White Paper: Multiplex ELISA Validation

Printer_Friendly

 

Introduction

Enzyme-Linked Immunosorbent Assays, or ELISAs, are sensitive bioanalytical tests that utilize 1) immuno-particles adsorbed to a surface (hence immunosorbent) and 2) an enzyme reporter molecule (hence enzyme-linked) to detect and quantify a specific substance, such as a protein, antibody, or hormone, out of a complex mixture, such as serum, urine, or saliva. Traditional ELISAs measure one analyte at a time. A multiplex ELISA, on the other hand, is a high-throughput ELISA platform in which multiple ELISA reactions can be performed at the same time with one kit, allowing the simultaneous quantification of multiple analytes. For example, a planar-based multiplex ELISA utilizes an array of capture protein spots deposited on the bottom of the wells of a microtitre plate; each spot is a unique assay.

Like all analytical techniques, careful design and validation of the multiplex ELISA method is required to ensure that the resulting data is valid. In the May 2001 Bioanalytical Method Validation Guidance for Industry [1], the USFDA Centers for Drug Evaluation and Research (CDER) and Veterinary Medicine (CVM) recommend that the fundamental parameters of method development for a bioanalytical ligand-binding assay should include validation of selectivity, accuracy, precision, recovery, calibration curves, and stability. Many other regulatory agencies have also published definitions and standard practice recommendations on this topic in general, as well as pertaining specifically to ELISA assay development. Thus, the purpose of this white paper is to clarify some common terms used throughout government and industry for method validation, as well as to explain the validations performed on Quansys Biosciences Q-Plex multiplex ELISA products.

Selectivity

Selectivity is defined by the International Union of Pure and Applied Chemistry (IUPAC) as “the degree to which a method can quantify the analyte accurately in the presence of interferents” [2]. The unique ability of antibodies to differentiate and bind the analyte of interest among the complicated milieu of a biological sample, such as serum, empowers ELISAs as useful bioanalytical tools.This said, antibodies are often not perfectly specific. Many target proteins, such as steroid hormones or drugs and their metabolites, have closely related structures and/or highly conserved epitopes, making differentiation between related proteins or slight differences in proteins, such as post-translational modifications, more difficult [3]. For this reason, it is essential to evaluate antibody selectivity in the first stage of ELISA design.

With ligand-binding assays such as ELISAs, it is recommended that two selectivity issues be checked: 1) interference from substances physiochemically similar to the analyte, also known as cross-reactivity and non-specific binding, and 2) “Matrix effects unrelated to the analyte [2].” With multiplexed ELISA products, there are several additional issues that can occur, for example (the following are phrased specifically for immunometric, or “sandwich”, ELISAs for clarity, but correlating cross-reactivity can also occur in assays of other formats):

Antigen-Solid Phase Antibody Cross-Reactivity: A solid phase antibody binds the wrong antigen, which was then either specifically detected by the antigen’s corresponding detection antibody or non-specifically detected by another detection antibody. If antigen-solid phase cross-reactivity is truly present, then the solid phase antibody cannot be used to multiplex under the conditions tested, and different solid phase antibodies or assay conditions must be screened.

Detection Antibody-Solid Phase Antibody Cross-Reactivity: A solid phase antibody binds a detection antibody directly, resulting in a signal where the target antigen was not present. Detection-solid phase antibody cross-reactivity can often be minimized via reagent or diluent optimization.
Antigen-Detection Antibody Cross-Reactivity: An antigen is captured by the appropriate solid phase antibody, but then detected by an antibody targeting a different antigen. When antigen-detection cross-reactivity is present one needs to take care to account for the added signal generation that may be present with two detector antibodies for a given antigen.

Solid Phase Antibody-Conjugate or Antigen-Conjugate Cross-Reactivity: The conjugate, such as streptavidin horseradish peroxidase, nonspecifically binds directly to a capture antibody or to an antigen bound by its respective capture antibody. This type of cross-reactivity is rare, and such signal is more often due to biotin contamination of the solid phase antibody. Although rare, solid phase antibody-conjugate and antigen-conjugate cross-reactivity must be evaluated, as this type of cross-reactivity is prohibitive and must be resolved for the assay to be validated.

A related but slightly broader category of interference in ELISAs is non-specific binding. Non-specific binding refers to undesired binding of any of the assay reagents to other assay components, such as binding of analytes to albumins or immunoglobulins in the sample, or to assay surfaces [4]. These components are often in great excess of the target analyte, such that although the non-specific binding may provide a minimal percentage of the total binding, it nevertheless negatively affects the assay. Non-specific binding often results in loss of sensitivity or in false positive or negative readings for a sample.
The combined effect of all components of the sample on the measurement of the analyte is known as matrix effects. This is a general term used when the exact molecular cause(s) of the interference is unknown. Matrix effects can include interference by sample components such as human anti-animal-antibodies, heterophilic antibodies, or by sample conditions such as viscosity, pH, or ionic strength. For example, heterophilic antibodies, or weak antibodies with binding capability for multiple proteins [5], are present in biological samples such as serum, plasma, and tissue homogenates, and can bind the capture antibodies on the solid phase of an ELISA plate and then present epitopes for binding by the secondary or detection antibody in the absence of antigen [6]. Essentially a bridge is formed between the pair of antibodies in the absence of antigen [5]. For example, rheumatoid factors are multi-specific IgM antibodies that bind to the Fc sections of human antibodies. In immunoassays, these interfere by binding to the Fc sections of the human antibodies used in the assay, and thus cause false positives.

When selectivity problems such as cross-reactivity, non-specific binding, or matrix effects are observed in multiplex ELISAs, there are two common approaches to resolving the issue. The first is to utilize a more specific antibody(ies) for the assays involved. Through controlled cross-reactivity experiments, the optimum antibodies for detection can be selected. The second common approach to resolving selectivity problems is to modify the sample itself during the assay via diluent additives. Polyethylene glycol precipitation[7], addition of mouse IgG [8], or other proprietary additives available from commercial sources which neutralize the effects of anti-animal and heterophilic antibodies are examples of diluent additives that may help improve assay selectivity.
The selectivity validation criterion for all ELISA products at Quansys Biosciences is that the percent cross-reactivity must be below 1% between all assays in an array. Tests to determine other potential interference and matrix effects are also performed, including the linearity testing explained below.

Accuracy

Accuracy is defined as “the closeness of mean test results obtained by the method to the true value (concentration) of the analyte [1].” For proteins, such as cytokines and growth factors, some standard reference materials for measuring accuracy are available from NIBSC [9], however, many biological analytes do not have an official reference material associated with them yet. When a standard is available, manufacturers may follow the CDER/CVM recommendation that the mean value of a minimum of five replicates of three concentrations must be within 15% of the actual values, except at the Lower Limit of Quantification (LLOQ), which can be within 20% [1]. When a standard for determination of accuracy is not available, ELISA developers often assign values to the calibrator(s) using one or more outside standards as a metric. This process is known as “referencing,” and is often performed by assay manufacturers before, during, and after producing a lot of kits to ensure the defined concentration of the standard is accurate. Manufacturers may also opt to check the recovery of a pure calibrator-spiked sample; this is known as a recovery test, and is further explained below. The disadvantage of the latter two methods is that reported sample values may vary from one manufacturer’s kit to another based on the method of determining the concentrations of the kit standards.

For many of the assays developed at Quansys Biosciences, there are not officially recognized standards, so referencing and recovery testing is typically used to determine accuracy. We reference many of our multiplex products to commercially-available ELISA kits such as R&D Systems ELISAs. An explanation of recovery testing in general and at Quansys is provided below.

Precision

Precision is the “closeness of agreement between the results of independent measurements of an analyte when the procedure is applied repeatedly to multiple aliquots of a single homogeneous volume of biological matrix” [10]. The Clinical and Laboratory Standards Institute (CLSI) recommends that the precision of a method should be tested at at-least two levels; each run in duplicate, with two runs per day over 20 days [11]. Additionally, there are several hierarchical estimates of precision: intra-batch, inter-batch, intra-laboratory, and inter-laboratory [1, 10].

Quansys addresses precision for all multiplex ELISA products by ensuring during development that the following is true for each assay of an array:

Edge Effects: The percent difference between replicates of a mid-level positive control run in the outer wells and inner wells of a 96-well plate must be below 10%.

Intra-Assay: The %CV for 20 replicates of a high-level, mid-level, and low-level positive control over 4 plates from the same Lot must be below 15%.

Inter-Assay: The %CV for 4 replicates on 20 plates from 3 different plate lots run by 3 users must also be below 15%.

Recovery and Linearity

Recovery and linearity experiments are often used to assess the accuracy, matrix effects, and the compatibility of a particular sample diluent to be used for assaying analytes from a particular sample type such as serum, plasma, saliva, urine, etc. Specifically, recovery is the determination of the concentration of an analyte added to a sample, measured in that sample [3], and is used to determine if the assay is affected by the difference between the diluent used to prepare the standard curve and the sample matrix. According to the May 2001 Bioanalytical method validation Guidance for Industry [1], “recovery of the analyte need not be 100%, but…should be consistent, precise, and reproducible.” Linearity of dilution, also known as parallelism, is a determination of the extent to which the dose-response of the endogenous analyte is “directly proportional to the concentration of analyte in the sample within the range of the standard curve” [12] in a particular diluent, and should ideally be 100%.

Multiplexing introduces many challenges to optimizing linearity and recovery. Recovery and linearity are affected by diluent composition, among other things, and optimizing a single diluent for many analytes can be challenging. Recovery can also be affected by the presence of other interfering molecules in a sample. For example, IL-2 frequently has poor recovery in some samples because of the presence of soluble IL-2 receptor. Analytes that form multimers also present challenges to optimization of linearity and recovery. Overall, although linearity of dilution can also be affected by diluents composition and multimer formation, it tends to be a more reliable measure of accuracy and for complex arrays is relied on more heavily at Quansys than is recovery.

The recovery and linearity criterion for Q-Plex assays is that the percent recovery of a positive control should preferably be between 80%-120% in a given sample type, and the percent linearity must be between 80%-120% in a given sample type for at least 5 points of an appropriate dilution series of a high-level positive control. Exceptions to this criterion are noted in the appropriate kit manual.

Curve-fitting

Quantitative immunoassays rely on the mathematical equation used to model how the signal of calibrators responds to changes in concentration in order to interpolate the unknown concentration of analytes in samples. The mathematical function that best matches the calibrator concentration-response data is called the calibrator curve, hence the term curve-fitting. Unlike many methods of chemical measurement, the calibrator curve of an immunoassay is not linear. When plotted on a normal x-y scale, the curve often looks like a single-site saturation model [13]. When plotted on an x-log scale, the curve resembles a sigmoidal-shaped curve.

There are many equations that can be used for immunoassay curve-fitting. Several of the most common are explained below, and are available in the Q-View Software, a tool provided by Quansys for the quantitative analysis of multiplexed chemiluminescent or infrared fluorescent assays, such as Q-Plex Arrays:

5 Parameter Logistic (PL): This non-linear regression model is generally considered the best choice for fitting sigmoidal immunoassay standard curves, and is recommended for Q-Plex kits. The 5PL regression should only be used for standard curves with 6 or more points.

4 Parameter Logistic (PL): This is another non-linear regression model that is slightly less complex than 5 PL, as it does not allow for asymmetry, but also considered a good model for analyzing ELISA standard curves.

Log-Log: This curve-fitting model uses log-based regression on both the x and y-axes.

Linear: This model takes the best fit line over the entire standard curve. It is most likely not the ideal data output for standards provided in the Q-Plex kits.

Point to Point: Also known as spline analysis, unknowns are determined by the straight lines between each standard point in this model.

Qualitative: This option generates pixel intensities only; no concentrations are calculated. Bar charts showing the average pixel intensity for each well type and dilution are also generated.

Auto-Select: This option fits standard curves for each system individually based on the lowest AIC value.

Once a calibrator curve has been fit, the quantitative range limits of the assay can be calculated. At Quansys Biosciences, three terms are used to define the range and sensitivity of our multiplex ELISA assays: Lower Limit of Quantification (LLOQ), Lower Limit of Detection (LLD), and Upper Limit of Quantification. The definitions of these terms as used at Quansys Biosciences and in science are discussed below.

Upper and Lower Limit of Quantification (ULOQ and LLOQ): The ULOQ and LLOQ are the highest and lowest standard curve points at which quantitative results may be obtained with a specified degree of confidence, or the highest/lowest concentration of an analyte that can be accurately measured. Together, the ULOQ and LLOQ define the range of quantification for the assay. Limits of quantitation are matrix, method, and analyte specific, and can be calculated as follows:

Equation 1. Calculation used in Q-View Software: ULOQ & LLOQ = Highest or Lowest Standard, respectively, with a %backfit of 120%-80%, a %CV of < 30%, and a positive mean pixel intensity difference between it and the negative control.

Lower Limit of Detection (LLD): The LLD is the lowest concentration level that can be determined to be statistically different from a blank at a 99% confidence level. In other words, it is the lowest quantity of a substance that can be distinguished from the absence of that substance (a blank value) within a stated confidence limit, generally 1%. The Limit of Detection is matrix, method, and analyte specific, and can be calculated as follows:

Equation 2. Calculation used in Q-View Software: LLD = 2*(StDev of negative control pixel intensities before ‘Negative Well Subtraction’)* LLOQ / (Difference between pixel intensity of lowest standard and negative control)

Not only are the ULOQ, LLOQ, and LLD valuable measures of range and sensitivity for an individual assay, they can also be used to compare one experimental run to another, or watched over time as a parameter of assay stability. For further reading on quantification limits, see the NCCLS publication EP-17A [18].

Stability

The United States Pharmacopeia (USP) defines stability as “the extent to which a product retains, within specified limits, and throughout its period of storage and use, i.e., its shelf life, the same properties and characteristics that it possessed at the time of manufacture” [19]. The procedures used to establish stability and expiration dating for immunoassay kits depend on the manufacturers, but generally include exposing kits to various conditions that simulate what the kit may experience before use. Quansys stability testing includes:

Overall Kit Stability: The standard curve and a high-level, mid-level, and low-level positive control must perform to within 2 standard deviations of Day 0 data when stored at 4oC. Target: 1 year.

Kit Shipping Stability: The standard curve and a high-level, mid-level, and low-level positive control must perform to within 2 standard deviations of Day 0 data after kits are exposed to 4 hours of either -20oC or 45oC followed by 16 hours at 25oC, then stored at 4oC. Target: 9 weeks.

Opened Kit Stability: The standard curve and a high-level, mid-level, and low-level positive control must perform to within 2 standard deviations of Day 0 data after kits are opened, all regents reconstituted, and opened components are stored at 4oC, -20oC, or -80oC. Target: 4 weeks.

Conclusion

Multiplex ELISAs are useful research tools to simultaneously detect and quantify a many specific substances within a complex biological sample. This paper offers explanations of several terms commonly used to describe method validation parameters associated with such bioanalytical ligand-binding assays, namely selectivity, accuracy, precision, recovery, calibration curves, and stability. In addition, the validation criterion for these parameters at Quansys Biosciences, a manufacturer of multiplex ELISAs, is also provided.

References

  1. Center for Drug Evaluation and Research (U.S.) and Center for Veterinary Medicine (U.S.). Guidance for industry bioanalytical method validation. 2001
  2. Harmonized Guidelines for Single Laboratory Validation of Methods of Analysis, IUPAC Technical Report. Pure Appl. Chem., Vol. 74, No. 5, 2002. p. 835–855,
  3. Anderson, N.L. and N.G. Anderson, The human plasma proteome: history, character, and diagnostic prospects. Mol Cell Proteomics, 2002. 1(11): p. 845-67.
  4. Kusnezow, W., Hoheisel, J.D., Solid supports for microarray immunoassays. J. Mol. Recognit. 16 (2003), 165-176.
  5. Kaplan, I.V. and S.S. Levinson, When is a heterophile antibody not a heterophile antibody? When it is an antibody against a specific immunogen. Clin Chem, 1999. 45(5): p. 616-8.
  6. Bjerner, J., O.P. Bormer, and K. Nustad, The war on heterophilic antibody interference. Clin Chem, 2005. 51(1): p. 9-11.
  7. 7. Ismail, A.A., A radical approach is needed to eliminate interference from endogenous antibodies in immunoassays. Clin Chem, 2005. 51(1): p. 25-6.
  8. Vaidya, H.C. and B.G. Beatty, Eliminating interference from heterophilic antibodies in a two-site immunoassay for creatine kinase MB by using F(ab’)2 conjugate and polyclonal mouse IgG. Clin Chem, 1992. 38(9): p. 1737-42.
  9. Tovey, M.G., M. Wadhwa, and R. Thorpe, WHO international cytokine standards and reference preparations. J Leukoc Biol, 2010. 88(3): p. 425-6.
  10. Chesher, D., Evaluating assay precision. Clin Biochem Rev, 2008. 29 Suppl 1: p. S23-6.
  11. NCCLS, Evaluation of Precision Performance of Quantitative Measurement Methods; Approved Guideline Second Edition., in NCCLS document EP5-A2. 2004.
  12. Bansal, S. and A. DeStefano, Key elements of bioanalytical method validation for small molecules. Aaps J, 2007. 9(1): p. E109-14.
  13. Watterson, T.L., C. PS3-70 Alternative non-linear models for fitting cytokine ELISA curves. in Cytokines 2010: Cancer in Infectious Diseases, Autoimmune Disorders and Cancer. 2010. Chicago, IL: ISICR.
  14. Oda, M., et al., Evidence of allosteric conformational changes in the antibody constant region upon antigen binding. International Immunology, 2003. 15(3): p. 417-426.
  15. Findlay, J.W. and R.F. Dillard, Appropriate calibration curve fitting in ligand binding assays. Aaps J, 2007. 9(2): p. E260-7.
  16. Polettini, A., Applications of LC-MS in Toxicology. 2006. p. 84.
  17. Armbruster, D., Pry, T. Limit of Blank, Limit of Detection and Limit of Quantitation. Clin Biochem Rev, 2008. 29 Suppl 1 p. S49-52.
  18. NCCLS, Protocols for Determination of Limits of Detection and Limits of Quantitation: Approved Guideline. EP-17A. 2004, NCCLS: Wayne, PA.
  19. USP-NF, Pharmaceutical Manufacturing Handbook: Regulations and Quality. 2008

About Us

"We are dedicated to the development of protein arrays that aid researchers and clinicians in better understanding, diagnosing, and treating disease. Our products and services provide value to customers by improving the accuracy and efficiency of their testing."

Quality

Quansys Biosciences is an ISO 9001 and ISO 13485 registered company. We pride ourselves in the quality of our product and our commitment to customer satisfaction.


Quansys Bioscience
Fax: 435-750-6869
888-QUANSYS (782-6797)