We obtained 264 S. enterica isolates including serovars London (n = 43), Kentucky (n = 35), Typhimurium (n = 34), Derby (n = 26), Rissen (n = 24), Indiana (n = 19), and others (n = 83) from 1,041 fecal samples from food-producing animals (pigs, chickens, and cattles) and 1,205 retail meat products (chicken meat, pork, and beef) in supermarkets or farmers’ markets in Gansu, Guangdong, Guizhou, Henan, Hubei, Jiangsu, Liaoning, Shandong, and Xinjiang provinces and Shanghai of China from July 2019 to June 2020 (Supplementary Table 2). Eighty-seven Salmonella isolates including serovars Typhimurium (n = 33), Rissen (n = 30), London (n = 14), Derby (n = 4), Reading (n = 4), and Pakistan (n = 2) were obtained from pig carcass swab samples and intestinal content samples in a slaughterhouse in Jiangsu province and nine Salmonella isolates including Wandsworth (n = 2), Lexington (n = 2), Reading (n = 1), Indiana (n = 1), Typhimurium (n = 1), London (n = 1), and Virchow (n = 1) from chicken meat from Chongqing, China, in 2016.

The presence of fosA7 was detected by polymerase chain reaction and sequencing with the primers fosA7-F (5′-ACTTAACGCTTGCTGTC-3′) and fosA7-R (5′-TGCTAAATCTCCCACAT-3′). All fosA7-positive isolates were tested for their susceptibility to ampicillin, cefotaxime, meropenem, gentamicin, amikacin, streptomycin, tetracycline, chloramphenicol, florfenicol, nalidixic acid, ciprofloxacin, fosfomycin (with 25 mg/L glucose-6-phosphate), and sulfamethoxazole/trimethoprim using the agar dilution method and colistin using the microbroth dilution method. The results were interpreted according to CLSI M100, 30th edition (Clinical and Laboratory Standards Institute [CLSI], 2020). Streptomycin (≥32 mg/L) and florfenicol (≥ 32 mg/L) were interpreted according to EUCAST¹ epidemiological cutoff values for S. enterica. Fosfomycin (> 32 mg/L) was interpreted according to EUCAST (see text footnote 1) breakpoint tables for Enterobacterales. E. coli American Type Culture Collection (ATCC) 25922 was used as quality control.

The whole genomes of all fosA7-positive Salmonella were sequenced by Illumina Hiseq, and sequence reads were assembled into contigs with SPAdes v.3.13.0. The whole-genome sequences were analyzed for resistance genes, chromosome mutations, and plasmids using Center for Genomic Epidemiology pipelines². Phylogenetic analysis of 30 fosA7-positive S. Derby isolates and ST71 fosA7-negative S. Derby 14C-S8N4 obtained from pork in Yangzhou, Jiangsu province, China, in 2014 based on 3002 cgMLST locus were performed using cgmlstfinder.py contained in the cgMLSTFinder service (Clausen et al., 2018). iTOL was used to make the phylogenetic tree visual (Letunic and Bork, 2021). Five fosA7-carrying S. Reading isolates in this study and S. Reading CVM N42528 (GenBank accession no. GCA_001481335.1) from pork chop in the United States were analyzed for the core genome regions by Parsnp software. The whole-genome sequencing data have been deposited in the GenBank under accession number PRJNA743999.

The fosA7 deletion in S. Derby HA2-WA5 that exhibits susceptibility to all tested agents was performed by double exchange of homologous recombination as previously described using primers listed in Supplementary Table 3 (Guo et al., 2020). The deletion mutant HA2-WA5ΔfosA7 was tested for minimum inhibitory concentration (MIC) of fosfomycin (25 mg/L glucose-6-phosphate).

The fosA7 of HA2-WA5 was amplified by polymerase chain reaction and cloned into the pBR322 vector. The vector pBR322 and the recombinant plasmid pBR322-fosA7 were transformed into E. coli DH5α, Salmonella Typhimurium SL1344, S. Pullorum C79-13, S. Enteritidis P125109, S. Derby 14C-S8N4 (ST71, fosA7-negative), Enterobacter cloacae CMCC45301, and Klebsiella pneumoniae CMCC46117 via electroporation, selected by 32 mg/L ampicillin. The function of fosA7 was determined by testing their MICs of fosfomycin (25 mg/L glucose-6-phosphate).

The growth curves of all transformants, HA2-WA5 and HA2-WA5ΔfosA7, were determined in LB broth for 14 h at 37°C and 180 rpm. The overnight cultures were adjusted to an OD₆₀₀ of 1 and then diluted 1:200 (10 μl/20 ml) in fresh LB broth. The OD₆₀₀ of the cultures was measured every hour using Tecan Sunrise Microplate Reader (TECAN, Switzerland). Experiments were performed in triplicate.

To investigate the stability of pBR322 and pBR322-fosA7 in different hosts, all transformants were maintained for 7 days in daily refreshed (100-fold dilution) LB broth without antibiotics. Cultures were streaked on LB agar plates, and 50 colonies were replica-plated on LB agar plates containing 32 mg/L ampicillin or 256 mg/L fosfomycin with 25 mg/L glucose-6-phosphate on each day.

To investigate the relative fitness (RF) of fosA7, Salmonella isolates SL1344-pBR322-fosA7, C79-13-pBR322-fosA7, 14C-S8N4-pBR322-fosA7, P125109-pBR322-fosA7, and HA2-WA5ΔfosA7 were competed against SL1344-pBR322, C79-13-pBR322, 14C-S8N4-pBR322, P125109-pBR322, and HA2-WA5, respectively, in daily refreshed (1,000-fold dilution) LB broth for 96 h as previously described (Wu et al., 2018). At 48 and 96 h, the total number of bacteria was determined by plating proper diluted cultures on LB agar, and all colonies were replica-plated on LB agar with 16 mg/L fosfomycin and 25 mg/L glucose-6-phosphate. All competition experiments were performed at least three replicates. The RF was calculated as the percentage of transformants or deletion mutant in all isolates. An RF value > 0.5 indicated that the transformants/deletion mutant had a selective advantage over the original strain, whereas a score < 0.5 indicated a fitness cost.

The results were analyzed with GraphPad Prism version 7.0 (GraphPad Software, Inc., La Jolla, CA, United States).

Among 360 Salmonella isolates, fosA7 was detected in 35 strains, with S. Derby (n = 30) being the dominant host. Five S. Reading were also positive for fosA7. The fosA7-carrying Salmonella isolates showed highest resistance to chloramphenicol (n = 31, 88.57%), followed by nalidixic acid (n = 30, 85.71%), tetracycline (n = 29, 82.86%), florfenicol (n = 27, 77.14%), sulfamethoxazole/trimethoprim (n = 27, 77.14%), and ampicillin (n = 27, 77.14%), but none of them were resistant to amikacin, meropenem, and colistin (Table 1). Notably, all fosA7-positive Salmonella isolates were still susceptible to fosfomycin, with MICs ranging between 1 and 32 mg/L (Table 1).

Whole-genome sequencing showed that all S. Derby belonged to ST40, with one being ST40 variant (isolate YZ19MBS1), and five S. Reading strains were classified as ST1628. The fosA7 gene was located on chromosomes of all Salmonella isolates in this study and differed from the firstly identified fosA7 in S. Heidelberg by one (S. Reading) or 12 (S. Derby) nucleotide differences, resulting in one or five amino acid changes. Agreement with the first detected fosA7 in S. Heidelberg ABB07-SB3031 (Rehman et al., 2017), the fosA7 gene was surrounded by putative open reading frames. The genetic environment of fosA7 and its neighboring region (∼24.3 kb) in all S. Derby and S. Reading strains obtained in this study were highly identical (> 98.0%) to the corresponding region of numerous S. Heidelberg strains, including ABB07-SB3031 (GAC_000973785) and S. Stanleyville strain CFSAN059881 (CP075116) (Supplementary Figure 1). In addition to fosA7, 35 Salmonella isolates carried one to 28 resistance genes, conferring resistance or decreasing susceptibility to β-lactams (bla_TEM, bla_OXA, bla_CTX–M, and bla_CMY), aminoglycosides [aac(3)-IV, aac(3)-IVa, aph(4)-Ia, aph(3)-Ia, aac(6′)-Iaa, ant(3″)-Ih, aac(6′)-IId, aadA, and strAB], quinolones [qnrS, qnrVC, aac(6′)-Ib-cr, and oqxAB], tetracyclines [tet(A), tet(B), and tet(M)], chloramphenicols (catA1, catB3, and floR), sulfonamides (sul1, sul2, and sul3), trimethoprims (dfrA12 and dfrA14), macrolides (mefB and mphA), and rifampin (arr-3). Mutation within parC (T57S) was found in all fosA7-bearing S. Derby; gyrA (D87N) mutation was also observed in two of them (Figure 1A). No plasmid replicons and mutations in gyrA or parC were detected in S. Reading. Among the 30 fosA7-positive S. Derby, plasmid replicons were identified in 10 of them (Figure 1A).

To further understand their genetic relationship, we performed a phylogenomic analysis on the fosA7-positive S. Derby and S. Reading. The results showed the division of 30 fosA7-positive S. Derby ST40 isolates into four clades (Figure 1A). Clade I only included one isolate, HA2-WA5, obtained from a pig slaughterhouse in 2016 with the least resistance genes. Twelve S. Derby isolates from pork were classified as clade II, and seven of them contained the same resistance genes. Clade III contained two isolates from pork and one isolate from cattle with few resistance genes. Fourteen isolates from various sources, including all S. Derby from cattle and beef, belonged to clade IV; numerous resistance genes were observed in this clade. Five fosA7-positive S. Reading were clustered in two clades (Figure 1B). The strains in the same clade had identical antimicrobial susceptibility profiles and resistance genes. The results indicate that clonal spread has occurred particularly in the same host.

Because fosA7 was located on the chromosome and S. Derby was the most common host, we constructed an in-frame fosA7 deletion mutant to assess the impact of chromosomally encoded FosA7 in S. Derby. After deletion of fosA7 in S. Derby HA2-WA5, its MIC of fosfomycin was reduced from 32 to 2 mg/L, further confirming the role of chromosomally located fosA7 in reducing the susceptibility of fosfomycin. The growth curves of HA2-WA5 and HA2-WA5ΔfosA7 were almost identical, suggesting that the fosA7 deletion in the chromosome did not affect the growth (Figure 2A). Additionally, deletion of fosA7 caused slightly decreased fitness at 48 h (P > 0.05) but enhanced fitness at 96 h with a biological advantage (0.01 < P < 0.05) (Figure 3A).

Chromosomal-encoding FosA7 in S. Derby exhibited low MIC values to fosfomycin. To investigate whether the chromosomal fosA7 in S. Derby could be mobilized to fosA7-negative bacteria and inactivated fosfomycin, we constructed recombinant plasmid pBR322-fosA7 and transformed it to different hosts. The recombinant plasmid pBR322-fosA7 was successfully transformed to E. coli DH5α, S. Typhimurium, S. Pullorum, S. Enteritidis, and S. Derby but failed to K. pneumoniae and E. cloacae after multiple electroporation experiments. Plasmid pBR322-fosA7 was stable for 7 days in S. Typhimurium, S. Pullorum, S. Derby, and E. coli but was lost (∼4%) in S. Enteritidis on the last day of passage (Supplementary Figure 2).

The acquisition of fosA7 significantly increased MICs of fosfomycin in Salmonella (1,024-fold) and conferred fosfomycin resistance, whereas E. coli DH5α was still susceptible after acquiring fosA7 with 16-fold elevated MIC of fosfomycin (Supplementary Table 4).

To assess the impact of fosA7 on the bacterial growth, growth curves of E. coli, S. Typhimurium, S. Pullorum, S. Enteritidis, and S. Derby carrying pBR322 or pBR322-fosA7 were compared. The acquisition of fosA7 rarely affected the growth of E. coli and Salmonella (Figures 2B–E).

The RF of fosA7 was determined by growth competition in vitro; E. coli DH5α was not included because FosA7 did not confer fosfomycin resistance in DH5α. The fosA7 gene significantly increased bacterial fitness, particularly in S. Pullorum (Figure 3B). S. Pullorum C79-13-pBR322-fosA7 completely replaced C79-13-pBR322 at 48 h, showing a significant biological benefit. In S. Typhimurium, fosA7 also exhibited fitness advantage (RF = 0.7555 ± 0.0168) at 48 h; no apparent difference was observed between 48 and 96 h (RF = 0.7920 ± 0.0710). Among four Salmonella serotypes, the least enhancement of fitness was shown in S. Derby (RF = 0.6924 ± 0.0435 at 96 h) after acquiring fosA7.