Comparative analysis of mRNA targets for human PUF-family proteins suggests extensive interaction with the miRNA regulatory system

Abstract

Genome-wide identification of mRNAs regulated by RNA-binding proteins is crucial to uncover post-transcriptional gene regulatory systems. The conserved PUF family RNA-binding proteins repress gene expression post-transcriptionally by binding to sequence elements in 3'-UTRs of mRNAs. Despite their well-studied implications for development and neurogenesis in metazoa, the mammalian PUF family members are only poorly characterized and mRNA targets are largely unknown. We have systematically identified the mRNAs associated with the two human PUF proteins, PUM1 and PUM2, by the recovery of endogenously formed ribonucleoprotein complexes and the analysis of associated RNAs with DNA microarrays. A largely overlapping set comprised of hundreds of mRNAs were reproducibly associated with the paralogous PUM proteins, many of them encoding functionally related proteins. A characteristic PUF-binding motif was highly enriched among PUM bound messages and validated with RNA pull-down experiments. Moreover, PUF motifs as well as surrounding sequences exhibit higher conservation in PUM bound messages as opposed to transcripts that were not found to be associated, suggesting that PUM function may be modulated by other factors that bind conserved elements. Strikingly, we found that PUF motifs are enriched around predicted miRNA binding sites and that high-confidence miRNA binding sites are significantly enriched in the 3'-UTRs of experimentally determined PUM1 and PUM2 targets, strongly suggesting an interaction of human PUM proteins with the miRNA regulatory system. Our work suggests extensive connections between the RBP and miRNA post-transcriptional regulatory systems and provides a framework for deciphering the molecular mechanism by which PUF proteins regulate their target mRNAs.

Figure 1. mRNAs specifically associated with human PUM proteins.

Rows represent unique transcripts ordered according to increasing FDRs determined by SAM analysis. Columns represent individual experiments. The colour code indicates the degree of enrichment (green-red log₂ ratio scale). (A) mRNAs associated with PUM1. Three experiments with PUM1 protein and three mock experiments both with dye-swap technical replicates are shown. (B) mRNAs associated with PUM2. Four experiments with PUM2 protein and three mock experiments are shown. (C). Venn diagram representing overlap between PUM1 and PUM2 targeted transcripts (right) and the corresponding genes (ENSEMBL, left).

Figure 2. PUM targets encode proteins acting in cancer related pathways.

Components whose mRNAs are associated with PUM1 are depicted in yellow, those bound by both PUM1 and PUM2 are shown in red. Messages that contain a PUF motif are shown with a thick black border. (A) Regulators of angiogenesis. PUM associated messages code for the tyrosine kinase receptors fms-related tyrosine kinase 1 (FLT1), which is the vascular endothelial growth factor receptor 1, fibroblast growth factor receptor 1 (FGFR1) and EPH receptor B4 (EPHB4) and its ligand ephrin-B1 (EFNB1). These receptors and their ligands can trigger signals that induce angiogenesis , , . ARAF (v-raf murine sarcoma 3611 viral oncogene homolog) and MAPK1 (mitogen-activated protein kinase 1) are part of the RAS/RAF/MAPK pathway that can activate ETS (E26 transformation specific sequence) family transcription factors that promote angiogenesis . Human PUM proteins commonly target messages of both canonical (Wnt/ß-catenin) and non-canonical (Wnt/calcium signaling and planar cell polarity) pathways: WNT5A (wingless-type MMTV integration site family, member 5A) activates non-canonical Wnt signaling , which induces proliferation of endothelial cells in vitro. WNT5A is thought to promote the expression of the angiogenic effectors MMP1 (matrix metallopeptidase 1), and TEK (endothelial TEK tyrosine kinase, TIE-2) . PUM1 and PUM2 commonly target components of the “ß-catenin destruction complex” consisting of the serine/threonine kinase GSK3A (glycogen synthase kinase 3 alpha), which phosphorylates ß-catenin marking the protein for ubiquitylation and rapid degradation by the proteasome, the tumor suppressor APC (adenomatous polyposis coli), and the scaffold protein AXIN1. PUM1 further associates with mRNAs coding for the co-factors TCF/LEF1 (transcription factor/lymphoid enhancer-binding factor 1) that become activated when ß-catenin enters the nucleus. This includes TCF4 and TCF7L27 (transcription factors 4 and 7-like 2), that stimulate the transcription of genes implicated in cell growth regulation . (B) Activators and effectors of RAS , . PUM bound messages code for EGFR (epidermal growth factor receptor), adaptor proteins GRB2 (growth factor receptor-bound protein 2) and SHC1 (Src homology 2 domain containing transforming protein 1), which activate Ras proteins upon recruitment of the guanine nucleotide exchange factor SOS (son of sevenless). RAS interacts specifically with ARAF, MAP3K1 (mitogen-activated protein kinase kinase kinase 1), PIK3CB (phoshoinositide-3-kinase, catalytic, beta polypeptide) and TIAM2 (T-cell lymphoma invasion and metastasis 2), which can initiate cascades of protein-protein interactions and further activate more specific signaling pathways. Components of the Raf/MEK/ERK and the MEKK/SEK/JNK pathways are covered by PUM1 targets encoding mitogen-activated protein kinases MAPK1 and MAPKAPK5 (mitogen-activated protein kinase-activated protein kinase 5), and RPS6AK3 (ribosomal protein S6 kinase, polypeptide 3). These pathways target the transcription factors JUN (jun oncogene), ATF2 (activating transcription factor 2) and STAT1 (signal transducer and activator of transcription), which commonly induce cell proliferation . The PI3K-mediated (PIK3CB) signal is further triggered by activation of protein kinase B (AKT1, v-akt murine thymoma viral homolog 1) and phosphorylates GSK3A. PUM1 also targets effectors downstream of TIAM such as the GTP-binding protein RAC1 (ras-related C3 botulinum toxin substrate 1) and the Ras homologs gene family members B, F and J (RHOB, RHOF, RHOJ), which are all components of the TIAM/RAC/RHO signaling pathway implicated in the reorganization of the actin cytoskeleton . The RAC effector PAK2 (p21 protein- activated kinase 2) is involved in cell migration and invasion , and EXOC2 (exocyst complex component 2) induces vescicle trafficking upon RAL (Ras-related) activation .

Figure 3. Analysis of an RNA consensus sequence associated with human PUM proteins.

(A) PUF consensus motif in 3′-UTR sequences associated with PUM1, PUM2, Drosophila Pum and yeast Puf3, Puf4 and Puf5 proteins , . Height of the letters indicates the probability of appearing at the position in the motif. Nucleotides with less than 10% appearance were omitted. (B) Distribution of PUF consensus motifs. (C) Number of PUF motifs in the 3′-UTRs of PUM bound messages. (D) Distances between double PUF motifs present in 3′-UTR. Represented bins are 50 nts (0–4000 nts distance) and 10 nts (0–200 nts distance). (E) Analysis of PUF motif conservation among PUM1 and PUM2 targets. The x-axis shows the position (relative to the middle of the PUF motif), and the y-axis shows the logarithm (base 10) of the p-value from the Wilcoxon test determining whether conservation scores come from the same distribution for PUM targets and non-targets. The vertical blue line is drawn at position 0 corresponding to the PUF motif. The dashed black line is drawn at a p-value of 0.05, the continuous black line at a p-value of 0.01, and the red line at a p-value of 10⁻⁵, which is the threshold for significance considering multiple testing.

Figure 4. Validation of human PUM mRNA targets.

RNA-protein complexes formed between biotinylated 3′-UTRs and extracts of HeLa S3 cells expressing PUM1-HD-TAP and PUM2-HD-TAP were purified on streptavidin magnetic beads and monitored for the presence of TAP-PUM-HD by immunoblot analysis with anti-PAP antibody. (A) Biotin-labeled 3′-UTR sequences for indicated genes (lanes 3 to 8) were incubated with PUM1-HD-TAP and PUM2-HD-TAP extracts (lane 1). Rps26 3′-UTR was used as negative control probe RNA (lanes 8/9). The supernatant after pull-down with INTS2 is shown in lane 2. (B) Validation of the PUF-binding motif. Biotinylated RNA corresponding to the Dll1 3′-UTR was combined with PUM1-HD-TAP extract (lane 2) and 100-fold excess of competitor RNA (R1; AUUGUAAAUA; lane 3) or control RNA where the core motif is mutated (R2; AUACAAAAUA; lane 4). A fragment of MET 3′-UTR bearing wild type (UGU) or mutant (ACA) PUF binding sites is shown in lanes 5 and 6.

Figure 5. miRNA binding sites are enriched among human PUM targets.

(A) Example of a motif identified using the Phylogibbs motif finding algorithm in the vicinity (400 nucleotides upstream) of high-confidence miR-30a target sites. The x-axis indicates the position of a nucleotide in the inferred motif, and the y-axis gives the information score (bits) at that position. The height of each letter is proportional to the frequency of the respective nucleotide at that particular position in the alignment of inferred sites. (B) Distribution of the density of high-confidence miRNA sites (sites per nucleotide of 3′-UTR) in the 3′ UTRs of PUM targets (IPed, red) and non-targets (non-IPed, black).