Mitochondrial amidoxime-reducing component 2 (MARC2) has a significant role in N-reductive activity and energy metabolism

The mitochondrial amidoxime-reducing component (MARC) is a mammalian molybdenum-containing enzyme. All annotated mammalian genomes harbor two MARC genes, MARC1 and MARC2, which share a high degree of sequence similarity. Both molybdoenzymes reduce a variety of N-hydroxylated compounds. Besides their role in N-reductive drug metabolism, only little is known about their physiological functions. In this study, we characterized an existing KO mouse model lacking the functional MARC2 gene and fed a high-fat diet and also performed in vivo and in vitro experiments to characterize reductase activity toward known MARC substrates. MARC2 KO significantly decreased reductase activity toward several N-oxygenated substrates, and for typical MARC substrates, only small residual reductive activity was still detectable in MARC2 KO mice. The residual detected reductase activity in MARC2 KO mice could be explained by MARC1 expression that was hardly unaffected by KO, and we found no evidence of significant activity of other reductase enzymes. These results clearly indicate that MARC2 is mainly responsible for N-reductive biotransformation in mice. Striking phenotypical features of MARC2 KO mice were lower body weight, increased body temperature, decreased levels of total cholesterol, and increased glucose levels, supporting previous findings that MARC2 affects energy pathways. Of note, the MARC2 KO mice were resistant to high-fat diet–induced obesity. We propose that the MARC2 KO mouse model could be a powerful tool for predicting MARC-mediated drug metabolism and further investigating MARC's roles in energy homeostasis.

The mitochondrial amidoxime-reducing component (MARC) is a mammalian molybdenum-containing enzyme. All annotated mammalian genomes harbor two MARC genes, MARC1 and MARC2, which share a high degree of sequence similarity. Both molybdoenzymes reduce a variety of N-hydroxylated compounds. Besides their role in N-reductive drug metabolism, only little is known about their physiological functions. In this study, we characterized an existing KO mouse model lacking the functional MARC2 gene and fed a high-fat diet and also performed in vivo and in vitro experiments to characterize reductase activity toward known MARC substrates. MARC2 KO significantly decreased reductase activity toward several N-oxygenated substrates, and for typical MARC substrates, only small residual reductive activity was still detectable in MARC2 KO mice. The residual detected reductase activity in MARC2 KO mice could be explained by MARC1 expression that was hardly unaffected by KO, and we found no evidence of significant activity of other reductase enzymes. These results clearly indicate that MARC2 is mainly responsible for N-reductive biotransformation in mice. Striking phenotypical features of MARC2 KO mice were lower body weight, increased body temperature, decreased levels of total cholesterol, and increased glucose levels, supporting previous findings that MARC2 affects energy pathways. Of note, the MARC2 KO mice were resistant to high-fat diet-induced obesity. We propose that the MARC2 KO mouse model could be a powerful tool for predicting MARC-mediated drug metabolism and further investigating MARC's roles in energy homeostasis.
Chemical bonds between nitrogen and oxygen are part of a large number of natural and synthetic compounds. They range from small inorganic molecules like nitric oxide, nitrite, and nitrate to physiological and nonphysiological organic molecules with N-O bonds as part of various functional groups (e.g. hydroxylamines, hydroxyamidines, hydroxyguanidines, oximes, N-oxides, hydroxamic acids, and sulfohydroxamic acids). These N-oxygenated functions are essential components of many drugs but are often first introduced by cytochrome P450 -catalyzed xenobiotic metabolism (1). Many of these formed metabolites are of high toxicological relevance (2). This also affects DNA base analogs. N-hydroxylated nucleobases have been shown to be mutagenic for prokaryotic and eukaryotic cells (3,4); thus, it is important for organisms to ensure detoxification strategies by reduction to the original nucleobases. In this context, the mitochondrial amidoxime-reducing component (MARC) 3 has been proven to be an extremely effective catalyst for the reduction, even under aerobic conditions of all of the functional groups mentioned above (5,6).
In 2006, MARC was discovered and identified as a molybdenum-containing protein (7). All mammalian genomes studied to date contain two MARC genes: MARC1 and MARC2. The proteins encoded by these genes are MARC1 and MARC2; they represent the simplest form of eukaryotic molybdenum enzymes, only binding the molybdenum cofactor. Beside sulfoxidase, aldehyde oxidase, and xanthinoxidoreductase, MARC is the fourth molybdenum-containing enzyme found in humans. In the presence of NADH, each MARC paralog exerts reductive activity toward N-oxygenated substrates together with the two electron transport proteins cytochrome b 5 B (CYB5B) and cytochrome b 5 reductase (CYB5R) (Fig. 1) (8). It is well-accepted that the MARC-containing enzyme system plays a major role in N-reductive drug metabolism; e.g. activation of amidoxime prodrugs like ximelagatran (9) and Mesupron (10) or reductive metabolism of drug metabolites like N-hydroxyvaldecoxib (11) or sulfamethoxazole hydroxylamine (12).
Although the N-reductive metabolism has been studied extensively, the physiological function of MARC is not completely known. The nitric oxide precursor N-hydroxy-L-arginine (13), nitrite (14) and the physiological metabolite trimeth-ylamine N-oxide (15) are currently being discussed as potential endogenous substrates. Involvement in, e.g., lipid metabolism (16,17) and association with diabetes mellitus (18) are also discussed. Up to now, it is unknown whether both proteins are exclusively involved in reductive pathways or whether one or both have additional functions.
The International Knockout Mouse Consortium is in the process of producing mutations in murine embryonic stem cells for all known protein-coding genes. The murine genome is comparable with the human genome, with a similar estimated number of genes, chromosomal DNA content, and informational complexity (19,20). MARC2 (MARC1) presents nearly 88% (89%) of nucleotide sequence similarity between humans and mice (identical sequences: MARC2, 74%; MARC1, 80%) (21). Therefore, it may be possible to use the MARC2 KO mouse model to imitate functions of this molybdenum enzyme in humans. In particular, deletion of the MARC2 gene can help us understand more about MARC2 phenotypes and roles under some physiological and pathological conditions as well as differences compared with MARC1.
A number of mouse mutant lines, including MARC2 (but not MARC1) KO, have already been produced and phenotyped in high-throughput screenings. By this means, decreased body weight, increased startle reflex, and decreased prepulse inhibition have been reported as significant phenotypes of the MARC2 KO mouse (22). In this report, the MARC2 KO mouse model was obtained from the European Mouse Mutant Archive (23). The absent expression of MARC2 was confirmed by gene sequencing and immunoblot analysis of tissue homogenates. In this study, the phenotype of MARC2 KO mice was characterized by their body characteristics, clinical chemistry, and histological investigation. In addition, protein expression profiling of the reductase enzyme system and metabolic studies using benzamidoxime (BAO; a marker substrate of MARC-mediated reductase activity (12)) and other N-hydroxylated substrates were conducted to further determine the importance of MARC2 in the N-reductive pathway.

MARC2 KO mice have altered body characteristics
An apparent phenotypical feature of MARC2 KO mice is a higher body temperature by, on average, one to two degrees (WT, 24.2°C Ϯ 0.3; KO, 26.0°C Ϯ 0.7) and a lower body weight. Of several anatomical features tested, only small intestine length and spleen and lung weight were significantly changed in both sexes when MARC2 KO was compared with WT mice (Table 1).

MARC2 KO mice are resistant to high-fat diet-induced obesity
A previous study indicated that MARC1 and MARC2 protein abundance is increased in livers of animals on an HFD, suggesting a functional link between the MARC complex and lipid metabolism (16). In this study, both WT and MARC2 KO mice were fed an HFD or a normal diet (ND) for 23 weeks with weekly body weight measurements. As expected, animals on the HFD significantly gained body weight (WT, ϩ49%; MARC2 KO: ϩ30%) compared with animals on the ND ( Fig. 2A). Interestingly, there was no significant difference in body weight between MARC2 KO mice on the HFD and WT mice on the ND. Clinical chemistry measurements in sera revealed that, irrespective of diet, MARC2 KO mice exhibited significantly decreased levels of total cholesterol and significantly increased levels of glucose. With the ND, but not with the HFD, triglycerides and high-density lipoprotein (HDL) were significantly decreased in MARC2 KO mice (Fig. 2C). Furthermore, the levels of ␥-glutamyltransferase and alanine aminotransferase were significantly elevated only in the WT (Fig. 2C), indicating obesity-related hepatocyte damage. To relate biochemical data to liver status, liver histopathology was performed in each experimental group, revealing microvesicular steatosis and hepatocellular ballooning in WT mice on the HFD (Fig. 2B).

MARC2 KO mice show decreased N-reductive metabolism in vivo
To assess whether N-reductive metabolism is decreased in KO mice, we performed in vivo studies. Mice were treated with BAO i.v., and after 15 min, blood from the orbital sinus was collected and processed for serum. For comparison, WT mice were treated identically.
HPLC analytics showed a significantly lower concentration of BAO in serum samples of WT mice compared with KO mice. 7-Fold increased levels of BAO in KO mice demonstrate a substantial decrease in N-reductive activity (Fig. 3).

Reductase activity correlates with the protein expression pattern of the N-reductive enzyme system
To verify that protein expression of the N-reductive enzyme system correlates with reductase activity, SDS-PAGE of murine tissue homogenates (liver, kidneys, and lung) was carried out, and protein expression was examined by Western blot analysis using specific antibodies against the proteins of interest (POI): Figure 1. Schematic of the MARC enzyme system. Shown is the N-reductive enzyme system consisting of CYB5R, CYB5B, and MARC as well as their cofactors FAD, heme, and the molybdenum cofactor (Moco). Catalysis is demonstrated by means of the reduction of an amidoxime. semiquantification of the ECL signal was not meaningful. In contrast, expression of CYTB5R does not vary massively among these investigated tissues (Fig. 4B). These results could be confirmed by semiquantitative determination of protein levels using ImageJ (Fig. 4C). To compare enzyme expression with

Knockout studies with MARC
reductase activity, homogenates were incubated with BAO. The expression pattern of MARC1/2 and CYB5B is well in agreement with high benzamidoxime reductase activity in the kidneys or liver and low reductase activity in the lungs (Fig. 4A).
To consider whether MARC2 KO influences the expression of the other components of the N-reductive enzyme system, Western blot analyses were performed. It was shown that only in the liver does MARC2 KO increase the expression of MARC1. There was no influence on the expression of CYB5B and CYB5R in all tested tissue samples, except for the lungs in the case of CYB5B (Fig. 5).
To investigate the influence of MARC2 on reductive metabolism, tissue from KO and WT mice was used for a reductase assay with several known substrates (Fig. 6A) and Western blot analysis. No or only residual reductase activity could be detected when murine KO tissue was used in reductase assays containing BAO, N 4 -hydroxycytidine, and guanoxabenz. Remarkably, among the tested substrates, only amitriptyline N-oxide was reduced in murine KO samples to the same extent as WT samples (Fig. 6B).

Discussion
It is well-accepted, that MARC plays a crucial role in N-reductive xenobiotic and drug metabolism (11,12). Both proteins can act as reductases in an enzyme system with CYB5B and CYB5R; strong similarities in amino acid levels and overlapping substrate specificities of both MARC proteins make discrimination between them difficult.
Highly abundant expression of porcine MARC1/2 was found in the kidneys, liver, thyroid, and pancreas (5). This is well in accordance with the findings of this study, which confirmed high MARC1/2 expression (Fig. 4B) as well as high benzamidoxime reductase activity in murine liver and kidney homogenates (Fig. 4A). Notably, expression of the heme protein CYB5B corresponds to the expression pattern of MARC, whereas expression of the flavin protein CYB5R did not vary among tested tissues (Fig. 4B).
MARC2 was selected by the International Knockout Mouse Consortium to generate a KO mouse model (24). In contrast to human cell lines with predominant MARC1 expression, high MARC2 and low MARC1 gene expression levels have been described in mice (25). To verify that N-reductive metabolism is decreased, mice were treated i.v. with BAO, a typical substrate of the molybdenum enzyme (10). As expected, serum samples of KO mice showed a 7-fold higher substrate concentration compared with control mice (Fig. 3). This proves that BAO reduction is catalyzed by the MARC-mediated enzyme system.
To investigate the biotransformation reactions in more detail, tissue homogenates were prepared and incubated with several N-oxygenated substrates. Homogenates originating from KO mice undergo a massive decrease up to total loss of reductase activity toward the N-hydroxylated amidine derivate BAO and the N-hydroxylated amidinohydrazone guanoxabenz (Fig. 6) compared with control samples of WT mice. As recombinant human MARC1 and MARC2 showed overlapping substrate specificities with similar K m values (MARC1, 0.5 mM; MARC2, 0.8 mM) (26) for BAO, our results indicate that MARC2 seems to be mainly responsible for the murine N-reductive pathway. The residual N-reductive activity detected in KO samples (Fig. 6B) is probably MARC1-mediated. MARC1 expression was only increased in the liver of KO mice. It is remarkable that the tissue with the highest reductase activity (kidney) shows nonaltered expression of MARC1 (Fig. 5). Compensatory up-regulation of xenobiotic enzymes (e.g. cytochrome P450) has been reported (27,28) and was therefore also expected to a great extent for MARC. Maybe MARC1 is not

Knockout studies with MARC
only a "backup reductase" for MARC2 but also involved in reactions besides the N-reductive pathway (especially in the kidneys). Nevertheless, both assumptions would explain why all annotated genomes of mammals so far appear to possess two copies of MARC genes (29,10), and it suggests the physiological need of these enzymes.
In the case of N 4 -hydroxycytidine, MARC1-mediated reductase activity still amounted to 71% in KO mice (Fig. 6B). This unexpectedly high residual activity can be explained by distinct kinetic parameters of recombinant human MARC1 (K m , 0.4 mM) and human MARC2 (K m , 5.7 mM), resulting in a 16-fold higher catalytic efficiency of MARC1 toward this hydroxylamine derivative (5).
So far, only N-oxides have been found to be exclusively reduced by recombinant human MARC1 (15,30). Therefore it was not surprising that no significant differences in reductase activities of WT and KO liver homogenates were detected when amitriptyline N-oxide was used as a substrate in in vitro incubations (Fig. 6B).
Our findings demonstrate that MARC2 KO mice are a sensitive model to measure reductive activity toward N-hydroxylated compounds. The reduction of N-oxygenated functional groups has been demonstrated in vivo for numerous compounds (31,32). However, the enzymatic basis of these reductions has remains unclear (33). Several enzymes tested were only active in the absence of oxygen; thus, the physiological function remained questionable. With the discovery of the MARC enzyme system in 2006, it became clear that this catalyst may play a major role in N-reductive metabolism, as this enzyme was unaffected by oxygen and able to reduce all N-O bonds studied so far. This has been shown using recombinant enzymes, cellular systems, and knockdown studies (34), explaining a lot of metabolites detected in vivo. The KO studies presented here finally confirm the major participation of MARC in N-reductive metabolism. MARC2 KO mice represent a novel and available animal model that can be used as a powerful tool to investigate the N-reductive metabolism of, for example, new drug candidates. The recently published structure of MARC1 (35) explains the ability of MARC enzymes to reduce such an incredible number of N-O substrates and, thus, its great role as a drug-metabolizing enzyme.
In addition, to study the role of MARC2 in N reduction, we used the animal model for further characterizations. Homozygous MARC2 KO mice were viable and fertile and did not exhibit any obvious physiological deficiencies, suggesting that deletion of the gene has no pivotal role in growth and development.
However, KO mice exhibit decreased body weight and are refractory to diet-induced obesity (Table 1 and Fig. 2A), Strikingly, the body weights of KO mice on the HFD were nearly the same as those of WT mice on the ND at the end of 23 weeks of feeding. Thus, MARC2 depletion protects against adipose tissue growth and concomitant body weight gain. Interestingly, KO mice exhibit significantly higher glucose serum levels (Fig.  2C). These traits together suggest that MARC2-depleted cells, which are inefficient to store fatty acids in lipids, rewire metabolism to cope with their excess by promoting fatty acid catabolism for glucose production. Then the surplus of glucose is consumed for higher energy production, which is supported by observed thermogeneration of KO mice. Remarkably, previous studies have reported that human MARC2 expression is upregulated by high-glucose treatment in cell culture and kidneys of diabetic Goto Kakizaki rats (a type 2 diabetes animal model) (18,36). Additionally, our observations are in agreement with our previous studies, which demonstrate that nutritional deficiency particularly down-regulates expression of MARC2 and CYB5B, which results in a lower in vitro benzamidoxime reductase activity of liver homogenates prepared from fasted mice (16). As MARC1 expression was only slightly increased in the liver and not altered significantly in the kidneys of KO mice (Fig. 5), this effect is probably solely due to MARC2.
Further clinical chemistry measurements revealed a significant decrease in total cholesterol, triglycerides, and HDL in sera of KO mice (Fig. 2C). This is in line with the findings of Neve et al. (37), who showed that siRNA-mediated MARC2 knockdown leads to a significant decrease in lipid and fatty acid levels during differentiation of murine 3T3-L1 cells into adipocytelike cells. It is striking that MARC2 is dually localized in peroxi-  A and B, tested N-oxygenated substrates (A) and N-reductive activity (B) of tissue homogenates (liver, kidneys, and lungs) of six individual WT and MARC2 KO mice. Every homogenate was incubated separately in multiple incubations (N Ն 4) over a time range of up to 90 min. The limit of quantification was 0.1 nmol⅐mg Ϫ1 ⅐min Ϫ1 for BAO and guanoxabenz and 0.02 nmol⅐mg Ϫ1 ⅐min Ϫ1 for N 4 -hydroxycytidine and amitriptyline-N-oxide. Statistical significance was proven by U test and t test. Negative controls without NADH were performed with each substrate, and none of the respective metabolites could be detected in these samples. In case of guanoxabenz, the detected residual N-reductive activity of KO samples (2.4 Ϯ 0.1 nmol⅐mg Ϫ1 ⅐min Ϫ1 ) meets the value of the negative control without NADH (2.1 Ϯ 0.1 nmol⅐mg Ϫ1 ⅐min) and was corrected by this value. #, below the limit of quantification. **, p Ͻ 0.01; ***, p Ͻ 0.001; n.s., not significant.

Knockout studies with MARC
somes and mitochondria, as both multifunctional organelles play cooperative roles in the metabolism of cellular lipids (38,39).
In conclusion, we envision the MARC2 KO mice usability to study the role of the molybdenum enzyme in N-reductive drug metabolism. In addition, this animal model should be used to investigate involvement of the MARC complex in metabolic pathways, as MARC2 depletion affects lipid, cholesterol, and glucose levels. This study provides the first in vivo evidence of a role of the MARC complex in lipid/energy metabolism; however, further research regarding the molecular and metabolome level utilizing both the MARC2 and MARC1 KO and the MARC1/2 double KO mouse model are warranted to reveal the bona fide role of the MARC complex in metabolism homeostasis.

Animals
For the in vivo and in vitro studies, KO mice provided by the European Mouse Mutant Archive were used, which lack the functional MARC2 gene. The MARC2 KO mice was developed on the background C57BL/6NTac strain. Homozygous MARC2 KO mice as well as homozygous WT mice were used, which were derived from the same heterozygous C57BL/6NTac strain. Genome sequencing was carried out at the Institute Clinique de la Souris (Illkirch Cedex, France).

Ethics statement
Mice were housed in the Department of Genetics and Laboratory Animal Breading at the Maria Sklodowska-Curie Memorial Cancer Center and Institute of Oncology. All mice were maintained in accordance with recommendations in the Guide for Care and Use of Laboratory Animals," from the Institute on Laboratory Animal Resources, National Research Council, and the experimental protocol was approved by the Second Local Ethical Committee for Animal Research (Warsaw, Poland, decision 25/2015).

Body temperature measurement
Measurement was performed with an IR noncontact thermometer around the anus. Results are the means of n ϭ 13 (WT) and n ϭ 10 (KO) mice on a regular diet.

HFD study design
Four groups of eight mice were fed and characterized over a period of 23 weeks. Two groups (WT and KO) were fed a regular diet (10% of calories from fat) containing 19.2% protein, 67.3% carbohydrate, and 4.3% fat (D12450B, Research Diets), and the other two (WT and KO) were fed an HFD (60% of calories from fat) containing 26.2% protein, 26.3% carbohydrate, and 34.9% fat (D12492, Research Diets). Body mass was measured weekly.

Histological examination of murine liver
Fragments of fresh liver tissue were formalin-fixed, paraffinembedded, and subjected to histopathological examination as described previously (40).

In vivo BAO biotransformation study
For the in vivo study, six homozygous female MARC2 KO mice and six female WT mice at the age of 10 -11 weeks were used. N-reductive metabolism was examined by means of reduction of the established model substrate BAO to benzamidine (BA). 20 mg/kg BAO was given i.v. to the animals, followed by blood collection and sacrifice. Before intravenous administration of BAO, KO and WT animals were anesthetized with an intraperitoneal injection of ketamine (80 mg/kg) and xylazine (10 mg/kg) solution. After 15 min, blood from the orbital sinus was collected for serum, and mice were sacrificed by severing the spinal cord, followed by organ collection. The blood was shaken for 60 min at room temperature and clotted. Serum was obtained through centrifugation, and N-reductive activity was determined by measuring the concentration of the model substrate BAO.

Preparation of tissue homogenates
Murine livers, kidneys, and lungs were cut into pieces and homogenized using a Potter-Elvehjem homogenizer (Kimble Chase Life Science and Research Products, LLC) in ice-cold buffer containing 0.25 M sucrose, 1 mM EDTA, and 10 mM potassium dihydrogen phosphate (pH 7.4). Samples were frozen and stored at Ϫ80°C. Every organ was worked up separately and analyzed further (reductase assay and Western blotting).

Determination of protein content
For murine homogenates, protein content was determined using the bicinchoninic acid protein assay kit (Pierce) according to the manufacturer's protocol.

Determination of BAO in murine serum
Serum was mixed with two aliquots of acetonitrile and shaken for 45 min at 4°C. After centrifugation for 30 min, the resulting supernatant was analyzed by HPLC.

In vitro N reduction assays
N-reductive activity of murine tissue homogenates (liver, kidneys, and lungs) of six WT and six KO mice were determined by measuring the reduction of BAO to BA, guanoxabenz to guanabenz, N 4 -hydroxycytidine to cytidine, and amitriptyline N-oxide to amitriptyline (Fig. 6A). Incubation mixtures contained 0.05 mg of protein (homogenate), 3 mM substrate, and 1 mM NADH in 100 mM potassium dihydrogen phosphate buffer (pH 6.0). To improve the solubility of guanoxabenz and amitriptyline N-oxide, 6% DMSO was added. Incubation was carried out at 37°C. The reaction was initiated by addition of NADH after 3 min of preincubation and stopped after different times (1-90 min) by addition of 100 l of ice-cold methanol. Samples were shaken and centrifuged, and the supernatant was analyzed via HPLC. Two controls were measured: incubation mixture without protein or without NADH.

Quantification of the metabolite BA by HPLC
BA was quantified using a LiChroCHART column with LiChrospher 60 RP-select B (5 m) and a RP-select B 464-m guard column (Merck kGaA). The mobile phase consisted of 80% 10 mM octyl sulfonate buffer and 20% (v/v) acetonitrile. The flow rate was isocratic at 1 ml/min at room temperature, and detection was carried out at 229 nm. The retention time was 15.9 Ϯ 0.1 min for BA and 7.1 Ϯ 0.2 min for BAO.

Quantification of the metabolite cytidine by HPLC
Cytidine was quantified using a LiChroCHART column with LiChrospher 60 RP-select B (5 m) and a RP-select B 464-m guard column (Merck KGaA). The mobile phase consisted of 75% 10 mM ocytl sulfonate (pH 2.0) and 25% (v/v) methanol. The flow rate was isocratic at 1 ml/min at room temperature, and detection was carried out at 281 nm. The retention time was 5.5 Ϯ 0.1 min for cytidine and 3.1 Ϯ 0.1 min for N 4 -hydroxycytidine.

Quantification of the metabolite guanabenz by HPLC
Guanabenz was quantified using a LiChroCHART column with LiChrospher 60 RP-select B (5 m) and a RP-select B 464 m guard column (Merck kGaA). The mobile phase consisted of 70% (v/v) 100 mM ammonium acetate (pH 4.0) and 30% (v/v) methanol. The flow rate was isocratic at 1 ml/min at room temperature, and detection was carried out at 272 nm. The retention time was 22.0 Ϯ 0.7 min for guanabenz and 15.5 Ϯ 0.4 min for guanoxabenz.

Quantification of the metabolite amitriptyline by HPLC
Amitriptyline was quantified using a Phenomenex Gemini NX C18 (5 pm) 150 ϫ 4.6 mm column (Phenomenex Inc., Aschaffenburg, Germany) with a Phenomenex security guard cartridge system (Gemini C18 4 ϫ 2 mm precolumn). The mobile phase consisted of 20 M dipotassium phosphate with 0.1% TFA (pH 2.5) and acetonitrile (65:35, v/v). The flow rate was isocratic at 1 ml/min at room temperature, and detection was carried out at 245 nm. The retention time was 9.9 Ϯ 0.1 min for amitriptyline and 11.6 Ϯ 0.1 min for amitriptyline N-oxide.

Statistical analyses
Statistical analyses were carried out using SigmaPlot 11 software (Systat Software Inc.). The significance of observed differences was evaluated by U test and t test. A p value of less than 0.05 was considered significant: *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001.
After washing with TBST, the specific band for each POI was visualized by the enhanced chemiluminescence method according to the instructions (GE Healthcare Biosciences). For the stripping procedure, we used either a commercially available stripping buffer according to the manufacturer's protocol (Candor Bioscience GmbH) or a stripping buffer containing 200 mM glycine, 3.5 mM SDS, and 900 mM Tween 20 (pH 2.2).
Electrochemiluminescence was visualized using both Amersham Biosciences Hyperfilms TM (GE Healthcare) and the Che-moStar Touch ECL and Fluorescence Imager (Intas Science Imaging Instruments GmbH, Göttingen, Germany). Digital signals were then quantified densitometrically using ImageJ software version 1.52a.