Assessment of Antibiotic Resistance and Efflux Pump Gene Expression in Neisseria Gonorrhoeae Isolates from South Africa by Quantitative Real-Time PCR and Regression Analysis

Introduction Treatment of gonorrhoea infection is limited by the increasing prevalence of multidrug-resistant strains. Cost-effective molecular diagnostic tests can guide effective antimicrobial stewardship. The aim of this study was to correlate mRNA expression levels in Neisseria gonorrhoeae antibiotic target genes and efflux pump genes to antibiotic resistance in our population. Methods This study investigated the expression profile of antibiotic resistance-associated genes (penA, ponA, pilQ, mtrR, mtrA, mtrF, gyrA, parC, parE, rpsJ, 16S rRNA, and 23S rRNA) and efflux pump genes (macAB, norM, and mtrCDE), by quantitative real-time PCR, in clinical isolates from KwaZulu-Natal, South Africa. Whole-genome sequencing was used to determine the presence or absence of mutations. Results N. gonorrhoeae isolates, from female and male patients presenting for care at clinics in KwaZulu-Natal, South Africa, were analysed. As determined by binomial regression and ROC analysis, the most significant (p ≤ 0.05) markers for resistance prediction in this population, and their cutoff values, were determined to be mtrC (p = 0.024; cutoff <0.089), gyrA (p = 0.027; cutoff <0.0518), parE (p = 0.036; cutoff <0.0033), rpsJ (p = 0.047; cutoff <0.0012), and 23S rRNA (p = 0.042; cutoff >7.754). Conclusion Antimicrobial stewardship includes exploring options to conserve currently available drugs for gonorrhoea treatment. There is the potential to predict an isolate as either susceptible or nonsusceptible based on the mRNA expression level of specific candidate markers, to inform patient management. This real-time qPCR approach, with few targets, can be further investigated for use as a potentially cost-effective diagnostic tool to detect resistance.


Introduction
Increasing antimicrobial resistance (AMR) to Neisseria gonorrhoeae is now a public health priority [1,2] as it threatens the current World Health Organization (WHO) recommended dual therapy (ceftriaxone and azithromycin) [1][2][3]. Molecular mechanisms of drug resistance have been well characterized [4,5] and are mainly due to mutational alterations of the drug target, plasmids, and efflux pumps [6,7].
Globally, 87 million new cases of the sexually transmitted infection (STI) gonorrhoea occur annually, where the highest prevalence has been reported in the WHO Africa region [8]. e estimated prevalence of gonorrhoea in African countries was reported to be 1.4%-15.2%, with higher prevalence in high-risk groups (sex-workers and participants recruited from venues considered to have a higher probability for acquiring infection, e.g., bars) [9]. South Africa has an estimated prevalence of ∼5% [9,10]. N. gonorrhoeae infections are usually localized to the mucosal surfaces of the hosts initial exposure to the organism [11,12]. Infection of the male urethra causes urethritis (inflammation of the urethra), the symptoms of which, include purulent discharge and dysuria [13]. While male urethritis commonly produces symptoms, gonorrhoea in women is often asymptomatic [14]. e sequelae of untreated gonorrhoea includes acute urethritis, cervicitis, pelvic inflammatory disease (PID), infertility, abortion, ectopic pregnancy, maternal death, and neonatal blindness [15][16][17][18][19].
In Africa and other resource-limited settings, syndromic management of patients remains the main STI management model. Syndromic and empiric treatment leads to overtreatment [20] and contributes to the development of resistance to currently recommended drugs in many parts of the world [21][22][23][24]. ere is no vaccine for gonorrhoea yet; thus, its prevention and control depends on an accurate diagnosis and appropriate antimicrobial therapy [25]. Currently, treatment options are few and antimicrobial stewardship programmes can reduce antibiotic resistance [26]. ere is an urgent need for rapid diagnostic tools to direct therapy [27,28]. e AMR mechanisms via which N. gonorrhoeae has developed resistance has been thoroughly reviewed [4,29]. ese include antimicrobial inactivation, alteration of target sites, increased export via efflux pumps, and decreased uptake via porins [29]. Resistance-to-penicillin and extended-spectrum cephalosporins (ESC) have been associated with modifications and recombination within penA, porB, ponA [30], and the presence of bla TEM plasmid (penicillin) [4,31]. Modifications in penA result in decreased affinity for penicillin, and recombination with penA genes from commensal Neisseria species has led to the development of mosaic penA alleles which causes resistance-to-penicillin, cefixime, and ceftriaxone [4,32,33]. e mutation L421P in ponA reduces the rate of acylation with penicillin [34]. Mutations in porB, which encode porinB, reduce the porin permeability, which then reduces penicillin influx [4]. e bla TEM -1 gene is responsible for plasmid-mediated resistance-to-penicillin, and a previous study from South Africa showed a prevalence of 66% in nonsusceptible isolates [35]. Mutations in mtrR, as well as its promoter region, can cause overexpression of the MtrCDE efflux pump, which has been associated with resistance-to-hydrophobic agents (penicillin, cefixime, ceftriaxone, and azithromycin) [36]. Pore formation in the outer membrane is encoded for by the pilQ gene, mutations in this gene result in reduced antibiotic influx and high-level resistance-to-penicillin [4,12,[37][38][39]. When treating patients with an antibiotic, low-level resistance means that an increased dose of the antibiotic can still overcome the resistance to it and clear the infection. Highlevel resistance, however, means that even an increased dose will not be able to clear the infection i.e., the antibiotic should not be used.
Resistance to tetracycline has been associated with the presence of tetM and mutations in rpsJ, mtrR, and porB [36]. e tetM gene confers high-level plasmid-mediated resistance-to-tetracycline by binding to the 30S ribosomal subunit, thus, releasing the tetracycline molecule and protein synthesis continues [40]. Chromosomally mediated resistance is due to the ribosomal subunit being modified, thus, increasing the efflux and decreasing the influx of tetracycline [4]. e rpsJ mutation V57M, alters the binding site, thus, reducing binding affinity of tetracycline for the ribosome [41]. As described for penicillin, modifications in mtrR and porB, which result in reduced drug accumulation, also contribute to resistance-to-tetracycline [4,12,42].
Resistance to ciprofloxacin is due to mutations in gyrA and parC [36], and mutations in the norM promoter results in overexpression of NorM efflux pump, which increases ciprofloxacin MICs [4,7,43]. Quinolones inhibit DNA gyrase (encoded by gyrA and gyrB) and topoisomerase IV (encoded by parC and parE), which are essential for DNA metabolism, resulting in bactericidal activity [4]. Mutations in these genes alter quinolone recognition of the enzymes and result in resistance [4,29]. Although many mutations have been identified in gyrA and parC, the key mutations responsible for quinolone resistance include gyrA_S91F, gyrA_D95N, parC_S88P, and parC_E91K [4,5,44]. Mutations in the gyrB and parE genes did not significantly impact resistance-to-ciprofloxacin. Mutation in the norM promoter results in overexpression of the NorM efflux pump, which decreases ciprofloxacin susceptibility MICs [4,45].
A range of molecular diagnostic approaches have been evaluated, each with its strengths and limitations [42,[47][48][49][50][51]. While most methods target specific mutations to infer resistance, our approach targets the gene and its expression levels to infer resistance. Genomics prediction tools and equations have been extremely effective in characterizing antimicrobial resistance mechanisms [52,53]. At present, ResistancePlus ® GC (SpeeDx Pty Ltd, Sydney, Australia), which detects resistance-to-ciprofloxacin, is the only commercially available genotypic resistance testing assay for N. gonorrhoeae. e assay uses real-time PCR to detect species-specific porA and opa genes for identification and differentiates the gyrA_S91 wild type 2 International Journal of Microbiology from the gyrA_S91F mutant to determine susceptibility or resistance-to-ciprofloxacin. In populations where the resistance-to-ciprofloxacin is high, assays which predict resistance to additional drugs would be beneficial. e aim of this study was to correlate mRNA expression levels in N. gonorrhoeae antibiotic target genes and efflux pump genes to antibiotic resistance in our population using real-time qPCR, a cost-effective alternative to WGS. A secondary objective was to determine from isolates, if any genes are expressed more in either gender.  [11,13,54]. In addition, a realtime PCR assay, N. gonorrhoeae TaqMan ® probe Ba046466252 ( ermo Scientific) was used for molecular identification. A subset of 61 male and female isolates with similar antibiotic profiles were selected for WGS. is data confirmed the identification of N. gonorrhoeae using Kraken [55] and Pathogenwatch [56].

RNA Extraction and cDNA Synthesis. RNA was extracted
from the N. gonorrhoeae isolates using TRIzol ™ reagent (Invitrogen) with the PureLink ™ RNA Mini Kit ( ermo-Fisher Scientific) and PureLink ™ DNase ( ermoFisher Scientific) as per the manufacturer's instructions. e total RNA concentration was quantified using a nanodrop spectrophotometer, and samples were used only if the optical density at 260 nm (OD 260/280 ) was ∼2.0. RNA integrity was confirmed using a bleach gel method [71]. One microgram of total RNA from each sample was reversed transcribed using the iScript ™ Reverse Transcription Supermix for RT-qPCR (Bio-Rad) as per the manufacturer's instruction and reaction protocol. e total cDNA concentration was quantified, and samples were used only if the optical density at 260 nm (OD 260/280 ) was >1.8.

RNA Quantification by Real-Time PCR.
Quantitative real-time PCR was performed for the primer sequences listed in Table 1 cDNA was diluted using a 1 : 10 ratio for real-time PCR analysis. Each PCR mixture (5 μl total volume) consisted of the respective primers (0.5 pmol/μl for farB and mtrD; 0.7 pmol/μl for 16S rRNA, penA and macA; ese served as standard curves for quantitative analysis. e Quant Studio 5 ( ermoFisher, CA, USA) analysis software version 3.3 was used for quantitative analysis. e expression levels were calculated after normalization to a housekeeping gene (16S rRNA).

Statistical Analysis.
Nonparametric statistical analysis and correlations were performed using GraphPad Prism v5.0 (Graphpad Software Inc. CA, USA) and IBM ® SPSS Statistics v27. e differences between the groups (susceptible/ nonsusceptible, mutations/no mutations, and males/females) were evaluated using t-tests. e regression analysis was performed to determine the relationship of susceptibility as the dependant variable to mRNA levels of antibiotic resistance-associated genes as the independent variables. For each drug, the regression model included all independent variables. A p -value <0.05 was considered to denote statistical significance. Using a receiver operating characteristic

Results
Resistance-to-penicillin, tetracycline, and ciprofloxacin were high in our isolates (supplementary Figure 2), also described in our previous paper [35]. All isolates were susceptible to spectinomycin, azithromycin, ceftriaxone, and cefixime. A total of 110 isolates were analysed to determine differences in mRNA levels between susceptible and nonsusceptible isolates. A subset of 61 isolates with similar MIC values (30 vaginal swabs and 31 urethral swabs) with similar MIC values were analysed to determine differences in mRNA levels between males and females and for differences in mRNA levels between isolates with and without resistanceassociated mutations.

Comparison of Gene Expression Levels between Males and
Females.
e median expressions of antimicrobial target genes and efflux pump genes (farA, farB, gyrA, macA, macB, mtrA, mtrC, mtrD, mtrE, mtrF, mtrR, norM, parC, parE, penA, ponA, pilQ, rpsJ, 16S rRNA, and 23S rRNA) associated with resistance were examined to determine any differences between isolates from males and females. We found no significant difference between the expression levels of isolates from males compared to females. is was confirmed when we subdivided groups into male susceptible, male nonsusceptible, female susceptible, and female nonsusceptible.
Binomial logistic regression was performed to ascertain the effects of penA, ponA, pilQ, mtrR, mtrC, mtrD, mtrE, mtrA, and mtrF on the likelihood that isolates were resistant to penicillin. e model explained 28.6% (Nagelkerke R2) of the variance in resistance-to-penicillin and correctly classified 81.0% of cases. Sensitivity was 96.3% and specificity Bold � significant resistance-associated marker for prediction of antimicrobial resistance in this setting. * e regression models for spectinomycin and ESC were not statistically significant due to a low number of data points in the resistant group, and therefore not included in this table.
International Journal of Microbiology was 29.2%. Of the predictor variables, mtrC was statistically significant (p � 0.024). e discrimination of this model, as determined by ROC curve analysis, is acceptable (AUC 0.8). e logistic regression model was statistically significant, p < 0.009. e regression analysis was used to ascertain the effects of gyrA, parC, parE, and norM on the likelihood that isolates were resistant-to-ciprofloxacin. e model explained 19% (Nagelkerke R2) of the variance in resistanceto-ciprofloxacin and correctly classified 65.8% of cases. Sensitivity was 90.1% and specificity was 30.6%. Of the predictor variables, gyrA and parE were statistically significant, p � 0.027 and 0.036, respectively. e discrimination of this model, as determined by the ROC curve analysis, is acceptable (AUC 0.7). e logistic regression model was statistically significant, p � 0.001.
Regression analysis was used to ascertain the effects of 23S rRNA, macA, macB, mtrR, mtrC, mtrD, mtrE, mtrA, and mtrF on the likelihood that isolates were resistant to azithromycin. e model explained 67.9% (Nagelkerke R2) of the variance in resistance-to-azithromycin and correctly classified 94.9% of cases. Sensitivity was 42.9% and specificity was 98.2%. Of the predictor variables, 23S rRNA was statistically significant (p � 0.042). e discrimination of this model, as determined by the ROC curve analysis, is outstanding (AUC 0.98). e logistic regression model was statistically significant, p ≤ 0.001.
For tetracycline, spectinomycin, and ESC, the discrimination of the models was excellent (AUC >9) and correctly classified >98% of cases. However, due to a low number of data points in either the susceptible or resistant groups, the regression models for these drugs were not statistically significant.

ROC Analysis of qPCR Data.
To determine the threshold (cutoff) of individual genes determined as drug-resistant, a ROC (receiver operating characteristic) analysis was performed. Using the AUC, cutoff, sensitivity, and specificity results (listed, respectively), we evaluated the qPCR assays as a tool to predict resistance to each drug (Table 3). Multiple resistance-associated genes for each antibiotic showed high sensitivities (82%-100%). e performance characteristics of

Discussion
DNA-based diagnostic approaches which detect resistanceassociated single nucleotide polymorphisms (SNPs) are commonly investigated for use in N. gonorrhoeae AMR diagnosis. ese approaches require the detection of multiple known mutations to infer resistance to a particular drug. In this study, we considered the whole gene rather than SNPs, and investigated the expression of known antibiotic target genes and efflux pump genes and correlated gene expression with AMR. To determine if sex-specific environments contribute to the transcription of AMR genes, we compared expression levels from isolates with similar susceptibility profiles from males and females. Regression analysis was used to determine the strongest predictors of drug resistance in our population, and using a ROC analysis; we estimated cutoff values. e antibiotic target genes in N. gonorrhoeae have been widely described [4,12]. Alterations in antibiotic target genes are associated with increased MICs and resistance. Our analysis shows that expression levels of antibiotic target genes are significantly higher in susceptible isolates compared to nonsusceptible isolates. A recent RNA-based study identified candidate markers from the transcriptomes which were highly expressed in the susceptible isolates only [51]. e mtrR gene, which represses the MtrCDE efflux pump, was higher in the susceptible group. is emphasizes the role of mtrR in AMR. MtrR has a more diverse regulatory role than only the MtrCDE efflux pump. It has also been shown to regulate MtrF (inner membrane accessory protein for efficient drug efflux) and FarR (repressor of the farAB efflux pump) [72]. MtrR has also been described to play a role in the expression of two other genes involved in susceptibility, ponA, and pilQ. By increasing the expression of ponA (encodes penicillin binding protein) and repressing the expression of pilQ (channel for entry for penicillin) [72].
Mutations can interrupt cellular processes and often hold the key to understanding gene function. Mutations in target genes are associated with resistance; however, for most mutations, we did not see any significant difference in gene expression between isolates with mutations and isolates without mutations. We found significant differences between the wildtype genes and isolates genes with mutations mtrF_V213I (higher expression), gyrA_S91F (lower), gyrA_D95G (lower), parC_S87N (lower), parC_V384I (higher), and macA_A8S (lower). A combination of mutations and other factors contribute to increased MICs; however, in our cohort, the majority of resistance to penicillin and tetracycline is attributed to bla TEM and tetM, respectively [35].
It was previously reported that N. gonorrhoeae transcriptional responses to infection differed in genital specimens from men and women, and AMR gene expression was increased in men, with a higher expression of MtrCDE efflux pump-related genes, suggesting that the expression of AMR genes is driven by sex-specific environments [49]. While overall gene expression signatures may be sex-specific, we found that in our cohort of South African patients, there were no significant differences in expression of resistanceassociated targets between isolates from men and women. In addition, the farA and farB genes which encode the FarAB efflux pump (export host-derived antimicrobials, including cationic antimicrobial peptides and long-chain fatty acids) [4], were similarly expressed. Based on this outcome, we found that while therapeutic strategies could be based on gender, when using a diagnostic assay, there was no need to streamline the gene target profile based on gender, and that the same targets can be used for AMR detection for specimens from males and females in our setting.
In this study, we determined thresholds of mRNA levels, which could be used for resistance prediction within a South African population. Using regression analysis, we then determined the strongest predictors associated with AMR status. N. gonorrhoeae AMR is associated with numerous resistance mechanisms. We found the most significant markers for AMR status prediction in this population to be mtrC, gyrA, parE, rpsJ, and 23S rRNA. Our approach provides thresholds with high sensitivity for each of the strongest predictors using ROC analysis and can be used as a rule-in test for resistance prediction.
A limitation of the study and proposed models was the lack of clinical isolates with resistance to spectinomycin, azithromycin, cefixime, and ceftriaxone. Future studies will include a larger data set. Another limitation is that this needs to be evaluated on clinical specimens direct from patients to establish sensitivity as well as time-toresult. Also, the strongest predictors for the antibiotic resistance detection were limited to the isolates data used to generate the regression equations and is valid for the local setting in which these strains were isolated. is enforces the need for continuous local surveillance of isolates. Similar regression analysis can be used to identify candidate markers for resistance prediction in different geographic areas.

Conclusion
Using real-time qPCR, we have identified that mRNA levels of potential candidate markers of resistance can be used for AMR testing of N. gonorrhoeae. Together with the ROC cutoff values, these can be explored further as a set of genetic markers of antimicrobial resistance in our setting. A larger-scale validation is required, and evaluation directly from clinical specimens. Identifying local candidate markers has the potential to be used as a nearpatient test in addition to NAATs identification.

Data Availability
Whole-genome sequence data is available in DDBJ/ENA/ GenBank with the BioProject number PRJNA681740.

Conflicts of Interest
e authors declare that they have no conflicts of interest.
Acknowledgments is study was funded by the DST-NRF Centre of Excellence (CoE) in HIV Prevention grant (96354) and the National Health Laboratory Service Research Trust grant (19762). e funders had no role in study design, data collection, and interpretation or the decision to submit for publication. e authors acknowledge Prof P. Moodley for donating a subset of isolates, V. Maseko and the staff at National Health Laboratory Service Microbiology Department (Durban).

Supplementary Materials
Supplementary Figure 1