Surveillance for antibiotic resistance is essential in recommending effective therapeutic regimens for gonorrhoea. Current surveillance depends on testing cultured organisms against a range of antibiotics to determine the minimum inhibitory concentration (MIC) for each antibiotic. The development of highly specific and sensitive nucleic acid amplification tests (NAAT) detecting both Chlamydia trachomatis and Neisseria gonorrhoeae now offers an alternative to culture diagnosis of N gonorrhoeae.1^–3

In Scotland a comprehensive surveillance system operates in which every culture of N gonorrhoeae is referred to the Scottish Bacterial Sexually Transmitted Infections Reference Laboratory throughout the year for antibiotic susceptibility testing and typing using the highly discriminatory N gonorrhoeae multi-antigen sequence typing (NG–MAST) method.4 The widespread adoption of the molecular diagnosis of gonorrhoea could compromise current antibiotic resistance surveillance programmes because cultures of the organism may not be available for every episode of gonococcal infection identified by NAAT. The molecular detection of antibiotic resistance is not practical as not all the genetic markers of resistance have been elucidated, nor can these be accurately related to the MIC measured. Also, existing assays are not always specific to gonococcal genes, so could not be used directly on clinical samples in which other organisms with a resistant phenotype may be present. NG–MAST,5 which is usually performed on DNA from cultured organisms, can, however, be successfully performed directly on clinical material.6 The aim of this study was to determine whether antibiotic resistance is uniform within any given sequence type and whether NG–MAST data might therefore serve as a useful predictor of antibiotic resistance for N gonorrhoeae when culture isolates are not available.

METHODS

Bacterial strains

All gonococcal isolates referred to the Scottish N gonorrhoeae Reference Laboratory between 1 April 2004 and 31 March 2006 were included in this study. Isolates were grown on modified New York City medium (Biomerieux, Basingstoke, UK) at 37°C, 5% carbon dioxide before antibiotic susceptibility testing and typing.

Antibiotic susceptibility testing and resistance categorisation

Antibiotic sensitivity to seven antibiotics (azithromycin 0.015–2 mg/l, cefixime 0.015–0.12 mg/l, ceftriaxone 0.015–0.12 mg/l, ciprofloxacin 0.004–1 mg/l, penicillin 0.015–2 mg/l, spectinomycin 16–128 mg/l and tetracycline 0.25–4 mg/l) was determined using the agar dilution method.7 E-tests (AB Biodisk, Sweden) were also performed according to the manufacturer’s instructions for those isolates with a ciprofloxacin MIC of 0.125 mg/l or greater and for any isolate with an MIC that exceeded the range of the antibiotic dilution series. All isolates were tested for penicillinase activity using a chromogenic penicillinase test;8 isolates with positive results were confirmed using the nitrocephin test and classified as penicillinase-producing N gonorrhoeae (PPNG). Isolates with a tetracycline E-test of 16 mg/l or greater were classified as tetracycline-resistant N gonorrhoeae (TRNG). Resistance defined as PPNG or TRNG is plasmid mediated. Chromosomal resistance categories were defined as previously:4 decreased susceptibility to azithromycin ⩾1 mg/l, cefixime and ceftriaxone ⩾0.5 mg/l, resistance to penicillin ⩾2 mg/l but not PPNG, resistance to spectinomycin ⩾128 mg/l, resistance to tetracycline ⩾2 mg/l but not TRNG and resistance to ciprofloxacin ⩾0.125 mg/l. Isolates with intermediate resistance to ciprofloxacin (MIC 0.125–0.5 mg/l) were included in the resistant category, because previous work has demonstrated that such isolates can result in treatment failure.9 Isolates with an intermediate resistance to penicillin (MIC 0.12–0.5 mg/l) or to tetracycline (MIC 0.5–1 mg/l) were included in the susceptible/intermediate category.

N gonorrhoeae multi-antigen sequence typing

Characterisation of all isolates by NG–MAST was performed as originally described5 with minor alterations as previously indicated.10

Analysis of discrepancies between suceptibility category and sequence type

The antibiotic susceptibility data for all isolates with non-unique sequence types were reviewed. The modal susceptibility category for each antibiotic within each sequence type was determined and an isolate was considered discrepant if it gave a susceptibility category that differed from the mode. Isolates with an MIC result greater than a twofold difference from the mode were re-tested to confirm that the MIC were correct. The antibiotic susceptibility was determined for all isolates within each sequence type. The percentage of isolates with the expected antibiotic susceptibility for a given sequence type was calculated and exact confidence intervals based on binomial distribution are given for sequence types represented by 10 or more isolates.

Opa-typing

Isolates from selected sequence types in which a discrepant antibiotic susceptibility category was observed were opa-typed11 with modifications to the method as previously described.12

RESULTS

Between 1 April 2004 and 31 March 2006 the Scottish N gonorrhoeae Reference Laboratory received gonococcal isolates from 1811 episodes of infection, of which 1765 (97.5%) were recovered and 46 isolates were non-viable. In all, 1762 (97.3%) isolates were tested for antibiotic susceptibility by an agar dilution method and typed by NG–MAST. Repeat MIC determination was performed for eight isolates for ciprofloxacin, 10 isolates for penicillin, 36 isolates for azithromycin and 24 for tetracycline. MIC for five isolates required adjustment (one each for ciprofloxacin, penicillin and azithromycin and two for tetracycline).

Of the 1762 isolates, 141 (8.0%) were PPNG (including 108 that were also TRNG); all PPNG had an MIC greater than 2 mg/l. There were 148 (8.4%) TRNG (including 108 that were also PPNG); 47 isolates (2.7%) had chromosomal penicillin resistance; 537 (30.5%) had chromosomal tetracycline resistance; 36 (2.0%) had decreased susceptibility to azithromycin; 446 (25.3%) were ciprofloxacin resistant (including 30 (1.7%) with intermediate resistance). Resistance to spectinomycin and decreased susceptibility to ceftriaxone or cefixime were not observed.

NG–MAST identified 405 different sequence types. There were 236 sequence types that occurred only once within the dataset; these were excluded from further analysis. There were 169 sequence types that were represented by two to 85 isolates accounting for 1526 isolates. Table 1 gives details of the consensus susceptibility to each antibiotic and any discrepancies observed for those sequence types represented by 10 or more isolates. Complete details of all sequence types represented by two or more isolates are available at the nucleic acid amplification testing web page of the Health Protection Scotland site (http://www.hps.scot.nhs.uk).

There were 125 sequence types that contained isolates without any discrepancy in their antibiotic susceptibility. The overall level of concordance between the sequence type and antibiotic susceptibility category for all seven antibiotics was 98.1% (10 474/10 682; 95% CI 97.8 to 98.3). The level of concordance for penicillin (chromosomal and plasmid-mediated) was 97.1% (95% CI 96.1 to 97.8), for ciprofloxacin it was 99.5% (95% CI 99.1 to 99.8), for azithromycin it was 97.8% (95% CI 96.9 to 98.5) and for tetracycline (chromosomal and plasmid-mediated) it was 92.0% (95% CI 90.5 to 93.3). There were 44 sequence types that contained isolates with discrepant susceptibility categories; details of the MIC for each antibiotic are given in table 2 for those sequence types represented by 10 or more isolates (details of sequence types represented by less than 10 isolates are available as appendices at the Health Protection Scotland website given above). When possible the modal MIC was identified and those isolates with an MIC within one dilution of the mode are highlighted.

The majority of isolates with discrepancies in penicillin (31/45, 68.9%) and tetracycline (90/122, 73.8%) susceptibility had an MIC at the threshold of the resistance category and were within one dilution of the modal MIC for the particular sequence type. In contrast, few ciprofloxacin and azithromycin discrepant strains had MIC values within one dilution of the modal MIC for the sequence type and a wide range of MIC values was observed (ciprofloxacin 0.004 to ⩾32 and azithromycin 0.03 to ⩾256 mg/l). There were no sequence types containing only strains with decreased susceptibility to azithromycin. For plasmid-mediated resistance there was a high level of concordance: 99.4% (95% CI 98.8 to 99.7) for PPNG and 99.4% (95% CI 98.9 to 99.7) for TRNG.

Opa-typing was performed on selected isolates from all sequence types with a ciprofloxacin susceptibility discrepancy and on isolates from ST340 (represented by two isolates), ST470 and ST649 (these accounted for the majority of azithromycin discrepancies). The majority of these sequence types contained isolates that were indistinguishable from one another by opa-typing: ST495, ST1146, ST1365 (each represented by three or four isolates), ST470 and ST649. Isolates from ST470 and ST649 had the same opa-type. Isolates of ST225 had minor band differences (0–3 bands) and were considered to be related strains. The two ciprofloxacin resistance isolates of ST292 had 12 band differences compared with the eight ciprofloxacin-sensitive strains and were considered to be unrelated strains. Two isolates of ST340 had 17 band differences and were also considered to be unrelated.

DISCUSSION

Overall, the antibiotic susceptibility profile for any given sequence type was remarkably consistent, with only 1.9% discrepancies observed. Consistency was greatest for plasmid-mediated resistance to penicillin and tetracycline and chromosomal susceptibility to ciprofloxacin. Discrepancies in chromosomal penicillin and tetracycline susceptibility categories were slightly higher but were commonly accounted for by a difference of one MIC dilution (an acceptable level of laboratory variation) at the threshold of the resistance category.

These data suggest that, with few exceptions, strains of gonococci as defined by NG–MAST are currently homogeneous with respect to their antibiotic susceptibility. NG–MAST data could thus predict the presence of clinical resistance with a high degree of accuracy for commonly circulating sequence types in which antibiotic susceptibility data have already been accumulated. This could complement culture-based antibiotic resistance surveillance in which a proportion of diagnoses of gonorrhoea are made on the basis of NAAT testing. Performing NG–MAST on clinical material would incur the additional cost of DNA preparation (and possibly nested PCR); this approach has been demonstrated previously on forensic material6 but warrants further validation for routine clinical specimens.

Discrepancies in the ciprofloxacin susceptibility category within sequence types were rare; five of the seven discrepancies identified were confirmed as being from indistinguishable or closely related organisms using a second highly discriminatory typing method, opa-typing. ST292 appears to be a sequence type containing two groups of unrelated strains, which differ in their ciprofloxacin susceptibility and opa-type.

Empiric treatment of gonorrhoea changed in 2003 when ciprofloxacin was no longer recommended;13 this has not resulted in the widespread loss of ciprofloxacin resistance during the period of this study and ciprofloxacin resistance has continued to increase, mainly as a result of the ongoing spread of some common strains such as ST225 and ST147.4 This would imply that the loss of this resistance phenotype offers no selective advantage in N gonorrhoeae and suggests that to date this has only occurred sporadically. This situation is similar to that reported by Corkill et al,14 in which high levels of endemic ciprofloxacin-resistant N gonorrhoeae were maintained in Liverpool beyond the change in antibiotic prescribing policy by the presence of one common strain.

Discrepancies in azithromycin susceptibility occurred mainly in two commonly occurring sequence types, ST470 and ST649, which are closely related to each other (indistinguishable by opa-typing and one nucleotide difference in the por allele by NG–MAST) and could be considered to be part of a single super-cluster. The majority of strains in these sequence types were sensitive to azithromycin, but a minority exhibited decreased susceptibility over a broad range of MIC up to levels of 256 mg/l or greater. One possible explanation for this could be the endemic development of azithromycin resistance in certain strains in response to selection pressure caused by the widespread use of this antibiotic to treat genital chlamydia infections. In this situation the development of resistance would offer a selective advantage for strains of N gonorrhoeae in patients with both infections. The recommended dose of azithromycin for the treatment of chlamydia is 1 g.15 A 2 g dose is, however, needed for reliable efficacy (⩾95%) in the treatment of N gonorrhoeae but the higher dose is not well tolerated by patients.16 Therefore, a 1 g dose of azithromycin could result in the exposure of N gonorrhoeae to sublethal levels of the antibiotic.

This study based on 1526 isolates tested over a period of 2 years for susceptibility to seven antibiotics shows excellent concordance between the sequence type and antibiotic susceptibility. Temporal and geographical variation may exist, however, particularly given that selective pressures, as a result of different national antibiotic prescribing policy, will inevitably vary. Antibiotic susceptibility testing of the cultured organism should continue to be the mainstay of any resistance surveillance scheme because only a phenotypic approach can detect the emergence of new antibiotic-resistant strains. There may, however, be a supplementary role for NG–MAST in predicting antibiotic susceptibility trends of common sequence types in a situation in which some episodes of infection are diagnosed by NAAT.