Doxycycline Changes the Transcriptome Profile of mIMCD3 Renal Epithelial Cells

Tetracycline-inducible gene expression systems have been used successfully to study gene function in vivo and in vitro renal epithelial models but the effects of the common inducing agent, doxycycline (DOX), on gene expression are not well appreciated. Here, we evaluated the DOX effects on the transcriptome of a widely used renal epithelial cell model, mIMCD3 cells, to establish a reference. Cells were grown on permeable filter supports in the absence and presence of DOX (3 or 6 days), and genome-wide transcriptome profiles were assessed using RNA-Seq. We found DOX significantly altered the transcriptome profile, changing the abundance of 1,549 transcripts at 3 days and 2,643 transcripts at 6 days. Within 3 days of treatment, DOX significantly decreased the expression of multiple signaling pathways (ERK, cAMP, and Notch) that are associated with cell proliferation and differentiation. Genes associated with cell cycle progression were subsequently downregulated in cells treated with DOX for 6 days, as were genes involved in cellular immune response processes and several cytokines and chemokines, correlating with a remarkable repression of genes encoding cell proliferation markers. The results provide new insight into responses of renal epithelial cells to DOX and a establish a resource for DOX-mediated gene expression systems.


INTRODUCTION
Gene expression systems utilizing drug-induced trans-activation provide the means to conditionally investigate gene function in a temporal manner. They can be used to activate expression of a target gene for study, or to induce Cre-recombinase for conditional knockout cell of floxed alleles. The tetracycline-inducible gene expression system has proven especially popular. Doxycycline (DOX), a synthetic derivative of tetracycline, is widely used in these systems (Kistner et al., 1996;Das et al., 2016) because of its long half live and low cell toxicity. However, effects of DOX on a global gene expression and cellular processes in renal epithelial cell models are not known.
The absence of DOX effects on global gene expression in Saccharomyces cerevisiae (Wishart et al., 2005) initially suggested DOX might be inert. However, several more recent studies revealed DOX can change gene expression in mammalian tissues. For example, in ex vivo studies with surgically removed pterygia tissue from the eye, RNA sequencing (RNA-Seq)-based transcriptomic analysis (Larrayoz et al., 2012) revealed DOX affected the expression of mitochondrial genes, the ER stress cascade, growth factors, interleukins, cell cycle regulators, integrins, and components of the extracellular matrix. In rat aortic tissue (Lu et al., 2017), DOX was found to change the expression of other sets of genes, and these were mainly enriched in pathways that control neutrophil chemotaxis, chronic inflammatory responses, and cellular responses to mechanical stimuli and negatively regulate apoptotic processes. In vivo effects of DOX have been documented in studies to characterize the trans-activator (rtTA2) system in the mouse liver (Reboredo et al., 2008), and DOXsensitive genes were found to enriched in cellular pathways related to cell growth and death and mitochondrial electron transport (Reboredo et al., 2008). Although the sample size still remains small, these few studies question the extent to which DOX may have generalizable responses; the heterogenous nature of responses suggest DOX action may be specifically governed by cell type or experimental condition. More studies are required with other cell types to identify common pathways. It remains unknown if DOX affects gene expression in renal epithelial cells.
Here we focus on characterizing the DOX response in the renal epithelial cell line, mIMCD3. The cells have been engineered to include a tetracycline-inducible gene expression system (Schlimpert et al., 2018;Lashhab et al., 2019;Kang et al., 2019), and the model has proven popular for studying genes in a renal epithelial environment. Here we examine how DOX affects the transcriptome profiles of mIMCD3 cells with RNA-Seq. The data should provide an informative resource for future studies with the tetracycline-inducible mIMCD3 cell line and similar in vitro renal epithelial models.

Cell Culture
Mouse kidney epithelial mIMCD3 cells derived from the inner medulla of a simian virus SV40 transgenic mouse were obtained from Maryland Polycystic Kidney Disease Research and Translation Core Center at the University of Maryland, and documented to be mycoplasma free. Cells were grown in T75 flasks with the RenaLife Epithelial Basal Medium (LM-0010, LIFELINE CELL TECHNOLOGY) supplemented with a RenaLife LifeFactors Kit (LS-1048, LIFELINE CELL TECHNOLOGY) and 5% FBS. This culture medium, which contains a nutrient blend of amino acids, vitamins, organic and inorganic supplements and salts, growth factors (0.5 µg/mL Insulin, 1 µM Epinephrine, 0.1 µg/mL Hydrocortisone, 10 nM Triiodothyronine, 10 ng/mL EGF, 5 µg/mL Transferrin, 2.4 mM L-Alanyl-L-Glutamine), and antibiotics (30 µg/mL Gentamicin, 15 ng/mL Amphotericin B), has become the choice for in vitro studies in the Maryland Cell engineering core of the NIH U54 funded Polycystic Kidney Disease Research Consortium. It was used here to provide a reference for on-going and future transcriptome profiling studies in the core. For RNA-Seq studies, cells were plated permeable filters (Polyester membrane) of a 6-well transwell plate (#3450, CORNING) and grown to confluence (10 days) before doxycycline treatment. Doxycycline (2 µg/mL) or vehicle (DMSO) added to both sides of the permeable filter and changed daily for 3 or 6 days.

Total RNA Isolation and RNA Sequencing-Based Transcriptome Profiling
Cells were lysed in Trizol reagent (15596018, Invitrogen). Total RNA was isolated from the Trizol lysate using Direct-Zol RNA Miniprep plus kit (R2070, ZYMO RESEARCH) and eluted in RNase-free water, and RNA concentration was measured using Qubit TM RNA HS Assay Kit (Q32852, Invitrogen). To enrich mRNA, 1-3 µg of total RNA was applied to oligo dT-based mRNA isolation using NEBNext R Poly(A) mRNA Magnetic Isolation Module (E7490, NEW ENGLAND BioLabs Inc.) according to manufacturer's instructions. mRNA (20 ng) was used to create the cDNA libraries, using NEBNext R Ultra II Directional RNA Library Prep Kit for Illumina (E7760, NEW ENGLAND BioLabs Inc.) and NEBNext R Multiplex Oligos (E7335; E7500, NEW ENGLAND BioLabs Inc.). cDNA libraries were sequenced on the Illumina HiSeq 4,000 platform. Sequence reads (2 × 75 bp, paired-end) were aligned on Ensembl genome GRCm38p6 using STAR (2.6.0c).

Differential Expression Analysis and Bioinformatic Analysis
To identify differentially expressed genes between vehicle-and doxycycline-treated cells, transcript abundance was quantified using salmon (Patro et al., 2017). Differential expression (DE) analysis was carried out using edgeR (Robinson et al., 2010) on R (3.6.0). Low abundant genes with CPM (Counts Per Million) less than 1 were removed from the data set for the DE analysis. Significance of DE was determined using a modified statistical test (edgeR "glmTreat") with a threshold of expression changes above 20% at FDR < 0.05 (Chen et al., 2016). Plots were generated using R package ggplot2. Pathway enrichment analysis of DE genes was carried out using Gene Ontology (Biological process) on Metascape platform 1 (Zhou et al., 2019). The GO term analysis was performed using NaviGO (Wei et al., 2017).

Statistics
Significance of DE was evaluated using edgeR and p-values were corrected using the Benjamini-Hochberg method. For significance of gene enrichment in the pathway analysis, q < 0.05 was considered as significant enrichment.

Data Availability
All fastq files and a raw count file from RNA-Seq were deposited in GEO (GSE171573 2 ).

Transcriptome Profiles of mIMCD3 on Different Culture Environments
RNA-Seq analysis was performed to provide a transcript database of fully polarized mIMCD3 cells on permeable filter supports 1 http://metascape.org/ 2 https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?&acc=GSE171573 FIGURE 1 | Transcriptomic characteristics of renal epithelial cell line mIMCD3 in different culture conditions. (A) Transcriptomic correlation was assessed in mIMCD3 cells grown on the permeable filter supports for 3 and 6 days. (B,C) Transcriptome profiles of mIMCD3 grown on the permeable filter support and the solid support were compared. The transcriptome dataset of mIMCD3 cells grown on the solid support was obtained from GSE97770 (Chan et al., 2018). (D-G) Box plots of gene expression associated with epithelial cell, mesenchymal cell, tight junction formation, and extracellular matrix (ECM) in mIMCD3 cells grown on the solid support (Chan et al., 2018) and the permeable filter support.
in the RenalLife medium which includes growth factors and antibiotics. For these studies, the cells were grown to confluence on the permeable filter supports for 10 days, and then were treated with DMSO or DOX in the RenalLife medium for 3 or 6 days.
To establish the baseline, we first examined the control cells, treated with DMSO. We found a high correlation between the transcriptome of cells treated with DMSO at 3 and 6 days (Pearson correlation: 0.9956, Figure 1A), consistent with stable gene expression and lack of transient DMSO responses. Based comparison of normalized expression values (transcripts per million, TPM), the transcriptome exhibited a high correlation with native mIMCD3 cells grown on plastic solid supports (GSE97770) as reported by Chan et al. (2018) (Figures 1B,C). Whether grown on permeable supports or plastic solid supports, mIMCD3 equally express the conventional epithelial cell markers (Jedroszka et al., 2017), such as Tjp1, Krt18, Dsp, Muc1, and Sdc1 ( Figure 1D). Other extracellular matrix components, Col4a2 and Sparc, were also more abundantly expressed in the mIMCD3 cells grown on the solid support compared to mIMCD3 cells on the filter support ( Figure 1G). Interestingly, mesenchymal cell marker genes, namely Cdh2, Fn1, Sparc, and Vim, were also more abundant in mIMCD3 cells grown on the solid support compared to mIMCD3 cells grown on the filter support ( Figure 1E). By contrast, in mIMCD3 cells grown on the filter supports, tight junction proteins, Cldn4, Cldn7, and Epcam, were more abundant than in cells on the solid support ( Figure 1F). Thus, transcriptomic signature of mIMCD3 cells is most compatible with a differentiated epithelial state when cells are grown on permeable supports compared to cells grown on plastic supports. Yet the transcriptome profiles of mIMCD3 cells grown on permeable and non-permeable are more similar than different. Interestingly, we did not identify the enrichment of specific transcripts of medullary collecting duct in this cell line. Nevertheless, the dataset of comprehensive transcriptome profile in mIMCD3 provides an informative resource for further utilization as the in vitro kidney epithelial cell model. To expand understanding of intracellular signaling pathways in mIMCD3 cells grown on solid supports (Valkova and Kultz, 2006;Chan et al., 2018) compared to cells grown on filters, we classified transcriptome profile into genes encoding transcription factors, ion channels, transporters, and G protein-coupled receptors (GPCRs) in Tables 1-4 (Supplementary Table 1). Many of the listed genes have been identified in previous studies. For example, Hnf1b (hepatocyte nuclear factor-1 beta, TPM: 43.0) (Aboudehen et al., 2017), Pax2 (Paired box 2, TPM: 20.8) (Torban et al., 2000;Cai et al., 2005), and Egr1 (Early Growth Response 1, TPM: 49.7) (Cohen et al., 1994(Cohen et al., , 1996 are transcription factors reliably expressed in the mIMCD3 cells (Supplementary Table 1, "Transcription factors"). In addition to known expression of transcription factors, expression of transcription factors that could be involved in the kidney nephron development, such as Emx2, Pax8, Tfap2a, Hmga2, Hmgb2, and Hoxa11, were identified (Ribes et al., 2003;Schwab et al., 2003;Chambers et al., 2019).

Doxycycline-Responsive Transcriptome Changes in mIMCD3 Cells
To characterize the DOX response in mIMCD3 cells, we compared the transcriptomes of DOX-treated mIMCD3 cells and DMSO-treated mIMCD3 cells (Figure 2 and Supplementary Table 2). We found DOX changed the gene expression profile, reflecting a change in abundance of 1,662 genes at 3 days ( Figure 2A) and 2,858 genes at 6 days of treatment ( Figure 2B). A total of 1,157 genes were consistently changed at both times. Downstream analysis identified GO Biological Processes (GOBPs) that are enriched in DOX-treated cells (Figure 2C). The NaviGO based GO term association analysis, which classifies similarity within GO term hierarchies, revealed three major clusters of GOBPs: (1) cell proliferation/differentiation; (2) signal transduction; and (3) immune responses ( Figure 2D). In the cells treated with DOX for 3 days, differentially expressed genes were enriched for signal transduction pathways and cell proliferation/differentiation (Figures 2C,D). In particular, this involved significant changes of genes associated with ERK, cAMP, and Notch signaling pathways, known to mainly involve cell proliferation and development processes (Figure 2C). Differentially expressed genes from the 6 days-dataset were also especially enriched in biological processes associated with cell proliferation/differentiation but were also enriched in immune response pathways (Figures 2C,D). As shown in the heatmap (Figure 2E), expression of genes involved in cell proliferation/differentiation processes and immune responses were mostly decreased at both times. The results indicate that DOX attenuates gene expression associated with cell proliferation and immune responses. Full list of genes associated with these pathways are provided in Supplementary Table 3.
Examination of genes encoding downstream factors in signal transduction processes, including transcription factors and kinases, revealed alterations in three different signaling pathways (cAMP, ERK, Notch). Remarkable downregulation of several transcription factors and protein kinases was found in DOXtreated cells (Figure 3A). Especially, Hnf1b (Chung et al., 2017), Egr1 (Sukhatme et al., 1988), Tead1 (Zhao et al., 2008), and members of the Fox family (Kume et al., 2000;Aschauer et al., 2013), which have been proposed previously as key players in epithelial cell proliferation. Additionally, several protein kinases downregulated in response to DOX, namely Tgfbr2 (LeBleu et al., 2013), Kit (Gomes et al., 2018), and protein kinase A (Amsler et al., 1991), are highly related to regulation of epithelial cell function ( Figure 3A). Furthermore, the effects of DOX to decrease the expression of genes involved in each step of the cell cycle and cell cycle progression suggest that DOX may have proclivity to inhibit cell proliferation ( Figure 3B). The transcriptome dataset showing the attenuated cell cycle progression also exhibited that several genes known as cell proliferation markers were consistently downregulated by DOX treatment at 6 days (Figure 4). The result confirmed that DOX treatment suppresses cell proliferation. In addition to cell proliferation markers, DOX treatment for 3 and 6 days induced significant changes of immune responseassociated cellular pathways (Figure 2), largely reflecting DOXresponsive reduction in cytokine production at day 3 and comprehensive alteration of cellular inflammatory response at day 6. In particularly, we identified several chemokines that were significantly downregulated at 3 days after DOX treatment from the literature-based gene sets of immunologic mediators (Commins et al., 2010 ; Figure 4), corresponding the pathways involved in the repression of epithelial cell proliferation, which are associated with epithelial responses to cytokines (Stadnyk, 1994).

DISCUSSION
The mIMCD3 cell line, an inner medullary collecting duct (IMCD) cell line derived from a simian virus 40 (SV40) T-antigen transgenic mice, has been widely used as an in vitro model to FIGURE 2 | DOX-responsive transcriptomic changes in mIMCD3 cells. (A,B) Identification of transcriptomic changes in mIMCD3 cells treated with DOX for 3 and 6 days using RNA-Seq-based differential expression analysis. Significantly changed genes (FDR < 0.05 over 20% expression change threshold) were labeled as green in volcano plots and used further downstream analysis. (C) Gene Ontology (GO) biological processes significantly enriched by DOX-responsive genes. Cellular processes (GO Biological Process) were identified from significantly changed genes by DOX treatment for 3 and 6 days. To avoid sample-source bias, whole transcriptome was used as a background gene set. Significant enrichment was considered as q < 0.05. (D) Bubble chart view of the GO term association using NaviGO. In the plot, the X-axis is the Resnik semantic similarity score (RSS) and the Y -axis chosen is Lin's similarity score (LSS). GO terms are colored according to time point when DE genes are found, day 3, pink; day 6: green; and day 3 and 6: blue. The number of GO terms are listed in (C). (E) Heatmap of gene sets associated with "cell proliferation/differentiation" and "immune responses." Full list of genes associated with GO terms are provided in Supplementary Table 4. study renal epithelial physiology and cell biology (Rauchman et al., 1993;Yu et al., 2013;Flannery et al., 2015;Siroky et al., 2017). With inclusion of a DOX-inducible gene expression system, the cell line has become an invaluable model to study gene function in renal epithelial background. Here, we applied next-generation sequencing technology to characterize how the mIMCD3 transcriptome is affected by DOX with the goal of creating a database to guide interpretation of these studies.
Appropriate differentiation of epithelial cells and development of apical-basolateral polarity in vitro is commonly believed to require growth on a permeable surface. Comparison of the mIMCD3 transcriptome of cells grown on permeable polyester (PE) filter membrane surface, as defined here, with the published the transcriptome profile of mIMCD3 cells grown on plastic (Chan et al., 2018) lends credence to this idea. Although we found transcriptome profiles of mIMCD3 cells grown on filters exhibit a surprising degree of similarity with the transcriptome profile of cells grown on a solid surface (Reboredo et al., 2008), the profile of cells grown on filters is consistent with a more differentiated state. We found genes encoding key junctional stability components (Cldn4, Cldn7, Epcam) are induced when cells are grown on filters, whereas mesenchymal markers and ECM components were less abundant in filter-grown cells. Cldn4 (Claudin-4) and Cldn7 (Claudin-7) are among claudin family proteins that are specifically expressed in the loop of Henle (LOH) and distal nephron (Leiz and Schmidt-Ott, 2019), dictating the unique pericellular solute permeation profile and the electrical resistance of these nephron segments. Epcam (Epithelial Cell Adhesion Molecule) is best known as a marker of epithelial oncogenesis, but it is highly expressed in the LOH FIGURE 4 | DOX-responsive cytokines and cell proliferation markers. Significantly changed genes encoding cytokines including chemokines were identified and listed in the plot. In addition, expressional changes of genes previously known as cell proliferation markers were listed in the plot. and collecting duct, where its natural role in cell adhesion and morphogenesis may determine the differentiated state of these nephron segments by interaction with Claudin-7 to negatively regulate epithelial migration by inhibiting ERK and actomyosin contractility (Barth et al., 2018). The increased abundances of ECM components may identify a gene expression signature of solid support grown cells, reflecting a futile cellular response to establish a permeable basement membrane, and thereby influence cell proliferation and differentiation (Matter et al., 2005;Ichikawa-Tomikawa et al., 2011;Bonnans et al., 2014;Diaz-Coranguez et al., 2019). Supporting this idea, the upregulated ECM are ubiquitous components of most basement membranes, including members of type IV collagen family proteins, Col4a1 and Col4a2, and the Secreted Protein Acidic and Rich in Cysteine (SPARC, aka Osteonectin).
We found the subtle growth substrate-dependent changes occur against an otherwise, remarkably stable mIMCD3 transcriptome profile. Constancy of the gene expression profile may explain why the phenotypic properties of the cell line have been so stable over time and numerous different studies. Comparison of the mIMCD3 transcriptome to the single cell RNA profiles of mouse cell kidney (Ransick et al., 2019) indicates that stability has come at the cost of lost nephron cell definition, however. The transcriptome profile of mIMCD3 does not uniquely resemble transcriptomes of any cell type in the kidney.
Instead, it is compatible with a generic renal epithelial cell model, presumably reflecting a dedifferentiated state.
Tetracycline drugs, like Doxycycline (DOX), are widely used in eukaryotic cell gene expression systems, in part, because they have been assumed to have limited toxicity profiles. They inhibit protein synthesis in a broad spectrum of bacteria by binding to rRNAs of ribosomal complexes and double-strand RNAs (Chukwudi, 2016;Chukwudi and Good, 2016), but do not usually accumulate enough in mammalian cells to affect protein synthesis. We found DOX, at least at the higher concentrations commonly used in vitro cell models to drive teto gene expression, has a larger than anticipated toxic profile. Differentially expressed genes in response to DOX were especially enriched in the suppression of cell proliferation processes, including a decrease in cell cycle progression genes (Figures 3, 4). DOX also has been reported to suppress cell proliferation in epithelial-derived cancers (Fife et al., 1998), providing reason to speculate that tetracycline derivatives might be developed as anticancer therapeutics (Kroon et al., 1984;van den Bogert et al., 1985van den Bogert et al., , 1986Mortison et al., 2018). In human epithelial cell lines, DOX reduces cell proliferation, associated with alterations in mitochondrial function (Sourdeval et al., 2006;Chang et al., 2010;Ahler et al., 2013).
Although future studies will be required to unravel the mechanisms by which DOX suppresses cell proliferation, the change in the mIMCD3 transcriptome profile provides important insights. Decreased expression of cytokines and chemokines (Colombo et al., 2018;Koga et al., 2019), together with the attenuation of intracellular signal transduction pathways (cAMP, ERK, and Notch) that are known to regulate cell proliferation (Yamaguchi et al., 2000;Stork and Schmitt, 2002;Kim and Shivdasani, 2011) provide reason to suggest that DOX suppresses the production of autocrine cell growth factors that are required for mIMCD3 proliferation. Additionally, DOX activated the interferon-beta pathway, involving guanylate-binding proteins (GBPs) and interferon activated genes (Lubeseder-Martellato et al., 2002;Klamp et al., 2003;MacMicking, 2004), which is known to inhibit cell proliferation and differentiation (Hertzog et al., 1994;Garrison et al., 1996;Vitale et al., 2006). It will be interesting to learn if either of these pathways can be manipulated to reduce the anti-proliferation effects of DOX. It will be important to corroborate that changes in protein abundance track changes in transcript levels.
The effects of DOX to rapidly induce an anti-proliferative transcriptomic signature in mIMCD3 cells should not be taken as a reason to discontinue use of the otherwise highly effective DOX-based gene induction system. Because it seems likely that all induction agents will have some off-target effects, it more practical to develop strategies to work around the off-target responses. Washing out DOX after gene induction is an obvious remedy. However, future studies will be required to determine the longevity of the DOX response, after withdrawal.
Growth and differentiation of cells in vitro can be profoundly influenced by the cell culture medium, nutrients, growth factors, hormones, and other supplements. They were not varied in the present study, and thus should not contribute to differences in the transcriptome profiles between vehicle and DOX treated cells. However, the basal transcriptome profile is likely to be influenced by the supplements in the media. Moreover, potential synergistic actions between DOX and any one of the many supplements is unknowable. Consequently, caution should be exercised in generalizing our results with other studies that use different culture media.
In summary, we have assessed the transcriptomic response to DOX in mIMCD3 at genome-wide resolution. The data base provides an atlas to guide future interpretation of studies using the DOX-inducible gene expression system in renal epithelial cells.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary Material.