P16-specific DNA methylation by engineered zinc finger methyltransferase inactivates gene transcription and promotes cancer metastasis

Background P16 DNA methylation is well known to be the most frequent event in cancer development. It has been reported that genetic inactivation of P16 drives cancer growth and metastasis, however, whether P16 DNA methylation is truly a driver in cancer metastasis remains unknown. Results A P16-specific DNA methyltransferase (P16-dnmt) expression vector is designed using a P16 promoter-specific engineered zinc finger protein fused with the catalytic domain of dnmt3a. P16-dnmt transfection significantly decreases P16 promoter activity, induces complete methylation of P16 CpG islands, and inactivates P16 transcription in the HEK293T cell line. The P16-Dnmt coding fragment is integrated into an expression controllable vector and used to induce P16-specific DNA methylation in GES-1 and BGC823 cell lines. Transwell assays show enhanced migration and invasion of these cancer cells following P16-specific DNA methylation. Such effects are not observed in the P16 mutant A549 cell line. These results are confirmed using an experimental mouse pneumonic metastasis model. Moreover, enforced overexpression of P16 in these cells reverses the migration phenotype. Increased levels of RB phosphorylation and NFκB subunit P65 expression are also seen following P16-specific methylation and might further contribute to cancer metastasis. Conclusion P16 methylation could directly inactivate gene transcription and drive cancer metastasis. Electronic supplementary material The online version of this article (doi:10.1186/s13059-015-0819-6) contains supplementary material, which is available to authorized users.


Background
P16 (CDKN2A or Ink4a) is one of the most frequently deleted genes in cancer genomes and has been studied extensively [1]. P16 germline mutation carriers have been shown to have a greatly increased predisposition to familial melanoma [2][3][4]. Recently, genetic inactivation of P16 has been proven to be a driver for cancer metastasis in mice [5].
It has been reported that artificial P16 DNA methylation induced through the insertion of alu motifs increased the susceptibility of mice to developing cancer [16]. However, whether P16 DNA methylation drives cancer metastasis has not been characterized. In the present study, a P16-specific DNA methyltransferase (P16-dnmt) was used to directly inactivate P16 transcription and the subsequent effects on proliferation, migration, and invasion of cancer cells were evaluated in vitro. These results were further confirmed in immuno-deficient mice. This study provides experimental evidence that strongly implicates P16 DNA methylation as a driver in cancer metastasis.

P16 DNA methylation directly inactivates gene transcription
In order to determine whether P16 DNA methylation directly inactivates gene transcription, a P16 promoterspecific DNA methyltransferase (P16-dnmt) was initially constructed using the pcDNA3.1_myc/His vector as described in the methods section (Fig. 1a). Western blot analysis confirmed that endogenous P16 was greatly reduced in HEK293T cells 48 h following transient transfection with the P16-dnmt vector (Fig. 1b). A dual-luciferase reporter assay further illustrated that P16 promoter activity was significantly inhibited in the P16-dnmt-transfected cells (Fig. 1c). Notably, methylation of CpG islands within both the P16 promoter and exon-1 regions was detected using denatured high performance liquid chromatography (DHPLC) and bisulfite-sequencing ( Fig. 1d and e). An additional control lacking approximately 80 % of the DNA methyltransferase activity (R882H mutant) was constructed to evaluate the impact of steric hindrance from P16-Dnmt DNA binding on gene transcription. As expected, chromatin-immunoprecipitation (ChIP)-PCR analysis showed that the mutant still bound with the P16 promoter DNA fragment (Fig. 2a), but did not induce P16 Fig. 1 P16-specific methyltransferase (P16-dnmt) induces methylation of P16 CpG islands and represses gene transcription in HEK293T cells. a Construction of P16-dnmt using P16 promoter-specific seven-zinc finger protein (7ZFP) and the catalytic domain of mouse DNMT3a; b Western blot analysis for P16-Dnmt and endogenous P16; c Reporter assay results following P16-dnmt transfection; d DHPLC methylation analysis of the P16 promoter in HEK293T cells; the 392-bp methylated (M) and unmethylated (U) P16 fragments were detected at the partial denaturing temperature of 57.0°C; e bisulfite clone-sequencing results from HEK293T cells transiently transfected with P16-dnmt DNA methylation (Fig. 2b). Furthermore, its capacity to repress P16 expression was sharply decreased in both HEK293T and BGC823 cells ( Fig. 2c and d). These data suggest that P16 DNA methylation is directly responsible for P16 repression as opposed to steric hindrance. Taken together, these results indicate that P16-dnmt encodes an active methyltransferase for P16 CpG islands, and P16 DNA methylation is sufficient to inactivate endogenous P16 expression.
In order to specifically methylate P16 CpG islands, the P16-dnmt coding-sequence was then integrated into the pTRIPZ lentivirus vector carrying a 'Tet-on' switch to allow the gene expression to be controlled. Expression of P16-Dnmt protein was induced in GES-1 cells stably transfected with the P16-dnmt pTRIPZ vector after treatment with 0.25 μg/mL doxycycline for 3 days (61KD; Fig. 3a). Significant inhibition of endogenous P16 expression was observed in Western blot and quantitative RT-PCR analysis when compared to GES-1 cells transfected with the dnmt3a and 7ZFP control vectors ( Fig. 3a and b). Confocal microscopy revealed that the average density of nucleic P16 gradually decreased in the P16-dnmt expressing cells (Fig. 3c). In fact, after treatment with doxycycline for 3 and 7 days, P16 expression levels were decreased by 21.4 % and 53.3 %, respectively (P <0.001). Most importantly, intensive methylation of P16 CpG islands was induced in the GES-1 cells stably transfected with P16-dnmt and treated with doxycycline, but not in cells transfected with the control vectors, nor in cells that did not receive doxycycline treatment (Fig. 3d). Similarly, P16 DNA methylation and subsequent repression of P16 expression was also induced by P16-Dnmt in the BGC823 cell line (Additional file 1: Figure S1).
ChIP-PCR analysis also showed that P16-Dnmt specifically bound the P16 promoter, but not the P14 promoter (Additional file 1: Figure S2). Similarly, ChIP-sequencing confirmed that the P16-Dnmt binding fragment was only detected in the promoter of P16-Dnmt/Myc antibody immunoprecipitated DNA from the P16-dnmt-expressing BGC823 cells, but not in the IgG control, nor the cells transfected with the control vector (Fig. 4a, red fragment; Additional file 2: File S1, Additional file 3: File S2, and Additional file 4: File S3). Although most of the P16-Dnmt binding fragments were found in intergenic and intron sequences (Additional file 1: Figure S3A), the main P16-Dnmt binding motif was found to closely match the antisense strand of the targeted fragment in the P16 promoter with a similarity of 21/23 (91.3 %) base pairs (Additional file 1: Figure S3C, red-framed motif ). Genome-wide methylation analysis of P16-Dnmt expressing BGC823 cells was performed using an Infinium Methylation 450 K array. The results illustrated that 647 of 481,615 informative CpG sites Interestingly, 229 of these 647 CpG sites were located in intragenic CpG islands and shores corresponding to 203 genes (Additional file 5: File S4). The targeted P16 CpG island was included in the list of differentially hypermethylated sites (Fig. 4a, blue arrow). Furthermore, DNA methylation was not induced in the CpG islands of two control genes, P14 (located within the same CDKN2A locus as P16) and ZNF382 (located on a different chromosome) (Fig. 4b). These results suggest that doxycycline-induced P16-dnmt expression could specifically methylate P16 CpG islands.

P16-specific DNA methylation promotes migration and invasion of cancer cells
Various assays were then conducted to further characterize the biological behaviors of cancer cells following P16-specific inactivation by DNA methylation. Transwell assays revealed that the migration ability of GES-1 and BGC823 cells was significantly increased following P16-specific DNA methylation ( Fig. 5a and b). Similarly, Matrigel assays showed that the invasion capacity of these cell lines was also significantly enhanced by P16-specific DNA methylation ( Fig. 5c and d).
Four weeks after BGC823 cells stably transfected with P16-dnmt were injected into the tail vein of the NOD SCID mice, metastatic nodules were observed in the lung (Fig. 6a). The average lung weight, which correlates with the number of metastatic cells, in the P16-dnmt group was 152.5 % that of the empty vector control group (Mann-Whitney test, P <0.001; Fig. 6b). The average proportion of metastatic nodule area to total lung area in the P16-dnmt group was also significantly higher than the control group (P <0.004, Fig. 6c).
In addition, P16-specific DNA methylation was found to slightly, but significantly, inhibit proliferation of GES-1 cells, while the proliferation of BGC823 cells was not affected (Additional file 1: Figure S4). However, growth inhibition of the P16-dnmt transfected GES-1 cells was not observed in the SCID mice despite detection of methylated-P16 alleles in the xenografts (Additional file 1: Figure S5).
In order to confirm whether the enhanced migration of cancer cells is P16 DNA methylation-specific, a rescue assay was carried out in the P16-dnmt expressing BGC823 cells through transient transfection of a P16 expression vector. Results of the Transwell assay demonstrated that enforced P16 overexpression significantly reversed the enhanced migration phenotype of these cells (Fig. 7a). Similar results were also observed in HONE-1 cells (Additional file 1: Figure S6). In contrast, downregulation of endogenous P16 expression through transient siRNA transfection significantly enhanced the migration of BGC823 and GES-1 cells (Fig. 7b). Furthermore, the migration capacity of A549 cells, which lack P16 alleles, was not changed following stable transfection of P16-dnmt and 7 days of doxycycline treatment (Fig. 7c). Taken together, these results imply that the enhanced migration and invasion phenotypes of cancer cells are P16-specific. To confirm that P16 DNA methylation affects its downstream signal pathway, P16-CDK4/6-RB, the phosphorylation level of RB protein was analyzed using Western blot analysis. As expected, increased levels of phosphorylated RB were detected in the P16-dnmt transfected BGC823 and GES-1 cells treated with doxycycline when compared to those without doxycycline induction and those transfected with the control vector. Total RB protein levels were not changed ( Fig. 8a and b). Furthermore, the expression level of nuclear factor NFκB subunit P65 was also increased in the P16-dnmt transfected cells.

Discussion
It is well known that methylation of CpG islands around transcription start sites is inversely correlated with the c Migration capacity of A549 cells (lacking P16 alleles) following stable transfection of P16-dnmt. After treatment with 0.25 μg/mL doxycycline for 1 week, these cells (4.0 × 10 4 ) were seeded into each well and incubated for 28 h. Western blot analysis of P16-Dnmt and P16 expression is also presented. Migration assays were independently repeated in triplicate expression level of genes in many cells and tissues. P16 DNA methylation may occur as a long-term mechanism to maintain gene suppression following transcriptional silence induced by repressive histone modifications [17]. Whether P16 promoter methylation alone is capable of silencing transcription has not been well studied. In the present study, we found that P16-Dnmt-induced methylation of P16 CpG islands could directly inactivate gene expression and promote metastasis of cancer cells.
Studies have shown that target-specific methylation/ demethylation is associated with transcriptional inactivation/re-activation of several other human genes [18][19][20]. Zhang et al. reported that P16-specific artificial transcription factor (P16-ATF) could induce P16 DNA demethylation and re-activate its expression [21]; however, it is not known whether DNA demethylation is essential for re-activation of this gene. Yu et al. successfully established a p16 DNA methylation model in mice through the insertion of alu motifs into the mouse p16 promoter and found subsequent gene inactivation [16]; however, the possibility that the insertion of alu motifs directly contributed to transcriptional repression cannot be excluded. In order to increase the targeting specificity in the present study, we employed the pTRIPZ vector that allowed for controllable expression of P16-Dnmt. Our results showed that P16-Dnmt-induced DNA methylation was sufficient to silence transcription in two human gastric epithelial cell lines. This result is consistent with a recent report which demonstrated that engineered transcription activator-like effector (TALE)-Dnmts induced P16 DNA methylation, inactivated gene expression, and increased replication in human fibroblasts [22]. Additionally, steric hindrance from P16-Dnmt DNA binding does not appear to play a significant role in repressing gene transcription as was demonstrated using the P16-dnmt R882H mutant control. Taken together, the evidence suggests that P16 DNA methylation primarily accounts for the inactivation of P16 transcription.
Genome-wide CRISPR screens in mouse models have shown that genetic p16 inactivation may be a driver for tumor growth and metastasis [5]. Luo et al. have reported that the proportion of methylated P16 alleles is significantly associated with metastasis of gastric cancers [13]. Zhang et al. have also suggested that re-activation of methylated P16 by P16-ATF inhibits migration and invasion in AGS and H1299 cancer cell lines [21]. Here, we have provided evidence to demonstrate that P16-Dnmt mediated DNA methylation might promote metastasis of cancer cells in vitro and in vivo. Moreover, we found that such an effect was not observed in A549 cells lacking P16 alleles, and siRNA downregulation of P16 expression also promoted the migration of cell lines, and overexpression of P16 reversed the cell migration phenotype. These facts strongly implicate inactivation of P16 by DNA methylation as a possible promoter of migration/invasion and metastasis of cancer cells.
Inactivation of the P16 gene results in higher cyclin D-dependent protein kinase activity and thus induces aberrant phosphorylation of RB protein. Therefore normal cell cycle checkpoints are bypassed allowing accelerated and GES-1 cell lines were tested 14 and 8 days following doxycycline treatment, respectively. a WB image; b Relative density of tested protein bands normalized against GAPDH cell growth and increased genomic instability [23,24]. We found that the induction of P16-specific DNA methylation could also increase the phosphorylation of RB.
NFκB subunit P65 is the master regulator in the senescence-associated secretary phenotype (SASP) [25]. In melanomas, the expression of P65 is increased while P16 expression is decreased [26]. P65 also regulates the transcription of a group of metastasis-related genes, including MMP-9/2 [27][28][29]. In this study, we also found that induction of P16 DNA methylation also increases the amount of P65 protein in cancer cells. Additional studies are required to determine other pathways involved in the P16 DNA methylation-related metastasis phenotype.

Conclusion
Engineered zinc finger protein-targeted P16 DNA methylation directly inactivates P16 expression and promotes invasion and metastasis of cancer cells.

Cell lines and cultures
HEK293T, BGC823, and GES-1 cell lines were kindly provided by Professor Yang Ke at Peking University Cancer Hospital and Institute. The A549 cell line lacking the P16 locus was kindly provided by Professor Zhiqian Zhang at the same institute. The HONE-1 cell line was kindly provided by Professor Zhen Sun at Capital Medical University School of Stomatology, Beijing. All of these cell lines were tested and authenticated using the Goldeneye20A STR Identifiler PCR Amplification Kit (Beijing Jianlian Genes Technology Co., Ltd.) before being used in this study [30]. These cell lines were cultured in RPMI1640 medium supplemented with 10 % FBS and maintained at 37°C in humidified air with 5 % CO 2 . Cell proliferation was analyzed using the Cell Counting Kit-8 (CCK-8) [21].

Construction of vectors and transfection
The P16-dnmt plasmid was constructed by fusing a SP1-like engineered seven-zinc finger protein (7ZFP) 6I capable of specifically binding the 21-bp fragment (5′-GAG GAA GGA AAC GGG GCG GGG-3′, including a Sp1-binding site) within the human P16 promoter [21] with the catalytic domain (approximately 608-908aa) of mouse dnmt3a in the pFast Bac HT A-dnmt3a vector (kindly provided by Professor Keith Robertson at Georgia Regents University, USA) [31]. Point mutation R882H in the catalytic domain of Dnmt3a is the most frequent somatic mutation in acute myeloid leukemia [32]. The methyltransferase activity of R882H DNMT3A is reduced by approximately 80 % compared with the wide-type [33]. Thus, a P16-dnmt R882H mutant control was constructed as a negative control. The P16-dnmt coding sequence was integrated into a pcDNA3.1 vector and an expression controllable pTRIPZ vector carrying a 'Tet-on' switch (Open Biosystem, USA), respectively. Control vectors for the Dnmt3a catalytic domain or 7ZFP(6I) were also constructed. The purified P16-dnmt plasmid DNA was mixed with VSVG and Δ8.9 (Addgene, USA) to prepare lentivirus transfection particles. The P16 expression vector was constructed using wild-type P16 coding sequence cDNA and integrated into the pIRES2-EGFP vector. The cells (4.5 × 10 4 ) were transiently transfected with the pIRES2-P16 expression vector, seeded into each well, and incubated for 43 h. P16 specific siRNAs (5′-CCGUA AAUGU CCAUU UAUAT T-3′ and 5′-UAUAA AUGGA CAUUU ACGGT T-3′) were synthesized (GenePharma) and used to transiently transfect cells at a final concentration of 1.0 μg/1 mL. The scramble siR-NAs (5′-UUCUC CGAAC GUGUC ACGUT T-3′ and 5′-ACGUG ACACG UUCGG AGAAT T-3′) were used as negative control. The fresh lentivirus particles were used to transfect human cells.
Bisulfite-DHPLC, −sequencing, MethyLight, and methylation-specific PCR (MSP) The 392-bp fragments isolated from the antisense-strand of P16 exon-1 in cultured cells were amplified with a CpG-free primer set and analyzed using DHPLC and clone sequencing as described previously [13,34]; however, the PCR annealing temperature was fixed at 57.0°C to avoid amplification bias between methylated and unmethylated P16 alleles. The 567-bp fragment in the antisense-strand of the P16 promoter was also amplified using a CpG-free primer set (forward, 5′-gaatt agggt ttttg attta gtgaa tt-3′; reverse, 5′-accct atccc tcaaa tcctc taaa-3′) at an annealing temperature of 65°C, analyzed at the partial denaturing temperature of 54°C in DHPLC analysis, and confirmed using clone sequencing. Methylated and unmethylated P16 were also analyzed by 150/151-bp MSP [35].
Quantitative RT-PCR, Western blot, and confocal analysis of P16 expression The P16 mRNA and protein level in cell lines were analyzed as described [21].

Genome-wide analysis of DNA methylation
Illumina Infinium HD Methylation450K arrays were used to perform differential CpG methylation analyses on BGC823 cells stably transfected with the P16-dnmt and pTRIPZ control vectors following 14 days of doxycycline treatment according to the Assay Manual. Two parallel samples were tested for each group. DNA methylation levels for each CpG site were computed as the ratio of normalized methylated signal intensity to the sum of methylated and unmethylated signal intensities using GenomeStudio software. Using the control vector as a reference, Δβ was calculated to represent differential methylation for each CpG site in the P16-dnmt expressing cells. The differential methylation was considered to be significant when the Δβ value was >0.50. The raw methylation data are available as Additional file 6: File S5.

Dual-luciferase reporter assay
The P16 promoter (approximately −597 to +155 nt) was integrated into the pGL3-Basic vector and used for promoter activity analysis as previously described [21].

Transwell migration and Matrigel invasion tests
The Transwell migration and Matrigel invasion tests were performed using GES-1 and BGC823 cells suspended in 150 μL serum-free medium (2 × 10 5 cells/ mL). The BGC823 cells were incubated for 36 h and 96 h at 37°C in 5 % CO 2 before quantifying their migration and invasion capacity, respectively. Similarly, the GES-1 cells were quantified following 48-h and 108-h incubation, respectively [21]. Wound healing status was dynamically recorded using the IncuCyte ZOOM™ livecell imaging platform. Each trial consisted of three independent samples, and all the assays were repeated two to three times.

GES-1 cells (1.4 × 10 6 cells in 200 μL
Matrigel suspension) were stably transfected with the P16-Dnmt or control vector, induced with 0.25 μg/mL doxycycline for 7 days, and then injected subcutaneously into the lower limb of NOD SCID mice (female, 5 weeks old, weight 10-20 g, purchased from Beijing Huafukang Biotech). The mice were provided distilled, sterile water containing 2 μg/mL doxycycline. These mice were sacrificed 48 days after transplantation. The weight and volume of tumors were then analyzed.
For the pneumonic metastasis assay, BGC823 cells stably transfected with the P16-dnmt or control vector were also induced with 0.25 μg/mL doxycycline for 7 days, and then injected into the tail vein of the SCID mice (1.5 × 10 6 cells in 0.15 mL medium) (10 randomly allocated mice per group). The lung weight was detected at the 19th experimental day for each mouse [40]. The lung organs were fixed with Bouin solution, paraffinembedded and cut into 5 μm slides along the maximum area, and examined microscopically following H&E staining. The lung area and total tumor nodule area were measured using INFINITY Analyze (Version 4.0, Lumenera Sci). The nodule area to the lung area ratio was calculated for each mouse.

Statistical analysis
Results were displayed by constituent ratios of enumeration or ranked data. All P values were two-sided, and a difference with P <0.05 was considered statistically significant.

Ethical approval
This study was approved by the institute's animal ethics committee (approval no. AE-2012-06).

Data and materials availability
The methylation array data have been deposited into the Gene Expression Omnibus under accession number GSE74233. The ChIP-sequencing raw data have been deposited into the bioproject database under accession number SRA306603.