The human leukocyte antigen (HLA) class I and class II loci are the most polymorphic genes in the human genome. Hematopoietic stem cell transplantation requires allele-level HLA typing at multiple loci to select the best matched unrelated donors for recipient patients. In current methods for HLA typing, both alleles of a heterozygote are amplified and typed or sequenced simultaneously, often making it difficult to unambiguously determine the sequence of the two alleles. Next-generation sequencing methods clonally propagate in parallel millions of single DNA molecules, which are then also sequenced in parallel. Recently, the read lengths obtainable by one such next-generation sequencing method (454 Life Sciences, Inc.) have increased to >250 nucleotides. These clonal read lengths make possible setting the phase of the linked polymorphisms within an exon and thus the unambiguous determination of the sequence of each HLA allele. Here we demonstrate this capacity as well as show that the throughput of the system is sufficiently high to enable a complete, 7-locus HLA class I and II typing for 24 or 48 individual DNAs in a single GS FLX sequencing run. Highly multiplexed amplicon sequencing is facilitated by the use of sample-specific internal sequence tags (multiplex identification tags or MIDs) in the primers that allow pooling of samples yet maintain the ability to assign sequences to specific individuals. We have incorporated an HLA typing software application developed by Conexio Genomics (Freemantle, Australia) that assigns HLA genotypes for these 7 loci (HLA-A, -B, -C, DRB1, DQA1, DQB1, DPB1), as well as for DRB3, DRB4, and DRB5 from 454 sequence data. The potential of this HLA sequencing system to analyze chimeric mixtures is demonstrated here by the detection of a rare HLA-B allele in a mixture of two homozygous cell lines (1/100), as well as by the detection of the rare nontransmitted maternal allele present in the blood of a severe combined immunodeficiency disease syndrome (SCIDS) patient.

Historically, association of disease with the major histocompatibility complex (HLA) genes has been tested with HLA alleles that encode antigen-binding affinity. The association with Parkinson disease (PD), however, was discovered with noncoding SNPs in a genome-wide association study (GWAS). We show here that several HLA-region SNPs that have since been associated with PD form two blocks tagged by rs3129882 (p = 9 × 10(-11)) and by rs9268515 and/or rs2395163 (p = 3 × 10(-11)). We investigated whether these SNP-associations were driven by HLA-alleles at adjacent loci. We imputed class I and class II HLA-alleles for 2000 PD cases and 1986 controls from the NeuroGenetics Research Consortium GWAS and sequenced a subset of 194 cases and 204 controls. We were therefore able to assess accuracy of two imputation algorithms against next-generation-sequencing while taking advantage of the larger imputed data sets for disease study. Additionally, we imputed HLA alleles for 843 cases and 856 controls from another GWAS for replication. PD risk was positively associated with the B(∗)07:02_C(∗)07:02_DRB5(∗)01_DRB1(∗)15:01_DQA1(∗)01:02_DQB1(∗)06:02 haplotype and negatively associated with the C(∗)03:04, DRB1(∗)04:04 and DQA1(∗)03:01 alleles. The risk haplotype and DQA1(∗)03:01 lost significance when conditioned on the SNPs, but C(∗)03:04 (OR = 0.72, p = 8 × 10(-6)) and DRB1(∗)04:04 (OR = 0.65, p = 4 × 10(-5)) remained significant. Similarly, rs3129882 and the closely linked rs9268515 and rs2395163 remained significant irrespective of HLA alleles. rs3129882 and rs2395163 are expression quantitative trait loci (eQTLs) for HLA-DR and HLA-DQ (9 × 10(-5) ≥ PeQTL ≥ 2 × 10(-79)), suggesting that HLA gene expression might influence PD. Our data suggest that PD is associated with both structural and regulatory elements in HLA. Furthermore, our study demonstrates that noncoding SNPs in the HLA region can be associated with disease irrespective of HLA alleles, and that observed associations with HLA alleles can sometimes be secondary to a noncoding variant.

The human leukocyte antigen (HLA) class I and class II loci are the most polymorphic genes in the human genome; distinguishing the thousands of HLA alleles is challenging. Next generation sequencing of exonic amplicons with the 454 genome sequence (GS) FLX System and Conexio Assign ATF 454 software provides high resolution, high throughput HLA genotyping for eight class I and class II loci. HLA typing of potential donors for unrelated bone marrow donor registries typically uses a subset of these loci at high sample throughput and low cost per sample. The Fluidigm Access Array System enables the incorporation of 48 different multiplex identifiers (MIDs) corresponding to 48 genomic DNA samples with up to 48 different primer pairs in a microfluidic device generating 2304 parallel polymerase chain reactions (PCRs). Minimal volumes of reagents are used. During genomic PCR, in this 4-primer system, the outer set of primers containing the MID and the 454 adaptor sequences are incorporated into an amplicon generated by the inner HLA target-specific primers each containing a common sequence tag at the 5' end of the forward and reverse primers. Pools of the resulting amplicons are used for emulsion PCR and clonal sequencing on the 454 Life Sciences GS FLX System, followed by genotyping with Conexio software. We have genotyped 192 samples with 100% concordance to known genotypes using 8 primer pairs (covering exons 2 and 3 of HLA-A, B and C, and exon 2 of DRB1, 3/4/5 and DQB1) and 96 MIDs in a single GS FLX run. An average of 166 reads per amplicon was obtained. We have also genotyped 96 samples at high resolution (14 primer pairs covering exons 2, 3, and 4 of the class I loci and exons 2 of DRB1, 3/4/5, DQA1, DQB1, DPB1, and exon 3 of DQB1), recovering an average of 173 sequence reads per amplicon.

The high degree of polymorphism at HLA class I and class II loci makes high resolution HLA typing challenging. Current typing methods, including Sanger sequencing, yield ambiguous typing results due to incomplete genomic coverage and inability to set phase for HLA haplotype determination. The 454 Life Sciences GS FLX next generation sequencing system coupled with Conexio ATF software can provide very high resolution HLA genotyping. High throughput genotyping can be achieved by use of primers with multiplex identifier (MID) tags to allow pooling of the amplicons generated from different individuals prior to sequencing. We have conducted a double blind study in which eight laboratory sites performed amplicon sequencing using GS FLX standard chemistry and genotyped the same 20 samples for HLA-A, -B, -C, DPB1, DQA1, DQB1, DRB1, and DRB3, DRB4 and DRB5 (DRB3/4/5) in a single sequencing run. The average sequence read length was 250 base pairs (bp) and the average number of sequence reads per amplicon was 672, providing confidence in the allele assignments. Of the 1280 genotypes considered, assignment was possible in 95% of the cases. Failure to assign genotypes was the result of researcher procedural error or the presence of a novel allele rather than a failure of sequencing technology. Concordance with known genotypes, in cases where assignment was possible, ranged from 95.3% to 99.4% for the eight sites, with overall concordance of 97.2%. We conclude that clonal pyrosequencing using the GS FLX platform and Conexio ATF software allows reliable identification of HLA genotypes at high resolution.

No abstract