Use of a New High Resolution Melting Method for Genotyping Pathogenic Leptospira spp.

Background Leptospirosis is a worldwide zoonosis that is endemic in tropical areas, such as Reunion Island. The species Leptospira interrogans is the primary agent in human infections, but other pathogenic species, such as L. kirschner and L. borgpetersenii, are also associated with human leptospirosis. Methods and Findings In this study, a melting curve analysis of the products that were amplified with the primer pairs lfb1 F/R and G1/G2 facilitated an accurate species classification of Leptospira reference strains. Next, we combined an unsupervised high resolution melting (HRM) method with a new statistical approach using primers to amplify a two variable-number tandem-repeat (VNTR) for typing at the subspecies level. The HRM analysis, which was performed with ScreenClust Software, enabled the identification of genotypes at the serovar level with high resolution power (Hunter-Gaston index 0.984). This method was also applied to Leptospira DNA from blood samples that were obtained from Reunion Island after 1998. We were able to identify a unique genotype that is identical to that of the L. interrogans serovars Copenhageni and Icterohaemorrhagiae, suggesting that this genotype is the major cause of leptospirosis on Reunion Island. Conclusions Our simple, rapid, and robust genotyping method enables the identification of Leptospira strains at the species and subspecies levels and supports the direct genotyping of Leptospira in biological samples without requiring cultures.


Introduction
Pathogenic Leptospira bacteria are agents of leptospirosis, one of the most widespread zoonoses worldwide [1]. Leptospira spp. can infect a wide range of animal species, which serve as maintenance hosts, but rodents are the most important reservoir of Leptospira spp. Humans typically become infected via contact with urine-contaminated soil or water or, less frequently, by direct contact with the tissues or urine of animals [2]. Leptospira spp. have been subdivided and classified into more than 260 serovars that are grouped into serogroups based on shared antigens [1]. Molecular studies have helped to determine a classification system based on genetic similarities which is currently used in conjunction with antigenic classification [3]. This molecular taxonomy system classifies Leptospira into 21 genomospecies [4,5]: nine pathogenic species, five intermediate species, and seven saprophytic species [3]. In regions of endemicity, the identification of the infecting strain can help to determine the infectious animal source, and thus aid in the implementation of control measures [2]. Various molecular techniques have been described for studying the molecular epidemiology of Leptospira, including pulsed-field gel electrophoresis (PFGE) [6], restriction enzyme analysis (REA) [7], ribotyping [8], randomly amplified polymorphic DNA (RAPD) [9], restriction fragment length polymorphism (RFLP) [10], fluorescent amplified fragment length polymorphism (FAFLP) [11], 16S rRNA sequencing [12] and several PCR-based approaches. Unfortunately, all of these methods suffer from significant drawbacks [13]. In recent years, two genotyping methods, multiple loci variable number of tandem repeats analysis (MLVA) [13] and multilocus sequence typing (MLST) [14], have emerged as noteworthy PCR-based genotyping techniques, but one major limitation of these methods is their reliance on bacterial cultures to generate a significant amount of highpure DNA [14,15]. Moreover, MLVA requires the determination of the PCR product size, which requires agarose gel electrophoresis, a time-consuming technique with a poor resolution of amplicon size. MLST requires high-throughput sequencing and bioinformatics, but provides a portable, reproducible, and scalable typing system [14,16].
High-resolution melting (HRM) analysis [17], which measures the melting temperatures of amplicons in real time using a fluorescent DNA-binding dye, is a closed-tube, non-sequencing-based system for genotyping and mutation scanning [17,18]. This method has emerged as a valuable tool for the rapid testing of diverse biological specimens and tissues for the presence of micro-organisms and for the differentiation of Brucella spp. genetic variants [19], for example, Chlamydiaceae [20], Mycoplasma pneumoniae [21], Leishmania [22], Bordetella pertussis [23], Staphylococcus aureus [24], Bacillus anthracis [25], Mycoplasma synoviae [26], Pseudomonas aeruginosa [27] adenovirus serotypes [28], and Aspergillus species [29]. HRM has even been used for the identification of members of the Anopheles funestus group [30]. Tulsiani et al. [31] developed a random amplification of polymorphic DNA (RAPD) method associated with HRM analysis (RAPD-HRM) using 13 previously published RAPD primers to genotype 10 Leptospira strains. Traditional HRM curves are difficult to interpret, thus, the interpretation of HRM results can be arbitrary. In this study, we evaluated the potential of an alternative method based on unsupervised high-resolution melting curve (HRM) analysis using Screen-Clust HRM (Qiagen, Courtabeuf, France) and examined the ability of this new method to type Leptospira from human Reunion Island specimens rapidly and easily.
The DNA extractions from the in vitro cultured strains and sera were performed using the NucliSens easyMAG system (BioMérieux, Marcy l'Etoile, France). The concentration of the Leptospira DNA was normalized using the quantification analysis module of a RotorGene 6000 cycler (Qiagen, Courtabeuf, France). To normalize the input template concentration for the strains, the Ct (threshold cycle) was set to be between 27 and 30 cycles.
The sensitivity of each PCR using the selected primer(s) was evaluated by performing PCRs on 10-fold serial dilutions of the DNA extracted from three in vitro cultured strains from Mayotte: L. interrogans sg. Pyrogenes, L. borgpetersenii sg. Mini, and L. kirchneri sg. Mini/Hebdomadis. The quantification of the leptospires was based on genomic DNA mass, taking into consideration that the size of the genome of the L. interrogans strain Fiocruz L1-130 is 4.6 Mb (1 genome is approximately 5 fg). The end-point detection limit was determined by 10 repetitions of the measurement of the last positive point with 100% amplification.
We also cloned a plasmid DNA carrying the G1/G2 gene of L. interrogans sv. Copenhageni strain Wijinberg into the pGEM-T Easy Vector (Promega, USA).
The reproducibility of our method (which refers to the variation in our results between runs) was evaluated by ten repetitions of PCRs with the selected primer(s) for three strains (L. kirschneri sv. Grippotyphosa str. Moskva V, L. borgpetersenii sv. Hardjobovis str. Sponselee, and L. interrogans sv. Icterohaemorrhagiae str. Verdun). For each strain, if the species identification was identical in the ten repetitions, then the reproducibility was determined to be 100%.
The PCRs were performed using a Type-IT HRM PCR Qiagen Kit (Qiagen, Courtabeuf, France) on a RotorGene 6000 system (Qiagen, Courtaboeuf, France). The 20 μl reactions contained 10 μl of mix 2X master mix HRM, with a 0.7 μM final concentration of each primer (TibMolBiol, Berlin, Germany), and 5 μl of the extracted nucleic acid solution. The following amplification protocol was used: denaturation at 95°C for 5 min, 45 cycles at 95°C for 10 s, 55°C for 30 s, and 72°C for 10 s. These conditions were used for all the primer pairs. For the species determination, a melting curve analysis determined the melting temperature (Tm). To test the reproducibility of the Tm determination, three strains of different species were tested in 10 separate runs with the selected primers. For the subspecies determination, after PCR cycling with the VNTR primers, the samples were heated from 65°C to 95°C with continuous acquisition. For each VNTR locus, a normalization region of the melting curve was selected to improve the analysis per the recommendations of the Rotor-Gene ScreenClust HRM Software User Guide (Qiagen, Courtabeuf, France). Two normalization regions were selected: one before and one after the melting curve transition. The highest fluorescence value was 100 and the lowest fluorescence value was zero. The data were analyzed using ScreenClust HRM (Qiagen, Courtabeuf, France) in unsupervised mode (i.e., there were no known controls for each cluster and the number of clusters was unknown). The unsupervised mode was also employed to analyze Leptospira DNA fluorescence data obtained with human samples, and the results from the control strains were employed as pseudo-unknowns. The data were analyzed by ScreenClust using the principal component analysis statistical method [41], which enabled the maximum separation of genotypes.
A cluster plot was built in three dimensions (principal components) and ellipses representing each cluster were drawn. The probability of each sample fitting into a specific cluster was calculated and required to be at least 0.7. The typicality, which measures how well a sample falls within the cluster to which it has been assigned, was also given for each sample and was required to be at least 0.05.
The genotype [42] of each Leptospira DNA was determined by combining the results of the two clusters that were obtained with the two VNTR primers selected by the species in the subspecies characterization.
To evaluate the ability of the tested primers to correctly identify the Leptospira species, a Mann-Whitney test was used to determine whether the difference in the mean Tm was significant between species.
The typeability of each set of primers for each species of Leptospira (i.e., the proportion of strains that were assigned a type by the typing system) was determined by using the following formula: T = Nt/N, where Nt is the number of strains that were assigned a type and N is the number of strains that were tested [5].
To evaluate the discriminatory power of the VNTR primers, the Simpson diversity index (DI) was used as described by Hunter and Gaston [43]. 1). The polymorphism is considered to be high when the Simpson diversity index is greater than 95% [42,44].

Determination of species
Our results are shown in Table 1 and Figs 1, 2 and 3. In all the runs, the negative control, a DNA extract from an in vitro culture of L. biflexa sv. Patoc, did not show amplification with lfb1F/R nor G1/G2. Seven primers were tested on 16 strains to characterize each species, including L. interrogans, L. borgpetersenii, L kirschneri, and L. noguchii. Only the association of the primer pairs lfb1F/R and G1/G2 resulted in a species-specific Tm (except L kirschneri, which had no amplification with G1/G2) and allowed for the discrimination of the four species (p<0.001); thus, these two primer pairs were selected for further analysis.
The reproducibility (evaluated on ten runs) was 100% for the three tested strains using the two primer pairs lfb1F/R and G1/G2.
The sensitivity was evaluated for the two selected pairs of primers, lfb1F/R and G1/G2, using three strains. For each PCR assay (with the exception of G1/G2, which did not amplify the L. kirschneri strain) the two sets of primers showed that 100% positive amplification was obtained for an average of 5 genome equivalents per PCR (Ct 37 cycles).
Because one leptospire contains an average of five genome equivalents [49], our analytical sensitivity was calculated to be 1 bacterium/μl. This result was confirmed by using the plasmid DNA of the G1/G2 gene of L. interrogans sv. Copenhageni strain Wijinberg because we had obtained a limit of detection of 25 copies per PCR for L. interrogans sg. Pyrogenes.
The limits of detection on plasma were assessed using DNA that was extracted, diluted from identical strains and spiked. The limit of detection on the spiked plasma was estimated to be 5 genome equivalents per PCR, which was similar to that of the two sets of primers on the different strains and the extracted unspiked DNA.

Genotype determination of Leptospira reference strains using ScreenClust HRM Software
Of the 11 primer sets tested, three VNTR primer sets were considered to be noteworthy for discriminating the strains and were considered further: VNTR-Lb4, VNTR-4 bis and VNTR-Lb5 (Table 1, Figs 1, 4 and 5). Two VNTR primer pairs per species were used to genotype Leptospira at the subspecies level. Using these three VNTR primers, the probability of a strain belonging to a specific cluster was 0.95, and the typicality was 0.05 for all the strains.
L. interrogans. Using VNTR-Lb4, the typeability value obtained for the 27 L. interrogans strains tested was low (0.074); consequently, VNTR-Lb4 was not employed further. The typeability values for VNTR-4bis and -Lb5 were both 1; thus, these two VNTRs were selected for further study, and their normalization temperatures were 83.7-89.3°C and 84.5-88.4°C, respectively. The reproducibility of the PCR runs with the two primers VNTR-4bis and -Lb5 was 100%, which was tested on L. interrogans sv. Icterohaemorrhagiae str. Verdun.
The HRM analysis of the amplicons obtained with primers VNTR-4bis and -Lb5 was highly discriminating, with 17 and seven clusters generated, respectively, for the 27 different L. interrogans strains tested. By combining the results using VNTR-4bis and -Lb5, we found 22 genotypes in the 27 strains tested, with 19 strains having a unique genotype (DI = 0.98) ( Table 1).
The HRM analysis of the amplicons obtained with primerVNTR-Lb4 generated five clusters for the ten L. borgpetersenii strains, while VNTR-Lb5 discriminated six different clusters. Thus, in the analysis of the L. borgpetersenii strains, the combination of Lb4 and -Lb5 generated eight out of 10 genotypes, with six strains having a unique genotype (DI = 0.98) ( Table 1). The method could not discriminate sv. Ceylonica str. Piyasena from sg. Javanica, and sv. Nona str. Nona from sg. Hebdomadis, which were grouped into cluster 5, nor sv. Castellonis str. Castellonis 3 sg. Ballum from sv. Jules str. Jules sg. Hebdomadis, which were grouped into cluster 1.
Our method generated 37 genotypes in 48 reference strains. Thirty strains had a unique genotype while seven clusters contained two to three strains. Thus, the global DI of our method was 0.98.

Application of ScreenClust software to biological samples (Table 2, Figs 2 and 3)
The 42 Leptospira DNAs that were extracted from patient serum samples showed a Tm corresponding to that of the species L. interrogans using the primer pairs lfb1F/R and G1/G2.
Using VNTR-4 bis and VNTR-Lb5, the genotypes of 17 out of 42 (40.5%) DNA samples were identified. These 17 Leptospira DNA samples exhibited the same genotype, which was consistent with the strains L. interrogans sv. Copenhageni or Icterohaemorrhagiae.
Using VNTR-4 bis and VNTR-Lb5, the genotype determination was 100% when the Ct was 30 cycles (8250 copies/ml, data not shown) and 28.5% (10/35) when the Ct was >30. We then evaluated the integrity of three genotypable DNAs and three non-genotypable DNAs by performing a standard PCR targeting the secY gene. Agarose gel electrophoresis revealed multiple bands of different sizes for the non-genotypable DNAs resulting from nonspecific amplification, and a 202-bp band for the three genotypable DNAs. To understand how the DNA lost its integrity, we subjected three good quality DNAs to several freezing-defrosting processes. After each freezing-defrosting process, the migration of the amplified fragments showed progressive DNA degradation.

Discussion
The main aim of this study was not to diagnose pathogenic leptospirosis infection but to describe a new genotyping method to identify Leptospira species and subspecies. For a better understanding of the epidemiology of leptospirosis in a region, information about the prevalent circulating serovars or genotypes is essential.
The first objective of the present study was to provide a rapid technique to discriminate between species of pathogenic Leptospira (only the four most frequent pathogenic leptospires were considered [45]), which is a necessary step for further levels of genotyping, i.e., at the serovar level. Using two sets of primers (G1/G2 and lfb1 R/F), the melting temperature profiles of the strains are able to accurately discriminate L. interrogans, L. kirschneri, L: borgpetersenii, L. noguchii, and the newly described L. mayottensis [33] with a specificity and reproducibility of 100% with less than 0.5% Tm coefficients of variation. Our method enabled us to accurately determine the infectious Leptospira species from DNA extracted from the blood of sick patients at Reunion Island. Previous methods for the identification of Leptospira species, such as PCR amplification followed by electrophoresis on a non-denaturing polyacrylamide gel [46] or the  sequencing of partial 16SrDNAs [47], are time-consuming. In contrast, this method allows for species discrimination in a short time period. Recently described PCR methods for the diagnosis of acute infection [48] may have a lower end-point detection than our PCRs, consequently, a PCR-positive sample that has been identified with these highly sensitive methods might be unsuitable for characterization by melting temperature analysis. In a previous study, Merien et al. [35] used a similar melting curve analysis with the primer pairs lfb1 F/R, which facilitated the differentiation of the pathogenic species L. interrogans, L. borgpetersenii, and L. kirschneri, but not L. noguchii and L. kirschneri. In our study, the use of a second set of primers provided improved specificity. In a recent study, Feirrera et al [49] developed sets of species-specific probes with primers designed from ompL1 and secY with 100% analytical specificity for the species L. interrogans, L. kirschneri, L. borgpetersenii, and L. noguchii. These probes were specifically designed for these four species and do not allow for the detection and identification of the new species L. mayottensis.
Tulsiani et al. [31] applied a high-resolution melting curve analysis of random-amplifiedpolymorphic DNA to reference collection strains. MLVA has been previously proposed for typing the L. interrogans, L. borgpetersenii, and L. kirschneri strains [40,13]. In our study, the HRM analysis of the VNTR sequences was performed using ScreenClust HRM software. To the best of our knowledge, our study signifies the first time that Leptospira strains have been genotyped using VNTR primers. Using our method, we obtained a high Hunter-Gaston's discriminatory index (0.98) compared to a serological classification of all the reference strains tested. This high index (> 0.95) shows the genetic polymorphism within each tested species and validates the high sensitivity and specificity of auto-calling genotypes from HRM data using ScreenClust software.
The discriminatory index was relatively lower for L. kirschneri (0.87), which is potentially due to the decreased polymorphism of this species. Using MLVA, Salaün et al. [40] showed that the genetic diversity of L. kirschneri serovars was lower than L. interrogans serovars.
Using the 'unsupervised analysis' mode, HRM analysis using ScreenClust HRM software permitted the determination of clusters without genotypic knowledge. HRM analysis using ScreenClust HRM software also enabled the automated genotyping of unknown samples.
Although the performance measurements of our typing method [41] were excellent (reproducibility 100%, typeability 100%, discriminatory index 0.98), not all the strains (37/48, 77.1%) could be discriminated at the serovar level. The identification of Leptospira serovars by serological methods relies on surface-exposed lipopolysaccharides (LPS) and there is a poor  [40,50] and HRM profiles but are distinguishable by PFGE [51]. For L. kirschneri, the genotype determined by HRM for serovars Bogvere and Djatzi was identical, but they had different profiles using MLVA and PFGE. MLST, PFGE, MLVA, and HRM cannot distinguish between strains belonging to the serovars Copenhageni and Icterohaemorrhagiae. Currently, no single molecular typing method is ideal and fulfils all the characteristics required to unambiguously determine if the isolates from numerous patients and/or animals are identical. However, HRM curve analysis is a simple and effective technique that does not require culture isolation and can be performed in a single test tube in less than two hours. Our molecular investigation of the Leptospira DNA extracted from blood samples revealed that the Leptospira species responsible for 100% of the clinical cases in humans on Reunion Island was L. interrogans. In contrast, on the sister island of Mayotte, only 8.5% of the isolated strains belong to L. interrogans, and no cases were attributed to the serogroup Icterohaemorrhagiae [52]. In our study, the Reunion Island patient infecting strains were all clustered by ScreenClust HRM software using the VNTR-4bis and -Lb5 primers into the L. interrogans strain serogroups Icterohaemorrhagiae serovar Copenhageni or Icterohaemorrhagiae. We can therefore conclude that only one genotype has been involved in human symptomatic leptospirosis on Reunion Island since 2008. This result reflects the persistence of a serovar over time in a geographical area. These results are also confirmed by our previous studies on human patients. Using a serological method, the strains we isolated in culture in 1998 and 2012 were identified as belonging to the L. interrogans serogroup Icterohaemorrhagiae. Between 1998 and 2009, the serogroup Icterohaemorrhagiae was reported to represent 59.3% of the human cases diagnosed on Reunion Island using MAT [53].

Conclusions
Our genotyping method was able to identify a change in genotype that was responsible for the clinical cases on Reunion Island. The unsupervised mode of ScreenClust HRM should also enable one to discover new genotypes that are present in clinical samples. Our genotyping method is limited by the number of leptospires in blood samples, wherein approximately 8250 copies/ml are necessary to obtain 100% typeability.
In this study, the use of HRM confirmed the predominance of a clonal subpopulation of L. interrogans as the cause of human leptospirosis on Reunion Island, suggesting that this method would be useful for epidemiological studies.