Tacrolimus Increases Nox4 Expression in Human Renal Fibroblasts and Induces Fibrosis-Related Genes by Aberrant TGF-Beta Receptor Signalling

Chronic nephrotoxicity of immunosuppressives is one of the main limiting factors in the long-term outcome of kidney transplants, leading to tissue fibrosis and ultimate organ failure. The cytokine TGF-β is considered a key factor in this process. In the human renal fibroblast cell line TK-173, the macrolide calcineurin inhibitor tacrolimus (FK-506) induced TGF-β-like effects, manifested by increased expression of NAD(P)H-oxidase 4 (Nox4), transgelin, tropomyosin 1, and procollagen α1(V) mRNA after three days. The macrolide mTOR inhibitor rapamycin had similar effects, while cyclosporine A did not induce fibrose-related genes. Concentration dependence curves were sigmoid, where mRNA expression was induced already at low nanomolar levels of tacrolimus, and reached saturation at 100–300 nM. The effects were independent of extracellular TGF-β as confirmed by the use of neutralizing antibodies, and thus most likely caused by aberrant TGF-β receptor signaling, where binding of tacrolimus to the regulatory FKBP12 protein results in a “leaky” TGF-β receptor. The myofibroblast marker α-smooth muscle actin was neither induced by tacrolimus nor by TGF-β1, indicating an incomplete activation of TK-173 fibroblasts under culture conditions. Tacrolimus- and TGF-β1-induced Nox4 protein upregulation was confirmed by Western blotting, and was accompanied by a rise in intracellular H2O2 concentration. Si-RNA mediated knock-down of Nox4 expression prevented up-regulation of procollagen α1(V) mRNA in tacrolimus-treated cells, but induced procollagen α1(V) expression in control cells. Nox4 knock-down had no significant effect on the other genes tested. TGF-β is a key molecule in fibrosis, and the constant activation of aberrant receptor signaling by tacrolimus might contribute to the long-term development of interstitial kidney fibrosis in immunosuppressed patients. Nox4 levels possibly play a regulatory role in these processes.


Introduction
The availability of the calcineurin inhibitors (CNIs) cyclosporine (CsA) [1] and tacrolimus (FK-506) [2] has revolutionized transplantation medicine. Currently more than 90% of all patients receiving a renal graft are treated post-transplant with CNIs [3]. However, CNI nephrotoxicity is a major problem, and lesions at least partly attributable to CNI nephrotoxicity can be seen in virtually all histological sections ten years after transplantation [4].
Fibrogenic effects of CNIs have been described in different compartments of the kidney, with main focus on the tubularinterstitial region. Already in 1990, procollagen secretion in murine epithelial cells and fibroblasts exposed to CsA was reported [5]. The knowledge about the role of tacrolimus in fibrosis is more diverse. Similar fibrogenic responses in patients receiving CsA or tacrolimus have been described six and twelve months after renal transplantation [6]. One year after transplantation, control biopsies from tacrolimus-treated patients with stable graft function show a significantly lower TGF-b1 expression compared to CsAtreated ones [7]. However, after a mean period of 22/28 months not only the expression of TGF-b mRNA is higher in the tacrolimus group, but also several markers of fibrogenesis are overexpressed [8]. As a further consequence of activation of TGFb signaling, interstitial fibrosis is promoted by an increasing production of extracellular matrix (ECM) proteins [9,10], and induction of epithelial-to-mesenchymal transition (EMT) [11]. In renal fibroblasts a conversion to a myofibroblastic cell type appeared after exposure to TGF-b [12].

Microarray analysis
Cells were exposed to 10 mM of tacrolimus for one day, or three days, resp. Experiments were performed in duplicates. RNA was isolated from cell culture samples by Qiagen RNeasy Mini Kit. In brief, cells were lysed in buffer containing mercaptoethanol, and the lysate was homogenized by passing it through a sterile syringe needle (0,9 mm) several times. RNA was purified over the columns and eluted in RNase-free H 2 O. Cy3-labeled cRNA was generated from total RNA with the LowInput QuickAmp Labeling Kit (Agilent Technologies, 5190-2331). 1,65 ng of cRNA from each sample were hybridized to one-color microarrays (Whole Human Genome 4644 K Oligo Microarray kit, Agilent Technologies, G4112F) for 17 hours at 65uC. Slides were scanned on an Axon GenePix 4000B scanner (Molecular Dynamics), and the obtained data were analyzed in the Agilent Feature Extraction software (Vers. 9.5.3.1). Data can be accessed online at www.ebi.ac.uk/ arrayexpress under accession number E-MTAB-778).

Intracellular ROS measurement
Cells grown in 96-well plates were loaded with the cellpermeable H 2 O 2 indicator 29,79-dichlorofluorescin diacetate (DCF-DA) by incubation in serum-free medium containing 3 mM DCF-DA (Sigma 6883; diluted from 10 mM stock in DMSO) for 60 minutes at 37uC/5% CO 2 . After loading, the cells were carefully washed two times with fresh medium within 10 minutes, and fluorescence at 485 nm/535 nm (excitation/emmission) was determined immediately afterward on a 96-well plate reader (Tecan GENios). Fluorescence intensity values were averaged over 12 wells for each condition and normalized to control = 1.

TGF-b neutralization by antibodies
TGF-b antibodies (anti-hTGFb-IgA, Invivogen) were added to the cell cultures at a concentration of 300 ng/ml, to neutralize extracellular TGF-b. In case of the addition of 5 ng/ml TGF-b1 as a positive control, the complete medium containing both TGF-b1 and antibodies was set aside for several minutes to allow complete neutralization before application to the cells.

si-RNA mediated Nox4 knockdown
TK-173 fibroblasts were lipofected with Nox4 siRNA (Dharmacon D-010194-01) utilizing Dharmacon transfection reagent No. 3. 37 ml of transfection reagent and 7 ml of siRNA (20 mM) were diluted separately in each 320 ml of serum-free medium. The two solutions were then mixed carefully, and the mixture was left standing for 20 minutes at 37uC. In parallel, a transfection solution with non-target control RNA was prepared the same way. Cells were washed with serum-free medium, and 100 ml of the solution were pipetted on each single well of 12-well plates with confluent TK-173 cells. Incubation with drugs was started the next day.

Statistics
One-way ANOVA followed by Bonferroni's post-hoc comparisons was performed in all statistical analyses, except for the time course experiments (Fig. 1), which were analysed by unpaired t-test combined with Levene's test for homogenity of variances. All tests were performed by PRISM software (GraphPad Software, Inc.).

Tacrolimus induces up-regulation of fibrosis-related genes in human renal fibroblasts
Human renal fibroblasts (TK-173) and human proximal tubular cells (RPTEC/TERT1) were exposed to 10 mM tacrolimus for one and three days. Statistical analysis of the microarray data identified 211 transcripts that were up-regulated, and 64 that were downregulated ($twofold, p#0.05) in TK-173 fibroblasts after three days. Results after one day were comparable, although with generally lower amplitudes. RPTEC/TERT1 cells showed a much weaker response, with only 25 genes up-regulated and 10 downregulated. Therefore, we focused further experiments on TK-173 cells. For detailed microarray results see www.ebi.ac.uk/ arrayexpress (login data in Methods section).
A subset of fibrosis-related genes was studied in greater detail by real-time quantitative PCR: The cytoskeletal components transgelin (SM22a), tropomyosin 1, alpha (Tpm-1), and alpha-smooth muscle actin (a-SMA, ACTA2); the extracellular matrix component procollagen a1(V); NAD(P)H oxidase 4 (Nox4); and transforming growth factor b-1 (TGF-b1). When treated with tacrolimus concentrations ranging from 1 to 1000 nM for three days, all genes (with the exception of a-SMA) showed a similar concentration-dependent response: the curves had a sigmoid shape, and effects became noticeable already at low nanomolar concentrations of tacrolimus (Fig. 2). From 100-300 nM of tacrolimus on, the effects became saturated, and the RT-qPCR results at high concentrations validated the values from the microarray experiments. The strongest positive reactions were observed on the expression of Nox4 (max. 7,3-fold), tropomyosin 1 (max. 7,3), and transgelin (max. 6,7). a-SMA was not regulated by exposure to tacrolimus (Fig. 2F), neither in RT-qPCR nor in microarray analysis. We chose a concentration of 100 nM tacrolimus for all subsequent studies, as this had a significant effect on most genes tested without reaching saturation.
We also monitored the time course of mRNA expression from day 0 (time point of drug application) to day 5 in the presence of 100 nM tacrolimus. Nox4, transgelin, and procollagen a1(V) mRNA expression levels were significantly elevated after one day (Nox4 and transgelin), or two days (procollagen a1(V)), resp. (Fig. 1). Although tacrolimus had no effect on a-SMA mRNA expression, we observed a constant increase from day 1 to day 5 in both tacrolimus-treated and untreated cells. This was most likely a reaction to the serum deprivation during the treatment period.
Tacrolimus activates the TGF-b/Smad pathway, but in a ligand-independent way As several fibrosis-related genes were positively regulated by tacrolimus, we hypothesized involvement of the TGF-b/Smad signaling pathway. When we treated TK-173 fibroblasts with the pro-fibrotic cytokine TGF-b1 (10 ng/ml), expression of Nox4, transgelin, and tropomyosin 1 was induced as in tacrolimustreated cells, but with higher amplitudes (Fig. 3, A-C). Blocking TGF-b signaling with the TGF-b type I receptor kinase inhibitors LY364947 (3 mM) omitted both the reaction to tacrolimus and . Cells were switched to serum-free medium on day -1, and were then exposed to 100 nM tacrolimus starting at day 0. Cell samples were collected from day 0 to day 5. All results were normalized to day 0. Cells showed a significant response to tacrolimus after one day (Nox4 and transgelin) or two days (tropomyosin-1), respectively. Alpha-smooth muscle actin mRNA showed a slow up-regulation in both treated and untreated cells, most likely as a reaction to serum deprivation. (*) denotes significant (p#0,05) difference between tacrolimus-treated and control cells at the same timepoint. doi:10.1371/journal.pone.0096377.g001  . Effects of TGF-b1 signaling blockade. TK-173 cells exposed to 100 nM tacrolimus (TAC), or 10 ng/ml TGF-b1 (TGF) for three days showed increased expression of Nox4 (A), transgelin (B), and tropomyosin-1 (C). Blockade of TGF signaling by TGF-b1 RI inhibitor LY364947 inhibited both the reactions to TGF-b1 and tacrolimus [p#0,05 (*), and p#0,001 (**), resp.]. In contrast, application of anti-TGF-b antibody to the medium left the reaction to tacrolimus unaffected (E), but prevented the effects of 5 ng/ml TGF-b1 on the expression of procollagen a1(V) (Col5a1), Nox4, transgelin (Tagln), and tropomyosin-1 (Tmp1) almost completely (D). Antibodies had no effect in unstimulated control cultures (F). All values were normalized to untreated control cultures. doi:10.1371/journal.pone.0096377.g003 TGF-b1, and suppressed gene expression below control levels, suggesting that tacrolimus might act through activation of the TGF-b receptor.
As we had also measured a slight increase in TGF-b1 mRNA in tacrolimus-treated TK-173 fibroblasts, we utilized antibodies against TGF-b to block possible autocrine TGF-b receptor activation. A concentration of 300 ng/ml of anti-TGF-b antibody in the medium was sufficient to almost completely suppress the effects of 5 ng/ml TGF-b1 (positive control) on the expression of procollagen a1(V), Nox4, transgelin, and tropomyosin 1 (Fig. 3D). In contrast, antibody-mediated depletion of extracellular TGF-b did not change the response of fibroblasts to tacrolimus exposure (Fig. 3E). Thus, TGF-b receptor activation by tacrolimus is not dependent on presence of the specific ligand in the medium.

Rapamycin, but not CsA has a similar action on gene expression as tacrolimus
CsA and rapamycin (sirolimus) are the two other most widely used immunosuppressants besides tacrolimus. The use of different concentrations (1 mM of CsA, and 100 nM of tacrolimus and rapamycin, resp.) in our experiments was intended to roughly reflect the differences in actual therapeutic doses. With respect to the three genes of interest, effects of rapamycin on mRNA expression in RT-qPCR experiments were comparable to tacrolimus, with significantly positive regulation of Nox4, transgelin and tropomyosin-1 (Fig. 5). CsA had no effect on the expression of the genes tested.

Tacrolimus-induced Nox4 protein expression is associated with increased H 2 O 2 levels
In Western blots, Nox4 protein was below detection level in the control cultures. Tacrolimus-treated fibroblasts (100 nM, 3 days) showed a weak, but clearly positive signal for Nox4 at approx. 60 kD, while TGF-b1 treatment (10 ng/ml, 3 days) induced a stronger response (Fig. 4B).
The increased levels in Nox4 protein resulted in increased intracellular ROS levels. NAD(P)H oxidases are generally thought to produce superoxide anion (26), but our initial attempts to detect intracellular Nox4-generated superoxide anion by nitroblue tetrazolium conversion [27] failed (data not shown). However, fluorometric measurement of intracellular H 2 O 2 by dichlorofluorescin diacetate conversion revealed increased peroxide concentrations in tacrolimus-treated (max. 1,2-fold) and TGF-b1-treated fibroblasts (max. 1,4-fold) (Fig. 4C). Concentration dependence followed a sigmoid curve, comparable to Nox4 mRNA expression curves in the RT-qPCR experiments. Treatment with the TGF-breceptor 1 kinase activity blocker LY364947 prevented the effects of both tacrolimus and TGF-b1 and completely abated the increased H 2 O 2 levels (Fig. 4D).

Nox4 expression levels regulate expression of procollagen a1(V)
There are as of yet no known specific inhibitors of Nox4, therefore we used siRNA-mediated gene knock-down to suppress Nox4 activity. Transfection of the cells with Nox4-targeted siRNA decreased Nox4 mRNA expression significantly by approx. two thirds compared to scrambled-RNA transfected cells, both in tacrolimus-treated cells and untreated cells (Fig. 6A). Nox4 knockdown had the most intense effect on procollagen a1(V), where mRNA expression was increased 2,6-fold in the control cells, and suppressed to control levels in the tacrolimus-treated cells (both compared to scrambled-RNA transfected cells) (Fig. 6E). Nox4 knock-down had no significant effects on the expression of cytoskeletal proteins (a-SMA, transgelin, and tropomyosin-1) both in tacrolimus-stimulated and control cells.

Discussion
Induction of ligand-independent (''aberrant'') TGF-b signaling by tacrolimus has been reported in literature [28,29]. The underlying mechanism was first described by Chen et al. in 1997 [30]: The FK506-binding protein 12 (FKBP12) binds to TGF-breceptor subunit I (TGF-b-R1) [31] and anticipates the spontaneous phosphorylation of Smad 2 and 3 in the absence of ligand. As tacrolimus (FK506) occupies the TGF-b-R1-binding pocket of FKBP12 (both share the same binding region on FKBP12), interaction of FKBP12 with the receptor is inhibited and its negative modulatory effect on TGF-b signaling is lost. This results in a ''leaky'' receptor, which constantly activates the Smad 2/3 signaling cascade by phosphorylation, even in the complete absence of ligand. The saturation of the tacrolimus-induced TGF-b signaling at .300 nM in our experiments could indicate the complete depletion of free FKBP12 and therefore maximum aberrant signaling of the TGF-b receptor. As 10 ng/ml TGF-b1 had a markedly stronger effect on gene expression than tacrolimus, it is evident that TGF-b receptor signaling cannot be fully activated by tacrolimus-mediated FKBP12 depletion alone. The similar response of the cells to rapamycin further supports the concept above as rapamycin binds to FKBP12 at the same macrolide binding site as tacrolimus [32]. In contrast, cyclosporine-treated TK-173 cells did not show any signs of TGF-b receptor activation. Pro-fibrotic effects of cyclosporine and tacrolimus are obviously, at least in part, triggered by different pathways, although both drugs act as calcineurin inhibitors.
In untreated fibroblasts, the TGF-b receptor I blocker LY364947, but not TGF-b neutralizing antibodies, attenuated the expression of transgelin and tropomyosin-1 well below control levels (Figs. 3B, 3C). This findings, together with basal SMAD2 phosphorylation in Western blots from control cells, indicate a constant level of aberrant TGF-b signaling in untreated TK-173 fibroblasts.
The pro-fibrotic characteristics of TGF-b are well accepted [33], and a constantly increased Smad 2/3 signaling in fibroblasts by aberrant TGF-b receptor activation might at least partly contribute to kidney fibrosis under tacrolimus regimen. TGF-b induces an ''activated state'' in fibroblasts, the so-called myofibroblast, which is characterized by a partially smooth-muscle-like cell type with contractile properties and increased matrix deposition. The differentiation into a smooth-muscle-like cell type with contractile properties requires reorganization of the cytoskeleton, and the cytoskeletal components smooth-muscle actin, transgelin, and tropomyosin are well defined markers for this transformation [34,35]. The actin-associated protein transgelin is one of the earliest markers in smooth muscle differentiation [36].
The renal fibroblast cell line TK-188, which was isolated from fibrotic kidney by the same group as the non-fibrotic TK-173 cells, had been shown to have a five-fold higher basal expression of a-SMA mRNA than TK-173 cells [37]. In our in-vitro system, we could not observe an increase in alpha smooth muscle actin (a-SMA) expression in TK173 cells, neither by tacrolimus, nor by TGF-b1 (TGF-b1 data not shown). However, cells slowly increased a-SMA mRNA expression as a reaction to serum deprivation. Taking the up-regulation of the cytoskeletal components tropomyosin and transgelin into account, the fibroblasts are likely missing an additional factor to fully differentiate into myofibroblasts under the given culture conditions. Desmouliere et al. proposed an intermediate ''proto-myofibroblast'' state induced by mechanical stress, which can then differentiate further into a fully activated fibroblast in the presence of TGF-b [38].
The increased release of extracellular matrix compounds by activated myofibroblasts contributes to histologic changes in fibrotic tissue. Interstitial fibrosis is characterized by accumulation of type I and III collagens, but no up-regulation of these collagens was observed in our microarray data. The ECM component most intensely regulated by tacrolimus in our experiments was procollagen a1(V). Although collagen V normally represents only 1-3% of collagen fibers, it plays a central role in the regulation of fibrillogenesis [39]. Increased deposition of collagen V in the interstitial ECM has been reported in patients with chronic renal disease [40]. Mozes et al. [41] found trace amounts of collagen V within the mesangium of kidneys from transgenic mice with TGFb overexpression, while it was essentially absent in wild-type mice.
Nox4 has been reported to play an important role in the activation of fibroblasts, including heart, lung, and kidney [12,14,42]. It is thought that Nox4 activity shifts the redox potential within the cell and thereby modulates redox-sensitive pathways (e.g. MAP kinases). The strong impact of Nox4 knockdown on procollagen a1(V) expression indicates that the corresponding signaling pathways are redox-sensitive and respond to Nox4-generated H 2 O 2 -levels. The other genes tested were not impaired by Nox4 knock-down, so we assume that Nox4 is not generally mandatory for pro-fibrotic effects of tacrolimus-mediated TGF-b signaling.
While the other NAD(P)H oxidases release superoxide anion, Nox4 produces H 2 O 2 as an intrinsic function of distinct regions in the protein's E-loop [43]. The tacrolimus-concentration depen-dent curve of intracellular H 2 O 2 (Fig. 4C) mirrored the corresponding Nox4 mRNA expression curve ( Fig. 2A) very well. This is consistent with data from literature, which suggest that Nox4 is constitutively active and its activity is regulated only by expression levels [13]. The DCF-DA assay only provides us with relative values for intracellular H 2 O 2 , as it is not possible to calibrate results. Taking the strong background into consideration, the increase in the ROS indicator fluorescence signal by max. 20% (tacrolimus) and 40% (TGF-b1) may in fact reflect a much stronger increase in H 2 O 2 concentrations. Furthermore since blocking TGF-b signaling resulted in the suppression of Nox4 mRNA below control levels without a corresponding reduction in DCF fluorescence signal, we postulate that in unstimulated cells the Nox4-generated H 2 O 2 levels are negligible.
In conclusion we demonstrate that tacrolimus induces aberrant TGF-b signaling in a human renal fibroblast cell line, leading to changes in cytoskeletal genes expression, intracellular H 2 O 2 levels, and ECM synthesis. The effect is cell-type specific, as proximal tubule epithelial cells did not show it. TGF-b-like effects of tacrolimus and the structurally related rapamycin had been first described more than a decade ago, but the consideration of ligandindependent TGF-b signaling is still somehow under-represented in literature, and not all recent publications on pro-fibrotic effects of macrolide immunosuppressants take aberrant signaling into account. Furthermore, our findings in a cell culture model indicate a possible role of tacrolimus-induced Nox4 expression in fibrotic processes through modulation of collagen V expression. The mechanism behind is unclear, as Nox4 knock-down had opposite effects on collagen V expression in tacrolimus-treated and control cells. Further research on the involvement of Nox4 and collagen V in the onset of fibrosis could deepen our understanding of fibrotic processes in transplanted kidney.