Efficient Sequencing, Assembly, and Annotation of Human KIR Haplotypes

Abstract

The homology, recombination, variation, and repetitive elements in the natural killer-cell immunoglobulin-like receptor (KIR) region has made full haplotype DNA interpretation impossible in a high-throughput workflow. Here, we present a new approach using long-read sequencing to efficiently capture, sequence, and assemble diploid human KIR haplotypes. Probes were designed to capture KIR fragments efficiently by leveraging the repeating homology of the region. IDT xGen^® Lockdown probes were used to capture 2-8 kb of sheared DNA fragments followed by sequencing on a PacBio Sequel. The sequences were error corrected, binned, and then assembled using the Canu assembler. The location of genes and their exon/intron boundaries are included in the workflow. The assembly and annotation was evaluated on 16 individuals (8 African American and 8 Europeans) from whom ground truth was known via long-range sequencing with fosmid library preparation. Using only 18 capture probes, the results show that the assemblies cover 97% of the GenBank reference, are 99.97% concordant, and it takes only 1.8 haplotigs to cover 75% of the reference. We also report the first assembly of diploid KIR haplotypes from long-read WGS. Our targeted hybridization probe capture and sequencing approach is the first of its kind to fully sequence and phase all diploid human KIR haplotypes, and it is efficient enough for population-scale studies and clinical use. The open and free software is available at https://github.com/droeatumn/kass and supported by a environment at https://hub.docker.com/repository/docker/droeatumn/kass.

Keywords: DNA; annotation; assembly; haplotype; killer-cell immunoglobulin-like receptor; natural killer.

Figure 1

Common KIR haplotype structures and their names. Blue regions represent the presence of a gene or the KIR3DP1-KIR2DL4 intervening intergenic region. Some genes are partially blue to indicate their portion of a fusion allele. The first column contains the informal name for the haplotype. Each haplotype name is a combination of its two regions: centromeric (proximal) regions are preceded with a “c” and telomeric (distal) regions are preceded with a “t”. Not all haplotype structures are shown.

Figure 2

Workflow. The workflow starts with fragment capture in vitro or in silico and PacBio assembly (A). The sequences are error corrected (B) and binned by KIR region and KIR gene (C) before de novo assembly of each bin (D). Finally, the assembled haplotigs/haplotypes are annotated by gene (E) and exon (F).

Figure 3

Reference haplotype structures in the two validation cohorts. Each haplotype represents one of the structures previously established via fosmid library preparation and long-read sequencing. The unofficial name of the haplotype is on the left. Lines connect genes with the same name in different structures. Solid lines connect the same gene in neighboring structures. Dashed lines connect the same gene in non-neighboring structures (i.e., the line goes through one or more neighboring haplotypes). cA02~tA03: 1 EUR. cA01~tA02: 1 AFA. cA01~tA01: 5 AFA, 9 EUR. cA03~tB02: 1 EUR. cA01~tB01: 1 AFA, 1 EUR. cA01~tB04: 1 EUR. cB01~cB01: 2 AFA, 1 EUR. cB04~tB03: 1 EUR. cB01~tA01: 3 AFA. cB02~tA01: 1 EUR. cB03~tA01: 1 AFA.

Figure 4

Alignment of assembled haplotigs with reference haplotype sequence MN167513 (cA01~tA01), whose length is 147,345. The gene features are annotated across the top. The haplotigs are stacked below and colored by nucleotide.

Figure 5

IGV depiction of the alignment of the 18 set capture probes across a 49 kb region of KP420440 (cB01~tB01). The locations of the probes are displayed by the vertical ticks with the red labels above them in the middle track. The locations of exons and repeat elements (horizontal blue bars) are on the bottom two tracks. Seven distinct probes align in this window. The probe pattern 4-3-12-10-2-7-13 repeats three times from left to right, except probe 10 does not align in the middle group.

Figure 6

Recombination hotspot. The top shows the 9 kb haplotype context of the repetitive region of the KIR3DP1-KIR2DL4 intergene region, whose self-alignment is shown in the bottom half. Red lines indicate alignment of the haplotype with itself in the same orientation, and blue lines indicate reverse complement orientation. The location of the repetitive elements is at the bottom, with red and blue again indicating orientation. The yellow boxes highlight three AluSx and one AluSq elements that align with each other, two in each direction.