Emergence of fosA3 and blaCTX–M–14 in Multidrug-Resistant Citrobacter freundii Isolates From Flowers and the Retail Environment in China

We examined the prevalence and transmission of the fosA3 gene among Citrobacter freundii isolates from flowers and the retail environments. We identified 11 fosfomycin-resistant C. freundii strains (>256 μg/mL) from 270 samples that included petals (n = 7), leaves (n = 2), dust (n = 1) and water (n = 1). These 11 isolates were multidrug-resistant and most were simultaneously resistant to fosfomycin, cefotaxime, ciprofloxacin and amikacin. Consistently, all 11 isolates also possessed blaCTX–M–14, blaCMY–65/122, aac(6’)-Ib-cr, qnrS1, qnrB13/6/38 and rmtB. These fosA3-positive isolates were assigned to two distinct PFGE patterns and one (n = 9) predominated indicating clonal expansion of fosA3-positive isolates across flower markets and shops. Correspondingly, fosA3 was co-transferred with blaCTX–M–14 via two plasmid types by conjugation possessing sizes of 110 kb (n = 9) and 260 kb (n = 2). Two representatives were fully sequenced and p12-1 and pS39-1 possessed one and two unclassified replicons, respectively. These plasmids shared a distinctive and conserved backbone in common with fosA3-carrying C. freundii and other Enterobacteriaceae from human and food animals. However, the fosA3-blaCTX–M–14-containing multidrug resistance regions on these untypable plasmids were highly heterogeneous. To the best of our knowledge, this is the first report of fosA3 and blaCTX–M–14 that were present in bacterial contaminants from flower shops and markets. These findings underscore a public health threat posed by untypable and transferable p12-1-like and pS39-1-like plasmids bearing fosA3-blaCTX–M–14 that could circulate among Enterobacteriaceae species and in particular C. freundi in environmental isolates.


INTRODUCTION
There are two major problems in the treatment of bacterial infections: the spread of multidrug-resistant (MDR) or extensively drug-resistant (XDR) pathogens and lack of development of new antibiotics active against these bacteria (Falagas et al., 2016). This situation has renewed interest in older antibiotics such as fosfomycin as alternatives or "last resort" therapies (Falagas et al., 2016;Sastry and Doi, 2016). Fosfomycin is viewed as a suitable empirical drug that has retained activity against resistant strains (Popovic et al., 2010). The World Health Organization has reclassified fosfomycin as a "critically important antimicrobial" based on its broad-spectrum bactericidal reactivity and good pharmacological properties (World Health Organization [WHO], 2011).
Clinical cases of fosfomycin resistance have increased in the last decade especially due to inactivation of the drug via plasmid-mediated fosfomycin-modification (fos) genes in Enterobacteriaceae (Diez-Aguilar et al., 2013). Fosfomycin covalently binds to a cysteine thiol in the active site of MurA and interferes with peptidoglycan synthesis at an earlier step than the action of β-lactams or glycopeptides. FosA is a glutathione S-transferase that covalently modifies fosfomycin for inactivation. There are currently more than 10 fos types, and fosA, its subtypes, and fosC2 are primarily found in the Enterobacteriaceae (Yang et al., 2017). In China, plasmidencoded fosA3 in Escherichia coli isolates from food and pets have been reported with high detection rates (Hou et al., 2012;Yang et al., 2014;Yao et al., 2016;Wang et al., 2017) although fosfomycin has not been approved for veterinary use. Interestingly, fosA3 is often found co-localized with bla CTX−M on epidemic plasmids and this most likely promotes the transfer and dissemination of fosA3 in humans and animals (Yang et al., 2014;Feng et al., 2015). Co-spread of fosA3 with other important antibiotic resistance genes (ARG) is concerning due to the potential to rapidly develop into MDR Enterobacteriaceae strains.
Additionally, E. coli strains carrying fosA3 and bla CTX−M−14 have been isolated from vegetables in Netherlands (Freitag et al., 2018), even more mcr-1 gene in E. coli was identified in fresh vegetables in Guangzhou, China (Luo et al., 2017). Literature also have claimed that plants could be contaminated by manure and wastewater from animal farming, contributing to the widely spread of AMR (Zeng et al., 2019;Sun et al., 2020). As a kind of plant, flowers are closely related to human life and possible vectors for AMR genes transformation. However, few reports have focused on the significance of flowers as a pathway for AMR spread.
Therefore, we investigated drug-resistant bacteria/drugresistant genes from flowers and retail environment including water and dust in florists in Guangzhou, China. There are relatively few reports of plasmid-borne fosA3 in Citrobacter. freundii, a bacterium associated with opportunistic nosocomial infections of the respiratory and urinary tracts and blood (Feng et al., 2015;. Herein, we present the first report of the emergence of fosA3 in C. freundii isolates from flowers and the retail environments.
We further investigated the molecular epidemiology of fosA3-carrying C. freundii isolates and characterized the fosA3-bearing plasmids.

Bacterial Strains and Detection of fos Genes Materials and Methods
A total of 270 samples were randomly collected from 3 flower markets and 6 flower shops in Guangzhou, China during March 2017. The samples included lily petals (n = 90), lily leaves (n = 90), dust (n = 45), and water (n = 45). Flowers and leaves were separately collected in sterile sealed bags. Water for watering flowers was collected in 50 mL centrifuge tubes. Dust samples were wiped with sterile cotton swabs in the surface dust from tables or floors (each 10-cm × 10-cm area) in retail shops, and then were rinsed in 2 mL sterile physiological saline solution. One petal and one leaf of each lily flowers were picked and washed with 10 mL of sterile saline solution. Then 100 µL of dust-resuspension, flower washed fluid and water samples were incubated in 4 mL drug-free LB broth for 12-16 h at 37 • C and plated on MacConkey agar plates containing 256 µg/mL fosfomycin plus 25 µg/mL glucose-6-phosphate. After 18 h incubation at 37 • C, 1-2 red colonies of different morphologies from each plate were selected. The bacterial species identification was performed using the MALDI-TOF MS (Shimadzu-Biotech, Japan) and 16S rRNA gene sequencing (16srRNA-F: AGAGTTTGATCATGGCTC; 16srRNA-R: GGTTACCTTGTTACGACTT).

Molecular Typing
Chromosomal DNA digested with XbaI restriction enzyme was used for PFGE (Gautom, 1997) to analyze the genetic relatedness of all isolates containing fos. PFGE patterns were analyzed with the Dice coefficient and the unweighted pair group method with average linkages (UPGMA) clustering method using BioNumerics software (Applied Maths, Sint-Martens-Latem, Belgium). PFGE types were defined with >90% similarity between clusters.

Transfer of fos Genes, Gene Location and Plasmid Replicon Typing
To determine the transferability of fosA3 genes, isolates positive for fosA3 were selected for conjugation experiments using the broth-mating method and streptomycin-resistant E. coli strain C600 (MIC > 2000 µg/mL) as the recipient. The donor and recipient strains were inoculated into 4 mL LB broth (Huankai Co., Ltd., Guangzhou, China) and shaken at 37 • C for 4 h, then the donors and recipients were mixed in a 1:4 (100 and 400 µL) ratio in a 2 mL EP tubes and incubated at 37 • C for 20 h. Transconjugants were selected on MacConkey agar plates supplemented with 2000 µg/mL streptomycin and 256 µg/mL fosfomycin (Yang et al., 2014). Antimicrobial susceptibility testing of the transconjugants and co-transfer of other resistance genes were determined as mentioned above. Incompatibility (Inc) groups were determined using PCR-based replicon typing (PBRT) (Carattoli et al., 2005).
The bacterial cell of the transconjugants were lysed with the ESP buffer (0.5 M EDTA, pH 9.0; 1% sodium lauroyl sarcosinate; 1 mg of proteinase K per mL) and then the bacterial DNA was embedded in the gel block. The S1-PFGE protocol is detailed in previous reports (Barton et al., 1995). The fosA3 gene genomic locations were identified by linearization of plasmids from transconjugants using S1 nuclease followed by PFGE (Yang et al., 2014). Southern blotting was carried out from S1-PFGE gels using a digoxigenin-labelled probe specific for fosA3. 1 http://www.ncbi.nlm.nih.gov/

WGS Sequencing and Characterization of fosA3-Bearing Plasmids
Based on the results of plasmid analysis and PFGE typing, the total genomic DNA was extracted from two C. freundii strains S39 and H12-3-2 using a TIANamp Bacteria DNA Kit (Tiangen) and DNA libraries were constructed with 250-bp paired-end whole-genome sequencing using the Illumina HiSeq system (Illumina, San Diego, CA, United States) (Ma et al., 2020). The obtained paired-end Illumina reads were assembled de novo using SPAdes v3.6.2 (default parameters except -careful and -k 21,33,55,77,99,127) (Bankevich et al., 2012). In addition, to obtain long reads sequence, selected strains were further sequenced using Oxford Nanopore MinION flowcell R9.4 (Li R. et al., 2018). De novo hybrid assembly was performed using a combined Illumina HiSeq and Nanopore sequencing approach (Nextomics). Genome assembly was performed with Unicycler version 0.4.1 (Wick et al., 2017) using a combination of short and long reads, followed by error correction with Pilon version 1.12 (Walker et al., 2014). Gene prediction and annotation were performed using RAST (Overbeek et al., 2014) 2 and BLAST 3 and rechecked manually. Alignments with highly homologous sequences (with >90% coverage and >90% nucleotide identity) from the NCBI database and generation of plasmid maps were performed with BRIG (Alikhan et al., 2011) and Easyfig (Sullivan et al., 2011).

Nucleotide Sequence Accession Numbers
The representative fosA3-bearing genome sequences S39 and H12-1-2 were submitted to NCBI with the accession numbers CP045555 and CP045837, respectively, and fosA3-bearing plasmids sequences pS39-1 and pH12-1 with the accession numbers CP045556 and CP045838, respectively.

Pulsed-Field Gel Electrophoresis (PFGE) Typing
We successfully performed PFGE typing for all 11 fosA3-positive C. freundii isolates. We found two different PFGE patterns that retained >90% similarity and were designated types I and II. Type I predominated and contained nine C. freundii isolates recovered from Lily petals, dust and water from two markets and three flower shops (Figure 1).

DISCUSSION
Fosfomycin plays a critical role against MDR and XDR Gramnegative pathogens and in particular, it commonly plays a synergistic role paired with β-lactams and aminoglycosides in treating urinary tract infections (Sastry and Doi, 2016). However, the occurrence and dissemination of plasmid-borne fosfomycin resistance genes, especially fosA3, has cast a shadow over the use of fosfomycin in clinics. In this study, we identified fosA3-carrying C. freundi isolates with a detection rate of 4.1% among 270 samples from flowers and the retail environments in Guangzhou, China. The prevalence of fosA3 among C. freundii strains from flowers was relatively low when compared with E. coli isolated from food animals (8. 8%, during 2009-2011; 10.5%, during 2015-2016) in China (Yang et al., 2014;Wang et al., 2017). However, our fosA3 detection rate was consistent with the recent report that fosA3 sporadically occurs in C. freundii and other Enterobacteriaceae including Salmonella spp., Proteus mirabilis, and Enterobacter fergusonii from humans, food animals, pets, retail meat as well as wild birds (Lin and Chen, 2015;Villa et al., 2015;Wong et al., 2016;Yao et al., 2016;Fang et al., 2019). Perhaps of great concern is the emergence of the fosA3 gene in different Enterobacteriaceae species from these diverse origins.
Plants including vegetables and flowers can be contaminated with ARGs via wastewater irrigation or manure application. Interestingly, an MDR C. freundii strain (WCHCF65) in sewage from a Chinese hospital carried multiple clinically significant ARGs including bla CTX−M−12 , bla CTX−M−14 , bla SHV−12 , bla NDM−1 and bla KPC−2 as well as fosA3 . Additionally, mcr-1 was present with fosA3 and bla CTX−M−14 in Raoultella ornithinolytica and E. coli isolates from retail vegetables in Guangzhou, China (Luo et al., 2017). MDR E. coli strains co-harboring fosA3 and bla CTX−M−14 also occurred in fresh vegetables in Netherlands (Freitag et al., 2018). These observations indicate that plants are potential ARG reservoirs including fosA3.
In the present study, we identified the presence of fosA3 on flowers and the retail environments. These fosA3-positive C. freundi isolates exhibited resistance to most of the tested antibiotics including cefotaxime, ciprofloxacin, and amikacin in addition to fosfomycin. Consistently we found that fosA3 coexisted with the ESBL bla CTX−M−14 , the pAmpCs bla CMY−65 and bla CMY−122 , the PMQR genes aac(6')-Ib-cr, qnrS1, qnrB13/qnrB6/qnrB38 as well as rmtB. Additionally, WGS analysis demonstrated that bla CMY−65 /bla CMY−122 and qnrB13/qnrB38 were located in the chromosome. This was consistent with the origin of plasmid-mediated qnrB and bla CMY−2 -like genes from the chromosome of Citrobacter spp. (Verdet et al., 2009;Jacoby et al., 2011;Liao et al., 2015). Importantly, flowers contaminated with MDR bacteria will most likely come into direct contact with humans complicating the treatment and management of disease.
In this study, the fosA3-carrying C. freundii isolates were genetically related as judged by their PFGE profiles. In particular, we found an epidemic PFGE type that was composed of isolates from diverse origins across flower markets and shops. This indicated possible clonal dissemination of MDR fosA3positive C. freundii isolates from flower markets and shops in a local region.
The spread of bacterial plasmids is an increasing global problem contributing to widespread ARG dissemination (San Millan, 2018). Several plasmid types are associated with the spread of fosA3 and in particular, the epidemic IncF33:A-:Band ST3-IncHI2 plasmids in Enterobacteriaceae from pets and food animals (Hou et al., 2012;Yang et al., 2014;Fang et al., 2019). However, we found that all our 11 fosA3-carrying plasmids including p12-1 and pS39-1 in C. freundi could not be assigned to any known Enterobacteriaceae incompatibility group. The single replicon gene repA possessed in plasmid p12-1 was highly similar to that in plasmid p112298-KPC where the repA was assigned to the IncFII RepA superfamily (Feng et al., 2015). Plasmid p12-1 showed a similar backbone region to the untypable fosA3-bearing plasmids in E. hormaechei and C. freundii isolates from humans in China and the United States ; Figure 3).
Interestingly, a similar scenario was also observed for the other untypable fosA3-bearing plasmid pS39-1. Linear genomic comparisons revealed that a conserved backbone, including the two unclassified replicons repA and repB, were identified between pS39-1 and another two untypable fosA3-bearing plasmids pTEM-2262 and pMH17-012N_4 in C. freundii isolates of pig and human origin from China Zhang et al., 2019; Figure 2). Noticeably, pTEM-2262 also shared a conserved backbone with another four untypable non-fosA3bearing plasmids from different species including C. freundii, Kosakonia radicincitans, and Citrobacter werkmanii with origins including the environment, vegetables and humans (Zhou et al., 2017;Zurfluh et al., 2017;Becker et al., 2018;Barry et al., 2019). These indicated that the untypable and conjugal p12-1-like and pS39-1-like plasmids could act as vectors for fosA3 Frontiers in Microbiology | www.frontiersin.org transmission between different Enterobacteriaceae from different ecological niches.
In contrast to the conserved backbones, the fosA3bla CTX−M−14 -containing MDR of these untypable plasmids from the GenBank were highly heterogeneous. This was primarily due to acquisition or deletion of resistance determinants mediated by mobile genetic elements and recombination. Plasmid pS39-1 possessed a large fosA3-containing MRR composed of 15 ARGs and diverse insertion sequences and transposons. Furthermore, the large fosA3-containing MRRs in plasmid pS39-1 partially resembled analogous regions from different plasmids indicating the MRR likely originated from the recombination of genetic contents from different plasmids as previously described (Hammerum et al., 2016;Jing et al., 2019;Maherault et al., 2019). Unlike pS39-1, an additional insertion region mainly composed of hypothetical proteins was also integrated into the variable region of plasmid p12-1 in addition to the fosA3-bla CTX−M−14containing resistance region. Interestingly, heterogeneous fosA3-containing multidrug resistance regions have been also identified on the epidemic ST3-IncHI2 and F33:A-:B-plasmids with a conserved backbone in Salmonella (Yang et al., 2014). These data indicated that diverse and flexible transmission of fosA3 was associated with heterogeneous MRRs and conserved backbones of a specific group of plasmids including ST3-IncHI2 and F33:A-:B-as well as untypable replicons in Enterobacteriaceae.
In conclusion, this study revealed the presence of fosA3 that co-existed with bla CTX−M−14 in MDR C. freundii isolates from flowers and the retail environments. Clonal expansion and horizontal transmission of untypable plasmids were involved in the spread of fosA3 and bla CTX−M−14 among these C. freundi isolates. To the best of our knowledge, this is the first report of the identification of fosA3 in bacteria isolated from flower shops and markets. The fosA3-and bla CTX−M−14 -bearing untypable and transferable p12-1-like and pS39-1-like plasmids could circulate among diverse Enterobacteriaceae species from diverse origins, including plants and humans. Future studies are necessary to monitor the prevalence and transmission of these plasmids in Enterobacteriaceae species especially C. freundi, to better understand the potential threat to public health.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/ Supplementary Material.

AUTHOR CONTRIBUTIONS
Y-HL, JS, and X-PL designed the study. KC, L-XF, DW, BH, Q-WG, J-QL, and Z-XZ performed the experiments and collected the data. KC, L-XF, DW, BH, Q-WG, J-QL, Z-XZ, X-LL, YY, and X-RW analyzed and interpreted the data. KC wrote the draft of the manuscript. L-XF, Y-HL, JS, X-PL, and X-LL edited and revised the manuscript. Y-HL and JS coordinated the whole project. All authors contributed to the article and approved the submitted version.