Histone Lysine Methyltransferase SETD8 Promotes Carcinogenesis by Deregulating PCNA Expression

Protein methylation is recently considered an important posttranslational modification and is predominantly found on lysine and arginine residues. Lysine methylation involves the addition of 1 to 3 methyl groups on the amino acid's ϵ-amine group, to form mono-, di-, or tri-methyllysine. Its function is best understood in histones (1). With the exception of Dot1/DOT1L, all histone lysine methyltransferases (HKMT) contain a SET domain of about 130 amino acids, and so far nearly 40 SET domain–containing HKMTs or potential HKMTs have been identified (2). While our knowledge of the physiologic functions of HKMTs is growing, their involvement in human diseases including cancer is still not well understood.

Proliferating cell nuclear antigen (PCNA) is an evolutionally well-conserved protein found in all eukaryotic species from yeast to humans, as well as in archaea. PCNA functions are related to vital cellular processes such as DNA replication, chromatin remodeling, DNA repair, sister chromatid cohesion, and cell-cycle control (3). PCNA was originally reported as an antigen for autoimmune disease in patients with systemic lupus erythematosus, detected only in the proliferating cell populations (4). Thereafter, it was shown that expression levels of PCNA during cell cycle are differential and associated with proliferation and transformation (5, 6). In the following years, a number of experiments have been done to uncover the role of PCNA in DNA replication, and one of the first functions clarified was a sliding clamp for DNA polymerase δ (7, 8). Meanwhile, the progress in the field not only strengthened the importance of PCNA, but also even placed PCNA at the crossroad of many essential pathways. Importantly, PCNA is posttranslationally modified in several ways, which affects its function. So far, it has been reported that PCNA is ubiquitinated, phosphorylated, acetylated, and even SUMOylated (3). One of the well-documented posttranslational modifications of PCNA is ubiquitination. In response to DNA damage, PCNA is monoubiquitinated at the lysine 164 residue by the E2 Ub–conjugated enzyme Rad6 and the E3 Ub ligase Rad18 (Rad6/Rad18 complex; ref. 9). Rad18 not only binds to Rad6 and PCNA, but also to DNA (10). Thus, Rad18 recruits the ubiquitination machinery to the chromatin-bound target, PCNA. In addition to ubiquitination, it is estimated that approximately 6% of chromatin-bound PCNA is subjected to phosphorylation on Tyr 211 (11). It has been considered that phosphorylation of Tyr 211 on PCNA may stabilize chromatin-bound PCNA, as opposed to polyubiquitination. Furthermore, acetylation is another modification detected on PCNA (12), and in yeast, a poly-SUMOylation on PCNA has been described (13). However, functions of lysine methylation on PCNA have never been elucidated.

In this study, we showed that the histone methyltransferase SETD8 methylates Lys 248 on PCNA and regulates functions of PCNA in cancer cells. This is the first report to describe the significance of lysine methylation on PCNA.

MRC-5, CCD-18Co, 5637, SW780, SCaBER, UMUC3, RT4, T24, HT-1376, A549, H2170, HCT116, LoVo, and 293T cells were from American Type Culture Collection in 2001 and 2003 and tested and authenticated by DNA profiling for polymorphic short tandem repeat (STR) markers, except for SW780. The SW780 line was established in 1974 by A. Leibovitz from a grade I transitional cell carcinoma. RERF-LC-AI and SBC5 cells were from Japanese Collection of Research Bioresources (JCRB) in 2001, and tested and authenticated by DNA profiling for polymorphic short tandem repeat (STR) markers. 253J and 253J-BV cells were from Korean Cell Line Bank (KCLB) in 2001, and tested and authenticated by DNA profiling for polymorphic STR markers. EJ28 cells were from Cell Line Service (CLS) in 2003, and tested and authenticated by DNA profiling for polymorphic STR markers. ACC-LC-319 cells were from Aichi Cancer Center in 2003, and tested and authenticated by DNA profiling for single-nucleotide polymorphism, mutation, and deletion analysis.

Bladder tissue samples and RNA preparation were described previously (14–17). Uroplakin is a marker of urothelial differentiation and is preserved in up to 90% of epithelially derived tumors (18). Use of tissues for this study was approved by Cambridgeshire Local Research Ethics Committee (Ref 03/018).

Specific primers for all human GAPDH (glyceraldehyde-3-phosphate dehydrogenase; housekeeping gene), SDH (housekeeping gene), SETD8, and PCNA were designed (primer sequences in Supplementary Table S1). PCR reactions were conducted with the LightCycler 480 System (Roche Applied Science) following the manufacturer's protocol.

siRNA oligonucleotide duplexes were purchased from Sigma-Genosys for targeting the human SETD8 transcript. siEGFP, siFFLuc, and siNegative control (siNC), which is a mixture of 3 different oligonucleotide duplexes, were used as control siRNAs. The siRNA sequences are described in Supplementary Table S2. siRNA duplexes (100 nmol/L final concentration) were transfected into bladder and lung cancer cell lines with Lipofectamine 2000 (Life Technologies) for 72 hours, and cell viability was examined by Cell Counting Kit-8 (Dojindo).

To investigate roles of a HKMT in human carcinogenesis, we had examined expression levels of several HKMTs in a small subset of clinical bladder cancer samples and found a significant difference in expression levels of SETD8 between normal and cancer cells (data not shown). We then analyzed 124 bladder cancer samples and 28 normal control samples and confirmed the significant elevation of SETD8 expression in tumor cells compared with normal cells (Supplementary Table S4). Expression levels partly correlated with the grade of malignancy in bladder cancer (Supplementary Fig. S1A). We also found overexpression of SETD8 in both non–small cell lung carcinoma (NSCLC) and small cell lung carcinoma (SCLC; Fig. 1A). Subsequent immunohistochemical analysis using anti-SETD8 antibody identified strong SETD8 staining mainly in the nuclei of malignant cells, but no staining in nonneoplastic tissues (Fig. 1B). In addition, our expression profiling analysis indicated the upregulation of SETD8 in chronic myelogenous leukemia, hepatocellular carcinoma, and pancreatic cancer (Supplementary Fig. S2 and Table S5). Furthermore, a high level of SETD8 was identified in various cancer cell lines than in a normal lung cell line SAEC (Supplementary Fig. S3).

To investigate the role of SETD8 in the growth of cancer cells, we conducted a knockdown experiment using 2 independent siRNAs against SETD8 (siSETD8#1 and #2) and 2 control siRNAs (siEGFP and siFFLuc). We transfected each of these siRNAs into SW780 bladder cancer cells and found that SETD8 expression was efficiently suppressed by either of the 2 different siRNAs targeting SETD8, compared with control siRNAs (Supplementary Fig. S1B). Using the same siRNAs, we conducted cell growth assays and found significant growth-suppressive effects on 1 bladder cell line (SW780) and 2 lung cancer cell lines (RERF-LC-AI and SBC5), whereas no effect was observed when we used control siRNAs (Fig. 1C). Detailed cell-cycle analysis using flow cytometry indicated that the cell populations of cancer cells lacking SETD8 had a significant increase in the amount of S-phase and sub-G₁ phase cells and a concomitant reduction in the proportion of G₁ cells (Fig. 1D). Furthermore, we showed that in bromodeoxyuridine (BrdUrd) incorporation analysis, the amount of newly incorporated BrdUrd in cancer cells was significantly decreased after treatment with siSETD8 (Fig. 1E), implying that knockdown of SETD8 results in the retardation of DNA replication in cancer cells. These results indicated that SETD8 might play an important role in the regulation of cancer cell growth, especially in S-phase, and knockdown of SETD8 would cause apoptosis of cancer cells.

As PCNA is known to be a key regulator of cell-cycle progression and SETD8 is a component of the PCNA complex (19, 20), we examined the functional relationship between SETD8 and PCNA. Immunoprecipitation assay showed that 3xFLAG-tagged SETD8 bound endogenous PCNA (Fig. 2A). We also confirmed the interaction between endogenous PCNA and SETD8 proteins (Fig. 2B); endogenous SETD8 and PCNA proteins were colocalized in HeLa cells (Fig. 2C). Immunoprecipitation using deletion mutants of SETD8 showed that its N-terminal region of SETD8 is essential for binding to PCNA (Fig. 2D), and this portion contains a PCNA-interacting protein (PIP) box (Supplementary Fig. S4A). Because histone methyltransferases have been found to methylate nonhistone substrates, we evaluated a possibility of PCNA to be a substrate of SETD8. First, we conducted an in vitro methyltransferase assay and confirmed that PCNA was methylated in a dose-dependent manner (Fig. 2E). The amino acid analysis detected a single lysine methylation site in PCNA following this reaction (Supplementary Fig. S5). To verify in vivo SETD8-dependent PCNA methylation, we labeled 293T cells after transfection with FLAG-PCNA (WT) and hemagglutinin (HA)-mock or HA-SETD8 (1-352) expression vectors with l-[methyl-³H] methionine and found that SETD8 could methylate PCNA in vivo (Supplementary Fig. S6A). Subsequent liquid chromatography/tandem mass spectrometry (LC/MS-MS) analysis identified monomethylation at lysine 248 on PCNA by SETD8 (Fig. 2F). To validate this result, we constructed the plasmid (PCNA-K248A) that was designed to substitute lysine 248 of PCNA protein to alanine and conducted in vitro methyltransferase assay (Supplementary Fig. S6B). The intensity of the band corresponding to PCNA methylation in PCNA-K248A was significantly diminished compared with that of the wild-type PCNA (PCNA-WT). These data show that lysine 248, which is highly conserved in the PCNA ortholog from green alga to human (Supplementary Fig. S4B), is the primary target of SETD8-dependent methylation (Fig. 2G). On the basis of this result, we generated an antibody against a methylated K248 synthetic peptide (Supplementary Fig. S7A) that showed high affinity and high specificity by ELISA (Supplementary Fig. S7B). Western blot analysis using this antibody confirmed that it specifically recognizes K248-methylated PCNA (Fig. 2H and Supplementary Fig. S7C and S7D), and this specific signal was dependent on the methyltransferase activity of SETD8 (Fig. 2I). Importantly, the methyltransferase activity of N-terminal–deleted SETD8 protein, which lacks the PIP box domain, was significantly low than that of wild-type SETD8 protein (Fig. 2J). This result indicates that the N-terminal region of SETD8 containing PIP box domain seems to be important for SETD8-dependent PCNA methylation. This antibody was used to examine the methylation status of PCNA in vivo after treatment with siSETD8 (Supplementary Fig. S8). Monomethylation of PCNA at lysine 248 diminished after knockdown of SETD8 in SBC5 cells, implying SETD8-dependent PCNA K248 methylation occurs both in vitro and in vivo.

To clarify the physiologic significance of PCNA methylation by SETD8, we examined protein expression levels of PCNA in SW780 cells 48 hours after knockdown of SETD8 using 2 independent siRNAs (Fig. 3A). Knockdown of SETD8 decreased PCNA protein, suggesting involvement of SETD8 in regulating PCNA stability in cancer cells. To further validate this result, we examined the cell-cycle dependency of SETD8 and PCNA protein expression levels after aphidicolin synchronization (Fig. 3B). Intriguingly, when we treated with SETD8 siRNAs, PCNA protein expression decreased in both G₁ and S-phases according to the levels of SETD8, indicating that SETD8 is likely to be a key regulator of PCNA protein expression at G₁ and S-phases. Because quantitative real-time PCR analysis implied that PCNA mRNA level was not affected by treatment with siSETD8 (Fig. 3B), the regulation of PCNA expression by SETD8 was not at the transcriptional level but at the protein level. To examine that this regulation is mediated by SETD8-dependent methylation, we examined PCNA (WT) or PCNA (K248A) protein expression levels in 293T cells transfected with mock or SETD8 expression vectors after cycloheximide treatment. Although wild-type PCNA was significantly stabilized by SETD8 expression, methylation-inactive mutant PCNA (PCNA-K248A) was unstable (Fig. 3C). Furthermore, we examined the PCNA stability in endogenous level after depletion of SETD8 and found that the degradation rate of PCNA in cells treated with siSETD8 more rapidly than siEGFP (Supplementary Fig. S9). Taken together, SETD8-dependent methylation is crucial for PCNA stabilization. Then, we validated the effect of SETD8-dependent methylation on ubiquitination of PCNA proteins. The PCNA (WT) or PCNA (K248A) expression vector was cotransfected into 293T cells with a vector expressing either the full-length or N-terminal region of SETD8, and ubiquitination and methylation status of PCNA was examined (Fig. 3D). As we expected, the status of ubiquitination and methylation on PCNA showed the inverse correlation. Hence, we consider that methylation of PCNA inhibited its ubiquitination. We also examined the phosphorylation status of Tyr 211 on PCNA, which is known to influence the stability of PCNA (11), but no significant relationship between methylation and phosphorylation status was observed (data not shown). These data show that PCNA protein is stabilized through inhibition of the ubiquitination by its SETD8-dependent methylation.

We conducted immunoprecipitation analysis to further investigate the significance of PCNA methylation, using wild-type and methylation-inactive mutant PCNA proteins, and identified a partner protein, FEN1, which interacted with PCNA in a methylation-dependent manner. Methylation of PCNA significantly enhanced the interaction between PCNA and FEN1 (Fig. 4A). To validate the effect of PCNA methylation on the interaction with FEN1 in more detail, we conducted an in vitro binding assay using methylated PCNA and unmethylated PCNA with FEN1 recombinant protein. SETD8-dependent lysine methylation of PCNA significantly enhanced the interaction between PCNA and FEN1 in vitro (Supplementary Fig. S10). FEN1 is a structure-specific nuclease with both 5′ flap endonuclease and 5′-3′ exonuclease activities (21). During DNA replication, this enzyme is responsible for RNA primer removal during Okazaki fragment processing and was identified as the factor responsible for the completion of replication in vitro (22). Yeast cells lacking the FEN1 gene (also called RAD27) are viable but are unable to grow at high temperatures, indicating defective DNA replication (23). To examine the effect of PCNA methylation on FEN1 function, we measured levels of the phosphorylated form of H2AX histone variant (γH2AX), an early marker of the cellular response to DNA breaks. In the absence of any exogenous source of DNA damage, basal levels of phosphorylated γH2AX in 293T cells expressing methylation-inactive mutant PCNA (PCNA-K248A) were higher than those in 293T cells expressing wild-type PCNA (Fig. 4B). This implies the accumulation of DNA double-strand breaks resulting from methylation-inactive mutant PCNA expression. During the S-phase of the cell cycle, FEN1 is recruited to DNA replication loci through the interaction with PCNA. Disruption of the FEN1–PCNA interaction impairs such localization (24). If methylation of PCNA were important for interacting with FEN1, failure in methylation would lead to a defect in FEN1′s localization to replication foci. FEN1 could be colocalized with PCNA at replication foci in cells when PCNA was a wild-type. However, FEN1 was unable to localize to the foci in cells in which methylation-inactive mutant was present (Fig. 4C). These data suggest that PCNA methylation is important for regulation of FEN1′s subnuclear localization. Because phosphorylation of FEN1 has been shown to abolish its PCNA interaction (25), we examined FEN1 phosphorylation status in cells expressing wild-type and methylation-inactive mutant PCNA, but found no significant difference in phosphorylation status of FEN1 between wild-type and methylation-inactive mutant PCNA–expressing cells (Fig. 4D). This implies that the different affinity between PCNA and FEN1 seems to be regulated not by phosphorylation status of FEN1 but by methylation status of PCNA (as shown in Fig. 4A and C). Furthermore, an Okazaki fragment maturation assay was conducted using the deoxynucleotide triphosphate mixture containing radiolabeled dCTP and a model substrate containing an RNA-DNA flap, which mimicked the Okazaki fragment maturation intermediate. The assay simulates the sequential reactions of gap filling, RNA primer removal, and DNA ligation during Okazaki fragment maturation. When the assay was conducted in vitro, nuclear extracts from PCNA-K248A–expressing cells showed significant decrease in removing RNA primer flaps and some extent of defect in DNA ligation (Fig. 4E), indicating that the methylation defect of PCNA retarded Okazaki fragment maturation. Defects in the Okazaki fragment maturation process during DNA replication or defects in ligation during DNA repair could lead to accumulation of DNA double-strand breaks (26, 27). To examine the levels of double-strand breaks, 293 cells expressing wild-type and methylation-inactive mutant PCNA were treated with H₂O₂ to determine the survival rate (Supplementary Fig. S11). Consistent with previous data, methylation-inactive mutant PCNA–expressing cells were more sensitive to H₂O₂.

We then compared the methylation of endogenous PCNA in normal and cancer cell lines. PCNA was significantly methylated in various types of cancer cell lines, whereas no detectable level of PCNA methylation was found in normal cell lines (Fig. 5A). We subsequently conducted the immunopathologic analysis on clinical lung tissues, analyzing the correlation between SETD8 and PCNA protein expression levels (Fig. 5B). Clinical information and staining patterns of clinical tissues are described in Table 1 and Supplementary Table S6. We found a correlation factor (ρ) of 0.55378 with P value of 2.068 × 10⁻⁶ (by Spearman correlation coefficient) in a cohort of 64 cases (Fig. 5C); lung adenocarcinoma showed a stronger correlation (ρ = 0.79596, P = 3.857 × 10⁻⁴), supporting our hypothesis that SETD8 overexpression stabilizes and increases the PCNA protein expression in cancer cells. Finally, we examined the effect of PCNA methylation on the growth of cancer cells (Fig. 5D and E). Methylation-inactive–type PCNA–expressing HeLa cells (HeLa-PCNA-K248A) showed a slower growth rate than those with wild-type PCNA–expressing HeLa cells (HeLa-PCNA-WT). Furthermore, to exclude the effect of endogenous PCNA proteins, we first knocked down PCNA gene expression, and then, conducted a clonogenecity assay of HeLa cells overexpressing wild-type PCNA and methylation-inactive-type PCNA (Supplementary Fig. S12). Consistent with our previous data, wild-type PCNA showed higher growth promoting effects than methylation-inactive type PCNA. Taken together, these results imply that methylation of PCNA is likely to play a crucial role in the growth promotion of cancer cells.

Histone lysine methylation plays a central epigenetic role in the organization of chromatin domains and the regulation of gene expression. We previously reported that the HKMT SMYD3 stimulates cell proliferation through its methyltransferase activity and plays a crucial role in human carcinogenesis (28, 29). Of the various posttranslational protein modifications, the role of protein methylation in signal transduction has not been well characterized. While the carboxyl group and arginine methylation have been implicated in several cellular responses, including receptor signaling, protein transport, and transcription (30), lysine methylation has been considered to be histone specific (31). In the present study, we found that the HKMT SETD8 is overexpressed in various types of cancer and regulates PCNA functions through the methylation of lysine 248. This is a new mechanism revealing the importance of lysine methylation in nonhistone proteins in human cancer.

PCNA was originally reported to be a DNA-sliding clamp for replicative DNA polymerases and is an essential component of the eukaryotic chromosomal DNA replisome (32, 33). It interacts with multiple partners including proteins involved in Okazaki fragment processing, DNA repair, DNA synthesis, DNA methylation, chromatin remodeling, and cell-cycle regulation (34). PCNA has been reported to be modified by ubiquitination, SUMOylation, phosphorylation, and acetylation (9, 11, 12, 35, 36) but its lysine methylation has never been. These kinds of protein modifications are vital for a wide variety of PCNA functions. As reported here, PCNA protein is stably overexpressed in various types of cancer cells, together with SETD8 protein, indicating that SETD8-dependent methylation of PCNA enhances its biologic activity. Knockdown of SETD8 significantly suppressed the growth of cancer cells by diminishing PCNA methylation and reduction of its protein levels. It has been recently reported that knockdown of SETD8 leads to several aberrant phenotypes, including DNA damage, S-phase arrest, and global chromosome condensation (20, 37, 38), consistent with our findings, which suggest that these abnormalities is likely to be caused by dysfunction of the PCNA protein.

PCNA is also considered to be the crucial factor in maintaining the balance between survival and cell death. For instance, PCNA displays an apoptotic activity through interaction with proteins belonging to the Gadd45 family (Gadd45, Myd118, and CR6), which was involved in growth control, apoptosis, and DNA repair (39, 40). Lack of SETD8 induces an increase of the sub-G₁ population of cancer cells (Fig. 1D), so it is possible that apoptosis may be induced by SETD8 depletion through dysfunction of PCNA. Furthermore, we clarified that methylation of PCNA is critical for the interaction with FEN1. It has been reported that FEN1 forms distinct protein complexes for DNA replication and repair. Through its interaction with PCNA, FEN1 is recruited to the replication foci for RNA primer removal and to repair sites for DNA base excision repair (41). Recently, the FEN1–PCNA interaction has been implicated in coordinating the sequential action of polymerase δ (Pol δ), FEN1, and DNA ligase 1 (Lig1) during Okazaki fragment maturation (24). Disruption of PCNA–FEN1 interaction impairs Okazaki fragment ligation (24). We showed that methylation-defective PCNA retards both Okazaki fragment maturation and DNA replication, and induces DNA damages. Cells expressing methylation-inactive mutant PCNA were more sensitive to DNA damage. Because deregulation of FEN1 nuclease has also been reported to be linked to human cancer (42), it is possible that abnormal interactions between FEN1 and PCNA may cause human carcionogenesis. Intriguingly, Guo and colleagues recently showed that methylation of FEN1 suppresses nearby phosphorylation and facilitates PCNA binding (43). Together with our result, this implicates methylation as the crucial player in the interaction between PCNA and FEN1 proteins.

In conclusion, as expression levels of SETD8 in normal tissues are significantly low (Supplementary Fig. S13), an inhibitor targeting its enzymatic activity might be an effective drug for cancer therapy. Further functional analysis will explore the SETD8-dependent PCNA methylation pathway as a therapeutic target in various types of cancer.

Y. Yamane, Y. Iwai, and K. Maejima are employed as Researchers in OncoTherapy Science, Inc. R. Hamamoto is a scientific advisor in OncoTherapy Science, Inc. The other authors disclosed no potential conflicts of interest.

Conception and design: M. Takawa, H.-S. Cho, S. Hayami, G. Toyokawa, M. Sugiyama, Y. Nakamura, R. Hamamoto

Development of methodology: M. Takawa, H.-S. Cho, S. Hayami, K. Ueda, N. Dohmae, R. Hamamoto

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Takawa, M. Kogure, K. Ueda, A. Masuda, N. Dohmae, B.A.J. Ponder, R. Hamamoto

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Takawa, H.-S. Cho, M. Kogure, Y. Yamane, Y. Iwai, K. Maejima, K. Ueda, A. Masuda, N. Dohmae, T. Tsunoda, T. Akasu, M. Sugiyama, R. Hamamoto

Writing, review, and/or revision of the manuscript: M. Takawa, H.-S. Cho, K. Ueda, H.I. Field, T. Akasu, S. Ohnuma, Y. Nakamura, R. Hamamoto

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Takawa, H.-S. Cho, Y. Yamane, Y. Iwai, K. Maejima, M. Sugiyama, R. Hamamoto

Study supervision: M. Takawa, M. Sugiyama, Y. Atomi, S. Ohnuma, B.A.J. Ponder, R. Hamamoto

The authors thank Kazuyuki Hayashi, Noriko Ikawa, and Haruka Sawada for technical assistance.

This work was supported by Grant-in Aid for Young Scientists (A; 22681030) from the Japan Society for the Promotion of Science.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.