Antibody Validation in Flow cytometry

My previous employer grew very quickly from manufacturing licensed clones of common antigens for flow cytometry to eventually undertaking the R&D to develop novel clones across the gamut of biological and technological applications.  The kind of validation required for these two efforts are as different as night and day.  To clarify, in this instance I’d define common antigens as basic phenotyping markers for cell subsets found in blood.  You start the validation having info on the characterizing features of the antigen and its expression. Novel clone development does not always have that benefit. Also, with newly determined antigens or those that were extracted under non-native conditions, often the antibodies we do have for comparison or use as a gold standard are actually not as good as the newly developed ones, leaving us to wonder which result to trust more?  And finally, blood is “easy” in that it’s readily accessible, even blood malignancies.  Other tissue types, especially if they are human and from a disease state, might be highly inaccessible in abundance and price.  Same with validation models that require a specific stimulation where degree of expression or simple human error or other factors that can be out of our control might make the model unreliable.  In those cases, when the antibody isn’t working, how can you be sure it’s the antibody and not the model? 

So let’s start with basic definitions, as always.  Sensitivity versus specificity in context of antibody performance itself, not the issues manifesting from the instrument, fluorescent conjugate or combination of conjugates in a flow panel. 

Sensitivity is the lowest limit of detection of an antigen (truly positive events).  In disease, this would be the lowest level of biomarker in a sample at which point the disease can be diagnosed confidently. 

So in flow cytometry, this metric describes conditions of high background not related to the specificity of the clone and poor affinity of the antibody for its antigen.  Non-specific binding of the antibody is the reason why isotype controls can be relevant especially in instances like detection of monocyte, macrophage or B-cell populations.  There can be non-specific binding due to Fc receptor binding, poor purification of the antibody conjugate or non-specific binding of the fluorophore itself.  None of those artifacts of the assay are a result of poor specificity of the antibody for its antigen.  On a plot, this would be a shift of the unstained or negative population that would reduce the ability to resolve the positive from negative.  Staining Index is a measure of sensitivity. 

Sensitivity could also be also be poor binding kinetics of the antibody for the antigen.  Not to be confused with specificity which is the selectivity of the antibody for the antigen (its detecting the thing it was designed to detect).  Every antibody also has kinetics of binding.  As we all know, you can’t use the same microgram amount of antibody for every antigen.  The molar ratio of antigen to antibody is not 1:1.  That’d be amazing.  More often than not, there is a molar excess of antibody to antigen to achieve saturation.  Sensitivity is a reflection of HOW MUCH of a molar excess of antibody is needed to achieve saturation.  This element of sensitivity manifests when we are looking at rare populations of cells in a sample or antigens of low abundance on the cell.

Specificity on the other hand is simply the question, does the antibody detect the antigen and only the antigen it was designed for?  As in, when there is no specific staining present, the sample is truly negative?  A good example is the detection of caspases which are important proteases involved in apoptosis.  There are antibodies against specific caspases, like caspase 3 or 8 which are pathway dependent or pan-caspases that will detect all enzymes in the caspase family.  Then there are antibodies for activated caspases, those that are poised to carry out their function and those that cannot differentiate the activated vs native caspase.  That’s specificity.  For example, does the antibody detect both activated and non-activated caspase-3?  

Sensitivity could also be also be poor binding kinetics of the antibody for the antigen.  Not to be confused with specificity which is the selectivity of the antibody for the antigen (its detecting the thing it was designed to detect).  Every antibody also has kinetics of binding.  As we all know, you can’t use the same microgram amount of antibody for every antigen.  The molar ratio of antigen to antibody is not 1:1.  That’d be amazing.  More often than not, there is a molar excess of antibody to antigen to achieve saturation.  Sensitivity is a reflection of HOW MUCH of a molar excess of antibody is needed to achieve saturation.  This element of sensitivity manifests when we are looking at rare populations of cells in a sample or antigens of low abundance on the cell.

Image from Cell Signaling Technologies: Red=PI, Green= phalloidin, Blue= anti-cleaved caspase 3

Some other factors involved in specificity:

1. Is the antibody polyclonal or monoclonal? Recombinant or traditionally produced?

2. What region of the antibody/ epitope is the antibody raised against? If you mutate that region, which mutations will lead to a loss of detection? If that site undergoes post-translational modification, does the antibody still detect?

3. Is the native structure of the epitope the only conformation the clone will detect or if the antigen is denatured, is it still antigenic to the antibody?

4. Is there a related epitope that you only find out later is more attractive to the antibody you thought was specific and competes for the antibody in solution?

5. Is there species-dependence in it’s specificity? Mouse, human? But what about macaque, pig, dog, etc. Whenever I go to New Zealand, inevitably I’m asked about cross-reactivity with sheep!

6. In what other methods of sample prep and cell-based assay methods is the clone able to be used? The specificity of the antibody gains a lot of clarity if the antibody can also be used in ICC or IHC or IP.

There are probably even more factors than that. But the question is, how far do manufacturers need to go to ensure specificity?  Here are methods of determining specificity.  Any single one of these methods is not sufficient.  What I want to drive home in this blog is that depending on the nature of the antigen/antibody/biology/platform, not all methods would be possible or even relevant.  So the question to consider is, how many different methods for specificity should be required before a clone is released for sale from the manufacturer and what can be determined by the user over time to add to the wealth of published info on that clone?

 1. Known expression patterns: In a normal control sample, expression patterns are one parameter for validation can be an expected range of % positive cell subset or a range of MFI on a particular fluor like PE.  My favorite resource for antigen expression is CDMaps: http://bioinformin.cesnet.cz/CDmaps/ and https://europepmc.org/backend/ptpmcrender.fcgi?accid=PMC6820661&blobtype=pdf for the publication (graphic example cited from this publication).  Using a backbone of common phenotypic markers to delineate subsets, you can now have a better understanding specific expression patterns to guide validation criteria.

2. Epitope blocking: Good specificity would be the absence of staining with the new conjugated antibody under a condition where the antigen is saturated already with an unconjugated antibody.  A “gold standard” clone to pre-block the sample would be ideal if it were raised against the same epitope as your antibody so the blocking might be 100%.  This isn’t usually likely.  I’m sure you can see the limitations of this method since it doesn’t really indicate the antibody is actually specific for its target in the larger sense.  There is also the possibility that the fluorophore conjugated to the antibody might exhibit non-specific binding itself, so specificity needs to be verified by using different types of fluorophores for detection.

3. Knock-out/ Genetic methods: Reliability of the testing model is imperative and knock-out or knock-in cell lines can be very useful towards this purpose.  However, it’s a laborious process to create a stable cell line and the expression may exhibit attrition over time.  Or the knock-in/knock-out product might produce a non-viable animal or cellular model.  You might say, just CRISPR it, however commercial antibody research using CRISPR requires a license that can be untenable for manufacturers.  

We often had web data showing the kind of clearly positive and negative staining that is so nice about an expression model.  However, from the users perspective, this data is not really relevant since it does not represent the endogenous expression level or variability.  It might errantly lead someone to believe they can use a less bright fluor on the antibody because the web data is showing them this artificially strong signal from over-expression.  For me, this question of antibody validation is really about transparency.  Show the data you acquired during validation to your customer.  All of it.  If the only web data you see as a customer is from an expression model, you are not likely to feel much confidence in the validation efforts of the antibody.

4. Sequencing epitopes: The most difficult method of validation for commercial antibodies is sequencing the antigen binding zone of the antibody.  This would lend itself to the creation and use of recombinant antibodies, which do in many ways ensure a more consistent performance of each antibody in the production lot.  There are many times that a well regulated hybridoma has shifted between lots over extended periods of time using conventional methods of antibody expression.  We simply identified that shift from performance QC parameters before it could be sold.  However, if you mass-spec’d every lot, that would be an extremely tight quality control criteria that is not open to interpretation.  Yet, I don’t know of a single antibody manufacturer that owns a mass spec, knows the sequence of every antigen recognition zone or epitope it recognizes and would be willing to take the time and cost involved in validating via this method for their whole catalog.  It’s a great idea, in concept, but not practical enough to keep the cost of antibody R&D, production and QC at a reasonable amount.

So, since we are so dependent on performance characteristics to qualify the specificity of the antibody, what can antibody manufacturers do to increase our confidence in their product?  These are just my personal suggestions based on what is reasonable and also accessible.  

5. Importance of disease state/ activation mode:

Let’s take an example of a moderately challenging antigen/ antibody like PD-1.  PD-1 has a low level endogenous expression on hematopoietic cell types.  However, with cell activation that expression is upregulated.  The up-regulation leads to an inhibitory pathway to suppress the killing activity of T-cells.  In cancer, blocking PD-1 and PD-L1 rescues anti-tumor immune activation (graphic taken from Salgia et al, Journal for ImmunoTherapy for Cancer).  There should be QC standards set for both the detection of endogenous expression and the activated expression for each fluorophore conjugated.

However, something familiar to people in Pharma, make sure that the antibody is suitable for its context of use, which is largely immuno-oncology in human samples/biopsies.  Disease state tissue or blood samples should be applied extensively in the antibody R&D.  They may be expensive, depending on the disease, but not so much that you’d actually lose profit margin on the product.  If you can’t buy them, collaborate! And the data that can be shared from that R&D testing is priceless in increasing the confidence one has in the validation of that antibody.

Certainly not comprehensive, of course.  There are a lot of excellent publications on this matter, though I wish more of them might take a practical attitude and approach to the solution.  We need antibodies to do research.  We need faith in the antibodies as an accurate tool that we can trust to give us a reliable result.  All of this is attainable if we understand the limitations  of the process and all manufacturers hold themselves to the same expectation of quality.