Identification of Comamonas testosteroni as an androgen degrader in sewage

Numerous studies have reported the masculinization of freshwater wildlife exposed to androgens in polluted rivers. Microbial degradation is a crucial mechanism for eliminating steroid hormones from contaminated ecosystems. The aerobic degradation of testosterone was observed in various bacterial isolates. However, the ecophysiological relevance of androgen-degrading microorganisms in the environment is unclear. Here, we investigated the biochemical mechanisms and corresponding microorganisms of androgen degradation in aerobic sewage. Sewage samples collected from the Dihua Sewage Treatment Plant (Taipei, Taiwan) were aerobically incubated with testosterone (1 mM). Androgen metabolite analysis revealed that bacteria adopt the 9, 10-seco pathway to degrade testosterone. A metagenomic analysis indicated the apparent enrichment of Comamonas spp. (mainly C. testosteroni) and Pseudomonas spp. in sewage incubated with testosterone. We used the degenerate primers derived from the meta-cleavage dioxygenase gene (tesB) of various proteobacteria to track this essential catabolic gene in the sewage. The amplified sequences showed the highest similarity (87–96%) to tesB of C. testosteroni. Using quantitative PCR, we detected a remarkable increase of the 16S rRNA and catabolic genes of C. testosteroni in the testosterone-treated sewage. Together, our data suggest that C. testosteroni, the model microorganism for aerobic testosterone degradation, plays a role in androgen biodegradation in aerobic sewage.

Steroid hormones of either natural or anthropogenic origin are ubiquitous in various environments such as manures, biosolids, soil, sediments, groundwater, and surface water 1,2 . These compounds typically occur at low concentrations (ng L −1 to μ g L −1 ) in surface water [3][4][5][6][7][8][9] . However, steroid hormones have attracted increasing attention because of their ability to act as endocrine disruptors and thus adversely affect wildlife physiology and behavior, even at picomolar concentrations 10,11 . The masculinization of aquatic vertebrates exposed to androgens has been comprehensively reported 12,13,14 . For instance, defeminization of female fish was observed when wild fathead minnows were exposed to cattle feedlot effluent 15 .
In developed countries, sewage treatment plants are crucial for removing steroid hormones produced by humans and livestock 9,16 . The degradation of testosterone by microbial activity has been observed in several environmental matrices such as soil 17 , biosolids in wastewater treatment plants 18 , manure-treated soil 19 , and stream sediments 20 . Numerous studies have reported the essential role of bacterial degradation in removing these endocrine disruptors from the environment [21][22][23] . Actinobacteria and proteobacteria capable of androgen degradation have been isolated and characterized [24][25][26][27] . For instance, various actinobacteria, including Rhodococcus spp., can use androgens as the sole source of carbon and energy 26,27 . A betaproteobacterium, Comamonas testosteroni, has received special attention and its androgen catabolic intermediates and genes have been studied in detail 25 . C. testosteroni is frequently present in polluted environments 28 . C. testosteroni strains can use various hydrocarbons, including steroids (e.g., androgens and bile acids), monoaromatic compounds, acetate, and lactate, as their sole carbon sources and show resistance to heavy metals and antibiotics [29][30][31][32] .
As shown in Fig. 1, the aerobic degradation of testosterone by C. testosteroni is considered to be initiated by the dehydrogenation of the 17β -hydroxyl group to androst-4-en-3,17-dione (AD), which is then converted to androsta-1, 4-diene-3, 17-dione (ADD). The degradation of the sterane structure begins with the introduction of a hydroxyl group at C-9 of the steroid substrate 25 . The resulting intermediate is extremely unstable and undergoes simultaneous cleavage of the B-ring accompanied by aromatization of the A-ring to produce a secosteroid, 3-hydroxy-9, 10-seco-androsta-1, 3, 5(10)-triene-9, 17-dione . The further cleavage of the core ring system proceeds through hydroxylation at C-4 33 , and the A-ring is then split via TesB-mediated meta-cleavage. The tesB-disrupted mutant does not grow on testosterone, indicating that dioxygenase TesB is essential for aerobic testosterone degradation 34 . The tesB gene is embedded in a gene cluster of C. testosteroni comprising 18 androgen catabolic genes 35 . The gene cluster is widely present in androgen-degrading proteobacteria, including species within the genera Burkholderia, Comamonas, Cupriavidus, Glaciecola, Hydrocarboniphaga, Marinobacterium, Novosphingobium, Pseudoalteromonas, Pseudomonas, Shewanella, and Sphingomonas 25,36 . In addition to the well-studied 9, 10-seco pathway, alternative catabolic pathways of androgens have been observed in bacteria. For instance, aerobically grown Sterolibacterium denitrificans adopts an oxygenase-independent pathway to degrade steroid substrates 37 .
The biochemical mechanisms underlying aerobic androgen biodegradation were studied in pure cultures 25,33,34,38,39 . However, studies on the catabolic mechanisms and agents of in situ androgen biodegradation are lacking. It is unknown which androgen biodegradation pathway is functional in polluted ecosystems. Moreover, the distribution and abundance of androgen-degrading bacteria in the environment are yet to be investigated. In the present study, we examined microbial androgen degradation in the aerobic sewage of the Dihua Sewage Treatment Plant (DHSTP), which treats domestic wastewater produced by the three million residents of Taipei City, Taiwan. We used the following approaches: (i) identification of androgen metabolites through ultra-performance liquid chromatography -tandem mass spectrometry (UPLC-MS/MS), (ii) phylogenetic identification of the androgen degraders through Illumina Miseq sequencing, and (iii) detection of the essential catabolic gene tesB through PCR.

UPLC-MS/MS identification of androgenic metabolites in DHSTP sewage.
Androgenic metabolites were extracted from various sewage treatment samples and identified through UPLC-atmospheric pressure chemical ionization (APCI)-MS/MS (Fig. 2). No testosterone was detected in the original DHSTP sewage ( Fig. 2A). Furthermore, testosterone was not degraded when incubated with autoclaved sewage (Fig. 2B). By contrast, testosterone was transformed to 1-dehydrotestosterone, AD, and ADD during the first two days of aerobic incubation of active sewage (Fig. 2CI). After 72 hours of incubation, the intensities of peaks corresponding to the androgens decreased considerably. We then used the extracted ion current at m/z 301.18 (the predominant ion peak of 3-HSA) to detect 3-HSA, the signature metabolite of the 9, 10-seco pathway, in the aerobic sewage (Fig. 2CIII). The UPLC retention time (5.10 min; Fig. 2CIII) and MS/MS fragmentation spectrum (Fig. 2CIV) of the extracted ion was comparable with that of the authentic standard.
We determined the androgenic activity of the initial intermediates of the 9, 10-seco pathway by using a lacZ-based yeast androgen assay. The results showed that testosterone, 1-dehydrotestosterone, AD, and ADD exhibited apparent androgenic activity. However, the secosteroid 3-HSA had no detectable androgenic activity even at a concentration of 500 μ M (Fig. 3A). The androgenic activity of the ethyl acetate extracts from the sewage treatment samples was then determined (Fig. 3B). The androgenic activity of the sewage extracts decreased over time, which is consistent with the results of the androgen metabolite analysis.
Phylogenetic identification of androgen-degrading bacteria in aerobic sewage. DNA was extracted from various sewage treatment samples. The V3-V4 hypervariable region of bacterial 16S rRNA gene was amplified through PCR, and the resulting amplicons were sequenced using an Illumina MiSeq sequencer (Illumina; San Diego, CA, USA). The sequences were analyzed using BaseSpace 16S Metagenomics App V1.01 (Illumina; San Diego, CA, USA) (Fig. 4). The nucleotide sequence data set was deposited in the National Center for Biotechnology Information (NCBI) Sequence Read Archive (accession number: SRP062202). From each sample, an average of 306 025 reads was obtained. The duplicates of individual sewage treatment samples exhibited high similarity in bacterial community structure (Fig. S2). Except the unclassified and other (individual genus with a relative percentage of < 1%) microorganisms, 35 genera were identified in the DHSTP sewage, among which Lewinella, with a relative abundance between 4% and 13%, was present in all sewage treatment samples, regardless of the incubation conditions (Fig. 4A). Moreover, we observed no apparent enrichment of Lewinella spp. in the testosterone-incubated sewage.
Members of the genus Pseudomonas were slightly enriched in the testosterone-treated aerobic sewage (Fig. 4B). In a replicate, the relative abundance of Pseudomonas spp. reached 6% after 96 hours of incubation. The bacterial community analysis suggested that Pseudomonas entomophila and P. panipatensis were the most  enriched species (Fig. S3). Pseudomonas spp. were not enriched in aerobic sewage incubated without testosterone (Fig. 4B), suggesting that Pseudomonas spp. might play a role in aerobic androgen biodegradation.
Comamonas spp. were enriched in the testosterone-treated aerobic sewage in the first two days, and their relative abundance accounted for approximately 20% of the total bacterial community (Fig. 4C). Further phylogenetic analysis suggested that after 48 hours of incubation with testosterone, most of the obtained sequences were associated with C. testosteroni (43% in the genus of Comamonas) and C. composti (23%) (Fig. 4C). The relative abundance of Comamonas spp. in the bacterial community then decreased with time and they were undetectable after 240 hours of incubation. Comamonas spp. were not enriched in the aerobic sewage incubated without testosterone (Fig. 4C). These results suggest that Comamonas spp. are active catabolic players in the testosterone-treated sewage. The enrichment of Comamonas spp. and Pseudomonas spp. were observed in the experimental replicates (Fig. S4). After 48 hours of aerobic incubation, the relative abundance of Comamonas spp. and Pseudomonas spp. reached 29% and 2%, respectively in the testosterone-treated sewage.
The specificity of degenerate primers was determined using genomic DNA isolated from several bacteria as templates. PCR products (approximately 700 bp) were amplified from the strictly aerobic C. testosteroni but not from the denitrifying Sterolibacterium denitrificans (Fig. 5A). No PCR products could be amplified from steroid-degrading actinobacteria (Fig. 5A). This could be because of the low sequence similarity of 3, 4-DHSA dioxygenases between actinobacteria and proteobacteria (Fig. S5).
When DNA isolated from the testosterone-treated sewage was used as a template, the time course changes in the amount of tesB sequences (Fig. 5B) corresponded with the temporal changes in Comamonas abundance in the aerobic sewage (Fig. 4C). No PCR products could be amplified from DNA isolated from sewage incubated without testosterone. PCR products amplified from the aerobic sewage (48 hours of incubation with testosterone) were cloned in Escherichia coli, and 20 clones were randomly selected for sequencing. All the obtained DNA fragments (Fig. S6, nucleotide sequences) exhibited the highest similarities (87-96%) to the tesB gene of C. testosteroni Quantitative PCR confirmed the remarkable increase of the 16S rRNA and catabolic genes of C. testosteroni in the testosterone-treated sewage. The metagenomic analysis ( Fig. 4) and PCR-based functional assay ( Fig. 5) could not robustly support an increase of the C. testosteroni population in the testosterone-treated sewage. Therefore, we performed a quantitative PCR study to examine the temporal changes in the 16S rRNA and tesB genes of C. testosteroni in different sewage treatments. The abundance of the C. testosteroni genes in each sample was normalized by the total eubacterial 16S rRNA gene. The duplicates of individual sewage treatment samples exhibited high similarity, and the apparent increase of the C. testosteroni genes was only observed in the testosterone-treated sewage (Fig. 6). The real-time quantitative PCR results were coherent with those of the conventional PCR assays. The relative abundance of the C. testosteroni 16S rRNA and tesB genes increased after 48 hours of incubation and decreased thereafter. After 48 hours of aerobic incubation, the abundances of the 16S rRNA (Fig. 6A) and tesB genes (Fig. 6B) in the duplicates reached 18.3~23.5% and 0.9~1.0%, respectively.

Discussion
Activated sludge processes are used to treat wastewater in most cities in developed countries. The basic activated sludge process involves using a microbial community to mineralize organic carbons and oxidize ammonia (through nitrification) under aerobic conditions. Microbial communities play a crucial role in bioprocesses such as wastewater treatment and soil remediation 42,43 . However, exploiting the microbial resources requires an understanding of not only their phylogeny, but also their metabolic functions and ecological roles. Steroid hormones have been recognized as a major group of endocrine-disrupting chemicals, and sewage treatment plants play a critical role in removing these highly bioactive compounds 9,16 . In the present study, to explore the underlying biochemical mechanisms and microorganisms involved in androgen degradation in aerobic sewage, we applied various isotope-independent approaches, including the UPLC-MS/MS-based detection of the signature metabolites, community structure analysis, and PCR-based functional assays.
The 9, 10-seco pathway has been demonstrated in proteobacteria 25 and actinobacteria 27,36 . We assigned 3-HSA as the signature metabolite of this aerobic degradation pathway because (i) 3-HSA does not possess a sterane    Fig. S4. The sequences were retrieved from the UniProtKB/TrEMBL database by performing standard BLASTP search with 3, 4-DHSA dioxygenase (the product of tesB) from C. testosteroni TA441 as a query. In known steroid-degrading proteobacteria, the dioxygenase gene is harbored in the conserved testosterone-degradation gene cluster 25 . * The biphenyl-2, 3-diol 1, 2-dixoxygenase of P. pseudoalcaligenes.
structure and exhibits extremely weak androgen activity. Thus, 3-HSA detection demonstrates androgen biodegradation in the studied ecosystems. (ii) Compared with other secosteroids in this aerobic pathway, 3-HSA often accumulates in androgen-degrading bacterial cultures 33,38,44 and could be detected through UPLC-MS/MS. Unlike sterane-containing androgens (e.g., testosterone, 1-dehydrotestosterone, AD, and ADD), the androgen metabolite 3-HSA cannot be easily ionized through APCI. Therefore, we used an extracted ion current method to detect this secosteroid. In the testosterone-treated aerobic sewage, the androgen activity decreased remarkably with time. However, the androgen activity did not decrease in the autoclaved sewage, indicating that testosterone degradation in sewage is exclusively due to microbial activity. We detected the common androgen metabolites, including 1-dehydrotestosterone, AD, and ADD, during the first two days. The biotransformation of testosterone to these androgens has been widely reported in natural and engineered ecosystems such as soils 17,19 , wastewater treatment plants 18 , and river sediments 20 . Furthermore, the corresponding redox enzymes have been identified in microorganisms including bacteria 45 , yeast, and fungi 46 . Although the detected androgen metabolites, namely 1-dehydrotestosterone, AD, and ADD, may be produced mainly by bacteria, our data do not exclude the role of eukaryotic microorganisms in the redox biotransformation reactions. The UPLC retention time and the MS/MS fragmentation spectrum of the identified metabolite are comparable to that of the authentic standard, suggesting the production of 3-HSA in the aerobic sewage incubated with testosterone for three days. 3-HSA, the key intermediate of the 9, 10-seco pathway 25,47 , has rarely been detected in environmental samples. Yang et al. 24 reported the detection of this secosteroid in bacterial cultures enriched from swine manure. An alternative steroid catabolic pathway, 2, 3-seco pathway, is adopted by aerobically grown Sterolibacterium denitrificans 37 , and 1-dehydrotestosterone, AD, ADD also serve as the initial metabolites in this alternative pathway. However, we did not detect 2, 3-SAOA, the representative androgen metabolite, in the aerobic sewage treatment samples. The results indicate that the 9, 10-seco pathway is functional in aerobic androgen biodegradation in sewage.
The 9, 10-seco pathway was described only in bacteria 25,36 . Thus, we studied the changes in the bacterial community in various aerobic sewage treatment samples. In testosterone-treated sewage, the appearance of 3-HSA and decrease in androgenic activity were accompanied by the enrichment of proteobacteria in the bacterial community. Although the 3, 4-DHSA dioxygenase-dependent 9, 10-seco pathway has been commonly identified in various bacteria, including proteobacteria 25 and actinobacteria 37,48 , our Illumina MiSeq data suggest that proteobacteria, including Comamonas and Pseudomonas, are the androgen degraders in aerobic DHSTP sewage. None of the actinobacterial genera showed a relative abundance of > 1% in the initial aerobic sewage samples collected from the DHSTP. Moreover, we observed no enrichment of actinobacteria in the testosterone-treated sewage. A recent investigation that analyzed the microbial communities in 13 Danish wastewater treatment plants also revealed betaproteobacteria as the predominant components, whereas members of actinobacteria exhibited extremely low relative abundance 49 . Bacteria belonging to Saprospiraceae (mainly Lewinella spp.) were predominantly present in the aerobic sewage of DHSTP; however, their abundance did not apparently increase in the testosterone-treated sewage. To our knowledge, Lewinella spp. have not been identified as steroid degraders. Nevertheless, our current data cannot exclude that Lewinella spp. and actinobacterial species play a role, directly or indirectly, in androgen degradation in aerobic sewage. This is because (i) androgen catabolism does not necessarily result in an enrichment of the populations of the degraders, especially in a short-term incubation; and (ii) in aerobic sewage, an increase of the degrader population could be counteracted by removal processes like predation and viral lysis.
The Illumina Miseq analysis of 16S rRNA genes enriched in the testosterone-treated sewage suggested that Pseudomonas spp. (likely P. entomophila and P. panipatensis) play a role in aerobic androgen degradation. These two Pseudomonas species were not described as testosterone-utilizing bacteria. Moreover, most steroid catabolic genes, including the tesB gene, were not found in the genomes of these two Pseudomonas species 36,50 . Our tesB gene probe was derived from several steroid-degrading proteobacteria including P. putida. However, the 40 sequenced tesB fragments did not exhibit high similarity to that of any Pseudomonas species. It is worth noting that the tesB sequences were amplified from the sewage incubated for two days, in which the bacterial community was dominated by Comamonas spp., but not Pseudomonas spp. Accordingly, the enrichment of Pseudomonas spp. might be due to the indirect involvement in the bioprocess (e.g., feeding on metabolites excreted by the androgen degraders). Nevertheless, further investigation is necessary to elucidate the androgen degradation potential of Pseudomonas species enriched in the aerobic sewage.
The Illumina Miseq analysis revealed an apparent dominance of Comamonas spp. (likely C. testosteroni and C. composti) in the testosterone-treated sewage. This is in line with the observed increase in the relative abundance the C. testosteroni tesB gene after 48 hours of incubation with testosterone. However, the result of quantitative PCR revealed a much lower relative abundance for the tesB gene as compared to the C. testosteroni 16S rRNA gene. The lower abundance of the C. testosteroni tesB gene in the testosterone-treated sewage may be due to (i) Figure 6. Real-time quantitative PCR indicated the temporal changes in the C. testosteroni 16S rRNA (A) and tesB (B) gene copies in the testosterone-treated sewage. Relative abundance of individual C. testosteroni genes was calculated as a proportion of the total number of bacterial 16S rRNA gene copies. Sewage treatments were performed in duplicate and the grey (Replicate 1) and black (Replicate 2) lines represent different replicates. Data are shown as the mean ± SE of three experimental measurements. Primer pairs CteA2, TesBq, and Eub were used to amplify the 16S rRNA and tesB genes of C. testosteroni as well as total eubacterial population, respectively. Real-time quantitative PCR standard curves obtained using the three primer pairs are shown in Fig. S11.
Comamonas spp. were also enriched in testosterone-amended swine manure 24 . Thus, Comamonas species might play an important role in aerobic androgen degradation in the environment, such as in sewage and agricultural soil treated with manure. Members of the Comamonas genus belong to betaproteobacteria, with versatile metabolic capacities and possess a wide spectrum of substrate utilization. To date, the genus Comamonas encompasses 11 species that have been validated: C. aquatica, C. badia, C. composti, C. denitrificans, C. kerstersii, C. koreensis, C. nitrativorans, C. odontotermitis, C. terrigena, C. testosteroni, and C. thiooxidans 51,52 . These species exhibit extremely different physiological and metabolic capabilities. For instance, C. thiooxidans can grow under anoxic conditions, whereas other species are strictly aerobes 51 . Among them, only C. testosteroni was reported to utilize steroids as the sole carbon source 53 . The comparative genomic analysis also indicated that steroid catabolic genes are only present in the C. testosteroni strains, but not in other Comamonas species 36 . In the present study, although C. composti was assigned as an enriched species during aerobic incubation with testosterone, our metabolite analysis indicated that this species cannot degrade androgens (Fig. S9). Moreover, we did not find a tesB-like gene in the draft genome of C. composti (accession: NZ_AUCQ00000000). Liu et al. 52 compared the genomes of 14 C. testosteroni strains, and steroid catabolic genes were found in all genomes. Considering that androgens are typically present at low concentrations (ng L −1 -μ g L −1 ) in the natural environment, it is unreasonable that C. testosteroni strains have evolved the unusual metabolic capability to use rare and structurally complex carbon sources such as testosterone. This bacterium can also grow on bile acids 32 , which often occur in significant amounts in the environment. One may thus envisage that bile acids could serve as the target substrates of the steroid catabolic enzymes of C. testosteroni.
C. testosteroni is the most widely studied microorganism for aerobic androgen degradation 25 . Here, for the first time, we provide strong evidence showing that C. testosteroni plays a role in removing androgens from the environment. Considering that the degradation of steroid hormones in anaerobic environments is typically slow 54 , aeration and introducing aerobic degraders could be efficient bioremediation options. Although C. testosteroni is not a dominant species (0.4% relative abundance in the initial sewage bacterial community) in the DHSTP sewage, which typically contains androgen concentrations of approximately 35 nM 16 , these bacteria can efficiently respond to changes in androgen input (1 mM in this study), suggesting that C. testosteroni could be used in the bioremediation of steroid-contaminated ecosystems.
The high-throughput sequencing of 16S rRNA gene has enabled a deeper understanding of bacterial diversity in complex environmental samples; however, the method also introduces ambiguity because of the limited taxonomic capability of short reads (450 bp in the present study). Moreover, the taxonomic assignments are inconsistent among different classification methods 55 . We identified several Comamonas species in the testosterone-treated sewage. However, C. composti showed no androgen-degrading ability. A recent genomic study also indicated that the distribution of the steroid degradation pathways among proteobacterial taxa is generally patchy, and only a few genomes from each proteobacterial genus appear to encode steroid catabolic genes 36 . Thus, our results suggest that the taxonomic assignment of bacteria based the high-throughput sequencing of 16S rRNA genes alone is insufficient for characterizing biodegradation events. The combination of 16S rDNA-based phylogenetic analysis, signature metabolite probing, and PCR-based functional detection proposed in this study may provide information on both the biochemical mechanisms and the active players in bacterial degradation. Future studies should include a kinetic analysis of substrate utilization by C. testosteroni and the immobilization of bacterial cells to improve the efficiency of androgen removal in sewage treatment plants. Moreover, systems biology approaches, such as metatranscriptomics or metaproteomics coupled with metagenomics, can elucidate the ecophysiological relevance of androgen-degrading microbes. This should also facilitate developing or engineering microbial consortia for the efficient removal of steroids from polluted ecosystems.

Conclusions
The application of an integrated approach comprising several culture-independent tools appears useful for investigating the microbiology and biochemistry of environmentally relevant processes such as steroid biodegradation. Under our experimental conditions, aerobic androgen biodegradation in the testosterone-treated sewage proceeds through the established 9, 10-seco pathway. However, UPLC-MS/MS analyses cannot help to identify the metabolites from undescribed degradation pathways. Our data thus do not exclude the operation of other degradation pathways in the testosterone-treated aerobic sewage. The metegenomic analysis, PCR-based functional assay, and quantitative PCR supported the catabolic role of C. testosteroni in the testosterone-treated sewage. However, our data did not indicate the crucial role of C. testosteroni in aerobic androgen degradation in the operating sewage treatment plants where androgens are present at much lower concentrations.

Materials and Methods
Chemicals and bacterial strains. The chemicals were of analytical grade and were purchased from Fluka, Mallinckrodt Baker, Merck, and Sigma-Aldrich. C. testosteroni ATCC 11996 was obtained from the American Type Culture Collection (Manassas, VA, USA). Gordonia cholesterolivorans DSMZ 45229, Mycobacterium smegmatis DSMZ 43277, and S. denitrificans DSMZ 13999 were purchased from the Deutsche Sammlung für Mikroorganismen und Zellkulturen (Braunschweig, Germany). S. denitrificans was anaerobically grown on testosterone 56 . The other aerobic bacteria were aerobically cultured in Luria-Bertani medium.
Collection of sewage samples. The DHSTP is the largest municipal wastewater treatment plant (500 000 m 3 day −1 ) in Taipei. Along with domestic water, the DHSTP receives industrial, medical, and livestock wastewater as well as groundwater 57 . The design of the DHSTP includes an anoxic/oxic process for removing carbon and nitrogen, and the hydraulic retention time is approximately 10 hours 57,58 . Sewage samples (10 L) were Scientific RepoRts | 6:35386 | DOI: 10.1038/srep35386 collected from the aerobic tank of the DHSTP in June 2014. The aerobic sewage was placed in a sterilized 20-L glass bottle, and delivered to the laboratory within 30 minutes.
Incubation of aerobic sewage with testosterone. The DHSTP aerobic sewage samples (0.5 L sewage in 2-L glass bottles) were incubated under the following conditions: autoclaved sewage with testosterone (1 mM), active sewage without testosterone, and active sewage with testosterone. The sewage treatments were performed in duplicate, and the bottles were incubated at 25 °C with stirring at 160 rpm for two weeks. Samples (10 mL) were withdrawn from the bottles every 12 hours and stored at − 80 °C before use. The androgenic activity and androgen metabolites in the sewage samples were detected using the yeast androgen assay and UPLC-APCI-MS/MS, respectively. The bacterial 16S rRNA and functional tesB genes in the sewage samples were analyzed through Illumina MiSeq sequencing and PCR-based functional assays, respectively.

UPLC-MS/MS identification of androgenic metabolites in sewage.
Aerobic sewage samples (1 mL) were extracted three times with the same volume of ethyl acetate. The extracts were pooled, the solvent was evaporated, and the residues were re-dissolved in 100 μ L of methanol. The ethyl acetate extractable samples were analyzed through UPLC-APCI-MS/MS, as described by Wang et al. 38 .
lacZ-based yeast androgen bioassay. The sewage samples (0.5 mL) were extracted three times using equal volume of ethyl acetate. After the solvent evaporated, the extracts were re-dissolved in the same volume of dimethyl sulfoxide (DMSO), and the androgenic activity in the sewage samples was determined using a lacZ-based yeast androgen assay. The yeast androgen bioassay was conducted as described by Fox et al. 59 with slight modifications. Briefly, the individual steroid standards or sewage extracts were dissolved in DMSO, and the final concentration of DMSO in the assays (200 μ L) was 1% (v/v). The resulting DMSO solutions (2 μ L) were added to yeast cultures (198 μ L, initial OD 600nm = 0.5) located in a 96-well microtiter plate. The β -galactosidase activity was determined after 18 hours incubation at 30 °C. The yeast suspension (25 μ L) was added to a Z buffer (225 μ L) containing o-nitrophenol-β -D-galactopyranoside (2 mM), and the reaction mixtures were incubated at 37 °C for 30 min. The reactions were stopped by adding 100 μ L of 1 M sodium carbonate, and the amount of yellow-colored nitrophenol product was determined spectrophotometrically at 420 nm on a plate spectrophotometer (SpectraMax M2e, Molecular Devices).
Illumina MiSeq sequencing of bacterial 16S rRNA amplicons. DNA was extracted from the frozen sewage samples by using the Powersoil DNA isolation kit (MO BIO Laboratories). A 16S amplicon library was prepared according to the Illumina 16S Metagenomic Sequencing Library Preparation Guide (/mentation/chemistry_documentation/16s/16s-metagenomic-library-prep-guide-15044223-b. pdf ) with slight modifications. Genomic sections flanking the V3-V4 region of the bacterial 16S rRNA gene were amplified from 24 sewage treatment samples by using HiFi HotStart ReadyMix (KAPA Biosystems) through PCR (95 °C for 3 minutes; 25 cycles: 95 °C for 30 s, 55 °C for 30 s, 72 °C for 30 s, and 72 °C for 5 min). A primer pair flanked by the Illumina Nextera linker sequence (forward primer: 5′ -TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGCCTACGGGNGGCWGCAG-3′ and reverse primer: 5′ -GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGACTACHVGGGTATCTAATCC-3′ ) was used. The PCR products were first separated on an agarose gel, and those with the expected size (approximately 445 bp) were excised from the gel and purified using the GenepHlow Gel/PCR kit (Geneaid). Next, Illumina Nextera XT Index (Illumina) sequencing adapters were integrated to the ends of the amplicons through PCR (95 °C for 3 min; 8 cycles: 95 °C for 30 s, 55 °C for 30 s, 72 °C for 30 s; and 72 °C for 5 min). The final libraries were purified using AMPure XP beads (Beckman Coulter) and quantified using the Qubit dsDNA HS Assay Kit (Life Technologies). The library profiles were randomly analyzed using the Agilent High Sensitivity DNA Kit on BioAnalyzer. To ensure consistency in pooling, all 24 libraries were subjected to quantitative PCR normalization by using KAPA Library Quantification Kits to derive the molar concentration, and the final library mixture was verified through quantitative PCR. The library pool was sequenced on the Illumina MiSeq V2 sequencer by using MiSeq Reagent Kit V3 for paired-end (2 × 300 bp). We analyzed the sequencing data in the Illumina BaseSpace cloud service by using the BaseSpace 16S Metagenomics App (Illumina) (http://support. illumina.com/content/dam/illumina-support/documents/documentation/software_documentation/basespace/ 16s-metagenomics-user-guide-15055860-a.pdf). The reads were classified against the Illumina-curated version of May 2013 Greengenes taxonomy database by using the Ribosomal Database Project (RDP) naïve Bayesian algorithm (http://rdp.cme.msu.edu/classifier/). PCR by using the tesB gene probe. Multiple alignment of the tesB-like genes from eight testosteronedegrading proteobacteria 25 was performed using Align/Assemble (Genious 8.1.4). A degenerate primer pair [forward primer (tesB-f1): 5′ -TAYYYSGCMTCBGGHTGGGA-3′ and reverse primer (tesB-r1): 5′ -WRAAR TCRTGBCCCCA-3′ ] were deduced from the conserved regions (Fig. 7). Furthermore, the tesB fragments were amplified through PCR (94 °C for 2 minutes; 30 cycles: 94 °C for 30 s, 48 °C for 30 s, 72 °C for 60 s; and 72 °C for 10 minutes). The partial tesB sequences (approximately 700 bp) amplified from the aerobic sewage were cloned in E. coli (One Shot TOP10; Invitrogen) by using the pGEM-T Easy Vector Systems (Promega). The tesB genes were sequenced on a ABI 3730xI DNA Analyzer (Applied Biosystems; Carlsbad, CA, USA) with BigDye terminator chemistry according to the manufacturer's instructions.