Resolution of a Protracted Serogroup B Meningococcal Outbreak with Whole-Genome Sequencing Shows Interspecies Genetic Transfer

A carriage study was undertaken (n = 112) to ascertain the prevalence of Neisseria spp. following the eighth case of invasive meningococcal disease in young children (5 to 46 months) and members of a large extended indigenous ethnic minority Traveller family (n = 123), typically associated with high-occupancy living conditions. Nested multilocus sequence typing (MLST) was employed for case specimen extracts. Isolates were genome sequenced and then were assembled de novo and deposited into the Bacterial Isolate Genome Sequencing Database (BIGSdb). This facilitated an expanded MLST approach utilizing large numbers of loci for isolate characterization and discrimination. A rare sequence type, ST-6697, predominated in disease specimens and isolates that were carried (n = 8/14), persisting for at least 44 months, likely driven by the high population density of houses (n = 67/112) and trailers (n = 45/112). Carriage for Neisseria meningitidis (P < 0.05) and Neisseria lactamica (P < 0.002) (2-sided Fisher's exact test) was more likely in the smaller, more densely populated trailers. Meningococcal carriage was highest in 24- to 39-year-olds (45%, n = 9/20). Evidence of horizontal gene transfer (HGT) was observed in four individuals cocolonized by Neisseria lactamica and Neisseria meningitidis. One HGT event resulted in the acquisition of 26 consecutive N. lactamica alleles. This study demonstrates how housing density can drive meningococcal transmission and carriage, which likely facilitated the persistence of ST-6697 and prolonged the outbreak. Whole-genome MLST effectively distinguished between highly similar outbreak strain isolates, including those isolated from person-to-person transmission, and also highlighted how a few HGT events can distort the true phylogenetic relationship between highly similar clonal isolates.

N eisseria species, most of which reside in the nasopharynx, are considered harmless human commensal bacteria with two important exceptions: Neisseria gonorrhoeae, which inhabits the urogenital tract, and Neisseria meningitidis, a commensal bacterium which very rarely invades the nasopharyngeal epithelia and causes life-threatening septicemia and less frequently meningitis. In common with most countries worldwide, in the Republic of Ireland (RoI), invasive meningococcal disease (IMD) is legally notifiable (1). The RoI continues to experience an elevated rate of IMD relative to the rest of Europe despite the declining incidence observed over the last decade (1, 2). The majority of IMD is sporadic, with incidence in 2014 of 1.79/100,000 and 1.44/100,000 (80%) for serogroup B (3). As elsewhere, clusters are reported occasionally, typically in densely populated living environments such as military or university dormitories, with the protracted Princeton University outbreak, where the outbreak strain belonged to the hyperinvasive ST-41/44 clonal complex (cc), representing one such recent example from the United States (4).
Meningococcal populations are composed of multiple distinct lineages known as clonal complexes, according to the globally accepted bacterial typing paradigm, multilocus sequence typing (MLST) (5). Lineages consist of many highly similar strains, or sequence types (STs), which arise as a consequence of both clonal decent and the high rates of horizontal gene transfer (HGT) observed in meningococci (6,45).
Meningococcal disease rates vary with age, with incidence rates peaking and declining within the first year of life until the mid to late teenage years when a second smaller incidence peak is observed (7), concomitant with the highest rates of carriage (8) and possibly a consequence of increased social activity. Many risk fac-tors for IMD have been reported, such as complement cascade deficiencies, smoking, and socioeconomic status (9)(10)(11). The most important factor is the frequency and closeness of social contacts.
Endemic strain disease incidence can be controlled by vaccines targeting the polysaccharide that defines the meningococcal serogroup, with the important exception of serogroup B, which is poorly immunogenic in humans (12). Close contacts of individuals with IMD are typically offered chemoprophylaxis to reduce the occurrence of secondary cases. The National Guidelines for the Early Clinical and Public Health Management of Bacterial Meningitis define an outbreak of meningococcal disease as "a minimum of four cases of definite meningococcal disease within a 3-month interval, or 40/100,000 in any age group in a 3 month period, in a geographical area that makes epidemiological sense AND where available microbiological characterization of the organism is the same" (13).
Meningococcal outbreaks are less frequently observed than instances of endemic sporadic disease and tend to occur in environments where population densities are high, facilitating increases in both transmission rates and carriage and increasing the likelihood of persistent carriage.
Between March 2010 and November 2013, eight cases of IMD occurred in young children (5 to 46 months), who were members of an extended Irish Traveller family. Irish Travellers exhibit a unique culture and lifestyle (14), and as an indigenous minority, they experience social and cultural marginalization (15). No studies describing meningococcal disease outbreaks in a Traveller population have been reported previously. One month after the final case, a carriage study was undertaken to ascertain meningococcal prevalence. We present the molecular typing results of disease and carriage isolates, characterized by whole-genome sequencing (WGS), as an exemplar demonstrating how housing density can drive meningococcal carriage and facilitate persistence. In compliance with the National Guidelines and paralleling the Princeton experience, chemoprophylaxis was administered to close contacts following each case, but this failed to clear the outbreak strain from carriage in the larger network. Both outbreaks were controlled but only following mass chemoprophylaxis and simultaneous vaccination.

MATERIALS AND METHODS
Cohort. This extended family lived in three separate midland towns, either in town housing or in trailers. Community members (n ϭ 123) were invited to special outbreak control clinics; 112 choose to attend. The overall community size is somewhat larger (approximately 140 people) as some members were in other parts of Ireland or the UK and were unavailable to attend the clinics.
N. meningitidis carriage prevalence was evaluated by collecting anonymized pharyngeal swabs from 112/123 family members who attended special outbreak control clinics in December 2013. Specific demographic data were collected, including age, housing type, and housing occupancy. Pharyngeal swabs were cultured for Neisseria spp. The presence of meningococcal nucleic acid was investigated from swab head solutions using a ctrA/porA duplex real-time N. meningitidis-specific PCR assay (16).
Whole-genome sequencing. Genomic DNA was purified using the Wizard kit (Promega) according to the manufacturer's instructions. DNA quantity was assessed using the Qubit device (Life Technologies). DNA integrity was checked using 1% agarose gel (Promega). Genomic DNA from case and pharyngeal carriage isolates were sequenced on an Illumina HiSeq sequencer (Wellcome Trust Center for Human Genetics, University of Oxford, UK). The resulting short-read sequences were assembled de novo using the VelvetOptimiser algorithm (23) as part of an in-house pipeline developed in Oxford and then added to the publicly accessible Neisseria PubMLST website (24) (http://pubmlst.org/neisseria/), which uses the Bacterial Isolate Genome Sequence Database (BIGSdb) platform (24).
Data analysis. The BIGSdb Genome Comparator tool was used to investigate the relationships among the isolates. A specified set of loci among the isolates was compared, generating a matrix of allelic related-ness, which resolved isolates into a phylogenetic network using the Neighbor-Net algorithm (25) viewed in SplitsTree4 (26).
To verify that the assembled N. meningitidis sequences appearing to contain recombinant fragments from Neisseria lactamica were not due to contamination and/or misassembly, the short-read mapping programs Bowtie (27) and SRST2 (28) were used. R version 3.2.0 was used for both statistical analyses and the Venn diagram, using the "venn" function from the "gplots" package (www.r-project.org).

Molecular typing results of suspected IMD cases.
Real-time PCR confirmed the presence of both meningococcal DNA and serogroup B capsular polysaccharide synthesis genes in all case-associated specimens received in the Irish Meningococcal and Meningitis Reference Laboratory (IMMRL). Case B and H isolates were identified as B:4:P1.7-2,4:ST-6697(cc41/44) with FetA VRs F1-21 and F5-12, respectively (Table 1). Case C and E isolates were identified as ST-6697(cc41/44). Incomplete MLST profiles were obtained for cases A (aroE and pgm) and F (abcZ and pgm) but were identical at the typed loci to those in the outbreak strain ST6697 (Table 1). All six strains shared the PorA VR type P1.7-2,4. A comparison of the ST6697 strains from this outbreak and the cc41/44 invasive isolates in the Meningitis Research Foundation Meningococcus Genome Library (MRF-MGL) revealed the rare nature of this ST, where all outbreak-related isolates clustered together and branched off a clade composed of ST154 or ST154 variants (Fig. 1).
Carriage: age and dwelling types. Pharyngeal swabs were obtained from 112 family members, yielding 14 cultures of N. meningitidis. One further N. meningitidis-positive carrier was identified by real-time PCR of swab head solutions, resulting in a carriage rate of 13.4% (n ϭ 15/112) (29) ( Table 2). N. lactamica was isolated from 31.25% (n ϭ 35/112) of pharyngeal swabs, and Neisseria mucosa was isolated once (0.9%). Both N. meningitidis and N. lactamica were isolated from 3.6% (n ϭ 4/112) of individuals. The age distribution of this family was notable in that 66% were younger than 20 years of age (52.2% for the general Traveller population [n ϭ 29,573] and 27.2% for the general Irish population [2011 census data]).
The average occupancy for those living in houses (n ϭ 67) was 6.7 people with carriage rates of 7.5% (n ϭ 5/67) and 19.4% (n ϭ 13/67) observed for N. meningitidis and N. lactamica, respectively. Higher carriage rates of 22.2% (n ϭ 10/45) for N. meningitidis and 48.9% (n ϭ 22/45) for N. lactamica were observed in trailers (n ϭ 45), where the average occupancy was 7.2 people per dwelling. An overall effect of housing type for the carriage of N. meningitidis (P Ͻ 0.05) and N. lactamica (P Ͻ 0.002) (2-sided Fisher's exact test) was observed. The effect of housing type and age-specific carriage was also investigated for both species (see Table S1 at http://imsrl.zohosites.com/files /Suplemental_Materials_MulhallRM_JCM00881-16.pdf). In general, carriage was more likely in trailers than in houses.
Dwelling density as a factor for carriage was also explored. Those dwelling in homes with Ն8 people were more likely to carry N. lactamica (odds ratio [OR], 3.79; 95% confidence interval [CI 95 ], 1.44 to 9.94; P ϭ 0.005), while a similar but weaker effect was observed for N. meningitidis (OR, 3.26; CI 95 , 0.84 to 12.63; P ϭ 0.067). Age-specific dwelling density was explored for both N. meningitidis and N. lactamica, where generally carriage was more likely for those in a dwelling with Ն8 occupants (see Table S2 at http://imsrl.zohosites.com /files/Suplemental_Materials_MulhallRM_JCM00881-16 .pdf). Higher average bedroom occupancies were observed for trailers (n ϭ 4.49 people) than for houses (n ϭ 1.95 people). As bedroom occupancy increased, carriage rates for both N. meningitidis and N. lactamica also increased (see Table S3 at http: //imsrl.zohosites.com/files/Suplemental_Materials_MulhallRM _JCM00881-16.pdf).
The high meningococcal carriage (45%, n ϭ 9/20) in the 25-to 39-year-olds may be a consequence of the especially crowded living conditions experienced by several individuals in this age group. Positive carriers (n ϭ 9) and negative carriers (n ϭ 11) had average occupancies per dwelling of 8.1 people and 5.2 people, respectively, and average per bedroom occupancies of 4.7 people and 2.1 people, respectively.
Although a likely confounding factor, smoking could not be considered because data on smoking were not collected. It should be noted that carriage studies are point estimates of carriage and may or may not be representative of carriage rates over longer periods of time. Care must be exercised when inferring from the associated data collected, as living arrangements may also be subject to variations over time. Further, several of the cohort had been offered chemoprophylaxis in the previous 4 months (n ϭ 39/112).
Comparison of ST6697 isolates. The relatedness of 11 ST6697 isolates, based on an allele-by-allele comparison of the core N. meningitidis genome (1,605 loci), was determined (Fig. 4). This included eight carriage isolates (December 2013), outbreak case B and H isolates (November 2010 and 2013, respectively), and one ST6697 isolate from an individual from the general population living in town B and predating the outbreak (February 2009). The latter isolate differed from the case B isolate at 41/1,605 loci, and both isolates were more related to each other than to the other meningococcal isolates. Carriage isolate M1 showed the greatest similarity to the most recent invasive isolate, case H, differing at 13/1,605 core loci (0.8%), representing 49 nucleotide differences. This very likely represents direct transmission among family members of the same strain that caused invasive disease 1 month earlier. Isolates M2, M13, and M34 represent three variants of another meroclone, grouped very closely to each other but not to any other isolates, again representing direct person-to-person transmission.
Case B and H isolate comparison. Whole-genome MLST comparison of isolates from case B (March 2010) and case H (November 2013) revealed that 8.5% of alleles were dissimilar (163/ 1,887), with 71.8% (117/163) of this allelic variation found in the core genome. Eight regions that had five or more consecutive variant alleles were identified, suggesting that these alleles were imported by HGT. Many of these alleles encoded gene products with known or presumptive metabolic functionality. The largest region  NEIS2153) and, on the basis of sequence comparison, was likely imported from cc162 meningococci (data not shown). Interspecies recombination. N. meningitidis and N. lactamica were cocultured from pharyngeal swabs on four occasions. Pharyngeal swabs M31, M35, M45, and M46 were volunteered by individuals dwelling in trailers, who were aged 32, 25, 6, and 3, and years, respectively. ST6697 isolates from pharyngeal swabs M45 and M46 were comparatively dissimilar to the other contemporaneously isolated meningococcal ST6697 strains as a result of cocolonization and HGT with N. lactamica (Fig. 4).
A whole-genome comparison (1,887 loci) of two N. lactamica isolates (ST10984) and two meningococcal isolates (ST6697) cultured from their respective pharyngeal swabs was undertaken (Fig. 5) (F5-43), a transferrin binding protein, tbpB, and a prepilin peptidase, pilD. Many of these shared alleles were observed adjacent to each other in the meningococcal genomes, suggesting acquisition via HGT. Short-read mapping-based sequence assembly tools Bowtie and SRST2 verified that neither contamination nor misassembly was the cause of this higher than usual sharing of alleles among the species.
Nineteen N. lactamica isolates were genome sequenced; 16 of these were identified as ST10984 and shared between three and five identical alleles with both the invasive ST6697 isolates and the four carried meningococcal ST6697 isolates not found cocolonizing with N. lactamica. All four cocolonizing ST6697 isolates contained a greater number of shared alleles (see Fig. S1 at http: //imsrl.zohosites.com/files/Suplemental_Materials_MulhallRM _JCM00881-16.pdf). Together these isolates likely represent donor and recipient strains before and after HGT.

DISCUSSION
Although rare in many settings, meningococcal serogroup B outbreaks can be both protracted in nature and highly variable in scale, some prompting the vaccination of millions of individuals, for example, the experiences in New Zealand (30), Chile (31), Cuba (32), and Norway (33). Outbreaks have been described in institutional settings, and these too can be protracted; for example, an outbreak in a crèche in Warwickshire, UK, lasted 4 months and was caused by B:ST-1194(cc41/44) meningococci (34). Other high-risk settings include military camps and universities, exemplified by the recent outbreak at Princeton, where the outbreak strain (B:ST-409:cc41/44) persisted for 12 months and spanned two academic years. In the current outbreak, the ST-6697 outbreak strain persisted in the Traveller cohort for at least 44 months, and 1 month after the final case, significantly higher carriage rates of N. meningitidis and N. lactamica were observed in the inhabitants of the trailers, which were more densely populated. This is consistent with many observations made since Glover described a meningococcal epidemic in 1917 at "X" Depot where, as the number of cases increased over the winter months, carriage rates increased dramatically (35). After the number of military personal beds per barrack room was decreased, the epidemic was much less severe 1 year later as this intervention impacted the carriage of and disease caused by the meningococcus (35).
In the Traveller cohort investigated here, the high carriage rates presumably facilitated persistence of the outbreak strain and subsequent transmission to infants and young children. The high compliance for chemoprophylaxis and vaccination (Bexsero) (36) probably cleared the strain, as no new cases have occurred since November 2013. It is known that chemoprophylaxis can prevent secondary cases, but controlling such outbreaks effectively requires mass chemoprophylaxis and simultaneous vaccination to protect individuals should chemoprophylaxis fail to purge the strain from the larger network. Before these measures can be implemented, the outbreak must first be recognized, which ideally requires real-time high-resolution characterization of IMD case isolates and a rapid and reproducible way of relating isolates to each other and to endemic disease strains. Specific national recommendations for high-risk groups might also be conducive to early outbreak recognition.
Conventional MLST, augmented by antigen typing (porA and fetA), has sufficient discriminatory resolution for both national and international epidemiological analysis and has been routinely implemented (37). However, isolates from outbreaks are typically indistinguishable by this approach, given the short periods of time involved to accumulate variation, and so this scheme is less useful for understanding the epidemiology of very closely related isolates. The application of WGS data allows the expansion of the MLST principle (38) to include more genes, for example, the core meningococcal gene scheme (n ϭ 1,605) (39). This provides high discrimination and greater robustness for epidemiological analysis. The present study illustrates the capacity of core genome  MLST (cgMLST) to unambiguously resolve a group of extremely similar outbreak strain isolates, including those recovered from person-to-person transmission. Consequently, whole-genome MLST may be applied to investigate the transmission dynamics of pathogenic bacteria in human cohorts. For example, the elucidation of transmission chains among permanent hospital staff may provide an opportunity for intervention, by revealing how staff or their working environments facilitate the persistence of antimicrobial-resistant strains and their subsequent transmission to patients.
Horizontal gene transfer (HGT) is common among meningo-cocci, and its complicating effects on strain characterization have long been recognized (6,39). Evidence of inter-and intraspecies HGT can be observed by the differential allele frequencies that exist within and among different Neisseria species (http://pubmlst .org/neisseria/). In addition, shared nucleotide sequence polymorphisms are present in numerous loci, including those encoding certain common antigenic loci such as the typing target FetA (40). Here we observed inter-and intraspecies HGT, including extensive introgression from an N. lactamica strain to a meningococcal strain present in the same individual. This is, to our knowledge, the first in vivo example of interspecies recombination and demonstrates how extensive HGT can be. At least 3% of the nucleotide content of the introgressed meningococcus isolate M46.3 had been replaced with N. lactamica DNA, including a large single event which contained 26 newly acquired alleles. This demonstrates that care is required in the investigation of outbreaks caused by highly recombinogenic organisms, as many genetic differences can be introduced rapidly by a few HGT events. Understanding the nature of genetic variation among such isolates is therefore important. Alleles can vary by a single nucleotide, which is likely the consequence of point mutation, whereas multiple nucleotide differences are likely the consequence of HGT with a different strain or species. Allelic variation may occur from nucleotide insertions/deletions resulting from HGT with sister cells, which was commonly observed among the ST6697 carried isolates that were studied here. The study further demonstrated the effect of housing density on the carriage of N. meningitidis and N. lactamica and the way that high carriage facilitated strain persistence. It also emphasized and trailer (C).

FIG 3
Neighbor-Net diagram of the 14 isolates recovered from nasopharyngeal swabs, generated from 1,605 core meningococcal genes. Included are eight ST-6697 isolates (cc41/44), three cc269 isolates, and one isolate each from cc22, cc461, and cc1157. the utility of WGS to augment conventional MLST typing and unambiguously discriminate extremely closely related isolates and illustrated the potential effectiveness of cgMLST as a novel public health tool for early outbreak detection. As this level of strain characterization becomes routine, further examples of HGT among Neisseria species will emerge. Understanding the dynamics of the antigenic surface structure exchange between Neisseria species has potential consequences for vaccine design and postimplementation monitoring, especially since the relationships between commensal bacteria and pathogenic species are poorly understood. N. lactamica carriage peaks in the first years of life when it is inversely correlated with low N. meningitidis carriage (41). Carriage of N. lactamica has also been demonstrated to confer protection against the development of IMD (42). Currently licensed me-ningococcal serogroup B substitute vaccines, the four-component vaccine Bexsero (36) and the bivalent Trumenba (43), utilize antigens which exhibit evidence of frequent interspecies HGT with commensal Neisseria (44), with the exception of the PorA component (Bexsero) which is unique among nasopharyngeal Neisseria species to the meningococcus. Neisserial heparin-binding antigen (Bexsero) is common to several Neisseria species, including N. lactamica (44).
In conclusion, this study shows how a single meningococcal strain with relatively high invasive potential can establish itself in a semiclosed population, causing a persistent disease outbreak. Although in this case the outbreak was very likely exacerbated by the particular setting, including individuals housed in trailers and the consequent close proximity of people of different age groups, this is very likely similar to what happens in larger population and on longer time scales, such as the Stonehouse outbreak in the UK in the 1980s and the New Zealand outbreak in the 1990s. The diversity of the outbreak strain over 4 years in spreading to different towns was limited, with the greatest degree of variation occurring by a limited number of HGT events, which resulted in the introduction of multiple variant alleles.