Figure 2.

Three effects that cause the allele-specific expression variability. In these plots, the y-axis quantifies the proportion of expression from the PWD allele (PWD percentage). The x-axis provides an arbitrary index for different individuals from the reciprocal crosses. The left panels show offspring from the PWD X AKR cross, and the left panels show offspring from the AKR X PWD cross. Different colors represent different X-linked genes. (a) A diagram to illustrate the allele-specific expression results when there is no sampling effect, no eQTL effect and no parent-of-origin effect. In this case, there is little variability of PWD allelic expression among individuals or among the two reciprocal crosses. The only source of variability is the pyrosequencing measurement error. This is the case for the autosomal genes and X-linked genes that escape X inactivation. (b) A diagram to illustrate the sampling effect caused by random X inactivation. In this diagram, the X-inactivation process itself is random, but the number of brain-forming cells is small during the time of X inactivation, resulting in sampling variation among individuals. Although individuals are expected to show a 1:1 expression ratio, if each cell randomly and independently inactivates one or the other X chromosome, then we expect to see a binomial distribution of counts of cells inactivating the maternal X versus the Xp. If the count of cells is small, the variance in expression ratios could be large, and a maternal bias observed in a small number of individuals might be explained by this sampling effect. The sampling effect of X inactivation also drives the observed co-variation of allelic bias in expression of all X-linked genes. (c) A diagram to illustrate the eQTL effect. If there is a cis-regulatory polymorphism near the respective gene, it may drive differential allelic expression yielding allelic expression counts different from 1:1. The regulatory variant might drive higher expression from the PWD or the AKR allele, so the mean PWD expression percentage is not 50%. Such an effect would be allele-specific (or strain-specific), and would not explain differences in expression between reciprocal crosses or a maternal bias. (d) A diagram of preferential Xp inactivation. Here the X inactivation is NOT random and the Xp is preferentially inactivated. In this case we will observe greater expression from the maternal allele. The bias is like that of a biased coin. For small numbers of tosses, not all samples will show a skewed ratio of heads to tails, but with a sufficiently large sample, the bias will appear as a shift in the mean. In this cartoon, a comparison of the two reciprocal crosses shows that the allele-specific expression profile is shifted.