Proteus mirabilis, widely distributed in the natural environment, is a member of Enterobacteriaceae family. As an opportunistic pathogen, P. mirabilis can cause urinary tract and wound infections. In addition, it can also contaminate meat (Kim et al., 2005; Wong et al., 2013), vegetables (Uzeh et al., 2009), and seafood (González-Rodríguez et al., 2002), and has been associated with food poisoning (Wang et al., 2010). Previous studies focused on the antimicrobial resistance and the distribution of resistance genes among foodborne P. mirabilis (Kim et al., 2005; Wong et al., 2013). Unlike other foodborne bacteria (Zou et al., 2014; Zhang et al., 2016), little data have been reported on the disinfectant resistance and molecular mechanisms underlying resistance in foodborne P. mirabilis. Disinfectants have the ability to co-select for antimicrobial resistance when these traits are genetically linked (Chapman, 2003). Moreover, much evidence has shown that disinfectant and antimicrobial resistance genes can be co-transferred between bacteria via horizontal gene transfer (Zhao et al., 2009; Call et al., 2010; Sáenz et al., 2010; Partridge et al., 2012), which poses a risk to public health.

Benzalkonium chloride (BC), an important representative of quaternary ammonium compounds (QACs), is used extensively as a disinfectant in the food processing environment to prevent the growth of microorganisms and to ensure the microbiological safety of food products. Its widespread use, however, may impose a selective pressure for resistant strains of bacteria (Cruz and Fletcher, 2012). In the past decades, resistance to BC (resistant breakpoints for BC were >30 μg/mL in Sidhu et al. (2002) and ≥20 μg/mL in Meier et al. (2017)) has been reported in many bacterial isolates from food and food processing plants (Sidhu et al., 2002; Meier et al., 2017).

Efflux pumps, QacE, QacEΔ1, QacF, QacG, QacH/I, and SugE(p), which contribute for BC resistance, have been identified in Gram-negative bacteria (Zou et al., 2014). They are members of the small multidrug resistance (SMR) family and are generally located on mobile genetic elements, such as integrons and plasmids (Zou et al., 2014). qac are closely associated with class 1 integrons. qacE is located in the 3′-conserved segment (CS) of class 1 integrons and qacEΔ1, a deletion derivative of qacE, confers increased resistance to BC (Kazama et al., 1998). qacF shows a high degree of similarity (67.8% identity) with qacE. Class 1 integrons carrying qacG have been found in Gram-negative bacteria (Chu et al., 2001; Partridge et al., 2012). qacH, which was first identified in Staphylococcus (Heir et al., 1998), has been observed frequently in Gram-negative bacteria (Hegstad et al., 2010) and it confers a broader resistance phenotype compared with qacG (Heir et al., 1998). qacH is usually recognized as an important component of class 1 integron that lacks the normal 3′-CS region (i.e., a “non-classic class 1 integron”), which has been detected in many species of Enterobacteriaceae (Antunes et al., 2007; Chang et al., 2009, 2011; Sáenz et al., 2010; Farkas et al., 2016). Several studies found qacH gene and β-lactamase genes (bla_IMP-15, bla_GES-1, bla_GES-5, and bla_OXA-2) linked to class 1 integrons from Pseudomonas aeruginosa clinical isolates (Garza-Ramos et al., 2008, 2010; Viedma et al., 2009). qacH-aadB has been reported from environmental bacteria, including Paracoccus versutus, Brevundimonas diminuta, Brachymonas denitrificans, Stenotrophomonas acidaminiphila, and Psychrobacter spp. (Li et al., 2009) and qacH-aadA8 from Vibrio cholerae (Ceccarelli et al., 2006). Additionally, qacH located on a novel transposon Tn6188 was found in Listeria monocytogenes (Müller et al., 2013). In several studies, qacH in Enterobacteriaceae was renamed as qacI to distinguish it from qacH from Staphylococcus (Naas et al., 2001; Curiao et al., 2011). For this paper, we still use the gene name “qacH”. qacH exhibits 91.6% similarity with the sequence of qacF. Finally, sugE(p) is frequently present on multidrug resistance plasmids, that have been reported in Escherichia coli and Salmonella (Zhao et al., 2009; Call et al., 2010).

Five chromosome-encoded efflux pump genes (sugE(c), emrE, mdfA, and ydgE/ydgF) have been reported to confer resistance to BC (Bay and Turner, 2009; Zou et al., 2014). In addition to mdfA encoding a multidrug resistance efflux pump belonging to the major facilitator superfamily (MFS), the remaining genes encode the SMR family efflux pumps.

The aims of this study were to assess the BC resistance and investigate the presence of disinfectant resistance genes and the association of qac genes with class 1 integrons among the P. mirabilis isolates from cooked meat products in China.

In the present study, fifty-two isolates of P. mirabilis were isolated from cooked meat samples. According to Wong et al. (2013), the rate of P. mirabilis in raw chicken carcass samples in Hong Kong was 86.2% (50/58), which was much higher than that (29.2%) in our study. Cooking procedures kill most microorganisms that colonize raw meats, however, survival of microorganisms due to improper processing or cross-contamination of food after cooking may occur. This could be an explanation for the lower incidence of P. mirabilis in cooked meat products compared to raw meat samples.

Proteus mirabilis isolates in our study showed the MICs of BC ranging from 4 to >32 μg/mL, with MICs of 24 μg/mL most frequently. As there is no standard resistant breakpoint of BC for P. mirabilis, it is difficult to classify the isolates as BC susceptible, intermediate or resistance in this study. Although data on the BC susceptibility in P. mirabilis were scarce, several studies have reported the MICs of BC in Enterobacteriaceae isolates from different sources (Sidhu et al., 2002; Aarestrup and Hasman, 2004; Zhang et al., 2016). In the study of Zhang et al. (2016), E. coli isolated from retail meat showed the MICs of BC in the range of 16–64 μg/mL. Aarestrup and Hasman (2004) have shown that the MICs of BC for Salmonella (n = 156) and E. coli (n = 202) from food animals ranged from 64 to 256 μg/mL and from 16 to 128 μg/mL, respectively. Compared with the studies mentioned above, P. mirabilis isolates in our study showed relatively lower MICs to BC. In another survey of Enterobacteriaceae from food, almost half of the isolates exhibited the MICs of BC with <10 μg/mL (Sidhu et al., 2002). Actually, the user concentrations of BC in food industry are usually 200–1000 μg/mL (Møretrø et al., 2017), which are much higher than the MICs of P. mirabilis in this study. In practical application, BC is commonly rinsed from surfaces (equipment, machines, floor etc.) with water after disinfection. This rinsing step, however, is not sufficient to remove all BC residues from surfaces and consequently, bacteria are likely to be exposed to diluted, sub-lethal BC concentrations (Buffet-Bataillon et al., 2012). Repeated exposure to sub-lethal BC concentrations may facilitate the development of resistance (Buffet-Bataillon et al., 2012). Therefore, it was not surprising that P. mirabilis isolates in the present study exhibited low-level of BC MICs.

In this study, the presence of QAC resistance genes was investigated. Our results showed that mdfA and ydgE/ydgF were the most prevalent among P. mirabilis, which was in agreement with the similar studies of E. coli (Zou et al., 2014; Zhang et al., 2016). Among the isolates tested in our study, qacE always occurred simultaneously with qacEΔ1. It was noted that the presence of sugE(p) was significantly associated with the higher MICs of BC (P < 0.05). Two qacH-positive isolates also exhibited relatively high MICs of BC (>32 μg/mL). In previous research, the higher MICs of BC were associated with plasmid-encoded genes (Zou et al., 2014). Because each isolate harbored more than one QAC resistance gene, it is difficult to assess what level of BC resistance was contributed by each QAC resistance gene.

A previous study reported that aadA1 gene cassette was observed most commonly among the P. mirabilis isolates from retail meat products (Kim et al., 2005). In contrast, our results showed that dfrA17-aadA5 was the most common cassette array, which is similar to other reports of foodborne P. mirabilis isolates in China (Shen et al., 2011). Interestingly, an uncommon integron gene cassette array dfrA32-ereA1-aadA2 was found in one P. mirabilis isolate in this study. To the best of our knowledge, this is the first report of dfrA32-ereA1-aadA2 in foodborne P. mirabilis. The integron gene cassette array dfrA32-ereA1-aadA2 in our study showed 99.6% identity to that of Salmonella enterica (Krauland et al., 2010; GenBank accession number GU067642), 99.7% identity to that of Laribacter hongkongensis (Feng et al., 2011; GenBank accession number GU726907) and 99.7% identity to that of Aeromonas hydrophila (unpublished; GenBank accession number KJ543558). Recently, this cassette array was detected in clinical P. mirabilis isolates in Zhejiang Province of China (Wei et al., 2014), which had 99.6% identity to our sequence. The high similarity of dfrA32-ereA1-aadA2 indicated that the class 1 integrons carrying this cassette array may be located on conjugative plasmids.

Classic class 1 integrons are composed of qacEΔ1 and sul1 genes in their 3′-CS. Recently, class 1 integrons without the typical 3′-CS have been found in some species of Enterobacteriaceae, including Escherichia coli, Salmonella, Shigella sonnei and Klebsiella pneumoniae (Antunes et al., 2007; Chang et al., 2009, 2011; Sáenz et al., 2010). In these previous studies, the 3′-CS of integrons was associated with qacH and sul3. Xu et al. (2009) reported that two clinical isolates of Pseudomonas aeruginosa carried a class 1 integron lacking 3′-CS, however, the genetic structure of these non-classic integrons was unclear. In our study, non-classic class 1 integrons lacking qacEΔ1 and sul1 genes were found in 2 of the 10 intI1-positive isolates (PM4 with dfrA32-ereA1-aadA2 and PM19 with aadA2). PCR amplifications revealed that the gene cassettes in the two isolates were followed by an unusual 3′-CS linked to qacH and sul3. The gene cassette organization dfrA32-ereA1-aadA2-cmlA1-aadA1-qacH-IS440-sul3 and aadA2-cmlA1-aadA1-qacH-IS440-sul3 were identified in PM4 and PM19, respectively. Both of them contained cmlA1 coding for chloramphenicol resistance, aadA1 coding for streptomycin resistance, qacH coding for QAC resistance, insertion element IS440, and sul3 coding for sulfonamide resistance. It seems that sul3 had replaced sul1 in addition to the replacement of qacEΔ1 by qacH in the 3′-CS region. Although the gene structure cmlA1-aadA1-qacH-IS440-sul3 was commonly found in non-classic integrons (Antunes et al., 2007; Sáenz et al., 2010), the combination of dfrA32-ereA1-aadA2-cmlA1-aadA1-qacH-IS440-sul3 was reported for the first time. This novel arrangement detected in P. mirabilis isolate from food has been included in GenBank with the accession number KY426918. A similar gene cassette array of qacH-dfrA32-ereA1-aadA2-cmlA1-aadA1 was detected in Salmonella enterica serovar Stanley isolates from Taiwan (Krauland et al., 2010). Both of the two arrays carried the same gene cassettes, but the position of qacH was different. PM4 and PM19 had similar qacH-IS440-sul3-integron platforms with different gene cassette arrays, suggesting evolution of the genetic background by different recombinatorial events.

Our results demonstrated that non-classic class 1 integrons were located on conjugative plasmids of 100 kb in two tested isolates. Previous studies showed that the class 1 integrons lacking the normal 3′-CS and associated with qacH and sul3 genes were usually located on large plasmids of different size (between 70 and 240 kb) and the conjugative plasmids of 100 kb were the most disseminated among the different isolates (Antunes et al., 2007; Sáenz et al., 2010). Pal et al. (2015) also found that plasmids with co-selection potential for resistance to disinfectants and antimicrobials tended to be large and conjugative. Conjugation experiments confirmed that the disinfectant and antimicrobial resistance genes of non-classic class 1 integrons in PM4 and PM19 were co-existed on the same conjugative plasmids and could be co-transferred to E. coli. When theses isolates are exposed to QACs, selection pressure from disinfectants could increase risks for the spread of QAC and antimicrobial resistance genes among the bacteria. Notably, two transconjugants containing qacH showed the same MICs of BC as the recipient in this study. Similar observations that the presence of QAC resistance genes didn’t increase the MICs of BC in transconjugants has also been reported in a previous study (Zhang et al., 2016). There was the possibility that the agar dilution method used for susceptibility testing was not sensitive enough to detect the differences of BC MICs between the recipient and transconjugants.

Quaternary ammonium compounds resistance genes, including mdfA, ydgE/ydgF, qacE, qacEΔ1, emrE, sugE(c), and sugE(p) were found in foodborne P. mirabilis isolates in this study. Our data demonstrated the presence of non-classic class 1 integrons with the gene structure qacH-IS440-sul3 among the isolates. Moreover, qacH-associated non-classic class 1 integrons were located on conjugative plasmids and therefore could constitute an effective way for co-dissemination of antimicrobial and disinfectant resistance genes.

TY, HW, and LS designed and supervised the study. XJ, LL, YL, and KZ performed the experiments. XJ analyzed data. XJ and TY drafted the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.