SNP-specific array-based allele-specific expression analysis

Abstract

We have developed an optimized array-based approach for customizable allele-specific gene expression (ASE) analysis. The central features of the approach are the ability to select SNPs at will for detection, and the absence of need to PCR amplify the target. A surprisingly long probe length (39-49 nt) was needed for allelic discrimination. Reconstitution experiments demonstrate linearity of ASE over a broad range. Using this approach, we have discovered at least two novel imprinted genes, NLRP2, which encodes a member of the inflammasome, and OSBPL1A, which encodes a presumed oxysterol-binding protein, were both preferentially expressed from the maternal allele. In contrast, ERAP2, which encodes an aminopeptidase, did not show preferential parent-of-origin expression, but rather, cis-acting nonimprinted differential allelic control. The approach is scalable to the whole genome and can be used for discovery of functional epigenetic modifications in patient samples.

Figure 1.

Optimization and validation of an array-based hybridization approach to allele-specific expression (ASE). (A) Optimization of feature design. (Left) For all cases in which a CEPH sample was heterozygous at a SNP, shown are the log₂ ratios of the average of the two alleles at the SNP to the average of the two nonalleles, as a function of the length and T_m of the feature on the array. Probe lengths ranged from 29 bases to 55 bases (white = 29–39 nt, green = 35–45 nt, red = 40–49 nt, yellow = 45–55 nt), and relative melting temperatures of 68°C to 120°C (melting temperature increased from left to right within each color; see equation in design of array features in the Methods). (Right) For all cases in which a CEPH sample was homozygous at a SNP, shown are the log₂ ratios of the allele present to the average of the two nonalleles, as a function of the length and T_m of the feature on the array. For both homozygotes and heterozygotes, the highest signal-to-noise was achieved by probe lengths of 40–49 nt. (B) Validation by allelotyping known alleles from cDNA of heterozygotes for a given SNP. (Top) The difference between the average log₂ signal for the two alleles at the SNP and the average log₂ signal for the two nonalleles, with results for all probes for a SNP combined. (Bottom) The estimated log₂ ASE ratio at each SNP, as measured from the probes on the “plus” strand, against that measured from the probes on the “minus” strand.

Figure 2.

Reconstitution experiment measuring output ASE as a function of varying input amounts and allelic ratios. Log₂ ratio of observed signal against the log₂ ratios spiked-in, as a function of the total amount of DNA (in micromolars) spiked-in. (A–C) Without background for three input concentrations; (D–F) with background correction, with background estimated by the average signal from the probes of the two nonalleles.

Figure 3.

Observed frequency of ASE skewing in six CEPH cell lines. Histogram of the absolute of log₂ ASE ratio for SNPs in the top quartile of gene-expression level. Among these, 10% showed an ASE ratio >4.

Figure 4.

Parent-of-origin analysis of genes showing ASE in a replicate set. (A) NLRP2 demonstrates preferential expression of the maternal allele in placentas of eight different individuals. (B) OSBPL1A demonstrates preferential expression of the maternal allele in placentas of seven (shown) out of 10 different individuals. In both A and B, the first row demonstrates the simplex files for the three possible genotypes of the pyrosequencing assay. The next two rows demonstrate the pyrosequencing raw data files for the gDNA of the mother (homozygous) and the offspring (heterozygous), respectively. The final rows demonstrate the ASE (in triplicate) from cDNA of the cases and the corresponding RT− control for each assay. In all cases, the maternal allele (red) is preferentially expressed compared with the paternal allele (blue). The columns represent individual sample sets.

Figure 5.

ERAP2 demonstrates preferential expression of the C allele in seven cases, but is not affected by parent of origin. Images are assembled as in Figure 4, with cDNA shown in the bottom four rows. Note that the sequence has TT following the (C/T) SNP. Thus, the C/C allele is 2 × C, followed by 2 × T in the simplex, C/T is 1 × C followed by 3 × T, and T/T is 4 × T.