HuR-TE binding sites stabilize genes, unless in Alu elements. (A) As expected for the known transcript stabilizer HuR, targeted genes were downregulated after HuR knockdown (KD) compared with unbound genes, shown here as the cumulative distributions of the Cuffdiff differential expression test statistic (CDF). (B) HuR’s stabilizing effect depended strongly on the number of binding sites in the gene span, shown through the decreasing medians and interquartile ranges of the test statistic distribution. (C) Unexpectedly, binding sites in Alu elements had the opposite effect; genes targeted only in Alu elements were upregulated after HuR knockdown. All other TE binding sites had the expected effect with similar magnitude as nonrepetitive sites. The Venn diagram indicates the wide scale of TE binding sites, depicting the number of genes bound only in Alu elements, only in non-Alu TEs, only in non-TEs, and in various mixtures. (D) A linear regression on the logarithm of the number of sites in each class verified that non-Alu TE and nonrepetitive sites predicted downregulation, but Alu sites did not. Error bars represent a 95% confidence interval. The Alu exon and TE exon coefficients have large error bars because there were few examples to learn from.