Ribotyping.

Ribotyping consists of cutting the target DNA with restriction enzymes, running the fragments with gel electrophoresis, transferring the separated fragments into a membrane, and then hybridizing them with labeled probes that are complementary to the targeted ribosomal RNA (rRNA) gene sequences (Southern blotting). This technique results in the visualization of the profile (ribotype) of the DNA fragments containing rRNA genes. For Y. pestis, EcoRI and EcoRV restriction enzymes were used to fragment the DNA. It was first used to assess global Y. pestis diversity in 1994 [19], and in 1997 the diversity of Y. pestis strains from Madagascar was investigated using this technique [20]. Nineteen ribotypes were identified among global Y. pestis strains [19], with 4 of these found in Madagascar: B, Q, R, and T [20]. Of these, 3 (Q, R, and T) were only identified from Malagasy strains.

The Q, R, and T ribotypes appear to have emerged in different locations in the Central Highlands foci. The Q ribotype was first discovered in 1983 in 2 different locations: Ambovombe in Manandriana District and Vohiposa in Ambohimahasoa District. The R ribotype was first reported in 1982 from Andina in the Northern part of the Ambositra district, and the first strains of the T ribotype were detected in Talata Vohimena in Manandriana District in 1994 [20]. Studies of 131 strains isolated in Ambositra and Ambohimahasoa between 1940 and 1996 revealed 58.02% of B ribotype, 18.32% of R ribotype, 21.37% of Q ribotype, 1.53% of T ribotype, and 0.76% coinfection by R and Q ribotypes [20]. By investigating the correlation between each ribotype and case outcomes, a link between the R ribotype and serious symptoms resulting in fatal outcome was suggested [20] without further investigation. Because ribotyping requires isolation and manipulation of live strains, a very limited number of strains have been tested with this technique. It is not used anymore for this reason and its low level of resolution.

Direct observation of plasmid content and plasmid restriction fragment length polymorphisms (RFLP).

Most Y. pestis strains contain 3 plasmids: pPla/pPCP1 (approximately 10 kb), pCD1/pYV (approximately 70 kb), and pFra/pMT1 (approximately 100 kb) [21–26]; some contain even more [23,27–32]. The 3 main plasmids play an important role in virulence of Y. pestis because of the virulence factors encoded by them (e.g., type III secretion system [33], capsular antigen [34], and plasminogen activator [35]), and differences regarding their presence have been investigated by examining plasmid content directly or by observing plasmid restriction profiles. These methods consist, respectively, of extracting plasmids from strains and analyzing them directly with electrophoresis [23], or subjecting them to the EcoRV restriction enzyme to generate fragments of different sizes prior to analysis with gel electrophoresis [20,36,37]. Only 1 strain from Madagascar (vaccine EV strain) has been tested using direct observation of plasmid content and the size of its pCD1 plasmid (45 MDa) was determined to be similar to that of other 1.ORI strains, but smaller than that of strains from other countries (47 MDa, 48 MDa, and 49 MDa) [23]; no additional plasmids were detected in this strain. Using plasmid RFLP, 5 Y. pestis strains from Madagascar have been reported to have lost at least one of the main plasmids as evidenced by the absence of some fragments compared to a reference Y. pestis strain (strain 6/69) [20], with the specific identification of the lost plasmid(s) having not been the subject of further study. On the other hand, additional bands present in the profile of 15 other Malagasy strains [20,36,37] suggested the presence of supplementary plasmids, which has been confirmed in only 2 isolates (16/95 and 17/95). These novel plasmids conferred resistance to streptomycin for 16/95 [37] and to multiple antibiotics (ampicillin, chloramphenicol, kanamycin, streptomycin, spectinomycin, sulfonamides, tetracycline, minocycline) for 17/95 [36]. The drawbacks of direct observation of plasmid content and plasmid-RFLP are their lack of resolution and the need to isolate and manipulate live strains.

Whole-genome sequencing (WGS).

Since the first Y. pestis genome became available in 2001 [38], it has been possible to compare multiple Y. pestis genomes. In Madagascar, WGS has been used mainly to identify genetic markers of evolution, such as single-nucleotide polymorphisms (SNPs), in order to use them to genotype strains that have not been sequenced [3,39,40]. Increasingly, all the strains in a study are sequenced, with the resulting whole genome sequences used to directly call SNPs and to infer the phylogeny [41,42]. In that case, the genotyping process is more exhaustive because it captures the diversity of the entire genome sequence of all the strains studied. This technique has also been used to identify genetic determinants that could support new phenotypic characteristics, such as antibiotic resistance [11]. Indeed, 3 AMR strains from Madagascar [11] acquired resistance to streptomycin via the same spontaneous point mutation in the 30S ribosomal protein S12 (rpsL) gene, thereby documenting that AMR can also arise in Y. pestis via point mutations. Two of these strains were isolated from the same pneumonic plague outbreak and were 100% identical with WGS, also documenting for the first time that AMR strains can spread person-to-person via primary pneumonic plague [11], which is a possible scenario that might occur if Y. pestis was utilized as a bioweapon [43]. More recently, WGS of Y. pestis isolates was used to characterize local plague activity in multiple endemic rural foci from August to November 2017 [44], with subsequent phylogenetic analyses assigning isolates to the same distinct Y. pestis lineages that have been persisting in those different foci for many years [39]. Overall, this technique has the advantage to reveal the diversity of the strains included in a study in their entirety. However, the downside is that sequencing all the strains in a given study is still quite expensive, whether for reagents and equipment, or for shipping samples elsewhere for sequencing. This is why in most of the studies undertaken in Madagascar until now, only a subset of the strains included in the studies have been sequenced.

Clustered regularly interspaced palindromic repeats (CRISPR).

A CRISPR locus is a genetic structure made up of conserved repeat regions separated by unrepeated elements called spacers. A 300 to 500 bp sequence called the “leader” is adjacent to and located at the end of CRISPR loci. The repeated sequences as well as the leader sequence are conserved within a bacterial species and are different between 2 distinct species [45]. The existence of 3 CRISPR loci—YP1, YP2, and YP3 (sometimes also referred to as loci A, B, and C, respectively)—with the same 28 bp repeat sequence in the Y. pestis genome was first described in 2002 [45]. For YP1 and YP2, the first repeat is truncated (19 bp for YP1 and 22 bp for YP2) [46]. The study of CRISPR in Y. pestis led to the formulation of 3 rules for the evolution of CRISPR loci [46], which were then subsequently used to study the evolution of the strains carrying them. First, deletion of one or more spacers can occur and is random. Second, it has been shown that the acquisition of new spacers occurs in a polarized fashion and is only on one side (between the leader sequence and the last repeated sequence), but before that, the repeated sequence directly upstream of the leader sequence is first duplicated and added just in front of the leader sequence. Third, the occurrence of the same spacers in a CRISPR allele indicates a common ancestral origin rather than arising from independent events.

For the study of CRISPR in Y. pestis, 2 different aspects of the same spacer nomenclature exist. The first consists of assigning a letter to each spacer according to the order of discovery. For YP2 and YP3 loci, the number 2 or 3 is added to the letter, respectively [46]. Each combination of spacers corresponds to an allele (for example, the allele of YP1 locus of Y. pestis strain CO92 is: abcdefgh). In 2005, based on this nomenclature, Pourcel and colleagues identified and named twenty-one, nine, and three alleles for YP1, YP2, and YP3, respectively, in Y. pestis strains isolated worldwide, and also reported the first CRISPR genotyping of Malagasy strains [46]. The second aspect consists in adding a number after a, b, or c depending on the locus considered (A, B, or C, respectively). Because letters are in finite number (26, a–z), this second aspect of the same nomenclature was used more often as new spacers continued to be described [47]. Seventeen other spacers (13 for locus A, 2 for locus B, and 2 for locus C) that appear to be specific to Madagascar were discovered during the analyses performed on clinical samples from plague patients during the 2007 plague season [48] and named using this nomenclature: a89, a90, a91, a92, a93, a94, a95, a96, a97, a98, a99, a100, a102 for locus A spacers; b5′ and b50 for locus B spacers; and c11 and c13 for locus C spacers. Taken together, the alleles of YP1, YP2, and YP3 constitute a genotype.

One advantage of CRISPR is in the correlation found between CRISPR genotypes/spacers and geographic regions, helping to discern population structures and phylogeography of Y. pestis [49]. Another advantage is that, unlike the above genotyping methods, CRISPR genotyping can be performed directly on DNA extracted from complex biological samples without the need to isolate a strain [48,50]. This method was therefore used to genotype Y. pestis-positive DNA extracts obtained from human and animal biological samples during the investigation of the 2011 pneumonic plague outbreak in the Ambilobe district (an area outside the known plague foci located in northern Madagascar) during which a new Madagascar-specific spacer of the YP3 locus was discovered at 2 different locations separated by around 400 km (Ambarakaraka in Ambilobe district and Bealanana in Bealanana district) [50]. Based on the discovery of 4 CRISPR genotypes circulating locally, it was concluded that Ambilobe was a natural plague focus long before 2011 [50]. The most frequent genotype, assumed to have caused the outbreak, was also found in Ankazobe, suggesting anterior expansion from the Central Highlands [50]. However, the disadvantages to this technique are its low discriminatory power, the cost of sequencing for spacers identification, and the possible exceptions to the polarized acquisition of spacers as described in other bacteria [51], thereby leading to the failure of CRISPR to properly reflect evolution.

Multiple-locus variable number tandem repeats (VNTR) analysis (MLVA).

MLVA, as indicated by its name, implies the analysis of multiple VNTR loci. VNTRs, also called minisatellites or microsatellites depending on the number of nucleotide(s) repeated, are tandem repeated sequences in which variability at a given VNTR locus concerns the number of repeats that differ among strains. Therefore, genotyping with this method consists in determining the number of repetitions at each locus for each strain, each number representing an allele. Consequently, a VNTR locus can have many different alleles. Starting in 2011, a hierarchical SNPs-MLVA system containing 43 VNTR loci identified from the early Y. pestis genomes [52–54] emerged as the pioneering technique to unveil the extensive diversity exhibited by Y. pestis in Madagascar, successfully identifying a total of 226 genotypes from the 262 examined strains [55].

MLVA genotyping is useful because, like CRISPR, isolation of Y. pestis strains is not mandatory, it can be performed with DNA extracted from different types of human and environmental samples [48]. The high mutation rate of VNTRs explains the discriminatory nature of MLVA, enabling different strains of a clonal species such as Y. pestis to be distinguished. It was also the first technique that allowed an extensive phylogeographic study of Y. pestis in Madagascar, providing insights into the distribution of different genetic groups across years and throughout the different plague foci [55]. Many discoveries have been made using this technique. Among them, the identification of 11 and 4 subclades, respectively, in major phylogenetic groups I (I.A-I.K) and II (II.A-II.D). These 15 subclades appeared to be geographically separated [55], suggesting that a specific genotype was initially introduced into a given area with subsequent local differentiation (founder effect). The previous studies on these MLVA subgroups concluded that strains assigning to group I were found across all the Malagasy plague foci (Central and Northern Highlands and Mahajanga), whereas strains assigning to group II were only found in the Central Highlands in plague hotspots such as the Betafo, Manandriana, Ambositra, and Ambatofinandrahana districts [55]. Within group I, subclades found in the Northern Highlands foci (I.C, I.G, and I.I) are different from those found in the Central Highlands (I.A, I.B, I.D, I.E, I.F, I.H, I.J, and I.K). Regarding the coastal focus of Mahajanga, Y. pestis found there is dominated by the I.A subclade, even forming a subcluster called the Mahajanga I.A subclade [55]. This high fidelity between geographic locations and specific subpopulations of Y. pestis means that genetic typing can be used to identify rare transfer events, such as multiple introductions of Y. pestis from the Central Highlands to the coastal city of Madagascar, where it persisted for some years before eventually dying out [40,41]. Finally, Ambositra was among the most diverse districts in terms of Y. pestis genotypes, with 6 MLVA subclades isolated from there. Based on this finding, it was suggested as the district of origin for the I.A subclade as the earliest strains assigned to this subclade to date were isolated there [55].

The high mutation rate of VNTRs is both a strength and a weakness of these markers. A strength because of the high resolution it provides among even very closely related strains, but also a weakness because, in some cases, its inherent high mutation rate may lead to mutational saturation and homoplasy distorting the reconstruction of the phylogeny [56] or yielding low bootstraps in phylogenies [55]. Mutational saturation is the state reached when additional mutations do not bring additional genetic diversity to the locus anymore meaning that studying that locus does not give comprehensive information on the evolution of the species because some mutations are not detectable and balanced out by recurrent mutations [56]. Due to these phenomena, MLVA also sometimes fails to identify deeper relations like those between the 2 major groups [55], and mutations in VNTRs loci may even happen during laboratory cultivation making the strains studied not completely identical to the original ones [57]. To overcome these problems, a hierarchical approach is often adopted: SNPs are used first because they have lower mutation rates and so can resolve the major groups, and then, MLVA is used to split strains assigned to these major groups into subgroups because of their higher resolution that can discriminate more closely related strains.

Single-nucleotide polymorphisms (SNPs).

SNPs are substitutions involving only a single nucleotide and, in general, Y. pestis genotyping using SNPs can be done in 2 different ways. The first consists in determining the allele carried by each strain studied by amplification with specific primers (polymerase chain reaction or PCR) [58]. This first technique can only be applied to SNPs already described [58] or newly described following the whole genome sequence comparisons of a subset of strains [39,40,55] and, in general, is not exhaustive in the sense that if other evolutionary informative SNPs are present in the strains not sequenced, those will not be considered during genotyping, creating what is called a “discovery bias” [59]. The second technique consists in WGS of each of the strains included in the study as reported above; however, a bias persists given that sequencing all the strains existing in Madagascar is often practically impossible.

SNPs genotyping has been particularly important in the case of Y. pestis in Madagascar because it is the only technique capable of reliably distinguishing the 2 major groups (group I and group II) circulating in the country [3] differentiated by the Mad-43 allele [position 1,348,724 in the Y. pestis strain CO92 genome (accession NC_003143.1), ancestral state (C) for group I and derived state (T) for group II] [55], and one of the techniques that really revealed the extensive diversity of Y. pestis in Madagascar. Indeed, new subgroups (y, z, ɑ, and β in the group I; t, u, v, w, and x in the group II) and new nodes within different subgroups could be identified with the identification of new SNPs as more strains were whole-genome sequenced [39,44]. This higher level of resolution allowed the subsequent studies of the distribution of populations across foci. Genotyping Y. pestis isolates obtained from the same locations across multiple years provided insights into plague persistence in the environment, including identification of the specific Y. pestis subpopulation that persists in a given location, and this background information is important for determining the likely geographic origins and spread of plague activity in humans. Although multiple hypotheses have been proposed to explain it (reviewed in [60,61]), plague persistence in the environment is still poorly understood in most global regions where it occurs. In Madagascar, plague is highly endemic in rodent populations in rural foci throughout the Central and Northern Highland [62], leading to regular spillover into humans [9], so it was possible to use MLVA and SNPs to genotype >750 Y. pestis isolates collected across multiple years from the same rural foci [39]. The results revealed that distinct Y. pestis subpopulations are found in different rural foci in the highlands and persist in those locations across years, documenting that—at least in Madagascar—plague persists locally in rat and flea populations [62]. And again, the invaluable insights into the phylogeography of Y. pestis in Madagascar provided by these studies enabled the detection of transfers with their directionality as described for the s subgroup [40]. It is thought to have emerged in the Ambositra district where s1 strains (the most ancestral node within s subgroup) were found and then spread successively to other districts: Soavinandriana (s2 and s3), Miarinarivo (s3 and s4), Arivonimamo (s3), Antananarivo (s3, s5, s8) and Manjakandriana (s3 and s5), Antanifotsy (s5), Fianarantsoa (s5), and Mahajanga (s5, s8 and s9) [40]. Following studies also found that MLVA and SNP subgroups were highly congruent (7 out of the 13 MLVA subgroups in group I showed congruence with SNP subgroups and 6 out of the 7 for group II [39]), making it possible to only screen SNPs for these congruent subgroups, which is easier. It can also be applied to clinical samples [48,50], although it has been reported to have lower discriminatory power compared to MLVA or CRISPR in a previous study [48]. A summary of the distribution of genetic diversity of Y. pestis (according to isolation time range and geographical location) in Madagascar using MLVA and SNPs techniques is provided in Table 2 below.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Distribution of genetic diversity of Y. pestis in Madagascar.

https://doi.org/10.1371/journal.pntd.0012252.t002

Finally, dating a phylogenetic tree generated using SNPs is possible when metadata, such as date of strain isolation, is available, and results can provide new insights into the epidemiology of the infectious disease of interest. Recent studies have focused on dating phylogenetic trees of Y. pestis to determine the divergence times of the different clades [39,41]. The authors studied the transmission events between Madagascar and other countries (from India to Madagascar, from Madagascar to Turkey) and inside Madagascar (between the Central Highlands and Mahajanga) [41]. For inter-countries transmissions, the median tMRCA (the time to the most recent common ancestor) of 1.ORI2a (Indian strains) and 1.ORI3 (Malagasy strains) is estimated at 1877 meaning that these 2 clades probably separated around that time [41]. Regarding the transmission event to Turkey, the tMRCA of the 1.ORI3 strains from Turkey and Madagascar has been estimated to be 1905 [41]. As 1.ORI3 is present exclusively in Madagascar and Turkey, the authors emitted the hypothesis that the lineage first introduced into the island was 1.ORI2a, and then 1.ORI3 emerged probably following transmissions and adaptation to local reservoirs and subsequently spread to Turkey from Madagascar [41]. Regarding transmissions between different foci in Madagascar, 5 transfers have been identified between the Central Highlands and Mahajanga, confirming the hypothesis formulated earlier based on the study of MLVA [40]; four of them from the Central Highlands to Mahajanga and 2 out of the 4 would have been responsible of the 1990s epidemics [41]. These findings also reinforced the hypothesis of a reintroduction as a source of plague in Mahajanga rather than a reemergence [40,41].

Whole-genome DNA capture and enrichment.

Whole-genome DNA capture and enrichment techniques are typically used to amplify low level DNA signals before sequencing. They consist in first designing biotinylated probes targeting the organism of interest using a reference genome, extracting the DNA from the samples, building libraries, hybridizing the libraries with the probes, capturing the hybridized libraries-probes with streptavidin-coated beads, and then sequencing [63]. This technique was extensively used in ancient DNA studies [63–65] but is increasingly used in contemporary studies using complex clinical samples as starting point. This takes advantage of the technique’s ability to specifically capture the target genome and increase the signal in samples composed of DNA from multiple origins, thereby increasing the probability of detection of the pathogen of interest. In the case of plague in Madagascar, the first description of the use of this technique was in the recent study of the 2017 urban pneumonic plague epidemic [44] for which only few Y. pestis strains were available. It was of key importance to determine that the epidemic was caused by multiple introductions of Y. pestis from several different rural foci via travel of infected individuals into urban areas [44]. The main drawback of the technique is its cost.

Discovery bias: We could only find what we were looking for.

WGS is a relatively new technique that is still expensive, particularly for a low-income country like Madagascar. This, combined with the fact that to date it has only been available outside Madagascar, led to only a limited number of Y. pestis isolates from Madagascar being entirely sequenced and serving as a basis to the discovery of informative SNPs. The SNPs discovered from a limited set of whole-genome sequences was then used to infer phylogeographic patterns within a larger set of strains. This method is very powerful and cost-effective compared to sequencing all strains included in a study. However, a problem remains: the identified SNPs represent the diversity of the strains sequenced only, not the entire set of strains (nor the entire collection) therefore, if other SNPs exist in any of the strains that have not been sequenced, they will not be included in the reconstruction of the phylogeny, this phenomenon is called “discovery bias” [59]. The consequence of that bias is primarily in the topology of the resulting tree, which would have some branches collapsed, but also an underestimation of the real diversity of the strains [59]. This could be rectified by establishing the capability to regularly perform WGS of all new strains collected in Madagascar, as well as archived strains.

The unknown contribution of horizontal gene transfer (HGT) to the local evolution of Y. pestis in Madagascar.

Recent studies on the evolution of Y. pestis in Madagascar focused mainly on SNPs and MLVA techniques. These techniques are very reliable and allowed to answer important questions, particularly the origin of the outbreaks in Mahajanga [40,41], the distribution of Y. pestis genotypes across different plague foci, and their local evolution [39,55,66]. However, one aspect of the evolution cannot be grasped by these approaches, namely HGT. HGT is a very important aspect of evolution in bacteria, particularly for the members of Enterobacteriaceae family and even in eukaryotes [67], because it allows the acquisition of new features, such as novel virulence mechanisms, resistance to antibiotics, resistance to biocides. Three mechanisms of HGT are known in bacteria: conjugation, transformation, and transduction.

Although plasmid genotyping, as we mentioned earlier in this review, has initiated investigations into the difference in plasmid content of different strains (conjugation), such research was not subsequently pursued. Even though Y. pestis has been considered clonal and HGT via conjugation is rare in it, this type of transfer has still been documented, and was even at the origin of the species itself when it acquired its two novel virulence plasmids, pPla/pPCP1 and pFra/pMT1. HGT has also allowed some Y. pestis strains in Madagascar to acquire plasmids with resistance genes to one (streptomycin or doxycyclin) [37,68] or multiple antibiotics [36]. Other features acquired through HGT have also been described in Y. pestis isolated in other countries, such as arsenic resistance in Java 9 strain [29]. YpfФ is a prophage of 8.7 kb [69], most likely acquired through HGT also, that has been described in the chromosome of Y. pestis and reported to affect the fitness and the pathogenicity of Y. pestis [70]. Therefore, HGT is important in the evolution of Y. pestis and deserves to be better investigated in future research.