Introduction

Triclosan is a broad-spectrum anti-bacterial agent that has been widely used in a variety of personal care products, household items, clinical settings, and medical devices (Fang et al. 2010; Rodricks et al. 2010). There is potential for humans to be exposed to triclosan through dermal contact with personal care products containing triclosan or through the consumption of drinking water contaminated with triclosan. Pharmacokinetic studies in humans and animals have shown that triclosan can be absorbed from the gastrointestinal tract and skin (Rodricks et al. 2010). In fact, triclosan has been detected in human body fluids, including breast milk (Dayan 2007), plasma (Hovander et al. 2002), and urine (Calafat et al. 2008).

In rodents, triclosan undergoes metabolism by cytochrome P450 (CYP), glucuronosyltransferase (UGT), and sulfotransferase (SULT) (Black et al. 1975; Fang et al. 2016a; Moss et al. 2000; Tulp et al. 1979). Glucuronide and sulfate conjugates of triclosan have been detected in the urine of mice, rats, and guinea pigs following dermal, oral, or intraperitoneal administration of triclosan (Black et al. 1975; Fang et al. 2016a; Moss et al. 2000; Tulp et al. 1979). Aromatic hydroxylation and scission of the ether bond of triclosan, which is believed to be mediated by CYP, also occurred in mice and rats (Fang et al. 2016a; Tulp et al. 1979). Several CYP-mediated metabolites of triclosan, including monohydroxy-triclosan and 2,4-dichlorophenol, have been detected in both mice and rats (Fang et al. 2016a; Tulp et al. 1979); however, very few studies have reported the CYP metabolism of triclosan in humans. The CYP isoforms that metabolize triclosan in humans and the extent of metabolism-associated toxicity are currently unknown.

We recently reported that dermal application of triclosan to B6C3F1 mice for 13 weeks resulted in an increase in liver weight, an elevation in the serum levels of alanine aminotransferase and alkaline phosphatase, and hepatocellular hypertrophy, indicating that triclosan is a potential liver toxicant in mice when applied topically (Fang et al. 2015). A number of metabolic enzymes, such as CYP, UGT, and SULT, are expressed in liver (Achour et al. 2014; Ohno and Nakajin 2009; Riches et al. 2009), and xenobiotic metabolism is closely related to toxicity. Metabolic intoxication or detoxification has been documented with many chemicals both in vitro and in vivo (Fang et al. 2016b; Guengerich 2001; Laine et al. 2009; Xiao et al. 2009; Xue et al. 2011).

In view of the CYP metabolism of triclosan observed in rodents and the potential hepatotoxicity of triclosan found in mice, we investigated the metabolism of triclosan by human CYPs. A panel of 18 HepG2-derived cell lines was established, each of which overexpressed a single isoform of human CYP. We evaluated the metabolic activity of each CYP isoform toward triclosan using reversed-phase high-performance liquid chromatography (HPLC), with online radiochemical detection, and examined the effects of CYP-mediated metabolism of triclosan on its cytotoxicity in HepG2 cells. In addition, the metabolic profile of triclosan in primary human hepatocytes and the role of glucuronidation and sulfonation of triclosan in triclosan-induced cytotoxicity were determined.

Materials and methods

Chemicals and reagents

Triclosan (2,4,4′-trichloro-2′-hydroxydiphenyl ether) was obtained from Alfa Aesar (Ward Hill, MA). Ammonium acetate, 4-chlorocatechol, 2,4-dichlorophenol, 2,2-bis(hydroxymethyl)-2,2′,2″-nitrilotriethanol (Bis–Tris), dimethyl sulfoxide (DMSO), β-nicotinamide adenine dinucleotide 2′-phosphate-reduced tetrasodium salt hydrate (NADPH), thiazolyl blue tetrazolium bromide (MTT), valproic acid, pentachlorophenol, α-naphthoflavone, ticlopidine, and SKF-525A were obtained from Sigma-Aldrich (St. Louis, MO). Acetonitrile, Dulbecco’s modified Eagle medium (DMEM), methanol, penicillin–streptomycin solution, blasticidin, sodium pyruvate, nonessential amino acids, and 2.5% trypsin were purchased from Thermo Fisher Scientific, Inc. (Pittsburgh, PA). Fetal bovine serum (FBS) was acquired from Atlanta Biologicals (Lawrenceville, GA). Bovine collagen-I solution was obtained from Advanced Biomatrix (Carlsbad, CA). [2,4-Dichlorophenyl-14C(U)]triclosan (radiochemical purity ≥97.0% by HPLC; specific activity 80 mCi/mmol) was purchased from Moravek Biochemicals (Brea, CA). [18O]H2O (97 atom% 18O) was obtained from Cambridge Isotope Laboratories, Inc. (Andover, MA).

Antibodies

Goat polyclonal antibody to human CYP2D6 and mouse monoclonal antibodies to human CYP1A2, CYP2A6, CYP3A4, CYP3A7, and β-actin were acquired from Santa Cruz Biotechnology (Santa Cruz, CA). Rabbit polyclonal antibody to human CYP4A11 was obtained from Thermo Fisher Scientific, Inc. Rabbit polyclonal antibodies to human CYP1A1, CYP2A13, CYP2C18, and CYP4B1, rabbit monoclonal antibodies to human CYP1B1, CYP2B6, CYP2C9, CYP2C19, CYP2E1, and CYP3A5, mouse polyclonal antibody to human CYP2A7, and goat polyclonal antibody to human CYP2C8 were obtained from Abcam Inc. (Cambridge, MA).

Cell culture

Human hepatoma HepG2 cells were obtained from American Type Culture Collection (ATCC, Manassas, VA) and cultured in William’s medium E supplemented with 10% FBS and penicillin–streptomycin solution. The 293T cell line used for lentivirus packaging was purchased from Biosettia (San Diego, CA) and maintained in DMEM supplemented with 10% FBS, 1 mM sodium pyruvate, and nonessential amino acids. Cryopreserved primary human hepatocytes from three donors were purchased from In Vitro ADMET Laboratories (Columbia, MD), recovered in Universal Cryopreservation Recovery Medium, seeded into collagen-I coated 96-well plates in Universal Primary Cell Plating Medium. All cells were maintained at 37 °C in a humidified atmosphere with 5% CO2.

Generation of HepG2 cells overexpressing human CYP isoforms

The coding sequence of 18 individual human CYP isoforms (Table 1) was cloned into the lentiviral expression vector pLV-EF1a-IRES-Bsd (Biosettia). The generated human CYP expression vectors or empty vector and viral packaging plasmids (Biosettia) were co-transfected into 293T cells to produce lentivirus stocks. The titrations of the lentivirus stocks were determined with a titer kit provided by Biosettia. HepG2 cells were then infected with the lentivirus carrying human CYP expression vectors or empty vector at a multiplicity of infection of 10. At 48 h post-transfection, cells were passaged and subsequently grown in media containing 5 μg/ml blasticidin for the selection of blasticidin-resistant cells that stably expressed individual human CYPs (HepG2/CYP) or the empty vector (HepG2/vector).

Table 1 GenBank accession numbers for individual human CYP coding sequences
Full size table

When cells reached 80–90% confluence, the culture medium was removed, and the cells were trypsinized and washed three times in phosphate-buffered saline (PBS). The cells were lysed in 5 mM Bis–Tris (pH 7.0) and 0.1 mM EDTA for 30 min on ice, sonicated three times for 10 s each time, and centrifuged at 14,000g for 30 min at 4 °C. The resulting supernatant fraction (cell lysate) was collected, and the protein concentrations of the cell lysates were determined using a BCA protein assay (Thermo Fisher Scientific, Inc.).

Western blot analysis

The levels of human CYP proteins in CYP-overexpressing cell lines were measured by Western blot analysis. Forty micrograms of cell lysate protein was separated by SDS-polyacrylamide gel electrophoresis and transferred onto a PVDF membrane. The membranes were blocked with 5% nonfat milk and incubated with specific primary antibodies against CYP1A1 (1:1000), CYP1A2 (1:1000), CYP1B1 (1:1000), CYP2A6 (1:1000), CYP2A7 (1:500), CYP2A13 (1:1000), CYP2B6 (1:1000), CYP2C8 (1:1000), CYP2C9 (1:1000), CYP2C18 (1:1000), CYP2C19 (1:1000), CYP2D6 (1:1000), CYP2E1 (1:1000), CYP3A4 (1:1000), CYP3A5 (1:1000), CYP3A7 (1:1000), CYP4A11 (1:500), CYP4B1 (1:1000), or β-actin (1:2000), followed by a secondary antibody conjugated to horseradish peroxidase. The blots were then detected by chemiluminescence using Immobilon Western Horseradish Peroxidase Substrate (Millipore Corporation, Billerica, MA), a UVP BioSpectrum AC Imaging System, and VisionWorks LSD Image Acquisition & Analysis Software (UVP LLC, Upland, CA).

Preparation of microsomes from HepG2 cells and HepG2/CYP-overexpressing cell lines

For each cell line, 2 × 108 cells were harvested, suspended in 6 ml of ice-cold microsome preparation buffer (250 mM sucrose, 10 mM Tris, pH 7.4), and sonicated three times for 10 s each. The homogenate was centrifuged at 10,000g for 10 min at 4 °C to remove debris and large organelles, and then the supernatant was centrifuged at 100,000g for 60 min at 4 °C. The pellet was resuspended in 1 ml of microsome preparation buffer, and aliquots were stored at −80 °C. The protein concentration of the microsomal preparation was determined using a BCA Protein Assay kit.

CYPs activity assay

Enzymatic activities of CYPs in microsomes isolated from the HepG2/CYP cell lines were determined using P450-Glo™ assays (Promega, Madison, WI) or Vivid® CYP450 screening assays (Thermo Fisher Scientific) following the manufacturer’s instructions. Briefly, for both assays, 20 μg of microsome protein isolated from HepG2/vector cells or HepG2/CYP-overexpressing cells was incubated with the corresponding CYP substrate (luciferin-CEE for CYP1A1 and CYP1B1, luciferin-1A2 for CYP1A2, luciferin-2B6 for CYP2B6, luciferin-ME for CYP2C8 and CYP4A11, luciferin-H for CYP2C9, luciferin-H EGE for CYP2C19, luciferin-ME EGE for CYP2D6, luciferin-IPA for CYP3A4, luciferin-PFBE for CYP3A5 and CYP3A7, Vivid® CC for CYP2A6, and Vivid® EOMCC for CYP2E1), in the presence of a NADPH regeneration system, in potassium phosphate buffer (pH 7.2). For P450-Glo™ assays, the reaction mixtures were incubated in white 96-well plates at 37 °C for 30 min, followed by the addition of a luciferin detection reagent, and the luminescence was measured with a BioTek Cytation 5 cell imaging multimode reader (BioTek Instruments, Inc., Winooski, VT). For Vivid® CYP450 screening assays, the reaction mixtures were incubated in black 96-well plates at 37 °C for 60 min, and the fluorescence was measured at 415 nm/460 nm excitation/emission. The luminescence and fluorescence values were normalized to that of microsomes isolated from HepG2/vector cells. Enzymatic activities of CYP2A7, CYP2A13, CYP2C18, and CYP4B1 were not determined due to lack of commercially available kits or identified substrates.

Enzymatic activities of CYP1A2, CYP2B6, CYP2C19, and CYP2D6 in HepG2/CYP-overexpressing cells and primary human hepatocytes treated with or without CYP inhibitors were determined using P450-Glo™ assays following the manufacturer’s instructions. At the end of treatment, cells were incubated with PBS (CYP1A2 and CYP2B6) or fresh culture medium (CYP2C19 and CYP2D6) containing the corresponding CYP substrate at 37 °C for 1–3 h, and then 80 μl of PBS or culture medium was transferred to white 96-well plates, followed by the addition of equal volume of luciferin detection reagent. The luminescence was measured with a BioTek Cytation 5 cell imaging multimode reader.

Metabolism of triclosan by human liver or HepG2/CYP-overexpressing cells microsomes

Metabolism of triclosan by pooled human liver microsomes from 150 donors (BD Biosciences, San Jose, CA) and microsomes isolated from HepG2/CYP-overexpressing cells was assayed in a 125-µl final reaction volume by incubating microsomes (1.0–1.8 mg of microsomal protein) with 0.5 mM triclosan, 4 mM MgCl2, 2 mM NADPH, and 100 mM potassium phosphate buffer (pH 7.2) at 37 °C for 30 min with gentle shaking. A screening assay of CYP activity was performed with 0.5 mM [2,4-dichlorophenyl-14C(U)]triclosan (diluted with unlabeled triclosan to a specific activity of 53.4 mCi/mmol). Negative controls for the assay included incubations conducted with microsomes from parent HepG2 cells and HepG2/vector cells as well as incubations conducted in the absence of triclosan or NADPH. All reactions were performed in the linear range with respect to the amount of microsomal protein and incubation time. Reactions were terminated by the addition of an equal volume of ice-cold methanol. Precipitated material was removed by centrifugation (5 min, 14,000g), and the supernatants were analyzed for triclosan metabolites by reversed-phase HPLC.

HPLC analyses were performed on a Waters HPLC system consisting of a 600 Controller, a 996 Photodiode Array detector, and a 717 Plus autosampler (Waters Corporation, Milford, MA). Samples were injected onto a 4.6 × 250 mm C18 (5 µm particle size) Luna column (Phenomenex, Torrance, CA). HPLC separations were performed with acetonitrile (solvent A) and 50 mM ammonium acetate, pH 5.0 (solvent B) as follows: 5 min with 80% solvent B, a 45-min linear gradient to 25% solvent B, and a final 3-min linear gradient to 0% solvent B. The HPLC flow rate was 1 ml/min. The column was washed with 100% solvent A for 10 min and equilibrated for 10 min with 80% solvent B/20% solvent A after every run. When [2,4-dichlorophenyl-14C(U)]triclosan was used, the HPLC system included a β-RAM radioactive flow detector (IN/US, Tampa, FL), with a scintillation fluid flow rate of 2 ml/min. The lower limit of quantification for 2,4-dichlorophenol was 5 pmol, based on radio flow detection and quantification of radioactivity within the 2,4-dichlorophenol HPLC peak as determined using a IN/US β-RAM radioactivity detection program. Intraday and interday precision variation (coefficient of variation) was <10%.

Mass spectrometry and 1H NMR analyses

Mass spectrometry system I was a Waters Acquity ultra-performance liquid chromatography (UPLC) system coupled with a Waters Quattro Premier XE tandem quadrupole mass spectrometer equipped with an electrospray ionization source (ESI–MS/MS). The chromatographic separations were achieved on a Waters BEH C18 column (2.1 × 50 mm, 1.7 µm particle size) maintained at 30 °C using a 2-min linear gradient of 5–100% acetonitrile in 15 mM ammonium acetate at a flow rate of 300 µl/min. The mass spectrometer was operated in the negative electrospray mode.

Mass spectrometry system II was a Waters ACQUITY I-class UPLC system coupled to a Waters Xevo TQ-S ESI–MS/MS. The mass spectrometer was operated in the positive electrospray mode. The chromatographic separations were achieved on a Waters BEH C18 column (2.1 × 50 mm, 1.7 µm particle size) maintained at 50 °C using the following gradient between eluent A (water/0.1% formic acid) and eluent B (acetonitrile): 0–0.3 min 50% B; 0.3- to 4.0-min linear gradient 50–95% B at a flow rate of 500 µl/min. The samples were analyzed in full scan and multiple-reaction monitoring modes.

1H NMR spectroscopy was conducted with a Bruker Avance III spectrometer, equipped with a Bruker BBFO Plus Smart Probe, at 500 MHz and 300°K (Bruker Instruments, Billerica, MA), using DMSO-d 6 as solvent.

Cytotoxicity determinations

Triclosan-induced cytotoxicity was assessed using an MTT assay as previously described (Wu et al. 2015). Briefly, HepG2/vector cells and HepG2/CYP-overexpressing cells were seeded into 24-well plates at 1 × 105 cells/well, allowed to attach to the bottom of the wells for 24 h, and then treated with 15, 30, or 60 μM triclosan for 48 h. Primary human hepatocytes were seeded into collagen-I-coated 96-well plates at 4 × 104 cells/well, allowed to attach to the bottom of the wells for 4 h, and then treated with 30, 60, or 120 μM triclosan in the presence or absence of the UGT inhibitor valproic acid (10 mM) in hepatocyte induction medium supplemented with 10% FBS for 48 h. Control cells were fed with complete culture medium containing 0.1% (v/v) DMSO, which had no effect on cell growth. At the end of the treatment, cells were incubated with fresh culture medium containing 1 mg/ml MTT for 2 h. The resulting formazan was dissolved in DMSO, and the absorbance at 540 nm was determined with a BioTek Cytation 5 cell imaging multimode reader.

Statistical analyses

Data are presented as the mean ± standard deviation of three independent experiments. Comparisons among concentrations were made by one-way analysis of variance, with pairwise comparisons versus control group being performed by Dunnett’s method. Pearson’s correlation analysis was performed using GraphPad Prism 6.0 (GraphPad Software Inc., La Jolla, CA). The results were considered significant at p < 0.05.

Results

Metabolism of triclosan by human CYPs

To determine whether or not triclosan can be metabolized by human CYPs, triclosan was incubated with pooled human liver microsomes, and the reaction mixture was analyzed by reversed-phase HPLC. Triclosan, 4-chlorocatechol, and 2,4-dichlorophenol were readily resolved by reversed-phase HPLC with UV detection (Fig. 1b). Two major metabolites were observed, one of which had an identical retention time to that of 2,4-dichlorophenol, indicating that 2,4-dichlorophenol was formed by human liver microsomes (Fig. 1c, d). The other metabolite, peak X at 33.6 min, was collected, purified, and subjected to structural characterization by mass spectrometry system I and 1H NMR spectroscopy.

Fig. 1
figure 1

HPLC analysis of triclosan metabolites formed by human liver microsomes. a Standard of [2,4-dichlorophenyl-14C(U)]triclosan as assessed by radiochemical detection. Inset chemical structure of [2,4-dichlorophenyl-14C(U)]triclosan. b Standards of triclosan, 2,4-dichlorophenol, and 4-chlorocatechol as assessed by UV detection at 280 nm. c, d Incubation of human liver microsomes with [2,4-dichlorophenyl-14C(U)]triclosan as described in “Materials and methods.” Formation of 2,4-dichlorophenol and peak X as assessed by radiochemical detection (c) and UV detection at 280 nm (d), respectively

Full size image

An analysis of the purified peak X by UPLC-ESI–MS in full scan mode revealed a prominent peak at 1.66 min, whose spectrum consisted of a complex set of anions in the range of m/z 302.9–306.9. The m/z and relative intensity distribution of this ion cluster (Fig. 2a) proved identical to that calculated for a monohydroxylated triclosan anion (C12H6Cl3O3) using the isotope modeling tool in Waters MassLynx software, taking into account the natural isotopic contributions of 35Cl and 37Cl (Fig. 2b). These data indicated that peak X was most likely a monohydroxylated metabolite of triclosan.

Fig. 2
figure 2

Mass spectral analysis of peak X. Experimental (a) and simulated (b) isotope distribution of ions obtained in full scan analysis of a compound with a molecular formula of C12H6Cl3O3, consistent with hydroxytriclosan. Modeling of the isotopic distribution was conducted using the isotope modeling tool in Waters MassLynx software

Full size image

A comparative analysis of the 1H and COSY NMR spectra of triclosan and its hydroxylated metabolite (Table 2) revealed that the oxidation led to a greatly simplified set of resonances in the affected ring system, affording only two prominent singlets at 6.56 and 6.94 ppm. While the absence of the ortho coupling pattern observed versus the parent compound could be justified by hydroxylation in either position 5′ or 6′, the loss of meta coupling observed in proton 3′ (singlet at 6.94 ppm) can only be justified by a hydroxylation in position 5′. The observed upfield shift in the resonance of proton 6′ (+0.45 ppm) is ascribable to the expected positive mesomeric effect in the position ortho to the newly introduced phenol function. These data identified peak X as 5′-hydroxytriclosan.

Table 2 1H NMR assignment of resonances in triclosan and its hydroxylated metabolite
Full size table

Upon the incubation of triclosan with human liver microsomes, 5′-hydroxytriclosan and 2,4-dichlorophenol were detected, suggesting that human CYPs catalyze 5′-hydroxylation and the cleavage of the diphenyl ether bond of triclosan.

The cleavage of the diphenyl ether bond of triclosan

While the detection of 2,4-dichlorophenol in the human liver microsomal incubations demonstrated that the metabolism of triclosan can lead to the cleavage of the diphenyl ether bond, the detailed oxidative mechanism underlying the cleavage remained unclear. In order to understand this mechanism, incubations using pooled human liver microsomes were conducted under the same conditions as described in Materials and Methods using [18O]H2O. After incubation at 37°C for 30 min, an equal volume of acetonitrile was added, and then the mixture was centrifuged to eliminate protein. An aliquot of the supernatant was derivatized with dansyl chloride according to the procedure described by de Ruiter et al. (1988). The derivatized sample was reconstituted in acetonitrile and analyzed by mass spectrometry system II.

As depicted in Fig. 3, the incubations conducted with [16O]H2O (Fig. 3a) and isotopically enriched [18O]H2O (Fig. 3b) afforded 2,4-dichlorophenol whose dansylated derivative isotope patterns were consistent with a structure incorporating a 16O phenol function (simulated in Fig. 3c), but not a 18O phenol function (simulated in Fig. 3d). Conversely, the incubations conducted in [18O]H2O (Fig. 3f) afforded 4-chlorocatechol whose di-dansylated derivative isotope pattern was clearly distinct from that stemming from the incubations conducted in [16O]H2O (Fig. 3e) and that corresponded to a combination of the spectra simulated for 16O (Fig. 3g) and 18O (Fig. 3h) di-dansylated 4-chlorocatechol, with clear enhancements in the ions at m/z 613, 615, and 617. In order to obtain a more quantitative insight into the incorporation of 18O in the incubations, analyses of the two metabolites were also performed in multiple-reaction monitoring mode. Consistent with the isotope profiles observed in full scan mode, while the ratios of the transitions 396.1 > 171.0 and 398.1 > 171.0 for dichlorophenol were identical in the metabolite obtained in [16O]H2O (1.45) and [18O]H2O (1.44), the ratio of the 4-chlorocatechol transitions 611.3 > 170.0 and 613.1 > 170.0 was lower in the metabolite obtained in [18O]H2O (1.15) than in [16O]H2O (2.29), demonstrating the incorporation of 18O in the structure. These data demonstrate that the enzymatic cleavage of the ether function of triclosan involves an oxidative process at carbon 1′ and that water is the source of oxygen for this process. It should be noted that the 4-chlorocatechol isotope pattern obtained in the [18O]H2O incubations (Fig. 3f) reflects the partial (but not complete) enrichment of the total water contained in the incubations with [18O]H2O.

Fig. 3
figure 3

Mass spectral analysis of the dansylated derivatives of the triclosan metabolites 2,4-dichlorophenol (a–d) and 4-chlorocatechol (e–h) by human liver microsomes. Incubations of human liver microsomes with triclosan were performed in [16O]H2O (a, e) or isotopically enriched [18O]H2O (b, f). The recorded transitions were as follows: dansylated dichlorophenol 396.1 > 171.0 (transition for 16O metabolite; cone voltage 40.0 V, collision energy 25 eV), 398.1 > 171.0 (transition for 16O and 18O metabolites; cone voltage 45.0 V, collision energy 25 eV); dansylated chlorocatechol 611.3 > 170.0 (transition for 16O metabolite; cone voltage 50.0 V, collision energy 35 eV), 613.1 > 170.0 (transition for 16O and 18O metabolites; cone voltage 60.0 V, collision energy 35 eV). The ratioα was determined from the peak area attained for transition 396.1 > 171.0 and peak area of transition 398.1 > 171.0. A decrease in this ratio would indicate the incorporation of 18O into the structure of 2,4-dichlorophenol. The ratioβ was determined from the peak area attained for transition 611.3 > 170.0 corresponding to 4-chlorocatechol and peak area of transition 613.1 > 170.0 corresponding to 4-chlorocatechol with incorporated 16O and 18O. A decrease in this ratio would indicate the incorporation of 18O into the structure of 4-chlorocatechol. Theoretical isotope distribution of 16O-containing dansylated 2,4-dichlorophenol (c) and dansylated 4-chlorocatechol (g). Theoretical isotope distribution of 18O-containing dansylated 2,4-dichlorophenol (d) and dansylated 4-chlorocatechol (h). Modeling of the isotopic distribution was conducted using the isotope modeling tool in Waters MassLynx software

Full size image

Metabolism of triclosan by individual human CYP isoform

To identify which human CYP isoforms metabolize triclosan, a panel of HepG2-derived cell lines was established, each of which overexpressed a single human CYP isoform, including CYP1A1, CYP1A2, CYP1B1, CYP2A6, CYP2A7, CYP2A13, CYP2B6, CYP2C8, CYP2C9, CYP2C18, CYP2C19, CYP2D6, CYP2E1, CYP3A4, CYP3A5, CYP3A7, CYP4A11, and CYP4B1. Western blot analysis showed that most of the CYPs examined in this study were not expressed in HepG2 or HepG2/vector cells, with the exception of CYP2C9 and CYP4B1, which were expressed at low levels (Fig. 4). The expression of individual CYP isoform was readily detected in the corresponding HepG2/CYP-overexpressing cells, and the levels were much higher than those in HepG2 or HepG2/vector cells (Fig. 4).

Fig. 4
figure 4

Western blot analysis of individual CYP isoform expression in HepG2/CYP-overexpressing cells. β-Actin served as a loading control

Full size image

Microsomes were isolated from HepG2 cells, HepG2/vector cells, and 18 HepG2/CYP-overexpressing cell lines, and the enzymatic activities of CYPs in these microsomes were validated using known substrates. The activities of four CYP isoforms (CYP2A7, CYP2A13, CYP2C18, and CYP4B1) were not determined due to lack of commercially available kits or known substrates. The other fourteen CYP isoforms were active toward their known substrates (Table 3).

Table 3 Metabolism of triclosan by microsomes isolated from HepG2/CYP-overexpressing cells
Full size table

To elucidate whether or not these human CYPs exhibited activity toward triclosan, a comprehensive screening was conducted. As summarized in Table 3, seven CYP isoforms were capable of metabolizing triclosan, with the order of activity being CYP1A2 > CYP2B6 > CYP2C19 > CYP2D6 ≈ CYP1B1 > CYP2C18 ≈ CYP1A1. The remaining 11 isoforms (CYP2A6, CYP2A7, CYP2A13, CYP2C8, CYP2C9, CYP2E1, CYP3A4, CYP3A5, CYP3A7, CYP4A11, and CYP4B1) had little or no activity toward triclosan. Typical HPLC chromatograms for the analysis of triclosan metabolites are shown in Fig. 5. Incubation of triclosan with microsomes isolated from HepG2/CYP1A2 cells led to the formation of 2,4-dichlorophenol and 5′-hydroxytriclosan (Fig. 5a, b), which is consistent with the data obtained using human liver microsomes. No such metabolites were detected in reactions using HepG2 cells microsomes (Fig. 5c, d) and HepG2/vector cells microsomes, or in incubations lacking triclosan or NADPH.

Fig. 5
figure 5

HPLC analysis of triclosan metabolites by microsomes isolated from HepG2 cells and HepG2/CYP1A2 cells. [2,4-Dichlorophenyl-14C(U)]triclosan was incubated with microsomes isolated from HepG2/CYP1A2 cells (1.0 mg protein, a, b) or HepG2 cells (1.8 mg protein, c and d) as described in “Materials and methods.” The formation of 2,4-dichlorophenol and 5′-hydroxytriclosan was examined by radiochemical (a, c) and UV detection at 280 nm (b, d), respectively

Full size image

4-Chlorocatechol was not detected by reversed-phase HPLC analyses in incubations with human liver microsomes (Fig. 1) and microsomes isolated from HepG2/CYP1A2 cells (Fig. 5). Nonetheless, mass spectrometry analyses clearly demonstrated the formation of 4-chlorocatechol in these incubations (Fig. 3).

Metabolism of triclosan metabolites by human CYPs

Three metabolites, i.e., 5′-hydroxytriclosan, 2,4-dichlorophenol, and 4-chlorocatechol, were formed upon human CYPs metabolism of triclosan. The stability of triclosan and the three metabolites was assessed using reversed-phase HPLC analyses as described above. 5′-Hydroxytriclosan in 100 mM potassium phosphate (pH 7.2) was unstable and decreased by 23% after two weeks at room temperature. The other metabolites were stable under same conditions.

To determine whether or not 5′-hydroxytriclosan, 2,4-dichlorophenol, and 4-chlorocatechol were further metabolized by human CYPs, incubations of human liver microsomes and microsomes isolated from HepG2 cells with each of triclosan metabolites (55 nM) were performed as described in Materials and Methods and then analyzed using reversed-phase HPLC. As shown in Fig. 6, the levels of 2,4-dichlorophenol and 4-chlorocatechol decreased in incubations with human liver microsomes in the presence of NADPH, but remained unchanged in the absence of NADPH, suggesting that both chemicals were further metabolized by human CYPs and were stable at 37 °C for 30 min. HepG2 cells microsomes were also capable of metabolizing 4-chlorocatechol but not 2,4-dichlorophenol (Fig. 6). Incubation of 5′-hydroxytriclosan with either human liver microsomes or HepG2 cells microsomes significantly reduced its amount in the presence of NADPH (Fig. 6). There was also a slight reduction in the level of 5′-hydroxytriclosan in the absence of NADPH (Fig. 6). These data indicate that 5′-hydroxytriclosan was further metabolized by human CYPs and was slightly unstable at 37 °C during the 30-min incubation.

Fig. 6
figure 6

Metabolism of 2,4-dichlorophenol, 4-chlorocatechol, and 5′-hydroxytriclosan by human liver microsomes and HepG2 cells microsomes. 2,4-Dichlorophenol, 4-chlorocatechol, and 5′-hydroxytriclosan were incubated with 1 mg of human liver microsomes or HepG2 cells microsomes as described in “Materials and methods.” The amount of each chemical was determined by reversed-phase HPLC and was compared to the standard without any incubation. The results shown are the mean and standard deviation of three independent experiments. *Significantly (p < 0.05) different from the standard

Full size image

Effect of CYP-mediated metabolism on triclosan-induced cytotoxicity

To investigate the effect of CYP-mediated metabolism of triclosan on its cytotoxicity, the cytotoxicity of triclosan was compared among HepG2/vector cells and 18 HepG2/CYP-overexpressing cell lines using an MTT assay. After 48 h of exposure, the culture media were collected for the analysis of triclosan and its metabolites. In accord with the in vitro screening data, triclosan was metabolized in HepG2 cells overexpressing CYP1A2, CYP1B1, CYP2B6, CYP2C18, CYP2C19, and CYP2D6. Typical HPLC chromatograms, shown in Fig. 7, indicated the presence of 2,4-dichlorophenol both in the culture media and in the HepG2/CYP1A2 cells (Fig. 7f, g), but not in HepG2/vector culture media or cells (Fig. 7c, e). This metabolism was accompanied by a substantial decrease in the level of triclosan in the culture media (Fig. 8a). The viability of these cells following a 48 h of exposure to triclosan was much higher than that of HepG2/vector cells (Fig. 8b), suggesting that CYP-mediated metabolism of triclosan attenuated its cytotoxicity. CYP1A1 also exhibited metabolic activity toward triclosan in the in vitro microsomal screening assay (Table 3); however, minimal levels of 2,4-dichlorophenol were detected in the culture media of HepG2/CYP1A1 cells.

Fig. 7
figure 7figure 7

HPLC analysis of triclosan metabolites formed in HepG2/vector cells and HepG2/CYP1A2 cells. a Standards of triclosan glucuronide, triclosan sulfate, 4-chlorocatechol, 2,4-dichlorophenol, and triclosan as assessed by UV detection. HepG2/vector cells (b–e) or HepG2/CYP1A2 cells (f, g) were treated with 0.1% DMSO (b, d) or 30 μM triclosan (c, e, f, g) for 48 h. The formation of triclosan glucuronide, triclosan sulfate, and 2,4-dichlorophenol in the culture medium (b, c, f) or intracellularly (d, e, g) was monitored by UV detection at 280 nm

Full size image
Fig. 8
figure 8

Effect of CYP-mediated metabolism on triclosan-induced cytotoxicity. a The levels of triclosan in the culture medium of HepG2/vector, HepG2/CYP1A2, HepG2/CYP1B1, HepG2/CYP2B6, HepG2/CYP2C18, HepG2/CYP2C19, and HepG2/CYP2D6 cells following a 48-h exposure to 30 μM triclosan. b The viability of HepG2/vector, HepG2/CYP1A2, HepG2/CYP1B1, HepG2/CYP2B6, HepG2/CYP2C18, HepG2/CYP2C19, and HepG2/CYP2D6 cells treated with 15–60 μM of triclosan for 48 h. c The viability of HepG2/vector cells treated with 15–60 μM of triclosan, 2,4-dichlorophenol, 4-chlorocatechol, or a mixture of 2,4-dichlorophenol and 4-chlorocatechol for 48 h. The results shown are the mean and standard deviation of three independent experiments. *Significantly (p < 0.05) different from HepG2/vector cells. #Significantly (p < 0.05) different from cells treated with triclosan

Full size image

The cytotoxicity of triclosan and its two CYP-mediated metabolites 2,4-dichlorophenol and 4-chlorocatechol was further compared in HepG2/vector cells. At the same molarity, both 2,4-dichlorophenol and 4-chlorocatechol were less toxic than triclosan (Fig. 8c). The cleavage of the diphenyl ether bond of triclosan leads to the formation of equimolar quantities of 2,4-dichlorophenol and 4-chlorocatechol. To mimic this situation, HepG2/vector cells were exposed to an equimolar mixture of 2,4-dichlorophenol and 4-chlorocatechol, and the cell viability was found to be significantly higher than that of cells exposed to the same concentration of triclosan (Fig. 8c). These data further corroborate the findings that CYP-mediated metabolism of triclosan attenuated its cytotoxicity.

Effect of UGT- and SULT-mediated metabolism on triclosan-induced cytotoxicity

Although CYP-mediated oxidation of triclosan did not occur in HepG2/vector cells (Fig. 7c, e), the formation of triclosan glucuronide and triclosan sulfate in the culture media of HepG2/vector cells (Fig. 7c) indicates that triclosan undergoes UGT- and SULT-mediated metabolism in HepG2/vector cells. This is in agreement with the fact that HepG2 cells have basal expression of several isoforms of UGT and SULT (Guo et al. 2011). As shown in Fig. 9a, there was a significant increase in the level of triclosan glucuronide and triclosan sulfate, with a concomitant decrease in the level of triclosan, in the culture media of HepG2 cells overexpressing CYP2A6, CYP2A13, CYP2C9, and CYP4A11 compared to HepG2/vector cells. The greater extent of UGT- and SULT-mediated metabolism led to greater viability of these cells compared with HepG2/vector cells following triclosan treatment (Fig. 9b), which suggested that UGT- and SULT-mediated metabolisms of triclosan are detoxification metabolic pathways. An opposite scenario applies to HepG2/CYP1A1 cells and CYP2A7 cells, in which a decreased extent of UGT- and SULT-mediated metabolism resulted in higher levels of triclosan and greater cytotoxicity (Fig. 9c, d). The effects of triclosan on cell growth were comparable between HepG2/vector cells and HepG2 cells overexpressing CYP2C8, CYP2E1, CYP3A4, CYP3A5, CYP3A7, and CYP4B1 (Fig. 9f). The levels of triclosan glucuronide and triclosan sulfate in the culture media of these cells were either unchanged or significantly changed to a very slight extent, which led to less than a 20% alteration in the levels of triclosan (Fig. 9e).

Fig. 9
figure 9

Effect of UGT- and SULT-mediated metabolism on triclosan-induced cytotoxicity. a, c, e The levels of triclosan, triclosan glucuronide, and triclosan sulfate in the culture medium of HepG2/CYP-overexpressing cells following 48-h exposure to 30 μM triclosan. b, d, f The viability of HepG2/CYP-overexpressing cells treated with 15–60 μM of triclosan for 48 h. The results shown are the mean and standard deviation of three independent experiments. *Significantly (p < 0.05) different from HepG2/vector cells

Full size image

To characterize further the effects of UGT- and SULT-mediated metabolism on triclosan-induced cytotoxicity, inhibitors of human UGTs and SULTs were used. Co-treatment of HepG2 cells with triclosan and valproic acid, a pan human UGT inhibitor (Braun Trapnell et al. 1998; Ethell et al. 2003), for 48 h led to a marked decrease in the levels of triclosan glucuronide and triclosan sulfate compared to cells treated with triclosan only (Fig. 10a). The addition of pentachlorophenol, a SULT inhibitor (Wang and James 2006), significantly decreased the levels of triclosan sulfate, while it had no significant effects on the levels of triclosan glucuronide (Fig. 10a). The inhibition of either UGTs or SULTs resulted in increased levels of triclosan (Fig. 10a), which was accompanied by decreased cell viability following triclosan exposure (Fig. 10b). Valproic acid or pentachlorophenol alone was not cytotoxic at the tested concentrations. These data further substantiate the observations that the glucuronidation and sulfonation of triclosan attenuated its cytotoxicity.

Fig. 10
figure 10

Effect of UGTs and SULTs inhibition on triclosan-induced cytotoxicity. HepG2/vector cells were treated with 0.1% DMSO or 30 μM triclosan in the presence or absence of the UGT inhibitor valproic acid (1 mM) or the SULT inhibitor pentachlorophenol (10 μM) for 48 h. The levels of triclosan, triclosan glucuronide, and triclosan sulfate in the culture medium (a) and cell viability (b) were determined. The results shown are the mean and standard deviation of three independent experiments. *Significantly (p < 0.05) different from cells treated without inhibitor

Full size image

Correlation analysis between cell viability and the levels of triclosan, triclosan glucuronide, or triclosan sulfate

An inverse correlation was observed between the extracellular levels of triclosan and cell viability (Fig. 11a) in all the 18 HepG2/CYP-overexpressing cell lines. Among the 12 HepG2/CYP cell lines that do not have CYP metabolic activity toward triclosan, cell viability was found to be positively associated with the levels of triclosan glucuronide (Fig. 11b) and triclosan sulfate (Fig. 11c). The correlation analysis implied that triclosan, rather than its metabolites, is responsible for the cytotoxicity observed in HepG2 cells.

Fig. 11
figure 11

Correlation analysis between cell viability and levels of triclosan, triclosan glucuronide, or triclosan sulfate. a Correlation between levels of triclosan and cell viability among 18 HepG2/CYP-overexpressing cell lines. b, c Correlation between levels of triclosan glucuronide (b) or triclosan sulfate (c) and cell viability among 12 HepG2/CYP-overexpressing cell lines that had no CYP metabolic activity toward triclosan

Full size image

Metabolism of triclosan in primary human hepatocytes

The metabolic profile of triclosan and the role of CYP, UGT, and SULT in triclosan-induced cytotoxicity were further investigated in primary human hepatocytes from three different donors. The demographic information of the three donors and the activity data of drug-metabolizing enzymes provided by the supplier are presented in Supplementary Tables 1 and 2. To understand better the levels of CYPs activities in primary hepatocytes compared to HepG2/CYP-overexpressing cells, the enzymatic activities of CYP1A2, CYP2B6, CYP2C19, and CYP2D6, the four major CYP isoforms that metabolized triclosan in vitro, were determined in primary human hepatocyte preparation HH1072 and in HepG2/CYP-overexpressing cells in the presence and absence of a CYP inhibitor. CYP1A2, CYP2B6, and CYP2C19 activities were much lower in HH1072 cells compared to the corresponding HepG2/CYP-overexpressing cells (Fig. 12). Although the luminescence value of CYP2D6 reaction mixture from HH1072 cells was higher than that of HepG2/CYP2D6 cells, there was only a 14% decrease in the luminescence value when HH1072 cells were treated with a CYP inhibitor SKF-525A, compared to a 94% decrease in HepG2/CYP2D6 cells (Fig. 12). This could be due to the fact that luciferin-ME EGE, the CYP2D6 substrate used in this assay, was not specific to CYP2D6. Similar results in the enzymatic activities of the four CYP isoforms were observed in the HH1051 and HH1076 primary hepatocyte preparations.

Fig. 12
figure 12

CYPs activities in primary human hepatocytes compared to HepG2/CYP-overexpressing cells. Enzymatic activities of CYP1A2, CYP2B6, CYP2C19, and CYP2D6 were determined in primary human hepatocyte preparation HH1072 and the corresponding HepG2/CYP-overexpressing cells in the presence or absence of CYP inhibitors (10 μM α-naphthoflavone for CYP1A2, 20 μM ticlopidine for CYP2B6, or 10 μM SKF-525A for CYP2C19 and CYP2D6) for 24 h. Luminescence values of reaction mixtures are shown as percentage of those from HepG2/CYP-overexpressing cells without any inhibitor

Full size image

HPLC analysis of triclosan metabolites in the culture medium from the primary human hepatocytes showed that triclosan was metabolized predominantly to triclosan glucuronide followed by triclosan sulfate, with 2,4-dichlorophenol being barely detectable. This was in agreement with the low CYP activities in these primary hepatocytes and suggested that UGT- and SULT-mediated conjugation of triclosan was the major metabolic pathway in primary human hepatocytes. Co-treatment of HH1051 cells with triclosan and valproic acid for 48 h led to a marked decrease in the levels of triclosan glucuronide and triclosan sulfate compared to cells treated with triclosan only (Fig. 13a). Similar treatment of HH1072 and HH1076 cells with valproic acid significantly decreased the level of triclosan sulfate, while this had no effect on the level of triclosan glucuronide (Fig. 13a). The inhibition of UGTs and/or SULTs resulted in increased levels of triclosan (Fig. 13a) in the three primary human hepatocyte preparations, which was accompanied by decreased cell viability following triclosan exposure (Fig. 13b). Valproic acid alone was not cytotoxic at the tested concentration. These data were in accordance with those obtained in HepG2 cells and corroborated the findings that the glucuronidation and sulfation of triclosan attenuated its cytotoxicity in primary human hepatocytes. Notably, triclosan at 30 μM did not exhibit any cytotoxicity in primary human hepatocytes (Fig. 13b), indicating that the greater extent of UGT-mediated metabolism in primary human hepatocytes compared to HepG2/vector cells made them less vulnerable to triclosan-induced cytotoxicity.

Fig. 13
figure 13

Metabolism of triclosan in primary human hepatocytes. a The levels of triclosan, triclosan glucuronide, and triclosan sulfate in the culture medium of human primary hepatocytes (HH1051, HH1072, and HH1076) following 48-h exposure to 60 μM triclosan in the presence or absence of valproic acid (10 mM). Values are shown as fold of HH1051 cells without valproic acid. b The viability of primary human hepatocytes treated with 30–120 μM of triclosan for 48 h in the presence or absence of valproic acid (10 mM). The results shown are the mean and standard deviation of triplicate wells. *Significantly (p < 0.05) different from 0 μM triclosan. #Significantly (p < 0.05) different from cells treated without valproic acid

Full size image

Discussion

Previous studies have reported the presence/detection of 2,4-dichlorophenol, hydroxytriclosan, and 4-chlorocatechol in the urine and/or feces of rats and mice exposed to triclosan (Fang et al. 2016a; Tulp et al. 1979), indicating the occurrence of both aromatic hydroxylation and cleavage of the diphenyl ether bond of triclosan. However, the present study is the first study to characterize the metabolism of triclosan by human CYPs. Our data demonstrated that human CYPs catalyze hydroxylation at the C5′ position and the cleavage of the diphenyl ether bond at the C1′ position of triclosan (Fig. 14) to give rise to 5′-hydroxytriclosan, 2,4-dichlorophenol, and 4-chlorocatechol. Among 18 human CYPs tested, six CYP isoforms (i.e., CYP1A2, CYP1B1, CYP2B6, CYP2C18, CYP2C19, and CYP2D6) metabolized triclosan, and their metabolic activities toward triclosan attenuated the cytotoxicity induced by triclosan in HepG2 cells.

Fig. 14
figure 14

Simplified scheme of triclosan metabolism

Full size image

In vitro incubation of microsomes isolated from HepG2/CYP1A2 cells with triclosan for 30 min yielded three triclosan metabolites, 2,4-dichlorophenol, 4-chlorocatechol, and 5′-hydroxytriclosan, whereas exposure of HepG2/CYP1A2 cells to triclosan for 48 h led to the formation of only low levels of 2,4-dichlorophenol, with the other two metabolites being undetectable. This could indicate that both 4-chlorocatechol and 5′-hydroxytriclosan undergo subsequent human CYP-catalyzed metabolism (Fig. 6) and/or conjugation reactions. In addition, 4-chlorocatechol and 5′-hydroxytriclosan may spontaneously oxidize to quinone derivatives, which could explain the instability of 5′-hydroxytriclosan in the incubations conducted in the absence of NADPH (Fig. 6).

HepG2 cells have been demonstrated to be of great value in toxicological studies (Dykens et al. 2008; Felser et al. 2013; Greer et al. 2010; Juan-García et al. 2013; Nguyen et al. 2013). Although low levels of CYP2C9 and CYP4B1 were detected in HepG2 cells, the other CYP isoforms examined in the current study were absent in HepG2 cells (Fig. 4). In addition, CYP-mediated metabolism of triclosan was not observed in either HepG2 cells or HepG2/vector cells (Fig. 7). The absence or low basal level of CYPs expression makes HepG2 cells an appropriate model for the overexpression of individual CYP isoforms.

CYPs represent the most important drug-metabolizing enzymes. The human genome contains as many as 57 CYP functional genes, which are divided into 18 families (Sim and Ingelman-Sundberg 2010). In our study, we selected 18 CYP isoforms from four families (CYP1, CYP2, CYP3, and CYP4) because these are the major CYPs that metabolize exogenous substances (Zanger and Schwab 2013). The expression levels and activities of CYPs vary to a great extent among different individuals due to a number of factors, such as genetic polymorphisms, xenobiotic induction or inhibition, hormone regulation, sex, and age (Zanger and Schwab 2013; Zhou et al. 2009). For instance, CYP1A2 is potently induced by smoking, and the mRNA level of CYP1A2 in human liver exhibits more than a 40-fold interindividual difference (Gunes and Dahl 2008). More than 20 genetic variant alleles have been identified in human CYP2C19 and CYP2D6, and a number of these are null alleles that led to the complete loss of enzymatic activity in individuals carrying these homozygous alleles (Desta et al. 2002; Zhou 2009). Our findings that triclosan is metabolized by CYP1A2, CYP1B1, CYP2B6, CYP2C18, CYP2C19, and CYP2D6 suggest that human populations may have different internal exposure burdens to triclosan depending on their metabolic activity. Nonetheless, caution should be exercised in extrapolating the findings obtained in HepG2/CYP-overexpressing cells because the extremely high level of CYP activity in these cells may not be readily achieved in vivo.

We have previously found that mice dermally treated with 100 mg triclosan/kg body weight have plasma levels of total triclosan and its metabolites of approximately 150 µM (Fang et al. 2016a), and mice topically applied with similar dose of triclosan for 13 weeks showed signs of liver injury (Fang et al. 2015). The present study demonstrated that triclosan exhibited greater cytotoxicity than its metabolites in hepatic cells. The blood levels of total triclosan in humans following use in either mouth rinses or dentifrices vary from 1.4 nM to 1.4 μM (Bagley and Lin 2000; Calafat et al. 2008; DeSalva et al. 1989; Sandborgh-Englund et al. 2006). Triclosan may not represent a significant hepatotoxic hazard to general human populations because the cytotoxic concentration of triclosan observed in our study is higher than general human exposure levels; however, individuals with lower expression levels or activities of triclosan-metabolizing enzymes might be exposed internally to higher levels of triclosan and, thus, be more prone to potential triclosan-induced hepatotoxicity.

In addition to CYP metabolism, UGTs- and SULTs-mediated metabolism of triclosan occurred in HepG2 cells and primary human hepatocytes. This is consistent with previous findings that triclosan was readily metabolized to triclosan glucuronide and triclosan sulfate (Fang et al. 2010, 2016a; Moss et al. 2000; Wang et al. 2004). The levels of triclosan glucuronide and triclosan sulfate varied among the 18 HepG2/CYP-overexpressing cell lines and HepG2/vector cells, which were all originally derived from HepG2 cells. This variation could be because (1) overexpression of CYPs has some effects on the expression levels or activities of UGTs and/or SULTs; (2) the lentivirus carrying the desired CYP gene may have been inserted into the regulatory elements of UGTs and/or SULTs, which could impact the expression levels of UGTs and/or SULTs; and/or (3) the levels or activities of UGTs and/or SULTs differed among different passages of cells. The primary human hepatocytes results further indicated that UGT- and SULT-mediated metabolism of triclosan was major metabolic pathway when CYP activities were low. Notably, valproic acid decreased the levels of triclosan glucuronide in HH1051 cells but not in HH1072 and HH1076 cells. This could be due to the differential expression of UGT isoforms in primary human hepatocytes from different donors since valproic acid only inhibits certain isoforms of UGT (Ethell et al. 2003). Collectively, these data point out the importance of assessing the contribution of UGT- and SULT-mediated metabolism when investigating the relationship between CYP metabolism of drugs and their toxicity.

In conclusion, we have identified the major human CYP isoforms that metabolize triclosan in vitro, and demonstrated that metabolism of triclosan by CYPs, UGTs, and SULTs attenuated its cytotoxicity in hepatic cells.