Simultaneous genotyping, gene-expression measurement, and detection of allele-specific expression with oligonucleotide arrays

Abstract

Oligonucleotide microarrays provide a high-throughput method for exploring genomes. In addition to their utility for gene-expression analysis, oligonucleotide-expression arrays have also been used to perform genotyping on genomic DNA. Here, we show that in segregants from a cross between two unrelated strains of Saccharomyces cerevisiae, high-quality genotype data can also be obtained when mRNA is hybridized to an oligonucleotide-expression array. We were able to identify and genotype nearly 1000 polymorphisms at an error rate close to 3% in segregants and at an error rate of 7% in diploid strains, a performance comparable to methods using genomic DNA. In addition, we demonstrate how simultaneous genotyping and gene-expression profiling can reveal cis-regulatory variation by screening hundreds of genes for allele-specific expression. With this method, we discovered 70 ORFs with evidence for preferential expression of one allele in a diploid hybrid of two S. cerevisiae strains.

Figure 1.

Schematic illustrating the methods for genotyping and detecting allele-specific expression. (A) Routine gene-expression analysis using mRNA. Signals from all probes interrogating the same ORF are averaged to produce an estimate of the gene-expression level. In this example, the probe signals suggest that the reference strain (red) expresses this ORF at a higher level than the test strain (blue). All probes hybridize equally well in the test strain, providing no evidence of sequence mismatches between the probe and target sequence. (B) Genotyping using genomic DNA. When DNA is hybridized to arrays, the intensities in the two strains are the same, except at probes interrogating sequences where the test strain bears a polymorphism. Here, the test strain bears a polymorphism in the region interrogated by “Probe 4,” and as a result, the test strain produces a weak signal at this probe. (C) Genotyping using mRNA. The gene is expressed at a different level in the test strain than in the reference strain, but the polymorphism in the sequence interrogated by “Probe 4” is still readily detected, because the signal at this probe is substantially less than the gene-expression level as estimated using the entire probe set. (D) Equal expression of both alleles in a diploid. A diploid strain formed by mating the reference and test strains (green) is heterozygous for the sequence polymorphism at “Probe 4.” The ratio of the intensity at “Probe 4” to the remaining probes in the heterozygous diploid is equal to the average of the ratios of “Probe 4” to the remaining probes in the reference and test strains, suggesting that both alleles are expressed in equal amounts. (E) Preferential expression of the reference strain allele. Here, the ratio of the intensity at “Probe 4” to the remaining probe is the same as the ratio in the reference strain, suggesting that only the allele derived from the reference strain is expressed. (F) Preferential expression of the test strain allele.

Figure 2.

Sensitivity of polymorphism discovery. The dotted line shows the probability of detecting a single base change as a function of position in the 25 base-pair probe when mRNA material is hybridized to the array. Vertical lines give the 95% confidence intervals for these probabilities. The dashed line shows the probability of detecting single base changes using genomic DNA. The solid line depicts the baseline rate of polymorphism.

Figure 3.

Comparison of genomic DNA and mRNA-derived genotypes. (A) Chromosome IV genotypes in haploid segregants. The x's represent locations of markers. In each box, the middle line (without x's) is the most likely set of genotypes, estimated using the HMM, producing the mRNA data. (B) Chromosome XIII genotypes in diploid hybrids formed by mating two segregants.

Figure 4.

Examples illustrating the use of the 1:1 mixture to estimate the false positive and sensitivity of the method for detecting allele-specific expression. In each plot, the x's represent the observed intensities (I) and the dashes represent the expected intensities (Î) for a marker probe located in FKS1. Note that the probe represents a good marker, because the I/Î ratio is much greater in BY than in RM. For this transcript, the expected probe intensities for the three BY replicates are approximately twofold higher than in the three RM replicates, suggesting that the transcript is expressed at a higher level in BY. A 1:1 mixture of total mRNA from BY and RM would therefore contain approximately two times as much of the BY FKS1 allele as the RM allele. (A) Model 1 analyzes the false-positive rate. Here, the solid lines in the 1:1 mixture represent the predicted range for the observed probe signals under a model that correctly accounts for the twofold overabundance of BY allele. The observed signals from the 1:1 mixture replicates fall within this range, providing no evidence for deviation from the model. (B) Model 2 analyzes the power. Here, the dashed lines represent the predicted range of the observed probe signals under a model assuming equal expression of both alleles. All six observed intensity signals in the 1:1 mixture fall above this range, correctly revealing that the 1:1 mixture contains more BY allele than RM allele for this transcript.

Figure 5.

Allele-specific expression analysis and quantitative PCR. (A,C) Array-based detection of allele-specific expression in TIP1 and SOD1, respectively. Note that the observed intensities fall above the predicted range for both TIP1 and SOD1, indicating that the observed to expected intensity ratio is more similar to BY than RM, suggesting that the BY allele is preferentially expressed. (B,D) Quantitative PCR experiments. Green lines indicate the diploid cDNA samples, and black lines indicate the diploid genomic DNA samples. Data points on the standard curve are labeled as parts BY allele: parts RM allele. Blue points represent excess RM allele, whereas red points represent excess BY allele. The estimated proportion of BY allele present in the diploid cDNA is given on the vertical axis in green.