Hypermethylation suppresses microRNA-219a-2 to activate the ALDH1L2/GSH/PAI-1 pathway for fibronectin degradation in renal fibrosis

Epigenetic regulations, such as DNA methylation and microRNAs, play an important role in renal fibrosis. Here, we report the regulation of microRNA-219a-2 (mir-219a-2) by DNA methylation in fibrotic kidneys, unveiling the crosstalk between these epigenetic mechanisms. Through genome-wide DNA methylation analysis and pyro-sequencing, we detected the hypermethylation of mir-219a-2 in renal fibrosis induced by unilateral ureter obstruction (UUO) or renal ischemia/reperfusion, which was accompanied by a significant decrease in mir-219a-5p expression. Functionally, overexpression of mir-219a-2 enhanced fibronectin induction during hypoxia or TGF-β1 treatment of cultured renal cells. In mice, inhibition of mir-219a-5p suppressed fibronectin accumulation in UUO kidneys. ALDH1L2 was identified to be the direct target gene of mir-219a-5p in renal fibrosis. Mir-219a-5p suppressed ALDH1L2 expression in cultured renal cells, while inhibition of mir-219a-5p prevented the decrease of ALDH1L2 in UUO kidneys. Knockdown of ALDH1L2 enhanced PAI-1 induction during TGF-β1 treatment of renal cells, which was associated with fibronectin expression. In conclusion, the hypermethylation of mir-219a-2 in response to fibrotic stress attenuates mir-219a-5p expression and induces the up-regulation of its target gene ALDH1L2, which may reduce fibronectin deposition by suppressing PAI-1.


Introduction
Chronic kidney disease (CKD) is a highly prevalent renal disease that affects about 15% of adults in the USA (https://nccd.cdc.gov/CKD/Default.aspx). CKD is featured by a gradual loss of renal function with various risk factors including hypertension, diabetes, autoimmune disorders, and acute kidney injury transition to CKD. Renal brosis, a pathological hallmark of CKD, is characterized by excessive accumulation of extracellular matrix (ECM) proteins including bronectin and collagens. Currently, effective therapy which can reverse or delay renal brosis is lacking. Various signaling pathways, including TGF-β signaling pathway and HIF pathway, play pivotal roles in renal brosis by promoting ECM protein production [1][2][3] . The ECM is continuously undergoing degradation and remodeling. The imbalance between ECM protein expression and degradation leads to excessive ECM deposition and brosis 4 . However, the understanding of dysregulation of ECM remodeling in renal brosis remains incomplete.
Epigenetic regulation includes the mechanisms that modulate gene expression inheritably without changing the primary DNA sequence. DNA methylation is one of the epigenetic mechanisms that broadly affect gene expression by adding a methyl group to the cytosine at a CpG site through a reaction catalyzed by DNA methyltransferases 5 . Functionally, the methylation of CpG sites in the promotor region of a gene usually suppresses the transcription of the gene. Aberrant DNA methylation has been implicated in various diseases, including CKD 6-13 . However, it remains poorly understood how DNA methylation contributes to the development and progression of CKD.
Non-coding RNAs, including microRNAs, represent another major mechanism of epigenetic regulation.
In this study, we examined DNA methylation changes in mouse models of renal ischemia/reperfusion and unilateral ureter obstruction (UUO). Our genome-wide DNA methylation analysis identi ed the hypermethylation of mir-219a-2 at its gene promotor region in these models, which was associated with a decrease in mir-219a-5p expression. Functionally, mir-219a-5p was shown to directly target and repress ALDH1L2, leading to the suppression of reduced glutathione (GSH) production, induction of plasminogen activator inhibitor 1 (PAI-1), and accumulation of bronectin. These ndings unveil hypermethylation of mir-219a-2 as a novel anti-brotic mechanism that is activated in renal tubular cells in renal brosis.
We analyzed genome-wide DNA methylation changes by reduced representation bisul te sequencing in models with signi cant renal brosis ( Supplementary Fig. 1), including 25 minutes of bilateral renal ischemia followed by 1-week reperfusion (I25/1wk) or 1 month reperfusion (I25/1M), and unilateral ureter obstruction for 7 days (UUO7D). Interestingly, in addition to protein-coding genes (data not shown), we identi ed multiple microRNA genes with differential methylation in brotic kidneys compared to control kidneys (Fig. 1A). Among them, mir-219a-2 was the only one showing signi cant hypermethylation (> 20%) (mean methylation values of 5 CpG sites) at the 5'-end promotor region of the gene in all models tested (Fig. 1A, 1B). We further con rmed the hypermethylation of two speci c CpG sites in the mir-219a-2 promotor region in these kidneys by pyrosequencing (Fig. 1B).
MicroRNAs repress target genes by binding to their mRNAs in the RNA-induced silencing complex (RISC) to suppress translation. In our experiments, mir-219a-5p enhanced bronectin expression (Figs. 2, 3), indicating that bronectin was not a direct target of mir-219a-5p. To elucidate the underlying mechanism, we systematically identi ed the direct targets of mir-219a-5p in renal cells (Fig. 4A). Speci cally, we performed RISC-immunoprecipitation (RISC-IP) to pull down the potential target mRNAs of mir-219a-5p for deep sequencing, revealing 197 mRNAs (Supplementary Table 1). We then analyzed the potential mir-219a-5p targeting sites in the 3'-UTR of these mRNAs by an online database (http://www.microrna.org) 24 and identi ed 10 mRNAs with conserved mir-219a-5p targeting sites in both human and mouse (Fig. 4B). We further searched the online database (The Human Protein Atlas, https://www.proteinatlas.org) to examine the protein expression and function of these 10 potential target genes 25 . Five of these genes mainly showed expression and function in neurons and were excluded from further study. For the other ve genes (CANX, ALDH1L2, SMG1, NR2C2, CRTC1), we veri ed that mir-219a-5p induced signi cant accumulation of these 5 gene mRNAs in RISC, while it did not induce obvious changes in cellular levels of these mRNAs (Fig. 4C, 4D). Thus, we considered them as the potential direct targets of mir-219a-5p in renal cells. To further identify the target gene that is responsible for the effect of mir-219a-5p in renal brosis, we examined the protein expression of these potential targets. Mir-219a-5p did not change CANX protein expression, although it induced the highest mRNA accumulation of CANX in RISC ( Supplementary   Fig. 7). The second most accumulated mRNA induced by mir-219a-5p in RISC was ALDH1L2 (Fig. 4B). In BUMPT cells, mir-219a-2 overexpression signi cantly suppressed ALDH1L2 protein expression ( Fig. 5A &  5B). Similarly, in HEK293 cells, mir-219a-5p mimics inhibited ALDH1L2 expression (Fig. 5D, 5E). In mouse kidneys, ALDH1L2 was mainly localized in renal tubular cells (Fig. 5C). Renal brosis in UUO was associated with a remarkable loss of ALDH1L2 in a large portion of renal tubules, which was prevented by anti-mir-219a-5p (Fig. 5C, supplementary Fig. 8). In luciferase microRNA target assay, the luciferase expression was signi cantly suppressed by mir-219a-5p in the presence of the predicted binding sequence at 3'-UTR, whereas it had no inhibitory effect when the binding sequence was mutated (Fig. 5F, Supplementary Fig. 9). Together, these data indicate that ALDH1L2 is a direct target of mir-219a-5p in renal brosis.
To determine the role of ALDH1L2 in regulating bronectin expression, we knocked down ALDH1L2 with speci c siRNAs in HEK293 cells (Fig. 6A, 6B). ALDH1L2 knockdown cells had higher levels of bronectin expression during TGF-β1 treatment than negative control (NC) sequence-transfected cells (Fig. 6C, 6D), indicating an anti-brotic role of ALDH1L2. Interestingly, ALDH1L2 knockdown increased bronectin protein without changing bronectin mRNA expression (Fig. 6E), suggesting ALDH1L2 regulates bronectin at the level of protein turnover. ALDH1L2 is a mitochondrial folate metabolic enzyme that mediates fatty acid metabolism, and its depletion leads to impaired production of GSH and enhanced oxidative stress 26,27 . Reduction of GSH can up-regulate PAI-1 to suppress plasmin-mediated bronectin degradation 28 . Therefore, we examined the possible connections between mir-219a-5p, ALDH1L2, and Fibronectin. Mir-219a-2 overexpression or mir-219a-5p mimics led to signi cant decreases in GSH in BUMPT cells and HEK293 cells, respectively (Fig. 7A, 7B). Similarly, the knockdown of ALDH1L2 decreased GSH in HEK293 cells (Fig. 7C). TGF-β1 induced PAI-1 as reported previously 28 , and this induction was markedly higher in ALDH1L2 knockdown cells (Fig. 7D). In vivo, UUO induced PAI-1 in speci c renal tubules, which was suppressed by anti-mir-219a-5p (Fig. 7E, 7F). To con rm the role of PAI-1 in renal brosis, we tested the effect of TM5441, a speci c PAI-1 inhibitor. As shown in Fig. 8, TM5441 reduced bronectin expression during TGF-β1 treatment in ALDH1L2 knockdown cells to the level of negative control siRNA-transfected cells. Collectively, these results indicate that mir-219a-5p may repress ALDH1L2 to induce PAI-1 resulting in bronectin accumulation.

Discussion
In this study, we deciphered a novel crosstalk between two epigenetic mechanisms, DNA methylation and microRNA, in renal brosis. The genome-wide DNA methylation sequencing revealed mir-219a-2 as the only hypermethylated microRNA gene in mouse models with renal brosis, including ureter obstruction and maladaptive repair after renal ischemia-reperfusion injury (Fig. 1A, 1B). Furthermore, this mir-219a-2 hypermethylation is associated with the suppression of mir-219a-5p expression ( Fig. 1C-1F) both in vivo and in vitro. All these results indicate the importance of mir-219a-2 hypermethylation in renal brosis.
Although various aberrant epigenetic regulations, including both DNA methylation and microRNA regulation, have been noted as critical pathological mechanisms, the crosstalk between different epigenetic mechanisms is not well understood. DNA methylation has been reported to regulate microRNA biogenesis 29 and DNA methylation-related microRNA expression change has been pro led in diabetic nephropathy 30 . However, a detailed functional analysis of such crosstalk is lacking. Notably, our study elucidated a crosstalk between DNA methylation and microRNA regulation by comprehensive examinations of the mir-219a-2 promotor hypermethylation, mir-219a-5p inhibition, and the downstream protein factors regulated by mir-219a-5p in renal brosis.
Moreover, the new pro-brotic function of mir-219a-5p was identi ed in renal brosis. We examined BUMPT cells with mir-219a-2 overexpression and identi ed that this overexpression led to more bronectin expression in vitro (Fig. 2). The pro-brotic effect of mir-219a-5p was further con rmed in a few other renal cell lines including rat proximal tubular cells (RPTC), human proximal tubular cells (HK2), and HEK293 cells ( Supplementary Fig. 5). In vivo, the inhibition of mir-219a-5p was mainly localized in renal tubules in UUO kidneys (Fig. 1E). If mice were treated with anti-mir-219a-5p LNAs, the interstitial bronectin accumulation in the kidneys after UUO injury was relieved (Fig. 3). Meanwhile, we did not detect an obvious effect on α-SMA (the broblast marker). Although renal broblast is one of the major contributors to ECM accumulation in renal brosis, emerging evidence indicates that renal tubular cells are playing critical roles in renal brosis initiation and progression 31 . All these data indicate that the suppression of mir-219a-2 gene expression by DNA hypermethylation may act as a self-protection mechanism for the kidney to prevent brosis development in CKD, and renal tubular cells are the major functional sites of mir-219a-5p.
The function of mir-219a-5p has been examined previously in neurons and cancer cells [20][21][22][23] . In those studies, multiple pathways were reported to be suppressed by mir-219a-5p, including the NMDA receptor pathway, Tau, Hedgehog signaling pathway, and Wnt/β-catenin pathway Therefore, mir-219a-5p may target different genes in different tissues or organs. To identify the target genes in renal brosis, we performed Ago2-IP with deep sequencing, followed by the examination of putative target protein expression and the luciferase microRNA target assay. These analyses identi ed ALDH1L2 as a direct target of mir-219a-5p in renal brosis, especially bronectin expression (Figs. 4 & 5). ALDH1L2 is a mitochondrial homolog of 10-formyltetrahydrofolate dehydrogenase, which controls the β-oxidation in fatty acid metabolism 26,27 . There is no report about ALDH1L2 in renal pathophysiology. Fatty acid oxidation is a major energy source in kidney proximal tubular cells, and its dysregulation is closely involved in renal brosis [32][33][34] . In our experiment, the knockdown of ALDH1L2 signi cantly enhanced bronectin expression during TGF-β1 treatment ( Fig. 6A-6D), suggesting that ALDH1L2 is a key downstream target of mir-219a-5p to regulate renal brosis. In vivo, anti-mir-219a-5p prevented the decrease of ALDH1L2 in UUO injured kidneys, which was associated with less bronectin induction (Fig. 5C). Altogether, these results demonstrate the important role of ALDH1L2 as a mir-219a-5p target in renal brosis.
How does ALDH1L2 suppression lead to renal brosis? In HEK293 cells, ALDH1L2 knockdown increased bronectin protein but not its mRNA (Fig. 6), suggesting the regulation of bronectin protein stability by ALDH1L2. Fibronectin is one of the major components of ECM that contributes to brosis 4 . Especially, bronectin polymerization is an essential step for other ECM proteins to deposit in the brotic niche 35 . The degradation of bronectin and related ECM remodeling involves two major families of metalloproteinases, matrix metalloproteinases (MMPs) and plasmin 4 . MMPs degrade a variety of ECM proteins, whereas plasmin is more speci c for degrading bronectin, brin, and laminin to regulate ECM remodeling. Our results showed that PAI-1 expression was induced after ALDH1L2 knockdown (Fig. 7D).
PAI-1 is the principal plasminogen activator inhibitor to promote plasminogen truncation to be plasmin 36 , and its expression can be regulated by GSH level 28 . In this regard, ALDH1L2 promotes the production of GSH in cells 27 , which may suppress PAI-1 expression and induce plasminogen to plasmin conversion for bronectin degradation 28 . Consistently, in our present study ALDH1L2 knockdown attenuated GSH production and induced PAI-1 in kidney cells (Fig. 7A, 7B, 7C).
In our study, inhibition of mir-219a-5p mainly suppressed bronectin expression, and its effect on other ECM proteins, such as collagens, was marginal (Fig. 3, Supplementary Fig. 6). Even though plasmin mainly mediates bronectin degradation and remodeling 4 , global knockout of PAI-1 attenuated both bronectin and collagen deposition in kidney brosis 37 . In addition, PAI-1 inhibitors including TM5441 reduced overall ECM accumulation in diabetic nephropathy 38 . There are a few possibilities that may explain why mir-219a-5p did not have signi cant effects on collagen deposition in our study. First, according to the histological examinations of mir-219a-5p and ALDH1L2 (Fig. 1E, Fig. 5C), mir-219a-5p inhibition to regulate ALDH1L2 and fatty acid oxidation is mainly localized in renal tubules. In kidney, because of the high energy requirement, proximal tubule is the major renal compartment using fatty acid as a fuel source 39 . Therefore, the effect of mir-219a-5p inhibition in ECM remodeling is relatively limited in renal proximal tubules. However, global PAI-1 de ciency or PAI-1 inhibitor treatment can function on all renal cells, such as myo broblasts and macrophages, to regulate ECM deposition. In fact, both PAI-1 knockout and inhibitor treatment were shown to suppress the in ammation as well as the transcription of ECM proteins in kidney 37,38 . Second, mir-219a-5p may have multiple potential target genes according to our microRNA target prediction and the Ago2-IP/RNA-seq result (Fig. 4). Although we con rmed ALDH1L2 as a direct target of mir-219a-5p in renal brosis, other target genes of mir-219a-5p may modulate ECM remodeling and collagen deposition as well. Finally, we only examined renal brosis up to two weeks after UUO injury. It is unclear whether the changes in bronectin may affect collagen deposition at later time-points or in the long term.
In conclusion, this study has demonstrated the regulation of mir-219a-2 by DNA methylation in renal brosis, unveiling a novel crosstalk between epigenetic mechanisms. Speci cally, brotic stress leads to hypermethylation of the mir-219a-2 gene, leading to the down-regulation of mir-219a-5p and a consequent increase in its target ALDH1L2 expression. Upon expression, ALDH1L2 promotes GSH to suppress PAI-1 expression, resulting in bronectin degradation and less renal brosis.

Methods
Animal models C57BL/6J mice were originally from The Jackson Laboratory (Bar Harbor, ME), bred and housed in Charlie Norwood VA Medical Center animal facility. Male mice of 8-12 weeks old were used for bilateral kidney ischemia surgery or unilateral ureter obstruction (UUO) surgery. All animal experiments followed the protocol approved by the Institutional Animal Care and Use Committee in Charlie Norwood VA Medical Center.
Bilateral kidney ischemia-reperfusion was conducted as described before 40 . Brie y, mice were anesthetized with 60mg/kg pentobarbital and kept on a homoeothermic blanket to maintain the body temperature at 36.5°C. Both kidney pedicles were clamped with micro-aneurysm clips for 25 minutes to induce kidney ischemia. The clips were released for kidney reperfusion and the mice were kept for 1 week or 1 month to collect kidneys. Sham operation was performed without renal pedicle clamping.
UUO was induced in mice as previously 41 . Brie y, mice were anesthetized with 60mg/kg pentobarbital and kept on a homoeothermic blanket for body temperature maintenance. The left ureter was ligated at two points with 4 − 0 silk suture to block urine and a cut was made between these two ligations. Sham operation was performed without ureter ligation and cut for comparison in in situ hybridization experiment. The contralateral kidneys were used in other experiments as the control for comparison purposes.

Genome-wide DNA methylation sequencing
The genomic DNA samples were extracted from mouse kidney cortex and outer medulla with QIAmp DNA Blood Mini kit from Qiagen (Germantown, MD) according to the manufacturer's manual. The Reduced Representative bisul te sequencing and reads alignment were conducted in Cancer Center Genomic Core Facility at Augusta University as described before 42 .

Pyrosequencing
The differentially methylated regions associated with the mir-219a-2 gene from genomic sequencing analysis were examined by pyrosequencing conducted by EpigenDx, Inc.(Hopkinton, MA) to con rm the methylation levels. The sequencing information was shown in Supplementary Table 2.

RNA Extraction
Total RNA from kidneys or cultured cells was extracted with miVana miRNA Isolation kit (Thermo Fisher Scienti c, Carlsbad, CA) or GeneJet RNA puri cation kit (Thermo Fisher Scienti c, Carlsbad, CA) following the manufacturer's instructions. Kidneys were ground in liquid nitrogen and immediately lysed in a lysis buffer. Cells were lysed with lysis buffer in the culture dishes.

Reverse transcription and Quantitative real-time PCR (RT and qPCR)
To quantify mRNA expression, 1µg of total RNAs were reversely transcribed into cDNA with iScript cDNA synthesis kit (Bio-Rad, Hercules, CA). qPCR was performed using iTaq Universal SYBR Green Supermix In situ hybridization Kidney samples were xed in 4% paraformaldehyde overnight at 4°C and balanced in 20% sucrose in PBS. Fresh cryo-sections of 7µm were air-dried for 10 minutes and brie y xed in 4% paraformaldehyde for 20 minutes. After PBS wash, the slides were treated with proteinase K (0.5µg/ml) for 10 minutes at room temperature and rinsed with PBS. The hybridization was then performed with IsHyb in situ hybridization kit (Biochain Institute Inc., Newark, CA) following the manufacturer's instructions. Brie y, the slides were incubated with a pre-hybridization solution at 63°C for 3hrs. The hybridization mixture with 0.5 nM DIG-labelled locked nucleic acid (LNA) probe (Exiqon, Germantown, MD) in 100ul hybridization solution was heated at 65°C for 5 min to linearize probes and then chill on ice. The slides were incubated with a hybridization mixture at 58°C overnight. After the SSC solution washes and blocking, the slides were exposed to anti-DIG alkaline phosphatase-conjugated antibody. Finally, the signal was developed with NBT/BCIP solution.

In vitro brosis models
The following renal cells were used for in vitro treatment to induce brosis: (1) The mouse proximal tubular cell (BUMPT) line was originally obtained from W Lieberthal and JH Shwartz at Boston University and maintained in DMEM culture medium with 10% FBS 43 . BUMPT cells were stably transfected with empty vectors (pCMV-MIR, OriGene, Rockville, MD) or mir-219a-2 overexpression plasmids (OriGene, Rockville, MD) to examine the role of mir-219a-5p in renal brosis. (2) rat proximal tubular cell (RPTC) was originally from Dr. Ulrich Hopfer 44 . RPTC cells were maintained in DMEM/F12 medium with 10% FBS and transiently transfected with miRIDIAN miRNA mimics negative control #1 or miRIDIAN mir-219a-5p mimics (Dharmacon, Lafayette, CO) to examine the role of mir-219a-5p. For hypoxia-induced brosis, the cells were incubated in a full culture medium in hypoxia (1% O 2 ) for 24-72 hours to induce brosis. Cells cultured in normoxia were used for comparison. To induce brosis by TGF-β1 (EMD Millipore, Burlington, MA), cells were cultured in a serum-free medium with 10 or 20 ng/ml TGF-β1 for 24-72 hours. Cells without TGF-β1 treatment were used for comparison.
In vivo LNA delivery LNAs of negative control sequence or anti-mir-219a-5p (Exiqon/Qiagen, Germantown, MD) were injected into mice through the tail vein. LNAs were dissolved in nuclease-free PBS at a concentration of 5mg/ml. Two injections of 20 mg/kg of LNAs were delivered two days before UUO surgery and three days after UUO surgery respectively. Animals were sacri ced at 2 weeks after UUO surgery. In this experiment, the contralateral kidneys were used as a control for comparison with kidneys with ureter obstruction.
Immunohistochemical Isolation kit for RNA sequencing. The high-through mRNA deep sequencing with a minimum of 6G raw data/sample and the standard quanti cation data analysis was performed by Novogene (Durham, NC).
microRNA target Luciferase assay HEK293 cells were co-transfected with luciferase DNA plasmids and 100nM RNA oligos from Dharmacon. Three DNA plasmids were used for comparison: Empty Vector (pMIR-REPORT Luciferase from Thermo Fisher Scienti c, Carlsbad, CA), ALDH1L2 (pMir-REPORT Luciferase plasmid inserted with predicted mir-219a-5p binding site in 3'UTR of human ALDH1L2), and Mutated (pMIR-REPORT luciferase plasmid inserted with mutated mir-219a-5p binding site) (Supplementary Fig. 9). At 24 hours after transfection, the cell lysates were collected with reporter lysis buffer (Promega, Madison, WI) and the luciferase activity was examined with Luciferase Assay System (Promega, Madison, WI) using a Tecan plate reader.

GSH Measurement
Cells were plated into a 96-well plate and GSH level was measured with GSH-Glo Glutathione Assay from Promega (Madison, WI) following the assay manual. Statistics.
Data were expressed as mean ± SD and analyzed with Microsoft Excel or GraphPad Prism 9. Student's ttest was used to show the signi cant difference between two groups. One-way ANOVA analysis with multiple comparisons speci ed in Figure Legends was used for multigroup difference analysis. A p-value less than 0.05 was considered statistically signi cant.

Declarations
Data Availability: All the data supporting the nding of this article are available upon request to the corresponding authors. Hypermethylation of Mir-219a-2 is accompanied by decreased mir-219a-5p expression in brotic kidneys.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. Mir219studyNCFiguresandsupplementary.pdf