Loss of Histone H3 K79 Methyltransferase Dot1l Facilitates Kidney Fibrosis by Upregulating Endothelin 1 through Histone Deacetylase 2

Significance Statement

If and how endothelin 1, a potent vasoconstrictor with proinflammatory and profibrotic properties, is upregulated to promote CKD is largely unknown. Emerging evidence has suggested that histone H3 K79 methyltransferase Dot1l exerts an antifibrotic effect by repressing the gene encoding endothelin 1 (Edn1) in the connecting tubule/collecting duct. The authors demonstrate in connecting tubule/collecting duct–specific Dot1l conditional knockout mice that disruption of Dot1l facilitates CKD development by upregulating endothelin 1, and that Dot1l and Edn1 double-knockout mice have a significantly attenuated CKD phenotype. Dot1l and histone deacetylase 2 mutually inhibit their association with the Edn1 promoter to regulate endothelin 1 production. This study is the first to define Dot1l as an epigenetic modifier of CKD, establishes a new CKD mouse model, and reveals a novel mechanism regulating Edn1 transcription.

Visual Abstract

Abstract

Background

The progression rate of CKD varies substantially among patients. The genetic and epigenetic contributions that modify how individual patients respond to kidney injury are largely unknown. Emerging evidence has suggested that histone H3 K79 methyltransferase Dot1l has an antifibrotic effect by repressing Edn1, which encodes endothelin 1 in the connecting tubule/collecting duct.

Methods

To determine if deletion of the Dot1l gene is a genetic and epigenetic risk factor through regulating Edn1, we studied four groups of mice: wild-type mice, connecting tubule/collecting duct–specific Dot1l conditional knockout mice (Dot1l^AC), Dot1l and Edn1 double-knockout mice (DE^AC), and Edn1 connecting tubule/collecting duct–specific conditional knockout mice (Edn1^AC), under three experimental conditions (streptozotocin-induced diabetes, during normal aging, and after unilateral ureteral obstruction). We used several approaches (colocalization, glutathione S-transferase pulldown, coimmunoprecipitation, yeast two-hybrid, gel shift, and chromatin immunoprecipitation assays) to identify and confirm interaction of Dot1a (the major Dot1l splicing variant in the mouse kidney) with histone deacetylase 2 (HDAC2), as well as the function of the Dot1a-HDAC2 complex in regulating Edn1 transcription.

Results

In each case, Dot1l^AC mice developed more pronounced kidney fibrosis and kidney malfunction compared with wild-type mice. These Dot1l^AC phenotypes were ameliorated in the double-knockout DE^AC mice. The interaction between Dot1a and HDAC2 prevents the Dot1a-HDAC2 complex from association with DNA, providing a counterbalancing mechanism governing Edn1 transcription by modulating H3 K79 dimethylation and H3 acetylation at the Edn1 promoter.

Conclusions

Our study confirms Dot1l to be a genetic and epigenetic modifier of kidney fibrosis, reveals a new mechanism regulating Edn1 transcription by Dot1a and HDAC2, and reinforces endothelin 1 as a therapeutic target of kidney fibrosis.

CKD is a growing public health burden in the United States, with age as a major risk factor. The rising rates of diabetes and hypertension, prevalent in older patients, account for up to two thirds of CKD cases. Repetitive injury remodels the structure of the kidney, contributes to a decline in organ function, and frequently progresses to overt fibrosis and functional decline, eventually leading to renal failure. Despite aggressive management, CKD often advances to ESKD, for which dialysis or transplantation are the only therapeutic options.

Tubular epithelial cells are particularly susceptible to injury and release factors that activate interstitial myofibroblasts, resulting in the deposition and accumulation of extracellular matrix components. Various signaling networks regulate myofibroblast function and renal pathogenesis, including the TGF-β1, Wnt, BMP, and Notch pathways. It remains largely unknown, however, why the progression rate of CKD varies substantially from patient to patient. Identification of genetic and epigenetic factors that modify the patients response to kidney injury may not only explain the individual susceptibilities to develop ESKD, but could also lead to development of novel therapeutic strategies and diagnostic tools.

The connecting tubule/collecting duct (CNT/CD) is the major site of the classic action of aldosterone to promote Na⁺ retention, leading to hypertension, a well established cardiac and renal injury risk factor (reviewed by Hostetter and Ibrahim¹). However, a role for CNT/CD in kidney scarring has not been established.

Evidence exists suggesting that histone H3 K79 methyltransferase Dot1l has an antifibrotic property by repressing Edn1, which encodes endothelin 1 (ET1), in the CNT/CD. First, Dot1a, a major splicing variant of Dot1l in the mouse kidney, regulates aldosterone target genes, including αENaC and Edn1.^2–5 Second, patients with diabetic nephropathy (DN) have decreased histone H3 dimethyl K79 (H3m2K79), which is presumably catalyzed by hDOT1L, in the CNT/CD cells.⁶ Third, complementary DNA array analyses of CNT/CD-specific Dot1l conditional knockout (Dot1l^f/f Aqp2Cre or Dot1l^AC) versus Dot1l^f/f mice at 2 months postpartum revealed upregulation of ET1.⁶ At this age, fibrosis is not detectable in both genotypes,⁷ indicating that ET1 upregulation is the potential cause rather than the consequence of CKD. Fourth, ET1 is secreted into the circulation and thus CNT/CD-derived ET1 has the potential to induce global effects. Finally, ET1 has a well established role in CKD (reviewed by Dhaun et al.⁸) and functions in an autocrine and paracrine manner.^6,9

We report here that Dot1l^AC mice recapitulated the renal pathologic features of patients with DN. Development of the Dot1l^AC phenotypes was inhibited by CNT/CD-specific disruption of Edn1. Dot1a and histone deacetylase 2 (HDAC2) mutually antagonize to modulate histone H3 K79 methylation and acetylation of the Edn1 promoter. Our study implicates, for the first time, a role for Dot1l as a genetic and epigenetic modifier of DN and CKD, establishes Dot1l^AC as a new CKD mouse model, reinforces ET1 as a therapeutic target for CKD management, and links loss of histone H3 K79 methylation and enhanced H3 acetylation to kidney fibrosis.

Methods

Reagents

The primary antibodies used were 15 rabbit antibodies specific for H3m2K79 (ab3594–100; Abcam), H3K9 methylation (07–212; Upstate), H3 acetylation (06–599; Millipore), histone H3 (ab1791; Abcam), HDAC2 (ab16032; Abcam), ET1 (18201; IBL International), endothelin A (ETA; ab85163; Abcam), ETB (ab117529; Abcam), RFP (632496; Clontech), GFP (632460; Clontech), FSP1 (A5114; Dako), fibronectin (F3648; Sigma), vimentin (ab92547; Abcam), collagen type 1 (ab292; Abcam), and collagen type 4 (LV1584915; Millipore); and four mouse antibodies against Aqp2 (sc-515770; Santa Cruz Biotechnology), α-tubulin (sc-8035; Santa Cruz Biotechnology), α-SMA (A2547; Sigma), and GAPDH (10R-G109a; Fitzgerald). Constructs expressing GFP-Dot1a, GFP-Dot1a 417–828, GST-Dot1a 479–659, BD-Dot1a, and AD-Af9 were as previously described.^3,10,11 pME18S-HDAC2 expressing FLAG-tagged mouse HDAC2 was kindly provided by Edward Seto (George Washington University).¹² A 1.5 kb EcoRI/XhoI fragment encoding full-length mouse HDAC2 was isolated from pME18S-HDAC2 and cloned into pGADT7, pGBKT7, and pDsRed-monomer-V5³ at EcoRI/XhoI sites for expressing HDAC2 as GAL4 AD, GAL4 BD, or RFP fusions, respectively. PCR fragments encoding mouse HDAC2 1–122 aa, 123–244 aa, 245–367 aa, and 368–489 aa were cloned into pGADT7 and pGBKT7 for expressing them as GAL4 AD and GAL4 BD fusions, respectively. All plasmids were verified by sequencing.

Generation and Characterization of Dot1l^f/f, Dot1l^AC, DE^AC, Edn1^AC, Dot1l^+/+ Aqp2Cre RFP^+/−, and Dot1l^f/f Aqp2Cre RFP^+/− Mice

Aqp2Cre,¹³ Dot1l^f/f, Dot1l^f/f Aqp2Cre (i.e., Dot1l^AC)^6,7,14 and Edn1^f/f15 mice have been previously reported. In this study, we studied four groups of mice: (1) wild-type (WT), which consisted of Dot1^f/f and Edn1^f/f; (2) Dot1l^AC; (3) Dot1l^f/f Edn1^f/f Aqp2Cre (DE^AC); and (4) Edn1^f/f Aqp2Cre (Edn1^AC). Inbreeding between offspring of Dot1l^ff Aqp2Cre and Edn1^f/f mice created Dot1l^f/f ET1^f/f, Edn1^AC, and DE^AC mice. Mating between Dot1l^f/f ET1^f/f and DE^AC mice produced these two genotypes at 1:1. Similarly, Rosa-RFP from Ai14¹⁶ was introduced into Aqp2Cre, and Dot1l^AC to create Dot1l^+/+ Aqp2Cre RFP^+/−, and Dot1l^f/f Aqp2Cre RFP^+/− mice. PCR-based genotyping was conducted to identify the genotypes. All mice were generated in a highly pure C57BL/6 background and used at the age of 2 months with free access to water and a normal Na⁺ diet unless otherwise indicated.

Streptozotocin-Induced Diabetes Model

Diabetes was induced by a single, intraperitoneal dose of freshly prepared streptozotocin (STZ; 150 mg/kg body wt) in 0.1 M citrate phosphate buffer (pH 4–5). Blood glucose levels were monitored every 2 weeks from tail vein blood samples. Diabetes was induced 14 days postinjection, with blood (glucose) reaching >250 mg/dl. Four months after STZ induction, urine and blood samples were collected. BP was measured using tail-cuff plethysmography. Mice were then euthanized for kidney harvest.

Unilateral Ureteral Obstruction Model

Unilateral ureteral obstruction (UUO) model was generated as previously described.¹⁷ In brief, mice were anesthetized. A lateral incision on the abdomen of the mouse was made to expose the left ureter, which was fastened at two points with a silk suture and continuously ligated. Mice were euthanized 14 days post-UUO for kidney collection.

Urine and Blood Analyses

Measurements of blood and urine parameters were conducted as we have previously reported.^5,7 Briefly, a flame photometer (PFP7; Jenway) was used to determine urine [Na⁺] and [K⁺]. Urine osmolality was measured with a vapor pressure (Wescor Vapro Vapor Pressure Osmometer 5520; Scimetrics, Houston, TX). Blood parameters were recorded using VetScan i-STAT (ABAXIS, Union City, CA) according to the manufacturer’s instructions. Plasma ET1 was measured, using a specific mouse ET1 ELISA kit (133030; Abcam). Systolic and diastolic BPs were determined with the CODA tail-cuff BP system (Kent Scientific, Torrington, CT) as previously reported.⁵ All mice were subjected to at least one cycle of measurement containing 20–30 individual readings for each parameter each day for 3 days.

Histologic Analyses and Tubular Injury Score

Paraffin-embedded kidney sections were subject to hematoxylin and eosin or Masson trichrome staining, following the standard protocols. The slides were examined by two renal pathologists (A.R.L and L.F.), who were unaware of the genotypes and experimental settings. Renal injury was scored using hematoxylin and eosin–stained slides, as described previously.¹⁸ Renal tubular damage was graded on the basis of tubular sloughing, cast formation, dilation, degeneration, atrophy, or tubulitis. At least ten nonoverlapping fields (×200 magnification) in the cortex were chosen randomly. Each field was scored from 0 to 4 (0, no injury; 1, 1%–20% of area; 2, 21%–50%; 3, 51%–75%; and 4, >75%). For each Masson trichrome–stained section, the whole section was scanned using NanoZoomer (Hamamatsu, Bridgewater, NJ) for quantification of kidney fibrosis. Kidney fibrosis was defined as the area fraction of the blue staining, using the ImageJ software (National Institutes of Health, Bethesda, MD) and normalized to WT.

Immunofluorescence Studies

Kidneys were fixed in 4% paraformaldehyde overnight at 4°C, embedded in paraffin, and cut into thin (4 μm) sections for epifluorescence and confocal microscopy. After boiling in antigen retrieval buffer (0.01 M sodium citrate, pH 6.0), paraffin sections were blocked with 5% BSA/0.5% Triton X-100 in PBS, followed by labeling with primary antibodies from different species (mouse and rabbit) in various combinations as detailed in each figure and figure legend. Primary antibodies were diluted in 1% BSA in PBS at 1:100–1000 and added to the slides overnight at 4°C. After three rounds of 5-minute washes in PBS, the slides were incubated with the corresponding combination of secondary antibodies (Alexa Fluor 488- and 594-conjugated donkey anti-mouse IgG, and anti-rabbit IgG), and mounted using Prolong Gold antifade mounting media (Invitrogen).

All slides were examined under an epifluorescence microscope (Olympus IX71) or confocal microscope (Nikon Eclipse Ti and Zeiss LSM 880 NLO confocal microscope with Airyscan).^4,6 Adobe Photoshop CS4 was used to assign each marker a color, which may be different from that observed under the microscopy for consistency between experiments.

Gel Shift Assay

The free oligo 5′-GATCCCGGAAGACTCTCCTCCGG-3′ was labeled with biotin, using LightShift EMSA Optimization & Control Kit (Thermo Fisher Scientific, Waltham, MA). Yeast were transformed with plasmids pBD-Dot1a 479–659 or pAD-HDAC2, and used for cell lysate preparation. Yeast cells were lysed using glass beads in 100 mM Tris-HCl buffer (pH 7.9) containing 250 mM ammonium sulfate, 1 mM EDTA, and 10% glycerol. Gel shift assays were conducted as detailed in our previous work.^19,20

Isolation, Immortalization, and Sorting of CNT/CD Primary Cells

CNT/CD cells deficient in Dot1l would be ideal to test if ET1 relays the Dot1l deletion–induced profibrotic effect to proximal tubule cells in vitro. Such cells, however, were unavailable. Hence, we isolated and immortalized CNT/CD primary cells with intact or disrupted Dot1l from Aqp2Cre Dot1^+/+ RFP^/+ and Aqp2Cre Dot1^f/f RFP^/+ mice. Kidney was dissected, minced, dissociated, and digested with collagenase type 1. Cells were grown in kidney culture medium for several days, which were replaced with retrovirus-containing supernatant from the PA317 LXSN HPV16E6E7 packaging cell line (CR-2203; American Type Culture Collection) in the presence of polybrene. Immortalized cells were selected with G418 and sorted on the basis of RFP at the Cytometry and Cell Sorting Core Facility, Baylor College of Medicine (Houston, TX).

Cell Culture, Transfection, Immunoblotting, Chromatin Immunoprecipitation, Coimmunoprecipitation, Real-Time Quantitative RT-PCR, Glutathione S-Transferase Pulldown, and Yeast Two-Hybrid Assays

All of these assays were conducted as we extensively detailed previously.^3–5,21 Whole kidneys were used in immunoblotting and real-time quantitative RT-PCR (RT-qPCR). For the production of conditioned medium, 1×10⁶/ml immortalized Dot1l^+/+ and Dot1l^−/− CNT/CD cells were plated in 10-cm dishes and cultured for 24 hours in 10 ml DMEM/F12 medium with 10% FBS, 100 U/ml penicillin, and 100 μg/ml streptomycin. After PBS washing, the cells were switched to DMEM/F12 medium with 1% FBS for 48 hours. The conditioned medium was then collected, sterile-filtered, and portioned. In parallel, HK2 cells were seeded in six-well plates and cultured in DMEM/F12 medium with 10% FBS overnight. After briefly washing with 1× PBS, HK2 cells were cultured in the freshly prepared conditioned medium of either immortalized Dot1l^+/+ or Dot1l^−/− CNT/CD cells for 48 hours. For neutralization experiments, the conditioned media were preincubated with 0.1, 0.5, and 2 μg/ml ED1 antibody (18201) at 4°C for 2 hours, before adding to HK2 cells. The treated HK2 cells were harvested and RNAs were extracted. For colocalization studies, IMCD3 cells were chosen because of their physiologic relevance to CNT/CD cells. For glutathione S-transferase (GST) and coimmunoprecipitation experiments, we used 293T cells because they are easily transfected, allowing high levels of expression of the proteins encoded by the transfected plasmids. For immunoblotting, blot densitometry was quantified, using ImageJ to measure background-subtracted region-of-interest intensities.

Statistical Analyses

Quantitative data are presented as mean±SEM. An unpaired t test was used to analyze two-group differences. Multigroup differences were assessed by one-way ANOVA followed by the Tukey post hoc comparison test, using Prism 5 software. The statistical significance was set at P<0.05.

Study Approval

All animal studies were performed in accordance with National Institutes of Health Guides for the Care and Use of Laboratory Animals and were approved by Institutional Animal Care and Use Committee at University of Texas at Houston and Albany Medical College.

Results

Dot1l Loss Facilitates Age-Dependent Development of Kidney Fibrosis by Upregulating ET1

We verified that Aqp2Cre drives Cre-mediated Dot1l ablation efficiently and specifically in the CNT/CDs in Dot1l^AC mice (Supplemental Figure 1). Although 5-month-old Dot1l^AC mice are apparently normal,⁷ 14-month-old Dot1l^AC animals developed severe kidney fibrosis. Hematoxylin and eosin staining revealed that WT mice lacked obvious abnormalities throughout the kidney. Dot1l^AC mice, in contrast, exhibited interstitial fibrosis with mononuclear inflammatory infiltrate, nonsclerotic glomeruli with dilated Bowman’s capsule in the cortex, and with attenuated tubular epithelium with hyaline casts in the outer and inner medulla. Because podocytes possess a fully functional endothelin system²² and represent an ET1 target, the glomerular shrinkage may result from the detrimental action of ET1. DE^AC mice showed no abnormalities in the cortex and reduced tubular epithelial cytoplasm without hyaline casts in the medulla. The renal injury score was significantly increased in Dot1l^AC and improved in DE^AC mice, compared with WT mice (Supplemental Figure 2).

Trichrome staining confirmed that Dot1l^AC kidneys developed more pronounced kidney fibrosis than WT mice, and that the phenotype was mitigated in DE^AC mice (Figure 1, A and B). Real-time RT-qPCR, immunofluorescence, immunoblot, and ELISA analyses confirmed that the expression levels of fibronectin, FSP1, α-SMA, vimentin, collagen type 1, collagen type 4, ET1, ETA, ETB, and HDAC2, and plasma ET1 levels, were significantly elevated in Dot1l^AC mice compared with WT or DE^AC mice. WT and DE^AC mice frequently had comparable levels (Figure 1, C–S).

Figure 1.

Dot1l deficiency facilitates development of severe kidney fibrosis at the age of 14 months by upregulating ET1. (A and B) Masson trichrome staining images (A) and ImageJ-based quantification of the staining (B) showing a significantly higher level of kidney fibrosis in Dot1^AC compared with WT or DE^AC mice. Boxed areas in 1× were sequentially 10× and 40× magnified. n=8–12 mice/group. Each point represents the average of 3–4 independent measurements of the entire kidney of the same mouse. The intensity was normalized to WT. (C–H) Real-time RT-qPCR showing mRNA expression of the genes in the whole kidney as indicated. n=4–6 mice/group. (I) Immunofluorescence showing expression of three fibrotic markers (green). (J–R) Immunoblot analyses of the proteins in the whole kidney as indicated. The 35-kd band in the ET1 blot was considered to be nonspecific and was excluded from quantification. GAPDH was used for normalization. n=7–8 mice/group. (S) ELISA showing plasma ET1 levels. n=6–18 mice/group. In all cases, *P<0.05 versus WT; ^#P<0.05 versus Dot1l^AC. In all cases, quantitative data were normalized to WT.

Metabolic analyses revealed that the pathologic deterioration in Dot1l^AC and its amelioration in DE^AC mice paralleled the corresponding changes in most of kidney functional parameters (Supplemental Figure 3). Dot1l^AC versus WT mice exhibited significantly increased water intake, urine volume, Na⁺ excretion, K⁺ excretion, BUN, diastolic BP, and systolic BP, and significantly decreased urine [Na⁺], [K⁺], osmolality, and blood [K⁺]. Many of these measurements, however, were completely or partially rescued in DE^AC mice. Therefore, ET1 is a principal contributor to the age-dependent kidney fibrosis induced by CNT/CD-specific Dot1l inactivation.

Dot1l Deficiency Exacerbates STZ- and UUO-Induced Kidney Fibrosis by Upregulating ET1

Dot1l^AC mice developed a more severe CKD phenotype than WT, Edn1^AC, and DE^AC mice 4 months after STZ injection (Supplemental Appendix 1, Supplemental Figures 4 and 5), suggesting that Dot1l inactivation confers susceptibility to DN by upregulating ET1. Similarly, more severe kidney fibrosis was found in Dot1l^AC mice than in WT and DE^AC mice 14 days after UUO (Supplemental Figure 6).

Dot1l^−/− Cells Secrete ET1 That Activates Fibrotic Genes in HK2 Cells

To investigate if ET1 relays the Dot1l deletion–induced profibrotic effect to proximal tubule cells in vitro, we isolated and immortalized CNT/CD primary cells with intact or disrupted Dot1l from Aqp2Cre Dot1^+/+ RFP^/+ and Aqp2Cre Dot1^f/f RFP^/+ mice. Cre-mediated recombination activated RFP expression specifically in CNT/CD cells in both genotypes and abolished Dot1l-mediated H3m2K79 in the latter (Figure 2A). After immortalization, primary CNT/CD cells were sorted on the basis of RFP to establish multiple single cell–derived Dot1l^+/+ and Dot1l^−/− clones. All cells within a given clone were RFP⁺ (Figure 2B). Immunoblot analyses verified robust H3m2K79 in the Dot1l^+/+ and abolished H3m2K79 in the Dot1l^−/− cells (Figure 2C). Dot1l^−/− cells had significantly higher ET1 mRNA expression and secreted more ET1 into the medium than Dot1l^+/+ cells (Figure 2, D and E).

Figure 2.

CNT/CD cells deficient in Dot1l secrete ET1 to stimulate expression of fibrotic markers in HK2 cells. (A) Immunofluorescence showing intact and abolished H3m2K79 (green) in the CNT/CD cells marked by RFP (red) in Dot1l^+/+ RFP^/+ Aqp2Cre and Dot1l^−/− RFP^/+ Aqp2Cre kidneys. (B) Autofluorescence (left) and phase (right) images showing a single cell–derived Dot1^−/− cell clone. CNT/CD primary cells were isolated, immortalized (see text), and sorted, on the basis of RFP, into single cells for clone formation. (C) Immunoblot showing presence and loss of H3m2K79 in Dot1l^+/+ and Dot1l^−/− clones, respectively. Total H3 and β-actin served as loading controls. (D) Real-time RT-qPCR showing upregulation of ET1 in Dot1l^−/− versus Dot1l^+/+ cells. n=4. *P<0.05. (E) ELISA showing that immortalized primary CNT/CD Dot1l^−/− cells secreted a higher level of ET1 into the media than Dot1l^+/+counterparts. n=6. P=0.0214 (F) Experimental scheme for (G–L). (G–I) Conditioned media from Dot1l^−/−, but not from Dot1l^+/+ cells significantly stimulated expression of three fibrotic markers in HK2 cells as indicated. n=6–12. *P<0.05 versus untreated; ^#P<0.05 versus Dot1l^+/+-treated. (J–L) The stimulatory effect was blunted by preincubation of the Dot1l^−/− conditioned medium with an ET1 antibody in a dose-dependent manner. n=4–8. *P<0.05 versus untreated; ^#P<0.05 versus Dot1l^−/−-treated.

Proximal tubule epithelial HK2 cells were treated with conditioned media from either Dot1l^+/+ or Dot1l^−/− cells in the absence or presence of increasing amounts of an ET1 antibody (Figure 2F). Real-time RT-qPCR revealed that Dot1l^−/−, but not Dot1l^+/+ conditioned media, significantly stimulated expression of FSP1, fibronectin, and collagen type 1, compared with untreated HK2 cells (Figure 2, G–I). Addition of the ET1 antibody largely abolished this effect in a dose-dependent manner (Figure 2, J–L). Thus, the Dot1l^−/− cells likely secrete ET1 to promote the expression of the fibrotic markers in HK2 cells.

Dot1l Loss Abolishes H3m2K79 and Increases H3 Acetylation at the Edn1 Promoter

To investigate if Dot1l represses Edn1 transcription through methylating H3 K79 at the Edn1 promoter, we performed chromatin immunoprecipitation (ChIP) assay. The 2-kb 5′ flanking region of Edn1 was arbitrarily divided into A–D subregions (Figure 3A). In Dot1l^+/+ cells, H3m2K79 ChIP signal was relatively high in C, mild in A and B, and weak in D. H3m2K79 became undetectable throughout the region in Dot1l^−/− cells, demonstrating the detection specificity of H3m2K79 (Figure 3B). Acetylated H3 (H3Ac), in contrast, was high in D, mild in A, low in B, and barely detectable in C in both genotypes. Surprisingly, Dot1l^−/− versus Dot1l^+/+ cells possessed significantly higher H3Ac, particularly in A and D, two subregions associated with little or no H3m2K79 in Dot1l^+/+ cells (Figure 3C). H3 methylation at K9 was not significantly different in each of the subregions between Dot1l^+/+ and Dot1l^−/− cells (Figure 3D), ruling out a nonspecific effect of Dot1l ablation on histone modifications. These data suggest that markedly reduced Dot1l-mediated H3m2K79 is linked to enhanced H3 acetylation at the Edn1 promoter.

Figure 3.

Dot1a and HDAC2 counterbalance each other to regulate Edn1 transcription by modulating H3m2K79 and H3 acetylation associated with the Edn1 promoter. (A) Diagram of the endogenous Edn1 promoter. (B–E) ChIP assays as indicated. H3m2K79 was associated with subregions A–C, barely with subregion D in Dot1l^+/+ cells, and abolished in Dot1l^−/− cells B. (C) H3Ac primarily occupied subregions A and D, with significantly higher levels in Dot1l^−/− than in Dot1l^+/+ cells. (D) H3 methylation at K9 (H3mK9) was mainly found in subregions A and D, with no difference between Dot1l^−/− and Dot1l^+/+ cells. (E) HDAC2 displayed a very similar pattern as H3Ac. n=4. (F) GFP-tagged Dot1a full-length and mutants harboring aa 479–659, but not aa 2–478, colocalized with RFP-HDAC2 in the nuclear compartments with barely detectable DAPI staining in IMCD3 cells. The DAPI-stained nucleus was highlighted with a broken line. *P<0.05 versus Dot1l^+/+.

Enhanced H3Ac Is Associated with Increased HDAC2 Recruitment to the Edn1 Promoter

On the basis of complementary DNA microarray analysis of 2-month-old Dot1l^AC versus Dot1l^f/f mice,⁶ it appears unlikely that Dot1l deletion significantly affects expression of acetyltransferases or deacetylases to affect H3Ac. Because HDAC2 can activate transcription in a deacetylase-independent manner and enhanced HDAC2 recruitment is associated with increased H3Ac,²³ we investigated if Dot1l disruption facilitated HDAC2 binding to the Edn1 promoter. ChIP with HDAC2 antibody yielded a very similar pattern to H3Ac across the region in both Dot1l^+/+ and Dot1l^−/− cells. HDAC2 binding was significantly elevated in A and D by Dot1l deletion (Figure 3E). These results suggest that Dot1l and HDAC2 may cooperate to regulate Edn1 transcription.

Dot1a Colocalizes with HDAC2 in the Nuclear Compartments Where DNA Staining Is Barely Detectable

To determine if Dot1a colocalizes with HDAC2, their GFP and RFP constructs were expressed in IMCD3 cells. Dot1a and HDAC2 fusions colocalized in the nuclear regions where DAPI staining was barely detectable (Figure 3F), suggesting that the complex is either associated with transcriptionally active euchromatin or does not bind DNA. Mutational analyses revealed that Dot1a 479–659 was sufficient for mediating this spatial overlap. RFP-HDAC2 colocalized with all GFP fusions harboring Dot1a 479–659. Dot1a 2–478, in contrast, did not apparently codistribute with HDAC2, although it formed large nuclear foci (Figure 3F). Hence, the colocalization is not simply because of overexpression of the fusion proteins.

Dot1a Interacts with HDAC2 in Multiple Assays

To confirm the interaction biochemically, we performed GST pulldown assays. GST fusions containing Dot1a 479–659, or two nonoverlapping subfragments within this region (e.g., aa 479–569 and 570–659), were expressed and purified from E. coli. The GST fusions were tested for their abilities to pull down RFP-HDAC2 in IMCD3 cell lysates. All GST-Dot1a fusions, but not GST alone, bound RFP-HDAC2 (Figure 4A, top panel). The observed results were not because of differences in the amounts of GST and GST-Dot1a fusions used in the assay (Figure 4A, bottom panel), as they were all comparable.

Figure 4.

HDAC2 interacts with Dot1a and impairs its DNA binding. (A) GST-Dot1a fusions, but not GST alone purified from E. coli, specifically binds RFP-HDAC2 from 293T cells. *A non-specific band (lane 1, top) in the lysates of un-transfected 293T cells. (B) GFP-Dot1a and RFP-HDAC2 coimmunoprecipitated in 293T cell lysates. Dotted line boxes are RFP-HDAC2 and GFP-Dot1a 417–828, respectively. (C and D) Dot1a and HDAC2 expressed as GAL4 BD and AD fusions, respectively, failed to activate GAL4 reporters in yeast two-hybrid assays. AD-Af9 served as the positive control. (E) Same as D, except BD and AD being switched. (F) Gel shift with GAL4 binding site as probe showing Dot1a binding to the Gal4 binding site through Gal4 BD, forming a BD-Dot1a/DNA complex (dotted line box). However, the increasing amount of AD-HDAC2 because of the interaction between Dot1a and HDAC2 impaired the formation of this protein/DNA complex.

The interaction was further confirmed by coimmunoprecipitation in 293T cells. RFP-HDAC2 coimmunoprecipitated with GFP-Dot1a 417–828 (using a GFP antibody), but not with GFP alone. Replacing the GFP antibody with normal rabbit IgG failed to immunoprecipitate both Dot1a and HDAC2 fusions, confirming the specificity of the coimmunoprecipitation and thus the Dot1a-HDAC2 interaction (Figure 4B). We did not use GFP-Dot1a 479–659 because it migrates closely with IgG heavy chain, complicating data analyses.

Dot1a-HDAC2 Interaction Impairs Their Association with DNA

An in vitro study reported strong, but apparently nonsequence-specific, binding of hDOT1L aa 1–416 with DNA.²⁴ HDACs do not have intrinsic DNA binding activity and are recruited to target genes via protein-protein interactions.²⁵ To test the hypothesis that the interaction prevents the Dot1a-HDAC2 complex from association with DNA, we used yeast two-hybrid assays. This approach was on the basis of the following findings. First, Dot1a fused to Gal4 DNA binding domain (GAL4-BD-Dot1a) interacts with GAL4 activation domain fusions harboring Af9 or Af17 (Gal4-AD-Af9 and GAL4-AD-Af17).^3,4 Second, GAL4-AD-HDAC2 interacts with GAL4-BD-IRS-1.²⁶ These findings demonstrate that GAL4 BD-Dot1a retains the GAL4-specific DNA binding activity and that GAL4 AD-HDAC2 has the ability to interact with DNA-bound GAL4-BD fusions to activate the Gal4 reporters. We reasoned if Dot1a-HDAC2 interaction prevents the complex from binding DNA, coexpression of GAL4-BD-Dot1a and GAL4-AD-HDAC2 will not lead to activation of the Gal4-dependent reporters. Indeed, the Gal4-dependent reporters were activated by co-expression of Gal4-BD-Dot1a with Gal4-AD-Af9, but not with Gal4-AD-HDAC2. Similar results were obtained with four nonoverlapping HDAC2 mutants tested as Gal4 AD or BD fusions (Figure 4, C–E).

We directly assessed the effect of HDAC2 on the ability of Gal4-BD-Dot1a to form DNA-protein complexes with the GAL4 binding site through gel shift assays. Gal4-BD-Dot1a 479–659 formed a specific DNA-protein complex with the biotin-labeled Gal4 binding site, leading to formation of a supershifted band in the presence of yeast lysate containing the Dot1a fusion. The formation of the Dot1a-DNA complex, nevertheless, was impaired with increasing amounts of Gal4-AD-HDAC2–containing yeast lysate, despite the fact that the amounts of Dot1a fusion were kept constant. The decreased formation of the Dot1a-DNA complex was coupled by increased release of the free probe (Figure 4F, lane 4). These results suggest that the interaction between Dot1a and HDAC2 impairs the DNA binding activity of the complex.

Dot1a and HDAC2 Are Mutually Antagonistic in the Regulation of Edn1

To test the hypothesis that Dot1a and HDAC2 mutually antagonize their ability to regulate target genes, we transiently transfected IMCD3 cells with constructs expressing GFP alone or GFP-Dot1a in the absence or presence of increasing amounts of RFP-HDAC2. Compared with GFP, GFP-Dot1a overexpression significantly decreased ET1 mRNA abundance. Such an effect, however, was attenuated by RFP-HDAC2 overexpression in a dose-dependent manner (Figure 5A). Reciprocally, overexpression of RFP-HDAC2 significantly upregulated ET1, compared with RFP vector–transfected cells. The upregulation of ET1 was impaired by increasing amounts of GFP-Dot1a (Figure 5B). Hence, our data suggest that Dot1a and HDAC2 counterbalance each other to coordinately regulate ET1.

Figure 5.

Dot1a and HDAC2 are mutually antagonistic in the regulation of Edn1. (A) Real-time RT-qPCR showing HDAC2 relieved Dot1a-mediated repression of ET1 in a dose-dependent manner. IMCD3 cells were transiently transfected with constructs expressing GFP alone or GFP-Dot1a without or with increasing amounts of RFP-HDAC2 plasmid. n=4. *P<0.05 versus GFP; ^#P<0.05 versus Dot1a alone. (B) As in A, Dot1a blunted HDAC2-mediated upregulation of ET1. IMCD3 cells were transiently transfected with constructs expressing RFP alone or RFP-HDAC2 without or with increasing amounts of GFP-Dot1a plasmid. n=4. *P<0.05 versus RFP; ^#P<0.05 versus HDAC2 alone.

Discussion

Mice with CNT/CD-specific inactivation of Dot1l recapitulated multiple molecular, cellular, and clinical features of patients with CKD. Upregulation of ET1 is the driving cause of kidney fibrosis induced by Dot1l inactivation. This study implicates Dot1l as a genetic and epigenetic modifier of CKD, reinforces ET1 as a potential therapeutic target, establishes Dot1l^AC mice as a new CKD model, identifies a new mode of Dot1a-HDAC2–mediated epigenetic regulation of transcription, and links abolished histone H3 K79 dimethylation and increased H3 acetylation in the CNT/CD to kidney fibrosis.

Patients with CKD differ substantially in the rate of disease progression, possibly because of genetic and epigenetic factors. The heritability of GFR, a CKD diagnostic marker, is 0.33 in a genome-wide study, 0.75 in patients with diabetes, and 0.41 in familial hypertension.^27–29 CKD is associated with single nucleotide polymorphisms at specific loci,³⁰ and certain inherited forms of CKD are linked to rare mutations in several genes (reviewed by Vehaskari³¹).

Dot1a is an integral component of aldosterone signaling network and represses αENaC by modulating H3K79 methylation in an aldosterone-sensitive manner. Under basal conditions, Dot1a complexes with Af9 at the αENaC promoter and promotes H3m2K79 hypermethylation to repress αENaC. Aldosterone relieves the repression by inhibiting the formation of Dot1a-Af9 complex.^2,3 Independent of aldosterone, Af17 derepresses αENaC by competing with Af9 to bind the same domain of Dot1a (aa 479–659) and facilitating Dot1a nuclear export.^4,5,32 Any or all of these mechanisms may be applicable to Edn1. Nevertheless, in this study we report a new mode of Dot1a regulation through its interaction with HDAC2. Similar to Af9 and Af17, HDAC2 also binds the same fragment of Dot1a.⁴ Unlike Af9, which recruits Dot1a to the promoters of target genes, and Af17, which facilitates Dot1a nuclear export, HDAC2 inhibits Dot1a association with DNA and vice versa. In this regard, relief of Dot1a-mediated repression can be achieved in both aldosterone-dependent and -independent pathways.

In addition, because hDOT1L plays a role in the pathogenesis of mixed-lineage leukemia–rearranged leukemia caused by the mixed-lineage leukemia fusion proteins,^14,33 hDOT1L inhibitors are currently being tested in clinical trials as a novel leukemia therapy.^34–36 It remains unknown, however, if such therapy will have any side effect on kidney histology and function.

Given the inhibitory effect of Dot1a on HDAC2 association with DNA, it is not surprising to detect increased HDAC2 ChIP signal at the endogenous Edn1 promoter in Dot1l^−/− cells. Enhanced HDAC2 occupancy, however, was associated with increased rather than decreased Edn1 H3 acetylation. HDAC2 can activate target genes independent of its deacetylase activity in part by recruiting paired box 5 to their promoters.²³ Accordingly, we propose a working model linking Dot1a-HDAC2–mediated H3m2K79 and H3Ac to kidney fibrosis via upregulating ET1 (Figure 6).

Figure 6.

A working model. (A) Diagram showing CNT/CD-specific Dot1l ablation facilitates development of kidney fibrosis by upregulating ET1. (B) Diagram showing Dot1a and HDAC2 counterbalance to regulate Edn1 transcription by modulating H3m2K79 and H3Ac associated with the Edn1 promoter, leading to more pronounced kidney fibrosis in Dot1l^AC than in other three groups. Note, HDAC2 activates Edn1 in a deacetylase-independent manner (see text).

Many cells including the vascular endothelium and tubular epithelium synthesize ET1, although to significantly lower levels compared with the CD. When produced in excess, ET1 promotes vasoconstriction, cellular injury, inflammation, proteinuria, and fibrosis through the ETA receptor. ETA antagonism alone, and/or combined ETA/B blockade, attenuate CKD progression in both preclinical models and in clinical trials.³⁷ Most recently, the Study of Diabetic Nephropathy with Atrasentan phase 3 trial^38,39 showed that atrasentan, a selective ETA antagonist, significantly decreased the risk of renal events in patients with type 2 diabetes mellitus and CKD.^38,39

Loss of Dot1l in the CNT/CD to increase ET1 levels appears to be the main mechanism for the CKD pathogenesis under the conditions tested (STZ, aging, and UUO). This notion is strongly supported by the severe phenotype of Dot1l^AC mice and its mitigation in DE^AC mice. Aged Dot1l^AC mice had significantly higher plasma ET1 levels than aged WT mice (approximately 3.8 versus 2.5 pg/ml; Supplemental Figure 1). Plasma ET1 in 14-month-old human EDN1 transgenic mice reached approximately 22 pg/ml versus controls (approximately 15 pg/ml). These transgenic animals displayed glomerulosclerosis, interstitial fibrosis, and renal cysts, but not hypertension.⁴⁰ Overexpression of mouse ET1 also mildly increased plasma ET1, from 1.5 pg/ml in WT controls to 2.7 pg/ml in 12-month-old transgenic mice. This change was accompanied by phenotypic and functional renal manifestations, including prominent interstitial fibrosis, renal cysts, glomerulosclerosis, elevated urinary protein excretion, and salt-dependent hypertension.⁴¹ Similarly, varying plasma ET1 concentrations have been reported in patients with CKD, ranging from approximately 3.7 to 60 pg/ml.^42,43 Although this wide range of plasma ET1 concentrations might be because of the different experimental settings and the use of different ELISA kits, our results suggest that mild, chronic overproduction of ET1 may confer susceptibility to the development of CKD.

Nevertheless, other mechanisms exist. Dot1l^AC versus WT mice had significantly higher expression of Aqp5, which interacts with Aqp2 and impairs its apical localization,⁶ contributing to extreme hypoosmotic polyuria (Supplemental Figure 3, D and E). Polyuria is proposed as a mechanistic driver of tubulo-interstitial injury and progression to renal failure in DN.⁴⁴ Chronic hypokalemia (Supplemental Figure 3R) can cause chronic renal injury. HDAC2 upregulation may also be contributory because HDAC2 activity was significantly augmented in STZ-induced diabetic rats and db/db mice.⁴⁵ We also found that HDAC2 was significantly increased in Dot1l^AC mice, compared with other groups in the STZ and aging settings. In the UUO model, however, DE^AC mice had even higher HDAC2 than Dot1l^AC mice (Supplemental Figure 6N), indicating that the role of HDAC2 in the development of kidney fibrosis may vary among the settings. We did not use endothelin receptor blockers to determine the contribution of ET1 to our observed phenotype—a limitation of this study. Collectively, Dot1l^AC mice recapitulate multiple features of patients with DN and CKD, including hypoosmotic polyuria, upregulation of ET1, and kidney fibrosis. Hence, the Dot1l^AC mouse is a new CKD model with potential for mechanistic discovery and amenable for development of new therapeutic strategies.

Disclosures

None.

Funding

This work was supported by National Institutes of Health grants DK080236 (to Wenzheng Zhang) and DK104073-01 (to Wenzheng Zhang), and startup funding from Albany Medical College (to Wenzheng Zhang).

Supplemental Material

This article contains the following supplemental material online at http://jasn.asnjournals.org/lookup/suppl/doi:10.1681/ASN.2019070739/-/DCSupplemental.

Supplemental Appendix 1. Supplemental results, discussion, and references.

Supplemental Figure 1. Dot1l is efficiently and specifically disrupted in the CNT/CD of Dot1l^AC mice.

Supplemental Figure 2. Dot1l inactivation facilitates development of severe kidney fibrosis at the age of 14 months by upregulating ET1.

Supplemental Figure 3. Dot1l inactivation facilitates kidney malfunction during normal aging by upregulating ET1.

Supplemental Figure 4. Dot1l ablation exacerbates STZ-induced kidney fibrosis by upregulating ET1.

Supplemental Figure 5. Dot1l loss exacerbates STZ-induced kidney malfunction by upregulating ET1.

Supplemental Figure 6. Dot1l ablation exacerbates UUO-induced kidney fibrosis by upregulating ET1.