Population dynamics of Neisseria gonorrhoeae in Shanghai, China: a comparative study

Gonorrhea is caused by Neisseria gonorrhoeae, a gram-negative aerobic diplococcus, and remains a major sexually transmitted disease (STD) in many countries worldwide [1–4]. The high prevalence of N. gonorrhoeae in developing countries and in some subpopulations in developed countries is related to poor accessibility to diagnostic tests and appropriate treatments [1, 5]. Furthermore, inappropriate use of antibiotics has led to the development of resistant strains, although the prevalence of drug resistance varies greatly among countries [6–11]. The emergence of resistance to antimicrobial agents in gonococci represents a major challenge for public health control strategies in both developing and developed countries [4, 12]. In fact, gonorrhea has proven difficult to control and remains a prime example of the influence that social, behavioral, and demographic factors, as well as microbial factors, can have on the epidemiology of an infectious disease.

In China, gonorrhea represents one of the most common sexually transmitted diseases and the most frequently reported STD in this country [9]. A total of 169,715 cases of gonorrhea were reported to the national Center for STD Control, CDC, China, in 2005 [13]. Even though the incidence of gonorrhea in this country is still increasing, population genetic and molecular epidemiologic studies remain limited [6, 9, 14].

Understanding the patterns of genetic diversity in microbial populations represents an important step for public health control strategies [15, 16]. In fact, the characterization of the sources of such heterogeneity in these populations is relevant for the design of targeted treatment strategies. Previous population genetic studies of N. gonorrhoeae pointed out the role of recombination, mutation and selective pressures as driving forces for the evolution of this pathogen [10, 17–19].

In order to understand the evolutionary processes acting on N. gonorrhoeae transmission dynamics in a community with a high incidence of gonorrhea, such as Shanghai, China, we estimated genetic diversity, recombination, growth, and selective pressure for the porB, seven housekeeping genes that constitute an MLST (MultiLocus Sequence Typing) scheme, and fluoroquinolone resistance genes from 96 Shanghai gonococcal isolates. We also inferred the evolutionary history and population dynamics of this population over the last 40 years. Finally, we compared the demographics of N. gonorrhoeae in Shanghai to that observed in another high incidence area, Baltimore, MD, USA [19]. In Shanghai, during the period of our study, gonorrhea incidence was 75 to 107 new cases per 100,000 inhabitants [20], while in Baltimore it was reported as 180 new cases per 100,000 inhabitants in 2005 [21]. However, in contrast with the Shanghai population that has a prevalence of fluoroquinolone resistant gonococci of 98% [6], in Baltimore such strains are still rare [22, 23]. This comparative study will therefore allow for inferring common patterns of evolution and transmission dynamics between a resistant gonococcal population from Shanghai and a fluoroquinolone sensitive gonococcal population isolated in Baltimore.

Among the 96 strains collected in Shanghai between 2001 and 2005, the phylogenies constructed by maximum likelihood and Bayesian methods did not show any clustering related to the date of isolation (data not shown). The evolutionary relationships assessed using network approaches also failed to reveal temporal clustering in this population, as illustrated by the network reconstruction for the housekeeping genes (Figure 1). The other genes showed also similar patterns in these analyses (Additional files 3, 4 and 5, Figures S2-S4). The quantitative analyses for detecting temporal structuring (F _ST analyses) were also not significant for all year comparisons and for all genes analyzed, except for the fluoroquinolone resistance genes in which very few of these analyses (only 5 among the total 26 analyses for these genes) showed significant P values. However, after Benferroni correction, among these five significant associations, only one year comparison (2002 - 2005) remains significant (Additional file 6, Table S2). This finding suggests therefore no temporal genetic differentiation between the isolates from Shanghai.

The network reconstructions for all the genes indicated a high level of genetic diversity among the isolates, with the greatest diversity observed in the housekeeping genes. In fact, 85 different genotypes were present in the housekeeping genes, with almost 90% of the total genotypes as unique (Figure 1). The networks of the fluoroquinolone resistance genes and the porB PIA and PIB genes showed 37, 7, and 40 different genotypes, respectively (Additional files 3, 4 and 5, Figures S2-S4). The unique genotypes among the fluoroquinolone resistance genes and the porB PIA and PIB genes comprised about 52%, 72%, and 65%, respectively, of the total genotypes in the Shanghai population.

We inferred the past population dynamics of N. gonorrhoeae in Shanghai using the BEAST program. The population dynamics of the gonococcal population in Shanghai over the past 40 years inferred from the housekeeping genes showed that the relative effective population size (Ne) increased dramatically and steadily from the early 1980s until 2002 and then leveled off (Figure 2). The inference made from the porB gene showed that relative Ne was level until the late 1990s and subsequently increased sharply from 1997 until 2002 and more slowly since that time. The confidence limits around the estimate are wide and it cannot be excluded that the rise began in the early 1990s and has been much steeper than the estimate derived from the mean value (Figure 2). In contrast, the fluoroquinolone resistance genes showed a different scenario for the population dynamics of N. gonorrhoeae with stable to slightly increasing relative Ne until 1985 followed by a sharp decline reaching its minimum size in 2005.

We also compared the Shanghai gonococcal population with that from a previous study we conducted using isolates from Baltimore, MD. The comparison between the gonococcal populations isolated in the two geographical areas showed different population dynamics over the past 35 years in each of these settings (Figure 3). Over the past 25 years (from 1980 until 2005), the relative Ne inferred from the housekeeping genes declined dramatically in Baltimore except for a transient rise and fall between 1990-1995, whereas it increased in Shanghai. A different population dynamics was also observed for the fluoroquinolone resistance genes in the Baltimore and Shanghai populations over the past 25 years (Figure 3). Relative Ne was stable in Baltimore, apart from a slight rise and fall, which paralleled the changes seen for the housekeeping genes in the 1990-1995 period, while in Shanghai, relative Ne declined between 1985 and the present, as noted above. In both populations the changes in relative Ne inferred from the housekeeping genes tracked trends in the total number of reported cases of N. gonorrhoeae. Thus, in Baltimore the general decline in cases from 1975 until the present corresponded to a decline in relative Ne, with the exception noted above for the period 1990-1995. Although reporting for Shanghai is incomplete, the dramatic increase in reported cases between 1994 and 1998 was reflected in a steep rise in relative Ne during this time period (Figure 3).

Finally, we analyzed recent evolutionary relationships among the housekeeping gene sequences in the Shanghai gonococcal population using the eBURST algorithm. The allelic profiles of the 96 isolates were classified into two groups with 76 STs in group 1 and three STs in group 2, in addition to six other unlinked STs. In the eBURST diagram, the three major clonal complexes did not include the isolates collected during the same period of time; and the primary founders in two clonal complexes (G793 and G799) were isolated in 2005 (Figure 4-A). The founder ST (G819) for the largest clonal complex was isolated in 2002. For this dataset, the level of recombination is moderate and the nodes in the largest clonal complex represent therefore true lines of descent. The eBURST analysis supports the findings of the phylogenetic and F _ST analyses that there is no temporal genetic differentiation between the isolates from the Shanghai population. We also compared the Shanghai and Baltimore populations using this algorithm in order to detect STs common to both populations. The STs for each population are colored differently in the diagram. No STs shared by both populations were detected (Figure 4-B). Additionally, F _ST analysis revealed significant differences between both populations (P < 0.001).

Fluoroquinolones have been widely used for the treatment of gonorrhea since the 1980s because of their strong antibacterial efficiency in addition to their lower cost and the convenience of oral administration. However, resistance to these drugs emerged quickly in many countries [47–51]. In China, the rate of fluoroquinolone resistance is very high, reaching even 100% in some areas[52]. However, these drugs are still used to treat gonorrhea in this country because all health providers do not follow official recommendations for antibiotics administration and antibiotics are widely available without a prescription. Despite the rapid spread of resistant gonococci in this region, information on population genetics and epidemiology of N. gonorrhoeae in China remain limited [6, 9, 14]. Our study represents the first study to focus on understanding the evolutionary processes acting on the transmission dynamics of resistant gonococci in Shanghai.

Our analysis showed that the evolutionary processes acting on the fluoroquinolone resistant gonococcal population from Shanghai shared many features with what we previously described for a fluoroquinolone sensitive population from Baltimore [17–19]. For both populations, mutation plays a larger role than recombination in the evolution of the porB gene, whereas the latter seems to be the main force driving the evolution of the housekeeping and fluoroquinolone resistance genes [18, 19]. While it is clear that recombinational hotspots occur in genomes (e.g., [53]), the causes of such hotspots remain a mystery [54]. The differences observed in this study seem to be related to differences in selection regimes for the different genes in question. The housekeeping genes are under purifying selection, which leads to a larger number of synonymous relative to nonsynonymous substitutions, and would select against rearrangements due to recombination that would cause fitness declines. For these genes, recombination would therefore play a larger role in creating genetic diversity by shuffling synonymous mutations into novel allelic combinations. The porB gene and fluoroquinolone resistance genes, on the other hand, are more likely to be impacted by diversifying selection and directional selection, respectively. The porB gene undergoes diversifying selection due to its interaction with the immune system to escape immune detection, and we can then expect, as observed in our analyses, a high impact of mutation for generating diversity in this gene. Interestingly, such selection pressure seems also to lead to a dominant role for recombination to shuffle existing variation into novel combinations. The impact of recombination should therefore be incorporated into vaccine strategies that are centered around the porB gene [55, 56]. The fluoroquinolone resistance genes, on the other hand, are exposed to directional selection to acquire substitution changes associated with drug resistance. Apparently, novel combinations produced through recombination aid also directional selection. In fact, this evolutionary process appears to be the dominant force relative to mutation for genes under both directional and diversifying selection. Indeed, the evolution of these genes is analogous to the env and protease genes (i.e., diversifying selection to avoid immune detection and evolution of drug resistance, respectively) in HIV where recombination has a major impact [57]. This interpretation is also supported by the selection analysis that showed a much higher level in the proportion of positively selected sites in the fluoroquinolone resistance genes relative to the housekeeping genes and an increase in selection in the porB gene as well (Table 2). Indeed, the porB gene shows an 8-fold increase in the proportion of sites under diversifying selection relative to the housekeeping genes.

Additionally, in both populations, estimates of growth rate (g) showed high positive values for all genes with the exception of PIA sequences. A likely explanation for the negative value for the PIA sequences is the small number of sequences included in the analysis since g estimation seemed to be very sensitive to sample size, as illustrated by the large confidence interval for this estimate in these sequences. However, it cannot be excluded that there is a decrease in expression of PIA alleles in the gonococcal population as the same observation was made in the Baltimore population. The phylogenetic analyses and the network reconstructions showed no temporal clustering among the Shanghai strains, similar to our previous findings for the gonococcal population from Baltimore [18]. The population genetic analyses of the Shanghai isolates showed similar levels of genetic diversity for time partition, providing further evidence for the absence of temporal clustering. One exception was that isolates collected in 2005 showed a higher level of genetic diversity for the fluoroquinolone resistance genes than isolates collected in the four previous years. Contrary to the Shanghai population, the gonococcal population from Baltimore showed evidence for temporal clustering in the Baltimore population based on the F _ST estimates. However, it is important to note that the sampling interval (1 year) and time span (5 years) of the Shanghai study population were less than those of the Baltimore study, where isolates were collected every fifth year over a 15 year time period. This difference in sampling strategy may have limited our ability to detect temporal clustering. However, it is important to emphasize that the sample size of the Shanghai population is appropriate and adequate for generating population genetic estimates [58].

Changes in relative Ne over time represent a valuable tool for inferring the current and past epidemiology of infectious diseases. Our analysis of population dynamics of N. gonorrhoeae over the past 40 years in Shanghai, inferred from the housekeeping genes, showed that relative Ne has been increasing since the early 1980s with a leveling off from 2002 until 2005. The inference made from the porB gene was similar to that observed with the housekeeping genes. In contrast, the fluoroquinolone resistance genes showed a declining relative Ne since 1985. The appearance of antibiotic resistant strains would be predicted to cause a decrease in relative Ne since a limited number of strains with resistance mutations would have a selective advantage. This expectation was born out by the BEAST analysis of the fluoroquinolone resistance genes. Additionally, a decrease in relative Ne estimated from the housekeeping and porB genes would also be predicted, due to a selective sweep. However, the opposite pattern was observed in this population: relative Ne estimated from these genes increased (Figure 2). The most likely explanation for this finding is that fluoroquinolone antibiotics continue to be used widely in China and their continued use led to rapid diversification as the resistance genes spread throughout the gonococcal population. The evolutionary pattern of a rising relative Ne in housekeeping genes in the presence of a declining relative Ne for resistance genes may have therefore broad application in clinical practice as an indication that antibiotics are being misused and the need for better strategies for managing the resistance. However, we cannot exclude the possibility that social and epidemiological factors also contributed to the increase in relative Ne inferred from the housekeeping and porB genes. The rise in incidence of gonorrhea in late 1990s has been mainly attributed to the influx of people into Shanghai and to economic development attracting more construction workers from the countryside to that area. This movement of people may have resulted in the importation of new strains into the city, fueling the increase in relative Ne. Conversely, the recent leveling off or decline in relative Ne observed in our analysis may be related to a concerted effort by the public health services to prevent and treat new N. gonorrhoeae cases.

The comparative analysis of past population dynamics of gonococcal populations from Shanghai and Baltimore revealed significant differences between these populations (Figure 3). Relative Ne inferred from housekeeping genes declined dramatically in Baltimore over the past 25 years, while relative Ne in Shanghai increased sharply. In both geographic areas, the changes in relative Ne appeared to largely track changes in the total number of reported cases of N. gonorrhoeae. The historical patterns of prevalence of gonorrhea influence therefore deeply its population dynamics. This finding is indeed reassuring that the BEAST analysis is robust despite the recombination level detected in this population. We observed similar trends in relative Ne from housekeeping and fluoroquinolone resistance genes in our analysis of Baltimore strains, compared to the results from the Shanghai gonococcal strains. This is not surprising since fluoroquinolone resistance was not reported in Baltimore until 2005 [19, 22], and thus gyrA and parC genes were under the same selective pressure as the housekeeping genes.

The eBURST analysis showed also that the Shanghai population is very different and unique, compared to the Baltimore population. No ST was shared between the two populations. The lack of identical STs shared by these two populations provides evidence for geographic structuring of gonococcal populations, as previously shown in analyses using porB sequences of strains from different geographic regions [17]. However, phylogenetic analysis of the combined Shanghai and Baltimore MLST dataset did not reveal segregation of gonococcal isolates between the two cities (data not shown). The eBURST methodology appears, therefore, to have greater sensitivity for identifying recent clonal populations, or at least applies a much more stringent definition for such populations. This approach emphasizes also the continuous emergence of resistant strains in Shanghai and the urgent need for better control and prevention strategies in this area.

In comparison to our previous study of sensitive gonococcal strains in Baltimore, this population genetic analysis of quinolone resistant strains revealed differences that may have microbiological and clinical implications. The more dominant role of recombination compared to mutation in the evolution of resistant strains raises the possibility that horizontal gene transfer is an important mechanism for the acquisition of resistance.

Future studies should investigate potential environmental sources for resistance conferring determinants. Although quinolone resistance genes function as housekeeping genes and among sensitive strains the population dynamics of these genes paralleled that of the MLST genes, in the resistant population in Shanghai we found a rising relative Ne in MLST genes in the presence of a declining relative Ne for resistant strains, indicating continuous diversification of resistant strains rather than spread of a clonal resistant population. We speculate that one explanation for this pattern is continued use of an ineffective antibiotic with acquisition of resistance by strains with novel genotypes. In our previous study of quinolone resistant strains from Israel [10], we saw a pattern characteristic of the spread of an epidemic strain with low levels of diversity at all tested loci. Given these different evolutionary pictures, monitoring the population dynamics of N. gonorrhoeae strains at MLST and antibiotic resistance loci could assist in the design of control measures that focus, for example, on core transmitters in the latter case or on poorly informed providers in the former scenario. In this study of gonorrhea in Shanghai as well as our previous study of gonorrhea in Baltimore, we noted that the pas t population dynamics of gonocooci mirrored the historical pattern of the prevalence of gonorrhea. Future prospective studies are needed to see if changes in population dynamics have predictive value and thus would assist in the design of control measures.