Introduction

The emergence and spread of antimicrobial resistance (AMR) are of growing importance worldwide. With a population of more than 96 million and a high burden of infectious diseases, Vietnam faces since a decade a significant increase of AMR [1, 2]. Bloodstream infections (BSI) caused by multidrug-resistant bacteria frequently leads to fatal treatment failure [3, 4]. In Vietnam, BSIs are mostly caused by multidrug-resistant pathogens listed in the Global Antimicrobial Resistance Surveillance System (GLASS) report, including those classified as critical and high priority levels by the World Health Organization [5], such as extended spectrum beta-lactamase- (ESBL) and carbapenemase-resistant Enterobacteriaceae (CREs) [6, 7]. The most common cause of both community- and hospital-acquired BSI in South-East Asia is Escherichia coli, a member of the Enterobacteriaceae family [8, 9].

Wide dissemination of beta-lactamase producing E. coli, and increased incidences of hospital-acquired ESBL-producing E. coli infections have been documented [10,11,12,13,14]. In Italy, the incidences of community- and hospital-acquired infections by ESBL-producing E. coli increased between 2007 and 2015 from 29 to 42% and 25 to 35%, respectively [10]. Reports for the years 2002 until 2011 suggest an exponential increase in ESBL resistances from 20 to 45% in the Asia–Pacific region and from 39 to 55% in South-East Asia [15]. Most resistant bacterial strains were identified in patients treated at intensive care units (ICUs). In South-East Asia, Philippines and Vietnam have reported a high burden of ESBL-producing E. coli infections of 59% and 81%, respectively [16]. The most common resistance factors in clinical settings worldwide are the CTX-M-type beta-lactamases [17]. Furthermore, CRE carrying the transmissible carbapenemase genes blaKPC, blaNDM and blaOXA-48 are well characterized across various geographical regions [18]. Local epidemiological data on AMR is essential in guiding empirical antibiotic therapies to improve clinical management and outcome of BSIs, while supporting antibiotic stewardship measures to combat the spread of AMR [9]. However, epidemiological data on AMR from low- and middle-income countries (LMIC) are scarce compared to industrialized nations with a well-functioning surveillance system, so that the AMR burden in LMIC may be underestimated.

The present study aims to describe the molecular epidemiology and resistance patterns of E. coli strains isolated from Vietnamese individuals with BSI using a retrospective study cohort. We also evaluated the concordance of genotypic and phenotypic resistance in E. coli causing BSI towards molecular detection of ESBL-genes to predict cephalosporin resistance in E. coli and to optimize the management of E. coli BSI in a high ESBL-prevalent setting.

Materials and methods

Patient recruitment

The study group included patients with BSI admitted to the 108 Military Central Hospital in Hanoi, Vietnam, between October 2014 and May 2016. The 108 Military Central Hospital is a hospital of maximum care for Northern Vietnam, providing medical care for patients in Hanoi and for those being referred from various regional hospitals. In the latter case patients´ previous antibiotic treatment scheme was provided by the referring hospital. 459 BSI patients with bacterial pathogens diagnosed by blood cultures were hospitalized in the 108 Military Central Hospital during the study period. Of those, 115 patients with blood cultures positive for E. coli were included in this study. Diagnoses of BSI followed the criteria of the Surviving Sepsis Campaign guidelines [19] and the sepsis-related organ failure assessment (SOFA) score.

Blood culture

Two independent venous blood samples of approximately 8 mL per blood culture bottle were collected from both arms of the patients for blood culture with the BACTEC™ Plus Aerobic/F system (Becton–Dickinson, Franklin Lakes, NJ, USA). Blood culture bottles were incubated in the BD BACTEC™ 9120 device (Becton–Dickinson, Franklin Lakes, NJ, USA) at 36 °C for 18–72 h. Positive cultures were subjected to bacterial species identification and antimicrobial susceptibility testing using the VITEK® 2 compact automated system (BioMérieux, Lyon, France). Bacterial strains were stored with 20% glycerol at − 80 °C for further molecular testing.

Antimicrobial susceptibility testing

The minimum inhibitory concentration (MIC) for a wide spectrum of antibiotics were tested using VITEK AST-N204 cards and the VITEK® 2 compact automated system (bioMérieux, Lyon, France). The interpretation of susceptibility patterns was done according to the Clinical and Laboratory Standards Institute (CLSI) guidelines [20]. The following antimicrobial substances were included: Ampicillin (AM), Amoxicillin/Clavulanic acid (AMC), Piperacillin/Tazobactam (TZP), Cefotaxime (CTX), Ceftazidime (CAZ), Cefepime (FEP), Ertapenem (ETP), Imipenem (IPM), Meropenem (MEM), Amikacin (AN) and Ciprofloxacin (CIP).

Phenotypic confirmation of ESBL production was achieved by double disk diffusion methods following the guideline of CLSI [20]. The CTX (30 µg) and CAZ (30 µg) disks alone and in combination with clavulanate (10 µg) were used for testing of ESBL phenotype E. coli on Mueller–Hinton agar. The tests were considered positive if the difference in the zone of clearance between CTX and CTX with clavulanate or CAZ and CAZ with clavulanate was ≥ 5 mm.

Detection of beta-lactamases

Genes encoding ESBLs and carbapenemases were screened using multiplex PCR assays described previously [21]. Six genes encoding ESBL (blaSHV, blaTEM, blaCTX-M [ESBL-1]; blaVEB, blaGES, blaPER [ESBL-2]) and nine genes encoding carbapenamases (blaNDM-1, blaSPM, blaVIM [CARBA-1]; blaIMP, blaAIM, blaKPC [CARBA-2]; blaOXA-23, blaOXA-48-like, blaOXA-58 [CARBA-3]) can be detected using the multiplex PCR panels (Additional file 1: Table S1). The primers were designed to detect the most common variants of the beta-lactamase genes irrespective of the potency of the beta-lactamase activity. Bacterial isolates were incubated at 35 ± 1 °C and colonies were picked and diluted in 300 μl TE buffer, followed by DNA extraction. In detail, the suspension was immersed in 300 μl universal lysis solution containing 200 mM NaOH and 1% SDS and incubated at 95 °C for 5 min. An equal volume of 1 M Tris–HCl was added to achieve a pH of 7.5. Subsequently, the solution was transferred to a 1.5 mL Eppendorf tube and an equal volume of phenol–chloroform-isoamyl alcohol mixture (25/24/1) (ThermoFisher Scientific, Invitrogen, Waltham, MA, USA) was added. The mixture was vortexed for 5 min and centrifuged at 13,200 rpm. The upper aqueous phase was transferred to an Eppendorf tube and an equal volume of isopropanol was added. After mixing, the solution was centrifuged at 13,200 rpm for 30 min to pellet the DNA. Precipitated DNA was washed twice with 70% ethanol and reconstituted in 150 μl TE buffer (25 mM Tris with 0.1 mM EDTA). Multiplex PCR assays were performed in 25 μl reaction volumes containing hot start master mix (2×) (Promega, San Luis Obispo, CA, USA) and individual primer pairs with varying concentrations (Additional file 1: Table S1). Thermal cycling conditions were denaturation at 95 °C for 2 min followed by 40 cycles of 94 °C for 30 s denaturation, 61 °C for 30 s annealing, followed by an extension at 72 °C for 40 s and a final extension step of 72 °C for 5 min. Amplicons were visualized on 1.2% agarose gels (Additional file 1: Figure S1).

Statistical analysis

Statistical analyses were performed using the SPSS software v.23.0 (IBM Corporation, Chicago, IL, USA). Continuous variables are presented as mean ± standard deviation. Categorical variables are given as frequencies with percentages and comparisons of categorical variables between groups were performed using chi-square and Fisher’s exact tests. The level of significance was set at p-values < 0.05.

Results

Clinical characteristics of study subjects

Of the 115 patients, 70 (61%) were males. The mean age of the patients was 62 years; 62/115 patients were > 60 years old. In 73 patients (64%) pre-existing conditions (solid cancer, hypertension, diabetes, liver cirrhosis) were recorded and eight of them were receiving immunosuppressive therapy (Table 1). A primary source of infection was identified in 96 (84%) patients, with urine tract infection (40/96; 42%) and infection of the bile duct system (36/96; 38%) as the main sources, followed by respiratory and post interventional infections. The mean SOFA score was 3.36 ± 3.0 and the median procalcitonin level was 10.19 (0.24–100) ng/L. Shock and mortality rates were 17% and 16%, respectively (Table 1).

Table 1 Baseline characteristics of BSI patients with Escherichia coli infections
Full size table

Phenotypic antimicrobial susceptibility

Of the 115 E. coli isolates, 67 (58%) produced ESBLs. The phenotypic resistances of all study isolates are summarized in Table 1. The most prevalent resistance to beta-lactam antibiotics was observed for AM (109/115; 95%), followed by CTX (81/115; 70%), CAZ (52/115; 45%), AMC (32/115; 28%), FEP (28/109; 26%) and TZP (8/115; 7%) (Table 2). Three E. coli strains exhibited reduced susceptibility to ETP; one of them resistant with a MIC value of 2 mg/L, and two with intermediate susceptibility with MIC values of 1 mg/L. All isolates were phenotypically susceptible to IPM and MEM. With regard to other antibiotics, 67% (77/115) of strains were resistant to CIP, but 99% (114/115) of strains were still susceptible to AN (Table 2).

Table 2 Phenotypic resistance of Escherichia coli causing BSI in Vietnam
Full size table

Detection of beta-lactamases

The proportion of strains carrying blaCTX-M, blaTEM, blaSHV was 70% (80/115), 58% (67/115) and 1% (1/115), respectively. Of the 115 strains, 105 (91%) carried at least one of the three resistance genes (blaCTX-M, blaCTX-M, blaSHV); none of the strains carried all three genes. The combination of blaCTX-M and blaTEM was detected in 42 (37%) of the 115 strains (Table 3). 9/115 (8%) strains were positive for blaNDM and 5/115 (4%) carried the blaVIM gene.

Table 3 Concordance of genotypic and phenotypic resistance in blaCTX-M-harbouring Escherichia coli
Full size table

The blaNDM gene was detected in 2 isolates carrying the blaCTX-M gene, in 2 isolates carrying the blaTEM, and in 5 isolates harbouring both blaCTX-M and blaTEM genes. The blaVIM gene was detected in 4 isolates harbouring the blaCTX-M, and in one isolate harboring both blaCTX-M and blaTEM genes.We did not detect strains carrying blaKPC, blaVEB, blaPER, blaGES, blaIMP, blaSMP, blaAIM, blaOXA-23, blaOXA-48-like and blaOXA-58.

Concordance of resistance genotypes and phenotypic susceptibility

The concordance of genotypic and phenotypic resistance is summarized in Table 3. 14 out of 81 phenotypically CTX-resistant isolates were non-ESBL-producers according to the double disk method. However, 5 out of the 14 harboured blaCTX-M, 2 harboured blaTEM, 6 harboured both blaCTX-M and blaTEM, and one isolate did not harbour any beta-lactamase genes in our test-panel. There was a discrepancy between the detection of blaCTX-M and the phenotypic testing for ESBL, 22 out of 80 (27.5%) of isolates with blaCTX-M were classified as non-ESBL-producers by the double disk method. In our study setting, the sensitivity and specificity of blaCTX-M detection to predict the ESBL phenotype was 87% (95% CI 76–93%) and 54% (95% CI 39–68), respectively. A summary of the sensitivities and specificity of blaCTX-M detection to predict phenotypic resistance to CAZ, CTX and FEP are summarized in Table 4.

Table 4 Performance of blaCTX-M detection to predict phenotypic resistance of Escherichia coli
Full size table

Strikingly, all isolates carrying blaNDM and blaVIM were phenotypically susceptible to all carbapenems (ETP, IPM and MEM), which was unusual. The prevalence of quinolone resistance was overall very high in our isolate collection. However, there was an over-representation of quinolone resistance in blaCTX-M-carrying isolates.

bla CTX-M in cephalosporin-susceptible E. coli associated with unfavorable clinical outcome

Since not all blaCTX-M-habouring isolates exhibited phenotypic resistance to third- and fourth-generation cephalosporins tested, we performed an analysis comparing the clinical outcome (shock and fatality) of BSI with cephalosporin-susceptible E. coli with blaCTX-M as an exposure variable. Our analysis suggested that patients with blaCTX-M-negative E. coli BSI had better clinical outcomes than patients with blaCTX-M-positive E. coli BSI. In particular, the presence of blaCTX-M in phenotypically cephalosporin-susceptible E. coli BSI is associated with fatal outcome of the infection for CTX-susceptible phenotype (4% versus 27%, p = 0.05), and with a statistical trend for ESBL-negative phenotype (8% versus 27%, p = 0.07), for CAZ-susceptible phenotype (4% versus 20%, p = 0.06) for ESBL-negative phenotype (4% versus 27%, p = 0.05), for FEP-susceptible phenotype (6 versus 19%, p = 0.1) (Fig. 1, Additional file 1: Table S2). In addition, the duration of hospital stay was significantly longer in patients with CTX-M-positive E. coli infections than in patients with CTX-M-negative E. coli BSI (Additional file 1: Table S2). For discrepant genotypic-phenotypic carbapenemase susceptibility, there was no association between the discrepancy and the clinical outcome (neither shock nor death). This analysis was performed with very small number of cases so that further investigations are needed before drawing any definitive conclusions.

Fig. 1
figure 1

The contribution of blaCTX-M to the outcome of BSI patients caused by cephalosporin susceptible Escherichia coli. a For ratio of shock and b for fatality. Red column is ratio of shock (a) or death (b) of BSI patients caused by blaCTX-M positive E. coli and green column is ratio of shock (a) or death (b) of BSI patients caused by blaCTX-M negative E. coli

Full size image

Discussion

In this study, we reported a high prevalence (58%) of ESBL-producers in E. coli causing BSI in Vietnamese patients. Our findings are in accordance with previously published data on the epidemiology of ESBL-producing E. coli from Vietnam, which suggested a comparable (48–55%) prevalence [9, 16, 22]. Similar findings have been reported from studies conducted in in China (56–67%) [22, 23] and other regions in South-East Asia, such as Thailand (50%) [22]. In comparison, studies from industrialized (middle to high income) countries reported lower rates of ESBL-producers in clinical E. coli strains. A study from Canada has reported a far lower rate of 4% ESBL-producing E. coli strains [24]. Another study conducted in Sweden reported similar low rates of ESBL-producing E. coli strains with 2%, 3% and 4% of cases in 2007, 2009 and 2011, respectively [25]. Taken together, these discrepancies illustrate and highlight the increasing AMR problematic in the Asian region and in LMIC, in general. Surveillance, antibiotic use and higher compliance with antibiotic stewardship might account for these differences [26]. In many South-East Asian countries including Vietnam, antibiotics, especially of broad-spectrum cephalosporins can be purchased over the counter without prescription. Thus, inappropriate indications for antibiotic therapy may have an influence on the high rates of resistance [27, 28].

Almost all (97%) ESBL-producing E. coli in our study harboured at least one beta-lactamase gene, with blaCTX-M and blaTEM as the most common, consistent with published studies from Thailand and Vietnam [29, 30]. Indeed, among the 300 ESBL subtypes [31], the TEM-, SHV- and CTX-M-types are predominant and of particular relevance [32]. The CTX-M-type beta-lactamases play a major role as emerging and the most widely spread resistance factors in Enterobacteriaceae on a global scale [17]. Strikingly, we found a significant proportion (46%) of non ESBL-producers, as determined by phenotypic testing for ESBL, carrying the blaCTX-M gene. Since our multiplex PCR was designed to detect the most common CTX-M types, we cannot rule out that this discrepancy is attributed to CTX-M subtypes with weaker beta-lactamase activity. Overall, the concordance between phenotypic and genotypic susceptibility was good for blaCTX-M-carrying isolates. However, the presence of blaCTX-M alone cannot predict phenotypic resistance to cephalosporins perfectly. Not all CTX-M subtypes are equally potent; depending on the expression, the beta-lactamase activity levels may be variable [17, 33]. Furthermore, the presence of insertion sequences, such as ISE cp1B and IS903D, can modify the expression of the ESBL genes [34]. This highlights the necessity of implementing high-resolution molecular typing methods, such as whole-genome sequencing, to study the molecular epidemiology of AMR and multidrug-resistant pathogens in high-prevalent settings.

Carbapenem-resistance is low in E. coli causing BSI in our study population. Our study shows 98% ETP and 100% IPM and MEM susceptibility of E. coli isolates. However, we detected the presence of the Ambler class B metallo-beta-lactamase in 12 isolates, 8 with blaNDM and 4 with blaVIM. Oddly, the presence of these genes did not correlate with the phenotypic resistance to any carbapenems tested, which we could not explain. Indeed, discrepancies between genotypic und phenotypic resistances have been described previously [35]. The primers for both blaNDM and blaVIM were not specific for a particular variant; therefore, it is possible that nucleotide variations leading to a non-functional protein cannot be ruled out. In addition, the expression of beta-lactamase genes may be affected by promoter regions or other genetic element of the plasmid carrying these AMR genes [36]. Nonetheless, the high susceptibility rates for carbapenems and TZP indicate that these substances may be appropriate for the empirical antibiotic therapy of E. coli BSI in Vietnam. Of note, CIP resistance rate of 67% can be considered high in comparison to other studies [37], which may have been the result of frequent use of CIP.

The molecular detection of blaCTX-M in non-ESBL and cephalosporin-susceptible E. coli was significantly associated with a worse clinical outcome (septic shock and fatality) of E. coli BSI. This finding was unexpected since phenotypic susceptibility is considered as the gold standard for guiding antimicrobial therapies. The increased expression of antimicrobial resistance genes under antibiotic pressure may be an explanation for this observation [38]. Exposure to an antimicrobial substance may induce stress response (SOS-response) in bacteria, which causes several genes including AMR genes, such as blaCTX-M to be up-regulated thus increasing tolerance to various substance classes (cross-resistance) [39, 40]. In such cases, antibiotic therapy guided by phenotypic resistance only may not be suitable (i.e., false susceptible) and, thus, may lead to a poor outcome as a consequence of undertreatment. In this study cohort, we did not observe any significant differences in the clinical outcome for cases with discrepant phenotypic and genotypic carbapenem susceptibility. Due to the small number of cases (n = 14), we would refrain from drawing any conclusions. Nonetheless, this finding is unexpected and warrants further scrutiny. The time to antibiotic therapy initiation may influence the outcome of an infection. Therefore, the acceleration of time to report is of particular interest in the routine microbiological diagnostics [41]. Conventional blood culture diagnostics often take more than 24–48 h until antibiotic susceptibility profiles are available for the attending physicians, the introduction of genotypic PCR-based AMR detection may accelerate this process and thus potentially increasing the accuracy of calculated empirical antimicrobial therapy [42]. However, further validation in a larger cohort is needed prior to implementation in the routine setting.

Our study has limitations, the small sample size, especially for the clinical outcome analysis, may limit the generalizability of our findings. Nonetheless, with our small cohort, we could show that the presence of blaCTX-M in cephalosporin-susceptible E. coli correlated with a poor outcome and warrant further investigations. Furthermore, our PCR panel encompasses multiple variants of the beta-lactamase genes. The primers were designed to detect the most common variants of the respective beta-lactamase genes, which may have resulted in discordance in the genotypic-phenotypic susceptibility for low-activity beta-lactamases, such as blaTEM-1, blaTEM-2 and blaSHV-1 [43]. Due to this limitation, the study focused on blaCTX-M.

Taken together, our data contributes to epidemiological data on the AMR burden in South-East Asia, which needs attention and should be closely monitored. The detection of AMR gene does not necessarily correlate with the phenotypic resistance in E. coli and warrant further scrutiny. Easy to implement PCR-based AMR detection method may have added benefit, especially in high-prevalent settings, to accelerate, optimize and guide antimicrobial therapy. The surveillance of the spread of multidrug resistance in LMIC is still suboptimal and access to high-resolution molecular typing methods may help combat the spread of AMR genes on a global scale.