Immune Tolerance-Adjusted Personalized Immunogenicity Prediction for Pompe Disease

Abstract

Infantile-onset Pompe disease (IOPD) is a glycogen storage disease caused by a deficiency of acid alpha-glucosidase (GAA). Treatment with recombinant human GAA (rhGAA, alglucosidase alfa) enzyme replacement therapy (ERT) significantly improves clinical outcomes; however, many IOPD children treated with rhGAA develop anti-drug antibodies (ADA) that render the therapy ineffective. Antibodies to rhGAA are driven by T cell responses to sequences in rhGAA that differ from the individuals' native GAA (nGAA). The goal of this study was to develop a tool for personalized immunogenicity risk assessment (PIMA) that quantifies T cell epitopes that differ between nGAA and rhGAA using information about an individual's native GAA gene and their HLA DR haplotype, and to use this information to predict the risk of developing ADA. Four versions of PIMA have been developed. They use EpiMatrix, a computational tool for T cell epitope identification, combined with an HLA-restricted epitope-specific scoring feature (iTEM), to assess ADA risk. One version of PIMA also integrates JanusMatrix, a Treg epitope prediction tool to identify putative immunomodulatory (regulatory) T cell epitopes in self-proteins. Using the JanusMatrix-adjusted version of PIMA in a logistic regression model with data from 48 cross-reactive immunological material (CRIM)-positive IOPD subjects, those with scores greater than 10 were 4-fold more likely to develop ADA (p<0.03) than those that had scores less than 10. We also confirmed the hypothesis that some GAA epitopes are immunomodulatory. Twenty-one epitopes were tested, of which four were determined to have an immunomodulatory effect on T effector response in vitro. The implementation of PIMA V3J on a secure-access website would allow clinicians to input the individual HLA DR haplotype of their IOPD patient and the GAA pathogenic variants associated with each GAA allele to calculate the patient's relative risk of developing ADA, enhancing clinical decision-making prior to initiating treatment with ERT. A better understanding of immunogenicity risk will allow the implementation of targeted immunomodulatory approaches in ERT-naïve settings, especially in CRIM-positive patients, which may in turn improve the overall clinical outcomes by minimizing the development of ADA. The PIMA approach may also be useful for other types of enzyme or factor replacement therapies.

Keywords: Pompe Disease (glycogen storage disease type II); Tregitope; acid alpha-glucosidase (GAA); anti-drug antibodies (ADA); cross-reactive immunological material (CRIM); enzyme replacement therapy (ERT); immune tolerance induction (ITI); personalized immunogenicity assessment (PIMA).

Figure 1

Overview and evolution of the PIMA scoring algorithms used to calculate the ADA risk assessment score. V1 was previously published. V2-V3 are intermediate steps to developing V3J as described here. V2 added subject-specific HLA DR epitope prediction, V3 added conservation with nGAA at the TCR face of epitopes, and V3J examine the potential for certain nGAA epitopes to be tolerogenic by comparing the sequence to other human genome epitopes. The most accurate for this cohort was V3J, which adjusted the prediction for T cell epitopes that are cross-conserved with other self-epitopes (not confined to nGAA). In contrast to V1, where epitopes mismatched between rhGAA and nGAA of individuals with homozygous nGAA mutations were counted once, V2-V3J included mismatched epitopes twice in the calculations, once for each allele. We then identified specific T cell epitopes in GAA that generated tolerance in vitro. Individual immune tolerance to nGAA sequences may diminish the risk of ERT-related ADA.

Figure 2

Epitope differences in therapeutic GAA from endogenously expressed native GAA predicted to drive ADA or be tolerized. CRIM-positive IOPD subjects who express residual nGAA may be tolerant to epitopes conserved, for their HLA, with the rhGAA replacement protein. (A) T cell epitopes contained within the rhGAA may be recognized as “foreign” if they are within the truncated or mutated portions of the patient-specific nGAA. (B) T cell epitopes within the rhGAA that contain T cell receptor (TCR)-facing residues that are different from those found in nGAA may be sufficient to generate a different T cell phenotype response. (C) T cell epitopes within the rhGAA that contains different MHC-facing residues but the same TCR-facing residues as epitopes found in nGAA are predicted to be tolerated by the immune system (this hypothesis was included in PIMA V3). (D) The presence of a T cell epitope in the rhGAA sequence with TCR-facing residues highly cross-conserved with several self-human proteins may not appear as foreign and would also be tolerated by the immune system. (This hypothesis was included in PIMA V3J).

Figure 3

Overall agreement of the four PIMA scoring algorithms as predictors of ADA status. Among the 48 IOPD subjects evaluated, 19 (40%) developed high ADA titers; this includes subjects with high and sustained antibody titers (HSAT) as well as sustained intermediate titers (SIT). The remaining 29 subjects (60%) developed low ADA titers. PIMA V1 (previously published) accurately predicted the ADA status for 56% of the IOPD subjects. Versions V2 and V3 accurately predicted 54% of subjects, thus further adjustment was explored. Adjusting for potential Treg epitopes with V3J improved accuracy to 64% of the subjects.

Figure 4

GAA-peptides bind to HLA DR1 as predicted. Selected GAA peptides were evaluated for HLA DRB1 binding in vitro and IC₅₀ values were calculated. hGAA-1a, hGAA-1b, hGAA-2, hGAA-6, hGAA-11 and hGAA-13 FV621 peptides bound with the multiple alleles tested (DRB1*0101, *0301, *0401, *0701, *1101, *1301 and *1501) whereas hGAA-12 was predicted to be more HLA-restricted and consequently had limited binding to HLA. A seven-point competition assay using a validated control peptide was performed; color coding reflects binding affinity IC₅₀ was determined by interpolation. Using a standard Z-score threshold of 1.64 (top 5%), overall positive predictive value for EpiMatrix predictions was 92%, with sensitivity of 79%. False negatives are not uncommon when testing peptides containing significant predictions for several alleles (EpiBars), as many contain “near-miss” Z-scores in the top 10% of predicted peptides. Note that peptide GAA-12 and 13 were not selected for promiscuity: they were designed for individualized testing in a specific patient for which both the mutation and the HLA DRB1 allele restriction concurred.

Figure 5

GAA-derived peptides inhibit memory CD4 T cell response to Tetanus Toxoid (TT) in healthy donors. (A) Representative flow cytometry dot plots show CD4 memory T cell proliferative response to TT and dose-dependent inhibition by hGAA-6 peptide. (B) Inhibition of CD4+T cell recall response by GAA-peptides in TTBSA. PBMCs from healthy donors were stimulated with 0.5 µg/ml of TT with or without FV621 or GAA-peptides and analyzed at six days post-stimulation by flow cytometry for inhibition of CD4+ T cell proliferation. Data are the representative donor from 5 donors in the experiments. Significant suppressive capacity of CD4+ T cell proliferation was observed for 4 putative Treg peptides in GAA confirming their regulatory potential across all donors tested. P values * ≤ 0.05, ** ≤ 0.005 and *** ≤ 0.0005 represents statistical significance between peptide stimulation vs TT using a two-tailed t test. GAA peptides 1a, 1b, 6 and 11 significantly suppressed the expansion of TT-memory T cells in this in vitro assay.

Figure 6

GAA-peptides modulate T regulatory to T effector cell ratio. Healthy donor PBMCs were stimulated with Tetanus Toxoid (TT) with or without GAA-selected peptides for 7 days and analyzed by flow cytometry. (A) Representative flow cytometry dot plots show the effect of hGAA-6 peptide on the inhibition of T effector (CD4+CD25^hiFoxP3^int) cells for a single donor. (B) The effect of hGAA-6 peptide on regulatory T cells (CD4+CD127^lowCD25^hiFoxP3^hi) in the representative donor is shown. (C) Representative histogram indicates the effect of selected GAA peptides on the Treg to Teff ratio in an individual donor. GAA peptides 1a, 1b, 6 and 11 significantly increased the Treg : Teff ratio similar to the FV621 Tregitope control. P values * ≤ 0.05, ** ≤ 0.005, *** ≤ 0.0005 and **** ≤ 0.00005 represents statistical significance between peptide stimulation at a given concentration vs TT using a two-tailed t test.

Figure 7

The Pompe PIMA user interface prototype. The upper screenshot shows the Pompe PIMA homepage with data management options. The lower two screenshots from the Upload New Patient page feature the required (*) input fields to generate the IOPD patient’s individualized ADA risk assessment score.