Ultraconserved elements are associated with homeostatic control of splicing regulators by alternative splicing and nonsense-mediated decay

Abstract

Many alternative splicing events create RNAs with premature stop codons, suggesting that alternative splicing coupled with nonsense-mediated decay (AS-NMD) may regulate gene expression post-transcriptionally. We tested this idea in mice by blocking NMD and measuring changes in isoform representation using splicing-sensitive microarrays. We found a striking class of highly conserved stop codon-containing exons whose inclusion renders the transcript sensitive to NMD. A genomic search for additional examples identified>50 such exons in genes with a variety of functions. These exons are unusually frequent in genes that encode splicing activators and are unexpectedly enriched in the so-called "ultraconserved" elements in the mammalian lineage. Further analysis show that NMD of mRNAs for splicing activators such as SR proteins is triggered by splicing activation events, whereas NMD of the mRNAs for negatively acting hnRNP proteins is triggered by splicing repression, a polarity consistent with widespread homeostatic control of splicing regulator gene expression. We suggest that the extreme genomic conservation surrounding these regulatory splicing events within splicing factor genes demonstrates the evolutionary importance of maintaining tightly tuned homeostasis of RNA-binding protein levels in the vertebrate cell.

Figure 1.

NMD pathway is blocked by Upf1 RNA interference (RNAi) and emetine treatment. (A) Western blot using antibodies against Upf1 and α-Tubulin. Lanes 1, 2, and 3 contain equal amounts of total cell protein 2 d after siRNA transfection. Lanes 4 and 5 contain 50% and 25% of lane 3. (B) RT–PCR of nPTB mRNA isoforms. nPTB primers are in constitutive exons 9 and 11 of nPTB mRNA. Exon 10 skipping will generate NMD-targeted nonsense isoforms. Nonsense isoforms are accumulated in Upf1 siRNA-transfected cells and emetine-treated cells. RT–PCR of β-actin mRNA was used as control.

Figure 2.

NOL5-splicing isoform with inclusion of stop codon exon is subject to NMD. (A) Microarray data of skip probe set intensity versus include probe set intensity for NOL5 stop codon exon before and after blocking NMD. The lines of the plot represent the robust regression coefficient (constrained to go through the origin) for NMD-blocked sample groups (open circles, emetine; open squares, Upf1 siRNA treated) or non-NMD-blocked sample groups (filled circles, untreated cells; filled squares, control siRNA treated). The log₂ difference in the slopes is 1.1922, indicating 2.3-fold inclusion in NMD-blocked cells relative to non-NMD-blocked cells for this exon (Sugnet et al. 2006). (B) RT–PCR validation of accumulation of NOL5 stop codon exon after blocking NMD. The stop codon exon-inclusion isoform (nonsense isoform) is specifically accumulated upon Upf1 siRNA transfection or emetine treatment as compared with control cell samples. Percentage of exon inclusion increases fourfold by Upf1 siRNA transfection or ninefold by emetine treatment, as measured by an Agilent Bioanalyzer. (C) NOL5 stop codon exon as seen in the UCSC Genome Browser (Karolchik et al. 2003). Base position shows three reading frames. In each reading frame, black squares represent stop codons and the white squares represent methionine codons, whereas the gray squares represent nonmethionine amino acid codons. The NOL5 exon has stop codons in all three reading frames. The conservation track at the bottom shows high conservation in exonic and flanking intronic sequences of the NOL5 stop codon exon.

Figure 3.

RT–PCR tests of stop codon exons. Twelve stop codon exons found by array and 30 conserved three-frame stop codon exons were tested by RT–PCR. The upper bands of each test are stop codon exon-inclusion isoforms. Bracket shows more than one stop codon exon-inclusion isoform was detected. Percentage of exon inclusion is labeled at the bottom of the corresponding exon and treatment and is a molar percent as determined using an Agilent Bioanalyzer. True validation result means that stop codon exon-inclusion isoforms increased when blocking NMD by emetine. All 12 stop codon exons found by array are true. Fourteen of 15 (94%) detectable bioinformatics-predicted stop codon exons are true.

Figure 4.

Model for homeostatic auto- or cross-regulatory maintenance of proper splicing factor levels in the cell. Expression of splicing activator genes (such as SR protein genes) that carry stop codon exons are regulated by AS-NMD. When the stop codon exon is skipped, functional activator protein mRNA is on. Too many splicing activator proteins can turn off their own expression by activating inclusion of their stop codon exon, triggering NMD. In addition, splicing activator protein level activates splicing of their many target exons in the genome to counteract the effect of the negative (hnRNP protein) splicing-repressor factors. Expression of splicing repressors (such as hnRNP protein) that carry a skipped coding exon can be regulated by AS-NMD. When the coding exon is included, functional hnRNP protein mRNA is on. Too many splicing repressor proteins can turn off their own expression by repressing the inclusion of the coding exon (not a multiple of three) that creates a frameshift and triggers NMD. In addition, splicing repressor protein levels also repress splicing at incorrect splice sites at many target exons in the genome and counteract the effects of the positive splicing factors (see text).