Immune Complex Formation Is Associated With Loss of Tolerance and an Antibody Response to Both Drug and Target

Abstract

AMG 966 is a bi-specific, heteroimmunoglobulin molecule that binds both tumor necrosis factor alpha (TNFα) and TNF-like ligand 1A (TL1A). In a first-in-human clinical study in healthy volunteers, AMG 966 elicited anti-drug antibodies (ADA) in 53 of 54 subjects (98.1%), despite a paucity of T cell epitopes observed in T cell assays. ADA were neutralizing and bound to all domains of AMG 966. Development of ADA correlated with loss of exposure. In vitro studies demonstrated that at certain drug-to-target ratios, AMG 966 forms large immune complexes with TNFα and TL1A, partially restoring the ability of the aglycosylated Fc domain to bind FcγRIa and FcγRIIa, leading to the formation of ADA. In addition to ADA against AMG 966, antibodies to endogenous TNFα were also detected in the sera of subjects dosed with AMG 966. This suggests that the formation of immune complexes between a therapeutic and target can cause loss of tolerance and elicit an antibody response against the target.

Keywords: AMG 966; TL1A; TNFα; anti-drug antibodies; immune complexes; immunogenicity; inflammatory bowel disease; tolerance.

Figure 1

In vitro T cell assays did not reveal sequence-based risk of immunogenicity for AMG 966. T cell assays were performed with naïve donors representative of global HLA allele frequencies. Results are shown as stimulation index, or test protein divided by the baseline condition (media alone). (A) PBMC from 50 donors were stimulated with each test protein for 6 days prior to assessment of CD4 T cell proliferation by flow cytometry. A control monoclonal antibody and an IgG1 Fc domain were tested both with charge pair mutations (CPM) and without (stnd). (B) Monocytes from 30 PBMC donors were differentiated into dendritic cells, loaded with test protein, and matured. Autologous CD4 T cells were isolated and co-cultured with mature dendritic cells presenting test protein agretopes for 6 days prior to assessment of CD4 T cell proliferation by flow cytometry. Fab domains were tested with and without CPM.

Figure 2

AMG 966 was highly immunogenic in all cohorts. (A) The magnitude of the anti-AMG 966 antibody response (signal to noise ratio) is shown over time. Every subject in the SAD cohorts developed anti-AMG 966 antibodies. The study duration varied with dose, and the end of study antibody sample was taken on day 85 for cohort 1, day 113 for cohorts 2-5, and day 141 for cohort 6. (B) S/N for all AMG 966 dosed subjects from the MAD cohorts is shown. 17 of 18 subjects developed anti-AMG 966 antibodies by end of study, day 169. Subjects in these cohorts were dosed Q2W with the last dose administered on day 71. (C) Pre-treatment of serum samples with protein G or protein G/L beads depleted signal in the immunoassay to background levels, confirming that signal in the immunoassay is antibody derived.

Figure 3

Anti-AMG 966 antibodies impacted exposure in MAD cohorts. (A) AMG 966 concentration is plotted together with the magnitude of the anti-AMG 966 antibody response, shown as signal to noise ratio. Representative subjects are shown from each MAD cohort. Subjects were dosed Q2W with the last dose administered on day 71. (B) AMG 966 area under the curve (AUC) is shown pre and post ADA onset in cohorts 7 and 8. ADA onset was defined as the first study timepoint with an ADA positive result. For cohort 9, all subjects maintained exposure throughout the dosing phase, and ADA onset was too late to assess impact on AUC.

Figure 4

Anti-AMG 966 antibodies bound to all domains of AMG 966 and became specific for TL1A charge pair mutations over time. Domain characterization was performed by pre-treating serum samples with various domains of AMG 966 and then re-testing in the ADA assay to determine the extent to which assay signal was depleted. Percent depletion was calculated by dividing the difference between treated and untreated samples by the untreated signal. (A) Antibody positive samples from 19 subjects from cohorts 1-4 were assessed for binding to each domain of AMG 966 or bulk human IgG negative control. The timepoint for each sample is indicated on the y-axis. (B) Paired antibody positive samples from the same subject were analyzed at day 15 and day 85/113 to explore how the specificity of the antibody response changes over time. Data shown represent 2 paired samples from cohort 1, 2 pairs from cohort 2, and 1 pair from cohort 3. (C) Antibody positive samples were pre-treated with either a wild-type domain or a domain containing the charge pair mutations to assess the extent to which antibody binding was dependent on charge pair mutations. Data shown were derived from 19 subjects from cohorts 1-4.

Figure 5

AMG 966 forms large complexes with TNFα and TL1A in vitro. SEC-MALS analysis was performed after overnight incubation of various therapeutic proteins and trimeric target at different ratios. (A) Etanercept, adalimumab, and AMG 966 were analyzed either alone, or after being mixed in a 1:1 ratio with TNFα. (B) AMG 966 was combined with TNFα and TL1A at various ratios and analyzed.

Figure 6

AMG 966 complex formation with targets restored binding to FcγRIa and FcγRIIa. Competitive binding assays were utilized to assess the impact of AMG 966 complex formation with targets on FcγR binding. An IgG1 positive control antibody was used in each assay and compared to AMG 966 alone and various ratios of AMG 966/target complex for ability to bind FcγRIa (A), FcγRIIa (B), or FcγRIIIa (C).

Figure 7

TNFα is bound to immunoglobulin and accumulates in serum over time. (A) Serum TNFα was measured for all subjects in the SAD cohorts throughout the course of the study. (B) Serum samples from cohort 1 at day 85 were pre-treated with protein G/L beads or blank sepharose beads and tested alongside untreated samples in the TNFα assay to assess whether TNFα was bound to immunoglobulin. (C) Serum samples were pre-treated with excess infliximab and re-tested in the ADA assay to confirm that the assay was not yielding false positive results due to 1:1 complexes of TNFα and endogenous anti-TNFα antibody. A high concentration of TNFα known to give a false positive was used as a control and tested alongside low and high positive controls (polyclonal anti-AMG 966 antibody).

Figure 8

AMG 966 elicited endogenous antibodies to TNFα but not TL1A. Subjects from cohorts 2 and 7 were tested for endogenous antibodies to TNFα and TL1A using an SPR approach. One AMG 966-treated subject from cohort 7 who withdrew from the study early was not included. Samples with reactivity for TL1A or TNFα based on anti-human IgG detection were further tested using the AMG 966-specific 1A3 reagent for detection. Asterisks indicate subjects in the placebo group. Cohort 2 (A) and cohort 7 (B) were assessed for TL1A binding antibodies and reactive samples were further tested for 1A3 binding (C, D). Cohort 2 (E) and cohort 7 (F) were also tested for TNFα binding antibodies and reactive samples were further tested for 1A3 binding (G, H).