Modeling and optimization of triclosan biodegradation by the newly isolated Bacillus sp. DL4: Kinetics and metagenomics analysis

Overusing triclosan (TCS) endangered ecological safety and human health, and the pandemic of COVID-19 aggravates the accumulation of TCS in the aquatic environment. Therefore, reducing residual TCS concentrations in the environment is an urgent issue. An aerobic bacterium, Bacillus sp. DL4 was isolated with the capability of TCS biodegradation. Response surface methodology (RSM) and arti�cial neural network (ANN) were carried out to optimize and verify the different condition variables. All the variables were linear and the interaction of the three factors signi�cantly affected TCS removal at the quadratic level (p < 0.001). Under the optimal conditions (35 ℃ , initial pH 7.31, and 5% strain DL4), the TCS removal rate of 95.89 ± 0.68% was observed and found to be consistent with the predicted values from RSM and ANN models. In addition, statistical comparisons between the models indicated that the ANN model had a stronger predictive capability than the RSM model. Kinetic studies showed that TCS degradation was consistent with a pseudo-rst-order kinetic model. Whole genome sequencing indicated that many functional genes were involved in and facilitated TCS degradation. Main metabolite products were detected and identi�ed during the biodegradation process by LC-MS, and a possible degradation pathway was tentatively hypothesized. Overall, this study provides a theoretical foundation for the characterization and mechanism of TCS biodegradation in the environment by Bacillus sp. DL4.


Introduction
The occurrence and fate of pharmaceuticals and personal care products (PPCPs) in the aquatic environment have been paid special attention in the last few decades due to their potential undesirable ecological and human health effects (Kümmerer et al., 2009;Ebele et al., 2017;Tan et al., 2021).Triclosan (2,4,4'-trichloro-2'-hydroxy diphenyl ether, TCS), a typical polychlorinated aromatic antimicrobial agent, has been widely added in PPCPs (e.g., body and facial cleansers, toothpaste, cosmetics, and plastics), which was frequently detected in human samples such as urine, plasma, and breast milk (Ruszkiewicz et al., 2017;Adolfsson-Erici et al., 2002;USEPA, 2008).The triclosan wastewater was released into wastewater treatment plants (WWTPs), leading to the concentration of triclosan in wastewater in uent up to µg/L (van Wijnen et al., 2018; Jia et al., 2020).However, TCS is characterized by lipophilic, acute toxicity, resistance to biodegradation, environmental persistence, and relatively high octanol-water partition coe cient (log Kow = 4.8) (Dong et al., 2021;Dai et al., 2017).Its bioaccumulation and biomagni cation in biota threaten the safety of organisms and may lead to higher concentrations of µg/L to mg/L (Stasinakis et al., 2007;Gao et al., 2019).Meanwhile, triclosan could react with free chlorine in the water to form toxic environmental chlorine-containing by-products.In addition, TCS has been proven to interfere with the endocrine system in animal and human breast cancer cells (Dann et al., 2011), showing the potential for bacterial resistance to antibiotics at critical concentrations.It can also cause skin irritation, immunotoxicity, and neurotoxic reactions when exposed to an extreme amount of TCS (Langsa et al., 2017;Li et al., 2018).
An attractive method for eliminating TCS is biodegradation to convert it into harmless compounds.Compared with the conventional methods (e.g., adsorption, photocatalysis, and chemical oxidation), biological treatment might be able to remove TCS effectively, cost-effective, and environmentally friendly (Maura et al., 2001;Kim et al., 2011;Sabaliunas et al., 2003).According to many reports, the removal of triclosan in wastewater treatment was mainly due to the adsorption and biodegradation of TCS by sludge, but biodegradation played a signi cant role (Chen et al., 2015a;Hena et al., 2021).Recently, researchers have isolated some microorganisms degrading TCS and its derivatives and elucidated the metabolic pathway of these compounds.For fungus, Aspergillus versicolor could tolerate up to 15.69mg/L and biodegrade 71.91% of TCS at an initial concentration of 7.5mg/L (Ertit et al., 2015).Tastan et al. (2016) evaluated that the maximum TCS degradation rates (2.7mg/L) of Rhodotorula mucilaginosa and Penicillium sp. when grown in a mineral salt medium, were 48% and 82%, respectively.For algae, Chlorella pyrenoidosa (Wang et al., 2018) and diatom Navicula sp.(Ding et al., 2018) absorbed TCS and accumulated it but could not convert it.Bai and Acharya (2016) explored how the green alga Nannochloris sp.removed TCS from freshwater and wastewater.They found that 74% of the TCS was removed straight away and 100% of the TCS was gone after 7 days.Besides, some bacteria have been found to degrade TCS, such as Sphingomonas sp.PH-07 (Kim et al., 2011), Sphingopyxis strain KCY1 (Lee et al., 2012), Sphingomonas sp.YL-JM2C (Mulla et al., 2016), Shewanella putrefaciens CN32 (Wang et al., 2017), Dyella sp.WW1 (Wang et al., 2018) and Providencia rettgeri MB-IIT (Balakrishnan and Mohan, 2021).Sphingomonas sp.Rd1 (Hay et al., 2001) and Nitrosomonas europaea (Roh et al., 2009) were used in nitrifying activated sludge for the biodegradation of TCS.A diphenyl ether-degrading bacterium, Sphingomonas sp.PH-07 had shown the ability to degrade TCS and produced three metabolites, including hydroxylated triclosan, 4-chlorophenol, and 2,4-dichlorophenol (Kim et al., 2011).
In addition, some bacteria used TCS as the sole carbon source, such as Pseudomonas putida and Alcaligenes xylosoxidans subsp.denitri cans (Liu et al., 2016;Maura et al., 2001).What's more, ammonia-oxidizing bacteria and heterotrophic microorganisms have demonstrated that they were able to remove TCS in activated sludge (Roh et al., 2009).Although several studies have been performed on TCS biodegradation, the resources of functional bacteria are still scarce and need to be further developed.
Since microbial biodegradation depends not only on the availability of microorganisms but also on different external environmental factors, such as the concentration of pollutants and the initial pH, etc. (Umar Mustapha et al., 2020), optimizing these factors can improve the biodegradation capability of the microorganisms.Response surface methodology (RSM) and arti cial neural network (ANN) statistical tools are widely used optimization tools, both have different perspectives for modeling different parameters.RSM is a multivariate optimization technique to investigate the interaction effects of independent operational parameters, verify experimental precision, identify uncertainty and generate statistical models (Ye et al., 2010).As one of the main platforms for arti cial intelligence, ANN has become increasingly popular with researchers in recent years due to its potential to handle partial data sets and non-linear tuning (Fan et al., 2018).
The primary purpose of this research was to obtain new functional bacteria and to investigate the effect of ambient conditions, including initial pH, temperature, and the inoculum size, on the biodegradation of TCS by the obtained strain.The removal kinetics of TCS were analyzed at different concentrations (5, 10, 15, 20, and 25 mg/L).The intermediates proposed a degradation pathway of TCS, which was identi ed by the liquid chromatography-mass spectrometric (LC-MS) analysis.Finally, the genome of strain DL4 was analyzed to reveal the functional genes associated with the TCS degradation process.The results may enrich the bacterial community and provide valuable information for the potential application of biological treatment TCS methods.

Chemical reagents and culture medium
Triclosan with purity > 98.5% (CAS:3380-34-5) was purchased from Sigma-Aldrich (China).HPLC grade solvents were acquired from Merck (Darmstadt, Germany).The stock solution of TCS (200mg/L) was prepared with methanol and stored in brown bottles at 4℃ before use.All solutions were prepared with ultrapure water from a Milli-Q reagent water system (Millipore, Billerica, MA, USA).

Isolation and identi cation of TCS-degrading bacteria
Triclosan-biodegrading microorganisms were isolated from pharmaceutical wastewater samples collected at Shenyang, Liaoning, China.The strains were puri ed by gradient dilution and streak plate methods (Ren et al., 2014).Cultures were carried out in 250mL Erlenmeyer asks containing 100mL LB medium and incubated aerobically at 30℃ on a rotary shaker (120rpm).Transferred 1mL of the enriched culture into a 10mL centrifugal tube containing 9mL of sterile water and shaken to mix.Serial dilutions obtained the bacterial suspensions of 10 − 2 to 10 − 9 , and approximately 1mL of three appropriate concentrations (10 − 7 , 10 − 8 , and 10 − 9 ) were inoculated in duplicate onto LB solid medium.Disparate mono-colonies selected based on morphological characteristics were subcultured by choosing and streaking three times to obtain isolated, puri ed colonies.
The genomic DNA of obtained strain was extracted using a DNA Mini extraction kit (Sangon Biotech, Shanghai, China).The 16S rDNA gene was ampli ed with universal bacterial primers 27F (5'-AGAGTTTGATCCTGGCTCAG-3') and 1492R (5'-TACCTTGTTACGACTT-3').The PCR was carried out with the previous work (Zhao et al., 2018).The sequence of the strain was summited to GenBank for BLAST analysis in NCBI, and phylogenetic analysis was performed with the algorithm against the database of sequences.Whole genome sequencing and gene function annotation based on De novo assembly was assisted by Sangon Biotech.

Effect of initial concentration of TCS on strain DL4
To investigate the strain DL4 tolerance to triclosan, the experiments were conducted at 5, 10, 15, 20, and 25mg/L initial triclosan concentrations.The experimental sample without TCS (at a concentration of 0 mg/L) was used as a control.5% (v/v) of the obtained strains at the logarithmic phase were cultured in a sterilized CF medium containing different initial TCS concentrations, respectively.The bacterial suspension was taken every six hours to monitor the growth of the strain DL4 and TCS concentration.
The calibration curve of TCS was generated over a wide range of concentrations with strong linearity (R 2 > 0.99).The triplicate samples' relative standard deviations (RSDs) were less than 1%.The detection limit of TCS was 0.01mg/L.
The kinetic parameters re ect the characteristics of pollutants in the mass transfer process, which can theoretically explain the relationship between pollutants and the removal rate of the strain.In this study, a pseudo-rst-order kinetics model has been applied to analyze the effect of different initial TCS concentrations on the degradation of TCS (Qiao et al., 2010).The kinetics equation was as follows Eq. ( 1): 1 Where was the initial concentration of TCS at time zero, was the initial concentration of TCS at time t, and k was the removal rate constant (d − 1 ).

The degradation of half-life (
) was determined using the algorithm.

2.4 Optimization of bacterial growth conditions by RSM
Response surface methodology (RSM) based on the Box-Behnken design was considered to combine the advantages of the simplex method and factorial design.The method greatly reduces the number of experiments, obtains the desired results, and generates useful information about critical variables (Devatha et al., 2019).To evaluate the impact of initial pH value, temperature, and inoculum size on the triclosan degradation performance of strain DL4, batch experiments were conducted with a sterilized CF medium, including the initial TCS of 5mg/L.Initial pH (6-8), culture temperature (25℃-35℃), and inoculum size (1%-5%) were the selected three principal variables in Table S1.The pretreatment experiments showed that the bacterium was di culty surviving under polar acid or alkali and extreme temperatures, so neutral pH and suitable temperature were selected for the experiments.Suspension samples from all batch experiments were extracted every 6 h to observe the change in TCS concentration until they remained stable.
ANN was regarded as an advanced and sophisticated tool for optimization, due to its robust prediction and evaluation capabilities (Ameer et al., 2017).In the present study, an ANN model was used to investigate the network architecture to achieve the TCS removal e ciency target model using Matlab R2021 software.At present, a feed-forward arti cial neural network strategy approach was used, where the input layer consisted of signi cant in uential independent variables, a hidden layer, and an outer layer on the response (Fig. 1).A successful ANN architecture and an appropriate network topology were essential for accurate prediction of target responses.In addition, the predictive potential of an arti cial neural network model can be in uenced by the number of hidden layer neurons (Reza et al., 2022;Gadekar et al., 2019).Many works followed a trial-and-error approach to xing the number of hidden layer neurons (Alemu et al., 2018).The number of hidden layer neurons ranged from 1 to 20 in most ANN-based research (Karri et al., 2018;Maghsoudi et al., 2015).Although increasing the number of hidden layer neurons can improve processing e ciency, it may be also causes over tting of the model.
Conversely, ANN with a small number of hidden layer neurons may be limited to approximately arbitrary accuracy and representation, and the learning ability is limited to approximately arbitrary accuracy due to a lack of su cient degrees of freedom (Kim et al., 2019).Therefore, different feedforward networks consisting of different hidden layer neurons were trained, and the network with the lowest mean square error (MSE) and the highest correlation coe cient (R 2 ) was selected (Table S2).Thus, 8 neurons were selected as the optimum number of neurons in the model topology.The feed-forward network topology of 3:8:1 was adopted in this work to model the removal e ciency of TCS.
Three input variables (initial pH, temperature, and inoculum size) were used by using the Levenberg-Marquardt (LM) backpropagation algorithm.And the output layer was the ability of TCS degradation.In Table 2, the data from 60 experimental runs were randomly divided into training (70%), validation (15%), and testing (15%) subgroups to apply to the model (Angeline et al., 2019).Non-speci ed weights were given to each single neuron connection between layers to set up the training procedure.The weights were modi ed until the minimum error between the actual and forecasted values of TCS removal e ciency was obtained.The obtained data via the ANN model were further veri ed using the validation process.Then, the simplicity of the ANN model used was evaluated using a testing process.With the validation and testing process completed, the ANN model was implemented to predict the effectiveness of triclosan removal.The performance of the trained network was evaluated by linear regression analysis of the experimental data against the predicted values.3 Where n was the number of points, Y predict was the predicted values of the models, Y actual was the experimental actual values, and the symbol " − " was the meaning of average values.

Analytical methods
The concentration of TCS was determined by high-performance liquid chromatography (HPLC, Agilent, USA) equipped with a UV-vis detector.The chromatographic separations were performed on a reversedphase Agilent Eclipse XDB-C18 column (5µm, 250mm×4.6mm).Methanol and ultrapure water were used as mobile phases with a ow gradient of up to 90% methanol-10% water at a wavelength of 220nm.The ow rate was 0.8mL/min and a volume injection of 20µL.All samples were thoroughly extracted with ethyl acetate and ltered by 0.45µm organic membrane before analysis by HPLC.
Furthermore, to identify the transformation products of TCS, the samples were analyzed using liquid chromatography/mass spectrometry (LC/MS, Agilent 6244, USA) equipped with an electrospray ionization source (ESI) operating in the negative-ion mode.A 2µL sample was injected into the source, and the mobile phase consisted of methanol and water (90:10 v/v) at 0.35mL/min.An Agilent Eclipse XDB-C18 column (2.4µm,150mm×2.1mm)was used with a column temperature of 45 ℃.The capillary voltage was 5.5kV, the temperature of 550 ℃, and the scan range of m/z = 100-2000.A multiple reaction monitoring (MRM) method was used to quantify TCS and its metabolites, and the data were acquired and processed using the MassLynx 4.1 software.
The strains' growth was monitored by measuring the optical density of culture bacterial suspensions at 600nm (OD600) using UV-vis spectrometry (UV-1240, Shimadzu, Japan).The pH was measured using a pH tester (S220, Mettler-Toledo, Switzerland).Ion chromatography detected chloride ion concentration (930 Compact IC Flex, Metrohm, Switzerland).All shake asks used in the experiment were wrapped in aluminum foil to exclude photodegradation.All experiments were conducted in triplicate.The degradation rate of TCS was calculated according to Eq. ( 6): Where and were the concentration of triclosan at time zero and t, respectively.

Isolation and identi cation of the strain DL4
Five separate colonies were selected on LB solid medium and transferred to CF liquid medium for 96 h to explore the potential of TCS degradation by these strains.An isolated strain DL4 was chosen for further study based on its highest triclosan-degradation capacity (Fig. S1).
Strain DL4 was a short, rod-shaped (0.5×1.7), gram-native bacterium.The cell morphology of strain DL4 in the CF medium was observed by scanning electron microscope (Fig. S2).Mono-colony was oval, yellowish-viscous, and opaque with a smooth surface.Fragment of 16S rDNA gene sequences (1491bp) of strain DL4 was ampli ed.Homology analysis was conducted by BLAST, and the results showed that the strain exhibited high similarity (up to 99%) to several strains of the genus Bacillus, such as Bacillus tropicus MCCC 1A01406 and Bacillus luti MCCC 1A00359, indicating strain DL4 belonged to Bacillus.Neighbor-joining phylogenetic analysis was performed using the 16S rDNA gene sequence of strain DL4 and its close relatives to form a phylogenetic tree (Fig. 2).Excellent performance of Bacillus sp. in the degradation of PPCPs (e.g., 4-nitroaniline, chlortetracycline, and Malathion pesticide) (

Effect of initial concentration of TCS on Bacillus sp. DL4
Strain DL4 could survive in CF-medium supplemented with triclosan, indicating that triclosan was used as the carbon source to support its growth.Therefore, the initial concentration of TCS was an important parameter.The experiment results showed that strain DL4 barely grew in the sample without TCS.When the concentrations were 5-15mg/L, the growth rate of the bacterium rose steeply and exponentially within 12 to 24 hours and reached their peak values after 96 h (Fig. S3).Among them, when the initial content of TCS was 10mg/L, strain DL4 achieved the fastest growth rate, and the peak value of OD600 (up to 0.92) was slightly higher than the others.However, as the initial concentration was increased up to and more signi cant than 20mg/L, the growth rate and cell biomass decreased gradually, and the maximum value of OD600 (less than 0.8) was signi cantly lower than that at relatively low concentrations.The highest biodegradation rate was up to 100% at 5mg/L, and 10mg/L initial TCS concentration (Fig. 3), 46.37% and 49.21% of TCS were removed in the rst 24 h, respectively.To be clear, when the concentration of TCS was 5mg/L, the biodegradation rate reached a constant rate after 72 h, and the speci c degradation rate was 0.069mg/L/h.The trends indicated that strain DL4 could utilize TCS as a substrate for growth and proliferation, and the growth rate gradually accelerated with substrate concentration.However, the proliferation rate was gradually inhibited, and the degradation rate was negatively correlated with the initial substrate concentration when the tolerance value of TCS was exceeded.This may be due to the inherent nature of TCS as a broad-spectrum bacterial inhibitor and therefore has a degree of toxic effect on bacteria.
The data were nonlinearly tted by Origin 9.1, and the results revealed that biodegradation of TCS in the exposure of different initial substrate concentrations by strain DL4 agreed well with the pseudo-rstorder equation.As shown in Table 1, the biodegradation rate constant (k) of each concentration of TCS calculated ranged from 0.0171 to 0.031h − 1 with a half-life (t 1/2 ) of 0.93 to 1.69 days, respectively.The correlation coe cient R 2 ranged from 0.9354 to 0.9775, which meant that biodegradation t well with the pseudo-rst-order kinetic model.The kinetic results revealed that microbial degradation effects of TCS were signi cantly higher when adding 5mg/L TCS to the media compared to that added other concentrations values because the half-life of TCS was only 0.93 days, which was far less than other experimental groups.Therefore, considering the degradation effect and economic cost, subsequent batch experiments were based on the initial TCS concentration of 5mg/L.

Optimization experiments of culture parameters
To study the signi cant impact of factors on TCS biodegradation, a three-factor three-level Box-Behnken design (BBD) was adopted.The experimental design matrix and the response of the dependent variable of TCS degradation were presented.The model predicted 20 trials of varied combinations (Table S3).These trials were experimentally performed, the nal TCS level was estimated by HPLC, and removal e ciency was reported accordingly.A quadratic polynomial function was mainly t to the experimental values, resulting in the following regression equation Eq. ( 7): 7 Where R was the TCS degradation rate (%) by the strain DL4, A, B, and C were the temperature, initial pH, and inoculum size, respectively.
SAS software package was used to conduct F-test and ANOVA analysis to evaluate the authenticity and goodness of t of the developed models for TCS degradation.The calculated model F was 21.51, indicating that the model was signi cant and the probability of the model F value due to noise was less than 0.1% ( The removal e ciency of TCS depended on all the variables considered.In each trial, the relationship between these factors was well established so that under all possible variations of the other factors, the optimum conditions were obtained and were re ected in the 3D representation as Fig. 4. The convexity of the response curve indicated that a better optimization condition exists after considering the factors.The interactions of different initial pH and temperatures, different inoculum sizes and temperature, different inoculum sizes, and initial pH on TCS biodegradation were shown in Fig. 4 (a), (b), and (c), respectively.The 3D graphs show a steeper response surface, indicating that the interactions between several variables were signi cant and had a greater effect on the TCS removal rate.The degradation e ciency increased from 70 ± 5.8% to 90 ± 2.5%, when the initial pH was enhanced from 5 to 7, and approximately 15% of the degradation e ciency was boosted when the inoculum size increased from 1-5%.Meanwhile, about a 10% reduction in TCS degradation occurred when the cultural temperature increased from 25℃ to 35℃.
Bacteria that can survive at different pH levels are more likely to successfully biodegrade in an environment with changing acid and alkaline conditions.In Fig. 4, the maximum reduction occurred at near-neutral pH, which may be attributed to its strong chemical hydrolysis.In addition, the higher TCS removal performances at 30℃ might be due to the increase in the activity of the enzymes relate to biodegradation.Inoculum size played a key role in TCS degradation.ANOVA results of the model demonstrated that temperature was a remarkable variant in TCS degradation (p < 0.0001).The increased removal e ciency with increasing inoculum may be due to the high density of the microbial cells.Similar observations were described by Li et al., (2013).

Arti cial neural network (ANN) model
To construct an optimal arti cial neural network topology, the RSM experimental sets and responses were used as targets in the model architecture and trained with the Levenberg-Marquardt algorithm to generate the ANN model.The results of the experimental and predicted values of the TCS removal (target response) in all network validation subsets were shown in Fig. 5.In all conditions, the data points were closely correlated with the regression line, indicating the precise and accurate predictive ability of the developed ANN model (Bayuo et al., 2020).A higher R 2 of 0.97257, 0.9897, and 0.98381 for training, validation, and testing of TCS biodegradation were observed, respectively.The higher R 2 proved that the developed ANN model had su cient accuracy and data prediction capability.In addition, another regression analysis of the results obtained in 20 experimental runs using RSM and ANN predictions also demonstrated the accurate predictive capability of the ANN model (Fig. 6).According to the results, it was inferred that all the selected input variables play key roles in the biodegradation of TCS and none of the variables could be ignored in this work.

Optimization and identi cation of the RSM and ANN model
The main purpose of the condition optimization study was to nd the optimal values of the independent parameters to maximize the bacterial degradation capacity.The independent variables were usually selected from the available options to obtain the preferred condition for a particular response.The preferred level and target response (TCS removal e ciency) of the independent operational variables were set to "within range" and "maximum", respectively.The optimal conditions for the biodegradation of TCS in wastewater by aerobic bacterium DL4 were listed in Table 3.The optimized conditions were further evaluated in triplicate experiments and were identi ed by comparing the actual values and predicted values by the model.95.89 ± 0.68% TCS reduction e ciency was obtained under the optimal conditions (initial pH = 7.31, temperature = 35 ℃, and inoculum size = 5%), while the predicted values of RSM (96.66%) and ANN (95.38%) were observed, respectively (Table 3).The experimental values obtained were within the 95% con dence interval.

Comparison analysis between the developed models
The RSM and ANN models were compared by analyzing the pattern of residual distributions of the two models to explain the accuracy and evaluative capability of their predictions.It could be seen from Fig. 6 that, compared with the RSM model, the ANN model had smaller changes in residuals and stronger stability.Further statistical evaluation of the performance of the RSM and ANN models based on R 2 , MSE, and AAD values.Larger R 2 values indicated a better model t, and conversely, smaller RMSE and AAD values demonstrated a superior performance (Igwegbe et al., 2019).In this study, both models can provide high-quality prediction with R 2 > 0.97 (Table 4).But ANN model had a signi cant advantage over the RSM model in terms of accuracy.Previous studies have also shown that the ANN model was more accurate in its predictive capability than the RSM model (Reza and Chen, 2021, Gadekar and Ahammed, 2019).Generally, speci c experimental design is not necessary for ANN models, but it is the most notable part of RSM models to guarantee robust prediction of target response and to illuminate the interactions between independent parameters.In addition, the ANN model is more adaptive, and new experimental data can be easily added to build accurate models.Therefore, the ANN model may be more reliable and provide a more plausible explanation for the removal of YCS from wastewater biodegraded by aerobic bacteria.The genome sequencing data of strain DL4 was assembled, and the general properties of the genome were shown in Table S5.Gene numbers were multiply annotated in the database: NR (5584), COG (3699), GO (3861), KEGG (2019), Pfam (4343), and Swiss-Prot (4055).As shown in Fig. S4, the COG database was annotated with 20 speci c functional genes and contents.There were 2963 genes with GO annotation function in strain DL4, accounting for 51.6% of the total number of genes, among which 18.37% were CC, 39.16% were MF, and 42.47% were BP (Fig. S5).Further analysis found that 18 genes were annotated to the cell-killing term, 485 genes to the response to stimulus term, and 43 genes to the locomotion term.
The KEGG database is an extensive knowledge base for systematically analyzing gene functions and linking genomic and functional information.Genome annotation through the KEGG database helps to understand the biological function of genes systematically.The BLAST algorithm was used to compare the predicted genes with the KEGG database.According to the ko numbers obtained, by comparison, the speci c biological pathways involved in the corresponding genes were obtained, and a total of 3491 genes were annotated to 216 pathways (Fig. S6).There were a large number of genes annotated in metabolic classi cation, which were typical of amino acid metabolism (444), metabolism of cofactors and vitamins (225), overview (429), xenobiotics biodegradation and metabolism (168), and carbohydrate metabolism (457) in Table 5.In addition, signal transduction (265) and membrane transport (223) in environmental information processing classi cation also played essential roles in gene annotation.

Identi cation of TCS transformation products
The analysis of the 24th and 72nd bacterial culture supernatant of strain DL4 grown in the CF medium with 5mg/L triclosan by LC-MS revealed the presence of three compounds.Mass spectra and fragmentation patterns identi ed the intermediate products generated during the experiments.The possible structure of the intermediate compounds and their corresponding m/z were shown (Fig. S7).Three major products (2,4,6-trichlorophenol, 2,4-dichlorophenol, and hydroquinone) were identi ed from the reaction of TCS with strain DL4.Based on identi ed metabolites, a biodegradation pathway of triclosan was proposed (Fig. 7).The parent compound was formed by hydroxylation into mono-hydroxyltriclosan or dihydroxyl-triclosan.Many previous studies have shown that mono-or dioxygenase owned a broad substrate speci city and could degrade a variety of organic compounds (Vainberg et al., 2006;Robrock et al., 2011).Introducing hydroxyl radicals into benzene rings may be the rst step in microbial degradation of contaminations (Parales & Resnick, 2006).The biomass increase demonstrated that triclosan as a substrate could induce strain DL4 to secrete the mono-or dioxygenase.As the biochemical reaction proceeded, monohydroxy-triclosan and dihydroxy-triclosan underwent the breaking of carbon-oxygen bond (C-O) to form 3,5-dichlorocatechol, chlorohydroquinon and hydroquinone (pathway1).In addition, monohydroxy-triclosan and dihydroxy-triclosan might be further linked by 2, 3dioxygenase and then underwent ether bond cleavage to produce 2, 4-dichlorophenol (Pfeifer et al., 1989;Wang et al., 2018).2,4-dichlorophenol could be converted to 2-chlorohydroquinone through the reaction of dichlorination, further forming hydroquinone (pathway 2).The degree of chlorination could be determined by the content of chlorine released during TCS degradation because TCS is a chlorinated organic compound.At the end of the triclosan degradation experiments (about 96 h), the nal concentration of chlorine ions was 1.79mg/L (Data not shown).Based on stoichiometry, the theoretical value of chloride ion was 1.84mg/L, which was approximately equal to the actual release value, indicating that strain DL4 was completely dechlorinated.

Conclusion
An isolated strain of Bacillus sp.DL4 could rapidly remove about 95% TCS (5mg/L).RSM and ANN models were utilized to assess the signi cant parameters affecting TCS biodegradation.Furthermore, the authenticity and signi cance of the models exhibited a signi cantly robust positive correlation between the actual and predicted values.The TCS-biodegradation by strain DL4 complied with pseudorst-order kinetics.Genomic analysis revealed the genes associated with hydroxylation, ether bond cleavage, and dichlorination in strain DL4.Functional annotations revealed functional gene terms and pathways that promoted strain DL4 biodegradation of TCS.TCS metabolic mechanism was also proposed tentatively.This was the rst report that Bacillus sp.became a functional bacterial genus to degrade triclosan.

Declarations Author
All authors contributed to the study conception and design.Material preparation, data collection and analysis were performed by Xuejie Li, Xiaomin Hu, Xin Zhao, Fan Wang, Yan Zhao.The rst draft of the manuscript was written by Xuejie Li, and all authors commented on previous versions of the manuscript.All authors read and approved the nal manuscript.

Figure 1 Architecture
Figure 1

Table 1
Kinetic parameters of TCS in the presence of different initial TCS concentrations

Table 3
Predicted and experimental values under optimal conditions for model validation