Elsevier

Human Immunology

Volume 81, Issue 8, August 2020, Pages 423-429
Human Immunology

Research article
Concordance between predicted HLA type using next generation sequencing data generated for non-HLA purposes and clinical HLA type

https://doi.org/10.1016/j.humimm.2020.06.002Get rights and content

Abstract

We explored the feasibility of obtaining accurate HLA type using pre-existing NGS data not generated for HLA purposes. 83 exomes and 500 targeted NGS pharmacogenomic panels were analyzed using Omixon HLA Explore, OptiType, and/or HLA-Genotyper software. Results were compared against clinical HLA genotyping. 765 (94.2%) Omixon and 769 (94.7%) HLA-Genotyper of 812 germline allele calls across class I/II loci and 402 (99.5%) of 404 OptiType class I calls were concordant to the second field (i.e. HLA-A*02:01). An additional 19 (2.3%) Omixon, 39 (4.8%) HLA-Genotyper, and 2 (0.5%) OptiType allele calls were first field concordant (i.e. HLA-A*02). Using Omixon, four alleles (0.4%) were discordant and 24 (3.0%) failed to call, while 4 alleles (0.4%) were discordant using HLA-Genotyper. Tumor exomes were also evaluated and were 85.4%, 91.6%, and 100% concordant (Omixon and HLA-Genotyper with 96 alleles tested, and Optitype with 48 class I alleles, respectively). The 15 exomes and 500 pharmacogenomic panels were 100% concordant for each pharmacogenomic allele tested. This work has broad implications spanning future clinical care (pharmacogenomics, tumor response to immunotherapy, autoimmunity, etc.) and research applications.

Introduction

Pharmacogenetic testing has historically been performed reactively at the time the patient will be prescribed a medication or, in some cases, to explain a toxicity that has already developed. There is increasing interest in preemptive testing of multiple pharmacogenes and storing that data in the patient’s electronic health record (EHR) so that the information can be readily available and used immediately to prescribe medications when needed without a delay for testing. At the same time, the cost of sequencing has dramatically decreased in recent years [1], such that exome sequencing (ES) and even genome sequencing (GS) are increasingly used in the diagnosis of hereditary disorders. In addition, patients are increasingly interested in obtaining genetic information, including pharmacogenomics. As a result, “healthy” exomes and genomes are beginning to be offered [2]. Therefore, exome data is available for rising numbers of patients, many of whom would opt for pharmacogenomic interpretation if available [3], [4], [5].

Several human leukocyte antigen (HLA) loci are currently known to be associated with increased risk for severe medication reactions and are the subject of guidelines [6], [7], [8], [9]. Therefore, when interpreting ES data for pharmacogenomic purposes, ideally, HLA alleles should also be included. However, genes encoding the HLA class I and class II molecules are the most polymorphic in the human genome, with 19,031 class I and 7183 class II alleles documented as of December 2019 [10]. The HLA genes are difficult to genotype, particularly with short-read next generation sequencing (NGS), due to the high degree of polymorphism as well as structural variation. Using traditional techniques, HLA typing can be performed at different resolution depending on which technique is utilized [11], [12], [13]. For example, sequence-specific oligonucleotide (SSO) typing can be used to generate results to the first field (i.e. HLA-A*02), which corresponds to allele groups by serological activity and semi-accurately to the second field (i.e. HLA-A*02:01), which corresponds to protein level resolution. Sequence-specific primer (SSP) amplification and sequence based typing (SBT) produces results accurate to the second field. These techniques are labor-intensive and are being replaced with NGS techniques [14], [15]. Recently, the use of targeted sequencing and/or long-range PCR methods coupled with NGS with read depths of >50–100 have been shown to produce accurate results, and allow for typing to the third and fourth field (i.e. HLA-A*02:01:01 and HLA-A*02:01:01:01), which corresponds to synonymous polymorphisms and non-coding variants [16], [17], [18]. The use of ES, in the absence of target enrichment and specific PCR amplification steps, for HLA typing has been less explored to date.

In addition to accurate HLA typing being critical for matching of transplant donors and recipients and for pharmacogenomics, it is also important in the diagnosis of autoimmune diseases and a role in prediction of response of solid tumors to immune checkpoint blockade is also emerging [19], [20], [21], [22], [23]. Therefore, the ability to generate accurate HLA typing from clinical ES and from existing research data sets would be valuable for both clinical and research purposes. Currently available targeted HLA sequencing by NGS typically requires amplification of individual loci by either long-range PCR or by multiplex PCR targeting exons related to the antigen recognition site, with read depths of at least 50 [24]. In addition, medium read length (150–1000 bp) instruments are typically used. Therefore, we explored the feasibility of accurate HLA typing by filtering to include only reads from the MHC region from existing ES data or a large NGS pharmacogenomic panel that was generated by sequence capture and short (101 bp) NGS reads for clinical diagnostic purposes or as part of a research study.

Section snippets

Clinical data sets

Three separate data sets were used for this study to reflect the possibility of HLA typing for pharmacogenomics and/or other applications from NGS data not originally generated for the specific purpose of HLA typing. These studies were reviewed and approved by the Mayo Clinic Institutional Review Board.

Our first data set consisted of 68 patients enrolled in BEAUTY (NCT02022202) at Mayo Clinic, which included adult women with newly diagnosed breast cancer [25]. Genomic DNA was extracted from

Data set 1: Germline whole exome sequencing of patients with breast cancer

The first data set was interrogated in detail at each of the HLA-A, B, C, DRB1, DQA1, and DQB1 loci. The command line filtering step recommended by the Omixon software developer was deemed to be necessary; when this step was omitted, the software was unable to process the data. Similarly, running RazerS3 was recommended prior to using OptiType as that would significantly reduce the run time of Optitype. The average coverage of each locus is shown in Fig. 2. Class I genes generally had higher

Discussion

The use of next-generation sequencing, including large panels, exome sequencing, and genome sequencing is increasing for both clinical diagnostic purposes, as well as for research purposes. At the same time, interest in pre-emptive pharmacogenomic testing is also increasing. Therefore, it is important to understand the feasibility of making pharmacogenomic calls from this type of data for future clinical and research use. Several large studies have explored some pharmacogenes from NGS data [27]

Funding

The BEAUTY study is funded in part by the Mayo Clinic Center for Individualized Medicine; Nadia’s Gift Foundation; John P. Guider; the Eveleigh Family; George M. Eisenberg Foundation for Charities; generous support from Afaf Al-Bahar; and the Pharmacogenomics Research Network (PGRN). Other contributing groups include the Mayo Clinic Cancer Center (MPG) and the Mayo Clinic Breast Specialized Program of Research Excellence (SPORE) (MPG and KK).

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgements

We would like to thank Laura Train, Kate Kotzer, Michelle Kluge, Susan Lagerstedt, Mary Beth Karow, Kimberley Harris, Amy Barthel, Charles Kremer, Brenda Moore, Sandra Peterson, Linnea Baudhuin, and the RIGHT study team for their contributions to the pharmacogenomics data. We would like to thank the BEAUTY study team for their contribution to the whole exome sequencing data.

References (36)

  • A.A. Lemke et al.

    Patient perspectives following pharmacogenomics results disclosure in an integrated health system

    Pharmacogenomics

    (2018)
  • M.S. Hershfield et al.

    Clinical Pharmacogenetics Implementation Consortium guidelines for human leukocyte antigen-B genotype and allopurinol dosing

    Clin. Pharmacol. Ther.

    (2013)
  • M.A. Martin et al.

    Clinical pharmacogenetics implementation consortium guidelines for HLA-B genotype and abacavir dosing

    Clin. Pharmacol. Ther.

    (2012)
  • E.J. Phillips et al.

    Clinical pharmacogenetics implementation consortium guideline for HLA genotype and use of carbamazepine and oxcarbazepine: 2017 update

    Clin. Pharmacol. Ther.

    (2018)
  • J.H. Karnes et al.

    Applications of immunopharmacogenomics: predicting, preventing, and understanding immune-mediated adverse drug reactions

    Annu. Rev. Pharmacol. Toxicol.

    (2018)
  • J. Robinson et al.

    The IPD and IMGT/HLA database: allele variant databases

    Nucl. Acids Res.

    (2015)
  • R. Holdsworth et al.

    The HLA dictionary 2008: a summary of HLA-A, -B, -C, -DRB1/3/4/5, and -DQB1 alleles and their association with serologically defined HLA-A, -B, -C, -DR, and -DQ antigens

    Tissue Antigens

    (2009)
  • S.D. Adams et al.

    High throughput HLA sequence-based typing (SBT) utilizing the ABI Prism 3700 DNA Analyzer

    Tumori

    (2001)
  • Cited by (4)

    View full text