Background
Poly(rC)-binding protein 1 (PCBP1) belongs to the heterogeneous nuclear ribonucleoprotein (hnRNP) family which is composed of hnRNP K/J and the alpha-complex proteins (PCBP1–4α or CP1–4) containing three hnRNP K homology (KH) domains for RNA-binding [
1]. Thus, PCBP1 and PCBP2 are also known as hnRNP E1 and hnRNP E2 [
2,
3], respectively, and share 89% of amino acid similarity [
4].
PCBP1 is ubiquitously expressed and functionally plays multiple roles in transcriptional and posttranscriptional regulations, including mRNA stabilization [
5,
6]. PCBP1 regulates gene expression via binding to specific elements of target mRNAs with AU-rich elements (AREs) or U-rich elements located in 3’-untranslated regions (3’-UTR) (e.g.
AR, p21, p63, eNOS) [
7‐
10] or in 5’-UTRs (e.g.
c-myc, PRL-3, EV71) [
6,
11,
12]. PCBP1 also inhibits the alternative spicing of CD44 [
13] or its alternative polyadenylation [
14]. Recent studies indicate that PCBP1 phosphorylation at Ser43 by Akt2 upon TGF-beta stimulation promotes epithelial to mesenchymal transition in various cancer cells [
15‐
19]. PCBP1 also functions as transport protein with PCBP2 together to deliver irons to ferritin for storage, or binds to prolyl and asparagyl hydroxylases to metalate the mononuclear iron center [
20‐
22]. In addition, PCBP1 modulates housekeeping degradation of mitochondrial antiviral signaling (MAVS) involved in antiviral immunity and anti-inflammation [
23]. Moreover, PCBP1 can bind with RACK1 to regulate MOR expression [
24].
Increasing evidence indicates that PCBP1 could be involved in repressing carcinogenesis, as downregulation of PCBP1 has been observed in multiple cancers, including cervical cancer [
25], liver cancer [
13], breast cancer cells [
26], colon cancer and lung cancer [
6]. Our previous study revealed that PCBP1 translationally represses metastatic PRL-3 and its overexpression inhibits tumorigenesis, whereas PCBP1 knockdown in turn enhances tumor formation. However, the simultaneous inverse correlation of PCBP1 protein level to that of PRL-3 is observed in only 37% lung and 24% colon carcinoma samples, as PCBP1 silence fails to provoke PRL-3 upregulation in some tumor samples [
6], indicating that PCBP1 could play multiple roles in tumor suppression, rather than only by delaying PRL-3 translation. So far, the underlying mechanism of PCBP1 in tumor suppression remains elusive.
Generally, protein level is not only determined by mRNA amount, but the posttranscriptional regulation. To thoroughly investigate PCBP1’s anti-tumorigenic character, here we first applied the transcriptome-wide RNA immunoprecipitation (RIP) to identify the stably bound mRNAs to PCBP1, and followed by high-throughout RNA sequencing. Of note, PCBP1 was bound to a class of mRNAs, in which cell cycle inhibitor p27Kip1 (p27) was notably identified as a novel PCBP1-bound transcript. We characterized that binding of PCBP1 to p27 mRNA 3’-UTR stabilized p27 mRNA, consequently increased p27 protein expression to induce cell cycle arrest, inhibit cell proliferation, and repress tumorigenesis both in vitro and in vivo. Eventually, we also demonstrated that loss of PCBP1 mRNA and protein were positively related to p27 downregulation in clinical tumor samples, indicating the general importance of p27 loss in tumorigenesis.
Methods
Cell culture
A2780 cells were cultured in DMEM (Hyclone), MDA-MB-231 cells were cultured in high-glucose DMEM, so as DLD-1 cells in RMPI-1640 medium. All media were supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin antibiotics. Cells were cultured at 37 °C with 5% CO2 in incubator (Thermo).
Immunoprecipitation and sequencing identification of GFP-PCBP1 bound mRNAs
A2780-GFP and A2780-GFP-PCBP1 cells were grown to 90% of cell confluency and washed with cold PBS, then treated for crosslinking of PCBP1 to the bound transcripts with 1% paraformaldehyde-PBS at 37 °C for 20 min on rocking platform and stopped with 0.2 M glycine-PBS at 37 °C for 5 min, followed by washing another 3 times with ice-cold PBS and harvesting cells by scraping in ice-cold PBS. Harvested cells were centrifuged at 2500 rpm for 5 min and lysed in 500 μl ice-cold hypotonic buffer (10 mM HEPES pH 7.0, 100 mM KCl, 5 mM MgCl2, 0.5% NP40 and 1 mM DTT, 1× Protease Inhibitor Cocktail (Roche), 100 units/ml RNase inhibitors) on ice for 5 min to swell, and subsequently aliquoted and frozen at − 80 °C overnight. The lysates were thawed quickly and centrifuged at 14,000 rpm for 10 min at 4 °C and the supernatants were collected and quantified by BCA protein assay using GENios Plus Microplate Reader (Tecan).
For RNA-Protein Immunoprecipitation (RIP), μMACS™ Epitope Tag Protein Isolation Kits (Miltenyi Biotec) were used. Briefly, 2 mg total protein from the above preparations were pre-cleared by incubating with Protein A sepharose (GE Healthcare) and 4 μg normal rabbit IgG (Santa Cruz) in NT2 buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 mM MgCl2 and 0.05% NP40) with 0.1 μg/ml BSA (Sigma) and 0.1 μg/ml yeast tRNA (Ambion) at 4 °C for 1 h on shaker. After centrifuging, the supernatant were incubated with 50 μl Anti-GFP MicroBeads (Miltenyi Biotec), 0.5 mg/ml yeast tRNA and 0.5 mg/ml BSA at 4 °C for 1 h with rotation. Bound protein-RNA complexes were enriched and the bound RNAs were then extracted with elution buffer containing 6 μl 5 M NaCl containing 20 μg proteinase K in 150 μl volume at 42 °C for 1 h, and 65 °C for 1 h more, followed by phenol:chloroform clearance and Na-acetate:lithium chloride precipitation. The mRNAs on the purified RNAs were amplified with TransPlex® Complete Whole Transcriptome Amplification Kit (Sigma) according to the manufacturer’s instruction for sequencing. RNA-sequencing was conducted by Ribo Bio Co., Ltd., China and the enriched mRNAs were confirmed by RT-PCR analysis.
Semi-quantitative and quantitative RT-PCR analyses
Total RNA was isolated from cells or clinical tumor samples with Ultrapure RNA kit (CWBiotech) and the complementary DNA (cDNA) was synthesized by using Transcriptor First Strand cDNA Synthesis System kit (Roche) with Oligo (dT)
18. For semi-quantitative analysis, PCR amplification was performed with GoTaq® DNA polymerase (Promega), 1 μM of each pair of the indicated primers (Additional file
1: Table S1), and 0.5 μl cDNA for a 2 min initial denaturation at 95 °C, followed by 20–27 cycles of 30 s at 95 °C, 30 s at the appropriate anneal time, and 1 min at 72 °C. Products were run in 1.5% agarose gel and the band intensity was scanned, and normalized based on their corresponding internal GAPDH control as the relative expression level. Real time PCR was performed by using a SYBR Green reaction kit (TIANGEN, China) in CFX 96 real time PCR instrument(BioRad,USA), according to the supplier’s protocol. All experiments were carried out in triplicate and analyzed using the comparative threshold cycle (2
−ΔΔCT) method, where the ΔΔCT is the difference between normalized target gene and internal control (ΔCT sample – ΔCT control = ΔΔCT).
Plasmid constructions
To obtaining the 3’-UTR of human p27 mRNA, total RNA was extracted from A2780 cells and cDNA (from 2 μg RNA) was synthesized with SuperScript III and primed with oligo (dT) in accordance with the manufacturer’s instructions (Invitrogen). Primers 5’-GC
TCTAGAACAGCTCGAATTAAGAATATGT-3′ with an
Xba I restriction site and 5’-
GGATCCAATAGCTATGGAAGTTTT-3′ with a
BamH I site were used to amplify the full length of p27 mRNA 3’-UTR. The full length fragment was double digested with
Xba I and
BamH I and subcloned into pGL3-Basic vector (Promega), which is located downstream of the firefly luciferase. The indicated primers (Additional file
1: Table S2) were used to generate deletions of p27 mRNA 3’-UTR (constructs a-m).
To obtain point mutation of human PCBP1, pEGFP-PCBP1 [
6] was used as template to make various point mutations with QuikChange® Site-Directed Mutagenesis Kit (Stratagene), according to the manufacturer’s instruction. Amino acid substitution mutations of G30A, KH2 G114A and G296A in GXXG motif of KH domain were produced to break KH1, KH2 and KH3 of PCBP1, respectively [
27]. Similarly, other indicated mutations were also produced to break the phosphorylation sites of PCBP1. All primers for the above mutations are listed in Additional file
1: Table S3.
Two p27 specific shRNA sequences are 5’-CGCAAGTGGAATTTCGACTTT-3' (shRNA1) and 5’-CCCGGTCAATCATGAAGAACT-3’ (shRNA2) [
28] which were cloned into pGFP-V-RS shRNA vector (OriGene). All above mentioned constructs were verified by DNA sequencing.
Cell cycle analysis
A2780 and DLD-1 cells stably expressed PCBP1 KD, GFP-PCBP1, GFP-PCBP1-p27 KD and GFP were seeded and cultured for 24 h, followed with serum-free starvation or combined with 0.08 μM methotrexate (MTX) to synchronize cell phase for 24 h. Cells were cultured for another 12 h or 16 h, then fixed with 70% ethanol in PBS and stained with 25 μg/ml propidium iodide (Beyotime) containing 50 μg/ml RNAse A for 30 min. The treated cells were analyzed with EPICS XL-MCL flow cytometry (Bechman Coulter), and the data were analyzed with FlowJo 7.6 software. In brief, the control samples were loaded first to optimize the conditions and then set the proper parameters to all samples for flowing and analyzing to generate graphical reports.
Apoptosis analysis
A2780 and DLD-1 cells stably expressed GFP-PCBP1 and GFP were seeded and cultured for 24 h, followed with treatment with 3 μM Paclitaxel for 24-36 h. Cells collected by detaching with 0.25% EDTA-free trypsin and washed with PBS, then stained with APC-conjugated Annexin V and 7-ADD (KeyGEN BioTECH, KGA1017) for 15 min at room temperature in dark. Cells were analyzed by flow cytometry (Beckman, Gallios).
Transient transfection and luciferase assays
A2780 cells were seeded at a density of 1 × 104 cells/96-well plate one day before and then co-transfected with each p27 mRNA 3’-UTR constructs (a-m) and internal control pRL-TK (Renilla) plasmid with X tremeGENE HP DNA transfect reagent (Roche). Luciferase activities were measured using the Dual-Luciferase-reporter assay system (Promega) at 24 h after transfection with Synergy 2 Muti-Mode Reader (Bio Tek). The relative luciferase activity was normalized based on that of the individually corresponding Renila luciferase, which was considered as the internal control. The presented data are from two independent triplicate experiments, shown as mean ± SD.
Cell proliferation assay
Cells were seeded 96-well plates with time-course incubations, and were then washed with phosphate-buffered saline (PBS) with MTT (0.5 mg/mL PBS) and incubated at 37 °C for 30 min. Formazan crystals were dissolved with dimethyl sulfoxide (40 μL/well) and detected at OD570 using an Emax Endpoint ELISA microplate reader (Molecular Devices).
Colony formation assay was performed using double-layer soft agar in 6-well plates with a bottom layer of 0.5% agar and a top layer of 0.35% agarose. In detail, the 0.5% base agar layer was prepared in 6-well plates first, then the assayed cells were harvested, counted and resuspended in 0.7% top agarose solution, followed by pouring the cell mix on top of the base agar layer and incubating for 10–14 days, while feed cells 2 times a week. At the end of the incubation, cells were fixed with 4% paraformaldehyde-PBS for 5 min and stained with 0.1% crystal violet. Colonies were photographed and counted under stereomicroscope in 5 random fields.
Tumorigenicity assays in immunodeficient mice
All mouse experiments were conducted under the Institutional Animal Care and Use Committee (IACUC) approved protocols by Sun Yat-sen University. The immunodeficient mice BALB/c-nu were purchased from Guangdong Medical Laboratory Animal Center. Five or six nude mice at 4-week-old in each group were subcutaneously injected with 5 × 106 A2780 cells or 1 × 107 DLD-1 cells overexpressing PCBP1 with additional p27 knockdown, or cells overexpressing PCBP1 alone with GFP control cells into the left and right sides of their armpit area to examine the tumorigenicity of the experimental cells in vivo. After 3 (A2780 cells) or 4 (DLD-1 cells) weeks, the induced tumor sizes were measured and recorded, then the tumors were dissected for weighing. GraphPad Prism 5 and paired t test were used for statistical analysis.
mRNA stability assay
A2780 cells stably expressed GFP-PCBP1 or GFP control treated with DMSO, 0.5 μg/ml Actinomycin D (Act D) (Biosharp), 20 μM MG132 (Sigma) or combination of Act D with MG132 for 8 h. Total RNA was extracted for cDNA synthesis, and transcript abundances of p27, c-myc mRNAs and controls were detected by semi-quantitative and real time RT-PCR as indicated.
Measuring the distribution of p27 mRNA by ribosome profiling
For polyribosome isolation [
6,
29], cells were incubated with 90 mg/ml cycloheximide (Sigma) for 10 min followed by trypsinization and harvest. Twenty million cells were resuspended in RSB (20 mM Tris-HCl [pH 7.4], 20 mM NaCl, 30 mM MgCl
2, RNasin, 100 mg/ml Heparin, and 5 mg/ml cycloheximide). An equal volume of lysis buffer (1.2% Triton X-100, 1.2% deoxycholate) was added, followed by incubating on ice for 5 min. The nuclei and cell debris were removed by centrifugation for 3 min at 12,000 rpm. The supernatant was then diluted with an equal volume of dilution buffer (25 mM Tris-HCl [pH 7.4], 25 mM NaCl, 25 mM MgCl
2, 0.05% Triton X-100, and 500 mg/ml heparin) and 400 ml of the extract was loaded onto 9.5 ml linear 10 to 50% sucrose gradients and centrifuged at 36,000 rpm for 2 h in a SW41 rotor (Beckman). Ten (1 ml each) fractions were collected with the BioComp piston gradient fractionator linked to an EM-1 UV Monitor (BioRad). The fractions were incubated in 1% SDS and proteinase K at 42 °C for 30 min. RNA was purified by Phenol Chloroform extraction followed by ethanol precipitation. p27 mRNA was detected by semi-quantitative RT-PCR and qRT-PCR as mentioned above.
Immunochemistry (IHC) analyses of PCBP1 and p27 expression in tumor samples
Ovary tissues (18 normal; 21 carcinoma tissues) and 10 pairs of freshly frozen colon tumor tissues as well as 8 pairs of renal tumor samples were collected from Sun Yat-sen University Cancer Center (SYSUCC), under their Standard Experimental Ethics Protocol. Tissues were formalin-fixed, paraffin-embedded and sliced into 6 μm thin sections using Leica BM 2135 microtome, which were subsequently stained with anti-PCBP1 (1:200 dilution, Abcam), anti-p27 (1:200 dilution, ExCell Bio) and Skp2 (1:200 dilution, CST) antibodies. IHC was developed with Polink-1 HRP DAB Detection System (ZSGB-BIO). Each sample was scored semiquantitatively using the immunoreactive score [
30] method that is in consideration of the values of immunoreaction intensity and the percentage of tumor cell staining, as described previously [
31,
32]. IRS calculation are presented as IRS = SI (staining intensity) × PP (percentage of positive cells), in which SI was determined as 0 = negative; 1 = weak; 2 = moderate; and 3 = strong. Likewise, PP was defined as 0, < 1%; 1, 1–10%; 2, 11–50%; 3, 51–80%; and 4, 80% positive cells. The relative score in Fig.
6b based on IRS are determined as -, 0; +, 1–3; ++, 4–8; +++, 9–12.
Statistics and GEO data analyses
SPSS 13.0 and Chi-square analysis were used to analyze the relationship between PCBP1 and p27 expression in IHC staining and the statistical significance was defined as p < 0.05.
To analysis p27 mRNA expression in ovarian epithelia and the ovary cancer epithelia cells, Gene Expression Omnibus (GEO) dataset GDS3592 was used. Likewise, 3 independent GEO datasets of the drug-resistant subclones of breast cancer cells treated with Tamoxifen (GSE26459), Doxorubincin (GSE24460), and Lapatinib (GSE16179) [
33‐
35] were analyzed to validate the relationship of PCBP1 and p27 mRNA expression in drug resistance, which were presented as the mean ± SD and compared statistically by Student’s t test, using Graphpad Prism (GraphPad Software), n indicates ovary sample number or the drug-resistant breast cancer cell subclones. A
p value of < 0.05 was considered statistically significant. All supplementary experimental procedures are availble in Additional file
2.
Discussion
Loss of PCBP1 is involved in various carcinogenesis [
6,
13,
25,
26]. Recently, PCBP1 has been shown as a negative regulator of thyroid carcinoma [
44], a negative regulator of EMT related to metastasis in NSCLC [
45], while PCBP1 expression is suppressed in peritoneal gastric cancer metastasis [
46]. PCBP1 also can sensitize colorectal cancer to anti-cancer drug [
47]. All these highlight the clinical significance of PCBP1 in cancers. In our previous work, we also disclosed that PCBP1 inhibits AKT activation through suppression of metastasis-associated PRL-3 [
6,
11,
12], but the correlation between PCBP1 and PRL-3 expression is not so tight, suggesting its PRL-3-independent ways. Thus, investigation of the alternative PCBP1 function in tumorigenesis would unravel other unknown events which would be of help to tumor diagnosis and therapy. Here we captured the PCBP1-bound mRNAs in transcriptome-wide with RIP-seq strategy. Our results exhibited that multiple mRNAs can be enriched in human A2780 cells (Fig.
1a and Additional file
1: Table S4), some of which have been identified as PCBP1 binding targets, including p21 and eIF4E [
36,
48] that are involved in tumorigenesis. Meanwhile, our results indicate the multiple binding capabilities of PCBP1 and hint the aberrant PCBP1 expression would affect a large spectrum of gene expression, resulting in many physiological abnormities in a specific context, including transformation.
Among the enriched transcripts, p27 mRNA is a novel PCBP1-bound transcript related to cell cycle proliferation, whose abnormal expression is putatively described in many types of cancers. Our results clearly demonstrated that PCBP1 at least binds to the 3’-UTR of p27 mRNA to evidently block the p27 mRNA degradation, leading to higher p27 protein level, although some other indirect possibilities are existed. PCBP1 can translationally repress PRL-3 expression to activate AKT [
6,
11,
12], which further to regulate p27 and its ubiquitin ligase Skp2. However, in a subset of breast carcinomas, it has shown that loss of p27 expression is an independent predictor of both overall survival and disease-free survival, as Skp2 is expressed at low levels despite low expression of p27 [
49]. Our results here may unveil the puzzle that loss of PCBP1 destabilizes p27 mRNA at the first step to lead to low p27 expression, irrespective of Skp2 levels and the posttranslational regulation via other manners (Fig.
5c, d), since Skp2 expression is rarely seen to be upregulated in the detected tumor samples (Fig.
6; Additional file
11: Figure S9C). As a RNA-binding protein, PCBP1 can directly regulate gene expression in posttranscription by binding to 3’-UTR (e.g.
AR, p21, p63, eNOS, POLH) [
7‐
10,
50] or 5’-UTRs (e.g.
c-myc, PRL-3, EV71) [
6,
11,
12] of various mRNAs, although the molecular details of such regulation remain to be structurally determined.
It has been reported that RNA binding proteins, including HuR [
51‐
53], DND1 [
54,
55], RBM38 [
56], CRD-BP [
57,
58] and hnRNP E2 [
59], can compete same binding sites located on their correspondingly targeted genes with microRNAs, subsequently antagonizing the microRNA-targeted gene silencing. DND1 binds 3’-UTR of p27 mRNA to block the binding sequences targeted by p27 specific microRNAs, and maintains p27 protein expression in a germ-cell tumor cell line [
54]. Thus, PCBP1 may work in the same manner to compete to the binding of some microRNAs with p27 mRNA and cause the less mRNA degradation.
Our results revealed that break of PCBP1 KH1 domain apparently mitigated p27 protein expression and the reporter activity, but not the KH2 and KH3 mutations, indicating that PCBP1 binds to p27 mRNA mainly through its KH1 domain, rather than KH2 and KH3 domains, which is consistent to that PCBP1 KH1 domain interacts with stem-loops I and IV of EV71 5’UTR, facilitating viral replication [
12]. Our results also pointed that mutation of PCBP1 phosphorylation sites at T60 and T127 led to lower p27 protein level, which is opposite to the mutation of S43 that can block PCBP1 mRNA-binding ability [
16]. Therefore, our results could also indicate the complicated regulatory roles of PCBP1 phosphorylation in its RNA-binding capability.
p27 is a putative inhibitor of cyclin E-CDK2 [
60‐
62] and binds cyclin E/A:CDK2 complex to inhibit G1-S transition [
63]. Usually, p27 accumulates on cell cycle exit, and is rapidly degraded when cells re-enter the cell cycle from quiescence [
64]. Our results showed that cell cycle was arrested in G1 phase of DLD1 cells, while in S-G2 phase of A2780 cells (Fig.
3a-c). These arrested states can be released by the additional knockdown of p27 in the both cell lines, manifesting that p27 mRNA stabilization by PCBP1 may instinctively maintain p27 protein at high level for longer time to thoroughly check the cell cycle stages and integrity. Our results also illuminate that cell cycle modulation by p27 associated to PCBP1 is independent of the previously recognized G1/S transition checkpoint, as in 2780 cells, PCBP1 clearly block cell cycle at S phase (Fig.
3a-c; Additional file
3: Figure S2). Indeed, it is already known that p27 has function in inhibiting both G1/S and G2/M cell cycle progression [
38,
39]. Therefore, it would be particularly interested in exploring what is the exact downstream event of PCBP1-p27 signaling in maintaining appropriate cell cycle progression.
It is believed that PCBP1 can bind to multiple transcripts, such as POLH (DNA polymerase η) to control its expression via mRNA stability [
50], and POLH loss is responsible for the human cancer-prone syndrome XPV, but there is no in vivo and clinical result to further support this notion. Here, our both in vitro and in vivo results implied that p27 could be likely the most effective downstream target of PCBP1 in inhibition of tumorigenesis, since the cell proliferation and tumorigenesis inhibition imposed by PCBP1 can almost be blocked by the additional knockdown of p27 in the same cells. Clinical patient samples also validated this PCBP1-p27 positive correlation, that is, loss of PCBP1 simultaneously accompanies the downregulation of p27 in tumors, which further confirms that the signaling of PCBP1 through p27 plays critical role in tumorigenesis inhibition. As a binding target of PCBP1 and cell cycle family member, p21 could be also involved in PCBP1-dependent cell cycle inhibition. But based on our results, additionally silencing p27 in PCBP1-overexpressing cells can almost neutralized PCBP1’s effect, indicates the dominant PCBP1-p27 signaling in tumor suppression.
Given that tumors with p27 downregulation have poor prognosis [
65,
66], including colon, ovary and breast cancers, and p27 is rarely inactivated in human cancers, our results can suggest that loss of PCBP1 most likely could be the key initial step for p27 downregulation, leading to eventual tumorigenesis, as we tentatively used various types of tumor cells as well as tumor datasets in this study to draw the similar conclusion, hinting the general role of PCBP1-p27 signaling in tumorigenesis suppression.