Diagnosis of Whooping Cough in Switzerland: Differentiating Bordetella pertussis from Bordetella holmesii by Polymerase Chain Reaction

Bordetella holmesii, an emerging pathogen, can be misidentified as Bordetella pertussis by routine polymerase chain reaction (PCR). In some reports, up to 29% of the patients diagnosed with pertussis have in fact B. holmesii infection and invasive, non-respiratory B. holmesii infections have been reported worldwide. This misdiagnosis undermines the knowledge of pertussis' epidemiology, and may lead to misconceptions on pertussis vaccine's efficacy. Recently, the number of whooping cough cases has increased significantly in several countries. The aim of this retrospective study was to determine whether B. holmesii was contributing to the increase in laboratory-confirmed cases of B. pertussis in Switzerland. A multiplex species-specific quantitative PCR assay was performed on 196 nasopharyngeal samples from Swiss patients with PCR-confirmed Bordetella infection (median age: 6 years-old, minimum 21 days-old, maximum 86 years-old), formerly diagnosed as Bordetella pertussis (IS481+). No B. holmesii (IS481+, IS1001−, hIS1001+) was identified. We discuss whether laboratories should implement specific PCR to recognize different Bordetella species. We conclude that in Switzerland B. holmesii seems to be circulating less than in neighboring countries and that specific diagnostic procedures are not necessary routinely. However, as the epidemiological situation may change rapidly, periodic reevaluation is suggested.


Introduction
Pertussis outbreaks are reported worldwide and Bordetella pertussis is detected by polymerase chain reaction (PCR) with increasing frequency in nasopharyngeal (NP) samples of symptomatic patients. Teenagers and young adults represent the main reservoir, and transmit the bacterium to the more vulnerable pediatric population. In Switzerland, clusters of pertussis must be reported to the government, and current incidence of pertussis is both high (2012: 94/100'000) and increasing [1].
Although the emerging pathogen Bordetella holmesii was initially described as causing bacteremia in immunodeficient patients [2], it has also been recovered in NP samples of immunocompetent individuals with pertussis-like symptoms [3]. Respiratory infection with B. holmesii can easily be misidentified as B. pertussis by PCR because both genomes contain the insertion sequence IS481 targeted by PCR assays routinely used in most laboratories. This misdiagnosis of B. holmesii as B. pertussis is common: in a recent large quality control test, only one out of 24 European laboratories properly identified a B. holmesii strain [4].
Accurate identification of B. holmesii is important for several reasons. One concerns the pertussis vaccine: some countriesincluding Switzerland -have implemented or are currently implementing booster pertussis vaccine doses for teenagers and/ or adults. These new recommendations were implemented in response to the increasing number of patients with pertussis-like symptoms, and a shift in the age of B. pertussis' reservoir. The primary aim of reducing the circulation of pertussis is to protect the at-risk population, namely the young infants. However, the pertussis vaccines -both the cellular and acellular vaccine-do not protect against B. holmesii, according to an animal study [5]. In humans, the estimated effectiveness of the adolescent acellular pertussis booster vaccination against B. holmesii during a pertussis outbreak in Ohio (USA) was 36% (95%CI: 233%; +69%), compared with 67% against B. pertussis (95%CI: 38%; 82%) [6]. Therefore, immunized patients are likely not protected against B. holmesii infection and subsequent misdiagnosing of B. holmesii for B. pertussis could falsely suggest vaccine failure. This could have a strong impact on both physicians' and patients' perception of the vaccine but also lead to inappropriate public health measures. In addition, it is unclear at this time if prophylactic antibiotic treatment of all contacts of index cases, such as in B. pertussis infection, is necessary for B. holmesii infection [7]. B. holmesii canunlike B. pertussis -cause invasive diseases, mainly in immunocompromised patients [2,8], but also in immunocompetent hosts [9][10][11]. A work-up and treatment should therefore be considered when encountered. Finally, macrolides -antimicrobials used to treat B. pertussis infections -seem to have a lower activity against B. holmesii [12]. In vitro studies report that the most effective antibiotics against B. holmesii are fluoroquinolones and carbapenems which are rarely used to treat pertussis [2,[13][14].
Respiratory infections caused by B. holmesii have been reported worldwide in different frequencies [3,5,[15][16][17][18][19][20]. A study showed that up to 20% of French adolescents and adults with pertussis-like symptoms were in fact infected with B. holmesii [17]. Switzerland shares borders with France and none of the Swiss laboratories routinely identifies B. holmesii in NP samples. The aim of this pilot study was to retrospectively re-analyze extracted DNA from Bordetella-positive's NP samples for the presence of B. holmesii in our region, and to see whether B. holmesii could account for part of the recent increase in laboratory-confirmed pertussis cases in Switzerland.

Materials and Methods
Stored DNA extracted from 196 NP specimens from patients in Switzerland identified as positive for B. pertussis or B. parapertussis by routine PCR in 2009-2012 were reanalyzed by a triplex speciesspecific quantitative PCR (qPCR).
For the routine test, extraction of DNA from clinical specimens was carried-out with an automated extraction system (easyMAGH, BioMérieux, Marcy-l'Etoile, France). The routine real-time quantitative duplex PCR targeted IS481 and IS1001 and was performed in a LightCycler 2.0 (Roche, Basel, Switzerland).
Target sequences of the species-specific triplex qPCR were IS481, IS1001 and hIS1001 ( Table 1). The IS481 target is highly sensitive for B. pertussis: its genome contains .50 copies of the IS481 sequence, but it is not species-specific: it is also present in the genome of B. holmesii (8-10 copies), and in some animal (occasionally human) isolates of B. bronchiseptica. The IS1001 target is commonly used to diagnose B. parapertussis, although it can also be present in some strains of B. bronchiseptica [21]. The hIS1001 target was selected to identify B. holmesii, as previously described [22]. This multiplex qPCR hence allowed the discrimination between B. pertussis (IS481+, IS10012, hIS10012), B. holmesii (IS481+, IS10012, hIS1001+), and B. parapertussis (IS4812, IS1001+, hIS10012). Randomly selected samples were then retested by a simplex qPCR using oligonucleotides recognizing the 59 end of the pertussis toxin target (ptxA), which is only present in B. pertussis.
Sequences of target genes were scanned for the design of sequence-specific oligonucleotides with the PrimerExpress 2.0 software (PE Biosystems, Foster City, CA, USA) using default settings. Conditions for the amplification on the CFX-96 (Bio-Rad, Hercules, CA, USA) were as follows: t1, 15 min at 95uC; t2, 15 s at 95uC; t3, 30 sec at 60uC (t2 and t3 were repeated 40 times). The volume of the PCR mixture (Bio-Rad, Hercules, CA, USA) was 20 ml and contained primers and probes at indicated concentrations (Table 2). Default analysis parameters were used with the CFX-96 device software; the standard deviation of fluorescence values recorded from cycles 3 to 15 was multiplied by 10 to define the cycle threshold line. Cycle thresholds were derived from the intercept between this line and the signal obtained during the qPCR. A reaction was considered positive when fluorescence levels exceeded the detection threshold between cycles 15 and 30. Sensitivity levels of the qPCR were assessed by using 10-fold serial dilutions of purified genomic DNA from control isolate, over a 6 log range. The detection sensitivities of B. pertussis and B. holmesii were 7-8 bacteria/sample.
After consultation of our institutional review board (Ethic Committee of Human Research of the University Hospitals of Geneva) patients did not have to provide informed written consent given that the samples had been irreversibly anonymized and kept for quality control. The ethics committee has waived the need to approve the study once the protocol was submitted.

Results
Of the 196 available NP samples, 194 had enough material left to be reanalyzed: 188 (97%, 95% Confidence Interval (CI): 94%-99%) were confirmed as B. pertussis, and 5 (3%, 95%CI: 0%-5%) as B. parapertussis. None contained B. holmesii (0%, 95%CI: 0%-1.5%). One sample was negative for all three targets. None of the positive samples showed multiple signals, meaning that there was no co-infection. These results were confirmed by the simplex qPCR targeting ptxA. Only B. pertussis, as determined by IS assay, yielded positive toxin signals. Generally the Ct of the IS signal appeared 4-8 cycles earlier than that obtained for the toxin, corresponding to the relative difference of abundance within cells. Patients had a median age of 6 years (interquartile range 3-15 years; minimum 21 days-old, maximum 86 years-old), with 52% of females (Table 3).

Discussion
We detected no B. holmesii in NP samples from Swiss patients with PCR-confirmed Bordetella infection reported as positive with a routine, non-discriminating, Bordetella PCR assay. Therefore it seems unlikely that B. holmesii is responsible for the reported increasing incidence of pertussis diagnosis in Switzerland. This result is in accordance with other studies from various geographical origins: B. holmesii was not recovered in samples from Tunisian infants [23], or patients from Finland [24], the Netherlands [24], Canada (Alberta) [25] and Australia [26]. In contrast, B. holmesii was identified in 0.4%-29% of Bordetella-positive NP samples from Canada (Ontario) [16], in children from Argentina and Chile [15,19], and adolescents and adults from France [17], Japan [18], and USA (Massachusetts, Ohio) [3,[5][6].
Our results were unexpected for several reasons. First, we recently identified B. holmesii in blood samples of two immunocompromised patients hospitalized in our institution and hence expected the ''endemic'' presence of B. holmesii in our population [27]. In addition, infection due to B. holmesii has been detected in several patients from different regions of Switzerland (Swiss laboratories network, personal communication). Hence, it is possible, that B. holmesii circulates in our country, but that its incidence is low or that it follows epidemic curves such as reported for B. pertussis [28]. Second, B. holmesii was reported in neighboring France [17]. Given the retrospective setting of our pilot study, we unfortunately lack specific information on the patients' clinical presentation, immunization history and socio-demographic characteristics. Therefore, it is unclear whether and how our patients differ from the French patients reported by Njamkepo et al [17]. Age difference could not explain this dissimilarity: the majority of our patients were 0-9 years-old children (113/196, 58%), but there were also 41 adolescents (9-17 years-old, 21%) and 42 adults (.17 years-old, 21%), which is comparable to the age distribution of the patients reported in the French study (119 children, 20 adolescents, 39 adults).
The highest frequency of B. holmesii has been described in adolescents and adults with pertussis-like symptoms. However, it is now known that B. holmesii can cause respiratory infections in all age groups since children -including seven ,6 months-old -were reported in recent publications from Argentina and Chile [15,19]. In our study, all age groups were represented: we were therefore surprised not to find any B. holmesii in patients with PCRconfirmed Bordetella infection. A limit in our qPCR's detection level could explain the fact that we had no positive patient. However, serial dilutions showed a sensitivity level around 8 bacteria, sufficient to identify B. holmesii in the specimen, even in a low amount. In all positive reactions, the fluorescent signal appeared quite early (earlier than cycle 30) reflecting a significant abundance of the amplified target gene.
Finally, little is known on B. holmesii epidemiology and its interaction with B. pertussis' epidemiology. It is possible that some other factors (genetic, pertussis immunization coverage, management following suspicion of B. pertussis index case, or local pertussis outbreaks) influence the transmission of B. holmesii, and differ between countries explaining the observed difference.
The two main limitations of this study are its retrospective aspect, as discussed earlier, and the small number of samples. It would have been interesting to have access to the patients' clinical data, and to compare them with other studies looking at signs and symptoms in patients with Bordetella holmesii versus non-holmesii infections. It is unclear at this time how to differentiate clinically patients presenting with whooping cough but infected with a different Bordetella. It is estimated that around 5000 pertussis cases occur per year in Switzerland [1]: we tested less than 200 samples. However, in our study all age groups are represented and the bacteria were sampled from symptomatic patients in one region sequentially. The difference in B. holmesii incidence between countries [3,[5][6][15][16][17][18][19][23][24][25] remains unknown, and suggests that active surveillance should be continued. However, since no B. holmesii was isolated in our samples and because of the higher costs of the Bordetella species-specific PCR, there is currently no urge to implement specific diagnostic procedures routinely. The priority remains to diagnose B. pertussis. Based on our results and given the importance to prevent misdiagnosis of B. holmesii as B. pertussis, we would highly recommend to all regions facing with an increasing incidence of pertussis diagnosis to use Bordetella species-specific tests in order to see whether and to what extend B. holmesii is contributing to this raise. Depending on the result, it should than be discussed if species-specific tests are needed routinely; they might not be indicated if the prevalence of B. holmesii is very low, given their higher cost. However, the epidemiology of both species can change rapidly and a second study with species-specific tests should be considered a few years later. Therefore, for the moment, in our opinion, in Switzerland, specific PCR diagnosis for B. holmesii should only be used for epidemiological reasons, in unusual patient situations, or in vaccine studies.