Project description:We sought to identify the transcripts that responded to changes in cellular SFPQ/PSF (polypyrimidine tract-binding protein (PTB)-associated splicing factor) protein levels by expression profile analysis of PSF knockdown cells. We constructed clonal HeLa cell lines stably expressing small hairpin RNAs (shRNA) targeted to various regions within the PSF CDS. For microarray studies, we chose two clones (#s 1_1 and 3_1) harboring independently targeted shRNAs that showed ~50% reduction of the endogenous PSF protein. As controls, we used two independent clones harboring a non-targeting shRNA (#s 6_2 and 6_4). Microarray analysis of total cellular RNA identified the transcripts that were reproducibly upregulated (>1.5 fold, n=284) and downregulated (< 0.66 fold, n=121) in both independent knockdown cell lines, and hence were likely enriched in the nonrandomly responding species. We next studied the effects of PSF depletion on particular gene categories. Analysis of the transcripts with unique RefSeq identifiers (n=15805) using PANTHER tools at http://www.pantherdb.org (Molecular Function) revealed significant nonrandom responses of transcripts encoding ribosomal proteins (RP) (downregulation, n=150, P=2.9E-04), while other subgroups exhibited higher P-values. We wondered whether the nonrandomly affected transcripts shared nucleotide sequence features that might account for their response to PSF depletion. To extract such features, we analyzed the mRNAs representing the top up- and downregulated species (n=100 each) as well as the “ribosomal protein” category retrieved from PANTHER database (n=150) for the presence of shared sequence motifs, by using Gibbs motif sampling and dscan software at http://bayesweb.wadsworth.org/gibbs/gibbs.html. The Gibbs sampling returned very similar shared signatures for the up- and downregulated species; and a very similar signature (motif #3) was elicited from the combined up and downregulated datasets (n=200). Notably, a related signature was also elicited from RP sequences. In order to assign unbiased probabilities to the nonrandomness of motif occurrence, we used nucleotide frequency matrices representing motif #3 and dscan utility to calculate the hit scores in the up and downregulated mRNAs. As a control, the same procedure was performed on samplings (n=100) randomly extracted from human EST sequence databases, and the hit scores were evaluated using the Mann-Whitney test. We found that the random samplings featured only low scores, while the PSF-repressed and, to even greater extent, the PSF-activated samplings exhibited significantly higher scores (P=0.01 and P=0.0001, respectively; α=0.05), and similar P-values were obtained when using different control samplings. These data revealed a correlation between PSF-responsiveness and high occurrence of a particular sequence signature, and suggested that the presence of such sequences could contribute to the response. Because PSF can directly bind mRNA in vivo, it was plausible that it recognized the motifs directly, as part of mRNA. Indeed, comparison of motif #3 to the direct PSF-binding sites in pre-mRNA revealed a similarity that included a common UGNAGC signature, whereas the SELEX-derived PSF aptamers shared the core GYYG signature. In addition, the motif #3 also exhibited, albeit to a different extent, direct repeats of a trinucleotide CUG that were not found in the CLIP and SELEX signatures. Since such sequences have not been previously noted in PSF-binding RNA, their significance was not clear. However, the pre-mRNA CLIPs were nonrandomly enriched in CUG‘s constituent dinucleotides CU and UG, possibly reflecting a shared recognition determinant. In summary, our results showed that PSF regulates specific subsets of human genes, and suggested that PSF-mRNA recognition could contribute to the regulation.

Dataset Information

Homo sapiens

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets