DNA methyltransferase 1 (DNMT1) promotes cyst growth and epigenetic age acceleration in autosomal dominant polycystic kidney disease

Abstract

Alteration of DNA methylation leads to diverse diseases, and the dynamic changes of DNA methylation (DNAm) on sets of CpG dinucleotides in mammalian genomes are termed “DNAm age” and “epigenetic clocks” that can predict chronological age. However, whether and how dysregulation of DNA methylation promotes cyst progression and epigenetic age acceleration in autosomal dominant polycystic kidney disease (ADPKD) remains elusive. Here, we show that DNA methyltransferase 1 (DNMT1) is upregulated in cystic kidney epithelial cells and tissues and that knockout of Dnmt1 and targeting DNMT1 with hydralazine, a safe demethylating agent, delays cyst growth in Pkd1 mutant kidneys and extends life span of Pkd1 conditional knockout mice. With methyl-CpG binding domain (MBD) protein-enriched genome sequencing (MBD-seq), DNMT1 chromatin immunoprecipitation (ChIP)-sequencing and RNA-sequencing analysis, we identified two novel DNMT1 targets, PTPRM and PTPN22 (members of the protein tyrosine phosphatase family). PTPRM and PTPN22 function as mediators of DNMT1 and the phosphorylation and activation of PKD associated signaling pathways, including ERK, mTOR and STAT3. With whole genome bisulfide sequencing in kidneys of patients with ADPKD versus normal individuals, we found that the methylation of epigenetic clock associated genes was dysregulated, supporting that epigenetic age is accelerated in the kidneys of patients with ADPKD. Furthermore, five epigenetic clock associated genes, including Hsd17b14, Itpkb, Mbnl1, Rassf5 and Plk2 were identified. Thus, the diverse biological roles of these five genes suggest that their methylation status may not only predict epigenetic age acceleration but also contribute to disease progression in ADPKD.

Graphical Abstract

Introduction

Autosomal dominant polycystic kidney disease (ADPKD), one of the most common inherited kidney diseases, is characterized by progressive renal cyst formation and kidney enlargement ultimately leading to end stage kidney disease (ESKD).¹ Symptoms of ADPKD vary in severity and age of onset, but usually develop between the ages of 40 and 50, and tend to get worse over time, eventually resulting in kidney failure. The role of genetic mutations in PKD1 or PKD2 in ADPKD development is well understood.^{2, 3} However, ADPKD cannot be fully understood in terms only of the genetic setting, especially for example, in families with the same genetic mutations but variable disease severity. Epigenetic mechanisms play an important role in human diseases, aging, and aging-related progressive diseases, acting as critical drivers of cell fate and survival by targeting many different pathways in a variety of ways ^{4, 5}, even in genetically identical humans ^{6, 7}. As such, epigenetic mechanisms may be an additional means of explaining individual differences in PKD patient severity, and are a focus of investigation in ADPKD.^{8, 9}

DNA methylation, the first epigenetic modification to be identified for more than half a century¹⁰, plays an important role in lots of biological processes, including repression of gene expression, embryonic development, genomic imprinting, inflammation, and stem cell differentiation^{11, 12}. DNA methylation is the post-replicative addition of methyl groups to the C5 position of cytosines catalyzed by DNA methyltransferases (DNMTs).¹³ In vertebrates, there are five known DNMTs which differ in structure and function. The DNMT3 subfamily, comprised of DNMT3a, DNMT3b and DNMT3L1, is active during DNA replication and DNA repair.^{13, 14} However, only DNMT1 functions to maintain DNA methylation patterns that are established by the other DNMTs. Abnormal DNA methylation has been associated with disease related gene expression.¹⁵ In general, altered DNA methylation patterns in tumor tissues may silence tumor suppressor genes and activate oncogenes through hyper/hypo methylation.^{16, 17} In addition, recent studies have reported that the dynamic changes of DNA methylation (DNAm) on sets of CpG dinucleotides in mammalian genomes, named as “DNAm age” and “epigenetic clocks”, is able to predict chronological age and severity of age-related diseases.¹⁸ ADPKD patients develop worsening clinical symptoms as they age ¹⁹, suggesting that ageing process seems to be one of the factors to promote cyst progression in ADPKD patients and that epigenetic changes associated with the aging process may contribute to disease progression. It is necessary to establish a mechanistic connection of Pkd1 mutations with abnormal DNA methylation and epigenetic age acceleration in ADPKD. In this study, we investigate the roles and mechanism of abnormal DNA methylation and its regulator, DNMT1, in promoting cyst growth and epigenetic age acceleration in ADPKD with methyl-CpG-binding domain sequencing (MBD-seq), ChIP-sequencing (ChIP-seq), RNA-seq and whole genome bisulfite sequencing (WGBS).

Methods

Cell culture and reagents.

PH2 and PN24 cells (provided by S. Somlo through the George M. O’Brien Kidney Center, Yale University, New Haven, Connecticut, USA) were cultured as described previously.²⁰ 5-azacytidine, hydralazine, and S3I-201 were purchased from Sigma-Aldrich.

Chromatin Immunoprecipitation Sequencing (ChIP-seq) and Methyl-binding domain sequencing (MBD-seq).

ChIP-seq was performed using an anti-DNMT1 antibody. Each sample with 10 ng of DNA generated from the ChIP assay was used to construct the sequencing libraries with the ChIP-Seq Sample Prep Kit (Illumina). MBD-seq was performed according to the protocol as described previously.²¹ Briefly, purified genomic DNA from PH2 and PN24 cells were fragmented to an approximate median fragment size of 125 bp by using ultrasonication (Covaris, Woburn, MA). The MethylMiner DNA enrichment kit (Invitrogen, Carlsbad, CA) was used, which employs the methyl-binding domain 2 (MBD2) protein to capture fragments with one or multiple methylated CpGs. The methylated DNA fragment was captured, eluted, and quantified for library construction and next generation sequencing.

The libraries of ChIP-seq and MBD-seq were sequenced using an Illumina HiSeq2500 Sequencing System at the Genome Sequencing Facility at KUMC and Mayo Clinic. The resulting tag sequence csfasta files were aligned to the mouse genome using the Bowtie mapping software. The sequencing data sets have been deposited in the NCBI’s Gene Expression Omnibus (GSE257535).

Mouse strains and treatments.

Pkd1^flox/flox mice (B6; 129S4-Pkd1^tm2Ggg/J; stock 010671; Jackson Laboratory) possess loxP sites on either side of exons 2–4 of Pkd1. Pkhd1-Cre mice Pkd1^flox/flox:Pkhd1-Cre mice express Cre recombinase under the control of the Pkhd1 promoter.²² Dnmt1^flox/flox mice with loxP sites flanking exons 4 and 5 of the Dnmt1 gene were described previously.^{20, 23} The Pkd1^flox/flox:Dnmt1^flox/flox:Pkdh1-Cre mice were generated by crossing female and male Pkd1^flox/+:Dnmt1^flox/+:Pkhd1-Cre mice. The kidneys and blood were harvested from PN25 Pkd1^flox/flox:Dnmt1^+/+:Pkdh1-Cre and Pkd1^flox/flox:Dnmt1^flox/flox:Pkdh1-Cre mice. Pkd1^flox/flox:Pkhd1-Cre mice were treated with 5-azacytidine (5 mg/kg, dissolved in normal saline) or normal saline (control) daily from P8 to P24 by intraperitoneal injection (IP). Serum and kidneys were harvested and analyzed at P25. Hypomorphic Pkd1^nl/nl mice ²⁴ were injected i.p. with hydralazine (10 mg/kg, dissolved in normal saline) or normal saline (control) daily from P7 to P27. Serum and kidneys were harvested and analyzed at P28. Pkd1^flox/flox:tamoxifen-Cre-ERT mice were injected i.p. with tamoxifen (125 mg/kg body weight, formulated in corn oil) on 2 sequential postnatal days (P31 and P32) to induce Pkd1 deletion. Pkd1^flox/flox:tamoxifen-Cre-ERT mice were treated with hydralazine (10 mg/kg, daily) or vehicle control from 3.5 month to 6 months of age. The blood samples and kidneys are harvested from 6-month-old hydralazine- and vehicle-treated mice for histopathological and biochemical analysis. Equal numbers of male and female animals were used for all experiments.

Other methods can be found in supplemental materials.

Results

DNMT1 is upregulated in Pkd1 mutant renal epithelial cells and tissues.

DNMT1 is the only DNA methyltransferase that functions in the maintenance of DNA methylation patterns in cells.^{13, 14, 25} We found that DNMT1 mRNA and protein expression were increased in postnatal (PN) Pkd1 homozygous mutant PN24 cells compared to postnatal Pkd1 heterozygous PH2 cells (Fig. 1a, b). The expression of DNMT1 was also increased in cystic kidneys from postnatal day 25 Pkd1 conditional knockout Pkd1^flox/flox:Pkhd1-Cre mice versus age-matched wild type kidneys as examined by western blot (Fig. 1c) and immunohistochemistry (Fig. 1d). In addition, DNMT1 protein was increased in cyst-lining epithelial cells in the kidneys from 6-month-old Pkd1^flox/flox:tamoxifen-Cre-ERT mice (Fig. S1a) and human ADPKD kidneys (Fig. S1b) compared to that in age-matched wild type kidneys and normal human kidneys, as analyzed by immunohistochemistry.

Figure 1. The expression of DNMT1 was increased in Pkd1 mutant renal epithelial cells and tissues.

(a) qRT-PCR analysis of relative DNMT1 mRNA in Pkd1-heterozygous PH2 cells and Pkd1-homozygous PN24 cells. p < 0.01. (b) Western blot analysis of DNMT1 expression from whole cell lysates in PH2 cells and PN24 cells. (c) Western blot analysis of DNMT1 expression in kidneys from PN25 Pkd1^+/+:Pkhd1-Cre (WT) and Pkd1^flox/flox:Pkhd1-Cre (Flox) mice. Relative DNMT1 expression in the kidneys (bottom panel) as standardized to actin. p < 0.05. (d) Immunohistochemistry analysis indicated that DNMT1 expression was increased in cyst-lining epithelia in kidneys from PN25 Pkd1^flox/flox:Pkhd1-Cre versus Pkd1^+/+:Pkhd1-Cre mice. Scale bars: 50 μm.

Double-conditional knockout of Dnmt1 and Pkd1 delays cyst formation.

In order to explore the in vivo function of DNMT1 in a Pkd1 knockout mice, we generated Pkd1 and Dnmt1 double-conditional knockout Pkd1^flox/flox:Dnmt1^flox/flox:Pkhd1-Cre mice. We found that cyst growth was significantly delayed in the absence of DNMT1 in Pkd1^flox/flox:Dnmt1^flox/flox:Pkhd1-Cre mice (n = 14) at postnatal day 25 compared to that in age matched Pkd1^flox/flox:Dnmt1^+/+:Pkhd1-Cre mice (n = 11) (Fig. 2a b). Cystic index (Fig. 2c), kidney weight (Fig. S2a) and kidney weight to body weight (KW/BW) ratios (Fig. 2d) from Pkd1^flox/flox:Dnmt1^flox/flox:Pkhd1-Cre mice were also dramatically reduced compared to those from Pkd1^flox/flox:Dnmt1^+/+:Pkhd1-Cre mice. Knockout of Dnmt1 did not affect the body weight (Fig. S2a, b). Meanwhile, the expression of DNMT1 could not be detected in cyst lining epithelial cells in kidneys from Pkd1^flox/flox:Dnmt1^flox/flox:Pkhd1-Cre mice as analyzed by immunohistochemistry (Fig. S2c). Additionally, the blood urea nitrogen (BUN) levels were significantly reduced in Pkd1^flox/flox:Dnmt1^flox/flox:Pkhd1-Cre mice versus Pkd1^flox/flox:Dnmt1^+/+:Pkhd1-Cre mice (Fig. 2e), which indicated that renal function was preserved in Pkd1^flox/flox:Dnmt1^flox/flox:Pkhd1-Cre mice. Importantly, Pkd1^flox/flox:Dnmt1^flox/flox:Pkhd1-Cre mice (n = 12) lived to a mean age of 121 days, while Pkd1^flox/flox:Dnmt1^+/+:Pkhd1-cre mice (n = 12) died of PKD at a mean age of 47 days (mean age 121 d versus 47 d; p < 0.01) (Fig. 2f). Furthermore, we found that the proliferation of cyst lining epithelial cells was significantly decreased as analyzed by proliferating cell nuclear antigen (PCNA) staining (Fig. 2g), and cyst-lining epithelial cell apoptosis was increased as analyzed by TUNEL assay (Fig. 2h) in Pkd1^flox/flox:Dnmt1^flox/flox:Pkhd1-Cre mice versus Pkd1^flox/flox:Dnmt1^+/+:Pkhd1-cre mice. Our in vivo studies suggest that DNMT1 plays an important role in cyst development through regulating cyst-lining epithelial cell proliferation and apoptosis in Pkd1 knockout mice.

Figure 2. Double conditional knockout of Dnmt1 and Pkd1 delayed cyst growth.

(a and b) Representative kidneys (a) and H&E staining of kidneys (b) from PN25 Pkd1^flox/flox:Dnmt1^+/+:Pkhd1-Cre (Dnmt1^+/+) and Pkd1^flox/flox:Dnmt1^flox/flox:Pkhd1-Cre (Dnmt1^fl/fl) mice. Scale bar: 2 mm. (c-e) Cystic index (c), KW/BW ratios (d), and BUN levels (e) were significantly decreased in kidneys of PN25 Pkd1flox/flox:Dnmt1flox/flox:Pkhd1-Cre (Dnmt1fl/fl) versus Pkd1flox/flox:Dnmt1+/+:Pkhd1-Cre (Dnmt1+/+) mice. p < 0.01. (f) Pkd1^flox/flox:Dnmt1^flox/flox:Pkhd1-Cre mice lived to a mean age of 121 days (n = 12), whereas Pkd1^flox/flox:Dnmt1^+/+:Pkhd1-Cre mice died of PKD at a mean age of 47 days (n = 12). p < 0.01. (g) Cell proliferation was decreased in kidneys from PN25 Pkd1^flox/flox:Dnmt1^flox/flox:Pkhd1-Cre mice versus Pkd1^flox/flox:Dnmt1^+/+:Pkhd1-Cre mice, as detected with PCNA staining. Scale bars: 50 μm. p < 0.01. (h) Knockout of Dnmt1 induced cyst-lining epithelial cell death in kidneys from PN25 Pkd1^flox/flox:Dnmt1^flox/flox:Pkhd1-Cre mice, while apoptosis was rare in kidneys from Pkd1^flox/flox:Dnmt1^+/+:Pkhd1-Cre mice, as detected by TUNEL assay. Scale bars: 50 μm. p < 0.01.

Treatment with the demethylating agents 5-azacytidine or hydralazine delays cyst growth in Pkd1 mutant mouse models.

To test if DNMT1 is a potential therapeutic target, we first treated Pkd1^flox/flox:Pkhd1-Cre mice with the demethylating agent 5-azacytidine (5 mg/kg) or vehicle control by daily intraperitoneal (IP) injection from PN8 to PN24 and sacrificed the mice at PN25. The dosage, route and time frames of 5-azacytidine administration in Pkd1 mutant mice were based on previous publication.²⁶ Treatment with 5-azacytidine decreased cystic index (Fig. 3a, b), KW/BW ratios (Fig. 3c) and BUN levels (Fig. 3d) in the Pkd1^flox/flox:Pkhd1-Cre mice (n = 10). Furthermore 5-azacytidine treatment decreased cyst lining epithelial cell proliferation as analyzed by Ki67 staining (Fig. 3e), and increased cyst lining epithelial cell apoptosis as analyzed by the TUNEL assay (Fig. 3f) in Pkd1^flox/flox:Pkhd1-Cre mice.

Figure 3. Treatment with 5-azacytidine delayed cyst growth in Pkd1 conditional knockout mice.

(a) Histological examination of kidneys from PN25 Pkd1^flox/flox:Pkhd1-Cre mice injected daily with 5-azacytidine or vehicle control from P8 to P24. Scale bar: 2 mm. (b) Percent cystic area relative to total kidney area of kidneys from Pkd1^flox/flox:Pkhd1-Cre treated with 5-azacytidine (n = 10) or vehicle control (n = 10). p < 0.01. (c and d) Treatment with 5-azacytidine compared with vehicle control decreased KW/BW ratios (c) and BUN levels (d) in Pkd1^flox/flox:Pkhd1-Cre mice. p < 0.01. (e) 5-azacytidine treatment reduced cyst-lining epithelial cell proliferation in kidneys from Pkd1^flox/flox:Pkhd1-Cre, as detected with Ki67 staining. Scale bars: 50 μm. p < 0.01. (f) 5-azacytidine treatment induced cyst-lining epithelial cell death in kidneys from Pkd1^flox/flox:Pkhd1-Cre mice as detected by TUNEL assay. Scale bars: 50 μm. p < 0.01.

Disease progression of ADPKD from cyst initiation to ESKD is a slow process which lasts for decades. The considerable cytotoxicity of 5-azacytidine, which as a nucleoside analogue can be incorporated into DNA and RNA, limited its application in diseases other than myelodysplastic syndrome and acute myeloid leukemia as approved by the FDA.²⁷ Therefore, a demethylating agent that targets DNA methyltransferases without being incorporated into DNA should provide safe and therapeutic benefits to ADPKD patients during a long-term treatment. For this, hydralazine as a DNA hypomethylating agent in cancer and chronic kidney disease (CKD) may be a potential candidate drug.²⁸ Thus, we further tested the short-term and long-term therapeutic effects of hydralazine in Pkd1 hypomorphic Pkd1^nl/nl (n = 10) and Pkd1^flox/flox:Tamoxifen-Cre-ERT (n = 10) mice, respectively. As renal cysts could be detected as early as PN7 and the cystic kidneys reached maximal size at PN28 in Pkd1^nl/nl mice ⁸, we treated Pkd1^nl/nl mice with hydralazine (10 mg/kg/day) from PN6 to PN27 by daily intraperitoneal injection and collected kidneys at PN28. Treatment with hydralazine significantly delayed cyst growth as showed by the decrease of cystic index, KW/BW ratios and BUN levels in Pkd1^nl/nl mice (Fig. 4, a–d). Treatment with hydralazine also decreased cell proliferation (Fig. 4e) but increased apoptosis (Fig. 4f) of cyst lining epithelia in Pkd1^nl/nl mice.

Figure 4. Treatment with hydralazine delayed cyst growth in Pkd1 hypomorphic Pkd1^nl/nl mice.

(a) Histological examination of kidneys from PN28 Pkd1^nl/nl mice injected daily with hydralazine or vehicle control from P7 to P27. Scale bar: 2 mm. (b) Percent cystic area relative to total kidney area of kidneys from Pkd1^nl/nl mice treated with hydralazine (n = 10) or vehicle control (n = 10). p < 0.01. (c and d) Treatment with hydralazine compared with vehicle control decreased KW/BW ratios (c) and BUN levels (d) in Pkd1^nl/nl mice. p < 0.01. (e) Hydralazine treatment reduced cyst-lining epithelial cell proliferation in kidneys from Pkd1^nl/nl mice, as detected with Ki67 staining. Scale bars: 50 μm. p < 0.05. (f) Hydralazine treatment induced cyst-lining epithelial cell death in kidneys from Pkd1^nl/nl mice as detected by TUNEL assay. Scale bars: 50 μm. p < 0.01.

Next, to test the effect of hydralazine in Pkd1^flox/flox:Tamoxifen-Cre-ERT mice, in which the Cre is driven by the Esr1+ promoter ²⁹, we induced Pkd1 deletion by intraperitoneal injection of tamoxifen on PN31 and PN32 days. Renal cysts can be detected in these mice at 4 months of age.³⁰ We treated these mice with hydralazine (10 mg/kg/day) or vehicle control from 3.5 months to 6 months. Hydralazine treatment delayed cyst growth as showed by the decrease of cystic index, KW/BW ratios, BUN levels, and the proliferation of cyst-lining epithelial cells; and the increase of cyst-lining epithelial cell apoptosis (Fig. 5, a–f) compared to those in kidneys of vehicle treated control mice. Treatment with hydralazine had no effects on the body weight in Pkd1^nl/nl mice, and slightly increased the body weight of Pkd1^flox/flox:Tamoxifen-Cre-ERT mice (Fig. S3a, b ). These results suggest that targeting DNMT1 with pharmacological inhibitor hydralazine may delay cyst growth in ADPKD patients.

Figure 5. Treatment with hydralazine delayed cyst growth in Pkd1^flox/flox:Tamoxifen-Cre-ERT mice.

(a) Histological examination of kidneys from 6 month Pkd1^flox/flox:Tamoxifen-Cre-ERT mice injected daily with hydralazine or vehicle control from 3.5 months to 6 months. Scale bar: 2 mm. (b) Percent cystic area relative to total kidney area of kidneys from Pkd1^flox/flox:Tamoxifen-Cre-ERT mice treated with hydralazine (n = 10) or vehicle control (n = 10). p < 0.01. (c and d) Treatment with hydralazine compared with vehicle control decreased KW/BW ratios (c) and BUN levels (d) in Pkd1^flox/flox:Tamoxifen-Cre-ERT mice. p < 0.05 or p < 0.01. (e) Hydralazine treatment reduced cyst-lining epithelial cell proliferation in kidneys from Pkd1^flox/flox:Tamoxifen-Cre-ERT, as detected with PCNA staining. Scale bars: 50 μm. p < 0.01. (f) Hydralazine treatment induced cyst-lining epithelial cell death in kidneys from Pkd1^flox/flox:Tamoxifen-Cre-ERT mice as detected by TUNEL assay. Scale bars: 50 μm. p < 0.01.

Hypermethylated genes identified by MBD-sequencing analysis and DNMT1 target genes identified by ChIP-sequencing in cystic renal epithelial cells.

To identify the changes in DNA methylation of specific genes in cystic renal epithelial cells, we performed MBD-seq to identify the genome-wide set of hypermethylated genes in Pkd1 mutant PN24 cells.^{21, 31} We identified 180192 and 154974 methylated CpG sites in PH2 and PN24 cells (Tables S1 and S2), corresponding to 21390 and 20797 genes, respectively (Tables S3 and S4). As one gene may be associated with more than one CpG site, the total numbers of identified CpGs were more than the identified gene numbers. With the comparison of the gene lists generated in PH2 and PN24 cells, we identified 2603 genes with hypermethylated CpGs specifically in PN24 cells (Table S5). To gain initial insight into the cellular pathways associated with the hypermethylated genes in Pkd1 mutant cells that may influence ADPKD development, we used the gProfiler analysis suite which revealed a number of significantly overrepresented Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways.³² We found that the identified hypermethylated genes are involved in signaling pathways associated with PKD, such as MAPK signaling, PI3K-Akt signaling, and Jak-STAT signaling etc., and some signaling pathways that have not been studied in PKD (Table S6).

To identify genes that are directly targeted by DNMT1, we performed ChIP-seq analysis with a DNMT1 antibody. We identified 1215 genes specifically targeted by DNMT1 in PN24 cells (Table S7). The distribution of DNMT1 binding regions was determined with the cis Regulatory Element Annotation System (CEAS) analysis (Fig. S4). The identified DNMT1 target genes are involved in signaling pathways that are associated with PKD or are not studied in PKD by KEGG pathway analysis (Table S8). We should point out that we also identified DNMT1 target genes in PH2 cells with ChIP-seq analysis (Table S9).

Protein tyrosine phosphatases, PTPRM and PTPN22, are novel DNMT1 target genes in cystic renal epithelial cells.

A total 1554 downregulated genes (more than two-fold) was identified in PN24 cells versus PH2 cells by RNA-seq analysis. We identified 65 genes within the intersection of the three gene lists that were identified by MBD-seq, ChIP-seq, and RNA-seq (downregulated genes in PN24 cells), which are potential DNMT1 targets (Fig. S5 and Table S10). Two important protein tyrosine phosphatases (PTPs), protein tyrosine phosphatase, receptor type, M (PTPRM) and protein tyrosine phosphatase, non-receptor type 22 (PTPN22) are among those 65 genes, which are key negative regulators of tyrosine phosphorylation by removing phosphate groups from intracellular proteins. Given that multiple PKD associated signaling pathways, such as STAT3, MAPK, and mTOR signaling, are activated as seen by the increase of their phosphorylation ^33–35, therefore, we focused on those two novel DNMT1 targets, PTPRM and PTPN22. We identified the binding sites and peaks of MBD and DNMT1 on the genes of PTPRM and PTPN22 in Pkd1 heterozygous PH2 and Pkd1 homozygous PN24 cells (Fig. S6). The binding of MBD on the genes of PTPRM and PTPN22 was increased in PN24 cells versus PH2 cells (Fig. S6a), and the binding sites and peaks of DNMT1 on the genes of PTPRM and PTPN22 was only identified in PN24 cells but not in PH2 cells (Fig. S6b). We confirmed that PTPRM and PTPN22 were indeed DNMT1 target genes with the following evidences: 1) the levels of PTPRM and PTPN22 mRNA and protein were decreased in PN24 cells compared to PH2 cells (Fig. 6a, b); and 2) DNMT1 and MBD2 bound to the promoters of PTPRM and PTPN22 in PN24 cells as examined by ChIP assay (Fig. 6c, d), indicating the methylation of these promoters in cystic renal epithelial cells. With the MSP analysis, we confirmed that the methylated /unmethylated ratios on the promoters of PTPRM and PTPN22 were increased up to 2.5-fold in PN21 Pkd1 homozygous kidneys compared to age matched Pkd1 heterozygous kidney (Fig. 6e). Additionally, deletion of DNMT1 decreased methylated/unmethylated ratios on the promoters of PTPRM and PTPN22 in kidneys from Pkd1^flox/flox:Dnmt1^flox/flox:Pkhd1-Cre mice versus Pkd1^flox/flox:Dnmt1^+/+:Pkhd1-Cre mice (Fig. 6e). Furthermore, inhibition of DNMT1 with 5-azacytidine and hydralazine, and knockdown of DNMT1 with siRNA increased the levels of PTPRM and PTPN22 mRNA and protein (Fig. 6, f–k), supporting that DNMT1 mediated the methylation on the promoters of PTPRM and PTPN22 should repress their transcription.

Figure 6. PTPRM and PTPN22 were novel DNMT1 targets in renal epithelial cells.

(a) The mRNA expression of PTPRM and PTPN22 in PH2 and PN24 cells by qRT-PCR. p < 0.01. (b) Western blot analysis of expression levels of PTPRM and PTPN22 in PH2 and PN24 cells. (c and d) The binding of DNMT1 (c) and MBD2 (d) on PTPRM and PTPN22 promoters as analyzed by CHIP-PCR. (e) The methylated (M) and unmethylated (UM) status of PTPRM and PTPN22 promoters in kidneys from postnatal day 21 Pkd1 heterozygous Pkd1^flox/+:Pkhd1-Cre mice (Pkd1-het), Pkd1 homozygous Pkd1^flox/flox:Dnmt1^+/+:Pkhd1-Cre mice (Pkd1-homo), and Pkd1/Dnmt1 double knock out Pkd1^flox/flox:Dnmt1^+/+:Pkhd1-Cre mice (Pkd1/Dnmt1-DKO) were analyzed by MSP (left). The M/UM ratios of PTPRM and PTPN22 promoters in kidneys from three mice in each group were quantified by band density (right). n = 3. p < 0.05. (f-i) 5-azacytidine (f and g) and hydralazine (h and i) treatment increased the mRNA (f and h) and protein expression (g and i) of PTPRM and PTPN22 in PN24 cells. p < 0.05 or p < 0.01. NS, not significant. (j-k) Knockdown of DNMT1 with siRNA increased the mRNA (j) and protein expression (k) of PTPRM and PTPN22 in PN24 cells. p < 0.01.

DNMT1 regulates PKD-associated pathways by targeting PTPRM and PTPN22.

As members of the PTP family, downregulation of PTPRM and PTPN22 might result in increased phosphorylation and activation of PKD associated pathways. To support this, we found that knockdown of PTPRM and PTPN22 with siRNA increased the phosphorylation of STAT3, ERK and S6 in PH2 cells (Fig. 7a). The co-immunoprecipitation analysis confirmed that PTPRM and PTPN22 directly interact with STAT3, ERK and S6 in renal epithelial cells (Fig. 7b, c).

Figure 7. PTPRM and PTPN22 regulated the phosphorylation of STAT3, ERK and S6 in renal epithelial cells.

(a) Western blot analysis of expression levels of p-STAT3, STAT3, p-ERK, ERK, p-S6 and S6 in PH2 cells transfected with control siRNA, PTPRM siRNA, or PTPN22 siRNA. (b and c) The association of PTPRM with STAT3, ERK or S6 as analyzed by co-IP in PH2 cells, and the interaction of PTPN22 with STAT3, ERK or S6 as analyzed by co-IP in PH2 cells.

Upregulated DNMT1 increases cystic epithelial cell proliferation by regulating PKD-associated signaling pathways.

Deletion of DNMT1 decreased the phosphorylation of STAT3, S6 and ERK in kidneys from Pkd1^flox/flox:Dnmt1^flox/flox:Pkhd1-Cre mice versus Pkd1^flox/flox:Dnmt1^+/+:Pkhd1-Cre mice (Fig. 8a). c-Myc functions as a transcription factor to regulate cell cycle progression, and Cyclin D1 drives cells through the G1/S transition during the cell cycle.³⁶ The expression levels of c-Myc and Cyclin D1 were elevated in cystic kidneys compared to wild type kidneys, whereas the levels of c-Myc and Cyclin D1 were decreased in Pkd1^flox/flox:Dnmt1^flox/flox:Pkhd1-Cre mice versus Pkd1^flox/flox:Dnmt1^+/+:Pkhd1-Cre mice (Fig. 8a). Knockdown of DNMT1 with siRNA, and inhibition of DNMT1 with 5-azacytidine and hydralazine decreased the phosphorylation of STAT3, ERK, and S6, and the expression of c-Myc and Cyclin D1 in PN24 cells (Fig. 8, b–d). In addition, treatment with 5-azacytidine only inhibited the cell growth of Pkd1 homozygous PN24 cells, but not that of Pkd1 heterozygous PH2 cells with a lower expression of DNMT1 (Fig. S7). These results indicated that the upregulated DNMT1 should increase cystic cell proliferation through these PKD associated pathways in vivo.

Figure 8. DNMT1 regulated PKD associated pathways in Pkd1 mutant cells and kidneys.

(a) Western blot analysis of the phosphorylation of ERK, S6 and STAT3 as well as the expression of c-Myc and cyclin D1 in PN25 kidneys from wild type, Pkd1^flox/flox:Dnmt1^+/+:Pkhd1-Cre and Pkd1^flox/flox:Dnmt1^flox/flox:Pkhd1-Cre mice. (b) Western blot analysis of the phosphorylation of ERK, S6 and STAT3 as well as the expression of c-Myc and cyclin D1 in PN24 cells transfected with DNMT1 or control siRNA. (c and d) Western blot analysis of the phosphorylation of ERK, S6 and STAT3 as well as the expression of c-Myc and cyclin D1 in PN24 cells treated with different concentrations of 5-azacytidine (c) or hydralazine (d).

DNMT1 regulates cystic epithelial cell apoptosis through p53.

For that deletion and targeting of DNMT1 induced cystic renal epithelial cell death in kidneys from Pkd1^flox/flox:Pkhd1-Cre mice (Fig. 2h and Fig. 3f), we examined how DNMT1 regulates cystic renal epithelial cell death. We found that 1) inhibition of DNMT1 with 5-azacytidine induced apoptosis of Pkd1 mutant PN24 cells, but not control PH2 cells (Fig. 9a); 2) treatment with 5-azacytidine or knockdown of DNMT1 increased the expression of p53 mRNA (Fig. 9b); 3) 5-azacytidine treatment increased the levels of p53 protein and its downstream pro-apoptotic protein Bax, as well as the apoptotic marker, cleaved PARP (Fig. 9c); 4) inhibition of DNMT1 with hydralazine increased the expression of p53 mRNA and protein (Fig. S8a, b); 5) DNMT1 bound to the p53 promoter in PN24 cells as analyzed with ChIP assay (Fig. 9d); and 6) knockdown of p53 with siRNA prevented 5-azacytidine induced cell death of PN24 cells (Fig. 9e). Together, these results suggest that DNMT1 regulates cystic renal epithelial cell apoptosis via downregulation of p53.

Figure 9. DNMT1 regulated Pkd1 mutant cell apoptosis through targeting p53.

(a) TUNEL assay in PH2 and PN24 cells treated with 5-azacytidine. (b) The expression of p53 mRNA in PN24 cells treated with 5-azacytidine and vehicle (control) and treated with DNMT1 siRNA and control siRNA. p < 0.05. (c) Western blot analysis of the expression of p53, Bax and Cleaved PARP in PN24 cells treated with 5-azacytidine and vehicle (control). (d) DNMT1 bound to the p53 promoter as examined by ChIP-PCR. (e) TUNEL assay of PN24 cells treated with 5-azacytidine, or 5-azacytidine plus p53 siRNA.

PKD1 mutation results in the alteration of genome wide DNA methylation in human genome, resulting in epigenetic age acceleration.

The modulation of DNA methylation of specific genes can either accelerate or decelerate the aging process. A number of ‘epigenetic clocks’ based on the degree of DNA methylation (DNAm) of specific CpG sites have been developed, and evidence also shows an accelerated age in specific tissues due to diseases.^{37, 38} The first DNAm age clock is generated based on DNAm at 353 CpG sites, which is highly predictive of chronological age.³⁹ We found that those 353 CpG sites were associated with 309 genes in the human genome (Tables S11).

We performed WGBS analysis in kidneys collected from 5 PKD1 mutant ADPKD patients and 5 non-ADPKD individuals and identified about 169,684 differentially methylated regions (DMRs) in the genome of ADPKD patients versus non-ADPKD individuals. With the alignment of these DMRs with epigenetic clock associated genes, the methylation of 217 out of the 309 human epigenetic age associated genes were dysregulated in promoter, transcription start site (TSS), transcription termination site (TTS), exons and introns (Table S12). In particular, the changes of methylation occurred at the TSS sites of 80 out of 217 human epigenetic age associated genes (Table S13). These results indicates that the epigenetic age is accelerated in ADPKD kidneys.

Furthermore, with the comparison of those 309 epigenetic clock associated genes with the genes identified by MBD-seq and ChIP-seq analysis, 5 epigenetic clock associated genes are both bound by DNMT1 and MBD in PN24 cells but not in PH2 cells (control), including hydroxysteroid (17-beta) dehydrogenase 14 (Hsd17b14), inositol trisphosphate 3 kinase (Itpkb), muscle blind like splicing factor 1 (Mbnl1), Ras associated domain family member 5 (Rassf5), and polo like kinase 2 (Plk2) (Fig. S9 and S10). The methylation of the promoters of these five genes were gradually changed along disease progression in Pkd1 mutant mouse kidneys as examined by MSP assays (Fig. S11). Additionally, a total of 343 genes (Table S14), including PTPRM and PTPN22 (Fig. S12), were identified to be hypermethylated in both mouse and human ADPKD by comparing the hypermethylated genes in PN24 cells identified by MBD-seq with those in human ADPKD kidneys identified by MGBS analysis.

Discussion

In this study, we investigated the roles of DNA methylation and DNMT1 in the regulation of disease progression and epigenetic age acceleration in ADPKD. We found that upregulation of DNMT1 did not result in the hypermethylation of whole genome, and instead, it only resulted in hypermethylation of specific genes as identified in Pkd1 mutant renal epithelial cells with MBD-seq and DNMT1 ChIP-seq analysis. We identify two novel DNMT1 target genes, PTPRM and PTPN22, which negatively regulate the phosphorylation of PKD-associated signaling pathways, including STAT3, ERK and mTOR (S6). The other genes identified with these techniques are: 1) categorized into signaling pathways that have been associated with PKD and others that have not yet been studied in PKD, and 2) associated with epigenetic age, including Hsd17b14, Itpkb, Mbnl1, Rassf5, and Plk2, which could be potential biomarkers to predict epigenetic age acceleration and may also contribute to disease progression in ADPKD. Knockout of Dnmt1 and targeting DNMT1 with hydralazine delayed renal cyst growth in Pkd1 knockout mice, supporting a potential therapeutic strategy for ADPKD by targeting abnormal DNA methylation with a safe demethylating agent (Fig. 10).

Figure 10. Working model of DNMT1 in regulation of cyst growth in ADPKD.

A schematic diagram depicting DNMT1 mediated pathways and processes in Pkd1 mutant renal epithelial cells and tissues. Pkd1 knockout or mutation results in the upregulation of DNMT1, which can be induced by cyst fluid TNF-α and further upregulated by the activation of STAT3 in Pkd1 mutant kidneys. Upregulation of DNMT1, 1) represses the expression of PTPRM and PTPN22, leading to the phosphorylation and activation of STAT3, ERK, and S6 to increase cystic renal epithelial cell proliferation; 2) represses the expression of p53, resulting in the inhibition of renal cystic epithelial cell apoptosis, and 3) may contribute to epigenetic age acceleration in cystic kidneys through methylation of specific epigenetic age associated genes. Targeting DNMT1 with a demethylating agent, hydralazine, delays cyst growth in Pkd1 mutant mouse kidneys.

“Epigenetic age” is an estimate of biological age based on changes in DNA methylation at particular locations along the genome.^{39, 40} The novel epigenetic clock can be used to address a lot of questions in developmental biology, cancer, and aging. ³⁹. For example, breast tissue shows evidence of significant DNAm age acceleration, whereas the DNAm age of sperm is significantly lower than the chronological age of the donor.³⁹ In this study, we found that PKD1 mutations resulted in dysregulation of the methylation of genes associated with the epigenetic clock (DNA methylation age) determined with MBD-seq, ChIP-seq and WGBS analysis, supporting an epigenetic age acceleration in ADPKD kidneys. The upregulation of DNMT1 was correlated with the dysregulation of epigenetic clock-associated genes in Pkd1 mutant renal epithelial cells and kidneys, suggesting that upregulated DNMT1 might promote epigenetic age acceleration in ADPKD kidneys. Furthermore, the functions of the five genes, including that 1) Hsd17b14 regulates sex steroid hormone metabolism, and is highly expressed in the kidneys and downregulated in the presence of kidney injury ⁴¹, 2) Itpkb is involved in inositol phosphate metabolism by phosphorylation of second messenger inositol 1,4,5-trisphosphate to release calcium from intracellular store in the endoplasmic reticulum ⁴², 3) Mbnl1 is an RNA-binding protein that regulates RNA alternative splicing, and plays important roles in cardiac fibroblasts and leukemia ^{43, 44}, 4) Rassf5 functions as a tumor suppressor gene that suppresses cell proliferation and promotes cells death ⁴⁵, and 5) Plk2 is an evolutionarily conserved serine/threonine kinase that are essential for mitotic centriole replication and cell cycle regulation ⁴⁶, suggest that the proteins encoded by these genes may also be involved in regulation of cyst progression in ADPKD.

In general, a global hypomethylation in ADPKD was observed compared to healthy individuals as examined with WGBS analysis, which is consistent with a recent study ⁴⁷. With reduced representation bisulfite sequencing (RRBS) analysis, ADPKD genomic DNA also exhibits global hypomethylation compared to non-ADPKD kidneys, and 13 differentially methylated regions are identified in ADPKD.⁴⁸ A greatest amount of variation of DNA methylation within CpG islands and gene bodies across each cyst of the ADPKD kidney has been identified, while intergenic fragments are comparatively stable.⁴⁹ With pyrosequencing analysis, about 91% over 13,000 unique fragments of the genome exhibited hypermethylation in human ADPKD kidneys versus non-ADPKD samples.⁵⁰ Our study focuses on the differentially methylated genes associated with epigenetic aging, which may provide a guidance to re-examine epigenetic aging associated genes in the databases generated in above studies.

MBD-seq analysis has been demonstrated to be highly specific, sensitive and applicable tool for identifying DMRs in methylome-wide association studies (MWAS).⁵¹ The genes identified by DNMT1 ChIP-seq analysis were fewer than that by MBD-seq analysis, suggesting that the binding of DNMT1 may be transient in the genome, or only specific genes are targeted by DNMT1 at specific time points during cyst progression.

The two novel DNMT1 target genes, PTPRM and PTPN22, can directly regulate the phosphorylation of PKD associated signaling pathways, including ERK, mTOR (S6) and STAT3, supporting that these tyrosine phosphatases are the mediators between DNMT1 and the PKD associated signaling pathways. STAT3 has been reported to regulate the transcription of DNTM1.⁵² In addition, TNF-α has been consistently detected in cyst fluid in ADPKD kidneys.⁵³ We found that the expression of DNMT1 is regulated by STAT3 and induced by TNF-α (Fig. S13), suggesting two mechanisms for the upregulation of DNMT1 in cystic kidneys: 1) Pkd1 mutation-mediated activation of STAT3 might partially contribute to the upregulation of DNMT1; and 2) during cyst initiation and development, accumulation of TNF-α in cyst fluid could further induce the expression of DNMT1 in cyst lining epithelial cells. A positive feedback loop should exist between DNMT1 and STAT3 in cystic renal epithelial cells and tissues (Fig. 10).

DNA methylation is reversible and thus can be altered by biological or biochemical manipulation, making it an attractive target for therapeutic intervention. Because 5-AZA has considerable cytotoxicity in addition to its demethylating activity ⁵⁴, there is a need of safer demethylating agents for clinical use. Hydralazine is a top candidate and highly safe. It has long been known to possess demethylating activity and has been extensively used for hypertension.⁵⁵ Hydralazine and its derivative dihydralazine were effective in demethylating and re-activating expression of abnormally silenced genes in cancer patients, leading to a delay in cancer progression.⁵⁶ The low-dose hydralazine has showed to decrease fibrosis and CKD progression.²⁶ The results that treatment with hydralazine slowed cyst growth in early-stage and late-stage Pkd1 knockout mice are important and exciting. The only concern is that hydralazine as a pan DNMT inhibitor may target both DNMT1 and other DNMTs. However, because DNMT1 is the only DNMT that functions to maintain DNA methylation established by DNMT3A/3B, and treatment with hydralazine has the same effect on PKD associated pathways in Dnmt1 knockout tissues and knockdown cells, any off-target effects on DNMT3A/3B should be limited. Considering particularly its high safety in human trials, our results that treatment with hydralazine slowed cyst growth in ADPKD animal models provide a rationale to initiate a clinical trial in ADPKD patients.

Supplementary Material

1

Figure S1. The expression of DNMT1 was increased in Pkd1^flox/flox:Tamoxifen-Cre-ERT mouse kidneys and human ADPKD kidneys.

Figure S2. The kidney weight and body weight as well as the expression of DNMT1 in Pkd1 and Dnmt1 double knock out mice.

Figure S3. Treatment with hydralazine had no effect on body weight of Pkd1^nl/nl mice but increased the body weight of Pkd1^flox/flox:Tamoxifen-Cre-ERT mice.

Figure S4. The distribution of important genomic features of ChIP-seq peaks by DNMT1 in PN24 cells.

Figure S5. DNMT1 target genes identified in PN24 cells by ChIP-seq, MBD-seq and RNA-seq (downregulated in PN24 cells versus PH2 cells) analysis.

Figure S6. PTPRM and PTPN22 were identified by MBD-seq and DNMT1 ChIP-seq in Pkd1 mutant cells.

Figure S7. Treatment with 5-azacytidine inhibited the growth of Pkd1 mutant cells.

Figure S8. Inhibition of DNMT1 with hydralazine increased the expression of p53 in Pkd1 mutant cells.

Figure S9. Epigenetic clock associated genes identified in PN24 cells.

Figure S10. Epigenetic aging associated genes were differentially methylated in Pkd1 mutant cells.

Figure S11. The methylation status of 5 epigenetic aging associated genes in kidneys from Pkd1 wild type and mutant mice.

Figure S12. PTPRM and PTPN22 were differentially methylated in human ADPKD kidneys.

Figure S13. The expression of DNMT1 was regulated by STAT3 and TNFα in renal epithelial cells.

2

Table S1. List of the 180192 methylated CpG sites identified by MBD-seq in PH2 cells.

3

Table S2. List of the 154974 methylated CpG sites identified by MBD-seq in PN24 cells.

4

Table S3. List of the 21390 genes identified by MBD-seq in PH2 cells.

5

Table S4. List of the 20797 genes identified by MBD-seq in PN24 cells.

6

Table S5. List of the 2603 genes with hypermethylated CpGs identified by MBD-seq in PN24 cells.

7

Table S6. KEGG enriched signaling pathways associated with hypermethylated genes identified by MBD-seq.

8

Table S7. List of the 1441 DNMT1 ChIP-seq peaks and associated 1215 genes identified in PN24 cells.

9

Table S8. KEGG enriched signaling pathways associated with DNMT1 target genes identified with ChIP-seq analysis.

10

Table S9. List of the 4874 DNMT1 ChIP-seq peaks and associated 3487 genes identified by DNMT1 ChIP-seq in PH2 cells.

11

Table S10. List of 65 genes identified by MBD-seq, DNMT1 ChIP-seq, and RNA-seq.

12

Table S11. List of the 309 genes associated with epigenetic clock (DNA methylation age) in human genomes.

13

Table S12. List of the 217 epigenetic clock (DNA methylation age) associated genes that were aligned with the differentially methylated regions (DMRs) identified by WGBS analysis in ADPKD patients.

14

Table S13. List of the 80 out of 217 human epigenetic clock (DNA methylation age) associated genes that were differentially methylated at TSS sites identified by WGBS analysis in ADPKD patients.

15

Table S14. List of the 343 hypermethylated genes identified by MBD-seq analysis in PN24 cells and by MGBS analysis in human ADPKD kidneys.

16

Table S15. The primers used for MSP analysis.

Translational Statement.

The key DNA methyltransferase, DNMT1, is upregulated in ADPKD kidneys. Understanding its role and mechanisms could lead to the development of therapeutic strategies to normalize dysregulated DNA methylation in ADPKD. To this end, the present study shows the identification of i) protein tyrosine phosphatases, PTPRM and PTPN22, as the novel DNMT1 targets that regulate PKD associated signaling pathways, ii) five genes associated with epigenetic age acceleration in ADPKD kidneys, which can be used as biomarkers to predict epigenetic age and disease progression based on their methylation status, and iii) hydralazine as a safe and efficient drug for ADPKD treatment.

Data Sharing Statement

The sequencing data sets have been deposited in the NCBI’s Gene Expression Omnibus (GSE257535).