RNA pull down was performed as described (Das et al., 2018) with some modifications. The hlncMGC (GenBank MW802745, MW802746, and MW802747) sense and antisense strands were biotin-labeled after in vitro transcription using Biotin RNA labeling Mix (Cat No. 11685597910, Roche) and T3 or T7 RNA Polymerase (Cat No. EP011, Stratagene) from pCR4-TOPO (ThermoFisher Scientific). Biotinylated lncMGC sense and antisense RNAs were treated with RNase-free DNase I and purified on G-50 Sephadex Quick Spin columns (Cat No.11274015001, Roche). Biotinylation efficiency for sense and antisense strands was determined using a Biotin Chromogenic Detection Kit (Cat No. K0661, ThermoFisher Scientific, Carlsbad, CA). 1 μg of biotinylated RNA was denatured by heating to 60°C for 10 min and slowly cooled to 4°C. RNA was mixed with 1 mg of nuclear extract prepared from human kidney cell line HK2 in RNA immunoprecipitation (RIP) buffer (150 mmol/L KCl, 25 mmol/L Tris pH 7.4, 0.5 mmol/L DTT, 0.5% NP40, 100 mmol/L PMSF, and 1x protease inhibitor) and incubated at 4°C for 2 h. Then, 60 μL of washed streptavidin agarose beads (Cat No. SA10004, Invitrogen) was added to each binding reaction and incubated at 4°C for 2 h. The beads were washed five times in spin columns (Cat No. 69725 Pierce, ThermoFisher Scientific). Proteins were eluted using SDS buffer and separated on a 4–15% precast SDS gel (Cat No. 5671084, Criterion, BioRad, Hercules, CA) and stained with SimplyBlue^TM SafeStain (Cat No. LC6065, Life Technologies). Bands were excised and subjected to protein identification by the Mass Spectrometry Core at City of Hope National Medical Center. Gel-separated proteins were reduced with DTT (Cat No. R0861, ThermoFisher Scientific, Rockford, IL), alkylated with iodoacetamide (Cat No. A3221-10VL, Sigma, St. Louis, MO), and digested with a mixture of trypsin and LysC (Promega). Peptides were extracted from the gel, evaporated to dryness in a vacuum centrifuge, and resuspended in 0.1% formic acid. The digested peptides were analyzed by LC/MS using an Orbitrap Fusion Mass Spectrometer with an EasyNano1000 nanoflow UHPLC (ThermoFisher Scientific, San Jose, CA). The peptides were loaded onto a 75 µm × 2 cm PepMap trapping column packed with 3 µm C18 silica particles, 100 Å pore size, and then, eluted through a 75 μm × 25 cm PepMap analytical column packed with 2 µm C18 silica particles, 100 Å pore size (both ThermoFisher Scientific), using an 85-min buffer A/B linear gradient from 8% to 25% buffer B (buffer A: 0.1% aqueous formic acid, buffer B: 0.1% formic acid in acetonitrile). The MS spectra of the intact peptides were acquired in the Orbitrap, and the CID MS/MS spectra were acquired in the ion trap. Data were searched using Sequest in Proteome Discoverer 2.1 (ThermoFisher Scientific). The database used was a concatenation of the Homo sapiens RefSeq proteome and a database of common laboratory contaminant proteins and was searched separately in the forward (target) and reverse (decoy) direction. The search assumed tryptic specificity with a maximum of two missed cleavages, a precursor ion tolerance of 5 ppm, and a fragment ion tolerance of 0.6 Da. It assumed the quantitative carbamidomethylation of cysteine and potential oxidation of methionine and acetylation of the protein amino terminus. The search results were loaded into Scaffold version 4.8.4 for probability assignment. For mouse lncMGC, all of the procedures used were the same as hlncMGC except that mouse lncMGC (GenBank MW802743 and MW802744) and the mouse kidney cell line TCMK1 were used. The interaction of proteins identified by mass spectrometry was analyzed using STRING DB (https://string-db.org/).

In vitro transcription/translation assay was performed to determine the coding potential of lncMGC using the T7 TNT quick-coupled transcription/translation system (Promega, Madison, WI, Cat. No. L1170) and transcend colorimetric non-radioactive translation detection system (Promega, Cat. No. L5070) following the manufacturer’s instructions. In brief, lncMGC lncRNA was subjected to in vitro transcription and translation from the full-length pCR4-TOPO construct. Then, 1 µL of the in vitro transcription/translation product was added to 15 µL of the SDS loading buffer, followed by protein denaturation at 90°C for 10 min and resolved using 4–15% SDS-PAGE gel. Protein products labeled with the biotinylated transcend tRNA were transferred and detected with streptavidin antibody and Western blue reagent following the manufacturer’s instructions. The T7 control DNA plasmid provided with the kit was used as a positive control.

All animal studies were conducted according to protocols approved by the Institutional Animal Care and Use Committee at the Beckman Research Institute of City of Hope National Medical Center. CRISPR gRNA design software (CRISPR direct) was used to choose the sgRNAs targeting lncMGC, as mentioned previously (Naito et al., 2015; Kato et al., 2021). A mixture of two CRISPR single-guide RNAs (sgRNA) flanking the transcription start site of lncMGC (5′-GAG​UUA​GUG​UGG​CCU​UCA​UC-3′ and5′-GCACGGUGCUGAAAGAGAGG-3′ in equal amounts, 50 ng/μL total) and 50 ng/μL Cas9 enzyme (IDT, Coralville, IA, United States of America) was microinjected into the pro-nuclei and cytoplasm of fertilized C57BL/6J 1-cell embryos. The microinjected embryos were implanted into pseudo-pregnant recipient female mice to produce mutant mice. Mutant candidate mice were screened by performing PCR with DNA extracted from the tails of the surviving mice. Several founders had shorter fragments relative to the WT by fragment analysis (Schuelke, 2000; Kato et al., 2021) and the deletion was confirmed by sequencing. The founders confirmed to have the anticipated deletion crossed with WT C57BL/6 mice and subsequent litters tested for germline transmission of the lncMGC deletion. The heterozygotes were crossed with each other to obtain homozygotes. A significant decrease of lncMGC expression in homozygous lncMGC-KO mice was confirmed using qPCR. Several lines of mutants were obtained.

Glomeruli were isolated from freshly harvested mouse kidneys (Kato et al., 2016; Kato et al., 2021). The renal capsules were removed, and the cortical tissue of each kidney was separated manually. The cortical tissue was gently strained through a stainless sieve with a pore size of 200 μm. The glomeruli were collected and filtered sequentially through sieves with pore sizes of 150 and 75 μm. After several washes with cold PBS, the glomeruli remaining on top of the sieve were collected. Pooled glomeruli were centrifuged, and the pellet was collected for RNA, protein extraction, or for preparing MMC according to our reported methods (Kato et al., 2021). MMCs were maintained in Roswell Park Memorial Institute 1640 Medium supplemented with 10% fetal bovine serum and treated with 10 ng/ml TGF-β (R&D Systems) for 24 h. Passages 5–7 were used for experiments (Kato et al., 2016; Kato et al., 2021).

RIP assays were performed as previously reported (Kato et al., 2010; Kato et al., 2021). In brief, MMC from WT and lncMGC-KO mice were plated in four 10-cm Petri dishes (70% confluence), washed with PBS, and cross-linked by exposure to 50 J/m² UV using a UV Stratalinker (Stratagene). The cross-linked cells were collected using 200 μL lysis buffer (provided in the miRNA Isolation kit, WAKO) per dish, gently mixed, incubated on ice for 10  min, sonicated (30 s × 5 with 30 s intervals) using a BIORUPTOR (Diagenode), and stored at −80°C. The lncMGC RNA-protein complex was immunoprecipitated with antibodies, IQGAP (Santa Cruz, sc-374307), NUMA1 (LSBio, LS-B12675), SMARCA5 (Boster Biol. Tech., A02687), DBC1 (Abcam, ab215852), BAT2 (Santa Cruz, sc-373747), MYBBP1A (Novus Biol., NB100-61050), YBX1 (LSBio, LS-B12352), KAP1 (Abcam, ab10484), Nucleolin (Abcam, ab134164), and SMARCC2 (Cell Signaling, #12760) by gently rocking at 4°C and washed with lysis buffer three times. RNAs were extracted using phenol-isoamyl alcohol and chloroform, precipitated with ethanol, washed with 70% ethanol, air dried, and dissolved in 10 μL nuclease-free water. The extracted RNAs were subjected to qPCR. The RIP-qPCR results (normalized enrichment) were calculated by the 2^−ΔΔCT method (Livak and Schmittgen, 2001) and normalized to input RNA (purified from 5% of the same cross-linked RNA–protein complex).

ChIP assays were performed as reported (Kato et al., 2013; Kato et al., 2016). MMCs were left untreated or treated with TGF-β (10 ng/ml) and then cross-linked with formaldehyde. These MMCs were cross-linked with formaldehyde (final concentration of 1% in phosphate-buffered saline for 20 min at room temperature) and then quenched with 125 mM glycine (5 min at room temperature). The cross-linked chromatin was sheared and immunoprecipitated with antibodies against H3K27Ac (Abcam, ab4729) and SMARCA5 (Boster Biol. Tech., A02687). The ChIP-qPCR results (normalized enrichment) were calculated by the 2^−ΔΔCT method (Livak and Schmittgen, 2001) and normalized to input DNA (purified from 5% of the same cross-linked chromatin).

RNA sequencing libraries were prepared with KAPA mRNA HyperPrep Kit (Kapa Biosystems, Cat KR1352) according to the manufacturer’s protocol (Kato et al., 2016; Kato et al., 2021). The libraries were validated with the Agilent Bioanalyzer DNA High-Sensitivity Kit and quantified with Qubit followed by cluster generation and sequencing performed on a HiSeq 2500 platform to generate 51 bp single-end reads at the Integrative Genomics Core of City of Hope National Medical Center. The raw sequences were aligned to the mouse reference genome mm10 using TopHat v2.1.1 (Trapnell et al., 2009) with default parameters, and gene-level expression levels of all the mouse RefSeq genes (downloaded from USCS genome Browser, https://genome.ucsc.edu) were counted using HTseq-count (Anders et al., 2015) v 0.6.0 with settings -s reverse (strand-specific) and -a 10 (filtrating poor alignment reads). The counts were normalized by the trimmed mean of M value (TMM) method (Robinson and Oshlack, 2010), and between-group comparisons on the expression of each expressed gene (RPKM >1 in at least one sample) were performed using the Bioconductor package edgeR v.3.20.9 (Robinson et al., 2010). For each comparison, a different score was assigned to each gene, which is a negative-log-transformed p-value followed by negating for the downregulated gene (fold change <1). Pre-ranked GSEA analysis was performed on genes ordered by the resulting score (high to low) using GSEA v. 4.0.3 to identify positive/negative-associated KEGG pathways defined in MSigDB. Bubble plots were generated to reflect normalized enrichment score (NES) and enrichment p-values of the significant signatures using R package “ggplot2” (v3.2.0). Upregulated genes (total 479 genes) induced by TGF-ß, namely, TGF-ß upregulated genes, in WT MMCs were identified from the expressed genes (with PRKM >1 in at least three samples) with fold change >2 at FDR <5% when comparing the TGF-ß WT vs. SD WT group. Among these genes, those whose expression upregulation by TGF-ß were reversed or attenuated by KO were identified by cluster analysis. IPA was applied to differentially expressed genes (DEGs) for gene ontology and pathway analyses. Motif analysis was performed by geneXplain (https://genexplain.com/).

A published Omni-ATAC protocol (Corces et al., 2017) was used for cell lysis, tagmentation, and DNA purification. The Tn5-treated DNA was amplified with 10 cycles of PCR in 50 μL reaction volumes. 1.8X AMPure XP bead purification was used for the PCR product cleanup. The libraries were validated with an Agilent Bioanalyzer DNA High Sensitivity Kit and quantified with qPCR. High-throughput sequencing on Illumina Hisq2500 was performed on the libraries followed by image analysis using real-time analysis (RTA) 2.2.38 software to generate paired-end sequences of 101 bp in-length. Raw reads were subsequently aligned to mouse genome mm10 using Novoalign (V3.02.07), followed by filtering reads located on mitochondria, not properly paired, of pair-end length > 2kb, mapping quality <20, or duplicated reads using Samtools. Genrich v0.5 was further applied to the resulting bam files of the three replicate samples in one group to identify the peak regions (open chromatin) for each group. To identify differentially enriched regions between the two groups, the peak regions of the two groups were combined (union) to generate common peak regions. Reads located in each common region were counted in each sample (three replicates per group) followed by quantile and TMM normalization and compared using edgeR. Differentially enriched regions including hyper-enriched regions and hypo-enriched regions for each comparison were identified as the common peak regions with fold change >1.5 and FDR <5%. To present the enrichment difference at the identified decreased–enriched regions between two groups of samples using heatmap, the coverage of each sample was calculated using all the paired reads using readGAlignmentPairs function provided in the GenomicAlignment R package after removing duplicate reads and scaled to a total of 50 million reads. For each hypo-enriched region between KO vs. WT under TGF-ß treatment, the 5 kb flanking region relative to its midpoint was divided into 500 bins (sub-regions), each of 20 bp in length. The average coverage of each bin was then calculated, resulting in a matrix with each row representing one region and each column representing one bin (which has the same location relative to the midpoint of the region). The average coverage of the triplicates in each group was calculated and represented by a heatmap. To get an aggregated profile of all the regions to present for each group, the coverage of each bin across all the regions was averaged and presented as a line plot. De novo motif analyses were applied to hypo/hyper-enriched regions to identify transcription binding sites enriched in the identified regions using the peak-motifs function in RSAT Metazoa (http://rsat.sb-roscoff.fr/).

ChIP-seq libraries were prepared with a KAPA DNA HyperPrep Kit (Kapa, Cat KK 8700) according to the manufacturer’s protocol. 5–10 ng of immunoprecipitated DNA underwent end-repaired A tailing and adaptor ligation. The final sequencing library was produced by 10 cycles of PCR reactions. The libraries were validated with the Agilent Bioanalyzer DNA High Sensitivity DNA Kit and quantified with Qubit and sequenced on Illumina NovaSeq to generate paired-end 101 bp sequences. Raw reads were subsequently aligned to mouse genome mm10 using BWA and peak calls. A bedgraph file for each sample was generated on the aligned reads after extending to 200 bp on each read and scaled to 50 million total reads. To make the heatmap and aggregate profile of H3K27ac signals at the ATAC hypo-enriched regions in KO-TGF-ß vs. WT-TGF-ß, the same approach, as described in ATAC-seq data analysis, was applied to obtain the average coverage of H3K27ac signals across each region (5 kb-flanking region relative to the midpoint) for the duplicate samples in each group and then the aggregated coverage across all the regions for each group.

Oligonucleotides for siRNAs and corresponding control oligos were obtained from Integrated DNA Technologies or ThermoFisher Scientific Inc. (Waltham, MA) and used as described (Kato et al., 2009; Kato et al., 2013). In brief, MMCs (∼10⁶/transfection) were transfected with siRNA oligos using an Amaxa Nucleofector (Lonza, Basel, Switzerland) according to the manufacturer’s protocols. siRNAs (double-stranded oligos) targeting mouse SMARCA5 and non-targeting siRNA controls were obtained from ThermoFisher Scientific. MMCs were trypsinized and resuspended in Basic Nucleofection Solution at 1 × 10⁷/ml. Subsequently, 100 μL of cell suspension (1 × 10⁶cells) was mixed with the siRNA or control oligonucleotides or ON-TARGET plus siRNA or NCs (ThermoFisher Scientific). The transfected cells were collected for RNA extraction. RNA was extracted from the cells, and the expression of coding genes, lncMGC, or miR-379 was examined using primers designed for each target.

RT-qPCR analysis was performed as described (Kato et al., 2016; Kato et al., 2021). RNA was extracted using an RNeasy Mini Kit (Qiagen, Valencia, CA). miR-379 quantification was performed using the qScript miRNA cDNA Synthesis Kit (Quanta Biosciences, Gaithersburg, MD) and amplified using PerfeCTa SYBR Green SuperMix (Quanta Biosciences). For miRNAs, specific mature miRNA sequences were used as forward primers, and the universal primer provided in the kit was used as the reverse primer. U6 was used as an internal control. A GeneAmp RNA PCR Kit (Applied Biosystems, Carlsbad, CA) and POWER SYBR Green Mix (Applied Biosystems) were used for mRNA quantification. mRNA expression was normalized to Cypa as an internal control. Sequences of primers used in this study are shown in Supplementary Table S6.

Statistical data analyses were performed using GraphPad Prism software (8.2.1). Normal distribution of each sample group was confirmed using the χ ²-test or Shapiro–Wilk test before comparison between groups. All data were expressed as the mean ± SEM, and statistical analyses were performed using Student’s t-tests (two-sided) to compare the two groups or analysis of variance (ANOVA) followed by post hoc Tukey’s test to compare multiple groups. For all experiments, the number of replicates is shown in the figure legends. Asterisks indicate significant difference (*p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001).

The genomic structure of hlncMGC (GenBank accession numbers MW802745, MW802746, and MW802747), which covers miR-379 miRNA cluster from miR-379 (most 5′) to miR-656 (most 3’) on human chromosome 14q32.2, is shown in Supplementary Figure S1A (Kato et al., 2016). lncRNA actions are often determined by their protein-interacting partners (Kato and Natarajan, 2019; Statello et al., 2021). Therefore, as a first step, to identify hlncMGC-interacting proteins, hlncMGC (MW802745) was cloned into the pCR4-TOPO expression vector to express lncMGC RNA in vitro (Supplementary Figure S1A). No potential lncMGC protein was detected from the vector unlike the positive control luciferase protein, suggesting that hlncMGC RNA is not protein-coding (Supplementary Figure S1B). Very low or no ribosome occupancy (GWIPS-viz, https://gwips.ucc.ie/) (Ingolia et al., 2009) was detected at lncMGC and also other non-coding RNA regions (MEG3), although significant ribosome occupancy was detected at three coding regions (DLK1, PPP2R5C, and DYNC1H1) (Supplementary Figure S1C). Those results again suggest that lncMGC is not a protein-coding RNA, as reported previously (Charlier et al., 2001; Hatada et al., 2001; Seitz et al., 2004; Kato et al., 2016). In vitro-transcribed biotinylated hlncMGC RNA was mixed with the human kidney cell (HK2) lysate, and RNA–protein complexes were isolated (Supplementary Figure S2A). Isolated proteins were separated by SDS-PAGE (Supplementary Figure S2B) and analyzed by mass spectrometry (MS). MS data revealed 135 proteins specifically interacted with sense hlncMGC RNA (Figure 1A). The STRING database (https://string-db.org/) suggested that the hlncMGC-interacting proteins identified by MS were grouped into RNA processing factors, ribosomal proteins, and nucleosome remodeling factors (Figure 1B). Interestingly, among the nucleosome remodelers, SMARCA5 was in the center and appeared as a strong candidate hub protein (Figure 1B). The top 10 candidate interacting proteins were displayed in a heatmap (Figure 1C), with the ranking based on the numbers of sense (S) RNA-interacting peptides detected by MS. In HK-2 cells, the interaction of lncMGC and candidate proteins was further validated by RIP and qPCR (Supplementary Figure S2C) and relative affinities depicted as a heatmap (Figure 1D). The relative affinity of hlncMGC for SMARCA5 was higher than that reported for non-coding lncTCF7 (positive control for SMARCA5 interaction) (Wang et al., 2015) (Figure 1D).

We also performed RNA pulldown–MS experiments to identify mouse lncMGC-interacting proteins. The genomic structure of mouse lncMGC (MW802743 and MW802744), which covers the miR-379 miRNA cluster from miR-379 (most 5′) to miR-3072 (most 3’) on mouse chromosome 12qF1, is conserved and displays synteny with the hlncMGC genomic region (Supplementary Figure S3A). Mouse lncMGC (MW802744) was cloned into the pCR4-TOPO expression vector (Supplementary Figure S3B). In vitro-transcribed biotinylated lncMGC RNA was mixed with mouse kidney cell (TCMK-1) lysates, and RNA–protein complexes were isolated and analyzed by MS. The STRING DB again suggested groups of RNA processing factors and nucleosome remodeling factors as mouse lncMGC-interacting proteins (Supplementary Figure S4), similar to hlncMGC. SMARCA5 was again selected as a strong candidate because it was in the center of mouse lncMGC-interacting nucleosome remodelers (Supplementary Figure S4). These results suggest that lncMGC regulates the chromatin structure in kidney cells and that the function of lncMGC RNA is conserved from human to mouse. The structure of SMARCA5 along with histone/DNA is depicted in Figure 1E (view from the front) and Figure 1F (view from the side) (Protein Data Bank, https://www.ebi.ac.uk/pdbe/entry/pdb/6ne3/analysis#assembly_1). Through these interactions, SMARCA5 is suggested to unwind nucleosome structures to promote chromatin accessibility and allow gene transcription (Corona and Tamkun, 2004; Erdel and Rippe, 2011). Therefore, hlncMGC–SMARCA5 interactions may regulate chromatin remodeling and enhance the transcription of the lncMGC/miR-379 miRNA cluster as well as other lncMGC targets.

To study the molecular actions and function of lncMGC, we used the primary renal mesangial cells isolated from lncMGC-knockout mice, which were generated using CRISPR-Cas9 editing. Two guide RNAs (gRNAs) were designed (Naito et al., 2015; Kato et al., 2021) to delete the upstream region of miR-379 and within lncMGC (Supplementary Figure S5A). Several lines of mutant mice were obtained, and two of them (KO1, 37bp deletion and KO5, 56bp deletion) are schematically shown in Supplementary Figure S5A. Deletion of the indicated genomic region was confirmed by Sanger sequencing (Supplementary Figure S5B). We verified that 56 bp just downstream of INR (initiator, TSS) was missing in lncMGC-KO5 mice. The potential structures of WT and partial lncMGC-KO5 RNAs were predicted by The Vienna RNA Websuite (http://rna.tbi.univie.ac.at/) (Gruber et al., 2008). WT lncMGC RNA, the deleted sequence in the KO5, and partially deleted KO5 lncMGC RNA are shown in Supplementary Figure S5C. Based on our subsequent data, it is likely that the stem-loop structure (partly deleted in the KO5 RNA) may be important for the interaction with nucleosome remodelers such as SMARCA5 and the regulation of chromatin structures. We established primary cultures of MMC from these KO mice and tested the candidate gene expression (Supplementary Figures S5D, S5E). A significantly lower expression of lncMGC, miR-379, and Mirg lncRNA MMC (overlapping with the 3’ part of mouse lncMGC) was confirmed in lncMGC-KO5 compared to the WT MMC. lncMGC was not expressed in lncMGC-KO1 and KO5 MMC. There was a significant increase in lncMGC in the WT MMC after TGF-β (10 ng/ml for 24 h) (Kato et al., 2009; Kato et al., 2011) treatment (Supplementary Figure S5D). Furthermore, the induction of miR-379 and Mirg by TGF-β was also lost or attenuated in lncMGC-KO MMC, whereas both miR-379 and Mirg were increased by TGF-β in the WT MMC (Supplementary Figures S5E, S5F). lncMGC-KO5 MMC, which showed a trend toward lower miR-379 expression than KO1 MMC (Supplementary Figure S5E), was used for further studies.

To test the global transcriptomic effects of lncMGC, RNA-seq was performed on MMC isolated and cultured from WT and lncMGC-KO5 mice (Figure 2). The RNA-seq results showed differentially expressed genes with a significant decrease of extracellular matrix (ECM) genes, such as Col4a3 and Col4a4, in KO5 MMC both in control (serum-depleted/SD) and TGF-β-treated cells, confirming that KO5 MMC have attenuated response to TGF-β (Figures 2A, B). Gene set enrichment analyses (GSEA) comparing KO vs. WT MMC identified several gene sets related to inflammatory response, such as RIG-I-like receptor, NOD-like receptor, and JAK-STAT signaling, which were attenuated in KO vs. WT, basal and TGF-β treated cells (Figure 2C). Furthermore, Ingenuity Pathway Analyses (IPA) of differentially expressed genes (upregulated by TGF-β in WT MMC but downregulated in KO5 MMC), to elucidate DKD related pathways, revealed inclusion of genes associated with kidney damages and fibrosis (Figure 2D) and inflammation (TNF in the center and SMAD3) (Figure 2E).

The heatmap shown in Figure 3A emphasized that genes involved in the pathology of DKD, including TGF-β, ECM, fibrosis, oxidative stress, mitochondria, inflammation, and ER stress pathways, are upregulated by TGF-β in WT MMC but attenuated in lncMGC-KO5 MMC, suggesting that lncMGC has widespread effects on gene expression regulated by TGF-β (Figure 3A). We further validated the differential expression of Col4a3, Col4a4, Ucp2, Nox4, and Nkd2, by RT-qPCR (Figures 3B–G). Significantly lower expression of these genes was detected in KO5 MMC compared to WT MMC in the TGF-β-treated condition by RT-qPCR, although no such decrease was detected in the control condition (SD). Pai1 expression was not significantly reduced in KO5 MMC (Figure 3G), suggesting not all DKD-related genes may be regulated by lncMGC.

To determine if SMARCA5 is involved in the regulation of key DKD-related genes by lncMGC, we treated WT MMC with SMARCA5 siRNA (siSMARCA5) or a negative control siRNA (siNC) (Figures 3H–N). siSMARCA5 significantly decreased the expression of TGF-β-induced Col4a3, Col4a4, Ucp2, Nkd2, and Nox4 in WT MMC (Figures 3H–M). However, the expression of some genes such as Pai1 was not affected by siSMARCA5 (Figures 3N) and mirrors data in KO MMC (Figure 3G).

Because of the connection of lncMGC to nucleosome remodeling factors, we next analyzed its impacts on the chromatin structure by genome-wide mapping of chromatin accessible regions (open chromatin) using Assay for Transposase-Accessible Chromatin with high-throughput sequencing (ATAC-seq). The ATAC-seq results identified peaks at 6,919 loci in WT MMC more than KO5 MMC in the control untreated condition. After treatment with TGF-β, 3,862 loci were more accessible in WT vs. KO5 MMC, suggesting lncMGC can confer widespread alterations in chromatin and that this process is TGF-β sensitive. There was a significant decrease in global ATAC-peaks (intensities and average densities) from -5 kb to +5 kb at transcription start sites (TSS) in KO5 MMC compared to WT MMC in control (Figures 4A, B) and TGF-β-treated conditions (Figures 4E, F). Since histone H3K27 acetylation (H3K27ac) is a chromatin mark of active enhancers and gene expression (Creyghton et al., 2010; Kundaje et al., 2015) in parallel, we performed chromatin immuprecipitation-Seq (ChIP-seq) with the histone H3K27ac antibody (Figures 4C, D, G, H & Supplementary Figure S6). Similar to ATAC-seq, we observed significant decreases in global H3K27ac-peaks (intensities and average densities) from -5 kb to +5 kb at TSS in KO5 MMC compared to WT MMC in control (Figures 4C, D) and TGF-β-treated conditions (Figures 4G, H).

Heatmap ATAC-seq peak analysis indicated that gene loci involved in DKD show higher signals in WT MMC compared to lncMGC-KO5 MMC even after TGF-β treatment. Thus, lncMGC may be involved preferentially in the opening chromatin at the loci of DKD-related genes regulated by TGF-β (Figure 5A). We next checked the signals of ATAC-peaks nearby candidate genes involved in DKD (Figure 3A) and observed a significant decrease of ATAC-seq signals in KO5 MMC (Figure 5A) at genes decreased in KO5 MMC treated with TGF-β (Figure 3A).

We also observed H3K27ac signals at H3K27-enriched regions (H3K27ac-peaks) in the promoter or gene body of the DEGs related to DKD (Figure 3A), which also confirmed the overall correlation of H3K27ac signals with gene expression (Figure 5B) with some exceptions. H3K27ac levels were increased even in KO5 MMC treated with TGF-β. It is likely some TGF-β-regulated genes, such as Col4a3, Col4a4, Ucp2, Nox4, and Nkd2, were suppressed in KO5 cells (Figure 3A) because their chromatin was closed in the KO5 MMC (Figure 5A) even though H3K27 was acetylated (Figure 5B). Then, we more closely examined the genome tracks of RNA-seq, ATAC-Seq, and H3K27ac ChIP-seq at Nox4(Oh et al., 2016), Ucp2, Col4a3, Col4a4, Nkd2, Cadm1, and Klhl1 representative loci (Figures 5C–H). The expression of Nox4 and Ucp2 genes (associated with oxidative stress) was upregulated by TGF-β in WT MMC, but the induction was clearly lower in KO5 MMC (RNA-seq, red). ATAC tracks showed that their promoter regions were open even before TGF-β treatment in WT MMC and closed in KO5 MMC even after TGF-β treatment (ATAC-seq, green). H3K27ac was slightly increased by TGF-β treatment in both WT and KO5 MMC (blue). IGV tracks show the overlap of H3K27ac peaks and ATAC-peaks at the Nox4 promoter (Figure 5C), confirming ATAC-seq marks at regulatory regions. H3K27ac is an epigenetic mark of active promoters and enhancers (Creyghton et al., 2010; Kundaje et al., 2015). Genes associated with kidney fibrosis, such as Col4a3, Col4a4, Nkd2 (Kuppe et al., 2021), Cadm1 (Hagiyama et al., 2019), and Klhl1, showed similar trends (Figures 5D–H). Differentially expressed genes (upregulated by TGF-β but downregulated in KO5 MMC), such as Xylt1 (Tziastoudi et al., 2020), Pcsk5 (Petra et al., 2022), and Ap1s2 (Woroniecka et al., 2011; Zhong et al., 2013; Gong et al., 2016), also showed similar patterns in ATAC and H3K27ac (Supplementary Figure S7). Nlrx1 (related to mitochondrial immunity and inflammation) (Moore et al., 2008; Tattoli et al., 2008; Arnoult et al., 2009; Nagai-Singer et al., 2019) showed significantly lower expression and ATAC-peaks in KO5 MMC (Supplementary Figure S8). This was in line with a more closed chromatin in KO5 MMC and in parallel significantly lower expression of Nlrx1, despite relatively high levels of H3K27ac. These results support the notion that the chromatin status dictates the expression of the TGF-β-regulated genes more than H3K27ac. Yet, other genes, such as Pai1, Ctgf, and Il11, showed high expression in KO5 MMC treated with TGF-β like WT, suggesting not all genes regulated by TGF-β are similarly regulated by lncMGC (Supplementary Figure S9A–S9C).

Interestingly, the promoter region of lncMGC showed ATAC-peaks in WT MMC even before treatment of TGF-β, but very low ATAC-peaks in KO5 MMC even after TGF-β treatment (Figure 6A and Supplementary Figure S10). TGF-β increased H3K27ac in WT but not in KO cells (Figure 6A). Therefore, an open chromatin status also appears critical for the expression of lncMGC.

Previously, we identified Smad and CHOP binding sites in the promoter of lncMGC (Kato et al., 2016). Thus, we determined H3K27ac enrichment at these transcription factor binding sites by ChIP-qPCR (Figures 6B–E). Interestingly, H3K27ac levels were increased at the Smad site and initiator (INR, TSS) in WT MMC treated with TGF-β, while no changes were detected in KO5 MMC, again confirming the loss of TGF-β response in the KO5 MMC (Figures 6B–E). TGF-β treatment of WT and KO5 MMC did not alter K27ac enrichment at the CHOP binding site, although levels were significantly lower in untreated KO cells vs. WT MMC. These results suggest that lncMGC regulates its promoter.

Given that SMARCA5 is an lncMGC-interacting protein, we performed ChIP assays to determine SMARCA5 enrichment at the Smad and CHOP sites in the lncMGC promoter. SMARCA5 enrichment was increased in WT MMC treated with TGF-β at the Smad site and INR (Figures 6F, H), but not in KO5 MMC even after TGF-β treatment, confirming our finding of increased K27ac at the lncMGC promoter in WT MMC treated with TGF-β but not in KO5 MMC (Figure 6A). At the CHOP site, SMARCA5 binding was decreased in KO5 MMC compared to WT MMC, and TGF-β treatment did not alter SMARCA5 change in either cell type (Figure 6G). Furthermore, the expression of lncMGC and miR-379, as well as the enrichment of H3K27ac at the Smad site in the lncMGC promoter, was significantly increased by TGF-β in siNC-treated control cells but significantly reduced in MMC treated with siSMARCA5 (Figures 6I–K). These results demonstrate that the chromatin around the lncMGC promoter is closed in KO cells and this is accompanied by lower H3K27ac levels and lower expression of lncMGC (and miR-379). These data suggest that lncMGC may regulate its own promoter and other target loci regulated by SMARCA5. Nucleosome remodelers such as SMARCA5 may be recruited to the lncMGC promoter through the interaction with lncMGC RNA to subsequently alter the chromatin structure.

Several transcripts are mapped close to the lncMGC locus, prompting a closer look at ATAC-peaks from Dlk1 to Rian (Figure 7A). Interestingly, several adjacent genes (Dlk1, Meg3, Rian, and Mirg) were significantly decreased in KO5 MMC compared to WT MMC (Figures 7B–E). Also, ATAC-peaks at these neighboring transcripts (Dlk1, Meg3, and Rian) were reduced in KO5 MMC compared to WT MMC, regardless of TGF-β treatment (Figure 7F). These results suggest that lncMGC KO affects the chromatin structures of neighboring genes. Similar to the lncMGC promoter, the chromatin of neighboring regions of lncMGC was open in WT MMC before TGF-β treatment but closed in KO5 MMC even after TGF-β treatment. In addition, the expression of neighboring genes (Dlk1, Meg3, Rian, and Mirg) was significantly decreased in WT MMC treated with siSMARCA5, although they were significantly increased by TGF-β in siNC control-treated cell (Figures 7G–J). These results support that SMARCA5 (opening chromatin) is essential to induce TGF-β-regulated genes including neighboring genes, and that its interaction with lncMGC RNA may enhance the opening of chromatin and the expression of neighboring genes and its promoter (Figure 7A). Hi-C and Virtual 4C data (http://3dgenome.fsm.northwestern.edu/virtual4c.php) support the interactions from the human DLK1 region to the lncMGC region (Supplementary Figure S11A, S11B). Overall, these results demonstrate the regulation of neighboring genes by lncMGC through DNA/DNA and DNA/RNA interactions.

As shown previously, not all DKD-related genes were regulated by lncMGC. To investigate the differences between regulated and unregulated lncMGC genes, we explored enriched motifs at ATAC-peaks and differentially expressed genes. Based on differential ATAC-peaks (Figures 8A–C), Smad binding sites were significantly higher in WT MMC treated with TGF-β (Figure 8A), in line with the fact that Smad proteins transduce TGF-β signaling (Massague et al., 2005; Kato and Natarajan, 2014; Meng et al., 2016; Kato and Natarajan, 2019). Smad motifs were also lower in KO5-TGF compared to WT-TGF (Figure 8B). This is in line with our observation that the promoter regions of key TGF-β-regulated genes were closed in KO5 MMC. Zinc finger (ZF), ZBTB26 (ZF and BTB domain 26), sites were increased in WT-TGF compared to WT-SD (Figure 8A). ZF, ZBTB32 (ZF and BTB domain 32), sites were less enriched in KO-TGF compared to WT-TGF MMC (Figure 8B). Interestingly, AT-rich sequences (ARID3b) were also lower in KO-SD vs. WT-SD (Figure 8C). Nucleosome remodelers have AT-rich interacting domains (ARIDs) (Becker and Hörz, 2002) and, thus, may be recruited to AT-rich sequences at ATAC-peaks. Based on DEG analysis (Supplementary Tables S1–S4), ZF sites and Smad motifs were enriched in certain genes that were increased in WT MMC treated with TGF-β (Supplementary Table S1). Interestingly, ZF motifs were also enriched in the downregulated genes in KO5 MMC compared to WT MMC (both in untreated and TGF-β conditions) (Supplementary Table S2–S4). These data suggest that the lncMGC-SMARCA5 complex may be preferentially recruited to the ZF, ARID, and Smad sites and facilitate chromatin opening and expression of the associated TGF-β-regulated genes (Figure 8D).

We also examined transcription factor (TF) motifs in the lncMGC DNA sequence and found ZF motifs (GGT/GGG/CCC/ACC repeats) and AT-rich motifs in both human and mouse lncMGC genes (Supplementary Figures S12A, S12B). Interestingly, the sequence that was deleted in lncMGC-KO5 vs. WT contains such important ZF and AT-rich motifs (Supplementary Figures S12C–S12E). This further supports the notion that such motifs in the lncMGC may be critical for recruiting SMARCA5 to specific binding sites (ZF and ARID) on target genes.