Molecular Epidemiology of Third-Generation-Cephalosporin-Resistant Enterobacteriaceae in Southeast Queensland, Australia

ABSTRACT

Third-generation cephalosporin-resistant (3GC-R) Enterobacteriaceae represent a major threat to human health. Here, we captured 288 3GC-R Enterobacteriaceae clinical isolates from 264 patients presenting at a regional Australian hospital over a 14-month period. In addition to routine mass spectrometry and antibiotic sensitivity testing, isolates were examined using rapid (∼40-min) real-time PCR assays targeting the most common extended-spectrum β-lactamases (ESBLs; bla_CTX-M-1 and bla_CTX-M-9 groups, plus bla_TEM, bla_SHV, and an internal 16S rRNA gene control). AmpC CMY β-lactamase (bla_CMY) prevalence was also examined. Escherichia coli (80.2%) and Klebsiella pneumoniae (17.0%) were dominant, with Klebsiella oxytoca, Klebsiella aerogenes, and Enterobacter cloacae infrequently identified. Ceftriaxone and cefoxitin resistance were identified in 97.0% and 24.5% of E. coli and K. pneumoniae isolates, respectively. Consistent with global findings in Enterobacteriaceae, most (98.3%) isolates harbored at least one β-lactamase gene, with 144 (50%) harboring bla_CTX-M-1 group, 92 (31.9%) harboring bla_CTX-M-9 group, 48 (16.7%) harboring bla_SHV, 133 (46.2%) harboring bla_TEM, and 34 (11.8%) harboring bla_CMY genes. A subset of isolates (n = 98) were subjected to whole-genome sequencing (WGS) to identify the presence of cryptic resistance determinants and to verify genotyping accuracy. WGS of β-lactamase-negative or carbapenem-resistant isolates identified uncommon ESBL and carbapenemase genes, including bla_NDM and bla_IMP, and confirmed all PCR-positive genotypes. We demonstrate that our PCR assays enable the rapid and cost-effective identification of ESBLs in the hospital setting, which has important infection control and therapeutic implications.

INTRODUCTION

Third-generation cephalosporin-resistant (3GC-R) members of the Enterobacteriaceae represent a significant threat to human health due to their ability to rapidly transmit antimicrobial resistance (AMR) determinants within and among bacterial populations (1). Both infection and colonization with these globally disseminated organisms are associated with poor clinical outcomes and death (2). 3GC-R Enterobacteriaceae prevalence rates vary considerably among hospitals and countries, with particularly high rates in India, Asia, and the Middle East (3). Australia has historically observed a relatively low frequency of 3GC-R Enterobacteriaceae. However, this pattern may be changing, with an extended-spectrum β-lactamase (ESBL) phenotype detected in 13.3% of Escherichia coli isolates and 9.8% of Klebsiella pneumoniae isolates in blood culture in 2018 (4), up from 2013 rates of 7.6% and 6.3%, respectively. Moreover, ceftriaxone resistance was seen in 13.4% of E. coli isolates and 9.4% of K. pneumoniae isolates, with 86.3% and 82.6% containing CTX-M-type ESBL genes, respectively (4).

AMR in 3GC-R Enterobacteriaceae is generally conferred by the presence of an ESBL or the overexpression of a chromosomal or plasmid-borne AmpC β-lactamase gene (5). CTX-M-type ESBLs are currently the most common AMR determinants found in 3GC-R Enterobacteriaceae isolated in Australian hospitals, with bla_CTX-M-1 and bla_CTX-M-9 groups collectively comprising 86.2% of identified CTX-M-type enzymes (4). Detection of ESBL subtypes is not standard clinical practice; instead, their presence is typically inferred from antimicrobial susceptibility patterns and phenotypic confirmatory tests, which are laborious, costly (∼$AUD25/isolate), and time-consuming (>24 h) to perform (6). Use of molecular-based diagnostics for AMR determinant detection in the clinical setting would improve patient outcomes by enabling more rapid, directed, and personalized treatment strategies, thereby improving antibiotic stewardship measures and reducing AMR burden. Their utility has already been demonstrated in a number of clinical settings, including for bloodstream infections, pneumonia, and infection control screening (7–9).

In the current study, we prospectively collected all 3GC-R Enterobacteriaceae identified over a 14-month period at the Sunshine Coast University Hospital, Queensland, Australia, a 450-bed, regional teaching and tertiary hospital. Using large-scale comparative genomics, we developed a novel quadriplex and two singleplex real-time PCR assays to rapidly detect the most commonly identified β-lactamase (including ESBL) genes in 3GC-R Enterobacteriaceae. In association with whole-genome sequencing (WGS), we used our diagnostic real-time assays to determine the molecular epidemiology of the β-lactamase genes in our isolate set and compared their performance with both genotypic and phenotypic typing methods to validate assay accuracy, sensitivity, and specificity. Finally, we examined the population structure of our 3GC-R Enterobacteriaceae population to identify potential cases of nosocomial spread or outbreak clusters.

RESULTS AND DISCUSSION

ESBL enzymes have established themselves as major contributors of AMR in Gram-negative bacteria and are increasing in prevalence, particularly in the community setting. Over 14 months, we collected a total of 288 isolates from all diagnosed cases of 3GC-R Enterobacteriaceae in a regional hospital in Southeast Queensland, Australia. Our patient cohort was predominantly female (64.0%) and aged over 65 years (60.9%). Specimens from intensive care unit (ICU) patients comprised 54 (18.8%) of the total collection and were obtained from 48 ICU patients over a 12-month period. Five ICU patients had multiple isolates collected over the course of their stay in the ICU. Single species with identical AMR profiles were isolated from three patients, and multiple species were isolated from two patients, one with identical AMR profiles and one with differing profiles (see Table S1 in the supplemental material). E. coli and K. pneumoniae were dominant, comprising 80.2% and 17.0% 3GC-R strains, respectively (Table 1 and Table S1). The remaining isolates were identified as Enterobacter cloacae (1.4%; n = 4), Klebsiella oxytoca (1.0%; n = 3), and Klebsiella aerogenes (0.3%; n = 1). Resistance to ceftriaxone, ceftazidime, or cefepime was common and observed in 96.2%, 93.4%, and 86.8% of isolates, respectively. In contrast, cefoxitin resistance was noted in only 81 (28.1%) isolates (Table 1). Meropenem resistance was rare, with only six resistant strains (2.1%) identified. The double disc diffusion method for ESBL phenotypic testing was negative in 32 (11.1%) isolates, indicating a lack of β-lactamase inhibition in the presence of clavulanic acid. The remaining 256 (88.9%) isolates were positive for this phenotypic test.

TABLE 1.

Summary of antimicrobial resistance mechanisms among 288 clinical Enterobacteriaceae isolates^a

Resistance mechanism	No. (%) of isolates
	E. coli (n = 231)	K. pneumoniae (n = 49)	K. aerogenes (n = 1)	K. oxytoca (n = 3)	E. cloacae (n = 4)
Phenotypic antimicrobial resistanceb
Ceftriaxone	223 (96.5)	46 (93.9)	1 (100)	3 (100)	4 (100)
Ceftazidime	215 (93.1)	47 (95.9)	1 (100)	2 (66.7)	4 (100)
Cefepime	201 (87.0)	43 (87.8)	1 (100)	2 (66.7)	3 (75.0)
Cefoxitin	64 (27.7)	12 (24.5)	1 (100)	0 (0)	4 (100)
Piperacillin-tazobactam	65 (28.1)	31 (63.3)	1 (100)	3 (100)	4 (100)
Ticarcillin-clavulanate	186 (80.5)	46 (93. 9)	0 (0)	3 (100)	3 (75.0)
Meropenem	3 (1.3)	2 (4.1)	0 (0)	0 (0)	3 (75.0)
ESBL phenotypic testc
Positive	204 (88.3)	46 (93.9)	0 (0)	2 (66.7)	4 (100)
Negative	27 (11.7)	3 (6.1)	1 (100)	1 (33.3)	0 (0)
ESBL and ampC gene prevalence
blaCTX-M-1	104 (45.0)	36 (73.5)	0 (0)	2 (66.7)	2 (50.0)
blaCTX-M-9	89 (38.5)	4 (8.2)	0 (0)	0 (0)	0 (0)
blaSHV	9 (3.9)	38 (77.6)	0 (0)	0 (0)	1 (25.0)
blaTEM	93 (40.3)	33 (67.3)	0 (0)	2 (66.7)	2 (50.0)
blaCMY	30 (13.0)	3 (6.1)	0 (0)	0 (0)	1 (25.0)
PCR negatived	7 (3.0)	1 (4)	1 (100)	1 (33.3)	1 (25.0)

^aIsolates were retrieved from 264 patients admitted to a Southeast Queensland hospital over a 14-month period.

^bAntimicrobial susceptibility testing was performed by an automated method (Vitek 2), and EUCAST breakpoints were applied.

^cESBL phenotypic testing was performed by the disc diffusion method.

^dDNA integrity was confirmed by 16S rRNA gene PCR (23), which forms part of the quadriplex assay. Three PCR-negative E. coli strains carried bla_DHA-1, one E. cloacae carried bla_ACT-17, one K. oxytoca carried bla_OXY-2-1, a second K. oxytoca (included as a mixture with K. pneumoniae) carried bla_OXY-2-4, and 5 of 11 isolates had no β-lactamase genes identified despite whole-genome sequencing.

Redesign of current ESBL PCR assays to improve specificity and detection rates.

We next set out to identify the abundance of 3GC-R AMR determinants in our isolate collection by using real-time PCR. The rapid expansion of WGS data in recent years has enabled improvements to be made upon assays designed in the pre-WGS era, which risk being either nonspecific or intolerant to sequence diversity within enzyme families. Given a sufficiently large and diverse data set, sequence diversity can result in a nontrivial proportion of false-negative calls, skewing epidemiological observations. Examination of previously published ESBL assays (10) identified several mismatches in primer/probe pairs, with the mismatches especially problematic for the CTX-M enzymes. To overcome probable issues with false-negative results using existing assays, we exploited the extensive volume of publicly available sequence data to identify conserved sequence regions suitable for assay design. Our redesigned assays target the genes encoding the most common ESBLs (CTX-M group 1, CTX-M group 9), along with SHV and TEM β-lactamases, which can evolve to become ESBLs through mutations in their active sites or contribute to clinically relevant AMR due to increased expression. Our redesigned assays thus represent a considerable improvement over the accuracy of existing ESBL assays by avoiding or accommodating sequence mismatches at primer- and probe-binding sites.

ESBL and β-lactamase assay performance.

Preliminary validation of our singleplex (bla_TEM, bla_CMY) and quadriplex (bla_CTX-M-1 group, bla_CTX-M-9 group, bla_SHV, 16S rRNA gene) PCR assays was performed to determine their accuracy and specificity. Following in silico verification, assays were tested on 12 previously characterized control strains (Table S2) (11) that represented a range of Enterobacteriaceae containing the gene targets of interest. All expected targets were identified in this control data set; no false positives or false negatives were detected, although both the bla_TEM and 16S rRNA gene assays yielded late (>30 cycles) amplification in no-template control wells (Fig. 1). This result was expected due to the presence of residual E. coli DNA in the recombinant SsoAdvanced Taq polymerase, which carries both bla_TEM and 16S rRNA genes.

FIG 1.

Singleplex and quadriplex PCR performance on third-generation cephalosporin-resistant Enterobacteriaceae. 1, Enterobacter cloacae SCHI0016.S.104 containing CMY; 2, Escherichia coli SCHI0016.S.205 containing TEM; 3A, E. coli SCHI0016.S.187 positive for CTX-M group 1; 3B, E. coli SCHI0016.S.60 positive for both CTX-M groups 1 and 9; 3C, E. coli SCHI0046.S.46; 3D, E. coli with only 16s rRNA gene amplification with no CTX-M or SHV detected. Note that both 16S rRNA gene and TEM assays have late amplification in the no-template controls (NTC) due to low levels of recombinant E. coli DNA present in the SsoAdvanced Taq polymerase.

We found that all genotypes could be accurately determined for the control strains by directly inoculating PCR wells with a very small amount of bacterial culture, thereby obviating the need for DNA extraction and leading to a faster time to diagnosis. However, this technique requires skill to ensure that the PCR is not inhibited by excess DNA and cellular constituents and can lead to greater PCR failures if not performed correctly. In addition, we recommend using a 10-μl rather than a 5-μl total PCR volume, as, in our hands, the smaller reaction volume led to a greater failure rate. Inclusion of the 16S rRNA gene assay enables true negatives to be differentiated from PCR failures. This method provides an attractive option where rapid diagnosis is paramount or where DNA extraction is not feasible.

Next, we screened our PCR assays across a diversity panel of non-Enterobacteriaceae human pathogens or colonizers (n = 48 strains). Our assays showed 100% sensitivity and specificity for all targets, with no amplification detected in these isolates, except for the 16S target in bacterial strains; as with the control data set, the bla_TEM PCR was inhibited by the presence of non-bla_TEM-harboring DNA and therefore did not amplify in any diversity panel strains.

LoQ and LoD determination of the ESBL and β-lactamase assays.

Using serial dilutions of previously characterized DNA samples ranging from 40 ng to 0.4 fg, the limit of quantitation (LoQ) values for each assay were identified (Fig. 2) as follows: bla_CMY, 400 fg (∼73 genome equivalents [GEs]); bla_TEM, 400 fg (∼65 GEs); 16S rRNA gene, 40 fg (∼3 GEs); bla_CTX-M-1 group, 4 pg (∼344 GEs); bla_CTX-M-9 group, 4 pg (∼344 GEs); and bla_SHV, 4 pg (∼344 GEs). Limit of detection (LoD) values for each assay were as follows: bla_CMY, 4 fg (∼7 GEs); bla_CTX-M-1 group, 4 pg (∼344 GEs); bla_CTX-M-9 group, 400 fg (∼34 GEs); and bla_SHV, 400 fg (∼34 GEs). Due to low levels of recombinant E. coli DNA present in the SsoAdvanced Taq polymerase, which results in late amplification in no-template control wells, LoD values for the bla_TEM and 16S rRNA gene assays could not be determined. As expected, these results indicate inferior performance of the quadriplex assay in comparison with the singleplex assays. Although not examined in this study, increased LoQ and LoD values may be achieved for the bla_CTX-M-1 group, bla_CTX-M-9 group, and bla_SHV assays by omitting the 16 rRNA gene assay or by running these assays in singleplex format. Although we have defined the limits of our assays by using pure bacterial templates, it remains to be determined what detection limits would be achieved from clinical specimens.

FIG 2.

Limits of detection (LoD) and quantitation (LoQ) for the extended-spectrum β-lactamase real-time PCR assays developed in this study. The bla_CMY and bla_TEM assays were examined as singleplex PCRs, whereas the 16S rRNA gene, bla_CTX-M-1 group, bla_CTX-M-9 group, and bla_SHV assays were examined in quadriplex format. RFU, relative fluorescence units.

ESBL and β-lactamase prevalence in clinical 3GC-R Enterobacteriaceae isolates.

Following specificity confirmation, the PCR assays were tested across our clinical set (n = 288). All assays performed well, with 100% concordance between PCR genotypes and WGS of 98 isolates and with no false negatives identified (Table S3). Overall, a β-lactamase genotype was identified in 277 (96.2%) isolates. Consistent with previously reported rates (11), the bla_CTX-M-1 group was most common, being identified in 144/288 (50%) isolates, including 105/231 (45.5%) E. coli and 36/49 (73.5%) K. pneumoniae isolates. The bla_CTX-M-9 group was next most frequent, being found in 93/288 (32.3%) isolates, including 89/231 (38.5%) E. coli and 4/49 (8.2%) K. pneumoniae isolates. The presence of both bla_CTX-M-1 group and bla_CTX-M-9 group enzymes was identified in five (1.7%) E. coli isolates and one (2.0%) K. pneumoniae isolate. Of the eight isolates belonging to other species, bla_CTX-M-1 and bla_CTX-M-9 group genes were present in 4/8 (50%) and 0/8 (0%) of these isolates, respectively. The acquired ampC gene, bla_CMY, was identified in 30/231 (13%) E. coli isolates, 3/49 (6.1%) K. pneumoniae isolates, and 1/8 (13%) isolates of other Enterobacteriaceae species. bla_CMY and bla_CTX-M-1 were both identified in seven isolates (five E. coli, one K. pneumoniae, and one E. cloacae isolate), and bla_CMY and bla_CTX-M-9 were present in two E. coli isolates only (Table 1).

Seven strains (five E. coli and two K. pneumoniae strains) had bla_SHV only; nine E. coli strains had bla_TEM only, and seven (five K. pneumoniae, one E. coli, and one E. cloacae strain) had both bla_SHV and bla_TEM. Although most SHV (e.g., bla_SHV-1) and TEM (e.g., bla_TEM-1, bla_TEM-2) enzymes do not typically cause 3GC-R, their presence in strains that lack CTX-M-type ESBLs should be investigated, as these enzymes can occasionally confer 3GC-R through enzyme active-site mutation or gene overexpression (5). Of the 14 strains carrying bla_SHV but not bla_CTX-M-1 group, bla_CTX-M-9 group, or bla_CMY enzymes, five had accompanying WGS data, which permitted assessment of bla_SHV ESBL status. In all cases, the bla_SHV enzyme variant was an ESBL (two bla_SHV-2, one bla_SHV-64, one bla_SHV-66, and one bla_SHV-134) (12). In contrast, none of the four strains carrying bla_TEM only were ESBLs, with all four strains carrying bla_TEM-1. Three of these were negative by the double disc diffusion method, indicating probable chromosomal mutations underpinning this phenotype. The basis for the ESBL phenotype in the fourth strain could not be elucidated.

Correlation between bla_CMY and cefoxitin resistance.

Of the 81 cefoxitin-resistant isolates, bla_CMY was detected in only 31 (38.3%) cases. Although imperfect, cefoxitin resistance is often used as a marker for AmpC presence (13). Correlation is also dependent on the MIC, with higher MICs more reliably predicting AmpC presence (13). Our findings concur with this prior work, with 61.7% of cefoxitin-resistant isolates failing to amplify bla_CMY. AmpC has previously been reported as an underappreciated mechanism of 3GC-R due to a lack of standardized testing (14). As such, molecular methods that can detect other common and emerging β-lactamase gene families (e.g., bla_DHA, bla_OXY, bla_MIR) may prove useful for infection control management and treatment of multidrug-resistant infections. This is particularly true for piperacillin-tazobactam, which may not be effective against AmpC producers (11, 15).

WGS identifies uncommon ESBL genes.

Eleven (3.8%) PCR-negative isolates that otherwise successfully amplified their 16S rRNA gene internal control (seven E. coli, one K. pneumoniae, one K. oxytoca, one K. aerogenes, and one E. cloacae isolate) were subjected to WGS to determine the molecular basis of 3GC-R. Six were found to carry uncommon β-lactamase enzyme genes, with three E. coli isolates carrying bla_DHA-1, one K. pneumoniae/K. oxytoca mixture harboring bla_OXY-2-4, one K. oxytoca carrying bla_OXY-2-1, and one E. cloacae carrying bla_ACT-17. The remaining five isolates had no discernible 3GC-R determinants, although misclassification via automated susceptibility testing or upregulation of chromosomal β-lactamases are both possible causes for this discrepancy.

Six isolates (three E. cloacae, two E. coli, and one K. pneumoniae) exhibiting meropenem resistance alongside 3GC-R were also subjected to WGS to determine the basis of resistance toward this carbapenem. As expected, meropenem resistance was mostly conferred by carbapenemases, with 4/6 (66.6%) isolates possessing one of these enzymes. bla_NDM-1 or bla_IMP-4 was identified in two E. cloacae isolates, bla_NDM-5 was identified in an E. coli strain, and a gene encoding an OXA-48-like enzyme, bla_OXA-232, was identified in a K. pneumoniae isolate. No carbapenemase gene could be identified in the remaining two isolates (one E. coli and one E. cloacae isolate). Although not investigated in this study, chromosomally encoded porin loss or efflux pump overexpression may form the basis of meropenem resistance in these strains.

Monitoring patient-to-patient spread in an ICU.

Fifty-four (18.8%) 3GC-R isolates obtained during this study originated from intensive care unit (ICU)-admitted patients. Due to their dominance, only E. coli and K. pneumoniae genomes were analyzed for potential transmission events. Among the E. coli population (Fig. 3), isolates SCHI0016.S.60 and SCHI0016.S.62 and isolates SCHI0016.S.265 and SCHI0016.S.46 were separated by 10 single nucleotide polymorphisms (SNPs) and 2 SNPs, respectively; however, both pairs came from individual patients, reflecting expected levels of within-host evolution of a single infecting strain. E. coli isolates SCHI0016.S.42 and SCHI0016.S.294 were very similar, being separated by 13 SNPs. Although both strains were isolated from ICU-admitted patients, SCHI0016.S.294 was isolated 306 days after SCHI0016.S.42, with no other linking epidemiology identified. Additionally, no similar strains were isolated from ICU patients over this time period, suggesting that these do not represent transmission. E. coli isolates SCHI0016.S.172 and SCHI0016.S.274 differed by ∼300 SNPs, with no epidemiological or temporal correlation. Approximately 300 SNPs separated SCHI0016.S.112 and SCHI0016.S.229, again with no demonstrable epidemiological link. Among K. pneumoniae isolates (Fig. 4), the closest isolates, SCHI0016.S.15 and SCHI0016.S.261, differed by 361 SNPs, with no epidemiological link. Additionally, no horizontal transfer of plasmids carrying ESBL determinants was observed. As standard practice, all ICU patients, and all patients in other wards infected or colonized with ESBL-producing E. coli and K. pneumoniae, have single rooms with an assigned nurse (1:1 ratio), with contact precautions exercised. In addition, hand hygiene rates in our hospital ICU are generally above 80% (16). Taken together, our results demonstrate no evidence of transmission in the ICU, suggesting that the implemented hygiene measures are effective.

FIG 3.

Whole-genome phylogeny of 79 third-generation cephalosporin-resistant Escherichia coli isolates and their β-lactamase gene complement. Sequence type 131 (ST131) is indicated by a gray-shaded box; strains isolated from ICU patients are highlighted in green.

FIG 4.

Whole-genome phylogeny and β-lactamase enzyme presence in Klebsiella pneumoniae. Green-shaded isolates were isolated from the ICU.

Population structure of E. coli and K. pneumoniae genomes.

We subjected an additional 36 strains to WGS (total n = 98) to capture a snapshot of genomic diversity and population structure (Fig. 3 and 4). In silico multilocus sequent typing (MLST) of the 79 E. coli strains found that ST131 was most common (n = 11; 13.9%). This finding is consistent with a recent study of the population structure of Australian, New Zealand, and Singaporean multidrug-resistant E. coli isolates (11). The highly successful ST131 clone has now spread worldwide, where it frequently causes urinary tract and bloodstream infections and accounts for 12 to 27% of all community-acquired E. coli infections (17, 18). Among our ST131 isolates, we found high frequencies of fluoroquinolone resistance-conferring variants (GyrA_S83L, GyrA_D87N, ParC_S80I, and ParC_E84V) alongside frequent carriage of plasmid-borne CTX-M-14 (CTX-M-9 group) and CTX-M-15 (CTX-M-1 group) ESBL genes. We also identified three strains belonging to the emerging multidrug-resistant pandemic clone ST1193. These isolates also possessed the same three fluoroquinolone-resistant SNPs as ST131. It has been proposed that ST1193 strains evolved fluoroquinolone resistance relatively recently through multiple recombination events with other E. coli strains (19).

Phylogenomic analysis of our 10 K. pneumoniae genomes identified several distinct lineages, which comprised two ST14 strains and one each of ST17, ST307, ST309, ST395, ST656, ST697, ST873, and ST969. Of note, the globally disseminated ST258 clone, which is responsible for a large number of hospital deaths (20), was not identified in our data set. However, we identified another globally disseminated clone, ST395, which has also been retrieved from clinical cases in Europe and Asia (21). Concerningly, this ST395 isolate, SCHI0016.S.137, carried both the ESBL genes bla_SHV-11 and bla_CTX-M-15 and the OXA-48-like carbapenemase gene bla_OXA-232, the latter of which confers low-level meropenem resistance. Additionally, we observed one K. pneumoniae ST307, which is an emerging multidrug-resistant, globally disseminated clone postulated to have emerged in the mid-1990s (22). Although reports of this clone from Australian populations are uncommon, it has been previously observed (22).

This study had several recognized limitations. First, isolates were collected from a nosocomial setting; no community sources were sampled. It is well established that differences in AMR patterns and their associated molecular mechanisms exist between hospital and community settings. Second, the absence of certain high-risk populations within our hospital (e.g., solid organ or hematopoietic stem cell transplantation) limits the generalizability of our findings to other settings. Third, our PCR assays can identify only the presence or change in copy number of the most common ESBL and ampC genes. Although these loci make up the majority (>90%) of β-lactamase enzymes present in 3GC-R Enterobacteriaceae, they do not capture all potential ESBLs, leaving less common ESBL and ampC genes unidentified. In addition, our PCR assays for bla_TEM or bla_SHV cannot discriminate between ESBLs and non-ESBLs, with additional assays or concurrent WGS required to confirm ESBL status. Our assays also cannot detect chromosomal changes leading to AMR (e.g., porin mutations or efflux pump upregulation), making these assays less useful for 3GC-R bacteria that frequently encode these AMR determinants (e.g., Pseudomonas aeruginosa). Fourth, our PCR was unable to distinguish chromosomal versus low-copy-number plasmid-borne β-lactamase genes, which has implications for AMR transmissibility between and among bacterial populations. Finally, ∼40% of samples were obtained from rectal swabs, which have limitations for detection of multidrug-resistant organisms due to the high microbial burden coupled with difficulties in sample collection. It is therefore likely that we did not capture all 3GC-R isolates during our study period and thus may have missed potential transmission events.

Conclusions.

Taken together, our findings identified a predominance of CTX-M (and to a lesser extent, CMY) β-lactamases among 3GC-R Enterobacteriaceae in the Sunshine Coast region. These findings are consistent with global findings of Gram-negative bacterial AMR determinants, particularly among E. coli isolates. Our study demonstrates the likely absence of 3GC-R Enterobacteriaceae transmission in our ICU setting, indicating effective infection control processes. We show that rapid molecular assays, such as our multiplex PCR, can detect common β-lactamase genes from bacterial culture within 40 min. These early molecular results would permit more informed selection of antibiotic therapy up to 1 to 2 days prior to 3GC-R confirmation with traditional culture methodologies. Further research is required to determine whether these molecular assays can be used directly on clinical specimens and whether the implementation of molecular assays and more advanced technologies (e.g., WGS) is associated with improved patient outcomes in the hospital setting.

MATERIALS AND METHODS

Ethics.

This study was approved by The Prince Charles Hospital Human Resources Ethics Committee (HREC/18/QPCH/110). A waiver of informed consent was granted due to the low to negligible risk associated with this study.

Isolate collection.

Clinical specimens from 264 adult and pediatric patients (including ICU patients [n = 48]) that were culture positive for 3GC-R Enterobacteriaceae were collected prospectively from July 2017 through September 2018 at the Sunshine Coast University Hospital microbiology laboratory (n = 288). Multiple isolates were collected from 23 patients, due to either reinfection over the course of the study (n = 9) or multiple samples taken during initial admission (n = 11) or collected during the course of their hospital stay (n = 3). Samples were inoculated onto ChromID ESBL agar (bioMérieux) and ChromID Carba/OXA agar (bioMérieux) and examined for colored colonies at 18 to 24 h. An ESBL screening disc test was performed on isolates growing on ESBL agar. Species was determined by using the Vitek 2 matrix-assisted laser desorption ionization–time of flight mass spectrometer (MALDI-TOF MS) with the GN ID card (bioMérieux, Murrarie, QLD, Australia). Isolate information is detailed in Table S1 in the supplemental material.

Susceptibility testing.

Antibiotic susceptibility testing and MIC values were determined using Vitek 2 or Etest (bioMérieux) according to the manufacturer’s instructions. EUCAST criteria (v.10.0) were used for susceptibility and resistance interpretation.

Culture DNA preparation.

For PCR, crude DNA was extracted from all isolates using 5% Chelex-100 resin (Bio-Rad Laboratories, Gladesville, NSW, Australia). For a subset of samples (n = 50), a pinhead-sized amount of culture collected using a sterile 10-μl pipette tip was placed directly into 4 μl of PCR master mix to facilitate rapid turnaround time. For WGS, DNA was extracted using the Quick-DNA miniprep kit (Zymo Research, Irvine, CA, USA).

PCR development and β-lactamase detection.

A quadriplex real-time PCR assay was developed to detect the two major CTX-M ESBL groups (bla_CTX-M-1 and bla_CTX-M-9) and bla_SHV (both ESBL and non-ESBL); a 16S rRNA gene probe-based assay (23) was included in this quadriplex assay to verify DNA integrity. Additionally, two singleplex assays were designed to amplify bla_TEM (both ESBL and non-ESBL) and bla_CMY β-lactamase genes (Fig. 1). Given the large number of mutations in bla_TEM and bla_SHV that can confer ESBL status, identification of bla_TEM and bla_SHV ESBL variants is impossible using a single real-time PCR assay. Therefore, our bla_TEM and bla_SHV assays were designed only to flag bla_TEM- or bla_SHV-harboring strains; secondary screening (e.g., WGS) is required for ESBL confirmation of bla_TEM or bla_SHV PCR-positive strains. To identify conserved regions for PCR assay design, all publicly available genetic sequences for bla_CTX-M-1, bla_CTX-M-9, bla_CMY, bla_SHV, and bla_TEM were downloaded from the NCBI nucleotide database (as of March 2019) and aligned using Clustal Omega v1.2.2. Aligned sequences were visualized in Geneious v11.0.4. Candidate primers and probes were identified in Primer Express v3.0 (Applied Biosystems, Foster City, CA, USA) or Primer3 v0.4.0 (24), followed by primer dimer and heterodimer assessment using NetPrimer (Premier Biosoft, San Francisco, CA, USA).

Black hole quencher probes were used to enable multiplexing in a single PCR well and robust target identification. For the quadriplex assay, the following primers and probes (5′ to 3′) (Macrogen Inc., Geumcheon-gu, Seoul, Republic of Korea) were designed: bla_CTX-M-1 (CTX1_F, CGAMACGTTCCGTCTCGAC; CTX1_R, CTTTCATCCAYGTCACCAGCTG; CTX1_P1, FAM-CGTGATACCACTTCACCTCGGG-BHQ1), bla_CTX-M-9 (CTX9_F, CGTCSCGCTCATCGATACC; CTX9_R, CAGCTGCTTTTGCGTTTCACTC; CTX9_P1, HEX-CGCTTTCCAATGTGCAGTACC-BHQ1), and bla_CTX-SHV (SHV_F, CCARCGTCTGAGCGCC; SHV_R, TATCGGCGATAAACCAGCCC; SHV_P1, Cy5-CAGCACGGAGCGGATCA-BHQ2). The previously described 16S assay (23) incorporated a different fluorophore on the probe (Texas Red-CACGAGCTGACGACARCCATGCA-BHQ2) to accommodate the quadriplex assay format. Two singleplex assays targeting bla_TEM and bla_CMY were also designed: TEM (TEM_F, TGTGGYGCGGTATTATCCCGT; TEM_R, TGCATAATTCTCTTACTGTCAWGCCA; TEM_P1, Cy5.5-CACCAGTCACAGAAAAGCATCT-BHQ2), CMY (CMY_F, TGGCGTATTGGYGATATGTA; CMY_R, TTATGCACCCATGAGGCTTT; CMY_Pr, FAM-TGGGAGATGCTGAACTGGCC-BHQ1). Coincidentally, the reverse primer sequence for our bla_CMY assay was identical to one designed previously (25).

The quadriplex PCR was carried out using 0.1 μM 16S probe and primers, 0.2 μM bla_CTX-M-1 and bla_CTX-M-9 probes and primers, 0.3 μM bla_SHV primers, and 0.4 μM bla_SHV probe. For the bla_TEM singleplex assay, 0.2 μM bla_TEM primers and probe were used, and for the bla_CMY singleplex assay, 0.2 μM primers and 0.4 μM probe were used. All reaction mixtures contained 1× SsoAdvanced universal probes supermix (Bio-Rad), primers, probe(s), 1 μl DNA template, and molecular grade H₂O, to a 5-μl total volume. PCRs were performed on a CFX96 Touch real-time PCR detection system using white Hard Shell 96-well PCR plates and Microseal “C” adhesive seals (Bio-Rad) or 0.2-ml white 8-tube PCR strips and optical caps (Bio-Rad). For bacterial cultures, thermocycling conditions comprised an initial 2 min of denaturation at 95°C, followed by 45 cycles of 95°C for 1 s and 60°C for 3 s (total run time, 39 min).

LoD and LoQ determination.

Three previously characterized strains (26) that represent the PCR assays developed in this study, namely, MER85 (K. pneumoniae), MER110 (E. coli), and MER132 (E. coli), were subjected to total DNA extraction using the Gram-positive bacterial extraction protocol of the Qiagen DNeasy kit (Qiagen, Chadstone Centre, VIC, Australia). DNA quantities were measured on a Qubit 4 fluorimeter (Thermo Fisher Scientific, Seventeen Mile Rocks, QLD, Australia) using the Qubit double-stranded DNA (dsDNA) BR kit and normalized to 40 ng/μl in molecular grade H₂O. LoD and LoQ values were determined across 1:10 serial dilutions (range, 40 ng to 0.4 fg) using MER85 (bla_TEM singleplex), MER110 (bla_CMY singleplex), and a 50:50 mix of MER85 (bla_CTX-M-1 group, bla_SHV) and MER132 (bla_CTX-M-9 group) for the quadriplex assay. For PCR thermocycling, denaturation and annealing times were increased to 5 s and 10 s, respectively, to ensure assay robustness under very low DNA input conditions. Genome equivalents were calculated by assuming E. coli and K. pneumoniae genome sizes of 4.984 Mbp and 5.638 Mbp, respectively.

Assay specificity.

The quadriplex and singleplex assay amplicons were first assessed for specificity in silico by using the microbial nucleotide discontiguous BLAST tool and the complete genomes database (http://blast.ncbi.nlm.nih.gov/; performed March 2019). Next, assay specificity was examined by laboratory testing against a range of both pathogenic and nonpathogenic microbial species found in humans. This diversity panel comprised Achromobacter spp. (n = 4), Bordetella bronchiseptica (n = 1), Candida albicans (n = 1), Cupriavidus sp. (n = 2), Enterobacter aerogenes (n = 1), Enterobacter cloacae (n = 3), Enterococcus faecalis (n = 1), Klebsiella pneumoniae (n = 2), Klebsiella oxytoca (n = 1), Lactobacillus spp. (n = 2), Prevotella spp. (n = 6), Pseudomonas aeruginosa (n = 10), Staphylococcus aureus (n = 4), Staphylococcus epidermidis (n = 1), Stenotrophomonas maltophilia (n = 5), and Veillonella spp. (n = 4).

WGS.

Whole-genome sequencing (WGS) was performed at the Australian Centre of Ecogenomics (University of Queensland, St. Lucia, QLD, Australia) on a NextSeq 500 system (Illumina, San Diego, CA, USA) with Nextera DNA Flex libraries (Illumina).

Genomic analysis.

All genomic data were lightly quality filtered using Trimmomatic v0.39 (27) using parameters described elsewhere (28). WGS data were screened for multiple species by using Kraken 2 v2.0.8 (29). Single species were assembled using MGAP v1.1 (https://github.com/dsarov/MGAP---Microbial-Genome-Assembler-Pipeline), whereas WGS data containing multiple species were assembled with metaSPAdes (30). To determine plasmid copy number and identity, plasmids were assembled using the plasmid module of SPAdes (31), with plasmid homology determined using BLAST (32). Assemblies were binned with MaxBin v2.0 (33), and species determined with Kraken 2 v2.0.8 (29). Antibiotic resistance genes were identified using the Resistance Gene Identifier (34). Phylogenomic analysis was performed using SPANDx v3.2.1 (35). Samples identified as containing multiple strains according to the method described by Aziz and colleagues (36) were excluded from phylogenetic analysis (n = 7); samples able to be split into single species using MaxBin were included in phylogenetic analysis (n = 4). E. coli O157:H7 Sakai (NCBI Assembly ID ASM886v2) (37) or K. pneumoniae HS11286 (NCBI Assembly ID ASM24018v2) (38) were used as read mapping reference genomes for species-specific analyses. To achieve the highest possible resolution, error-corrected assemblies of strains with matching multilocus sequence types (STs) were used as reference genomes for transmission analyses. ST profiles were determined using SRST2 v0.2.0 (39). Phylogenetic trees were reconstructed using PAUP v4.4.7 (40) and visualized in iTOL v5.5 (41).

Data availability.

Raw sequence reads generated in this study are available in the NCBI Sequence Read Archive database under BioProject accession number PRJNA606985.