Applying molecular algorithms to predict decreased susceptibility to ceftriaxone from a report of strains of Neisseria gonorrhoeae in Amsterdam, the Netherlands

In a recent article, de Korne-Elenbaas et al.¹ described a group of ceftriaxone-decreased-susceptible and ceftriaxone-susceptible Neisseria gonorrhoeae strains in Amsterdam, the Netherlands, collected from 2014 to 2019 at a sexually transmitted infection (STI) clinic. Resistance has been reported in every class of antibiotics used for gonococcal treatment. The prevalence of antimicrobial resistance in N. gonorrhoeae is rising globally.² The current treatment recommended by the WHO is dual therapy with ceftriaxone and azithromycin, although several countries have moved to ceftriaxone monotherapy.^3,4 In the Netherlands, the recommended treatment for uncomplicated gonorrhoea is monotherapy with ceftriaxone 500 mg intramuscular as a single dose.⁵ It is therefore critical to be able to predict the susceptibility of N. gonorrhoeae strains to ceftriaxone in order to guide treatment and reduce the spread of resistant strains.

The mechanisms for the development of resistance to ceftriaxone in N. gonorrhoeae are complex, making prediction of ceftriaxone susceptibility difficult.⁴ The four primary genes that have been associated with resistance to ceftriaxone are penA, ponA, penB and mtrR.⁶ The penA gene encodes PBP2; multiple alterations in penA can result in resistance through decreased binding affinity of β-lactam drugs, such as ceftriaxone. The ponA gene encodes PBP1 and the amino acid alteration L421P results in a similar but less influential effect on drug binding affinity. The penB gene encodes the PorB porin protein and amino acid alterations at the 120 and 121 positions result in decreased permeability of antimicrobials. Finally, the mtrR gene encodes the transcriptional repressor of the MtrCDE efflux pump and the deletion of an adenine residue in the promoter region results in increased efflux of antimicrobials. Despite our knowledge of these genes, accurate prediction of ceftriaxone susceptibility phenotype has been impeded by the multiple mechanisms of resistance and genetic heterogeneity of N. gonorrhoeae strains globally.⁴

Coupling ceftriaxone susceptibility and genetic data in a global set of N. gonorrhoeae strains published through 15 October 2019, we proposed four molecular algorithms to predict decreased susceptibility to ceftriaxone using an MIC of >0.064 mg/L.⁶ Our algorithms vary in whether they (i) include penA mosaicism and (ii) include penA or non-penA genes (mtrR, penB, ponA). Some of the algorithms resulted in high sensitivity or specificity with low to moderate complementary specificity and sensitivity, depending on the genetic targets used. The proposed algorithms could offer flexibility in the genes targeted, depending on the setting and prevalence of genetic markers.

In the report by de Korne-Elenbaas et al.,¹ there were 318 N. gonorrhoeae strains with MICs ranging from <0.002 to 0.125 mg/L. Using the MIC breakpoint of >0.064 mg/L, there were 80 ceftriaxone-decreased-susceptible strains and 238 ceftriaxone-susceptible strains.¹ Using the genomic sequence data in their report, we applied the previous molecular algorithms to these strains to determine their performance in predicting decreased susceptibility.

First, using the algorithm in Figure 1(a), the arm with the highest sensitivity included wild-type penA A311 and non-wild-type penA A510V, resulting in a sensitivity and specificity of 94% and 25%, respectively. Applying that algorithm to the Amsterdam strains, all ceftriaxone-decreased-susceptible isolates were captured (100% sensitivity) with a specificity of 7%.

Figure 1.

Four molecular algorithms predicting decreased susceptibility to ceftriaxone: (a) an algorithm utilizing penA amino acid alterations without mosaicism determination; (b) an algorithm utilizing penA amino acid alterations with mosaicism determination; (c) an algorithm utilizing non-penA (ponA, penB) amino acid alterations without mosaicism determination; and (d) an algorithm utilizing non-penA (ponA, mtrR) amino acid alterations with mosaicism determination. A plus symbol indicates the presence of the corresponding genetic alteration, while a minus symbol indicates the absence of the corresponding genetic alteration. Sensitivity and specificity values are for decreased susceptibility to ceftriaxone and are from our previous work. Testing for all genetic loci in each algorithm is intended to be done simultaneously and not necessarily in a step-wise fashion. An asterisk indicates that no ceftriaxone-decreased-susceptible strains have been found with this combination of genetic alterations. Adapted from Lin et al. J Infect Dis 2021; 223: 1232–40.

Using the algorithm in Figure 1(b), the estimated sensitivity and specificity were 95% and 62%, respectively, in the absence of mosaicism with non-wild-type penA L447V and the presence of any of the following additional penA mutations: G542S, P551L/S or A501V/T. Applying that algorithm to the Amsterdam strains, the sensitivity was 100%, while the specificity was 77%.

Using the algorithm in Figure 1(c), the estimated sensitivity and specificity in the presence of ponA L421P and at least one of penB G120 or A121 were 92% and 61%, respectively. Applying that algorithm to the Amsterdam strains was associated with 100% sensitivity and a specificity of 84%.

Finally, using the algorithm in Figure 1(d), the estimated sensitivity and specificity were 89% and 72%, respectively, in the absence of mosaicism and the presence of mtrR promoter adenine deletion and ponA L421P. Once again, all ceftriaxone-decreased-susceptible strains were captured (sensitivity 100%) with a specificity of 75%.

Next, using the clinic’s reported prevalence of ceftriaxone decreased susceptibility of 1.1%, we estimated the positive and negative predictive values for each of the four algorithms. All algorithms resulted in a positive predictive value <10% with a 100% negative predictive value. The algorithm utilizing non-penA genes without mosaicism resulted in the highest positive predictive value of 6.5%. Therefore, given the low prevalence of decreased susceptibility, the low positive predictive values indicate that these algorithms, if developed into molecular tests, would have limited utility within the Amsterdam STI clinic. However, developing tests to detect decreased susceptibility using a lower MIC breakpoint of >0.064 mg/L would help identify concerning isolates. For example, positive tests might help to identify isolates needing additional antibiotic susceptibility testing, modified treatment regimens (e.g. increased ceftriaxone dosage or alternative antimicrobials) or closer follow-up with a test of cure, ultimately reducing the chance for treatment failure, overall cost of treatment and spread of resistant N. gonorrhoeae strains.

Furthermore, as the prevalence of ceftriaxone decreased susceptibility increases, so too will the positive predictive values of these algorithms. For example, at a prevalence of 30%, the positive predictive values of three out of the four algorithms are over 60% with no compromise in the negative predictive values. Therefore, while utilization of these algorithms as molecular tests would incur additional lab work and cost, and have minimal benefit at the current prevalence of ceftriaxone decreased susceptibility in the Amsterdam STI clinic, they would be significantly more useful as the rates of ceftriaxone-resistant N. gonorrhoeae increase.

Using the report by de Korne-Elenbaas et al.,¹ we demonstrate the potential application of the proposed molecular algorithms for the prediction of decreased susceptibility to ceftriaxone in N. gonorrhoeae. Not only were all ceftriaxone-decreased-susceptible strains captured using all four algorithms, but it appears that three of the four algorithms perform well in the Netherlands with higher specificity values than those predicted using a global set of isolates. Moreover, we recently validated two of our algorithms in a report of a ceftriaxone-resistant N. gonorrhoeae strain from Portugal.^7,8 The genetic distinctness of the Amsterdam and Portugal strains with respect to sequence-type combinations (MLST-MAST) of N. gonorrhoeae isolates and different countries of origin further illustrates the value of incorporating several genetic markers into molecular tests to predict decreased susceptibility to ceftriaxone. In light of the urgent global health threat posed by antibiotic resistance in N. gonorrhoeae, utilizing genetic targets for the rapid prediction of ceftriaxone resistance is critical. Continual global monitoring of genetic loci associated with ceftriaxone resistance will enable further improvements in molecular testing algorithms.

Funding

This work was supported by the National Institutes of Health (R21 AI157817 to J.D.K. and T32MH080634 to P.C.A.).

Transparency declarations

None to declare.