Introduction

Steroids are ubiquitous and abundant in nature. In contrast to the structural function of sterols in cell membranes, steroid hormones are extremely bioactive and regulate various physiological processes, including development and reproduction, in animals (Ghayee and Auchus, 2007). The increasing number of reports on intersex organisms found in global aquatic ecosystems has raised concerns regarding the occurrence and persistence of steroid hormones in the environment (Safe, 2004; Wise et al., 2011). Several studies have reported the masculinization of freshwater wildlife exposed to androgens in polluted rivers (Howell et al., 1980; Bortone et al., 1989; Parks et al., 2001; Orlando et al., 2004). Furthermore, steroids were identified as pheromones and adversely affect fish behavior even at extremely low concentrations (Kolodziej et al., 2003). For example, androgens at picomolar concentrations elicit both odorant and pheromonal responses in fish (Adams et al., 1987; Moore and Scott, 1991; Serrano et al., 2008; Katare et al., 2011).

Steroid hormones are discharged into the environment through various routes, including the agricultural application of livestock manure and municipal sewage biosolids as fertilizers (Hanselman et al., 2003; Lorenzen et al., 2004). Moreover, phytosterols in pulp and paper mill effluents were transformed to androgens by microorganisms in river sediments, consequently influencing fish development (Jenkins et al., 2001, 2003). Steroid hormones have been frequently detected in various environmental matrices, including soil, groundwater, surface water and sediments (Ying et al., 2002). These substances were often found at nanograms to micrograms per liter concentrations in surface water (Ternes et al., 1999; Baronti et al., 2000; Kolodziej et al., 2003; Yamamoto et al., 2006; Chen et al., 2010; Chang et al., 2011; Fan et al., 2011). Quantifying five classes of steroid hormones in the surface waters of Beijing (China) revealed that androgens (up to 1.9 μg l−1) were the most abundant steroids in urban rivers (Chang et al., 2009).

Biodegradation has been considered a crucial mechanism for removing steroids from natural and engineered ecosystems (Johnson and Sumpter, 2001; Andersen et al., 2003; Khanal et al., 2006). Sewage treatment plants are essential in removing steroid hormones, as shown by a comparison between sewage influents and effluents (Fan et al., 2011). Several androgen-degrading aerobes have been isolated and characterized (Donova, 2007; Yang et al., 2011; Horinouchi et al., 2012). Among them, Comamonas testosteroni has received special attention, and catabolic intermediates, genes, and enzymes have been identified (see Horinouchi et al. (2012) for a review). By contrast, literature on the biochemical mechanisms and the phylogenetic diversity involved in anaerobic steroid biodegradation is extremely limited. Hanselman et al. (2003) indicated that anoxic sediments and soil may be reservoirs for steroids because microbial degradation occurs slowly in these matrices. So far, two anaerobic species (Steroidobacter (Sdo.) denitrificans and Sterolibacterium denitrificans) capable of growing on androgens have been reported (Tarlera and Denner, 2003; Fahrbach et al., 2008). We recently elucidated the anaerobic catabolic pathway (2,3-seco pathway) of testosterone by using S do. denitrificans DSMZ 18526 (gammaproteobacterium) and Sterolibacterium (Stl.) denitrificans DSMZ 13999 (betaproteobacterium) as model microorganisms (Wang et al., 2013b, 2014) (Figure 1b). However, information regarding the catabolic genes and enzymes involved in the anaerobic catabolic pathway is scant.

Figure 1
figure 1

Proposed catabolic pathways of testosterone. (a) Aerobic 9,10-seco pathway established in Comamonas testosteroni TA441. (b) Anaerobic 2,3-seco pathway established in Steroidobacter denitrificans DSMZ18526. The compound in bracket is presumed. Suggested signature metabolites are enclosed in boxes.

Full size image

The biochemical mechanisms of androgen biodegradation have been studied using a few pure cultures. However, whether these established catabolic pathways are functional in the environment remains unknown. In addition, the distribution, abundance, and ecological relevance of culturable steroid-degraders in the ecosystem remain unexplained. Conventional culture-independent approaches, such as stable-isotope probing and fluorescence in situ hybridization-microautoradiography, are rarely used in the studies of microbial steroid degradation because (i) completely 13C-labeled steroids are not commercially available and (ii) hydrophobic compounds, such as steroids, easily attach to or pass through bacterial membranes. Therefore, distinguishing the metabolic activities (passive diffusion or active uptake; redox transformation or complete degradation) of radiolabeled bacterial cells is difficult. The lack of molecular markers and suitable culture-independent techniques has impeded in situ investigations of microbial steroid degradation, thus necessitating efficient and specific tools for monitoring the environmental fate of steroids.

In this study, we first identified the androgen catabolic genes and enzymes of Sdo. denitrificans. We then investigated androgen biodegradation in the anoxic sewage of a municipal sewage treatment plant designed to treat domestic wastewater produced by three million residents of Taipei City, Taiwan, for which we used the following culture-independent approaches: (i) ultra-performance liquid chromatography—tandem mass spectrometry identification of signature metabolites; (ii) identification of major catabolic players through next-generation sequencing techniques; and (iii) PCR-based identification of functional genes. This study is the first integrated ‘omics’ investigation on the biochemical mechanisms and phylogenetic diversity of steroid biodegradation in the environment.

Materials and methods

General materials and methods

The materials and methods section in the Supplementary Information provide complete descriptions of the following: bacterial strains and bacterial cultivation; steroid extraction, detection, and quantification; DNA/RNA extraction; basic molecular biological methods; genome sequencing, assembly and gene prediction; RNA-Seq; Illumina MiSeq sequencing of 16S rRNA gene amplicons; 16S rDNA-based taxonomic analysis; lacZ-based yeast androgen bioassay; and the purification and characterization of 1-testosterone hydratase/dehydrogenase (AtcABC: anaerobic testosterone catabolism).

DCPIP-based spectrophotometric assays

AtcABC activity was spectrophotometrically assayed (SpectraMax M2e spectrophotometer; Molecular Devices, Sunnyvale, CA, USA) using 2,6-dichlorophenol-indophenol (DCPIP) as the electron acceptor. The decrease in absorption caused by the substrate-dependent reduction of DCPIP occurred at 578 nm (ɛ=16.8 cm−1mmol−1 (Jin et al., 2011)). To determine the hydratase/dehydrogenase activity of individual protein fractions, assay mixtures (0.9 ml) containing 20 mm Tris-HCl (pH 7.5), 50 μm DCPIP, 125 μm 1-testoterone and 1.25% (v/v) 2-propanol were aerobically incubated at 30 °C. The reaction was initiated by adding 0.1 ml of enzyme solution (0–500 μg of proteins).

Sewage sample collection

Dihua sewage treatment plant (DHSTP) is the largest municipal wastewater treatment plant (500 000 m3 day−1) in Taipei. Along with domestic water, DHSTP receives industrial, medical and livestock wastewater as well as groundwater (Lin et al., 2009). The hydraulic retention time is ~10 h, and the effluent is discharged into Tamsui River, the largest in Taipei and its suburbs (Lin et al., 2010). The plant is designed as an anoxic/oxic process for carbon and nitrogen removal (see Lin et al. (2009 and 2010) for detailed information). Influent and effluent water samples (5 l) were collected in triplicate in June 2014. Formaldehyde (final concentration=1% v/v) was added to the glass bottles before sample collection. After 10 min of formaldehyde fixation, 5 μg of [2,3,4C-13C]testosterone (Isosciences, King of Prussia, PA, USA) was added to the samples. Androgens in water samples were quantified using liquid chromatography-mass spectrometry as described in Supplementary Information. Sewage samples (20 l) were collected from the anoxic tank (denitrification tank; range of dissolved oxygen levels=0.2–0.5 mg l−1) of DHSTP in April and June 2014. A sterilized 20 l glass bottle was completely filled with the anoxic sewage and was tightly capped to avoid headspace formation. The samples were delivered to the laboratory within 30 min.

Incubation of anoxic sewage with testosterone

Anoxic sewage (1 l) was transferred into 1.2 l glass bottles in a laminar flow hood. The bottles containing sewage were sealed with a rubber stopper, and sterile nitrogen gas (~1200 ml; passed through a sterile cotton plug) was used as the carrier gas (flow rate, 40 ml min−1) to expel residual oxygen from the bottles. Anoxic sewage was treated under the following conditions: autoclaved sewage with testosterone (1 mm) and nitrate (10 mm), native sewage with nitrate and without testosterone, native sewage with testosterone and without nitrate and native sewage with testosterone and nitrate. Ascorbate (a reducing agent) and resazurin (a redox indicator) were not added to the anoxic sewage treatments because these compounds could be used as carbon and energy sources for sewage bacteria. Sewage treatments were performed in duplicate and the bottles were incubated at 25 °C with stirring at 160 r.p.m. for 2 weeks. Sewage samples (10 ml) were withdrawn from the bottles by using a 10 ml sterile syringe every 12 h and were stored at −80 °C before use. Androgenic activity and metabolites in sewage samples were detected using the yeast androgen assay and ultra-performance liquid chromatography (UPLC)—atmospheric pressure chemical ionization (APCI)—MS/MS, respectively. Bacterial 16S rRNA gene and the androgen catabolic gene (atcA) in sewage samples were analyzed using the Illumina (San Diego, CA, USA) MiSeq platform and PCR-based functional assay, respectively.

PCR-based functional assay by using the atcA gene probe

Search for atcA-related gene sequences in the GenBank Database was conducted on protein–protein BLAST (http://www.ncbi.nlm.nih.gov/blast/) with Sdo. denitrificans AtcA as a query. Multiple alignment of the atcA-related gene sequences was performed using Align/Assemble (Geneious 8.1.4). A degenerate primer pair was deduced from regions conserved in the atcA genes but not in the gene coding for the large subunit of 3-hydroxycyclohexanone dehydrogenase (MhyADH) from Alicycliphilus denitrificans (Jin et al., 2011) (Supplementary Figure S1). The atcA fragments were amplified using the degenerate primers (forward primer: 5′–GGCASCGYYSAGTTCATCGACAA–3′ and reverse primer: 5′–GCCGCTGTCRTAYTCRTTSCCGCTSGG–3′) through PCR (94 °C for 2 min; 25 cycles: 94 °C for 30 s, 55 °C for 30 s, 72 °C for 90 s; and 72 °C for 10 min). The atcA fragments (~1100 bp) amplified from the anoxic DHSTP sewage were cloned in Escherichia coli (One Shot TOP10; Invitrogen, Waltham, MA, USA) by using the pGEM-T Easy Vector Systems (Promega, Fitchburg, WI, USA).

Results

Sdo. denitrificans genome

The 3.47 Mb chromosome (61.7% G+C; accession number: CP011971) of Sdo. denitrificans contains 2986 predicted protein-coding genes (Supplementary Table S1), 1 rRNA operon (5S, 16S and 23S), 47 tRNA genes, 1 CRISPR gene cluster and 62 pseudogenes. Altogether, 2022 protein-coding genes were classified into COG categories, and lipid metabolism genes (197 genes; COG I; Supplementary Table S1) were the most abundant group of metabolism-related genes. A high correlation was observed between the phenotype and genotype of Sdo. denitrificans. First, this denitrifying bacterium cannot use ferric or perchlorate as the alternative electron acceptor. Second, Sdo. denitrificans has a narrow substrate spectrum and cannot utilize carbohydrates, monoaromatic compounds and long-chain fatty acids (Fahrbach et al., 2008). Third, this bacterium depends on the supply of various vitamins (for example, cobalamin, pyridoxamine and thiamine) for growth (DSMZ Medium 1116). No genes related to these metabolic features were found in the genome of Sdo. denitrificans (Supplementary Table S1). We identified the genes encoding the ABC transporters for the uptake of ferric and molybdate ions (Figure 2a; Supplementary Table S1). Molybdenum serves as a cofactor of the essential enzymes for the denitrification and anaerobic degradation of hydrocarbons (Dermer and Fuchs, 2012). The genes (ACG33_02275~ACG33_02290) encoding the molybdopterin biosynthesis proteins (Moe) were identified (Figure 2a). In addition, genes for the bacterial secretion systems, including type II, type IV, Sec and Tat, are present on the Sdo. denitrificans chromosome (Supplementary Table S1; highlighted in green).

Figure 2
figure 2

(a) Structure and distribution of the putative catabolic genes for androgen degradation on the chromosome of Sdo. denitrificans. The genes involved in molybdopterin biosynthesis and molybdate transport are also shown. Proposed gene products are enclosed in gray boxes. (b) Genes coding for a steroid C25 dehydrogenase-like enzyme are found in the genomes of Sdo. denitrificans, Stl. denitrificans and Thauera terpenica. The latter was identified as an anaerobic androgen-degrader in this study. Homologous open reading frames (colored arrows) between different bacterial strains are connected with dotted lines. Numbers (%) indicate the identity of the deduced amino acid sequences with those of the genes of Sdo. denitrificans. (c) Genes encoding β-oxidation enzymes are conserved in androgen-degrading anaerobes and aerobes (e.g., C. testosteroni). The numbers (%) indicate the identity of the amino acid sequences with those of the functionally confirmed genes of C. testosteroni TA441. For the detailed catabolic functions of individual genes of the strain TA441, refer Horinouchi et al. (2012). (d) Genes coding for the molybdoproteins of the xanthine oxidase family. The numbers (%) indicate the identity of deduced amino acid sequences with those of the corresponding genes of Sdo. denitrificans, whose product was characterized as 1-testosterone hydratase/dehydrogenase in this study.

Full size image

The androgen metabolite analysis of aerobically and anaerobically grown Sdo. denitrificans cells suggested the involvement of various redox enzymes in the initial steps of both catabolic pathways (Wang et al., 2013b). In the Sdo. denitrificans chromosome, the genes ACG33_00240, ACG33_09105 and ACG33_10715, respectively coding for 3-ketosteroid Δ1-dehydrogenase, 3α(20β)-hydroxysteroid dehydrogenase and 3β(17β)-hydroxysteroid dehydrogenase, were identified. The other putative catabolic genes involved in steroid degradation were also identified in this genome. For example, four clustered genes (ACG33_10670~ACG33_10685) whose products were similar (>38% sequence identity) to the steroid C25 dehydrogenase of Stl. denitrificans (Figure 2b). This molybdoenzyme catalyzes the anaerobic hydroxylation of steroid substrates (Dermer and Fuchs, 2012). In addition, a cluster of 14 genes (ACG33_00310~ ACG33_00375) was identified within a 15 kb stretch of the circular chromosome. The encoding proteins shared up to 88% amino acid sequence identity with the β-oxidation enzymes involved in the degradation of steroid C/D rings by C. testosteroni (Figure 2c). In the genome of C. testoteroni TA441, tesB encoding a meta-cleavage dioxygenase is located next to ORF1. Similar gene organization was observed in the genomes of C. testosteroni strains CNB-2 (Figure 2c) and KF1 (Horinouchi et al., 2012). However, a tesB-like gene was not present in the chromosomes of Sdo. denitrificans and Stl. denitrificans (Figure 2c). Stl. denitrificans uses the 2,3-seco pathway (see Figure 1b for detailed information) to degrade androgens regardless of oxygen conditions (Wang et al., 2013a; Wang et al., 2014), and no genes encoding steroid catabolic oxygenases are found in the Stl. denitrificans genome (Dermer and Fuchs, 2012). Therefore, the β-oxidation genes (ACG33_00310~ ACG33_00375) of Sdo. denitrificans may be involved in the degradation of steroid C/D rings during anaerobic growth on testosterone. In the genome of C. testosteroni, the genes responsible for the oxygenase-dependent cleavage of the steroid A/B rings are organized into a single cluster (ORF18,17, tesIHA2A1DEFG). By contrast, the corresponding genes of Sdo. denitrificans are scattered throughout its chromosome (Figure 2a). The putative steroid catabolic genes are highlighted in red in Supplementary Table S1.

Comparative transcriptomics of Sdo. denitrificans and the identification of genes expressed during anaerobic testosterone degradation

Depending on oxygen availability, Sdo. denitrificans degrades testosterone through two distinct pathways (Wang et al., 2013b). Therefore, we proposed that testosterone catabolic genes were regulated on the basis of the prevailing oxygen tension. To identify the genes involved in anaerobic testosterone catabolism, transcriptomic investigations were conducted on Sdo. denitrificans grown under three conditions: (i) aerobic growth with testosterone (2.5 mm); (ii) anaerobic growth with testosterone; and (iii) anaerobic growth with heptanoic acid (10 mm). Sdo. denitrificans transcriptomes were sequenced using RNA-Seq (Supplementary Table S2). Transcriptomic analysis revealed 232 genes (highlighted in orange in Supplementary Table S2) that were at least twofold upregulated during anaerobic growth on testosterone, compared with anaerobic growth on heptanoic acid and aerobic growth on testosterone (Figure 3a). Many of the upregulated genes were involved in transcription (COG K; 14 genes), translation (COG J; 14 genes) and post-translational modification (COG O; 18 genes). Moreover, we detected the upregulation of 11 genes involved in secondary metabolites biosynthesis and catabolism (COG Q), 12 in lipid metabolism (COG I); 12 in energy production (COG C), 15 genes in cell wall and membrane biogenesis (COG M), and 21 associated with coenzyme transport and metabolism (COG H). This reflects a general physiological adaptation of Sdo. denitrificans to anaerobically grow on a hydrophobic compound with a complex structure like testosterone. In addition, 78 upregulated genes coded for hypothetical proteins and could not be classified into any COG categories. We observed the apparent upregulation (threefold) of most genes in cluster 1 (ACG33_4765~ ACG33_4815; 11 genes) and cluster 2 (ACG33_10725~ ACG33_10770; 10 genes) during anaerobic growth with testosterone (Figure 3b). Most genes in cluster 1 encoded proteins involved in membrane and cell wall biogenesis. Most genes in cluster 2 exhibited high sequence identity (65–86%) to the uncharacterized proteins of Azoarcus toluclasticus, and some genes in this cluster (for example, ACG33_10730 encoding an acetyl-CoA acetyltransferase and ACG33_10735 coding for a short-chain dehydrogenase) were proposed to be involved in lipid metabolism (Table 1). Furthermore, the specific expression of genes involved in the uptake and biosynthesis of the enzyme cofactors, for instance, genes encoding a TonB-dependent receptor (for siderophore uptake; ACG_02125), cobalamin-bind protein (ACG_04760), and Fe-S cluster assembly protein (ACG_06585) was observed. Moreover, genes encoding the molybdopterin biosynthesis proteins MoeA (ACG_02290) and MoeC (ACE_2280) and the molybdenum ABC transporter (ACG_10875) were upregulated in cells anaerobically grown on testosterone (Supplementary Table S2).

Figure 3
figure 3

Global gene expression profiles (RNA-Seq) of Sdo. denitrificans grown under different conditions. (a) The COG classification of the upregulated (greater than twofold) genes during anaerobic growth on testosterone, compared with anaerobic growth on heptanoic acid and aerobic growth on testosterone. Note that some genes are assigned into multiple COG groups. (b) Aerobic versus anaerobic growth on testosterone. (c) Anaerobic growth on testosterone and heptanoic acid. Each spot represents a gene. Diagonal lines indicate the identical thresholds. The house keeping genes include glucose-6-phosphate 1-dehydrogenase (ACG33_00890), chaperone htpG (ACG33_01365), isocitrate dehydrogenase (ACG33_02930), enolase (ACG33_09515), pyruvate dehydrogenase (ACG33_11195), malate dehydrogenase (ACG33_14070) and type II citrate synthase (ACG33_14495).

Full size image
Table 1 Expression profile of the Sdo. denitrificans genes which are probably involved in anaerobic testosterone degradation
Full size table

The transcriptomic analysis revealed the apparent upregulation (greater than threefold), compared with aerobic growth on testosterone, of three clustered genes (ACG_03370, 3375 and 3390) during anaerobic growth on testosterone (Figure 3b). Compared with Sdo. denitrificans grown anaerobically on heptanoic acid (Figure 3c), these three genes were upregulated, albeit to a lesser extent, in anaerobically testosterone-grown cells. Deduced amino acid sequences of these upregulated genes exhibited sequence identities of 38–64% to those of the 3-hydroxycyclohexanone dehydrogenase genes of the cyclohexanol-degrading Ali. denitrificans (Jin et al., 2011). We previously hypothesized that a MhyADH-like molybdoenzyme might catalyze the hydration of 1-testosterone (Wang et al., 2013b). Similar genes were harbored by Stl. denitrificans and closely related betaproteobacteria (Figure 2d). The majority of the clustered β-oxidation genes (ACG33_00310~ ACG33_00375) were slightly upregulated (less than twofold) in the anaerobically testosterone-grown cells, compared with those aerobically grown on testosterone. We suggest that Sdo. denitrificans uses the β-oxidation enzymes to degrade the C- and D-rings of testosterone regardless of the oxygen conditions. This suggestion is based on (i) the highly conserved primary structures (up to 88%) of the β-oxidation enzymes between the denitrifying Sdo. denitrificans and the strictly aerobic C. testosteroni, and (ii) the common expression of these β-oxidation genes under both anoxic and oxic conditions. The expression profile of the putative steroid C25 dehydrogenase genes and the 3-ketosteroid Δ1-dehydrogenase gene in the Sdo. denitrificans cells aerobically and anaerobically grown on testosterone was indistinguishable (less than twofold) (Figure 3b).

Purification and characterization of 1-testosterone hydratase/dehydrogenase (AtcABC) from Sdo. denitrificans

The androgen metabolite profile (Wang et al., 2013b) and genomic and transcriptomic (this study) analyses suggested the involvement of a MhyADH-like molybdoenzyme in anaerobic testosterone catabolism by Sdo. denitrificans. To prove our hypothesis, AtcABC was purified from the cell extract of testosterone-grown Sdo. denitrificans in three chromatographic steps (Figure 4a). A spectrophotometric assay with 1-testosterone as substrate and DCPIP as artificial electron acceptor was developed for activity measurements during purification of this bifunctional enzyme. The enzyme assay was performed aerobically at 30 °C and the transformation of the substrate to 17-hydroxy-androstan-1,3-dione was monitored at 578 nm. The identity of the reaction product was confirmed through UPLC-APCI-MS (Figure 4b). SDS–polyacrylamide gel electrophoresis analysis of the active gel filtration pool showed three protein bands with the following molecular masses: large protein (AtcA), ~85 kDa; medium protein (AtcB), ~30 kDa; and small protein (AtcC), ~18 kDa (Figure 4a). Liquid chromatography-tandem mass spectrometry analysis revealed that tryptic peptides originating from the protein bands AtcA, AtcB and AtcC are identical to the predicted tryptic products of genes ACG33_03390, ACG33_03375 and ACG33_03370, respectively.

Figure 4
figure 4

Purification and characterization of 1-testosterone hydratase/dehydrogenase (AtcABC) from Sdo. denitrificans grown anaerobically on testosterone. (a) SDS–polyacrylamide gel electrophoresis (4% to 12%) of active pools during the fast protein liquid chromatography purification of AtcABC. (b) APCI-MS spectrum of the reaction product 17-hydroxy-androstan-1,3-dione. (c) Substrate preference of AtcABC determined using a DCPIP-based enzyme assay. The assay mixtures (0.9 ml) contained 20 mm Tris-HCl (pH 7.5), 5 μg purified enzyme, 50 μm DCPIP, 125 μm of individual substrates and 1.25% 2-propanol. 2-Propanol was used as the solvent. Abbreviations: DEAE, DEAE sepharose; Butyl, butyl sepharose; Sephacryl, Sephacryl S-300.

Full size image

Purified AtcABC had a pH optimum of 6.5 (Supplementary Figure S2A) and an apparent Km of 52±9 μm for the natural substrate 1-testosterone. The Vmax of 1-testosterone oxidation was 169±12 nmol min−1 mg protein−1 (see Supplementary Figure S2B for the Michaelis–Menten plot). The substrate preference of AtcABC was tested by screening various compounds (125 μm) with an α,β-unsaturated ketone structure (Figure 4c). The molybdoenzyme exhibited the highest dehydrogenation activity toward 3-keto-1-en-steroids. The 3-keto-4-en-steroids could not serve as substrates, suggesting that an unsaturated double bond at the C-4/C-5 positions of steroids hinders dehydrogenase activity. Unlike MhyADH, AtcABC could not recognize the monocyclic substrate 2-cyclohexenone (Figure 4c). The apparent substrate preference of AtcABC enables the application of atcABC genes as molecular biomarkers to assess anaerobic androgen biodegradation in the environment.

Phylogenetic analyses of essential catabolic enzyme involved in the anaerobic pathway

We used the AtcA sequence of Sdo. denitrificans as a query to blast the protein database of NCBI. The most similar sequences were harbored by denitrifying betaproteobacteria (Figure 5 and Supplementary Table S3), which showed an identity of ~60% with the Sdo. denitrificans sequence. The AtcA-like proteins of Thauera terpenica and Azo. toluclasticus are completely uncharacterized. By contrast, the activity of 1-testosterone hydratase/dehydrogenase was detected in the cytoplasm of Stl. denitrificans (Lin et al., 2015), although the corresponding genes were not identified. The Sdo. denitrificans AtcA shows a low sequence identity (38%) with the large subunit of 3-hydroxycyclohexanone dehydrogenase of Ali. denitrificans; this result is consistent with the substrate preference of AtcABC (Figure 4c). Moreover, we used the sequence of Sdo. denitrificans AtcA as a query to blast UniProtKB/Swiss-Prot, in which the protein functions are experimentally confirmed. Members of the xanthine oxidase family were the most similar proteins extracted from this database (see Supplementary Table S3 for detailed information). These proteins introduce a hydroxyl group into substrates and use water as the oxygen atom source (Leimkühler et al., 2004; Okamoto et al., 2004). Phylogenetic analysis (Figure 5) revealed that AtcABC and its highly similar proteins formed a distinct clade and were separated from other members of this enzyme family. The distribution of AtcABC-like proteins in bacteria therefore suggested that the anaerobic androgen degradation ability is conserved in proteobacteria (mainly beta-subgroup).

Figure 5
figure 5

Phylogenetic tree of the members of the xanthine oxidase family based on the amino acid sequences of molybdopterin-containing subunits. Supplementary Table S3 provides detailed information on these molybdoenzymes. The phylogenetic trees were constructed using the neighbor-joining method with Jukes–Cantor parameter and a bootstrap value of 1000. Abbreviations: CHO, aldehyde; CO, carbon monoxide; DH, dehydrogenase; 1-t, 1-testosterone.

Full size image

UPLC-MS/MS identification of androgenic signature metabolites in DHSTP sewage

The water and sewage samples were collected in June 2014. Four androgens (testosterone; androst-4-en-3,17-dione; androsta-1,4-diene-3,17-dione; and epiandrosterone) in the influent and effluent water of DHSTP were quantified through liquid chromatography-mass spectrometry (Supplementary Table S4). Our data indicated that androgens (approximately 10 μg l−1, mainly testosterone) in the domestic wastewater of Taipei City were efficiently removed (>99%). DHSTP sewage incubated under different conditions was sampled every 12 h, extracted using ethyl acetate, and the androgenic metabolites were identified through UPLC-APCI-MS/MS (Figure 6 and Supplementary Figure S3). The androgenic activity of the ethyl acetate extracts was determined using the lacZ-based yeast androgen assay (Supplementary Figure S4).

Figure 6
figure 6

UPLC-APCI-MS/MS analysis of the ethyl acetate extracts of the anoxic DHSTP sewage treatments. (a) Total ion chromatograms of anoxic sewage incubated with testosterone but without nitrate. (b) Total ion chromatograms of anoxic sewage incubated with testosterone and nitrate. (c) Extracted ion chromatograms (m/z=305.21 for 2,3-SAOA) of anoxic sewage incubated with testosterone and nitrate. (d) MS/MS spectra of the authentic standard (top) and 2,3-SAOA extracted from anoxic sewage incubated with testosterone and nitrate (bottom). See Figure 1 for the abbreviations of androgens.

Full size image

Testosterone transformation was not observed in autoclaved anoxic sewage incubated with testosterone and nitrate (Supplementary Figure S3). The biotransformation of testosterone to androst-4-en-3,17-dione was observed in anoxic sewage incubated with testosterone (1 mm) but without nitrate (Figure 6a). However, androgens in this treatment were not apparently consumed. By contrast, testosterone was exhausted after 72 h of incubation under denitrifying conditions (Figure 6b). The drastic decrease in androgenic activity with time confirmed androgen biodegradation in the denitrifying sewage (Supplementary Figure S4). UPLC-APCI-MS analysis revealed the biotransformation of testosterone to other androgens through redox reactions within the first 2 days. Unlike androgenic metabolites with a sterane structure, the ring-cleavage product 17-hydroxy-1-oxo-2,3-seco-androstan-3-oic acid (2,3-SAOA) cannot be easily ionized. Thus, we employed extracted ion current for m/z 305.21 (the most dominant ion peak of 2,3-SAOA) to detect 2,3-SAOA (Figure 6c). The UPLC retention time (5.0 min) and the MS/MS spectrum of the extracted ion were the same as those of the authentic standard (Figure 6d). In addition, 2,3-SAOA was detected in anoxic sewage collected in April 2014 (Supplementary Figure S5) during denitrifying incubation with testosterone. In contrast, the signature metabolite of the aerobic 9,10-seco pathway, 3-hydroxy-9,10-seco-androsta-1,3,5(10)-triene-9,17-dione (3-HSA), was not detected in anoxic sewage. UPLC-APCI-MS/MS analysis indicated that denitrifying bacteria in anoxic sewage used the 2,3-seco pathway to degrade testosterone.

Phylogenetic identification of androgen-degrading bacteria in anoxic sewage

DNA was extracted from sewage samples, and the V3-V4 hypervariable region of bacterial 16S rDNA sequences was amplified through PCR. The resulting amplicons were sequenced using an Illumina MiSeq sequencer. For each sewage treatment, an average of 341 585 reads was obtained and the sequences were analyzed at the genus level by using the BaseSpace 16S Metagenomics app V1.01 (Illumina). The nucleotide sequence data set was deposited in the NCBI Sequence Read Archive under the accession number SRP062202. Except for the unclassified and other (individual genus with a relative percentage of <1%) organisms, overall, 35 genera were identified in anoxic DHSTP sewage (Supplementary Figure S6). Among them, Lewinella was common in all sewage treatments, regardless of the incubation conditions. The duplicates of individual sewage treatments exhibited high similarity in bacterial community structures (Supplementary Figure S6).

The relative abundance of Thauera spp. was extremely low (~0.5%) in the initial anoxic sewage. During denitrifying incubation with testosterone, Thauera was apparently enriched with time in anoxic sewage (14% abundance after 96 h) (Figure 7a). The growth of Thauera spp. did not improve in anoxic sewage incubated with testosterone or nitrate alone.

Figure 7
figure 7

(a) Genus-level phylogenetic analysis (Illumina MiSeq) revealed the enrichment of Thauera spp. in anoxic sewage incubated with testosterone and nitrate. See Supplementary Figure S6 for detailed information. (b) PCR-based functional assay using degenerate primers (see Supplementary Figure S1 for sequences) derived from the atcA genes. (bI) Agarose gel electrophoresis showed that the atcA genes are harbored only by androgen-degrading anaerobes. (bII) Temporal increase in atcA-specific PCR products was observed only in anoxic DHSTP sewage incubated with testosterone and nitrate. (bIII) Neighbor-joining tree of atcA gene fragments obtained from anoxic DHSTP sewage incubated with testosterone and nitrate for 120 h. atcA-like sequences are shown in Supplementary Figure S8. The gene encoding the large subunit (MhyADHL) of 3-hydroxycyclohexanone dehydrogenase from Ali. denitrificans served as an outgroup sequence.

Full size image

PCR amplification of atcA-like genes in anoxic sewage

Degenerate gene-specific primers were used for detecting atcA in anoxic sewage. Chromosomal DNA was isolated from androgen-degrading anaerobes, Sdo. denitrificans, Stl. denitrificans and T. terpenica 58Eu. We found that T. terpenica is indeed able to degrade testosterone under denitrifying conditions (Supplementary Figure S7). The negative controls were DNA from strictly aerobic androgen degraders, C. testosteroni, Gordonia cholesterolivorans (Wang et al., 2013a), and Sphingomonas sp. KC8 (Roh and Chu, 2010). PCR products with the expected size of ~1100 bp were amplified from the androgen-degrading anaerobes but not from the androgen-degrading aerobes (Figure 7bI). Note that the atcA gene probe could not be used to detect the related genes (33–39% sequence identity), including the 3-hydroxycyclohexanone dehydrogenase gene of Ali. denitrificans, the gene encoding the molybdopterin-binding protein (accession number: WP 011229495) of Geobacillus kaustophilus, and the gene coding for the carbon monoxide dehydrogenase (accession number: P19913) of Hydrogenophaga pseudoflava, under the PCR conditions used.

To investigate the phylogeny of the atcA-harboring bacteria enriched in anoxic sewage incubated with testosterone and nitrate, DNA extracted from three treatments of anoxic DHSTP sewage was used as the nucleotide template. During denitrifying incubation with testosterone, the amount of PCR products in anoxic sewage increased with time (Figure 7bII). By contrast, the amount of PCR products did not increase in anoxic sewage incubated with testosterone or nitrate alone. PCR products amplified from denitrifying sewage (120 h of incubation with testosterone) were cloned in E. coli. Twenty clones were randomly selected for sequencing. The DNA fragments revealed minor differences in nucleotide sequences (Supplementary Figure S8). According to the results of phylogenetic analysis, all partial atcA sequences showed the highest similarities to that from T. terpenica (Figure 7bIII).

Discussion

We adopted a holistic approach to investigate anaerobic steroid biodegradation in the environment. This included the detection of signature metabolites, profiling of the community structure, and detection of specific functional genes. It was thus necessary to begin by investigating the genome and transcriptome of Sdo. denitrificans anaerobically grown on testosterone. This enabled the identification of relevant biomarkers that could be used in our environmental study. We purified and characterized 1-testosterone hydratase/dehydrogenase (AtcABC) and the corresponding genes were identified in the Sdo. denitrificans genome. The phylogenetic analysis of the sequences of AtcABC suggested that this enzyme belongs to the xanthine oxidase family containing molybdopterin, FAD, and iron-sulfur clusters. Nonetheless, the prosthetic groups of AtcABC are yet to be experimentally confirmed. The transcriptomic analysis revealed the upregulation of the atc genes during anaerobic growth on testosterone, indicating that AtcABC is crucial for anaerobic testosterone catabolism. We also identified the genes coding for a putative molybdoenzyme that is highly similar (39–68%) to a putative steroid C25 dehydrogenase (S25dA4B4C4D4) of Stl. denitrificans. This bacterium exhibits an extremely narrow substrate spectrum (limited to a few sterols, androgens, and long-chain fatty acids) and its genome harbors genes coding for eight C25 dehydrogenase-like enzymes, suggesting their roles in the anaerobic hydroxylation of different intermediates of anaerobic steroid catabolism (Dermer and Fuchs, 2012). Moreover, highly similar genes were found in the T. terpenica genome, implying their catabolic role in anaerobic testosterone degradation. However, we could not detect the specific expression of the putative steroid C25 dehydrogenase genes in the Sdo. denitrificans cells anaerobically grown on testosterone. Nevertheless, the catabolic function of this gene cluster remains to be unraveled. The genes encoding the redox enzymes (for example, 3-ketosteroid Δ1-dehydrogenase) were also found on the Sdo. denitrificans chromosome. Their similar expression during aerobic and anaerobic growth may indicate that these genes function in both aerobic and anaerobic testosterone catabolism. The tesB gene was not be found in the chromosome of Sdo. denitrificans. The detection of 3,4-DHSA in the aerobically grown cells (Wang et al., 2013b) suggested that Sdo. denitrificans adopts a meta-cleavage dioxygenase to open the steroid A-ring. This bacterium might use an enzyme exhibiting a low sequence similarity with the tesB gene product of C. testosteroni to aerobically cleave the steroid A-ring. Nonetheless, the possibility that a tesB-like gene is harbored in an extra-chromosomal genetic element of Sdo. denitrificans cannot be excluded. The presence of catabolic genes in plasmids has been reported in numerous bacteria (Roselli et al., 2013; Stolz, 2014).

In the chromosome of Sdo. denitrificans (and other androgen-degrading anaerobes, Stl. denitrificans and T. terpenica), we identified β-oxidation genes whose products were highly similar (62–88% sequence identity) to the enzymes of C. testosteroni involved in the degradation of the C- and D-rings of testosterone (Horinouchi et al., 2014). A similar β-oxidative mechanism was demonstrated in the degradation of steroid C/D rings by actinobacteria (Casabon et al., 2013a, 2013b). β-oxidative enzymes are active under both oxic and anoxic conditions, and a relevant example is the side-chain degradation of cholesterol by Stl. denitrificans (Lin et al., 2015). Moreover, the β-oxidation genes of Sdo. denitrificans were expressed during anaerobic and aerobic growth with testosterone, but they were slightly upregulated under anoxic conditions. One may thus envisage that Sdo. denitrificans uses the products of the β-oxidation genes to degrade the C/D rings of testosterone, regardless of oxygen availability. The biochemical mechanism of anaerobic cleavage of the B-ring by Sdo. denitrificans is completely unknown. The androgen metabolite profiles (Wang et al., 2013b) and genomic and transcriptomic analyses suggested that in the anaerobic pathway of androgen catabolism by Sdo. denitrificans, the cleavage of the B-ring may precede the degradation of C/D rings.

The transcriptomic analysis revealed the apparent upregulation of various membrane transporters and at least one gene cluster involved in cell wall and membrane biogenesis during the anaerobic growth of Sdo. denitrificans on testosterone. The lipopolysaccharide leaflet on the outer surface of the outer membrane impedes steroid uptake by proteobacteria through passive diffusion (Plésiat and Nikaido, 1992). The biochemical mechanisms involved in susbstrate accession and uptake are thus crucial for the survival of steroid-degrading proteobacteria in the environment. The purification and characterization of the membrane transporters for the uptake of hydrophobic compounds is difficult, and the alternative strategy includes the identification of the transporter genes through mutagenesis. For example, by mutational analysis, Mohn et al. (2008) identified the actinobacterial genes encoding the cholesterol transporter. Sdo. denitrificans and Stl. denitrificans cannot not grow on the media solidified with agar or gelrite, therefore preventing the use of conventional molecular biological techniques. Molecular biology experiments have been performed using various Thauera strains (Bhandare et al., 2006). Thus, T. terpenica 58Eu can be used as another model organism for anaerobic steroid degradation.

The signature metabolite 2,3-SAOA and the 1-testosterone hydratase/dehydrogenase genes were identified in anoxic sewage incubated with testosterone and nitrate, indicating that denitrifying bacteria in anoxic sewage adopt the 2,3-seco pathway to degrade testosterone. Although the relative abundance of Steroidobacter spp. did not apparently increase during the denitrifying incubation of DHSTP sewage with testosterone (1 mm), our data based on the enrichment experiments cannot exclude the role of Steroidobacter in in situ androgen degradation. Considering the highly conserved catabolic strategy for anaerobic androgen degradation, the genetic information obtained from this model microorganism serves as a basis for environmental investigations.

The quantification of androgens in the influent and effluent water samples from DHSTP suggested that most androgens (>99%) were degraded by the sewage microorganisms. Bacteria belonging to Saprospiraceae (mainly Lewinella) were dominant microorganisms in the anoxic sewage of DHSTP; however, their abundance appeared stable in different treatments. Although most members of Saprospiraceae are marine species, some strains were isolated from activated sludge (Kragelund et al., 2008; Xia et al., 2008; Nielsen et al., 2009). The members of Saprospiraceae are capable of hydrolyzing complex carbon sources, including carbohydrates and proteins (Kämpfer, 1995; Xia et al., 2007). Based on our results, we posit that the members of Saprospiraceae may be involved in the degradation of organic compounds other than androgens in sewage.

The redox transformations of testosterone to other androgens were observed in denitrifying sewage incubated with testosterone for 30 min. However, Thauera was not enriched until 72 h of the treatment. In addition, the transformation of testosterone to 1-dehydrotestosterone was detected in anoxic sewage incubated without nitrate, where the abundance of Thauera spp. was extremely low. Our data thus suggested that in the denitrifying DHSTP sewage, testosterone was first transformed to other androgens via redox reactions by various microorganisms, including the initial Thauera spp. The redox transformations of steroids have been reported in numerous microorganisms, including bacteria (Kisiela et al., 2012) and fungi (Kristan and Rižner, 2012). For example, the 3-ketosteroid Δ1-dehydrogenase and cholesterol dehydrogenase of Stl. denitrificans can function under both oxic and anoxic conditions (Chiang et al., 2008a, 2008b). After the initial redox transformations of testosterone, T. terpenica cleaved the steroidal core ring, extracted the carbons and energy from steroids, and began proliferating. T. terpenica can degrade bicyclic terpenoids (Foss and Harder, 1998); however, its ability to degrade steroids has not been reported. Our data indicated that T. terpenica 58Eu (DSMZ 12139) anaerobically degrades testosterone also through the 2,3-seco pathway (Supplementary Figure S7). Moreover, the genome of T. terpenica harbors the clustered genes that are highly similar to those involved in anaerobic testosterone catabolism by Sdo. dentirificans. Although T. terpenica is not a dominant organism in the DHSTP sewage, which usually contains androgens at approximately 10 μg l−1 (35 nm), this bacterium can efficiently respond to the pulsed androgen input (1 mm in this study), suggesting that T. terpenica can be used in the bioremediation of androgen-contaminated ecosystems.

A noticeable feature of Sdo. denitrificans is its narrow substrate spectrum, which is restricted to a few steroids (natural estrogens and androgens), fatty acids (C2–C7), and glutamate (Fahrbach et al., 2008). Moreover, it cannot grow on complex media composed mainly of peptides and carbohydrates. Analyzing the Sdo. denitrificans genome reveals the lack of catabolic genes of alkanes, aliphatic alcohols, long-chain fatty acids, monoaromatic compounds and carbohydrates, which are abundant in sewage (Harrison et al., 2006). By contrast, T. terpenica can utilize a number of organic compounds as carbon and energy sources (Foss and Harder, 1998), which is reflected by its ability to grow on complex media. The versatile metabolic capability of T. terpenica may explain why this organism, but not Sdo. denitrificans, was apparently enriched in the testosterone-incubated anoxic sewage. Investigating the utilization kinetics of Sdo. denitrificans and T. terpenica toward testosterone may also provide an explanation for the overwhelming enrichment of T. terpenica in the anoxic sewage incubated with testosterone.

Catabolic genes involved in the biodegradation of complex compounds have been used as molecular markers in environmental investigations (Kuntze et al., 2011). We propose the use of the atcA gene as the biomarker for anaerobic steroid biodegradation because (i) AtcABC catalyzes a crucial hydration/dehydrogenation reaction in anaerobic steroid catabolism, including androgens and cholesterol, (ii) AtcABC exhibits a narrow substrate spectrum and (iii) the large subunit (AtcA) contains the unusual molybdopterin-binding domain. Although AtcABC exhibits some sequence similarity (>38%) to 3-hydroxycyclohexanone dehydrogenase from Ali. denitrificans, it cannot transform the monocyclic compounds. This observation is reflected by the inability of Sdo. denitrificans to grow on cyclohexanol (Fahrbach et al., 2008). We observed a temporal increase in atcA PCR products during the denitrifying incubation of sewage samples with testosterone. This increase was absent in sewage incubated with testosterone or nitrate alone. These results show that AtcABC is involved in anaerobic testosterone biodegradation. Furthermore, they validated atcA as a competent and reliable biomarker for monitoring anaerobic steroid biodegradation. Androgens serve as intermediates in both aerobic and anaerobic catabolism of cholesterol (Van der Geize et al., 2007; Wang et al., 2013a). For example, in the anaerobic catabolic pathway, Stl. denitrificans degrades the side chain of cholesterol to produce androgens, whose A-ring is then cleaved through the 2,3-seco pathway (Lin et al., 2015). Cholesterol and other sterols are highly abundant in the environment. The determination of steroid contents and substrate utilization patterns of the microorganisms in the environmental samples may help to elucidate the metabolic capacity and ecological roles of atcA-harboring microorganisms.

Isotope-independent approaches based on the UPLC—tandem mass spectrometry detection of signature metabolites, community structure analysis, and PCR-based functional assays can be used for investigating the bacterial biodegradation of complex and hydrophobic compounds like steroids, for which conventional culture-independent techniques, such as stable-isotope probing and fluorescence in situ hybridization-microautoradiography, are unavailable. Our integrative omics approach is particularly useful for compounds that are highly abundant (for example, sterols, alkanes and aromatic compounds) in the environment because then substrates with a concentration close to detection levels in most natural environments can be used to enrich the potential degraders.