having identity to the antigen.
> >
> But they mimic the structure of the epitope, which is the point I was 
> trying to argue.

Yes but the mimicry may not have any primary sequence homology at all. i.e.
they can be two completely different protein sequences that mimic a similar
tertiary epitope. So the antibody is not specific to one antigen.

> Most cross reacting antibodies I have run into cross  reacted because
> they were identifying a shared epitope on different  proteins (i.e. have
> the same amino acid sequence and identical 3D  structure).  A perfect
> example of this is the phospho-tyrosine antibody. 
>  This antibody identifies epitopes on many different proteins, but it is 
> the SAME epitope - phorphylated tyrosine.  So at the point of 
> identification there is "100% homology" between the proteins.  As I said 
> previously, I've never heard of an antibody cross reacting with another, 
> completely different epitope.

But there are other examples of 3D mimicry with no 1D homology

[snip] Must rush now. I'll deal with your other points tomorrow, but the
Tour de France beckons.........

Mike Clark,                        <URL:>
 o/ \\    //        |\   ,_ o      Mike Clark
<\__,\\  //   __o   | \ /  /\,   "A mountain climbing, cycling, skiing,
 ">    ||   _`\<,_  |__\  \> |  immunology lecturer, antibody engineer and
  `    ||  (_)/ (_) |   \corn computer user"

More information about the Immuno mailing list