Escherichia coli (E. coli) is a major causative organism of complicated urinary tract infections, bloodstream infections, and pneumonia. Due to the advent of antimicrobial resistance, therapeutic options have become limited. The carbapenems are sometimes the only effective regimens against the severe infection caused by drug-resistant E. coli. However, with the widespread use of antimicrobial agents, the prevalence of carbapenem resistance in E. coli has increased. The reported occurrence of carbapenem-non-susceptible or -resistant E. coli varies by region. The prevalence rate of carbapenem-non-susceptible E. coli was 0–3% among the 12 Asia-Pacific nations (Mendes et al., 2013), 4.3% among hospitals in the US (McCann et al., 2018; Castanheira et al., 2020), and 0–7% in Europe (Kostyanev et al., 2019). According to surveillance conducted in Taiwan in 2009, the rate of carbapenem non-susceptibility in E. coli was 1.6–7.1% (Jean et al., 2013). The proportion of carbapenem-non-susceptible and -resistant E. coli has been increasing (Martirosov and Lodise, 2016). Moreover, patients infected with carbapenem-resistant pathogens, when compared to those infected with susceptible pathogens, showed increased morbidity and mortality (Nordmann and Poirel, 2019). The crude mortality of patients with carbapenem-resistant Enterobacterales infections is estimated to be 70% (Friedman et al., 2017). Carbapenem resistance increases the medical cost and causes an enormous economic burden (Bartsch et al., 2017).

Fluoroquinolone remains a choice in carbapenem-resistant E. coli (CREc) if the bacteria is found to be susceptible to the drug. Despite robust studies on fluoroquinolone-resistant mechanisms of E. coli, few studies specifically addressed the CREc group.

Two important mechanisms of resistance are observed in CREc. One is mediated by the carbapenemase genes, and the isolates are termed as carbapenemase-producing carbapenem-resistant E. coli (CP-CREc). They are classified according to the Ambler classification into three groups: class A, class B, and class D enzymes (Ambler, 1980). The mechanism of resistance in non-carbapenemase-producing carbapenem-resistant E. coli (non-CP-CREc) is mediated by ESBL or AmpC beta-lactamase associated with an overexpression of efflux pumps (such as AcrAB) or a loss of expression of porin (such as OprD and OprM) (Chetri et al., 2019; Theuretzbacher et al., 2020).

The prevalence of CP-CREc has increased in recent years in Taiwan (Jean et al., 2013), and the reported types of carbapenemases in CP-CREc is also increasing (Dortet et al., 2017). However, some mechanisms that are responsible for carbapenem resistance in E. coli are not well understood. In addition, the phylogenetic relationship, clinical phenotypes, and resistome characteristics still warrant further investigation. Whole-genome sequencing (WGS) provides a comprehensive method to identify the mechanisms underlying carbapenem resistance in E. coli, in particular non-CP-CREc, such as porin-encoding gene alternation along with ESBL or AmpC (Larkin et al., 2020). In this study, we aim to use WGS to characterize the molecular resistance mechanisms, virulence determinants, and phylogenetic relatedness of CREc.

Carbapenems are the last resort to treat multidrug-resistant Gram-negative E. coli. With the increasing use of carbapenems, the prevalence of carbapenem-resistant E. coli is growing and poses a threat to public health and challenges to physicians. Only a few antibiotics are left for physicians to treat severe infectious diseases (Huang et al., 2021a). An understanding of the resistance mechanisms is essential to therapy and improved control measures. In this study, we demonstrated a high level of genetic diversity in CREc in both community and hospital settings. A previous study found that 0.04–29.5% of the CRE isolates were associated with the community (Kelly et al., 2017). Another report from Taiwan identified that 29.5% of CRE cases were community-acquired infections (Tang et al., 2016). In line with this study, CRE can be acquired from the community and hence poses an urgent public health threat.

In our study, the majority of CREc are not non-carbapenemase-producing strains. Two main molecular mechanisms contribute to carbapenem resistance in Enterobacteriaceae. The resistance is mediated either by carbapenemase production or by ESBL or AmpC in combination with increased efflux pump activity or porin alternation, as observed in the cases of non-carbapenemase-producing CRE. Among carbapenem-resistant and non-susceptible E. coli, the proportion of carbapenemase-producing strains varies across different regions of the world, as shown in Figure 3. Despite the fact that the definition of carbapenem non-susceptibility or resistance slightly varied between the studies, a trend toward an increasing number of CP-CREc cases is observed. The predominant carbapenemase varies in different geographic areas, as shown in Figure 3. Among the E. coli isolates that were non-susceptible/resistant to carbapenem in Taiwan, the rate of CP-CREc occurrence increased from 1.4% during 2010–2012 (Wang et al., 2015), to 7.59% during 2012–2015 (Chang et al., 2019), and to 29.5% during 2016–2018 (Huang et al., 2021a). According to previous studies in Taiwan, NDM-1 is the most common carbapenemase and accounts for 39.1% of CP-CREc (Chang et al., 2019; Huang et al., 2021a). In this study, 30% of the CREc are CP-CREc. Among the CP-CREc strains, OXA-48 and NDM-1 were presented. The data provide insight into the molecular characteristic of CREc in Central Taiwan.

In this study, we detected OmpC D192G and OmpF 264-269 deletions using comparative genomics in non-CP-CREc. Porins, which are outer membrane proteins, block antibiotic permeation into bacteria and contribute to antibiotic resistance. Mutation in porins mediates antibiotic resistance by three different mechanisms: an alteration in the electrostatics of the channel, a reduction in the size of the porin channel, or a decrease in the expression level of porins or loss of porins (Fernández and Hancock Robert, 2012). In E. coli, two modifications of porin were observed: loss of OmpC and OmpF and mutation in OmpC (Vergalli et al., 2020). Lou et al. (2011) and Bajaj et al. (2016) proved that a single mutation in OmpC proteins could decrease the permeability of an antibiotic, by using liposome permeation assays and single-channel electrophysiology. OmpC D192G has been identified in a previous study and verified by PROVEAN (Al-Farsi et al., 2020; Tian et al., 2020). To our best knowledge, OmpF 264-269 deletion has not been reported before. OmpF 264-269 deletion might contribute to carbapenem resistance. Further studies regarding the impact of these mutations on the expression level are warranted.

In the present study, 10 isolates belonged to eight different ST types, indicating polyclonal spreading. Previous studies reported that 55.6% of NDM-positive E. coli isolates were classified as ST410 type (Huang et al., 2021a), and among carbapenem non-susceptible E. coli, ST131 is the predominant ST type in Taiwan (Ma et al., 2013; Mathers et al., 2015). The NDM-1-producing E. coli in the present study belonged to ST167, which differed from the results of previous studies. These findings suggest that pathogenic E. coli in Taiwan is polyclonal and highly genetically diverse.

The NDM-1 gene identified in this study is located within the chromosome, and the isolate belongs to ST167. Usually, bla_NDM genes are located in various types of plasmids, such as IncA/C, IncF, IncNIncL/M, and untypable/IncR (Poirel et al., 2011; Khan et al., 2017). Shen et al. reported the first case of a chromosomally integrated bla_NDM−1 E. coli which originated in China. The bla_NDM−1 gene was inserted between two tandem copies of an insertion sequence common region 1 element, which originated from the plasmid of Proteus mirabilis. The chromosomally encoded bla_NDM−1 E. coli strain belonged to ST167, which is identical to the bla_NDM−1 E. coli strain of this study (Shen et al., 2017). In this study, we demonstrated that bla_NDM−1 in E. coli can be chromosomally integrated and transmitted vertically. Close surveillance and monitoring are warranted to prevent the spread of the bla_NDM−1 chromosomally integrated E. coli strains.

We also searched for the mechanism of quinolone resistance among the five quinolone-resistant CREc, and no qnr-like, horizontally transferred resistant genes were detected by BLAST in all the plasmids. We found GyrA D87N mutation within the chromosome in all the quinolone-resistant CREc strains in this study.

Whole-genome sequencing is an emerging technology to investigate molecular resistance mechanisms. While most previous studies used PCR to identify ESBL and AmpC beta-lactamase (Chia et al., 2009; Dupont et al., 2016; Ho et al., 2016; Tian et al., 2020), some studies employed WGS to identify molecular resistant mechanisms in non-carbapenemase-producing CREC (non-CP-CREc)/carbapenem non-susceptible E.coli (non-CP-CNEc) (Dagher et al., 2018; Senchyna et al., 2019; Al-Farsi et al., 2020). The major mechanisms of the non-CP-CRE/CNEC are AmpC or ESBL β-lactamase in combination with alternation or loss of an outer membrane porin (OmpC and OmpF), as shown in Figure 4. In Taiwan, plasmid-encoded AmpC β-lactamase CMY-2 and DHA-1 combined with OmpC/F loss was the primary mechanism (Chia et al., 2009; Ma et al., 2013). Despite intensive surveys, an average of 16% of the non-CP-CREc/CNEc isolates in the literature cannot be fully explained by the above-mentioned mechanism. In the present study, we found OmpF 264-269 deletion and verified via the online PROVEAN platform that this deletion was deleterious. There are still knowledge gaps regarding the mechanism underlying non-CP-CREc. Further studies on the efflux pump and regulatory protein, which regulate the expression of outer membrane porin, may improve our understanding of these issues.

A previous study demonstrated that CP-CRE strains had greater potency to hydrolyze carbapenems than non-CP-CRE strains and were more likely to resist other non-β-lactam antibiotics (Tamma et al., 2017). However, in our study, we could not identify differences in the MIC distribution of carbapenems. We did find that the susceptible rate of ciprofloxacin was lower in CP-CREc (33.3%) than in Non-CP-CREc (57.1%). Previous studies have proved that the mode of CREc acquisition differs between CP-CREc and non-CP-CREc isolates. The mode of acquisition of non-CP-CREc was through an endogenous pathway, which involves selective pressure of antibiotics on the intestinal microbiota and is primarily associated with recent exposure to antibiotics. While for CP-CREc, the mode of acquisition was primarily exogenous through “patient-to-patient” horizontal spread (Goodman et al., 2016; van Duin and Doi, 2017; Bouganim et al., 2020). By investigating the molecular mechanisms of CREc, we can better predict the risk factors among patients who are at risk of being infected with CREc and prevent CREc.

All E. coli strains in the present study harbor gad, terC, and traT. In this study, we found that uropathogenic E. coli strains (n = 5, totally contain 95 virulence factors), compared with non-uropathogenic E. coli strains (n = 5, totally contain 69 virulence factors), had higher frequencies of virulence genes. Extra-intestinal pathogenic E. coli (ExPEC) are multifaceted pathogens that can cause infections at extra-intestinal sites. The ExPEC group includes avian pathogenic E. coli, neonatal meningitis E. coli, uropathogenic E. coli (UPEC), and sepsis-associated E. coli (SEPEC). ExPEC are pathogenic to humans due to the presence of a constellation of virulence genes. A previous study discovered that E. coli that caused IAIs had fewer virulence genes than those that caused UTIs or primary bacteremia (Chen et al., 2021). It was postulated that during IAI, enteric E. coli may inadvertently translocate from the intestine to the bloodstream. A previous study had suggested that CP-CRE may be more virulent than non-CP-CRE and are associated with poorer outcomes (Tamma et al., 2017). Analysis has indicated zoonotic origin for ExPEC, as evident by the fact that avian and human E. coli isolates encompass similar virulence genes and that they belong to the same phylogenetic groups. This finding reminds us that to control the spread of ExPEC, poultry product surveillance and monitoring are important steps (Sarowska et al., 2019).

This study has several limitations. First, a genotype is not equal to a phenotype. Isolates harboring carbapenemase-producing genes are not identical to the isolates that express carbapenemase. Second, we did not conduct a drug susceptibility test for tigecycline and colistin. Third, the efflux pump was not analyzed in this study. Fourth, no function assessment of porin mutation was done in this study. Fifth, we defined community- and hospital-acquired CRE according to the previous studies (Garner et al., 1988; Weinstein et al., 1997). Thus, a patient with multiple exposures to the healthcare system can be classified into community-acquired CRE. Sixth, we could not explain the carbapenem-resistant mechanisms underlying three CREc isolates.

Complete genome sequencing detected that gyrA D87N mutation mediated fluoroquinolone resistance in all the CREc isolates in this study, suggesting a common mechanism. We also identified that OmpF 264-269 deletions might contribute to carbapenem resistance in CREc isolates. This study further provides insight into the clonal diversity, resistant mechanisms, and virulence factors of CREc strains isolated from patients with bacteremia in Taiwan. In E. coli, Bla_NDM−1 mutation can be observed in patients with community-acquired bacteremia and in patients without risk factors for CRE infection. Combined clinical and genomic analyses indicate the highly diverse nature of the CREc in Taiwan.

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 below: https://www.ncbi.nlm.nih.gov/genbank/, CP084378.

W-TY, P-YL, and Y-TH contributed to the conception and design of the study and wrote the first draft of the manuscript. W-TY and P-YL contributed to the collection of clinical isolates. P-YL and Y-TH contributed to the experiments. Y-TH performed the bioinformatic analysis. All authors contributed to revising the manuscript and have approved the submitted version.

Y-TH was supported in part by the Ministry of Science and Technology (109-2221-E-194-038-MY3). P-YL was supported in part by the Ministry of Science and Technology (110-2314-B-075A-011) and the Taichung Veterans General Hospital (TCVGH-1113901C, TCVGH-1113901D, TCVGH-NK1109004, TCVGH-NK1099002, and TCVGH-1093902D).

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.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.