Next-generation sequencing for HLA typing of class I loci

Abstract

Background: Comprehensive sequence characterization across the MHC is important for successful organ transplantation and genetic association studies. To this end, we have developed an automated sample preparation, molecular barcoding and multiplexing protocol for the amplification and sequence-determination of class I HLA loci. We have coupled this process to a novel HLA calling algorithm to determine the most likely pair of alleles at each locus.

Results: We have benchmarked our protocol with 270 HapMap individuals from four worldwide populations with 96.4% accuracy at 4-digit resolution. A variation of this initial protocol, more suitable for large sample sizes, in which molecular barcodes are added during PCR rather than library construction, was tested on 95 HapMap individuals with 98.6% accuracy at 4-digit resolution.

Conclusions: Next-generation sequencing on the 454 FLX Titanium platform is a reliable, efficient, and scalable technology for HLA typing.

Figure 1

HLA Class I amplification strategy. PCR amplification of the polymorphic exons 2 and 3 in the class I HLA loci was performed (using primers within introns surrounding each exon) prior to 454 sequencing.

Figure 2

Schematic of the HLA Caller Algorithm. The HLA calling algorithm determines the most likely pair of HLA types at each locus by systematically evaluating all possible pairs of 4-digit HLA types. A) The genotyping algorithm within the GATK calculates the probability of observing particular genotypes in the data given a pair of HLA alleles. Probabilities were combined multiplicatively across base positions to obtain the cumulative probability based on genotypes. B) A binomial distribution function was used to calculate of probability of observing particular haplotypes in the data given a pair of HLA alleles. Probabilities were combined multiplicatively across pairs of polymorphic positions to obtain the cumulative probability based phase information. C) Prior probabilities for specific allele pairs were calculated as the product of allele frequencies in a specific population. Probabilities based on genotypes, phase information, and allele frequencies were combined multiplicatively to obtain the posterior probability for each HLA allele pair.

Figure 3

Depth of coverage for 95 HapMap individuals. Sequenced using the barcoded-PCR method: (a) coverage by individual, and (b) coverage across each of 6 exons.

Figure 4

Integrative Genomics Viewer (IGV) sequence data. IGV snapshot of sequence data at HLA-B for one HapMap sample (NA18507), and the corresponding calls made from the data. Each horizontal bar represents one read, and colored bars indicate where DNA base pairs differ from the reference.

Figure 5

IGV snapshot: Example 1. An example of HLA-C*02:10 mislabeled as C*02:02 in the gold standard (NA15822).

Figure 6

IGV snapshot: Example 2. IGV snapshot: an example of HLA-A*02:07 mistyped as A*02:01, and A*26:02 mistyped as A*26:01 in the same individual (NA18987) in the gold standard. The orange and green arrows indicate the polymorphisms that differ between the respective types.

Figure 7

IGV snapshot: Example 3. IGV snapshot of a sample (NA18970) miscalled due to presence of 4 distinct alleles/chromosomes in the data. Other loci also showed similar patterns.