Written by: George Parsons, Ph.D.
Antibody selection is key to creating a robust lateral flow immunoassay that consistently provides reliable analytical information. As discussed in a previous blog, proper antibody selection calls for a rigorous design process in which the assay’s intended use is defined with enough detail so that specific information is gathered and considered by all stakeholders. Some “give and take” during this process is normal and stretch goals should be included and clearly identified. No one really knows what they can do unless they try.
Before venturing into the lab to begin the antibody selection process, one should be immersed in the literature of the assay in question. There is no need to reinvent wheels that already exist and that are documented in the literature. This should include the peer-reviewed literature and competitors’ package inserts, material safety data sheets, and the patent literature.
This process yields parameters that are useful in antibody selection. The parameters could include molecular weight of the analyte and the desired analytical range, sample size, reaction volume, and assay time.
It is important to consider the analyte’s molecular weight when deciding whether the assay is a competitive or a sandwich one. Antibody binding sites typically recognize six to eight amino acids, so a compound such as Angiotensin I (which has 10 amino acids) can only be addressed with a competitive assay simply because there is not enough space on the molecule to accommodate more than one antibody. Small molecules, such as steroids and drugs, can also only be addressed with competitive assays. A 30 amino acid peptide such as GLP-1 can give rise to a whole family of antibodies that recognize different regions of the peptide. Proteins that are bigger than GLP-1 obviously have the space to also bind to multiple antibodies.
Once the likely assay architecture (competitive or sandwich) is established, it is time to consider the desired analytical range. There are often one or more medically relevant concentrations that are encompassed in the analytical range. For example, a lateral flow assay for Thyroxine or T4 should have a range such that both the hypothyroid cutoff (4.6 ug/dL) and the hyperthyroid cutoff (12 ug/dL) are in areas of the assay where adequate reproducibility is possible.
Competitive assays typically span a range of concentrations of about a factor of 10 (1 log), so a good T4 assay would have a range of 2–20 ug/mL with a midpoint of about 8 ug/mL. Sandwich assays have a better range with about a factor of 30 (1.5 logs) from bottom to top, so a screening assay for prostate-specific antigen (PSA) with a medical decision point of 4 ng/mL ought to go from 0.5 ng/mL to 15 ng/mL with a midpoint around 4 ng/mL.
Two parameters determine how an antibody behaves. One is the affinity constant, which is fixed. The other is the concentration of the antibody in the reaction, which can be varied. The best way of deciding which concentration of antibody to use is to do a titration of antibody concentration while holding all the other parameters (e.g., sample size, total assay volume, label concentration, timing) constant. While this can be done with the entire standard curve range of calibrators, it is often sufficient to probe the system with just a few calibrators such as the medical decision point, a zero point, and a high calibrator point. For a T4 assay, this would be 0 ug/dL, 4 ug/dL, and 16 ug/dL. For a PSA assay, these concentrations would be zero ng/mL, 4 ng/mL, and 15 ng/mL. The abbreviated three-point standard curve is quite useful for screening multiple antibodies.
A few words of caution are in order about varying the antibody concentration. Closest packing of antibodies on a surface provides a saturation coating at about 1 ug/cm2.1 On plastic surfaces, this coated concentration can be achieved with coating concentrations of about 3 ug/mL. Exceeding this concentration can leave loosely bound antibodies that can interfere in the desired reaction and give rise to paradoxical low-dose hook effects. This appears as diminished assay sensitivity in a sandwich assay or as increased signal with increased analyte concentration in the lower end of the standard curve.
An antibody concentration that is too low can also cause problems. In addition to the obvious problem of low binding, trying to coat antibodies at less than 0.5 ug/mL can lead to bizarre antibody stability problems that manifest as increasing antibody activity over time. Spiking the coating antibody with an irrelevant antibody at 0.5 ug/mL can obviate this problem without affecting binding levels.
Screening antibodies on a platform other than the one the assay is intended to be ultimately used on is often a waste of time. For reasons that are not well understood, antibodies can behave differently on different platforms. Antibodies that behave well in a microtiter plate platform often do not perform as well on magnetic particles.
Once a reliable standard curve has been established, the real work of selecting an antibody begins. Since all other parameters depend on having a reproducible result, rigorous reproducibility studies must be conducted. A look at almost any FDA approved immunoassay’s package insert will provide examples of what data is needed.
The real test of any antibody is how it behaves with real world samples, preferably ones with relevant assay values already known from a well-validated reference method. A minimum set of at least 100 samples spanning the range of interest should be assembled. Adequate volumes of these samples will ensure the ability to test multiple antibodies several times, if necessary. Smaller subsets of samples (10–20) can be used in an initial screen.
There is no perfect assay, and that includes any putative reference method. In almost any method comparison, there will be discrepant results in which the reference method yields a result in one bin (e.g., above 12 ug/dL T4) and the test method yields a result in another bin (e.g., below 12 ug/dL T4). A “tie-breaker” assay from another source can be useful in this case to determine which assay is correct. If the tie breaker can be another technology (such as liquid chromatography-mass spectrometery [LC-MS]), so much the better.
A good correlation between the new method and the reference method—but with a slope greater or less than unity—can indicate problems with calibration or with a binding partner in the sample. Spiking and recovery experiments can help determine if a binding partner in the sample is in play. Binding partner problems can be solved with the addition of releasing agents such as anilonaphthalene sulfonic acid salts2, salicylate, or synthetic analogs such as danazol3 that are not recognized by the antibody in the new method.
Selection of the right vendor for the antibody is also crucial. Since a lot of time and money are going to be expended in developing an assay, it makes sense to ensure that the selected antibody is going to be available over time. A shorthand question to ask is, “Is this antibody supply or cell line auditable?” If the answer is “No,” then selecting this vendor increases the risk that the antibody supply could be compromised over time. Auditing the vendor either in person or remotely with a questionnaire can identify weak spots in the vendor’s practices that can be addressed.
The issues of monoclonal antibody versus polyclonal antibody and whether to make or buy the antibody are also important and will be discussed in future blogs.
Our team of engineers and scientists can help you determine the specification for antibody selection for your test. Feel free contact us with any questions and antibody selection.
1 Binding of protein to polystyrene in solid-phase immuno assays
Pesce AJ, Ford DJ, Gaizutis M, and Polak VE, Biochim. Biophys. Acta 492, 399 (1977)
2 Chopra IJ, Ho RS, Lam R. An improved radioimmunoassay of triiodothyronine in serum: its application to clinical and physiological studies. J Lab Clin Med;80:729-739 (1972)
3 Enzyme Immunoassay of Estradiol in Serum of Women Enrolled in an In Vitro Fertilization and Embryo Transfer Program; Bouve J, De Bouver J, Leyseele D, Bosmans E, Dubois P, Kohen F, and Vandekerckhove D; CLIN.CHEM.38/8, 1409-1413(1992)