ABSTRACT
We describe a multilocus short sequence repeat (MLSSR) sequencing approach for the genotyping of Mycobacterium avium subsp. paratuberculosis (M. paratuberculosis) strains. Preliminary analysis identified 185 mono-, di-, and trinucleotide repeat sequences dispersed throughout the M. paratuberculosis genome, of which 78 were perfect repeats. Comparative nucleotide sequencing of the 78 loci of six M. paratuberculosis isolates from different host species and geographic locations identified a subset of 11 polymorphic short sequence repeats (SSRs), with an average of 3.2 alleles per locus. Comparative sequencing of these 11 loci was used to genotype a collection of 33 M. paratuberculosis isolates representing different multiplex PCR for IS900 loci (MPIL) or amplified fragment length polymorphism (AFLP) types. The analysis differentiated the 33 M. paratuberculosis isolates into 20 distinct MLSSR types, consistent with geographic and epidemiologic correlates and with an index of discrimination of 0.96. MLSSR analysis was also clearly able to distinguish between sheep and cattle isolates of M. paratuberculosis and easily and reproducibly differentiated strains representing the predominant MPIL genotype (genotype A18) and AFLP genotypes (genotypes Z1 and Z2) of M. paratuberculosis described previously. Taken together, the results of our studies suggest that MLSSR sequencing enables facile and reproducible high-resolution subtyping of M. paratuberculosis isolates for molecular epidemiologic and population genetic analyses.
Mycobacterium avium subsp. paratuberculosis (M. paratuberculosis) is a slowly growing, acid-fast, mycobactin J-dependent bacterium. Infection with this bacterium leads to a chronic granulomatous enteritis, termed Johne's disease, in cattle and other ruminants and occurs worldwide (11). Clinical signs of the disease include diarrhea, weight loss, fatigue, decreased milk production, and mortality. Infection with this pathogen results in considerable economic losses in the dairy production industry, with estimated annual costs of $40 to $227 per year per cow, totaling industry-wide annual losses of $1.5 billion (20, 21). In addition to the serious health and economic impacts of the pathogen to the dairy industry, several reports suggest a possible link between M. paratuberculosis and Crohn's disease in humans (5, 7, 10, 30).
Methods for differentiation or subtyping of bacterial strains provide important information for molecular epidemiologic analysis and assist in providing an understanding of the population genetics of the species. DNA-based molecular subtyping techniques such as multiplex PCR for IS900 integration loci (MPIL) (4, 20), restriction fragment length polymorphism (RFLP) analysis (6, 8, 23), and amplified fragment length polymorphism (AFLP) analysis (20) have been previously applied to investigate genetic variation in M. paratuberculosis. However, the MPIL, AFLP, and RFLP techniques are generally unable to resolve M. paratuberculosis isolates into meaningful epidemiologic groups due to the apparently restricted genetic diversity within the subspecies. Furthermore, the data generated by these techniques are biallelic and, hence, are able to provide only limited information regarding the overall genetic diversity and evolutionary mechanisms within the species.
Short sequence repeats (SSRs) or variable-number tandem repeats (VNTRs) in bacterial DNA have been used as markers for the differentiation and subtyping strains of several bacterial species, including Mycobacterium tuberculosis (9, 16), Yersinia pestis (1), Salmonella enterica subsp. enterica serovar Typhimurium (17), and Bacillus anthracis (15). SSRs consist of simple homopolymeric tracts of a single nucleotide (mononucleotide repeats) or multimeric tracts (homogeneous or heterogeneous repeats), such as di- or trinucleotide repeats, which can be identified as VNTRs in the genome of the organism (35). The variability of the repeats is believed to be caused by slipped-strand mispairing (31); the genetic instability of polynucleotide tracts, especially poly(G-T) (12); and DNA recombination between homologous repeat sequences (32).
The complete genome sequence of M. paratuberculosis strain K10 (GenBank accession number AE016958) has recently been characterized in the Department of Microbiology and Biomedical Genomics Center, University of Minnesota(L. L. Li et al., unpublished data). Preliminary bioinformatic analyses led to the identification of numerous SSRs in the M. paratuberculosis genome. We evaluated the utility of a multilocus SSR (MLSSR)-based typing approach for differentiating among isolates of M. paratuberculosis. The results of our studies suggest that MLSSR is a useful approach for strain differentiation and enables the rapid and facile discrimination of epidemiologically and geographically distinct strains of M. paratuberculosis.
MATERIALS AND METHODS
Bacterial isolates and DNA isolation.A total of 33 M. paratuberculosis isolates from different host species and geographic locations were used in this study, as shown in Table 1. M. paratuberculosis isolates were grown on Middlebrook 7H9 broth or 7H11 agar (Difco Laboratories, Detroit, Mich.) with oleic acid-albumin-dextrose-catalase supplement (Becton Dickinson, Sparks, Md.) and mycobactin J (2 mg/100 ml). In some instances the bacterial cultures were incubated at 37°C for 4 to 6 months until colonies were observed. DNA was isolated from the bacterial culture by use of the QIAamp DNA Mini kit (Qiagen Inc., Valencia, Calif.), as described previously (20).
M. paratuberculosis isolates examined in this study
Database search for SSRs and primer design.The whole-genome sequence of M. paratuberculosis strain K10 was analyzed for SSRs with Tandem Repeat Finder (version 2.02) software (3). The coordinates of the SSRs were then matched for the regions upstream and downstream to locate the repeats and open reading frame (ORF) flanking the repeat by use of the DNA sequence viewer and annotation software Artemis (28). Primers specific for sequences flanking these repeat sequences were designed with Primer 3 software (27) to yield an average amplification product of ∼250 bp for each sequence (Table 2).
Primers for 78 sequence repeat loci used for polymorphism analysis of the region
MLSSR.A total of 78 loci were amplified by PCR with specific primers, and the amplification products were sequenced to identify sequence polymorphisms in each locus among six strains of M. paratuberculosis (reference strain MAP-K-10 and isolates from cattle [isolates MAP-08 and MAP-09], sheep [isolates MAP-06 and MAP-11], and a human [isolate MAP-14]) (Table 1). These six M. paratuberculosis isolates were selected because they represent the extent of genetic diversity in the species, as previously identified by MPIL and AFLP analyses (20).
The 25-μl PCR amplification reaction mixture for each SSR comprised 1× PCR buffer II (Perkin-Elmer, Applied Biosystems Division, Foster City, Calif.), 2.0 mM MgCl2 (Perkin-Elmer), 200μ M each deoxynucleoside triphosphate (Roche Diagnostic Co., Indianapolis, Ind.), 0.6 μmol of each primer (Integrated DNA Technologies, Coralville, Iowa), 0.5 U of AmpliTaq Gold (Perkin-Elmer), 5% dimethyl sulfoxide (Sigma Chemical Co, St. Louis, Mo.), and 1 μl of DNA. The amplification conditions consisted of an initial denaturation at 94°C for 15 min, followed by 35 cycles of denaturation at 94°C for 45 s, annealing at 60°C for 1 min, and extension at 72°C for 2 min 30 s, with a final extension step at 72°C for 7 min. A 2-μl volume of the PCR products was mixed with 2μ l of loading buffer (0.2% Orange G in 50% glycerol), and the mixture was electrophoresed in 1% agarose with 0.5 μg of ethidium bromide per ml. The gels were photographed under UV light with an Eagle Eye II gel documentation system (Stratagene, La Jolla, Calif.). The PCR amplicons were then sequenced with an ABI 3100 automated fluorescent DNA sequencer (Perkin-Elmer) at the University of Minnesota's Advanced Genetic Analysis Center (www.agac.umn.edu).
MLSSR data analysis.The sequences of each SSR locus of 33 isolates were aligned, and the numbers of tandem repeats were identified by use of the MegAlign program (DNASTAR Inc., Madison, Wis.). The nucleotide sequences of 11 polymorphic SSR loci were analyzed for each isolate, and allele numbers were assigned to reflect the number of copies or the number of nucleotide substitutions represented in the SSR sequence for each locus. Statistical analysis for genetic diversity and overall relationships among the isolates was performed with the computer programs ETDIV and ETCLUS, which were modified for use with the SSR data (2). MLSSR types were then assigned on the basis of the unique combination of alleles for each locus. Genetic diversity (D) was calculated by using the following equation: 1 − Σ(allele frequency)2(22, 29). The unweighted pair group method with arithmetic averages-based cluster analysis and bootstrap analysis with 1,000 replications were performed with the program PAUP (version 4.0; Sinauer Associates, Inc. Sunderland, Mass.), and the index of discrimination (D) was determined as described previously (13).
RESULTS
SSRs in M. paratuberculosis genome.Analysis of the whole-genome sequence of M. paratuberculosis strain K10 (4.83 Mbp) identified 185 SSRs consisting of three or fewer nucleotides per repeat unit. Of these, 78 mono-, di-, and trinucleotide repeats with perfect matches between adjacent copies were identified and were included as candidate polymorphic loci for further analysis (Table 2). These 78 SSR loci were also selected for inclusion in our analysis because they were short (1 to 3 bp), as is common in prokaryotes, and each locus had at least five copies. Dinucleotide repeats were the most frequently identified SSRs in the M. paratuberculosis genome and were present at 63 distinct loci, with the copy numbers varying between 5 and 5.5 per repeat. Mono- and trinucleotide repeats were represented at 2 and 13 loci, respectively.
MLSSR analysis revealed that 11 of the 78 loci were polymorphic in the six isolates examined. The ORFs or genes flanking each locus were also identified (Table 3). For example, locus 2 is located within ORF 210_MAP.128, which is unique to M. paratuberculosis. Locus 3 was identified in an intergenic region between two ORFs: a 5′ ORF encoding 6-aminohexanoate-cyclic dimer and a 3′ ORF encoding alpha/beta-hydrolase (Table 3). The functional consequences of the presence of the loci and the influence of the locus copy number on the expression of the adjacent genes deserve further investigation.
SSRs used in MLSSR analysis
MLSSR.The 11 polymorphic SSR loci identified in the preliminary screening were characterized in 27 additional M. paratuberculosis isolates that were previously characterized by MPIL and AFLP analyses (20). The analysis identified 20 MLSSR types among the 33 M. paratuberculosis isolates recovered from different host species and geographic areas (Tables 1 and 4). The D value for each SSR locus was calculated on the basis of the allele frequency and the number of alleles and revealed an average number of alleles per locus of 3.20, with an average D value of 0.393 and a range of D values of 0.100 to 0.700 (21, 28) (Table 3). While the allelic variation observed in this study focused on the number of copies of the SSRs (Fig. 1A), it is noteworthy that some loci also revealed one or two base substitutions in some isolates (Fig. 1B). For instance, the analysis revealed a single polymorphic site each at SSR loci 4 and 10 and four and five polymorphisms at loci 5 and 9, respectively (Fig. 2). It is interesting that the vast majority of the nucleotide substitutions were found in MAP-06, an isolate recovered from a sheep.
Sequence analysis of two representative SSR loci. (A) Locus 8 with (GGT)5 repeats. Strain MAP-K10 contains five copies of GGT, while MAP-08 and MAP-11 contain four and three copies of GGT, respectively. (B) Locus 10 with (GCC)5 repeats. Strain MAP-K10 contains five copies of GCC, while MAP-11 and MAP-07 have a A-to-C substitution in a single copy of GCC.
Allelic variation at 11 SSR loci among 33 M. paratuberculosis isolates. The aligned nucleotide sequences of each of the alleles at the 11 SSR loci discovered and characterized during this investigation, along with adjacent conserved sequences, are shown. The SSRs and polymorphic sites are boxed. The locations of single-nucleotide polymorphisms are indicated with asterisks below the aligned sequences.
Profiles of alleles at 11 SSR loci for 20 clones of M. paratuberculosis
Genetic relationships among M. paratuberculosis isolates based on MLSSR analysis.The unweighted pair group method with arithmetic averages-based cluster analysis of M. paratuberculosis identified 20 distinct MLSSR types among the isolates that were grouped into two major clusters, clusters M and N (Fig. 3). Cluster M contained 87.88% (29 of 33) of the isolates in the sample, including isolates recovered from bovine, caprine, murine, deer, rabbit, and human sources. The isolates in cluster M with the most common MPIL and AFLP fingerprints, A18 and Z1 and Z2, respectively, were further divided into three groups, clusters M1, M2, and M3. Cluster M1 contained one isolate (isolate MAP-06), which was recovered from a sheep and which had the A1 MPIL fingerprint. Three of the five isolates from caprine sources were assigned to cluster M2. A total of 13 unique genotypes, including a majority (10 of 15) of the bovine isolates included in this study, were represented in cluster M3. In addition, the three isolates from human sources included in the sample used in this study were also found in cluster M3. Interestingly, two isolates recovered from humans (isolates MAP-14 and 0003) were clustered into the same clade as an isolate of bovine origin (isolate 0180). Isolates that were recovered from a mouse (isolate 0012), rabbits (isolates 0237 and 0160 to 0162), a deer (isolate 0883), and soil (isolate 0560) were also grouped along with the M3 genotype.
Dendrogram depicting genetic relationships among 33 M. paratuberculosis isolates on the basis of the 11 SSR loci determined by MLSSR analysis. The dendrogram was generated by the unweighted pair-group method with arithmetic averages with the PAUP program. The results of the bootstrap analysis are represented as percentages and are indicated adjacent to the major nodes when the branch order was supported by> 50% of the 1,000 replicate trees. Genetic distance is indicated at the top of the dendrogram. Isolate identifications, sources, geographic locations, and MILP and AFLP types are shown to the right of the dendrogram.
In contrast to cluster M, which consisted of isolates recovered from a variety of animal species, all four isolates that were included in cluster N were recovered from sheep. Strains of ovine origin (four of five) also showed a distinct allelic profile compared with the profiles of strains from cattle, goats, or humans.
Discriminatory power of subtyping methods.The discriminatory power (D) of MLSSR in comparison with those of other subtyping methods was determined as described previously (13). MPIL analysis differentiated the 27 M. paratuberculosis isolates for which MPIL typing information was available into 6 subtypes with a D value of 0.50, indicating only limited discriminatory power, while MLSSR differentiated 27 M. paratuberculosis into 17 subtypes with a D value of 0.96. In contrast, AFLP analysis differentiated the 24 M. paratuberculosis isolates for which AFLP typing information was available into 15 subtypes, with a D value of 0.92 (20). In comparison, MLSSR differentiated the 24 M. paratuberculosis isolates into 14 subtypes, with higher D value of 0.95. Overall, MLSSR distinguished 20 subtypes among the 33 isolates in the sample with a D value of 0.96, indicating that it has a relatively high index of discrimination (Tables 3 and 4).
DISCUSSION
SSRs have been used to type many bacterial pathogens associated with human and animal infections (32). Within the genus Mycobacterium, VNTR or mycobacterial interspersed repetitive units have been used for the subtype-specific differentiation of several Mycobacterium species (19, 26, 35). In the present study we have identified polymorphic SSRs by genomic analysis of M. paratuberculosis and used this information to develop a highly discriminatory method for the typing of M. paratuberculosis isolates.
The SSRs discovered during our preliminary screening of the M. paratuberculosis genome were similar to the repeats that have previously been described in other bacteria, including Haemophilus, Mycoplasma, and Mycobacterium spp. (24, 32, 33). It has previously been recognized that regions of mono-, di-, and trinucleotide tandem repeats are often the most diverse in a bacterial genome, while complex longer repeats generally have lower levels of diversity (14). This is thought to result from slipped-strand mispairing (or replication slippage events) of the DNA polymerase that occurs with greater frequency on the SSRs, a hypothesis that remains to be tested for the SSRs that we have identified in M. paratuberculosis (32).
Several important attributes of a strain differentiation assay determine its utility in a clinical or epidemiologic setting. Especially for organisms such as M. paratuberculosis that have restricted levels of genetic diversity, the discriminatory power of an assay is a particularly important attribute. Assays such as MPIL and RFLP analysis have been shown to have only moderate abilities to differentiate among epidemiologically distinct isolates of M. paratuberculosis and therefore have limited applicabilities in molecular epidemiologic studies (4, 5, 33). The recently described AFLP technique has been shown to have a greater resolving power than the other two approaches but suffers from the limitation that allelic variation is indexed at anonymous biallelic markers (20). In contrast, the MLSSR assay described herein is far more discriminatory, being able to differentiate 33 M. paratuberculosis from distinct geographic localities and host species into 20 subtypes on the basis of allelic variation at the 11 SSR loci examined, with a notably high D value of 0.96. Consistent with its high discriminatory power, MLSSR enabled the differentiation of seemingly monomorphic M. paratuberculosis strains that were indistinguishable by MPIL and AFLP analyses (20). An important advantage of the MLSSR approach is that it also indexes variations at known genetic loci and has the ability to identify multiple alleles per locus. Together, these attributes not only allow an increase in the strain-resolving power of the assay but also enable an understanding of the genetic mechanisms driving strain diversification and evolution within the species.
Another key attribute of a strain differentiation assay is its ability to identify epidemiologically and genetically related strains of a bacterial species. In this context, MLSSR analysis clearly showed that some isolates that are of sheep origin (cluster N) are genetically distinct from those of bovine, caprine, and human origin (cluster M), a finding consistent with those of previous studies (4, 6, 20). It is noteworthy, however, that the five isolates of sheep origin examined during this study were represented by three distinct MLSSR types (MLSSR types 1, 19 and 20), and four isolates clustered together in cluster N. Interestingly, all four of these phylogenetically linked M. paratuberculosis isolates were recovered from sheep in South Dakota, suggesting that they are both genetically and epidemiologically related and well distinguishable from the other isolates in the collection. The same isolates were also grouped into four distinct MPIL genotypes (A1, A8, A16, and A17) and three AFLP genotypes (Z7, Z8, and Z18), suggesting that they are indeed genetically distinct from the other isolates in the collection. However, by the MPIL and the AFLP approaches, these isolates do not cluster together as closely as they do by MLSSR analysis (20). Hence, these results suggest that MLSSR analysis may enable molecular epidemiologic investigations that will lead to a better understanding of strain transmission and the spread of M. paratuberculosis in natural populations and across host species.
In contrast to the relatively close clustering of the sheep M. paratuberculosis strains in the samples examined, far greater diversity was observed in isolates of bovine origin. The analysis showed that while a majority of the M. paratuberculosis isolates of bovine origin clustered together in the M3 subgroup, 60% (three of five) of the caprine isolates were represented by the closely related cluster M2, suggesting that caprine isolates bear greater genetic resemblance to cattle strains than to isolates of ovine origin, a finding that is consistent with the findings of previous studies (34). Similarly, deer and cattle strains also appeared to be more closely related to each other by MLSSR analysis, suggesting a sharing of strains of M. paratuberculosis in wildlife species that graze or that may otherwise come into close contact with cattle, as hypothesized previously (25).
Our studies demonstrate that MLSSR analysis offers several advantages over other methods for differentiating among M. paratuberculosis isolates. First, as described above, the technique has a high discriminatory power for known multiallelic genetic loci, an essential attribute for the effective differentiation of genetically distinct isolates. Second, MLSSR results are based upon DNA sequencing and, hence, are unambiguous and reproducible and can likely be obtained for most loci of all M. paratuberculosis isolates, even those recovered from sheep or wildlife species, as demonstrated by our studies described herein. However, we note the formal possibility that mutations or deletions at the primer sites may render some strains untypeable at some loci, such as loci 10 and 11 in MAP-06. Third, MLSSR analysis is based on the amplification of SSR loci by PCR and thus not only is rapid but also may be performed directly with bacterial colonies without DNA extraction. Fourth, due to the considerable advances in automated DNA sequencing technologies and the falling prices of DNA sequencing, the MLSSR method is amenable to adaptation for high-throughput analysis and can be performed relatively inexpensively as well. Finally, a key advantage of the approach is that the data are sequence based and, hence, enable accurate interlaboratory comparisons to be made and the information used in the development of SSR databases for further molecular epidemiologic studies, which are greatly required in this field (18). While it must be recognized that sequence errors due to strand slippage during either PCR or sequencing reactions may result in an erroneous assignment of genotype, the occurrence of such slippage errors is minimized by increasing the amount of sequence coverage at the locus (by confirming both the forward and the reverse sequences or testing duplicate samples), as is routinely practiced in our laboratory.
In conclusion, we have described here the development of MLSSR-based typing for the subtype-specific differentiation of M. paratuberculosis isolates. Our preliminary analyses suggest that this approach will be of considerable utility in enabling detailed molecular epidemiologic and population genetic analyses of this important animal pathogen.
ACKNOWLEDGMENTS
Research in the laboratory of V. Kapur is funded by grants from the U.S. Department of Agriculture, the National Institutes of Health, and the Minnesota Agricultural Experimental Station. Research in the laboratory of S. Sreevatsan is funded through a seed grant from the Ohio Agricultural Research and Development Center's research enhancement competitive grants program.
FOOTNOTES
- Received 14 August 2003.
- Returned for modification 23 October 2003.
- Accepted 2 January 2004.
- American Society for Microbiology
