ABSTRACT
Alternative splicing contributes greatly to the diversification of mammalian proteomes, but the molecular basis for the evolutionary emergence of splice variants remains poorly understood. We have recently found a novel class of splicing regulatory elements between the polypyrimidine tract (Py) and 3′ AG (REPA) at intron ends in many human genes, including the multifunctional PRMT5 (for protein arginine methyltransferase 5) gene. The PRMT5 element is comprised of two G tracts that arise in most mammals and accompany significant exon skipping in human transcripts. The G tracts inhibit splicing by recruiting heterogeneous nuclear ribonucleoprotein (hnRNP) H and F (H/F) to reduce U2AF65 binding to the Py, causing exon skipping. The resulting novel shorter variant PRMT5S exhibits a histone H4R3 methylation effect similar to that seen with the original longer PRMT5L isoform but exhibits a distinct localization and preferential control of critical genes for cell cycle arrest at interphase in comparison to PRMT5L. This report thus provides a molecular mechanism for the evolutionary emergence of a novel splice variant with an opposite function in a fundamental cell process. The presence of REPA elements in a large group of genes implies their wider impact on different cellular processes for increased protein diversity in humans.
INTRODUCTION
Alternative precursor mRNA (pre-mRNA) splicing greatly increases the proteomic diversity in metazoans (1–3). In particular, splice variants have reached the highest complexity in humans and other primates (4, 5), with about 90% of human genes alternatively spliced (6, 7). Aberrant splicing causes a large fraction of human genetic diseases (8, 9). However, the molecular basis for the evolutionary emergence of alternative exons that impact protein functions and cellular processes remains largely unknown, although several models have been proposed (10).
The 3′ end of introns between the polypyrimidine tract (Py) and 3′AG is highly constrained in sequence and length, with a consensus sequence of PyNYAG (Y, pyrimidine; N, any nucleotide) (11). However, we have found a CA-rich splicing regulatory element called CaRRE1 at this location (12–15), suggesting relaxation of the constraint in some transcripts and the potential existence of other, similar elements. In particular, a purine-rich (G-rich or A-rich) element such as a G tract at this location is expected to strongly disrupt the 3′ splice site (3′SS).
G tracts with a minimal functional GGG motif are splicing regulatory elements bound by heterogeneous nuclear ribonucleoprotein (hnRNP) H (H1) or its paralogues, including hnRNP F (16–24). They are enhancers or silencers of splicing, depending on their location in the pre-mRNA (17, 22, 24–28). We have identified G tracts between the Py and 3′ AG in more than a thousand human genes, including PRMT5 (for protein arginine methyltransferase 5). We call elements at this location REPA (regulatory elements between the Py and 3′AG) (29). These REPA G tracts appear to have mostly emerged in mammalian ancestors to act as splicing silencers involving hnRNP H and F (H/F). However, critical questions regarding the molecular mechanisms with respect to their role in the emergence of splice variants as well as their functional consequences in cellular processes remain to be answered.
PRMT5 catalyzes the symmetrical dimethylation of protein arginines (30), including histones 2A, 3, and 4 (H2A, H3, and H4) (31–34), spliceosomal Sm proteins (30, 35), and tumor suppressors PDCD4 (programmed cell death 4) and p53 (36, 37). It controls gene transcription/RNA processing (32, 38–40), maintains circadian rhythm and stem cell pluripotency or proliferation (34, 40–43), accelerates cell cycle progression (44, 45), and promotes tumorigenesis (37, 46). However, the molecular basis of these diverse roles in cellular processes has not been fully understood.
Here we use PRMT5 as an example to elucidate a molecular mechanism for the emerged REPA G tracts to mediate human exon skipping and its consequences in the cell cycle.
MATERIALS AND METHODS
Plasmid construction.Splicing reporters with 3′SS of different species or human G-tract mutant were constructed as described previously (13, 29) (see Fig. 2B). To make PRMT5 exon 3 splicing reporters, the ApaI and BglII fragment of DUP175 was replaced by human PRMT5 exon 3 with partial upstream (62-nucleotide [nt]) and downstream (43-nt) introns. To make pET28a-hnRNP H for bacterial expression of His-hnRNP H, the reverse transcription-PCR (RT-PCR) product of HeLa RNA amplified with upstream 5′-TTGGATCCATGATGTTGGGCACGGAAGGTGGAGAG-3′ and downstream 5′-GGCTCGAGCTATGCAATGTTTGATTGAAAATCACTG-3′ primers was cloned into pET28a between the BamHI and XhoI restriction sites. The Myc-PRMT5L or -PRMT5S expression plasmids were made by PCR amplification of full-length PRMT5 or of short open reading frames of PRMT5 using Phusion High-Fidelity DNA polymerase, from a PRMT5 cDNA clone (identification no. [ID] 3833019; Open Biosystems), and cloned into the pCMV-Myc vector between EcoRI and BglII restriction sites. For lentivirus-based expression, Myc-PRMT5L or -PRMT5S was subcloned into lentiviral vector cppt2E (47). All constructs were confirmed by sequencing. Lentiviral plasmid pLKO.1 containing short hairpin RNA (shRNA) against the 3′ untranslated region (3′UTR) of human PRMT5 (shPRMT5; clone ID TRCN0000107085 [mature antisense sequence TATTCCAGGGAGTTCTTGAGG]) was purchased from Open Biosystems.
Cell culture and transfection.The cells were grown and transfected as described previously (29). We used 0.4 μg of expression plasmids for HeLa cells in 24-well plates. For virus preparation, we carried out calcium phosphate-mediated transfection of HEK293T cells as described previously (48).
RNA interference (RNAi) and rescue.Vesicular stomatitis virus glycoprotein-pseudotyped lentiviral vectors were produced and transduced as described previously (48). HeLa cells were transduced with shPRMT5-containing virus for 3 h in 24-well plates. After 36 h, the cells were transduced again with shPRMT5 virus only or with shPRMT5 virus and either of the rescue protein-expressing viruses for 12 h followed by splitting. Five days after first transduction, cells were harvested for Western blot analysis, RNA extraction, or immunostaining. The knockdown of hnRNP H/F was carried out as described previously (19, 29).
RT-PCR and primer extension.We performed semiquantitative RT-PCR for endogenous PRMT5 exon 3 splicing and for validation of differential expression in transcriptome sequencing (RNA-Seq) analysis, as previously described (13, 29). For human or zebrafish PRMT5, we used upstream primer 5′-CAGGAACCTGCTAAGAATCG-3′ or 5′-GACCCTGCAAAGTCACGTCCTG-3′ and downstream [γ-32P]ATP-labeled primer 5′-GCCAGTGTGGATGTGGTTG-3′ or 5′-GATTCGAGCCAGGTTGGCACAG-3′, respectively (see Fig. 2A), and amplified for 24 cycles. Primer extension procedures were carried out as described previously (13) but with a [γ-32P]ATP-labeled reverse primer, DUP10 (5′-CAAAGGACTCAAAGAACCTCTG-3′), annealed at 50°C. RT-PCR or primer extension products were resolved on 6% urea PAGE gels and exposed to phosphorimager plates after drying. We quantified band intensities using ImageJ (National Institutes of Health). The splicing efficiency or exon inclusion/skipping level is expressed as the variant intensity relative to the total level of transcripts (included plus skipped products).
UV cross-linking and immunoprecipitation.For UV cross-linking, HeLa nuclear extract was prepared as described previously (15, 49). RNA probes were in vitro transcribed in the presence of [α-32P]UTP with T7 RNA polymerase as described previously (50) from PCR products of human wild-type or G tract mutant splicing reporter minigenes using T7 promoter-tagged upstream primer 5′-TAATACCGACTCACTATAGGGAAGACTCTTGGGTTTCTG-3′ or 5′-TAATACGACTCACTATAGGGCTCAAACAGACACCATGCATGG-3′ and downstream primer 5′-CATGGTGTCTGTTTGAGGTTG-3′. UV cross-linking and immunoprecipitation were carried out as described previously (15). For immunoprecipitation, we used protein G-Sepharose beads (Pierce) coated with 1.5 μg of anti-hnRNP H/F (1G11-Santa Cruz Biotechnology) or 2.0 μg of anti-U2AF65 (MC3-Sigma-Aldrich). Immunodepletion of hnRNP H/F from a HeLa nuclear extract containing 0.5 M NaCl was carried out based on a published procedure (51). Immunodepleted HeLa nuclear extract was dialyzed 3 times against DG buffer (12) (20 mM HEPES-KOH [pH 7.9], 20% glycerol, 80 mM potassium glutamate, 0.2 mM EDTA, 0.2 mM phenylmethylsulfonyl fluoride [PMSF], 1.0 mM dithiothreitol [DTT]).
Purification of recombinant His-hnRNP H.To purify the recombinant N-terminal histidine-tagged hnRNP H, Escherichia coli Rosetta-gami 2 (DE3) pLysS (Novagen) bacteria transformed with pET28a-hnRNP H were grown overnight in LB medium containing 50 μg/ml kanamycin and induced with 0.3 mM isopropyl 1-thio-β-d-galactopyranoside at 37°C for 3 h. Protein purification was carried out as described previously (17) except that phosphate-buffered saline (PBS) was used instead of 50 mM Tris-HCl (pH 8.0) containing 0.1 M NaCl.
Deep sequencing of RNA transcripts (RNA-Seq).HEK293T cells expressing Myc-PRMT5L or -PRMT5S were used for RNA-Seq due to their high transfection efficiency (∼90%, confirmed by immunostaining of Myc-tagged epitopes). The total RNAs of control (nontransfected) and PRMT5L- or PRMT5S-expressing groups (each processed in triplicate) were reverse transcribed using random primers for library preparation and subjected to Illumina HiSeq protocols at the McGill University and Genome Quebec Innovation Centre using a HiSeq 2000/2500 sequencer. We obtained on average 67 ± 1.9 million uniquely mapped paired HiSeq reads for each sample, with a total of ∼2 million reads and 23,342.33 ± 168.56 (42.41% ± 0.3%) mapped genes per treatment group. The differential gene expression analysis was done using the DEseq and edgeR Bioconductor package (52, 53). The edgeR-adjusted P values (<0.05) were used to filter for the significantly changed transcripts.
Western blot analysis and immunostaining.Western blot analyses were performed as previously described (50). Anti-U2AF65, anti-hnRNP F/H (1G11), anti-hnRNP H, antinucleolin, anti-β-actin, and anti-cMyc were purchased from Santa Cruz Biotechnology Inc. Anti-PRMT5 (EPR5772) and anti-H4 symmetric dimethyl R3 (ab5823) were purchased from Abcam and rabbit anti-U2AF65 and fluorescein isothiocyanate (FITC)-conjugated anti-α-tubulin from Sigma. Immunostaining was carried out as described previously (50), using rabbit anti-cMyc or mouse FITC-conjugated anti-α-tubulin primary antibodies at a dilution of 1:100 or 1:1,000, respectively. Texas Red-conjugated anti-rabbit IgG secondary antibody was used at a dilution of 1:1,000. We used DAPI (4′,6-diamidino-2-phenylindole) to visualize the DNA at a 1:5,000 dilution.
Human genome search and pathway analysis.A human genome search and the following analysis were carried out as described previously (29).
RESULTS
Evolutionary “invasion” of G5–8 into the upstream 3′SS of a group of human exons.Of the approximately 1,000 REPA G tracts identified from the human genome (29), 130 contain G5–8 tracts that are significantly associated with alternative splicing (Fig. 1A; see also Table S1 in the supplemental material) (P = 9E–7 in hypergeometric testing). Most of them (82%) are G pentamers, and 12 of the 14 randomly examined pentamers (86%) were present only in mammalian genes in a sequence alignment of the 3′SS of multiple vertebrate species (29) (Fig. 1B; see also Table S1). Thus, the G pentamers appear to have emerged mainly in mammalian ancestors.
Identification and functional clustering of human genes containing REPA G5–8 between Py and 3′ AG of 3′SS. (A) Diagram showing the position of evolved REPA G5–8 (red) between Py and 3′AG, in comparison with the corresponding location of the constitutive 3′SS. Possible locations of the potential trans-acting factors U2AF65/35 (green ovals) and hnRNP H/F (red oval) binding to the 3′SS are indicated. (B) Functional clustering of genes that contain alternative exons with REPA G5–8. In red within the clustered functions are 7 of the 12 genes with their REPA G tracts found in mammalians only and not in lower vertebrates, from 14 genes that can be verified by sequence alignment. PRMT5 is highlighted in bold.
The host genes of these G tracts are significantly enriched in several categories of cellular functions (Fig. 1B). The biggest functional cluster contains 30 genes that control cell growth and proliferation, including the PRMT5 gene that is known to promote cell cycle progression (Fig. 2). We chose the G tracts upstream of its exon 3 for mechanistic and functional studies.
Alternative splicing of PRMT5 exon 3 and evolutionary emergence of REPA G tracts between the Py and 3′AG. (A) Diagram of pre-mRNA around human PRMT5 exon 3 (not to scale). Alternative splicing of exon 3 with concomitant usage of an alternative 3′SS within exon 4 creates longer or shorter PRMT5 splice variants, PRMT5L or PRMT5S. Red bar, location of the G tracts; arrowheads, positions of PCR primers. (B) Alignment of the upstream 3′SS sequences of PRMT5 exon 3 of 29 species. The G tracts between Py and 3′ AG are indicated above the aligned sequences. Dotted lines and blue nucleotides, corresponding position of the G tracts in birds and other lower chordates. At the bottom is the consensus sequence of constitutive 3′SS for comparison. (C) Usage of PRMT5 exon 3 in human cell lines and zebrafish tissues. A representative denaturing PAGE gel of semiquantitative RT-PCR products with exon 3 included (PRMT5L) or excluded (PRMT5S), as confirmed by sequencing, is shown. The asterisk indicates a product from a potential cryptic splice site.
Exon 3 is skipped in some human transcripts in the University of California, Santa Cruz (UCSC), Genome Browser (e.g., GenBank accession no. AK302240.1). In all 15 cases of the skipping, a 5′ fragment of exon 4 is also skipped due to alternative usage of a 46-nt downstream 3′SS within the exon (Fig. 2A), producing a previously uncharacterized shorter isoform, PRMT5S. The involvement of exon 4 suggests competition or cooperative interactions between the splice sites, as observed by others (54–58); however, this mode of regulation is not common among the 130 exons. Therefore, we focused on exon 3 skipping to illustrate the role of the G tracts and their mechanism of action.
The human 3′SS upstream of exon 3 harbors a G3TG5 element (Fig. 2B). Similar G tracts of various lengths are present in most mammals but absent in chick or other lower chordates, including zebrafish, as shown in the sequence alignment of the corresponding 3′SS of 29 species (G nucleotides within the element in red in Fig. 2B). Intermediate G0–2[T(A/C)T]TG2–6 sequences are also present in a group of mammalian species within the region corresponding to the REPA G tract (Fig. 2B). The alignment thus suggests that the G tracts emerged in a mammalian ancestor but have since evolved differently through accumulated insertions/deletions/point mutations.
Semiquantitative low-cycle-number PCR after reverse transcription (RT-PCR) showed that the PRMT5S isoform comprises about 30% of transcripts in human HeLa, MDA-231, and HEK293T cells whereas the proportion is about 5-fold lower in zebrafish muscle, egg, and brain tissues (Fig. 2C). This dramatic difference in exon skipping levels between human and fish is consistent with the splicing silencer role of the REPA G tracts in PRMT5 and other genes (29). We thus went on to further characterize its effect on exon usage and mechanism of action.
The evolved PRMT5 G tracts inhibit U2AF65 binding to Py and splicing by mainly recruiting hnRNP H.We first assessed the effect of the PRMT5 REPA G tracts of different species on splicing by transferring each corresponding 3′SS to the upstream sequence of a heterologous constitutive exon of the DUP175 splicing reporter (13) (Fig. 3A). The vector itself showed almost no exon skipping (Fig. 3A, lanes 1 and 2) in a primer extension assay of the transiently expressed transcripts in HEK293T cells. Replacing the vector 3′SS with the zebrafish 3′SS, which contains no G tracts, did not change this splicing pattern (lane 3). The opossum 3′SS, containing a G tetramer, caused 15% exon skipping plus a substantial amount of cryptic splicing (indicated by an asterisk in Fig. 3A) at a downstream 3′SS in the middle exon (lane 4), a product verified in our previous reports (15, 59). In contrast, the 3′SS of platypus, dog, and humans, each containing a G4 plus a G3, a G6 plus a G2, and a G5 plus a G3, respectively (Fig. 2B), caused almost complete skipping of the exon along with a substantial level of cryptic splicing (Fig. 3A, lanes 5 to 7, >90% skipping). Importantly, mutating each G of the human element to an A strongly reduced the percentage of exon skipping as well as cryptic splicing and increased the full-length product (lane 8, 43% skipping), as seen in RT-PCR assays (29). Therefore, the 3′SSs of different species exhibit different effects on splicing, and the splicing inhibition is detected only in the presence of REPA G tracts.
Role of the PRMT5 G tracts as splicing silencers. (A) Splicing of the PRMT5-3′SS splicing reporter minigenes, containing 3′SS of different species upstream of middle exon in the DUP175 vector. Upper panel, diagram of the minigenes. Black bar, upstream 3′SS of PRMT5 exon 3 of different species that replaced the corresponding 3′SS of DUP175 (not to scale). Splicing patterns are indicated by slanted lines linking exons (boxes). *, cryptic splicing; arrowhead, reverse primer for primer extension. Lower panel, representative denaturing PAGE gels of primer extension assays of minigenes containing the upstream 3′SS of vector (lane 2) or of PRMT5 exon 3 of different species (lanes 3 to 8). C, mock-transfected sample. The percentages (%) of exon skipping are calculated as the intensity of the 204-nt product relative to the total of the 204-nt and full-length (379-nt) products in each lane. S.D., standard deviation; nt, nucleotides; WT, wild type; Mut, mutant. (B) Upper panel, diagram of the minigene for human PRMT5 exon 3 (E3; 86 nt) and its splicing products (not to scale). Black box and long bars, human PRMT5 exon 3 and partial flanking introns, respectively. Short bar, wild-type G tracts or mutant. Arrowhead, primer. Lower panel, primer extension assay performed as described for panel A. (C) UV cross-linking assay of the binding of hnRNP H/F to the human PRMT5 G tracts. Upper panel, probes used in UV cross-linking. Lower panel, phosphorimages of PAGE gels of proteins from HeLa nuclear extract cross-linked to wild-type or mutant RNA probes (lanes 2 and 3) and immunoprecipitated hnRNP H(H1) from the reactions (IP; lanes 4 to 6). (D) Effect of hnRNP H/F knockdown on endogenous PRMT5 exon 3 skipping in HeLa cells. Upper panel, Western blot analysis of nontreated or mock-, control siRNA (scrambled)-, or hnRNP H/F siRNA-transfected HeLa cell lysates (lanes 1 to 4), showing knockdown of hnRNP H/F. Lower panel, denaturing PAGE gels of semiquantitative RT-PCR of endogenous PRMT5 variant products and GAPDH (glyceraldehyde-3-phosphate dehydrogenase). (E) Effect of hnRNP H/F immunodepletion and of His-hnRNP H add-back on U2AF65 binding to the Py of the upstream 3′SS of PRMT5 exon 3. Upper panel, Western blot analysis of normal (lane 1), hnRNP H/F-depleted (lane 2), or His-hnRNP H add-back HeLa nuclear extracts (lanes 3, 4, and 5). *, His-hnRNP H fragment. Lower panel, diagram of RNA probe (top), phophorimage of immunoprecipitated U2AF65 cross-linked to the G-tract-containing PRMT5 RNA probe in HeLa nuclear extracts (middle), and Western blot analysis of the same gel showing equal loading of U2AF65 (bottom).
To confirm that the human G tracts also inhibit splicing of their original downstream exon, we cloned PRMT5 exon 3 with partial flanking introns into DUP175 and tested its splicing with wild-type or G tract mutant 3′SS in HEK293T cells. A total of 57% (±1.8%, n = 3) of the transcripts from wild-type reporter showed exon 3 skipping in primer extension assays (Fig. 3B, lane 2). Importantly, mutating the G tracts nearly abolished the exon skipping (lane 3, 5% skipping). Thus, the evolved human G tracts are indeed splicing silencers of PRMT5 exon 3.
To elucidate the molecular mechanisms underlying the splicing inhibition by the G tracts, we first verified their trans-acting factors using UV cross-linking assays with wild-type or mutant RNA probes and HeLa nuclear extracts. In these assays, a cross-linked protein band of about 50 kDa was specifically abolished by the G-to-A mutations (Fig. 3C, lanes 2 and 3), reminiscent of the G tract-binding proteins hnRNP H/F (16, 18–24, 29). Therefore, we immunoprecipitated the cross-linked products with anti-hnRNP F/H (1G11) or anti-hnRNP H (H1) antibody. The result indicates that both hnRNP H and hnRNP F are cross-linked to the wild-type but not to the mutant probes (lanes 4 and 5) and that the major band contains hnRNP H (lane 6).
Previously, we observed a slight increase of the exon 3 inclusion upon hnRNP H/F knockdown in a screening of several exons using agarose gels (29). To more accurately measure the molar changes of the splice variants, we carried out the more sensitive assay using 32P-labeled primers for RT-PCR and denaturing polyacrylamide gels. The result showed that the proportion of exon 3-skipped transcripts was about 50% in nontreated cells under this culture condition and that the proportion decreased significantly to 28% upon hnRNP H/F knockdown (Fig. 3D, compare lane 4 to lanes 1 to 3) (P = 7.6E–05, n = 3). Taken together with the mutagenesis and cross-linking data (Fig. 3A to C), this result supports the hypothesis that G tract-binding hnRNP H and hnRNP F are indeed splicing repressors of endogenous PRMT5 exon 3.
To determine the mechanism of splicing inhibition by the element/trans-acting factors, we examined the binding of U2AF65 to wild-type or G tract mutant RNA probes in HeLa nuclear extracts. G-to-A mutation of G tracts increased the amount of U2AF65 immunoprecipitated after UV cross-linking of nuclear extract to the RNA probe (see Fig. S1 in the supplemental material), suggesting an inhibitory effect of G tracts on U2AF65 binding. Since the inhibition of splicing by G tract requires the presence of hnRNP H/F in cells (29), we also examined the effect of the trans-acting factors on U2AF65.
We examined U2AF65 binding to the wild-type RNA probe in hnRNP H/F-immunodepleted or hnRNP H add-back HeLa nuclear extracts (Fig. 3E). The depletion of both hnRNP H and F from HeLa nuclear extracts enhanced (2.13-fold) cross-linking of U2AF65 to the RNA (lanes 1 and 2). Importantly, adding back bacterially expressed His-hnRNP H reversed the depletion effect in a dose-dependent manner (lanes 3 to 5). Thus, hnRNP H binding to the G tracts inhibits U2AF65 binding to the Py.
Taken together, these data demonstrate that the evolved human REPA G tracts recruit mainly hnRNP H to inhibit U2AF65 binding and splicing of PRMT5 exon 3. This provides a molecular basis for the mechanism of splicing inhibition at an early stage of spliceosome assembly.
The splicing inhibitory effect of these G tracts is consistent with their atypical location disrupting the 3′SS consensus but is completely opposite to the reported enhancing effect seen when they are positioned at other regions of mammalian introns (17, 23, 24, 26). The evolutionary emergence of the alternative splice site by “invasion” of a de novo element between the Py and 3′AG is also distinct from the creation of other alternative splice sites by simple point mutations (60).
The G tract-mediated exon skipping creates a shorter PRMT5 protein (PRMT5S) that exhibits distinct localization and promotes cell cycle arrest at interphase.To determine the functional consequences of the emerged splice variant of PRMT5, we examined the resulting PRMT5S protein. Skipping of exon 3 and concomitant usage of the alternative 3′SS within exon 4 result in deletion of 44 amino acids from the NH2 terminal region of the original full-length (637-amino-acid [aa]) PRMT5 protein (Fig. 4A, PRMT5S and PRMT5L). However, PRMT5S (593 aa) still contains the intact C-terminal methyltransferase domain, suggesting that it may also methylate proteins, as does PRMT5L.
Domains of PRMT5S and PRMT5L proteins and rescue of H4R3 methylation by either of them. (A) Diagram of the domains of human PRMT5S (upper panel) and PRMT5L (lower panel) and the region of G tract-regulated exon 3 as well as exon 4 (middle panel). Red nucleotides, G tracts; aa, amino acids; AdoMet-MTase, S-adenosylmethionine-dependent methyltransferases; red amino acids and red bar, the peptide deleted in PRMT5S because of exon 3 and partial exon 4 skipping. The arrowhead in exon 4 indicates an alternative 3′ splice junction used for PRMT5S. (B) Western blot showing knockdown with lentiviral short hairpin PRMT5 (shPRMT5) and rescue with MycPRMT5L (L) or MycPRMT5S (S) in HeLa cells. (C) Western blot showing the histone H4R3 methylation in HeLa cells after PRMT5 knockdown/rescue. # and *, bands of unknown identity, likely symmetrically dimethylated arginine epitopes of unknown proteins that are also recognized by the antibody.
To determine if the PRMT5S has an effect on the methylation status of target proteins in cells, we looked for such an effect by expressing either PRMT5L or PRMT5S in HeLa cells in which the total PRMT5 proteins were knocked down by lentivirus-mediated RNA interference. The PRMT5 proteins were efficiently knocked down, and the exogenous isoforms were expressed in the knockdown cells (Fig. 4B). We probed the cell lysates with an antibody against histone 4 symmetric dimethyl R3 (H4R3me2s), a known target of PRMT5 (31, 32, 34, 61). The antibody recognizes two bands of ∼12 kDa and 15 kDa in SDS-PAGE gels consistent with the known sizes of the H4 protein (31, 61) (Fig. 4C). The H4R3me2s level decreased upon PRMT5 knockdown. Importantly, it was restored to the original level upon expression of either PRMT5L or PRMT5S. Therefore, PRMT5S is capable of maintaining H4R3 methylation, as is PRMT5L. This effect is consistent with the intact methyltransferase domain in PRMT5S (Fig. 4A).
To determine the effect of the PRMT5S protein on cells, we examined it during the cell cycle, where PRMT5L plays an important role (41, 44–46, 61). We differentiated the interphase and mitotic phases by the typical tubulin and DAPI staining patterns in HeLa cells that had been left untreated (control), subjected to PRMT5 knockdown (shPRMT5), or rescued with either one of the splice variants (Fig. 5A and B). Immunostaining of the endogenous PRMT5 in control cells showed diffuse localization to the nucleus and cytoplasm, as well as enrichment in the cytoplasm near the nucleus (Fig. 5A, upper panel, red), consistent with a previous report (62). The immunosignals were abolished by shPRMT5 expression as expected (second panel from top, PRMT5 row). The expressed MycPRMT5L mainly localized near the nucleus, a pattern similar to its colocalization with the Golgi apparatus (63), whereas MycPRMT5S localized more diffusely throughout the cell (lower two panels, red). A Flag-tagged PRMT5L showed localization similar to that of the MycPRMT5L (data not shown). Moreover, though their localization patterns changed during the cell cycle, the major difference with respect to diffusing distribution or punctate distribution remained. We thus conclude that PRMT5S exhibits localization patterns distinct from those of PRMT5L in cells.
Effect of PRMT5S opposite to that of PRMT5L on the cell cycle. (A) Representative images of immunostained HeLa cells at different stages of the cell cycle, subjected to total PRMT5 knockdown or rescued with MycPRMT5L or MycPRMT5S as described for Fig. 4B, using anti-α-tubulin or anti-PRMT5. DAPI staining of the nuclei is also shown below each sample. Arrowheads indicate representative cells in the images with multiple cells; dotted lines indicate the diving nuclei. (B) Upper panel, percentages of HeLa cells in interphase or prophase to telophase (mitosis) after knockdown of PRMT5/rescued as described for panel A and in the Fig. 4B legend. ***, P < 0.001. Lower panel, representative lower-magnification images of PRMT5 knockdown/rescue HeLa cells. The patterns of tubulin or DAPI staining typical of the cell cycle stages as described for panel A indicate that the effect of PRMT5S on the cell cycle is opposite that of PRMT5L. (C) Index of mitosis versus the relative levels of PRMT5S protein. The relative levels of PRMT5S [PRMT5S/(PRMT5S + PRMT5L)], calculated from Western blots of HeLa cells, are shown in red and total PRMT5 protein in blue. For the index of mitosis, the percentage of cells in mitotic phases (prophase to telophase) of the MycPRMT5L rescue group is taken as 100; all others are normalized to this group. The upper and lower bands of PRMT5 in Fig. 4B were measured as PRMT5L and PRMT5S, respectively.
To determine the effects of the two variants on the cell cycle, we examined the number of cells at the interphase or mitotic phases of the cell cycle among these cells (Fig. 5A and B). Upon knockdown of endogenous total PRMT5, the percentage of interphase cells increased from 55.3% in nontransduced controls to 68.2% (±5.09%, P = 0.01, n = 3 groups, 300 cells/group, same in the following cell phase analysis in this panel) (Fig. 5B). This change is consistent with the predominant expression of PRMT5L in HeLa cells (Fig. 2C and 4B) and its mitosis-promoting effect (41, 44–46, 61). Importantly, as anticipated, rescue with MycPRMT5L significantly reduced the proportion of interphase cells to 28.7% (±6.6%, P = 0.002) (Fig. 5B). In contrast, rescue with MycPRMT5S could not reverse the effect of knockdown on mitosis but instead further enhanced it by increasing the proportion of interphase cells to 76.2% (±3.8%, P = 0.001). Furthermore, the level of PRMT5S [relative to the total level of PRMT5 protein: PRMT5S/(PRMT5S + PRMT5L)], calculated from Western blot analyses of HeLa cells, is inversely correlated with the mitosis index whereas the relative level of the total PRMT5 protein alone (PRMT5S + PRMT5L) did not exhibit such a correlation (Fig. 5C). This suggests that the relative level of the PRMT5S isoform, but not the total level of the PRMT5 protein, corresponds to the level of mitosis inhibition in HeLa cells. Therefore, the human PRMT5S has an effect on cell cycle progression opposite that seen with its full-length counterpart, PRMT5L.
PRMT5S preferentially regulates expression of a group of genes involved in cell cycle arrest at interphase.To obtain insight into why the evolved PRMT5S has an effect on the cell cycle opposite from that seen with PRMT5L, beyond their obvious differences in subcellular localization, we carried out deep sequencing of mRNA transcripts to identify target genes that PRMT5S may regulate differently from PRMT5L.
The transcriptomes of HEK293T cells overexpressing either of the PRMT5 variants were sequenced using paired-end RNA HiSeq (see Materials and Methods). We identified 50 genes that are preferentially regulated by PRMT5S (regulated specifically by PRMT5S or more strongly by PRMT5S than by PRMT5L, P < 0.05). We were able to validate the trends of changes observed in RNA-Seq analysis by semiquantitative RT-PCR analysis of all 22 genes that gave expected products among the 25 genes examined (see Fig. S2 in the supplemental material).
To determine the impact of preferential gene expression regulation by PRMT5 variants on potential cellular functions, we carried out gene ontology analysis. Consistent with the known function of PRMT5 in promoting cell proliferation (41, 44–46, 61), PRMT5L-preferred genes were significantly enriched in functional clusters of cell growth and proliferation. Strikingly, the functions of the 50 PRMT5S-preferred genes clustered most significantly in cell death and cell cycle arrest pathways (Fig. 6A) (P < 0.0001). In particular, 8 of them are known to arrest cell cycle at different stages of interphase (G0 or G1, S, or G2).
Identification of a group of cell cycle-arresting genes preferentially regulated by PRMT5S in RNAi and rescue assays. (A) Functional clusters with most significant enrichment of PRMT5S-preferred genes identified by RNA-Seq analysis. (B) Genes preferentially regulated by PRMT5L or PRMT5S in PRMT5 knockdown and rescue assays. Left panel, representative agarose gels of the RT-PCR products of genes that have been confirmed to be preferentially regulated by PRMT5L or PRMTS in RNAi/rescue assays in HeLa cells, in alphabetical order for the PRMTL targets and in order of their description in the text for the PRMT5S targets. Four genes not reported to affect cell cycle but also regulated by PRMT5S are shown below the cell cycle-regulating ones. GAPDH, RNA loading control. Right panel, bar graphs (averages ± standard deviations [SD], n = 3) of the changes of the corresponding genes presented in the same order as in the left panel. Numbers above the dotted lines on top of bars represent the fold change values beyond the vertical axis scale. *, P < 0.05; **, P < 0.01; ***, P < 0.001.
To confirm the preferential regulation of cell cycle-arresting genes by PRMT5S, we again used HeLa cells with RNAi-mediated knockdown of total PRMT5 followed by rescue with either PRMT5S or PRMT5L to examine the expression of 19 genes identified in RNA-Seq analysis as preferentially regulated targets. The changes in the expression of 16 (84%) of these genes upon PRMT5 knockdown were rescued by either one or both PRMT5 isoforms, consistent with the trends observed in RNA-Seq analysis, and 15 of them showed significantly preferred regulation of expression (Fig. 6B). Of these, 5 genes were preferentially rescued by PRMT5L and 10 by PRMT5S (P < 0.05). Consistent with the gene ontology analysis, the PRMT5S-preferred group includes 6 genes that are known to induce cell cycle arrest when regulated in this manner. The products of these genes function at different phases of the cell cycle. Specifically, cysteine-rich angiogenic inducer 61 (CYR61), dual-specificity phosphatase 1 (DUSP1), and growth arrest and DNA-damage-inducible 45 beta (GADD45B) act at the G1 phase (64–66); NIMA (never in mitosis gene A)-related kinase 11 (NEK11) acts at the S phase (67); and DUSP1, GADD45B, NEK11, and connective tissue growth factor (CTGF) act at the G2 phase (66, 68–72). Moreover, PRMT5S in particular reversed the elevation of the levels of transcripts of proto-oncogene ETS1 (transcription factor E-26 transforming sequence 1) which were observed in the PRMT5 knockdown cells. ETS1 is also an enhancer of the G1/S-phase transition (73). Therefore, PRMT5S preferentially controls the expression of multiple genes that cause cell cycle arrest at different stages of interphase, providing a reasonable explanation for its different effect on the cell cycle compared to that of PRMT5L.
Taken together, our data demonstrate that the REPA G tracts emerged in mammalian ancestors to recruit mainly hnRNP H for the inhibition of U2AF65 binding to the Py, resulting in splicing repression. A novel PRMT5 isoform was thus generated to have a different subcellular localization and effects on cell cycle progression, possibly through preferential control of cell cycle-regulating genes, from those of the original isoform. This report therefore provides a molecular mechanism for the REPA G tracts to help generate functional diversity of a protein over the course of evolution.
DISCUSSION
Several models have been proposed for the evolutionary emergence of an alternative exon (10), including transition from constitutive to an alternative exon during evolution, which appears to be the case for PRMT5 exon 3. It transited from a mainly constitutive exon in fish and birds to an alternative exon in mammals during evolution. The existence of weakly detectable exon 3 skipped transcripts in fish (Fig. 2C) is perhaps a consequence of other elements or factors affecting the splice site strength. However, the critical factor for this transition is apparently the insertion of REPA G tracts in mammalian ancestors. This report provides a mechanism by which they help generate a novel splice variant that has a distinct functional consequence in cells.
The reported roles of PRMT5 are quite diverse, but the molecular basis of its functional diversity has been largely unclear. The novel PRMT5S variant that emerged in mammals adds to the protein and functional diversity of PRMT5 by exhibiting a different localization and differential effects on the expression of critical genes in cell cycle control (Fig. 5 and 6). The Western blot analysis of endogenous PRMT5 also showed a shorter protein corresponding to the expected size of PRMT5S (Fig. 4B), consistent with previous observations in different cell lines (29, 74). Moreover, PRMT5S is differentially expressed among human tissues (M. Sohail and J. Xie, unpublished data), perhaps as a tissue-specific modulator of cell cycles. Interestingly, PRMT5S enhanced the differentiation of dendritic cells whereas PRMT5L showed an inhibitory effect in ectopic expression assays (M. Sohail, S. Kung, and J. Xie, unpublished data). The different effects on cellular processes along with spatially and temporally regulated alternative splicing of the two variants thus likely contribute to the diverse roles of PRMT5, including but not limited to the maintenance of stem cell pluripotency and promotion of cancer cell growth as well as cell differentiation (34, 41–46, 62). Understanding these effects could be facilitated by revealing the potentially different profiles of methylation targets and enzyme activities of the PRMT5 isoforms in subcellular locations and compartments in future investigations.
The unusual invasion of the REPA G tract into the space between the Py and 3′AG and its mechanism of action revealed here clearly demonstrate that this region of mammalian 3′SS has been remarkably permissive for the evolutionary emergence of de novo regulatory RNA elements. It is possible that other REPA elements, in particular, purine-rich ones, emerged similarly at this location as well. Thus, more REPA elements are expected to be discovered in the mammalian genome in future investigations. The evolved element allows trans-acting factors to bind (Fig. 3) and, moreover, is likely to relay upstream signals for further regulation of alternative splicing of the mammalian genes, as in other cases (75). The presence of such elements in a group of genes implies that the impact of this molecular mechanism likely goes beyond the genes involved in cell cycle regulation, reaching a wider range of cellular processes in different tissues (Fig. 1B; see also Table S1 in the supplemental material) (29).
ACKNOWLEDGMENTS
We thank Doug Black for providing polyclonal hnRNP H/F and hnRNP F antibodies for the initial tests and for insightful comments on and editing the manuscript, Vincent Lobo for zebrafish, Niaz Mahmood for help with some of the gene expression experiments, the McGill University and Genome Quebec Innovation Centre, particularly Pascale Marquis, for their RNA-Seq service and gene expression analysis, Mike Myschyshyn and Irene Xie for editing, and Sika Zheng for helpful comments on the manuscript.
This work was supported by a discovery grant from the Natural Science and Engineering Research Council of Canada (NSERC) and in part by a CIHR grant (FRN106608) and a Manitoba Research Chair fund provided to J.X.
FOOTNOTES
- Received 19 February 2015.
- Returned for modification 16 March 2015.
- Accepted 6 April 2015.
- Accepted manuscript posted online 13 April 2015.
Supplemental material for this article may be found at http://dx.doi.org/10.1128/MCB.00190-15.
- Copyright © 2015, American Society for Microbiology. All Rights Reserved.
