Human Kidney Methoxyflurane and Sevoflurane Metabolism : Intrarenal Fluoride Production as a Possible Mechanism of Methoxyflurane Nephrotoxicity

Sevoflurane and methoxyflurane were supplied by Abbott Laboratories (Abbott Park, IL). Sulfaphenazole was kindly provided by Dr. William F. Trager (University of Washington). Sulfamic acid (> 99.5%) was obtained from Fluka (Ronkonkoma, NY), and sodium hydroxide (50% weight in weight, low in carbonate) from Fisher (Pittsburgh, PA). Microsomes containing individual complementary deoxyribonucleic acid (cDNA)-expressed cytochrome P450 isoforms were purchased from Gentest (Woburn, MA). Unless specified, all other reagents were obtained from Fluka or Sigma Chemical (St. Louis, MO) and were of the highest purity available. All buffers and reagents were prepared with high-purity (greater or equal to 17.5 M Omega *symbol* cm) water (Milli-Q, Millipore, Bedford, MA).

All animal experiments were approved by the institutional Animal Care and Use Committee. All human tissue experiments were approved by the institutional Human Subjects Review Committee.

Human kidneys were obtained from the National Disease Research Interchange (Philadelphia, PA). Cadaveric kidneys were retrieved within 4 h of asystole, frozen in liquid nitrogen, and shipped on dry ice to our laboratory. Donor kidneys were obtained from organ donors, perfused with University of Wisconsin solution, shipped on wet ice to our laboratory, and frozen in liquid nitrogen. Tissue was stored at -80 degrees Celsius until used. Preliminary experiments showed similar rates of anesthetic metabolism by microsomes from cadaveric or donor kidneys.

Microsomes were prepared by differential ultracentrifugation by using the following modifications of methods described previously for liver microsomes. [20] Tissue thawed in 10 mM potassium phosphate buffer (pH 7.4) containing 0.9% sodium chloride was homogenized in three to five volumes of 10 mM potassium phosphate buffer (pH 7.4)/0.25 M sucrose in a blender (Waring) followed by five strokes with a motorized glass-polytetrafluorethylene (Teflon) homogenizer. The homogenate was centrifuged at 12,500 g for 30 min, and the supernatant centrifuged at 110,000 g for 70 min. The supernatant produced by the 110,000 g centrifugation was saved as the cytosolic fraction. The microsomal pellet was resuspended in 10 mM potassium phosphate buffer (pH 7.4)/0.15 M potassium chloride and again centrifuged at 110,000 g for 70 min. The washed pellet was resuspended in 0.25 M sucrose and stored at -80 degrees Celsius. Microsomal protein concentrations were determined by the method of Lowry et al. with bovine serum albumin as the standard, as described previously. [20].

Preparation of P450 antibodies and Western immunoblot techniques were identical to those described previously, [20,25] except that 75–100 micro gram microsomal protein were loaded onto each lane. Because purified kidney P450 isoforms were not available as standards to permit numerical quantification, P450 immunoblots were graded exclusively by optical density and P450 amounts expressed on a scale of zero to two.

Anesthetic defluorination was determined in 2 *symbol* ml screw-capped polypropylene vials in a reaction mixture (1.0 ml) containing 2.5 mg microsomal protein, 2 mM reduced nicotinamide adenine dinucleotide phosphate (NADPH), and 2 micro liter anesthetic in 50 mM Tris-chloride buffer (pH 7.4 at 37 degrees Celsius) unless otherwise indicated. For experiments with cDNA-expressed P450s, incubations contained 0.7 mg microsomal protein. Reactions (37 degrees Celsius with reciprocal shaking) were initiated by the addition of anesthetic agent and NADPH, terminated after 60 min by immersion in a boiling water bath for 5 min, and centrifuged at 16,000 g for 15 min. Blank tubes from which NADPH was omitted were treated similarly.

Fluoride concentrations were measured with a fluoride-specific combination electrode (9409, Orion, Boston, MA) and pH/ISE meter (920A, Orion) by a significant modification of previously described techniques. [20] An 850-micro liter aliquot of the above supernatant was transferred to a clean polypropylene tube containing a miniature stir bar and 250 micro liter sulfamic acid buffer (0.88 M sulfamic acid, 0.22 M sodium hydroxide). After 10 min, with continuous stirring of the mixture (100–300 rpm), millivolt readings were obtained and compared with a standard curve to determine fluoride concentrations. Standard curves were prepared daily by adding known amounts of sodium fluoride to Tris buffer (final fluoride concentration 50–1,000 nM) and analyzing as described. The ionic strength of the reaction mixtures and standards was carefully controlled. The electrode was conditioned daily by soaking in ultrapure water for 30 min followed by soaking in 50 nM sodium fluoride-Tris-sulfamic acid buffer for 30 min. Between samples the electrode was washed with ultrapure water for 5 min with stirring and then allowed to reequilibrate in 850 micro liter Tris buffer plus 250 micro liter sulfamic acid buffer for 5 min. Samples and standard solutions were read in randomized order.

Experiments with isoform-selective chemical P450 inhibitors were conducted at the following final concentrations: coumarin (P450 2A6, 50 micro Meter), orphenadrine (P450 2B6, 5 micro Meter), sulfaphenazole (P450 2C9/10, 50 micro Meter), quinidine (P450 2D6, 5 micro Meter) diethyldithiocarbamate (P450 2E1, 100 micro Meter), troleandomycin (P450 3A, 100 micro Meter), midazolam (P450 3A, 50 micro Meter), and lauric acid (P450 4A, 200 micro Meter). [26–30] Lauric acid, a substrate (though nonselective) for P450 4A, [30] was used to probe the involvement of this isoform in defluorination. All inhibitors were added in Tris-hydrochloride buffer except troleandomycin, which was added in methanol (1 micro liter 100 mM stock, final methanol concentration 0.1%). Substrate and inhibitor concentrations were chosen to suppress more than 50% of isoform activity. For the competitive inhibitors sulfaphenazole, quinidine, and lauric acid, 0.2 micro liter methoxyflurane and 0.5 micro liter sevoflurane were added to the reaction mixture to maximize the inhibitor-substrate concentration ratio. For the noncompetitive mechanism-based inhibitors, in which the inhibitor-substrate concentration ratio is immaterial, 2 micro liter methoxyflurane or sevoflurane was added to the reaction mixture. Incubations containing sulfaphenazole, quinidine, or lauric acid were initiated by simultaneous addition of NADPH and anesthetic agent. Reaction mixtures containing orphenadrine, diethyldithiocarbamate, or trolcandomycin were preincubated at 37 degrees Celsius for 10 min with NADPH, and the defluorination reaction initiated by adding anesthetic agent. The defluorination reactions were terminated after 60 min and fluoride production determined as described above. Fluoride production in the presence of inhibitor was compared with appropriate controls (no preincubation and 10-min preincubation) and results expressed as a percentage of the rate without inhibitor.

Human kidney microsomes were found to catalyze the defluorination of methoxyflurane and, to a lesser extent, sevoflurane. Defluorination was dependent on time, protein concentration, anesthetic concentration, and NADPH, indicating that the defluorination was an enzyme-mediated metabolic process, rather than a nonspecific phenomenon (Figure 1). Kidney cytosol did not catalyze measurable anesthetic defluorination, despite the addition of multiple cofactors (Table 1). Therefore, all subsequent experiments were performed exclusively with kidney microsomes.

Defluorination of methoxyflurane and sevoflurane was compared in a population of human kidneys obtained from cadavers and organ donors. Anesthetic metabolism by several kidneys is shown in Figure 2. Sevoflurane defluorination was minimal and only slightly greater than blanks in all kidneys examined, with net fluoride production rates of 10–80 pmol *symbol* mg sup -1 *symbol* h sup -1. In contrast, methoxyflurane defluorination was substantially greater than that of sevoflurane in several kidneys studied, and there was considerable variability in the net rate of methoxyflurane defluorination, which ranged from 40 to 420 pmol *symbol* mg sup -1 *symbol* h sup -1.

Human renal microsomal methoxyflurane and sevoflurane defluorination was compared with that by human liver microsomes (Table 2). For both agents, defluorination by renal microsomes was substantially less than that by liver microsomes, in accordance with the lower specific content of cytochrome P450 in kidney microsomes. The differential between average rates of hepatic and renal metabolism appeared greater for sevoflurane than for methoxyflurane, although there was considerable variability in the individual rates and the organs were not obtained from the same donor.

The cytochrome P450 enzymes are a superfamily of multiple P450 isoforms, with discreet but overlapping substrate specificities and differential organ expression. [31] At least 14 P450s have been identified in human liver, although substantially less is known about the expression and activity of P450 isoforms in human kidney. To identify the P450 isoform(s) responsible for human renal anesthetic defluorination, a panel of inhibitors each selective for a P450 isoform was used (Figure 3). [26–29] This approach was used previously to identify P450 isoforms catalyzing anesthetic metabolism in human liver. [20,21] Renal methoxyflurane and sevoflurane defluorination were inhibited by diethyldithiocarbamate, a mechanism-based inhibitor of P450 2E1. Evidence suggests that diethyldithiocarbamate also inhibits P450 2A6. [32] Coumarin, a substrate of P450 2A6, also inhibited metabolism of methoxyflurane and, to a lesser extent, sevoflurane. Midazolam, a substrate and competitive inhibitor of P450 3A, [29] diminished metabolism of methoxyflurane but not sevoflurane. Troleandomycin, a mechanism-based P450 3A inhibitor, decreased methoxyflurane and to a lesser extent sevoflurane metabolism. No other P450 isoform-selective inhibitor significantly diminished methoxyflurane or sevoflurane metabolism.

To confirm the above identification of P450 isoform(s) responsible for human renal anesthetic defluorination. P450s present in the renal microsomal samples were probed by Western blot analysis. Antibodies against hepatic cytochrome P450 isoforms 2C, 2E1, and 3A4 were used. Rabbit antihuman P450 3A4 antibodies cross-reacted with a renal microsomal protein co-migrating with a hepatic P450 3A protein (Table 3). This protein was found in highest abundance in HK 109, and in variable amounts in the other kidneys studied. Rabbit polyclonal antihuman P450 2C9 antibodies cross-reacted with a renal microsomal protein co-migrating with hepatic P450 proteins in the 2C family, however no more selective isoform identification was possible. We detected no kidney microsomal proteins that comigrated and cross-reacted with antibodies against P450 2E1.

The ability of microsomes, containing individual human P450 isoforms expressed from a hepatic cDNA library, to deflourinate methoxyflurane and sevoflurane was compared (Figure 4). P450 isoforms 2B6, 2E1, 2A6 and 3A4 catalyzed methoxyflurane and sevoflurane defluorination, in descending order of activity, whereas P450s 1A2 and 2D6 showed no activity. All isoforms catalyzing anesthetic metabolism demonstrated significantly greater activity toward methoxyflurane compared with sevoflurane, particularly P450s 2B6, 2E1 and 2A6.

Results of the current investigation demonstrate that human kidney microsomes catalyze the metabolism of methoxyflurane, resulting in the liberation of free fluoride ion. Sevoflurane also undergoes metabolism, but at a rate substantially less than that of methoxyflurane. The rates of anesthetic defluorination reported herein may represent a conservative measurement that underestimates the true rate of metabolism on a per milligram basis. This is because microsomal P450 content is highest in the renal cortex and minimal in the inner medulla, [33] whereas our microsomes were prepared from whole kidney.

There was considerable interindividual variability in renal methoxyflurane metabolism. This variability mirrored that observed previously in human hepatic anesthetic metabolism. [20,21] The highest rate of defluorination was found in HK109, obtained from a donor chronically treated with hydrocortisone and levothyroxine. No other organ donors were identified as taking medications known to alter drug metabolizing activity. Thus renal anesthetic metabolism demonstrates considerable individual variability and potential susceptibility to the effects of inducing agents.

We found renal methoxyflurane defluorination confined to the microsomal tissue fraction, with no activity in cytosol. Previous investigations have shown that methoxyflurane was defluorinated by rat liver cytosol, catalyzed by one or more of the glutathione S-transferase enzymes. [34] The current data do not provide an explanation for the apparent difference between rat hepatic and human renal cytosolic methoxyflurane metabolism, which may relate to differences between species, tissue-specific patterns of isoenzyme expression, and substrate specificities of the various glutathione S-transferase enzymes. [35,36].

Anesthetic metabolism by human kidney microsomes was substantially less than that by rabbit kidney microsomes reported previously. For example, methoxyflurane defluorination by rabbit kidney microsomes [22] was 1,860 plus/minus 360 pmol *symbol* mg sup -1 *symbol* h sup -1 compared with 190 plus/minus 50 pmol *symbol* mg sup -1 *symbol* h sup -1 by human kidney microsomes in the current investigation. This is consistent with known differences in renal mixed function oxidase activity between rodent species and humans. [37] However, in both rabbit and human kidney microsomes, methoxyflurane defluorination was significantly greater than that of the other anesthetics studied, sevoflurane and enflurane.

Anesthetic metabolism by human liver microsomes was considerably greater than that by human kidney microsomes. This is congruent with prior reports of significantly lower total renal P450 content and mixed function oxidase activity compared with that of total liver. [37] Nevertheless, specific renal proximal tubular cells may contain P450 activity that is equal to that of hepatocytes. [38] Thus rates of anesthetic defluorination in proximal tubular cells, the target site of methoxyflurane injury, may be greater than rates reported for whole kidney microsomes.

Initial identification of the P450 isoforms catalyzing renal anesthetic metabolism was based on the effects of P450 isoform-selective inhibitors. Additional evidence was based on Western blot analysis of renal microsomes and published reports of renal P450 isoform contents, in conjunction with anesthetic metabolism by human hepatic cDNA-expressed P450 isoforms. Together these suggested that P450s 2E1, 2A6, and 3A are the most likely isoforms catalyzing renal methoxyflurane defluorination. Identifications for sevoflurane are less certain, however, because of the overall very low rate of sevoflurane metabolism. P450 2E1 and/or 2A6 appear to be the major isoforms responsible for renal sevoflurane metabolism, with a possible contribution from P450 3A. This interpretation requires qualification, however, because the P450s used herein were expressed from hepatic rather than renal cDNA libraries, and little information is available concerning the catalytic activity of renal P450 isoforms or the inhibitory activity or selectivity of hepatic P450 isoform-selective inhibitors against their renal counterparts.

Several findings indicate that a member of the P450 2A family participated in human kidney microsomal metabolism of methoxyflurane, and to a lesser extent, sevoflurane. Human renal defluorination was inhibited by the P450 2A6 substrate coumarin, and human hepatic cDNA-expressed P450 2A6 was a highly effective catalyst of the defluorination reaction. Previous investigations have suggested the presence of a P450 2A in human kidneys, by the use of ethoxycoumarin deethylase activity as a metabolic marker (albeit nonspecific) for P450 2A6 activity. [23,27] Although no information is available regarding the catalytic activities of human renal P450 2A6, animal studies do suggest strong structural and catalytic similarities between hepatic P450 2A6 and extraheptic P450s of the 2A class. [39].

P450 2E1 participation in methoxyflurane and sevoflurane metabolism is suggested from inhibition by diethyldithiocarbamate, an effective inhibitor of hepatic P450 2E1 [40] and human hepatic volatile anesthetic metabolism. [20,21] This was supported by the catalytic activity of expressed P450 2E1. Human renal P450 2E1 has not been purified or expressed, thus there is no information regarding the catalytic activity of human renal, compared with hepatic, P450 2E1. Nevertheless, P450 2E1 is highly conserved between species, and animal studies do suggest strong similarities between hepatic and renal P450 2E1. [41–43] In contrast, Western blots with our antirat P450 2E1 antibody did not, however, detect a renal P450 comigrating with P450 2E1, consistent with a previous report. [21] Interestingly, cytochrome P450 2E1 has a substrate selectivity profile that partially overlaps that of P450 2A6 [44,45] and the inhibitory activity of diethyldithiocarbamate toward P450 2A6 has recently been recognized. [32] Thus the apparent paradox between inhibition by diethyldithiocarbamate despite the absence of detectable P450 2E1 protein may be rationalized if some diethyldithiocarbamate-inhibitable renal defluorination represents activity of P450 2A6 in addition to or instead of P450 2E1.

A P450 protein of the 3A class* was the third P450 isoform apparently catalyzing renal anesthetic metabolism. Methoxyflurane metabolism was inhibited by the P450 3A inhibitors troleandomycin and midazolam. Western blot analysis detected a kidney microsomal protein that comigrated with P450 3A and that cross-reacted with an antihuman P450 3A4 antibody. This was consistent with the findings of Schuetz et al., who found P450 3A5 in all kidneys they examined, and P450 3A4 in only one of seven kidneys they studied. [46] In the current investigation, the content of P450 3A varied widely among the seven kidneys examined and was much less than the P450 3A4 and 3A5 content in human liver observed previously by us [25] and by others. [24,31] Expressed human hepatic P450 3A4 did metabolize methoxyflurane and sevoflurane, at rates that were lower than for the other active P450 isoforms. It is unknown whether expressed hepatic P450 3A4 is an accurate model for human renal P450 3A activity, although catalytic similarities between human hepatic P450 3A4, hepatic P450 3A5 and extrahepatic P450 3A5 have been observed. [46,47].

No evidence was obtained for the participation of other P450 isoforms in anesthetic metabolism. P450 4A is the most abundant human renal P450 isoform, but the P450 4A substrate lauric acid had no effect on methoxyflurane or sevoflurane metabolism. Although hepatic cDNA-expressed P450 2B6 exhibited the highest intrinsic rate of methoxyflurane defluorination, the selective 2B6 inhibitor orphenadrine was without inhibitory activity in renal microsomes. The absence of orphenadrine inhibition is congruent with the observation that P450 2B6 is infrequently expressed in human liver, constitutes less than 1% of total P450 when it is expressed, and is likely expressed in very low amounts, if at all, in human kidney. [48] These experiments do not, however, completely exclude the possible catalytic participation of other P450 isoforms.

Identification of the P450 isoforms catalyzing renal anesthetic metabolism provides a mechanistic basis for understanding the greater renal metabolism of methoxyflurane compared with sevoflurane. P450s 2E1, 2A6 and 3A4 catalyzed defluorination of methoxyflurane three to ten times faster than that of sevoflurane. Differences between rat and human renal anesthetic metabolism may also be rationalized by the substantially greater content of P450 2E1 in rat, compared with human kidneys, particularly in relation to P450 2E1 content in the liver. [24,43].

The last decade has witnessed tremendous expansion and evolution toward understanding the biochemical mechanisms of chemically induced organ-specific nephrotoxicity, the subject of a comprehensive review. [49] Some renal toxins undergo prior metabolism in the liver to protoxicants, which are then further activated in the kidney to an ultimate reactive metabolite causing selective renal damage. Nephrotoxicity of other compounds such as chloroform, acetaminophen, and p-aminophenol results from bioactivation by P450 in situ in the kidney and local action of reactive metabolites. Renal P450 2E1 [50] and P450 3A4 [51] contents are highest in proximal tubular cells, the most common site of toxicity. [49] Thus the importance of renal metabolism as an etiologic factor in renal drug toxicity has become increasingly apparent.

In this regard it is interesting to note that strain differences in methoxyflurane nephrotoxicity mirror those of acetaminophen and p-aminophenol. Renal function and histopathologic abnormalities were significantly greater in Fischer 344 rats compared with other strains exposed to the three agents. [5,52] The enhanced susceptibility of Fischer 344 rats to acetaminophen and p-aminophenol nephrotoxicity resulted from greater intrarenal activation of p-aminophenol in this, compared with other, strains. Greater in situ renal methoxyflurane metabolism might similarly explain the strain differences in nephrotoxicity of this anesthetic agent.

There is incontrovertible evidence that methoxyflurane nephrotoxicity is intimately related to biotransformation to one or more toxic metabolites. The severity of human nephrotoxicity is proportional to the degree of systemic methoxyflurane biotransformation. [1] In animals, suppression of methoxyflurane biotransformation, either by inhibition of P450 activity, [8] or by substitution of deuterium for hydrogen on methoxyflurane to retard metabolism, [53] resulted in diminished biochemical and structural evidence of nephrotoxicity. Conversely, augmentation of methoxyflurane metabolism by induction of P450 activity in animals worsened the severity of renal damage. [8].

It is classically accepted that fluoride ion is the toxic methoxyflurane metabolite, with a plasma concentration of 50 micro Meter the threshold for renal damage. In this paradigm, hepatic anesthetic defluorination liberates the nonselective metabolic toxin fluoride anion, which is distributed systemically yet causes selective renal toxicity when plasma concentrations exceed 50 micro Meter.

There is some evidence to support the hypothesis that increased systemic fluoride concentrations cause nephrotoxicity: First, the severity of methoxyflurane nephrotoxicity in humans was directly related to plasma fluoride concentrations. [1] Second, increases or decreases in plasma fluoride concentrations after methoxyflurane anesthesia in animals, caused by induction or inhibition of metabolism, resulted in increased or diminished renal toxicity, respectively. [8] Finally, injection of fluoride in rats mimics the renal pathologic changes caused by methoxyflurane. [15].

There is also compelling evidence to reevaluate the hypothesis that increased systemic fluoride concentrations alone cause nephrotoxicity. First, in experiments in which exogenous fluoride was administered to replicate renal toxicity, [15] plasma fluoride concentrations were not measured and shown comparable to those found after methoxyflurane administration. Second, numerous investigations have demonstrated the safety of enflurane, isoflurane, and sevoflurane despite plasma fluoride concentrations exceeding 50 micro Meter. Prolonged sevoflurane anesthesia, resulting in plasma concentrations exceeding 50 micro Meter, had no effect on renal function assessed by provocative testing or by clinical indexes. [11,13,14] Enflurane anesthesia in isoniazid-treated humans resulted in peak plasma fluoride concentrations that were clearly in the nephrotoxic range (as high as 130 micro Meter) and that exceeded those seen with methoxyflurane, yet there was no evidence of renal dysfunction. [16] These fluoride concentrations remained increased longer than those seen with sevoflurane. Prolonged isoflurane anesthesia (18–20 MAC-h), during which plasma fluoride concentrations exceeded 50 micro Meter for 2 or 3 days, had no deleterious effect on any measure of renal function. [17] Similarly, prolonged sedation with isoflurane in which plasma fluoride concentrations were increased for days (as high as 93 micro Meter), [18] had no adverse effects on renal function. [18,19,54] Children undergoing prolonged isoflurane sedation, in which fluoride concentrations remained increased for as long as 32 days, exhibited no significant changes in renal function. [55] In contrast, a comparatively brief (8-h) exposure to enflurane, resulting in moderately increased serum fluoride concentrations (mean 33 micro Meter) that remained increased for 1 day, did cause clinical nephrotoxicity evidenced by diminished urine osmolality. [15] Interestingly, during the period in which urine concentrating ability was diminished 25%, the average serum fluoride concentration was only 15 micro Meter. [15] Together these investigations strongly suggest that neither peak systemic fluoride concentration nor duration of fluoride increase alone can be applied nonselectively to all anesthetic agents to explain or predict nephrotoxicity. Rather, the mechanism(s) of anesthetic nephrotoxicity may be agent-specific.

In summary, we have demonstrated that microsomes from human kidney, the target organ of anesthetic toxicity, can metabolize methoxyflurane to the putative nephrotoxin, fluoride ion. If fluoride (or any other toxic metabolite) produced locally, at the site of toxicity, exerts a local toxic effect, then renal metabolism may contribute to methoxyflurane nephrotoxicity. The relative paucity of renal sevoflurane metabolism may explain the absence of clinical sevoflurane nephrotoxicity, despite plasma fluoride concentrations that may exceed 50 micro Meter. The role of renal anesthetic metabolism, in addition to or in contrast to hepatic anesthetic metabolism, in the pathogenesis of methoxyflurane renal toxicity clearly merits further investigation.

The authors are indebted to Dr. Martin Frant of Orion for his invaluable assistance in the development of the high-sensitivity fluoride assay.

*Human P450s in the 3A class include P450 3A4, the predominant form in adult liver: P450 3A3, which differs from P450 3A4 by a few amino acids and is catalytically and electrophoretically indistinguishable from P450 3A4; P450 3A5, which is expressed extrahepatically and in approximately 20% of adult livers; and P450 3A7, expressed exclusively in fetal tissues. [31].