Effects of Broad-Spectrum Antibiotic (Florfenicol) on Resistance Genes and Bacterial Community Structure of Water and Sediments in an Aquatic Microcosm Model

This study evaluates the effects of a broad-spectrum antibiotic (florfenicol) on antibiotic resistance genes (ARGs) and bacterial community structure in aquatic environments. We constructed an indoor aquatic microcosm model, adding different concentrations of florfenicol (0.1, 1, 10, 100 mg L−1), and water and sediment samples were collected after 0, 7, 30, and 60 days. qPCR and 16S rDNA amplicon sequencing were used to study the changes in the ARGs and bacterial community structure of the collected samples. The results show that the inclusion of florfenicol resulted in an increased abundance of the floR and optrA genes. Adding 100 mg L−1 florfenicol to the water increased the abundance of optrA gene copies with the maximum on the Day 7, and increased the abundance of floR gene copies with the maximum on Day 30. Adding 100 mg L−1 florfenicol to the sediment increased the abundance of floR and optrA genes by one order of magnitude on Day 60. Meanwhile, the average number of operational taxonomic units (OTUs) in the water samples was 257, and the average number of OTUs in sediment samples was 823. The bacterial community diversity and richness in sediments were higher than those in water. The difference between the maximal and minimal values of the Shannon diversity index in the water and sediment samples was 4.36 and 1.95, respectively. The effect of florfenicol on the bacterial community structure in water was much higher than that in sediment. At 30 days, the diversity index and richness index of the florfenicol treatment groups with 1 and 10 mg L−1 concentrations began to increase; at 60 days, the diversity and richness indices of the 100 mg L−1 florfenicol treatment group began to increase. The samples at the same sampling time in the sediments clustered closer together. The results of this study provide a scientific basis for guiding the rational use of florfenicol in aquaculture, maintaining a healthy and stable microecological environment in aquaculture, and provide theoretical data for environmental ecological risk assessment and safety management caused by microbial resistance under the abuse of florfenicol.


Introduction
Antibiotics have played a vital role in human medical care and animal production in the past few decades. China is a large country in animal husbandry, and most of the produced antibiotics are used on animals [1][2][3]. According to the FAO report in 2020, the scale of global aquaculture is increasing, and China's aquaculture industry is also at a stage of rapid development [4,5]. The use of antibiotics in intensive farming is currently universal. Florfenicol is one of the antibiotics approved by FAO for use in aquaculture

Abundance of Florfenicol ARGs
As shown in Figure 1, the abundance of the floR gene in the water body ranged from 3.24 × 10 5 to 1.33 × 10 9 copies·µL −1 . On Day 7, after the addition of florfenicol, the abundance of the floR gene was one order of magnitude higher than that of the control group, and 0.1 mg·L −1 of florfenicol could significantly increase the abundance of the floR gene. The abundance of the floR gene decreased with time. The abundance of the floR gene in sediments ranged from 1.63 × 10 5 to 9.86 × 10 6 copies·µL −1 . The abundance of the floR gene was one order of magnitude higher on Day 60 after the addition of 10 and 100 mg·L −1 florfenicol. The abundance of the optrA gene in water ranged from 3.34 × 10 4 to 1.62 × 10 7 copies·µL −1 . On Day 7, after 10 and 100 mg·L −1 florfenicol supplementation, the Antibiotics 2022, 11, 1299 3 of 14 abundance of the optrA gene increased by 1-2 orders of magnitude. With the change in time, the abundance of the optrA gene showed a downward trend, and the abundance of the optrA gene in the samples added with florfenicol at 60 days was basically the same as that in the control group. The abundance of the optrA gene in the sediments ranged from 4.24 × 10 4 to 3.81 × 10 5 copies·µL −1 , and the abundance of the optrA gene increased by one order of magnitude on Days 30 and 60 after the addition of 100 mg·L −1 fluorobenzene. According to the statistical analysis results, concentrations of florfenicol and time significantly affected the abundance of ARGs in an aquatic microcosm model (p < 0.001).
The abundance of the floR gene decreased with time. The abundance of the floR sediments ranged from 1.63 × 10 5 to 9.86 × 10 6 copies·μL −1 . The abundance of the fl was one order of magnitude higher on Day 60 after the addition of 10 and 100 florfenicol. The abundance of the optrA gene in water ranged from 3.34 × 10 4 to 1. copies·μL −1 . On Day 7, after 10 and 100 mg·L −1 florfenicol supplementation, the abu of the optrA gene increased by 1-2 orders of magnitude. With the change in ti abundance of the optrA gene showed a downward trend, and the abundance of th gene in the samples added with florfenicol at 60 days was basically the same as tha control group. The abundance of the optrA gene in the sediments ranged from 4. to 3.81 × 10 5 copies·μL −1 , and the abundance of the optrA gene increased by one o magnitude on Days 30 and 60 after the addition of 100 mg·L −1 fluorobenzene. Ac to the statistical analysis results, concentrations of florfenicol and time significa fected the abundance of ARGs in an aquatic microcosm model (p < 0.001).

Analysis of OTUs
After the quality optimization of the original data, the number of valid seq obtained in water samples was 31,046,103,252, and the number of valid sequence sediment samples was 31,570-74,821. After OTU cluster analysis, a total of 580 OTU obtained from the water samples, with an average of 257 OTUs per sample, of wh W4D7 group had the fewest OTUs, 91; a total of 1235 OTUs were obtained from t ments, and the average number of OTUs per sample was 823, of which the S4D6 had the fewest OTUs, 590. According to the out cluster analysis results, a Venn d was drawn, and there were 7 OTUs in water samples and 224 OTUs in sediment s As shown in Figure 2, the number of common OTUs was the smallest when the tration of florfenicol in the water samples was 100 mg·L −1 (Figure 2a), and the nu common OTUs in the sediment samples was the smallest when the concentra florfenicol was added at 1 mg·L −1 (Figure 2b). The 30-day water samples shared the OTUs (Figure 2c), and the 60-day sediment samples shared the fewest OTUs (Figu Sequences were randomly sampled, and a dilution curve was constructed w number of valid sequences drawn from the sample (sequences per sample) as the a and the number of (observed) OTUs as the ordinate. As presented in Figure 3, as th ber of extracted sequences increased, the number of detected OTUs increased and ally became flat, indicating that the sequencing tended to be saturated, the sampl basically reasonable, and that the sequencing results contained the most microbial

Analysis of OTUs
After the quality optimization of the original data, the number of valid sequences obtained in water samples was 31,046,103,252, and the number of valid sequences in the sediment samples was 31,570-74,821. After OTU cluster analysis, a total of 580 OTUs were obtained from the water samples, with an average of 257 OTUs per sample, of which the W4D7 group had the fewest OTUs, 91; a total of 1235 OTUs were obtained from the sediments, and the average number of OTUs per sample was 823, of which the S4D60 group had the fewest OTUs, 590. According to the out cluster analysis results, a Venn diagram was drawn, and there were 7 OTUs in water samples and 224 OTUs in sediment samples. As shown in Figure 2, the number of common OTUs was the smallest when the concentration of florfenicol in the water samples was 100 mg·L −1 (Figure 2a), and the number of common OTUs in the sediment samples was the smallest when the concentration of florfenicol was added at 1 mg·L −1 (Figure 2b). The 30-day water samples shared the fewest OTUs (Figure 2c), and the 60-day sediment samples shared the fewest OTUs (Figure 2d).
Sequences were randomly sampled, and a dilution curve was constructed with the number of valid sequences drawn from the sample (sequences per sample) as the abscissa and the number of (observed) OTUs as the ordinate. As presented in Figure 3, as the number of extracted sequences increased, the number of detected OTUs increased and gradually became flat, indicating that the sequencing tended to be saturated, the sampling was basically reasonable, and that the sequencing results contained the most microbial groups. This can more realistically reflect the microbial community structure in the sample through the addition of 100 mg·L −1 fluorobenzene. This can more realistically reflect the microbial community structure in the sample through the addition of 100 mg·L −1 fluorobenzene.

Alpha Diversity Index Analysis
The community richness index mainly includes the Chao1 and ACE indices, and the bacterial community diversity index mainly includes the Shannon and Simpson indices As shown in Table 1, with the increase in drug concentration, the species richness and diversity of the water samples showed a downward trend; with the change in time, the community richness and diversity were first suppressed and then gradually recovered The diversity index of the W2 and W3 groups increased on Day 30, while the diversity index of the W4 group began to increase on Day 60, and the minimal value of the diversity index appeared in the W4 group. As shown in Table 2, among the sediment samples, only the S4 and D60 groups exhibited the inhibitory effect of florfenicol on the bacterial species This can more realistically reflect the microbial community structure in the sample through the addition of 100 mg·L −1 fluorobenzene.

Alpha Diversity Index Analysis
The community richness index mainly includes the Chao1 and ACE indices, and the bacterial community diversity index mainly includes the Shannon and Simpson indices. As shown in Table 1, with the increase in drug concentration, the species richness and diversity of the water samples showed a downward trend; with the change in time, the community richness and diversity were first suppressed and then gradually recovered. The diversity index of the W2 and W3 groups increased on Day 30, while the diversity index of the W4 group began to increase on Day 60, and the minimal value of the diversity index appeared in the W4 group. As shown in Table 2, among the sediment samples, only the S4 and D60 groups exhibited the inhibitory effect of florfenicol on the bacterial species

Alpha Diversity Index Analysis
The community richness index mainly includes the Chao1 and ACE indices, and the bacterial community diversity index mainly includes the Shannon and Simpson indices. As shown in Table 1, with the increase in drug concentration, the species richness and diversity of the water samples showed a downward trend; with the change in time, the community richness and diversity were first suppressed and then gradually recovered. The diversity index of the W2 and W3 groups increased on Day 30, while the diversity index of the W4 group began to increase on Day 60, and the minimal value of the diversity index appeared in the W4 group. As shown in Table 2, among the sediment samples, only the S4 and D60 groups exhibited the inhibitory effect of florfenicol on the bacterial species richness and diversity, and the smallest value of the diversity index appeared in the 60-day sediment samples when the concentration of florfenicol was added at 100 mg·L −1 .

Bacterial Community Structure Analysis
Species relative abundance analysis was performed at the phylum level, and a histogram of species information is indicated in Figure 4. A total of 24 gates were obtained for the water bodies, and 40 gates for sediments. The dominant bacteria in the water were Proteobacteria, Actinobacteriota and Bacteroidota, and the dominant bacteria in the sediment were Proteobacteria, Bacteroidota, Desulfobacteria (Desulfobacterota), and Acidobacteriota, Myxococcota. Proteobacteria were the dominant bacteria in all samples, with relative abundances ranging from 39% to 92% in water, and 33% to 57% in sediments. Compared with the control group, the relative abundance of Proteobacteria increased significantly, while the relative abundance of Actinobacteria and Bacteroidetes decreased significantly in the water samples. Significant changes in the microbial community structure of the sediments appeared on Days 30 and 60. The relative abundance of Proteobacteria and Myxococcota increased, and the relative abundance of Bacteroidota, Desulfobacterota, Acidobacteriota, and Chloroflexi decreased. The relative abundance of Firmicutes in the water samples of the control group was 0, while the relative abundance in the W4D7 and W4D30 samples was 6.1% and 8.9%, respectively.
At the genus level, a total of 257 genera were obtained from the water samples, with the number of each group ranging from 47 to 181; a total of 435 genera were obtained from the sediments, with the number of each group ranging from 270 to 367. The top 30 genuslevel species were selected to draw a histogram. As shown in Figure 5, the main dominant bacterial genera in the water without florfenicol were hgcI-clade, CL500-29-marine-group, PeM15, Bacteroidota, and Kapabacteriales. On Day 7, florfenicol had a significant effect on the bacterial community structure in the water. With the increase in florfenicol concentration, the relative abundances of Vogesella, Hydrogenophaga, Comamonadaceae, and Cellvibrio in Proteobacteria were decreased. The dominant bacterial genera in the sediments were Dechloromonas, Geothrix, Anaeromyxobacter, Bacteroidetes-vadinHA17, Geobacteraceae, Steroidobacteraceae, Arenimonas, and Ellin6067. Compared with water, the bacterial diversity in the sediments was higher, the community structure was more complex and stable, and it was stronger in resisting antibiotic stress.  At the genus level, a total of 257 genera were obtained from the water samples, with the number of each group ranging from 47 to 181; a total of 435 genera were obtained from the sediments, with the number of each group ranging from 270 to 367. The top 30 genus-level species were selected to draw a histogram. As shown in Figure 5, the main dominant bacterial genera in the water without florfenicol were hgcI-clade, CL500-29-marinegroup, PeM15, Bacteroidota, and Kapabacteriales. On Day 7, florfenicol had a significant effect on the bacterial community structure in the water. With the increase in florfenicol concentration, the relative abundances of Vogesella, Hydrogenophaga, Comamonadaceae, and Cellvibrio in Proteobacteria were decreased. The dominant bacterial genera in the sediments were Dechloromonas, Geothrix, Anaeromyxobacter, Bacteroidetes-vadinHA17, Geobacteraceae, Steroidobacteraceae, Arenimonas, and Ellin6067. Compared with water, the bacterial diversity in the sediments was higher, the community structure was more complex and stable, and it was stronger in resisting antibiotic stress.

Multisample Comparative Analysis
The nonmetric multidimensional scaling method (NMDS analysis) and unweighted pair group method with arithmetic mean (UPGMA) clustering were used to express the degree of similarity and difference between different samples. The results are shown in Figure 6. When stress is less than 0.2, NMDS analysis can accurately reflect the degree of difference between samples; water samples with similar microbial community structure were clustered together in cluster analysis. W4D7 and W4D30W, and W3D30 and W3D60 were clustered together, and NMDS analysis was farther away from other treatment groups. S2D60, S3D60, and S4D60 in the sediment samples were significantly separated from other treatment groups, cluster analysis shows obvious time differences, and the treatment groups with different sampling times had farther clustering distances regarding antibiotic stress.

Multisample Comparative Analysis
The nonmetric multidimensional scaling method (NMDS analysis) and unweighted pair group method with arithmetic mean (UPGMA) clustering were used to express the degree of similarity and difference between different samples. The results are shown in Figure 6. When stress is less than 0.2, NMDS analysis can accurately reflect the degree of difference between samples; water samples with similar microbial community structure were clustered together in cluster analysis. W4D7 and W4D30W, and W3D30 and W3D60 were clustered together, and NMDS analysis was farther away from other treatment groups. S2D60, S3D60, and S4D60 in the sediment samples were significantly separated from other treatment groups, cluster analysis shows obvious time differences, and the treatment groups with different sampling times had farther clustering distances regarding antibiotic stress.

Multisample Comparative Analysis
The nonmetric multidimensional scaling method (NMDS analysis) and u pair group method with arithmetic mean (UPGMA) clustering were used to degree of similarity and difference between different samples. The results ar Figure 6. When stress is less than 0.2, NMDS analysis can accurately reflect th difference between samples; water samples with similar microbial communit were clustered together in cluster analysis. W4D7 and W4D30W, and W3D30 a were clustered together, and NMDS analysis was farther away from other groups. S2D60, S3D60, and S4D60 in the sediment samples were significantly from other treatment groups, cluster analysis shows obvious time differenc treatment groups with different sampling times had farther clustering distance antibiotic stress.

Correlation Analysis between Bacterial Communities and ARGs
Taking ARGs as environmental factors, combined with the microbial community structure of each sample, bacteria with higher relative abundance were selected as samples, and redundancy analysis (RAD) was used to study the correlation between bacterial community structure and ARGs. The results of redundancy analysis of water samples are shown in Figure 7a. Spindles 1 and 2 explain 73.72% and 26.28% of the bacterial community structure and ARGs parameters, respectively. Among them, genera Comamonas, Acinetobacter, Chryseobacterium, Pesudomonas, and other genera in the bacterial community were positively correlated with the optrA resistance gene, and TC1, Rivicola, Acinetobacter, Chryseobacterium, Pseudomonas, Alcaligenaceae, and other genera were positively correlated with the floR resistance gene. The results of the redundancy analysis of sediment samples are shown in Figure 7b. Spindles 1 and 2 explain 80.62% and 19.38% of the bacterial community structure and ARGs parameters, respectively. In the bacterial community, genera Roseomonas, Rhodobacter, Flavihumibacter, Acetoanaerobium, Macellibacteroides were positively correlated with the optrA resistance gene, and genera such as Hydrogenophaga, Kaistia, Brevundimonas, Methylomonas, Nitrospira, and Diaphorobacter were positively correlated with the floR resistance gene.

Correlation Analysis between Bacterial Communities and ARGs
Taking ARGs as environmental factors, combined with the microbial commu structure of each sample, bacteria with higher relative abundance were selected as s ples, and redundancy analysis (RAD) was used to study the correlation between bacte community structure and ARGs. The results of redundancy analysis of water samples shown in Figure 7a. Spindles 1 and 2 explain 73.72% and 26.28% of the bacterial com nity structure and ARGs parameters, respectively. Among them, genera Comamonas, netobacter, Chryseobacterium, Pesudomonas, and other genera in the bacterial commu were positively correlated with the optrA resistance gene, and TC1, Rivicola, Acinetoba Chryseobacterium, Pseudomonas, Alcaligenaceae, and other genera were positively correla with the floR resistance gene. The results of the redundancy analysis of sediment sam are shown in Figure 7b. Spindles 1 and 2 explain 80.62% and 19.38% of the bacterial c munity structure and ARGs parameters, respectively. In the bacterial community, gen Roseomonas, Rhodobacter, Flavihumibacter, Acetoanaerobium, Macellibacteroides were p tively correlated with the optrA resistance gene, and genera such as Hydrogenophaga, stia, Brevundimonas, Methylomonas, Nitrospira, and Diaphorobacter were positively co lated with the floR resistance gene.

Discussion
Under antibiotic pressure selection, bacteria in aquacultural environments deve drug resistance and ARGs, thereby destroying the environmental microecological bala [23]. Aquatic environments are vulnerable to contamination through the use of antibio in aquacultural wastewater, and the increased selective pressure caused by the exten use of florfenicol in aquaculture has accelerated the development and spread of bacte resistance to this antibiotic, resulting in potential public health risks [24,25]. Among reported resistance genes of florfenicol, the floR gene mainly mediates the resistanc Gram-negative bacteria to florfenicol [26]. The optrA gene is involved in bacterial sistance to amide alcohols and oxazolidinones; oxazolidinones (e.g., linezolid and t zolid) are currently only approved for human infection by superbugs such as methici resistant Staphylococcus aureus and vancomycin-resistant Enterococcus, so they are the " line of defense" against pathogenic bacteria [27]. In this experiment, the change florfenicol resistance genes floR and optrA were determined by qPCR. The results sh that the addition of florfenicol had a greater impact on the abundance of ARGs in wa and 0.1 mg·L −1 florfenicol could increase floR gene abundance by 1-3 orders of magnitu

Discussion
Under antibiotic pressure selection, bacteria in aquacultural environments develop drug resistance and ARGs, thereby destroying the environmental microecological balance [23]. Aquatic environments are vulnerable to contamination through the use of antibiotics in aquacultural wastewater, and the increased selective pressure caused by the extensive use of florfenicol in aquaculture has accelerated the development and spread of bacterial resistance to this antibiotic, resulting in potential public health risks [24,25]. Among the reported resistance genes of florfenicol, the floR gene mainly mediates the resistance of Gram-negative bacteria to florfenicol [26]. The optrA gene is involved in bacterial resistance to amide alcohols and oxazolidinones; oxazolidinones (e.g., linezolid and tedizolid) are currently only approved for human infection by superbugs such as methicillin-resistant Staphylococcus aureus and vancomycin-resistant Enterococcus, so they are the "last line of defense" against pathogenic bacteria [27]. In this experiment, the changes in florfenicol resistance genes floR and optrA were determined by qPCR. The results show that the addition of florfenicol had a greater impact on the abundance of ARGs in water, and 0.1 mg·L −1 florfenicol could increase floR gene abundance by 1-3 orders of magnitude, while 10 and 100 mg·L −1 florfenicol increased optrA gene abundance by 1-2 orders of magnitude. With the deposition at higher florfenicol concentration and longer Antibiotics 2022, 11, 1299 9 of 14 time, the abundance of ARGs began to increase. In addition, the abundance of both ARGs in the water body decreased with time, and the abundance of the optrA drug resistance gene recovered to the control level at 60 days. A previous study showed that the floR gene is found on both chromosomes and plasmids, and plasmids carrying the floR gene can be spread between bacteria [28]. Miranda et al. [26] also isolated a large number of florfenib-resistant strains from two commercial scallop hatcheries in Chile carrying the floR gene. Furthermore, Fan et al. [29], and Torres et al. [30] found that the optrA resistance gene exists in both Staphylococcus and Enterococcus. Wang et al. [31] observed that the addition of florfenicol to soil significantly affected the four resistant genotypes of aminoglycosides, β-lactamases, tetracyclines, and trimethoprim, while the resistance to the amide alcohol genotype had no significant effect. However, other relevant findings suggested that the abundance of resistance genes did not decrease after antibiotics had been discontinued in aquaculture, but remained at a high level [32]. The inconsistencies with the present study may have been caused by the environmental differences and different host species of ARGs. They may also be related to the amount and period of used antibiotics [33].
Water environmental safety is closely associated with the bacterial community structure; when the water environment changes, the bacterial community structure is also altered [34]. The effects of antibiotics on microbial communities are mainly community diversity, abundance, composition, and function [35]. According to a previous report, the impact of antibiotics on the bacterial community structure can be roughly divided into three stages: the first stage has no obvious effect on bacterial abundance, but affects its physiological activity; the second stage has a significant inhibitory effect on bacterial abundance, and the bacterial community difference increases; in the third stage, the bacterial community structure basically returns to the control level [36]. The in vitro minimal inhibitory concentration (MIC) of florfenicol against common pathogens is 0.3-1.6 µg·mL −1 . In the present study, 0.1 mg·L −1 florfenicol in the water environment did not affect the diversity and abundance of bacterial communities, but promoted the growth of some specific bacteria; 100 mg·L −1 florfenicol had a dramatic effect on the bacterial community structure, significantly reducing the diversity and abundance of the bacterial communities. The effect of florfenicol on bacterial diversity gradually decreased with time. When the addition of florfenicol was 100 mg·L −1 , the bacterial community structure in the sediment was disturbed, and the community diversity and richness decreased. The relative abundance of Acinetobacter strains dominated after supplementing 100 mg·L −1 florfenicol, including Acinetobacter variabilis, Acinetobacter soli, and Acinetobacter baylyi. All three Acinetobacter species are resistant to carbapenems [37][38][39]. The study of Yong-Ji et al. [40] showed that florfenicol significantly affects the environmental bacterial community structure, leading to a complete subversion of the microbial community. Liu et al. [41] suggested that florfenicol is mainly degraded and eliminated in the form of a prototype in an aquacultural environment, most of the drugs persist in the water until degraded, and the content of the drugs in the sediment is less. This is consistent with the changing trend of the results in our experiment. In addition, some studies indicated that sediments have a strong adsorption capacity for certain antibiotics. With changes in the external environment, antibiotics adsorbed in sediments may be resolved and released into waters [42], which can also explain our findings.
ARGs affect the diversity index of soil bacterial communities [43]. Meanwhile, microbes serve as hosts for ARGs, and changes in their community structure are related to the expression of ARGs [44]. The correlation analysis of this study shows that Acinetobacter, Chryseobacterium, and Pseudomonas were positively correlated with the optrA and floR genes in water; these three bacteria are common opportunistic pathogens and may also be the potential host bacteria of florfenicol resistance genes. Acinetobacter has the ability to acquire OXA-type carbapenemases and metallo-β-lactamases, and develop rapid resistance to new antibiotics; therefore, it can become the dominant genus under antibiotic stress [45]. Chryseobacterium is widespread in water, soil, and hospital environments. Chryseobacterium is resistant to a variety of antibiotics, such as lactams, aminoglycosides, and quinolones [46][47][48]. Fu et al. [49] isolated multidrug-resistant Chryseobacterium sp. POL2 from livestock and poultry farming wastewater. The integrated engagement element of this strain can confer new drug resistance and pathogenicity to the host bacteria, allowing for the host bacteria to rapidly adapt to changing ecological niches, increasing the risk of transmission and infection. Pseudomonas has a variety of environmental adaptation mechanisms and is widely present in the environment [50]. In both veterinary and human medicine, resistance to commonly used antipseudomonal antibiotics such as carbapenems, cephalosporins, fluoroquinolones, and aminoglycosides was reported [51][52][53]. Pseudomonas has multiple drug-resistance mechanisms, including target gene mutations, changes in the expression of genes encoded by the efflux pump system, the production of inactivating enzymes, and the formation of biofilms [54,55]. The selection pressure of antibiotics makes these potential pathogenic bacteria acquire resistance genes and become dominant genera, accelerating the spread of ARGs and the risk of disease outbreaks. Hence, avoiding the misuse of broad-spectrum antibiotics such as florfenicol in aquaculture is imperative.

Reagents
Lake surface water (1-10 cm) and lake surface sediments (1-10 cm) in Guangdong Ocean University (Zhanjiang, China) were collected, and the debris was removed to construct an aquatic microcosm. Florfenicol (13021322)

Experimental Design and Sampling
The indoor aquatic microcosm model was constructed to simulate the aquatic ecosystem. The collected lake water and sediment were picked out and packed in a transparent 50 × 40 × 30 cm plastic box. A total of 15 aquatic microcosm models were constructed. The sediment thickness of the group was about 10 cm, and the water body was about 30 L. After the aquatic microcosm system had basically stabilized, a florfenicol solution was added to the water body to concentrations of 0, 0.1, 1, 10 and 100 mg·L −1 (the numbers are water samples W0-W4 and sediment samples S0-S4), placed at room temperature of 25 ± 3 • C, and the water samples were collected at four points at 0, 7, 30, and 60 days (D0, D7, D30, D60) at a distance of about 10 cm from the water surface (W group) and surface sediment (1-5 cm) samples (group S); the sediment samples were mixed and centrifuged to remove the supernatant. The experiment was divided into 5 groups, each with 3 replications.

Detection of ARGs
The SYBR-primer method was used to determine the abundance of ARGs on the CFX Connect Real-Time System instrument (BIO-RAD). Primer-related information is shown in Table 3 and was based on the previous studies [56,57]. The purified product after PCR amplification was ligated into a 19-T vector, and the ligated product was transformed into competent E. coli. Positive clones were selected, plasmids were extracted from them, and the plasmid concentration was determined using a NanoDrop microspectrophotometer. The known plasmid concentration was diluted with fivefold gradient concentration, the logarithmic value of copy number was used as the abscissa, and the C T value was used as the ordinate to draw a standard curve. Three repetitions were performed for each group, and the average was taken.
DNA copy number (copies·µL −1 ) = [C × 10 −9 × 6.02 × 10 23 where C is the plasmid concentration (ng·µL −1 ), and L is the number of bases of the cloned product (bp). The qPCR reaction system was: 10 µL SYBR, 1 µL DNA template, 1 µL upstream and downstream primers (10 mol·L −1 ) each, and 8 µL ddH 2 O. The qPCR reaction program was: 94 • C, 5 min; 94 • C for 30 s, annealing for 30 s (annealing temperature is shown in the table), 72 • C for 30 s, 39 cycles; fluorescence was collected after each cycle. From 60 to 95 • C, the fluorescence was collected every 0.5 • C to generate the melting curve to test the specificity of the amplification results. Table 3. Primer information of antibiotic resistance genes (ARGs) for qPCR.

16S rDNA Sequencing
The extraction of the total microbial genome DNA from water and sediments was according to the instructions of the commercial kits. The concentration and purity of the extracted DNA were determined with a nanodrop UV-vis spectrophotometer. The DNA samples extracted from each group of the 3 replicates were mixed into one sample and sent to Suzhou Jinweizhi Biotechnology Co., Ltd. (Suzhou, China) for the Illumina MiSeq paired-end sequencing of the V3-V4 amplicon regions of 16S rDNA. The sequencing details were based on a former report [58].

Statistical
Analysis QIIME 1.9.1 was used to optimize the high-throughput sequencing raw data: splicing the overlapping regions at the ends of the sequences, removing sequences with a length of less than 200 bp, and removing the chimeric sequences to obtain valid data. Operational taxonomic unit (OTU) cluster analysis was performed according to 97% similarity. Species taxonomic annotation was performed using the Silva 138 16S rDNA database. Excel 2016 was used to analyze the abundance of ARGs, QIIME 1.9.1 software was used for α diversity index analysis, R language was used for β diversity analysis, and CANOCO 5.0 software was used to analyze the correlation between ARGs and the community structure. GLMs followed by multiple comparisons using Duncan's test were applied to test the effects of concentrations of florfenicol, time, and their interactions on the abundance of ARGs. Statistical analyses were performed with SAS 9.2 (SAS Institute Inc., Cary, NC, USA).

Conclusions
In this study, the addition of florfenicol increased the abundance of the floR and optrA genes by orders of magnitude. The bacterial community diversity and richness in sediment were higher than those in water. The addition of florfenicol reduced the diversity and richness of the bacterial community, and the effect on the bacterial community structure in water was much higher than that in the sediment. The richness and diversity of the bacterial communities gradually recovered over time. High concentrations of florfenicol break the original community structure and significantly increase the relative abundance of drug-resistant bacteria. Therefore, to reduce the risks of environmental safety and public health that are posed by florfenicol, the standardized use of broad-spectrum antibiotics such as florfenicol in aquaculture, and closely monitoring and unraveling alternatives to antibiotics are necessary.