Genetic Analysis of Bordetella pertussis Isolates from the 2008–2010 Pertussis Epidemic in Japan

A large pertussis epidemic occurred between 2008 and 2010 in Japan. To investigate epidemic strains, we analyzed 33 Bordetella pertussis isolates from the epidemic period by sequencing virulence-associated genes (fim3, ptxP, ptxA, and prn) and performing multilocus variable-number tandem repeat analysis (MLVA), and compared these results with those of 101 isolates from non-epidemic, earlier and later time periods. DNA sequencing of the fim3 allele revealed that the frequency of fim3B was 4.3%, 12.8%, 30.3%, and 5.1% within isolates in 2002–2004, 2005–2007, 2008–2010, and 2011–2012, respectively. The isolation rate of the fim3B strain therefore temporarily increased during the epidemic period 2008–2010. In contrast, the frequencies of the virulence-associated allelic variants, ptxP3, ptxA1, and prn2, increased with time during overall study period, indicating that these variants were not directly involved in the occurrence of the 2008–2010 epidemic. MLVA genotyping in combination with analysis of allele types showed that the prevalence of an MT27d strain temporarily increased in the epidemic period, and that this strain carried virulence-associated allelic variants (fim3B, ptxP3, ptxA1, and prn2) also identified in recent epidemic strains of Australia, Europe, and the US. Phenotypic analyses revealed that the serotype Fim3 strain was predominant (≥87%) during all the periods studied, and that the frequency of adhesion pertactin (Prn) non-expressing B. pertussis decreased by half in the epidemic period. All MT27d strains expressed Prn and Fim3 proteins, suggesting that B. pertussis MT27d strains expressing Prn and Fim3B have the potential to cause large epidemics worldwide.


Introduction
Bordetella pertussis, a highly communicable Gram-negative coccobacillus, is the cause of pertussis (whooping cough), a major acute respiratory infection resulting in severe childhood illness and infant death [1]. Although universal immunization programs have contributed to significant reductions in the morbidity and mortality rates associated with pertussis, the incidence of pertussis has increased in several countries despite high vaccination coverage [2][3][4][5][6]. In Japan, acellular pertussis vaccines (ACVs) were introduced in 1981 and are used to control pertussis with a schedule of 3 primary doses and single booster dose at ages 3, 4, 6, and 18223 months, respectively. This vaccination schedule has been followed since 1994. The incidence of pertussis cases in adolescents and adults, however, has significantly increased since the early 2000s [7]. The waning of vaccine-acquired immunity and the decrease in opportunities of natural immune boosting owing to reduced levels of B. pertussis circulation have been proposed as explanations for the recent resurgences of pertussis [2,8,9]. Another possible underlying factor is the adaptation of the B. pertussis population to vaccine-induced immunity [2,10,11].
Antigenic and genetic shifts in B. pertussis circulating strains have been identified within virulence-associated genes encoding serotype 3 fimbriae (fim3), pertussis toxin S1 subunit (ptxA), pertactin (prn), and the pertussis toxin promoter (ptxP). Allele frequencies of the virulence-associated allelic variants, fim3B, ptxA1, prn2, and ptxP3, have increased within the B. pertussis population in several countries [10,[12][13][14][15][16][17]. B. pertussis strains carrying ptxP3 are more capable of producing pertussis toxin (PT) than are ptxP1 strains, and the emergence of ptxP3 strains was associated with pertussis resurgence in the Netherlands [18]. Similarly, a significant correlation was observed between an increase in fim3B strains and pertussis notifications in the US [13]. Strains with fim3B have a single amino-acid substitution (A87E) within the surface epitope of Fim3, which interacts with human serum [19,20]. Furthermore, multilocus variable-number tandem repeat analysis (MLVA) has revealed that the B. pertussis population has changed over the past 50 years worldwide. In Australia, the frequency of B. pertussis MLVA type 27 (MT27) and MT70 strains increased after the introduction of an ACV, and subsequently, the MT27 strain became predominant in 200822010 [21,22]. An increase in the frequency of B. pertussis MT27 strain was also observed in Europe and the US [13,16,23].
Pertussis epidemics still occur worldwide, and epidemic strains have been characterized by Fim serotyping and/or genotyping within some regions [24][25][26]. In a Dutch epidemic, significant changes in Fim serotypes and MTs were observed during a period when the pertussis vaccine dose was lowered [24]. Besides phenotypic variants of Fim, B. pertussis variants that do not express adhesion pertactin (Prn) have been recently identified in Japan as well as in other countries [27][28][29][30]. Since Prn is a component of ACVs, it is reasonable to hypothesize that Prn-negative variants have increased fitness in humans immunized with ACV. To date, the relationship between the prevalence of Prn-negative variants and pertussis epidemics has not been evaluated.
In Japan, a large pertussis epidemic occurred in 200822010 despite high vaccination coverage with ACVs. To elucidate the causes of the epidemic, we determined temporal trends in the frequencies of virulence-associated genes (fim3, ptxP, ptxA, and prn) and genotypes in the B. pertussis population from 2002 to 2012. In addition, phenotypes of epidemic isolates were characterized by their expression of Fim and Prn proteins.

Pertussis surveillance data
National surveillance data were obtained from the Ministry of Health, Labor and Welfare of Japan Infectious Disease Surveillance data. Each week, the number of pertussis cases was reported from approximately 3,000 sentinel clinics and hospitals within Japan. Diagnosis was based on bacterial culture, clinical symptoms, and/or the results of a serologic test. The reporting criteria did not change during the study period, 200222012.

Bacterial strains
We examined 134 clinical B. pertussis isolates collected in Japan from 2002 to 2012 (Table S1). Thirty-three of those isolates were collected during the 2008-2010 pertussis epidemic, while 101 isolates were collected from non-epidemic periods: 23 in 2002-2004, 39 in 2005-2007, and 39 in 2011-2012 ( Fig. 1). All isolates were epidemiologically unrelated cases of pertussis. The isolates were cultured on Bordet-Gengou agar (Difco) or cyclodextrin solid medium (CSM) agar [31], and incubated at 36uC for 223 days. DNA was extracted from B. pertussis isolates by boiling, and stored at 220uC.
Sequence analysis of fim3, ptxP, ptxA, and prn Four virulence-associated genes, fim3, ptxP, ptxA and prn, were analyzed using PCR-based sequencing [14,18,32,33]. Sequence reactions were carried out with a BigDyeH Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems), and resultant products were sequenced using an Applied Biosystems 3130xl Genetic Analyzer or 3730 DNA Analyzer. Subsequent sequencing of the variable region 2 (R2) of prn was performed where necessary, to distinguish between prn1 and prn7 alleles. Primer sets used in this study are listed in Table S2.

MLVA
MLVA typing was performed as previously described [22,28]. MTs were assigned using the MLVA typing tool found at http:// www.mlva.net. Novel MLVA types were assigned by the webmasters, Drs. H. van der Heide and F. Mooi, from the Centre for Infectious Disease Control Netherlands, within National Institute for Public Health and the Environment, in the Netherlands.

Immunoblotting and serotyping analysis
Prn expression in B. pertussis isolates was analyzed by immunoblotting [28]. Briefly, protein samples (1 mg) were first subjected to 10% SDS-PAGE, then transferred onto nitrocellulose membranes (Bio-Rad), and finally incubated with anti-Prn antiserum. Antigen-antibody complexes were visualized using a horseradish peroxidase-conjugated secondary antibody (Bio-Rad) and the Western Lightning ECL Pro (PerkinElmer, Inc.). Resultant blots were imaged using a LAS-3000 (Fujifilm, Tokyo, Japan).
Serotyping of B. pertussis isolates was performed with indirect whole-cell ELISA using anti-Fim2 and anti-Fim3 monoclonal antibodies as previously described [34,35], with some minor modifications. Briefly, bacterial cells cultured on Bordet-Gengou agar were resuspended in phosphate-buffered saline (PBS) to an optical density of 0.01 at 620 nm, and then inactivated at 56uC for 1 h. The wells of 96-well ELISA plate (Nunc Immuno Plate Maxisorp) were coated with 100 ml of this suspension to each well and allowing it to evaporate overnight at 36uC. Anti-Fim2 (NIBSC 04/154) and anti-Fim3 (NIBSC 04/156) antibodies were used at a 1:1,000 dilution in PBS. Antibody binding to bacterial cells was detected following the addition of a 1:4,000 dilution of alkaline phosphataselabeled goat anti-mouse IgG (Southern Biotechnology Associates, Inc.) and with the use of p-nitrophenylphosphate as a substrate. The optical density was measured at 405 nm with 650 nm as a reference using a Multiskan FC microplate reader (Thermo Fisher Scientific Inc.). B. pertussis strain 18323 that expresses both Fim2 and Fim3 was used as a positive control.

Temporal changes in the frequencies of Prn and Fim variants
The Prn and Fim expression phenotypes of the 134 B. pertussis study isolates were determined with immunoblotting and indirect whole-cell ELISA, respectively. Figure 3A shows the temporal trend of the frequencies of the 2 identified Prn variants, Prnexpressing and Prn-negative strains, with the frequency of Prnnegative strains at 43.5%, 41.0%, 21.2%, and 25.  (Table S1). Prn-negative strains carrying prn1 were previously found in Finland [30], and those carrying prn2 were found in France and the US [27,29]. On the other hand, 4 Fim variants, Fim2, Fim3, Fim2,3, and Fim(2), were identified among the B. pertussis isolates. The Fim2,3 strain was detected by both Fim2 and Fim3 antigens, while the Fim(2) strain was not detected by either. As shown in Figure 3B (2) strain was previously identified in Ontario, Canada [12].

Temporal changes in MTs and genetic diversity
Among the 134 B. pertussis study isolates, 24 different MTs were identified, of which 2 were novel (MT251 and MT253). MT27 strains could be further classified into 5 subtypes (MT27a to MT27e) based on their allele types (Table 1). MT27a strains carried fim3A, ptxP3, ptxA1, and prn2, and were the predominant subtype. Notably, they were collected in both non-epidemic and epidemic periods. In contrast, MT27d strains carrying fim3B, ptxP3, ptxA1, and prn2 were predominantly isolated during the epidemic period, with 10 of 12 isolates (83%) being of this subtype in 200822010. An MT27c strain carrying fim3A, ptxP3, ptxA1, and prn9 was also isolated during the epidemic period. The other MT27 subtypes, MT27b and MT27e, were found only in 201122012.
The genetic diversity of MTs and the 5 MT27 subtypes was subsequently determined. As shown in Figure 5

Discussion
In the present study, we demonstrated that the prevalence of B. pertussis strains carrying fim3B, which were classified as MT27d, temporarily increased during the 2008-2010 pertussis epidemic in  Japan. The MT27d strains of the epidemic period were collected in several areas (Table S1). All MT27d strains carried fim3B, ptxP3, ptxA1, and prn2, and expressed Prn and Fim3 proteins. Although Prn-negative strains have recently increased in their prevalence within Japan and in other countries [27][28][29][30], here, a lowered frequency of Prn-negative strains was observed during the epidemic period specifically, indicating that Prn-negative strain types were not involved in the epidemic. Similarly, we found that the prevalence of the virulence-associated allelic variants, ptxP3, ptxA1, and prn2, has increased with time from the early 2000s, indicating that the variants were also not directly associated with the epidemic. Taken together, our findings demonstrate that B. pertussis strains carrying fim3B (i.e., MT27d) were associated with the 2008-2010 pertussis epidemic.
We evaluated the regional effect on B. pertussis population because of the low number of samples of isolates. When the number of isolates was compared, the difference in regional population between epidemic (2008-2010) and post-epidemic (2011-2012) periods was statistically significant (P , 0.01), possibly because of the high number (18/39) of isolates collected in the Kinki district (including Osaka) during the post-epidemic period. A sampling bias cannot be excluded from the analysis of the trend in 2011-2012. However, no significant difference was observed between pre-epidemic (2005-2007) and epidemic periods (P . 0.05). The regional effect was therefore small or negligible for detection of the emergence of strain MT27d in the 2008-2010 epidemic.
In the past decade, the prevalence of B. pertussis strains carrying fim3B has increased worldwide [12][13][14]25]. In Ontario, Canada, 1 predominant B. pertussis clone was identified in the 2000s [12]. This strain carried the same virulence-associated allelic variants (fim3B, ptxP3, ptxA1 and prn2) as our epidemic strains within MT27d and was identical to the strains involved in recent pertussis resurgences within Europe and Australia. Similarly, fim3B strains carrying ptxP3, ptxA1, and prn2 were predominant in the US during the 2000s, and most were genotyped to MT27 [13]. Interestingly, the pertussis resurgence in the US was correlated with the emergence and predominance of the fim3B allele, but not with the ptxP3 allele. On the other hand, in the Netherlands, the prevalence of fim3B strains temporarily increased in the early 2000s, although the strains disappeared in 2010 [14]. Likewise, MT27d strains (carrying fim3B) were not identified after the 200822010 epidemic period, and the reasons behind the disappearance of this strain type are unclear. Our findings along with those of previous studies, therefore, suggest that the MT27d strain is a recent epidemic strain that is found worldwide, and that this strain is not only associated with pertussis resurgence but can also be correlated with pertussis epidemics. Fimbriae of B. pertussis are composed of Fim2 and/or Fim3 and FimD. The minor fimbrial subunit FimD forms the adhesive tip [38]. Fim2 and Fim3 are encoded by the single-copy genes fim2 and fim3, respectively [39,40], and are serologically distinct [41]. Fim resulting from the fim3B strain is Fim3B, which differs from Fim3A by a single amino-acid substitution (A87E) at the surface epitope. To date, the biological differences between Fim3A and Fim3B are unknown. In an effort to address this issue, we recently observed B. pertussis clinical strains and found that most strains carrying fim3B had strong autoagglutination capability, unlike those carrying fim3A, following the suspension of CSM agar cultures into saline (Otsuka and Kamachi, unpublished data).
Surprisingly, autoagglutination was not observed when the strains were cultured on Bordet-Gengou agar. Bacterial autoaggregation is a phenomenon associated with pathogen virulence in many Gram-negative bacteria [42][43][44][45][46], suggesting Fim3B strains are more virulent than Fim3A strains because of their ability to autoagglutinate. Further study is necessary to fully elucidate the relationship between autoagglutination and Fim3B. Attempts to identify the molecular mechanism that regulates autoagglutination within fim3B strains are currently underway.
In many countries, a shift from serotype Fim2 to Fim3 in B. pertussis circulating strains was observed after mass vaccination, and antigenic differences (in Fim, PT, and Prn) have been since noted between B. pertussis vaccine strains and circulating strains. In Japan, the B. pertussis strain Tohama carrying ptxA2, prn1, and fim2 has been used as a vaccine strain to produce ACVs from 1981. Among the 4 currently used Japanese ACVs, 2 contain Fim2 and all do not contain Fim3 [47]. A recent study demonstrated that Fim2 and Fim3 are immunogenic antigens, and that individuals recently infected with pertussis had greater anti-Fim3 IgG concentrations than anti-Fim2 IgG concentrations, consistent with the current predominance of Fim3 strains [41]. Based on this finding, there is a clear need for the improvement of currently used ACVs. Specifically, the inclusion of Fim3 (Fim3A and/or Fim3B) in ACVs may be an effective way to reduce the number of current circulating B. pertussis strains, including Fim3B strains. Although Fim2 has been shown to be a protective antigen, the protective immunogenicity of Fim3 is still unknown [48]. Further study of this topic will be required to evaluate the effectiveness of Fim3 as a protective antigen.
In the present study, the genetic diversity of the B. pertussis population decreased after the 200822010 pertussis epidemic. This decrease reflects the expansion of the MT27a strain type and the disappearance of MT27d strains. The MT27a strains carried the same ptxP3, ptxA1, and prn2 alleles as the MT27d strains, but additionally carried fim3A. Significant changes in the B. pertussis population were previously observed in a pertussis epidemic in the Netherlands, and this study suggested that strain surveillance may serve as an early detector of pertussis epidemics [24]. Here, we observed significant changes within the B. pertussis population during the 200822010 epidemic, lending further support for an early warning system of future pertussis epidemics.  In conclusion, the prevalence of B. pertussis MT27d strains temporarily increased during the 200822010 pertussis epidemic in Japan. The MT27d strains carried the same virulence-associated allelic variants (fim3B, ptxP3, ptxA1, and prn2) as recent epidemic strains observed in Australia, Europe, and the US. B. pertussis MT27d strains, therefore, likely have the potential to cause large epidemics in other countries, and, hence, further study and strain surveillance of the MT27d strain type is warranted.