ChloS-HRM, a novel assay to identify chloramphenicol-susceptible Escherichia coli and Klebsiella pneumoniae in Malawi

Abstract Objectives Chloramphenicol is a broad-spectrum antimicrobial widely available in sub-Saharan Africa. With susceptibility re-emerging among Enterobacteriaceae in Blantyre, Malawi, we designed and evaluated a new high-resolution melt (HRM) RT-PCR assay, ChloS-HRM, to identify chloramphenicol-susceptible infections in a hospital setting. Methods Seventy-two previously whole-genome sequenced isolates of Escherichia coli and Klebsiella pneumoniae from the Queen Elizabeth Central Hospital, Malawi, were subjected to determination of chloramphenicol MICs. Primers were designed to detect 18 chloramphenicol resistance genes that produce seven distinct peaks correlating with different gene groups (catA1, catA2, catA3, catB2, catB group 3, cmlA and floR) following HRM analysis. ChloS-HRM results were compared with MIC and WGS results. Results ChloS-HRM correctly identified 15 of 17 phenotypically susceptible isolates and 54 of 55 resistant isolates, giving an accuracy of 88% in identifying susceptibility and 98% in identifying resistance. WGS identified 16 of 17 susceptible and 54 of 55 resistant isolates, giving an accuracy of 94% in identifying susceptibility and 98% in identifying resistance. The single false-susceptible result had no detectable gene by ChloS-HRM or WGS. Compared with WGS, ChloS-HRM had 100% sensitivity and specificity for catA (catA1–3), cmlA and floR, and 96% specificity for catB; sensitivity could not be estimated due to the lack of catB in the clinical sample collection. The overall agreement between MIC and HRM was 96% and between MIC and WGS it was 97%. Conclusions ChloS-HRM could support antimicrobial stewardship in enabling de-escalation from third-generation cephalosporins by identifying chloramphenicol-susceptible infections. This would be valuable in areas with chloramphenicol-susceptible MDR and XDR Enterobacteriaceae.


Introduction
The greatest burden of antimicrobial resistance (AMR) is in the developing world where severe bacterial infection is an important cause of morbidity and mortality 1 and third-generation cephalosporins are frequently both first-and last-line antimicrobials.
Although AMR prevalence data from sub-Saharan Africa are limited, available data consistently reveal that AMR is increasing, particularly in the Enterobacteriaceae, 2 with resistance to thirdgeneration cephalosporins encountered in up to 46% of isolates depending on the setting. 3 Between 1996 and 2016 in Malawi 68.3% of Gram-negative infections were resistant to first-line drugs amoxicillin or penicillin, chloramphenicol and co-trimoxazole compared with just 6.6% of Gram-positive infections. 4 With the current paucity of new drugs in the pipeline, 5 reimplementing older 'forgotten' drugs is a potential solution, particularly in resource-limited settings. 6 Many of these drugs, such as chloramphenicol, were introduced in the 1940s-70s and increasing resistance has led to their restriction or removal from clinical use. 7 In their absence susceptibility has returned, increasing their utility in treating susceptible infections. 8,9 This is particularly important with the increasing prevalence of ESBLs and carbapenemases reducing the effectiveness of first-line agents.
Chloramphenicol was discovered in 1947 10 and became widely used due its broad activity against many Gram-positive and -negative species; however, by the 1960s it had largely been abandoned in high-income countries due to its toxicity profile. 11 Despite this it V C The Author(s) 2019. Published by Oxford University Press on behalf of the British Society for Antimicrobial Chemotherapy. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/ by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. remains a key drug in low-and middle-income countries due to its low price and ease of production.
Reports of returning chloramphenicol susceptibility are increasing. 9 Studies in Indian hospitals have found that 68% of MDR Gramnegative bacteria 12 and 62.5% colistin-resistant Gram-negative bacteria 13 were susceptible to chloramphenicol. In Malawi chloramphenicol was used in combination with benzylpenicillin as the firstline therapy in the empirical management of suspected sepsis, until 2015. 4 Irrespective of the national guidelines, ceftriaxone has been widely used in Malawi since 2005, partly due to ease of administration (once daily) and partly because the most common cause of bloodstream infection was MDR Salmonella enterica, 14 reducing chloramphenicol use. Susceptibility to chloramphenicol has been gradually returning; resistance rates in 2016 were 61% in Escherichia coli and 48% in Klebsiella pneumoniae compared with 80% of E. coli 15 and 81% of K. pneumoniae in 2001. 4 The regulated reintroduction of chloramphenicol for the treatment of susceptible isolates could reduce the duration of courses of third-generation cephalosporins, protecting these vital agents. 8 Chloramphenicol resistance is most commonly due to enzymatic inactivation by chloramphenicol acetyltransferases (CAT), of which there are two structurally distinct types: CAT-A and CAT-B. 16 There are at least 16 catA genes, with catA1-3 found in Enterobacteriaceae, [16][17][18] alongside catB2 and catB3 group 3 (catB3-6 and catB8). 16 Resistance also occurs due to efflux mechanisms mediated by cml and floR genes. 16 Additional chloramphenicol resistance genes found in Gram-positive species include fexA and fexB from Staphylococcus, the ABC-F gene optrA from Enterococcus spp. 19 and Staphylococcus spp., 20 and the multiresistance genes cfr, cfr(B) and cfr(C), which have been sporadically detected in Proteus spp. and E. coli. 21 Typically, phenotypic testing requires an overnight culture step to isolate the organism, followed by additional culture-based drug susceptibility testing with results available within 48-72 h. This delay results in the prolonged use of empirical broad-spectrum antimicrobials and drives resistance. Molecular assays, however, can be performed from the initial overnight culture, with a result available within 2-3 h, including the DNA extraction process. We report here the design of an RT-PCR assay to detect chloramphenicol susceptibility in E. coli and Klebsiella spp. employing highresolution melt (HRM) analysis, which enables a high degree of multiplexing, without expensive fluorescent probes. 22

Ethics
The collection of the isolates during routine surveillance was approved by the University of Malawi College of Medicine Research and Ethics Committee (COMREC), Blantyre, under study number P.08/14/1614.

Bacterial isolates
Seventy-two previously whole-genome sequenced bacterial isolates, 15 all from separate patients, were collected in Queen Elizabeth Central Hospital, Blantyre, Malawi between 1998 and 2016. The collection consisted of 39 E. coli and 33 K. pneumoniae, and included 61 isolated from blood culture, 10 from CSF and 1 from a rectal swab.
Isolates were transported to the Liverpool School of Tropical Medicine (LSTM), Liverpool, UK on microbiological beads (Biobank) on dry ice where they were stored at #80 C.

Identifying resistance genes from WGS data
The sequencing of the isolates using the Illumina Hiseq 2000 platform (Illumina, Inc., San Diego, CA, USA), de novo assembly and sequence annotation was reported by Musicha et al. 15 The final annotated sequences were screened for acquired AMR genes using NCBI BLAST against a bespoke database of genes curated by the Wellcome Trust Sanger Institute based on the ResFinder database. A cutoff of .95% for nucleotide identity and .90% for coverage against the database was used to confirm gene identity.

DNA extraction
Cultures were revived from microbiological beads (BioBank) using Mueller-Hinton agar (Sigma) plates and incubated at 37 C overnight. Single colonies were selected and DNA was extracted using a DNeasy Blood and Tissue kit (Qiagen, Germany) following the protocol for Gram-negative bacteria.

Chloramphenicol MICs
Chloramphenicol MIC assays were performed following EUCAST 2018 guidelines. 23 Briefly, a single colony from each isolate was added to 5 mL of Mueller-Hinton broth and incubated at 37 C overnight. The culture was diluted to 1%10 5 cfu/mL and added to a 96-well plate containing triplicates of 512 to 0.2 mg/L chloramphenicol in Mueller-Hinton broth and incubated at 37 C overnight. A concentration of 8 mg/L was used as the breakpoint for resistance as per the EUCAST guidelines. 23

Primer design
Primers were designed for all the major chloramphenicol resistance genes found in Enterobacteriaceae: catA1, catA2, catA3, catB2, catB3 group 3 (including catB3, catB4, catB5, catB6 and catB8), cmlA (cml, cmlA, cmlA1, cmlA3, cmlA4, cmlA5, cmlA6 and cmlA7) and floR. 16,24 Sequences for each gene were downloaded from NCBI GenBank and aligned using ClustalX in MEGA 7.0.14 to identify conserved sites. 15 The 16S rRNA primers were designed with a mismatch on the 3 0 end of the forward primer to any non-Enterobacteriaceae species preventing amplification, 25 and produced separate peaks for K. pneumoniae and E. coli. Primers were designed using Primer3 (http://primer3.ut.ee/) to give each amplicon a distinct melting temperature (Tm) for identification following HRM analysis. Amplicon Tm was estimated using the nearest neighbour method on OligoCalc (http://bio tools.nubic.northwestern.edu/OligoCalc.html). Primer sequences, concentrations and amplicon details are displayed in Table S1 (available as Supplementary data at JAC Online).

ChloS-HRM assay
The ChloS-HRM assay was performed in a total volume of 12.5 lL, consisting of 6.25 lL of 2% HRM Type-It mix (Qiagen), variable concentrations of each primer (Table S1), molecular-grade water and 2.5 lL of the DNA template. All HRM development and evaluation runs were performed on a Rotor-Gene Q 6000 (Qiagen, Germany) with the following conditions: Taq activation at 95 C for 5 min, then 30 cycles of 95 C for 10 s, 62 C for 30 s and 72 C for 20 s. The subsequent HRM step consisted of melting from 73 C to 89 C, reading at every 0.1 C step with a 2 s stabilization. Data analysis was carried out using the Rotor-Gene Q software.
Isolates carrying chloramphenicol resistance genes according to WGS data were used to confirm the Tm and location of the peak for each gene. catA3 and catB were not available in the Malawian isolates and therefore synthesized sequences of catA3 (NCBI reference sequence: NG_052661.1) Chloramphenicol-susceptible E. coli and K. pneumoniae identification JAC and catB2 (NCBI reference sequence: NG_047602.1) (Eurofins) and a K. pneumoniae isolate from the LSTM archive carrying the catB3 gene were used as positive controls. Calling bins were set for each gene, including 16S rRNA, and a positivity cut-off of 10% maximum peak fluorescence was set at 0.5 dF/dT ( Figure S1). Any isolate with a peak within a defined calling bin over the positivity cut-off value was classified as positive for the corresponding gene. Any isolate that generated no resistance peaks but produced a 16S rRNA peak was classified as susceptible to chloramphenicol.

Pilot study
The 72 bacterial isolates were tested with the ChloS-HRM assay using the pre-defined calling bins and positivity cut-off value by an operator blinded to the MIC and sequencing results. The ChloS-HRM results were compared against the MIC results as a reference standard and the detection of specific genes was compared with WGS data.

Hypothetical cohorts
To estimate the performance of ChloS-HRM we determined assay outcomes with hypothetical cohorts of 1000 patients in areas with differing levels of chloramphenicol resistance prevalence (90% to 10%). The negative predictive value (NPV), positive predictive value (PPV) and an estimation of the number of cases where treatment could change were calculated.

Pilot study
The ChloS-HRM results were compared with MIC data to determine its accuracy in identifying susceptible isolates ( Table 1). The HRM assay correctly identified 15 of 17 chloramphenicolsusceptible isolates, giving an accuracy of 88% to identify susceptibility (95% CI: 63.56%-98.54%). Of the two isolates falsely classified as resistant by ChloS-HRM, one had a peak for catB group 3 and the other had a peak for floR; this isolate was also positive for floR by WGS. A total of 54 of 55 MIC resistant isolates were classified correctly by the HRM assay, giving an accuracy of 98% (95% CI: 90.28%-99.95%) in identifying resistant isolates. The single false-susceptible result had an MIC of 64 mg/L but did not have any resistance genes detected by HRM or WGS.
Compared with the WGS data the ChloS-HRM assay had 100% sensitivity and 100% specificity for detecting catA (including catA1-3), cmlA and floR ( Table 2). The assay had a specificity of 96% for catB group 3; we were unable to estimate sensitivity due to the lack of this gene in the sample collection. All four isolates with two resistance genes by WGS were correctly identified by ChloS-HRM.
The overall agreement between MIC testing and HRM was 96% in classifying isolates, while the overall agreement between MIC testing and WGS was 97%. See Table S2 for full MIC, WGS and ChloS-HRM data for each isolate.
A 16S peak was detected in 65 of the 72 isolates (90%); all 7 isolates without a 16S peak had at least one resistance marker peak, indicating a successful RT-PCR. A total of 16 of 16 HRM susceptible isolates produced a 16S peak and the correct species was identified in 64 of 64 isolates (100%).

Hypothetical cohorts
The hypothetical cohort indicated that as susceptibility prevalence increases, so too does the PPV in identifying susceptibility, while the number of false-susceptible cases decreases (Table 3). Thus, the higher the prevalence of susceptibility, the more effective the assay is. Once susceptibility drops ,30%, where the PPV and NPV are both 95%, there are too many false-susceptible results for the assay to be effective. The most recent data in Malawi from 2016 indicated a susceptibility prevalence of 39% in E. coli giving a PPV of 97% and an NPV of 93%, and a susceptibility prevalence of 52% in K. pneumoniae giving a PPV of 98% and an NPV of 89%. 4

Discussion
We have developed a highly multiplexed HRM RT-PCR assay capable of detecting 18 different chloramphenicol resistance genes within seven melt peaks, along with a bacterial 16S rRNA control enabling discrimination of E. coli and K. pneumoniae. These are the two most prevalent causes of Gram-negative bacteraemia, excluding Salmonella spp., in Malawi, and the pathogens most strongly associated with ESBL production. The assay was used to detect chloramphenicol-susceptible isolates in Malawi, where rapid de-escalation from broad-spectrum b-lactams such as ceftriaxone could be possible. The assay was 88% accurate in identifying phenotypically susceptible isolates and 98% accurate in identifying resistant isolates. Accurately identifying resistant infections is vital; false-susceptible results would lead to the incorrect use of chloramphenicol on a resistant infection, likely leading to treatment failure. The lower accuracy in detecting susceptible infections would have less impact on treatment success, as a false-resistant result would lead to the continuation of broad-spectrum empirical therapy.
Molecular testing of AMR genes has been shown to correlate well with phenotypic resistance in Enterobacteriaceae, 26,27 and our comparison of WGS data with MIC testing showed a 97% agreement. However, as PCR-based methods detect the presence, but not expression level or copy number of a gene, the effect of these changes of transcription levels is not reflected in the result. 28 PCR methods also cannot distinguish silenced resistance genes, in these cases predicting resistance in susceptible isolates. Silent resistance genes are sometimes reactivated in response to antibiotic pressure but often remain silent. 29 All catA, cmlA and floR genes were correctly detected; however, the degenerate catB group 3 primers produced 3 of 72 falsepositive results; two of these isolates contained other resistance genes and so did not affect the overall result. Degenerate primers enable the detection of multiple catB genes, maximizing sensitivity, but can lead to a higher likelihood of non-specific binding. 30 In the two isolates containing only the floR resistance gene, one had an MIC of 64 mg/L and the other an MIC of 8 mg/L, making it chloramphenicol susceptible. WGS analysis did not reveal any SNPs in the gene from this isolate 15 ; however, the floR gene has previously been shown to confer variable resistance. 31,32 All chloramphenicol resistance genes reported in the Enterobacteriaceae were included in the assay, with some rarely encountered exceptions. 16,24 All genes identified in Malawi during a previous project sequencing resistant invasive isolates were included. 15 The MDR gene cfr, found predominantly in Grampositive species, but also rarely reported in E. coli and Proteus, 21 was excluded. Data on cfr in E. coli and K. pneumoniae are still very scarce, particularly in sub-Saharan Africa.
The Enterobacteriaceae 16S internal control was successfully amplified in 65 of 72 (90%) isolates. 16S primer concentrations were deliberately lower than the other sets to prevent amplification of the 16S gene occurring to the detriment of more important resistance genes. All seven of the samples where the peak was absent contained at least one resistance peak.
Estimation of test outcomes on hypothetical cohorts illustrated that as the prevalence of susceptibility increases, the number of falsesusceptible infections decreases, thus increasing the PPV and enhancing the potential for de-escalation of therapy. Chloramphenicol resistance in the Enterobacteriaceae ranges between 31% and 94.2% depending on the setting; 3,12 therefore, the implementation of this test must be considered alongside local resistance rates.
Recent WGS studies in Malawi found that 56 of 134 (41.8%) of ESBL-containing E. coli 15 and 41 of 262 (15.6%) ESBL-containing K. pneumoniae 39 were chloramphenicol susceptible, highlighting that the ChloS-HRM assay would prove effective in de-escalating to a more effective treatment option, particularly in E. coli. Furthermore, reports of chloramphenicol-susceptible S. enterica are also increasing. [33][34][35][36][37] As resistance is mediated by the genes that the ChloS-HRM assay identifies, 16 it could potentially identify susceptibility in this species. The identification of chloramphenicol susceptibility could also provide a welcome addition to the diminishing options available for treating for MDR and XDR Enterobacteriaceae.
Affordability of antibiotics is critical for low-and middle-income countries, particularly in countries such as Malawi where the public healthcare system provides treatment for free. 38 Chloramphenicol is highly affordable in comparison with carbapenems, the next line of antimicrobial chemotherapy, and increased use would likely be cost saving.
The ChloS-HRM assay enables the sensitive detection of chloramphenicol-susceptible E. coli and K. pneumoniae, allowing Chloramphenicol-susceptible E. coli and K. pneumoniae identification JAC the de-escalation to chloramphenicol and enhancing antimicrobial stewardship. De-escalation to chloramphenicol would also prevent escalation to carbapenems and so would also be protective of this class of drugs as well as cephalosporins. The test would be a powerful tool for molecular epidemiology and surveillance studies on chloramphenicol resistance, where WGS is not feasible.  Williams et al.