Analysis of Bordetella pertussis clinical isolates circulating in European countries during the period 1998–2012

Despite more than 50 years of vaccination, pertussis is still an endemic disease, with regular epidemic outbreaks. With the exception of Poland, European countries have replaced whole-cell vaccines (WCVs) by acellular vaccines (ACVs) in the 1990s. Worldwide, antigenic divergence in vaccine antigens has been found between vaccine strains and circulating strains. In this work, 466 Bordetella pertussis isolates collected in the period 1998–2012 from 13 European countries were characterised by multi-locus antigen sequence typing (MAST) of the pertussis toxin promoter (ptxP) and of the genes coding for proteins used in the ACVs: pertussis toxin (Ptx), pertactin (Prn), type 2 fimbriae (Fim2) and type 3 fimbriae (Fim3). Isolates were further characterised by fimbrial serotyping, multi-locus variable-number tandem repeat analysis (MLVA) and pulsed-field gel electrophoresis (PFGE). The results showed a very similar B. pertussis population for 12 countries using ACVs, while Poland, which uses a WCV, was quite distinct, suggesting that ACVs and WCVs select for different B. pertussis populations. This study forms a baseline for future studies on the effect of vaccination programmes on B. pertussis populations.


Introduction
Pertussis, or whooping cough, is a highly contagious human infection of the upper respiratory tract caused by the Gramnegative bacterium Bordetella pertussis [1]. Despite the introduction of vaccination against pertussis more than 50 years ago, the disease is still a public health problem worldwide, with epidemic outbreaks occurring every 3-5 years. In the last 10 years, a resurgence of pertussis was observed in countries with highly vaccinated populations, including Australia [2], Norway [3], Poland [4], the Netherlands [5], the United Kingdom [6] and the United States [7]. One of the hallmarks of the pertussis resurgence is the shift in prevalence from young infants to older persons with waning vaccine-induced immunity [8,9]. Several causes have been proposed for the resurgence of pertussis, including improved diagnosis, increased awareness [10,11], decreased vaccine efficacy, waning immunity and pathogen adaptation [12,13]. The contribution of these factors to the resurgence of pertussis may differ between countries.
Monitoring changes in the European B. pertussis populations and studying the impact of these changes on the disease burden are important in order to establish the most effective pertussis vaccines and vaccination strategies. To study these changes, a European network, designated EUpertstrain, was created in 2001, with the main aim of collecting and typing of European B. pertussis isolates. The number of participating countries increased from five in 2001 to nine in 2013 [40][41][42][43]. Members of the EUpertstrain group collected and typed isolates on a regular basis between 2001 and 2013. Typing involved multi-locus antigen sequence typing (MAST), fimbrial serotyping, multi-locus variable-number tandem repeat analysis (MLVA) and pulsed-field gel electrophoresis (PFGE). However, direct comparison of these isolates collected during the study period was only performed by PFGE analyses [41][42][43].
This present study is part of a European Centre for Disease Control and Prevention (ECDC)-funded network, EUpertlabnet, which focuses on the laboratory surveillance of whooping cough in EU Member States and European Economic Area (EEA) countries. In this study, we present the typing results of isolates collected from 13 countries  Table S1). Isolates collected between 1998 and 2009 were previously analysed by PFGE [41][42][43].
Here, we extended these analyses by typing more isolates from this period. Further, we present novel data from five Multi-locus antigen sequence typing (MAST) Polymorphisms in the genes for proteins used in the ACVs (PtxA, Prn, Fim2 and Fim3) were analysed as described previously [5,15,44,45]. The pertussis toxin promoter, ptxP, was also included, as previous studies have shown that the ptxP3 allele is an important characteristic of successful isolates [5, 20-22, 25, 27-30]. For DNA isolation, bacterial cells were lysed in Tris-EDTA buffer (Sigma-Aldrich, Zwijndrecht, the Netherlands, 1.0 M Tris-HCl, containing 0.1 M EDTA, 100× concentrated) at 95°C for 5 min, centrifuged briefly and used in a polymerase chain reaction (PCR) assay.

Serotyping
A bacterial suspension was mixed on a glass slide with monoclonal or polyclonal antibodies against Fim2 or Fim3 (National Institute for Biological Standards and Control (NIBSC), South Mimms, UK). Agglutination was determined after a maximum of 30 s to avoid false-positive agglutination [44]. Bacterial suspensions were mixed with a physiological salt solution to determine auto-agglutination.
Pulsed-field gel electrophoresis (PFGE) All EUpertstrain isolates were previously analysed by PFGE at the Swedish Institute for Communicable Disease Control (SMI) [43]. PFGE of isolates from Belgium, Ireland and Poland was performed by the respective countries, whereas PFGE of isolates from the Czech Republic and Spain was performed at the Finnish National Institute for Health and Welfare (THL). Gel images of all isolates from the five countries were also analysed at the Finnish THL. The PFGE protocol, reference strains used and profile analysis were the same as those previously described [43]. The profiles were analysed by using BioNumerics software version 4.61 (Applied Maths, Sint-Martens-Latem, Belgium). The Swedish nomenclature was used and based on cluster analysis with the group method with 1 % band tolerance and 1 % optimisation settings. The resulting profiles were designated BpSR1, BpSR2, BpSR3 etc. for those isolates with profiles first detected in Sweden. Isolates first identified in a country other than Sweden, such as Finland, were designated BpFINR1, BpFINR2 etc.

The EUpertstrain and EUpert-labnet collections
One of the aims of the EUpertstrain network is to collect isolates from different European countries in order to assess the emergence and spread of new variants of B. pertussis. A secondary aim is to determine the effect, if any, of different vaccination strategies on the pertussis burden and the emergence of new variants. Analyses of the three EUpertstrain collections isolated between 1998 and 2009 by PFGE and analysis of the EUpert I collection isolated between 1998 and 2001 by MLVA have already been published [40][41][42][43].
Here, we extend this work to the EUpert-labnet collection by including five more EU countries: Belgium, the Czech Republic, Ireland, Poland and Spain (Table 2). Further, additional typing was performed on previously collected isolates. The old and new data were integrated to give an overview of (country-specific) changes of the B. pertussis populations in the participating European countries. Isolates collected between 1998 and 2012 were aggregated in three periods to obtain a comparable number of isolates per year; 1998-2001 (n=106), 2002-2006 (n=165) and 2007-2012 (n=195). As PCR is replacing culture for pertussis diagnosis, obtaining sufficient strains for population studies has become difficult. The number of participating countries for the three periods was 7, 11 and 12, respectively ( Table 2). For Finland, France, the Netherlands and Sweden, 17 to 23 isolates were available for each period. In contrast, for Norway and the Czech Republic, isolates were only available for period 2007-2012. Table 1 shows the pertussis vaccines currently used in European countries.

Changes in allele frequencies
We investigated changes in frequencies of the alleles for four proteins used in ACVs: ptxA, prn, fim2 and fim3. The pertussis toxin promoter, ptxP, was also included, as several previous studies have shown that the ptxP3 allele is an important characteristic of recent clinical isolates [5,20,28]. Isolates were also serotyped to assess expression of the fim2 and fim3 genes. All allelic variants discussed here are associated with changes in protein structure. An overview of allelic and protein variants is provided in previously published reviews [12,47].
For our analysis, we included data only if at least five isolates were available in a particular period. In general, Poland was found to be distinct, while the remaining 12 countries showed minor differences in their B. pertussis populations. Therefore, except for the ptxA alleles and serotypes, we limited the comparisons to Poland and pooled the remaining 12 countries. All data are represented in Supplementary  Table S1.

ptxA alleles
The ACVs currently used in the European countries contain ptxA2 and ptxA4 [12] and the Polish WCV contains only ptxA2 [48]. Essentially, no differences were found between the 13 countries with respect to ptxA alleles. The non-vaccine type allele ptxA1 was identified in 428 out of the 429 isolates analysed. One isolate, isolated in Sweden in 1998, harboured the ptxA2 allele.

prn alleles
The ACVs currently used in the European countries contain prn1 and prn7 [12] and the Polish WCV contains only prn1 [48]. In the group of 12 countries, four prn alleles were observed in this study, prn1, prn2, prn3 and prn13. The Two fim2 alleles have been found worldwide, fim2-1 and fim2-2 [47]. Only fim2-1 is present in the currently used ACVs [12] and the Polish WCV [48]. In this collection, fim2-2 isolates were only detected in Poland and Belgium, while the remaining isolates harboured the fim2-1 allele. In Belgium, one fim2-2 isolate was observed in the period 2002-2006. In Poland, the fim2-2 frequencies were 55 % in 1998-2001 and 43 % in 2002-2006 (Fig. 3). Only two Polish isolates were available for the period 2007-2012: one isolate harboured the fim2-1 allele and one harboured the fim2-2 allele; therefore, the prevalences for the period 2007-2012 were not included in the figure.

Multi-locus variable-number tandem repeat analysis (MLVA)
Thirty-four MLVA types (MTs) were found in the 13 countries. MTs that were found less than five times in the group of 12 countries or in Poland were combined and designated as group R in Fig. 6a

Pulsed-field gel electrophoresis (PFGE)
Eighty-seven PFGE types were found in the 13 countries. PFGE types that were found less than five times in the group of 12 countries were combined and designated as group R in   (Fig. 7a). In Poland, BpSR23 predominated in the period 1998-2001 (frequency 46 %). The predominating PFGE type in Europe, BpSR11, was found once in Poland, in the period 1998-2001 (Fig. 7b).
The ptxA gene was found to be highly monomorphic and, with one exception, all circulating isolates carried the non-vaccine type allele ptxA1. Three ptxP alleles were observed in this collection of isolates, ptxP1, ptxP3 and ptxP20. A notable observation was the low percentage of ptxP3 isolates in Poland (10 %, n=20) compared to the other countries (average 72 %, range 37-100 %). Interestingly, the two Polish ptxP3 isolates were found in the first period, 1998-2001, but not in the later periods. Possibly, ptxP3 isolates do not have a fitness advantage in Poland, in contrast to the other 12 countries. Further, although the non-vaccine type allele prn2 was found at frequencies of 45 and 57 % in the periods 1998-2001 and 2002-2006 in Poland, respectively, these percentages were lower than those observed in the remaining 12 countries (average 83 %, range 37-100 %). Comparing the clinical outcome between ptxP1and ptxP3-infected individuals, especially in Poland, would be useful.
In Poland, two fim2 alleles circulated in approximately equal frequencies, fim2-1 (the vaccine type in Poland) and fim2-2. In contrast, in the remaining countries, only fim2-1 was observed, with the exception of Belgium, where a single fim2-2 isolate was identified. The Belgian fim2-2 isolate was linked to the same MLVA and PFGE type (MT70 and BpBGR4, respectively) as 30 % of the Polish fim2-2 isolates, suggesting that the fim2-2 isolate was introduced in Belgium from Poland. The fim2-2 allele was also observed in the pre-   Fig. 1 for further details vaccination era in the Netherlands (57 %, n=7), while all isolates from 1965 or later harboured fim2-1 [45]. More recently, fim2-2 isolates were observed in Moscow (89 % in the period 1990-2004) and Australia (36 % in the period 1998-2008) [50,51]. In both Poland and Australia, a significant percentage of the fim2-2 isolates were linked to MT70 (80 and 88 %, respectively), possibly suggesting a common origin. However, independent genesis and fixation of the fim2-2 single nucleotide polymorphism (SNP) by different B. pertussis lineages may also be possible. The latter would suggest that the mutation has a significant effect on fitness.
The limited distribution of fim2-2 isolates in Europe compared to Moscow and Poland could be due to geographic isolation, as travelling to Poland and/or Moscow from Western Europe was restricted until the late 1980s. However, in the Czech Republic, which was also located in the former Eastern Europe, no fim2-2 isolates have been observed. Another cause for the observation that the distribution of fim2-2 isolates was limited to Poland and Russia could be the use of the pertussis vaccines, as in Poland and Moscow, a WCV is still used, while an ACV has already been introduced in the Czech Republic in 2007.
The Polish and Irish populations showed higher Fim2 frequencies compared to the other 11 countries (80 and 30 %, respectively). An association between Fim2 and the prevalence of fim2-2 isolates has been found previously in the United Kingdom, where isolates collected between 1920 and 2002 were studied [52]. All isolates with the fim2-2 allele (n= 20) were Fim2. This was also observed in Poland, where 90 % of the fim2-2 isolates were Fim2. However, no fim2-2 alleles have been found in Ireland in this study.
MT27 is the predominating MT in Europe. Previous European studies showed a linkage between ptxP3 and MT27 [17,18,46]. Indeed, 97 % of MT27 isolates in this study harboured the ptxP3 allele. MT27 was also the predominating MT in countries outside Europe, including Australia [2], Japan [25] and the United States [22]. One of the two Polish ptxP3 isolates was also typed as MT27.
The Fim and Prn types of Danish isolates were similar to the isolates found in other countries, except Poland. Since the Danish vaccine contains only Ptx, the changes to the Fim and Prn types in Danish isolates may be the result of overflow from neighbouring countries, rather than selective pressure from the Ptx vaccine.
Five predominant PFGE types have been found in Europe, BpSR3, BpSR5, BpSR10, BpSR11 and BpSR12 [43]. These PFGE types were also predominant in at least one Canadian province, where 317 out of 434 (73 %) isolates collected between 1998 and 2006 belonged to the PFGE types BpSR5, BpSR11 and BpSR12 [24]. However, no isolate with PFGE types BpSR3 and BpSR10 have been found in Canada so far, suggesting some geographic differences in isolates.
The present study showed minor differences in European countries which used ACVs based on MAST, fimbrial serotyping, MLVA and PFGE. However, Poland, the only country included in this study where a WCV is still used, showed a distinct population.
A distinct population compared to Europe has also been found in Serbia, where a WCV has been continued to be used since 1957 [53]. Both countries, Poland and Serbia, have a vaccination history including several changes of strain compositions since the 1960s, which could explain the distinct B. pertussis populations compared to the European countries where an ACV is used. The Polish and Serbian B. pertussis populations both show high prn1 frequencies, 43 % in 2002-2006 and 47 % in 1985-2000, respectively [43], compared to high prn2 frequencies (>90 %) in European countries, where ACVs have been used for many years.
The immune response induced by the two vaccines is different. WCVs induce a broad immune response, but with relatively low titres, whereas ACVs induce high titres against only a few antigens. Theoretical studies suggest that a change from a broad to a narrow immune response will favour the emergence of escape variants [54]. Indeed, several studies have shown the emergence of Prn-deficient strains in populations where ACVs are used [7,[31][32][33][34][35][36][37][38][39]. This is in contrast with Poland and Serbia, where a WCV has been used for many years. No Prn-deficient strains have been detected in both countries [48,53].
This study has some limitations. Isolates were not sampled randomly and the sample size from some countries was small. However, as culture is being replaced by PCR, less and less isolates are available for typing and most members of this study sent all isolates available. Further, it was not possible to establish relationships between strain type and vaccination status, as very limited vaccination data were available. Further, FHA, also a component in some ACVs, was not included, as the large size of the gene precluded analyses. Interestingly, FHA-deficient isolates have been described [32]. In the future, targeted sequencing of genes for vaccine antigens may be replaced by whole genome sequencing, which may reveal other relevant loci to be included in MAST [51,55]. This study was not set up to identify isolates not producing Prn in Europe, as a recent study has already addressed this issue [38]. Overall, low prevalences (<4 %) of Prn-deficient strains were found, except in Norway and France (25 % in 2007-2009 and 15 % in 2013, respectively [56]). Centre for pertussis supported by the Belgian Ministry of Social Affairs through a fund within the Health Insurance System.

Conflict of interest
The authors declare that they have no conflict of interest.
Open Access This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.