Multilocus variable-number tandem repeat analysis as a molecular tool for subtyping and phylogenetic analysis of bacterial pathogens

Tandem repeat (TR) loci are abundant in both prokaryotic and eukaryotic genomes. TRs could be polymorphic in the number of repeats across strains. The polymorphic loci, referred to as variable-number TR (VNTR) loci, may have two to tens of alleles across a population. Thus, combined multiple VNTRs can, theoretically, generate a large number of genotypes. VNTRs mutate at a wide range of rates, ranging from 10^-6 to 10^-4 mutations per generation for the highly mutable VNTRs in Escherichia coli O157:H7 and Yersinia pestis[1,2]. A combined-locus mutation rate for both organisms reaches up to 10^-3 mutations per generation. This mutation rate is close to the mutation rate of a bacterial genome with a size of 5 × 10⁶ base pairs, as nucleotides change at approximately 10^-10 mutations per replication [3], resulting in a rate of 10^-3 mutations per genome per generation. The mutation rate of VNTRs also varies across organisms. VNTRs in Mycobacterium tuberculosis change at an average rate of 2.3 × 10^-8 mutations per generation, which is 350-times slower than the rate in E. coli O157:H7 [4]. The wide range of mutation rates across VNTRs produces a similarly wide range of allelic diversity, making these markers useful for assessing genetic relatedness among strains evolved over various timescales [5,6]. Accordingly, multilocus VNTR analysis (MLVA) can be an appropriate molecular tool for the fine typing and phylogenetic analysis of bacterial pathogens.

MLVA as a subtyping tool for short-term epidemiology

Multilocus VNTR analysis possesses the advantages of extreme resolving power, robustness, high throughput data portability, ease of data interpretation and concordance with epidemiological data [7]. Another advantage is that, unlike pulsed-field gel electrophoresis (PFGE), MLVA is a PCR-based approach that requires only a small amount of DNA for analysis. Thus, a live bacterial isolate is not obligatory for a MLVA assay. MLVA has been developed for a variety of bacterial pathogens in an attempt to provide a resolving power better than that of PFGE for the discrimination of closely related isolates. PFGE, a discriminative universal subtyping tool for bacteria, has been adopted as a common subtyping means for the international surveillance network for food-borne diseases, also known as PulseNet International [8]. Despite the success of PFGE in detecting disease clusters, PFGE is, at times, insufficient to distinguish some epidemiologically unrelated isolates of highly homogenous organisms, including E. coli O157:H7 [9] and Shigella sonnei[10]. With its extreme resolving power for closely related isolates, MLVA has been widely applied for purposes of short-term epidemiology, including studies focused on disease outbreak investigation, forensic investigation, pathogen-source tracking and disease surveillance [11–14]. Since highly mutable or variable VNTRs change rapidly, single-locus variants could be seen occasionally among isolates from a common outbreak [9,10].

The resolving power of MLVA is mainly attributed to highly variable VNTRs. An MLVA assay based on a small set of highly variable VNTRs can have a discriminatory power parallel to, or higher than, PFGE [10,15]. Studies on E. coli O157:H7 [9] and S. sonnei[10] have demonstrated the satisfactory capability by MLVA in discriminating PFGE-indistinguishable but epidemiologically unrelated isolates. MLVA has also been demonstrated to be very useful in routine subtyping of S. enterica serovar Typhimurium for detecting and confirming outbreaks [15]. A study conducted in our laboratory reinforced the observation by Torpdahl et al. that an MLVA assay based on four highly variable VNTRs was very successful to resolve isolates of a predominant PFGE type, which accounted for 30% of 2840 S.enterica Typhimurium isolates recovered in 2004–2008 in Taiwan. Accordingly, MLVA based on a small set of highly variable VNTRs can be sufficient to provide better resolving power than most current molecular-typing methods, making MLVA an ideal subtyping tool for the routine analysis of large numbers of isolates for disease surveillance.

MLVA for phylogenetic analysis

Variable-number tandem repeats with higher diversity values are more useful in establishing phylogenetic relationships among closely related isolates, whereas less-variable VNTRs are more suitable in discerning phylogenetic lineages among distantly related isolates [6]. Theoretically, a larger set of VNTRs would be more favorable to obtain a clearer separation of distinct groups. In two studies, Neisseria meningitidis isolates were classified using 12 VNTRs, which yielded groupings similar to those obtained by multilocus sequence typing (MLST) [16,17]. However, some sequnce typing (ST) groups or ST clonal complexes were not clearly separated, which may have been a result of the limited number of VNTRs applied. By contrast, Pourcel and colleagues effectively grouped panels of Y. pestis isolates that were from different biovars using MLVA profiles based on 25 VNTRs [18]. In another study, a panel of 29 VNTRs was applied to establish clear groups for different serovars of E. coli[19]. In our laboratory, we successfully used MLVA data based on 36 VNTRs to establish clear phylogenetic patterns for different S. flexneri serotype groups [Wang YW et al., Unpublished Data]. One important consideration is how to establish accurate phylogenetic patterns using genetic markers such as VNTRs, as these markers have a wide range of diversity values. If VNTRs have high mutation rates, then they can differentiate closely related isolates, as well as increase the likelihood of homoplasy. To correct this problem, a nested hierarchical approach, known as progressive hierarchical resolving assays using nucleic acids (PHRANA), proposed by Keim and colleagues, can be useful [5]. PHRANA employs markers with increasing resolution in a stepwise fashion, beginning with low-resolution markers, the canonical (can)SNPs, then moving on to high-resolution markers, the MLVA-15 panel, and finally using the highest resolution markers, the signal-to-noise ration (SNR)-4 system. This concept of nested hierarchical approach can also be applied to other bacteria using data by stepwise subsets of VNTRs with differential levels of diversity values.

Development of MLVA

A bacterial genome may contain thousands of TRs, all of which are potentially polymorphic across populations. The process of exploring VNTRs involves in silico searching for TRs within the genome(s) of interest, selecting highly potent loci from the thousands of TRs, and then examining the polymorphisms of the selected loci in a panel of diverse strains. Examining polymorphisms of TRs is considerably laborious. In general, only a few dozen of the initial TRs, which bear a high number of repeats with greater sequence identity and lack indels within repeats, are selected. Highly variable VNTRs are typically easier to identify through examining their polymorphism in a small number of diverse strains. Conversely, VNTRs with low variability are difficult to identify, as they could only be identified in highly diverse populations.

As more and more genomic sequences become available, the work of developing a novel MLVA method could be alleviated through the use of a computer program, such as VNTRDB [20]. The VNTRDB program has been used very successfully in our laboratory to facilitate exploring VNTRs from multiple genomes [Wang YW et al., Unpublished Data] [10]. Although developing a novel MLVA method can be facilitated by the in silico searching for potential VNTRs from multiple genomes, it has at least two limitations. First, a common VNTR seldom exists in two or more organisms that are distantly related. Loci may be highly variable in an organism but invariable in others. It is, therefore, very unlikely that one could identify a set of highly variable VNTRs that could work for a variety of serovars or species. Attempts in developing a universal panel of VNTRs that could be used to analyze various serovars and pathovars of E. coli and Shigella species were not successful [21,22]. In addition, an effort to develop a universal set of VNTRs for subtyping and phylogenetic analysis of all serotypes within S. flexneri was not fully satisfactory [Wang YW et al., Unpublished Data]. Similarly, efforts to identify universal VNTRs with high resolving power for S. enterica serovars would likely be futile as the highly variable VNTRs identified for S. Typhimurium genomes were rarely found in the other 12 S. enterica serovars analyzed [Chiou et al., Unpublished Data]. Second, highly discriminative MLVA may not be obtainable for all bacterial species, as highly variable VNTRs are rare in some species, an example being Streptococcus pyogenes. VNTRs in 11 S. pyogenes genomes had been explored in silico in our laboratory and the searches found only one locus bearing a high number of repeats [Chiou et al., Unpublished Data].

Conclusion & future perspectives

Multilocus VNTR analysis will probably not fade out in the near future unless whole-genome sequencing, the ultimate typing method, becomes both cost effective and increasingly high throughput. Since MLVA is a cost-effective and high-throughput subtyping tool, it can be expected that an increasing number of MLVA protocols will be standardized to complement PFGE for the purpose of disease surveillance. MLVA is particularly applicable to the investigation of the global transmission of homogenous organisms that have currently no discriminative means for phylogenetic analysis. Although the development of MLVA can be facilitated by in silico exploration of VNTRs from multiple genomes, it has its limitations. MLVA with high-resolving power may not be obtainable for all bacteria because of the rarity of highly variable VNTRs in some organisms, including S. pyogenes. Unlike PFGE, a MLVA protocol may only work on a specific species or serovar. Attempting to develop a generic MLVA protocol that can process a number of serovars within a species, such as S. enterica, may be futile. Thus, MLVA is a ‘customized’ molecular subtyping and phylogenetic analysis tool for bacterial species but it cannot fully replace PFGE for subtyping analyses.

Financial & competing interests disclosure

The author has no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

No writing assistance was utilized in the production of this manuscript.