Introduction
Electrochemiluminescence (ECL) based ligand binding bridging assays to support immunogenicity testing have been widely published (1-5). These assays typically require biotherapeutic-specific (TS) anti-biotherapeuticor drug antibody (ADA) positive control generation and labeling of a capture and detection reagent (ie, biotinylation and ruthenylation). Preparation and characterization of these TS critical reagents can be time consuming and costly, making their application limited for support of time sensitive preclinical studies during early biotherapeutic development. The assay proposed in this manuscript eliminates the need for TS reagents by implementing a universal positive control and antibody detection reagent. Surface plasmonresonance (SPR)-based immunoassays utilizing Biacore technology are also widely used in the industry (6, 7). Although this platform does not require labeled reagents, the method typically has to be optimized for immobilization and regeneration of each biotherapeutic, often not amenable to early research support. Up-front assay development time can be eliminated by implementing a universal method applicable across programs. The universal indirect species-specific assay (UNISA) supports utilization of one method across all biotherapeutics/ species that require immunogenicity assessments while retaining the sensitivity and dynamic range associated with the readout of the ECL based assay format on the MSD Sector Imager 6000.
This supplemental text will highlight the methods and results for the qualification of the UNISA across all species and the further robust validation performed for the cynomolgus monkey-specific UNISA.
Materials and Methods
Chemical and Reagent Preparation:
Species-specific Isotype/ subclass antibodies: Mouse IgG1, IgG2a, IgG2b, and IgG3 isotype controls (R&D Systems; catalog numbers MAB002, MAB003, MAB004, and MAB007 respectively), rat IgG1, IgG2a, and IgG2bisotype controls (R&D Systems; catalog numbers MAB005, MAB006, and MAB0061 respectively), cynomolgus monkey IgG1, IgG2, IgG3, and IgG4 isotype controls (Amgen Inc., described by Jacobsen et al. (8)) were utilized to determine the specificity of the secondary detector. All other chemical and reagent preparation captured in the materials and methods section of the main text of manuscript.
ADA Immunoassay: Universal Indirect Species-specific Assay (UNISA)
The UNISA is a species-specific indirect sandwich assay utilizing the ECL technology as the readout (Figure 1). Reagents were purchased from MSD Inc, Gaithersburg, MD (MSD Sulfo-TAG NHS-Ester, standard-bind bare plate, and 4× read buffer T) and KPL, Gaithersburg, MD (20× milk block/ diluent and 20× wash solution). Briefly, the MSD plate was coated overnight with 35 µL of a biotherapeutic antibody candidate diluted to 1 µg/mL in PBS. The serum samples were diluted to 1% in 5× KPL block, either untreated (screening test) or treated with excess relevant or irrelevant biotherapeutic (specificity analysis or competitive binding test). The coated and blocked plates (blocked with 5× KPL, 200 µL/well overnight) were washed on Day 2 (wash procedure for all wash steps; 1× KPL wash, 3x300 µL/well), and 35 µL of 1% serum sample was added to the plate well and incubated for approximately 1 hour. Plates were washed and 35 µL of ruthenylated species-specific detection antibody at 0.5 µg/mL in 5xKPL was added and incubated for approximately 30 minutes. Following another wash, 2× MSD T read buffer was added (150 µL/well). The plates were finally read using the SECTOR® Imager 6000 Instrument (MSD, Gaithersburg, MD, USA) plate reader, where an electrical current was placed across the plate-associated electrodes, resulting in a series of electrically induced oxidation-reduction reactions involving ruthenium (from the bound secondary detector antibody) and tripropylamine (from the MSD T read buffer). The resulting electrochemiluminescence was measured. No acid dissociation was performed for this assay format.
Figure 1. Depiction of the potential binding events in the Universal Indirect Species-specific Assay (UNISA). The stoichiometry of the biotherapeutic coated on the plate (MSD) allows the detection of both anti-ID and anti-Fc or framework (non-ID) specific antibody detection. The signal of the sample over the background response of the pooled species-specific serum (S/N) is proportional to the amount of anti-biotherapeutic antibody (ADA) present in the sample.
Tests and reporting criteria: Samples were reported based on the testing strategy summarized in Figure 2.
Figure 2. UNISA result flow diagram. Note: Diagram does not necessarily reflect the assay flow as detailed in the method.
Cut Point Determination
The assay cut point (ACP) is the S/N value that is used to differentiate between negative and potentially ADA containing samples. The depletion cut point (DCP) is the percent S/N reduction that is used to define the specificity of samples. A combination of the 2 cut points identifies a positive ADAsample.
Qualification Across all Species: To be consistent across programs and aid the rapid turnaround of results required during early discovery support, universal cut points were applied after evaluating a subset of animals in the UNISA specificity test across 1 fully human monoclonal antibody (hMab1) based biotherapeutic as follows: ACP (S/N > 1.5); DCP (%Depletion > 50). These universal cut points were then applied to all study support and monitored to ensure sensitivity in detecting potential ADA positive animals. Subsequent to the case studies shown in this manuscript, a retrospective analysis of the historical data led to the DCP creation of 20%. This value was in better alignment with the assay cut point and enabled a more conservative assessment of the response specificity in determining the final antibody conclusion status. All current UNISA support across species utilizes an ACP of 1.5 and a DCP of 20%.
Validation of the Cynomolgus Monkey-specific UNISA: Twenty eight normal cynomolgus monkeys were split into 2 group sequences with gender as a stratification factor. Group sequence 1 (S1) contained 9 male and 9 female animals. Group sequence 2 (S2) contained 5 male and 5 female animals plus an 8-point ADA titration curve (see sensitivity section below for further details). All samples were then tested according to a statistically derived experimental design model to evaluate the assay cut point (ACP) and depletion cut point (DCP) for 4 hMabs (A, B, C, and D) in the UNISA specificity test (Table I). In addition, ruggedness of the UNISA across all 4 hMabs could be explored using a mixed effect model applied to the S/N to study the effect of analyst, secondary detector, plate lot, and plate coating and their interactions on the assay performance with all sample types and assay controls. To determine the cut points, the following statistical methods were employed:
a)ACP was calculated as the upper bound of a one-sided 99% prediction interval for the distribution of the assay values (S/N). The form of the equation utilized was:
U99 = LS-mean + TINV(.99, DF)*SQRT(Variancetotal + VarianceLS-mean)
b)DCP was calculated using equation: 100% - L99 of %T/U, where L99 is the lower bound of a one-sided 99% prediction interval for the distribution of the %T/U values. The form of the equations utilized are:
%T/U: (S/N of untreated/ S/N of treated )*100
L99 = LS-mean - TINV(.99, DF)*SQRT(Variancetotal + VarianceLS-mean)
Table I. Cynomolgus Monkey Specific UNISA Validation Design of Experimentsa
Assay Run / Analyst / Secondary Detector / Plate Lot / Plate Coating / Group Sequence1 / 1 / 1 / 1 / 1 / S1
2 / S2
2 / 1 / 2 / 2 / 1 / S2
2 / S1
3 / 2 / 2 / 1 / 1 / S2
2 / S1
4 / 2 / 1 / 2 / 1 / S1
2 / S2
aDesign of experiments utilized to validate UNISA across 4 biotherapeutics while studying the ruggedness of the assay. All 4 assay runs were repeated across each biotherapeutic in the specificity analysis. Plate coating refers to 1 MSD 6000 bare plate lot that was coated with the appropriate biotherapeutic in 2 different preparations by 2 different analysts. Group sequence refers to a unique combination of cynomologus monkey animals and pooled cynomolgus monkey serum spiked with cyno-specific positive control antibody.
Assay Sensitivity
The assay sensitivity is defined as the lowest ADA concentration that gives aS/N response equivalent to the ACP.
Qualified Across all Species: The species-specific (SS) positive control was titrated from 7.8 to 1000 ng/mL in SS-pooled serum and analyzed once against hMab1. The S/N values were then analyzed against the SS-positive control concentrations in GraphPad Prism v5.04 using the following 4PL regression model: log(agonist) vs. normalized response -- Variable slope. The interpolated concentration equal to the universal ACP (S/N of 1.5) was captured as the assay sensitivity (Table I, main manuscript).
Validation of the Cynomolgus Monkey-specific UNISA: The cyno-specific UNISA universal positive control (mouse anti-human IgG/cynomolgus monkey Fc chimeric antibody or cyno-ADA) was titrated from 1000 to 0.0078 ng/mL (2-fold dilution) in cynomolgus monkey pooled serum (PNCS). This titration curve was then analyzed following the design model captured in supplemental table 1 against all 4 hMabs as part of group sequence 2. The S/N values were then analyzed against the cyno-ADA concentrations by a biostatistician. Whenever the non-linear regression model, such as 4PL, was not well defined for back-calculating the assay sensitivity, the first degree polynomial regression was applied to the log transformed S/N ratios vs. log transformed cyno-ADA concentration using the data in the range as noted.
Biotherapeutic tolerance
Biotherapeutic tolerance is defined as the highest drug concentration that still gives a S/N response above the ACP for each ADA concentration evaluated.
Qualified Across all Species: The hMAb1 was titrated from 1000 to 10 µg/mL in SS-pooled serum containing 500 ng/mL of the SS-positive control and analyzed once against hMab1. The S/N values for each hMab were then analyzed in GraphPad Prism v5.04 using the following 4PL regression model: log(inhibitor) vs. normalized response -- Variable slope. The interpolated concentration equal to the universal ACP (S/N of 1.5) was captured as the biotherapeutic tolerance level of the assay. In addition, once the tolerance level was determined, the molar ratio (ADA to biotherapeutic) could be determined (Table I, main manuscript).
Validation of the Cynomolgus Monkey-specific UNISA: The 4 hMabs (A, B, C, and D) were titrated from 1000 to 15.625 µg/mL (2-fold dilution) in PNCS containing 500 ng/mL of the cyno-ADA and analyzed once against their respective hMab. The S/N values for each hMab were then analyzed in GraphPad Prism v5.04 using the following 4PL regression model: log(inhibitor) vs. normalized response -- Variable slope. The interpolated concentration equivalent to the statistically derived ACP for each hMAb was captured as the biotherapeutic-specific tolerance level of the assay. In addition, once the tolerance level was determined, the molar ratio (ADA to biotherapeutic) was expressed.
Specificity
Qualified Across all Species - Secondary detector specificity: A test was performed to examine specificity of the ruthenylated SS-secondary detectors for their ability to bind to different SS-IgGisotype subclasses. MSD 6000 bare plates were coated with either mouse-IgG1, IgG2a, IgG2b or IgG3 (mouse-UNISA test) or rat-IgG1, IgG2a or IgG2b (rat-UNISA test) each at 0.5 and 1.0 µg/mL and processed per the method for incubating, blocking and washing. No samples were added to the plates. After washing the coating material, SS-detector was added and tested at concentrations of 250 and 500 ng/mL. Plates were processed per the method for detector incubation, washing and plate reading. ECL signals from a minimum of four wells were averaged for each condition. For purposes of this assessment, the cynomolgus monkey specific subclasses were tested using the Biacore 3000 by immobilizing the UNISA cyno-detector on a biosensor chip (immobilization range aim for 5000 response units) then passing each subclass (at 5 µg/mL) over the immobilized surface (flow rate of 5 µL/ min; 3 minutes,
100 mM HCL regeneration solution between cycles). The response units for each subclass (1 per cycle) were then captured.
Figure 3. Evaluation of the subclass specificity of the UNISA secondary antibody detectors. Specificity against IgG4 and IgG3/ IgG4 not evaluated for the mouse and rat detection antibodies respectively. a For the anti-cyno IgG2 subclass, a or b was not specified.
Validation of the Cynomolgus Monkey-specific UNISA - Biotherapeutic Target Interference Assessment: To evaluate the specificity of this assay format for antibody detection, impact from excess soluble target was assessed in hMab C. A titration curve (0 to 10 µg/mL) of hMab C soluble ligand and soluble target was spiked into PNCS with and without the universal cyno-ADA (0, 100 and 500 ng/mL, levels 1-3 respectively) and tested once in the UNISA screening assay against hMab C. For comparison purposes, the same soluble ligand/ receptor was spiked into PNCS with 0, 250, and 1000 ng/mL(levels 1, 4, and 5 respectively) of TS-positive control antibody (rabbit anti-hMab C polyclonal antibody) and tested once in the validated TS-bridging immunoassay. Level 1 (0 ng/mL of ADA) should produce a response less than or equal to the ACP, regardless of soluble ligand or soluble target concentration. For the PNCS spiked with ADA (levels 2-5), all samples should produce a response greater than the ACP, regardless of soluble target/ ligand concentration.
Results
Qualified Parameters
A summary of the qualified parameters are provided as Table I in the main manuscript text. In addition, Figure 3 represents the specificity of the detector across the subclass evaluated. As all combinations of isotype control coating (0.5 or 1 µg/mL) and ruthenylated detector (0.25 and 0.5 µg/mL) yielded similar results, only results for 0.5 µg/mL of both coating and detector are shown. The mouse-specific detector demonstrated different reactivity for the isotypes tested with the strongest specificity to the IgG2bsubclass. Overall, the detector was sensitive enough to detect acceptable levels of IgG1, IgG2aand IgG2b; however, the detector demonstrated weak specificity to IgG3. As a result, IgG3 may not be detected in an antibody assessment using the current detector and conditions. The rat-specific detector had similar reactivity for both IgG2aand IgG2b, but was about 50% less reactive to IgG1. Overall, the detector was sensitive enough in detecting all isotypes and subclasses evaluated. The cyno-specific detector had similar reactivity for both IgG1 and IgG2, was about 50% less reactive to IgG3 and demonstrated week specificity for IgG4. As a result, IgG4 may not be detected in an antibody assessment using the current detector and conditions. At this time, IgM, IgE, and IgA were not evaluated for specificity, as the scope for this assay is to detect high affinity IgG responses.
Cynomolgus Monkey-specific UNISA Validation
The UNISA validation results for hMab A-D showed the following: ACP range of 1.25 to 1.38 (99% prediction limit), DCP range of 23 to 45% (99% prediction limit), a sensitivity range of 6 to 8 ng/mL and a biotherapeutic tolerance ranging from 272 to 403 µg/mL at 500 ng/mL of cyno-ADA (Table II, Figure 4).
Figure 4. Assay sensitivity and biotherapeutic tolerance levels evaluated across 4 hMabs. Assay sensitivity utilized the cyno-specific UNISA positive control (cyno-ADA) titrated in pooled normal cynomolgus monkey serum (PNCS) from 7.8 to 1000 ng/mL; Biotherapeutic tolerance utilized the cyno-ADA spiked at 500 ng/mL in PNCS with an 8 point dose-response curve of each individual hMab from 15.625 to 1000 µg/mL. Responses from the screening UNISA test (signal to noise or S/N) were plotted against the ADA (ng/mL, solid line) or biotherapeutic (µg/mL, dashed line) concentration to evaluate the variability between the 4 hMAbs and interpolate the biotherapeutic tolerance levels. Observe that the slope and range of the ADA dose-response between all 4 hMabs is consistent and the slope of the biotherapeutic tolerance dose-response curve is similar for hMabs B-D.
Table II. Summary of Universal Validation Parametersb
Fully Human Monoclonal Antibody Biotherapeutic: / A / B / C / DAssay Cut Point (ACP); Response: Signal to Noise (S/N)
Number of Donors (Used in Analysis*) / 28 (26) / 28 (27) / 28 (28) / 28 (27)
Number of Samples (Used in Analysis*) / 112 (103) / 112 (107) / 112 (111) / 112 (108)
Min, Max* / 0.75,1.49 / 0.77, 1.35 / 0.77, 1.36 / 0.80, 1.35
Upper Bound on:
95% Prediction Limit / 1.23 / 1.20 / 1.23 / 1.15
99% Prediction Limit / 1.35 / 1.30 / 1.32 / 1.25
Depletion Cut Point (DCP); Response: %
Number of Donors (Used in Analysis*) / 28 (26) / 28 (27) / 28 (28) / 28 (27)
Number of Samples (Used in Analysis*) / 112 (103) / 112 (103) / 112 (110) / 112 (108)
Min, Max* / -36, 31 / -27, 13 / -20, 22 / -29, 20
Lower Bound on:
95% Prediction Limit / 21% / 14% / 16% / 10%
99% Prediction Limit / 30% / 20% / 21% / 16%
Sensitivity (ng/mL) at 99% ACP
Model: Linear Regression (7.8125 – 125 ng/mL); N=4 / 8.06 / 6.73 / 7.41 / 6.52
L95, U95 / 4.95, 13.11 / 4.70, 9.65 / 4.37, 12.55 / 4.24, 10.02
Biotherapeutic Tolerance (µg/mL) with 500 ng/mL ADA at 99% ACP
Model: 4PL Regression; N=1 / 403 / 302 / 272 / 311
ADA to Biotherapeutic Molar Ratio / 1:806 / 1:604 / 1:544 / 1:622
*After outlier removal.
bThe ACP, DCP, and assay sensitivity were established based on the data generated following the design of experiments in supplemental table I and analyzed by a biostatistician according to the appropriate methods section. Biotherapeutic tolerance was determined by running an 8 point dose-response curve (15.625 to 1000 µg/mL) of each hMab in pooled cynomolgus monkey serum containing 500 ng/mL of cyno-specific UNISA positive control antibody. The concentration corresponding to the ACP at a 99% prediction limit for each biotherapeutic was then interpolated on GraphPad Prism v5.04.
The variability assessment across the 4 hMabs for normal cynomologus monkeys demonstrated that the animal ID was the highest contributor to variability across all variables tested, which is due to biological variability and has no relevance on assay variability. Additionally, for hMabA, 22% of the variability could be attributed to analyst and plate lot, compared to negligible contribution for all variables against hMab B, C, and D(Table IIIa). For PNCS spiked with increasing concentrations of the cyno-ADA, percent contribution to total assay variability was evaluated for the combined data set (all curves across all hMab products; N=16) and does not take into account impact of the hMAb on all interactions. Thus, for the combined cyno-ADA titration curves, the main contributor to assay variability was observed for analyst to analyst, where each analyst consistently and purposefully used different incubation parameters to include flexibility in the final validated method (i.e. room temperature controlled incubator vsbenchtop; orbital vsmicromix shaking) (Table IIIb). For both the animals and the cyno-ADAtitration curve, the total %CV never exceeded 20%, thus demonstrating that the assay was in control and robust for the variance components evaluated during validation.
Table III. Ruggedness Demonstrated Across Variables and Biotherapeuticsc