In Silico Identification of Bioremediation Potential: Carbamazepine

Dec 15, 2016 - While there are many examples of success, the explosion of new ...... S.; Kim , C. G. Degradation of the long-resistant pharmaceutical ...
0 downloads 0 Views 1MB Size
Article pubs.acs.org/est

In Silico Identification of Bioremediation Potential: Carbamazepine and Other Recalcitrant Personal Care Products Kelly G. Aukema,†,‡ Diego E. Escalante,‡,§ Meghan M. Maltby,†,‡ Asim K. Bera,†,‡,∥ Alptekin Aksan,‡,§ and Lawrence P. Wackett*,†,‡ †

Department of Biochemistry, Molecular Biology and Biophysics, ‡BioTechnology Institute, and §Department of Mechanical Engineering, University of MinnesotaTwin Cities, Minneapolis, Minnesota 55455, United States S Supporting Information *

ABSTRACT: Emerging contaminants are principally personal care products not readily removed by conventional wastewater treatment and, with an increasing reliance on water recycling, become disseminated in drinking water supplies. Carbamazepine, a widely used neuroactive pharmaceutical, increasingly escapes wastewater treatment and is found in potable water. In this study, a mechanism is proposed by which carbamazepine resists biodegradation, and a previously unknown microbial biodegradation was predicted computationally. The prediction identified biphenyl dioxygenase from Paraburkholderia xenovorans LB400 as the best candidate enzyme for metabolizing carbamazepine. The rate of degradation described here is 40 times greater than the best reported rates. The metabolites cis-10,11-dihydroxy-10,11-dihydrocarbamazepine and cis-2,3-dihydroxy-2,3-dihydrocarbamazepine were demonstrated with the native organism and a recombinant host. The metabolites are considered nonharmful and mitigate the generation of carcinogenic acridine products known to form when advanced oxidation methods are used in water treatment. Other recalcitrant personal care products were subjected to prediction by the Pathway Prediction System and tested experimentally with P. xenovorans LB400. It was shown to biodegrade structurally diverse compounds. Predictions indicated hydrolase or oxygenase enzymes catalyzed the initial reactions. This study highlights the potential for using the growing body of enzyme−structural and genomic information with computational methods to rapidly identify enzymes and microorganisms that biodegrade emerging contaminants.



INTRODUCTION There is increasing concern about rising levels and elevated effects from the presence of personal care product chemicals in waters used by society, especially for those compounds that resist biodegradation and are biologically active in humans.1−3 Pharmaceuticals are perhaps most problematic as they are designed for metabolic stability and to have strong effects on the circulatory, hormonal, nervous, or immune systems. Carbamazepine is heavily prescribed for the treatment of epilepsy, schizophrenia, and sporadic pain, making it one of the most commonly identified pharmaceuticals in rivers, wastewater treatment plants, and irrigation waters.4,5 Alarmingly, carbamazepine is now being detected in the urine of humans who have never taken the drug, a potential concern for vulnerable individuals such as the elderly and children.6−8 Carbamazepine, in particular, is an issue because it strongly resists microbial biodegradation in wastewater treatment plants2,4,9 and is sequestered into crops irrigated with recycled waters.7,8,10 The problem can be exacerbated by water treatment methods that remove other contaminants, both chemical oxidation and biological treatments, that have been shown to transform carbamazepine into persistent and carcinogenic acridines.11−14 The ideal treatment for carbama© XXXX American Chemical Society

zepine would be a rapid biological transformation that removes its biological activity and prevents the formation of acridines, but despite more than 30 years of study, why carbamazepine shows such poor biodegradability has remained unclear. Carbamazepine is just one example in a long list of pharmaceuticals, detergents, plasticizers, medical imaging agents, fragrances, and other commonly used chemicals that are found entering wastewater treatment systems and, increasingly, in the treated water because of their poor biodegradability.1−5 The traditional approach for discovering new bioremediation potential is to study mixed cultures, enrich for pure cultures, or screen known organisms for new activities. While there are many examples of success, the explosion of new compounds and their dissemination in drinking water supplies suggest that new approaches are necessary to keep pace with the introduction of new chemicals. Computational approaches to identify enzymes and microorganisms capable of transforming specific compounds are Received: Revised: Accepted: Published: A

August 26, 2016 November 30, 2016 December 15, 2016 December 15, 2016 DOI: 10.1021/acs.est.6b04345 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

ethanol), sertraline and cotinine (1 mg/mL of methanol), and trimethoprim (0.5 mg/mL of methanol). Serum bottles (125 mL) were crimp-sealed with polytetrafluoroethylene (PTFE)backed silicone seals and shaken at 100 rpm and 25 °C for 24 h. HPLC/DAD Analysis. For samples analyzed by highperformance liquid chromatography (HPLC), sample bottles were unsealed and 1.5 mL of assay mixture was spun at 14000g for 1 min to remove cells. The supernatant was passed through a 0.2 μm PTFE filter prior to analysis by UV absorption on a Hewlett-Packard HP 1090 liquid chromatograph (Agilent Technologies, Santa Clara, CA); 100 μL samples were injected on an Agilent Eclipse plus C18 column. Carbamazepine, lamotrigine, sulfamethoxazole, and metoprolol were analyzed with isocratic elution of a 60% water/0.1% trifluoroacetic acid (TFA)/40% acetonitrile mixture at a rate of 1 mL/min for 10 min. The ratio of the water and TFA/acetonitrile mixture was changed to 80/20 for atenolol, trimethoprim, and sulfthiazole and to 20/80 for gemfibrozil. UV detection wavelengths and elution times are provided in the Supporting Information. HPLC/MS/MS Analysis. The reaction mixtures from resting cell assays with 150 ppm carbamazepine were analyzed by HPLC−MS/MS on an Acquity UPLC BEH C18 column (1.7 μm, 2.1 mm × 100 mm) (Waters Corp., Milford, MA) with a Q Exactive Hybrid Quadrupole Orbitrap Mass Spectrometer (Thermo Fisher, Waltham, MA). Samples were eluted with a constant flow rate of 0.3 mL/min and a shallow gradient of A (water with 0.1% formic acid) and B (acetonitrile with 0.1% formic acid): T (minutes)/%B, 0.5/2%, 1/35%, 7.5/38%, 8/ 95%. MS data were acquired by dynamically selecting the top five most abundant not-yet-collected precursor ions from the survey scans (m/z 200−500). Fragmentation was achieved with high-energy collisional dissociation at 70 V. GC/MS/FID Analysis. Resting cell assay samples analyzed by gas chromatography/mass spectrometry/flame ionization detection (GC/MS/FID) were extracted by injection of 2 mL of ethyl acetate and then vortexed for 10 s. The ethyl acetate layer was analyzed by GC/MS/FID. Separation was achieved on an HP-1ms column (100% dimethylsiloxane capillary; 30 m × 250 m × 0.25 m), a helium flow rate of 1.75 mL/min, and an injection port temperature of 250 °C. The sample was split at the column outlet between a flame ionization detector (HP7890A, Agilent Technologies, Santa Clara, CA) and a mass spectrometer (HP5975C). Electron impact mass spectra were collected using positive polarity and 70 eV. Computational Methods. The X-ray structures for the four different enzymes were obtained from the Protein Data Bank using the following accession codes: 3EN1 (TDO from Ps. putida F1),21 1O7G (NDO from Pseudomonas sp. strain 9816-4),22 2GBX (BPDO from S. yanoikuyae B1),23 and 2XRX (BPDO from P. xenovorans LB400).24 All of the enzyme structures were prepared using the Schrodinger Protein Preparation Wizard software package.25 Missing amino acid side chains and hydrogen atoms were added. The partial charges for each of the amino acid atoms were assigned on the basis of the OPLS_2005 force field, except for the iron and 2His-1-carboxylate facial triad (see the Supporting Information for additional details). The prepared files were then used for docking using the Glide application in the Schrodinger suite of software.25 The nonbonded interaction energy (van der Waals and Coulomb) was recorded for all the docking poses. Additional docking parameters and methods are provided in the Supporting Information. The channel access algorithm has been described elsewhere.26 Briefly, molecular dynamics

needed to expedite the discovery of bioremediation potential given millions of biodegradative enzymes available in genomic databases. Current computational tools in the field of biodegradation such as EAWAG-PPS and EPA-BIOWIN are designed to predict the extent of biodegradation and possible metabolites formed via general enzyme classes.15,16 In pharmaceutical research, molecular docking and molecular dynamics simulation of enzyme movement are methods routinely used to rationally design small molecule−enzyme interactions. Mammalian P450s have been extensively studied in drug metabolism, including for carbamazepine and other emerging pollutants.17,18 The study presented here aims to extend substrate prediction capabilities beyond P450s to other oxygenases known to be important for bioremediation and to catalog substrate predictions in a publically available database, RAPID.19 In this work, computational analysis was used to assess the biodegradability of polycyclic compounds like carbamazepine, and a potential explanation for its lack of biodegradability is suggested. Other compounds were tested computationally for their metabolic pathways and the types of enzymes that would react with them. Computational methods led to the identification of a specific biphenyl dioxygenase that rapidly oxidized carbamazepine to non-biologically active products. Paraburkholderia xenovorans LB400, which naturally harbors the reactive biphenyl dioxygenase, also contains other oxygenases and hydrolases that biodegrade other recalcitrant personal care products. Some of these compounds have not previously been demonstrated to be biodegradable with a single bacterial strain.



MATERIALS AND METHODS Bacterial Strains and Growth. The Rieske dioxygenasecontaining strains Sphingobium yanoikuyae B1, P. xenovorans LB400, and Pseudomonas sp. strain NCIB 9816-4 were grown at 28 °C in minimal medium containing 0.075% finely ground biphenyl or naphthalene according to the oxygenase substrate preference. P. xenovorans LB400 cultures also contained 0.025% yeast extract. S. yanoikuyae B1 and P. xenovorans LB400 express biphenyl 2,3-dioxygenases (BPDO), and Pseudomonas sp. strain NCIB 9816-4 expresses naphthalene 1,2-dioxygenase (NDO). Pseudomonas putida F1 was grown at 28 °C in minimal salts buffer (MSB) with toluene supplied in a vapor bulb to express toluene 2,3-dioxygenase (TDO). Additional growth details are provided in the Supporting Information. Escherichia coli DH5α was grown in LB medium at 37 °C. E. coli BL21 (DE3) expressing LB400 biphenyl dioxygenase (BPDO) from pT7-6a was grown at 37 °C in LB medium containing 100 mg/mL ampicillin to maintain the plasmid.20 Protein expression was induced with 0.5 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) and performed overnight at 28 °C. Resting Cell Assays. The majority of unused biphenyl or naphthalene was removed by filtering cultures through glass wool prior to the cultures being harvested at 6000g for 10 min. Cells were suspended at 0.1 g (wet weight) per milliliter in phosphate-buffered saline (PBS), 10 mM PO43−, 137 mM NaCl, and 2.7 mM KCl (pH 7.4). For each resting cell assay, 0.5 mL of the appropriate cell suspension was incubated with 3 mL of each compound at 10 ppm in PBS. Carbamazepine was also tested at 10 and 150 ppm in larger volumes for removal rate calculations as indicated in the Supporting Information. All chemicals and suppliers are listed in the Supporting Information. Stock solutions (10 mg/mL) were made in ethanol with the exception of primidone (0.7 mg/mL of B

DOI: 10.1021/acs.est.6b04345 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

compounds are typically biodegraded via initial oxidation by microbial Rieske oxygenases (Figure 1b).23 Rieske oxygenase genes have been identified in thousands of bacterial genomes and exist at levels on the order of 107−109 copies/g of sediment.29−31 Moreover, many well-studied Rieske dioxygenases have very broad substrate specificity that we are cataloguing on the RAPID online database to better use natural enzymes for biocatalysis and biodegradation.19,32,33 While Rieske enzymes oxidize hundreds of aromatic compounds, carbamazepine has, to the best of our knowledge, not been tested directly with these enzymes. With thousands of potential Rieske proteins to screen, carbamazepine was investigated with representative dioxygenases computationally first (Figure 2). The study was limited to Rieske dioxygenases for which X-ray crystal structures have been determined with and without a substrate in the active site. No homology models were used. Of the eight available X-ray structures that meet these criteria, four enzymes with well-characterized substrate specificity were used for computational docking and active site tunnel access. The enzymes and structures are toluene 2,3dioxygenase (TDO) from Ps. putida F1 (3EN1),21 naphthalene 1,2-dioxygenase (NDO) from Pseudomonas sp. strain 9816-4 (1O7G),22 biphenyl 2,3-dioxygenase (BPDOB1) from S. yanoikuyae B1 (2GBX),23 and biphenyl 2,3-dioxygenase (BPDOLB400) from P. xenovorans LB400 (2XRX).24 The two key energy values that determine if a compound fits properly in the active site of the dioxygenase enzymes are the Coulomb and van der Waals (nonbonded) interaction energy. Therefore, if a compound docked in the active site with a positive interaction energy value (E > 0 kcal/mol) or was unable to be docked by the software because of an excessively high interaction energy, these results would indicate a compound will not be oxidized by the dioxygenase. By this criterion, TDO was excluded, and BPDOB1 appeared to be unlikely to act on carbamazepine (Figure 2a). Rieske dioxygenases have a buried active site, likely providing some control of substrate specificity, and carbamazepine, unlike naphthalene, was deemed to be unable to gain entry through the access channel to the NDO using an algorithm developed for Rieske oxygenases (Figure 2b).26 The algorithm calculates the free energy of interaction between the enzyme and the compound as it traverses the channel connecting the solvent region (>20 Å from iron) and the active site pocket (3−10 Å from iron). In contrast, BPDOLB400 passed all criteria of active site access and productive docking. The results of molecular docking simulations were used to predict if the presence of the amide functionality might exclude carbamazepine from reacting with many Rieske dioxygenases. The amide group of benzamide, a known inhibitor of NDO, has been proposed to coordinate via the nitrogen through a water molecule to the active site iron atom. Such iron coordination could preclude oxygen binding and thus prevent the catalytic cycle from proceeding. This type of interaction has been observed directly in the 1.65 Å resolution X-ray structure of NDO with benzamide (4HKV).34 While a nonproductive docking orientation with the nitrogen near the iron is favored for BPDOB1 (Figure 2c), this is not the case with BPDOLB400. The favored docking pose is consistent with a carbamazepine orientation allowing oxidation of a carbon−carbon double bond (Figure 2d). The distance between the iron in BPDOLB400 and C3 of carbamazepine is 4.3 Å. This distance is in the same range as the equivalent Fe−C distances observed in NDO (1O7G), TDO (3EN1), BPDOB1 (2GBX), and BPDOLB400 (2XRX) structures with a substrate bound. Furthermore, the

simulation of the dioxygenase was performed for 40 ns. Then using 10% of the static frames or snapshots of the enzyme from the simulation, electrostatic interactions were calculated for the nonbonded interactions of naphthalene and carbamazepine with the tunnel residues as the entrance trajectory into the active site was simulated. After the MD simulations and structure preparations are complete, docking and tunnel simulations can be completed in minutes. The molecular dynamics (MD) simulation of carbamazepine used the docking pose obtained as the initial input structure. The simulation was run using Desmond to generate an ensemble of energetically accessible structures.27 The MD system was first relaxed by a series of energy minimizations and short MD simulations, where the temperature of the system was gradually increased from 0 to 300 K, using the default equilibration protocol in Desmond. The production simulation was run for 50 ns at constant temperature and pressure (NPT), where the temperature was maintained at 300 K and the pressure at 1 atm.



RESULTS AND DISCUSSION Computational Analysis of Rieske Dioxygenases with Carbamazepine. Carbamazepine consists of the puckered tricyclic dibenzazepine ring structure with a carboxamide functionality appended to the ring nitrogen atom (Figure 1).

Figure 1. (a) Carbamazepine and common Rieske dioxygenase substrates. The numbering system for dibenzazepine ring atoms follows the accepted numbering as indicated in the figure. (b) Typical reaction of a polyaromatic hydrocarbon with a Rieske dioxygenase, naphthalene dioxygenase (NDO).

It is well established to be poorly, if at all, biodegraded in conventional municipal wastewater treatment.2,4,9 The resistance of carbamazepine to biodegradation has been considered curious because many tricyclic aromatic ring compounds are highly biodegradable, such as anthracene and the structurally analogous carbazole ring system (Figure 1a).28 The latter C

DOI: 10.1021/acs.est.6b04345 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

Figure 2. Computational analysis of the interaction between Rieske dioxygenases and carbamazepine to predict enzymatic reactivity. (a) Nonbonded interaction energy between carbamazepine and the active site of four Rieske dioxygenases. (b) Free energy of interaction between carbamazepine and the NDO channel leading into the active site pocket as a function of the position of the compound in the channel. Distances are measured from iron at the distal end of the active site. Naphthalene, the natural NDO substrate, is shown for comparison. (c and d) Docking position of carbamazepine in BPDOB1 and BPDOLB400, respectively. Only BPDOLB400 passed all computational requirements for reacting with carbamazepine.

Figure 3. LC/MS analysis of carbamazepine transformation by P. xenovorans LB400 resting cells. (a) LC/MS total ion chromatogram. Peak labels correspond to subsequent mass spectra in panels b−f: (b) cis-10,11-dihydroxy-10,11-dihydrocarbamazepine, (c) carbamazepine-dihydrodiol, (d) carbamazepine-diol, (e) 2-hydroxycarbamazepine, and (f) carbamazepine.

In direct biological experiments with 10 ppm carbamazepine, only P. xenovorans LB400 oxidized carbamazepine. In 24 h, degradation was complete. The initial degradation rate of 1600 μg (g of wet cells)−1 day−1 was linear over 8 h (Figure S1), which is very high in comparison to previous reports.9,35−37 Increasing the level of carbamazepine to 150 ppm gave an initial removal rate of 16800 μg (g of wet cells)−1 day−1 (Figure S2). This rate is >300 times better than anything reported

equilibrated portion of the molecular dynamics simulation shows that carbamazepine in the active site BPDOLB400 was not found to drastically change orientation relative to the initial docking. Experimental Demonstration of Rapid Carbamazepine Bio-Oxidation. Following computational predictions, we subsequently tested the four naturally occurring bacterial strains expressing the respective Rieske dioxygenases described above. D

DOI: 10.1021/acs.est.6b04345 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

Figure 4. Predicted pathway to observed products of carbamazepine transformation by P. xenovorans LB400 BPDO (BphA) and dihydrodiol dehydrogenase (BphB). Acid-catalyzed dehydration of the dihydodiol of the benzenoid ring is predicted to give rise to the 2-hydroxycarbamazepine.

previously. For example, a fungus, Trametes versicolor, is reported to degrade 44 μg (g of mycelia)−1 day−1.38 Degradation of carbamazepine has also been observed with Aspergillus niger, Rhodococcus rhodocrous, Pleurotus ostreatus, and Pseudomonas sp. CBZ-4, albeit at lower rates.35,36,39 Demonstration of Rapid Oxidation Due to Biphenyl Dioxygenase. To confirm that the oxidation in vivo was due to the P. xenovorans LB400 biphenyl dioxygenase, we incubated carbamazepine with a recombinant E. coli strain expressing the bphAEFG genes from P. xenovorans LB400. All four of those genes are required to express a fully functional biphenyl dioxygenase consisting of the Rieske dioxygenase that binds and reacts directly with the substrate, and accessory reductase and ferredoxin proteins that pass electrons from NADH to the Rieske protein (Figure S3). The recombinant E. coli was able to oxidize carbamazepine (10 ppm), degrading 27 ± 8% in 4 h and 89 ±1% in 24 h, which equate to 8 and 27 μg, respectively. A control E. coli strain without the bphAEFG genes showed no demonstrable degradation. Identification of Carbamazepine Metabolites. To further confirm biodegradation, and to determine the course of degradation and whether carcinogenic acridines form, P. xenovorans LB400 resting cell cultures were incubated with carbamazepine and subsequently analyzed by LC/MS/MS and GC/MS. By LC/MS, four product peaks were identified in addition to the starting material, carbamazepine (Figure 3). The dihydroxylated products were cis-10,11-dihydroxy-10,11dihydrocarbamazepine, along with a carbamazepine-dihyrodiol and a carbamazepine-diol putatively identified as cis-2,3dihydroxy-2,3-dihydrocarbamazepine and 2,3-dihydroxycarbamazepine, respectively. A hydroxylated product was also observed. The identification of cis-10,11-dihydroxy-10,11dihydrocarbamazepine (Figure 3b) and 2-hydroxycarbamazepine (Figure 3e) was confirmed by comparison to authentic standards. LC/MS/MS spectra of the carbamazepine dihyrodiol and carbamazepine-diol are consistent with dioxygenation of a benzenoid ring (Figure S4). Products oxcarbazepine, 10,11dihydro-10,11-epoxycarbamazepine, and 3-hydroxycarbamazepine were ruled out with authentic standards. Importantly, the production of toxic metabolites acridine and acridine 9carboxaldehyde was ruled out by comparison of GC/MS data of P. xenovorans LB400 extracted resting cell assay samples to authentic standards (Figure S5).

The identification of the major metabolites as cis-10,11dihydroxy-10,11-dihydrocarbamazepine and the combination of cis-2,3-dihydroxy-2,3-dihydrocarbamazepine with carbamazepine-2,3-diol indicates that the biphenyl dioxygenase is the primary oxidizing enzyme, which had been also demonstrated with the recombinant E. coli expressing the biphenyl dioxygenase genes. The toxicity of carbamazepine to animals is largely dependent on activation to reactive epoxides (mainly at the 10,11 position) and the depletion of glutathione.40,41 While fungi present in wastewater treatment typically oxidize aromatic hydrocarbons to epoxides, which can be problematic, bacterial Rieske dioxygenases typically generate relatively unreactive intermediates, such as the cis-dihydrodiols described here.19,32−34,42 Carbamazepine biotransformation by P. burkholderia LB400 is initiated by BPDO as shown in Figure 4. The production of two dihydrodiols indicates a competition between the reactivity of the anti-aromatic bond between carbons 10 and 11,43 favoring dioxygenation of the central ring, and the energetically favorable docking orientation of carbamazepine with the C3 atom oriented toward the active site iron (Figure 2d). The benzenoid dihydrodiol was then further transformed to a carbamazepine-diol and 2-hydroxycarbamazepine. The dihydrodiol dehydrogenase enzyme, encoded by the bphB gene, Bxe_C1192 in the genome (Figure S3c), dehydrogenated the benzenoid dihydrodiol but not, at any detectable level, the 10,11-dihydrodiol. The 2-hydroxycarbamazepine likely forms spontaneously via dehydration of the dihydrodiol to the phenol.44 The presence of 2-hydroxycarbamazepine and the favored docking orientation of carbamazepine in BPDO strongly suggest that the benzenoid dihydrodiol formed by BPDO is 2,3-dihydroxy-2,3-dihydrocarbamazepine. It follows that the diol would be 2,3-dihydroxycarbamazepine. Dibenzazepine Is Shown To Be More Readily Oxidized by Rieske Dioxygenases. To explore the hypothesis that the carboxamide functionality of carbamazepine is a primary cause of the compound’s resistance to biodegradation, parallel experiments were conducted with dibenzazepine, the base ring structure lacking the carboxamide group. Computational analysis suggested that dibenzazepine would dock favorably in the active site of three of the dioxygenases, the exception being toluene dioxygenase (Table S1). Unlike carbamazepine, dibenzazepine is predicted to access the active site of NDO, BPDOB1, and BPDOLB400. Therefore, dibenzazepine was E

DOI: 10.1021/acs.est.6b04345 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

Figure 5. (a) Biotransformation of emerging pollutants by P. xenovorans LB400 after 24 h. The percent removal of each compound relative to an E. coli control is indicated after the name of the compound. The commercial application of the compound is shown in parentheses. Pathway Prediction System (PPS) models were used to determine the likely initiating metabolism, denoted by esterases/amidase to the left of the dotted lines and by oxygenases to the right. (b) Example esterase- and oxygenase-mediated degradation pathways predicted by the PPS for two representative emerging pollutants.

Biodegradation of Personal Care Products That Are Problematic in Contemporary Wastewater Treatment. Carbamazepine is only one of many chemicals escaping current wastewater treatment facilities. In light of the unique ability of P. xenovorans LB400 to metabolize carbamazepine rapidly, its unusually large genome (9.73 Mb), its unusually large number of predicted oxygenases and degradative enzymes (Table S2), and its broad catabolic activities with PCBs, dioxins, and terpenoids,45−47 other chemicals of emerging concern were examined here. These included other poorly degradable pharmaceuticals, endocrine-disrupting alkyl phthalates, a

computationally predicted to be a substrate of these enzymes. In wet laboratory experiments, dibenzazepine was accepted as a substrate for three of the strains, the exception being Ps. putida F1 expressing toluene dioxygenase. These data were consistent with the idea that the carboxamide functionality imposed a steric hindrance to being metabolized more readily in microorganisms containing the widely prevalent Rieske dioxygenases. In resting cell assays with 10 ppm dibenzazepine, >90% was removed by each of the strains in 24 h: Pseudomonas sp. strain NCIB 9816-4, S. yanoikuyae B1, and P. xenovorans LB400. F

DOI: 10.1021/acs.est.6b04345 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

biosponge materials designed to trap and concentrate pollutants prior to bioremediation. A large number of companies sell bioaugmentation products, but most of those consist of bacteria expressing amylases, lipases, and proteases that act to decrease the levels of nonproblematic organic materials that nonetheless contribute to chemical oxygen demand (COD). Only more recently have larger biotechnology companies been tracking and adjusting populations of bacteria in wastewater treatment with the goal of degrading emerging pollutants.50 The results of this study are consistent with a growing contemporary approach of knowledge-based bioaugmentation.51−54 The use of P. xenovorans LB400 in PCB bioremediation, its ease of growth, its constitutive and inducible expression of BPDO, and the enormous range for biodegradation via BPDO and additional enzymatic pathways suggest the potential for its broad application (Table S2).20,55−58 Coupled with the ability to categorize and implement such naturally occurring strains is the expanding use of computational methods to predict biodegradation capacity to allow wastewater treatment facilities to keep pace with the rapid introduction of personal care products that increasingly enter municipal water supplies. Certain biodegradative enzymes are known to react with hundreds of chemicals, and computational predictions might identify thousands. In this context, the ability to identify enzymes and microorganisms that can biodegrade highly bioactive and persistant personal care products will become increasingly important.

suncreen agent, fragrance compounds, and detergents, many of which are increasingly appearing in municipal water supplies. Each compound was incubated at a concentration of 10 ppm in resting whole cell assays for 24 h with P. xenovorans LB400 and E. coli. The latter served as a negative control to rule out adsorption or other potential effects leading to false positives. Compounds degraded by P. xenovorans LB400 and the percent removal compared to an E. coli control are shown in Figure 5a. The full list of compounds tested and the GC/FID and HPLC data are listed in Table S3. Chromatogram peak identity was confirmed by comparison of the retention time to those of authentic standards. For compounds volatile enough to be analyzed by GC, further confirmation of peak identity was provided by the mass spectrum. Because P. xenovorans LB400 is known to express a large number of biodegradative enzymes (Table S2), it is unlikely that all of these compounds are substrates of BPDO. While a full analysis of the biotransformation route of each compound is beyond the scope of this study, each compound was computationally analyzed to predict the likely initiation enzyme(s). The EAWAG-Pathway Prediction System (PPS) is a prediction approach that computationally predicts metabolic pathways using a rule-based system generated by expert knowledge and gene frequencies.15,48,49 The PPS shows the most likely initiating reaction types and is complementary to the RAPID prediction algorithm being developed because the PPS is not capable of predicting a specific enzyme. The compounds degraded were clustered by percent removed and by the enzyme types initiating the metabolism of the compounds as predicted by the PPS (Figure 5). Two observations can be made from the clustering. First, clustering of the compounds by percent removal alone clearly shows that chemical structure rather than contaminant source determines the extent of degradation. Second, clustering by the two parameters shows that those compounds transformed >50% are computationally predicted to undergo biotransformation via esterase or amidase enzymes. Likewise, the more recalcitrant compounds, those degraded by