Mechanisms of fluoroquinolone side effects

Fluoroquinolones (FQ) are powerful broadspectrum antibiotics whose side effects include renal damage and, strangely, tendinopathies. The pathological mechanisms underlying these toxicities are poorly understood. Here we show that the FQ drugs Norfloxacin, Ciprofloxacin, and Enrofloxacin are powerful iron chelators comparable to Deferoxamine, a clinically-useful iron chelating agent. We show that iron chelation by FQ leads to epigenetic effects through inhibition of α-ketoglutarate-dependent dioxygenases that require iron as a co-factor. Three dioxygenases were examined in HEK293 cells treated with FQ. At sub-mM concentrations these antibiotics inhibited Jumonji domain histone demethylases, TET DNA demethylases, and collagen prolyl 4-hydroxylases, leading to accumulation of methylated histones and DNA, and inhibition of proline hydroxylation in collagen, respectively. These effects may explain FQ-induced nephrotoxicity and tendinopathy. By the same reasoning, dioxygenase inhibition by FQ was predicted to stabilize transcription factor HIF1α by inhibition of oxygen-dependent HIF prolylhydroxylation. In dramatic contrast to this prediction, HIF-1α protein was eliminated by FQ treatment. We explored possible mechanisms for this unexpected effect and show that FQ inhibit HIF-1α mRNA translation. Thus, FQ antibiotics induce global epigenetic changes, inhibit collagen maturation, and block HIF-1α accumulation. We suggest that these mechanisms explain the classic renal toxicities and peculiar tendinopathies associated with FQ antibiotics. Food and Drug Administration (FDA) approved antimicrobial drugs are designed to target pathogenic microorganisms with minimal effects on the host. However, non-antibiotic effects of antimicrobial agents are well-known (1), due to unexpected interactions with cellular pathways. Generalized adverse effects (2-4) are common to most antimicrobials, balancing against benefits (511). Here we investigate the interaction of relevant concentrations of fluoroquinolone (FQ) antibiotics Ciprofloxacin (CIPRO; Fig. 1A), Norfloxacin (NOR), and Enrofloxacin (ENRO) with a cultured human embryonic kidney cell line, revealing previously unreported enzyme inhibition effects http://www.jbc.org/cgi/doi/10.1074/jbc.M115.671222 The latest version is at JBC Papers in Press. Published on July 23, 2015 as Manuscript M115.671222 Copyright 2015 by The American Society for Biochemistry and Molecular Biology, Inc. by gest on July 6, 2017 hp://w w w .jb.org/ D ow nladed from Mechanisms of fluoroquinolone side effects 2 that may explain toxicities associated with FQ treatment. FQ are popular synthetic broad-spectrum antibiotics that exert their antimicrobial effect by preventing energy-dependent negative supercoiling of bacterial DNA through gyrase inhibition (12). FQ are effective agents that target both gram-negative and gram-positive bacteria and are recommended for severe bacterial infections, including multidrug resistant infections (13). FQ side effects have been widely studied (14-19). However, the molecular mechanisms underlying these toxicities remain to be elucidated. One such peculiar FQ side effect is tendinopathy (15,20). The majority (>85%) of FQ-associated tendinopathies occur within a month of initial FQ therapy, with three-fold higher chance of tendon rupture within the first 90 days of exposure (21). In rare cases of patients with pre-existing musculoskeletal disorders, FQ therapy has been linked to tendinopathy as early as a few hours after administration to as late as 6 months after discontinuing medication (22). Although compromised collagen integrity after FQ treatment is well recognized in animal models (17,22,23) the underlying mechanism is unknown. Some studies report association of enhanced matrix metalloprotease (MMP) (23,24) or collagenase (25) expression associated with FQ-induced tendinopathy. However, a direct link to defects in collagen, a protein that accounts for greater than 6% of muscle mass (26), is still obscure. FQ-associated nephrotoxicity is also welldocumented (27-35). Past clinical studies on patients receiving FQ therapy have revealed a strong association with acute renal failure involving interstitial nephritis (27,32,34), acute tubular necrosis (29), and more recently, crystalluria (33,35). These complications are often attributed to immune-mediated allergic hypersensitivity to FQ antibiotics, with reversal after discontinuation of drug treatment (31,35). Although considerable clinical evidence for FQassociated nephropathy is available, detailed cellular effects of these antibiotics leading to nephritis are not well understood. Appreciating the mechanism of pathological side effects is important for improving our understanding of FQassociated nephrotoxicity, and for illuminating potential complications. Here, we provide evidence for new mechanisms of FQ toxicity involving renal cell epigenetics, impaired collagen maturation, and suppression of hypoxia-inducible factor, HIF-1α. We show that at least some of these effects are due to the powerful iron-chelating property of FQ drugs. An intrinsic FQ characteristic is the propensity to bind to metal cations (36-38). This is due to the electronegative oxygen atoms in the adjacent pyridone and carboxylate moieties (Fig. 1) of all quinolone derivatives (39). The potential for metal chelation by FQ suggests multiple toxic effects on cells. Here we focus on FQ effects on a class of Fe(II)-dependent enzymes known as 2ketoglutarate (2-KG)-dependent dioxygenases (40). The first and best-characterized 2-KG dioxygenase is prolyl-4 hydroxylase, which catalyzes the post-translational hydroxylation of proline residues in collagen (41,42). Other Fe(II)dependent dioxygenases include HIF-1α-prolyl hydroxylase dioxygenase (PHD), Jumonji domain histone demethylases (JMHD), and TET methylcytosine dioxygenase 1 (TET1), responsible for hydroxylation of the HIF-1α transcription factor, histone demethylation, and DNA demethylation, respectively. Here we test the hypothesis that all of these dioxygenases are subject to inhibition by the iron chelating properties of FQ antibiotics. In contrast to these dramatic epigenetic changes consistent with the predicted effects of iron chelation on dioxygenases, we report an unpredicted result in the case of HIF-1α. Here dioxygenase inhibition should stabilize HIF-1α by protecting it from prolyl hydroxylation (43). In fact, FQ treatment has the opposite effect, strongly suppressing HIF-1α accumulation. Thus, we suggest that iron chelation by FQ antibiotics inhibits α-KG-dependent collagen prolyl-4-hydroxylase and other dioxygenase enzymes, perhaps explaining FQ side effects, including spontaneous tendon ruptures (44). In addition, FQ-induced epigenetic modifications uncovered here may explain aspects of FQ nephrotoxicity. Finally, our unexpected observation of FQ-induced HIF-1α loss suggests the possible use of FQ drugs in cancer therapy (45-48). EXPERIMENTAL PROCEDURES Cell culture Human embryonic kidney (HEK293) cells were cultured under by gest on July 6, 2017 hp://w w w .jb.org/ D ow nladed from Mechanisms of fluoroquinolone side effects 3 physiologically relevant oxygen conditions (49): (37 ̊ C, 90% humidity, 5% CO2, 2% oxygen balanced by N2) in DMEM medium (Gibco) containing 10% FBS and 1% penicillin/streptomycin. Iron competition assay The universal siderophore assay of Schwyn and Neilands (50) was used to measure the iron-chelating activity of FQ antibiotics. Deferoxamine mesylate (DFO; Calbiochem), a siderophore produced by Streptomyces pilosus, was used as the positive control. Chrome Azurol S (CAS) assay solution (100 mL) was prepared with the following final concentrations: CAS (Sigma, 199532; 0.15 mM), hexadecyltrimethylammonium bromide (Sigma, 1102974; 0.6 mM), iron (III) chloride hexahydrate (Sigma, 236489; 0.015 mM from a stock dissolved in 10 mM HCl), and 4.3 g anhydrous piperazine (Sigma, P45907) dissolved in 6.25 mL of 12 M HCl and adjusted the pH to 5.6. The solution was stored in the dark at 4 ̊ C. Iron binding reactions were conducted in triplicate with 0.5 mL aliquots of the CAS assay solution and various concentrations of antibiotics to a total volume of 700 μL. Test samples were incubated at room temperature with slow rotation for 30 min, and then transferred to a 96-well plate for absorbance measurement (630 nm) using a microplate spectrophotometer. The apparent half maximal inhibitory concentration (IC50) for iron complexes was estimated as described in the supplemental materials. Stoichiometry determination was conducted based on Schwyn and Neilands assay as described above. In a typical assay, the concentration of iron (Fe) was held constant and increasing concentrations of test compounds required to quench the absorbance of CAS-Fe complex was added. Because tested drug concentrations are all well above the equilibrium dissociation constant for complex formation, the [compound]/[Fe] ratio required for complete quenching in such titrations gives the compound:Fe binding stoichiometry. Cell culture treatment with FQ, DFO, CoCl2, and ferric citrate 10 mM stock solutions of CIPRO, ENRO, and NOR (Sigma, 17850, 17849, N9890, respectively) were prepared by dissolving FQ in 0.01 N HCl. HEK293 cells were cultured in 10cm dishes to 70% confluence prior to treatment with FQ at the indicated final concentrations for the indicated times. In some experiments cells were treated with either 100 μM DFO (Sigma, D9533) or CoCl2 for 4 h. An equal volume of 0.01 N HCl (NT) was added to a separate dish of cells as negative control. In co-treatment experiments, cells were first treated with CIPRO for 30 min followed by addition of the indicated concentrations of ferric citrate (Sigma, F3388) for 4 h. Cells viability was determined using trypan blue dye exclusion. Immunoblotting Western blot analyses were performed by growing cells in 10-cm dishes, harvesting, and lysis with RIPA buffer containing 1x protease inhibitor cocktail (Roche) and 1x phosphatase inhibitor (Thermo Scientific). Cell lysate was agitated on ice for 20 min prior to centrifugation at 14,000 rpm for 15 min at 4 ̊ C in a microcentrifuge. Extracts were analyzed by electrophoresis through 10% BIS-tris polyacrylaminde gels under reducing conditions with detection by Western blotting using anti-HIF-1α antibody (BD 610958, 1:1000), anti-HIF-2α antibody (NB 100-122, 1:1000), anti-H3K9me2 antibody (Abcam 1220, 1:1000), anti-H3K9me3 antibody (Millipore 07-442, 1:500), antiH3K27me2 antibody (Abcam 24684, 1:1000), anti-H3K27me3 antibody (Millipore 05-1951, 1:15,000), anti-H3 antibody (SC 10809, 1:1000), anti-HDAC6 antibody (CS 7558S, 1:1000), antiJMJD2D antibody (Abcam 93694, 1:200), antiTET1 antibody (Abcam 156993, 1:1000), or antiactin antibody (Sigma A2066, 1:500). Secondary antibodies were anti-rabbit and anti-mouse conjugated to a horseradish peroxidase (Promega, 1:15000) and signals were developed using an ECL plus kit (Pierce). Genomic DNA extraction and hydrolysis A Qiagen genomic DNA extraction kit was used to harvest genomic DNA from cells. The manufacturer’s instructions were followed with minor changes as described (51). Briefly, the cell pellet was lysed with C1 buffer and subjected to centrifugation at 1,000 rpm for 10 min at 4 ̊ C in a clinical centrifuge. Pelleted nuclei were resuspended in C1 buffer with centrifugation for 5 min at 4 ̊C. G2 buffer was used to lyse the nuclear membranes. RNase A solution (Thermo Scientific, final concentration 10 mg/mL) and by gest on July 6, 2017 hp://w w w .jb.org/ D ow nladed from Mechanisms of fluoroquinolone side effects 4 Proteinase K solution (Sigma, final concentration 10 mg/mL) were added to the lysed nuclei and incubated overnight at 55 ̊ C. Subsequent purification steps were according to the manufacturer’s instructions. Genomic DNA was washed with 70% ethanol, re-suspended in water, and stored at -20 ̊ C. Three microgram of genomic DNA was hydrolyzed to mononucleosides as described (52). The resulting 40 μL mixture contained 3 μg DNA, 1x micrococcal nuclease buffer (New England BioLabs), 400 mM MgCl2, 4 mM ZnCl2, 20 U deoxyribonuclease I (New England BioLabs), 2000 U micrococcal nuclease I (New England BioLabs), 5 U antarctic phosphatase (New England BioLabs), and 0.4 U snake venom phosphodiesterase. Reactions were incubated overnight at 37 ̊ C. LC-MS analysis of nucleosides LC-MS was performed by loading 0.6 μg mononucleosides from digested genomic DNA onto a C18 analytical reverse phase column (Phenomenex-C18 1.0x250 mm) using an Agilent series 1100 instrument (Agilent Technologies) with mobile phase A (0.05 M ammonium formate, pH 5.4; Sigma, 17843) and mobile phase B (methanol) at a flow rate of 0.05 mL/min and absorbance at 277 nm. The following gradient program was used: 0 min: 2% B; 18 min: 10% B; 30 min: 25% B; 35 min: 2% B, and 60 min: 2% B. Mass spectrometry was performed as previously described (52). Briefly, the HPLC effluent was connected in-line to a mass spectrometer (MSD-TOF, Agilent Technologies) operated in positive ion mode. The MS conditions were: nebulizer 20 psi, dying gas 7 L/min, gas temperature 325 ̊ C; fragmentor 45 V, Oct 1 DC 37.5 V; Oct RF 250 V. All data were analyzed using Agilent MassHunter Quantitative Analysis