Genotypic and phenotypic characterization of Escherichia coli isolated from mollusks in Brazil and the United States

Abstract The aim of this study was to determine the serogroups, antimicrobial resistance and genetic diversity of Escherichia coli isolates from samples of bivalve mollusks collected along Santa Catarina coast, Brazil, and from the Chesapeake Bay, Maryland, USA. One hundred forty‐one E. coli isolates were characterized for serogroups with 181 specific O antisera and antimicrobial susceptibility using the disk diffusion method. The genetic diversity was assessed using pulsed‐field gel electrophoresis (PFGE). The results showed that among the isolates, 19.9% were classified as multi‐drug resistant (MDR) and resistance was most frequently observed to cephalothin, nitrofurantoin, and ampicillin. The predominant serogroups were O6, O8, and O38. Some serogroups were recognized as pathogenic E. coli. PFGE dendrograms indicated extensive genetic diversity among the isolates. Although characteristics of the E. coli isolates were highly variable, it is important to note that E. coli belonging to pathogenic serogroups and MDR isolates are present in mollusks of both study areas. This is the first report on the phenotypic and genotypic characterization of E. coli from mollusks from Santa Catarina and the Chesapeake Bay that should encourage studies focusing on comparison of isolates across countries.

and has also been suggested as a possible indicator to assess the antimicrobial resistance status in environmental settings (Berendonk et al., 2015).
The presence of pathogenic strains of E. coli in seafood is a public health concern and may lead to serious health risks to consumers (Costa, 2013). Consequently, authorities in various countries, such as Brazil (Brazil, 2012), the United States (FDA, 2015), and the European Union (EU, 2004), have established regulatory limits and monitoring programs using E. coli counts, fecal coliform levels of bivalves, or fecal coliform levels of bivalve growing areas. Shiga toxin-producing E. coli (STEC), especially E. coli O157:H7, has been widely implicated in outbreaks of foodborne illnesses (CDC, 2018).
However, no study was conducted to characterize E. coli recovered from mollusks by genotypic and phenotypic methods in the United States and in Brazil. Serotyping is an important tool for the differentiation of E. coli strains, but it does not comprehensively characterize a strain. In recent years, various genotyping methods such as pulsed-field gel electrophoresis (PFGE) have been used to differentiate E. coli and determine the genetic relationships of strain. PFGE is considered as the gold standard method because of its high level of discrimination and it has also been reported that this technique could be the most discriminatory genotypic method to provide a reproducible DNA fingerprinting .
Antimicrobial resistance (AMR) is another tool for characterization of bacterial isolates. With the wide use of antimicrobials in humans and in the environment, AMR E. coli have been reported from different sources and countries (Dou et al., 2016;Kao et al., 2016;Rabbia et al., 2016;Zhang et al., 2016). The presence of resistant organisms in the environment is an emerging concern around the world (Watkinson, Micalizzi, Graham, Bates, & Costanzo, 2007), and since AMR bacteria can accumulate in bivalve mollusks (Barkovskii, Green, & Hurley, 2010), it has been suggested that bivalves may be useful in assessing environmental contamination by AMR bacteria (Berendonk et al., 2015;Rees et al., 2015).
Until now, little information is available on the genotypic and phenotypic characteristics of E. coli isolated from samples of mollusks in Santa Catarina, Brazil, and the Chesapeake Bay, Maryland, USA, and no data have been published regarding the genetic relatedness between E. coli isolates from Brazil and the United States. Considering that both Santa Catarina and the Chesapeake Bay are known for mollusk production (NOAA, 2012;Santos, Marchiori, & Della Giustina, 2017), research is important to improve the knowledge about the E. coli strains in samples from those harvesting areas. Since bivalves can accumulate micro-organisms, including E. coli, present in surrounding waters by their filter-feeding activities and may present a risk to public health, studies on the characterization of those strains should be addressed. The aim of this study was to determine the serogroups, antimicrobial resistance and genetic diversity of E. coli isolates recovered from mussels and oysters collected in two distant global regions.

| Collection of samples (mussels and oysters)
In Santa Catarina, South Brazil coast, a total of 100 samples were collected from 10 different localities (São Francisco do Sul, Balneário Barra do Sul, Penha, Balneário Camboriu, Bombinhas; Porto Belo, Gov. Celso Ramos, Florianópolis, São José and Palhoça) from January to July of 2015 in an interval of three weeks. They were comprised of 40 samples of oysters (Crassostrea gigas and Crassostrea rhizophorae) and 60 samples of mussels (Perna perna). Each sample was comprised of 12 mollusks. In the Chesapeake Bay, Maryland, USA, a total of 18 samples of oysters (Crassostrea virginica) were taken from five localities (Oxford, Manokin, Chester, Broad Creek, and Tangier Sound) during October and November of 2015. Sampling frequency was every 3 weeks, and each sample was comprised of 12 mollusks.
Mussel samples were collected only from Santa Catarina coast because mussels represent 82% of total bivalve mollusk production while oysters represent 15% (Santos et al., 2017).
The samples in Brazil were collected by hand, and in the United States, an oyster dredge was used. Immediately after harvesting, oyster samples were bagged and placed in insulated chests. Bubble wraps were placed between the oyster bags and ice bags to prevent direct contact with ice and water. The shipping temperature was monitored by data loggers (ACR Systems, Inc., Data Logger Store, Contoocook, NH, USA) to ensure that it was maintained between 2 and 10°C. All microbiological analyses were initiated within 4 hr of sample collection.

| Microbiological analysis
All 118 samples were examined quantitatively for E. coli by a five tube most probable number (MPN) method using minerals modified glutamate broth (MMGB) (Oxoid Ltd, Basingstoke, Hampshire, UK) and the chromogenic medium Tryptone Bile 5-bromo-4-chloro-3indolyl-β-d-glucuronide agar (TBX) (Oxoid), in accordance with ISO 16649-3 method (ISO, 2015). The results were given as the number of E. coli in MPN/100 g. The lowest detectable concentration of E. coli when applying this method was 20 MPN/100 g.

| Antimicrobial susceptibility testing
The susceptibility to antimicrobials was tested by the disk diffusion  (Pagadala et al., 2012).
Escherichia coli isolates were classified as susceptible, intermediate resistant, or resistant according to the CLSI criteria for Enterobacteriaceae (CLSI, 2014). A strain was considered multi-drug resistant (MDR) when demonstrating resistance to three or more antimicrobial classes (Schwarz et al., 2010).
Escherichia coli isolates were tested for AMR index with the intent to find a correlation with pollution sources since this method was shown to be useful in differentiating human from nonhuman pollution sources in previous studies (Watkinson et al., 2007;Webster et al., 2004). The AMR indices were calculated as follows: isolate AMR index = no. of antimicrobials to which the isolate was resistant/total no. of antimicrobials tested (Parveen et al., 1997).

| Serogrouping
All E. coli isolates were referred to E. coli Reference Center located at Pennsylvania State University, Pennsylvania, USA, for serogrouping. The isolates were serogrouped for "O" antigen according to the methodology described by Orskov, Orskov, Jann, and Jann (1977

| Molecular typing
Genetic diversity of the isolates was assessed using PFGE according to a standard protocol developed by the United States Centers for Disease Control and Prevention (CDC) for E. coli. Briefly, agarose plugs were prepared with E. coli cell suspension, lysed with proteinase K and digested using XbaI enzyme. DNA fragments were separated on 1.5% agarose gel by electrophoresis on a CHEF DR-III system (Bio-Rad Laboratories, USA). The gel was stained with ethidium bromide (40 mg/ml) and then de-stained with deionized water and visualized with ultraviolet light. A molecular size standard (Salmonella enterica serotype Braenderup H9812) was used.
The results were evaluated using Bionumerics software (AppliedMaths, Austin, TX, USA). PFGE patterns were established based on the number and arrangement of fragments and computationally based on the levels of relatedness using the Dice similarity coefficient and unweighted pair group method using arithmetic averages (UPGMA) with 1.5% optimization and 1.5% tolerance and Pearson correlation coefficient (0.5%). A phylogenetic dendrogram was constructed based on PFGE fingerprint profiles. Isolates sharing at least 80% similarity were considered genetically related, and those sharing 100% similarity were classified as clones (Balière, Rincé, Blanco, et al., 2015;Kao et al., 2016;Rabbia et al., 2016).

| Statistical analysis
The distribution of pathogenic serogroups between the sample sites was compared using the chi-square test. A two proportion test was conducted to evaluate the incidence of MDR isolates and pathogenic serogroups between isolates from Brazil and the United States. The E. coli counts and the incidence of pathogenic serogroups were compared with the Mann-Whitney U test and logistic regression. Logistic regression analysis was also applied to evaluate relationships between pathogenic serogroups and MDR profiles; pathogenic serogroups and sites; pathogenic serogroups and E. coli counts and MDR profiles. These analyses were applied separately to Brazil and US isolates. A p value <0.05 was considered statistically significant. The statistical analysis was performed using Minitab Software, version 16.2.4.0 (Minitab Inc., State College, PA, USA).

| Prevalence and concentration of E. coli
All samples (100%) collected from Brazil (BR) and eleven (61%) of US samples were positive for E. coli, the counts varied widely, with concentrations ranging from 20 to 18,000 MPN/100 g among samples from Brazil and among samples from the United States, the concentrations ranged from <20 to 130 MPN/100 g. Fortynine percent of the samples from Brazil presented a concentration less than 230 MPN/100 g, 43% contained between 230 and 4,600 E. coli MPN/100 g and eight percent presented concentration higher than 4,600 MPN/100 g. In 54% of the oyster samples from the United States that were positive for E. coli, the bacterial counts were 20 MPN/100 g. No E. coli was found in samples from Manokin site.
A total of 141 E. coli isolates were recovered from samples collected in Brazil (n = 100) and in the United States (n = 41) for genotypic and phenotypic characterization. One typical blue colony of β-glucuronidase-positive E. coli was selected from each positive sample in Brazil (60 isolates from mussels and 40 isolates from oysters). Three to five colonies were selected from each positive sample in the United States due to small sample size.

| Antimicrobial susceptibility testing
Escherichia coli isolates were tested for susceptibility to 15 antimicrobial agents of veterinary and human health significance. Of the total number of isolates, 83% (n = 117) were resistant to at least one antimicrobial agent and 19.9% (n = 28) were classified as MDR.
The lowest AMR index observed was 0, and the highest was 0.47.
Regarding the isolates harvested in the United States, all of them were susceptible to chloramphenicol, cefepime, norfloxacin, and tetracycline and 90.2% were resistant to cephalothin, 43.9% were resistant to nitrofurantoin, 31.7% to cefoxitin, and 22% were resistant to ampicillin. Ninety-five percent of the isolates were resistant to at least one antimicrobial agent, and 26.8% were classified as MDR (Table 1). There were a total of 12 antimicrobial resistance profiles observed among US isolates (Table 2).

| Serogrouping
A total of 141 E. coli isolates were serogrouped, and 81% were typeable. The typed isolates displayed 49 different serogroups that are shown in Table 3   clusters, and at a 100% similarity, four clusters were clearly differentiated with between two and three isolates in each. These clusters showed concordance with the serogroups, isolation site, and antimicrobial profile, with few exceptions.

| D ISCUSS I ON
Since E. coli is a well-known indicator of recent fecal contamination, and it was demonstrated that oysters can bioaccumulate fecal coliforms to a concentration four times greater than surrounding water (Burkhardt & Calci, 2000), isolation and characterization of E. coli from bivalve mollusks are essential. Moreover, some serogroups of E. coli can be highly pathogenic to humans (Ramos et al., 2012) Of the 15 tested antimicrobials, E. coli were most resistant to cephalothin, nitrofurantoin, and ampicillin (Table 1). These results are consistent with previous findings (Dou et al., 2016;Parveen et al., 1997;Rees et al., 2015;Ryu, Lee, et al., 2012). Authors have reported that E. coli isolates are generally resistant to antimicrobials which have been in use the longest time in human and veterinary medicine, like nitrofurantoin, ampicillin, and tetracycline (Dou et al., 2016). Our study revealed that a high percentage (31.7%) of E. coli isolates from the United States were resistant to cefoxitin while none of the Brazilian isolates were resistant. On the other hand, 16% of isolates from Brazil were resistant to tetracycline while no US isolate presented resistance to that antimicrobial (Table 1). This might be due to geographical variation and different practices of antimicrobial use in the two countries. tries that cause influence on the water temperatures (Parveen et al., 2017;Ramos et al., 2012;Raszl, Froelich, Vieira, Blackwood, & Noble, 2016) and also the difference between the sample sizes may explain that observation. Differentiation between human and agricultural sources of E. coli strains has been limited; however, the use of AMR index has demonstrated potential for differentiating E. coli sources (Parveen et al., 1997;Watkinson et al., 2007;Webster et al., 2004).
Of the total number of isolates, 19.9% (n = 28) were MDR strains which is lower than the results observed by Van, Chin, Chapman, Tran, and Coloe (2008) Rhodes et al., (2000) provided direct evidence that related tetracycline resistance-encoding plasmids have disseminated between different Aeromonas species and E. coli and between human and aquaculture environments in distinct locations in England. Results obtained by Furushita et al. (2003) suggest that resistance genes from fish farm bacteria have the same origins as those from clinical strains in a study conducted in Japan.
The serogroup O15 that was identified in one E. coli isolate recovered from an oyster sample in Brazil was found by Balière, Rincé, Thevenot, et al. (2015) in EPEC E. coli isolate from a mussel sample collected from a shellfish-harvesting site in France and by Regua-Mangia, Gomes, Vieira, Irino, and Teixeira (2009) (Table 3), were identified among EPEC isolates recovered from shellfish samples collected in shellfish-harvesting sites in France (Balière, Rincé, Blanco, et al., 2015).
Based on statistical analysis, the presence of pathogenic serogroups was not associated with a specific site of collection, country of study, or incidence of MDR strains. However, in Brazil, the pathogenic serogroups were most frequently observed in mussel samples rather than in oysters (p < 0.05). In the United States, at a significance level of 5%, E. coli counts were higher in oyster samples where pathogenic serogroups were identified. It cannot be confirmed statistically in samples from Brazil.
The genetic relatedness of the 141 E. coli isolates was analyzed by PFGE, and the dendrograms revealed a high degree of genetic diversity ( Figure 1). Regarding the six clusters with isolates from Brazil and the United States considered genetically related with 80% similarity (Figure 1), the isolates belonged to different serogroups and no similarity in antimicrobial resistance profile was found, ex-  (Koornhof, Keddy, & McGee, 2001;Tatem, Rogers, & Hay, 2006). The spread of micro-organisms across countries was also documented too (Ruiz et al.., 2000). Those observations, and also the food trade between Brazil and the United States (Azevedo, Chaddad, & Farina, 2004), may be considered hypotheses for the presence of genetically related E. coli strains in Brazil and the United States as it has been previously demonstrated for V. cholerae, Shigella species, and enterohemorrhagic E. coli using PFGE (Koornhof et al., 2001).
In this study, many tested isolates showed a tendency to cluster based on their date of collection, serogroup, and antimicrobial resistance profile but collected from distant sites which could be observed in six different clusters (data not shown). For US isolates, four clusters were identified at 100% similarity index, we observed that in each cluster the E. coli isolates were recovered from the same sample and belonged to the same serogroup with the same antimicrobial susceptibility profile (data not shown).
The genetic diversity was expected in Santa Catarina where mollusks were collected from 10 different sites within 7 months.
In the United States, genetic variability was clearly observed too, even though the collection of samples was over a shorter period of time than Brazil. This was also observed in a previous study (Balière, Rincé, Blanco, et al., 2015) with mollusks with fewer collection sites than the present study and also with E. coli isolates from other food samples . It has been observed that the genetic variability among E. coli strains may be due to adaptive mutations that occur, for example, under stress conditions and that leads these strains to selective advantage (Foster, 2005). Furthermore, insertions and deletions in specific regions of the genome enhance the genetic variability and it could be reflected in the PFGE dendrogram (Kaas, Friis, Ussery, & Aarestrup, 2012). These findings could substantiate the genetic variability found in the present study and also the diversity of antimicrobial patterns and serogroups found.
This study contributed to the knowledge of genotypic and phenotypic characteristics of E. coli from mollusks samples. Some serogroups found in the present study are among the serogroups that could cause significant human illnesses; however, even the presence of non-pathogenic E. coli in mollusks should alert the public health since this bacterium is recognized as an indicator of fecal contamination. The antimicrobial resistance profiles of E. coli isolates from mollusks are a cause of concern, especially considering the MDR strains. Surveillance of environmental samples must be encouraged to comprehensively assess antimicrobial resistance in environmental bacteria. An extensive genetic diversity among the isolates from Brazil and the United States was observed; however, on the other hand, some isolates related were found. Those results encourage surveillance of environmental samples from different countries and perspective studies of genetic relatedness of the isolates.

CO N FLI C T O F I NTE R E S T
No conflict of interest declared.

AUTH O R S CO NTR I B UTI O N
MM and SP contributed to the conception and design of the study;

E TH I C S S TATEM ENT
Not required.

DATA ACCE SS I B I LIT Y
The authors declare that all data are included in the main manuscript.