AM251 Suppresses Epithelial-Mesenchymal Transition of Renal Tubular Epithelial Cells

Epithelial-mesenchymal transition (EMT) of renal tubular epithelial cells is one of the causative mechanisms of kidney fibrosis. In our study, we screened lipophilic compounds using a lipid library including approximately 200 lipids to identify those that suppressed EMT induced by a transforming growth factor (TGF)-β1 stimulus. Initial screening was performed with the immortalized HK-2 renal tubule epithelial cell line. The most promising compounds were further tested in RPTEC primary renal tubule epithelial cells. We found that the synthetic lipid AM251 suppressed two hallmark events associated with EMT, the upregulation of collagen 1A1 (COL1A1) and downregulation of E-cadherin. Though AM251 is known to act as an antagonist for the cannabinoid receptor type 1 (CB1) and an agonist for the G protein-coupled receptor 55 (GRP55), the suppression of EMT by AM251 was not mediated through either receptor. Microarray analyses revealed that AM251 inhibited induction of several EMT transcription factors such as SNAIL1, which is the key inducer of EMT, and the AP-1 transcription factors FOSB and JUNB. Activation of SMAD2/3 and p38 mitogen-activated protein kinase (MAPK) was inhibited by AM251, with greater inhibition of the latter, indicating that AM251 acted upstream of SMAD/p38 MAPK in the TGF-β signaling pathway. Our findings regarding the effects of AM251 on the TGF-β signaling pathway may inform development of a novel therapeutic agent suppressing EMT, thus preventing kidney fibrosis.


Introduction
Progression of chronic kidney disease (CKD) can cause renal fibrosis, which impairs kidney function and can lead, in the worst-case scenario, to renal failure [1]. In renal fibrosis, myofibroblasts overproduce extracellular matrix proteins such as collagens and fibronectin [2,3]. Myofibroblasts are derived from multiple sources [1], such as by activation of interstitial fibroblasts and pericytes [4], conversion of renal tubular epithelial cells through the epithelial-mesenchymal transition (EMT) [5,6], conversion of renal endothelial cells through endothelial-

Cell Culture
The immortalized proximal tubule epithelial cell line HK-2 (American Type Culture Collection, Manassas, VA, USA) was grown in serum-free keratinocyte medium (Thermo Fisher Scientific, Waltham, MA, USA) containing 5 ng/ml recombinant human epidermal growth factor (Thermo Fisher Scientific), 0.05 mg/ml bovine pituitary extract (Thermo Fisher Scientific), 100 units/ml penicillin, and 100 μg/ml streptomycin. The primary renal proximal tubule epithelial cells RPTEC (Lonza, Basel, Switzerland) were grown in the renal epithelial growth media (REGM; Lonza) containing REGM Single Quots (Lonza). REGM Single Quots includes a mixture of supplements and growth factors (hydrocortisone, epidermal growth factor, fetal bovine serum, epinephrine, insulin, triiodothyronine, transferrin, gentamicin and amphotericin B). RPTEC cells were used at passage 3 or 4. EMT was induced by replacing the growth medium with REGM lacking REGM Single Quots but containing 2 ng/ml TGF-β1 (R&D Systems, Minneapolis, MN, USA). However, in the experiment involving TGF-β1 pretreatment, REGM containing REGM Single Quots and 2 ng/ml TGF-β1 was used to prevent cytotoxicity.

RNA Interference
RNA interference was carried out using RNAiMAX (Thermo Fisher Scientific) by the reverse transfection method, according to the manufacturer's instructions. The Silencer 1 Select Negative Control No. 1 siRNA (Thermo Fisher Scientific) and siRNAs for the CB1 gene CNR1 (s3260 and s3262; Thermo Fisher Scientific) were used at 20 nM.

Screening for EMT-inhibiting Compounds
HK-2 cells were plated at 2 × 10 4 cells per well on poly-D-lysine-coated 96-well plates overnight. TGF-β1 (2 ng/ml) and one of the compounds from SCREEN-WELL 1 Bioactive lipid library (10 or 1 μM; 1/100 of the original concentrations in the library; Enzo Life Sciences, Farmingdale, NY, USA) were incubated with the cells for 24 h. Expression of collagen 1A1 (COL1A1) and, as a housekeeping gene, peptidylprolyl isomerase (PPIA) was then measured by real-time RT-PCR as described below.

Real-Time Quantitative RT-PCR
To detect EMT inhibiting compounds, cells were lysed using Fastlane Cell SYBR Green Kit (Qiagen, Hilden, Germany) and subjected to real-time RT-PCR using QuantiTect SYBR Green RT-PCR Kit (Qiagen) and primers (for COL1A1, COL1A1-F and COL1A1-R; for PPIA, PPIA-F and PPIA-R; Table 1), according to the manufacturer's instructions.
In other assays, total RNA was isolated from cells using the RNeasy Plus Micro Kit (Qiagen) and converted to cDNA using the High Capacity RNA-to-cDNA Kit (Thermo Fisher Scientific). The resulting cDNAs were then subjected to real-time RT-PCR using SYBR Premix Ex Taq II (Takara, Shiga, Japan) on the 7500 Real-Time PCR System (Thermo Fisher Scientific). Primers used are listed in Table 1 (-F, forward primer; -R, reverse primer).

Primer
Nucleotide sequence

Microarray
Total RNA was prepared using the RNeasy Plus Micro Kit (Qiagen). RNA quality was evaluated using Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). Total RNA samples with an RNA integrity number >8.0 [25] were converted to cRNAs and labeled with Cy3 using a Low Input Quick Amp Labeling Kit, One-Color (Agilent Technologies). Cy3-labeled cRNA was hybridized to a SurePrint G3 Human GE Microarray G4851B (Agilent Technologies) at 65˚C for 17 h using a Gene Expression Hybridization Kit (Agilent Technologies). The microarray slide was washed, dried, and scanned with a SureScan microarray scanner G2600D (Agilent Technologies). The raw microarray image data were read and processed by Feature Extraction software 11.5.1.1 (Agilent Technologies). Data were analyzed using Gene-Spring GX 12 software (Agilent Technologies). The 95 th percentile signal value was used to normalize microarray signals for inter-array comparisons. Genes with !2-fold change (increase or decrease), upon treatment of cells with TGF-β1, were subjected to hierarchical clustering analyses and enrichment pathway analyses. Hierarchical clustering analyses were performed by centroid distance method using GeneSpring GX 12.5. Enrichment pathway analyses were conducted using the MetaCore gene regulatory network database (Thomson Reuters, New York, NY, USA).

Identification of AM251 as an EMT-inhibiting Compound
We performed screening to identify lipophilic compounds that suppressed EMT induced by TGF-β1. For this screening, we used the renal tubular epithelial cell line HK-2 and a lipid library consisting of 196 natural and synthetic lipids. As an indicator of EMT, expression of type I collagen COL1A1 was examined. Expression levels of COL1A1 in cells treated with TGF-β1 were more than 10-fold higher than in those without treatment. Those compounds that reduced COL1A1 expression to 50% were regarded as hit compounds (Fig 2A, red circles). We also examined expression levels of peptidylprolyl isomerase (PPIA) to monitor cytotoxicity of the library compounds. Compounds causing a !4-fold decrease in PPIA expression were excluded from the hit compounds (Fig 2A, black circles). We obtained nine potential hit compounds, 1-stearoyl-2-arachidonoyl-glycerol (SAG), docosatrienoic acid (DTA), carbaprostacyclin (CA), ciglitazone (CI), 24, 25-dihydroxyvitamin D 3 (DHVD3), N-palmitoyl dopamine (PALDA), N-arachidonoyl dopamine (NADA), N-oleoyl dopamine (ODA), and AM251. Reproducibility testing by real-time RT-PCR confirmed that all except SAG, which caused only 20% reduction in TGF-β1-induced COL1A1 expression, met the criteria for hit compounds ( Fig 2B). Next, the remaining eight hit compounds were subjected to secondary screening in primary renal tubular epithelial cells (RPTEC). Four compounds, DTA, CA, CI, and DHVD3, exhibited little to no EMT suppressing activity ( Fig 2C). Three compounds, PALDA, NADA, and ODA, exhibited high toxicity (Fig 2D), though they also caused significant EMT suppression. Only AM251 exhibited EMT suppressing activity without toxicity and was therefore used for further analyses. The structure of AM251 is shown in Fig 2E. At the first step of EMT, cell-cell contacts between epithelial cells were disassembled. Consistent with this, downregulation of the component of adherens junctions, E-cadherin, is a well-known indicator of EMT [6,23]. TGF-β1 treatment indeed decreased E-cadherin protein levels to 34.2% of those in control cells (Fig 3A and 3B). AM251 caused an increase in E-cadherin levels both in the presence and absence of TGF-β1. Similar results were obtained for mRNA levels of the E-cadherin gene (CDH1) examined by real-time RT-PCR ( Fig 3C). Thus, AM251 reversed the downregulation of E-cadherin caused by TGF-β1.
In the experiments described above, cells were treated with TGF-β1 and AM251 for 24 h and then subjected to gene expression analyses. We also examined effects of longer AM251 treatment times (48, 72 and 96 h) on TGF-β1-induced COL1A1 expression. Cell toxicity was also evaluated by measuring expression of the housekeeping gene PPIA. AM251 strongly suppressed COL1A1 induction at all treatment times, although there was some cytotoxicity with TGF-β1 and AM251 with the 72 and 96 h treatment conditions (Fig 4A and 4B). We next examined whether AM251 could suppress COL1A1 induction caused by TGF-β1 pretreatment for 96 h. Treatment with AM251 for 24 h following TGF-β1 pretreatment strongly suppressed COL1A1 expression (Fig 4C and 4D). These results indicate that the suppression by AM251 was sustained for long periods and was effective even when COL1A1 expression had already been induced.

AM251 Suppresses EMT Independent of CB1 or GRP55
AM251 is known as an antagonist of cannabinoid receptor type 1 (CB1) or an agonist of the G protein-coupled receptor 55 (GPR55) [26,27]. Both CB1 and GPR55 are known as cannabinoid receptors [27,28,29] and are activated by exogenous plant cannabinoids, as well as by the endocannabinoids 2-arachidonoylglycerol and anandamide [27,28,29]. GPR55 is also activated by lysophosphatidylinositol [30]. To examine whether the effects of AM251 on EMT suppression were mediated through these receptors, we first measured mRNA levels of the CB1 gene (CNR1) in RPTEC cells. Real-time RT-PCR revealed that CNR1 mRNA levels were decreased in cells treated with TGF-β1 ( Fig 5A). When cells were treated with anandamide (ANA), the agonist of CB1 and GPR55, expression of COL1A1 mRNA was unchanged in either the presence or absence of TGF-β1 (Fig 5B). We next performed CB1 knockdown analyses using two independent siRNAs specific for the CB1 gene CNR1 (siCB1-1 and siCB1-2). Both siRNAs significantly reduced CNR1 mRNAs levels (to 3% by siCB1-1 and to 25% by siCB1-2) (Fig 5C). However, these siRNAs had no effect on TGF-β1-induced COL1A1 expression, either in the presence or absence of AM251 (Fig 5D).
Expression of GRP55 mRNA was slightly increased in cells treated with TGF-β1 (Fig 6A). When cells were treated with the selective GRP55 antagonist CID16020046 (CID) [31] at increasing concentrations, the expression of COL1A1 induced by TGF-β1 was not affected at any concentration, with or without AM251 (Fig 6B). From these results, we conclude that the EMT suppressing activity of AM251 was not mediated through CB1 or GRP55.
To compare gene expression patterns among cells subjected to different conditions, we performed hierarchical clustering analyses using genes increased by !2-fold upon treatment with TGF-β1 (Fig 7B). The gene expression pattern under untreated conditions (AM251(−) TGF-β1(−)) was most similar to that with only AM251 treatment (AM251(+) TGF-β1(−)). This result suggests that AM251 treatment did not globally affect gene expression. The gene expression pattern under TGF-β1 treatment conditions (AM251(−) TGF-β1(+)) differed the most from that under the untreated conditions, indicating that many genes observed were under Values are means ± SD of the ratio of COL1A1 to PPIA mRNA levels, expressed relative to the ratio in the control (no treatment) (A) or PPIA mRNA levels relative to the control (no treatment) (B), from three independent experiments. Statistically significant differences from the control (A, without AM251; B, no control of TGF-β1 signaling. Co-treatment of cells with AM251 and TGF-β1 caused suppression of many of the TGF-β1-induced gene expression changes, that is, the gene expression pattern of AM251(+) TGF-β1(+) was more similar to that of AM251(−) TGF-β1(−) than AM251(−) TGF-β1(+).
We next performed pathway analyses using the MetaCore gene regulatory network database, in which approximately 1,500 pathways are registered, to reveal which pathways were treatment) are indicated (** P < 0.01, Student's t-test). (C and D) RPTEC cells were incubated with 2 ng/ml TGF-β1 for 96 h in REGM medium containing REGM Single Quots and incubated with 2 ng/ml TGF-β1 and/or 10 μM AM251, as indicated, for the following 24 h in REGM medium. Total RNA was prepared and subjected to real-time RT-PCR to measure COL1A1 and PPIA. Relative mRNA levels of COL1A1 (C) and PPIA (D) were determined as described for (A) and (B), respectively. doi:10.1371/journal.pone.0167848.g004 EMT-or TGF-β-related pathways (Fig 7C). Others in this list included a pathway involved in adhesion and that involved in cytoskeleton remodeling. Because changes in adhesion and the cytoskeleton are induced by EMT, these pathways can also be considered to be EMT-related. These results suggest that AM251 did not affect pathways globally, but instead affected EMT/ TGF-β-related pathways relatively specifically.
We next confirmed the results of the microarray analyses by real-time RT-PCR. Expression of the key transcription factor involved in EMT, SNAIL1, was increased 6.5-fold in cells treated with TGF-β1, whereas AM251 suppressed its expression to nearly basal levels (Fig 9A). SNAIL2 expression was also induced by TGF-β1, albeit to a lesser extent than SNAIL1 (Fig 9B). AM251, similarly, reversed SNAIL2 expression to basal levels ( Fig 9B). The AP-1 components (JUNB and FOSB), transcription factors upstream from SNAIL1, were increased by 5-6 fold in cells treated with TGF-β1. AM251 treatment decreased their expression to about 50%  Fig 9C and 9D). The expression level of the TGF-β receptor TGFBR1 was increased 2.9-fold by TGF-β1 treatment and reduced to 79% by AM251 (Fig 9E). TGF-β1 treatment caused a large increase in expression of TGF-β family members TGFB2 (5.7-fold) and TGFB3 (4.4-fold); AM251 almost completely reversed these effects (Fig 8F and 8G). Thus, the results of microarray analyses are consistent with those of real-time RT-PCR.

AM251 Inhibits Activation of the SMAD2/3 and p38 MAPK Pathways
To determine the site of action of AM251, we next focused on the most upstream components of TGF-β signaling. TGF-β signaling begins with activation of SMAD2/3 and MAPKs by phosphorylation (Fig 1) [16,17,18,19,20]. Microarray analyses indicated that AM251 did not cause downregulation of SMAD2, SMAD3, MAPK11 (p38β), MAPK12 (p38γ), MAPK13 (p38δ), or MAPK14 (p38) (Fig 10A). Because SMAD2/3 and MAPKs are often regulated at the level of activity rather than at the level of expression, we next examined the activation status of SMAD3 and p38 MAPK by detecting their phosphorylated (active) forms. TGF-β1 treatment caused a 5.7-fold increase in phospho-p38 MAPK levels (Fig 10B and 10C). Simultaneous treatment of cells with AM251 decreased phospho-p38 MAPK levels to nearly half of this stimulated value. SMAD3 phosphorylation was increased by 27-fold in cells treated with TGF-β1  (Fig 10B and 10D). AM251 reduced phospho-SMAD3 levels, but only slightly (an approximately 10% decrease). Therefore, AM251 inhibited activation of both SMAD2/3 and p38 MAPK pathways, though its effect on p38 MAPK was greater.

Discussion
In our study, we screened lipid compounds to identify those that suppressed EMT induced by a TGF-β1 stimulus in renal tubular epithelial cells. We identified AM251, a known CB1 antagonist and GPR55 agonist, as an EMT suppressor. To our knowledge, our study is the first to report that AM251 affects EMT in renal tubular epithelial cells. However, protection against hyperglycemia-induced glomerular hypertrophy and fibrosis by AM251 was shown in a streptozotocin-induced diabetic rat model [32]. In that study, administration of AM251 caused decreased fibronectin, TGF-β1, TNFα, and IL-6 levels, as well as PPARγ activation. However, in another report, administration of AM251 in an obese rat model did not affect kidney fibrosis, although it improved albuminuria and renal tubular structure [33]. Thus, the effects of AM251 seem to depend on the fibrosis model. It is possible that the differences among models with respect to myofibroblast conversion, such as in EMT, EndoMT, or other processes, are responsible for this discrepancy. In our study, however, we found that AM251 was able to suppress fibrosis via EMT.
Regarding the relationship between CB1 and renal fibrosis, it was reported that, in a renal fibrosis model induced by unilateral ureteral obstruction (UUO), the fibrotic area was reduced to about 2/3 that in wild-type mice by CB1 (CNR1) gene knockout [34]. In this renal fibrosis model, CB1 and its ligand 2-arachidonoylglycerol were increased, suggesting that CB1 signaling was stimulated. CB1 expression was increased in myofibroblasts upon TGF-β1 treatment, suggesting that CB1 was involved in myofibroblast activation [34]. Furthermore, in the streptozotocin-induced diabetic rat model, the effect of AM251 on kidney fibrosis was suggested as being mediated by CB1, because antisense RNA against CB1 (CNR1) had effects similar to those of AM251 treatment [32]. However, the involvement of CB1 in EMT had not been previously examined. In our study, we demonstrated that neither treatment with anandamide nor knockdown of CB1 affected COL1A1 expression in renal proximal tubule epithelial cells ( Fig  5B and 5D). Furthermore, CB1 itself was downregulated by TGF-β1 treatment (Fig 5A). These results indicate that CB1 was not involved in EMT. It is possible that CB1 is involved in renal fibrosis at step(s) other than EMT. with or without 2 ng/ml TGF-β1 in the presence or absence of 10 μM AM251 for 1 h. (B) Total protein lysates were prepared, and equal amounts of protein (5 μg per sample) were separated by SDS-PAGE, followed by immunoblotting with anti-phopho-p38 (P-p38), anti-p38, anti-phospho-SMAD3 (P-SMAD3), or anti-SMAD3 antibodies. (C and D) The results from (B) were quantified. Values are means ± SD of phopho-p38 (C) or phospho-SMAD3 (D) levels relative to those in cells with no treatment (TGF-β1(−) AM251(−)), from three independent experiments. Statistically significant differences are indicated (** P < 0.01, Student's t-test).
doi:10.1371/journal.pone.0167848.g010 AM251 Suppresses Epithelial-Mesenchymal Transition AM251 acted upstream of SMAD2/3 and p38 MAPK (Fig 10), although its exact site of action is unclear. AM251 had a stronger effect on p38 MAPK than on SMAD2/3. Therefore, we speculated that inhibition of the p38 MAPK pathway by AM251 contributed more to its EMT suppression than did inhibition of the SMAD2/3 pathway. Downstream of p38 MAPK, expression of the AP-1 components JUNB and FOSB was decreased by AM251 treatment (Figs  8 and 9). In addition, expression of the key transcription factor for EMT, SNAIL1, was also decreased (Figs 8 and 9). In SNAIL1 knockout mice, renal fibroses (using the UUO model) was decreased [11]. The decrease in SNAIL1 may have caused downregulation of many of the EMT genes observed here (Figs 8 and 9).
There are currently no efficient drugs to prevent tubular fibrosis and EMT. AM251 suppressed pre-induced COL1A1 expression (Fig 4C). Therefore, AM251 may have potential value in both prevention and mitigation of fibrosis. In addition to the kidney, fibrosis can occur in other tissues such as liver and lung, and EMT is involved not only in fibrosis but also in cancer progression [22,23]. In a liver fibrosis model, SNAIL1 was activated and liver fibrosis was suppressed in SNAIL1 knockout mice [35]. Thus, it is possible that inhibition of SNAIL1 can reduce not only kidney fibrosis but also that occurring elsewhere. Our findings may inform development of future drugs for treating fibrosis in the kidney and other tissues, and AM251 is a suitable lead compound for these efforts.

Author Contributions
Conceptualization: TY KU SN KH TN YN AK.