Rapid, scalable and highly automated HLA genotyping using next-generation sequencing: a transition from research to diagnostics

Abstract

Background: Human leukocyte antigen matching at allelic resolution is proven clinically significant in hematopoietic stem cell transplantation, lowering the risk of graft-versus-host disease and mortality. However, due to the ever growing HLA allele database, tissue typing laboratories face substantial challenges. In light of the complexity and the high degree of allelic diversity, it has become increasingly difficult to define the classical transplantation antigens at high-resolution by using well-tried methods. Thus, next-generation sequencing is entering into diagnostic laboratories at the perfect time and serving as a promising tool to overcome intrinsic HLA typing problems. Therefore, we have developed and validated a scalable automated HLA class I and class II typing approach suitable for diagnostic use.

Results: A validation panel of 173 clinical and proficiency testing samples was analysed, demonstrating 100% concordance to the reference method. From a total of 1,273 loci we were able to generate 1,241 (97.3%) initial successful typings. The mean ambiguity reduction for the analysed loci was 93.5%. Allele assignment including intronic sequences showed an improved resolution (99.2%) of non-expressed HLA alleles.

Conclusion: We provide a powerful HLA typing protocol offering a short turnaround time of only two days, a fully integrated workflow and most importantly a high degree of typing reliability. The presented automated assay is flexible and can be scaled by specific primer compilations and the use of different 454 sequencing systems. The workflow was successfully validated according to the policies of the European Federation for Immunogenetics. Next-generation sequencing seems to become one of the new methods in the field of Histocompatibility.

Figure 1

Raw amplicon read count statistics. Boxplots show raw amplicon read count for 17 amplicons. For plotting, DRB exon 2 readcounts were downscaled (1/5) for an improved presentation. Readcounts for amplicons without amplification are not included. All amplicons together exhibit a mean of 282 ± 143.

Figure 2

Sequence read output after down-stream processing. Representation of trimmed high quality reads generated by the GS Junior compared to the used sequence reads for allele assignment by the ATF software. Next to the bars percentage of ATF used reads are given.

Figure 3

Residual ambiguities of HLA-A, -B, -C, -DQB1, -DRB1 and -DPB1. Ambiguities for the Sanger typing method are given in blue, red bars represent the number of ambiguities of the Genome Sequencer typing method, e.g. samples with 20 ambiguities for Sanger typing in locus A have 2 ambiguities with GS typing. Mean and standard deviation are given per locus, the first line represents the ambiguities from Sanger typing, the second line from GS typing.

Figure 4

ATF 4.7 user interface. HLA-A genotype assignment showing the “master layer”. The forward and reverse sequence reads are aligned to exon 2, 3 and 4 with partial coverage of intronic regions. Read counts and sequencing direction are displayed in white bars. For instance, exon 4 is represented by two (allele 1 and 2) forward sequences with coverage of 90, respectively 81 sequence reads and two reverse sequences including 82 and 86 sequence reads. Underneath the exon map, the consensus sequence and the genotype assignment are shown. The phased sequences at the exon 2/intron 2 boundary are illustrated and propose the combination HLA-A*01:01:01:01, 24:02:01:01 with zero mismatches (master layer MM) compared to the database. The non- expressed A*01:01:01:02 N and the low expressed A*24:02:01:02 L are excluded by intron 2 variations as shown in the non coding (N-C) column. Mismatches in the phase layer 3 and 4 (MM3, MM4) demonstrate cis/trans ambiguities that would arise if the phase of SNPs in exon 3 and exon 2 cannot be defined (generic Sanger sequencing). The remaining alleles A*24:02:01:03 (variation in intron 3) and A*24:02:40 (synonymous substitution in exon 5) cannot be excluded and are therefore displayed likewise with zero mismatches.

Figure 5

Schematic illustration of the HLA typing workflow for ten samples.