**3. Results**

#### *3.1. Distribution, Classification and Abundance of ARGs in Different Environments Associated with C. semilaevis Farming Industry*

The results of the high-throughput PCR indicated that a variety of ARGs could be found in different kinds of samples (Figure S1 in Supplementary Materials). The 243 detected ARGs included genes of eight distinct and predominant classes, conferring resistance to antibiotics such as aminoglycosides, multidrug, beta-lactamases, macrolide lincosamide streptomycin B (MLSB), chloramphenicol, sulfonamide, tetracycline and vancomycin, with a few of them related to other antibiotics (Figure S1 in Supplementary Materials). The proportion of the eight classes in each sample depended on both the sample type and location.

In *C. semilaevis* samples, ARGs were mostly abundant in A2 (P2), followed by A1 (P1) and A3 (P3). The lowest number of ARGs was found in A0 (P0). The copy number of almost all of the ARGs was higher in the A2 sample too (Figure S1A in Supplementary Materials). In A2, ARGs related to aminoglycosides (26 genes), beta-lactamases (26 genes), MLSB (23 genes), multidrug (20 genes) and tetracycline (14 genes) were the most abundant, among which the aph3ia (aminoglycoside), tetk (Tetracycline), ermK (MLSB) and bl3\_cpha (beta-lactamase) genes were the ARGs with the highest copy numbers (Figure S1A in Supplementary Materials). ARGs responsible for resistance to chloramphenicol could not be detected in A1–A3, while ARGs related to aminoglycosides and tetracycline resistance were absent in A0. In general, in type A samples, the abundant ARGs were multidrug, MLSB, beta-lactamases, aminoglycoside and tetracycline-related ones.

In sea water samples, tetracycline, aminoglycosides and multidrug-related ARGs were the most abundant in B1 (P1), while tetracycline, aminoglycosides and MLSB in B2 (P2) (Figure S1B in Supplementary Materials). In B3, ARGs involved in the resistance to beta-lactamase, multidrug and tetracycline were the most abundant (Figure S1B in Supplementary Materials). The number of ARGs was lower in B0 (natural seawater collected from P0) compared to the culture water samples (Figure S1B in Supplementary Materials), among which multidrug, tetracycline and aminoglycosides-related ARGs were dominant. In type

B samples, tetracycline, aminoglycoside, multidrug, beta-lactamases and MLSB-related ones were the main ARGs.

ARGs were detected in the soil samples C0-C3, among which multidrug, beta-lactamases, aminoglycoside, MLSB and tetracycline-related ARGs were in the majority (Figure S1C in Supplementary Materials). These ARGs were abundant in C1 (P1) and C2 (P2) compared to other soil samples (Figure S1 in Supplementary Materials).

#### *3.2. Differential Distribution of ARGs between Fish Farms and the Adjacent Natural Sea Area*

Based on the statistical analysis with the one-way ANOVA test in SPSS version 26.0, the comparison between samples from the fish farms (P1, P2 and P3) and the natural sea area (P0) indicated that ARGs were highly enriched in the fish farms, with more genes detected (Figure S1, Table S1 in Supplementary Materials). In addition, ARGs with a high abundance were recorded in P2. Some ARGs were detected only in the samples from the natural sea area, such as bacA-02, tetL-01 and ampC-06. Water samples from four locations shared more ARGs in common.

#### *3.3. Mechanisms of ARGs in Different Type of Samples*

The mechanisms of ARGs detected in all type of samples were antibiotic deactivation, cellular protection and efflux pump, with little other or unknown mechanisms (Figure S1 in Supplementary Materials). These ARG mechanisms were predominant in the A2 sample, but they were found in lesser extent in the other fish samples. In soil samples, antibiotic deactivation and efflux pump were the most representative mechanisms, especially in C1 and C2.
