Whole-genome molecular haplotyping of single cells

Abstract

Conventional experimental methods of studying the human genome are limited by the inability to independently study the combination of alleles, or haplotype, on each of the homologous copies of the chromosomes. We developed a microfluidic device capable of separating and amplifying homologous copies of each chromosome from a single human metaphase cell. Single-nucleotide polymorphism (SNP) array analysis of amplified DNA enabled us to achieve completely deterministic, whole-genome, personal haplotypes of four individuals, including a HapMap trio with European ancestry (CEU) and an unrelated European individual. The phases of alleles were determined at ∼99.8% accuracy for up to ∼96% of all assayed SNPs. We demonstrate several practical applications, including direct observation of recombination events in a family trio, deterministic phasing of deletions in individuals and direct measurement of the human leukocyte antigen haplotypes of an individual. Our approach has potential applications in personal genomics, single-cell genomics and statistical genetics.

Figure 1

Microfluidic device designed for the amplification of metaphase chromosomes from a single cell. A single metaphase cell is recognized microscopically and captured in region 1. Protease (pepsin at low pH) is introduced to generate chromosome suspension in region 2. Chromosome suspension is partitioned into 48 units (region 3). Content in each partition is individually amplified (region 4). Specifically, chromosomes at low pH are first neutralized and treated with trypsin to digest chromosomal proteins. Chromosomes are denatured with alkali and subsequently neutralized for multiple strand displacement amplification to take place. As reagents are introduced sequentially into each air-filled chamber, enabled by the gas permeability of the device’s material, chromosomes are pushed into one chamber after the next and finally arrive in the amplification chamber. Amplified materials are retrieved at the collection ports (region 5). In the overview image of the device, control channels are filled with green dye. Flow channels in the cell-sorting region and amplification region are filled with red and blue dyes, respectively.

Figure 2

Whole-genome haplotyping. (a) Determining the chromosomal origin of amplification products in a microfluidic device using 46-loci PCR. This table represents results from an experiment using a single metaphase cell of P0’s cultured whole blood. A row represents the content inside a chamber on the microfluidic device, and a column represents a locus, with specified chromosome and coordinate (NCBI Build 36.1). Each locus, except those on chromosomes 17 and 20, was found in two chambers. The two alleles of a SNP are highlighted in red and green. Heterozygous loci are labeled in blue. Chamber numbers labeled yellow were pooled together and genotyped on one HumanOmni1-Quad array, and chamber numbers labeled orange were pooled together and genotyped on another array. Genomic DNA extracted from cultured whole blood was also tested with the same 46-loci PCR. (b) Statistics of whole-genome haplotyping. The fraction of SNPs present on the array phased for each chromosome of each individual (GM12891, GM12892, GM12878 and a European individual ‘P0’) is shown as a colored bar. (c) Fraction of SNPs phased as a function of the number of pairs of homologous chromosomes assayed. This is based on the results from four single-cell experiments of P0. Each point represents the coverage of an autosome. The error bars represent s.e.m.

Figure 3

Comparison of statistically determined phases with experimentally determined phases. (a) Comparison of experimentally determined phases of ~160,000 heterozygous SNPs of GM12878 (child of the trio) and those determined by phase III of the HapMap project. Unambiguous SNPs refer to those that are homozygous for at least one parent and are deterministically phased using family data in HapMap. This comparison shows the accuracy of DDP. Ambiguous SNPs refer to those that are heterozygous for all members of the trio and statistical phasing is used in HapMap. This comparison provides an evaluation of statistical phasing. (b) Comparison of experimentally determined phases of P0 and those determined by PHASE. Seventy-six regions on the autosomal chromosomes were randomly selected and statistically phased three times. Each region carried 100 heterozygous SNPs and spanned an average of ~2 Mb. Switch error rate was calculated as the proportion of heterozygous SNPs with different phases relative to the SNP immediately upstream. Single-site error rate was calculated as the proportion of heterozygous SNPs with incorrect phase. A SNP was considered correctly phased if it had the dominant phase. For each region, the average values from the three runs were reported. Presented here are the average switch error and single-site error per region. The deterministic phases measured by DDP are taken as the ground truth.

Figure 4

Direct observation of recombination events and deterministic phasing of heterozygous deletions in the family trio. Each allele with DDP data available for the child and the parent is represented by a colored line (blue, alleles transmitted to the child from the father; red, alleles transmitted to the child from the mother; black, untransmitted alleles). Centromeres and regions of heterochromatin are not assayed by genotyping arrays and are thus in white. Heterozygous deletions in the parents are represented as triangles along each homologous chromosome. A solid triangle represents one copy and a hollow triangle represents a null copy. The phases of deletions are determined for each parent independently. The triangles are color coded according to the state of transmittance as determined by the location of the deletion relative to spots of recombination. The phases of the deletions in the child are determined independent of the parents and are shown on top of the parental chromosomes. The integers on the left are the IDs of each region given by HapMap phase III. The numbers on the right are the copy number of a region in the child as determined by HapMap. Chromosomes are plotted with the same length.

Figure 5

HLA haplotypes of P0 determined using DDP. At each of the six classical HLA loci, the experimentally phased SNP haplotypes of P0 and 176 phased SNP haplotypes of CEU trios available from HapMap phase III were placed on a neighbor-joining tree. The two haplotypes of P0 are labeled in red and blue. For haplotypes in the CEU panel with HLA typing data, the four-digit HLA allele is presented next to the sample label. Most of each tree is compressed. Each compressed subtree is labeled with the HLA allele associated with members inside the subtree, if HLA allele information is available. The allelic identities of HLA-B and HLA-C on haplotype 1 were not determined with DDP because CEU individuals with similar SNP haplotypes as P0’s SNP haplotypes did not have HLA typing data at these loci but could be inferred from the results of direct HLA typing of genomic DNA (first row of table). HLA-DQA1 was not directly typed.