In 1985, Jacks and Varmus described the first -1 ribosomal frameshifting, from which they established the canonical model of eukaryotic -1 frameshifting site (Jacks et al., 1985 and 1988). Today, several tens of viruses and one mouse nuclear gene (Shigemoto et al., 2001) have been identified as bearing such a -1 frameshifting site. A typical site contains a “slippery” heptamer in 5', where both A and P site tRNAs slip by one nucleotide upstream, followed by a stimulatory structure (stem loop, or pseudoknot) downstream (Brierley et al., 1989). The “slippery” heptamer is separated from the stimulatory structure by a short sequence (3 to 11 nucleotides), the so-called “spacer”. Based on this model, several studies have been undertaken to identify frameshifting sites in the nuclear genome of the yeast Saccharomyces cerevisiae (Hammell et al., 1999 and Liphardt, 1999). However, none of these allowed to identify with certainty authentic expressed genes controlled by -1 frameshifting. Two reasons might be proposed: first, the model might not be precise enough, leading to the identification of too many false positive candidates (Bekaert et al., 2003); conversely the model might be too rigid, failing to identify true positive candidates. This would be the case, for example, if -1 frameshift could be directed by a more degenerated structure, or by mechanisms that rely on other types of signals.
Checkpoint inhibitors (CPIs) have significantly enhanced cancer treatment, yet formation of antidrug antibodies (ADA) can reduce drug efficacy and lead to increased immune toxicity.1 Identifying biomarkers predictive of ADA formation is crucial to optimize administration of CPIs. Human leukocyte antigen (HLA) genes are responsible for presentation of antigens to T cells and harbor substantial inter-patient variation. A recent study2 identified allelic variation in the HLA-DRB1 gene as a risk factor for ADA formation in CPI. For non-CPI-directed cancer immunotherapies (CITs), aggregated clinical data is scarce, and the role of HLA in ADA formation remains largely unexplored.
Methods
To investigate associations between HLA alleles and ADA formation, we established a harmonized data mart from 23 early-phase CIT trials, encompassing 12 molecules with diverse mode of actions and various cancer indications. This comprehensive dataset includes a total of 3568 patients, both CPI-naïve and CPI-experienced, and features clinical and two-field class I and II HLA alleles imputed from genotyping arrays.3 Treatment-induced ADA status was determined using a standardized set of definitions based on therapeutic antibody-specific titers collected longitudinally.4 We adopted an HLA-wide approach5 to assess associations with persistent ADA formation, focusing on a patient population of European ethnicity to account for differences in HLA allele frequencies and correcting for potential confounders of ADA such as age and gender.
Results
Associations with the HLA alleles are summarized on a per molecule basis. After Bonferroni correction for HLA alleles, no statistically significant association was identified. However, the results for CPI as a standard-of-care combination therapy showed an odds ratio of 1.84 (CI 95% 1.02–3.21) for the DRB1*01:01 variant, consistent with the main finding previously reported in a set of larger atezolizumab monotherapy trials.2 Furthermore, consistency in the ranking of HLA alleles associated with ADA across the different therapeutic antibodies was low, with Spearman correlations ranging from 0.02 to 0.15.
Conclusions
This dataset has enabled a systematic investigation of the HLA region's role in ADA formation linked to non-standard CITs. Low correlations amongst rankings of the associations suggest that genetic predisposition to ADA formation is likely molecule-specific. However, these results may also be attributed to low statistical power and other confounding factors. As a next step, we aim to integrate in-silico prediction algorithms6 for T cell epitopes to factor in the immunogenicity cascade and elucidate the differences seen across the therapeutic antibodies.
Acknowledgements
Members of the Enhanced Data Insights and Sharing community at Genentech and Roche who developed, curated, and integrated data for this work in the Cancer Immunotherapy Data Mart.
References
van Brummelen EM, Ros W, Wolbink G, Beijnen JH, Schellens JH. Antidrug antibody formation in oncology: clinical relevance and challenges. Oncologist 2016 Oct;21(10):1260–1268.doi: 10.1634/theoncologist.2016-0061. Epub 2016 Jul 20. PMID: 27440064;PMCID: PMC5061540. Hammer C, Ruppel J, Kamen L, Hunkapiller J, Mellman I, Quarmby V. Allelic variation in HLA-DRB1 is associated with development of antidrug antibodies in cancer patients treated with atezolizumab that are neutralizing in vitro. Clin Transl Sci 2022 Jun;15(6):1393–1399. doi: 10.1111/cts.13264. Epub 2022 Apr 8. PMID: 35263013; PMCID: PMC9199883. Zheng X, Shen J, Cox C, Wakefield JC, Ehm MG, Nelson MR, Weir BS. HIBAG--HLA genotype imputation with attribute bagging. Pharmacogenomics J 2014 Apr;14(2):192–200. doi: 10.1038/tpj.2013.18. Epub 2013 May 28. PMID: 23712092; PMCID: PMC3772955. Shankar G, Devanarayan V, Amaravadi L, Barrett YC, Bowsher R, Finco-Kent D, Fiscella M, Gorovits B, Kirschner S, Moxness M, Parish T, Quarmby V, Smith H, Smith W, Zuckerman LA, Koren E. Recommendations for the validation of immunoassays used for detection of host antibodies against biotechnology products. J Pharm Biomed Anal 2008 Dec 15;48(5):1267–81.doi: 10.1016/j.jpba.2008.09.020. Epub 2008 Sep 19. PMID: 18993008. Migdal M, Ruan DF, Forrest WF, Horowitz A, Hammer C. MiDAS-meaningful immunogenetic data at scale. PLoS Comput Biol 2021 Jul 6;17(7):e1009131.doi: 10.1371/journal.pcbi.1009131. PMID: 34228721; PMCID: PMC8284797. Reynisson B, Alvarez B, Paul S, Peters B, Nielsen M. NetMHCpan-4.1 and NetMHCIIpan-4.0: improved predictions of MHC antigen presentation by concurrent motif deconvolution and integration of MS MHC eluted ligand data. Nucleic Acids Res 2020 Jul 2;48(W1):W449-W454. doi: 10.1093/nar/gkaa379. PMID: 32406916; PMCID: PMC7319546.
Economic choices involving pollution, like those concerning common resources, relate to the emergence of cooperation among actors. Since pollution propagates in space, the temporal dynamics of economic choices is coupled to the spatial dynamics of pollution. We start from a simple description of the internal representations of the agents proposed by Arthur and Lane (1993) to describe information contagion. The simultations done in this paper allow us to discuss the maximum price that the agents agree to pay for non-polluting devices as a function of pollution, propagation of information and memory characteristics of the agents. We also characterize the spatio-temporal dynamics of choices, market share and pollution.
