The prolyl hydroxylase OGFOD1 promotes cancer cell proliferation by regulating the expression of cell cycle regulators

OGFOD1, a prolyl‐hydroxylase, has been reported to translocate from the nucleus to the cytoplasm in response to cellular stress. Here, we demonstrate that OGFOD1 regulates the transcription and post‐transcriptional stabilization of cell cycle‐related genes. OGFOD1 knockdown in lung cancer cells induced cell cycle arrest through the specific depletion of cyclin‐dependent kinase (CDK) 1, CDK2 and cyclin B1 (CCNB1) mRNAs and the nuclear accumulation of p21Cip1. Analysis of the mRNA dynamics in these cells revealed that CDK1 decreased in a time‐dependent manner, reflecting post‐transcriptional regulation by OGFOD1 and the RNA‐binding protein HuR. In contrast, the depletion of CDK2 and CCNB1 resulted from decreased transcription mediated by OGFOD1. These results indicate that OGFOD1 is required to maintain the function of specific cell cycle regulators during cancer cell proliferation.

Edited by Lukas Alfons Huber OGFOD1, a prolyl-hydroxylase, has been reported to translocate from the nucleus to the cytoplasm in response to cellular stress. Here, we demonstrate that OGFOD1 regulates the transcription and post-transcriptional stabilization of cell cycle-related genes. OGFOD1 knockdown in lung cancer cells induced cell cycle arrest through the specific depletion of cyclin-dependent kinase (CDK) 1, CDK2 and cyclin B1 (CCNB1) mRNAs and the nuclear accumulation of p21 Cip1 . Analysis of the mRNA dynamics in these cells revealed that CDK1 decreased in a time-dependent manner, reflecting posttranscriptional regulation by OGFOD1 and the RNA-binding protein HuR. In contrast, the depletion of CDK2 and CCNB1 resulted from decreased transcription mediated by OGFOD1. These results indicate that OGFOD1 is required to maintain the function of specific cell cycle regulators during cancer cell proliferation.
Keywords: CDK; cyclin; HuR; OGFOD1; post-transcription 2-Oxoglutarate and Fe (II)-dependent oxygenase domain-containing protein 1 (OGFOD1) is a member of the prolyl-hydroxylase family. Based on sequence alignment, the N-terminal catalytic domain of OGFOD1 was predicted to contain a prolyl 4hydroxylase (P4H) domain and a nuclear localization signal [1,2]. OGFOD1 preferentially catalyses trans-C-3 prolyl hydroxylation. It has also been suggested that its yeast homologues, Saccharomyces cerevisiae termination and polyadenylation protein 1 (Tpa1) and Schizosaccharomyces pombe 2-oxoglutarate and Fe (II)-dependent oxygenase domain-containing protein 1 (Ofd1), are catalysts of trans-C-3 and/or C-4 prolyl hydroxylation [3,4]. Both OGFOD1 and Tpa1 hydroxylate the proline residues of the small ribosomal subunit protein RPS23 [5]. These hydroxylases have been shown to participate in protein translation and stress granule formation [5,6]. Furthermore, OGFOD1 has been demonstrated to control the cell cycle in MDA-MB-231 breast cancer cells, and the overexpression of OGFOD1 has been proposed to be a predictive marker for a poor clinical prognosis in patients with breast cancer [7]. Nevertheless, the molecular mechanism(s) whereby OGFOD1 contributes to cancer progression remains largely unknown.
Defects in cell cycle checkpoint genes in cancer cells have been shown to lead to the progression of malignancy. The functions of cyclin-dependent kinase (CDK) inhibitor proteins (CKIs) and tumour suppressor proteins are impaired by mutations in cell cycle checkpoint genes, resulting in the modification and degradation of various proteins, which then leads to the acceleration of the cell cycle in cancer cells [8,9]. CKIs can be categorized into two major families: inhibitors of kinase 4 (INK4) and CDK-interacting proteins/kinase inhibitory proteins (CIP/KIP). INK4 family proteins include p16 INK4a , p15 INK4b , p18 INK4c and p19 INK4d , which target cyclin-dependent kinase 4 and 6 proteins (CDK4 and CDK6), resulting in G1-phase arrest [10]. CIP/KIP family proteins, which include p21 Cip1 , p27 Kip1 and p57 Kip2 , are universal inhibitors of all cyclins (CCNs) and CDK complexes [11]. In addition, the transcription of genes encoding CCNs and CDK subunits varies depending on the cell cycle phase [12,13]. Therefore, cell cycle transitions are governed by both the expression and activity of CCNs, CDK complexes and CKIs.
The mRNA stabilities of transcripts encoding cell cycle regulators are also critically controlled by RNAbinding proteins. Human antigen R (HuR), a member of the embryonic lethal abnormal vision-like family of proteins, is well-known as one such RNA-binding protein. HuR localizes predominantly to the nucleus, and shuttles between the nucleus and cytoplasm, facilitating the translation of stabilized mRNAs in the cytoplasm [14][15][16]. Indeed, HuR is often found in complexes with mRNAs containing AU-rich elements (AREs) in their 3 0 -untranslated regions (3 0 UTRs), enhancing the translation of these transcripts into proteins in cells [17,18]. In several malignant cancers, HuR accumulation has been observed in cancer progression, and it results in increased stability of the mRNAs encoding CCNs and CDKs [19]. Thus, posttranscriptional regulation of CCNs and CDKs by HuR is also involved in the control of the cell cycle and growth of cancer cells.
In this study, we investigated whether the levels of the prolyl-hydroxylase OGFOD1 correlate with the molecular dynamics of cell cycle regulators in lung adenocarcinoma cells. Our results indicated that OGFOD1 contributes to the progression of non-small cell lung adenocarcinoma by accelerating the cell cycle and growth.

Small interfering RNA transfection
Small interfering RNA (siRNA) oligonucleotides corresponding to the OGFOD1 sequences (siOGFOD1) were purchased from Dharmacon/Thermo Fisher Scientific (Lafayette, CO, USA) and Sigma-Aldrich (St. Louis, MO, USA). The ON-TARGETplus non-targeting pool, which was used as control siRNA, was purchased from Dharmacon/Thermo Fisher Scientific. Cells were cultured in 24-well plates and transfected with siRNAs (30 nM) using Lipo-fectAMINE RNAiMax (Invitrogen Life Technologies) according to the manufacturer's instructions. The siRNA sequences used in this study are shown in Table S1.

Cell growth and cell cycle assay
Cells were transfected with siRNAs (30 nM) for 72 h, then the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay for cell growth was performed using Cell Count Reagent SF (Nacalai Tesque, Kyoto, Japan) according to the manufacturer's instructions. The concentrations of formazan products were determined by measuring the absorbance at 490 nm with an iMark microplate reader (Bio-Rad Laboratories, Hercules, CA, USA). For the cell cycle assay, cells were analysed using the Cell-Clock Cell Cycle Assay kit (Biocolor, County Antrim, UK) according to the manufacturer's protocol. After staining, cells were observed using an Olympus IX71 fluorescence microscope (Olympus, Tokyo, Japan). Images were analysed using IMAGEJ software (US National Institutes of Health; http://imagej.nih.gov/) to determine the percentage of cells in each cell cycle phase.

Immunoblotting
Cells were lysed with 2 9 sodium dodecyl sulfate (SDS) sample buffer (250 mM Tris-HCl, pH 6.8, 4% SDS, 20% glycerol, 10% beta-mercaptoethanol and 0.01% bromophenol blue), then the contents of the wells were harvested and sonicated. The resulting lysates were boiled for 2 min, then the proteins were resolved on SDS-polyacrylamide gels, and blotted onto polyvinylidene fluoride membranes. Nonspecific binding sites were blocked with a suspension of 5% skim milk powder in TBST (50 mM Tris-HCl, pH 7.2, 140 mM NaCl and 0.05% Tween-20) for 1 h. The primary antibodies, diluted in 5% skim milk in TBST, were added to the membranes and they were incubated overnight at 4°C. Next, the membranes were washed three times (10 min per wash) with TBST, incubated for 1 h with the HRP-conjugated secondary antibodies diluted in 5% milk in TBST and washed three times (10 min per wash) with TBST. Chemiluminescent signals were detected using ImmunoStar Zeta chemiluminescence reagent (Wako Pure Chemical, Osaka, Japan) and a ChemiDoc TM Touch Imaging System (Bio-Rad Laboratories).

mRNA stability assay
After 48 h of transfection with OGFOD1 siRNA, cells were treated with actinomycin D (0.5 lgÁmL À1 for PC-9 cells and 2.5 lgÁmL À1 for A549 cells) for 0, 3, 5 and 8 h. After 60 min of preincubation with actinomycin D to ensure transcriptional shutoff (time 0), samples were collected and processed using RNeasy Plus Mini kits. Reverse transcription was performed to synthesize cDNAs using ReverTra Ace qPCR RT Master Mix. Diluted cDNA (1 : 10) was subjected to PCR, and the levels of CDK1 were detected using a ChemiDoc TM Touch Imaging System; the intensity of bands was semi-quantified by IMAGE LAB software version 5.2 (Bio-Rad Laboratories).

Ribonucleoprotein complex immunoprecipitation assay
An RNA-protein immunoprecipitation assay using formaldehyde was performed as described previously [21]. PC-9 cells were transfected with siControl or siOGFOD1 for 48 h, then the cells were suspended in phosphatebuffered saline (PBS) containing 1% formaldehyde. Crosslinked RNA-protein complexes were immunoprecipitated using anti-HuR antibody (Santa Cruz Biotechnology, Santa Cruz, CA, USA). Following the immunoprecipitation of HuR and crosslink reversal, RNA was isolated using Trizol reagent (Invitrogen Life Technologies) according to the manufacturer's protocol. cDNA was generated from the isolated RNA using ReverTra Ace qPCR RT Master Mix. CDK1 sequences were amplified from the resulting cDNA by PCR. The PCR products were visualized on a 3% agarose gel containing ethidium bromide. The primer sets used in this study are shown in Table S1.
Dual-luciferase reporter assay PC-9 cells were seeded at a density of 5 9 10 4 cells per 0.5 mL in 24-well dishes. After 24 h of transfection with 500 ng of the luciferase reporter plasmid (psiCHECK-2 vector including Renilla and Firefly luciferases; Promega, Madison, WI, USA) containing CDK1 3 0 UTR cloned downstream of the Renilla luciferase gene, cells were transfected with siControl or siOGFOD1 (30 nM) for 40 h using LipofectAMINE RNAiMax (Invitrogen Life Technologies) according to the manufacturer's instructions. Luciferase activity was measured with the dual-luciferase reporter assay system (Promega) according to the manufacturer's protocol.

Suppression of cell growth in lung cancer cells with OGFOD1 knockdown
To investigate the role of OGFOD1 in non-small cell lung cancer cells, we first confirmed the expression of OGFOD1 and the silencing efficiency of siOGFOD1 in lung cancer cell lines. In all tested lung cancer cell lines (PC-9, H1975, A549 and HCC827), OGFOD1 was observed to localize primarily to the nucleus, as seen by immunofluorescence staining with anti-OGFOD1 antibody (Fig. S1). However, we observed that the OGFOD1 levels in HCC827 were lower than those in the other cell lines (Fig. S1A,B). To evaluate the efficacy of OGFOD1 knockdown in these lung cancer cells, we used three siRNAs designed to target different regions of the OGFOD1 gene. Since siOGFOD1 (#1), (#2) and (#3) all appeared to exhibit a similar knockdown efficiency (Fig. 1A), the results of siOG-FOD1 (#1) are focused upon in the following siRNA experiments. Transfection of lung cancer cells (PC-9, H1975 and A549) with siOGFOD1 resulted in a 50% to 80% decrease in cell growth (Fig. 1B). Furthermore, PC-9 and H1975 cells with OGFOD1 knockdown also exhibited stronger PI staining than the respective control cells transfected with the siControl, indicating increased cell death (Fig. S1C). In A549 cells with OGFOD1 knockdown, although the intensity of PI staining was lower than that observed in the nuclei of PI-stained PC-9 and H1975 cells with OGFOD1 knockdown, retardation of cell growth was still seen, as demonstrated by a reduction in the staining intensity for Ki-67, a marker of cell proliferation (Fig. S1D). In contrast, silencing of the OGFOD1 gene did not affect the cell growth of HCC827 cells ( Fig. 1B and Fig. S1D), which, as mentioned above, expressed OGFOD1 at lower levels than the other cell lines. Results obtained using a cell cycle staining kit that detects the cell cycle phase based on colour changes due to the cellular redox state revealed that there was a larger number of dark blue-coloured cells among the PC-9 and H1975 cells with OGFOD1 knockdown in comparison to the cells with siControl ( Fig. 1C,D), indicating G2/M cell cycle phase arrest.
The accumulation of yellow-coloured cells was seen in the A549 cells with OGFOD1 knockdown, suggesting the induction of G1-phase arrest. In contrast, no colour change due to OGFOD1 knockdown was seen in the HCC827 cells in comparison to the siControltransfected cells, indicating no cell cycle arrest in this particular cell line. Thus, we inferred that OGFOD1 modulates cell growth and cell cycle-related molecules.

Expression profile of cell cycle regulators in lung cancer cells with OGFOD1 knockdown
Based on the above findings, we used RT-PCR to examine whether the expression of tumour suppressor genes was altered in response to OGFOD1 knockdown (Fig. S2A). Notably, the expression of p53, p16 INK4A and p14 ARF mRNAs did not differ significantly between the cells transfected with siOGFOD1 or siControl. However, an increase in p21 Cip1 mRNA in the A549 and HCC827 cell lines and a decrease in the p27 kip1 transcript levels in all cell lines due to OGFOD1 knockdown were observed (Fig. S2A). When expression was assessed at the protein level, no significant changes were observed in the p27 Kip1 , p16 INK4A and p14 ARF tumour suppressor protein levels between the cells transfected with siOGFOD1 or siControl (Fig. S2B). On the other hand, the protein levels of p21 Cip1 were significantly increased in the PC-9, H1975 and A549 cells with OGFOD1 knockdown when compared to the control cells; however, the p21 Cip1 level did not increase in the HCC827 cells with OGFOD1 knockdown ( Fig. 2A and Fig. S2B). As for the p53 protein levels, we found an elevated p53 level in the siOGFOD1 (#1 and #2)-treated A549 and HCC827 cells (Fig. S2B,C). Notably, PC-9, H1975 and A549 cells with OGFOD1 knockdown exhibited increased accumulation of nucleus-localized p21 Cip1 ( Fig. 2A and Fig. S2D), and the fluorescence intensity of the nuclear p21 Cip1 was approximately two-to fourfold higher than that of the cytoplasmic p21 Cip1 in the cells with OGFOD1 knockdown, as assessed by immunofluorescence analysis (Fig. S3). Thus, these results implied that an increase in nuclear p21 Cip1 levels via p53dependent or -independent mechanisms might be involved in the regulation of cell growth arrest in PC-9, H1975 and A549 cells with OGFOD1 knockdown. Nuclear p21 Cip1 is known to target CCN/CDK complexes, and thereby block the cell cycle in cancer cells [22,23]. To confirm whether the expression of cell cycle proteins that interact with p21 Cip1 was changed by OGFOD1 knockdown, the mRNA and protein levels of CCNs and CDKs were examined by RT-PCR and immunoblotting (Fig. 2B,C and Fig. S4A). The mRNA and protein expression levels of CCNA1, CCNE1, CCND1, CDK4 and CDK6 in cells with OGFOD1 knockdown did not differ significantly from those of the cells transfected with siControl (Fig. S4A). However, PC-9, H1975 and A549 cells with OGFOD1 knockdown showed the depletion of both CDK1 mRNA and CDK1 protein, with the protein level falling twofold when compared to that in the control cells (Fig. 2B,C). In all cell lines, depletion of CDK2 and CCNB1 was also observed in the siOGFOD1-transfected cells, with CCNB1 protein levels decreasing approximately 10-fold in siOGFOD1transfected cells when compared to the control cells. The same results were obtained in the siOGFOD1 (#2)-treated cells (Fig. S4B). Thus, OGFOD1 knockdown appeared to cause the accumulation of nuclear p21 Cip1 and the depletion of CDK1, CDK2 and CCNB1 in PC-9, H1975 and A549 cells.

Stability of CDK1 mRNA in siOGFOD1-transfected cells
The translocation of p21 Cip1 from the cytoplasm to the nucleus has been reported to control the signalling of a variety of protein kinases, including Rho familyalpha serine/threonine protein kinase/protein kinase B, protein kinase A, protein kinase C, proviral insertion site in Moloney murine leukaemia virus 1 and glycogen synthase kinase 3 beta [24][25][26][27]. In practice, it is difficult to identify the specific kinases that regulate p21 Cip1 in response to OGFOD1 signalling. Therefore, we next focused on the depletion of CDK1, CDK2 and CCNB1 mRNAs that leads to decreases in the levels of the corresponding proteins. Specifically, we repeated the above RT-PCR analysis in the presence of actinomycin D, an inhibitor of transcription, which enabled the analysis of mRNA stability. We found that the mRNA stabilities of CDK2 and CCNB1 did not differ significantly between the cells transfected with siOG-FOD1 or siControl when cultured in the presence of actinomycin D (Fig. 3A,B). In contrast, the levels of CDK1 mRNA decreased at 5 h after the addition of actinomycin D in the cells with OGFOD1 knockdown; the CDK1 mRNA exhibited a half-life of approximately 7 to 8 h in the PC-9 and A549 cells with OGFOD1 knockdown (Fig. 3A,B). These results suggested that the decrease in the CDK2 and CCNB1 mRNA levels observed in siOGFOD1-transfected cells reflected a decrease in the transcription level, i.e. regulation at the transcription level, while the decrease in the CDK1 mRNA levels resulted from posttranscriptional regulation, e.g. the degradation of transcripts. We also confirmed that the CDK1 mRNA levels tended to decline in siOGFOD1 (#2)-treated cells (Fig. S4C). Furthermore, the interaction between CDK1 mRNA and HuR, a major mRNA stability factor [14][15][16][17][18], was demonstrated using the ribonucleoprotein complex immunoprecipitation assay (Fig. 3C). Under these conditions, the affinity of HuR for CDK1 mRNA in the siOGFOD1-transfected cells was reduced by twofold when compared to that in the siControl-transfected cells (Fig. 3D). Thus, CDK1 mRNA levels appeared to be regulated posttranscriptionally, an effect that was mediated by the loss of interaction between the transcripts and the RNA-binding protein HuR.

Depletion of HuR by OGFOD1 knockdown in lung cancer cells
To gain further insights into the depletion of CDK1 mRNA, we confirmed the localization and expression of HuR in cells with OGFOD1 knockdown following treatment with actinomycin D. In actinomycin D-treated siControl-transfected cells, HuR localized to the cytoplasm and nucleus, as seen by immunofluorescence staining with anti-HuR antibody (Fig. 4A). However, actinomycin D-treated siOGFOD1-transfected cells exhibited decreased levels of HuR in both the cytoplasm and nucleus when compared to the siControl-  transfected cells and the siControl-transfected cells (Fig. S5B). In contrast, the level of HuR protein in HCC827 cells was unaffected by OGFOD1 knockdown.
In the immunocytochemical analyses of PC-9 and A549 cells, forced expression of a mutant OGFOD1 protein harbouring substitutions (His155Ala and Asp157Ala) at sites predicted to form a hydroxylation active site [1,5] resulted in decreased levels of HuR when compared to the cells transfected with a construct-encoding wild-type OGFOD1 (Fig. S6A). Furthermore, the phosphorylation levels of HuR were impaired in the PC-9, H1975 and A549 cells with OGFOD1 knockdown. This reflected the reduction in HuR stability and affinity for RNA binding (Fig. 4B and Fig. S5C). By performing a luciferase reporter assay using the Renilla luciferase gene harbouring the CDK1 3 0 UTR (Fig. 4C,D and Fig. S6B), we confirmed that the luciferase activity of the reporter-CDK1 3 0 UTR constructs, RL-877 (five for the HuRbinding motif), was decreased by the siOGFOD1 when compared to the siControl. On the other hand, the CDK1 3 0 UTR construct, RL-296, which lacks the putative HuR-binding site, showed similar luciferase activity in both the siControl-transfected and siOGFOD1transfected cells (Fig. 4D). These results suggested that while HuR may be produced persistently in cancer cells, the protein may require OGFOD1 for stabilization and affinity for CDK1 mRNA in some lung cancer cell lines.

Discussion
Although the biological role of OGFOD1 remains poorly defined, recent studies have suggested that this protein has diverse functions and that it serves as a critical regulator in translation termination, stress granule formation and histone methylation [1][2][3][4][5][6]. Furthermore, OGFOD1 knockdown using a gene-specific siRNA (siOGFOD1) in breast cancer cells has been reported to impede cell cycle progression, resulting in growth arrest [7]. In the present study, we revealed a novel function for OGFOD1: the protein acts as a stabilizer for multiple cell cycle regulators. Specifically, OGFOD1 appears to facilitate the accumulation of CDK1, CDK2 and CCNB1 mRNAs; for the CDK1 transcript, this effect reflects the post-transcriptional stabilization of the mRNA in the OGFOD1/HuRmediated signalling axis. Based on the shift in the cell cycle phase, effects on cell proliferation and changes in the expression of genes encoding cell cycle regulators observed in the OGFOD1-knockdown experiments (Figs 1 and 2, Figs S1 and S2), we focused on genes encoding specific CDKs and cyclins, namely CDK1, CDK2 and CCNB1, all three of which are critically involved in the regulation of p21 Cip1 as a target molecule. We found that p21 Cip1 localized to the cytoplasm and nucleus in native and siControl-transfected PC-9, H1975 and A549 cells, whereas the level of the cytoplasmic form was significantly decreased and the level of the nuclear form was significantly increased, following OGFOD1 knockdown. Notably, the cellular proliferation of HCC827 cells, which showed lower levels of OGFOD1 protein than the other tested lung adenocarcinoma cell lines (Fig. S1), was not affected by OGFOD1 knockdown. Indeed, HCC827 cells exhibited very low endogenous levels of p21 Cip1 mRNA and protein; thus, proliferation in this cell line may occur in a manner that is essentially p21 Cip1 independent ( Fig. 1 and  Fig. S2). Cytoplasmic p21 Cip1 is degraded by the activated ubiquitin E3 ligase complex, eventually inducing cell cycle progression and the inhibition of apoptosis in cells producing phosphorylated p21 Cip1 [28,29]. Taken together, these results indicated that OGFOD1 may be critically involved in the translocation of p21 Cip1 from the cytoplasm to the nucleus. Furthermore, to verify the effects of the increased p53 protein levels in A549 and HCC827 cells (Fig. S2), we performed rescue experiments again with knockdown of p53 in siOGFOD1-treated cells. The results showed that double knockdown of p53 and OGFOD1 in A549 cells resulted in decreased p21 Cip1 mRNA levels, and restored cell growth, while the cells with single knockdown of OGFOD1 fell into growth arrest. Single knockdown of p53 also triggered a reduction in p21 Cip1 mRNA to some extent in comparison to the siControl. Thus, in A549 cells bearing wild-type p53, the induction of p21 Cip1 occurred in a p53-dependent manner (Fig. S7). The G1-phase arrest and decrease in cells in the G2 phase (Fig. 1D) seemed to be caused by the elevation of p21 Cip1 and p53. Meanwhile, the growth of HCC827 (in-frame deletion mutant p53 and V218 deletion) cells [30] was not affected by p53. In addition, in both the PC-9 (p53-R248Q) [20] and H1975 (p53-R273H) [31] cell lines, double knockdown of mutative p53/OGFOD1 resulted in p21 Cip1 mRNA levels and cell growth similar to those of siOGFOD1treated cells. Of note, a slight increase in the nuclear accumulation of p21 Cip1 protein was observed by immunoblotting with a long exposure time, but the increase did not affect cell growth (Fig. S7). In previous reports, CDK2 was shown to be required not only for centrosome duplication in the S phase but also for compensating for the function of CDK1/CCNA or the CCNB complex. CDK1 and CDK2 double knockdown induced G2-phase arrest [32]. G2/M-phase arrest in siOGFOD1-treated PC-9 and H1975 cells may be caused by the downregulation of CDK1, CDK2 and CCNB1, although the detailed regulation mechanism of p21 Cip1 by mutative p53 remains unclear.
Given that siRNA-mediated knockdown of OGFOD1 accelerated the nuclear accumulation of p21 Cip1 , we next investigated the changes in the expression of functional regulators of p21 Cip1 , namely CDKs (CDK1, CDK2, CDK4 and CDK6) and cyclins (CCNA1, CCNB1, CCND1 and CCNE1). Among these CDKs and cyclins, CDK1, CDK2 and CCNB1 showed selective decreases in the mRNA and protein levels in response to OGFOD1 knockdown. Further analyses revealed that the mRNA levels of CDK2 and CCNB1 were maintained in the cells with OGFOD1 knockdown when they were treated with actinomycin D; this indicated that the decrease in the levels of these two transcripts upon OGFOD1 knockdown reflected decreases in transcription. In contrast, the CDK1 mRNA level gradually fell after the cells with OGFOD1 knockdown were treated with actinomycin D; this suggested that the decrease in the level of this transcript upon OGFOD1 knockdown reflected a posttranscriptional process, e.g. degradation (Fig. 3B). The ribonucleoprotein complex immunoprecipitation assay revealed that the CDK1 mRNA bound to HuR in both the siControl-transfected and siOGFOD1transfected PC-9 cells, and that the siOGFOD1transfected cells exhibited a decrease in the level of HuR-bound CDK1 mRNA (Fig. 3C,D). Furthermore, we found that the levels of both nuclear and cytoplasmic HuR proteins were decreased in response to OGFOD1 knockdown ( Fig. 4A and Fig. S5A) even though the levels of HuR mRNA did not significantly change (Fig. S5B). Previous work has demonstrated that HuR functions to protect (stabilize) the mRNAs encoding several cell cycle regulators, such as p21 Cip1 , CDKs and cyclins, from rapid degradation, an effect that is mediated by the binding of HuR to the AREs in the 3 0 UTR of mRNAs [17][18][19][20]28]. Thus, the relationship between CDK1 mRNA and HuR protein reported in the present study is consistent with the results of previous studies indicating that the depletion of HuR protein triggers the instability of CDK1 mRNA [33,34]. However, previous studies have not examined the role of OGFOD1 in the interaction between CDK1 and HuR.
Although the present study did not address the mechanism whereby HuR preferentially binds to and stabilizes CDK1 mRNA in the presence of OGFOD1, HuR phosphorylation for binding RNA [35][36][37] was impaired in the cells with OGFOD1 knockdown (Fig. 4B). We postulate that this binding/stabilization may reflect the presence of multiple (five in total) AREs (consensus sequences of AUUUA/UUUUU) in the 3 0 UTR region of the CDK1 mRNA, which suggests a high affinity of HuR for the CDK1 transcript [28,29]. Moreover, phosphorylation at multiple sites of HuR, including the hinge region (amino acids 190-244), by various kinases, such as MAPK, Chk2, Cdk1, PKCa and PKCd, has been reported to affect the binding ability of HuR to its target mRNAs [38]. Based on previous reports, HuR phosphorylation at Ser 221 and 318 by PKCd is indispensable for the stabilization of COX-2 mRNA-containing AU-rich elements [39]. HuR phosphorylation (S88, S100 and T118) by Chk2 was also found to prevent ubiquitination-mediated HuR degradation [40]. Thus, the ARE content and cluster sequence or a structural change in HuR is required for mRNA-HuR binding [41]. In addition, the stability of p21 Cip1 mRNAcontaining AREs was also enhanced by the RNAbinding protein RNPC1 [42]. Other RNA-binding proteins, such as TTP, AUF1 and RNPC1, are also involved in the stability of mRNAs [42,43]. In our study, although no interaction of OGFOD1-CDK1 Fig. 5. Scheme of the function of OGFOD1 in controlling cell cycle regulators. Schematic representation summarizing the proposed intracellular dynamics of cell cycle regulators in response to OGFOD1 in non-small cell lung adenocarcinoma. The results obtained from the cells with OGFOD1 knockdown revealed that OGFOD1 affects the stability of CDK1 mRNA by interacting with HuR. CDK2 and CCNB1 are regulated at the transcriptional level, and nuclear p21 Cip1 is observed in the cells with OGFOD1 knockdown. P, phosphorylation; nuc., nuclear; cyt., cytoplasmic; ?, unidentified molecules. mRNA was observed in the RNA-protein immunoprecipitation assay (data not shown), the activity of the luciferase reporter constructs harbouring the AREs of CDK1 3 0 UTR was attenuated by OGFOD1 knockdown (Fig. 4C). Nevertheless, it is unclear why depletion of the HuR protein occurred in siOGFOD1transfected lung cancer cells; clarification of this point will require further detailed investigations of the OGFOD1-HuR interaction and/or the involvement of other RNA-binding proteins [44,45]. We hypothesize that the hydroxylase activity of OGFOD1 is needed to modify HuR, and thereby enhance the stability of the protein. Indeed, we showed that forced expression of an inactive form of OGFOD1 (bearing H155A and D157A substitutions in the predicted catalytic domain) resulted in alterations in the intracellular localization of HuR (Fig. S6A), suggesting that OGFOD1mediated protein modification is required for the intracellular function of HuR. These unclear details will need to be addressed in future investigations.
Here, we summarize the results of the present study, and our model is shown schematically in Fig. 5. In non-small cell lung adenocarcinoma cells, such as PC9, H1975 and A549, OGFOD1 facilitates cellular proliferation by ensuring the maintenance of the CDK1, CDK2 and CCNB1 mRNAs, and cytoplasmic p21 Cip1 protein. OGFOD1 maintains these transcripts by facilitating the transcription of CDK2 and CCNB1, and by stabilizing CDK1 mRNA through an interaction with HuR (Fig. 5, left panel). Furthermore, OGFOD1 may also play a critical role in stabilizing the phosphorylation of HuR, which is required to stabilize a variety of transcripts encoding cell cycle regulators. When OGFOD1 was subjected to siRNA-mediated knockdown, all three of the cell cycle regulators (CDK1, CDK2 and CCNB) were depleted, resulting in prominent nuclear accumulation of p21 Cip1 , especially in A549 cells (Fig. 5, right panel). In future studies, further details on the molecular mechanism of OGFOD1 in the phosphorylation of HuR, p21 cip , and transcriptional factors will need to be clarified. Clearly, OGFOD1 plays diverse roles in cellular functions, roles that appear to extend beyond simply its physiological prolyl-hydroxylase activity. Thus, while many aspects of the functions of OGFOD1 remain unclear, the protein seems to play a critical role as a cell cycle accelerator in cancer progression.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. Cell death in tumour cells with OGFOD1 knockdown. Fig. S2. Expression of tumour suppressors and nuclear p21 Cip in response to OGFOD1. Fig. S3. Cellular localization of p21 Cip1 by OGFOD1 knockdown. Fig. S4. Expression of cell cycle regulators in response to OGFOD1. Fig. S5. Expression of HuR in response to OGFOD1. Fig. S6. Intracellular localization of HuR due to mutant OGFOD1. Fig. S7. Induction of p21 Cip1 in cells with OGFOD1 and p53 knockdown. Data S1. Supplementary methods (plasmid construction, cell death assay, cellular fractionation, etc.) Table S1. Oligonucleotides.